Enthalpy-Entropy Compensation in the Binding of Modulators at Ionotropic Glutamate Receptor GluA2

Article abstract of DOI:10.1016/j.bpj.2016.04.032

The 1,2,4-benzothiadiazine 1,1-dioxide type of positive allosteric modulators of the ionotropic glutamate receptor A2 (GluA2) are promising lead compounds for the treatment of cognitive disorders, e.g., Alzheimer's disease. The modulators bind in a cleft

Full text of DOI:10.1016/j.bpj.2016.04.032

Article
Enthalpy-Entropy Compensation in the Binding of
Modulators at Ionotropic Glutamate Receptor GluA2
1
2
1
1
1
1
Christian Krintel, Pierre Francotte, Darryl S. Pickering, Lina Juknait e_ , Jacob Pøhlsgaard, Lars Olsen,
1
2
2
Karla Frydenvang, Eric Goffin, Bernard Pirotte, and Jette S. Kastrup
Department of Drug Design and Pharmacology, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark;
1,
*
1
2
and Department of Medicinal Chemistry, Center for Interdisciplinary Research on Medicines (CIRM), University of Li e` ge, Li e` ge, Belgium
ABSTRACT The 1,2,4-benzothiadiazine 1,1-dioxide type of positive allosteric modulators of the ionotropic glutamate receptor
A2 (GluA2) are promising lead compounds for the treatment of cognitive disorders, e.g., Alzheimer’s disease. The modulators
bind in a cleft formed by the interface of two neighboring ligand binding domains and act by stabilizing the agonist-bound open-
channel conformation. The driving forces behind the binding of these modulators can be significantly altered with only minor
substitutions to the parent molecules. In this study, we show that changing the 7-fluorine substituent of modulators BPAM97
(
2) and BPAM344 (3) into a hydroxyl group (BPAM557 (4) and BPAM521 (5), respectively), leads to a more favorable binding
enthalpy (DH, kcal/mol) from ꢀ4.9 (2) and ꢀ7.5 (3) to ꢀ6.2 (4) and ꢀ14.5 (5), but also a less favorable binding entropy (ꢀTDS,
kcal/mol) from ꢀ2.3 (2) and ꢀ1.3 (3) to ꢀ0.5 (4) and 4.8 (5). Thus, the dissociation constants (K , mM) of 4 (11.2) and 5 (0.16) are
d
similar to those of 2 (5.6) and 3 (0.35). Functionally, 4 and 5 potentiated responses of 10 mM L-glutamate at homomeric rat
i
GluA2(Q) receptors with EC50 values of 67.3 and 2.45 mM, respectively. The binding mode of 5 was examined with x-ray
crystallography, showing that the only change compared to that of earlier compounds was the orientation of Ser-497
pointing toward the hydroxyl group of 5. The favorable enthalpy can be explained by the formation of a hydrogen bond from
the side-chain hydroxyl group of Ser-497 to the hydroxyl group of 5, whereas the unfavorable entropy might be due to desolva-
tion effects combined with a conformational restriction of Ser-497 and 5. In summary, this study shows a remarkable example of
enthalpy-entropy compensation in drug development accompanied with a likely explanation of the underlying structural
mechanism.
INTRODUCTION
2
-Amino-3-(3-hydroxy-5-methylisoxazol-4-yl)propionic
toward D1 (3). This movement creates a conformational
strain on the TMD and promotes channel opening. Two
mechanisms can lead to channel closure: dissociation of
glutamate from the LBD leads to deactivation, whereas re-
arrangement of the D1:D1 interface with L-glutamate still
present in the LBD leads to desensitization (4). Positive
allosteric modulators bind in a crevice created by the
D1:D1 interface and act by strengthening the agonist-bound
state of the receptor attenuating deactivation or by strength-
ening the interface and thereby slowing desensitization
(4,5). Stabilization of the D1:D1 interface can also be
accomplished by mutation of a key leucine located at the
dimer interface (Leu-483 in the AMPA receptor ionotropic
glutamate receptor A2 (GluA2); numbering without signal
peptide) to tyrosine, which dramatically slows receptor
desensitization (6). Furthermore, this mutation produces a
dimeric LBD in solution without altering the modulator
binding pocket (4,7).
acid (AMPA) receptors mediate most fast excitatory synap-
tic transmission in the central nervous system (1). They are
tetrameric ligand-gated ion channels with each subunit con-
sisting of a flexible intracellular C-terminal domain, a heli-
cal transmembrane domain (TMD), an extracellular ligand
binding domain (LBD), and an N-terminal domain (2).
Although, in the intact receptor, the TMD has fourfold sym-
metry, the LBDs and N-terminal domains are arranged with
twofold symmetry as dimers of dimers. Each LBD consists
of the D1 and D2 subdomains with adjacent D1 domains
creating a dimeric interface. The receptor is activated by
L-glutamate binding in a pocket between the D1 and D2 do-
mains, leading to a conformational change where D2 moves
Submitted December 1, 2015, and accepted for publication April 18, 2016.
*Correspondence: jsk@sund.ku.dk
Christian Krintel and Pierre Francotte contributed equally to this work.
Editor: Axel Brunger.
The modulator BPAM97 (2) (Fig. 1) was developed on
the basis of the earlier weaker modulators diazoxide and
http://dx.doi.org/10.1016/j.bpj.2016.04.032
Ó 2016 Biophysical Society.
Biophysical Journal 110, 2397–2406, June 7, 2016 2397

Krintel et al.
FIGURE 1 Key structures of 1,2,4-benzothiadia-
zine 1,1-dioxide modulators.
IDRA21 (1) (Fig. 1) and showed a 70-fold increase in
the binding affinity toward the dimeric LBD construct
of ionotropic glutamate receptor A2 (GluA2) (GluA2
LBD-L483Y-N754S) over 1 (7–10). As for 1, two
molecules of 2 bind in the D1:D1 interface where each
molecule interacts with Pro-494, Phe-495, Met-496,
and Ser-497 across the dimer interface (7,11). Structural
analyses suggested that the major reason for the increase
in affinity is van der Waals interactions created by
the 4-ethyl moiety to Pro-494, Phe-495, and Met-496.
This was corroborated by the later development of
BPAM344 (3) (see Fig. 1, in which the ethyl moiety
was altered to a cyclopropyl, further increasing the affin-
GluA2 LBD-L483Y-N754S, while retaining binding affin-
ities similar to those of 2 and 3.
