Role of aromatic rings in the molecular recognition of aminoglycoside antibiotics: Implications for drug design

Published on Web 08/10/2010

Role of Aromatic Rings in the Molecular Recognition of

Aminoglycoside Antibiotics: Implications for Drug Design

Tatiana Vacas,^†Francisco Corzana,^‡Gonzalo Jime´nez-Ose´s,^§Carlos Gonza´lez,^|

Ana M. Go´mez,^†Agatha Bastida,^†Julia Revuelta,^†and Juan Luis Asensio^†,*

Instituto de Qu´ımica Orga´nica General (CSIC), Juan de la CierVa 3, 28006 Madrid, Spain,

Departamento de Qu´ımica, UniVersidad de La Rioja, UA-CSIC, Logron˜o, Spain, Departamento

de Qu´ımica Orga´nica y Qu´ımica F´ısica. UniVersidad de Zaragoza-CSIC, Zaragoza, Spain, and

Instituto de Qu´ımica F´ısica Rocasolano (CSIC), Madrid, Spain

Received May 28, 2010; E-mail: iqoa110@iqog.csic.es

Abstract: Aminoglycoside antibiotics participate in a large variety of binding processes involving both RNA

and proteins. The description, in recent years, of several clinically relevant aminoglycoside/receptor

complexes has greatly stimulated the structural-based design of new bioactive derivatives. Unfortunately,

design efforts have frequently met with limited success, reﬂecting our incomplete understanding of the

molecular determinants for the antibiotic recognition. Intriguingly, aromatic rings of the protein/RNA receptors

seem to be key actors in this process. Indeed, close inspection of the structural information available reveals

that they are frequently involved in CH/π stacking interactions with sugar/aminocyclitol rings of the antibiotic.

While the interaction between neutral carbohydrates and aromatic rings has been studied in detail during

past decade, little is known about these contacts when they involve densely charged glycosides. Herein

we report a detailed experimental and theoretical analysis of the role played by CH/π stacking interactions

in the molecular recognition of aminoglycosides. Our study aims to determine the inﬂuence that the antibiotic

polycationic character has on the stability, preferred geometry, and dynamics of these particular contacts.

With this purpose, different aminoglycoside/aromatic complexes have been selected as model systems.

They varied from simple bimolecular interactions to the more stable intramolecular CH/π contacts present

in designed derivatives. The obtained results highlight the key role played by electrostatic forces and the

desolvation of charged groups in the molecular recognition of polycationic glycosides and have clear

implications for the design of improved antibiotics.

resistance processes.³These proteins represent primary phar-

macological targets.

Introduction

Aminoglycosides are highly potent, broad-spectrum antibiot-

ics widely used in clinics. These drugs bind speciﬁcally to the

bacterial decoding site, in the 16S rRNA (A-site), and thereby

interfere with the accuracy of protein synthesis, leading to

bacterial cell death.¹Additional targets such as the DIS kissing-

loop complex, the Tat-responsive element, and the Rev-

responsive element have been identiﬁed within the HIV genomic

RNA and are also located in different functionally relevant RNA

fragments, including self-splicing ribozymes and tRNAs.²In

addition to interacting speciﬁcally with RNA receptors, ami-

noglycosides are also recognized by the enzymes involved in

antibiotic inactivation that play a central role in bacterial

The emergence during the past decade of high-resolution

structural data for aminoglycoside/protein⁴and aminoglycoside/

RNA⁵complexes has greatly stimulated the structure-based

design of new bioactive derivatives with improved properties.

(3) (a) Magnet, S.; Blanchard, J. S. J. Chem. ReV. 2005, 105, 477–497.

(b) Smith, C. A.; Baker, E. N. Curr. Drug Targets 2002, 2, 143–160.

(4) (a) Pedersen, L. C.; Benning, M. M.; Holden, H. M. Biochemistry

1995, 34, 13305–13311. (b) Fong, D. H.; Berghuis, A. M. EMBO J.

2002, 21, 2323–2331. (c) Vetting, M. W.; Park, C. H.; Hedge, S. S.;

Jacoby, G. A.; Hooper, D. C.; Blanchard, J. S. Biochemistry 2008,

47, 9825–9835. (d) Maurice, F.; Broutin, I.; Podglajen, I.; Benas, P.;

Collatz, E.; Dardel, F. EMBO Rep. 2008, 9, 344–349. (e) Nurizzo,

D.; Shewry, S. C.; Perlin, M. H.; Brown, S. A.; Dholakia, J. N.; Fuchs,

R. L.; Deva, T.; Baker, E. N.; Smith, C. A. J. Mol. Biol. 2003, 327,

491–506.

^†Instituto de Qu´ımica Orga´nica General.

(5) (a) Carter, A. P.; Clemons, W. M.; Brodersen, D. E.; Morgan-Warren,

R. J.; Wimberly, B. T.; Ramakrishnan, V. Nature 2000, 407, 340–

348. (b) Vicens, Q.; Westhof, E. Chem. Biol. 2002, 9, 747–755. (c)

Francois, B.; Rusell, R. J. M.; Murray, J. B.; Aboul-ela, F.; Masquida,

B.; Vicens, Q.; Westhof, E. Nucleic. Acids. Res. 2005, 33, 5677–5690.

(d) Mikkelsen, N. E.; Johanson, K.; Virtanen, A.; Kirsebom, L. A.

Nat. Struct. Biol. 2001, 8, 510–514. (e) Freisz, S.; Lang, K.; Micura,

R.; Dumas, P.; Ennifar, E. Angew. Chem., Int. Ed. 2008, 47, 4110–

4113. (f) Ennifar, E.; Paillart, J. C.; Marquet, R.; Ehresmann, B.;

Ehresmann, C.; Dumas, P.; Walter, P. J. Biol. Chem. 2003, 278, 2723–

2730. (g) Tereshko, V.; Skripkin, E.; Patel, D. J. Chem. Biol. 2003,

10, 175–187.

^‡Universidad de La Rioja.

^§Universidad de Zaragoza.

^|Instituto de Qu´ımica F´ısica Rocasolano.

(1) (a) Talaska, A. E.; Schacht, J. In Aminoglycoside Antibiotics: from

Chemical Biology to Drug DiscoVery, Arya, D. P., Ed.; Wiley.: New

York, 2007. (b) Vicens, Q.; Westhof, E. ChemBioChem 2003, 4, 1018–

1023.

(2) (a) Schroeder, R.; Waldsich, C.; Wank, H. EMBO J. 2000, 19, 1–9.

(b) Walter, F.; Vicens, Q.; Westhof, E. Curr. Opin. Chem. Biol. 1999,

3, 694–704. (c) Sucheck, J. S.; Wong, C. H. Curr. Opin. Chem. Biol.

2000, 4, 678–686.

9

12074 J. AM. CHEM. SOC. 2010, 132, 12074–12090

10.1021/ja1046439  2010 American Chemical Society

Molecular Recognition of Antibiotics

A R T I C L E S

Figure 1. (Top) Schematic representation of kanamycin-B (left panel), kanamycin-A (middle panel), and ribostamycin (right panel). The numbering

employed for the different sugar/aminocyclitol units is shown in black. In addition, the numbering employed for the different positions of the I/II

fragment is shown in red on the kanamycin-B structure (left panel). (Bottom) CH/π stacking interactions present in their respective complexes with

the resistance enzyme AAC(6′) (pdb code 2QIR), the ribosomal A-site RNA (pdb code 2ESI) and the viral HIV DIS kissing-loop (pdb code 2FCZ),

according to X-ray data. In the former case (left) the antibiotic unit III is represented in white for clarity. The position of the aminoglycoside NH₂/

+

NH₃groups is highlighted.

More speciﬁcally, research has been focused on the preparation

of tighter, more speciﬁc, and less toxic RNA binders,⁶enzymatic

inhibitors,⁷and new drugs not susceptible to enzymatic inactiva-

tion.⁸Unfortunately, design efforts have frequently met with

limited success, which, in our opinion, partially reﬂects our

limited understanding of the molecular forces that stabilize the

aminoglycoside complexes at a fundamental level.

Intriguingly, a detailed inspection of the structural information

available reveals that both protein and RNA receptors employ

common strategies for aminoglycoside recognition. Thus,

whereas complexes are, in most cases, stabilized by a signiﬁcant

number of salt-bridges and polar contacts, the speciﬁcity seems

to be provided by CH/π stacking interactions involving aromatic

residues of the protein/RNA and the ligand pyranose/aminocy-

clitol rings. Key examples of these contacts have been provided

by the recently described X-ray structures of kanamycin/

ribostamycin in complex with the resistance enzyme AAC(6′),^4d

the ribosomal A-site RNA^5cand the HIV DIS kissing-loop

complex^5e(Figures 1 and S1). Thus, in the former case, two

different tryptophan side-chains participate in contacts with

aminoglycoside rings I/II (the numbering used along the

manuscript for the different aminoglycoside units and positions

is shown in Figure 1). In a similar way, the aminoglucose ring

I stacks on a highly conserved guanine ring in both RNA

complexes.

(6) (a) Zhao, F.; Zhao, Q.; Blount, K. F.; Han, Q.; Tor, Y.; Hermann, T.

Angew. Chem., Int. Ed. 2005, 44, 5329–5334. (b) Blount, K. G.; Zhao,

F.; Hermann, T.; Tor, Y. J. Am. Chem. Soc. 2005, 127, 9818–9829.

(c) Revuelta, J.; Vacas, T.; Corzana, F.; Gonzalez, C.; Bastida, A.;

Asensio, J. L. Chem.sEur. J. 2010, 10, 2986–2991. (d) Francois, B.;

Szychowski, J.; Sekhar, S.; Pachamuthu, A. K.; Swayze, E. E.; Griffey,

R. H.; Migawa, M. T.; Westhof, E.; Hanessian, S. Angew. Chem., Int.

Ed. 2004, 43, 6735–6738. (e) Nudelman, I.; Rebibo-Sabbah, A.;

Cherniavsky, M.; Belakhov, V.; Hainrichson, M.; Chen, F.; Schacht,

J.; Pilch, D. S.; Ben-Yosef, T.; Baasov, T. J. Med. Chem. 2009, 52,

2836–2845. (f) Pokrovskaya, V.; Belakhov, V.; Hainrichson, M.;

Yaron, S.; Baasov, T. J. Med. Chem. 2009, 52, 2243–2254. (g)

Luedtke, N. W.; Baker, T. J.; Goodman, M.; Tor, Y. J. Am. Chem.

Soc. 2000, 122, 12035–12036.

Carbohydrate/aromatic CH/π stacking interactions have been

shown to play key roles in the recognition of neutral oligosac-

charides by proteins.^9-11They stabilize the complexes by

promoting the desolvation of the sugar hydrophobic patches.

Additional favorable contributions arise from the interaction

between the polarized pyranose CH groups and the aromatic

electron density. Moreover, these contacts are extremely sensi-

tive to the axial/equatorial orientation of the different pyranose

OH groups and, as a result, they increase the speciﬁcity of the

binding process.

