Simple amphiphilic α-hydrazido acids: Rational design, synthesis, and in vitro bioactivity profile of a novel class of potential antimicrobial compounds

Article abstract of DOI:10.1016/j.ejmech.2020.112072

Pursuing the search for a new class of structurally simple mimics of antimicrobial peptides, we optimized a short, cheap and high-yielding synthesis of mono-charged amphiphilic α-hydrazido acid derivatives. The most active derivatives furnished MICs that are among the best values reported in literature for synthetic amphiphilic membranolytic compounds. They exhibited a broad-spectrum in vitro activity against a variety of Gram-positive and Gram-negative bacteria, including two multidrug-resistant strains. In spite of the minimal cationic charge, the best compounds demonstrated to be selective toward bacterial cell membranes over mammalian cell membranes. The relationship between either the antibacterial or the hemolytic activity and the overall lipophilicity furnished an easy way to individuate the best dimensional range for the hydrophobic portions. The importance of a non-disrupted amphiphilicity was also demonstrated. Considering the bioactivity profile and the ease of synthesis, these chemically and proteolitically stable hydrochlorides are suitable for development of a new class of wide-spectrum antibiotics.

Full text of DOI:10.1016/j.ejmech.2020.112072

Journal Pre-proof
Simple amphiphilic α-hydrazido acids: Rational design, synthesis, and in vitro
bioactivity profile of a novel class of potential antimicrobial compounds
Paolo Amabili, Francesca Biavasco, Andrea Brenciani, Barbara Citterio, Dario
Corbisiero, Lucia Ferrazzano, Simona Fioriti, Giulia Guerra, Mario Orena, Samuele
Rinaldi
PII:
S0223-5234(20)30039-8
DOI:
https://doi.org/10.1016/j.ejmech.2020.112072
EJMECH 112072
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 25 November 2019
Revised Date: 10 January 2020
Accepted Date: 13 January 2020
Please cite this article as: P. Amabili, F. Biavasco, A. Brenciani, B. Citterio, D. Corbisiero, L. Ferrazzano,
S. Fioriti, G. Guerra, M. Orena, S. Rinaldi, Simple amphiphilic α-hydrazido acids: Rational design,
synthesis, and in vitro bioactivity profile of a novel class of potential antimicrobial compounds, European
Journal of Medicinal Chemistry (2020), doi: https://doi.org/10.1016/j.ejmech.2020.112072.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
©
2020 Published by Elsevier Masson SAS.

Simple Amphiphilic
α-Hydrazido Acids: Rational Design, Synthesis, and In Vitro
Bioactivity Profile of a Novel Class of Potential Antimicrobial Compounds
a, 1
b
c
b, *
d
Paolo Amabili , Francesca Biavasco , Andrea Brenciani , Barbara Citterio , Dario Corbisiero ,
d
c
a, 1
a
a, *
Lucia Ferrazzano , Simona Fioriti , Giulia Guerra , Mario Orena , Samuele Rinaldi
a
Department of Life and Environmental Sciences, Polytechnic University of Marche, Via Brecce
Bianche, 60131 Ancona, Italy.
b
Department of Biomedical Sciences and Public Health, Polytechnic University of Marche, Via Tronto
1
0/a, 60020 Torrette di Ancona, Italy.
c
Department of Biomolecular Science, Biotechnology Section, University of Urbino “Carlo Bo”, Via
Santa Chiara 27, 61032 Urbino, Italy.
d
Department of Chemistry ‘‘G. Ciamician’’ - Alma Mater Studiorum University of Bologna, Via
Selmi 2, 40126 Bologna, Italy.
*
Corresponding authors
E-mail addresses: s.rinaldi@staff.univpm.it (S. Rinaldi), barbara.citterio@uniurb.it (B. Citterio).
1
Paolo Amabili and Giulia Guerra contributed equally to this work. The manuscript was written
through contributions of all authors. All authors have given approval to the final version of the
manuscript.
ABSTRACT
Pursuing the search for a new class of structurally simple mimics of antimicrobial peptides, we
optimized a short, cheap and high-yielding synthesis of mono-charged amphiphilic α-hydrazido acid
1

derivatives. The most active derivatives furnished MICs that are among the best values reported in
literature for synthetic amphiphilic membranolytic compounds. They exhibited a broad-spectrum in
vitro activity against a variety of Gram-positive and Gram-negative bacteria, including two multidrug-
resistant strains. In spite of the minimal cationic charge, the best compounds demonstrated to be
selective toward bacterial cell membranes over mammalian cell membranes. The relationship between
either the antibacterial or the hemolytic activity and the overall lipophilicity furnished an easy way to
individuate the best dimensional range for the hydrophobic portions. The importance of a non-disrupted
amphiphilicity was also demonstrated. Considering the bioactivity profile and the ease of synthesis,
these chemically and proteolitically stable hydrochlorides are suitable for development of a new class
of wide-spectrum antibiotics.
KEYWORDS
Rational design
Cationic amphiphilic α-hydrazido acids
Wide-spectrum antibacterials
Overall lipophilicity
Non-disrupted amphiphilicity
Membranolytic compounds
ABBREVIATIONS
AMPs, antimicrobial peptides; HDPs, host defense peptides; CSAs, cationic steroid antimicrobials;
PNA, peptide nucleic acid; α-AApeptides, N-acylated-N-aminoethyl α-peptides; DFT, density
functional theory; IEF-PCM, integral equation formalism version of polarizable continuum model
method; NBO, natural bond orbital; AIM, atoms in molecules; DIPEA, N,N-diisopropylethylamine;
anh., anhydrous; DCM, dichloromethane; rt, room temperature; Boc, tert-butyloxycarbonyl protecting
group; Gly, glycine; EDCI, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride; TFA,
trifluoroacetic acid; LR-MRSA, linezolid- and methicillin-resistant Staphylococcus aureus AOUC-
0
915; GR-CREc, gentamicin- and colistin-resistant Escherichia coli 288328; HPLC, high performance
liquid chromatography; RP-HPLC, reverse-phase HPLC; MIC, minimum inhibitory concentration;
HC50, 50% hemolytic concentration; TI, therapeutic index; mcr, mobilized colistin resistance; aac(3)-
IIa, aminoglycoside N(3)-acetyltransferase; rRNA, ribosomal ribonucleic acid; LPS,
2

lipopolysaccharide; PBP, penicillin-binding protein; PBS, phosphate buffered saline; PI, propidium
iodide; NPN, N-phenyl-1-naphthylamine THF, tetrahydrofuran; iPrOH, iso-propanol.
1
. Introduction
The exponentially rising number of multidrug resistant bacteria, together with the slowing down of
the discovery of new antibacterial compounds, is a well-known world public health priority. The non-
judicial use or abuse of antibiotics, as well as their widespread use in livestock [1], are further favoring
the spread of antibiotic resistance [2] and the emergence of bacterial strains even resistant to last-resort
antibiotics [3].
In the search for new classes of antibiotics, the natural cationic antimicrobial peptides (AMPs), have
been extensively taken into consideration in the last decades [4-6]. Their amphiphilic active secondary
structure is essential to cause lysis of the bacterial cell membrane, even if occasionally they can also act
on intracellular targets [7-10] and modulate the innate immune response [11]. It is usually claimed that
bacteria can hardly develop resistance toward AMPs [8], but there are growing reports both on intrinsic
and acquired resistance, mainly due to membrane modifications [11].
Even if AMPs have a high net positive charge and are moderately selective in vitro toward bacterial
cells over mammalian cells, due to the different contents of anionic and uncharged/zwitterionic lipids
in membranes [12], their in vivo low efficacy and high toxicity has hampered the clinical use. Only
polymyxins have been systemically administered since the 1950s [13], but their use has been strongly
limited by the high incidence of nephrotoxicity and neurotoxicity [14], and the vast majority of
antimicrobial peptides that are currently in clinical development are for topical use only [11]. Other
important drawbacks, such as high cost of manufacture, instability toward proteases, and low
bioavailability, have further hindered the application of natural AMPs in clinical therapeutics [5,15].
In order to circumvent these problems, several groups have been working on synthetic mimics of
AMPs (Figure 1). Apart from some rationally designed short α-helices [16,17] and α-peptidic
dendrimers [18,19], most of the oligomeric structures with a number of residues between 5 and 20 units
are based on the diverse non-natural stable helices (foldamers) formed in solution by β-peptides [20-25]
(
Figure 1), oligoureas [26,27], and α-peptoids [28-31]. However, foldamers rely mainly on expensive
chiral monomers joined together by a multistep solid phase peptide synthesis followed by a demanding
purification, which are not suitable for industrial applications. Stable amphipathic conformations also
3