MATERIALS AND METHODS
Synthesis of the target compounds BPAM557 (4)
and BPAM521 (5)
All commercial chemicals and solvents were reagent grade and used
without further purification. Melting points were determined on a Stuart
SMP3 apparatus (Barloworld Scientific France SAS, Nemours, France) in
open capillary tubes and are uncorrected. NMR spectra were recorded on
1
13
a Bruker Avance 500 spectrometer ( H: 500 MHz; C: 125 MHz; Bruker
Daltonik, Bremen, Germany) using dimethylsulfoxide-d (DMSO-d ) as
6
6
solvent and tetramethylsilane as the internal standard; chemical shifts are
reported in d values (ppm) relative to internal tetramethylsilane. The abbre-
viations s, singlet; d, doublet; t, triplet; q, quadruplet; m, multiplet; and bs,
broad signal are used throughout. Elemental analyses (C, H, N, S) were car-
ried out on a Thermo Flash EA 1112 series elemental analyzer (Thermo
Electron, Milan, Italy). All reactions were followed by thin layer chroma-
tography (silica gel 60F254, Merck, Darmstadt, Germany) and visualization
was accomplished with ultraviolet light (254 or 366 nm).
ity almost 20-fold to a K of 350 nM (12)). Intriguingly,
d
in the structures of both 2 and 3 in complex with GluA2
LBD-L483Y-N754S it was clear that Ser-497 of the re-
ceptor adopts two different side-chain conformations of
which one interacts with the 7-fluorine of the modulators.
In addition, a recent study showed that Ser-497 in GluA2
LBD-L483Y-N754S with BPAM37 bound only adopts a
single conformation pointing toward the modulator,
despite the lack of a substituent in the 7-position of the
modulator (13).
4
1
-Cyclopropyl-7-methoxy-4H-1,2,4-benzothiadiazine
,1-dioxide (9)
A
solution of 2-amino-5-methoxybenzenesulfonamide (6) (1.0 g;
In this study, the 7-fluorine of 2 and 3 was substituted to
a hydroxyl group to create the new modulators BPAM557
4.9 mmol) (14) in methanol (20 ml) was supplemented with (1-ethoxycy-
clopropyloxy)trimethylsilane (4 ml) and glacial acetic acid (4 ml) and re-
fluxed for 18 h. The reaction mixture was then evaporated to dryness
under reduced pressure. The resulting oily residue was treated with
water (30 ml) and extracted with chloroform (3 ꢁ 30 ml). The organic
(
4) and BPAM521 (5), respectively (see Fig. 1). It was
hypothesized that the hydroxyl group would form favorable
hydrogen bonding to Ser-497. Following synthesis, the
two new modulators were characterized using two-electrode
voltage clamp (TEVC) electrophysiology, x-ray structure
determination, and isothermal titration calorimetry (ITC).
We show that the introduction of a hydroxyl group
instead of fluorine in the 7-position of the modulator
dramatically alters the thermodynamics of binding at
4
layers were collected and dried over anhydrous MgSO . The filtrate
was evaporated to dryness and the resulting oily residue, consisting of
crude 5-methoxy-2-(1-methoxycyclopropylamino)benzenesulfonamide (7)
(
transacetalization reaction in methanol leading to the methoxy-substituted
compound), was used in the next step without further purification.
The solution of 7 in tetrahydrofuran (50 ml) was supplemented with so-
dium borohydride (2.0 g, 52.9 mmol) and boron trifluoride diethyl etherate
(2 ml) and refluxed for 18 h. The solvent was then removed by distillation
2
398 Biophysical Journal 110, 2397–2406, June 7, 2016

GluA2: Enthalpy/Entropy Compensation
under reduced pressure and the residue was treated with water (30 ml). The
aqueous suspension was adjusted to pH 4 by means of 6N HCl and then
extracted with chloroform (3 ꢁ 30 ml). The organic layers were collected
3,4-Dihydro-4-ethyl-7-hydroxy-2H-1,2,4-benzothiadiazine
1
,1-dioxide (4)
and dried over anhydrous MgSO
4
. The filtrate was concentrated to dryness
The title compound was prepared as described for 5 starting from 3,4-dihy-
dro-4-ethyl-7-methoxy-2H-1,2,4-benzothiadiazine 1,1-dioxide obtained as
under reduced pressure and the resulting oily residue, consisting of crude
-cyclopropylamino-5-methoxybenzenesulfonamide (8), was used in the
next step without further purification.
ꢂ
1
2
previously described (8). White solid: mp 201–203 C; H NMR (DMSO-
) d 1.06 (t, J ¼ 7 Hz, 3H, CH CH ), 3.36 (m, 2H, CH CH ), 4.57 (s,
2H, 3-CH
), 6.80 (d, J ¼ 8.8 Hz, 1H, 5-H), 6.88-6.91 (m, 2H, 6-H/8-H),
7.85 (bs, 1H, NH), 9.25 (bs, 1H, OH). C NMR (DMSO-d
(CH CH ), 43.7 (CH CH ), 60.4 (C-3), 109.3 (C-8), 116.2 (C-5), 121.6
(C-6), 123.5 (C-8a), 136.1 (C-4a), 148.1 (C-7). Anal. (C S) theo-
d
6
2
3
2
3
The mixture of 8 in triethyl orthoformate (10 ml) was heated in an open
ꢂ
2
1
3
vessel at 150 C for 6 h. The resulting suspension was cooled on an ice
6
) d 11.8
bath and the insoluble material was collected by filtration, washed with
diethyl ether, and dried. The solid was dissolved in a hot mixture of
acetone and methanol and the solution was treated with charcoal,
filtered, and concentrated to dryness. The residue of the title compound
2
3
2
3
9 12 2 3
H N O
retical: C, 47.35; H, 5.30; N, 12.27; S, 14.04. Found: C, 47.45; H, 5.33; N,
12.34; S, 13.72.
(
9) was purified by crystallization in methanol (overall yields: 20–30%).