(7) (a) Boehr, D. D.; Farley, A. R.; Wright, G. D.; Cox, J. R. Chem. Biol.

2002, 9, 1209–1217. (b) Gao, F.; Yan, X.; Baettig, O. M.; Berghuis,

A. M.; Auclair, K. Angew. Chem., Int. Ed. 2005, 44, 6859–6862. (c)

Gao, F.; Yan, X.; Auclair, K. Chem.sEur. J. 2009, 15, 2064–2070.

(d) Lombs, T.; Begis, G.; Maurice, F.; Turcaud, S.; Lecourt, T.; Dardel,

F.; Micouin, L. ChemBioChem 2008, 9, 1368–1371. (e) Boehr, D. D.;

Draker, K.; Koteva, K.; Bains, M.; Hancock, R. E.; Wright, G. D.

Chem. Biol. 2003, 10, 189–196.

(9) (a) Quiocho, F. A. Biochem. Soc. Trans. 1993, 21, 442–448. (b)

Lemieux, R. U. Acc. Chem. Res. 1996, 29, 373–380.

(10) (a) Jime´nez-Barbero, J.; Asensio, J. L.; Can˜ada, F. J.; Poveda, A. Curr.

Opin. Struct. Biol. 1999, 9, 549–555. (b) Cha´vez, M. I.; Andreu, C.;

Vidal, P.; Aboitiz, N.; Freire, F.; Groves, P.; Asensio, J. L.; Asensio,

G.; Muraki, M.; Can˜ada, F. J.; Jime´nez-Barbero, J. Chem.sEur. J.

2005, 11, 7060–7074. (c) Terraneo, G.; Potenza, D.; Canales, A.;

Jime´nez-Barbero, J.; Baldridge, K. K.; Bernardi, A. J. Am. Chem. Soc.

2007, 129, 2890–2900. (d) Ferna´ndez-Alonso, M. C.; Can˜ada, F. J.;

Jimenez-Barbero, J.; Cuevas, G. J. Am. Chem. Soc. 2005, 127, 7379–

7386. (e) Ram´ırez-Gualito, K.; Alonso-R´ıos, R.; Quiroz-Garc´ıa, B.;

Rojas-Aguilar, A.; D´ıaz, D.; Jime´nez-Barbero, J.; Cuevas, G. J. Am.

Chem. Soc. 2009, 131, 18129–18138.

(8) (a) Roestamadji, J.; Graspas, I.; Mobashery, S. J. Am. Chem. Soc.

1995, 117, 11060–11069. (b) Haddad, J.; Vakulenko, S.; Mobashery,

S. J. Am. Chem. Soc. 1999, 121, 11922–11923. (c) Bastida, A.;

Hidalgo, A.; Chiara, J. L.; Torrado, M.; Corzana, F.; Can˜adillas, J. M.;

Groves, P.; Garc´ıa-Junceda, E.; Gonza´lez, C.; Jime´nez-Barbero, J.;

Asensio, J. L. J. Am. Chem. Soc. 2006, 128, 100–116. (d) Asensio,

J. L.; Hidalgo, A.; Bastida, A.; Torrado, M.; Corzana, F.; Junceda,

E. G.; Can˜ada, J.; Chiara, J. L.; Jime´nez-Barbero, J. J. Am. Chem.

Soc. 2005, 127, 8278–8279.

(11) Laughrey, Z. R.; Kiehna, S. E.; Riemen, A. J.; Waters, M. L. J. Am.

Chem. Soc. 2008, 130, 14625–14633.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12075

A R T I C L E S

Vacas et al.

Figure 2. Schematic representation of the different roles that positive charges could play in the stability of the aminoglycoside/aromatic complexes. (a) The

polarization of the antibiotic C-H groups is expected to be larger for the protonated ligand. Consequently, an enhancement in the stability of the CH/π

+

bonds would be expected under neutral or acidic conditions. The direct, attractive or repulsive, electrostatic interaction between the different NH₃groups

of the drug and the aromatic quadrupole might also have an inﬂuence on the complex stability. (b) Aromatic units might affect the solvation of the charged

ammonium functions providing a destabilizing contribution at neutral and acid pH values.

Interestingly, little is known about the way densely charged

glycosides interact with aromatic rings. In principle, the

aminoglycoside polycatonic character could have three dif-

ferential effects on the stability of the resulting complexes

(Figure 2). First, polarization of the pyranose/aminocyclitol CH

groups will be more important in the protonated antibiotic, and

consequently, CH/π bonds should be pH-sensitive and stronger

under neutral or acidic conditions. Second, those ammonium

groups located on the stacked sugar/aminocyclitol units could

interact electrostatically with the quadrupole of the aromatic

ring. Moreover, ammonium groups present on distant sugar units

of the aminoglycoside could also be involved in long-range

attractive or repulsive NH₃⁺/π interactions¹²with the aromatic

system. Finally, aromatic units might negatively affect solvation

of the charged ammonium functions, providing a destabilizing

contribution at neutral and acidic pH values. Indeed, under these

particular conditions, the removal of water molecules from the

proximity of the antibiotic would have two opposite inﬂuences

whose net result is unclear; the desolvation of the hydrophobic

CH functions, favorable for the formation of the aminoglycoside/

aromatic complexes, would be opposed by the unfavorable

and have clear implications for the design of improved

bioactive derivatives.

Results and Discussion

First Evidence for the Unfavorable Inﬂuence of the Aminogly-

coside Positive Charges on the CH/π Stacking Interactions.

NMR Analysis of Simple Bimolecular Complexes. In order to have

a ﬁrst indication of the inﬂuence that ammonium groups have

on the stability of the aminoglycoside/aromatic complexes, we

analyzed by NMR the interaction between natural kanamycin-B

(see Figure 1) and benzyl alcohol. This later represents the

simplest aromatic system soluble in water and non ionizable in

the 1-11 pH range.

The chemical shift perturbations induced in the drug by the

presence of this particular aromatic probe were monitored to

map out those antibiotic regions involved in specially favored

stacking interactions under different pH conditions. A similar

methodology has been successfully employed in the analysis

of stacking interactions involving neutral monosaccharide

units.¹⁰The results obtained are represented in Figure 3.

According to these data, neutral kanamycin-B (pH 10.3) interacts

with benzyl alcohol mainly through ring II. Indeed, the axial

protons H1-II/H3-II/H5-II (labeled in Figure 3 with red, cyan

and yellow circles, respectively) seem to participate in specially

favored CH/π bonds with the aromatic probe, as revealed by

their negative ∆δ values. Signiﬁcantly, lower chemical shift

changes were measured for unit I, whereas for ring III the

observed perturbations were almost negligible. This tendency

seems to be in agreement with the structural information

available for kanamycin-B/RNA or kanamycin-B/protein com-

plexes according to which, aminoglycoside units I/II are clearly

preferred to stack on aromatic systems (see Figure 1).^4,5As a

second point, this simple experiment strongly suggests that the

aminoglycoside/aromatic interaction is weakened or disrupted

by the protonation of the former. Indeed, the chemical shift

perturbations detected at pH 10.3 vanish under acidic conditions.

Moreover, the remaining minor ∆δ values exhibited, in most

cases, opposite sign to that expected for a CH/π bond (an upﬁeld

shift).

+

desolvation of the highly polar -NH₃groups.

Herein, we report a detailed experimental and theoretical

analysis of aminoglycoside/aromatic interactions. More specif-

ically, our study aims to determine the inﬂuence that the

antibiotic polycationic character has on the stability, preferred

geometries and internal mobility of the CH/π contacts. With

this purpose, the experimental behavior of several aminogly-

coside/aromatic interactions has been studied by NMR under

a variety of pH and temperature conditions. They varied from

simple bimolecular contacts to the more stable intramolecular

CH/π bonds present in synthetic derivatives (see below),

speciﬁcally designed to exhibit different degrees of internal

mobility. NMR experiments have been complemented with

extensive MD simulations and state of the art ab initio

calculations. The obtained results are, in our opinion, of

signiﬁcant relevance for the understanding of the driving

forces behind aminoglycoside recognition by RNA or proteins

The destabilization of the aminoglycoside/aromatic complexes

at low pH values probably reﬂects the unfavorable inﬂuence of

(12) Ma, J. C.; Dougherty, D. A. Chem. ReV. 1997, 97, 1303–1324.

9

12076 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

Figure 3. (a) Key regions of HSQC experiments measured at 25 °C for kanamycin-B (20 mM) in the absence (in red) and presence (in green) of 200 mM

benzyl alcohol at pH 10.3 (up) and 3.3 (down). Only the outer contour levels are represented for clarity. (b) Proton chemical shift perturbations (ppm/1000)

induced at the different positions of kanamycin-B (left) and 2-deoxy-streptamine (right) by the aromatic probe at pH 10.3 (up) and 3.3 (down).

+

speciﬁc NH₃groups. These might include not only those

attached to antibiotic units directly involved in CH/π bonds (as

the aminocyclitol unit II), but also those located on noninter-

acting sugar rings. In order to reduce the system complexity,

and as a next step, kanamycin-B was replaced by 2-deoxy-

streptamine (2-DOS). The bimolecular interaction between

benzyl alcohol and the aminocyclitol alone was tested employing

the same experimental protocol previously described for the

complete aminoglycoside. The results obtained are also repre-

sented in Figure 3. In this case, minor upﬁeld shifts of particular

aminocyclitol signals were detected under acidic conditions.

However, as previously observed for kanamycin-B, the ∆δ

values measured at pH 10.0 are signiﬁcantly larger than those

observed in the fully charged aminocyclitol. In conclusion,

aminoglycoside/aromatic complexes seem to be destabilized by

the protonation of the antibiotic amino groups. This result might

seem surprising considering that the polarization of the drug

C-H groups is more important in the fully charged compound.

Unfortunately, the labile and presumably highly dynamic

character of the previously described bimolecular interactions

prevents a precise quantiﬁcation of the energy penalty imposed

contacts are strongly favored by the aminoglycoside covalent

scaffold, so that relatively minor conformational ﬂuctuations

around the stacked geometries are allowed (see below). There-

fore, 2 provides a stable framework for the analysis of

aminocyclitol/aromatic interactions under a variety of pH and

temperature conditions.

Considering these general properties, derivative 1 was

employed to obtain quantitative information about the inﬂuence

of the aminoglycoside charges on the stability of the CH/π

bonds, whereas 2 was used to asses their effect on the stacking

geometry and dynamics.

It should be noted that in both 1 and 2 the aminoglycoside is

expected to interact with the aromatic ring through the most

favored face of the aminocyclitol unit II. Indeed, these contacts

represent the intramolecular version of those between benzyl

alcohol and kanamycin-B or 2-deoxystreptamine, previously

analyzed (see Figure 3).

