From Multistep Enzyme Monitoring to Whole-Cell Biotransformations: Development of Real-Time Ultraviolet Resonance Raman Spectroscopy

Article abstract of DOI:10.1021/acs.analchem.7b03742

Process analytical technologies (PAT) are used within industry to give real-time measurements of critical quality parameters, ultimately improving the quality by design (QbD) of the final product and reducing manufacturing costs. Spectroscopic and spectrophotometric methods are readily employed within PAT due to their ease of use, compatibility toward a range of sample types, robustness, and multiplexing capabilities. We have developed a UV resonance Raman (UVRR) spectroscopy approach to quantify industrially relevant biotransformations accurately, focusing on nitrile metabolizing enzymes: nitrile hydratase (NHase) and amidase versus nitrilase activity. Sensitive detection of the amide intermediate by UVRR spectroscopy enabled discrimination between the two nitrile-hydrolyzing pathways. Development of a flow-cell apparatus further exemplifies its suitability toward PAT measurements, incorporating in situ analysis within a closed system. Multivariate curve resolution-alternating least-squares (MCR-ALS) was applied to the UVRR spectra, as well as off-line HPLC measurements, to enable absolute quantification of substrate, intermediate, and product. Further application of hard modeling to MCR-ALS deconvolved concentration profiles enabled accurate kinetic determinations, thus removing the requirement for comparative off-line HPLC. Finally, successful quantitative measurements of in vivo activity using whole-cell biotransformations, where two Escherichia coli strains expressing either NHase (transforming benzonitrile to benzamide) or amidase (further conversion of benzamide to benzoic acid), illustrate the power, practicality, and sensitivity of this novel approach of multistep and, with further refinement, we believe, multiple micro-organism biotransformations.

Full text of DOI:10.1021/acs.analchem.7b03742

Subscriber access provided by READING UNIV
Article
From multi-step enzyme monitoring to whole-cell
biotransformations: development of real-time UVRR spectroscopy
Heidi Fisk, Yun Xu, Chloe Westley, Nicholas J Turner, Jason Micklefield, and Royston Goodacre
Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b03742 • Publication Date (Web): 27 Oct 2017
Downloaded from http://pubs.acs.org on October 29, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth
Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
From multi-step enzyme monitoring to whole-cell
biotransformations: development of real-time UVRR
spectroscopy
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Heidi Fisk, Yun Xu, Chloe Westley, Nicholas J. Turner, Jason Micklefield and Royston
Goodacre
School of Chemistry, Manchester Institute of Biotechnology, University of Manchester, 131
Princess Street, Manchester, M1 7DN, UK
*
Corresponding author roy.goodacre@manchester.ac.uk
Keywords: UV resonance Raman, biocatalysis, whole-cell biotransformation, MCR-ALS, hard
modelling.
ACS Paragon Plus Environment

