Correlation of carboxylic acid pKa to protein binding and antibacterial activity of a novel class of bacterial translation inhibitors

Article abstract of DOI:10.1016/j.bmcl.2007.05.006

Previously we reported the discovery and initial optimization of a novel anthranilic acid derived class of antibacterial agents which suffered from extensive protein binding. This report describes efforts directed toward understanding the relationship of the acidity of the carboxylic acid with the extent of protein binding. The pK_a of the acid was modified via the synthesis of a number of anthranilic acid analogs which vary the aromatic ring substituent at the 4-position. The pK_a and HSA binding constants have been determined for each of the analogs. Our results indicate a correlation between pK_a and HSA K_d. The physical properties and antibacterial activities will be discussed as well as how these results help address the protein binding issue with this series of compounds.

Full text of DOI:10.1016/j.bmcl.2007.05.006

Bioorganic & Medicinal Chemistry Letters 17 (2007) 5479–5482
Correlation of carboxylic acid pKa to protein binding
and antibacterial activity of a novel class of bacterial
translation inhibitors
Cory M. Stiff,a Min Zhong,c Ronald W. Sarver,b Hua Gao,e Andrea M. Ho,b
Michael T. Sweeney,d Gary E. Zurenkod and Donna L. Romeroa,*,
aMedicinal Chemistry, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, MI 49001, USA
bStructural Analytical and Medicinal Chemistry, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, MI 49001, USA
cPharmaceutical Sciences, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, MI 49001, USA
dInfectious Diseases Biology, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, MI 49001, USA
eStructural and Computational Chemistry, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, MI 49001, USA
Received 14 February 2007; revised 3 May 2007; accepted 4 May 2007
Available online 10 May 2007
Abstract—Previously we reported the discovery and initial optimization of a novel anthranilic acid derived class of antibacterial
agents which suffered from extensive protein binding. This report describes efforts directed toward understanding the relationship
of the acidity of the carboxylic acid with the extent of protein binding. The pKa of the acid was modified via the synthesis of a num-
ber of anthranilic acid analogs which vary the aromatic ring substituent at the 4-position. The pKa and HSA binding constants have
been determined for each of the analogs. Our results indicate a correlation between pKa and HSA Kd. The physical properties and
antibacterial activities will be discussed as well as how these results help address the protein binding issue with this series of
compounds.
Ó 2007 Published by Elsevier Ltd.
Emergence of bacterial resistance continues to be a grow-
ing problem in the treatment of bacterial infections.1–3
Thus, Pharmacia invested significant resources in an
attempt to discover antibiotics that targeted novel mech-
anisms of action, with the thought that this would be the
most expedient way to overcome resistance. We recently
reported the discovery of a novel class of protein tran-
scription/translation inhibitors4 through a combination
of high throughput screening and medicinal chemistry
optimization (compounds 1 and 2). While compound 2
exhibited good in vitro antibacterial activity, there was
no in vivo activity in a standard mouse bacteremia model
of infection. Since antibacterial activity is greatly reduced
in the presence of human serum, the lack of in vivo activ-
ity was attributed to the high protein binding of these
compounds. As a consequence, subsequent optimization
efforts focused on further improving the potency and
reducing the protein binding of these leads.5,6
By focusing on reducing lipophilicity and molecular
weight, several analogs with improved antibacterial
activity and/or reduced protein binding, relative to ini-
tial leads (e.g., 1 and 2), were designed and synthesized
(Fig. 1, compound 3; Fig. 2, compounds 4–6). Protein
binding was routinely assessed by adding human serum
to the in vitro antibacterial assay. Compounds 3–6 have
MICs of 0.125–1 lg/mL against Staphylococcus aureus,
but the addition of 10% serum raises the MICs to the
4–8 lg/mL range; a 4- to 64-fold difference. Even the
level of reduced protein binding afforded by compound
6 did not translate into in vivo activity (ED50s >200 mg/
kg in a mouse bacteremia model). Since human serum
albumin (HSA) represents 60% of the total mass of plas-
ma proteins and is known to bind aromatic carboxylic
acids (e.g., salicylates, ibuprofen),7 we hypothesized that
further reduction of protein binding was required in or-
der to attain in vivo activity.
Keywords: Translation inhibitors; Protein binding; Antibacterial;
Capillary electrophoresis; Acidity.
*
Corresponding author. Tel.: +1 636 247 9321; fax +1 636 247 5260;
e-mail: donna.l.romero@pfizer.com
Present address: Pfizer Inc., 700 Chesterfield Parkway West, BB 273-2
Chesterfield, MO 63017, USA.
0960-894X/$ - see front matter Ó 2007 Published by Elsevier Ltd.
doi:10.1016/j.bmcl.2007.05.006