characterize shorter oligomers, such as de novo designed small arylamide foldamers [34-36] (Figure 1).
In all of these cases, the preorganization of monomers is at the basis of inherently stable secondary
structures [21,24,35-37], which do not have to spend free energy in order to arrange into an
amphiphilic folding. Examples of amphipathic antimicrobials based on an extremely rigid backbone
can also be found in synthetically-produced cationic steroid antimicrobials (CSAs) [38-41], as well as
in triaryl derivatives [42] where the facially amphiphilic topology is constitutionally guaranteed by the
proper 1,3,5-substitution of aryl scaffolds (Figure 1).
Figure 1. Representative synthetic mimics of antimicrobial peptides.
The opposite approach has also been used, thus mimicking AMPs like indolicidin, which is
unstructured in aqueous environment, but shows a stable polyproline II-like structure upon interaction
with liposomes and is a membrane permeabilizing peptide [43]. Synthetic oligomeric species not based
on preorganized monomers, thus having much more flexible backbones, may penetrate the bacterial
cell wall barrier in an easier way, compared to conformationally rigid antimicrobials, but they can still
easily assume amphiphilic conformations and then have a potent disrupting action on membranes [44].
4

Members of these flexible oligomeric antimicrobials are α-AApeptides [45], acyl-lysine oligomers
46], a few dicationic oligo-α-peptide analogues with a minimalist design [47], and various
[
polycationic ammonium amphiphiles [48-50] (Figure 1). Interesting examples of apparently
unstructured antibacterials with a very small size have been recently reported in literature, such as
dicationic α/β-dimers [51] and extremely simple, but still potent, lysine-based monomeric peptoids [52]
(
Figure 1).
Polycationic antimicrobial polymers have also been extensively investigated and reviewed on all
aspects [53-55]. Remarkable examples of polymers having a good bioactivity profile are the
isobutylene-maleimide [56] and the guanidinium-oxanorbornene [57] polymers (Figure 1).
Herein we report the synthesis and the in vitro antibacterial and hemolytic activities of small
amphiphilic achiral α-hydrazido acids, which were demonstrated to be membranolytic. In view of a
possible practical implementation, other important features (i.e. stability toward enzymatic and
chemical degradation) were also evaluated. In their design, the hydrophobic portion is composed of
both the N- and C-terminal lipophilic groups, whereas the single charge is furnished by the ammonium
ion of the glycine side chain (Figure 2). The α-hydrazido acids are mimics of β-amino acids and have
been chosen because oligomers of their conformationally constrained version are prone to form the
peculiar intramolecular hydrogen bond pattern of the hydrazido-turn secondary structure [58]. Thus,
this folding could also be assumed by the present non-preorganized acyclic α-hydrazido acids and used
to segregate polar and charged molecular portions from the lipophilic moieties. However, the lack of
conformational constrictions in their flexible backbone should facilitate the penetration through the
bacterial cell wall barrier and give rise to a more potent activity, as demonstrated for amphiphilic
compounds of different dimensions [46,59,60]. In this initial investigation we devoted our attention
only to mono-charged compounds, which are likely to be less potent as antimicrobial and less selective
toward mammalian cells, so that this choice should also ensure the possibility for easy future
improvements, by simply using a multiple charge-bearing side chain.
Figure 2. General structure of antibacterial α-hydrazido acids, with the common nomenclature for
nitrogen atoms in these compounds.
5

2
. Results and discussion
2
.1 Design of the amphiphilic lead structure
The idea at the basis of the choice of this particular backbone, that is its innate tendency to form
amphiphilic structures, was the subject of a high-level density functional theory (DFT) analysis. After a
1
2
complete conformational search at a lower theory level on a model compound with R = R = Me, all
the previously found energy minima were refined using the well-performing hybrid functional ωB97X-
D3(0) [61], together with the very large 6-311++g(3df,3pd) basis set, describing the solvent bulk with
the integral equation formalism version of polarizable continuum model method (IEF-PCM) [62]. The
highly predominating conformer of the model compound was then used for constructing selected
conformers of an actually used compound, namely 4Cb, all having the aliphatic chains in the most
stable all-antiperiplanar conformation (see Supplementary Data for all the other structures and
additional computational details). The theoretical approach indicated that the preferred conformation in
water for this monomeric compounds is the same hydrazido-turn assumed by their conformationally
constrained oligomeric versions [58] (Figure 3).
+
Figure 3. (A) Structure of most stable conformer of ammonium cation C H CO-HydrGlyH -NHC H
7
15
8
17
1
2
(
4Cb, R = n-C H , R = n-C H ), computed at ωB97X-D3(0)/6-311++g(3df,3pd)/IEF-PCM level in
7 15 8 17
water. (B) Front and (C) rear views of electrostatic potential surface.
6

The Natural Bond Orbital (NBO) analysis [63,64] pointed out that this structure is mainly stabilized
by the C=O‧ ‧ ‧ H-N hydrogen bond (2.08 Å), which is characterized by an important covalent
α
contribution and forms an eight-membered pseudocycle, and, to a lesser extent, by the N ‧ ‧ ‧ H-N H-
bond (2.38 Å), which is instead essentially electrostatic in nature (Tables S3 and S5). The study
exploiting the Atoms In Molecules (AIM) theory [65,66] confirmed the existence of a bond critical
point associated with a quite strong C=O‧ ‧ ‧ H-N hydrogen bond, whereas no critical point was found
α
in the case of N ‧ ‧ ‧ H-N interaction, thus describing this latter hydrogen bond as purely electrostatic
(
(
Table S6 and Figure S20). It is easy to see from the electrostatic potential surfaces of 4Cb as a cation
Figure 3) or considering the chloride counterion (Figure S24) that the only hydrogen atom belonging
to hydrocarbon chains that does not have an almost neutral potential is the one experiencing
hyperconjugation with the N-terminal C=O (light blue).
Apart from the obvious contribution of the charged glycine side chain, within this hydrophilic
molecular portion the C-terminal carbonyl and the N-terminal NH functionalities must play an
important role in forming hydrogen bonds with water and charged lipid heads while interacting with
phospholipid bilayers. The ability of hydrazide NHs to form quite strong hydrogen bonds is also readily
apparent considering that their experimental proton chemical shifts in diluted deuterochloroform
1
solutions range from 7.81 to 8.03 ppm when R = OBn, and reach very high values (8.74-9.87 ppm)
1
when R is a hydrocarbon chain, thus experimentally confirming their high positive charge (see
Supplementary Data). Calculations of atomic charges of ammonium cation with NBO, AIM, Merz-
Singh-Kollman [67,68], CHelpG [69], and Hu-Lu-Yang [70] schemes (Figures S19-S23), as well as the
computed electrostatic potential and charges taking also into account the chloride anion (Figures S24-
S25), plenty confirmed these findings about the segregation of polar/charged and apolar portions. It
must be noted here that only inherently very stable structures can display amphipathic conformations in
water, whereas the use of organic co-solvents or the presence of liposomes and vesicles greatly
facilitates the formation of the amphiphilic secondary structure even for compounds which are
substantially unstructured in water or aqueous buffers [37,71,72]. These observations strongly point
toward an amphiphilic arrangement of our α-hydrazido acid hydrochlorides during the action on
phospholipidic bilayers.
2
.2 Synthesis
7

