Article Information

• PDF (543 kb)
• Full Text with Large Figures
• Supporting Material
• Cited by in Scopus (7)

PubMed

• Articles by Geert van den Bogaart
• Articles by Bert Poolman

Related Articles

• The differential actions of cortisol on the accumulation of ...

  The differential actions of cortisol on the accumulation of α-lactalbumin and casein in midpregnant mouse mammary gland in culture Cell, Volume 19, Issue 2, 1 February 1980, Pages 473-480 Masayoshi Ono and Takami Oka Summary The dose-response relationship between cortisol and the accumulation of the two milk proteins, casein and α-lactalbumin, was studied in organ culture of mammary gland from midpregnant mice. The accumulation of casein was low in culture with insulin but was enhanced by the further addition of prolactin. Further increases in casein were effected by the addition of cortisol in increasing concentrations up to 3 × 10 M, which was optimal for the accumulation of this protein. The content of α-lactalbumin in explants was similarly low in culture with insulin alone, but, in contrast, was increased to a maximal level by the addition of insulin and prolactin. The addition of cortisol up to 3 × 10 M with insulin and prolactin did not further increase the level of α-lactalbumin; in fact, at concentrations above 3 × 10 M the steroid caused progressive inhibition of the accumulation of this protein in cultured explants. Studies of the appearance of casein and α-lactalbumin in incubation medium during organ culture revealed the presence of substantial amounts of these milk proteins. During the first 2 days of culture with insulin, prolactin and 3 × 10 M cortisol, the amount of α-lactalbumin in culture medium was almost equal to the level found in tissue, whereas in the presence of 3 × 10 M cortisol, or in the absence of exogenous steroid, over 70% of total α-lactalbumin was retained in tissue. The observed difference in the amount of α-lactalbumin in culture medium can, however, only partially account for the inhibitory effect of high doses of cortisol on the accumulation of α-lactalbumin in cultured mammary explants. In contrast to α-lactalbumin, the relative amount of casein in culture medium containing insulin and prolactin was smaller—19% of total casein synthesized—and was further reduced to 16% and 11% of the total in the presence of 3 × 10 M and 3 × 10 M cortisol, respectively. The above results indicate that cortisol exerts dose-dependent differential actions on the accumulation of casein and α-lactalbumin in mouse mammary epithelium in vitro. Summary | PDF (865 kb)

• High-Sensitivity Fluorescence Anisotropy Detection of Protei...

  High-Sensitivity Fluorescence Anisotropy Detection of Protein-Folding Events: Application to α-Lactalbumin Biophysical Journal, Volume 80, Issue 4, 1 April 2001, Pages 1996-2003 Denis Canet, Klaus Doering, Christopher M. Dobson and Yves Dupont Abstract An experimental procedure has been devised to record simultaneously fluorescence intensity and fluorescence anisotropy. A photoelastic modulator on the excitation beam enables the anisotropy signal to be recorded in one pass using a single photomultiplier tube and eliminates the need for a polarizer on the emission path. In conjunction with a stopped-flow mixer, providing a time-resolved capability, this procedure was used to study the refolding of -lactalbumin following dilution from guanidinium chloride. Although the fluorescence intensity does not change detectably, the fluorescence anisotropy was found to resolve the conformational changes occurring between the initial unfolded state and the molten globule state formed either kinetically during refolding at pH 7.0 or at equilibrium at pH 2.0 (A-state). This result provides further evidence that fluorescence anisotropy is a valuable probe of protein structural transitions and that the information it provides concerning the rotational mobility of a fluorophore can be complementary to the information about the local environment provided by fluorescence intensity. Abstract | Full Text | PDF (158 kb)

• Solution structure of a peptide fragment of human α–lactalbu...