As a second consideration, we would like to emphasize that

derivatives 1 and 2 were speciﬁcally designed to promote the

stacking between units II and naphthyl or phenyl rings,

respectively. These complexes might exhibit slightly different

geometries, which can be broadly classiﬁed as parallel-type and

nonparallel-type (see Figure 4b), depending on the angle formed

by the interacting rings. In the former case, up to 3 axial CH

groups of the aminocyclitol moiety (positions H1-II, H3-II, and/

or H5-II) could be involved in CH/π bonds with the aromatic

system. For nonparallel arrangements, an equatorial position

(H2-II) might also participate in such contacts. Both stacking

geometries have been well characterized in neutral oligosac-

charides.¹⁰In contrast, those conformations that would allow a

direct interaction of the NH₃⁺groups with the aromatic electron

density, a pure cation/π bond,¹²are clearly disfavored. Conse-

quently, the inﬂuence of the aminoglycoside charges on the

stacking could be analyzed without the inﬂuence of competing

alternative interacting modes.

+

by the different NH₃groups.

Design of Kanamycin Derivatives for the Analysis of the

Aminoglycoside/Aromatic CH/π Stacking Interactions. The

requirement for alternative model systems, able to form more

stable aminoglycoside/aromatic complexes, led us to design and

synthesize derivatives 1 and 2 (Figures 4 and S2-S3).

Derivative 1 includes a naphthyl ring connected to a R-altrose

unit (ring III) by a ﬂexible ethylene linker. The aromatic system

is optimally positioned to stabilize hairpin-like structures by

means of a stacking interaction with the 2-DOS unit (II).

However, transitions between hairpin and disordered conforma-

tions would also be feasible. A benzylidene moiety between

positions 4 and 6 of the inherently ﬂexible altrose ring (unit

III) was incorporated in order to prevent its inversion.

Regarding derivative 2, its dimeric structure was conceived

to promote an intermonomer double stacking between the phenyl

rings, and the 2-DOS moieties (units II). Molecular dynamics

simulations (MD) demonstrate that these aminocyclitol/aromatic

The synthesis of derivatives 1 and 2 is described in detail in

the Supporting Information.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12077

A R T I C L E S

Vacas et al.

Figure 4. (a) Schematic representation of aminoglycosides 1 (left) and 2 (right) together with the corresponding molecular mechanics minimized models.

The stacking interactions between the aminocyclitol unit II and the aromatic rings are highlighted. The colored circles/squares show the code employed to

designate the relevant positions of the aminocyclitol and the aromatic system throughout the manuscript. (b) Schematic representation of the different

interacting modes between unit II and the corresponding aromatic ring accessible for both 1 and 2. A maximum of three CH/π contacts are feasible for

parallel and nonparallel stacking geometries. The aminocyclitol positions involved are highlighted.

Experimental NMR Analysis of Derivatives 1 and 2 in

Solution. Preliminary Experimental Considerations. Our analy-

sis of 1 and 2 involved the measurement of key aminocyclitol/

aromatic NOEs and chemical shifts under different pH and

temperature conditions. In addition, the pK_avalues of the

different amino groups were derived from NMR pH-titration

experiments. In order to estimate the chemical shift and pK_a

perturbations induced in the I/II fragments (see Figure 4) by

the aromatic systems (∆δ and ∆pK_a), natural kanamycin-B

(represented in Figure 1), together with several natural and

synthetic derivatives, all of them including the common neamine

core (rings I/II), were employed as reference compounds (Figure

S4).

Both ∆δ and ∆pK_avalues contained relevant information

about the aminoglycoside/aromatic complexes. Thus, the chemi-

cal shift perturbations measured in 1 provided a ﬁrst indication

of the hairpin population present in solution. In addition, the

pattern of ∆δ values observed in 1 and 2 was employed to

estimate the preferred stacking geometries. This parameter is

extremely sensitive to the structural details of the complexes,

providing a simple way to identify those aminocyclitol positions

involved in preferred CH/π contacts with the aromatic rings

(quantum mechanic calculations performed on simple models

demonstrate that, in all cases, the corresponding proton signals

should be signiﬁcantly upﬁeld shifted as a result of the

interaction, Figure S5).

and 2 renders the NMR analysis of pK_avalues a nontrivial task.

The presence of aromatic rings covalently linked to the

aminoglycoside scaffold implies that the conformational adjust-

ments that accompany the protonation of the different amino

functions may cause signiﬁcant chemical shift perturbations at

distant positions. In order to unambiguously determine pK_a’s,

all of the ¹H and ¹³C signals were assigned employing 2D-HSQC

experiments and their respective changes in chemical shift were

monitorized as a function of pH. In addition, heteronuclear ¹J_CH

constants for the HC groups directly attached to amino moieties

+

(HC-NH₂/NH₃fragments) were measured at different pH

values. This parameter has been shown to be sensitive to the

protonation state of the amino groups, and it is not affected by

the ring current effects of aromatic systems.¹⁴

Second, amino basicities are, to some extent, dependent on

the sample conditions. More speciﬁcally, salt tends to stabilize

the antibiotic protonated form, causing a general increase in

the pK_a’s.¹⁵Taking this into account, all of the experiments

were performed in absence of added salts. Moreover, to

guarantee that the measured pK_aperturbations do not result from

minor differences in the sample conditions, the protonation state

of 1 and 2 was also checked in the presence of reference

aminoglycoside kanamycin-B, employing 1:1 mixtures.

(13) (a) Wang, J.; Li, J.; Cheng, H.; Chang, H.; Tanifum, C. T.; Liu, H.;

Czyryca, P. G.; Chang, C. T. J. Med. Chem. 2004, 6, 1381–1384. (b)

Li, J.; Wang, J.; Czyryca, P. G.; Chang, H.; Orsak, T. W.; Evanson,

R.; Chang, C. T. Org. Lett. 2004, 6, 1381–1384.

The pK_aperturbations measured in 1 and 2 provided an

estimation of the energy contribution of the different aminogly-

coside charges to the stability of the stacked geometries. In our

opinion, several experimental aspects of the pK_aanalysis deserve

special attention. First, the peculiar structure of derivatives 1

(14) Freire, F.; Cuesta, I.; Corzana, F.; Revuelta, J.; Gonza´lez, C.; Hricovini,

M.; Bastida, A.; Jime´nez-Barbero, J.; Asensio, J. L. Chem. Commun.

2007, 2, 174–176.

(15) Barbieri, C. M.; Pilch, D. S. Biophys. J. 2006, 90, 1338–1349.

9

12078 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

Figure 5. (a) Key region of NOESY spectra collected for derivative 1 at pH 9.6 (up) and 5.0 (down). NOE cross-peaks connecting the naphthyl moiety and

the aminocyclitol proton signals are highlighted (color code indicated in Figure 4). (b) A superimposition of HSQC spectra collected for 1 (in red) and

kanamycin-B (in black), at basic (left) and acid (right) pH values. (c) Maximum chemical shift perturbations (with respect to kanamycin-B) induced by the

1

naphthyl ring in the aminocyclitol unit at the extreme pH values 9.5 and 3.5. (d) Key region of 1D H NMR spectra collected for derivative 1 at pH values

9.5, 7.0, and 4.8 (from top to bottom) and 40 °C. The chemical shifts variations (ppm) measured in the 5-40 °C temperature range are represented above

(in gray).

Finally, it should be considered that naphthyl and benzylidene

rings present in 1, confer to this aminoglycoside a certain

hydrophobic character and, consequently it was of key impor-

tance to rule out its oligomerization/aggregation at the concen-

tration employed in the NMR experiments (0.5 mM). With this

objective, control DOSY¹⁶and dilution experiments were

performed under different pH (in the 3-10 range) and temper-

ature (in the 5-40 °C range) conditions. These data unambigu-

ously proved that derivative 1 is monomeric (see Figure S6).

Therefore we proceeded with its analysis.

H3-II/H5-II remained sharp even at 5 °C. Moreover, they

exhibited small temperature coefﬁcients (Figures 5d and S9).

In fact, maximum chemical shift differences in the temperature

range 5-40 °C were, in all cases, below 0.1 ppm. Both

observations indicate that under these pH conditions, the

population of disordered species is very small.

A different behavior was observed at acidic pH values. Thus,

at pH 5.0 only extremely weak aminocyclitol/aromatic cross

peaks could be detected in NOESY experiments (Figures 5a

and S7).

Experimental Analysis of Derivative 1. First, compound 1 was

analyzed. NMR data collected at pH >9.0 (the neutral form of

the aminoglycoside) conﬁrmed the speciﬁc interaction of the

naphthyl unit with aminocyclitol ring II. Thus, several cross

peaks in NOESY spectra (Figures 5a and S7) demonstrated the

proximity between both regions. The presence of a large

population of hairpin conformers was further supported by the

analysis of the chemical shift perturbations induced by the

aromatic ring in the aminoglycoside scaffold (Figures 5b,c and

S8). Both, NOE and chemical shift data were consistent with a

dominant parallel-type stacking geometry, with three aminocy-

clitol axial protons (H1-II/H3-II/H5-II) involved in CH/π

bonds with the aromatic system. Thus, the most intense NOE

contacts (shown in Figure 5a) involved this triad, whose signals

were also signiﬁcantly upﬁeld shifted as a result of the

interaction (Figures 5b,c and S8). Indeed, even at 40 °C,

chemical shift perturbations measured at this region of 1 were

larger than 1 ppm (up to 1.6 ppm for H5). In contrast, the

equatorial proton H2-II exhibited only moderate ∆δ values and

did not present NOE contacts with the naphthyl unit.

These contacts were absent from the data acquired at pH 3.5

(the fully protonated state of 1). In addition, under these

experimental conditions, chemical shift perturbations at the

aminocyclitol moiety were drastically reduced (Figures 5b,c and

S8). Finally, in the pH range 4.0-6.0, temperature coefﬁcients

of the 2-DOS proton signals were signiﬁcantly larger than those

measured for the neutral aminoglycoside (Figures 5d and S9).

Indeed, under these pH conditions, a temperature decrease (from

40 to 5 °C) was accompanied by a signiﬁcant broadening of

the proton signals H1-II, H3-II, and H5-II, reﬂecting the

progressive increase in the hairpin population and the lower

exchange rate between hairpin and disordered structures (see

Figure S9). An intermediate behavior between those previously

described was observed at neutral pH. In conclusion, our analysis

demonstrates that, although remarkably stable at pH >9.0, the

aminocyclitol/naphthyl stacking is progressively destabilized

under neutral and acidic conditions by the protonation of the

aminoglycoside.

