Development of a practical biocatalytic process for (R)-2-methylpentanol

Article abstract of DOI:10.1021/op9002246

(R)-2-Methylpentanol is an important chiral intermediate for the synthesis of certain medicinally important compounds, natural products, and liquid crystals. Here we describe the development of a practical kinetic resolution utilizing an enantiospecific bio - catalytic reduction of racemic 2-methylvaleraldehyde. The process utilizes an evolved ketoreductase enzyme to selectively reduce the (R)-enantiomer of racemic 2-methylvaleraldehyde to the desired product with high volumetric productivity. A scaleable method for separating the desired product from the off-enantiomer of the starting material is also described. The process is cost-effective, green, and amenable to manufacturing scale.

Full text of DOI:10.1021/op9002246

Organic Process Research & Development 2010, 14, 119–126
Development of a Practical Biocatalytic Process for (R)-2-Methylpentanol
Owen W. Gooding,* Rama Voladri, Abigail Bautista, Thutam Hopkins, Gjalt Huisman, Stephan Jenne, Steven Ma,
Emily C. Mundorff, and Megan M. Savile
Codexis, Inc., 200 Penobscot DriVe, Redwood City, California 94063, U.S.A.
Susan J. Truesdell and John W. Wong
Biocatalysis Center of Emphasis, Chemical R&D, Pfizer Global Research and DeVelopment, Eastern Point Road,
Groton, Connecticut 06340, U.S.A.
Abstract:
R)-2-Methylpentanol is an important chiral intermediate for the
synthesis of certain medicinally important compounds, natural
products, and liquid crystals. Here we describe the development
of a practical kinetic resolution utilizing an enantiospecific bio-
catalytic reduction of racemic 2-methylvaleraldehyde. The process
utilizes an evolved ketoreductase enzyme to selectively reduce the
reductive cleavage of the auxiliary. More recently, a lipase-
catalyzed resolution of racemic 2-MP was described (Scheme
(
6
1
d). In this case high ee product could only be isolated in low
yield due to relatively poor enantiospecificity of the enzyme.
Application of microbial oxidation in a resolution reaction
7
afforded the desired enantiomer with only 40% ee (Scheme 1e).
Although effective at small scale, the procedures outlined in
Scheme 1 are all unsuitable for industry because each suffers
from one or more of the following: high cost of starting
materials, large number of processing steps, poor yields, poor
selectivity, or difficult purification procedures.
(R)-enantiomer of racemic 2-methylvaleraldehyde to the desired
product with high volumetric productivity. A scaleable method
for separating the desired product from the off-enantiomer of the
starting material is also described. The process is cost-effective,
green, and amenable to manufacturing scale.
The first industrially viable synthesis of R-2-MP was
8
developed at BASF in 2006 (Scheme 2). In this hybrid
biocatalytic-chemocatalytic process, the product was prepared
by selective hydrogenation of the unsaturated aldehyde to the
allylic alcohol followed by asymmetric hydrogenation of the
double bond at 200 bar. Due to the relatively low enantiose-
lectivity of the chemocatalyst (∼75% ee), an additional step
was required to upgrade the product’s chiral purity. To that end,
a lipase resolution was employed to increase the enantiopurity
of the product to 98% ee. Despite the drawbacks of high
pressure and the requirement for an extra step to upgrade the
ee of the product, this process was effective as demonstrated
Introduction
(R)-2-Methylpentanol (R-2-MP) is an important intermediate
for organic synthesis that has been utilized for the production
1
2
of certain pharmaceuticals and liquid crystals. Several different
synthetic approaches to this material based on resolution or
asymmetric syntheses have been previously described (Scheme
1). The first report utilized a resolution of racemic 2-MP through
crystallization of the tosylate salt of L-valine-2-methylpentyl
ester (Scheme 1a). Although somewhat effective based on
optical rotation, this procedure was low-yielding, and the
enantiomeric excess was not determined. In 1985 Oppolzer
described the first asymmetric synthesis via a diastereoselective
ester-enolate alkylation of a chiral sultam (Scheme 1b). This
procedure was high-yielding but provided material with only
modest enantiopurity. A year later Danishefsky utilized R-2-
MP in the synthesis of the antibiotic Zincophorin. In this work
3
3
9
by successful scale-up in a 3 m reactor.
Enzymes that catalyze the reduction of ketones (or alde-
hydes) to corresponding alcohols are known as ketoreductases
1
0
(
KRED). The use of KREDs for organic synthesis has been
growing rapidly because these biocatalysts can be highly
4
11
enantioselective in the formation of chiral alcohols and they
are becoming more widely available. With the advance of
12
5
an Evans diastereoselective alkylation of a chiral oxazolidinone-
derived imide enolate was employed (Scheme 1c). This alky-
lation went with 8:1 diastereoselectivity, and enantiopure R-2-
MP was obtained following separation of the diastereomers and
(
(
6) Barth, S.; Effenberger, F. Tetrahedron: Asymmetry 1993, 4, 823.
7) Clark, D. S.; Geresh, S.; DiCosimo, R. Bioorg. Med. Chem. Lett. 1995,
5
, 1383–1388.
(
8) J a¨ kel, C.; Heydrich, G.; Sturmer, R.; Paciello, R. World Patent
Application WO 2006/034812, 2006.
(
9) J a¨ kel, C.; Paciello, R. Chem. ReV. 2006, 106, 2912–2942.
*
Author for correspondence. E-mail: owen.gooding@codexis.com.
(10) This enzyme type is sometimes referred to as alcohol dehydrogenase
(ADH).