ꢂ
1
White solid: mp 216–219 C; H NMR (DMSO-d
CH(CH ), 1.15 (m, 2H, CH(CH ), 3.37 (m, 1H, CH(CH
), 7.30 (d, J ¼ 2.9 Hz, 1H, 8-H), 7.41 (dd, J ¼ 9.3 Hz/2.9 Hz,
6
) d 1.00 (m, 2H,
Pharmacology
2
)
2
2
)
2
2 2
)
), 3.87 (s,
3
H, OCH
H, 6-H), 7.80 (d, J ¼ 9.3 Hz, 1H, 5-H), 8.07 (s, 1H, 3-H). C NMR
) d 7.2 (CH(CH ), 32.1 (CH(CH ), 56.0 (OCH ), 106.0
C-8), 118.9 (C-5), 121.0 (C-6), 123.2 (C-8a), 130.1 (C-4a), 150.6 (C-3),
57.7 (C-7).
3
TEVC functional responses
1
3
1
(
(
DMSO-d
6
2
)
2
2
)
2
3
Recombinant rat GluA2(Q) cRNA was transcribed in vitro (AmpliCap-
i
Max T7, Cellscript, Madison, WI) and microinjected into Xenopus lævis
1
oocytes. Oocytes were used for TEVC recordings 2–5 days postinjection.
Recordings were made at room temperature at holding potentials in the
range of ꢀ15 to ꢀ70 mV, where the oocytes were continuously superfused
4-Cyclopropyl-3,4-dihydro-7-methoxy-2H-1,2,4-benzothia-
diazine 1,1-dioxide (10)
2
þ
with Ca -free frog Ringer’s solution (115 mM NaCl, 2 mM KCl, 1.8 mM
2
BaCl , 5 mM HEPES, pH 7.6). Potentiating compounds were prepared
2
þ
The solution of 4-cyclopropyl-7-methoxy-4H-1,2,4-benzothiadiazine 1,1-
dioxide (9) (0.2 g, 0.79 mmol) in isopropanol (5 ml) was supplemented
with finely divided sodium borohydride (0.1 g, 2.64 mmol) and the
in Ca -free frog Ringer’s solution containing 10 mM L-glutamate.
DMSO was used to dissolve 4, which was then diluted in Ringer’s/Glu so-
lution to 0.2% (v/v) DMSO at 1 mM compound. Compounds were applied
stepwise by bath application for 30–60 s until a plateau response was ob-
tained at each concentration. Data were fit to a logistic equation to deter-
ꢂ
mixture was heated at 50 C for 5–10 min. The solvent was removed by
distillation under reduced pressure and the residue was treated with
water (10 ml) and adjusted to pH 4 by means of 6N HCl. The suspension
was extracted with methylene chloride (3 ꢁ 15 ml). The organic layers
mine the EC50 and n
H
using GraphPad Prism v6 (GraphPad Software,
San Diego, CA).
4
were dried over MgSO and filtered. The filtrate was evaporated to dryness
and the solid residue of the crude compound 10 was crystallized in meth-
ꢂ
1
anol/water 1:1 (yields: 70–80%). White solid: mp 154–156 C; H NMR
Expression, purification, and x-ray structure
determination
(
(
DMSO-d
6
) d 0.62 (m, 2H, CH(CH
)
2
)
2
), 0.87 (m, 2H, CH(CH
), 3.73 (s, 3H, OCH ), 4.59 (s, 2H, 3-CH ), 7.04 (d,
2 2
) ), 2.42
m, 1H, CH(CH
2
2
3
2
J ¼ 3.0 Hz, 1H, 8-H), 7.12 (dd, J ¼ 9.2 Hz/3.0 Hz, 1H, 6-H), 7.27 (d,
Protein expression, purification, and cocrystallization
1
3
J ¼ 9.2 Hz, 1H, 5-H), 7.89 (s, 1H, NH). C NMR (DMSO-d
CH(CH ), 29.8 (CH(CH ), 55.7 (OCH ), 61.2 (C-3), 107.4 (C-8),
16.5 (C-5), 120.7 (C-6), 123.5 (C-8a), 138.6 (C-4a), 151.1 (C-7). Anal.
S) theoretical: C, 51.95; H, 5.55; N, 11.02; S, 12.61. Found:
C, 51.76; H, 5.54; N, 10.83; S, 12.39.
6
) d 8.3
The rat GluA2 LBD-L483Y-N754S protein was expressed and purified
as previously described (7). Cocrystals with 5 were obtained by the
(
2
)
2
2
)
2
3
1
ꢂ
hanging drop vapor diffusion technique at 7 C. The drops consisted of 1
11 14 2 3
(C H N O
ml of protein solution and 1 ml of reservoir solution. The protein
solution contained 5.7 mg/ml GluA2 LBD-L483Y-N754S in 10 mM
Hepes, 20 mM NaCl, 1 mM EDTA, and 5 mM L-glutamate. For cocrystal-
lization experiments 0.5 mg of 5 was added to 150 ml of protein
solution and the mixture was incubated for 24 h before crystallization.
Crystallization conditions were 20% PEG4000, 0.3 M lithium sulfate,
and 0.1 M phosphate-citrate, pH 4.5. Crystals were briefly transferred to
reservoir buffer containing 20% glycerol and then flash cooled in liquid
nitrogen.
4-Cyclopropyl-3,4-dihydro-7-hydroxy-2H-1,2,4-benzothiadia-
zine 1,1-dioxide (5)
The solution of 4-cyclopropyl-3,4-dihydro-7-methoxy-2H-1,2,4-benzothia-
diazine 1,1-dioxide (10) (0.1 g, 0.39 mmol) in chloroform (6 ml) was cooled
on an ice bath and then supplemented with boron tribromide (0.3 ml). After
20 h stirring, the reaction mixture was carefully poured on water (10 ml)
and the organic solvent was removed by distillation under reduced pressure.
Data collection and structure determination
The aqueous layer was extracted with ethyl acetate (3 ꢁ 40 ml). The
collected organic layers were dried over MgSO
4
and filtered. The filtrate
Diffraction data were collected at beamline I911-3 at MAX-lab
(Lund, Sweden) (15). The data were processed using iMosflm and
Scala (16,17) and the structure solved by molecular replacement using
Phaser (18) with the structure of GluA2 LBD-L483Y-N754S with 2
(Protein Data Bank (PDB): 3TDJ, molA) as the search model (7).
