Novel Apidaecin 1b Analogs with Superior Serum Stabilities for Treatment of Infections by Gram-Negative Pathogens

Abstract

Proline-rich antimicrobial peptides (PrAMPs) from insects and mammals have recently been evaluated for their pharmaceutical potential in treating systemic bacterial infections. Besides the native peptides, several shortened, modified, or even artificial sequences were highly effective in different murine infection models. Most recently, we showed that the 18-residue-long peptide Api88, an optimized version of apidaecin 1b, was efficient in two different animal infection models using the pathogenic Escherichia coli strains ATCC 25922 and Neumann, with a promising safety margin. Here, we show that Api88 is degraded relatively fast upon incubation with mouse serum, by cleavage of the C-terminal leucine residue. To improve its in vitro characteristics, we aimed to improve its serum stability. Replacing the C-terminal amide by the free acid or substituting Arg-17 with l-ornithine or l-homoarginine increased the serum stabilities by more than 20-fold (half-life, ∼4 to 6 h). These analogs were nontoxic to human embryonic kidney (HEK 293), human hepatoma (HepG2), SH-SY5Y, and HeLa cells and nonhemolytic to human erythrocytes. The binding constants of all three analogs with the chaperone DnaK, which is proposed as the bacterial target of PrAMPs, were very similar to that of Api88. Of all the analogs tested, Api137 (Gu-ONNRPVYIPRPRPPHPRL; Gu is N,N,N′,N′-tetramethylguanidino) appeared most promising due to its high antibacterial activity, which was very similar to Api88. Positional alanine and d-amino acid scans of Api137 indicated that substitutions of residues 1 to 13 had only minor effects on the activity against an E. coli strain, whereas substitutions of residues 14 to 18 decreased the activity dramatically. Based on the significantly improved resistance to proteolysis, Api137 appears to be a very promising lead compound that should be even more efficient in vivo than Api88.

INTRODUCTION

The discovery and subsequent use of antibiotics has revolutionized medicine and dramatically reduced the mortality and morbidity of bacterial infections in humans. It was assumed that humans had overcome bacterial epidemics, despite first reports about bacterial resistance mechanisms, indicating that such claims might be too optimistic. Although these resistance mechanisms already existed in bacteria even 30,000 years ago (1), it is obvious that only the broad use of antibiotics has provided such high evolutionary pressure on pathogens that resistant and, more recently, multi- or pan-resistant pathogens could develop (2, 3). This has triggered much research to find novel antibiotics that use novel modes of action and are directed toward new targets.

One class of antibiotics that have attracted a lot of interest first in immunology and later in pharmaceutical research are antimicrobial peptides (AMPs). AMPs are encoded in the genome of virtually all higher organisms as an important component of innate immunity to microbial infections (4). At least in higher organisms, AMPs perform a dual role by both modulating cells of the host immune system and killing the bacteria directly (5). Most AMPs are positively charged at neutral pH and possess an amphipathic topology that favors their binding to and insertion into anionic phospholipids, which rapidly disrupts the bacterial membrane (membranolytic mechanism) (6). Proline-rich AMPs (PrAMPs) represent an important class of peptides that act by a different, completely nonlytic mechanism, mostly on Gram-negative bacteria, such as Enterobacteriaceae (e.g., Escherichia coli, Klebsiella pneumoniae, and Enterobacter cloacae) and nonfermenting species (e.g., Acinetobacter baumannii and Pseudomonas aeruginosa) (7). These peptides contain a high content of proline (∼30%) and arginine residues and are typically 40 to 60 residues long in mammals (8, 9). Insects express shorter PrAMPs, typically around 20 residues long (10). Both classes act by the same mechanism, i.e., they freely diffuse through the outer membrane, invade the periplasmic space, enter the cytoplasm via an irreversible permease/transporter-mediated uptake (e.g., SbmA transporter), and ultimately inhibit their bacterial targets (e.g., chaperone DnaK) (11–14). This mechanism is highly specific for bacteria, as PrAMPs do not enter mammalian cells (except for some cells of the immune system) and even when artificially introduced are only slightly toxic (15, 16).

The high antimicrobial activity and low to zero toxicity toward mammalian cells have stimulated intense research to evaluate native PrAMPs or optimized analogs for therapeutic applications (17). All these studies have confirmed a very low toxicity in mice and high efficacies in different murine infection models, with peptide doses typically below 10 mg/kg of body weight (BW) (8, 18). Recently, we were able to extend the activity of apidaecin 1b, an 18-residue-long PrAMP expressed in honey bees, toward K. pneumoniae and P. aeruginosa, by modifying two residues and both termini (19). The resulting lead compound, Api88, was highly effective in two murine E. coli infection models without modulating the immune system of the animals (19).

Here we report sequence modifications that increased the serum stability of Api88 to prevent its inactivation in blood by proteolysis at cleavage sites in the C-terminal region. This was accomplished by replacing the arginine in position 17 or replacing the C-terminal amide by the free acid. Thus, we obtained three promising compounds that were much more stable in mouse serum and only slightly less active against the tested pathogens. Importantly, these peptides were neither toxic toward mammalian cell lines nor showed any hemolytic activity. Alanine and d-amino acid scans of the new lead compound Api137 did not indicate further substitutions that might improve its antimicrobial properties.