In order to provide further evidence for the unfavorable

inﬂuence of the aminoglycoside charges on the stability of the

CH/π interactions, pK_aperturbations (∆pK_a) were determined

employing NMR (Figures 6, S10, and S11). Typical titration

proﬁles obtained for 1 at 40 °C are shown in Figure 6a. The

pK_afor the amino group at position 2-I was obtained indepen-

dently from four titration curves (signals C2-I, H2-I, C3-I, and

H3-I). To determine the basicity of those amino groups located

at ring II, proton and carbon chemical shift changes at positions

The temperature dependence of the ¹H NMR spectra provided

additional evidence for the stability of the aminoglycoside/

aromatic complexes. Thus, it could be observed that, despite

their remarkable ∆δ values at basic pH, proton signals H1-II/

(16) Grooves, P.; Rasmussen, M. O.; Molero, M. D.; Samain, E.; Can˜ada,

F. J.; Driguez, H.; Jime´nez-Barbero, J. Glycobiology 2004, 14, 451–

456.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12079

A R T I C L E S

Vacas et al.

Figure 6. (a) Typical NMR titration proﬁles obtained for 1 at 40 °C. The monitorized ¹H/¹³C signals are indicated. (b) Heteronuclear ¹J_CHconstants measured

+

in 1 for the HC-NH₂/NH₃fragments located at positions 2-I and 1-II, under different pH conditions. The information provided by these values was fully

consistent with that derived from the chemical shift changes. (c) pK_avalues at 40 °C for the amino groups of neamine fragment I/II in 1 (in black) and

several related aminoglycosides (including kanamycin-B, see Figures S11 and S14) employed as reference. For these later derivatives, pK_aranges (in gray),

reﬂecting the variation in the measured basicities, are represented.

2-II and 6-II (highlighted with a red circle in Figure 6a) were

monitored. These sites are relatively unaffected by the proximity

of the naphthyl system. For position 2-II, two different transi-

tions were apparent, reﬂecting the sequential change in the

protonation state of the amino groups 1-II and 3-II. Therefore,

the corresponding pK_avalues could be determined independently

from a single titration proﬁle. In addition to chemical shifts,

heteronuclear ¹J_CHconstants were also measured for all the HC-

NH₂/NH₃⁺fragments present in the aminoglycoside at different

pH values (Figure 6b). The obtained results conﬁrmed the

destabilizing inﬂuence of the aminoglycoside positive charges.

shifts perturbations (up to -1.8 ppm, Figures 7b and S12) were

detected for the aminoclyclitol protons in both, neutral and

protonated 2. Interestingly, the pattern of ∆δ values observed

under different pH conditions conﬁrmed that aminoglycoside

protonation/deprotonation is accompanied by subtle adjustments

of the aminocyclitol/phenyl complex. Thus, at pH 9.6 axial

protons H1-II, H3-II, and H5-II, were the most affected (see

Figure 7b), which is consistent with their participation in CH-π

contacts with the aromatic system. This particular interaction

mode requires a nearly coplanar disposition of the phenyl and

aminocyclitol rings (parallel-type). In contrast, under acidic

conditions, the ∆δ value for position 5-II was basically

maintained, whereas those corresponding to axial H1-II and H3-

II and equatorial H2-II proton signals increased, suggesting that,

to some extent, this triad was also involved in contacts with

the benzylidene moiety (nonparallel complex). Particularly

signiﬁcant is the remarkable upﬁeld shifting measured for the

proton signal H1-II at pH 3.5. Overall, the chemical shift

perturbations observed in unit II are consistent with a dynamic

exchange between different parallel and nonparallel stacking

modes (described in Figure 4b), with the ﬁrst type being

predominant at basic pH values and the later under acidic

conditions. Moreover, the aminoglycoside protonation promotes

a slight shifting of the aromatic system toward the aminocyclitol

position H1-II, which seems to participate in especially persistent

CH/π contacts (both conclusions were fully supported by the

theoretical ∆δ values calculated from MD trajectories in

solution, see below). Finally, the increased magnitude of the

+

According to the pK_adata, three speciﬁc -NH₃groups of 1,

those at positions 2-I, 1-II, and 3-II, oppose the aminocyclitol/

phenyl stacking interaction. The total pK_adifferences with

respect to the reference compounds (Figure 6c) amounted, at

least, to 1.3 units, which implies that ammonium groups

destabilize the stacking interaction by more than 1.9 kcal/mol.

This result is fully consistent with the conformational behavior

previously described.

Experimental Analysis of Derivative 2. As a next step, the

conformational properties of 2 in solution were carefully

analyzed. Its structure was designed to strongly favor the

aminocyclitol/phenyl interaction, providing a more stable model

system for the analysis of CH/π contacts. In agreement with

these expectations, the NMR data proved the presence of the

predicted intermonomer double stacking under all the pH

conditions tested. Thus, clear NOE cross peaks connecting the

phenyl ring with the 2-DOS unit were apparent at both, basic

and acid pH values (Figure 7a). In addition, signiﬁcant chemical

9

12080 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

Figure 7. (a) Key region of NOESY spectra collected at pH values 8.8 (left) and 3.8 (right) and 5 °C for derivative 2. NOE cross-peaks connecting the

phenyl moiety and the aminoglycoside proton signals H1-II (red circle), H2-II (green circle), H3-II (cyan circle), and H5-II (yellow circle) are highlighted.

(b) Maximum chemical shift perturbations with respect to kanamycin-B, induced in the aminocyclitol unit at 40 °C and the extreme pH values 9.6 (left) and

3.5 (right). A superimposition of HSQC spectra collected for 2 (in red) and kanamycin-B (in black), at acid and basic pH values is also represented.

Figure 8. (a) Temperature dependence exhibited by the selective NOEs measured in 2 upon inversion of the benzylidene protons a (in red), under basic

(left) and acidic (right) conditions. Contacts with protons b (in cyan) and c (in green) are represented. (b) pK_avalues for the amino groups located in neamine

I/II fragment of 2 (in black) and several related aminoglycosides (including kanamycin-B, see Figures S13 and S14) at 40 °C. For the reference compounds,

a range (in gray), reﬂecting the variation in measured basicities is represented. (c) Key region of coupled HSQC spectra, acquired for a derivative 2:kanamycin-B

1

+

1:1 mixture at pH 7.0 and 40 °C. The heteronuclear J_CHconstants measured for the HC-NH₂/NH₃fragments located at positions 1-II and 3-II in both, 2

and kanamycin-B, are shown. This parameter is sensitive to the protonation state of the amino groups (as indicated in the upper panel) and demonstrates that,

under identical conditions, the amino group 1-II in derivative 2 exhibits a reduced pK_a.

chemical shift perturbations measured at pH 3.5 is also

suggestive of a less dynamic aminocyclitol/phenyl interaction.

The apparent decrease in the complex internal mobility at

low pH values might reﬂect the presence of stronger attractive

electrostatic forces between the phenyl system and the charged

aminocyclitol. Considering the potential relevance of this

ﬁnding, we looked for additional experimental evidence for the

stiffening of the complex under acidic conditions. In principle,

this could also be reﬂected in the rotation rate of the phenyl

rings. It should be noted that this conformational transition

implies the transient disruption of the aminocyclitol/phenyl

CH/π contacts and, consequently, should be specially affected

by an enhancement of the electrostatic interactions occurring

between both entities. To test this point, we performed selective

NOE experiments at different temperatures. Our analysis was

focused on two speciﬁc contacts, schematically represented in

Figure 8a. The benzylidene and altrose protons, Ha and Hb

(also referred as H4-III), are located at a ﬁxed distance (2.4 Å).

In contrast, that between Ha and phenyl protons Hc, should be

affected by the internal motion of the aromatic system. Figure

8a shows the temperature dependence exhibited by Ha/Hb and

Ha/Hc NOEs under basic and acidic conditions. It can be

observed that at pH 10.0 and 30 °C Ha/Hb contact is negative

while that between Ha and Hc almost undetectable. Interest-

ingly, at 35 °C the former NOE remains negative while the later

exhibits a clear positive sign. Finally, at 40 °C the Ha/Hb

contact was basically zero. Overall, these data demonstrate that

the Ha/Hc pair exhibits a shorter correlation time (and

consequently, larger mobility) reﬂecting the internal motions

of the phenyl system. In conclusion, the experimental data

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12081

A R T I C L E S

Vacas et al.

measured for neutral 2 fully supported the dynamic character

of the phenyl/aminocyclitol interaction.

A different behavior was observed at low pH values. Indeed,

under these conditions Ha/Hb and Ha/Hc contacts were both

negative in the 30-40 °C range. Moreover, both NOEs become

zero at the same temperature, around 45 °C (larger temperatures

could not be reached with the employed cryoprobe). According

to these data, the effective correlation time for both proton pairs

are similar, implying that the rotation of the phenyl rings is

signiﬁcantly frozen by the protonation of the aminocyclitol units.

This conclusion is in agreement with the chemical shift

perturbations observed in charged and neutral 2 and also with

the results of the theoretical analysis described in the following

sections (see below).

Despite the reduced internal mobility deduced for the

aminocyclitol/phenyl complexes under acidic conditions, the

analysis of the pK_aperturbations promoted by the aromatic ring

demonstrates that aminoglycoside charges are essentially de-

stabilizing (Figures 8b, S13, and S14). The pK_afor the amino

group at position 2-I was derived independently from the

analysis of four experimental curves (corresponding to C2-I,

H2-I, C3-I, and H3-I signals). To determine the basicity of those

aminos located on the aminocyclitol ring II, we focused our

attention on positions 2-II and 6-II, not involved in CH/π

contacts with the phenyl ring.

Figure 9. (Left) Ensembles of hairpin (down) and disordered (up)

conformers of 1 (total charge +4) obtained from a single 0.1 µs MD

simulation performed using explicit TIP3P water. (Right) Population of

conformers stabilized by stacking interactions in aminoglycoside 1,

derived from independent 0.1 µs MD trajectories. Three different

protonation states were considered: neutral, total charge +4 (neutral

linker) and total charge +5. Free energy differences between hairpin

and disordered conformers (∆G ) G_hairpin- G_disordered) estimated from

their respective populations are also represented (in black).

As previously observed for 1, three speciﬁc NH₃⁺groups of

2 (those at positions 2-I, 1-II, and 3-II) oppose the aminocyclitol/

phenyl stacking interaction. Indeed, these moieties exhibited a

signiﬁcantly reduced basicity (out of the range observed for the

reference compounds, see Figures 8b,c and S14). Even the

amino function 1-II, directly attached to the aminocyclitol

position involved in tighter CH/π contacts with the aromatic

system (as judged by its large ∆δ value), presents a reduced

pK_a. This conclusion was further conﬁrmed by the heteronuclear

¹J_CHconstants measured in derivative 2:kanamycin-B mixtures

(see Figure 8c). The total pK_adifferences amounted, at least, to

1.0 units implying that positive charges destabilize the stacking

interaction by more than 1.4 kcal/mol. In this particular case,

the inﬂuence of ammonium groups is not sufﬁcient to promote

the disruption of the aminocyclytol/phenyl contact, strongly

favored by the aminoglycoside covalent scaffold.

included in our analysis to separate the inﬂuence of the linker

protonation from that of the rest of amino groups on the

aminocyclitol/naphthyl stacking interaction. The obtained results

were fully consistent with the experimental observations de-

scribed in the previous section (Figure 9). Thus, for neutral 1,

hairpin conformations were dominant (95% population). Pro-

tonation of all the amino functions, except that on ring III (the

linker nitrogen), promoted a severe destabilization of the

naphthyl/2-DOS stacking interaction, which amounted to ∆∆G

) 1.6 kcal/mol.