Further model building was done in Coot (19) and refinements were per-
formed using Phenix (20). Topology and parameter files for ligands
were obtained using the Schr o¨ dinger software (Maestro, version 9.1;
Schr o¨ dinger, LLC, New York, NY). D1-D2 domain closures were
calculated relative to the apo structure (PDB: 1FTO, molA) (3), using
DynDom (21). Three-dimensional-structure figures were prepared using
The PyMOL Molecular Graphics System (version 1.7.2.3, Schr o¨ dinger,
LLC).
was evaporated to dryness and the residue was dissolved in ethyl acetate
2 ml). The addition of hexane (10 ml) gave rise to the precipitation of
the title compound, which was collected by filtration, washed with hexane,
(
ꢂ
1
and dried (yields: 80%). White solid: mp 187–189 C; H NMR (DMSO-d
d 0.60 (m, 2H, CH(CH ), 0.84 (m, 2H, CH(CH ), 2.37 (m, 1H,
CH(CH ), 4.54 (s, 2H, 3-CH
), 6.91 (d, J ¼ 2.7 Hz, 1H, 8-H), 6.93 (dd,
6
)
2
)
2
2 2
)
2
)
2
2
J ¼ 8.9 Hz/2.9 Hz, 1H, 6-H), 7.18 (d, J ¼ 8.9 Hz, 1H, 5-H), 7.83 (bs,
13
1
2
1
H, NH), 9.34 (bs, 1H, OH). C NMR (DMSO-d
9.9 (CH(CH ), 61.3 (C-3), 109.2 (C-8), 116.5 (C-5), 121.1 (C-6),
23.8 (C-8a), 137.3 (C-4a), 149.1 (C-7). Anal. (C10 S) theoretical:
6 2 2
) d 8.2 (CH(CH ) ),
2 2
)
12 2 3
H N O
C, 49.99; H, 5.03; N, 11.66; S, 13.34. Found: C, 49.83; H, 5.07; N, 11.69; S,
3.02.
1
Biophysical Journal 110, 2397–2406, June 7, 2016 2399

Krintel et al.
FIGURE 2 Synthesis of compounds. Reagents: i,
1-ethoxycyclopropyloxy)trimethylsilane, HOAc,
MeOH; ii, NaBH , BF .Et O, THF; iii, HC(OEt)
iv, NaBH , isopropanol; v, BBr , CHCl . Com-
pound 5 corresponds to BPAM521.
(
4
3
2
3
;
4
3
3
ITC
cyclopropyloxy)trimethylsilane to form the hemiaminal
ether intermediate 7. A transacetalization reaction in meth-
anol occurred leading to the methoxy-substituted com-
pound 7 instead of its ethoxy-substituted counterpart.
Compound 7 further reacted with boron trifluoride diethyl
etherate to provide the key intermediate 8, which was
engaged in the ring closure reaction with triethyl orthofor-
mate to form the 4H-1,2,4-benzothiadiazine 1,1-dioxide 9.
Saturation of the double bond at the 2,3-positions of 9
was performed by the reaction of the latter with sodium
borohydride in isopropanol giving rise to the 3,4-dihy-
dro-2H-1,2,4-benzothiadiazine 1,1-dioxide 10. The me-
thoxy group at the 7-position of 10 was converted into
the hydroxy group after its reaction with boron tribromide
in chloroform, leading to the final target compound 5. To
ꢂ
ITC was performed at 25 C using an ITC200 microcalorimeter (GE
Healthcare) with a cell volume of 200 ml for 4 and a VP-ITC microcalo-
rimeter (GE Healthcare) with a cell volume of 1.4 ml for 5. Before the ex-
periments, the protein was dialyzed twice against 100 mM Hepes, 100 mM
NaCl, 2 mM KCl, and 5 mM L-glutamate, pH 7.0 and ligands were dis-
solved in the same buffer. For measurement of 4 binding, the dimeric pro-
tein concentration was 45 mM and a concentration of 1.5 mM of 4 was
used for titration. For measurement of 5 binding, the dimeric protein con-
centration was 30 mM and a concentration of 0.6 mM of 5 was used for
titration. For 4, the experiments were performed as 20 injections with
3
-min intervals, using 0.4 ml in the first injection and 2 ml in the following
injections. For 5, the experiments were performed as 20 injections with
-min intervals, using 2 ml in the first injection and 15 ml in the following
3
injections. Data analysis was performed with the Origin 7.0 software (Ori-
ginLab, Northampton, MA) using a single binding site model. The first
data point was discarded. The reported DH, ꢀTDS, and K
d
are mean
values of three experiments. The protein concentration was determined
by ultraviolet absorption.
Calculation of solvation energy
Structures of 2–5 were optimized in the gas phase with B3LYP/6-31G**
and the solvation energies determined with a single point calculation using
the SM8 solvation model (22). The calculations were carried out with
Jaguar (Jaguar, version 8.8, Schr o¨ dinger, LLC, 2015).
RESULTS AND DISCUSSION
Synthesis and functional characterization of
BPAM557 (4) and BPAM521 (5)
The synthetic pathway to the 7-hydroxy-substituted 3,4-di-
hydro-2H-1,2,4-benzothiadiazine 1,1-dioxide BPAM521
FIGURE 3 Potentiation of 10 mM L-glutamate responses by BPAM557
(
i
4) and BPAM521 (5) at homomeric rat GluA2(Q) expressed in X. lævis
(
5) is illustrated in Fig. 2. Starting from 2-amino-5-me-
oocytes and recorded by TEVC. Shown are mean 5 SE of pooled data
from 7 to 8 experiments normalized to the maximum current response of
each oocyte. Insets: sample traces from representative experiments for 4
thoxybenzenesulfonamide (6) obtained as previously des-
cribed (14) the introduction of a cyclopropyl group on
the amine function of 6 was achieved by using (1-ethoxy-
(V
h
¼ ꢀ20 mV) and 5 (V ¼ ꢀ25 mV). The scale bars ¼ 200 nA and 1 min.
h
2
400 Biophysical Journal 110, 2397–2406, June 7, 2016

GluA2: Enthalpy/Entropy Compensation
FIGURE 4 Structure of GluA2 LBD in complex
with the neurotransmitter L-glutamate and the
positive allosteric modulator BPAM521 (5). (A)
Position of 5 (cyan at dimer interface) and L-gluta-
mate (green in the middle of the GluA2 LBD) in
the GluA2 LBD-L483Y-N754S dimer (gray). (B)
Ser-497 of molA (upper) and molB (lower) shown
with the interacting modulator molecule. A simu-
lated annealing 2 F
1
o
-F
c
OMIT map carved at
˚
.5 A is shown at sigma level 1. (C) Possible polar
contacts to Pro-494 and Ser-497 (salmon in sticks
presentation) are indicated by black dashed lines.
