Identification and application of threonine aldolase for synthesis of valuable α-amino, β-hydroxy-building blocks

Article abstract of DOI:10.1016/j.bbapap.2019.140323

Chiral β-hydroxy α-amino acid structural motifs are interesting and common synthons present in multiple APIs and drug candidates. To access these chiral building blocks either multistep chemical syntheses are required or the application of threonine aldolases, which catalyze aldol reactions between an aldehyde and glycine. Bioinformatics tools have been utilized to identify the gene encoding threonine aldolase from Vanrija humicola and subsequent preparation of its recombinant version from E. coli fermentation. We planned to implement this enzyme as a key step to access the synthesis of our target API. Beyond this specific application, the aldolase was purified, characterized and the substrate scope of this enzyme further investigated. A number of enzymatic reactions were scaled-up and the products recovered to assess the diastereoselectivity and scalability of this asymmetric synthetic approach towards β-hydroxy α-amino acid chiral building blocks.

Full text of DOI:10.1016/j.bbapap.2019.140323

Journal Pre-proof
Identification and application of threonine aldolase for synthesis
of valuable α-amino, β-hydroxy-building blocks
Mathieu Ligibel, Charles Moore, Robert Bruccoleri, Radka
Snajdrova
PII:
S1570-9639(19)30208-0
DOI:
https://doi.org/10.1016/j.bbapap.2019.140323
BBAPAP 140323
Reference:
To appear in:
BBA - Proteins and Proteomics
Received date:
Revised date:
Accepted date:
28 July 2019
31 October 2019
1 November 2019
Please cite this article as: M. Ligibel, C. Moore, R. Bruccoleri, et al., Identification and
application of threonine aldolase for synthesis of valuable α-amino, β-hydroxy-building
blocks,
BBA
-
Proteins
and
Proteomics(2019),
https://doi.org/10.1016/
j.bbapap.2019.140323
This is a PDF file of an article that has undergone enhancements after acceptance, such
as the addition of a cover page and metadata, and formatting for readability, but it is
not yet the definitive version of record. This version will undergo additional copyediting,
typesetting and review before it is published in its final form, but we are providing this
version to give early visibility of the article. Please note that, during the production
process, errors may be discovered which could affect the content, and all legal disclaimers
that apply to the journal pertain.
©
2019 Published by Elsevier.

Journal Pre-proof
Biochimica et Biophysica Acta - Proteins and Proteomics (BBA-Proteins and
Proteomics)
Identification and Application of Threonine Aldolase for Synthesis of
Valuable α-amino, β-hydroxy-Building Blocks
1
1
2
1,
Mathieu Ligibel, Charles Moore, Robert Bruccoleri, Radka Snajdrova *
1Global Discovery Chemistry, Novartis Institute for BioMedical Research, Basel,
Switzerland
2Congenomics, LLC, Glastonbury CT USA
*corresponding author: radka.snajdrova@novartis.com
Keywords: biocatalysis, threonine aldolase, enzymatic synthesis, amino acids, in-silico
gene mining
Abstract
Chiral β-hydroxy α-amino acid structural motifs are interesting and common synthons
present in multiple APIs and drug candidates. To access these chiral building blocks
either multistep chemical syntheses are required or the application of threonine aldolases,
which catalyze aldol reactions between an aldehyde and glycine. Bioinformatics tools
have been utilized to identify the gene encoding threonine aldolase from Vanrija humicola
and subsequent preparation of its recombinant version from E. coli fermentation. We
planned to implement this enzyme as a key step to access the synthesis of our target
API. Beyond this specific application, the aldolase was purified, characterized and the
substrate scope of this enzyme further investigated. A number of enzymatic reactions
were scaled-up and the products recovered to assess the diastereoselectivity and
scalability of this asymmetric synthetic approach towards β-hydroxy α-amino acid chiral
building blocks.
1
. Introduction
Biocatalysis is gaining ever more attention in the pharmaceutical industry, in part due
to recent break-through applications of diverse enzymes offering shorter, greener and
1
more atom- and cost-efficient routes to drug candidates or APIs. Industrial biocatalysis
is largely recognized for its power to interconvert functional groups, however bond

Journal Pre-proof
2
formation, especially C-C and C-N, remain under-represented. As mentioned in a recent
2
publication , aldolases are discrete enzymes capable of catalyzing C-C bond formation,
however, broader use of these biocatalysts has yet to become commonplace industrially.
Despite a handful of successful industrial scale processes applying aldolases, the limited
number of compatible reactants has prevented wide spread uptake of this technology.
Accelerated progress in enzyme engineering has further advanced aldolase application
3
4
by broadening the scope for acceptors and donors, increasing their stereospecificity
and improving their thermostability.5
In our case, threonine aldolases (TAs) were of particular interest because of their potential
to synthesize β-hydroxy α-amino acids in an enantio and diastereoselective manner. This
6
class of aldolases has been studied extensively by multiple academic groups. The
advantage of TAs is their inherent potential to generate two chiral centers in one step
starting from two prochiral substrates; inexpensive glycine and an aldehyde (Figure 1).
The resulting β-hydroxy, α-amino acid structural motif is found in many natural products
7
(
vancomycin, cyclosporin, polyoxin D) and represents an interesting synthon of many
8
9
pharmaceutical intermediates (e.g. DOPS , a peptidic mimetic of RNA , monocyclic β-
10
lactams ). In our case, we were looking to find an asymmetric route towards such a
synthon with trans configuration (target compound 1a). Despite the risk that
stereochemistry at the β-center might not be perfect, we were looking for initial starting
activity to form our target compound 1a. Such an enzyme could be further improved by
enzyme evolution if required. High-throughput screening and even selection to assist
enzyme engineering of PLP-dependent enzymes, such as TAs, has been described
previously.11
Unfortunately, during the course of our early work the target compound was deselected
from our internal program. However, we recognized the utility of the enzymatic work
carried out for this project and sought to develop it further for future applications.
Figure 1: retrosynthetic planning for synthesis of β-hydroxy α-amino acids using threonine aldolase towards target
compound 1a.
A number or reports have been published on using glycine-dependent aldolases in the
synthesis of such L-anti analogues; mainly by threonine aldolases (TAs) or serine
hydroxymethyltransfereases (SHMT).12 One such catalyst with good biochemical
characterization, L-TA from Vanrija humicola, was reported, purified and crystallized
13
almost 50 years ago, however no gene accession number or DNA sequence is known,
limiting its application.

