Activation of antibacterial prodrugs by peptide deformylase

Article abstract of DOI:10.1016/S0960-894X(00)00167-0

5'-Dipeptidyl derivatives of 5-fluorodeoxyuridine (FdU) (1a-d) were synthesized. These compounds are biologically inactive but can be activated by peptide deformylase, which removes the N-terminal formyl group of the dipeptide, to release the active drug FdU via an intramolecular cyclization reaction. Because the deformylase is ubiquitous among bacteria but absent in mammalian cells, 1a-d provide a novel class of potential antibacterial agents. (C) 2000 Elsevier Science Ltd. All rights reserved.

Full text of DOI:10.1016/S0960-894X(00)00167-0

Bioorganic & Medicinal Chemistry Letters 10 (2000) 1073±1076
Activation of Antibacterial Prodrugs by Peptide Deformylase
Yaoming Wei and Dehua Pei*
Department of Chemistry and Ohio State Biochemistry Program, The Ohio State University, 100 West 18th Avenue, Columbus,
OH 43210, USA
Received 6 January 2000; accepted 10 March 2000
AbstractÐ50-Dipeptidyl derivatives of 5-¯uorodeoxyuridine (FdU) (1a±d) were synthesized. These compounds are biologically
inactive but can be activated by peptide deformylase, which removes the N-terminal formyl group of the dipeptide, to release the
active drug FdU via an intramolecular cyclization reaction. Because the deformylase is ubiquitous among bacteria but absent in
mammalian cells, 1a±d provide a novel class of potential antibacterial agents. # 2000 Elsevier Science Ltd. All rights reserved.
The emergence of bacterial pathogens that are resistant
to multiple classes of existing antibiotics has created an
urgent demand for new antibacterial agents with novel
mechanisms of action.1 Recent studies from several
laboratories have suggested that peptide deformylase
(PDF), an enzyme responsible for removing the N-
terminal formyl group from newly synthesized polypep-
tides, may be a suitable target for antibacterial drug
design.2,3 While PDF is present in all bacteria and
essential for bacterial survival,2 it is apparently absent
in mammalian cells.3 Indeed, PDF inhibitors have
recently been shown to act as broad-spectrum anti-
bacterial agents.4 Here we report a novel, alternative
approach to antibacterial drug design, in which a phar-
macologically inactive prodrug is selectively processed
by PDF to release an active drug inside the bacterial
pathogen.
kinase (Scheme 1). Because of the unique presence of
PDF in bacterial cells, active drugs should be produced
only inside the bacterial cells, thus minimizing the cyto-
toxicity of these prodrugs to the host cells.
Prodrugs 1a±d were prepared from the t-BOC-protected
amino acids 5a±d, which were selectively coupled to the
50-OH of FdU (4) via Mitsunobu condensation7 to
a€ord carboxylic esters 6a±d (Scheme 2). Treatment of
6a±d with tri¯uoroacetic acid (TFA) followed by con-
densation with N-formylmethionine (for 1c) or its acti-
vated ester 7 led to prodrugs 1a±d in 25±73% overall
yields (from 5).
Prodrug 1a, which contains an l-proline as the penulti-
mate residue, is somewhat unstable under aqueous con-
ditions (20 mM sodium phosphate, pH 7.4); incubation
for 18 h at 37 ꢀC led to 16% degradation (Fig. 1, tracing
a), due to hydrolysis of the ester linkage and back-
ground deformylation followed by cyclization. Prodrug
1a was then tested for enzymatic degradation by adding
5% (®nal) fetal bovine serum to the phosphate bu€er.
Signi®cant degradation (55%) into FdU and two other
unidenti®ed species was observed after 18 h of incuba-
tion (Fig. 1, tracing b). Judging from the elution pattern
of the products, the cleavage occurred primarily at the
ester linkage to produce FdU and formyl-Met-Pro.