  Solution structure of a peptide fragment of human α–lactalbumin in trifluorethanol: a model for local structure in the molten globule Structure, Volume 2, Issue 8, 1 August 1994, Pages 703-712 Lorna J Smith, Andrei T Alexandrescu, Maureen Pitkeathly and Christopher M Dobson Summary The peptide conformation differs from that previously determined for residues 101–110 in crystal structures of native -lactalbumin determined at both high and low pH, particularly in the relative orientations of the side chains. The series of turns seen in the peptide could, however, be related to the-helical structure seen for residues 104-111 in crystals at high pH, and may be important in the molten globule state for bringing the peptide chain into a compact conformation where favourable interactions between the side chains can occur. Summary | Full Text | PDF (950 kb)

• …more

Copyright © 2007 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 92, Issue 4, 1233-1240, 15 February 2007

doi:10.1529/biophysj.106.088708

Channels, Receptors, and Electrical Signaling

Dual-Color Fluorescence-Burst Analysis to Probe Protein Efflux through the Mechanosensitive Channel MscL

Geert van den Bogaart^*, Victor Krasnikov^† and Bert Poolman^*, ,  

^* Biochemistry Department, Groningen Biomolecular Science and Biotechnology Institute & Materials Science Centreplus, University of Groningen, Groningen, The Netherlands
^† Ultrafast Laser and Spectroscopy Laboratory, Groningen Biomolecular Science and Biotechnology Institute & Materials Science Centreplus, University of Groningen, Groningen, The Netherlands

Address reprint requests to Bert Poolman, Tel.: 31-50-363-4190; Fax: 31-50-363-4165.

Mechanosensitive channels (MS) are found in bacteria, fungi, plants, and animals and play a vital role in cell physiology 1. Upon a dilution of the external osmolyte concentration, bacterial cells eject cytoplasmic molecules, including ions, metabolites 2, and possibly proteins 3. For instance, Escherichia coli releases a subset of proteins, up to 10% of total protein content, when cells are dispersed in water 4. At least part of the cytoplasmic molecules released has been attributed to membrane channel proteins gated by mechanical forces 5. For a basic understanding of protein release by microorganisms and for biomedical applications, it is important to establish whether channel proteins of large conductance can facilitate the passage of macromolecules. For instance, liposomes containing mechanosensitive channels have been proposed as drug delivery vehicles for the controlled release of therapeutic peptides and proteins 6. In the case of the MS channel of large conductance (MscL) from E. coli, the protein or membrane environment has been engineered to allow the channel to open and close in response to light or pH 7,8.

MscL is the biophysically and biochemically best characterized MS channel protein and has a large conductance of 2.5–4nS 1. On the basis of the crystal structure of the homologous MscL protein from Mycobacterium tuberculosis9 and a protein fusion study 10, MscL is predicted to form a pentamer. From molecular modeling 11,12, patch-clamp studies 11, and in vivo and in vitro flux measurements 5,13,14,15, the diameter of the pore of MscL has been estimated to be between 30 and 40Å. This diameter should be sufficient for the release of small proteins, but it is not known how large a molecule can actually pass through MscL 4,5,8,11,14,15.

Here, we used dual-color fluorescence-burst analysis, to monitor efflux of a broad range of (macro)molecules through a membrane pore. This analysis is based upon the fluorescent DiO labeling of MscL-containing liposomes and encapsulating macromolecules labeled with a second fluorophore Alexa fluor 633 (Fig. 1). The fluorescent lipid analog DiO and Alexa fluor 633 have nonoverlapping excitation and emission spectra and can be detected separately. For both labels, the fluorescence intensity fluctuations, resolving from the liposome diffusing through the observation volume of a confocal microscope, are correlated with each other. To activate MscL, a cysteine was engineered at the constriction site of the pore (G22C) for labeling with thiol-specific reagents. MscL (G22C) is pushed open when a charge is introduced at the Cys-22 site 16. Labeling of MscL G22C with MTSET introduces five (positively charged) choline moieties at the constriction site of the pore 8, whereas MTSES introduces (negatively charged) sulfonaethyl moieties 16.

Display large version of this figure

Figure 1

Principle of the dual-color fluorescence-burst analysis. (a) The focal geometry of a two-channel confocal microscope and diffusion of DiO-labeled liposomes (blue), containing Alexa fluor 633-labeled macromolecules (red). Note that the diffusion of the two fluorophores coincides. (b) Like panel a but after opening of MscL and release of the macromolecules. Note that the diffusion no longer coincides. (c) Fluorescence count traces resulting from the diffusion of DiO-labeled liposomes (blue trace, I_DiO) with Alexa fluor 633-labeled glutathione encapsulated (lower trace, I_AF633). Between t₁ and t₂ the intensity of the DiO signal is above the offset.