To see this figure in color, go online.
prepare compound BPAM557 (4) (the 4-ethyl-substituted
analog of 5), the previously described 3,4-dihydro-4-
ethyl-7-methoxy-2H-1,2,4-benzothiadiazine 1,1-dioxide
of binding sites, which is in agreement with recent studies
showing a second modulator binding site to produce both
an increase in positive cooperativity and a decrease in the
EC50 for dimerization (23).
1
7 (the 4-ethyl-substituted analog of 10) (8) was treated
with boron tribromide according to the experimental condi-
tions applied to compound 10.
The synthesized compounds 4 and 5 were evaluated as
AMPA receptor potentiators in a voltage clamp assay
Structure determination of GluA2 LBD-L483Y-
N754S in complex with BPAM521 (5)
(
GluA2(Q) . In the potentiation of 10 mM L-glutamate res-
TEVC) on Xenopus oocytes expressing recombinant rat
The purpose of cocrystallizing BPAM521 (5) with the
GluA2 LBD was to investigate the impact of exchanging
the fluorine atom in the previously published modulator
BPAM344 (3) with a hydroxyl group on the binding mode
and in particular on the orientation of Ser-497.
GluA2 LBD-L483Y-N754S in complex with L-glutamate
and 5 crystallized as a dimer (Fig. 4 A; Table 1) identical to
other glutamate bound GluA2 LBD structures. This mutant
i
ponses, 5 was the more potent compound (mean 5 SE)
EC50 ¼ 2.45 5 0.23 mM (n ¼ 8), whereas 4 was 27-fold
less potent, EC ¼ 67.3 5 5.1 mM (n ¼ 7). Both
5
0
compounds exhibited steep concentration-response curves
(
Fig. 3) with Hill slopes near 2 (5, n ¼ 2.15 5 0.25; 4,
H
H
n ¼ 1.90 5 0.03). Thus, the Hill slope indicates two pairs
Biophysical Journal 110, 2397–2406, June 7, 2016 2401

Krintel et al.
TABLE 1 X-Ray Data Collection and Refinement Statistics of
GluA2-LBD-L483Y-N754S in Complex with L-glutamate and
BPAM521 (5)
Data Collection
Space group
˚
a, b, c (A)
1 1
P2 2 2
98.79, 122.48, 47.54
34.25–6.07 (2.02–1.97)
5.1 (5.1)
a
˚
Resolution (A)
Redundancy
Completeness (%)
99.9 (99.9)
8.8 (26.0)
5.2 (1.9)
b
Rmerge (%)
I/sigmaI
Refinement
c
R
work/Rfree (%)
16.2/21.0
4836
263/263
No. of nonhydrogen atoms
No. of residues (molA/B)
No. of L-glutamate/BPAM521(5)/
water/glycerol/acetate/chloride/
sulfate/PEG
2/2/444/4/9/3/4/2
RMS Deviations
˚
Bond lengths (A)
0.007
1.0
100
Bond angles (deg)
No. residues in allowed areas
d
of Ramachandran plot (%)
2
Average B Values (A )
˚
molA/B
18.5/14.3
9.9/4.5/24.9/39.0/41.2/50.5/41.2/43.4
L-glutamate/BPAM521(5)/
water/glycerol/acetate/
chloride/sulfate/PEG
a
Values for the outermost resolution shell are denoted in parentheses.
b
R
intensity of an individual measurement of the reflection (I
merge ¼ Shkl
S
i
jI
i
(hkl) ꢀ I(hkl)j/Shkl
S
i
jI (hkl)j, with Miller indices hkl, the
i
i
(hkl)), and the in-
tensity from multiple observations (I(hkl)).
c
R
work ¼ ShkljF
o
ꢀ F
c
j/ShkljF
o
j, with observed (jF
o
j) and calculated (jF
c
j)
structure factor amplitudes. Rfree is calculated with 5% reflections omitted
from the refinement process.
d
PROCHECK (26) was used to calculate the Ramachandran plot.
construct was used as it has previously been shown to form a
dimer in solution without altering the modulatory site (7).
Two molecules of 5 could be unambiguously fitted into elec-
tron densities found at the dimeric interface between the two
D1 neighboring subdomains (Fig. 4 B). The D1-D2 domain
closure was found to be 20.5 (molA) and 19.2 (molB),
which is similar to other structures with L-glutamate (3).
The modulator molecules adopt a binding mode almost
identical to the one seen for the previously published com-
pound 3 (12). As with 3, the N4 cyclopropyl moiety of 5
makes nonpolar contacts with backbone atoms of Phe-495
and Met-496 and a hydrogen bond is formed between the
N2 secondary amide of 5 and the carbonyl oxygen of Pro-
ꢂ
ꢂ
FIGURE 5 Comparison of GluA2 LBD structures with modulators. (A)
Conformations of Ser-497 in GluA2 with different modulators. An overlay
of the D1 subdomains of the GluA2 LBD in complex with BPAM521 (5;
cyan), BPAM97 (2; magenta; PDB: 3TDJ), BPAM344 (3; orange; PDB:
4N07), BPAM37 (12; dark gray; PDB: 4U4X), BPAM25 (11; salmon;
PDB: 4U4S), and IDRA21 (1; green; PDB: 3IL1). Modulators and the cor-
responding Ser-497 are shown in stick representation. (B) Overlay of the D1
subdomains of GluA2 LBD in complex with 5, 3, and 12 (black). Modula-
tors and their corresponding Ser-497 are shown in stick representation.
˚
Water molecules at a distance up to 4.0 A of the Ser-497 hydroxyl group
in both conformations and the hydroxyl group of 5 are shown as spheres.
4
94 (Fig. 4 C). The most important difference between these
(
C) Overlay of GluA2 LBD structures with 1 and 5. Only the two modula-
tors and Ser-497 are shown. To see this figure in color, go online.
two structures is the orientation of Ser-497. In the complex
with 3 the side chain of Ser-497 adopts two different confor-
mations, one pointing toward and one away from 3. However,
in the GluA2 LBD structure with 5 the side chain of Ser-497
is seen predominantly in the conformation pointing toward 5
from 5 was refined to occupancy of 0.13 (molA) and 0.17
(molB) and was therefore not included in the structure.
