W. Matthew
Leevy
a,
Seth T.
Gammon
b,
Tatiana
Levchenko
e,
David D.
Daranciang
a,
Oscar
Murillo
a,
Vladimir
Torchilin
e,
David
Piwnica-Worms
ab,
James E.
Huettner
c and
George W.
Gokel
*ad
aDepartment of Molecular Biology & Pharmacology, Washington University School of Medicine, Campus Box 8103, 660 S. Euclid Avenue, St. Louis, MO 63110, USA. E-mail: ggokel@wustl.edu; Tel: +1 314 362 9297
bDepartments of Radiology, Washington University School of Medicine, Campus Box 8103, 660 S. Euclid Avenue, St. Louis, MO 63110, USA
cDepartments of Cell Biology and Physiology, Washington University School of Medicine, Campus Box 8103, 660 S. Euclid Avenue, St. Louis, MO 63110, USA
dDepartment of Chemistry, Washington University, 1 Brookings Drive, St. Louis, MO 63130, USA
eDepartment of Pharmaceutical Sciences, Bouve College of Health Sciences, Northeastern University, Boston, MA 02115, USA
First published on 26th August 2005
Hydraphile compounds are shown to be cytotoxic to Gram-negative and Gram-positive bacteria, yeast, and mammalian cells. Their cellular toxicity compares favorably with other synthetic ionophores and rivals that potency of natural antibiotics. The effects of structural variations on toxicity are described. The effects of these variations correlate well with previous studies of ion transport in liposomes. Whole cell patch clamping with mammalian cells confirms a channel mechanism in living cells suggesting that this family may comprise novel and flexible pharmacological agents.
Membrane-spanning proteins have garnered attention for their roles in regulating numerous cellular processes, including ion balance, cell signaling, and the uptake of organic substrates. Of particular interest have been protein channels that mediate the passage of ions through membranes. Channel proteins play the remarkable role of moving charged species through the approximate 30 Å thickness of membrane hydrocarbons. The “hydrocarbon slab” or “insulator regime” is thought to have a dielectric constant of 2–3. Nature's rigorous maintenance of cellular salt concentrations demonstrates the importance of ion transport, which must occur through the insulating bilayer. The protein ion channels that mediate ion conductance have been studied for decades, but structural details have only recently become available.2 Moreover, mechanistic details remain obscure in most cases.
Because channel proteins are so complex, chemists as early as the 1980s responded to the challenge of developing synthetic ionophores that are working models of these remarkable natural systems.3 Early contributions to the field came from the groups of Tabushi,4 Lehn,5 Fyles,6 and our own laboratory.7 More recent contributions include the peptide nanotubes of Ghadiri,8 the peptide-linked crown ethers of Voyer,9 our chloride-selective heptapeptide channels,10 and others.11 Many of the synthetic ion channel compounds have demonstrated classic open/closed channel behavior in planar bilayer voltage clamp experiments3,12,13 and have shown activity in synthetic liposomes.14–16 The peptide back-boned crown system9 is thought to form an α-helix in the membrane, which then aligns the macrocycles to form tubes through the bilayer. The cyclic D,L-amino acid nanotubes8 stack to form an extended, hydrogen-bonded tunnel that permits the passage of ions or sugars across the membrane. The hydraphiles mimic the structure of known protein channels by having an entry and exit portal connected to a centrally hydrated relay. Fig. 1 shows a schematic representation of these three ion conducting compounds or assemblies as they are thought to function in the bilayer.
![]() | ||
Fig. 1 Presumed channel conformations for (left) peptide nanotubes, (center) hydraphiles, and (right) multi-crown peptide. |
Natural ionophores are known to exhibit toxic effects in cellular systems. Examples include valinomycin,17 the cecropins,18 gramicidin,19 and melittin.20 The synthetic ionophores described above are also biologically active. Specifically, the peptide-linked crown ethers cause red blood cell haemolysis but are inactive against E. coli,21 while the cyclic peptide nanotubes and the synthetic hydraphiles act as potent antimicrobial agents.22,23 The cyclic peptide nanotubes are quite active, killing bacteria at concentrations as low as 2.5 µM.24 The hydraphiles, studies of which are reported here, are also cytotoxic to bacteria at concentrations as low as 0.6 µM. Other synthetic channels25 presumably hold this potential, although, to our knowledge, it is currently unrealized.