Journal Pre-proof
Here we report on the identification of the gene encoding this aldolase from Vanrija
humicola by in-silico gene mining. Vanrija humicola was originally taxonomically
classified as Candida humicola and later changed to Cryptococcus humicola with its
current classification residing in the Genus Vanrija; making these organism names
homotypic. We describe the subsequent gene cloning, purification and application to the
synthesis of β-hydroxy, α-amino acids which are valuable building blocks in the
pharmaceutical industry. Publishing this gene gives other researchers the ability to tap
into the biocatalytic potential of this aldolase or use it as starting point for subsequent
enzyme evolution for a specific application.
2
. Materials and Methods
2
.1. General
All chemicals and solvents were purchased from commercial suppliers (Sigma-Aldrich,
VWR and Brenntag) and were used without further purification.
All NMR spectra were recorded on a Bruker 400 MHz Avance III spectrometer in
Methanol-d
4
or DMSO-d and are referenced to the residual solvent peak.
6
High-resolution mass spectra were recorded on an electrospray ionization quadrupole
time-of-flight (ESI-Q-TOF) MaXis system (Bruker Daltonics, Bremen, Germany) coupled
to an Agilent 1200 HPLC system (Agilent technology, California, USA). Analytical HPLC
was performed using an Agilent 1260 HPLC system with a photo-diode array UV/Vis
detector and a single quadrupole MSD system (Agilent technology, California, USA) or
UPLC Acquity (Waters, Massachusetts, USA).
2
.2. Gene mining
14
DNA genomic assembly was performed using MIRA. Gene calling was performed using
15
EXONERATE. Annotation of genes was performed using BLASTx taking the top hit to
16
explain the likely function of each gene. Codon optimization was performed using JCat.
All sequence visualization was conducted in Geneious (Biomatters Ltd. Auckland, NZ).
The genome assembly used for this work was: Vanrija humicola
JCM_1457_assembly_v001 (NCBI). Multiple sequence alignments were performed to
17
confirm BLAST results of identified genes and was performed with Clustal Omega.
2
.3. Gene cloning and expression
A gene encoding wild-type L-TA from E. coli was purchased from Genewiz (South
Plainfield, NJ, USA) in pET28a (NdeI and HindIII restriction enzyme cloning). The plasmid
was transformed into E. coli BL21(DE3) and the cells were cultivated on an LB agar plate
containing 50 μg/mL kanamycin. A pre-culture from a single colony of freshly transformed
cells was cultured in 20 mL of LB medium containing 50 μg/mL kanamycin at 37 °C. After
the culture reached OD600=0.5 it was induced by addition of 1 mM IPTG (final
concentration), and further cultivated for 3 hours. The cells were pelleted by centrifugation

Journal Pre-proof
at 4000 x g, 4 °C, for 10 min, and the supernatant was discarded. Protein expression was
confirmed by SDS-PAGE.
Five genes encoding for different variants of the L-TA from Vanrija humicola ATCC 20265
(
reclassified into the Genus Vanrija at a later date) were purchased from Genscript in
pET28a. NcoI and XhoI with a stop codon to the 3’ side of the XhoI site for all five variants
which are non-his tagged variants. Each plasmid was transformed into E. coli BL21(DE3)
and the cells were plated on a LB agar plate containing 50 μg/mL kanamycin. A pre-
culture from single colony of freshly transformed cells was cultured in 10mL of Terrific
Broth medium containing 50 μg/mL kanamycin at 37 °C. After the culture reached
OD600=0.4 expression was induced by the addition of 1 mM IPTG (final concentration).
The cultivation temperature was then reduced to 30 °C, and the culture was incubated
overnight. The cells were pelleted by centrifugation at 7000 x g, at 4 °C, for 10 min, and
the supernatant was discarded. Protein expression was determined by SDS-PAGE.
Proposed gene sequences of five potential aldolases in pET28a (NcoI and XhoI restriction
enzyme cloning) were purchased to avoid N- and C-terminal HisTag for initial expression
study). After the protein expression (as above) and activity study, the gene encoding L-
TA_1 was recloned with either N- or C-terminal HisTag for protein purification -NdeI and
XhoI restriction enzyme cloning was adopted for the N-his tagged Ald_1. NcoI and XhoI
restriction enzyme cloning was adopted for the C-his tagged Ald-1 variant.
2
.4.
Protein purification
From 1 L culture grown in LB medium 10 g of cell pellet was obtained and suspended at
°C in 250 mL of the lysis buffer (100 mM NaH PO , 300 mM NaCl and 10 mM
Imidazole, pH 8.0). The cell suspension was disrupted using a French press twice from
00 to 80 bar.
4
2
4
8
The suspension was then centrifuged at 10,000 x g for 45 min to obtain a clear
supernatant which was subsequently filtered through a 0.22 µm filter. Clarified lysate was
loaded onto a HisTrap exel column (GE) at 1 mL/min, previously equilibrated with buffer
containing 250 mM imidazole. Protein was eluted with 350 mL of buffer containing 250
mM imidazole buffer and washed with 500 mM imidazole buffer. The fractions containing
the aldolase were dialyzed at 4 °C against 2 L of PBS buffer (pH=7.3) with an YM-30
membrane. Protein concentration after dialysis was estimated as 2.9 mg/mL (at A280, f:
0
.37). Aldolase was diluted to a final concentration of 0.60 mg/mL with buffer (100 mM
phosphate buffer pH=7.0, 0.05 mM PLP and 25% glycerol), aliquoted into 0.5 mL portions,
o
frozen on dry ice and stored at -80 C in a freezer until further use.
Large scale purification from 4 L of culture has been carried out by a third party (BLIRT)
and purification method is described in supplementary information.