A set of peptides and stably folded proteins were chosen as substrates for efflux through MscL (Table 1). The number of DiO molecules per liposome is independent of the state (open or closed) of MscL, whereas the number of Alexa fluor 633-labeled macromolecules does depend on this state. With MscL in the open state, the labeled molecules, depending on their size, might be released. When DiO-labeled liposomes, with encapsulated Alexa fluor 633-labeled molecules, diffuse through the observation volume of the confocal microscope, the DiO and the Alexa fluor 633 signals correlate (Figure 1a). If the macromolecules are small enough to efflux through MscL, opening of MscL G22C by chemical modification with MTSET or MTSES 16 will reduce this correlation (Figure 1b). Molecules smaller than bovine pancreas trypsin inhibitor ((BPTI) molecular weight of 6511) effluxed through MscL, whereas histidine-containing protein ((HPr) molecular weight of 9119) and larger molecules did not. These observations were confirmed using fluorescence correlation spectroscopy (FCS). Furthermore, we show that only a fraction of the MscL channel molecules participated in the protein transport across the membrane.

MscL G22C 16 was expressed in the E. coli mscL-knockout strain PB104 17 using vector pB10b 18. Cells were grown to midexponential growth phase (optical density at 660nm of 0.5) and expression of mscL was induced by adding 1mM isopropyl-β-d-thiogalactopyranoside (IPTG). Membrane vesicles were then prepared by rupturing the cells with a high-pressure homogenizer (Kindler Maschinen AG, Zürich, Switzerland) and solubilized by 3% octyl-β-glucoside. MscL was purified using nickel affinity chromatography as described previously 19.

Purified MscL G22C was inserted into Triton X-100 destabilized liposomes as described 7,21. Briefly, 4mg ml⁻¹ liposomes were extruded 13× through 200-nm pore size polycarbonate filters (Avestin, Ottawa, Canada) and titrated with Triton X-100 until saturation. The lipid mixtures were composed of 1,2-dioleoyl-sn-glycero-3-phosphatidylcholine (DOPC), 1,2-dioleyl-sn-glycero-3-phosphatidylserine ((DOPS) Avanti Polar-Lipids, Alabaster, AL) and the fluorescent lipid analog 3,3′-dioctadecyloxacarbocyanine perchlorate ((DiO) Invitrogen, Carlsbad, CA; excitation and emission wavelengths of 484 and 499nm, respectively) in a 30,000:10,000:1mol ratio. MscL was added to the mixture in a protein to lipid ratio of ∼1:50,000 (mol pentameric MscL/mol lipid) as estimated by a Folin reagent-based DC-protein assay (Bio-Rad, Hercules, CA). Subsequently, the detergent was removed by incubating the mixture overnight at 4°C with stepwise addition of 100mg ml⁻¹ Bio-Bead SM-2 polystyrene beads (Bio-Rad). Membrane reconstitution was performed in 50mM potassium phosphate, pH 7.0.

To measure the upper size of molecules that can diffuse through MscL, several proteins were labeled with the fluorophore Alexa fluor 633 C5-maleimide (Invitrogen; excitation and emission of 621 and 639nm, respectively). The hormone insulin from bovine pancreas (Sigma, St. Louis, MO), trypsin inhibitor from bovine pancreas (BPTI, Sigma), HPr of the phosphotransferase system (PTS) from E. coli (purified as described in Reizer et al. 20), thioredoxin from Spirulina sp. (Sigma), and the calcium metalloprotein α-lactalbumin from bovine milk (Fluka, Milwaukee, WI) were labeled by adding 1ml 100μM protein solution in 30mM K-Hepes, pH 8.5, to 40 nmol Alexa fluor 633, and the mixture was incubated for 30min at room temperature. At this pH, maleimides not only react with sulfhydryl groups but also with primary amines such as the side chain of lysines and N-terminus of proteins. Reaction conditions (2.5-fold excess of protein over Alexa fluor 633; pH 8.5) were chosen to preferentially label the N-terminus. The buffer was exchanged to 50mM potassium phosphate, pH 7.0 by gel filtration, using a Sephadex G-25 column (Nap10 column, Amersham Biosciences, Buckinghamshire, UK). Glutathione and the peptide RPPGFSPFC (bradykinin R9C, Genscript, Piscataway, NJ) were labeled by adding 1ml 500μM solution in 50mM potassium phosphate, pH 7.0, to 40 nmol Alexa fluor 633 C5-maleimide and incubating for 30min at room temperature.