Finally, protonation of the whole structure further destabilized

the folded state leading to a minor hairpin population (19%).

In order to understand the origin of this effect, we had a

thorough look at the solvation of 1 along a selected MD

trajectory. More speciﬁcally, our analysis was restricted to the

system with 4 positive charges (neutral linker). The ﬂuctuations

observed in the water content of the aminoglycoside ﬁrst and

second solvation shells are shown in Figure S15a. As expected,

transitions from disordered to the more compact hairpin

conformations were accompanied by release of water molecules

to the bulk. Interestingly, this effect was especially pronounced

for the second solvation shell, which, on average, exhibited

15-20 water molecules less for folded structures. We propose

that this effect leads to a poorer solvation of the aminoglycoside

Theoretical Analysis of Aminoglycoside/Aromatic Interactions:

Molecular Dynamics Simulations. Aminoglycoside Charges

and Complex Stability: Molecular Dynamic Simulations of

Derivative 1. In order to gain insights into the origin of the

peculiar behavior exhibited by 1, 0.1 µs MD simulations were

performed with the AMBER 10.0 package,¹⁷in the presence

of explicit TIP3P water,¹⁸periodic boundary conditions, and

Ewald sums for the treatment of long-range electrostatic

interactions.¹⁹These data allowed a detailed analysis of the

water content around 1 (ﬁrst and second solvation shells) in

both the hairpin and disordered conformations. Three different

protonation states of the aminoglycoside were considered: the

fully protonated and unprotonated structures (5 and 0 positive

charges, respectively) and that in which only nitrogen on ring

III remains neutral (4 positive charges). This later case was

+

NH₃functions. In agreement with this hypothesis, the radial

pair distribution functions²⁰measured for the amino groups 2-I

and 1-II showed that hairpin formation promotes an 18-20%

reduction in the water content of the second solvation shell

(Figure S15b). Intriguingly, for position 3-II a more minor effect

was observed. However, it should be considered that the extent

of ammonium desolvation might depend critically on the precise

geometrical features of the aminocyclitol/naphthyl complexes

(see below).

(17) (a) Kollman, P. A.; et al. AMBER 10; University of California: San

Francisco, 2008. (b) Pearlman, D. A.; Case, D. A.; Caldwell, W. J.;

Ross, W. R.; Cheatham, T. E.; DeBolt, S.; Ferguson, D.; Seibel, G.;

Kollman, P. Comput. Phys. Commun. 1995, 91, 1–41.

(18) Mahoney, M. W.; Jorgensen, W. L. J. Chem. Phys. 2000, 112, 8910–

8922.

(20) Corzana, F.; Motawia, M. S.; Herve´ du Penhoat, C.; Perez, S.;

Tschampel, S. M.; Woods, R. J.; Engelsen, S. B. J. Comput. Chem.

2004, 25, 573–586.

(19) Darden, T.; Perera, L.; Li, L.; Pedersen, L. Structure 1999, R55–R60.

9

12082 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

Figure 10. (a) Population of conformers (in red) stabilized by stacking interactions in simpler disaccharide versions of 1 (1a and 1b), estimated from 0.1

µs MD trajectories collected in implicit water. In all cases, two different protonation states were considered: neutral and total charge +2 (linker neutral).

Interacting CH groups are highlighted with a red circle. The estimated free energy differences between hairpin and disordered conformers (∆G ) G_hairpin

-

G_disordered) are indicated in black. (b) The aminocyclitol face involved in CH/π contacts with the naphthyl ring in neutral and charged 1a/1b is indicated.

+

Equatorial NH₃moieties render both faces of the aminocyclitol unit nonequivalent.

Aminoglycoside Charges and Facial Selectivity: Molecular

Dynamic Simulations of Simpliﬁed Derivative 1 Models. As a

next point, the inﬂuence of the positive charges on the facial

selectivity of the aminocyclitol/naphthyl interaction was theo-

retically tested. Previous studies on neutral carbohydrates have

shown that stacking complexes are extremely sensitive to the

axial/equatorial orientation of the pyranose substituents.¹⁰More

speciﬁcally, they seem to be severely destabilized by axial polar

groups. In this context, it should be noted that the aminocyclitol

unit II, present in 1/2, includes only equatorial substituents in

the most stable ring conformation. Consequently, CH/π contacts

established through both faces of this particular ring would be

expected to be energetically equivalent. The results previously

described for 1 strongly suggest that this hypothesis might be

wrong at neutral or acid pH values. Indeed, detailed inspection

of minimized models indicates that the charge desolvation

penalty associated to both complexation modes could be

signiﬁcantly different. To test this point, 0.1 µs MD trajectories

in solution were collected for the simple pseudodisaccharide

models, 1a and 1b, represented in Figure 10. The former

represents a truncated version of derivative 1. In contrast, model

1b incorporates an aminocyclitol ring glycosylated at position

5. This alternative pattern of substitution promotes the interaction

of the aromatic system with the opposite face of the aminocy-

clitol unit, which exposes axial protons H2/H4/H6. Conse-

quently, in this model compound, amino groups of ring II are

located at positions not directly involved in CH/π bonds. In

model with the Bashford modiﬁcation was employed to simulate

the water environment.²¹

As expected, for the neutral aminoglycosides the MD

trajectories predicted similar hairpin populations. Moreover, in

both cases, they were destabilized by the protonation of the

aminocyclitol unit. Interestingly, this effect was signiﬁcantly

more pronounced in charged model 1b (Figure 10). Considering

the particular location of the positive charges in this system (at

positions not involved in CH/π contacts and far from the altrose

ring), the observed effect should reﬂect, almost exclusively, the

higher energy penalty associated to their partial desolvation. In

+

conclusion, the presence of equatorial NH₃moieties renders

both faces of the aminocyclitol unit nonequivalent. This example

illustrates how charged groups can contribute to increase the

speciﬁcity of the antibiotic/aromatic complexes, favoring dis-

crimination between otherwise equivalent stacking arrangements.

Inﬂuence of the Aminoglycoside Charges on the Stacking

Geometry and Dynamics: Molecular Dynamic Simulations of

Derivative 2. To test the inﬂuence of the aminoglycoside charges

on the preferred geometry and dynamics of the aminocyclitol/

aromatic complexes, we performed MD simulations with

derivative 2, employing implicit water.²¹First, two different

protonation states of 2 were considered. They included the

neutral form, and that with rings I and II fully charged (total

charge +8). Given that, in this particular case, the aminogly-

coside covalent scaffold should allow only relatively minor

conformational ﬂuctuations, simulation lengths were limited to

20 ns. The obtained results are represented in Figure 11. It can

be observed that, in all cases, stacking interactions between the

2-DOS and the phenyl rings were maintained during the whole

trajectories. Interestingly, their precise geometries varied de-

pending on the aminoglycoside charge. Thus, for the neutral

molecule, the phenyl unit adopted mainly a parallel arrangement,

+

addition, this particular arrangement would place charged NH₃

moieties closer to the aromatic system, which might determine

a larger desolvation penalty. For our theoretical analysis, only

the neutral state of the models (1a and 1b) and that with the

aminocyclitol charged (the linker remains neutral) were con-

sidered. On this occasion, the less demanding generalized Born

(21) Onufriev, A.; Bashford, D.; Case, D. A. Proteins 2004, 55, 383–394.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12083

A R T I C L E S

Vacas et al.

Figure 11. Conformational ensembles obtained from 20 ns MD simulations of 2 in solution. The aminocyclitol and phenyl moieties are represented with

sticks and the nitrogen atoms at position 2-I with spheres. Other regions of the molecule are omitted for clarity. Two different views related by a 90° rotation

around the x axis are shown. Three different protonation states of the dimer were considered; (from left to right) neutral, only positions 6-I/1-II/3-II charged

and ﬁnally, rings I/II charged. Protonated and neutral amino groups are shown in blue and cyan, respectively. Average chemical shift perturbations (∆δ) at

key positions of the aminocyclitol unit calculated from the MD ensembles are represented on the top together with experimental NMR values (in brackets).

In addition, average distances between nitrogen atoms 2-I, 1-II, and 2-II and the geometrical center of the phenyl rings are represented in the lower panels

(also in blue or cyan). Finally, the experimental stacking geometry observed in the kanamycin-B/AAC(6′) complex according to X-ray studies (pdb code

2QIR) is shown in the right panel for comparison.

in which the aminocyclitol ring interacts with the aromatic

system through axial protons H1-II/H3-II/H5-II. In contrast, in

protonated 2, the aromatic system was clearly displaced toward

position 1-II, ﬂuctuating between parallel and nonparallel

stacking modes. This conformational rearrangement implies that

the aminocyclitol equatorial proton H2-II was transiently

involved in CH/π contacts with the aromatic system. In order

to validate the MD simulations, chemical shift perturbations at

key aminocyclitol positions H2-I/H1-II/H3-II/H5-II, in neutral

and protonated 2, were theoretically evaluated from the corre-

sponding conformational ensembles and compared with the

experimental values.

It should be noted that this represents a demanding test, given

the extreme sensitivity of the ∆δ values to minor structural

variations and the dynamic character of the aminocyclitol/phenyl

complex. As shown in Figure 11, the variations in ∆δ with the

pH deduced from the simulations were in a reasonable agree-

ment with the experimental trends. More speciﬁcally, the MD

data correctly reproduced the remarkable increase in ∆δ for the

aminocyclitol position H1-II detected at low pH. Also in

agreement with the experimental observations, the theoretical

∆δ value for the equatorial position H2-II was signiﬁcantly

increased in charged 2. Both ∆δ changes are consistent with a

larger population of nonparallel complexes under acidic

conditions.

solvation of position 2-I promoted by the conformational change

might overcompensate the more pronounced desolvation of the

aminocyclitol charged functions. In order to theoretically test

this hypothesis, we considered a different protonation state of

dimer 2. Starting from the previously analyzed charged deriva-

tive, NH₃⁺groups at key positions 2-I were selectively neutral-

ized (rendering a total charge +6) and a third 20 ns MD

trajectory was collected. As shown in Figure 11 (middle panels),

this single modiﬁcation had a large effect on the aminocyclitol/

phenyl stacking. It can be observed that, on this occasion, the

obtained conformational distribution was quite similar to that

predicted for the neutral system. This result supports the

conclusion that NH₃⁺at position 2-I has a signiﬁcant inﬂuence

on the stacking geometry. Interestingly, the precise geometrical

features of the aminocyclitol/aromatic interaction deduced for

protonated 2, closely resemble those observed in the kanamycin-

B/AAC(6′) complex, recently described by X-ray methods (see

Figure 11, right panel), suggesting that similar NH₃⁺/aromatic

interactions might determine the preferred stacking arrangement

in natural systems.^4d

As a second point, according to the MD simulations, the

dynamic character of the aminoglycoside/aromatic stacking

interaction is signiﬁcantly reduced upon protonation of the

aminocyclitol unit. The stiffening of the structure is mainly

illustrated by the reduced rotation rate exhibited by the two

aromatic moieties. Thus, for neutral 2, several transitions,

consisting in a 180° rotation of the phenyl ring (to generate

equivalent stacking modes), were observed. It should be noted

that these conformational changes require the transient disruption

of the CH/π contacts. In contrast, only few transitions were apparent

in the MD simulations collected with ring II being protonated

(Figure S16). This behavior is fully consistent with the experimental

NMR analysis of 2 previously described (see the quantum

mechanics analysis of simple models described in the following

sections for a detailed explanation of this effect).