Journal Pre-proof
2
.5. Enzymatic assays on analytical scale
Calculation of conversions using UPLC- or HPLC-UV analysis:
The conversion was calculated as follows:
Conversion = TAC peak area of the target compound / (TAC UV peak area of the target
compound + TAC UV peak area of starting material)
TAC: Total Absorbance Chromatogram (210 – 450 nm)
The HPLC-MS estimated relative conversion was calculated as follows:
MS Estimated relative Conversion = MS peak area of the target compound with the
chosen reaction conditions / (MS peak area of the target compound in the reference
reaction conditions)
Pig liver serine hydroxymethyltransferase:
To a 2 mL Eppendorf tube was added 400 µL of phosphate buffer 100 mM (pH=7.0, 7.4
or 8.0), 5 µL of the aldehyde solution in a 0.1 M solution in DMSO (1 mM final
concentration), 5 µL of glycine (1 M solution, 10 equiv.) and 100 µL of liver S9
homogenate (protein concentration 11.3 mg/mL estimated by Bradford method in
18
triplicates). The tube was incubated overnight at 30 °C and 1100 rpm in an Eppendorf
Thermomixer. After 20 hours, the reaction was quenched with 5 mL ACN/MeOH (1:1),
stirred 5 min at 1100 rpm and centrifuged at 21,130 x g for 3 min. The clear supernatant
was analyzed using the UPLC monitoring method (see supplementary information).
L-TA from E. coli:
To a 2 mL Eppendorf tube was added 500 µL of the cell free extract (protocol described
in Supplementary information) in 200 mM Tris buffer (pH=7.5), 5 µL of the aldehyde
solution in a 0.1 M in DMSO stock (1 mM final concentration) and 5 µL of glycine 1 M
(
10 equiv.). The tube was incubated overnight at 37 °C and 1100 rpm on an Eppendorf
Thermomixer. The reaction was quenched with 500 µL ACN/MeOH (1:1), stirred for 5
min at 1100 rpm and centrifuged at 21,130 x g for 3 min. The clear supernatant was
analyzed using HPLC monitoring method (see supplementary information).
Crude lysate of Vanrija humicola:
To a 2 mL Eppendorf tube was added 500 µL of the cell free extract (protocol described
in Supplementary information) in the reaction buffer (phosphate buffer 100 mM, pH=7.0,
0
.05 mM PLP), 2.5 µL of the aldehyde solution from 0.1 M in DMSO stock (0.5 mM final
concentration) and 12.5 µL of glycine 1 M (50 equiv.). The tube was incubated overnight
at 30 °C and 1100 rpm on an Eppendorf Thermomixer. The reaction was quenched with
5
00 µL ACN/MeOH (1:1), stirred for 5 min at 1100 rpm and centrifuged at 21,130 x g for

Journal Pre-proof
3
min. The clear supernatant was analyzed using HPLC monitoring method (see
supplementary information).
2
.6. General procedure for screening aldehyde acceptors
To a 2 mL Eppendorf tube was added 10 µL of the aldehyde stock solution in DMSO
(
0.1 M , final concentration 10mM), 90 µL of buffer (100 mM phosphate buffer, 0.1 mM
PLP, 2 M glycine, pH=7.0) and 2 µL of a 10.16 mg/mL enzyme solution in phosphate
buffer (100 mM, 1 mM PLP, pH=7.0, 0.2 mg/mL final protein concentration). The mixture
was incubated at 40 °C and 1100 rpm on an Eppendorf Thermomixer. After 1.5 hours,
the reaction was quenched with 200 µL ACN/MeOH (1:1), stirred for 5 min at 1100 rpm
and centrifuged at 21,130 x g for 3 min. The clear supernatant was analyzed using an
UPLC monitoring method (see supplementary information).
The reaction with trifluoroacetone 13 was monitored using LC/CAD/MS (see
supplementary information).
2
.7. Preparative synthesis of the Cbz derivatized β-hydroxy α-amino acids
To a 15-50 mL conical centrifuge tube was added 1.5-15 mL of the reaction buffer (100
mM phosphate buffer, 0.1 mM PLP, 2 M glycine, pH=7.0) and 166-1570 µL of the
aldehyde 1 M solution in DMSO (0.166-1.570 mmol, final concentration 100 mM). The
solution was vortexed at RT. In the case that the aldehyde precipitated out of solution,
another equivalent volume (1.5-15 mL) of the reaction buffer was added and the solution
was once more vortexed at RT and/or dissolved with ultrasound (final aldehyde
concentration 50 mM). L-TA was then added (16-320 µL of a 10.16 mg/mL stock solution
in phosphate buffer (100 mM, 1 mM PLP, pH=7.0, to 0.2 mg/mL final concentration)). The
reaction mixture was incubated at 40 °C, 750 rpm in a Eppendorf Thermomixer. After 0.5-
1
7 hours, a conversion in range of 32 % to higher than 90 % was reached. The reaction
mixture was quenched by the addition of one reaction volume of ethyl acetate under
vigorous stirring over 2-5 minutes. The aqueous layer was concentrated on a rotary
evaporator (40 °C, 10 mbar), filtered and directly injected on the reverse-phase column
of preparative HPLC.
Each fraction was monitored using an UPLC monitoring method. Fractions containing
pure product were collected and lyophilized.
To 1-3 mL of a saturated NaHCO
magnetic stirrer was added 20-288 mg of the α-amino, β-hydroxy-building block and THF
-3 mL. The reaction mixture was stirred at RT and 1 equiv. of Cbz-Cl was added
3
aqueous solution in a glass vial equipped with a
1
dropwise and continued with stirring at room temperature during next 0.5-2 hours. After
this period, a conversion higher than 90 % was observed.