The Alexa fluor 633-labeled macromolecules were encapsulated into MscL-containing liposomes by at least three fast freezing (liquid nitrogen), and slow thawing (at 4°C) steps, at concentrations ranging from 1 to 10μM and sample volumes of 200μl. After this, 1M of the membrane permeable osmolyte glycerol was added to slow down diffusion. The liposomes were extruded through 200nm pore filters at a total lipid concentration of 2.5mg ml⁻¹. The liposomes were diluted to a final concentration of 1mg lipid ml⁻¹ into 50mM potassium phosphate, pH 7.0, plus 1M glycerol in the presence or absence of 1mM of either MTSET ([2-(trimethylammonium)ethyl] methanethiosulfonate bromide) or MTSES ([2-sulfonatoethyl] methanethiosulfonate sodium salt, Anatrace, Maumee, OH) for opening of the MscL channels and incubated for 30min at 18°C. FCS or fluorescence-burst analysis was then measured for 5–10min at 18°C in a 2-μl volume.

Fluorescence-burst analysis and FCS measurements were carried out on a laser scanning confocal microscope 19, based on an inverted microscope Axiovert S 100 TV (Zeiss, Jena, Germany), in combination with a galvanometer optical scanner (model 6860, Cambridge Technology, Watertown, MA) and a microscope objective nano-focusing device (P-721, PI). For excitation of the fluorescent lipid analog DiO, an argon ion laser (488nm, Spectra-Physics, Irvine, CA) and for excitation of Alexa fluor 633 a He-Ne laser (633nm, JDS Uniphase, Milpitas, CA) were focused by a Zeiss C-Apochromat infinity-corrected 1.2N.A. 63× water immersion objective. The laser power for both channels did not exceed 10μW at the back aperture. Emission was collected through the same objective, separated from the excitation beams by a beam pick-off plate (BSP20-A1, ThorLabs, Newton, NJ), and split into two channels by a dichroic beam splitter (585dcxr, Chroma Technology, Brattleboro, VT) and finally directed through emission filters (HQ 535/50 and HQ675/50, Chroma Technology) and pinholes (diameters of 30μm) onto two avalanche photodiodes (SPCM-AQR-14, EG&G, Gaithersberg, MD). The fluorescence signals were digitized and auto- and cross-correlation curves were calculated using a multiple τ-algorithm. The autocorrelation curves were fitted with a two-step analytical model (Supplementary Material ; 21). For the fluorescence-burst analysis, the fluorescence count traces for both colors were recorded for >300s with a binwidth of 1μs.

The setup was calibrated by measuring the known diffusion coefficients of Alexa fluor 488 and 633 in water (Invitrogen; D=300μm² s⁻¹). The lateral radii ω_xy, defined as the point where the fluorescence count rate per molecule decreased e² times, for 488 and 633nm excitation were 180 and 240nm, respectively, corresponding to detection volumes of ∼0.20 fl and ∼0.45 fl. For the dual-color fluorescence-burst analysis, the overlap between the two focal volumes was optimized by maximizing the cross-correlation signal of a double-labeled, double-stranded oligonucleotide. For this, the oligonucleotides 5′-ATT ATT GAG TGG TCA CTT TAA A-3′ and 5′-TTT AAA GTG ACC ACT CAA TAA T-3′, labeled with Alexa fluor 488 and 633, respectively (Invitrogen), were boiled together for 5min at a final concentration of 0.5μM and subsequently slowly cooled to 4°C to allow annealing. Afterwards, 1mM NaN₃ was added as a preservative.