Analytical Chemistry
Page 2 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ABSTRACT
Process Analytical Technologies (PAT) are used within industry to give real-time measurements of
critical quality parameters, ultimately improving the quality by design (QbD) of the final product and
reducing manufacturing costs. Spectroscopic and spectrophotometric methods are readily employed
within PAT due to their ease of use, compatibility towards a range of sample types, robustness and
their multiplexing capabilities. We have developed a UV resonance Raman (UVRR) spectroscopy
approach to quantify industrially-relevant biotransformations accurately, focusing on nitrile
metabolising enzymes: nitrile hydratase (NHase) and amidase versus nitrilase activity. Sensitive
detection of the amide intermediate by UVRR spectroscopy enabled discrimination between the two
nitrile-hydrolysing pathways. Development of a flow-cell apparatus further exemplifies its suitability
towards PAT measurements, incorporating in situ analysis within a closed system. Multivariate curve
resolution-alternating least squares (MCR-ALS) was applied to the UVRR spectra, as well as off-line
HPLC measurements, to enable absolute quantification of substrate, intermediate and product.
Further application of hard modelling to MCR-ALS deconvolved concentration profiles enabled
accurate kinetic determinations, thus removing the requirement for comparative off-line HPLC.
Finally, successful quantitative measurements of in vivo activity using whole-cell biotransformations,
where two Escherichia coli strains expressing either NHase (transforming benzonitrile to benzamide)
or amidase (further conversion of benzamide to benzoic acid), illustrates the power, practicality and
sensitivity of this novel approach for multi-step and with further refinement we believe multiple
micro-organism biotransformations.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 3 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
INTRODUCTION
Nitrile hydrolysing enzymes, namely, nitrilases, nitrile hydratases (NHase) and amidases, are
extensively used within chemical synthesis in industry owing to their high selectivity and activity as
1
well as their broad substrate specificity. These nitrile metabolising enzymes work at ambient
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
temperature(s) and pressure(s) and can offer enantioselective control, meaning they are attractive
alternatives within pharmaceutical processes. At present, acrylamide (>30,000 tons per year),
nicotinamide (Vitamin B
3
, >6,000 tons per year), (R)-mandelic acid (a drug precursor), pyrazinamide
(an antituberculosis agent) and 5-cyanovaleramide (a herbicide intermediate) are largely synthesised
2
-7
using nitrile hydrolysing biocatalytic routes. There has also been increasing interest in the
detoxification of wastewater containing toxic nitriles of anthropogenic origins using nitrile
7
,8
hydrolysing enzymes.
Due to the appealing nature of these biocatalysts, different techniques to monitor the progression of
these biotransformations have been investigated. On-line techniques (in contrast to off-line) are
often more desirable as they directly monitor the reaction in real-time, providing rapid, continuous
feedback. As such, sample manipulation, extraction or (partial) purification is no longer required
prior to analysis, reducing the error by minimizing the need for sample transfers and handling.
Furthermore, real-time reaction monitoring greatly improves the efficiency and accuracy of the
overall process within PAT (process analytical technology), an essential component of QbD (quality
9
,10
by design). Fluorometric and colorimetric assays are examples of on-line techniques that have
been widely studied to monitor nitrile hydrolysis. Examples of colorimetric assays include those that
1
1-14
rely on changes in pH,
whereas spectrophotometric assays have involved monitoring ferrous and
1
5,16
ferric ions,
as well as NADH consumption when paired with an amidase and glutamate
1
7
dehydrogenase. Whilst these techniques are advantageous over other more time consuming and
labour-intensive off-line techniques, such as HPLC, NMR or GC-MS or LC-MS, a major limiting factor
is the requirement of a fluoro- or chromo-genic substrate, along with limited diagnostic and
1
8,19
structural information.
Consequently, there is an increased effort towards the emergence of new
and improved high-throughput screening (HTS) methods within biocatalysis and in situ PAT
techniques. Raman spectroscopy offers considerable potential as a physicochemical technique for
monitoring real-time enzymatic catalysed biotransformations. However, ‘normal’ Raman scattering
3
5
is often too weak, limiting its application, yet enhancements of 10 – 10 are attainable when deep-
UV (244 nm in this case) excitation is employed. In UV resonance Raman (UVRR) spectroscopy the
laser frequency coincides with the electronic transitions of the molecules under investigation,
20,21
specifically aromatic and conjugated systems, thus enhancing the Raman response.
ACS Paragon Plus Environment

Analytical Chemistry
Page 4 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
We recently demonstrated that UVRR could successfully monitor the real-time progression of
biotransformations using either NHase or xanthine oxidase (conversion of hypoxanthine to xanthine,
2
2
followed by xanthine to uric acid). UVRR achieved results that were in agreement with off-line
HPLC analysis, but with a >30-fold reduction in acquisition time (20 s measurement time for UVRR
versus >10 min for each HPLC run). Ultimately, rapid and sensitive detection of these analytes was
possible by exploiting aromatic functionality which is inherent to their structure.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 1. Enzymatic conversion of benzonitrile to the corresponding carboxylic acid (benzoic acid) and/or
carboxamide (benzamide).
In this study, we further investigate the application of UVRR to discriminate quantitatively between
nitrile metabolising enzymes. Nitrile catabolism comprises of two distinct pathways: (1) direct
generation of carboxylic acids via nitrilase; and (2) a two-step process whereby nitrile hydratases
catalyse the formation of an amide, which is subsequently hydrolysed by an amidase to give the
1
corresponding carboxylic acid (Figure 1). Consequently, discrimination between these pathways
until now has been problematic as they generate the same final product, resulting in only semi-
quantitative or qualitative analysis. We describe how UVRR spectroscopy can be used in real-time to
easily distinguish between the two pathways, without the requirement of additional fluoro- or
chromo-genic substrates, enabling full recovery of the reaction sample whilst affording quantitative
analysis. Furthermore, a flow-cell apparatus has been developed to reduce signal-to-noise
interference, further demonstrating its amenability for use in industrial processes (closed system
enabling direct in situ measurements). Finally, investigations were extended to include whole-cell
biotransformations (in vivo), omitting protein purification steps thus further increasing the overall
speed and utility of this technique.
ACS Paragon Plus Environment