(
1) Evans, M. C.; Franklin, L. C.; Murtagh, L. M.; Nanninga, T. N.;
Pearlman, B. A.; Saenz, J. E.; Willis, J. U.S. Patent Application
,714,168A1, 2007.
2) Morita, K.; Hachiya, S.; Moriwaki, F.; Endo, H. U.S. Patent 5,281,685,
994.
(11) (a) Kosjek, B.; Nti-Gyabaah, J.; Telari, K.; Dunne, L.; Moore, J. C.
Org. Process Res. DeV. 2008, 12, 584–588. (b) Lavandera, I.; Kern,
A.; Ferreira-Silva, B.; Glieder, A.; de Wildeman, S.; Kroutil, W. J.
Org. Chem. 2008, 73, 6003–6005. (c) Moore, J. C.; Pollard, D. J.;
Kosjek, B.; Devine, P. N. Acc. Chem. Res. 2007, 40, 1421–1419.
(12) A wide selection of KRED biocatalysts are commercially available
from Sigma-Aldrich (Milwaukee, WI), IEP GmbH (Wiesbaden,
Germany), Enzysource (Hangzhou, China), and Codexis (Redwood
City, CA).
2
(
1
(
(
3) Jermyn, M. A. Aust. J. Chem. 1967, 20, 2283.
4) Oppolzer, W.; Dudfield, P.; Stevenson, T.; Godel, T. HelV. Chim. Acta
1
985, 68, 212.
(
5) Zelle, R. E.; DeNinno, M. P.; Selnick, H. G.; Danishefsky, S. J. J.
Org. Chem. 1986, 51, 5032.
1
0.1021/op9002246  2010 American Chemical Society
Vol. 14, No. 1, 2010 / Organic Process Research & Development
•
119
Published on Web 12/01/2009

Scheme 1. Known small-scale routes to (R)-2-methylpentanol
Scheme 2. BASF process for the production
of(R)-2-methylpentanol
dependent on the cofactor ꢀ-nicotinamide adenine dinucleotide
phosphate, in its reduced form (NADPH), for effecting hydride
transfer to the carbonyl moiety. This cofactor should be required
in only catalytic amounts when its regeneration is coupled with
the oxidation of isopropyl alcohol (IPA) to acetone. Moreover,
the reduction of 2-methylvaleraldehyde and the oxidation of
IPA are both effected by a single enzyme with IPA serving as
14
(
13) For selected reviews of biocatalysis in a process chemistry environ-
ment, see: (a) Ran, N.; Zhao, L.; Chen, Z.; Tao, J. Green Chem. 2008,
1
2
0, 361–372. (b) Tao, J.; Zhao, L.; Ran, N. Org. Process Res. DeV.
007, 11, 259–267. (c) Sylvestre, J.; Chautard, H.; Cedrone, F.;
Delcourt, M. Org. Process Res. DeV. 2006, 10, 562–571. (d) Leresche,
J. E.; Meyer, H.-P. Org. Process Res. DeV. 2006, 10, 572–580. (e)
Wells, A. Org. Process Res. DeV. 2006, 10, 678–681. (f) Chikusa,
Y.; Hirayama, Y.; Ikunaka, M.; Inoue, T.; Kamiyama, M.; Moriwaki,
M.; Nishimoto, Y.; Nomoto, F.; Ogawa, K.; Ohno, T.; Otsuka, K.;
Sakota, A.; Shirasaka, N.; Uzura, A.; Uzura, K. Org. Process Res.
DeV. 2003, 3, 289–296. (g) Gotor, V. Org. Process Res. DeV. 2002,
6, 420–426. (h) Lye, G. J.; Dalby, P. A.; Woodley, J. M. Org. Process
Res. DeV. 2002, 6, 434–440. (i) Wandrey, C.; Liese, A.; Kihumbu, D.
Org. Process Res. DeV. 2000, 6, 286–290. (j) Bommarius, A. S.;
Riebel, B. R. Biocatalysis; Wiley-VCH: New York, 2003. (k) Liese,
A.; Seelbach, K.; Wandrey, C. Industrial Biotransformations, 2nd ed.;
Wiley-VCH: New York, 2006. (l) Drauz, K.; Waldmann, H. Enzyme
Catalysis in Organic Synthesis; Wiley-VCH: New York, 2002; (m)
Patel, R. StereoselectiVe Biocatalysis; Marcel Dekker: New York,
modern protein engineering techniques enzymes can be tailored
to accept specific substrates and function within commercially
relevant process constraints. In comparison with typical chemi-
cal catalysts for asymmetric reduction, enzymes have an added
benefit of being relatively low cost, easy to manufacture,
renewable, and enabling environmentally friendly manufacturing
13
processes.
Here we report on the development of an efficient process
for preparing R-2-MP from racemic 2-methylvaleraldehyde
based on enantiospecific enzymatic reduction (Scheme 3). In
this process, a highly active KRED biocatalyst was engineered
to convert only (R)-2-methylvaleraldehyde to the desired product
leaving the (S)-enantiomer unreacted. The KRED enzyme is
2
000.
14) (a) Bradshaw, C. W.; Hummel, W.; Wong, C. H. J. Org. Chem. 1992,
7, 1532–1536. (b) Bastos, F. M.; Dos Santos, A. G.; Jones, J., Jr.;
Oestreicher, E. G.; Pinto, G. F.; Paiva, L. M. C. Biotechnol. Tech.
999, 3, 661–664. (c) Bastos, F. M.; Franca, T. K.; Machado, G. D. C.;
(
5
1
Pinto, G. F.; Oestreicher, E. G.; Paiva, L. M. C. J. Mol. Catal. B:
Enzym. 2002, 19-20, 459–465.
