Kemp eliminase activity of ketosteroid isomerase

Article abstract of DOI:10.1021/acs.biochem.6b00762

Kemp eliminases represent the most successful class of computationally designed enzymes, with rate accelerations of up to 10⁹-fold relative to the rate of the same reaction in aqueous solution. Nevertheless, several other systems such as micell

Full text of DOI:10.1021/acs.biochem.6b00762

Subscriber access provided by UNIV TORONTO
Article
Kemp Eliminase Activity of Ketosteroid Isomerase
Vandana Lamba, Enis Sanchez, Lauren Rose Fanning, Kathryn Howe,
Maria Alejandra Alvarez Gonzalez, Daniel Herschlag, and Marcello Forconi
Biochemistry, Just Accepted Manuscript • DOI: 10.1021/acs.biochem.6b00762 • Publication Date (Web): 03 Jan 2017
Downloaded from http://pubs.acs.org on January 4, 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.
Biochemistry 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 58
Biochemistry
1
2
3
4
5
6
7
8
9
Kemp Eliminase Activity of Ketosteroid Isomerase
FUNDING
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
This work was supported by grants from the National Science Foundation to D.H.
(MCBꢀ1121778), from the Howard Hughes Medical Institute to the College of Charleston
as part of their 2012 Undergraduate Science Education Competition, from the National
Center for Research Resources (5 P20 RR016461) and the National Institute of General
Medical Sciences (8 P20 GM103499) from the National Institutes of Health to the
College of Charleston, and from Research Corporation for Science Advancement to
M.F. (Cottrell College Science Award #22490). The NMR spectrometer at the College of
Charleston is supported by the National Science Foundation under Grant No. 1429308.
AUTHORS
†,a
‡,a
‡,a
§
Vandana Lamba, Enis Sanchez, Lauren Rose Fanning, Kathryn Howe, Maria
ǁ
†, #
‡,
Alejandra Alvarez, Daniel Herschlag, and Marcello Forconi *
AFFILIATIONS
‡
Department of Chemistry and Biochemistry, College of Charleston, Charleston, South
Carolina, 29424
§
Palmetto Homeschool Association, Rock Hill, South Carolina, 29730
ǁ
R.B. Stall High School, North Charleston, South Carolina, 29418
1
ACS Paragon Plus Environment

Biochemistry
Page 2 of 58
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
†
#
Department of Biochemistry, Stanford University, Stanford, California, 94305
Department of Chemistry, Department of Chemical Engineering, Stanford ChEMꢀH,
Stanford University, Stanford, California, 94305
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
a
These authors contributed equally to this paper.
*
Corresponding author. Telephone 843ꢀ953ꢀ3616. Fax: 843ꢀ953ꢀ7811. Eꢀmail:
forconim@cofc.edu
2
ACS Paragon Plus Environment

Page 3 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
Abbreviations and Textual Footnotes
KSI = ketosteroid isomerase
tKSI = Comamonas testoteroni ketosteroid isomerase
wt = wild type
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
5
5
4
4
NBI = 5ꢀnitrobenzisoxazole
BrBI = 5ꢀbromobenzisoxazole
,6ꢀCl BI = 4,6ꢀdichlorobenzisoxazole
2
ClPh = 4ꢀchlorophenol
FOOTNOTE 1. The pHꢀrate profiles for the HG3 and HG3.17 enzymes are bellꢀshaped,
5
, 6
with inflections at pK values of ~ 6.0 and 7.9. It has been assumed that the first pKa
a
value (6.0) corresponds to the deprotonation of the catalytic aspartate (D127) while the
second pK value (7.9) corresponds to an unknown anticatalytic deprotonation event.
a
Nevertheless, a kinetically equivalent model in which the pK value of 6.0 corresponds
a
to the unknown deprotonation event and the pK value of 7.9 corresponds to the
a
5
0
deprotonation of D127 would give exactly the same profile. This alternative model has
not been tested or ruled out. Thus, it is uncertain whether the pK value of D 127 is 6.0
a
or 7.9 in the HG3 enzymes.
3
ACS Paragon Plus Environment

Biochemistry
Page 4 of 58
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
Kemp eliminases represent the most successful class of computationally designed
9
enzymes, with rate accelerations up to 10 ꢀfold relative to the same reaction in aqueous
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
solution. Nevertheless, several other systems, such as micelles, catalytic antibodies,
and cavitands are known to accelerate the Kemp elimination by several orders of
magnitude. We found that the naturally occurring enzyme ketosteroid isomerase (KSI)
also catalyzes the Kemp elimination. Surprisingly, mutations of D38, the residue that
acts as a general base for its natural substrate, produced variants that catalyze the
Kemp elimination up to 7,000ꢀfold better than wildꢀtype KSI, and some of these variants
accelerate the Kemp elimination more than the computationally designed Kemp
eliminases. Analysis of the D38N general base KSI variant suggests that a different
active site carboxylate residue, D99, carries out the proton abstraction. Docking
simulations and analysis of inhibition by active site binders suggest that the Kemp
elimination takes place in the active site of KSI and that KSI uses the same catalytic
strategies of the computationally designed enzymes. In agreement with prior
observations, our results strengthen the conclusion that significant rate accelerations of
the Kemp elimination can be achieved with very few, nonꢀspecific interactions with the
substrate if a suitable catalytic base is present in a hydrophobic environment.
Computational design is able to fulfill these requirements, and the design of more
complex and precise environments represents the next level of challenges for protein
design.
[end of abstract]
4
ACS Paragon Plus Environment