Journal Pre-proof
The crude sample was filtered and purified with the corresponding HPLC preparative
method.
For both steps, each fraction was monitored using an UPLC monitoring method.
The overall yield after two steps was 6-37 %
3
. Results and Discussion
3
.1. Biocatalyst identification
When looking for a potential biocatalyst to access our target compound, we considered,
among other enzymatic approaches, an aldolase to perform the C-C bond formation. This
strategy would greatly improve process efficiency by generating two chiral centers in one
step. Furthermore, this was a particularly appealing approach since an aldolase would
utilize two achiral precursors; the corresponding aldehyde and glycine. In addition, glycine
is an inexpensive commodity chemical, which makes this approach cost-attractive.
Our search for an enzyme catalyzing the aldol reaction initiated with the screening of
commercially available aldolase kit (threonine aldolase kit, Almac). However, none of the
commercial enzymes showed activity towards our target compound (1a). Disappointed
by this finding, we looked to the literature and searched for structurally similar compounds
prepared by known aldolases, microbial strains or other biocatalysts. This revealed three
aldolase substrate compounds below and their corresponding biocatalytic agents (Figure
). Compound 2a and 3a were prepared by Vanrija humicola AKU45861 and product
9,20
2
4
2
1,22
a by L-TA from E. coli and serine hydroxylmethyltransferase, respectively.
Figure 2: Structurally similar compounds to target building block 1a identified from literature search
Encouraged by these findings, we proceeded to test the biocatalysts against target
compound 1a and the relevant literature substrate as positive control.
Testing of serine hydroxymethyltransferase from pig liver was carried out in the form of a
liver homogenate. In the case of a positive hit, a recombinant form would be prepared
since three gene sequences were available. Unfortunately, no activity was observed
towards formation of product 1a and this approach was discontinued.

Journal Pre-proof
The gene encoding L-threonine aldolase (L-TA) from E. coli was DNA synthesized and
cloned into an inducible expression vector for protein expression from E. coli BL21 (DE3).
Clear bands with the expected size of L-TA (~45 kDa) were observed in the soluble
protein fraction (see supplementary information). Unfortunately, we have not obtained
any clear indication based on the activity data that this approach would be successful and
so terminated any further effort at this point.
After we exhausted the two options immediately available to us we purchased the yeast
23
strain of Vanrija humicola AKU4586 from ATCC (referenced as ATCC 20265, ). Upon
receipt of the strain it was cultivated according to a patent publication wherein L-threonine
24
was added to the minimal medium as the sole carbon source. An activity assay with
whole cells was performed, however, no formation of 1a was detected, nor was activity
against its natural substrate, 2a, observed. In parallel, we prepared cell-free extract to
test if the absence of activity could be due to poor transport of either the aldehyde or
glycine into the yeast cells. This appeared to be the case, since cell-free extract gave
clear conversion to the target product as evidenced by LC/MS.
In order to achieve higher production of enzyme and subsequently increased yield to
target product 1a, and to make this approach scalable, it became clear that recombinantly
expressed L-TA would be required. Since no DNA or protein sequence of the
corresponding L-TA from Vanrija humicola has been published to date, we initiated
identification of the the L-TA gene from its genome sequence.
3
.2. In-silico mining for L-TA
Since we knew that the aldolase of interest was present in Cryptococcus humicola we
searched in the genome of Vanrija humicola ATCC 20265; the final taxonomic name of
this organism that had undergone two Genus level reclassifications. This genome entry
was found as an unannotated draft quality genome. We were then challenged with
identifying the coding exons in an unannotated genome with uncharacterized intron
boundaries. By using protein sequences for aldolases which have a PLP binding domain,
we used the Exonerate program to identify the genomic regions where the translated
genomic regions match well and which have expected intron donor and acceptor sites.
The search of the Vanrija humicola genome using the 41 aldolase sequences only found
two genomic regions with high scoring alignments of putative coding regions. We
14
reassembled the genomic reads into a higher quality genome using MIRA and
15
annotated the ORFs with the gene caller Exonorate . Next we applied BLASTx to each
ORF using the threonine aldolase from Cryptococcus neoformans var. neoformans
JEC21 as the query (see supplementary information). Two potential ORFs for the
threonine aldolase were identified within these data with high sequence identity to
threonine aldolases from phylogenetically related organisms (see supplementary
information). Inspection of these ORFs revealed four possible start codons for one
example (L-TA_2.1 through L-TA_2.4 are derived from the same gene with different
possible start codons) and a singular likely start codon for the second example (L-TA_1).

Journal Pre-proof
Each of these potential full length ORFs was codon optimized for E. coli BL21 using
16
JCat , which returned an optimized gene with a codon adaptability index of 1.0 in all
cases, and ordered the genes as synthetic DNA cloned into pET28a using the NcoI and
XhoI restriction sites. This approach provided LacI mediated induction of the cloned gene.
3
.3. Expression and activity assays of identified enzyme variants
After receiving the 5 plasmids, the transformation with E. coli BL21 (DE3) strain was
carried out and initial enzyme expression tested, initially in its native non-tagged version.
Most of the constructs were expressed in a soluble form with the expected protein size
based on SDS-PAGE (see supporting information). Regardless of expression quantity,
all five enzyme variants were submitted to activity assays in the form of crude lysate. To
our delight, all five of them showed activity at various levels. The most active variant was
L-TA_1 whose DNA sequence was predicted with highest probability based on BLASTx
results. The enzymes from expression of the second gene with various length at the N-
terminus, were generally less active (approximately one third of activity, data not shown).
No dramatic difference in activity was observed in relation to various length of N-terminal.
To characterize the L-TA_1 enzyme better and to probe its substrate scope beyond the
target compound (product 1a), protein purification was necessary to assist with accurate
analytical detection and also to facilitate efficient recovery of the product.
3
.4. Purification of L-TA_1
In order to simplify purification of the enzyme, the gene with His-tag on either its N- or C-
terminus was recloned. Initial expression trials of both tagged enzyme variants indicated
good expression of soluble protein. Interestingly, the C-terminal His-tagged variant (C-
His L-TA_1) gave slightly higher conversion compared to its untagged version and also
higher activity in comparison to its N-terminal His-tag counterpart (1). Both tagged variants
were cultivated on larger scale (1 L culture) for protein purification purposes.
Since the initial enzyme expression trials showed a good level of soluble expression, no
significant optimization was conducted prior to larger scale enzyme purification, with the
exception of a slight modification of the cultivation medium and we directly proceeded
using the initial parameters (see supplementary information). Protein purification on Ni-
NTA and a subsequent dialysis afforded 72.8 mg of N-His L-TA_1 and 18.5 mg of C-His
L-TA_1 as pure protein. The size of the protein was also confirmed by MS (see
supplementary information). Based on this successful purification outcome, subsequent
purification was outsourced to third party (BLIRT) using their slightly modified protocol
(
see supplementary information).
Biocatalyst
Relative conversion [a]