To gain insights into the origin of this effect, the distances

+

from the phenyl rings to the NH₂/NH₃groups located at the

key positions 2-I, 1-II, and 3-II were monitorized along the MD

trajectories (Figure 11). It can be observed that the subtle

conformational change, induced by the protonation of rings I/II,

places the aromatic systems closer to the aminocyclitol am-

monium group 1-II, whereas its average distance to the

ammonium group 2-I is increased. A simple explanation for

this behavior is that the position of the phenyl rings in the

charged aminoglycoside is carefully adjusted to globally

+

minimize unfavorable contacts with the surrounding NH₃

functions. In this particular case, the improvement in the

Two conclusions can be derived from this analysis:

9

12084 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

Figure 12. (a) Quantum mechanics optimized structures for aminocyclitol(protonated form)/naphthalene pair in the different interaction geometries considered.

(b) Counterpoise-corrected interaction energies, related to the sum of isolated monomers, calculated at the SCS-MP2/6-311G(2d,p)//PCM/M06-2X/TZVP

level in the gas phase (red) and water (C-PCM-corrected values, in green). Different protonation states of the aminocyclitol moiety were considered. Interaction

energies derived for the analogous cyclohexane/naphthalene complexes are also shown for comparison. (c) Diverse contributions (electrostatic, polarization,

exchange, repulsion, and dispersion) to the interaction energies calculated through the localized molecular orbital energy decomposition analysis (LMO-

EDA) are schematically represented. Noteworthy, attractive dispersion forces are the responsible for the stacking interactions for cyclohexane and neutral

aminocyclitol. Any attempt to locate stacked geometries employing methods in which dispersion forces are poorly considered (i.e. B3LYP) failed.

First, the precise geometrical features of the aminocyclitol/

aromatic stacking seem to result from a delicate balance between

different NH₃⁺/aromatic interactions involving not only the

aminocyclitol charges but also those located in the vicinal sugar

units. This observation has implications for the molecular

recognition of aminoglycosides by protein or RNA receptors.

Most antibiotics of this family include a variable number of

positive charges, in the 4-6 range, at physiological pH. By

minimizing unfavorable contacts with speciﬁc aromatic func-

tions of the receptor, the antibiotic NH₃⁺groups might determine

a unique binding mode contributing to the speciﬁcity of the

recognition process.

Second, despite their globally destabilizing inﬂuence, ami-

noglycoside charges reduce de dynamic character of the CH/π

complexes probably reﬂecting an enhancement in the electro-

static interactions occurring between the aromatic and the

aminosugar/aminocyclitol rings.

Theoretical Analysis of Aminoglycoside/Aromatic Interactions:

Quantum Mechanics Calculations. Aminoglycoside Charges

and Complex Stability: Quantum Mechanics Calculations of

Simpliﬁed Derivative 1 Models. To further understand the

theoretical basis of the observed behavior, quantum mechan-

ics (QM) calculations were performed on a very simple

version of derivative 1, an isolated aminocyclitol/naphthalene

pair. The neutral, intermediate (charge +1), and fully

protonated (charge +2) states of the aminocyclitol were

considered. Several stacked structures were located and

energy-minimized. They comprised both parallel and non-

parallel geometries showing clear CH/π interactions. In

addition, a distorted T-shaped arrangement biased by NH₃⁺/π

interactions (only for the fully charged model) was also

considered. The aminocyclitol/naphthalene interaction ener-

gies were derived at the SCS-MP2/6-311G(2d,p) level on the

energy-minimized geometries calculated at the PCM/M06-

2X/TZVP level.^23,24Bulk solvent effects were considered

implicitly through different variants of the polarizable

continuum model for both geometry optimizations (IEF-

PCM²⁵as implemented in Gaussian 09) and single-point

energy calculations (C-PCM²⁶as implemented in Gamess).

For comparison purposes, the analogous cyclohexane/

naphthalene complexes were also evaluated employing the

same theoretical protocol.

The counterpoise-corrected interaction energies obtained are

represented in Figure 12. As can be seen through these data,

neutral aminocyclitol/naphthalene complexes are stabilized by

4.1/4.6 kcal/mol in the gas phase (∆Ei_gas), depending on the

particular geometry considered. It can be observed that the ∆Ei_gas

values derived for the protonated models (charge +1 and +2)

are much more favorable than those corresponding to the neutral

pairs, which is likely due to the enhancement of the CH/π

interactions by the presence of ammonium groups. While

attractive dispersion forces are of similar magnitude in all cases,

(22) Frisch, M. J.; et al. Gaussian 09, revision A.2; Gaussian, Inc.:

Wallingford, CT, 2009.

(23) Zhao, Y.; Truhlar, D. G. Theor. Chem. Acc. 2008, 120, 215–241.

(24) (a) Schaefer, A.; Horn, H.; Ahlrichs, R. J. Chem. Phys. 1992, 97, 2571–

2577. (b) Schaefer, A.; Huber, C.; Ahlrichs, R. J. Chem. Phys. 1994,

100, 5829–5835.

(25) (a) Cossi, M.; Barone, V.; Mennucci, B.; Tomasi, J. Chem. Phys. Lett.

1998, 286, 253–260. (b) Tomasi, J.; Mennucci, B.; Cance´s, E. J. Mol.

Struct. 1999, 464, 211–226. (c) Cossi, M.; Scalmani, G.; Rega, N.;

Barone, V. J. Chem. Phys. 2002, 117, 43–54. (d) Tomasi, J.; Mennucci,

B.; Cammi, R. Chem. ReV. 2005, 105, 2999–3093.

(26) (a) Barone, V.; Cossi, M. J. Phys. Chem. A 1998, 102, 1995–2001.

(b) Cossi, M.; Rega, N.; Scalmani, G.; Barone, V. J. Comput. Chem.

2003, 24, 669–681.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12085

A R T I C L E S

Vacas et al.

Figure 13. (a) Quantum Mechanics optimized structures for an aminocyclitol (protonated form)/benzene pair in the different interaction geometries considered

(labeled as a-d). Counterpoise-corrected interaction energies, related to the sum of isolated monomers, calculated for the fully charged and neutral complexes,

at the SCS-MP2/6-311G(2d,p)//PCM/M06-2X/TZVP level, both in gas phase (red) and water (C-PCM-corrected values, in green) are shown. Geometry b

corresponds to a transition state. (b) Quantum mechanics optimized structures for the pseudodisaccharide fragment I/II in complex with a simple aromatic

ring (toluene). The aromatic and aminocyclitol moieties are represented with sticks while the amino function at position 2-I is represented with a sphere.

Two different protonation states of the system were considered (total charge +3 and +4, see the main text). Interaction energies in the gas phase and in

water are shown in red and green, respectively. (c) BSSE-corrected dissociation potentials for a neutral and charged aminocyclitol/phenyl complex (minimum

c), calculated at the SCS-MP2/6-311G(2d,p) level, both in the gas phase (red curves) and in solution (C-PCM, green curves). Expansions of the curves

(2-4.5 Å) in the gas phase and in water are shown on the right.

the observed extra stabilization of the stacked geometries could

be related, hence, to greater long-range electrostatic and

induction (polarization) interactions occurring between the

dipole moment of the aminocyclitol and the quadrupole moment

of naphthalene. This view is supported by an analysis of the

diverse contributions to the interaction energies, calculated

through the localized molecular orbital energy decomposition

analysis, LMO-EDA²⁷(Figure 12c). Strikingly, this scenario is

completely inverted when bulk solvation effects are taken into

account. Thus, the enhanced interaction energies derived in the

gas phase for the protonated complexes are overcompensated

by signiﬁcantly less favorable solvation energies with respect

to the isolated molecules.

complex is negligible, which was attributed to the energy penalty

associated to ammonium desolvation.²⁸

Inﬂuence of the Aminoglycoside Charges on the Stacking

Geometry and Dynamics: Quantum Mechanics Calculations

of Simple Derivative 2 Models. As a next step, the inﬂuence of

the aminocyclitol charges on the stacking geometry was

analyzed. Because of the reduced size of the aromatic moiety

(benzene) with respect to the naphthalene system, the localized

minima, represented in Figure 13, were somewhat different to

those previously described. Thus, a number of parallel and

nonparallel stacked geometries, with one to two axial protons

of the aminocyclitol involved in CH/π bonds with the aromatic

ring, were found and characterized as local minima at the PCM/

M06-2X/TZVP level (labeled as a, c, and d in Figure 13). In

contrast, parallel arrangements with three interacting aminocy-

clitol axial CH positions, were characterized as local transition

states in the corresponding potential energy surface (b). This

observation is in agreement with previous results reported for

carbohydrate-phenyl interactions in neutral systems.¹⁰The

corresponding counterpoise-corrected aminocyclitol/phenyl in-

teraction energies, both in the gas phase and in water, are also

represented in Figure 13. Regarding the neutral system, it can

be observed that nonparallel arrangements (c and d) are always

Overall, full protonation of the aminocyclitol ring is predicted

to destabilize the stacked geometries by 0.7-1.5 kcal/mol, which

is in good agreement with the experimental analysis previously

described. It should be mentioned that, according to these

calculations, the unfavorable inﬂuence of ammonium desolva-

tion is more pronounced for a pure NH₃⁺/π interaction. This

might explain why, in contrast to stacking interactions, such

contacts are highly infrequent in protein/aminoglycoside or

RNA/aminoglycoside complexes. In agreement with this idea,

a recently reported study shows that the inﬂuence of direct

NH₃⁺/π bonds on the stability of a speciﬁc protein/ligand

(28) Salonen, L. M.; Bucher, C.; Banner, D. W.; Haap, W.; Mary, J.-L.;

Benz, J.; Kuster, O.; Seiler, P.; Schweizer, W. B.; Diederich, F. Angew.

Chem., Int. Ed. 2009, 48, 811–814.

(27) Su, P.; Li, H. J. Chem. Phys. 2009, 131, 1–15.