The toxicity of synthetic channels is presumably the result of rapid, unregulated ion flux through the plasma membrane that disrupts cellular ion gradients causing osmotic stress and, ultimately, cell death. Two mechanisms have been proposed to describe this process.26 Previous studies using membrane potential sensitive dyes showed that hydraphiles act as ionophores in the E. coli membranes.23 The hydraphiles were designed to be modular and are available in a range of lengths and polarities. We have assayed the biological activity of 13 synthetic ionophores and report here the observed structure–activity relationships. In addition, evidence for the mechanism of action is presented for the hydraphile family of compounds.
Compounds 1–11 and 13 contain three macrocycles linked covalently. The distal macrocycles of 2–11 are 4,13-diaza-18-crown-6, but in 1 the corresponding macrocycle is aza-18-crown-6. The sidearms in structures 4–10 are benzyl and n-dodecyl in 11. Compounds 4–10 differ in spacer chain length. The shortest chain is octylene (4) and the longest is eicosylene (10). The spacer chains in tris(macrocycle)s 1, 2, 6, 11 and 13 are all dodecylene. Compound 12 differs from the other structures in this study because a fourth macrocycle is present and linked covalently to the distal macrocycles. This symmetrical structure is thought to form a “tunnel” within the bilayer and is the most active sodium ion transporter reported in this study. Finally, compound 13 is identical to 6 except that para-fluoro substituents are present on the benzyl groups.
Structural and fluorescence studies indicate that the hydraphiles arrange in the bilayer with the distal macrocycles at opposite ends of the membrane's insulator regime while the hydrocarbon chains (e.g. the dodecyl groups of 11) align with the fatty acid chains.29 Compound 1 does not possess a side chain; instead the distal macrocycles are aza-18-crown-6 residues. In this compound, the hydraphile's typical N–R side chain is replaced by a single oxygen atom. Compound 1 showed no detectable sodium transport activity in 23Na-NMR experiments conducted in liposomes.14 In these same studies, compounds 2, 6, and 11 had relative rates of 28, 39, and 28 (compared to the transport activity of gramicidin (arbitrarily set to 100) determined simultaneously). Compound 12, the dodecyl chain “tunnel” structure, shows a relative rate for sodium ion transport of approximately 81 compared to the gramicidin standard.30 These sodium transport results indicate that an
N–H (not an ether link) is the minimum sidearm commensurate with transport activity. Other side chains and spacer chain lengths alter transport activity. These results are mirrored in the biological results described below.
Compound | |||||||
---|---|---|---|---|---|---|---|
4 | 5 | 6 | 7 | 8 | 9 | 10 | |
a Minimum bactericidal concentration (in µM). b Not determined. c Percent release of Na+ from liposomes determined by using an ISE technique.15 | |||||||
Spacer chain length | (CH2)8 | (CH2)10 | (CH2)12 | (CH2)14 | (CH2)16 | (CH2)18 | (CH2)20 |
E. coli | 170 | 80 | 9.4 | 2.3 | 4.6 | 75 | 160 |
B. subtilis | 42 | 10 | 1.2 | 0.56 | 0.6 | 1 | 2 |
S. cerevisiae | 170 | 160 | 38 | 4.6 | 2.3 | —b | —b |
Na+ releasec | ∼2 | 23 | 84 | 100 | 95 | 39 | 14 |
Replacing the macroring oxygen of 1 by NH results in 2, which shows a significant increase in toxicity to both E. coli and B. subtilis (MBC = 22 µM and 11 µM), while the yeast remains resistant. “Benzyl channel”, 6, is formed when a benzyl group replaces the hydrogen of the N–H group of 2. The activity of 6 increases compared to 2 by about 2-fold for E. coli (MBC = 9.4 µM) and 10-fold for B. subtilis (MBC = 1.2 µM). A moderate increase in activity is also observed for S. cerevisiae (MFC = 39). The toxicity of these compounds rates well as antibiotics are generally considered effective at MBCs of 10 µM and below.33 When the distal macrocycles of the hydraphile framework are terminated by 12-carbon side chains (11), activity increases again by about two-fold to 2.1 µM for E. coli and 0.53 µM for B. subtilis. The yeast sustain an approximate 5-fold increase in activity to 8.4 µM.