Benzyl carbazate was reacted with methyl bromoacetate in the presence of N,N-
diisopropylethylamine, obtaining compound 1 in excellent yield (Scheme 1). Benzyl carbazate
demonstrated to be a poor nucleophile, thus concentrated dichloromethane solutions had to be used in
order to obtain reasonable reaction times. Moreover, DIPEA was necessary to avoid the competition in
the nucleophilic substitution from less hindered amines, such as triethylamine, which invariably led to
unsustainable decreases in yields, due to the substantial formation of the corresponding quaternary
ammonium salts with methyl bromoacetate.
Scheme 1. General Synthetic Scheme for the Preparation of Amphiphilic α-Hydrazido Acid
a
Derivatives
8

a
(a) N,N-diisopropylethylamine (DIPEA), anh. DCM (0.5 mL per mmol of benzyl carbazate), rt, 48
h, 96%; (b) Boc-Gly-OH, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDCI), anh.
2
DCM (0.5 mL per mmol of 2), -20 °C, 2 h, 86%; (c) R NH
2
(R = alkyl group), anh. MeOH, 7 h at rt for
H, Pd/C, anh. DCM, rt, 1 h,
3
Aa, 18, 24 or 36 h at reflux for 3Ab, 3Ac, and 3Ad, 88-94%; (d) HCO
then pyridine, R COCl, anh. DCM, rt, 1 h, 65-95%; (e) Trifluoroacetic acid (TFA)/DCM 1:3, rt, 20
min, chromatographic purification of free amine, then 3 M HCl in anh. MeOH, 68-98%; (f) Lithium
2
1
bis(trimethylsilyl)amide (LHMDS), anh. THF, Ar, 0 °C, 10 min, then methyl iodide, rt, 24 h, 5Ab:
1
8
7% 5Ac: 90%; (g) HCO H, Pd/C, anh. DCM, rt, 1 h, then pyridine, R COCl, anh. DCM, rt, 1 h for
2
5
Bb-c, 20 h for 5Gb-c, 81-92%; (h) Trifluoroacetic acid (TFA)/DCM 1:3, rt, 20 min, chromatographic
purification of free amine, then 3 M HCl in anh. MeOH, 75-98%.
Then compound 1 was reacted with Boc-Gly-OH, using EDCI as the coupling agent, obtaining the
desired common precursor 2 in very good yield, but even in this case it was necessary to circumvent
the scarce reactivity of 1 by using concentrated solutions and also optimizing all the other experimental
conditions. After reaction of 2 with the suitable alkyl amines, the first four compounds ready to be
1
1
deprotected, 3Aa-d (R = OBn), where successfully synthesized, and the change of R group by means
of the carboxybenzyl removal/free hydrazide acylation sequence furnished the remaining compounds
3
B-Ha-d in good to excellent overall yields. Hydrochlorides 4 were then obtained in pure form by Boc
removal with trifluoroacetic acid, followed by an easy chromatographic purification of free amines on
silica gel and re-salification with hydrochloric acid. It must be pointed out that the achievement of the
best possible antibacterial efficacy was out of the scope of this initial evaluation, so a coarse grid
approach was voluntarily chosen, and only alkyl chains differing one from another for a large number
1
of methylenic units were chosen. Moreover, when R groups different from linear alkyl chains were
1
2
taken into account, not all the possible combinations among R and R groups were synthesized and
1
tested. In fact, as correctly reported in Scheme 1, this was done when the N-terminal chain, R , was the
1
parent carboxybenzyl protecting groups, and for all the alkyl chains R = CH -C H . After this
3
17 35
2
screening, the best performing C-terminal chains (R = C
8
H17, C12
H25) were chosen to synthesize
1
selected compounds with R = t-Bu, Ph, and 1-naphthyl.
In order to verify the hypothesis that the well-defined amphiphilic character of the preferred
conformations should be determining in favoring the destabilizing action on phospholipid bilayers, the
N-methylated versions of six selected compounds were also synthesized (Scheme 1). Starting from 3Ab
1
2
1
2
(
R = OBn, R = n-C
8
H17) and 3Ac (R = OBn, R = n-C12H25), deprotonation with lithium
bis(trimethylsilyl)amide followed by reaction with iodomethane directly gave the corresponding N-
1
2
methylated compounds 5Ab and 5Ac in very good yields. Compounds 5Bb (R = CH , R = n-C H ),
3
8
17
1
2
1
2
1
2
5
Bc (R = CH , R = n-C H ), 5Gb (R = t-Bu, R = n-C H ), and 5Gc (R = t-Bu, R = n-C H )
3 12 25 8 17 12 25
9

were then obtained in good overall yields following the same methodology reported above for the
corresponding unmethylated compounds, but using a prolonged reaction time when the acylating
reagent was pivaloyl chloride. Exploiting the sequence acidic Boc removal/purification of free
amine/re-salification, hydrochlorides 6 were eventually obtained in pure form in yields ranging from
7
5% to 98%.
2
.3 Antibacterial and hemolytic activities
The antibacterial activity was first tested against three Gram-positive (Enterococcus faecalis ATCC
9212, Enterococcus faecium 135562 (35C), and Staphylococcus aureus ATCC 29213) and two Gram-
2
negative (Escherichia coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853) collection strains
Table 1). The minimum inhibitory concentrations (MICs) were evaluated using the broth
microdilution method in 96-well microtiter plates, following the CLSI guidelines [73], while hemolytic
(
1
2
activity was determined according to a reported procedure [28]. Compounds 4Ae (R = n-C H , R =
1
7
35
1
2
n-C H ) and 4Hc (R = 1-naphthyl, R = n-C H ) were too insoluble to be tested. In addition, two
1
8
37
12 25
among the most active compounds were subsequently tested against multidrug-resistant bacteria,
namely the linezolid- and methicillin-resistant Staphylococcus aureus AOUC-0915 (LR-MRSA) and
the gentamicin- and colistin-resistant Escherichia coli 288328 (GR-CREc). MICs and hemolysis results
were reproducible between three independent experimental replicates. These results, as well as the
reverse-phase HPLC-derived overall lipophilicity, are reported in Table 1.
From the results in Table 1, many general deductions can be drawn. First, the hydrochlorides of α-
hydrazido acids are somewhat more active against Gram-positive bacteria, E. faecium being the most
sensitive, with respect to Gram-negatives, especially Pseudomonas aeruginosa. This observation
makes readily apparent that the thick peptidoglycan layer of Gram-positive bacteria cannot prevent the
1
entry of these flexible molecules, even when a large and rigid 1-naphthyl group is present as R (4Hb),
because of the nano-sized pores in the membrane of Gram-positive bacteria [74]. Moreover, the lower
sensitivity of Gram-negatives is common in literature for cationic amphipathic membranolytic
compounds, and is usually claimed as mainly caused by the reduced active concentration that can reach
and disrupt both the outer and inner membranes of Gram-negative bacteria. This in turn is due to
inability to permeabilize their primary barrier for hydrophobic compounds, which is the electrostatic
network of negatively charged lipopolysaccharide molecules bound to divalent cations in the outer
leaflet [75]. However, the lower activity toward Gram-negative bacteria could also be due to their high
1
0

content of zwitterionic phosphatidylethanolamine, whereas Gram-positives are usually richer in
negatively charged components [39], or to the substantial amounts of negative curvature-inducing
lipids into their phospholipid bilayers [76].
Table 1. In Vitro Antibacterial and Hemolytic Activities of α-Hydrazido Acids
-
1 a
MIC vs drug sensitive bacteria (µg mL )
b
Retention
timec
1
2
E.
E.
S.
aureus
E.
coli
P.
aeruginosa (µg mL ) S. aureus E. coli
HC50
TI vs TI vs
Compd
R
R
-1
faecalis faecium
(
min)
4
Aa
Ab
OBn
OBn
OBn
OBn
C
3
H
7
1024
128
16
512
64
512
64
512
64
512
128
128
512
2048
256
64
775±19
189±7
1.5
3.0
11
1.5
3.0
2.7
0.5
1.3
0.9
2.0
0.5
0.4
8.2
1.4
<1.0
2.6
3.8
<3.4
<0.7
0.5
0.4
0.2
N.D.
2.9
49
3.4
4
C
8
H
17
14.3
21.2
30.9
0.3
4
Ac
Ad
Ba
Bb
Bc
C
12
18
H
25
37
7
8
8
32
86±5
4
C
H
512
2048
256
32
256
2048
128
32
512
1024
128
64
512
2048
256
64
243±13
2700±43
223±9
0.5
2.6
1.7
2.0
0.9
0.4
8.2
5.7
<1.0
2.6
7.6
<3.4
1.4
3.7
1.6
0.5
N.D.
2.9
49
4
CH
CH
CH
CH
3
3
C H
4
3
C
8
H
17
9.3
4
3
C
12
18
H
25
37
7
127±8
17.5
28.0
5.2
4Bd
4Ca
4Cb
3
C
H
H
32
32
256
1024
16
512
1024
16
512
2048
32
228±18
433±10
131±8
C
C
C
C
H
15
H
15
H
15
H
15
C
1024
32
1024
16
7
7
7
7
3
C
8
H
17
17.8
24.0
33.1
16.6
23.5
29.1
35.9
26.5
32.2
35.8
41.8
12.3
19.7
11.7
19.8
14.4
21.6
4
Cc
C
12
18
H
25
37
7
32
16
64
256
>256
64
512
>256
64
367±24
267±18
167±16
486±21
438±12
364±11
234±4
4Cd
4Da
4Db
C
H
H
>256
64
>256
64
>256
64
C
C
C
C
C
C
C
C
11
11
11
11
17
17
17
17
H
23
H
23
H
23
H
23
H
35
H
35
H
35
H
35
C
3
C
8
H
17
16
8
64
128
>128
>512
512
512
512
N.D.
128
4
128
>128
512
512
512
512
N.D.
128
64
4
Dc
Dd
Ea
Eb
Ec
Ed
Fb
Fc
Gb
Gc
Hb
Hc
C
12
18
H
25
37
7
128
512
32
128
512
16
>128
256
64
4
C
H
H
4
C
3
4
C
8
H
17
512
512
64
128
256
N.D.
128
4
200±31
124±2
4
C
12
18
H
25
37
256
N.D.
128
4
d
4
C
H
N.D.
128
4
N.D.
4
Ph
Ph
C
8
H
17
368±6
4
C
12
H
25
17
195±13
775±22
394±28
311±31
N.D.
4
t-Bu
t-Bu
C
8
H
256
8
256
8
512
4
256
8
512
16
1.5
99
3.0
49
4
C
12
H
25
4
Naph
Naph
C
8
H
17
32
16
32
16
64
9.7
N.D.
19
4
C
12
H
25
N.D.
N.D.
N.D.
N.D.
N.D.
N.D.
1
1