Journal Pre-proof
L-TA_1 CFE
1
C-His L-TA_1 CFE 1.10
N-His L-TA_1 CFE 0.83
[
a]
Conversion after 3 hours, based on the MS Peak
intensity. CFE – cell free extract
Table 1: Conversion to 1a with N- or C-terminal His Tag or no His tag version of CFE containing L-TA_1 after 3
hours.
3
.5.
Substrate scope of L-TA_1
Having a pure enzyme in hand, we next examined the substrate scope of the enzyme.
Specifically, substrates leading to desirable building blocks or reflecting, to some extent,
intermediates of interest to a pharmaceutical portfolio. The search for suitable aldehydes
was conducted based on β-hydroxy α-amino acids or acid derivative motifs. In this case,
neither a comprehensive or systematic substrate finger printing was conducted, however,
we examined opportunities where the aldolase enzymatic step fit in pharmaceutically
relevant chemical space. The focus was on identifying situations that could impact the
overall synthetic strategy and perhaps offer a shorter route to a drug or drug candidate.
The second criteria was to have an aldehyde substrate with a chromophore for easy
reaction monitoring, unlike in the case of target compound 1a, where due to polarity and
absence of a chromophore, analytical monitoring proved to be challenging or required
derivatization of an amino group.
Based on these two criteria, approximately 30 aldehydes were selected for testing,
including the literature known substrate 10 as a positive control. In addition to the set of
aldehydes, two ketones were chosen to probe if the aldolase could give access to α-
amino acids with tertiary alcohols in β-positions, which are otherwise difficult to obtain
stereoselectively using classical chemical methods, but would open a door to many novel
applications of aldolases.
Trifluroacetone 13 was selected as one of the ketones due to its high reactivity. Indeed,
in this case, formation of a new product with the expected mass was observed with
LC/CAD/MS as it was not possible to follow the reaction using our standard LC/MS
method. To confirm that the new product formed was indeed the expected aldol product,
an authentic sample of 13a was spiked into the LC sample. This procedure revealed that
the new product observed in LC/CAD/MS was not the anticipated product 13a even
though it had the same mass as the target compound. It would be difficult to prove the
identity of the product from the enzymatic reaction since the conversion was very low and
further investigation was deemed unwarranted.
In summary, this methodology has been shown to be beneficial towards accessing a
range of molecules relevant to our discovery program. There is clearly a dependence on
conversion to the steric and electronic environment of the aldehyde and it is our intention
to further study this system to develop conditions that provide a more robust reaction with

Journal Pre-proof
broader substrate scope. For now however, the success generated has clearly
demonstrated the potential that aldolases have for accessing complex, ‘real world’
molecules in a medicinal chemistry setting.
Accepted substrates
No conversion
Substrate
[
a]
Conv. [a]
~70%
Substrate
Conv.
Substrate
1.4%
5
1
4
5
24
3
0%
3%
~80%
~50%
6
1
25
26
4
1
6
7
7
8
25%
N.M.
1
27
87%
~90%
1
8
9
9
2
8
9
>
95%
~50%
1
2
10

Journal Pre-proof
9
5
0
1%
N.M.
N.M.
~10%
1
1
2
20
3
0
%
2
1
2
3
1
2
1
.7%*
2
3
~
10%
2
3
3
33
N.M.*
1
[
a]
based on peak area in UPLC-UV. In some cases peaks were not fully resolved and in these cases
the conversion is prefixed by approximation sign ~
Reaction monitored by LC/CAD/MS, ketone could not be detected using this method, only formation
*
of new product.
N.M. not measured because analyte peak co-elutes with other chromatographic artefacts
Table 2: Substrates tested with L-TA_1. Conversion estimated after 1.5 hours of reaction by UPLC-UV.
It was previously demonstrated by many groups that selectivity for the β-center is usually
very low for threonine-dependent aldolase, however, this situation can be improved by
4
enzyme engineering. In our case, an initial hint of activity in the screening stage, would
be considered as sufficient encouragement to progress, with activity and selectivity being
improved later. For similar reasons, negative data (column III) are also reported to provide
an overview and demonstrate where the scope of this aldolase ends with the wild type
enzyme, but which could be perhaps expanded during further evolution programs.
3
.6. Scale up and isolation of novel β-hydroxy α-amino acids
Based on the screening data, the reactions with the highest conversion (based on
chromatography) were scaled up for isolation of novel products and determination of the
diastereoselectivity of the enzyme. The well-known downside of this enzymatic reaction