1
20
•
Vol. 14, No. 1, 2010 / Organic Process Research & Development

Scheme 3. Enantiospecific reduction of 2-methylvaleraldehyde
the stoichiometric reductant in the catalytic cycle. This makes
process economics quite favorable, since a second enzyme is
not required to regenerate the cofactor. We also report a
purification and isolation method for separating the product from
the unreacted substrate. The process was developed at the 2-L
scale and has been successfully scaled to produce 100 kg of
material suitable for use in the production of an API.
Table 1. Process targets for evolved biocatalyst performance
parameter
targets
-1
substrate] (g L )
g220
e2
e24
e10
g45
g98
-1
[enzyme] (g L )
reaction time (h)
NADP cost contribution ($/kg)
conversion (%)
ee (%)
Results and Discussion
0.5 g L- NADP cofactor to give a calculated selectivity factor
1
+
Kinetic vs Dynamic Kinetic Resolution. The reaction as
outlined in Scheme 3 is a kinetic resolution where (without
recycle) the possible maximum yield is 50%. If this type of
reaction can be run under conditions where the aldehyde is
undergoing racemization at a rate comparable to or faster than
reduction, then a dynamic kinetic resolution is possible. This
is of course highly desirable because the maximum theoretical
yield increases to 100%. For this to be successful, the racem-
ization reagents and conditions must be compatible with the
enzyme. It is well documented that racemization of chiral
18
of E ) 20. This data suggested that both enzyme enan-
tiospecificity and volumetric productivity needed to be improved
to achieve an economically viable chemical process. On the
basis of this analysis, an enzyme evolution project was initiated
utilizing the L. kefir KRED as a starting point.
Enzyme Evolution. For successful evolution of an enzyme
it is critical to screen libraries of enzyme variants under
conditions that resemble the target process conditions. Table
lists the performance targets required to achieve a com-
mercially desirable process that were defined at inception of
this project. Volumetric productivity (g L day ) is a critical
metric that is directly related to substrate loading (g L ) and
19
1
2-alkyl carbonyl compounds (including aldehydes) is rapid in
15
polar solvents with either acid or base catalysis. In our hands
racemization of (R)-2-methylvaleraldehyde was incomplete at
pH ) 11, and rapid at pH ) 12. Unfortunately, our enzymes
were not stable above pH 9 so high pH conditions supportive
of racemization were precluded. Other reagents known to
catalyze aldehyde racemization at lower pH, such as 4-dim-
ethylaminopyridine and L-proline were tested without success.
On the basis of these findings we did not pursue the dynamic
-1
-1
-1
-1
reaction time (day ). Our analysis indicated that a substrate
loading of >220 g L and a reaction time of <24 h giving a
product volumetric productivity of >100 g L day would
meet the requirement. Other critical targets included enzyme
-1
-1
-1
-1
charge (e2 g L ), cofactor cost contribution (e$10 per kg),
20
16
conversion (g45%), and selectivity (g98% ee). We estimated
that our commercialization goals could be realized through
achievement of these targets.
Full details of our directed-evolution approach are beyond
the scope of this work, and thus, only a brief summary is
kinetic resolution approach further.
Initial Survey of Potential Biocatalysts. The first step in
the development of a biocatalytic process for 2-methylvaleral-
dehyde resolution was the identification of a KRED enzyme
with some level of activity for carbonyl reduction. Screening
of our internal collection of wild-type enzymes revealed activity
for a KRED isolated from Lactobacillus kefir (GenBank
accession #AAP94029.1) that gave R-2-MP of 85% ee at 40%
(
(
18) The selectivity factor (E) takes both conversion and selectivity into
account, see: Kagan, H. B.; Fiaud, J. C. In Topics in Stereochemistry;
Eliel, E. L., Wilen, S. H., Eds.; Interscience: New York, 1988; Vol.
1
7
conversion under initial screening conditions. The screening
reaction conditions utilized 0.225 M (23 g L ) substrate and
1
8, pp 249-330.
-
1
19) This is a most critical aspect of the Codexis enzyme evolution
technique. In a classical process development program available
catalysts are screened to discover the best performers. Then the process
is adjusted to fit the catalyst. In our method the catalyst is adjusted/
optimized to fit the desired chemical process, rather than the other
way around. This represents a new paradigm in catalytic process
development.
(
15) (a) Badin, E. J.; Pascu, E. J. Am. Chem. Soc. 1945, 67, 1353–1356.
b) Braddock, D. C.; Brown, J. M. Tetrahedron: Asymmetry 2000,
1, 3591–3607. (c) Cram, D. J.; Rickborn, B.; Kingsbury, C. A.;
(
1
Haberfield, P. J. Am. Chem. Soc. 1961, 83, 3678–3687. (d) Ebbers,
E. J.; Ariaans, G. J. A.; Houbiers, J. P.; Bruggink, A.; Zwanenburg,
B. Tetrahedron 1997, 53, 9417–9476.
(20) The selectivity factor (E) takes both conversion and selectivity into
account. For a kinetic resolution with 45% conversion and 98% ee an
E ≈ 250 is required, see: Kagan, H. B.; Fiaud, J. C. In Topics in
Stereochemistry; Eliel, E. L., Wilen, S. H., Eds.; Interscience: New
York, 1988; Vol. 18, pp 249-330. Extremely high E values are
required when the desired product is formed from the enantiomer
undergoing reaction rather than the other way around. Thus, the present
type of kinetic resolution is the most difficult case.