e
e
e
f
MSI-78
---
64
8-16
0.12-1
0.12-1 0.25-1
0.5-2 ---
16-32
---
8-32
0.5-2
---
120
---
7.5-15 3.8-7.5
---
---
---
---
g
Tetracycline
8-32
4-16
>16
---
---
---
0.5-2
---
---
---
---
---
---
g
Gentamicin
---
g
Methicillin
---
-
1
MIC vs multidrug-resistant bacteria (µg mL )
TI vs TI vs Retention
LR- GR- time
MRSA CREc (min)
h
i
HC50
µg mL )
LR-MRSA
GR-CREc
-1
(
4
Fc
Ph
t-Bu
C
12
12
H
25
25
4
4
4
8
195±13
394±28
---
49
99
---
---
49
49
---
---
19.7
19.8
---
4
Gc
C
H
j
Gentamicin
---
128
j
Methicillin
>1024
---
---
---
a
b
Conservative estimates of at least three independent trials. HC50 is defined as the concentration of
compound that kills 50% red blood cells. Standard errors obtained by nonlinear regressions of average
c
data (three independent trials for each compound) are reported. Adjusted retention time, computed as
d
e
the instrumental retention time minus the hold-up time. N.D. stands for “not determined”. Values
f
g
h
taken from Ref. 78. Value taken from Ref. 34. Values taken from Ref. 73. Linezolid- and
methicillin-resistant S. aureus AOUC-0915. Gentamicin- and colistin-resistant E. coli 288328.
i
j
Values determined in this work.
As expected from the wide dimensional range of substituents used, MICs vary to a very large extent.
1
2
As an example, while compound 4Ba (R = CH , R = C H ) having the shortest chains is almost
3
3
7
-1
-1
completely inactive (e.g. MICs = 1024 µg mL vs S. aureus and 2048 µg mL vs E. coli), likely due to
1
2
a poor interaction with the apolar portions of phospholipid bilayer, compound 4Cb (R = n-C H , R =
7
15
-1
n-C
8
H17) shows a good activity (MICs = 16 µg mL vs S. aureus and E. coli), and the MIC increases
1
2
-1
again for too long chains, as for 4Cd (R = n-C H , R = n-C H , MICs >256 µg mL vs S. aureus
7
15
18 37
1
2
-1
-1
-1
and E. coli), 4Dd (R = n-C H , R = n-C H , MICs = 256 µg mL vs S. aureus and >512 µg mL
1
1
23
18 37
1
2
-1
vs E. coli), and 4Ec (R = n-C17H35, R = n-C12H25, MICs = 256 µg mL vs S. aureus and 512 µg mL
vs E. coli). Even in this case, the necessity to avoid either too small or too large apolar fragments
parallels the findings for amphiphilic antibacterials with different dimensions, such as polymers [77],
1
2
-1
foldamers [23], and small peptoids [52]. The best MICs (4Fc, R = Ph, R = n-C H , 4 µg mL vs S.
12
25
1
2
-1
-1
aureus and E. coli, and 4Gc, R = t-Bu, R = n-C H , 4 µg mL vs S. aureus and 8 µg mL vs E.
12
25
coli), are equal to or better than the values for the magainin derivative MSI-78 (pexiganan, MIC = 8-16
-
1
-1
µg mL vs S. aureus and 16-32 µg mL vs E. coli) [78], which is also more hemolytic (HC = 120 µg
5
0
-1
mL ) [34]. Despite the lack of any thorough optimization, these extremely simple and mono-charged
1
2

compounds are competitive toward much more complex and expensive 12- and 14-helix foldamers
-1
(
best MICs = 3.1-3.2 µg mL vs S. aureus) [23,24], and either isobutylene-maleimide or guanidinium-
-1
oxanorbornene polymers (best MICs = 3-12 µg mL vs S. aureus) (Figure 1) [56,57]. Taking into
consideration small molecules, α-hydrazido acids outperform tricationic tripeptides based on 2,5,7-tri-
tert-butyl tryptophan in terms of both antimicrobial activity and therapeutic window [32], whereas only
-1
the most effective among tetracationic triaryl derivatives (best MICs = 0.78 µg mL vs S. aureus, 3.13
-1
µg mL vs E. coli; best therapeutic indices = 406 vs S. aureus, 172 vs E. coli) give better overall results
[
42] (Figure 1). Even some among the structurally very simple lysine-based monomeric peptoids show
-1
-1
a slightly more powerful antimicrobial efficacy (best MICs = 2.2 µg mL vs S. aureus, 2.9 µg mL vs
E. coli), but despite the double positive charge they are less selective toward erythrocytes and have
worst therapeutic indices [52] (best TIs = 34 vs S. aureus, 23 vs E. coli) (Figure 1).
It is worth noting that these α-hydrazido acids show identical activities towards two multidrug
resistant strains, the linezolid- and methicillin-resistant S. aureus AOUC-0915 (LR-MRSA, MICs = 4
-
1
µg mL for both 4Fc and 4Gc) and the gentamicin- and colistin-resistant E. coli 288328 (GR-CREc,
-1
-1
MICs = 4 µg mL for 4Fc and 8 µg mL for 4Gc, Table 1). This is not surprising for LR-MRSA,
whose resistances to methicillin and linezolid rely, respectively, on the presence of PBP2a, a penicillin
binding protein with extremely low affinity for all β-lactams, except for last-generation cephalosporins
[
79], and an rRNA methyltransferase that catalyzes post-transcriptional methylation to the C8 position
of nucleotide A2503 in 23S rRNA, causing a decreased binding affinity for linezolid [80]. Thus, the
phospholipidic bilayers of LR-MRSA membrane are not substantially changed in comparison to the
drug sensitive counterpart, S. aureus ATCC 29213. On the other hand, the conservation of efficacy of
these cationic α-hydrazido acids toward GR-CREc was not predictable. In fact, whereas the resistance
to gentamicin of GR-CREc is due to the aac(3)-IIa gene, which causes the covalent modification of
gentamicin and lead to a poor binding to the ribosome target [81], the resistance to colistin is due to the
MCR-1 protein, which lead to addition of phosphoethanolamine to lipid A. Consequently, the binding
between the less negatively charged lipopolysaccharide (LPS) and the positively charged colistin is
much less effective [82], and the same reduced binding affinity was demonstrated to be at the basis of
an intrinsic resistance to cationic AMPs [83]. However, in the present case the possible decrease in
electrostatic interaction between the cationic α-hydrazido acids and the outer membrane external
leaflet, that is the first mandatory step for the following insertion and damaging action on phospholipid
bilayers, does not appear to be detrimental for the antimicrobial activity. This observation suggests that
1
3