Fluorescence-burst analysis was used to probe efflux through MscL. First, Alexa fluor 633-labeled macromolecules were encapsulated in MscL containing liposomes labeled with a second fluorophore DiO. Then, the fluorescence bursts emerging from diffusion of the liposomes through the focal volume of the dual-color confocal laser-scanning microscope was measured (Figure 1c). For all the bursts in the DiO fluorescence above a certain offset, the area of the i^th peak (L_i) was then calculated:

(1)

(2)

Because the fluorescence intensities are linearly proportional to the number of molecules, the concentration of macromolecules in arbitrary units, C_arb, could now be calculated:

(3)

In Eq. (3), the signals are normalized because DiO is associated to the surface of the liposomes and Alexa fluor 633 is present inside the vesicles; as a function of liposome radius, L_i and S_i thus scale to the power 2 and 3, respectively. From C_arb, the average concentration over all the liposomes (C_av) could be calculated:

(4)

To assess the diffusion of macromolecules through MscL, MscL G22C was reconstituted in liposomes that were labeled with the fluorescent lipid analog DiO. A series of Alexa fluor 633-labeled macromolecules was encapsulated in the liposomes. SDS-PAGE gels of Alexa fluor 633-labeled insulin, BPTI, HPr, thioredoxin, and α-lactalbumin, are shown in Fig. 2, where the proteins are visualized by ultraviolet (UV) illumination or Coomassie brilliant blue (CBB) staining. All the tested macromolecules were present in a monomeric form 22,23,24,25, which was confirmed by FCS and size-exclusion chromatography (not shown). The oligomeric state of insulin depends on its concentration and is monomeric at the concentration used here 26.

Display large version of this figure

Figure 2

The size markers. SDS-PAGE gels after UV-illumination (UV) or Coomassie brilliant blue staining (CBB) of insulin, BPTI, HPr, thioredoxin, and α-lactalbumin, all labeled with Alexa fluor 633. The three-dimensional structures of the macromolecules encapsulated in the liposomes were taken from the RCSB protein data bank. Shown are the NMR structure of human insulin 38, and the x-ray structures of BPTI 39, HPr from E. coli40, thioredoxin from E. coli41, and human α-lactalbumin 42. The scale bar represents 2nm.

For the dual-color fluorescence-burst analysis, the fluorescence intensity fluctuations resulting from the liposomes diffusing through the confocal volume was measured for 5–10min. During this time, 100–1000 liposomes passed through the focal volume, estimated from the fluorescence trace, where each burst corresponds to a liposome (Figure 1c). To verify whether the dual-color fluorescence-burst analysis was quantitative, Alexa fluor 633-labeled glutathione at 1, 5, and 10μM was encapsulated inside liposomes containing DiO. The fluorescence intensities were accumulated for 10min and the C_arb-values were calculated according to Eq. (3) (Figure 3a). The C_av values (Eq. (4)) were 1.2±0.3 for 1μM, 3.8±0.6 for 5μM, and 7.9±0.7 for 10μM (Figure 3b), which is in fair agreement with the expected ratio of 1:5:10. The percentages of empty liposomes were 53±12% for 1μM, 24±3% for 5μM and 5.5±1% for 10μM (Figure 3c). These values are close to the values of 76%, 26%, and 6.6%, respectively, calculated from the average size of the liposomes (Supplementary Material ), assuming a Poisson distribution of the number of molecules contained in the liposomes. Deviations between the measured and expected values are likely due to pipetting errors and uncertainties in the concentration of Alexa fluor 633-labeled glutathione, which were estimated from the absorption at 630nm. Spatial differences in the observation volumes of the channels could also have contributed to the observed deviations.

Display large version of this figure

Figure 3

Dual-color fluorescence-burst analysis: proof of principle. (a) Histogram shows the distribution of liposomes loaded with 0, 1, 5, or 10μM of glutathione as a function of C_arb (Eq. (3)). (b) The average concentration C_av (Eq. (4)) of the Alexa fluor 633-labeled molecules in the liposomes. (c) The percentages of empty liposomes. Error bars were calculated from three measurements.

