Molecular understanding of sterically controlled compound release through an engineered channel protein (FhuA)

Background Recently we reported a nanocontainer based reduction triggered release system through an engineered transmembrane channel (FhuA Δ1-160; Onaca et al., 2008). Compound fluxes within the FhuA Δ1-160 channel protein are controlled sterically through labeled lysine residues (label: 3-(2-pyridyldithio)propionic-acid-N-hydroxysuccinimide-ester). Quantifying the sterical contribution of each labeled lysine would open up an opportunity for designing compound specific drug release systems. Results In total, 12 FhuA Δ1-160 variants were generated to gain insights on sterically controlled compound fluxes: Subset A) six FhuA Δ1-160 variants in which one of the six lysines in the interior of FhuA Δ1-160 was substituted to alanine and Subset B) six FhuA Δ1-160 variants in which only one lysine inside the barrel was not changed to alanine. Translocation efficiencies were quantified with the colorimetric TMB (3,3',5,5'-tetramethylbenzidine) detection system employing horseradish peroxidase (HRP). Investigation of the six subset A variants identified position K556A as sterically important. The K556A substitution increases TMB diffusion from 15 to 97 [nM]/s and reaches nearly the TMB diffusion value of the unlabeled FhuA Δ1-160 (102 [nM]/s). The prominent role of position K556 is confirmed by the corresponding subset B variant which contains only the K556 lysine in the interior of the barrel. Pyridyl labeling of K556 reduces TMB translocation to 16 [nM]/s reaching nearly background levels in liposomes (13 [nM]/s). A first B-factor analysis based on MD simulations confirmed that position K556 is the least fluctuating lysine among the six in the channel interior of FhuA Δ1-160 and therefore well suited for controlling compound fluxes through steric hindrance. Conclusions A FhuA Δ1-160 based reduction triggered release system has been shown to control the compound flux by the presence of only one inner channel sterical hindrance based on 3-(2-pyridyldithio)propionic-acid labeling (amino acid position K556). As a consequence, the release kinetic can be modulated by introducing an opportune number of hindrances. The FhuA Δ1-160 channel embedded in liposomes can be advanced to a universal and compound independent release system which allows a size selective compound release through rationally re-engineered channels.


Introduction
A channel protein that is embedded in an impermeable membrane offers the possibility to develop novel triggered drug release systems with potential applications in synthetic biology (pathway engineering), and medicine (drug release). So far only FhuA [1], OmpF [2][3][4], Tsx [5] and MscL [6] have been reconstituted functionally into synthetic block copolymers or lipid membranes.
FhuA is a large monomeric transmembrane protein of 714 amino acids located in the E. coli outer membrane folded into 22 anti-parallel β-strands and two domains [7]. By removing the "cork" domain (deletion of amino acids 5-160 [8,9]) the resulting deletion variant behaves as a large passive diffusion channel (FhuA Δ1-160) [1]. FhuA and engineered variants have a significantly wider channel than OmpF (elliptical cross section of OmpF is 7*11 Å [10] whereas FhuA is 39*46 Å [1]) allowing the translocation of even single stranded DNA [11]. Recently we reported an exclusive translocation of calcein through an engineered transmembrane FhuA Δ1-160 which had been embedded in a tri-block copolymer membrane PMOXA-PDMS-PMOXA; where PMOXA = poly(2methyl-2-oxazoline) and PDMS = poly(dimethyl siloxane); and could be opened up through a reduction triggered system [12]. The reported calcein release kinetics were strongly modulated by the size of employed lysinelabeling reagents [12]. Twenty nine lysines are present in the FhuA Δ1-160; 19 lysines located on the protein surface, 6 are inside the channel, and 4 are at the barrel rim [12]. The 19 lysines on the FhuA surface point into the outer membrane and are after purification covered by oPOE rendering pyridyl-labeling unlikely.
An average of four lysine residues per FhuA Δ1-160 was determined to be pyridyl labeled [12]. Based on the hypothesis that the 6 lysine inside the channel might mainly be responsible for restricting compound fluxes, two subsets of FhuA Δ1-160 variants were generated. In the six subset (A) variants only one of the six lysines in channel interior was substituted by alanine and in the six subset (B) variants only one lysine remained in the channel interior whereas all other five were substituted to alanine. For the in total 12 investigated FhuA Δ1-160 variants a HRP based colorimetric TMB (3,3',5,5'-tetramethylbenzidine) detection system [13,14] was employed for quantifying the sterical hindrance of pyridyl-labeled lysines on the TMB substrate. The colorimetric HRP/ TMB detection was preferred over the previously reported calcein detection system due to a higher reproducibility [1,12]. Furthermore liposomes instead of a polymeric nanocontainer system were selected for characterizing the 12 FhuA Δ1-160 variants due to more simple and rapid assay procedures [15], despite drawbacks like leakiness, stability over time [16] and undesired biomolecule adsorption on the surface [17].
However, the better kinetic results reproducibility using liposomes compared to polymersomes, where the FhuA Δ1-160 insertion can be affected by block co-polymer poly-dispersity and traces of residual chemicals, suggested us to use liposomes correcting the kinetic results by the small leakage contribution (see Table 1). To our best knowledge we report a first detailed mutational study on a transmembrane channel protein to gain, on the molecular level, first insights on the sterically controlled diffusion of TMB through the FhuA channel interior modulated by labeled lysines. Interestingly only one single lysine position is the main responsible of the TMB diffusion.