Page 5 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
1-3
The Kemp elimination involves abstraction of the proton bonded to the C3 atom of the
aromatic benzisoxazole substrate (Figure 1A). Concurrently with proton transfer, bond
cleavage of the weak OꢀN bond opens the fiveꢀmembered ring and yields 2ꢀ
cyanophenolate as the product. This reaction represents, at least in principle, an ideal
candidate for the de-novo computational design of enzymes because it is not present in
any metabolic pathway, thereby eliminating biases that might arise from inclusions of
alreadyꢀknown catalytic motifs in the computational design.
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
At least two distinct classes of artificial Kemp eliminase have been designed in
the past decade. The first class, called KE, was developed using a quantum mechanical
4
approach to position residues around the substrates. This arrangement was then
compared to known scaffold in proteins using the RosettaMatch algorithm, and the
scaffolds obtained were further optimized through computation. Most of the designed
variants were inactive, but the most active variant (KE59), which was derived from a
triose phosphate isomerase barrel scaffold, was found to use a catalytic glutamate in
the active site and to carry out the Kemp elimination of 5ꢀnitrobenzisoxazole six orders
4
of magnitude faster than acetate in water. The second class of Kemp eliminases, the
HG eliminases, was designed using an iterative approach from a xylanase binding
5
pocket and optimized using molecular dynamics. The most active variant, HG3, uses a
catalytic glutamate and accelerates the reaction seven orders of magnitude relative to
4
, 6
acetate in water. For both classes, subsequent directed evolution resulted in ~100ꢀ
5
ꢀ1 ꢀ1
fold rate increase, with k /K for HG3.17 equal to 2.3 x 10 M s , which is nine orders
cat
M
of magnitude higher than the corresponding secondꢀorder rate reaction with acetate in
6
water. However, while the design led to considerable rate accelerations, these values
5
ACS Paragon Plus Environment

Biochemistry
Page 6 of 58
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
fall at the lower end of rate enhancements observed for naturally occurring enzymes,
27
7, 8
which range up to ~10 ꢀfold.
Although computational design yielded active enzymes, the Kemp elimination is
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
9-14
also accelerated in very simple systems.
In particular, negatively charged species,
such as acetate, carry out the Kemp elimination 7ꢀ9 orders of magnitude faster in
15
organic solvents than they do in water. These significant rate accelerations have been
ascribed to electron delocalization in the reaction’s transition state and poorer solvation
of the anionic base in aprotic solvents, which destabilizes the base and makes it more
15-17
available for proton abstraction.
Approaches such as evolution of catalytic
18-20
21-24
antibodies
and rational redesign of cavitands
have also produced proteins that
use an active site carboxylate and these provide rate accelerations that rival those of
computationallyꢀdesigned enzymes.
Thus, evolution, rational design, and computational design have provided
researchers with active Kemp eliminases. In addition, it is known that some proteins
possess promiscuous Kemp eliminase activity. Bovine serum albumin catalyzes the
ꢀ1 ꢀ1
Kemp elimination of 5ꢀnitrobenzisoxazole with a secondꢀorder rate constant of 40 M s
25, 26
at pH 8.0, but it does so using a catalytic lysine.
As amines are much better
27
catalysts than carboxylates for the Kemp elimination in aqueous solution, this value
represents a rate acceleration of only three orders of magnitude. Other proteins known
to accelerate the Kemp elimination are the putative oxidoreductase ydbC and the purine
28
nucleoside phosphorylase xapA, although their mechanisms are unknown.
Given the literature examples, it seemed likely that enzymes that possess a
carboxylate residue in a relatively hydrophobic active site that can accommodate Kemp
6
ACS Paragon Plus Environment

Page 7 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
elimination substrates might be able to accelerate their elimination reaction. To
determine if this were the case, we turned to ketosteroid isomerase from Comamonas
29
testosteroni (tKSI), an enzyme that binds hydrophobic steroids and catalyzes proton
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
5
migration from C4 to C6 of 3ꢀoxoꢀ∆ ketosteroids using an active site aspartate (Asp38,
30
Figure 1B). This reaction is facilitated by two additional residues, Asp99 and Tyr 14,
30
which constitute an oxyanion hole (Figure 1B) We have asked whether and to what
extent wild type and mutant KSI variants accelerate the Kemp elimination. We found
that tKSI is a good Kemp eliminase. Unexpectedly, a single mutation in the active site of
tKSI produced an even more active variant that accelerates the Kemp elimination more
than the computationally designed Kemp eliminases and is only one order of magnitude
less reactive than their evolved variants.
7
ACS Paragon Plus Environment

Biochemistry
Page 8 of 58
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
Materials and Methods.
Materials. 5ꢀnitrobenzisoxazole, 6ꢀnitrobenzisoxazole, 5ꢀbromobenzisoxazole, and 6ꢀ
bromobenzisoxazole were from Ark Pharm (Libertyville, IL). Benzisoxazole, 4,6ꢀ
dichlorosalicylaldehyde and hydroxylamine were from Alfa Aesar. Triphenylphosphine
and 2,3ꢀdichloroꢀ5,6ꢀdicyanoꢀ1,4ꢀbenzoquinone (DDQ) were from SigmaꢀAldrich.
Buffers and salts were from Alfa Aesar or SigmaꢀAldrich. UV spectra and
spectrophotometric kinetic assays were recorded using a Varian Cary 50 UV/visible
spectrophotometer.
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
Synthesis of 4,6-dichlorobenzisoxazole. 4,6ꢀDichlorobenzisoxazole was prepared
according to a modified literature procedure, as described below.
(a)
Synthesis of 4,6-dichloro-2-hydroxybenzaldoxime
4
,6ꢀDichlorosalicylaldehyde (0.500 g, 2.62 mmol) was dissolved in 7 ml of ethanol and
mixed with 12 ml of ethanolic NH OHꢁHCl (242 mg, 3.48 mmol) The solution was
2
refluxed at 65 °C under stirring and monitored by TLC (70:30 hexanes/ethyl acetate).
After 18 hours the solution was cooled to room temperature and cold water was added.
The yellow solid precipitate that crashed out of solution was collected by filtration and
1
dried under high vacuum for 4 hours. Yield: 398 mg (74%). The H NMR spectra in
31
CD OD was consistent with the one reported in the literature.
3
(b)
Synthesis of 4,6-dichlorobenzisoxazole
To a solution of 2,3ꢀdichloroꢀ5,6ꢀdicyanoꢀ1,4ꢀbenzoquinone (DDQ, 332 mg, 1.46 mmol)
and triphenylphosphine (Ph P, 383 mg, 1.46 mmol) in CH Cl (5 mL) was added 4,6ꢀ
3
2
2
8
ACS Paragon Plus Environment