The efflux from MscL-containing liposomes was studied by encapsulating Alexa fluor 633-labeled macromolecules; 4μM glutathione, 8μM insulin, 2μM BPTI, 2μM HPr, 2μM thioredoxin, or 8μM α-lactalbumin were encapsulated into the DiO-labeled liposomes. The mass of Alexa fluor 633 was determined to be 1089Da using mass spectrometry (Supplementary Material ). The macromolecules did not associate measurably with the membrane, as was verified by the addition of the Alexa fluor 633-labeled macromolecules to DiO containing liposomes and performing dual-color fluorescence-burst analysis (data not shown). The complete histograms of BPTI and thioredoxin are shown in Figure 4ab. Figure 4c shows the average concentration C_av of the tested compounds in the liposomes, and Figure 4d shows the percentages of empty liposomes. Upon opening of the MscL G22C channel by modification with MTSET or MTSES, the concentration of glutathione, insulin, and BPTI decreased significantly, whereas in the case of HPr, thioredoxin, and α-lactalbumin there was no significant change (Figure 4c). Furthermore, the fraction of empty liposomes increased for glutathione from 37.5±0.5% to 62±5%, for insulin from 11±2% to 36±2%, and for BPTI from 54.5±2% to 71.5±2%. In the case of insulin and BPTI, modification of MscL with the negatively charged MTSES resulted in an increase of empty liposomes comparable to MTSET. When HPr, thioredoxin, or α-lactalbumin were incorporated into liposomes, there was no increase in the fraction of empty liposomes upon MscL-modification with either MTSET or MTSES (Figure 4d). Addition of MTSET and MTSES to the labeled macromolecules in solution did not result in a change in fluorescence. Furthermore, MTSET and MTSES did not affect efflux of Alexa fluor 633-labeled glutathione or BPTI, encapsulated in liposomes without MscL, indicating that efflux occurred through MscL. Finally, steady-state levels of efflux were reached and the liposomes were not leaky due to, for instance, residual detergent, since the concentration distributions were stable for >24h.

Display large version of this figure

Figure 4

Dual-color fluorescence-burst analysis: protein passage through MscL. Histogram shows the distribution of MscL G22C containing liposomes loaded with (a) 2μM BPTI and (b) 2μM thioredoxin, unmodified (control) or modified with MTSET or MTSES, as a function of C_arb (Eq. (3)). (c) The average concentration C_av of the Alexa fluor 633-labeled molecules in the liposomes (Eq. (4)). (d) The percentages of empty liposomes. Error bars were calculated from three measurements.

To verify the results obtained by dual-color fluorescence-burst analysis, the diffusion of macromolecules was assessed using FCS. With FCS, one can determine diffusion constants and concentrations 21 for ensemble averages and not individual particles as with dual-color fluorescence-burst analysis. The fluorescence autocorrelation signals of the DiO and the Alexa fluor 633 channels were measured and the data were fitted with a two-step analytical model to quantify the release of macromolecules (Supplementary Material ). In the case of glutathione, the nonameric peptide bradykinin R9C, insulin, and BPTI, opening of the channel resulted in an increase of the fraction of fast component F, which corresponds to the released fluorophore (Fig. 5). No increase was observed for HPr, thioredoxin, and α-lactalbumin. These results are in perfect agreement with the results obtained by the dual-color fluorescence-burst analysis. The fluorescence-burst analysis, however, has the advantage over FCS of yielding a distribution of macromolecules over the liposomes, whereas FCS only provides qualitative information.

Display large version of this figure

Figure 5

FCS measurements. Fluorescence autocorrelation curves of MscL containing liposomes, before (black line/squares) and after (gray line/triangles) opening of the channel by modification with MTSET. Shown are typical autocorrelation measurements for liposomes containing (a) insulin and (b) thioredoxin, both labeled with Alexa fluor 633. Curves were fitted with a two-step analytical model (Supplementary Material ) and residuals are shown (Res). (c) The percentage of particles with a high mobility (F), corresponding to released substrates, of all tested macromolecules. Error bars were calculated from three measurements.