Linking the dodecyl side chains through an additional macrocycle (12) leads to another two-fold enhancement in activity to ∼2 µM for E. coli, while B. subtilis remains at the 500 nM level of inhibition. We note that 12 was the most active sodium transporter studied in liposomes.30 Yeast are resistant to the action of 12 but they are susceptible to 11 (8.4 µM). Some of the difference in activity may result from differences in the environments in which the organisms were grown. S. cerevisiae are grown in YPD media that has pH = 6.5, compared to LB media (bacteria) that has pH = 7. Previous studies showed that higher acidity reduced the toxicity of 2, 6, and 12 to bacteria. Surprisingly, 11 was found to be more toxic at lower pH. The yeast are grown at a lower pH and, as noted for bacteria, 11 is the most active compound, while 2, 6, and 12, are less cytotoxic.
A more recent study expanded the range from 4–10; these compounds differ in spacer chain length from 8 to 20 methylene units. In this work, an ion selective electrode (ISE) method was used to monitor Na+ efflux in synthetic liposomes.15 The phospholipids used to prepare different vesicles were identical, including a single cis-unsaturation. The data obtained for 4–10 confirmed the trend previously observed by NMR for 4–8. In addition, hydraphile length was found to generally correlate with phospholipid fatty acid chain length, i.e., membrane thickness.15 This critical correspondence between hydraphile length and membrane thickness was explored in the biological context using E. coli, B. subtilis, and S. cerevisiae.
The benzyl channel compounds (4–10) that were studied had spacer chains of 8, 10, 12, 14, 16, 18, and 20 methylenes. Table 2 records the response of microorganisms to the presence of these compounds. The C8 benzyl channel (4) is the least active compound to the bacteria. This result mirrors those of previous studies conducted in phospholipid vesicles. A modest increase in toxicity to both E. coli and B. subtilis is observed for 5 (C10 benzyl channel), which is approximately 5 Å longer overall than is 4. Addition of two more methylenes in each spacer chain gives 6 (C12 benzyl channel) and engenders a sizable increase in activity. This further 5 Å of channel length affords a ∼9-fold increase in toxicity to E. coli and to B. subtilis. Maximal toxicity is reached for E. coli in the C14–C16 range and activity declines as length increases further (9, C18 and 10, C20 hydraphiles). These results correlate well with Na+ transport activity monitored in synthetic liposomes.34 Maximal toxicity to B. subtilis, 0.56 µM, is observed for the C14 and C16 channels (7, 8). This value of ∼0.6 µM compares with a toxic concentration for penicillin of ∼8 µM and is among the most potent examples of ionophore-mediated toxicity reported, natural or synthetic.
![]() | ||
Fig. 2 Kinetics of killing bioluminescent E. coli with 3, 6, and 7 at the indicated concentrations. Also shown are data for kanamycin. |
These data highlight the potency of hydraphiles. When tested at their MBCs, 6, 7, and 3 killed half of the E. coli population in 8.5, 9.1, and 12.5 minutes. This compares with a halftime of the known antibiotic kanamycin (MBC = 1.3 µM) of 44.8 minutes at 10–20 times the concentration of hydraphile. Although less active than the hydraphiles, kanamycin shows high selectivity for bacteria over mammalian cells, while the hydraphiles do not (see below).
![]() | ||
Fig. 3 Toxicity of hydraphiles 3, 6, and 7 to HEK 293 cells. |
This MTS assay was further used to assess the antitumor activity of compounds 4, 6, and 11 against CaCo2 cells. Compounds 6 and 11 showed LD50 values of ∼12 µM while the C8 benzyl hydraphile, 4, was inactive at concentrations as high as 20 µM (Fig. 4). These CaCo2 cancer cells are about 6-fold more resistant to 6 than are HEK cells. A possible explanation is that the apoptosis machinery normally present in non-cancer cells such as HEK 293 is suppressed.