Journal Pre-proof
is the necessity to add an excess of glycine to push the reaction towards completion over
short reaction times. This equilibrium driving solution is useful on an analytical scale, but
complicates the work up procedure on preparative scale. This is especially true if a crude
enzyme preparation is used as the isolation of aldol product from reaction mixture is then
even more complex due to the presence of additional components resulting from the
microbial lysate. To make the preparative synthesis cost effective, a cell free extract was
chosen as the biocatalyst formulation in the first instance. We observed a very quick
reaction with phtalimidoacetaldehyde 10, unfortunately the reverse reaction proceeded
rapidly. For this reason, the reaction needed to be closely monitored by LC/MS, to
determine the stop point of the reaction and hence minimizing the reverse reaction, when
using crude lysate. Interestingly, when using purified protein, less of the reverse reaction
was observed. Full conversion with purified protein has been reached in less than 20
minutes with concentration of substrate up to 0.95 g/l (5 mM) and no reverse reaction was
observed following additional 40 minutes of incubation.
Time points
Conversion with CFE
with L-TA_1 (%)
Conversion with pure L-
TA_1 at 0.2mg/mL (%)
[a]
[b]
2
4
6
0min
0min
0min
88
82
69
100
100
100
[
a]
based on the peak area in UPLC-UV. Due to difficult chromatography
monitoring, conversions are only estimates.
[
a]
based on peak area in UPLC-UV.
Table 3: Time course of the aldol reaction with pthalimidoacetaldehyde and glycine catalysed with CFE of L-TA_1
and purified L-TA_1
A second complication occurring in the reaction using cell-free extract was noticed during
structure confirmation of the aldol product 10a. Despite a high degree of purity according
to chromatography under normal and reverse phase modes, a new product has been
observed during NMR measurement. A mixture of byproduct (10d) and the expected aldol
1
product (10a) was detected (byproduct is proposed to be diol base on H-NMR and
LC/MS analysis, Figure 3). Due to their very similar physico-chemical properties, the
separation of both compounds proved elusive in our hands. The ratio between the target
product and byproduct was determined as 38:62 (byproduct:amino alcohol) by NMR and
4
0:60 by chiral HPLC (byproduct:amino alcohol).
In the negative control, when using lysate prepared from the E. coli BL21 (DE3) strain
which did not bear any plasmid for L-TA enzyme, no activity toward substrate 10 has
been observed after even prolonged incubation time to 16.5 hours (Table 4). It is
unclear how the diol was formed, since this unusual byproduct has not been observed

Journal Pre-proof
when using purified enzyme and the host background does not seem to contribute to
this observed phenomenon
Figure 3: Aldol reaction between pthalimidoacetaldehyde and glycine catalysed by CFE of L-TA_1. Identity of by-
product-10d in the reaction mixture is proposed to be diol.
Conditions
.2 mg/mL L-
TA_1
0 mg/mL of E.
Conversion (%) [a]
>95
0
1
0
coli BL21 (DE3)
CFE
Buffer only
0
[
a]
based on peak area in UPLC-UV
Table 4: Negative control experiment under screening conditions (10 mM aldehyde 10, 2 M glycine, 40 °C)
Because of these two reasons, reverse reaction and byproduct formation, preparative
reactions were carried out using purified enzyme. Prior to the actual preparative synthesis
starting from aldehyde 9, 10, 11, 16, 18 and 19, reaction conditions (temperature, time,
substrate and enzyme concentration) and the work up procedures were optimized.
o
The final conditions applied were; 30 or 40 C, 50-100 mM substrate concentration (based
on substrate solubility in reaction buffer), 0.14-0.27 mg/mL purified aldolase concentration
in 100 mM phosphate buffer (pH=7.0) supplemented with 0.1 mM PLP and 2 M glycine.
The highest conversion was observed in less than one hour in most cases. In the case,
when conversion stalled, additional and an identical portion of enzyme has been added
to boost product formation.
For isolation and purification of the β-hydroxy α-amino acids, initial extraction with ethyl
acetate helped to remove unreacted aldehyde. The crude product, in the aqueous layer,
was purified by preparative HPLC with reverse phase and fractions containing the product
were combined and concentrated. The pure product was then derivatized to the CBz-
protected aldol product for easier determination of diastereoisomeric excess values by
either HPLC or NMR, except for 16a, which decomposed during derivatization. The low
isolated yields in some cases were attributed to losses in the work up procedure (6).
Shortly after finalizing our scale up experiments, an elegant method for an easier work up

Journal Pre-proof
6b
was published. Here a second enzyme, glycine oxidase from Bacillius subtilis, has been
used to degrade the excess of glycine to glyoxylate after the reaction is completed and
thus enabled higher isolated yields of aldol products. This approach will be very
interesting to test in our future application.
In conclusion, threonine aldolases represent an appealing alternative to chemical
synthesis of β-hydroxy α-amino acid chiral building blocks. Modern bioinformatic methods
help to identify genes of novel or putative aldolases, which can be further improved by
enzyme engineering. Thanks to their relatively broad substrate scope, these enzymes
can serve as an ideal starting point for evolution in development of a particular chemical
process and subsequent tailoring to a specific need (enhancement of activity, stability or
enantioselectivity).
.
Entry Aldehyde
1
Product
Conversion [a]
87%
Reaction
conditions
Overnight,
Isolated
yield
72% (227
mg)
d.r.[b]
[
1] [2]
2.2
[
3] [4]
40°C,
100mM
substrate,
.2mg/mL
0
9
0
9a
enzyme
2.5 hours,
2
3
>90 %
84 %
8 % (29.5
mg)
2.0 [2]
40°C,
50mM
substrate,
0.14mg/mL
enzyme
1
10a
[
1] [2]
3 hours,
58 % (97.4 1.5
mg)
[
3]
30°C,
50mM
1.6 [4]
substrate,
.2mg/mL
0
enzyme
1
1
6
11a
4
5
68 %
80%
1.5 hours,
22 % (11.1 1.0 [1]
mg)
40°C,
27mM
substrate,
.27mg/mL
0
1
enzyme
1
6a
8a
1 hour,
66% (288
mg)
1.9 [1]
40°C,
100mM
substrate,
.2mg/mL
enzyme
0
18
1