This allows the formation of a hydrogen bond between the
(
Fig. 4, B and C). The conformation of Ser-497 pointing away
2
402 Biophysical Journal 110, 2397–2406, June 7, 2016

GluA2: Enthalpy/Entropy Compensation
FIGURE 6 ITC studies of BPAM521 (5) and BPAM557 (4) binding to the GluA2 LBD-L483Y-N754S. Raw data (top panels) and isotherms (bottom
panels) are presented. The graphs show that heat is developed after each injection and that the signal is diminished when the protein becomes saturated
with ligand. (A) Titration of 5 into GluA2 LBD-L483Y-N754S. (B) Titration of 4 into GluA2 LBD-L483Y-N754S.
hydroxyl group of 5 and the hydroxyl group of Ser-497. Of all
available structures of GluA2 LBD in complex with 1,2,4-
benzothiadiazine 1,1-dioxides it is only in the complexes
with 5 and BPAM37 (12) (see Fig. 7) that Ser-497 adopts
this single conformation pointing toward the modulator
bond from the hydroxyl group in 5 to the side-chain hydroxyl
group of Ser-497. Polar contacts are likely to be reflected in
enthalpic contributions to binding energies and we therefore
measured the thermodynamics of binding using ITC (Fig. 6).
The GluA2 LBD-L483Y-N754S was used for studying
modulator binding to GluA2 as it is dimeric in solution and
presents a preformed modulator binding pocket, thereby
isolating the event of modulator binding from that of
dimerization (7). Even though the binding affinity of 5
(K ¼ 0.16 mM) was found to be similar to that of 3 (K ¼
(
Fig. 5 A). This conformation of Ser-497 is also the predom-
inant conformation seen when glutamate and no modulator is
bound (GluA2 flop: PDB: 1FTJ and 3DP6; GluA2 flip:
2
UXA). Despite different conformations of Ser-497 in
GluA2 upon binding of either 12 or 5 versus 3, a similar
network of solvent molecules are present in the modulator
binding pocket in all three structures (Fig. 5 B). Interestingly,
in this context compound 12 is very different from 5 by lack-
ing a 7-position substituent and it is therefore not able to
directly interact with Ser-497. Therefore, the observation
that the orientation of Ser-497 is the same in the complex
with 12 as well as in structures void of modulators, suggests
that this orientation of Ser-497 is likely to be due to lack of
repulsion by the modulator.
d
d
0.35 mM (12)), dramatically altered enthalpic and entropic
contributions to the binding energy were seen (Table 2).
Compared to 3, the new modulator 5 provides a strong
gain in binding enthalpy (5: DH ¼ ꢀ14.5 kcal/mol, 3:
DH ¼ ꢀ7.5 kcal/mol (12)), but also an equally strong unfa-
vorable entropy compensation (5: ꢀTDS ¼ 4.8 kcal/mol, 3:
ꢀTDS ¼ ꢀ1.3 kcal/mol (12)). This pattern was found to
be the same for BPAM557 (4) (K ¼ 11.2 mM, DH ¼
d
ꢀ6.2 kcal/mol, ꢀTDS ¼ ꢀ0.5 kcal/mol) and the related com-
pound BPAM97 (2) (K ¼ 5.6 mM, DH ¼ ꢀ4.9 kcal/mol,
d
ꢀ
TDS ¼ ꢀ2.3 kcal/mol) (7). This shows that introduction
Thermodynamics of binding using isothermal
titration calorimetry
of the 7-hydroxyl group alters the thermodynamics so that
binding of 4 and 5 is mainly driven by enthalpy.
The major difference in binding mode of BPAM521 (5) and
BPAM344 (3) in GluA2 LBD is the formation of a hydrogen
An overview of the thermodynamic contributions to the
binding energy of all BPAMs reported to date is shown in
Biophysical Journal 110, 2397–2406, June 7, 2016 2403

Krintel et al.
TABLE 2 Isothermal Titration Calorimetry Using GluA2 LBD-
L483Y-N754S
fold reorientation of 1 compared to the BPAMs, leading to a
location of the chlorine atom of 1 in the direction of the
6-position of BPAMs (Fig. 5 C). However, Ser-497 is also
likely to be less fixed with a lesser loss of disorder.
A comparison of the binding mode of 5 to that of 3 shows
that the favorable enthalpy most likely arises from the
hydrogen bond between the 7-hydroxyl group and Ser-
DH
(kcal/mol)
ꢀTDS
a
Compound
K
d
(mM)
(kcal/mol)
N
b
BPAM521 (5) 0.16 5 0.01 ꢀ14.5 5 0.2 4.8 5 0.2 2.2 5 0.03
c
BPAM344 (3) 0.35 5 0.02 ꢀ7.5 5 0.2 ꢀ1.3 5 0.2 2.7 5 0.1
BPAM557 (4) 11.2 5 0.01 ꢀ6.2 5 0.3 ꢀ0.5 5 0.3 3.6 5 0.08
d
BPAM97 (2)
5.6 5 0.9
ꢀ4.9 5 0.4 ꢀ2.3 5 0.4 3.0 5 0.2
4
4
97. On the other hand, differences in binding affinities of
and 5 versus those of their fluorinated counterparts 2
a
Sites per LBD dimer.
b
Standard deviation.
Values from (12).
and 3 are marginal. This is due to less favorable binding en-
tropies for 4 and 5. Two major terms are important for the
entropy of binding: 1) change in solvent entropy arising
from desolvation of the modulator and receptor binding
site upon binding and 2) change in conformational degrees
of freedom in the modulator and receptor upon binding.
The change in desolvation entropy will be most favorable
if surfaces buried upon binding are predominantly hydro-
phobic. Because 5 was found to be ~eightfold more soluble
in water compared to 3 (2.4 5 0.1 mM vs. 0.29 5 0.04 mM
at room temperature (24)), from a desolvation point of view
it is presumably more favorable to bury the 7-fluorinated
compound 3 compared to 5 containing a 7-hydroxyl group.
A similar conclusion can be deduced with 2 and 4 based on
their respective water solubility at room temperature (4:
2.90 5 0.04 mM; 2: 1.65 5 0.04 mM; data not shown).
c
d
Values from (7).
Fig. 7. Generally, it appears that the binding of BPAMs with
an electronegative atom or substituent in the 7-position
(
BPAM25 (11), 2, 3, 4, and 5) is mainly driven by enthalpy,
whereas an electronegative atom or substituent in the 5- or
-position (BPAM408 (13) and BPAM429 (15)) changes
6
the profile so that binding is mainly entropy driven.