Both dual-color fluorescence-burst analysis and FCS revealed that the tripeptide glutathione, the nonameric peptide bradykinin R9C, insulin, and BPTI, all labeled with Alexa fluor 633, are able to efflux through MscL, whereas HPr, thioredoxin, and α-lactalbumin are not. The molecular weights and dimensions of these macromolecules, derived from the three-dimensional structures and presented by the Research Collaboration for Structural Bioinformatics Protein Data Bank ((RCSB PDB); Fig. 2), are given in Table 1. These dimensions do not include the bound Alexa fluor 633 label of 1089Da. As a result of this labeling, the dimensions of glutathione and bradykinin are dominated by the Alexa fluor 633 label. Because of the dimensions of insulin and BPTI (Table 1), the diameter of MscL must be larger than ∼3nm. This diameter holds for MscL opened by MTSES or MTSET, which bind a moiety of ∼5–6Å to each of the subunits of the channel 16 in an unknown conformation. Patch-clamp experiments showed that the conductances of MscL channels opened by modification with MTSET and opened by stretching of the membrane by a negative pipette pressure are comparable 16, suggesting a similar pore diameter. The pore diameter estimated from the passage of insulin and BPTI is consistent with calculated estimates of 29–46Å based on the conductance of the channel 11,12.

It has been suggested that some small organic osmolytes like ATP, glycine betaine, and trehalose are hindered in their diffusion through MscL 27. In view of the data presented here on protein efflux, and reports on fluorophore 6,8 and glycine betaine efflux via MscL 28, the earlier claims that cytoplasmic molecules are rejected from entering the pore 5,27 seem questionable. It is possible, however, that the rates at which different types of molecules diffuse through MscL vary greatly. Owing to the limited time resolution of the dual-color fluorescence-burst analysis, we could not obtain kinetic information on (macro)molecule diffusion through the pore. The reason is that liposomes diffuse very slow due to their relatively large size. The measured average diffusion speed of liposomes (γ/R, Eq. 5) is ∼60 times smaller than that of free Alexa 633 dye. To measure enough liposomes to obtain an acceptable autocorrelation curve, the measurement time should typically be >1min and information about rates of diffusion are not readily obtained. Thus, the diffusion through MscL of gluthathione, bradykinin, insulin, and BPTI (and possibly that of ATP; 27) may be slow compared to that of ions.

It has been proposed 27 that the S3 domain of the associated C-terminal helices of the MscL pentamer forms a prefilter for metabolites at the cytoplasmic entrance, excluding high-molecular weight substances from entering the pore. We clearly show that macromolecules up to at least 6.5 kDa can pass through MscL, but we cannot distinguish between the prefilter or the actual pore in determining the size exclusion limits of the channel. Efflux experiments with the Δ110–136 mutant, lacking this S3 domain 29, might elucidate the role of the S3 domain in determining the pore size. It has been claimed 5 that wild-type E. coli cells release thioredoxin upon an osmotic downshift, whereas thioredoxin is retained by mscL⁻ cells, suggesting that thioredoxin is able to diffuse through MscL. This observation contradicts our work. Our estimates should be regarded as lower limits, because proteins may be hindered in their passage due to binding to the surface of the pore or interaction with the methanethiosulfonate label. We consider the latter less probable because MscL modification with positively charged (MTSET) and negatively charged (MTSES) labels yielded identical results. Lastly, our pore diameter matches the value obtained by patch-clamp studies, where it was found that poly-L-lysines with a diameter >37Å blocked the conductance of the channel, whereas smaller poly-L-lysines had no effect 11.

The 52-kDa elongation factor Tu, the 41-kDa heatshock protein DnaK, and the 142-kDa enterobactin synthase polypeptide EntF have been reported to be released upon osmotic downshock 13,14,15, and these macromolecules have been hypothesized to pass through MscL. Due to the high molecular weight and large dimensions of these molecules, this seems rather unlikely. Moreover, a more recent study with a MscL knockout strain suggests that these molecules might be released via a mechanism independent of MscL 4. The genome of E. coli K12 contains 262 open reading frames coding for proteins <9.2kDa, of which 46 are predicted transmembrane proteins and 26 are secreted proteins. Of the remaining 190 open reading frames, >100 are proteins of unknown functions and for a large number it is not known whether they self-associate or complex with other proteins. Therefore, the number of proteins, potentially able to diffuse through MscL upon an osmotic downshift, will most likely be <100. Nevertheless, the release of this relatively small number of proteins could be physiologically significant.