Page 9 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
dichloroꢀ2ꢀhydroxybenzaldoxime (200 mg, 0.971 mmol) at room temperature. The
resulting solution was stirred for an additional 10 minutes. Solvent was removed via
rotary evaporation, and the crude reaction mixture was purified using SiO column
2
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
chromatography with 85% hexane / 15% ethyl acetate as the eluent. Solvents were
removed in vacuo, and the product was isolated as a white solid (100 mg, 55% yield).
1
32
H NMR in CDCl was consistent with the literature.
3
Protein expression and purification of KSI WT and mutants: BL21 cells were
transformed with KSI constructs and a single colony was picked up to grow an overnight
culture. The colony was sequenced to confirm the identity of the mutant. From the
overnight culture, cells were grown in Luria broth media containing 50 ꢂg/ml carbenicillin
at 37 °C to an O.D. of 0.6ꢀ0.8. After that, IPTG was added to a final concentration of 0.5
mM to induce the protein expression. Cells were grown for additional 5ꢀ6 hrs and
harvested by centrifuging at 4042 × g for 20 min and purification was carried out using
33-35
previously reported protocols.
To prevent contamination, the deoxycholate affinity
column as well as FPLC loops and gel filteration column Superose 12 were washed 6 M
guanidinium, 40 mM potassium phosphate, pH 7.2 buffer followed by 40 mM potassium
35
phosphate, 1 mM EDTA, pH 7.2 buffer prior to each protein purification as described.
Phosphate buffer was removed via a Zeba Spin Desalting column (Thermo Fisher).
Identity of the mutant was confirmed by mass spectrometry and purity of > 95% was
determined using polyacrylamide SDS gel electrophoresis.
9
ACS Paragon Plus Environment

Biochemistry
Page 10 of 58
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
Kinetic assays of KSI-catalyzed reactions. Reactions were carried out at 20 °C in 25
mM buffer and with the appropriate variant and concentration of tKSI (0ꢀ400 ꢃM). The
following buffers were used: sodium acetate, pH 4.3ꢀ5.9; sodium 2ꢀ(Nꢀ
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
morpholino)ethanesulfonic acid (MES), pH 5.4ꢀ6.4; sodium bisꢀtris propane (BTP), pH
6.0−9.5. Components were mixed in semiꢀmicro plastic cuvettes (BrandTech Scientific)
to a total volume of 500 ꢃL and reactions were initiated by addition of the benzisoxazole
substrate (2ꢀ5 ꢃL). Appearance of product was monitored at the following wavelengths:
ꢀ1
ꢀ1
5ꢀnitrobenzisoxazole, 380 nm (ε = 18,400 M cm ); 5ꢀbromobenzisoxazole, 349 nm (ε =
ꢀ1
ꢀ1
ꢀ1
ꢀ1
8
,500 M cm ), 4,6ꢀdichlorobenzisoxazole, 333 nm (ε = 28,290 M cm ); 6ꢀ
ꢀ1
ꢀ1
nitrobenzisoxazole, 398 nm (ε = 6,600 M cm ); 6ꢀbromobenzisoxazole, 327 nm (ε =
ꢀ1
ꢀ1
ꢀ1
ꢀ1
6
,100 M cm ); benzisoxazole, 323 nm (ε = 5,080 M cm ). The extinction coefficients
of the reaction products were determined by addition of concentrated sodium hydroxide
to their parent substrates solutions. Kinetic parameters were determined by initial
velocities using Kaleidagraph (Synergy Software). Reported errors are from fitting.
Inhibition reactions by 4ꢀchlorophenol and equilenin were performed in the same way,
except that the inhibitor was present in the reaction mixtures. Varying the incubation
time with the inhibitor did not affect the initial velocities. Product inhibition in D38N tKSI
was determined by allowing 0.30 mM or 0.50 mM substrate to react to form the product
(because of the high concentration of product, reactions were monitored at 460 nm
ꢀ1
ꢀ1
using a measured ε value of 480 M cm ). After completion, additional substrate was
added and the reaction was monitored again to determine the decrease in rate. This
process was repeated one more time. Controls were run with D38N tKSI left in buffer
1
0
ACS Paragon Plus Environment

Page 11 of 58
Biochemistry
1
2
3
over the same period of time and showed no timeꢀdependent decrease in reactivity,
suggesting that product formation was the sole responsible for the measured inhibition.
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
Docking Simulations. Docking simulations of benzisoxazole binding to tKSI were
performed using the online server of Rosetta Ligand Docking
(http://rosie.rosettacommons.org/ligand_docking). The chain A structure of D38N tKSI
with bound 5αꢀextranꢀ3,17ꢀdione (PDB ID 1OHP) was used as the template. Chains B,
C, and D, the ligands, and all of the water molecules were manually removed before
submitting the structure to the online server. The benzisoxazoles were drawn using
ChemDraw 2015 (PerkinElmer) and converted into .sdf format with OpenBabel 2.3.2
(http://openbabel.org/). Generated structures were visualized and analyzed using UCSF
Chimera (http://www.rbvi.ucsf.edu/chimera/) and Pymol (Schrödinger Software).
1
1
ACS Paragon Plus Environment