5480
C. M. Stiff et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5479–5482
sulfonamide substitution on the central aromatic ring
Br
CO2H
NC
CO2H
CO2H
either meta or para to the amide carbonyl was opti-
mal.4,6 In addition, the potency of compound 6 was
one of the least impacted upon the addition of HSA.
Therefore, for the purposes of this study, we elected to
focus on the meta-substituted series of compounds
(Scheme 1, compounds 6 and 9–15).
NH
O
NH
O
NH
O
N
O
O2S
N
O2S
N
BnS
3
The desired analogs were prepared following the route
shown in Scheme 1. 3-Chlorosulfonylbenzoic acid was
reacted with morpholine to provide the sulfonamide 7,
which upon conversion to its acid chloride was coupled
to an anthranilic acid methyl ester (8). Basic hydrolysis
of the methyl ester provided the desired carboxylic acids
(6 and 9–15). In the few instances where the anthranilic
esters were not commercially available, the desired
methyl ester was obtained via esterification of the avail-
able acid or reduction of the corresponding nitro
derivative.
1
2
Cl
MIC
MIC
SAUR 9218
ClogP
MW
SAUR 9218 + 10% serum
>64 μg/mL
64 μg/mL
4 μg/mL
32 μg/mL
0.125 μg/mL
0.125 μg/mL
1
2
3
454
536
380
6.0
6.2
3.8
Figure 1. Discovery and optimization of translation inhibitors.
O
O
O
The pKa of each compound was determined using pres-
sure-assisted capillary electrophoresis (PACE). The
measurement of pKa by PACE is based on the ionic
mobility of the solute over a range of buffer pH values.8
Mobility and pH data can then be fit to equilibrium
expressions and the pKa of the solute determined (Table
1, column 3). The accuracy of this method has been
found to be 0.2 pH units. Additionally, pKas were pre-
dicted computationally using the computer program
PALLAS (Table 1, column 4).
NC
NC
NC
OH
NH
OH
NH
OH
NH
O
O
O
O2S
Cl
O2S
O2S
N
N
N
O
5
6
4
MIC
SAUR 9218
MIC
ClogP
MW
SAUR 9218 +10% serum
The binding of small molecules to HSA can be measured
using a variety of methods. Previous experiments
demonstrated that this class of translation inhibitors
competitively displaces the fluorescent probe dansylsar-
cosine from site II on subdomain IIIA of HSA. While
this result does not preclude the binding of the molecules
to another site on HSA, it does indicate a high affinity
for an area close to site II. Using a previously developed
method,9 the dissociation constant, Kd, for each com-
pound was calculated from the percentage of dansylsar-
cosine displaced from HSA (Table 1, column 5). Finally,
the MICs in the presence and absence of 5% human
serum were determined to provide another measure of
the impact of protein binding on activity.
8 μg/mL
8 μg/mL
4 μg/mL
4
5
6
0.125 μg/mL
1μg/mL
1μg/mL
482
399
415
5.1
3.1
2.4
Figure 2. Meta-sulfonamide-substituted translation inhibitors with
good antibacterial activity, but no in vivo activity.
Prior SAR studies had demonstrated that the carboxylic
acid was an essential part of the pharmacophore
required for activity. Substituting the acid with other
simple functional groups generally resulted in decreased
activity, except in cases where the functional group also
possessed an acidic proton. In such cases protein
binding was still a problem. Since we were limited to
acids or similar groups, it occurred to us that the acidity
of the acid might impact the extent to which the
compounds were protein bound. Therefore, we under-
took an investigation into the impact of the acidity of
the carboxylic acid on the protein binding, in order to
gain additional insight into ways in which protein
binding might be reduced.
COOH
COOH
(COCl)2/DMF
SO2
N
pyridine
THF 50 oC
NH
+
O
CH2Cl2
SO2Cl
7
O
R
COOH
NH
COCl
O
Since previous studies had demonstrated that substitu-
tion on the 5-position of the aryl ring was detrimental
and substitution on the 4-position of the aryl ring was
beneficial4 we targeted the synthesis of analogs in which
the 4-aryl substituent was varied from electropositive
substituents to electronegative substituents. Addition-
ally, in the course of the exploration of the structure–
activity relationships (SAR), it had been determined that
1. CH2Cl2/pyridine
R
+
OCH3
O
2. LiOH
THF/MeOH/H2O
SO2
N
NH2
6, 9-15
O2S
8
N
O
O
Scheme 1. Synthesis of anthranilic acid analogs.