9

12086 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

slightly favored. Interestingly, protonation of the aminocyclitol

ring has a remarkable inﬂuence, not only on the complex

stability but also on its preferred geometry. In the gas phase,

both properties seem to be determined by attractive electrostatic

interactions between the aromatic system and the charged

aminocyclitol ring. In contrast, the energy cost associated to

objective, an arbitrary geometry of a simple aminocyclitol/

benzene complex (that of minimum c, see Figure 13) was

selected and the dissociation potential was evaluated both, in

the gas phase and implicit water. Potential energy curves were

calculated at the SCS-MP2/6-311G(2d,p) level through nonre-

laxed scans of the distance between the H1 hydrogen of the

aminocyclitol moiety (both in neutral and protonated states) and

the center of the benzene ring. The obtained results are

represented in Figure 13c. Two speciﬁc features of the obtained

proﬁles deserve special attention. First, in water, formation of

the charged complex proceeds through an energy barrier, which

can be related to the penalty associated with charge desolvation.

On the contrary, these processes are barrierless in the neutral

state at this theoretical level. Second, in all cases, the potential

energy curve derived for the charged complex exhibits a more

pronounced slope in the proximity of the stacked geometry,

which represents narrower energy minima. These results might

+

NH₃desolvation is the dominant effect, when water is

implicitly considered. In agreement with this view, nonparallel

arrangements (c and d), which place the phenyl ring closer to

+

the charged NH₃functions, are clearly preferred in the gas

phase. In water, these stacking modes are more severely

destabilized by the charge desolvation penalty, leading to a

complete reversal in the geometrical preferences of the ami-

nocyclitol/phenyl complex. Indeed, arrangement a, which place

the aromatic ring further from the NH₃⁺groups, represents the

most stable minima in aqueous solution. It should be noted that

both, in the gas phase and in water, energy differences between

the alternative stacking modes considered are larger for the

charged system, which implies that the protonation of the

aminocyclitol ring enhances the speciﬁcity of the complex.

The analysis previously described allowed an identiﬁcation

of the preferred arrangements for isolated aminocyclitol/benzene

pairs. However, in the context of a complete aminoglycoside,

NH₃⁺groups located in vicinal sugar units could also modulate

the stacking. Indeed, both, the experimental NMR data and the

MD trajectories collected for 2, indicate that position 2-I of this

derivative has a remarkable inﬂuence on the complex geometry.

In order to test this particular point, we also analyzed the

interaction of the (more realistic) complete pseudodisaccharide

I/II fragment with a toluene molecule. Minimum energy

geometries for the complexes, together with the corresponding

interaction energies in the gas phase and in water, were derived

considering two different protonation states of the model: that

with rings I/II fully charged (total charge +4) and that with

position 2-I selectively neutralized (total charge +3). The results

obtained are represented in Figure 13b. According to these data,

the fully protonated minima exhibit enhanced interaction

energies (-16.93 vs -12.10 and -20.42 vs -16.90 kcal/mol

for geometries a′ and c′, respectively) in the gas phase, implying

that long-range NH₃⁺/aromatic electrostatic interactions estab-

lished through the aminoglycoside position 2-I are globally

attractive. Moreover, this particular contact has a signiﬁcant

inﬂuence on the preferred geometry of the complexes. Thus,

while the nonparallel geometry (c′) is always favored (reﬂecting

the dominant role of the aminocyclitol/aromatic electrostatic

interactions) its relative stability with respect to the parallel

arrangement (a′) is reduced by 1.31 kcal/mol in the fully charged

model.

+

seem surprising considering that in water NH₃solvation

opposes CH/π interactions. However, it should be noted that,

for low amplitude ﬂuctuations, the empty space generated

between the stacked entities might not be solvent accessible so

that the disruption of the CH/π bonds cannot be compensated

+

by an improvement in the NH₃solvation. As a result, low

amplitude ﬂuctuations of the aromatic ring along this particular

reaction coordinate, are signiﬁcantly hampered by the proton-

ation of the aminocyclitol unit, even in presence of water. This

conclusion is in excellent agreement with the reduced rotation

rate exhibited by phenyl rings in charged derivative 2.

To summarize, in the gas phase or low dielectric environ-

ments, a delicate balance among different CH/aromatic interac-

+

tions enhanced by adjacent NH₃groups determines the

preferred geometry of the aminoglycoside/aromatic complexes.

This scenario is drastically altered when water solvation is taken

into account. Under these circumstances, the position of the

aromatic ring within the complex seems to be optimized to

globally minimize desolvation of the surrounding ammonium

+

functions. Finally, both in the gas phase and water, NH₃

functions seem to increase the complex speciﬁcity and reduce

its dynamic character, as a result of the enhanced electrostatic

attraction between the aminocyclitol and the aromatic ring.

Role of Aromatic Rings in the Molecular Recognition of

Aminoglycoside Antibiotics: Implications for Drug Design.

Finally, some comments regarding the implications of the results

described along the manuscript for the molecular recognition

of aminoglycoside antibiotics and the design of improved

derivatives should be made.

From a molecular recognition perspective, the apparent

conﬂict between the highly polar nature of these antibiotics and

the hydrophobic character of the aromatic side-chains frequently

found in aminoglycoside receptors is intriguing. Ammonium

groups are known to contribute importantly to the stabilization

of aminoglycoside complexes with RNA or proteins. Thus, they

participate in salt bridges or hydrogen bonds with the receptor

polar groups and interact electrostatically with the usually

strongly anionic binding pockets. In fact, aminoglycoside

binding to RNA/proteins is, in most cases, coupled to the

protonation of several amino functions of the ligand.²⁹It could

be hypothesized that under these circumstances the contribution

of stacking interactions to the complex stability might be reduced

It can be observed that the former geometry places the

+

aromatic ring further from the NH₃2-I, leading to a weaker

contact.

On the other hand, in water, NH₃⁺/aromatic interactions

become globally unfavorable, so that position 2-I strongly

destabilizes the parallel arrangement a′. In contrast with the

behavior predicted for the isolated aminocyclitol/phenyl pair

(see Figure 13a), now the nonparallel geometry (minimum c),

+

with the aromatic ring proximate to the aminocyclitol NH₃

groups, is preferred. These results are in good agreement with

the conclusions of both the NMR and MD analysis of 2

described in previous sections.

As a ﬁnal point, the inﬂuence of aminoglycoside charges on

the energy cost for low amplitude ﬂuctuations, implying the

transient disruption of CH/π bonds, was tested. With this

(29) (a) Kaul, M.; Barbieri, C. M.; Kerrigan, J. E.; Pilch, D. S. J. Mol.

Biol. 2003, 326, 1373–1387. Kaul, M.; Pilch, D. S. Biochemistry 2002,

41, 7695–7706.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12087

A R T I C L E S

Vacas et al.

Figure 14. (a) CH/π interactions involving CH-NH₃fragments can be stabilized by the N-methylation of the amino groups.²⁹The inﬂuence of this

particular modiﬁcation on the binding equilibrium is schematically represented. (b) N-Methyl groups are present in a minor fraction of naturally occurring

aminoglycosides, as streptomycin (shown on the left). The recently described structure of this antibiotic in complex with the resistance enzyme APH(2′)

(pdb code 3HAV) reveals a clear interaction between the antibiotic N-methylated glucose unit and a tryptophan side-chain.

+

with respect to that usually observed for neutral oligosaccha-

rides. However, this will largely depend on the extent of

ammonium desolvation promoted by the hydrophobic aromatic

rings in the complexed state. The desolvation penalty would

ꢀ-hairpin peptides and speciﬁc protein/ligand complexes.^28,30

Despite the evident potential of this strategy within the

aminoglycoside ﬁeld, its scope and limitations have not been

explored yet.

Finally, it should be mentioned that, N-methyl groups are

present in a minor fraction of natural aminoglycosides, as

gentamycin, spectinomycin or streptomycin. Interestingly, the

recently described X-ray structure of streptomycin in complex

with resistance enzyme APH(2′) reveals a unique aminoglyco-

side/aromatic contact that precisely involves the antibiotic

N-methylated aminoglucose unit (Figure 14b). Probably, this

speciﬁc N-methyl group plays a signiﬁcant role in the stabiliza-

tion of the enzyme/substrate complex.

+

be expected to be relatively minor for those particular NH₃

groups already hidden to solvent by extensive polar contacts

with the receptor. In these cases, stacking and polar interactions

might operate synergistically to stabilize the complexes (Figure

S17). Thus, by neutralizing the positive charge of the ammonium

+

groups and promoting their total desolvation, the NH₃/receptor

contacts would favor the stacking of the aminoglycoside rings

with aromatic systems. In turn, the hydrophobic environment

provided by the aromatic residues might cooperatively enhance

the strength of the former polar interactions. On the contrary,

Conclusions

+

desolvation of unpaired NH₃groups of the antibiotic by the

We have performed a detailed experimental and theoretical

analysis of the role played by CH/π stacking interactions in

the molecular recognition of polycationic glycosides. With this

purpose, different aminoglycoside/aromatic pairs have been

selected as model systems. These comprised totally uncon-

strained bimolecular complexes together with several intramo-

lecular versions of the former, with decreasing degrees of

conformational freedom (derivatives 1 and 2). The results

obtained through this study demonstrate that:

(a) Aminoglycoside/aromatic CH/π interactions in the gas

phase (and presumably also in low dielectric environments) are

signiﬁcantly enhanced by the protonation of the antibiotic.

Favorable electrostatic forces occurring between the drug and

the aromatic quadrupole are at the origin of this effect.

receptor aromatic side-chains might be signiﬁcantly unfavorable.

However, even in these circumstances stacking CH/π interac-

tions could be essential to discriminate between alternative

binding modes, increasing the complex speciﬁcity. Indeed, our

analysis demonstrates that the ligand charges impose strict

requirements on the preferred geometry of the aminoglycoside/

aromatic complexes and reduce their dynamic character.

From a medicinal chemistry perspective, the obtained results

have interesting implications. Thus, they strongly suggest that

aminoglycoside binding to both RNA and protein receptors

might be signiﬁcantly enhanced by the N-methylation of speciﬁc

amino functions (Figure 14a). This modiﬁcation (originally

proposed by Waters et al. for the stabilization of CH/π

contacts)^30awould limit the energy penalty associated to

ammonium desolvation, while preserving the intense ligand/

receptor electrostatic interactions. Indeed, N-methylation has

been shown to stabilize lysine/tryptophan interactions in model

(30) (a) Hughes, R. M.; Waters, M. L. J. Am. Chem. Soc. 2005, 127, 6518–

6519. (b) Hughes, R. M.; Waters, M. L. J. Am. Chem. Soc. 2006,

128, 13586–13591. (c) Tatko, C. D.; Waters, M. L. J. Am. Chem.

Soc. 2004, 126, 2028–2034.