Biochemistry
Page 12 of 58
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
tKSI catalyzes the Kemp elimination
We first tested whether wildꢀtype tKSI (wt tKSI) catalyzes the Kemp elimination.
At pH 7.0 in the presence of 5.8 ꢃM wt tKSI, the Kemp elimination of 5ꢀ
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
nitrobenzisoxazole (5NBI) was faster than the same reaction in the absence of the
enzyme, with no apparent saturation observed up to 1.0 mM 5NBI (Figure 2). After
correction for the rate of the reaction without enzyme, we calculated a secondꢀorder rate
ꢀ1 ꢀ1
constant for the wt tKSIꢀcatalyzed reaction, k /K , of 2.5 M s (Figure 2, inset).
cat
M
Strikingly, this value is very similar to the secondꢀorder rate constants for the Kemp
elimination of 5NBI catalyzed by some of the nonꢀoptimized designed Kemp eliminases
(Table 1). If wt tKSI uses a carboxylate residue as the general base, the measured
4
secondꢀorder rate constant would correspond to an overall rate acceleration of 4.2 × 10
ꢀ5
ꢀ1 ꢀ1
6
fold relative to the acetateꢀcatalyzed reaction in water (k = 5.8 × 10 M s , ref ).
2
1
2
ACS Paragon Plus Environment

Page 13 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
Table 1. Second order rate constants for the Kemp elimination of 5-nitrobenzisoxazole catalyzed by some proteins at their
respective pH optima.
-1
-1
a
2
Protein
Description
Proposed Catalytic
k /K (M s )
(k /K )/k
ref
cat
M
cat
M
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
a.a.
4
tKSI
Natural enzyme
Asp
2.5 ± 0.6
4.2 ×10
7.2 ×10
this work
4
b
L99A/M102H Engineered cavitand from T4 His
1.80 ± 0.06
23
lysozyme
4
AlleyCat
Engineered allostericallyꢀ
regulated protein
Glu
5.8 ± 0.3
9.9 ×10
21
5
5
5
6
7
7
7
1
THF-2
Computationallyꢀdesigned
Natural lipidꢀcarrier
Asp
Lys
Glu
Glu
Glu
Asp
7.9 ± 0.1
1.3 ×10
2.0 ×10
2.1 ×10
2.7 ×10
2.2 ×10
2.2 ×10
5.5 ×10
5
a
BSA
6,500 ± 400
12.2 ± 0.1
160 ± 20
25, 36
4
KE07
KE59
AlleyCat7
HG3
Computationallyꢀdesigned
Computationallyꢀdesigned
Evolved AlleyCat
4
1,280 ± 10
1,300 ± 200
5,200 ± 200
24
Computationallyꢀdesigned
5, 6
20, 36
3
4E4
Optimized catalytic antibody Glu
1
3
ACS Paragon Plus Environment

Biochemistry
Page 14 of 58
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
8
9
9
D38N tKSI
KE59.13
HG3.17
ydbC
Single mutation in KSI
Evolved KE59
Asp
Glu
Asp
17,000 ± 1,000
59,000 ± 15,000
230,000 ± 40,000
430 ± 10
2.9 ×10
1.0 ×10
3.9 ×10
this work
32
6
Evolved HG3
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
Natural enzyme
Natural enzyme
Unknown
Unknown
Unknown
Unknown
28
28
xapA
1,000 ± 30
a
Rate accelerations were calculated using the values for the background reactions reported in the corresponding
b
references, which match the nature of the enzymatic catalytic base (carboxylate, amine, or imidazole). The rate
ꢀ5
ꢀ1 ꢀ1
constant for the background imidazoleꢀcatalyzed reaction (2.5 × 10 M s ) was derived from the data reported in Figure
S2 of this reference.
1
4
ACS Paragon Plus Environment

Page 15 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
To determine whether a carboxylate, such as the general base for the normal
reaction of tKSI (D38, Figure 1B), was the general base in the Kemp elimination, we
first determined the pHꢀrate profile for the tKSIꢀcatalyzed Kemp elimination of 5NBI. The
value of the rate constant for this enzymeꢀsubstrate pair increases to reach a plateau in
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
ꢀ
1 ꢀ1
the order of 2.5 M s . Fitting the data to a single deprotonation event, using Equation
1
, gives an apparent pK of 5.2 ± 0.8 (Figure 3, circles). This value is within error of the
a
37
value for the wt tKSI reaction of 4.6 which has been attributed to D38.
ꢉꢊꢋꢌꢍꢎ
ꢉꢊꢋꢌꢏꢐꢉꢊꢋꢌꢍꢎ
(ꢀꢁꢂꢃ/ꢄ ) =ꢆ(ꢀ /ꢄ ) ꢆ×
ꢅ ꢇꢂꢈ
(1)
ꢅ
ꢁꢂꢃ
The data presented in Figure 3 are consistent with a single catalytic
deprotonation event, such as the deprotonation of D38, raising the possibility that D38
could be the general base. Nevertheless, simple geometrical considerations suggest
that D38 might not be the general base for the tKSIꢀcatalyzed Kemp elimination. The
3
carbon center of reaction in the normal tKSI reaction is sp hybridized (Figure 1B),
which means that the general base is located outꢀofꢀplane with respect to the planar
38
steroid ring. Indeed, crystal structures of tKSI show D38 in such position. In contrast,
2
the carbon center of reaction in the Kemp elimination is hybridized sp , which requires
the base responsible for proton abstraction to lay in the same plane of the system
formed by the two aromatic rings. If the benzisoxazole system arranges itself with the
same geometry of the ketosteroid ring, then D38 would not be in a suitable position for
proton abstraction from the C3 of benzisoxazoles. To determine whether D38 is the
general base in the tKSIꢀcatalzed Kemp elimination, we performed mutagenesis in the
active site and we used docking simulations, as described below.
1
5
ACS Paragon Plus Environment