MATERIALS AND METHODS

Peptide synthesis.

Peptides were synthesized by 9-fluorenylmethoxycarbonyl/tert-butyl (Fmoc/^tBu) chemistry by using either in situ activation with N,N′-diisopropylcarbodiimide (DIC) in the presence of 1-hydroxybenzotriazole (HOBt) or 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU) in the presence of N,N-diisopropylethylamine (DIPEA) (19). Side chains of trifunctional amino acids were protected with 2,2,4,6,7-pentamethyl-2,3-dihydrobenzofuran-5-sulfonyl (Pbf) for Arg; tert-butyl (^tBu) for Tyr; tert-butyloxycarbonyl (Boc) for ornithine (Orn); trityl (Trt) for His, Asn, and Gln (MultiSynTech GmbH, Witten, Germany, and Orpegen Pharma GmbH, Heidelberg, Germany); and methyltrityl (Mtt) for l-ornithine (Iris Biotech, Marktredwitz, Germany). The solid supports used were Rink amide 4-methylbenzhydrylamine (MBHA) and 4-benzyloxybenzyl alcohol resins (Wang MultiSynTech) to obtain C-terminal amides and free acids, respectively.

The N,N,N′,N′-tetramethylguanidino group was incorporated onto the unprotected N terminus by using 10 equivalents of HBTU and DIPEA or N-methylmorpholine in dimethylformamide (19, 20). For the fluorescence polarization studies, methyltrityl-protected ornithine was selectively deprotected after guanidation on the resin by using 2% (vol/vol) trifluoroacetic acid (TFA) in dichloromethane and 5% (vol/vol) triisopropylsilane as scavenger. 5(6)-Carboxyfluorescein was coupled to the unprotected δ-amino group by using DIC/HOBt as described above.

The final peptide resins were washed with dichloromethane, dried, and cleaved with TFA containing 12.5% (vol/vol) of a scavenger mixture (ethanedithiol, m-cresol, thioanisole, and water; 1:2:2:2 [vol/vol/vol/vol]). After 2 h, the peptides were precipitated with cold diethyl ether, washed twice with cold diethyl ether, dried, and purified on a C₁₈ phase column by using a linear aqueous acetonitrile gradient in the presence of 0.1% (vol/vol) TFA as ion pair reagent (Jupiter C₁₈ column; 21.2-mm internal diameter, 250-mm length, 15-μm particle size, 30-nm pore size; Phenomenex Inc., Torrance, CA). The peptide purities were judged by reverse-phase high-performance liquid chromatography (Jupiter C₁₈ column; 4.6-mm or 2-mm internal diameter, 150-mm length, 5-μm particle size, 30-nm pore size; Phenomenex). The molecular weights were confirmed by matrix-assisted laser desorption ionization–time-of-flight mass spectrometry (MALDI-TOF MS; 4700 Proteomic Analyzer; AB Sciex, Darmstadt, Germany).

Antibacterial activities.

The MIC values were determined in triplicate by a liquid broth microdilution assay in sterile 96-well plates (Greiner Bio-One GmbH, Frickenhausen, Germany) using a total volume of 100 μl per well. Aqueous peptide solutions (1 g/liter) were serially 2-fold diluted in 1% (wt/vol) tryptic soy broth (TSB; corresponding to 33% TSB medium), starting at a concentration of 128 μg/ml and diluted down to, typically, 0.5 μg/ml in 8 steps. The overnight cultures in nutrient broth (Carl Roth GmbH, Karlsruhe, Germany) were diluted with 33% TSB medium to 1.5 × 10⁷ cells/ml. Fifty-microliter aliquots of these solutions were then added to each well, giving a starting cell concentration of 7.5 × 10⁵ cells/well.

Alternatively, serial 2-fold peptide dilution series from 32 μg/ml to 0.063 μg/ml (final concentration) were prepared in 1% TSB medium (corresponding to 0.03% [wt/vol] TSB) in a 96-well master plate, and 10 μl was transferred with the Liquidator⁹⁶ (Steinbrenner Laborsysteme GmbH, Wiesenbach, Germany) into sterile 96-well microtiter plates (Greiner Bio-One GmbH, Solingen, Germany). After overnight growth in nutrient broth (Becton, Dickinson, Sparks, MD) at 37°C, bacteria were first diluted in 1% TSB medium to an optical density at 578 nm of 0.1, before being further diluted 285-fold in the same medium. Aliquots of 190 μl of this suspension were added to each well to a final titer of approximately 1 × 10⁵ to 2 × 10⁵ CFU/ml. The plates were incubated for 18 h at 37°C. The MIC was defined as the lowest peptide concentration preventing visible growth of the bacteria.

Time-kill assay.