these new amphiphilic organic compounds might also be active against other MDR strains with the
same compositions of lipid bilayers.
As far as hemolytic activity is concerned, it is evident from data in Table 1 that the smallest and less
1
2
-1
active compounds show larger HC values (e.g. 4Ba, R = CH , R = C H , has HC = 2700 µg mL ),
5
0
3
3
7
50
1
2
while the most active species with medium-sized R and R chains are also more active against red
1
2
blood cells. However, the two most powerful antibacterial compounds (4Fc, R = Ph, R = n-C H ,
12
25
-1
1
2
-1
HC50 = 195 µg mL , and 4Gc, R = tBu, R = n-C H , HC = 394 µg mL ) have hemolytic
12
25
50
concentrations that are much higher than their MICs, thus leading to substantially better therapeutic
indices than all the other compounds. In addition, the percent hemolysis computed at a concentration
equal to the MIC for either S. aureus or E. coli was also evaluated using the parameters obtained from
the non-linear regression of experimental hemolysis data by the Hill equation [84,85] (Table S2). From
these data, it is easy to appreciate that also some less selective compounds with poorer therapeutic
1
indices actually show extremely low percent hemolysis at MICs. As an example, compound 4Ab (R =
2
OBn, R = C H ) has low TIs toward both S. aureus and E. coli (3.0, Table 1), but the computed
8
17
-1
-3
percent hemolysis at MICs (64 µg mL ) are almost null (9 10 %, Table S2). Interestingly, contrarily to
antibacterial efficacy, the hemolytic activity does not decrease when large N- and C-terminal
1
2
-1
substituents are present (e.g. 4Ec, R = C H , R = C H , HC = 124 µg mL ), and has a more
17
35
12 25
50
erratic general behavior. Both of these characteristics, taken together with the wide dimensional range
of substituents considered, are directly involved in the very large interval of therapeutic indices
obtained for all bacteria (e.g. from 0.5 to 99 for S. aureus).
It is easy to find the first structure-activity relationship considering only compounds 4A-Ea-d, for
1
2
which all the possible twenty combinations of R (OBn, methyl, heptyl, undecyl, or heptadecyl) and R
1
(
propyl, octyl, dodecyl, or octadecyl) were synthesized. For the shortest R tail (Me), there is a striking
2
improvement in antibacterial efficacy toward E. faecium passing from R = propyl (MIC = 2048 µg
-1
-1
-1
mL ) to the octyl (MIC = 128 µg mL ), and then the dodecyl (MIC = 32 µg mL ) and octadecyl (MIC
-1
=
32 µg mL ) chains (Figure 4A, see Figures S1-S5 for the other bacteria). On the contrary, for the
1
2
-1
longest R (n-C H ), the best-performing R fragment is the propyl (MIC = 16 µg mL ), whereas a
1
7
35
2
1
continuous increase in MICs is observed as the length of R increases. For all the other R , there is a
2
1
clear decrease of antibacterial potency at both sides of the R that best matches the particular R . The
1
same findings can be deduced considering how MICs vary according to different R moieties, taking
2
constant R group (Figure 4B). Analyzing the complete series of graphs (Figures S1-S5), it is also
worth noting that enterococci are more susceptible to compounds having N- and C-terminal chains
1
4

longer than those required to inhibit growth of S. aureus, which in turn is more sensitive to chains
1
longer than those necessary against Gram-negative bacteria. For example, when R = n-C H , the best
11
23
2
activities toward E. faecalis and E. faecium are obtained for R = n-C
8
H17, while in the case of S.
2
aureus R = n-C H and n-C H show the same potency, and for E. coli and P. aeruginosa the shortest
3
7
8
17
2
R furnished the lowest MICs (Figures S1-S5).
1
2
Figure 4. Variation of antimicrobial activity toward E. faecium with R and R substituents, for
1
compounds 4A-Ea-d. (A) bar graphs colored taking constant R groups, highlighting the dependence of
2
2
antimicrobial activity on R length. (B) bar graphs colored taking constant R groups, highlighting the
1
1
dependence of antimicrobial activity on alkyl R length (compounds 4Aa-d, with R = OBn , are not
reported).
Figure 4 was useful for comparing homogenous series of compounds with the same type of
1
substituents, but after the initial investigation we introduced ad hoc variations in R groups, taking
2
constant the two most active C-terminal R groups, the octyl and dodecyl chains (Table 1). A possible
way to develop a more general model for antibacterial activity, and also for hemolytic activity, is to use
the reverse-phase HPLC-based (RP-HPLC) overall lipophilicity, in the form of either the retention time
[
32,42,52,86] or the percentage of less polar eluent at retention time [29,30]. This kind of structure-
activity relationship highlights that, for all the bacteria tested, there is a parabolic trend of MICs toward
retention time (Figures 5A and 5B for S. aureus, see Figure S6 for other bacteria).
1
5

Figure 5. Variation of (A,B) antibacterial activity toward S. aureus (two different vertical scales), (C)
hemolytic activity, and (D) therapeutic index toward S. aureus with adjusted retention time
(
instrumental retention time minus the hold-up time; Hewlett-Packard Lichrosorb RP 18 column, 5 µm,
-1
L × I.D. 200 × 4.6 mm, flow rate = 0.5 mL min ; elution gradient: from water:2-propanol = 70:30 to
water:2-propanol = 10:90 in 40 minutes).
Of course, due to the fact that MICs determined with the adopted protocol were measured in powers
of 2 [73], and that in some cases only minimum theoretical values corresponding to the highest
2
concentration tested were available, coefficient of determination, r , indicating a very good parabolic
2
interpolation could not be expected. However, r of 0.88, 0.80, 0.68, 0.76, and 0.63 for E. faecalis, E.
1
6

faecium, S. aureus, E. coli, and P. aeruginosa, respectively, still confirm the visual impression that
MIC varies with retention time with an approximately parabolic trend and, most important, that overall
lipophilicity is by far the main parameter governing the antimicrobial activity for these compounds.
Albeit this finding is of obvious practical importance for directing the future synthesis of most active
compounds in a more effective way, a simple interpretation of the observed raise of MICs after the
optimal retention time ranges cannot be done. Of course, it is striking that the reason must relies on the
differences between the ability of these compounds to simply bind to covalently linked monolayers of
linear C18 hydrocarbon chains in a RP-HPLC column, and their capability to be attracted to the
external bacterial surface, reach the fluid double layers of differently composed membranes, insert and
then weaken or damage them by one or more of the known mechanisms. In fact, while the former
clearly appears to be a monotonically raising additive function of van der Waals interactions, the latter
is a much more complex phenomenon that encompasses many different molecular movements and
interactions. Depending on the particular mechanism, different dispositions of amphiphilic compounds
are required for either a regular pores formation, a generalized membrane weakening/permeabilization,
or a less ordered membrane dissolution/destruction with a detergent-like effect [7-10]. Thus, it is not
obvious that the ability of these compounds to kill bacteria must always increase as the interactions
with the apolar stationary phase in a RP-HPLC strengthen. In addition, even if overall lipophilicity is
evidently the main parameter for quantitatively describing the growth-inhibiting effect of these
amphiphilic compounds, there must also be secondary variables, such as substituent-specific effects
relying on shape and/or electronic features, which can hardly be described at the moment. In fact, for
all the bacteria there are some cases in which compounds with very close retention times have different
antibacterial activities, as well as cases of compounds with fairly different retention times that show the
same MICs (Figure 5 and Figure S6).
The higher sensitivity of Gram-negative bacteria to shorter alkyl chains than those required for
Gram-positives, previously highlighted analyzing complete homogeneous series in Figure 4 and
Figures S1-S5, can be deduced here in a more comprehensive way taking into consideration the ranges
of retention times in which compounds show good antimicrobial efficacy. In fact, retention time ranges
-1
where MICs drop below a given value (e.g.
and globally shift toward lesser values passing from E. faecium (14.3-32.2 minutes), to E. faecalis
14.4-28.0 minutes), to S. aureus (14.3-26.5 minutes), to E. coli (14.3-21.2 minutes), and eventually to
≤64 µg mL ) in Figure 5 and Figure S6, become narrower
(
P. aeruginosa (14.4-19.7 minutes). To the best of our knowledge, in all cases available in literature of
RP-HPLC-derived overall lipophilicity measured for cationic antibacterial amphiphiles, such a clear
1
7