Biochemistry
Page 16 of 58
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
Mutation of D38 leads to faster Kemp eliminase activity
If D38 were the general base in the Kemp elimination, we expected the reaction
to return at basal levels upon mutation of D38 to residues that cannot accept a proton.
Contrary to this prediction, the tKSI variant with this residue mutated to neutral
asparagine (D38N), a residue not effective in proton abstraction, is about four orders of
magnitude more reactive than wt tKSI in the plateau region, with a maximum value of
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
4
ꢀ1 ꢀ1
k /K = 1.7 × 10 M s for reactions of 5NBI (Figure 4, filled circles). Similar results
cat
M
were obtained with variants with D38 mutated to leucine, glycine, or lysine (Table 2).
Thus, these results strongly imply that, D38 is not the general base in these mutants,
and suggest that the negative charge on the oxygen atom of D38 might be inhibitory for
the tKSIꢀcatalyzed Kemp elimination, either directly or because it suppresses ionization
of another group needed for reaction (see below).
Table 2. Maximum values of k /K for the Kemp elimination of 5NBI by tKSI variants
cat
M
with different residues at position 38.
-1 -1
Residue 38
k /K (M s ) Relative k /K
cat M cat M
D
G
K
L
2.5 ± 0.6
(1)
420 ± 30
170
1,000
2,500 ± 300
16,000 ± 2,000 6,400
17,000 ± 1,000 6,800
N
1
6
ACS Paragon Plus Environment

Page 17 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
The pHꢀrate profile for reactions of 5NBI with D38N tKSI revealed an apparent
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
pK value of 7.8 ± 0.2 (Figure 4, filled circles). To determine whether this value
a
corresponds to a real pK or if it reflects a nonꢀchemical rateꢀlimiting step, we used a
a
slower substrate, 5ꢀbromobenzisoxazole (5BrBI). As the bromo group is less
electronegative than the nitro group, the negative charge developed in the transition
state of the reaction (Figure 1A) will be less stabilized and hence a slower rate is
2, 39
expected with this substrate.
The pHꢀrate profile for 5BrBI (Figure 4, filled squares)
shows the same trend as with 5NBI, with an apparent pK of 7.9 ± 0.2 that is identical to
a
2
ꢀ1 ꢀ1
the value measured for 5NBI. However, the maximum k /K value of 6.0 × 10 M s is
cat
M
about 30–fold lower than that for 5NBI (Figure 4, plateaus). These results strongly
suggest that the saturation in the pHꢀrate profile corresponds to a deprotonation event
of an enzyme residue with a pK of about 7.8.
a
D99 is likely to be the general base in the D38N tKSI-catalyzed Kemp elimination
Because D38 cannot be the catalytic base in D38N tKSI, we looked at other
residues in the active site of tKSI. The other residue in the active site of tKSI that can
act as the general base base is D99 (see Figure S1 for a diagram of the residues
surrounding the steroid substrate in KSI), which normally stabilizes the oxyanion hole
developing on the keto oxygen of the substrate by donating a hydrogen bond (Figure
1
B).This residue is relatively inꢀplane with the ketosteroid ring, making it a better
candidate for the general base role in the tKSIꢀcatalyzed Kemp elimination. In wild type
1
7
ACS Paragon Plus Environment

Biochemistry
Page 18 of 58
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
40
41
tKSI, the pK of D99 is 9.5, but it was proposed to drop to 8.5 in the D38H mutant.
a
The value of this pK in the D38N mutant is not known, but it is likely to be below 8.5
a
because of the reduced ability of asparagine (relative to aspartate and histidine) to
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
destabilize the anionic form of D99 or stabilize protonated D99.Thus, the measured pKa
of 7.8 in the Kemp elimination is consistent with the expected pK of D99 in the D38N
a
mutant.
To test the role of D99 in the D38N mutant, we mutated this residue to
asparagine to yield the double mutant D38N/D99N. In this case, we used the substrate
4,6ꢀdichlorobenzisoxazole (4,6ꢀCl BI), which gives a larger rate acceleration relative to
2
the background reaction than 5NBI. The pHꢀrate profile using this substrate with D99N
tKSI shows a trend analogous to those of 5NBI and 5BrBI (Figure 5, empty circles),
with similar pK , suggesting that the chemical step is also rateꢀlimiting in this case and
a
that the reaction mechanism is the same. Reactivity of D38N/D99N (Figure 5, empty
squares) was ~200ꢀfold slower than the single mutant (D38N) in the pHꢀlinear region.
Intriguingly, this mutant was not completely inactive, reacting ~30ꢀfold faster than the
background, hydroxideꢀcatalyzed reaction. It is not unusual that mutations of the
general base in enzymes that catalyze the Kemp elimination produce variants with
reduced, but not abolished activity. For example, the E101A mutation in KE07 and the
E100A mutation in KE61 resulted in only 4ꢀ8 fold decreases in the secondꢀorder rate
4
constants of these enzymes, despite mutation of the general base. The apparent pK
a
for reactions of the double mutant was > 9, suggesting that a different residue or solvent
ꢀ
(
OH ) is responsible for proton abstraction. These results, combined with the analysis of
KSI active site and with docking simulations (see below), suggest that D99 acts as the
1
8
ACS Paragon Plus Environment