Results
FhuA Δ1-160 based compound release system Figure 1 shows a FhuA Δ1-160 based compound release system where FhuA Δ1-160 is embedded in a lipid membrane (left) together with the colorimetric HRP/TMB reporter used for quantifying TMB translocation (right).

HRP based colorimetric TMB detection system
TMB as chromogen has been developed and widely used in enzyme immunoassays (EIA) employing horseradish peroxidase [13,14]. Besides, the colorimetric HRP/TMB detection system proved to be more reproducible than the previously employed calcein assay which generates a fluorescence signal upon release of self-quenching calcein from liposomes into the surrounding solution.
The HRP/TMB detection system is based on a two step consecutive oxidative reactions ABC (A = TMB; B and C = first and second TMB oxidation products, see Figure 1) catalyzed by HRP in presence of hydrogen peroxide. Each single step is a pseudo-second order rate reaction with a reported second order rate constant (myeloperoxidases) [14] of: k AB = 3.6*10 6 M -1 s -1 and k BC = 9.4*10 5 M -1 s -1 . The final TMB oxidation product C is unstable out of very acidic conditions [13] and the intermediate based on the first oxidation product B is used as reaction reporter, explaining the absorbance drop in time (see Additional file 1). The total amount of encapsulated HRP was not detectable though using the Soret absorption band. However kinetic data reproducibility was confirmed basing on a three data set for each measurement. Figure 2 shows the six lysine residues in the FhuA Δ1-160 inner channel which upon labeling might be responsible to modulate sterically the diffusion through the channel protein. In total 12 FhuA Δ1-160 variants were generated to identify the lysine(s) which might limit TMB flux through FhuA Δ1-160 inner channel. Two subsets of six FhuA Δ1-160 variants were generated. Subset A) contains FhuA Δ1-160 variants in which one of the six lysines in the interior of FhuA Δ1-160 was substituted to alanine; subset B) contains six FhuA Δ1-160 variants in which only one lysine was not changed to alanine. Table 1 summarizes for these two subsets the TMB conversions. TMB conversions were determined by diffusion limited translocation through the FhuA Δ1-160 [12] inner channel (Additional file 1: Figure S1 and S2) using a previously reported colorimetric HRP/TMB detection system [1,13].

FhuA Δ1-160 lysine positions and diffusion limited TMB translocation
HRP has been entrapped in the liposome harboring FhuA Δ1-160 variants by using film hydration method coupled with extrusion. In this method, the lipid amphiphile is brought in contact with the aqueous medium containing HRP and FhuA Δ1-160 in its dry state and is subsequently hydrated to yield vesicles. After homogenization and purification of the resultant liposomes, the TMB conversion was initiated by supplementing TMB (10 μl) to the aqueous solution. Background conversions of TMB due to liposome instabilities or translocation through the membrane in absence of FhuA Lambert Beer law was used with an extinction coefficient of 3.9 × 10 -4 M -1 cm -1 for the first TMB oxidation product. Two subsets (A & B) of FhuA Δ1-160 variants were apart from controls analyzed. FhuA Δ1-160 variants of subset A contain a single lysine to alanine substitution while subset B contain five lysine to alanine substitutions. All FhuA Δ1-160 variants are pyridyl-labeled except two controls (liposome lacking FhuA Δ1-160 and the unlabeled FhuA Δ1-160). "*": The true TMB conversion is calculated from the TMB-conversion of FhuA Δ1-160 variant subtracted by the TMB conversion of the background lacking FhuA Δ1-160; "**": TMB conversion ratio represents a ratio between TMB conversions of pyridyllabeled FhuA Δ1-160 variants and the liposome control lacking FhuA Δ1-160.
Δ1-160 were determined to be 13 [nM]/s ( Differences in the absolute values between the two experimental data sets can likely be attributed to pyridyl labeling efficiencies, i.e. inner channel sites have a lower probability to get labeled once lysines on the protein ores are labeled or when multiple sites are labeled (Additional file 1: Figure S3). Further experimental details on the TMB conversion calculations and controls (Additional file: Figure S1 and S2), CD-spectral measurements on secondary structure stability of FhuA Δ1-160 (Additional file 1: Figure S3), size measurements of liposomes (Additional file 1: Figure S4 and S5) and simulation details, can be found in the Additional file 1.