The time-kill kinetics were determined for E. coli BL21 AI, E. coli DSM 10233, and K. pneumoniae DSM 681 in sterile polypropylene tubes (33% TSB medium, 2 ml). The peptide concentrations were 10-fold or equal to the corresponding MIC values. The positive control did not contain any antibiotic. The inocula to be tested were prepared by adjusting the turbidity of an actively growing broth culture in 33% TSB medium to 5 × 10⁶ CFU/ml. Tubes were continuously shaken on an orbital incubator at 37°C; aliquots of 100 μl were taken in triplicate after 0 and 24 h and for E. coli BL21 AI additionally after 1, 2, 4, and 6 h. These aliquots were then spread in triplicate directly, or after appropriate dilution, onto an agarose plate (1.2%, wt/vol) containing 1% (wt/vol) TSB. Colonies were counted after an incubation period of 24 h at 37°C.

Cytotoxicity.

Cell viabilities were determined in a methyl thiazolyl diphenyl-tetrazolium bromide (MTT) cell proliferation assay (21, 22) for human embryonic kidney (HEK 293), human hepatoma (HepG2), differentiated SH-SY5Y, undifferentiated SH-SY5Y, and HeLa cells. The cell lines were cultured in Dulbeccco's modified Eagle's medium/Ham's F-12 medium (PAA Laboratories GmbH, Coelbe, Germany) with 10% (vol/vol) fetal bovine serum containing 1% (vol/vol) neomycin (10 mg/ml), penicillin, and streptomycin (5 mg/ml; Invitrogen, Karlsruhe, Germany), seeded (2 × 10⁴ cells/well) in the same medium into 96-well plates, and incubated (overnight, 37°C, 5% CO₂) or differentiated with trans-retinoic acid (SY5Y; 10 μmol/liter for 5 days). Cells were washed with phosphate-buffered saline (PBS), and the peptide solutions (0.6 g/liter or 2 g/liter) were added to fresh medium before incubated again under identical conditions as above for 24 h. The cell viability was determined with MTT (10 μl, 5 g/liter). Thus, after incubation (4 h, 37°C, 5% CO₂), a sodium dodecyl sulfate (SDS) solution (10% [wt/vol] in 10 mmol/liter hydrochloric acid; 100 μl) was added and incubated again under the same conditions for 16 h. The absorbance was measured at 590 nm relative to a reference wavelength of 650 nm (Paradigm microplate reader; Beckman Coulter, Wals, Austria) to estimate the viability of the cells. Dimethyl sulfoxide (DMSO) and PBS (both 12% [vol/vol]) were used as positive and negative controls, respectively.

Hemolytic activity.

Concentrated human erythrocytes were washed, suspended in PBS (2% [vol/vol]; 50 μl), added to a serial peptide dilution series from 600 to 5 μg/ml in PBS (50 μl; 96-well polypropylene plates, V bottom; Greiner Bio-One GmbH), and incubated (37°C, 1 h) (23). After centrifugation (1,000 × g), the absorbance of the supernatants was determined in a 384-well plate (flat bottom; Greiner Bio-One GmbH) at 405 nm in a Paradigm microplate reader. The positive controls were 0.1% Triton X-100 and melittin (75 to 0.6 μg/ml), and the negative control was PBS.

Generation of murine BMDC.

If not stated otherwise, all cell culture reagents were purchased from PAA Laboratories GmbH. Bone marrow of both femora and tibiae from female C57BL/6J mice (8 to 9 weeks old) was flushed with ice-cold PBS containing 5% fetal bovine serum (FBS; Gibco Life Technologies GmbH, Darmstadt, Germany). Erythrocytes were lysed using Gey's solution. Bone marrow cells were counted, seeded (1.75 × 10⁶ cells/ml), and incubated (37°C, humidified atmosphere containing 5% CO₂) in bone marrow-derived dendritic cells (BMDC) culture medium containing RPMI 1640 medium supplemented with 10% heat-inactivated FBS, penicillin (100 units/ml), streptomycin (100 μg/ml), β-mercaptoethanol (50 μmol/liter; Sigma, Taufkirchen, Germany), and human fms-like tyrosine kinase ligand (Flt3-ligand; 200 ng/ml; Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). The BMDC culture obtained after 7 days consisted of a mixture of two DC subsets: (i) conventional dendritic cells (cDC) positive for CD11c, highly positive for CD11b, and negative for B220, and (ii) plasmacytoid dendritic cells (pDC), positive for both CD11c and B220 and intermediately positive for CD11b.

BMDC stimulation.

Following two washing steps, BMDC were seeded in BMDC differentiation medium (5 × 10⁵ cells/ml; 24-well plates). Cells were stimulated with Api137 (400 μg/ml) or the Toll-like receptor ligands lipopolysaccharide (LPS; 5 μg/ml), CPG-ODN (1 μmol/liter), and poly(I·C) (100 μg/ml) as positive controls. After 24 h, the cells were collected in 1.5-ml tubes, centrifuged (300 × g, 4°C, 8 min), and the cell supernatants were harvested to determine the cytokine levels in an enzyme-linked immunosorbent assay (ELISA). Additionally, stimulated cells were collected for flow cytometric analysis of their activation marker expression.

Quantification of cytokines.