Efflux through MscL was previously assessed by two techniques: patch-clamp and calcein dequenching 8,11,16. Both methods have a number of limitations compared to the fluorescence-based approach used in this work. In the case of the patch-clamp technique, direct translocation of proteins through membrane pores is not readily observed because of limited charges on the proteins relative to the free ions in the solution. Protein translocation, however, can be observed indirectly through blocking of the channel conductance 11,30,31. The calcein effux assay is based upon the self-quenching of fluorophore 8 at very high concentrations (typically 25–100mM), unlike the low concentrations used here (1–10μM). Because fluorophore-labeled proteins cannot be incorporated at concentrations needed for self-quenching, calcein dequenching is not feasible for monitoring effux of macromolecules. The fluorescence-based approach used in this work has further advantages of relatively short measurement times (typically 5min) and small volumes (∼1μl). In this study, the diffusion through MscL was studied both by dual-color fluorescence-burst analysis and FCS. The FCS results, however, provide no information on the distribution of the macromolecules over the liposomes.

Similar to the fluorescence-based techniques used in this study, the macromolecule efflux from liposomes could also be studied with size exclusion chromatography. However, such measurements do not yield information on the distribution of macromolecules over the liposomes. Furthermore, much more material (∼100×) is needed for size exclusion chromatography. In addition to fluorescence autocorrelation spectroscopy, efflux could be measured using dual-color fluorescence cross-correlation spectroscopy 32, since DiO and the Alexa fluor 633-labeled macromolecules have distinctive emission and excitation wavelengths. However, according to our experience, the cross-correlation curve is difficult to quantify and prone to substantial errors because a large fraction of the liposomes does not contain an active channel or labeled macromolecule. Thus, the change of the cross-correlation curve upon opening of the channels is small. Therefore, in this particular case, dual-color fluorescence cross-correlation spectroscopy is expected to be less accurate than fluorescence autocorrelation spectroscopy.

Using fluorescence-burst analysis, it was shown that after opening of the channels, a large fraction of the liposomes still contained the same concentration of fluorescent-labeled macromolecule as before opening, even in the case of glutathione, insulin, and BPTI, which were able to diffuse through MscL. This indicates that a large fraction (between 40 and 70%) of the liposomes did not contain a functional MscL channel. Based on the reconstitution ratio of MscL pentamer/lipid of 1:50,000 (mol/mol), it is expected that each liposome should contain ∼10 MscL channels. Because about half of the liposomes did not contain an active channel, ∼90% of the channel must have been lost during the reconstitution and/or not functionally incorporated into the liposomes. This agrees with the reconstitution efficiencies reported in the literature. For instance, for MscL pentamer/lipid ratios of from 1:50,000 to 1:10,000 (mol/mol) typically only one to seven active channels are found in a patch with a diameter of ∼1μm, instead of the expected 30–150 8,33,34. On the basis of thiol-labeling with membrane-impermeant reagents, the majority of “active” MscL seems to be oriented right-side-in, but always a fraction of inside-out protein is present (35; A. Kocer, Biomade Technology Foundation, personal communication, 2006). When these preparations were frozen-thawed up to eight times, we did not see a change in the efflux activity (or specificity) (G. van den Bogaart, unpublished data). Although these data are consistent with the notion that MscL is active in both orientations, it was technically not possible to rule out the possibility that protein efflux occurred via the fraction of inside-out oriented molecules.

In conclusion, this is the first time that efflux of macromolecules through MscL has been unequivocally demonstrated. Dual-color fluorescence-burst analysis provides an approach of analyzing the permeability properties of membrane proteins, with a number of merits compared to existing methods. It can be used as a universal method to study translocation of fluorophore-labeled macromolecules by membrane pores, expressed in cells or reconstituted in liposomes. The method should be also of particular importance for studying the effective pores of antimicrobial peptides, for which channel formation and membrane permeabilization is still elusive 36,37. Preliminary experiments have demonstrated the usefulness of fluorescence-burst analysis in determining the pore properties of magainin 2, melittin, and indolicidin (G. van den Bogaart and J. Mika, unpublished data).