Page 5 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
EXPERIMENTAL
Materials. All chemical reagents were of analytical grade and used with no additional purification
unless otherwise stated. A nitrile hydratase (EC 4.2.1.84) construct was kindly provided by the Uwe
2
3
Bornscheuer group (Greifswald University), and an amidase (EC 3.5.1.4) synthetic gene was
purchased from Genewiz (New Jersey, USA). Full details of cloning, expression and protein
purification is described within the Supporting Information.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Reaction sample preparation. The starting reaction mixture contained either benzonitrile or
-2
-2
benzamide (final concentration 1.25 x 10 M) in potassium phosphate buffer (2.5 x 10 M, pH 7.2).
Reactions were performed using either purified enzyme (in vitro) or Escherichia coli whole-cells
(resting state) which expressed either NHase or amidase protein (in vivo). For reactions involving
-
6
-6
purified enzymes, NHase + amidase (6.0 x 10 M and 3 x 10 M, respectively) were added to the
reaction mixture to initiate the biotransformation. For whole-cell biotransformations, NHase or
amidase (500 µL or 250 µL respectively, of concentrated whole-cells in buffer) were introduced into
the reaction mixture or a combination of the two (375 µL of NHase and 25 µL of amidase of whole-
cells in buffer). See Supporting Information, supplementary methods for further information.
UVRR Instrumentation. 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
sample using a Lexel Model 95 ion laser (frequency doubled) emitting at 244 nm.
Reaction set-up. UVRR monitoring of biotransformations were performed using a flow-cell set-up.
In brief, a quartz flow-cell was focussed beneath the UVRR microscope objective, which was
connected to the reaction reservoir using tubing (Figure S1). Continuous stirring of the reaction
mixture was achieved using a peristaltic pump, along with a stirrer bar agitating the reaction
reservoir (the location at which enzyme is added and HPLC samples removed). A total reaction
volume of 10 mL was used, UVRR analysis was performed using a 20 s acquisition time at various
intervals (mainly every minute). In addition, every 2-3 min 20 µL of the reaction mixture was
removed and quenched with MeOH (180 µL), before being subjected to comparative HPLC analysis.
ACS Paragon Plus Environment

Analytical Chemistry
Page 6 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
RESULTS AND DISCUSSION
Multi-step enzyme biocatalysis
To explore if UVRR could distinguish between the two nitrile metabolising pathways, we investigated
the more complex, two-step pathway combining both a nitrile hydratase and an amidase. A flow-cell
apparatus was developed and optimized to reduce signal-to-noise interference by incorporating a
quartz flow-cell at the site of UVRR analysis (images of the set-up are shown in Figure S1).
Benzonitrile was investigated as the starting material (SM), which produces both benzamide as the
intermediate (I) and benzoic acid as the final product (P) in the presence of NHase and amidase
(Figure 1). Characteristic UVRR peaks for each analyte are highlighted in Figure 2 (see Table S1 in the
Supporting Information for tentative band assignments). Benzamide and benzoic acid produce very
similar UVRR responses yet subtle differences are seen, most noticeably, one can observe the peak
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
-
1
-1
shift from 1413 to 1389 cm (benzamide to benzoic acid) and the weak peak at 847 cm which is
absent in both benzonitrile and benzamide but present within the benzoic acid product.
Figure 2. Average UVRR spectra (n = 2) of each analyte under investigation: benzonitrile (blue), benzamide
-2
(
(
red) and benzoic acid (green). Spectra were obtained at 1.1 x 10 M in potassium phosphate buffer
2.5 x 10 M, pH 7.2), using conditions and concentrations that were representative of the initial
-2
reaction mixtures, with characteristic peaks identified (see Table S1 for assignments). UVRR spectra
were obtained for 20 s with baseline correction and normalisation applied (see the Supporting
Information, “Data processing” for full details).
ACS Paragon Plus Environment