Interleukins (IL) IL-12p40 and IL-6 were quantified in the supernatants of stimulated BMDC by a sandwich ELISA following the manufacturer's protocol. Briefly, the assay used 96-well round-bottom ELISA plates (Thermo Fisher Scientific/Nunc, Roskilde, Denmark) and capture and detection antibodies (Abs) for IL-6 (BD Pharmingen, Heidelberg, Germany) or IL-12p40 (capture Ab, anti-msIL-12p40, clone 5C3, 25 μg/ml; detection Ab, biotinylated goat anti-msIL-12/23p40, 1:1,000; both Abs were kind gifts from Hoffmann-La Roche Ltd., Basel, Switzerland). The standards used were recombinant IL-6 (Peprotech, Hamburg, Germany) and recombinant murine IL-12p70 (provided by M. Gately, Hoffmann-La Roche, Nutley, NJ). Samples and standards were diluted in serum diluent, i.e., PBS containing bovine serum albumin (0.5%, wt/vol), gelatin (0.1%, wt/vol), and Tween 20 (0.05%, vol/vol). Biotinylated antibodies were detected by horseradish peroxidase-linked streptavidin (streptavidin-HRP; 1:3,000; Southern Biotechnology Associates, Birmingham, AL) and the 3,3′,5,5′-tetramethylbenzidine (TMB) microwell peroxidase system (KPL, Gaithersburg, MD). The reaction was stopped with phosphoric acid (50 μl; 1 mol/liter), and the absorbance was recorded at 450 nm and 630 nm with a Spectra-max 340 ELISA reader using SoftMax Pro (Molecular Devices, Biberach a.d. Riss, Germany).

Flow cytometry (FACS).

All fluorescence-activated cell sorting (FACS) staining steps were performed at 4°C. Stimulated BMDC were collected, washed with PBS, and stained with ethidium monoazide bromide (0.5 μg/ml; Invitrogen GmbH, Karlsruhe, Germany) according to the manufacturer's instruction manual to detect dead cells within the sample. Subsequently, cells were washed with PBS and twice with FACS buffer, i.e., PBS containing FBS (3%, wt/vol) and NaN₃ (0.1%, wt/vol). Following blocking of FcγRΙΙ/III by preincubation with purified CD16/CD32 monoclonal antibody (clone 2.4G2; 2 μg/ml for 1 × 10⁶ cells; kindly provided by Frank Brombacher, ICGEB, Cape Town, South Africa), cells were stained using various antibodies, i.e., rat IgG2a isotype control conjugated to phycoerythrin (PE; eBioscience, Frankfurt, Germany), rat IgG2b isotype control conjugated to fluorescein isothiocyanate (FITC; BD Pharmingen), hamster IgG1 isotype control conjugated to allophycocyanin (APC; BD Pharmingen), anti-mouse CD11b–FITC (clone M1/70; eBioscience), anti-mouse CD86–PE (clone GL1; BD Pharmingen), anti-mouse B220–PE (clone RA3-6B2; eBioscience), or anti-mouse CD11c–APC (clone N418; biolegend, San Diego, CA). After incubation (20 min, 4°C), cells were washed twice with FACS buffer, once with PBS, fixed with paraformaldehyde (2% [wt/vol] in PBS), and washed once with PBS and FACS buffer. The stained samples were analyzed on a BD FACSCalibur (Becton, Dickinson, Heidelberg, Germany) flow cytometer and using the BD CellQuest software (Becton, Dickinson).

Serum stability.

Peptides (15 μg) were dissolved either in 25% (vol/vol) aqueous or undiluted pooled mouse serum (100 μl; PAA Laboratories GmbH) and incubated at 37°C (23). Aliquots were taken in duplicate or triplicate after 0, 30, 60, 120, 240, and 360 min and precipitated by addition of trichloroacetic acid to a final concentration of 3% (wt/vol). After 10 min on ice, the samples were centrifuged, and the supernatant was neutralized with sodium hydroxide solution (1 mol/liter) and stored at −20°C. The samples were analyzed on an analytical Jupiter C₁₈ column with a linear aqueous acetonitrile gradient containing 0.1% (vol/vol) TFA. Metabolites were identified by MALDI-TOF MS. The half-lives were determined by using an exponential fit (MS Excel).

Fluorescence polarization.

The binding constants were determined with full-length DnaK and 5(6)-carboxyfluorescein-labeled peptides in polarization buffer (20 mmol/liter Tris, 0.15 mol/liter KCl, 5 mmol/liter MgCl₂, 1 mmol/liter NaN₃, and 2 mmol/liter dithiothreitol; pH 7.5) at 28°C, as described recently (24). Briefly, black 384-well plates (flat bottom; Greiner Bio-One GmbH) were blocked with 0.5% (wt/vol) casein in washing buffer (10 mmol/liter sodium phosphate, 0.3 mol/liter NaCl, pH 7.4, and 0.05% [vol/vol] Tween 20) at room temperature for 1 h and washed three times with washing buffer. A 2-fold dilution series of full-length DnaK (40 μl/well) in polarization buffer containing a labeled peptide (10 nmol/liter) was added and incubated (28 ± 1°C). After 2 h the fluorescence anisotropy was measured on a Paradigm microplate reader in the top read position (λ of 485 nm for excitation and λ of 535 nm for emission) in triplicate on at least two different days. The data were fitted to a nonlinear, dose-response logistical transition equation [y = a₀ + a₁/(1 + x/a₂)^a₃] by using the Levenberg-Marquardt algorithm with the dissociation constants (K_D) being represented by the a₂ coefficients (SlideWrite, Encinitas, CA).