Page 19 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
general base in the D38N mutant; in the double mutant, the residual catalysis may arise
from proton abstraction from a different residue and/or from placement of the substrate
in nonꢀaqueous environment with solvent abstracting the proton.
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
To test whether D99 acts as the general base in wild type tKSI, we used the
D99N mutant. This mutant is only marginally less reactive then wt tKSI in the flat pH
ꢀ1 ꢀ1
region (k /K value of 1.3 ± 0.5 M s ), suggesting that D38, not D99, is the general
cat
M
base in wt tKSI. Further analysis suggests that anionic D99 is at least 100ꢀfold less
affective as a general base when anionic D38 is also present in the active site (Figure
S2), a result consistent with suppression of D99 ionization by the presence of ionized
41
D38.
Docking simulations and inhibition experiments place benzisoxazoles in the active site
of tKSI
As mutation of active site residues decreases or enhances activity, the tKSIꢀ
catalyzed Kemp elimination likely occurs in the active site. Further, the results
presented above suggest that D99 is the general base for this reaction in the D38N
mutant and that anionic D38 provides an inhibitory effect on D99 reactivity. A possible
reason for this inhibitory effect is unfavorable interactions between the negatively
42
charged oxygen of D38 and the substrate, as observed for the natural KSI substrate.
In order to obtain more information about (i) the KSI region in which the Kemp
elimination occurs, (ii) the nature of the general base, and (iii) the reason of the
inhibition of the Kemp elimination by D38, we performed docking simulation using
1
9
ACS Paragon Plus Environment

Biochemistry
Page 20 of 58
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
RosettaDocking. In these simulations, we used a crystal structure of D38N tKSI (PDB
ID: 1OHP) and docked various benzisoxazoles to it. In all of the simulations, the
structures with the lowest energy contained benzisoxazoles in the active site of tKSI.
Figure 6A shows the structure with the lowest energy for docking of 5NBI. In this
structure, 5NBI is placed within the active site and, with the exception of D99, it is
surrounded by either neutral or hydrophobic residues. N38 sits above the benzisoxazole
ring, and its nitrogen atom is 3.7 Å from the proton attached to the C3 of 5NBI (Figure
6B). The angle between the nitrogen atom of N38, the proton attached to C3 of 5NBI,
and C3 of 5NBI is 73°, which is significantly off from linearity (i.e., 180°). In contrast,
both of the oxygen atoms of D99 are within 2.9 Å from the acidic proton of 5NBI, and
the angles between these oxygen atoms, the proton attached to C3 of 5NBI, and C3 of
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
5
NBI are 136° and 125°. In agreement with the relatively weak binding affinity of 5NBI
(
K is > 1 mM for wt tKSI and >200 ꢃM for D38N tKSI) the only specific interaction
M
between the benzisoxazole and the active site of tKSI is the potential hydrogen bond
between D99 and the acidic proton of 5NBI.
Thus, docking simulations support the hypothesis that tKSI binds the Kemp
elimination substrate through nonꢀspecific interactions and that places it in the normal
active site. Because of the different positions of residues 38 and 99, these simulations
also support a model in which D99 abstracts the proton from the C3 of the substrate
and that the negative charge on residue 38, when present, alters substrate positioning
because of unfavorable interactions with the aromatic ring.
Active site binders inhibit the Kemp elimination.
2
0
ACS Paragon Plus Environment

Page 21 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
The mutational and computational analyses presented above strongly suggest
that the Kemp elimination takes place in the active site of tKSI. Thus, we expected that
inhibitors of the normal KSI reaction would inhibit the Kemp elimination. Equilenin, a
strong inhibitor of KSI reaction, inhibits the Kemp elimination of wt tKSI with an apparent
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
4
3
K of (5 ± 1) ꢃM (see Figure S3), which is similar to the literature value of 2ꢀ3 ꢃM.
I
Equilenin also inhibits reactions of the D38 mutants (data not shown) but we could not
determine its inhibition constant because it binds these mutants very tight (K < 1 nM at
I
43
pH values higher than 7). To determine an inhibition constant in D38 mutants, we
used a much weaker inhibitor, 4ꢀchlorophenol (4ClPh). This compound inhibits reactions
of D38N KSI and 4,6ꢀCl BI with a K of (8 ± 1) ꢃM (Figure 7), similar to the reported
2
I
44
values of 8ꢀ20 ꢃM for the cognate KSI reaction catalyzed by D38N. The very similar
values for the inhibition constants of the normal KSI reaction and of the Kemp
elimination, in conjunction with results from mutation of active site carboxylates and
docking simulations, strongly suggest that the Kemp elimination takes place at the KSI
active site. Additional experiments were consistent with the possibility of effects from
binding of a second molecule at high concentration of inhibitor (data not shown).
The product of the Kemp elimination is a phenolate (Figure 1A) and phenolates
34, 44, 45
are known KSI inhibitors.
In agreement with such observations, we measured a
K value for the reaction product 2ꢀnitroꢀ5ꢀhydroxybenzonitrile of 0.4ꢀ0.6 mM in D38N
I
tKSI (Figure S4). Thus, the product weakly inhibits the reaction.
2
1
ACS Paragon Plus Environment

Biochemistry
Page 22 of 58
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
D38N tKSI catalyzes the Kemp elimination of several benzisoxazoles.
The results reported above show that D38N tKSI catalyzes the Kemp elimination
of benzisoxazoles with a 5ꢀnitro (5NBI), 5ꢀbromo (5BrBI), and 4,6ꢀdichloro (4,6ꢀCl BI)
2
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
substituents. We included other benzisoxazoles to determine whether the trend in
reactivity parallels those observed for other systems in which a linear free energy
relationship (LFER) between the secondꢀorder rate constant of the reaction and the pKa
2, 6, 27, 46
of the cyanophenol product was established.
As shown in Figure 8, the rate
constant decreases as the pK increases, as observed in all the other systems. The
a
simplest interpretation is that the electronꢀwithdrawing groups on the ring facilitate
2
delocalization of charge in the transition state, thereby speeding up the reaction.
The calculated global slope (β ) of ꢀ0.96 in reactions of D38N tKSI is different
LG
6
from that of HG3.17 (ꢀ1.36) and more similar to the reported value for reaction
46
catalyzed by acetate in acetonitrile (ꢀ0.85). However, the latter β was calculated
LG
using the pK values of cyanophenols in water, and it has been pointed out that
a
correction of these values for the different solvent (acetonitrile) resulted in a much
47
shallower βLG of ꢀ0.51, which is instead similar to the β of ꢀ0.60 for reactions
LG
39
catalyzed by acetate in water. Although we cannot make any conclusion about the
nature of the environment in D38N tKSI active site by simply comparing βLG measured
in different systems and different solvents, our data show that D38N tKSI reacts with a
wide range of benzisoxazoles, suggesting a hydrophobic active site that can
accommodate all the tested substituents.
2
2
ACS Paragon Plus Environment