![]() | ||
Fig. 4 Cytotoxicity of hydraphiles to CaCo2 cells. [Hydraphile] is micromolar. |
![]() | ||
Fig. 5 Cellular electrical response during a brief application of 13 to an HEK 293 cell (black bar) under whole cell patch clamp conditions. |
The left panel of Fig. 5 shows the membrane current of HEK 293 cells during rinse with saline solution, followed by a brief application of p-fluorobenzyl hydraphile 13. This compound is identical to 6 (illustrated) except that a fluorine atom is present in the para- or 4-position of the benzyl group. After application of 13, the cell is rinsed with saline solution. Compound 13 was tested because it had previously shown long open times during planar bilayer conductance experiments.38
The right panel of Fig. 5 shows the current–voltage (I–V) plot of the cell during its first rinse with saline solution (open circles), during treatment with 13 (black circles), and during extended rinsing (grey circles). Membrane conductance increases from about 1 nS to 7 nS upon treatment with 4-fluorobenzyl compound 13. The cell exhibited a quick return to homeostasis after hydraphile application was stopped. Even after prolonged rinsing, a residual conductance of 1.8 nS persisted, possibly indicating stable inclusion of the compound in the cellular bilayer. On average, application of the compound caused a 3.7 ± 1.2 fold (n = 4) increase in membrane conductance. Based on these results, the use of hydraphiles at subtoxic concentrations may hold utility in the recently posited pharmaceutical approach known as channel replacement therapy.39
The HEK 293 cells are bathed in Tyrode's solution, which contains 150 mM NaCl as the highest concentration salt. The cells are clamped at an inside potential of −20 mV with respect to the outside. Thus, positively charged Na+ ions pass into the cell through a hydraphile, while negatively charged particles depart. The concentrations of Na+ and Cl− are low within the cell but the K+ concentration is high. Glucuronate, a large organic anion, is also present. The latter presumably cannot pass through the hydraphile channel. We therefore reason that Na+ is the primary ion driving the sizable currents noted at negative holding potentials. When the cell is clamped at positive potentials, we observe sizable currents equal to those just noted. These currents may be carried by K+ leaving the cell, and/or by Cl− entering the cell.
N-4-Fluorobenzyl-4,13-diaza-18-crown-6. Diaza-18-crown-6 (3.55 g, 13.5 mmol), 4-fluorobenzyl bromide (2.20 g, 11.6 mmol), Na2CO3 (28.8 g, 272 mmol), KI (0.32 g, 1.9 mmol), and CH3CH2CH2CN (135 mL) were heated under reflux (72 h) in a 250 mL round bottomed flask. The reaction mixture was evaporated and the residue dissolved in CH2Cl2 (100 mL), washed with H2O (3 × 50 mL), dried (MgSO4), and evaporated. Chromatography over Al2O3 (2% 2-PrOH in CH2Cl2) gave the title compound (1.14 g, 41%) as a white solid (mp 83–84 °C). NMR: 2.8–2.9 (m, 8H); 3.6–3.7 (m, 19H); 6.95–7.05 (pseudo-T, 2H); 7.3–7.4 (pseudo-T, 2H).
7,16-Bis-{12-[16-(4-Fluorobenzyl)-1,4,10,13-tetraoxa-7,16-diazacyclooctadec-7-yl]-dodecyl}-1,4,10,13-tetraoxa-7,16-diazacyclooctadecane, 13. N-4-Fluorobenzyldiaza-18-crown-6 (0.8 g, 2.2 mmol), N,N′-di-(n-12-bromododecyl)-4,13-diaza-18-crown-6 (0.82 g, 1.08 mmol), Na2CO3 (2.3 g, 21.7 mmol), KI (14 mg), and CH3CH2CH2CN (14 mL) were heated at reflux for 25 h. The reaction mixture was evaporated to a thick oil (1.98 g). Column chromatography (Al2O3, 10–15% 2-propanol in hexanes) gave 13 (0.41 g, 28%) as a waxy solid. NMR: 1.2–1.6 (m, 40H), 2.5–2.6 (broad S, 8H), 2.7–3.0 (m, 24H), 3.5–3.7 (broad s, 52H), 7.0, 7.3–7.4 (m, m, 8 H). Calculated for C74H132F2N6O12, High resolution mass spectrometry, exact mass: 1334.99. Found, (M + 1)+m/z 1335.996.