Statistical analysis.

Different groups were compared using the unpaired Mann-Whitney test, and significant differences are expressed at P levels between ≤0.001 and ≤0.05.

RESULTS AND DISCUSSION

Serum stability.

Based on the recently reported in vitro properties and in vivo efficacies of the lead compound Api88 (19), we tested the degradation of Api88 in mouse serum as the first part of a more detailed pharmacokinetic study. Api88, native apidaecin 1b, and C-terminally amidated apidaecin 1b were degraded in 25% aqueous mouse serum at 37°C, with half-lives of 10, 254, and 19 min, respectively (Table 1; Fig. 1A). Surprisingly, both of the peptides with amidated C termini showed similarly short half-lives, whereas the native sequence with the C-terminal acid was around 15 times more stable, indicating that the serum proteases might cleave the peptide near the C terminus. The chromatograms displayed two signals besides that of Api88, with retention times of 29.0 min and 29.6 min (see Fig. S1 in the supplemental material). Neither signal was present at the beginning of the incubation period. Mass spectrometry indicated mass losses of 112 m/z and 268 m/z relative to Api88 for these metabolites (see Fig. S1), which points to the C-terminally truncated peptides ^1–17Api88 and ^1–16Api88, respectively. The peak area of ^1–17Api88 increased very rapidly during the first 30 min to 70% of the initial peak area of Api88 and then showed a further slight increase to 80% within the next 30 min, before decreasing slowly afterwards (Fig. 1A). As only a single nonaromatic residue was cleaved at the C terminus, the extinction coefficients of this metabolite and Api88 should be very similar. Thus, the relative peak areas should reflect the peptide molar ratios very well. In contrast, the signal corresponding to ^1–16Api88 increased continuously over the observation period, but its peak area was still below 10% after 2 h (Fig. 1A). Thus, only the two C-terminal residues were cleaved off Api88, whereas the remaining sequences ^1–17Api88, and especially ^1–16Api88, were relatively inert toward serum proteases and peptidases. Since both truncated peptides were significantly less active against E. coli (Table 1), it was necessary to modify the C-terminal part of Api88 to stabilize it against proteolysis.

Table 1.

Sequences, MICs, and half-lives in mouse serum of apidaecin 1b and related designer peptides^a

Peptide no.	Code	Sequence	MIC (μg/ml) for E. coli BL21 AI	Half-life (min) in:
				25% aqueous mouse serum	Undiluted mouse serum
1	Apidaecin 1b	GNN...PHPRL-OH	1	254	161
2	Apidaecin 1b amide	GNN...PHPRL-NH2	2	19	12
3	Api88	Gu-ON...PHPRL-NH2	0.5	10	5
4	1–17Api88 amide	Gu-ON...PHPR-NH2	64	ND	152
5	1–16Api88 amide	Gu-ON...PHP-NH2	>128	ND	>360
6	Api134	Gu-ON...PHPOL-NH2	2	>360	244
7	Api137	Gu-ON...PHPRL-OH	0.5	>360	345
8	Api155	Gu-ON...PHP-Har-L-NH2	2	>360	173
9	Api161	Gu-ON...PHPRL-NH-C3H7	4	<30	ND
10	Api162	Gu-ON...PHPRL-NH-CH3	4	<30	ND
11	Api137 (R4A)	Gu-ONNAPVYIPRPRPPHPRL-OH	0.5	ND	307

^aGNN…PHP and Gu-ON…PHP are abbreviations for the N-terminal sequences of apidaecin 1b and Api88 from residues 1 to 16, i.e., GNNRPVYIPQPRPPHP and Gu-ONNRPVYIPRPRPPHP, respectively. The MICs were determined in 33% TSB medium, and the half-lives were determined in undiluted and 25% aqueous mouse serum at 37°C. Gu, N,N,N′,N′-tetramethylguanidino; ND, not determined; O, l-ornithine; Har, l-homoarginine.

Fig 1.

Degradation of Api88 in 25% aqueous mouse serum (top) and of Api88, Api134, Api137, and Api155 in undiluted mouse serum (bottom). The peptide quantities were determined based on the peak areas obtained in the corresponding RP chromatograms and are shown here relative to the initial peptide quantities of each peptide at an incubation time of 0 min (set to 100%). The two major degradation products of Api88 in diluted mouse serum were identified by MALDI-TOF MS as ^1-17Api88 and ^1-16Api88 (top).