9

12088 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Molecular Recognition of Antibiotics

A R T I C L E S

+

MD Simulations. To get insights into the origin of the peculiar

behavior deduced for the different aminoglycoside/aromatic com-

plexes analyzed along the manuscript, we performed molecular

dynamics (MD) simulations with the AMBER 10.0 package.¹⁷The

conformational and dynamical properties of 1 (together with simpler

disaccharide models) and 2 were tested assuming different proton-

ation states of the aminoglycosides (see above). For highly ﬂexible

derivatives, such as 1, the simulation lengths were adjusted to 0.1

µs in order to ensure an adequate sampling of the different

conformational populations present in solution. In 2 the confor-

mational ﬂuctuations should be severely restricted by the ami-

noglycoside covalent scaffold. Consequently, simulation lengths

were limited to 20 ns.

Two different treatments were employed for solvation. Thus,

derivative 1 was analyzed in the presence of explicit TIP3P water,¹⁸

counterions, periodic boundary conditions and Ewald sums for the

treatment of long-range electrostatic interactions,¹⁹following a

protocol identical to that previously described.³²This methodology

was rather expensive from a computational perspective, specially

for 0.1 µs trajectories. However, it gave us the opportunity to

determine the changes in the aminoglycoside solvation shells that

accompany transitions between hairpin and disordered conforma-

tions. Indeed, this analysis provided valuable clues about the origin

of the hairpin destabilization in protonated 1.

For aminoglycoside 2 and simpler 1 models (1a and 1b), MD

trajectories were collected employing the Born model with the

Bashford modiﬁcation.²¹Continuous algorithms have been shown

to provide reliable estimates of solvations energies at a reasonable

computational cost. Moreover, lengthy equilibration periods are not

required and the simulations are free from the artifacts derived from

the use of periodic boundary conditions. Finally, this methodology

is closely related to that used in the Quantum Mechanics analysis

of simple aminocyclitol/aromatic complexes, described in the last

section of the manuscript. Therefore, results derived employing both

approaches could be easily compared.

In all cases, RESP atomic charges for the analyzed derivatives

in the different protonation states considered, were derived by

applying the RESP module of AMBER 10.0 package¹⁷to the HF/

6-31G(d) ESP charges calculated with Gaussian 09.²²The AM-

BER94 force ﬁeld was implemented with GLYCAM04³³and

GAFF³⁴parameters to accurately simulate the conformational

behavior of these molecules. The time step was 1 fs in all of the

simulations.

(b) On the contrary, in water, aminoglycoside NH₃groups

destabilize the complexes. The energy penalty associated to

desolvation of the antibiotic charged functions, promoted by

the aromatic systems explains this behavior.

(c) Ammonium groups increase the speciﬁcity of the ami-

noglycoside/aromatic complexes and reduce their internal

mobility both, in the gas phase and in water. More speciﬁcally,

low amplitude ﬂuctuations implying the transient disruption of

the CH/π bonds are signiﬁcantly hampered by the protonation

of the antibiotic. The stiffening of the complex reﬂects the

enhancement of the electrostatic interactions between the

charged aminosugar/aminocyclitol rings and the aromatic

system. This conclusion is in good agreement with results

previously reported by Waters et al. for lysine/tryptophan CH/π

interactions in model peptides.^30b,c

(d) Electrostatic forces and charge desolvation are also the

dominant factors that determine the facial selectivity and

preferred geometry of the aminoglycoside/aromatic complexes,

in gas phase and water, respectively.

The results reported herein contribute to the understanding

of the driving forces behind aminoglycoside recognition and

should be carefully considered in the design of improved

aminoglycoside based ligands.

Materials and Methods

The Synthesis of derivatives 1 and 2 together with a detailed

characterization of the synthetic intermediates is included in the

Supporting Information.

NMR Experiments. NMR experiments were recorded on Bruker

Avance 600 and Varian Unitiy 500 spectrometers equipped with

cryoprobes at 278-313 K.

The interaction of benzyl alcohol (200 mM) with kanamycin-B

(20 mM) and 2-deoxy-streptamine (30 mM) was tested at 25 °C

using HSQC experiments. The chemical shift perturbations induced

at the different positions of these aminoglycosides by the aromatic

probe were monitored at pH 10.3 and 3.3.

Samples of 1 and 2 were prepared in D₂O at 0.1-1.0 mM

concentrations and several pH values in the 3.0-10.5 range.

Dilution experiments were performed (up to 20 µM) in order to

rule out the presence of a signiﬁcant aggregation of the aminogly-

coside under the employed experimental conditions. For derivative

1, DOSY experiments¹⁶were also collected under different

experimental conditions to prove its monomeric nature (Figure S6).

The ¹H and ¹³C assignments of 1 and 2 were derived from a set

of 2D-NOESY, TOCSY and DQF-COSY experiments carried out

in the phase-sensitive mode using the TPPI method. For the NOESY

experiments, mixing times from 100 to 800 ms were employed.

The basicity of the amino moieties present in 1, 2 and several

reference aminoglycosides (including kanamycin-B, see Figures 6,

8, S4, S10, S11, S13, and S14), at 40 °C were derived by NMR. In

order to unambiguously determine the pK_a’s, carbon chemical shifts

For derivative 2, the theoretical chemical shift perturbations (∆δ)

induced at key aminocyclitol positions by the phenyl rings were

calculated from the MD trajectories and compared with those

measured experimentally. A set of 40 conformations (including 80

aminocyclitol/phenyl pairs) was extracted from each trajectory and

the theoretical ∆δ values, evaluated using the Johnson-Bovey

model³⁵as implemented in the program MOLMOL,³⁶averaged

over the entire ensemble.

Quantum Mechanics Calculations. The interaction energies of

the different aminocyclitol/aromatic pairs considered were evaluated

employing Quantum Mechanics calculations, assuming different

protonation states of the aminoclyclitol and different interaction

geometries (including parallel and nonparallel stacking and NH₃⁺/π

interactions). Full geometry optimizations were carried out with

the Gaussian 09 package²²through the M06-2X functional recently

developed by Thrular²³and the triple-ꢁ TZVP basis set of

1

together with the heteronuclear J_HCconstants for the HC-N

fragments, were also measured at different pH values from coupled

¹H-¹³C HSQC experiments. This later parameter has been shown

to be sensitive to the protonation state of the amino functions and

is not affected by the ring current effect of the aromatic systems.¹⁴

As a ﬁnal control, the protonation state of 1 and 2 was checked in

the presence of reference aminoglycoside kanamycin-B, employing

equimolecular mixtures. In this way, any inﬂuence of the sample

conditions on the pK_aperturbations detected could be ruled out.

Selective 1-D NOE experiments were recorded at different

temperatures and pH values employing the 1D-DPFGSE NOE pulse

sequence.³¹

(32) Corzana, F.; Busto, J. H.; Engelsen, S. B.; Jime´nez-Barbero, J.;

Asensio, J. L.; Peregrina, J. M.; Avenoza, A. Chem.sEur. J. 2006,

12, 7864–7871.

(33) Woods, R. J.; Dwek, R. E.; Edge, C. J.; Fraser-Reid, B. J. Phys. Chem.

B 1995, 99, 3832–3846.

(34) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A.

J. Comput. Chem. 2004, 25, 1157–1174.

(35) Case, D. A. J. Biomol. NMR 1995, 6, 341–346.

(36) Koradi, R.; Billeter, M.; Wuthrich, K. J. Mol. Graph. 1996, 14, 51–

55.

(31) Stott, K.; Stonehouse, J.; Keeler, J.; Hwang, T. L.; Shaka, A. J. J. Am.

Chem. Soc. 1995, 117, 4199–4200.

9

J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010 12089

A R T I C L E S

Vacas et al.

Ahlrichs.²⁴Solvent effects were included in the optimization trough

the integral equation formalism variant of the polarizable continuum

model (IEF-PCM),²⁵using UFF radii and the internally stored

parameters for water, as implemented in Gaussian 09. Frequency

analyses were carried out at the same level used in the geometry

optimizations, and the nature of the stationary points was determined

in each case according to the appropriate number of negative

eigenvalues of the Hessian matrix. Gas-phase single-point energies

at the spin-component scaled MP2 (SCS-MP2) level as proposed

by Grimme³⁷with the triple-ꢁ 6-311G(2d,p) basis set were

calculated on the optimized geometries using Gamess (v.2009).³⁸

Basis set superposition errors (BSSE) were corrected by the

Boys-Bernardi counterpoise method.³⁹Solvation free energies

(∆G_solv) were calculated at the ab initio level through the conductor-

like polarized continuum model (C-PCM),²⁶which is based on the

COSMO⁴⁰method, using the internally stored parameters for water

as implemented in Gamess. Counterpoise-corrected electronic

energies were used for the discussion on the interaction energies

between the fragments both in the gas phase (∆Ei_gas) and in solution

(∆Ei_{H O}) ∆Ei_gas+ ∆G_solv). Diverse contributions (electrostatic,

2

exchange, repulsion, polarization and dispersion) to the interaction

energies were calculated through the localized molecular orbital

energy decomposition analysis (LMO-EDA)²⁷as implemented in

Gamess.

Acknowledgment. This investigation was supported by the

Spanish “Plan Nacional” (MCYT) (Grant CTQ2007-67403/BQU)

and the Comunidad de Madrid (Grant S2009/PPQ-1752). We also

thank CESGA (Santiago de Compostela) for computer support.

Supporting Information Available: Experimental procedures

for the synthesis of derivatives 1 and 2. Characterization of

synthetic intermediates and products. Figures S1-S17, showing

details of the experimental and theoretical analysis of aminogly-

coside/aromatic interactions described along the manuscript.

Complete refs 17 and 22. This material is available free of

charge via the Internet at http://pubs.acs.org.

(37) Grimme, S. J. Chem. Phys. 2003, 118, 9095–9102.

(38) (a) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.;

Gordon, M. S.; Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen,

K. A.; Su, S. J.; Windus, T. L.; Dupuis, M.; Montgomery, J. A.

J. Comput. Chem. 1993, 14, 1347–1363. (b) Gordon, M. S.; Schmidt,

M. W. In Theory and Applications of Computational Chemistry,

the ﬁrst forty years; Dykstra, C. E., Frenking, G., Kim, K. S., Scuseria,

G. E., Eds.; Elsevier: Amsterdam, 2005; Chapter 41, pp 1167-

1189.

JA1046439

(39) Boys, S. F.; Bernardi, F. Mol. Phys. 1970, 19, 553–566.

(40) Klamt, A.; Schu¨u¨rmann, G. J. Chem. Soc. Perkin Trans 2 1993, 799–

805.

9

12090 J. AM. CHEM. SOC. VOL. 132, NO. 34, 2010

Article Doi

Role of aromatic rings in the molecular recognition of aminoglycoside antibiotics: Implications for drug design

DOI: 10.1021/ja1046439

Source and publish data:

Authors:

Article abstract of DOI:10.1021/ja1046439

Full text of DOI:10.1021/ja1046439

Products guided by the article

R&D Labs maybe for 1241744-89-7

Relevant to this article

Hot Product