Introduction of an electronegative atom or substituent in
the 8-position (BPAM37 (12) and BPAM442 (16)) seems
to render the binding equally enthalpy and entropy driven.
Interestingly, compound IDRA21 (1) containing a different
substitution pattern in the thiadiazine scaffold is mainly
driven by entropy. This observation is consistent with a scaf-
FIGURE 7 Thermodynamics of binding of BPAM modulators measured by ITC. (A) Structures of IDRA21 (1) and 10 BPAMs investigated to date. (B)
Enthalpy, entropy, and free energy contributions to binding are shown as black, white, and gray columns, respectively. Listed from left are compounds for
which binding is mainly driven by enthalpy: BPAM521 (5) (this study), BPAM344 (3) (12), BPAM557 (4) (this study), BPAM411 (14) (27), BPAM97 (2) (7),
and BPAM25 (11) (13). The compounds are listed according to the difference between DH and –TDS, with the largest difference to the left and smallest to the
right, and then compounds mainly driven by entropy are listed: BPAM408 (13) (27), IDRA21 (1) (7), and BPAM429 (15) (27). Finally, compounds equally
driven by enthalpy and entropy are listed: BPAM37 (12) (13) and BPAM442 (16) (27).
2
404 Biophysical Journal 110, 2397–2406, June 7, 2016

GluA2: Enthalpy/Entropy Compensation
Moreover, the 4-ethyl-substituted compounds (2 and 4) are
more soluble than their 4-cyclopropyl counterparts (3 and
caused by an entropy penalty arising from desolvation ef-
fects and conformational constraints imposed on Ser-497
and possibly also on the hydroxyl group of the modulator.
5
). We calculated SM8 solvation energies (22) for com-
pounds 2–5, showing that the estimated solvation energy
of 4 and 5 (ꢀ18.1 and ꢀ18.4 kcal/mol, respectively)
is more favorable than for 2 and 3 (ꢀ14.3 and
ACCESSION NUMBERS
ꢀ
14.6 kcal/mol, respectively). The influence of cyclopropyl
The accession number for the structure coordinates and corresponding
structure factor file of GluA2 LBD-L483Y-N754S with BPAM521 (5) re-
ported in this paper is Protein Data Bank: 5ELV.
(
3 and 5) versus ethyl (2 and 4) seems to be negligible based
on the estimated solvation energies. Therefore, the observed
difference in experimental water solubility between ethyl
and cyclopropyl (2 vs. 3 and 4 vs. 5) seems not to be re-
flected in the SM8 solvation energies. However, the uncer-
tainty of 0.5–0.8 kcal/mol on the SM8 solvation energies
should be kept in mind. Another likely contribution to
the less favorable binding entropy with the exchange of
the 7-fluorine with a hydroxyl group is introduction of an
entropy penalty for the higher ordering of the Ser-497 side
chain and the modulator hydroxyl group in the 4 and 5 com-
plexes compared to the structures with 2 and 3. Therefore,
the unfavorable entropy might be partly due to desolvation
of more soluble compounds 4 and 5 and partly to conforma-
tional restriction of Ser-497 by the added hydroxyl group of
compound 5 and possibly also 4. The presence of the cyclo-
propyl group on 3 has previously been shown to increase
favorable enthalpy and affinity, over the ethyl substituted
compound 2 due to pi-interactions of the cyclopropyl group
with Met-496 (12). This is corroborated by this study in
comparing the thermodynamic data obtained for 5 with
those of 4. Interestingly, comparing 2 with 4 we observe a
lesser gain in favorable enthalpy and lesser loss of favorable
entropy from the exchange of F with OH than when
comparing 3 with 5. Intuitively, one would have expected
additivity of the substituents. However, other examples on
nonadditivity of substituents have previously been reported,
e.g., see (25).
In conclusion, we have reported the synthesis of two new
positive allosteric modulators of AMPA receptors (4 and 5)
that are 7-hydroxyl substituted analogs of the previously re-
ported compounds 2 and 3, respectively. We have investi-
gated the binding mode of 5 using x-ray crystallography
and shown that this substitution enables a hydrogen bond
to the side chain of Ser-497. We suggest that the conforma-
tion of Ser-497 in the GluA2 LBD complex with 5 is due to
lack of repulsion from the modulator and that this favors an
interaction between Ser-497 and the hydroxyl group of 5. In
addition, we investigated the effects of the substitution on
the affinities and thermodynamics of binding using ITC.
This revealed that compared to their fluorinated counterparts
AUTHOR CONTRIBUTIONS
C.K., P.F., D.S.P., J.P., L.O., E.G., B.P., and J.S.K. designed research; C.K.,
P.F., D.S.P., L.J., and E.G. performed research; C.K., P.F., D.S.P., L.J., J.P.,
L.O., K.F., E.G., B.P., and J.S.K. analyzed data. C.K., P.F., D.S.P., B.P., and
J.S.K. wrote the first draft of the article. B.P. and J.S.K. equally supervised
the work.
ACKNOWLEDGMENTS
Heidi Peterson is thanked for help with expression and purification of
GluA2 LBD-L483Y-N754S and MAX-Lab, Lund, Sweden is thanked for
providing beamtime.
We thank the Danish Council for Independent Research j Medical Sciences
(
to C.K. and J.S.K.), the Lundbeck Foundation (to J.P., L.O., and J.S.K.),
GluTarget (to C.K., L.J., K.F., and J.S.K.), and Danscatt (to C.K., L.J.,
K.F., and J.S.K.) for financial support.
REFERENCES
1
2
3
. Traynelis, S. F., L. P. Wollmuth, ., R. Dingledine. 2010. Glutamate
receptor ion channels: structure, regulation, and function. Pharmacol.
Rev. 62:405–496.
. Sobolevsky, A. I., M. P. Rosconi, and E. Gouaux. 2009. X-ray structure,
symmetry and mechanism of an AMPA-subtype glutamate receptor.
Nature. 462:745–756.
. Armstrong, N., and E. Gouaux. 2000. Mechanisms for activation and
antagonism of an AMPA-sensitive glutamate receptor: crystal struc-
tures of the GluR2 ligand binding core. Neuron. 28:165–181.
4
5
6
. Sun, Y., R. Olson, ., E. Gouaux. 2002. Mechanism of glutamate re-
ceptor desensitization. Nature. 417:245–253.
. Jin, R., S. Clark, ., K. M. Partin. 2005. Mechanism of positive allosteric
modulators acting on AMPA receptors. J. Neurosci. 25:9027–9036.