(
(
16) The off-line recycling of the unreacted substrate is viable but was
deemed impractical for our scale in light of its relatively low cost of
∼
5 U.S.$/kg at metric ton level.
17) The initial screening conditions utilized glucose/glucose dehydrogenase
GDH) cofactor recycling. See: Faber, K. Biotransformations in
Organic Chemistry, 5th ed.; Springer-Verlag: Berlin, Heidelberg, NY;
004; p 179.
(
2
Vol. 14, No. 1, 2010 / Organic Process Research & Development
•
121

Table 2. Evolution of L. kefir ketoreductase: the variants
listed were generated by various directed-evolution methods
Table 3. Summary of activity and selectivity data for
variants 202 and 302 at 25 °C and 10 °C
active mutations
(compared to wild-type) acid mutations
number of amino
temperature reaction conversion specificity
variant
E
variant
(°C)
time (h)
(%)
(% ee)
E
wild-type <50
none
L153Q; K236R
G82S; E145S;
L153Q; I223 V
G82S; S96A;
0
2
4
202
302
202
302
202
302
25
25
10
10
10
10
24
24
19
19
45
45
40
52
30
38
42
50
97.2
92.9
98.9
98.1
98.2
96.2
138
-
275
193
235
206
0
42
02
51-100
2
101-200
3
74
201-300
5
E145S; L153Q; I223 V
2
1
evolution were required to generate a biocatalyst capable of
meeting all of the targeted process conditions.
provided. Libraries of mutant enzymes were generated through
standard molecular biology techniques and screened for activity
and selectivity under the prescribed reaction conditions in the
high throughput format of 96-well plates. The reactions were
assayed in a tiered format: the first tier screen was an active/
inactive spectrophotometric assay for cofactor turnover; the
second tier screen was a GC analysis which was applied to the
active hits to determine activity (conversion) and enantiospeci-
ficity. These measurements were made relative to the control
enzyme so that improved variants could be culled out. The hits
were then validated under more representative process condi-
tions in small-scale chemical reactors where product isolation
was also studied. The genes encoding validated protein hits were
sequenced, and their translated protein sequences were analyzed
via statistical modeling [protein structure-activity relationships
Process Development
Effect of Temperature. In parallel with the identification
of improved enzyme variants during enzyme optimization, gram
quantities of the best variants in each round of evolution were
produced for process development studies. Enzyme optimization
and process development studies were performed concomitantly
to ensure that improved variants identified in high throughput
screening indeed exhibited improvements under process condi-
tions. In addition, these hits were used to fine-tune the process
conditions utilizing laboratory reactors. For example, the effect
of temperature on selectivity was evaluated utilizing two
biocatalysts at two different temperatures, 25 and 10 °C (Table
3). Biocatalyst 202 was the top performer from the second round
(
ProSAR)] methods. On the basis of this treatment, follow-on
of evolution (40% conversion in 24 h), while 302 was identified
as more active than 202 in round 3 screening (48% conversion
in 24 h) albeit with lower selectivity at 92.9% ee. In an effort
to improve selectivity, both variants were further tested at 10
libraries were designed, prepared, and screened. This process
was repeated in an iterative fashion until activity and selectivity
targets were reached.
The L. kefir KRED exhibited reasonable activity towards
°
C. The reactions were repeated in duplicate at 10 °C for two
different reaction times. Similar to what was observed at 25
C, 302 showed higher reactivity but less selectivity than 202.
2MP, but the product was formed in only 85% ee; thus
enantiospecificity became the key target for enzyme evolution.
The first improved variant, known internally as 042, contained
two mutations compared to the natural enzyme from L. kefir
°
Both variants provided R-2-MP with higher enantiopurity at
lower temperature; however, this increased selectivity came at
the expense of reaction rate.
(Table 2). This enzyme displayed better selectivity and was also
amenable to the preferred IPA cofactor recycling conditions.
This enzyme was capable of reaching 45% conversion giving
product of 95% ee in 23 h. This performance was achieved at
Stopping the Reaction. As described above, reactions run
with 302 showed that S-2-MP was detected as the reaction
approached 50% conversion. This was of concern because in a
kinetic resolution where the reaction product is desired, it is
imperative to stop the reaction before conversion exceeds 50%.
-1
-1
-1
+
1
10 g L substrate, 2 g L KRED, and 0.15 g L NADP
in 35% IPA/buffer. Although substantially improved over that
of the wild-type, this performance fell short of our process
targets so enzyme evolution was continued. The next substan-
tially improved enzyme variant was 202. This variant contained
four mutations relative to the natural enzyme. The performance
was improved in terms of both selectivity and activity, affording
2+
Since the L. kefir-based KRED variants require Mg to
22
maintain activity, it seemed plausible to use a chelating agent,
such as EDTA, to inactivate the enzyme so that the reaction
could be stopped at a desired conversion. A series of experi-
ments were conducted in which various amounts of EDTA
3
8% conversion and 97.8% ee in 40 h at the target substrate
2+
-1
(
25-85 equiv based on 2 mM Mg added during reaction set
loading of 220 g L ; however, reaction time and selectivity
were still below the performance goals. The next round of
evolution yielded 374, the first variant that met all the process
targets. This enzyme contained five mutations relative to the
natural enzyme. The performance was improved both in terms
of selectivity and activity affording product of 98.2% ee in a
up) were added to a 202-catalyzed reaction. The enzyme was
completely inactivated within minutes when at least 50 equiv
of EDTA was added at a point where the reaction had proceeded
to 25% conversion. This procedure was effective for stopping
the reaction at desired conversion and as such was incorporated
into the final process.