Page 23 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
DISCUSSION
The Kemp elimination reaction has fascinated researchers for decades.
47
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
Because of its absence from any metabolic pathway, it has been used as a benchmark
for the computational design of enzymes, with the assumption that high rate
accelerations corresponded to a success in design. Nevertheless, much simpler
systems such as organic solvents, micelles, and designed cavitands accelerate the
Kemp elimination by several orders of magnitude.
The D38N variant of tKSI contains a single carboxylate in a relatively
hydrophobic environment. This enzyme has been neither designed nor evolved to
catalyze the Kemp elimination, but nevertheless accelerates this reaction by several
orders of magnitude. At pH 7.0, 5NBI is converted to products by D38N with k /K =
cat
M
ꢀ1 ꢀ1
ꢀ1 ꢀ1
2,000 M s , while at pH 9.5 this value is 17,000 M s . By comparison (see Table 1),
the two most successful designed Kemp eliminases catalyze the reaction of the same
ꢀ1 ꢀ1
4
ꢀ1 ꢀ1
5
substrate with k /K values of 160 M s (KE59) and 1,300 M s (HG3). Importantly,
cat
M
all of these enzymes likely use a carboxylate to abstract the benzisoxazole proton,
making comparison between the three reactions meaningful.
At pH 7.5, the value of K for 5NBI is equal to (150 ± 40) ꢃM for reactions with
M
D38N (see Figure S5). This observation implies that D38N binds the Kemp elimination
6
substrate tighter than HG3 (K = 2.4 mM) and KE59 (K >1 mM) perhaps because of a
M
M
higher extent of hydrophobic interactions. The value of kcat for the same reaction of
ꢀ1
D38N KSI is 0.43 s (see Figure S5). This value is not very different from those
2
3
ACS Paragon Plus Environment

Biochemistry
Page 24 of 58
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 with the more active substrate 5ꢀNBI in the presence of the catalytic antibody
ꢀ1 20
ꢀ1 5, 6
3
4E4 (0.66 s ) and the designed Kemp eliminase HG 3 (1.7ꢀ3.0 s ) , but about
three orders of magnitude lower than the value measured for the evolutionallyꢀoptimized
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 6
HG3.17 (700 s ). This large difference likely reflects a significantly better alignment
between the base and the substrate achieved through the evolution process. It will be
interesting to determine whether KSI, an enzyme specialized to catalyze a different
reaction, possesses enough plasticity to be evolved to improve its turnover number in
the Kemp elimination.
In terms of k , the designed HG3 and HG3.17 provide rate enhancements of 3
cat
6
8
ꢀ6 ꢀ1 4
×
10 and 6 × 10 , respectively (k 1 × 10 s ). The HG3 rate enhancement is similar
non
to those measured for enzymes such as chorismate mutase, carbonic anhydrase, and
48
human cyclophilin, which lie in the lower portion of rate enhancements measured in
5
6
enzymes (10 ꢀ10 ), whereas the rate enhancement provided by HG3.17 starts to
9
approach those of more efficient enzymes such as triosephosphate isomerase (10 ).
The natural enzymes mentioned above accelerate reactions that are intrinsically fast,
with half lives of seconds to minutes for carbonic anhydrase and cyclophillin (knon = 10
ꢀ1
ꢀ1
ꢀ2 ꢀ1
s and 10 s , respectively) and hours for chorismate mutase and triosephosphate
ꢀ5
ꢀ1
ꢀ6 ꢀ1
isomerase (knon = 10 s and 10 s , respectively). As suggested for the Kemp
eliminase, it is possible that these enzymes are less precise and complex than the
enzymes that catalyze reactions with half lives longer than weeks. Alternatively (and
perhaps more likely) these enzymes may needed to select against alternative
substrates, so even enzymes with low rate enhancements may still be precise and
complex. The evolved HG3.17 seems to be able to discriminate between substrates
2
4
ACS Paragon Plus Environment

Page 25 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
with and without an electronꢀrich group at position 5 (ref. 6 and discussion below),
suggesting that evolution might have somewhat increased complexity of this enzyme.
However, because the tested functional groups at position 5 (NO , CN, Cl, and Br) differ
2
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
in shape and in charge distribution, this discrimination is likely to arise from nonꢀspecific
medium effects rather than from specific interactions.
Intriguingly, D38N tKSI catalyzes the Kemp elimination of the unsubstituted
ꢀ1 ꢀ1
benzisoxazole (BI) with a secondꢀorder rate constant of 39 M s (Figure 8), which is
ꢀ1 ꢀ1 6
about 100ꢀfold faster than the value reported for HG3.17 (0.34 M s ). This difference
is not dictated only by the difference in the βLG values of the two enzymes, which
predicts a ~3ꢀfold faster reaction for D38N tKSI. Instead, this difference is mainly due to
the outlier position of BI (and other substrates) in the LFER of HG3.17. As mentioned
above, it is possible that HG3.17 forms nonꢀspecific interactions with the 5ꢀposition of
the substrate, which would exclude or differentially position substrates without an
6
electronꢀrich group at position 5 in the active site of HG3.17. Thus, it seems that
HG3.17 possesses at least some binding pocket specificity that either was designed or
that arose through evolution, while D38N tKSI remains more promiscuous.
Nevertheless, the nonꢀactivated substrate BI would still react at least 3ꢀfold
intrinsically slower in HG3.17 than in D38N tKSI, suggesting that catalysis of nonꢀ
activated substrates is still a problem for the designed enzymes. Indeed, large and
negative β values, which are indicative of dramatic differences in reactivity between
LG
6
32
activated and nonꢀactivated substrates, were observed in both the HG and KE Kemp
eliminases (these βLG values range from ꢀ1.2 to ꢀ1.8). These differences are somehow
attenuated in D38N tKSI (β = ꢀ0.96, Figure 8), perhaps because of some general acid
LG
2
5
ACS Paragon Plus Environment