Among the many strategies published in recent years to stabilize peptides against proteases, we tested only modifications at single positions that (i) could stabilize both peptide bonds, and (ii) could be realized with commercial reagents without the need to synthesize special reagents or amino acid derivatives. Substitution of ¹⁸Leu in Api88 or apidaecin 1b amide with tert-leucine, cyclohexyl-l-alanine, d-leucine, or N-methylleucine significantly improved the serum stabilities but also abolished the antibacterial activity against E. coli (MIC, >128 μg/ml). The same was true for analogs containing β-homoarginine, d-arginine, or N^α-methyl-l-arginine at position 17 instead of arginine. Only two substitutions were identified that improved the serum stability without abolishing the antibacterial activities against E. coli BL21 AI, i.e., replacing ¹⁷Arg with either l-ornithine (Api134) or l-homoarginine (Api155) (Table 1). Alkylation of the C terminus with a propyl (Api161) or methyl (Api162) group reduced the activities against E. coli at least 8 times relative to Api88, whereas replacement of the C-terminal amide with the acid (Api137) did not affect the antibacterial activity (MIC, 0.5 μg/ml) (Table 1). Thus, only Api134, Api137, and Api155 appeared to be promising candidates. Their half-lives in full mouse serum were around 4 h, 6 h, and 3 h, respectively (Fig. 1B; Table 1), which was at least 35 times more than for Api88 (∼5 min) and even more than for native apidaecin 1b (∼161 min). The half-lives in horse serum and porcine serum were similar or slightly higher (see Table S1 in the supplemental material). Considering the time-kill kinetics of Api88 (19), which show that the number of CFU is reduced in vitro by at least 3 orders of magnitude within 1 h, the half-lives of the new analogs matched the pharmaceutical requirements very well.

Antibacterial activities.

Api137 was as active as Api88 against E. coli strains ATCC 25922 and DSM 10233, K. pneumoniae strain DSM 681, and P. aeruginosa strain DSM 3227, and only slightly less active against K. pneumoniae strain DSM 11678 and P. aeruginosa strain DSM 9644 (Table 2). The antibacterial activity decreased further for Api155 and Api134. Api134 was almost inactive against K. pneumoniae DSM 11678 and P. aeruginosa DSM 3227 (Table 2), and its antibacterial activities were similar to apidaecin 1b.

Table 2.

MICs of apidaecin 1b, Api88, Api134, Api137, and Api155 determined for three E. coli, two K. pneumoniae, and two P. aeruginosa strains in 33% TSB medium

Peptide	MIC (μg/ml) for species and strain
	E. coli		K. pneumoniae		P. aeruginosa
	ATCC 25922	DSM 10233	DSM 681	DSM 11678	DSM 9644	DSM 3227
Apidaecin 1b	4	0.25	8	>64	>128	>128
Api88	2	1	2	4	8	32
Api134	16–32	2	32	>64	16	128
Api137	4	0.5	2	32	16	32
Api155	8	2	4	32	16	64

The time-kill curves obtained for Api88, Api137, and Api155 concentrations corresponding to the respective 10-fold MIC values (e.g., 10 μg/ml for Api88) reduced the CFU of E. coli strains BL21 AI and DSM 10233 within the first 30 min, by approximately 10-fold (Fig. 2, top, and Table 3). This can be considered fast, as apidaecins inhibit intracellular targets and do not kill the bacteria by lytic mechanisms. After 24 h the CFU was reduced for both E. coli strains by more than 3 orders of magnitude, which we considered bactericidal activity. For K. pneumoniae DSM 681, only Api88 was bactericidal, whereas both Api137 and Api155 reduced the CFU only during the first 6 h before the bacteria could grow again (bacteriostatic activity). Ten-fold-lower peptide concentrations (1× MIC) altered the time-kill curves only slightly during the first 6 h (Fig. 2, bottom). All three peptides were still bactericidal for E. coli DSM 10233, but only Api88 remained bactericidal for E. coli BL21 AI (Table 3). The CFU count for K. pneumoniae DSM 681 after 24 h was basically identical to the values obtained at the higher peptide concentrations (10× MIC) (Table 3).

Fig 2.

Time-kill curves for E. coli BL21 AI based on peptide concentrations corresponding to the 10× MIC (top) and the MIC (bottom) of Api88 (▲), Api137 (■), and Api155 (⧫). No antibiotic was added to the positive control (●). Bactericidal activity (dotted line) was defined as a >1,000-fold (Δlog of −3) decreased number of viable CFU after 24 h. The CFU counts were determined in triplicate on agarose plates. Shown are the results of one representative experiment.

Table 3.

Time-kill assay

Peptide	Changes in CFU valuesa (Δlog) for strain and MIC strength and conclusion
	E. coli BL21 AI		E. coli DSM 10233		K. pneumoniae DSM 681
	10× MIC	1× MIC	10× MIC	1× MIC	10× MIC	1× MIC
Control	2.9 (2.0)	2.4 (0.1)	1.8	2.9	2.1	1.4
Api88	< −3 (−2.3), bactericidal	< −3 (−2.2), bactericidal	< −3, bactericidal	< −3, bactericidal	< −3, bactericidal	−3, bactericidal
Api137	<-3 (−2.1), bactericidal	2.4 (−2.9), bacteriostatic	< −3, bactericidal	< −3, bactericidal	−3, bactericidal	∼1, bacteriostatic
Api155	< −3 (−2.8), bactericidal	0.8 (−2.6), bacteriostatic	< −3, bactericidal	< −3, bactericidal	∼1, bacteriostatic	∼1, bacteriostatic

^aCFU count after an incubation time of 24 h (6 h).

Cellular toxicity.