C. M. Stiff et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5479–5482
5481
Table 1. Measured pKa, calculated pKa, measured protein binding, and antibacterial activity for the compounds with the 5-aryl ring substituent
varied
Compound
R
Clog P
pKa PACE
pKa Pallas
Protein binding Kd (lM)
MIC (lg/mL)a
SAUR 5% serumc
SAURb
9
10
11
12
13
14
6
H
2.6
3.1
2.6
2.8
3.4
3.6
2.4
2.7
3.18
3.36
4.00
4.06
3.89
3.66
3.63
3.61
3.38
3.26
6.0 1.2
7.5 1.6
8.1 1.7
6.6 1.9
6.5 1.2
4.8 0.7
5.3 1.2
4.4 0.3
>128
>128
>128
128
4
>128
>128
>128
128
32
CH3
OCH3
F
3.12
2.92
Cl
Br
2.81
2.79
4
32
CN
NO2
2.56
3.44 (2.46)d
1
2
16
15
4
a Minimum inhibitory concentration.
b Staphylococcus aureus UC9218.
c Staphylococcus aureus UC9218 + 5% pooled human serum. Human serum (male, from Sigma) was thawed at room temperature, then placed in a
56 °C water bath for 30 min. The serum was filtered using a 0.2-lm filtration system.
d Expected value of pKa given 0.8 units difference between calculated and experimental values.10
Compounds in Table 1 are arranged from least acidic to
most acidic. It is interesting to note that the computa-
tional predictions (PALLAS) of pKa for this class of
compounds were consistently 0.7–0.8 pH units less
acidic than those determined experimentally by PACE.
This difference can be attributed to an intramolecular
hydrogen bond between the amide NH and the carbonyl
of the carboxylic acid enhancing the acidity of the acid.
Small molecule X-ray structures of compounds in this
analog series confirm the presence of such a hydrogen
bond.11 The computational prediction is not able to take
this interaction into account, and thus returns a less
acidic pKa.
Compounds (9–11) containing a carboxylic acid with a
pKa P 2.92 (measured) show little or no antibacterial
activity against S. aureus. For the compounds that exhi-
bit antibacterial activity against S. aureus, the addition
of just 5% serum reduces the potency. Calculating the
ratio between potency with and without 5% serum also
gives an indication of the extent of protein binding.
For compounds 6 and 13–15 the ratio runs from 2- to
8-fold and generally trends with the Kd measurement.13
Lipophilicity has also been shown to impact the degree
of protein binding in this series of compounds.6 Its im-
pact is likely to be observed in the MIC assays with ser-
um protein present as this system presents a more global
view of the total extent of protein binding (i.e., in cases
where the compounds bind to more than one site, or to
other proteins in the media, vide infra). It is clear that a
strategy built upon reducing pKa to minimize protein
binding will require balancing with the requirement to
retain the degree of acidity needed to retain the potent
antibacterial activity.
The data collected in Table 1 indicate a correlation be-
tween the acidity of the acid (either calculated or mea-
sured) and the protein binding to HSA (Kd). Stated
another way: generally, the more acidic the acid, the
more highly protein bound the analog. A graphical rep-
resentation of the data is shown in Figure. 3, which plots
both the measured and calculated pKa of the analogs
versus the affinity for HSA (Kd). As the pKas increase
(either experimentally or calculated), the dissociation
constant for the dansylsarcosine site increases.
Compound 6 was selected for a more extensive evalua-
tion of its protein binding characteristics. Isothermal
titrating calorimetry (ITC) indicated that two or more
molecules of compound 6 bound to fraction V of
HSA, and that there is a 2 lM high affinity binding site
and a 19-lM lower affinity site(s). Kds of ꢀ14 and 2 lM
were measured using ultra filtration with human serum
and stopped-flow fluorescence quenching, respectively.14
Since the ITC measurement indicates more than one
binding site, the Kds in Table 1 determined by dansyl
sarcosine are only representative of a portion of the
affinity for HSA.
The biological data in Table 1 also indicate that the anti-
bacterial activity12 is dependent on the pKa of the acid.
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
In conclusion, a study in which the pKa of the carboxylic
acid moiety was varied to determine impact on protein
binding demonstrated that acidity of the carboxylic acid
correlates with affinity to HSA as measured by dansyl
sarcosine displacement. It also appears that there is a
certain threshold of acidity required to retain the anti-
bacterial activity. The fact that compound 6, which is
both highly acidic and highly protein bound, was one
of the most potent analogs in the presence of human ser-
um at the time of this study underscores the complexity
of this issue. Correlation of antibacterial activity of
2.00
2.50
3.00
3.50
4.00
4.50
pKa
CE Pallas
Figure 3. Anthranilic acid pKa versus HSA Kd.