Saccharomyces cerevisiae MIC/MFC experiments were conducted in similar fashion. Cells were grown in YPD media (10 g L−1 yeast extract, 20 g L−1 peptone, 20 g L−1 dextrose) at 30 °C. An inoculum of 5 × 103 cells was used for MIC studies, and allowed to grow for 48 hours to achieve turbidity.
After being freshly split, cells were counted on a hemacytometer and plated at a density of 20000 cells per well in a 96-well plate and grown for 24 hours. Ethanol stocks of each compound were diluted 1 : 100 into DMEM supplemented with 10%
ΔFBS and 1% Gln in triplicate. They were then serially diluted ½ into DMEM supplemented with 10%
ΔFBS, 1% Gln, and 1% ethanol. The original media was then removed from the cells and replaced with media containing the desired concentration of compound. As a positive control for growth, three wells containing cells were treated with DMEM supplemented with 10%
ΔFBS, 1% Gln, and 1% ethanol. As a negative control, three wells without cells were treated with DMEM supplemented with 10%
ΔFBS, 1% Gln, and 1% ethanol. The remaining empty wells on the plate were filled with 100 µL of phosphate buffered saline to minimize evaporation. After another 24 hours of culture, 20 µL of Cell Titer 96 Aqueous One (Promega) was added to each well and then developed in the tissue culture hood between 1 and 2 hours as per the manufacturer's protocol. The absorbance was measured at 490 nm and at 630 nm to correct for nonspecific absorbance. The data were plotted as OD490–OD630 and fit to a sigmoidal dose response curve using GraphPad 4 software.
To study the dose response of hydraphile toxicity, each hydraphile was dissolved in neat ethanol to a concentration of 10 mg mL−1 for 3 and 2.5 mg mL−1 for 6 and 7. The hydraphile ethanol stocks were then diluted 1:50 into LB-Amp (100 µg mL−1) in triplicate. As a vehicle control, ethanol was also diluted into LB-Amp (100 µg mL−1) in triplicate. The initial 1 : 50 dilution of hydraphile into LB was then further serially diluted 1 : 2 into LB-Amp (100 µg mL−1) and 2% ethanol (to maintain a constant concentration of vehicle). Finally, the procedure was repeated for kanamycin (20 mg mL−1 dissolved milli-Q water) except that the vehicle control was a 1 : 50 dilution of milli-Q water into LB-Amp (100 µg mL−1), and the media for the serial dilution was LB-Amp (100 µg mL−1) 2% milli-Q water.
The diluted compound (50 µL) was then added using a multichannel pipette to the appropriate wells containing the transformed E. coli. The result was that each well contained: 100 µL of lux operon transformed E. coli at OD = 0.05 in LB-Amp (100 µg mL−1), 1% ethanol, and the specified concentration of either hydraphile or Kanamycin.
The plates were then imaged for photon emission from the transformed E. coli by using a commercial IVIS 100 (Xenogen) CCD camera and computer system (settings: no filter, binning 8, f-stop = 1, exposure time = 30 s at the indicated time points). For the first 30 min, the plates remained in the IVIS at 37 °C, and then were transferred to an incubator in which the bacteria were grown at 37 °C with gentle shaking (∼210 rpm) between time points.
Data were analyzed by normalizing to the pre-drug photon flux from each well and then expressed as a percent of the vehicle treated wells at any given time point [eqn. (1)]. The standard error of the mean was propagated through eqn (1).
![]() | (1) |
This journal is © The Royal Society of Chemistry 2005 |