Page 7 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
Initially we looked at the conversion of benzonitrile to benzamide and subsequently to benzoic acid
using NHase and amidase in vitro (purified enzymes), with comparative HPLC measurements
acquired to confirm that the in situ UVRR measurements were accurate. Multivariate curve
resolution-alternating least squares (MCR-ALS) is a soft-modelling method, used mathematically to
deconvolve an instrumental response into the pure contributions of individual components present
within a mixture. The input of pure spectra of each component of the mixture is an initial
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
requirement, and the iterative method can then deconvolve the complex mixture and provide
24,25
concentration profiles of each component.
To benchmark this approach HPLC measurements of
the same samples allows the comparison of the known analyte concentrations with the predicted
UVRR concentrations via MCR-ALS modelling, thus serving as external validation (an overview of the
spectral pre-processing and MCR-ALS process is shown in Figure S2). One can also compare how well
the pure analyte spectra (as initially inputted into the model) and the resolved MCR-ALS spectra
agree with one another and in this case the real and deconvolved were congruent (Figure S3).
It is important to note that UVRR spectra were acquired every minute throughout the monitoring
process, whereas HPLC aliquots were taken less frequently (2-3 minute intervals). Time points
consisting of both HPLC and UVRR measurements were used as the training set for the MCR-ALS
model, with only UVRR measurements used as the test set. Figure 3 shows the resolved
concentration profiles, illustrating very good agreement between the HPLC and UVRR predictions.
The MCR-ALS model correctly recognized all three components within the reaction mixture, despite
their spectral similarities. Thus, enabling distinction between the two-step pathway (NHase +
2
amidase) versus nitrilase catalysis (no intermediate). Regression-coefficient values (
R
) demonstrate
the proportion of variability within a data set, as accounted for by the statistical model (MCR-ALS).
2
Generally,
R
values closer to 1 demonstrate excellent fit and correlation between the HPLC
measured and UVRR predicted concentrations. Results are provided in Figure S4 for all three analyte
2
components – benzonitrile (SM), benzamide (I) and benzoic acid (P) – with their corresponding
R
values: 0.9637, 0.9076 and 0.9895, respectively, indicating very good agreement between the two
techniques.
ACS Paragon Plus Environment

Analytical Chemistry
Page 8 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. An MCR-ALS model was applied to the UVRR data where it successfully deconvolved spectra into its
pure components for the biotransformation using pure enzymes. This figure shows the reaction
dynamics from the real-time UVRR measurements (denoted by circular symbols) and off-line HPLC
data (denoted by cross symbols) as a function of time for the conversion of benzonitrile (SM; blue)
to benzamide (I; red) to benzoic acid (P; green), catalysed by NHase and amidase, correspondingly.
To complement the incorporation of UVRR within on-line PAT processes further, we wanted to
2
6,27
explore the application of hard modelling (HM),
as this would remove the requirement for
external calibration (i.e., the additional HPLC measurements). The multivariate curve resolution-
alternating least squares-hard modelling (MCR-ALS-HM) process is shown in Figure S5. In brief,
assuming that the two steps of the biotransformations both followed a first-order reaction, the
resolved concentration profiles from MCR-ALS were used to determine the reaction rate constants
k and k using the kinetic models as shown in Figure S6. The concentrations of the reactants at any
1
2
time points during the reaction were then derived using the estimated k , k and the known initial
1
2
concentration of the starting material (i.e., benzonitrile). The MCR-ALS-HM results are shown in
Figure 4 along with comparative HPLC results to highlight their consistency. One can see that the
results are once again in good agreement, particularly for the benzonitrile (SM). However,
benzamide and benzonitrile HM concentration profiles do not agree with the HPLC calibration
concentrations quite as well. The MCR-ALS-HM approach assumes that 100 % of the SM (i.e.,
benzonitrile) is converted to product(s) (i.e., benzamide + benzoic acid). Upon further inspection of
ACS Paragon Plus Environment