parabolic relationship has never been found, and only a few cases of increases in MICs for large
lipophilic chains have been reported so far. However, this lack of previously reported similar behaviors
could be due to the usually much smaller dimensional span of hydrophobic moieties investigated, thus
generally leading to a monotonic increase in antibacterial activity as the overall lipophilicity increases
[
30,32,42,86]. Only in one case for lysine-based monomeric peptoids [52] (Figure 1), and in two cases
for N-terminal alkylated oligopeptoids [29], a sharp decrease of antimicrobial efficacy with extended
retention times was evidenced for structurally homogenous backbones bearing too large lipophilic
portions.
In contrast to the approximately parabolic trend of minimum inhibitory concentrations, HC values
5
0
show a sharp decrease as soon as the overall lipophilicity starts increasing and then, for the remaining
range of retention times, they have a quite erratic behavior but without any overall reduction or
increase (Figure 5C). This difference with respect MICs is also readily apparent directly analyzing only
1
2
data for compounds belonging to the complete series with all the combinations of R and R used in the
1
initial investigation (Table 1 and Figure S8). As an example, within the series of compounds 4Ca-d (R
1
1
=
n-C H ), 4Da-d (R = n-C H ), and 4Ea-c (R = n-C H ), the variation of hemolytic
7 15 11 23 17 35
2
concentration with R groups is strikingly different from the homogenous trends of MICs toward any of
the bacteria tested, as reported for E. faecium in Figure 4 (see also Figures S1-S5). The same applies to
1
2
the dependence of HC50 on R groups, taking into consideration compounds having the same R chains
Figure S8), thus confirming a somewhat more unpredictable behavior of hemolytic activity. In
(
addition, there are no other evident variables, such as shape or electronic features, identifying
subgroups of clearly more (or less) hemolytic compounds.
Due to the peculiar and extremely different trends of MIC and HC with respect to retention time,
5
0
the resulting pointed graphs of computed therapeutic indices (TI = HC50/MIC) as a function of overall
lipophilicity are not surprising (Figure 5A for S. aureus, Figure S7 for other bacteria), and exactly the
same behavior has already been found for analogs of antimicrobial peptide gramicidin S [87]. For all
bacteria, compounds having too short or too long retention times invariably show extremely low TIs,
thus being completely unsafe. Only within the range of optimal overall lipophilicity, corresponding to
the range of moderate to good antimicrobial activity (approximately from 14 to 27 minutes toward S.
aureus), α-hydrazido acids can show a better in vitro selectivity, confirming the usefulness of this
quantitative evaluation of overall lipophilicity by means of adjusted retention times. Obviously, due to
the definition of TI itself, the wider or narrower ranges of best therapeutic indices toward other bacteria
must parallel the wider intervals of overall lipophilicity furnishing good antimicrobial activity toward
1
8

enterococci, as well as the narrower intervals toward Gram-negative strains. In addition, based on the
general order of sensitivity of bacteria to these compounds reported above (Table 1), TI values for
many amphiphilic
P. aeruginosa. However, mainly due to the fluctuating behavior of hemolytic activity, even within
the optimal range for each bacterium there are some compounds showing a fairly reduced therapeutic
α-hydrazido acids decrease in the order E. faecium > E. faecalis > S. aureus > E. coli
>
1
2
window (Figures 5A and S7, Tables 1 and S1). The two most powerful antibacterials, 4Fc (R = Ph, R
1
2
=
n-C H , TIs = 49 for both S. aureus and E. coli) and 4Fc (R = t-Bu, R = n-C H , TIs = 99 for S.
12 25 12 25
aureus and 49 for E. coli), are by far the most selective compounds and have almost identical retention
times, thus they appear as a sort of spike in most of the graphs. They also show the same good
selectivity when MDR bacteria are taken into consideration (Table 1), and 4Fc is the only one having a
quite good therapeutic index for Pseudomonas aeruginosa (25), toward which all the other compounds
show much poorer TIs (<5). The only remarkable exception to their superiority is
α
-hydrazido acid
17) toward E. faecium, whose highest therapeutic index, 61 (Table S1),
is caused by both its high activity toward that bacterium and its inherently low hemolytic potency.
1
2
4
8
Db (R = n-C11H23, R = n-C H
2
.4 Permeabilization of outer and inner membranes
To ascertain beyond doubt the permeabilizing and destabilizing action on bacterial membranes of α-
hydrazido acid amphiphiles, we slightly modified previously reported procedures [49] and used
1
2
compound 4Gc (R = t-Bu, R = n-C H ) at its MIC as a model system toward both the susceptible
12
25
(
ATCC 25922) and the MDR (GR-CREc, 288328) E. coli strains (Figure 6). N-phenyl-1-
naphthylamine (NPN) was chosen as a fluorescent probe for the outer membrane permeabilization, due
to its gain in fluorescence when passing from an aqueous solution to a hydrophobic environment (i.e.
the membrane lipid bilayers), whereas propidium iodide (PI) was chosen as an indicator of inner
membrane permeabilization, because its fluorescence is greatly enhanced when it binds to nucleic
acids. The fluorescent probes showed a null (NPN) or extremely reduced (PI) uptake in absence of
4
Gc, while rapid increases in normalized fluorescence were observed in both cases by addition of the
amphiphile to suspensions containing either the sensitive (Figure 6 A) or the MDR strain (Figure 6 B),
thus plenty confirming the permeabilizing and damaging action of
bacterial membranes.
α-hydrazido acid hydrochlorides on
Even if the raise in normalized fluorescence for the susceptible ATCC collection strain seems to be
extremely reduced for both probes in comparison to the gentamicin- and colistin-resistant GR-CREc
1
9

strain, this is mainly due to substantially different starting points in terms of absolute fluorescence
intensity (see Figures S12-S15). The computed permeabilization rate constants must therefore be
considered in order to effectively compare the behavior of compound 4Gc toward the two strains. At
least with these experimental conditions and for the time considered (10 minutes), the permeabilization
2
of both membranes always demonstrated to follow an exponential raise to maximum (r ≥0.99), and the
actual rate constants computed for outer membrane (Figures S12-S13) and inner membrane (Figures
S14-S15) permeabilization are quite close for the two strains.
Figure 6. Variation of (A) NPN and (B) PI fluorescence with time, as a measure of outer and inner
membrane permeabilization, respectively, caused by compound 4Gc [red: E. coli ATCC 25922; black:
GR-CREc (E. coli 288328)].
Interestingly, the outer membrane permeabilization rate constant measured by NPN uptake for GR-
-1
-1
CREc (0.31 min ) resulted to be about 1.2 times greater than for the sensitive bacterium (0.25 min ),
whereas for the inner membrane permeabilization the rate constant of PI uptake for GR-CREc (0.24
-1
-1
min ) was about 1.1 times greater than for E. coli ATCC 25922 (0.22 min ). Thus, not only the less
negatively charged LPS of GR-CREc was unable to prevent the action of these amphiphiles more
effectively than that of susceptible E. coli, but it resulted even more permeable to α-hydrazido acid
hydrochlorides. This is in agreement with the identical MIC values of compounds 4Fc and 4Gc
obtained against either susceptible or resistant S. aureus and E. coli strains (Table 1).
2
0

2
.5 Stability toward enzymatic and chemical degradation
One of the main issues of AMPs, that is their sensitivity to proteolytic degradation [88], could not be
excluded a priori for α-hydrazido acids, due to the presence of three different potentially hydrolysable
α
carbonyl-nitrogen bonds, one of which linking the N atom to a glycine side chain. Thus, the minimum
1
2
1
inhibitory concentrations toward S. aureus of compounds 4Cb (R = n-C
7
H
15, R = n-C
8
H17), 4Fc (R =
2
1
2
Ph, R = n-C H ), and 4Gc (R = t-Bu, R = n-C H ), were evaluated after preincubation in fresh
12
25
12 25
5
0% blood plasma solution for 0, 3 and 6 hours at 37 °C [52].
Albeit MICs of both compounds 4Fc and 4Gc were subjected to a two-fold increase in presence of
blood plasma, whereas for 4Cb no variation was observed, it is very remarkable that all antimicrobial
activities remained constant even after 6 hours preincubation (Figure 7), therefore demonstrating the
resistance to proteolytic degradation for physiologically relevant time intervals.
Figure 7. (A) MICs toward S. aureus of compounds 4Cb, 4Fc, and 4Gc, after preincubation in 50%
blood plasma solution for 0, 3 and 6 h at 37 °C. Conservative estimates of three trials for each
compound. (B) Percent areas of samples of compound 4Cb, stored in different conditions at room
temperature, unless otherwise indicated (Hewlett-Packard Lichrosorb RP 18 column, 5 µm, L × I.D.
2
00 × 4.6 mm, flow rate = 0.5 mL/min; eluent water:2-propanol = 50:50).
The stability toward chemical degradation in different conditions was also important in view of a
potential practical implementation, and some decomposition had already been noted for compounds
stored for prolonged periods (>3 months) at -18 °C as free amines. Without re-salification after the
chromatographic purification, the main side reaction in diluted conditions was demonstrated to be the
transacylation, with transfer of N-terminal acyl group to the glycine amine functionality by means of an
2
1