Journal Pre-proof
6
>70%
1.5 hours,
37% (159
mg)
2.1[1] [4]
40°C,
100mM
1
9
substrate,
.2mg/mL
enzyme
0
19a
[
[
a]
b]
Conversion based on peak area in UPLC-UV.
[1]
[2]
[3]
Diastereoisomeric ratio determined by H-NMR , by chiral-HPLC after Cbz derivatization , by H-NMR after Cbz derivatization or by
[
4]
UPLC-UV after Cbz derivatization
Table 5: Preparative scale aldol reaction catalyzed by purified L-TA_1 and 2 M glycine
Acknowledgment:
The authors would like to thank Liliane Porchet-Zemp for expression trials and SDS-
PAGE analysis and David Plattner for developing method for purification of enzymes
and supplying initial quantities of enzymes. BLIRT S.A. provided assistance on the
purification of enzymes based on in-house protocol, which is gratefully acknowledged.
Thanks to Finton Sirockin for his assistance on chemoinformatics analysis. We are
also very thankful to our colleagues supporting us with analytical (Chiral separation:
Dan Huynh, Christophe Bury, Marie-Anne Lozach. Achiral separation: Céline
Patissier) and spectroscopic methods (NMR: Thomas Lochmann, Emine Sager,
Alexandre Luneau, Harald Schroeder. HRMS: Etienne Richard. LC/CAD/MS: Nicolas
Lambert, Etienne Richard) and X-ray crystallography (Piechon Philippe)
Supporting information
Can be found as a separate word file
References:

Journal Pre-proof
Highlights for review
We would like to thank the reviewers for their thorough reading of our manuscript! We believe that thanks
to their dedication, comments and suggestion, we could improve the manuscript’s quality to the benefit or
the readers.

Journal Pre-proof
Biochimica et Biophysica Acta - Proteins and Proteomics (BBA-Proteins and
Proteomics)
Identification and Application of Threonine Aldolase for Synthesis of
Valuable α-amino, β-hydroxy-Building Blocks
1
1
2
1,
Mathieu Ligibel, Charles Moore, Robert Bruccoleri, Radka Snajdrova *
1Global Discovery Chemistry, Novartis Institute for BioMedical Research, Basel,
Switzerland
2Congenomics, LLC, Glastonbury CT USA
*corresponding author: radka.snajdrova@novartis.com
Conflict of interest
All authors declare no conflict of interest
[
1] a. A. Crameri, M.L. Hill, S. L. Lovelock, M. Schober, D.G. Tew, P. J. Thomas, NOVEL PROCESSES
FOR THE PRODUCTION OF OLIGONUCLEOTIDES, US 20180023122 A1. b. F.K. Kleinbeck-Riniker, B.
Martin, G. Penn, F. Venturoni, T. Schlama, T. Ruch, T. Allmendinger, B. Wietfeld, P. Filipponi, NEW
PROCESS AND INTERMEDIATES, WO2017098430. c. C. Plass, A. Hinzmann, M. Terhorst, W. Brauer,
K. Oike, H. Yavuzer, Y. Asano, A. J. Vorholt, T. Betke, H. Groeger. Approaching Bulk Chemical Nitriles
from Alkenes: A Hydrogen Cyanide-Free Approach through a Combination of Hydroformylation and
Biocatalysis. ACS Catal. 96 (2019) 5198-5203.
[
2] J.P. Adams, M.J.B Brown, A. Diaz-Rodriguez, R.C. Loyd, G.-D. Roiban, Biocatalysis: a pharma
perspective. Adv. Synth. Catal. 361 (2019), 2421-2432.
3] a. R. Roldán, I. Sanchez-Moreno, T. Scheidt, V. Hélaine, M. Lemaire, T. Parella, P. Clapés, W. D.
[
Fessner, C. Guérard-Hélaine, Breaking the Dogma of Aldolase Specificity: Simple Aliphatic Ketones and
Aldehydes are Nucleophiles for Fructose‐6‐phosphate Aldolase, Chem. Eur. J. 2017, 23, 5005–5009; b.
D. Güclü, A. Szekrenyi, X. Garrabou, M. Kickstein, S. Junker, P. Clapés, W.-D. Fessner, Minimalist
Protein Engineering of an Aldolase Provokes Unprecedented Substrate Promiscuity ACS Catal. 6 (2016)
1848–1852;c. R. Roldán, K. Hernandez, J. Joglar, J. Bujons, T. Parella, I. Sánchez-Moreno, V. Hélaine,
M. Lemaire, C. Guérard-Hélaine, W.-D. Fessner, Biocatalytic Aldol Addition of Simple Aliphatic
Nucleophiles to Hydroxyaldehydes, ACS Catal. 8 (2018) 8804–8809. d. S. Junker, R. Roldan, H. J.
Joosten, P. Clapés, W. D. Fessner, Complete Switch of Reaction Specificity of an Aldolase by Directed
Evolution In Vitro: Synthesis of Generic Aliphatic Aldol Products. Angew. Chem. 130 (2018) 10310–
1
[
0314.
4] H. J. Gwon, S.H. Balik, Diastereoselective synthesis of L: -threo-3,4-dihydroxyphenylserine by low-
specific L-threonine aldolase mutants, Biotechnol Lett 32 (2010) 143-149.
[
5] L. Wieteska, M. Ionov, J. Szemraja, C. Fellerc, A. Kolinskid, D. Gront, Improving thermal stability of
thermophilic l-threonine aldolase from Thermotoga maritima, J. Biotechnol 199 (2015) 69-76.