As expected from the data obtained for Api88 (19), Api134, Api137, and Api155 were not toxic to HEK293, HepG2, HeLa, differentiated SY5Y, or undifferentiated SY5Y cells (see Fig. S2 in the supplemental material). Similarly, none of the three compounds showed any hemolytic activity against human red blood cells (see Fig. S3 in the supplemental material).

Activation of dendritic cells.

As several antimicrobial peptides can stimulate cells of the innate immune system, such as dendritic cells, monocytes, and macrophages (25–29), we studied Api137 for stimulatory effects on murine Flt3 ligand-generated BMDC, which comprise cDC and pDC (30–33). The identities of the differentiated BMDC were confirmed by their expression of DC-specific markers (i.e., CD11c, CD11b, and B220), which indicated that they contained 61 to 79% cDC and 6.2 to 7.4% pDC (data not shown). The BMDC secreted the proinflammatory interleukins IL-12p40 and IL-6 in response to stimulation with the Toll-like receptor ligands LPS, CpG-ODN, and poly(I·C), whereas incubation with Api137 did not induce the secretion of inflammatory cytokines (Fig. 3A and B). Additionally, Api137 did not upregulate the expression of the activation marker CD86 on cDC or pDC, whereas the Toll-like receptor ligands LPS, CPG-ODN, and poly(I·C) clearly activated both cDC and pDC (Fig. 3C and D). Both findings are consistent with a recent report on PrAMPs, including apidaecin 1b and Api88 (34).

Fig 3.

Effect of Api137 on the stimulation of murine dendritic cells. BMDC were generated in vitro and stimulated for 24 h with Api137 (400 μg/ml) or the Toll-like receptor ligands LPS, CpG-ODN, or poly(I·C) as positive controls for cell activation. Cell supernatants were collected, and the concentrations of IL-12p40 (A) and IL-6 (B) were determined by ELISA. The activation statuses of conventional dendritic cells (C) and plasmacytoid dendritic cells (D) were determined based on the mean fluorescence intensity of the CD86 signal. Pooled data from two individual experiments are shown (means ± standard deviations; each experiment was performed with duplicate samples which were pooled for flow cytometry staining). For statistical analysis, an unpaired Mann-Whitney test was used. n.d., cytokine levels were not detectable (below the detection limit).

DnaK binding.

The DnaK binding constants were determined for fluorescently labeled Api88, Api134, Api137, and Api155 in a fluorescence polarization assay using full-length DnaK. Api88 and the three analogs were all labeled with 5(6)-carboxyfluorescein at the δ-amino group of Orn-1, as the N terminus was already guanidated. The K_D values determined for Api134 (6.7 ± 1.3 μmol/liter [mean ± standard deviation]), Api137 (8.7 ± 1.0 μmol/liter), and Api155 (2.5 ± 0.2 μmol/liter) (see Fig. S4 in the supplemental material) were very similar to the recently reported K_D value of Api88 (5.0 ± 1.2 μmol/liter), clearly indicating that the new compounds bound equally well to the substrate binding domain (SBD) of DnaK. This is in full agreement with the X-ray data for Api88, which demonstrated that residues 3 to 11 interact with DnaK (19).

Selection of a lead compound.

Based on the aim of this study to improve significantly the serum stability of Api88, Api137 appeared to be the best lead compound, with half-lives in mouse, horse, and porcine serum of more than 6 h. Moreover, its antibacterial activities against the six tested bacterial strains were only slightly reduced, whereas Api134 and especially Api155 were significantly less active. Although K. pneumoniae DSM 11678 was around 8-fold less susceptible to Api137 than to Api88, the 24-fold-higher serum stability should compensate for this activity loss in vivo due to the longer exposure time. At the same time, Api137 was not toxic to mammalian cells, showed no hemolytic activity, and bound to DnaK as well as Api88 did. Thus, we selected Api137 for further studies. When evaluated for its antibacterial activities against a broader panel of bacteria (Table 4), which was also used in a previous study to evaluate Api88 (19), it appeared highly active against several clinically relevant Gram-negative pathogens (Table 4). The MIC values of Api137 and Api88 were very similar (see Table S2 in the supplemental material). Specifically, their MIC values were identical for 26 of the tested pathogens. Api137 was twice as active against two E. coli and two K. pneumoniae strains but only half as active against five P. aeruginosa and two P. vulgaris strains. Thus, the reduced net charge of Api137 with its free C terminus (compared to the amide in Api88) did not negatively affect its antibacterial activity in general, although the slightly higher MIC values for five of the eight tested P. aeruginosa strains might indicate that the C-terminal charge is important for this pathogen.

Table 4.

Antimicrobial activities of Api137 and Api137 (R4A) against clinically relevant Gram-negative pathogens^a

Pathogen	No. of tested strains and isolates	MIC (μg/ml) range
		Api137	Api137 (R4A)
E. coli	8	0.5	0.25–0.5
K. pneumoniae	6	0.25–0.5	0.25–1
P. aeruginosa	8	2–8	4–16
A. baumannii	5	1–2	0.5–1
E. cloacae	5	0.25	0.25–0.5
P. vulgaris	5	0.125–2	0.25–4

^aFor further details, see Table S2 in the supplemental material.