Process Optimization Utilizing the Final Variant. Once
variant 374 had been shown to meet the project targets of 45%
conversion and >98% ee further process development was
undertaken prior to validation and scale-up. Statistical design
2
3-h reaction at which point conversion was 46% at the target
-1
substrate loading of 220 g L . Thus, a total of three rounds of
(
21) (a) Fox, R. J.; Huisman, G. W. Trends Biotechnol. 2007, 3, 132–138.
b) Fox, R. J.; Davis, S. C.; Mundorff, E. C.; Newman, L. M.;
(
Gavrilovic, V.; Ma, S. K.; Chung, L. M.; Ching, C.; Tam, S.; Muley,
S.; Grate, J.; Gruber, J.; Whitman, J. C.; Sheldon, R. A.; Huisman,
G. W. Nat. Biotechnol. 2007, 25, 338–344.
(22) Hummel, W. Appl. Microbiol. Biotechnol. 1990, 34, 15–19.
1
22
•
Vol. 14, No. 1, 2010 / Organic Process Research & Development

concentration of 1.3 g L- would achieve 45% conversion,
giving product of 98.3% ee. This model prediction was tested
in duplicate reactions using the lowest amount of cofactor (0.05
1
Table 4. Variables, constants, and responses in the DoE
design utilizing variant 374
variables
constants
responses
-1
+
.75 g L-1 e KRED
g L NADP ). It was found that the reaction was slower than
expected, reaching only 41% conversion in 22 h (Table 5,
entries 2 and 3). Reactions using the original (pre-DoE)
conditions reached 45% conversion in 7 h (Table 5, entry 1).
However, under the new conditions, the enantiopurity of the
product was excellent at 99.0% ee.
0
0
1
[substrate] )
conversion (%)
-1
-1
e 2.5 g L
220 g L
.04 g L- e NADP+
1
IPA )
45% (v/v)
pH ) 7.5
selectivity (% ee)
e 0.13 g L-
1
5 °C e temperature
e 25 °C
of experiments (DoE) methods were employed to identify the
Product Isolation. The final reaction mixture of this kinetic
resolution contained at least 50% 2-methylvaleraldehyde, in
addition to isopropyl alcohol, water, acetone, and the product.
The separation of the product R-2-MP from the starting material
was challenging as both are low molecular weight liquids with
similar boiling points. Calculations indicated that ∼9 theoretical
plates would be required for their separation by fractional
distillation. In theory this is straightforward; however, in practice
it was quite difficult. Thus, several alternative procedures were
explored, including: (1) base treatment to convert unreacted
aldehyde to higher molecular weight products; and (2) bisulfite
adduct formation to capture the unreacted aldehyde as a water-
soluble compound.
Base Treatment Approach. Aliphatic aldehydes readily
undergo aldol condensations upon treatment with strong base.
A simple aldol condensation of 2-methylvaleraldehyde should
consume the aldehyde to afford a product with about twice the
molecular weight and a higher boiling point, thereby facilitating
the recovery of R-2-MP via a simple distillation. A drawback
to this approach is the destructive nature of the workup,
precluding the possibility to recycle the (S)-aldehyde. A
concentrated reaction mixture was treated with excess 50%
aqueous NaOH at 75 °C for 2 h. This led to complete
consumption of aldehyde without noticeable decomposition of
the alcohol. Subsequent fractional distillation afforded R-2-MP
in 30% isolated yield. Surprisingly however, the enantiopurity
of R-2-MP had decreased from 97.3 to 87.8% ee. Because of
the degradation in enantiopurity this approach was abandoned.
Co-distillation with IPA. During the course of isolation
experiments it was discovered that 2-methylvaleraldehyde
codistilled with IPA at atmospheric pressure, yielding a distillate
of the aldehyde/IPA mixture with a composition of 17:83 w/w.
This property was exploited to reduce the levels of 2-meth-
ylvaleraldehyde prior to the bisulfite adduct formation described
below.
23
key reaction parameters. The screening design described in
Table 4 was carried out to determine which variables were most
+
significant to the reaction. Three key variablessNADP
concentration, reaction temperature, and biocatalyst concentra-
tionswere included in the design. The reason was that
concentrations of KRED and cofactor were known to affect
activity, as indicated by historical data, while temperature had
been shown to affect both activity and enantiospecificity as
suggested by Table 3. The substrate concentration, pH, %IPA,
and mixing rate were kept constant.
A total of 19 runs were conducted in a single experiment
24
utilizing a manual parallel synthesis device and the response
data points were collected at 4 h, 8 h, and 23 h. Analysis of
variance and regression analysis of the 8-h sample data for
conversion provided the calculated response surface graphs
shown in Chart 1. In Chart 1a, it can be seen that the most
significant variable affecting the conversion was the KRED
concentration, while the cofactor concentration had no effect.
A minor variable affecting the conversion was temperature
+
(
Chart 1d). It was noteworthy that the NADP concentration
did not affect conversion within this design space, which
suggested that it might be lowered without substantially
affecting the process performance and thereby providing cost
savings. Analysis of the 4- and 23-h data led to similar
conclusions.
Analyses of the selectivity response data for 8-h samples
indicated that the KRED concentration and the temperature were
both significant variables as shown in b and c of Chart 1.
Consistent with previous observations, lowering the reaction
temperature resulted in the product with increased ee at a given
%
conversion. Lowering the KRED concentration gave an
apparent increase in specificity, which could be completely
25
+
explained by the lower conversion. Again the NADP
concentration had no significant effect on the specificity, which
suggested that the amount could be lowered.