Page 9 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
the total concentration throughout the monitoring process, we experience significant fluctuations in
total mass, consequently causing a disparity between the HPLC and MCR-ALS-HM results. The total
analyte concentration at various timepoints (as calculated by HPLC) is shown in Table S2. After the
initial timepoint, a substantial drop in the total concentration occurs and a third of the starting
concentration is unaccounted for, which subsequently increases again with increasing time (final
timepoint concentration = 96 % of starting concentration). We hypothesise that these variations in
the total concentration is a consequence of reduced solubility of the intermediate (benzamide),
resulting in partial insolubility before being converted to benzoic acid (product) where it is soluble
once again due to the reduced concentration of benzamide. As the MCR-ALS and HPLC results
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(
Figure 3) were in such good agreement with one another, we can conclude that both techniques
UVRR and HPLC) are detecting this reduction in concentration and it is a true effect. Unfortunately,
(
the HM results cannot adjust for this mass imbalance, thus we see a discrepancy between the MCR-
ALS-HM and HPLC results (Figure 4). Nevertheless, we can calculate reaction rate constants using the
1 2
MCR-ALS-HM model and these constants were k = 0.3316 and k = 0.0797 for NHase and amidase,
respectively (the equations for kinetic calculations are shown in Figure S6).
k
k
1
2
= 0.3316
= 0.0797
Figure 4. Multivariate curve resolution-alternating least squares-hard modelling (MCR-ALS-HM) results show
good agreement with HPLC measured concentrations. Calculated kinetic rates of each enzyme were
k
1
= 0.3316 and k
2
= 0.0797 for NHase and amidase, respectively.
ACS Paragon Plus Environment

Analytical Chemistry
Page 10 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Whole-cell biotransformations incorporating multi-step enzyme biocatalysis
In addition to the in vitro pure enzyme assessment above, we wanted to perform this
biotransformation using in vivo conditions using whole-cells. Omitting protein purification steps
would increase the overall efficiency and speed of this real-time monitoring approach, further
proving its suitability within industrial processes. NHase and amidase were expressed within E. coli
BL21 (DE3) cells using optimised expression conditions (see Supporting Information section ‘Protein
expression and purification’). Once cells were induced for protein expression and allowed to grow
overnight, the cells were harvested by centrifugation and re-suspended in potassium phosphate
buffer prior to use within the reaction. To ensure that bacterial cells were in the resting state and no
further growth was experienced, optical density (OD) measurements at 600 nm were performed
throughout the monitoring process. OD600 measurements were relatively consistent throughout the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
6
0 min period and thus we conclude that there was no significant further bacterial growth (see
Supporting Information section ‘OD600 measurements’ and Table S3). Initially we performed two
separate whole-cell biotransformations: the first contained E. coli expressing NHase for conversion
of benzonitrile (SM) to benzamide (P), and the second biotransformation with benzamide (SM) used
E. coli expressing amidase, generating benzoic acid (P). In order to avoid photo-damage of these E.
coli, as shown in Figure S1, a peristaltic pump was used to flow the bacteria reaction mixture
continuously through the quartz flow-cell that was used for collection of UVRR spectra.
The first one-step, in vivo biotransformation studied was the conversion of benzonitrile to
benzamide using E. coli cells expressing NHase activity. Consistent with our previous investigations,
both UVRR and HPLC measurements were taken and MCR-ALS analysis performed. The MCR-ALS
resolved profiles for starting material and product can be observed in Figure S7, along with the
predicted concentration profiles and known HPLC concentrations comparisons of each analyte
(Figure S8). The resolved MCR-ALS spectrum of benzamide (P) was less successful than benzonitrile
2
(
SM), reflected in their R values of 0.7286 and 0.9514, correspondingly. This is likely due to the E.
coli background response coinciding with the benzamide product peaks within the region of 1200 –
-1
1
800 cm . Despite this the MCR-ALS modelled concentrations versus HPLC results (Figure 5) are
evidently in very good agreement with one another and the UVRR approach can successfully
quantify the whole-cell catalysed hydrolysis of benzonitrile to benzamide.
As with the pure enzyme reactions hard modelling was further applied to these results, however as
previous, the reaction experienced a reduction in the total concentration (assuming benzamide
insolubility) which had a substantial effect on MCR-ALS-HM. The final reaction timepoint (t = 18 min)
ACS Paragon Plus Environment