Molecular Dynamics simulations to investigate the key modulating position 556
A working hypothesis for controlling the compound flux in the inner FhuA Δ1-160 precisely is a defined and rigid conformation of the blocking lysine residue. Lysine fluctuations of all six FhuA Δ1-160 have been directly correlated to the B factors deduced from Molecular Dynamics MD trajectories in a first simulation (see Additional file 1). The B factor analysis indicates the dynamic mobility of an atom or group of atoms. The concept is derived from the X-ray scattering/crystallography theory, alternatively known as "temperature-factor" or "Debye-Waller factor" [18]. Table 1 suggests that the amino acid position 556 is a key residue in modulating the compound flux through the inner channel. Interestingly a general trend between low B factors values of the unlabeled FhuA Δ1-160 (Figure 3) and translocation importance (experimental results; Table 1), has been found. Again the most important position 556 is in the B factor analysis the least mobile one. In detail, FhuA Δ1-160 is a β-barrel with a cross-section of 39 Å and 46 Å on the "top" part and a reduced crosssection on the "lower" exit of the barrel, 29 Å and 19 Å. K556 is placed in a rigid β-barrel at the "lower" cross-section ( Figure 2) which originally interacts with the ferrichrome peptide and TonB protein [19] for further iron translocation. Other lysines are located on the opposite site K167, K537 and K586 (placed near the top cross-section) or underneath (K344; lower cross-section). As indicated by B values the positions K556, K167 and K537 have the higher blocking effects which are in accordance to the experimental results in corresponding subset A and B variants though no simulations on the pyridyllabeled starting variant FhuA Δ1-160, has been performed (see Additional file 1 for further discussion).
In summary, experimental results and first computational simulations indicate that the rigidity of the labeled positions play an important role in generating FhuA Δ1-160 channels with a defined and "non-fluctuating" pore size. Fluctuations in pore sizes of FhuA Δ1-160 will reduce the discriminating power to control compound fluxes and are therefore an important prerequisite for a universal compound release system that can rapidly be re-engineered to match the compound size. Following up computational simulations are required to investigate in detail the roles of the pyridyl-label, to investigate cooper-ative effects of labeled lysine residues and taking labeling efficiency and perturbations of protein structure after labeling into account. Further FhuA Δ1-160 engineering efforts will be based on subset B) variant K556 to further advance the control of compound fluxes through the FhuA Δ1-160 channel, especially for low molecular weight compounds.

Conclusions
Molecular understanding of the sterically controlled diffusion in FhuA Δ1-160's inner channel is an important prerequisite to develop a universal compound release system that can rapidly be re-engineered for a "time and dose-dependent" compound release.
Six lysine residues were systematically analyzed in two subsets of engineered FhuA Δ1-160 channels. Analysis of 12 variants identified position K556 as a key substitution to sterically control compound fluxes through the inner channel of FhuA Δ1-160 embedded in liposome membrane. A first B-factor analysis based on MD simulations identified position K556 as the least fluctuating lysine among the six investigated lysines suggesting a correlation between flexibility and steric control of TMB compound translocation through the inner FhuA Δ1-160 channel. The subset B variant K556 of FhuA Δ1-160 represents therefore an excellent starting point to under- stand channel dynamics and to sterically control compound flux through engineered FhuA Δ1-160. Based on these results it seems promising that the reduction triggered release system can be advanced to a universal and compound independent release system which allows a size selective compound release through rationally reengineered FhuA Δ1-160 channels.

Methods
All chemicals used were of analytical reagent grade or higher quality, purchased from Sigma-Aldrich Chemie (Taufkirchen, Germany) and Applichem (Darmstadt, Germany) if not stated otherwise. A thermal cycler (Mastercycler gradient; Eppendorf, Hamburg, Germany) and thin-wall PCR tubes (Mμlti-ultra tubes; 0,2 ml; Carl Roth, Karlsruhe, Germany) were used in all PCRs.

Site-directed mutagenesis
Six lysines located in the FhuA Δ1-160 channel were substituted by alanine using QuikChange (developed by Stratagene; La Jolla, CA, USA) [20] derived SDM protocol generating two subsets (A & B; Table 1) of FhuA Δ1-160 variants. Table 2 lists the primers employed for SDM.
The SDM was performed by using a two-stage PCR protocol [21]: first stage: one cycle (95°C, 1 min), three cycles (95°C, 30 s; 55°C, 1 min; 68°C, 2 min) and a second stage: one cycle (95°C, 1 min), 15 cycles (95°C, 30 s; 55°C, 1 min; 68°C, 2 min) and one cycle (68°C, 25 min). In each reaction were employed template FhuA Δ1-160 (25 ng), a primer set (see Table 2; 200 nM each), dNTP mix (200 μM) and Pfu DNA polymerase (1 U) in Pfu reaction buffer (2 × 25 μl total volume, for stage one). In stage one for each primer the extension reaction is performed in a separate PCR tube and subsequently pooled for the stage 2 PCR. In Stage 2 additional Pfu DNA polymerase (0.02 U) is supplemented before starting the second PCR. For digestion of parental DNA, DpnI (10 U; 1 h, 37°C) is supplemented to the PCR mix. All 12 FhuA Δ1-160 variants were fully sequenced to assure lysine to alanine substitutions and lack of additional mutations. Amount of DNA after PCR was quantified using a NanoDrop photometer (NanoDrop Technologies, Waltham, Massachusetts, USA).