Bisulfite Addition Product Approach. Unhindered alde-
hydes readily react with sodium bisulfite to form addition
A summary of the model for this design space is shown in
Chart 1d. From this response surface model the % conversion
at 23 h for various combinations of KRED concentration and
27
products which are highly insoluble in organics. The reaction
is reversible on treatment with either acid or base and is useful
for separation of aldehydes from other organic-soluble com-
pounds. The addition of one equivalent of sodium bisulfite to
a concentrated reaction mixture led to the formation of a water-
soluble adduct. This allowed the separation of R-2-MP from
the aldehyde-bisulfite adduct through extraction with MTBE.
This approach was successfully implemented into the
production process. The final workup and isolation strategy
involved: (1) low temperature distillation to remove part of the
26
temperature may be predicted directly. Accordingly, it was
predicted that a 23-h reaction run at 20 °C with a KRED
(
23) For a review of parallel synthesis and design of experiment (DoE)
methods in pharmaceutical chemistry research see: Gooding, O. W.
Curr. Opin. Chem. Biol. 2004, 8, 297–304.
(
(
24) Two 12-place Radley’s Carousel instruments were utilized (Radleys
Discovery Technologies, U.K.).
25) It is typical for an inverse relationship to exist between conversion
and % ee in resolution processes where the reacting enantiomer is the
desired product.
(
26) The prediction is made by projecting values for the x-axis (amount
KRED) and y-axis (temperature) onto the z-axis response surface (%
conversion).
(27) (a) Soffer, M. D.; Bellis, M. P.; Gellerson, H. E.; Stewart, R. A.
Organuic Syntheses; Wiley and Sons: New York, 1963; Collect.Vol.
4, pp 903-906.
Vol. 14, No. 1, 2010 / Organic Process Research & Development
•
123

Chart 1. Predicted response surface diagrams from the model generated through statistical analysis of the DoE data
Table 5. Testing of model predictions: conversion and selectivity in MPH374-catalyzed reactions at reduced biocatalyst loadings
+
and NADP concentrations
amount of KRED
amount of NADP+
reaction time
(h)
temp
(°C)
conversion
(%)
selectivity
(% ee)
-1
-1
entry
(g L )
(g L )
1
2
3
2
1.3
1.3
0.1
0.05
0.05
7
22
22
25
20
20
45
41
41
98.0
98.9
99.1
unreacted aldehyde as a codistillate with IPA, (2) treatment with
sodium bisulfite to render the remaining residual aldehyde
water-soluble, (3) product extraction with MTBE, (4) concen-
tration and final simple distillation to isolate the product. A flow
diagram describing the final process is shown in Chart 2. The
process was validated at 2-L scale, affording product of excellent
quality in 67% theory as described in the Experimental Section.
Following this a pilot run was conducted at 400 kg substrate
charge. The process scaled nearly directly, requiring 28 h to
reach 45% conversion (8 h longer). Isolation of the product
through distillation was accomplished in two batches due to
equipment size limitations. The distillation proceeded as
expected, affording a total of 96.4 kg of product, 53.6% theory
shown in Table 1 were all met or exceeded, and the final product
quality was within specification for API production.
Conclusions
An effective process for the preparation of R-2-MP from
racemic 2-methylvaleraldehyde was developed. Directed en-
zyme evolution was applied to create a process-fit KRED
biocatalyst capable of operating at high substrate loading and
low enzyme and cofactor loading. The same KRED enzyme
was able to utilize IPA to regenerate the reduced form of the
cofactor so that the need for a second auxiliary enzyme, such
as GDH, could be circumvented. The biocatalytic reaction was
characterized and optimized utilizing DoE and parallel synthesis
techniques. Two approaches to product isolation were explored,
and an effective procedure was developed based on separation
of the product via bisulfite adduct formation of the unreacted
(assuming 45% conversion). The production batch showed
chemical purity of 99.93% (GC/FID area %) and enantiomeric
purity of 98.22% ee. Based on this the initial process targets
1
24
•
Vol. 14, No. 1, 2010 / Organic Process Research & Development

Chart 2. Flow diagram showing the operations employed in
the final process
First-Tier High Throughput NADPH Fluorescence Pre-
screen to Identify Enzymes for the Oxidation of 2-Propanol.
This first-tier NADPH fluorescence-based prescreen was de-
veloped to identify active variants and to assess the thermal
stability of KRED variants. Plasmid libraries obtained by
directed evolution and containing evolved ketoreductase genes
were transformed into Escherichia coli and plated on Luria-
Bertani (LB) broth containing 1% glucose and 30 µg/mL
chloramphenicol (CAM). After incubation for at least 16 h at
3
(
0 °C, colonies were picked using a Q-bot robotic colony picker
Genetix U.S.A., Inc., Beaverton, OR) into 96-well shallow-
well microtiter plates containing 180 µL of Terrific broth (TB),
% glucose, and 30 µg/mL CAM. Cells were grown overnight
1
at 30 °C with shaking at 200 rpm. Twenty microliters of this
culture was then transferred into 96-deep-well plates containing
3
50 µL of Terrific broth (TB), 2 mM MgSO , and 30 µg/mL
4
CAM. After incubation of deep-well plates at 30 °C with
shaking at 250 rpm for 2-3 h recombinant gene expression by
the cell cultures was induced by addition of isopropyl thioga-
lactoside (IPTG) to a final concentration of 1 mM. The plates
were then incubated at 30 °C with shaking at 250 rpm for
aldehyde. The developed process was validated at the 2-L scale
and was subsequently applied to produce 100 kg of material
suitable for use in production of an API. Compared with
previous routes, this one-step process is shorter and utilizes a
15-17 h.