Journal Pre-proof
[
6] a. S.L. Goldberg, A. Goswami, Z.Guo, Y. Chan, E.T. Lo, A. Lee, V. C. Truc, K. J. Natalie, C. Hang, L.
T. Rossano, M. A. Schmidt. Preparation of β-hydroxy α-amino acids using recombinant D-Threonine
aldolase. b. S. F. Beaudoin, M. P. Hanna, I. Ghiviriga, J.D. Stewart. Progress in using threonine aldolase
for preparative synthesis. Enzyme and Microbial Technology 119 (2018) 1-9. c. K. Baer, N. Dueckers, T.
Rosenbaum, C. Leggewie, S. Simon, M. Krausser, S. Osswald, W. Hummel, H. Groeger. A study towards
efficient L-threonine aldolase-catalyzed enantio- and diastereoselective aldol reations of glycine with
substituted benzaldehydes: biocatalyst production and process development. Tetrahedron Asym. 22
(2011) 925-928. d. N. Dueckers, K. Baer, S. Simon, H. Groeger, W. Hummel, Threoning aldolases-
screening, properties and applications in the synthesis of non-proteinogenic β-hydroxy α-amino acids,
Appl. Microbiol Biotechnol 88 (2010) 409-424.
[
(
7] a. G. Yim, M.N. Thaker, K. Koteva, G. Wright, Glycopeptide antibiotic biosynthesis. J. Antibiot. 67
2014) 31-41. b. S. Suzuki, K. isono, J. Nagatsu, T. Mizutami, Y. Kawashim, T. Mizuno, A new antibiotic
pyloxin A. J. Antibiot. 18 (1965) 131. c. M.C. Pirrung, D.S. Nunn, A.T. McPhail, R.E. Mitchell, Synthesis
and absolute configuration of hydroxythreonine, a biosynthetic precursor of rhizobitoxine in
Pseudomonas, Bioorg. Med. Chem. Lett. 3 (1993) 2095-2098.
[
8] S.H. Baik, H. Yoshioka, H. Yukawa, S. Harayama, Synthesis of L-threo-3,4-dihydroxyphenylserine (L-
threo-dops with thermostabilised low-specific L-threonine aldolase from Streptomyces coelicolor A3(2), J.
Microbiol. Biotechnol. 17 (2007) 721-727.
[
9] T. Miura, M. Fujii, K. Shingu, I. Koshimizu, J. Naganoma, T. Kajimoto, Y. Ida, Application of L-
Threonine Aldolase-Catalyzed Reaction for the Preparation of a Peptidic Mimetic of RNA: A Leading
Compound of Vero-toxin Inhibitors, Tetrahedron Lett 39 (1998) 7313-7316.
[
10] V. P. Vassilev, T. Uchiyama, T. Kajimoto, C.-H. Wong, An Efficient Chemo-Enzymatic Synthesis of α-
Amino-β-Hydroxy- γ-Butyrolactone, Tetrahedron Lett 36 (1995) 5063-5064.
11] L. Giger, M.D. Toscano, M. Bouzon, P. Marliere, D. Hilvert. A novel genetic selection system for PLP-
dependent threonine aldolases. Tetrahedron 68 (2012) 7549-7557.
12] Review and references therein: K. Fesko, M. Gruber-Khadjawi, Biocatalytic Methods for C-C Bond
Formation, ChemCatChem 5 (2013) 1248-1272.
13] a. H. Yamada, H. Kumagai, T. Nagate, H. Yoshida, Formation of threonine aldolase by bacteria and
[
[
[
yeasts, Agric. Biol. Chem. 35 (1971) 1340-1345. b. H. Kumagai, H. Yoshida, H. Yamada, T. Nagate,
Threonine aldolase from Candida humicola 2. Purificaton, crystalization and properties, Biochim Biophys
Acta 258 (1972) 779-790.
[
14] B. Chevreux, T. Pfisterer, S. Suhai, Automatic assembly and editing of genomic sequences, in S.
Suhai (Eds), Genomics and proteomics—functional and computational aspects, Kluwer
Academic/Plenum Publishers, New York. 2000, pp. 51–65.
[
15] G.S.C. Slater, E. Birney, Automated generation of heuristics for biological sequence comparison.
BMC Bioinformatics 6 (2005) 31.
16] A. Grote, K. Hiller, M. Scheer, R. Munch, B. Nortemann, D.C. Hempel, D. Jahn, JCat: a novel tool to
adapt codon usage of a target gene to its potential expression host. Nucleic Acids Res. 33 (2005) 526-
31.
17] F. Sievers, A. Wilm, D. Dineen, T.J. Gibson, K. Karplus, W. Li, R. Lopez, H. McWilliam, M. Remmert,
[
5
[
J. Söding, J.D. Thompson D.G. Higgins, Fast, scalable generation of high-quality protein multiple
sequence alignments using Clustal Omega. Mol. Syst. Biol. 7 (2011) 539.
[
18] Preparation of liver homogenate: The liver S9 fractions were prepared as described in Kittelmann M.
et al. (2003). M. Kittelmann, U. Reinegger, A. Espigat, L. Oberer, R. Aichholz, E. Francotte, O. Ghisalba,
Preparative enzymatic synthesis of the acylglucuronide of mycophenolic acid., Adv. Synth. Catal., 345
(
[
2003)s 825-829.
19] T. Tanaka, C. Tsuda, T. Miura, T. Inazu, S. Tsuji, S. Nishihara, M. Hisamatsu, T. Kajimoto.Design
and synthesis of peptide mimetics of GDP-fucose: targeting inhibitors of fucosyltransferases. Synlett 2
2004) 243-246.
20] T. Tanaka, M. Ozawa, T. Miura, T. Inazu, S. Tsuji, M. Hisamatsu, T. Kajimoto. Synthesis of novel
mimetics of CMP-sialic acid as the Inhibitors of sialyltransferases. Synlett 9 (2002) 1487-1490.
21] A. Saeed, D. W. Young. Synthesis of L- β-hydroxyaminoacids using serine hydroxymethytransferase.
Tetrahedron 48 (1992) 2507-2514.
22] T.Kimura, V. P. Vassilev, G.-J. Shenn, C.-H. Wong. Enzymatic synthesis of β-hydroxy-α-amino acids
based on recombinant D- and L-Threonine Aldolases. J. Am. Chem. Soc. 119 (1997) 11734-11742.
23] H. Yamade, Process for preparing L-serine. US 3755081 A patent, 1971.
(
[
[
[
[

Journal Pre-proof
[
24] H. N. Kawasaki, H. E..; Zushi, K. K. Kawasaki, S. Okumura, Biological method of producing serine
and serinol derivatives., S. O. US 3871958 patent, 1975.

Figure 1

Figure 2

Figure 3