Page 11 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
had a 38 % mass loss when compared to the starting concentration, thus rendering the MCR-ALS-HM
results inaccurate (MCR-ALS-HM kinetic equations are shown in Figure S9).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5. MCR-ALS model applied to the whole-cell biotransformation of benzonitrile (SM, blue) to benzamide
(
P, red) using NHase-containing E. coli cells. Figure shows the reaction dynamics from the real-time
UVRR measurements (denoted by circular symbols) and off-line HPLC data (denoted by cross
symbols) as a function of time.
Next, the second step of the biotransformation (benzamide to benzoic acid) was monitored in the
presence of amidase-expressing E. coli cells. The increased expression of amidase in vivo, relative to
NHase, enables a much lower quantity of cells to catalyse the reaction (which we estimated to be
1:15), hence the E. coli background does not overtly dominate the UVRR response and characteristic
peaks are observed. Thus, the MCR-ALS soft modelling approach was applied to the UVRR and HPLC
2
data and were shown to be in very good agreement with one another (R = 0.9439 and 0.9862 for
benzamide and benzoic acid, Figure S10 and S11). Figure 6A illustrates the similarity between the
two methods. The total concentration remains stable throughout, so MCR-ALS hard modelling was
applied to the data, removing the requirement of comparative HPLC (MCR-ALS-HM kinetic models
shown in Figure S9). Results of the MCR-ALS-HM are shown in Figure 6B, where one can see that the
hard modelling results (represented by the solid lines) are highly comparable to the off-line HPLC
ACS Paragon Plus Environment

Analytical Chemistry
Page 12 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
measured concentrations (dashed lines), enabling an accurate measurement of the kinetic rate,
identified as k = 0.1986.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
k = 0.1986
Figure 6. Modelling results for amidase-expressing bacterial whole-cells catalysing the conversion of
benzamide (SM) to benzoic acid (P). (A) MCR-ALS soft modelling results and (B) MCR-ALS hard
modelling results, kinetic rate was calculated as k = 0.1986.
Finally we combined the two different E. coli together. As discussed above the relative expression
levels of NHase were very low in comparison to that of amidase, and this required a much larger
quantity of NHase-containing E. coli cells within the reaction mixture. We established using HPLC
(data not shown) that the 15:1 ratio of NHase:amidase enabled detectable concentrations of
benzamide before its subsequent hydrolysis to benzoic acid. Unfortunately, despite this optimisation
to see the intermediate, in vivo UVRR reaction monitoring in the presence of NHase + amidase (15:1
ratio) was unsuccessful due to the high density of E. coli cells. Figure S12a shows the corresponding
UVRR spectra over the reaction time course, one can easily identify the decreasing nitrile peak at
-
1
2
234 cm , indicative of benzonitrile (SM) consumption. However, the bacterial cells produce a UVRR
response which has peaks coinciding with those characteristic of benzamide and benzoic acid (1200
-
1
– 1800 cm , see Figure S12b) which meant that subsequent MCR-ALS modelling was unsuccessful.
Despite trying to include the biomass peak profile within the MCR-ALS model (as an additional
individual component) the deconvolution into four individual components (SM, I, P and biomass)
was still unsuccessful. We believe that if the NHase expression had been higher that we would have
been able to achieve an in vivo biotransformation using multiple organisms; however we are yet to
find a clone that has sufficiently high expression at this time.
ACS Paragon Plus Environment