intramolecular nucleophilic acyl substitution (see synthesis of compound S1 in Supplementary Data).
On the other hand, solid free amines intentionally stored at room temperature for more than one month
also underwent intermolecular attacks and gave complex mixtures of products. Conversely, solid
hydrochlorides demonstrated to be completely stable at -18 °C for at least 6 months. To better evaluate
1
2
resistance to chemical degradation, samples of compound 4Cb (R = n-C H , R = n-C H ) as
7
15
8
17
-1
hydrochloride (1 µg mL ) in PBS (pH = 7.4), pure water, and an aqueous solution mimicking the
lowest value for the stomach pH range (1.5), were stored at room temperature for 8 weeks, together
with a methanolic solution of the free amine (Figure 7). Starting with a freshly prepared compound
having a 99.7% purity, based on HPLC areas, the hydrochloride in water did not show any
decomposition up to the maximum time, whereas only marginal changes were experienced by the
samples in PBS and acidic solution (97.1% and 97.2% final purities, respectively). As expected, the
methanolic solution of free amine was subjected to a pronounced degradation, with a loss of purity of
more than 6% within the first week and about 40% after 8 weeks.
2
.6 Importance of hydrazide NH and non-disrupted amphiphilicity
As initially reported, a series of six N-methylated compounds, whose parent unmethylated
hydrochlorides showed a wide range of antimicrobial activities, was synthesized with the aim of
verifying if the effective facial segregation of hydrophilic and lipophilic faces and the presence of
highly positive hydrazide NH hydrogens were decisive in favoring a good antimicrobial activity.
Comparing the antibacterial activities of NMe hydrochlorides in Table 2 with the MICs of the
corresponding unmethylated compounds in Table 1, it is easy to see that in only one case the N-
-1
methylated version is the most active (hydrochlorides 4Ac, MIC = 16 µg mL , and 6Ac, MIC = 8 µg
-1
mL , against E. faecalis), and in few other cases NMe derivatives have the same potency than their
parent NH compounds.
Table 2. In Vitro Antibacterial Activities of N-Methylated α-Hydrazido Acids
-
1 a
MIC vs drug sensitive bacteria (
µg mL )
1
2
Retentionb
Compd
Ab
Ac
R
R
E. faecalis E. faecium S. aureus E. coli P. aeruginosa
time (min)
6
OBn
OBn
C
8
H
17
25
128
8
128
8
128
8
128
64
256
512
13.8
20.7
6
C
12
H
2
2

6
Bb
Bc
Gb
Gc
CH
CH
3
3
C
8
H
17
1024
32
1024
32
2048
64
2048
128
256
64
2048
256
512
512
9.1
6
C
12
H
25
17.1
11.7
18.9
6
t-Bu
t-Bu
C
H
8
17
512
16
512
16
512
16
6
C H
12 25
a
b
Conservative estimates of at least three independent trials. Adjusted retention time, computed as
the instrumental retention time minus the hold-up time.
However, in most cases there is a great increase (from two-fold to sixteen-fold) in minimum
inhibitory concentrations, and the overall decrease in activity toward Gram-negative bacteria is more
pronounced than that against Gram-positives. For all NMe compounds there is also a reduction up to
0
.9 minutes in retention times, with respect to their NH counterparts. This decrease in the RP-HPLC
measured overall hydrophobicity could be due to the contribution of a low energy conformation with
an unconventional C=O‧ ‧ ‧ H-C hydrogen bond between the C-terminal hydrazide carbonyl and one
of the slightly positive N-methyl hydrogens, as computationally demonstrated for the model N-
+
methylated ammonium cation mAc-Hydr(Me)GlyH -NHMe (see Supplementary Data). This
1
2
conformation has a worst directionality of the hydrophobic R and R groups, compared to the
hydrazido-turn arrangement, thus probably leading to a poorer interaction with both the C18
hydrocarbon chains of the stationary phase in the HPLC column, and the lipid portions of membrane
bilayers.
Despite the observed reductions in overall lipophilicities, it must be pointed out that retention times
for the most active compounds in Table 2 (6Ac, 6Bc, and 6Gc) are well within the optimal ranges
determined for unmethylated hydrochlorides 4. Thus, overall hydrophobicity still remains the main
parameter for describing the antimicrobial activity even in the case of N-methylated hydrochlorides, but
other concepts must be invoked in order to explain the deleterious effect of such an apparently small
structural change. To this end, the two most stable unmethylated and N-methylated hydrazido-turn
1
2
1
2
conformers of either the simplified model compound (R = R = Me) or 4Cb (R = n-C H , R = n-
7
15
8
C H17), and their corresponding N-methylated versions, were analyzed comparing the electrostatic
potential surfaces and the atomic charges (see Supplementary data). All the charge calculations
schemes confirmed that, despite the electronwithdrawing hydrazide nitrogen, the three hydrogens of the
NMe group in the methylated species only have very small positive charges, therefore being essentially
hydrophobic in nature, which was experimentally demonstrated for Boc-protected compounds 5 in
deuterochloroform (see proton NMR spectra in Figures S59-S64). On the contrary, for the hydrazide
NH of unmethylated compounds high positive charges were invariably obtained. Considering also the
2
3

chloride counterion for both 4Cb and its NMe version allowed to better highlight the partial disruption
of amphiphilicity in N-methylated compounds, which arises from the substantially lipophilic character
α
of the methyl group on the N atom (Figure 8, see also Figures S29-S30). Even if it is the smallest alkyl
group, the methyl on hydrazide nitrogen behaves like an “hydrophobic bulge”, which very likely
negatively affects the interaction with phospholipid heads and water molecules during the destabilizing
and damaging action on membrane bilayers.
Figure 8. Rear views of electrostatic potential surfaces for the most stable structures of hydrochlorides
1
2
(
A) C H CO-HydrGly-NHC H ‧ HCl (4Cb, R = n-C H , R = n-C H ), and (B) its N-methylated
7 15 8 17 7 15 8 17
version C H CO-Hydr(Me)Gly-NHC H ‧ HCl, computed at ωB97X-D3(0)/6-311++g(3df,3pd)/IEF-
7
15
8
17
PCM level in water.
These findings about the decrease of activity induced by a partly disrupted amphiphilic topology are
similar and complementary to the results obtained for triaryl derivatives whit an additional pendant
lipophilic group R [42] (Figure 1). In both cases, amphiphilicity seems to have a greater importance for
the activity against Gram-negative bacteria. In α-hydrazido acids amphiphilicity is ruined by addition
of a small lipophilic group in the hydrophilic face, with the concomitant loss of a hydrogen bond-
donating functionality, whereas in the case of tetracationic triaryl derivatives the negative effect is
2
4

obtained by insertion of an amide linker in an otherwise completely hydrophobic region. However, for
those triaryl derivatives with facially disrupted amphiphilicity, the almost complete loss of activity
against Gram-negative bacteria was ascribed to a much less efficient insertion of the negative
curvature-inducing hydrophobic portions, caused by the high content in phosphatidylethanolamine (PE)
of Gram-negatives, which is itself a negative-inducing lipid. On the contrary, for the present α-
hydrazido acids the lipophilic portions are unchanged with respect their unmethylated versions,
whereas the N-methylation should reduce the overall ability of the polar and charged groups to act as a
positive curvature-inducing components. The correct balance between components inducing both
positive and negative curvature was demonstrated to be crucial to induce negative curvature in
phospholipid bilayers [89], which in turn is necessary for the formation of pores in bacterial
membranes [90]. Thus, complementary reasons could be at the basis of the more pronounced decrease
in antibacterial efficacy toward Gram-negative bacteria for tetracationic triaryl derivatives and α-
hydrazido acids.
3
. Conclusions
Starting from the idea that an α-hydrazido acid skeleton with suitable derivatizations has an inherent
propensity to the formation of the hydrazido-turn motif, simple mono-charged amphiphilic derivatives
were first computationally analyzed. Then a short, cheap and high-yielding synthesis furnished a
number of these novel mimics of antimicrobial peptides, which were submitted to biological assays.
The membranolytic action of amphiphilic α-hydrazido acids was ascertained, and the most active
compounds exhibited a broad-spectrum in vitro activity against a variety of Gram-positive and Gram-
negative bacteria, which did not change when multidrug resistant strains were employed. Structure-
activity relationships demonstrated that the overall lipophilicity is the main parameter governing the
bacteriostatic activity and the selectivity with respect hemolysis, whereas the change of a hydrazide NH
in favor of a methyl group caused both the loss of a good hydrogen bond donor and the partial
disruption of amphiphilicity, thus reducing the antibacterial activity. Considering that the results
reported here were obtained without performing any optimization in the search for the best bioactivity
profile, we believe that these chemically and proteolytically stable α-hydrazido acids are promising
lead compounds for the development of a new class of wide-spectrum antibiotics. Further studies are
ongoing to evaluate the effect of di- and tricationic side chains, as well as different N- and C-terminal
substituents, in order to find safer and more active compounds.
2
5