Positional alanine and d-amino acid scans of Api137.

Having identified a promising lead compound with good serum stability, we probed each position of the Api137 sequence for its influence on the antibacterial activity against E. coli BL21 AI. Thus, each position of the Api137 sequence was individually replaced with either the corresponding d-amino acid or l-alanine, e.g., l-ornithine (Orn) in position 1 was replaced by d-ornithine (Orn) or alanine. The MIC values of these Api137 analogs indicated that the N-terminal sequence tolerated substitutions without major effects on the antibacterial activity (Fig. 4A), although apidaecins and Api88 are presumed to inhibit DnaK by binding of residues 3 to 11 to the substrate binding domain of DnaK (19, 35, 36). Substitutions of C-terminal residues (Pro-11 to Leu-18) significantly reduced the antibacterial activity in most cases. Interestingly, replacing a given residue with the corresponding d-amino acid or alanine changed the MIC values for most positions to a similar degree, except for positions Pro-9, Arg-10, and His-15 to Leu-18, where the MIC values differed by more than a factor of 2. Extreme differences were obtained when His-15 and Leu-18 were replaced with d-His and d-Leu (activity loss against E. coli was larger than 32-fold), or with alanine (a 2- to 4-fold loss in activity, respectively) (Fig. 4A). This might indicate that the positive charge at position 15 and the hydrophobicity of position 18 are less important for the mode of action than the orientation of the side chains and/or the peptide backbone. In contrast, replacement of Arg-17 with alanine abolished the antibacterial activity, whereas replacement of d-Arg reduced it only by a factor of 8. Although these observations will be important in guiding future efforts to further optimize the structure of Api137, a detailed interpretation of these results has to be delayed until the exact function of the C-terminal sequence and its possible bacterial target or interaction partner (e.g., transporter proteins) are identified.

Fig 4.

MIC values of Api137 and Api137 analogs resulting from the positional l-alanine and d-amino acid scan (A), as well as with trans-4-hydroxy-l-proline and cis-4-hydroxy-l-proline (B), as indicated on the abscissa. The MIC values were determined for E. coli BL21 AI in a microdilution assay. Hash symbols on tops of some columns indicate that the MIC values were ≥32 μg/ml. Asterisk, peptide Api137.

As it is generally accepted that the positive charge of antimicrobial peptides directly influences their activity, it was interesting to see that both Arg-4 and Arg-10 could be replaced by Ala without a loss in efficacy against E. coli. Thus, we tested Api137 (R4A) against a broader panel of pathogens (Table 4; see also Table S2 in the supplemental material). The reduced positive charge basically did not influence the MIC values for E. coli, K. pneumoniae, or E. cloacae, but it reduced them 2-fold for A. baumannii and increased them 2-fold for P. vulgaris and even 4-fold for P. aeruginosa.

Replacement of proline with hydroxyproline.

Work with the Ala and d-Pro analogs of Api137 already indicated that substitution of the proline residues in positions 5, 9, or 13 reduced the activity only slightly (MIC, ≤4 μg/ml), whereas positions 11 (8 μg/ml), 14 (8 μg/ml for d-Pro and 16 μg/ml for Ala), and 16 (≥32 μg/ml for d-Pro and 8 μg/ml for Ala) were more sensitive. Thus, for all six Pro residues, we additionally investigated the effect of substitution with trans-4-hydroxy-l-proline (^tHyp) or cis-4-hydroxy-l-proline (^cHyp). The effect of these conservative substitutions on the antibacterial activity was rather minor (Fig. 4B), except for Pro-16. The MIC values of those ^tHyp and ^cHyp analogs increased by 2-fold and 4-fold, respectively. This confirmed again the importance of this position for the antibacterial activity of Api137, although the effect was much weaker than for the Ala and d-Pro substitutions. This indicated that the polarity of these residues and most likely also the cis/trans isomerization of the C-terminal peptide bond influence the antibacterial activity only weakly. The Hyp substitutions did not affect the serum stability (data not shown). Thus, we did not further investigate the Hyp analogs, although they might increase the stability against bacterial proteases (23) or improve the pharmacokinetics.

Conclusions.

The fast degradation of Api88 at the C terminus was prevented by replacing Arg-17 with Orn or Har or by replacing the C-terminal amide with the free acid. Whereas the substitution of Arg-17 reduced the antibacterial activity, especially toward K. pneumoniae and P. aeruginosa, the presence of the free acid had only a minor effect on the activity. The serum stabilities of Api137 increased more than 20 times, to around 6 h, which is probably much longer than the clearance rates in mice through kidneys and liver. Importantly, Api137 was not toxic to various cell lines and did not stimulate murine bone marrow-derived dendritic cells, which is a first indication that it does not stimulate the immune system.

DnaK represents an interesting target for antibiotics, as it is universal in bacteria and structurally highly conserved, especially in the SBD (37, 38). Thus, mutations in the SBD are rather unlikely, as they would probably disturb its chaperone activity. Studies with ΔdnaK mutants have shown that DnaK is only essential at elevated temperatures (>37°C). Interestingly, ΔdnaK mutants are still susceptible to PrAMPs, which indicates further bacterial targets that have not been identified yet (19).