Page 13 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
CONCLUSIONS
We have successfully demonstrated how UVRR can be used in a real-time, label-free and rapid (20 s)
manner to discriminate between nitrile-metabolising pathways (NHase + amidase versus nitrilase) by
observation of the amide intermediate. Combining on-line UVRR measurements with off-line HPLC
analysis, MCR-ALS modelling could be implemented to quantify each of the components accurately
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
within the reaction. High R values demonstrate that the two analytical techniques are in very good
agreement with one another. The similarity of the UVRR responses of the intermediate (benzamide)
and product (benzoic acid) does not appear to impact the model’s ability to distinguish them, further
indicating the sensitivity and selectivity of this combined UVRR + MCR-ALS approach. In contrast to
22
previous open pot reactions where products and substrates may evaporate, the development of a
flow-cell apparatus enables collection of UVRR measurements through a quartz flow-cell,
highlighting its suitability within PAT processes and facilitating measurements within a closed
system. To complement the application of a closed system further, hard modelling was combined
with MCR-ALS (MCR-ALS-HM) as to remove the requirement of external HPLC measurements, with
results shown to be in good agreement with off-line HPLC (validation). However, fluctuations in the
total concentration (arising from reduced solubility of the intermediate) reduced the accuracy of the
model as the mass of the system was no longer in equilibrium.
Finally, to extend our investigations and the applicability of this approach towards industrial
development, we introduced in vivo conditions by using E. coli whole-cells which expressed either
NHase or amidase activity. Omitting protein purification is attractive as it increases the overall
speed, enabling faster turnover of the desired product. Low-level expression of NHase in vivo and
the consequent requirement of a high density of bacterial cells led to unsuccessful detection and
discrimination of the individual analytes when performed in the presence of both NHase and
amidase bacterial cells. However, separating the two individual biotransformations proved to be
very successful. MCR-ALS soft modelling and hard modelling approaches could be applied to the
amidase-catalysed reaction, owing to its high-level expression (in vivo) and avoidance of solubility
interference. We are confident that these results showcase the sensitivity and accuracy of UVRR
spectroscopy for in situ, continuous measurements of enzyme activity, and its suitability for
implementation within industry.
ACS Paragon Plus Environment

Analytical Chemistry
Page 14 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
SUPPORTING INFORMATION
Details of: (i) supplementary methods; (ii) instrumentation and data processing, (iii) optical density
OD600) measurements; and (iv) protein expression and purification are provided in the SI. Provide
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
are the UVRR instrument set-up (Figure S1) and UVRR band assignments for analytes (Table S1). For
enzyme biocatalysis: the overall MCR-ALS processes (Figure S2) along with deconvolved spectra
(Figure S3) and results of modelling (Figure S4) are included as is the extension to MCR-ALS-HM
(Figure S5) with concentration constraints (Table S2) and the implementation of the kinetic
modelling (Figure S6). For whole-cell biotransformations: OD600 measurements are included (Table
S3), results of MCR-ALS deconvolution (Figure S7) and quantifications (Figure S8), the extension to
kinetic MCR-ALS-HM modeling is proposed (Figure S9), along with the results of deconvolution
(Figure S10), quantification (Figure S11) and UVRR spectra (Figure S12). Finally, details of cloning,
expression and protein purification protocols are provided (Figure S13-14).
ACKNOWLEDGMENTS
CW and HF are grateful to BBSRC for their PhD studentships. HF thanks Dr Jonathan Latham for
NHase insights, Dr Brian Law for his guidance about amidase cloning and expression and the Uwe
Bornscheuer group (Greifswald University) for the NHase construct. YX and RG thank the Wellcome
Trust for funding (grant 202952/Z/16/Z). NJT, JM and RG are also indebted to BBSRC and GSK for
financial funding (grant BB/K00199X/1). NJT thanks the Royal Society for a Wolfson Research Merit
Award.
REFERENCES
(1) Zhou, Z.; Hashimoto, Y.; Kobayashi, M. Actinomycetologica 2005, 19, 18-26.
(2) Asano, Y.; Yasuda, T.; Tani, Y.; Yamada, H. Agric. Biol. Chem. 1982, 46, 1183-1189.
(3) Mauger, J.; Nagasawa, T.; Yamada, H. J. Biotechnol. 1988, 8, 87-95.
(4) Nagasawa, T.; Mathew, C. D.; Mauger, J.; Yamada, H. Appl. Environ. Microbiol. 1988, 54, 1766-
1769.
(5) Kobayashi, M.; Nagasawa, T.; Yamada, H. Trends Biotechnol. 1992, 10, 402-408.
(6) Ghisalba, O.; Meyer, H.-P.; Wohlgemuth, R.; Flickinger, M. C. In Encyclopedia of Industrial
Biotechnology; John Wiley & Sons, Inc., 2009, pp 1-34.
(7) Gong, J.-S.; Lu, Z.-M.; Li, H.; Shi, J.-S.; Zhou, Z.-M.; Xu, Z.-H. Microb. Cell Fact. 2012, 11, 142.
ACS Paragon Plus Environment