Biochemistry
Page 26 of 58
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
catalysis by an unknown residue or because of the differences in the active site
environment.
The pHꢀrate profiles for reactions of D38N tKSI (Figures 4 & 5) can be fit well to
Equation 1, which describe a single deprotonation event. The simplest interpretation is
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
that a single residue with pK ~ 7.8 needs to be deprotonated, probably to act as the
a
catalytic base. Based on mutational analysis and the simulations presented herein, we
propose that this residue is D99. Interestingly, the pK of this carboxylate is ~1ꢀ2 units
a
20
higher than the pK values measured for the catalytic antibody 34E4 of 6.0 and for the
a
5, 6, 49
designed Kemp eliminases.
(footnote 1) Because the Kemp elimination is very
3
sensitive to base strength (β ≈ 0.7), this difference in pK might contribute up to one
a
order of magnitude to the relative efficiency of D38N tKSI.
The pHꢀrate profile for wt tKSI (Figure 3) is also consistent with a single
deprotonation event, such as deprotonation of D38. In agreement with the model of
suboptimal positioning of this residue derived from docking simulations, the maximum
value of k /K is ~ 7000ꢀfold smaller than that of D38N tKSI (Table 2). In addition, the
cat
M
negative charge present on D38 might interfere with binding of electronꢀrich substrates
such as benzisoxazoles, as suggested by comparing the binding data in Figure 2 (wt
tKSI) and Figure S5 (D38N tKSI). A direct interaction between D38 and D99 seems
51, 52
less likely, as these residues are more than 5 Å apart in crystal structures.
Regardless of the mechanism(s) used to catalyze the Kemp elimination, the
observation that many proteins possess Kemp eliminase activity (Table 1) implies that
this reaction is easier to accelerate than many reactions catalyzed by enzymes. The
2
6
ACS Paragon Plus Environment

Page 27 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
high secondꢀorder rate constant for the Kemp elimination using D38N tKSI further
indicates that design is not necessary to obtain efficient Kemp eliminases. We suspect
that detection of Kemp eliminase activity of many proteins is limited by the high rates of
the background hydroxideꢀ and bufferꢀcatalyzed reactions, which can obscure detection
when protein concentrations are in the commonly used nanomolar and low micromolar
ranges. Nevertheless, not all hydrophobic proteins with cavities are suited for the Kemp
elimination. For example, the xylanase from Thermoascus aurantiacus, which was used
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
5
as the initial scaffold for the computationallyꢀdesigned HG Kemp eliminases, the T4
23
21
lysozyme, and calmodulin have no detectable Kemp eliminase activity. Thus, a
combination of the right base in the right environment is needed; nevertheless, these
requirements still lack the specificity requirement that define catalysis for most naturally
occurring enzymes.
In vitro evolution of designed Kemp eliminases has produced significantly better
variants, with higher rates and diminution of product inhibition. For example, the evolved
KE59.13 and HG3.17 are about two orders of magnitude more efficient than their parent
6
variants (Table 1), and product inhibition is at least qualitatively minimal in HG3.17.
Although no quantitation of this product inhibition was provided, we estimate the K to be
i
~1.9 mM (see Figure S4, panel c, which is derived from Figure 2d of reference 6), a
6
value similar to the measured K of 5NBI in this enzyme. Remarkably, we have shown
M
that a single mutation in the active site of tKSI produced a variant that catalyzes the
Kemp elimination with a maximum rate only one order of magnitude lower than that of
the best computationally designed enzyme, although product inhibition is somewhat
higher and the value of kcat is significantly lower than that of HG3.17. As mentioned
2
7
ACS Paragon Plus Environment

Biochemistry
Page 28 of 58
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
above, it will be intriguing to determine to what extent this enzyme can be evolved in
vitro.
In summary, we have shown that ketosteroid isomerase is a remarkable Kemp
eliminase, in spite of lack of design. Our results, in conjunction with many examples in
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
53
the literature, showed that the Kemp elimination is a simple reaction that needs
minimal catalytic features of a hydrophobic cleft with a positioned general base.
Although some specificity arose in course of the inꢀvitro evolution of the Kemp
eliminases, this specificity seems to be related to nonꢀspecific effects rather than to
precise interactions. Further, the reactions of nonꢀactivated substrates remain
problematic. For future challenges, computational design should seek more
sophisticated benchmark systems that increase specificity and catalytic complexity, or
use reactions which are slow in intrinsic nature and difficult to catalyze.
2
8
ACS Paragon Plus Environment

Page 29 of 58
Biochemistry
1
2
3
4
5
6
7
8
9
ACKNOWLEDGEMENTS.
We thank Carson Reed for the synthesis of 4,6ꢀdichlorobenzisoxazole.
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
SUPPORTING INFORMATION
Supporting information available: Figure S1, which represents a schematic view of KSI
active site; Figure S2, which shows the expected pHꢀrate profiles for wt tKSI in case
D99 is equally or less efficient than it is in D38N tKSI; Figure S3, which shows equilenin
inhibition of wt tKSI; Figure S4, which shows the effect of product inhibition on D38N
tKSI and on HG3.17, and Figure S5, which shows binding of 5ꢀNBI to D38N tKSI.
2
9
ACS Paragon Plus Environment