28
Cells were pelleted via centrifugation, resuspended in 300
renewable biocatalyst.
µL of lysis buffer, and lysed by shaking at room temperature
for at least 1 h. The lysis buffer contained 100 mM triethano-
lamine(chloride) buffer (pH 7.0-7.2), 1 mg/mL lysozyme, 200
Experimental Section
Materials and Methods. ꢀ-nicotinamide adenine dinucle-
otide phosphate in its oxidized form (NaNADP) was obtained
from Oriental Yeast Company (Andover, MA). Screening
panels of enzyme variants and lyophilized enzyme powders
were obtained through standard molecular biology techniques
4
µg/mL polymixin B sulfate, and 2 mM MgSO . The plates were
then spun at 4000 rpm for 10 min at 4 °C, and the clear
supernatant was used in the fluorescent assay. In 96-well black
microtiter plates 20 µL of clear supernatant was added to 180
µL of an assay mixture consisting of 90 mM triethanolami-
29
described elsewhere. Screening panels and lyophilized enzyme
powders of KRED enzymes similar to those described here are
available from Codexis (Redwood City, CA).
4
ne(chloride) buffer pH 7.0, 1 mM MgSO , 1 g/L NADP, and
2% 2-propanol. Reaction progress was measured by following
All gas chromatographic analysis was conducted utilizing
an Agilent 6850 series II gas chromatograph (injection port
temperature, 200 °C; detector temperature, 300 °C; split ratio
the increase in fluorescence of NADPH at 445 nm after
excitation at 330 nm in a Flexstation (Molecular Devices,
U.S.A.). Variants with initial rates higher than the negative
control (cells transformed with empty vector) were consolidated
and tested in the cofactor/IPA recycling system.
100:1; He carrier gas, 0.6 mL/min). Specific methods are as
follows: Conversion of 2-methylvaleraldehyde to 2-methylpen-
tanol was determined using an Agilent 19091J-433E column
Second-Tier High Throughput Screening to Identify
Improved Enzymes for the Stereospecific Reduction of
(
phenyl methyl siloxane, capillary 30.0 m × 250 µm × 0.25
µm nominal); ramp profile: 60 °C for 3 min, ramp from 60 to
2
1
2-Methylvaleraldehyde Using Isopropyl Alcohol for Cofac-
00 °C at 20 °C per minute, hold for 1 min, injection volume,
µL; sample concentration, 15 mg/mL in IPA or MTBE;
tor Regeneration. Fifty microliters of cell lysate obtained as
described above was transferred to a deep well plate (Costar
retention time: 2-methylvaleraldehyde, 1.84 min; 2-methylpen-
tanol, 2.53 min. Calibration curves for substrate and product
were developed utilizing standard serial dilution techniques for
precise quantitation of conversion. The enantiomeric purity of
#3960). Forty microliters of 1 M triethanolamine(chloride)
buffer pH 7.0, 50 µL of 0.875 mg/mL Na-NADP (Oriental
Yeast), 70 µL of 2-methylpentanal, and 140 µL of isopropyl
alcohol were added to each well. After sealing the plates, the
reaction samples were shaken for at least 16 h at ambient
temperature.
(R)-2-methylpentanol was determined using a Chiraldex B-DP
column (30 m × 0.25 µm, cat no. 78023: Astec Inc., Whippany,
NJ), oven temperature 65 °C (isothermal), injection volume 3
µL, sample concentration 15 mg/mL in IPA or MTBE, retention
times: (R) 2-methylpentanol, 9.6 min, (S) 2-methylpentanol, 10.9
min.
After centrifugation (4000 rpm, 1 min, ambient temperature)
1
mL of MTBE was added per well. Plates were resealed and
shaken for 2 to 3 min, and the organic and aqueous layer were
separated by centrifugation (4000 rpm, 5 min, ambient tem-
perature). Five hundred microliters of the organic layer was
transferred into a new deep-well plate containing ∼100 mg of
(
28) The process mass intensity (PMI) for this process is 29.3, 54% of
which is recyclable solvent (IPA and MTBE), 30% is water, 12% is
substrate, and 5% is reagents.
(
29) Sambrook, J.; Fritsch, E. F.; Maniatis, T. Molecular Cloning: A
Laboratory Manual; Cold Spring Harbor Laboratory Press: Plainview,
New York, 1989; Vol. 1-3. (b) Provisional U.S. Patent application
submitted; filing date August 27, 2008, 61/092,647, 2008.
2 4
Na SO . The plates were sealed and shaken for 2-3 min. The
plates were resealed and shaken for 2-3 min. After centrifuga-
tion (4000 rpm, 5 min, ambient temperature), 100 µL of the
Vol. 14, No. 1, 2010 / Organic Process Research & Development
•
125

organic layer from each well was added to 100 µL of ethyl
acetate in wells of a fresh microtiter plate. After sealing the
plates, samples were subjected to chiral GC analysis described
above. The E value was calculated from the ee and %
conversion. Potential improved variants over the parent variant
were regrown in quadruplicates and retested under the described
conditions to identify the hit variants. In order to validate the
HTP results shake flask powders of improved variants were
prepared and tested in small-scale reactions according to the
procedures described below.
Production of Ketoreductase Powders; Shake Flask
Procedure. A single microbial colony of E. coli containing a
plasmid with the ketoreductase gene of interest was inoculated
into 50 mL of Tryptic broth (12 g/L bacto-tryptone, 24 g/L
yeast extract, 4 mL/L glycerol, 65 mM potassium phosphate,
pH 7.0) containing 30 µg/mL chloramphenicol and 1% glucose
in a 250 mL Erlenmeyer flask. Cells were grown overnight (at
least 16 h) in an incubator at 30 °C with shaking at 250 rpm.