5482
C. M. Stiff et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5479–5482
Sweeney, M. T.; Marotti, K. R. Bioorg. Med. Chem. Lett.
2006, 16, 6173.
5. Li, J.; Wakefield, B. D.; Ruble, J. C.; Stiff, C. M.; Romero,
D. L.; Marotti, K. R.; Sweeney, M. T.; Zurenko, G. E.;
Rohrer, D. C.; Thorarensen, A. Bioorg. Med. Chem. Lett.
2007. doi:10.1016/j.bmcl.2006.12.055.
these compounds in the presence and absence of serum
with the Kd measured via the dansylsarcosine displace-
ment is complicated by experimental data that indicate
binding at a second lower affinity site. Still, it appears
that efforts to reduce acidity would be well placed.
6. Thorarensen, A.; Li, J.; Wakefield, B. D.; Romero, D. L.;
Marotti, K. R.; Sweeney, M. T.; Zurenko, G. E.; Sarver,
R. W. Bioorg. Med. Chem. Lett. 2007. doi:10.1016/
j.bmcl.2007.03.036.
7. Peters, T., Jr. All About Albumin: Biochemistry, Genetics, and
Medical Applications; Academic Press: New York, 1996.
8. Zhongliang, J.; Ramstad, T.; Zhong, M. Electrophoresis
2001, 22, 1112.
9. (a) Epps, D. E.; Raub, T. J.; Kezdy, F. J. Anal. Biochem.
1995, 227, 342; (b) The assay described above was
modified to use a fluorescence plate reader instead of the
ISS K2 spectrofluorimeter.
In addition to protein binding, it is likely that a combi-
nation of factors such as unbound clearance and volume
of distribution impacts the in vivo efficacy of this class of
compounds. The correlation between protein binding
and acid pKa presented herein guided the strategy for
further optimization of this series of compounds with
the aim of identifying a novel antibacterial agent. Future
reports will describe efforts to replace the carboxylic acid
with bioisosteres.11,15
10. See Ref. 8 for a thorough discussion of assay limitations.
Compound 15 appears to be an experimental outlier
possibly due to pH shift of the CE buffers during the
experiment.
11. Stiff, C. M.; Graber, D. R.; Marotti, K. R.; Melchior, E.;
Sweeney, M. T.; Han, F.; Rohrer, D. C.; Zurenko, G. E.;
Romero, D. L. Biorg. Med. Chem. Lett. 2007, 45, accepted
for publication.
Acknowledgment
The authors thank Laura Holtzman for retrieving legacy
Pharmacia antibacterial data.
References and notes
12. National Committee for Clinical Laboratory Standards,
Approved Standard. Methods for Dilution Antimicrobial
SusceptibilityTests for Bacteria that Grow Aerobically, 5th
ed.; NCCLS document M7-A4: Wayne, Pennsylvania,
2000.
13. MIC measurements have a 2-fold error. Calculating ratios
between MIC measurements increases the error range,
making it difficult to draw firm conclusions. It is clear that
improving inherent potency has a positive impact on
potency in the presence of serum.
14. Wang, Zhigang; Epps, Dennis; Rogers, Joseph. Unpub-
lished results.
15. Ruble, J. C.; Wakefield, B. D.; Kamilar, G. M.; Marotti,
K. R.; Melchior, E.; Sweeney, M. T.; Zurenko, G. E.;
Romero, D. L. Bioorg. Med. Chem. Lett. 2007, 89.
doi:10.1016/j.bmcl.2007.04.074.
1. Cohen, M. L. Science 1992, 257, 1050.
2. (a) Allen, N. E. In Progress in Medicinal Chemistry;
Ellis, G. P., Luscombe, D. K., Eds.; Elsevier Science:
New York, 1995; 32, pp 157–238; (b) Hayes, J. D.;
Wolf, C. R. Biochem. J. 1990, 272, 281; (c) Spratt, B. G.
Science 1994, 264, 388; (d) Nikaido, H. Science 1994,
264, 382.
3. (a) Russell, A. D. In Progress in Medicinal Chemistry;
Ellis, G. P., Luscombe, D. K., Oxford, A. W., Eds.;
Elsevier Science: New York, 1998; 35,, Chapter 4 (b)
Fuchs, T. M. Naturwissenschaften 1998, 85, 99; (c) Cunha,
B. A. Drugs Today 1998, 34, 691.
4. Larsen, S. D.; Hester, M. R.; Ruble, J. C.; Kamilar, G.
M.; Romero, D. L.; Wakefield, B.; Melchior, E. P.;