Communications
affinity to the therapeutic target.[3] The development of high-
throughput methods to identify compounds that weakly bind
to resistance-causing proteins and strongly bind to therapeu-
tic targets would facilitate the discovery of improved anti-
biotics. Such methods also allow for a better understanding of
the interactions of antibiotics with therapeutic targets and
resistance-causing proteins. Microarrays,[4–8] created by the
immobilization of small molecules onto glass surfaces,
provide a versatile platform for rapidly screening several
thousand potential antibiotics in parallel for binding to
therapeutic targets and resistance-causing proteins. These
screens are particularly attractive since only miniscule
amounts of potential antibiotics, therapeutic targets, and
resistance-causing proteins are needed, and thus the limita-
tions of current screening methods can be overcome.
Here we report the construction of aminoglycoside
microarrays and their use in probing the binding of amino-
glycosides to resistance-causing proteins. Two aminoglycoside
acetyltransferases that cause antibiotic resistance, 2’-acetyl-
transferase (AAC(2’)) from Mycobacterium tuberculosis[9]
and 6’-acetyltransferase (AAC(6’)) from Salmonella enter-
ica[10] were used as examples. Hybridization of these enzymes
to aminoglycoside arrays show that each immobilized amino-
glycoside interacts with both AAC(2’) and AAC(6’). Fur-
thermore, delivery of as little as picomoles of aminoglycoside
to slides was sufficient for binding to be observed (Figure 1).
The signal from mannose, a negative control, was much lower,
indicating that specific interactions were detected.
Binding of aminoglycoside antibiotics to resistance
enzymes can be grouped into two categories based on the
fluorescence signal from the highest concentration spot
(Figure 1). For AAC(6’), the strongest fluorescence arose
from binding to amikacin, tobramycin, kanamycin B, livido-
mycin, neomycin, and ribostamycin. Kanamycin A, apramy-
cin, paromomycin, gentamycin, and neamine resulted in
lower signals. For AAC(2’), the strongest signals were
observed for amikacin, paromomycin, tobramycin, and ribos-
tamycin while kanamycin B, apramycin, kanamycin A, nea-
mine, neomycin, lividomycin, and gentamycin displayed
weaker signals.
Figure 1. Top: Aminoglycoside microarray after hybridization with
AAC(6’) (green) and AAC(2’) (blue). Bottom: Fluorescence intensities
for the arrays of antibiotic hybrids after binding to the aminoglyco-
sides.
To determine how hybridization of AAC(6’) to the
aminoglycoside arrays correlates with protein–aminoglyco-
side binding measurements in solution, we made comparisons
In an effort to find inhibitors of antibiotic-resistance
enzymes, a library of aminoglycoside mimetics was synthe-
sized, arrayed, and screened for tight binding to by AAC(2’)
and AAC(6’). Guanidinoglycosides[12,13] are an attractive set
of aminoglycoside analogues for several reasons: 1) They are
readily synthesized from aminoglycosides. 2) Their increased
positive charge may allow them to bind more tightly to the
aminoglycoside binding pocket present in resistance-causing
enzymes that contain several negatively charged amino
acids.[14] 3) The difference in the pKa values of guanidino
(pKa ~ 12.5) and amino groups (pKa ~ 8.8) suggests that
guanidinoglycosides may not be substrates for AAC(2’) and
AAC(6’).
to
a calorimetric study of aminoglycoside binding to
AAC(6’).[11] This analysis shows a strong correlation between
these different types of measurements. Calorimetry showed
that ribostamycin, tobramycin, lividomycin, and neomycin
have the strongest affinities, whereas kanamycin B, paromo-
mycin, gentamycin C, kanamycin A, and amikacin had lower
affinities to AAC(6’). Generally the array data and the
calorimetry data correlate well when the affinities are
classified as strong and stronger with the exception of
amikacin, which had one of the strongest fluorescence signals
with both enzymes. Amikacin, unlike most of the other
aminoglycosides, contains two primary amino groups; the
sterically least encumbered amino group is six carbon atoms
removed from the 2,4-deoxystreptamine core. Thus, amikacin
may be immobilized onto the surface differently than the
other aminoglycosides, giving rise to the more intense signal.
A diverse set of guanidinoglycosides was synthesized
(Scheme 1) by reacting each aminoglycoside with Boc-b-Ala-
OSu to introduce a primary amino linker for immobilization,
which was used to normalize surface loading for each library
component. Guanidinylation using N,N’-di(Boc)-N’’-triflyl-
1592
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2004, 43, 1591 –1594