. Stern-Bach, Y., S. Russo, ., C. Rosenmund. 1998. A point mutation in
the glutamate binding site blocks desensitization of AMPA receptors.
Neuron. 21:907–918.
7
8
. Krintel, C., K. Frydenvang, ., J. S. Kastrup. 2012. Thermodynamics
and structural analysis of positive allosteric modulation of the iono-
tropic glutamate receptor GluA2. Biochem. J. 441:173–178.
. Francotte, P., Pd. Tullio, ., B. Pirotte. 2007. Design, synthesis, and
pharmacology of novel 7-substituted 3,4-dihydro-2H-1,2,4-benzothia-
diazine 1,1-dioxides as positive allosteric modulators of AMPA recep-
tors. J. Med. Chem. 50:3153–3157.
2
and 3, the new positive allosteric modulators 4 and 5 show
a favorable decrease in binding enthalpy but also equally
nonfavorable binding entropy, leaving the affinity and po-
tency of the compounds essentially unchanged. This more
favorable binding enthalpy can be explained by the forma-
tion of a hydrogen bond between 5 and Ser-497 in GluA2.
The decrease in the favorable binding entropy is most likely
9
. Yamada, K. A., and S. M. Rothman. 1992. Diazoxide blocks glutamate
desensitization and prolongs excitatory postsynaptic currents in rat hip-
pocampal neurons. J. Physiol. 458:409–423.
1
0. Yamada, K. A., and C. M. Tang. 1993. Benzothiadiazides inhibit rapid
glutamate receptor desensitization and enhance glutamatergic synaptic
currents. J. Neurosci. 13:3904–3915.
Biophysical Journal 110, 2397–2406, June 7, 2016 2405

Krintel et al.
1
1. Ptak, C. P., A. H. Ahmed, and R. E. Oswald. 2009. Probing the allo-
steric modulator binding site of GluR2 with thiazide derivatives.
Biochemistry. 48:8594–8602.
20. Adams, P. D., P. V. Afonine, ., P. H. Zwart. 2010. PHENIX: a compre-
hensive Python-based system for macromolecular structure solution.
Acta Crystallogr. Sect. D: Biol. Crystallogr. 66:213–221.
1
2. Nørholm, A. B., P. Francotte, ., J. S. Kastrup. 2013. Synthesis, phar-
macological and structural characterization, and thermodynamic
aspects of GluA2-positive allosteric modulators with a 3,4-dihydro-
21. Hayward, S., and H. J. Berendsen. 1998. Systematic analysis of domain
motions in proteins from conformational change: new results on citrate
synthase and T4 lysozyme. Proteins. 30:144–154.
2
5
H-1,2,4-benzothiadiazine 1,1-dioxide scaffold. J. Med. Chem.
6:8736–8745.
2
2. Marenich, A. V., R. M. Olson, ., D. G. Truhlar. 2007. Self-consistent
reaction field model for aqueous and nonaqueous solutions based on
accurate polarized partial charges. J. Chem. Theory Comput. 3:2011–
2033.
1
3. Francotte, P., A. B. Nørholm, ., B. Pirotte. 2014. Positive allosteric
modulators of 2-amino-3-(3-hydroxy-5-methylisoxazol-4-yl)propionic
acid receptors belonging to 4-cyclopropyl-3,4-dihydro-2h-1,2,4-pyri-
dothiadiazine dioxides and diversely chloro-substituted 4-cyclo-
propyl-3,4-dihydro-2H-1,2,4-benzothiadiazine 1,1-dioxides. J. Med.
Chem. 57:9539–9553.
2
3. Ptak, C. P., C. L. Hsieh, ., R. E. Oswald. 2014. Role of stoichiometry
in the dimer-stabilizing effect of AMPA receptor allosteric modulators.
ACS Chem. Biol. 9:128–133.
2
4. Larsen, A. P., P. Francotte, ., J. S. Kastrup. 2016. Synthesis and
pharmacology of mono-, di- and trialkyl-substituted 7-chloro-3,4-dihy-
dro-2H–1,2,4-benzothiadiazine 1,1-dioxides combined with X-ray
structure analysis to understand the unexpected structure-activity rela-
tionship at AMPA receptors. ACS Chem. Neurosci. 7:378–390.
1
1
1
1
4. Girard, Y., J. G. Atkinson, and J. Rokach. 1979. A new synthesis of
1,2,4-benzothiadiazines and a selective preparation of o-aminobenze-
nesulfonamides. J. Chem. Soc., Perkin Trans. 1. 1043–1047.
5. Ursby, T., J. Unge, ., M. M. Thunnissen. 2013. The macromolecular
crystallography beamline I911–3 at the MAX IV laboratory.
J. Synchrotron Radiat. 20:648–653.
2
5. Biela, A., M. Betz, ., G. Klebe. 2012. Water makes the difference: re-
arrangement of water solvation layer triggers non-additivity of func-
tional group contributions in protein-ligand binding. ChemMedChem.
6. Battye, T. G., L. Kontogiannis, ., A. G. Leslie. 2011. iMOSFLM:
a new graphical interface for diffraction-image processing with
MOSFLM. Acta Crystallogr. D Biol. Crystallogr. 67:271–281.
7
:1423–1434.
2
6. Laskowski, R. A., M. W. MacArthur, ., J. M. Thornton. 1993.
PROCHECK: a program to check the stereochemical quality of protein
structures. J. Appl. Cryst. 26:283–291.
7. Collaborative Computational Project, Number 4. 1994. The CCP4
suite: programs for protein crystallography. Acta Crystallogr. D Biol.
Crystallogr. 50:760–763.
2
7. Nørholm, A. B., P. Francotte, ., C. Oostenbrink. 2014. Thermody-
namic characterization of new positive allosteric modulators binding
to the glutamate receptor A2 ligand-binding domain: combining exper-
imental and computational methods unravels differences in driving
forces. J. Chem. Inf. Model. 54:3404–3416.
1
1
8. McCoy, A. J., R. W. Grosse-Kunstleve, ., R. J. Read. 2007. Phaser
crystallographic software. J. Appl. Cryst. 40:658–674.
9. Emsley, P., B. Lohkamp, ., K. Cowtan. 2010. Features and develop-
ment of Coot. Acta Crystallogr. D Biol. Crystallogr. 66:486–501.
2
406 Biophysical Journal 110, 2397–2406, June 7, 2016