The culture was diluted into 250 mL of Terrific Broth (1 mM
After the desired temperature and pH were reached, the reaction
was initiated by adding the following in order: 0.154 g of Na-
NADP dissolved in 15 mL of deionized water, 3.05 g of KRED
enzyme 374 dissolved in 100 mL of deionized water. The
conversion was monitored by GC as described above. Samples
were prepared for analysis as follows: 50 µL of reaction mixture
was partitioned between 0.5 mL of water and 1.0 mL of MTBE.
The MTBE layer was withdrawn and injected onto the GC.
When the reaction reached 45% conversion (20 h), the enzyme
was inactivated by charging with 7.5 g of EDTA and 15 g of
Celite 545. The mixture was stirred for 15 min prior to
proceeding. The reaction mixture was filtered through an “M”
sintered glass funnel, and the cake was washed with 75 mL of
isopropyl alcohol. The biphasic mixture was transferred to a 2
L distillation flask and the pH was adjusted to 7 ( 0.5 using
20% NaOH solution. The distillation flask was fitted with a
thermocouple connected to a temperature controller, a heating
mantle, and a jacketed distillation column (1.5 cm diameter ×
15 cm length) packed with glass Rasching rings (6 mm diameter
× 6 mm length, Chemglass p/n CG-1283-01). The distillation
was allowed to proceed until a total of 810 mL of distillate
was collected. The residue was cooled to ∼25 °C, and 95 g of
sodium bisulfite dissolved in 200 mL of deionized water was
added over a 6 min period such that the temperature remained
e35 °C. The reaction mixture was stirred for 2 h at 35 °C during
which time it became monophasic. The pH was adjusted to 6.5
using 20% NaOH. GC analysis (sample preparation as described
above) determined a product-to-substrate ratio of 96:4. The
solution was extracted with 3 × 300 mL of MTBE, followed
by washing of the combined organic extract with 75 mL of
deionized water, 75 mL of brine, 75 mL of 0.5 N NaOH, 50
MgSO
density at 600 nm (OD600) of 0.2 and allowed to grow at 30
C. Expression of the ketoreductase gene was induced with 1
4
, 30 µg/mL chloramphenicol) in 1 L flask to an optical
°
mM IPTG when the OD600 of the culture was 0.6-0.8 and the
culture was incubated overnight (at least 16 h). Cells were
harvested by centrifugation (5000 rpm, 15 min, 4 °C), and the
supernatant was discarded. The cell pellet was resuspended with
an equal volume of cold (4 °C) 100 mM triethanolamine
(chloride) buffer, pH 7.0, and harvested by centrifugation as
above. The washed cells were resuspended in 12 mL of the
cold triethanolamine (chloride) buffer and passed through a
French Press twice at 12000 psi while the temperature was
maintained at 4 °C. Cell debris was removed by centrifugation
mL of saturated NaHCO
3
, and 25 mL of brine. The organic
for 3 h and filtered
(9000 rpm, 45 min, 4 °C). The clear lysate supernatant was
solution was dried over 30 g of MgSO
4
collected and stored at -20 °C. Lyophilization of the frozen
clear lysate provided a dry powder of crude ketoreductase
enzyme.
Small-Scale Reduction of 2-Methylvaleraldehyde by
Engineered Ketoreductases Derived from L. kefir. Improved
KRED variants were evaluated on a small scale with magnetic
stirring. A 25 mL glass vial was charged as follows: to 11 mL
of 100 mM triethanolamine hydrochloride buffer (pH 7) were
added 31 mg of lyophilized KRED powder, 32 µL of 1 M
into a 2 L distillation flask fitted with a thermocouple connected
to a temperature controller, heating mantle, and a jacketed
distillation column. MTBE and residual IPA were removed by
atmospheric distillation until the pot temperature reached 130
°C. The product was obtained by vacuum distillation at 275
mmHg pressure. The fraction with a boiling range of 109-110
°C was collected, affording 101 g of colorless liquid (67% of
theory). The chemical purity was 97.7%, and the enantiomeric
purity was 98.5% ee. All physical and spectroscopic properties
were identical to those of an authentic sample of racemic
2-methylpentanol obtained commercially (Aldrich 214019-1 L).
4
MgSO , 3 mg of Na-NADP, 2.2 g of 2-methylvaleraldehyde,
and 7 mL of isopropyl alcohol. Reactions were run at 25 °C
for 18-24 h. Reaction samples were analyzed by the GC
methods described above.
Acknowledgment
Preparative-Scale Production of (R)-2-Methylpentanol.
To a 2 L jacketed reaction vessel fitted with a mechanical
stirring paddle, pH probe, nitrogen inlet, and thermocouple were
added 684 mL of isopropyl alcohol, 285 mL of 100 mM
We thank Stephanie Hartouni and Justin Kittell for technical
assistance in screening and analysis during the project. We also
thank James Lalonde, John Grate, and Chris Davis for helpful
discussions during the course of this work.
4
triethanolamine (pH 9.5), 4 mM MgSO , and 342 g of
2
-methylvaleraldehyde. Stirring was started at ∼100 rpm, and
Received for review August 21, 2009.
OP9002246
the temperature was set to 20.0 °C. The pH was adjusted to
pH ) 7.4 + 0.1 by addition of 10 vol % hydrochloric acid.
1
26
•
Vol. 14, No. 1, 2010 / Organic Process Research & Development