4
. Experimental section
4
4
.1 Procedures for biological assays
.1.1 General important notes
All the hydrochlorides of α-hydrazido acids are surfactants, and their tendency to form foams very
negatively affected the first preliminary evaluations of MICs and HC s. This was especially true for
5
0
more concentrated solutions, and was ascertained to be related to the speed of withdrawal and, to a
lesser extent, of the addition, causing evident dispersion and unreliability of results. These observations
also applied to withdrawal and addition of 0.2 vol % Triton X-100. As an example, when both the
withdrawal and the addition of nominal values of 150 µL of 0.2 vol % Triton X-100 solution with a
Gilson P200 pipette were conducted employing about 0.5-1 seconds, the weights of solutions actually
transferred into the vials for five trials were much less than expected and not constant (126.9±12.5 mg,
mean±standard error), as determined with an analytical balance. On the contrary, employing about
three seconds for both the withdrawal and the addition steps, the reliability and reproducibility of
weights greatly improved (150.3±2.3 mg). To make much more reliable and reproducible the results, a
time of at least three seconds was always used in definitive assays for all the withdrawal and addition
steps. For the same reasons, the clear and homogeneous stock solutions were not vortexed immediately
before the experiments. Another variable that is usually underestimated, that is the rigorous mixing in
each well during the serial dilutions, was ascertained to be a secondary factor of error during the
preliminary trials.
4
.1.2 General procedure for the evaluation of minimum inhibitory concentrations
All tested compounds were dissolved in sterile water at the maximum possible concentration.
Bacterial strains were grown for 6 hours in Brain Heart Infusion (BHI) broth and diluted in Mueller-
6
Hinton II (MHII) broth (Oxoid spa, Milan, Italy) to give a final concentration of 1×10 cfu/mL. Serial
dilutions of the tested compounds in MHII broth were prepared in 96-well microtiter plate (Cellstar,
Greiner bio-one, Kremsmünster, Austria) (50 µL per well) and 50 µL of diluted bacterial suspension
were added into each well. The wells with bacteria alone were used as positive growth control wells.
Tetracycline was used as internal control, starting from a 1024 µg/mL stock solution made from a
freshly prepared 10000 µg/mL solution. The plate was aerobically incubated at 37 °C for 24 hours. All
2
6

tests were performed in triplicate. The MICs were defined as the lowest concentrations of compounds
inhibiting visible growth after 24 hours of incubation.
4
.1.3 General procedure for the evaluation of hemolytic concentrations
Hemolysis experiments were performed with a slight modification of a reported procedure [28]. 4
mL of freshly drawn heparinized human blood were diluted with 25 mL of phosphate buffered saline
PBS) pH 7.4, centrifuged at 1000 g for 10 minutes and resuspended in 25 mL of PBS for three times.
After washing, the pellet was resuspended in PBS to 20 vol % and, in a 96-well microtiter plates, 100
µL of erythrocyte suspension were added to 100 µL of different concentrations of the tested compounds
1:2 serial dilutions in PBS) and incubated for 1 hour at 37 °C. The negative and positive controls were
00 µL of PBS and 100 µL of 0.2 vol % Triton X-100, respectively. After incubation, each well was
(
∼
(
1
supplemented with 150 µL of PBS and the plate centrifuged at 1.200 g for 15 minutes. The supernatant
was diluted 1:60 (5 µl of supernatant in 295 µl of PBS), transferred in a new plate, and its absorbance
at a wavelength of 350 nm (A ) was measured using the Synergy HT microplate reader
3
50
spectrophotometer (BioTek, Winooski, VT, USA). The percent hemolysis was determined as follows:
(A − A )] × 100, where A is the absorbance of the test well, A is the absorbance of the
[
0
)/(Atotal − A
0
0
negative control, and Atotal is the absorbance of the positive control. Evaluation of HC and b slope
50
were carried out by nonlinear regression of the four-parameter logistic model of Hill [56], in all cases
were 100% hemolysis was reached in the experiments conducted using stock solutions of the tested
compounds at the maximum possible concentration. For compounds for which the 100% hemolysis
could not be obtained, the three-parameter logistic model of Hill with the constrain of 100% hemolysis
as the final value was used. The errors of the experiments were always less than 10%. The mean values
of three replicates were reported for HC50 and b slope.
4
.1.4 General procedures for the evaluation of outer and inner membrane permeabilization
Outer Membrane Permeabilization Assay. The Perkin Elmer LS 50 spectrometer was used,
operating with the following parameters: 350 nm (slit width 10 nm) for excitation and 420 nm (slit
width 10 nm) for emission, with a measure every 0.1 seconds for 10 minutes. All measurements were
performed at 25 °C on freshly prepared samples, using quartz cuvettes with 10 mm path length. Stock
solution of N-phenyl-1-naphthylamine (NPN, 0.5 mM) in acetone was prepared weekly and stored at 4
8
°
C in a dark place. Midlog-phase E. coli cells (grown for 6 h, 10 cells/mL) were harvested (4000 rpm,
1
0 min at room temperature), washed, and resuspended in PBS buffer at pH 7.2. The bacterial
2
7

suspension was stored at 4 °C and used within 3 hours. Then, the cuvette was supplemented with 2.97
mL of bacterial suspension and, after 15 minutes thermostatation at 25 °C, 6 µL of a 0.5 mM solution
-1
of NPN in acetone (working concentration 1 µM) and 23.4 µL of a 1024 µg mL solution of compound
-1
4
Gc in sterile water (working concentration 8 µg mL ) were added. After a rapid mixing, the outer-
membrane permeabilization was measured by the increase in fluorescence of NPN. Control
experiments were performed with (i) 3 mL of bacterial suspension and 6 µL of 0.5 mM solution of
NPN, (ii) 2.97 mL of PBS buffer at pH 7.2, 6 µL of a 0.5 mM solution of NPN and 23.4 µL of a 1024
-1
µg mL solution of compound 4Gc, (iii) 3 mL of PBS buffer at pH 7.2 and 6 µL of 0.5 mM solution of
NPN, and (iv) 3 mL of PBS buffer at pH 7.2 (see Supplementary Data).
Inner-Membrane Permeabilization Assay. The Perkin Elmer LS 50 spectrometer was used,
operating with the following parameters: 535 nm (slit width 10 nm) for excitation and 617 nm (slit
width 10 nm) for emission, with a measure every 0.1 seconds for 10 minutes. All measurements were
performed at 25 °C on freshly prepared samples, using quartz cuvettes with 10 mm path length.
8
Midlog-phase E. coli cells (grown for 6 h, 10 cells/mL) were harvested (4000 rpm, 10 min at room
temperature), washed, and resuspended in PBS buffer at pH 7.2. The bacterial suspension was stored at
4
°C and used within 3 hours. Then, the cuvette was supplemented with 2.95 mL of bacterial
suspension and, after 15 minutes thermostatation at 25 °C, 30 µL of a 1.5 mM solution of propidium
-1
iodide (PI, 1 mg mL in water, ready to use solution stored at 4 °C in a dark place, working
-1
concentration 15 µM) and 23.4 µL of a 1024 µg mL solution of compound 4Gc in sterile water
-1
(
working concentration 8 µg mL ) were added. After a rapid mixing, the inner-membrane
permeabilization was measured by the increase in fluorescence of PI. Control experiments were
performed with (i) 2.97 mL of bacterial suspension and 30 µL of 1.5 mM solution of propidium iodide,
(
ii) 2.95 mL of PBS buffer at pH 7.2, 30 µL of 1.5 mM solution of propidium iodide and 23.4 µL of a
-1
1
024 µg mL solution of compound 4Gc, (iii) 2.97 mL of PBS buffer at pH 7.2 and 30 µL of 1.5 mM
solution of propidium iodide, and (iv) 3 mL of PBS buffer at pH 7.2 (see Supplementary Data).
4
.1.5 General procedure for the evaluation of antibacterial activity in plasma
S. aureus ATCC 29213 was grown for 6 hours in brain heart infusion (BHI) broth and diluted in
6
Mueller Hinton II broth (Oxoid spa, Milan, Italy) to give a final concentration of 1×10 cfu/mL. Fresh
human blood cells were centrifuged at 3000 rpm for 5 minutes to separate the plasma from the red
blood cells. Three aliquots for each tested compound were dissolved in water at a concentration of 512
µg/mL and diluted twofold in the plasma to reach the final concentration of 256 µg/mL. The aliquots
2
8