To demonstrate the viability of this approach, the 50-
OH group of 5-¯uorodeoxyuridine (FdU) was acylated
with an N-formylated dipeptidyl unit to give com-
pounds 1a±d (Scheme 1) as potential prodrugs. In vivo,
FdU is converted by thymidine kinase into FdU 50-
phosphate, which is cytotoxic by inhibiting thymidylate
synthetase.5 As acylation of the 50-OH prevents 50-
phosphorylation of FdU, 1a±d should have little or no
inhibitory activity. However, when bacterial cells take
up 1a±d, their PDF would remove the N-formyl group,
thereby triggering the onset of a cyclization reaction6 to
release FdU, which would be converted into the active
thymidylate synthetase inhibitor in situ by thymidine
To improve the prodrug stability against protease/
esterase action, unnatural amino acids were used to
construct prodrugs 1b±d (Scheme 1). l-Methionine is
still used as the N-terminal residue, because PDF
strongly prefers an l-methionine at this position.8 11 At
the second position, d-proline (1b), N-methyl-l-leucine
(1c), and a-aminoisobutyrate (1d) were substituted for
*Corresponding author. Tel.: +1-614-688-4068; fax: +1-614-292-
1532; e-mail: pei@chemistry.ohio-state.edu
0960-894X/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0960-894X(00)00167-0

1074
Y. Wei, D. Pei / Bioorg. Med. Chem. Lett. 10 (2000) 1073±1076
Prodrug 1d, which contains an a-methylalanine at the
penultimate position, shows dramatically improved sta-
bility against both chemical and enzymatic hydrolysis
(Fig. 1, c and d). After 18 h of incubation at 37 ꢀC in the
presence or absence of 5% fetal bovine serum, only a
trace amount of FdU (<1%) was observed. Also, when
1d was treated with a crude Escherichia coli cell lysate,
deformylation and cyclization were the only reactions
observed (vide infra).
Compounds 1a±d were then tested for activation by
puri®ed PDF in vitro. In the presence of E. coli PDF,
1a±d were rapidly converted into deformylated inter-
mediates 2a±d (Fig. 2A), with kcat=KM values ranging
1
1
from 750 to 2.9Â104 M
s
(Table 1). These activities
Scheme 1.
are comparable to those reported for short N-for-
mylmethionyl peptide substrates.8 10 Since in vivo
deformylation is believed to be co-translational and
therefore the physiological substrates of PDF are likely
short unfolded peptides,12 prodrugs 1a±d should com-
pete favorably with the endogenous PDF substrates in
bacterial cells. Following the PDF reaction, the inter-
mediates 2a±d underwent facile, spontaneous intramo-
lecular aminolysis to release FdU (retention time t=7.3
min) and diketopiperazines 3a±d (t=17.0 min for 3a)
(Fig. 2A). The release of FdU displayed ®rst-order
l-proline. Previous studies have shown that PDF exhibits
broad substrate speci®city at the penultimate position and
can tolerate d-amino acids.9,10 Unfortunately, while 1b
and 1c are much more resistant to enzymatic degrada-
tion than the parent compound (1a), their formyl group
is susceptible to chemical hydrolysis. Incubation in the
20 mM phosphate bu€er for 18 h at 37 ꢀC resulted in
64 and 54% degradation for 1b and 1c, respectively.
Scheme 2.
Figure 2. (A) HPLC analysis of activation of prodrug 1a. At time 0,
67 mg of E. coli PDF was added to ꢁ1 mM 1a in 100 mL of 20 mM
sodium phosphate (pH 7.4). At various times, 10-mL aliquots were
withdrawn, quenched by the addition of an equal volume of 100 mM
TFA, and loaded onto an HPLC instrument (C18 column; monitored
at 214 nm). a, t=60 min control without PDF; b, t=1 min; c, t=15
min; d, t=60 min; and e, FdU standard. The retention times of 1a, 2a,
3a, and FdU were 19.7, 18.2, 17.0, and 7.3 min, respectively. The
small peak resistant to deformylation (t=20.1 min) is f-(d-Met)-Pro-
FdU. (B) First-order kinetics of cyclization of 2a±d. A0, concentration
of 2a±d at time zero (before cyclization); A, concentration of 2a±d at
time t.
Figure 1. HPLC analysis (monitored at 214 nm) of prodrug degrada-
tion (18 h at 37 ꢀC). a, 1a+20 mM phosphate bu€er (pH 7.4); b,
1a+phosphate bu€er+5% fetal bovine serum; c, 1d+phosphate
bu€er; d, 1d+phosphate bu€er+5% serum. The column (C18) was
eluted with a linear gradient of 0±40% acetonitrile in water over 20
min. Retention times for FdU, formyl-Met-Pro, 1a and 1d were 6.5,
13.9, 14.3, and 14.5 min, respectively. The smaller peak with retention
time of 14.7 min (in c and d) was due to racemization of 1d at the
methionyl residue during synthesis.

Y. Wei, D. Pei / Bioorg. Med. Chem. Lett. 10 (2000) 1073±1076
Table 1. Kinetic data of prodrug activation (pH 7.4)
1075
reversed-phase HPLC. All four compounds were
rapidly deformylated by the endogenous PDF followed
by cyclization and FdU release. Since 1d was most
stable against undesired chemical and enzymatic degra-
dation, it was characterized in more detail. As shown in
Figure 3, treatment of 1d with the crude cell lysate for 1
h resulted in complete disappearance of the 1d peak
(t=19.0 min). At the mean time, a new peak appeared
at t=15.1 min when monitored at 214 nm, which was
invisible when monitored at 260 nm. This species was
assigned as diketopiperazine 3d, as it was not derived
from the cell lysate, co-eluted with 3d derived from
treatment of 1d with puri®ed PDF, and had a molecular
mass of 216.09 Da (determined by mass spectrometry
after HPLC puri®cation). Interestingly, the peak corre-
sponding to free FdU (retention time t=7.3 min under
the conditions) was barely detectable (Fig. 3, tracing b).
Examination of the HPLC chromatograms revealed an
increase in intensity for a peak at t=2.8 min (peak 3) as
1d was hydrolyzed by PDF. Analysis of peak 3 by fast
atom bombardment mass spectrometry gave two peaks
at m/z=327.17 (MH+) and 371.22 Da (M++2Na),
respectively, consistent with the structure of FdU 50-
phosphate. Peak 3 and its neighboring peaks (peaks 1,
2, 4, and 5) were then individually collected, treated
with alkaline phosphatase, and again analyzed by
HPLC. While phosphatase treatment of peaks 1, 2, 4,
and 5 resulted in no change in the unknown species,
treatment of peak 3 led to a peak at t=7.3 min, the
characteristic retention time of FdU under the condi-
tions (the four natural nucleosides all have di€erent
retention times) (Fig. 3, tracing c). Thus, as FdU was
generated from the cyclization reaction, it was con-
verted by thymidine kinase in the crude lysate into the
more hydrophilic FdU 50-monophosphate, the active
thymidylate synthetase inhibitor. Note that there was
no degradation of 1d by any other cellular enzymes.
Taken together, the above results demonstrate that the
endogenous PDF activity is adequate for the activation
of prodrugs 1a±d and inecient deformylation is unli-
kely the reason for the low potency of 1a±d for cell
growth inhibition. Rather, it is the poor transport of
prodrugs 1a±d across the bacterial cell membrane(s)
Prodrug
Deformylation kcat=KM
Cyclization t1=2 (min)
1
(Â103 M 1 s
)
1a
1b
1c
1d
12.1Æ0.4
0.75Æ0.03
1.65Æ0.06
28.9Æ1.4
15.2
4.68
2.07
51.1
kinetics (t1=2=2±51 min), consistent with an intramole-
cular reaction (Fig. 2B and Table 1). Diketopiperazine
3a was collected from HPLC runs and its identity was
con®rmed by mass spectrometry.
Incorporation of unnatural amino acids into 1b±d exer-
ted some interesting e€ects on the kinetics of deformy-
lation and cyclization reactions. The use of a d-proline
(1b) reduced the deformylation rate by 16-fold (relative
to 1a) but increased the cyclization rate by 3.2-fold
(t1=2=4.7 min).6a Similarly, N-methylleucine (1c) slowed
down the PDF reaction by 7.3-fold, while increasing
the cyclization rate by 7.3-fold (t1=2=2.1 min).6b On
the other hand, the use of a-aminoisobutyrate (1d)
increased the PDF reaction rate by 2.4-fold, but reduced
the cyclization rate by 3.4-fold (t1=2=51 min). These
results suggest that it may be possible to ®ne tune the
rate of active drug release by varying the structure of the
penultimate amino acid.
In order to test their ability to inhibit bacterial growth,
compounds 1a±d were added directly to a culture of E.
coli BL21(DE3) cells in Luria broth. They resulted in
only modest inhibition of cell growth at high con-
centrations (IC50 > 100 mM). This poor antibacterial
activity could be caused by several factors. First, the
prodrugs may be poorly transported into the bacterial
cells. Second, the prodrugs may not be eciently defor-
mylated by PDF due to competition with the endogen-
ous substrates. Third, the prodrugs may be rapidly
degraded by intracellular enzymes other than PDF. We
therefore tested the prodrugs with BL21(DE3) cells
transformed with the overproducing plasmid pET-22b-
def.9 In the presence of 100 mM inducer, isopropyl-b-d-
thiogalactopyranoside, these cells produce PDF as
ꢁ50% of their total cellular protein; any prodrugs that
have entered these cells should be rapidly and com-
pletely deformylated. However, the growth inhibition
pro®le of these overproducing cells was similar to that
of the control BL21(DE3) cells. Prodrug 1d was also
tested against Staphylococcus epidermidis, a Gram-posi-
tive bacterium which is generally more sensitive to anti-
biotic treatment. Indeed, prodrug 1d resulted in stronger
growth inhibition with an IC50 value of ꢁ50 mM. Since
1d is stable against both chemical and enzymatic degra-
dation (other than PDF), these results suggest that
inecient drug transport across the cell membrane(s) is
likely the cause of poor antibacterial activity.
Figure 3. HPLC analysis of activation of prodrug 1d by crude E. coli
lysate (C18; 214 nm). a. 1d only; b. 1d treated with cell lysate for 1 h,
followed by 3 h incubation at room temperature (for cyclization to
complete); and c. treatment of peak 3 derived from b with alkaline
phosphatase. The small peak with retention t=19.5 min (in a and b) is
due to racemization of 1d at the methionyl residue; the resulting iso-
mer is resistant to PDF action.
To further examine this issue, 1a±d were treated with a
crude bacterial cell lysate prepared from wild-type E.
coli BL21(DE3) cells that had been grown to the mid-
log phase. The cleavage products were analyzed by

1076
Y. Wei, D. Pei / Bioorg. Med. Chem. Lett. 10 (2000) 1073±1076
that has limited the intracellular concentration of the
prodrugs.
agent. Although the compounds described here lack
potent in vivo antibacterial activity due to poor mem-
brane permeability, redesign with more hydrophobic
drugs or those that can be actively taken up by the
transport systems will likely provide a new class of
potent antibiotics. Such studies are currently under way.
The prodrug approach o€ers several advantages over
the inhibitor approach. First, the active drug is gener-
ated only inside a bacterial pathogen where active PDF
is present. Such drugs should have lower toxicity to a
eukaryotic host. Second, the prodrug approach is
extremely versatile. In principle, any agent that blocks
cell growth and requires a free nucleophile (e.g., a
hydroxyl group) for toxicity can be masked by an acyl
group to eliminate or reduce its toxicity. Once taken up
by a bacterial pathogen, the acyl group is removed to
release the cytotoxic agent, which may target any one of
the bacterial proteins essential for survival. Third, pro-
drugs of this type are more likely to be broad-spectrum
antibiotics relative to PDF inhibitors. Molecular clon-
ing and genomic sequencing have revealed a few dozen
PDF genes; their protein sequences show a wide range
of homologies (20% to 65%).13 Although the active-site
structure seems to be well conserved, a subtle di€erence
at or near the active site could render a potent inhibitor
for PDF from one organism totally inactive against a
di€erent pathogen. The sequence diversity should be
less of a concern for the prodrugs, because PDF is a
broad-speci®city enzyme that deformylates thousands
of bacterial proteins and the prodrugs are close mimics
of these natural substrates. Finally, bacterial cells are
less likely to develop resistance to the prodrugs than to
a PDF inhibitor. The sequence diversity of PDF pre-
dicts that a mutation in the substrate binding site (not
the active site) may be tolerated, although such a muta-
tion might change the substrate speci®city. Such a
mutation will likely a€ect or even abolish the inhibitor
binding, causing drug resistance. Cells with such a
mutation would still, however, be sensitive to the pro-
drugs, which are substrates of the deformylase. If a cell
encounters a mutation that prevents prodrug activation,
the mutant enzyme would likely fail to deformylate the
cell's own proteins, resulting in a fatal consequence.
Acknowledgement
Support of this work was provided by the National
Institutes of Health (Grant 1R01AI40575).
References
1. (a) Neu, H. C. Science 1992, 257, 1064. (b) Davis, J. D.
Science 1994, 264, 375. (c) Spratt, B. G. Science 1994, 264,
388.
2. (a) Mazel, D.; Pochet, S.; Marliere, P. EMBO J. 1994, 13,
914. (b) Meinnel, T.; Blanquet, S. J. Bacteriol. 1994, 176, 7387.
3. Rajagopalan, P. T. R.; Yu, X. C.; Pei, D. J. Am. Chem. Soc.
1997, 119, 12418.
4. Huntington, K. M.; Yi, T.; Wei, Y.; Pei, D. Biochemistry, in
press.
5. Santi, D. V.; McHenry, C. S.; Raines, R. T.; Ivanetich, K.
M. Biochemistry 1987, 26, 8606.
6. (a) Gisin, B. F.; Merri®eld, R. B. J. Am. Chem. Soc. 1972,
94, 3100. (b) Khosla, M. C.; Smeby, R.; Bumpus, F. M. J. Am.
Chem. Soc. 1972, 94, 4721. (c) Rothe, M.; Mazanek, J. Liebigs
Ann. Chem. 1974, 439. (d) Pedroso, E.; Grandas, A.; de las
Heras, X.; Eritja, R.; Giralt, E. Tetrahedron Lett. 1986, 27,
743. (e) Wipf, P.; Li, W.; Sekhar, V. Bioorg. Med. Chem. Lett.
1991, 1, 745.
7. Mitsunobu, O. Synthesis 1981, 1.
8. (a) Adams, J. M. J. Mol. Biol. 1968, 33, 571. (b) Livingston,
D. M.; Leder, P. Biochemistry 1969, 8, 435.
9. Rajagopalan, P. T. R.; Datta, A.; Pei, D. Biochemistry 1997,
36, 13910.
10. Hu, Y.-J.; Wei, Y.; Zhou, Y.; Rajagopalan, P. T. R.; Pei,
D. Biochemistry 1999, 38, 643.
11. Meinnel, T.; Patiny, L.; Ragusa, S.; Blanquet, S. Biochem-
istry 1999, 38, 4287.
In conclusion, we have conceptually demonstrated a
novel approach to antibacterial drug design which takes
advantage of the unique bacterial PDF to convert a
biologically inactive prodrug into an active antibacterial
12. Pine, M. J. Biochim. Biophys. Acta 1969, 174, 359.
13. (a) Mazel, D.; Coic, E.; Blanchard, S.; Saurin, W.; Marliere,
P. J. Mol. Biol. 1997, 266, 939. (b) Meinnel, T.; Lazennec, C.;
Villoing, S.; Blanquet, S. J. Mol. Biol. 1997, 267, 749.