Page 15 of 16
Analytical Chemistry
1
2
3
4
5
6
7
8
9
(8) Wang, C.-C.; Lee, C.-M.; Chen, L.-J. J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ.
Eng. 2004, 39, 1767-1779.
(9) Hinz, D. C. Anal. Bioanal. Chem. 2006, 384, 1036-1042.
(10) Undey, C.; Low, D.; Menezes, J. C.; Koch, M. PAT Applied in Biopharmaceutical Process
Development And Manufacturing: An Enabling Tool for Quality-by-Design; CRC Press, 2011.
(
(
11) Banerjee, A.; Kaul, P.; Sharma, R.; Banerjee, U. C. J. Biomol. Screening 2003, 8, 559-565.
12) Santoshkumar, M.; Nayak, A. S.; Anjaneya, O.; Karegoudar, T. B. J. Ind. Microbiol. Biotechnol.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
(
4
(
009, 37, 111.
13) Angelini, L. M. L.; da Silva, A. R. M.; Rocco, L. d. F. C.; Milagre, C. D. d. F. Braz. J. Microbiol. 2015
6, 113-116.
14) Black, G. W.; Brown, N. L.; Perry, J. J. B.; Randall, P. D.; Turnbull, G.; Zhang, M. Chem. Commun.
,
2015, 51, 2660-2662.
(15) He, Y.-C.; Ma, C.-L.; Xu, J.-H.; Zhou, L. Appl. Microbiol. Biotechnol. 2011, 89, 817-823.
(16) Lin, Z.-J.; Zheng, R.-C.; Lei, L.-H.; Zheng, Y.-G.; Shen, Y.-C. J. Microbiol. Methods 2011, 85, 214-
2
20.
17) Reisinger, C.; van Assema, F.; Schürmann, M.; Hussain, Z.; Remler, P.; Schwab, H. J. Mol. Catal. B:
Enzym. 2006, 39, 149-155.
(
(
(
(
(
(
18) Kumar, R. A.; Clark, D. S. Curr. Opin. Chem. Biol. 2006, 10, 162-168.
19) Acker, M. G.; Auld, D. S. Perspect. Sci. 2014, 1, 56-73.
20) Asher, S.; Johnson, C. Sci. 1984, 225, 311-313.
21) Asher, S. A. Anal. Chem. 1993, 65, 59A-66A.
22) Westley, C.; Fisk, H.; Xu, Y.; Hollywood, K. A.; Carnell, A. J.; Micklefield, J.; Turner, N. J.;
Goodacre, R. Chem. Eur. J. 2017, 23, 6983-6987.
23) Rzeznicka, K.; Schätzle, S.; Böttcher, D.; Klein, J.; Bornscheuer, U. T. Appl. Microbiol. Biotechnol.
(
2010, 85, 1417-1425.
(24) Tauler, R. Chemom. Intell. Lab. Syst. 1995, 30, 133-146.
(25) Jaumot, J.; Gargallo, R.; de Juan, A.; Tauler, R. Chemom. Intell. Lab. Syst. 2005, 76, 101-110.
(
(
26) Molloy, K. J.; Maeder, M.; Schumacher, M. M. Chemom. Intell. Lab. Syst. 1999, 46, 221-230.
27) de Carvalho, A. R.; Sánchez, M. d. N.; Wattoom, J.; Brereton, R. G. Talanta 2006, 68, 1190-1200.
ACS Paragon Plus Environment

Analytical Chemistry
Page 16 of 16
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
For TOC only
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment