Exchange of Gramicidin between Lipid Bilayers: Implications for the Mechanism of Channel Formation  Kevin Lum, Helgi I. Ingólfsson, Roger E. Koeppe, Olaf S. Andersen  Biophysical Journal  Volume 113, Issue 8, Pages 1757-1767 (October 2017) DOI: 10.1016/j.bpj.2017.08.049 Copyright © 2017 Biophysical Society Terms and Conditions

Figure 1 Two different proposals for gA channel formation. (A) Given here is the canonical model in which gA channels form by reversible transmembrane dimerization of two nonconducting subunits, one from each bilayer leaflet (as emphasized by the subunit coloring). (B) Given here is the model of Jones et al. (12), in which gA channels exist as dimers that can switch between nonconducting (gray) and conducting (colored) bilayer-spanning states. (In model B, the gA subunits do not dissociate, or dissociate very slowly, once two monomers join together.) To see this figure in color, go online. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 2 The four vesicle types used in this study. Types A and B are not loaded with ANTS, and therefore are invisible in the fluorescence quench experiments; types C and D are loaded with ANTS, and therefore are visible in the fluorescence quench experiments. Types A and C lack gA in the membrane; types B and D have gA incorporated into the membrane. The gramicidin-free type-A LUVs serve as sinks for the removal of gA from the ANTS-loaded gA-containing type-D LUVs. The gramicidin-containing type-B LUVs serve as donors for transfer of channels to the gA-free type-C LUVs. To see this figure in color, go online. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 3 gA exchange between DC22:1PC LUVs, as monitored by changes in the time course of Tl+-induced fluorescence quenching. Four different LUV mixtures were tested; each LUV type or mixture was incubated at 25°C for the indicated period, and the fluorescence quench rates were determined using stopped-flow spectrofluorometry. The quench rates were normalized to the rates determined for the standard (type-D) LUVs after 10 min incubation; the type-C and -D LUVs served as control for membrane integrity. To monitor the removal of gA subunits, type-D (+ANTS, +gA) LUVs were incubated with type-A (−ANTS, −gA) or type-B (−ANTS, +gA) LUVs at a 1:4 ratio of D to A or B; after 10 min, there is a twofold reduction in the quench rate for the D+4A mixture, with little further change (see also Fig. 4). (The mixture of types D and B served as positive control because any exchange of gA should not affect the average number of gA subunits in type-D.) To monitor the insertion of gA subunits (and formation of gA channels), type-C (+ANTS, −gA) LUVs were incubated with type-A (−ANTS, −gA) or -B (−ANTS, +gA) LUVs at a 4:1 ratio of types C to A or B; after 16 h, there was a 25% increase in the quench rate for the C+4B mixture (see also Fig. 4). (The mixture of types A and C served as negative control, as there was no gA that could partition into type-C.) To see this figure in color, go online. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 4 Summary of results on gA exchange between DC22:1PC vesicles. The fluorescence quench rates, determined in experiments like those in Fig. 3, were normalized to the rates determined for the standard vesicle (type-D) after 10 min incubation, where the average quench rate was 37 ± 10 s–1, mean ± SD (n = 5). For the experiments in the top row, the D+4A mixture monitors the removal of gA subunits (bilayer-spanning channels) from type-D LUVs. The D+4B mixture served as positive control, and type-D by itself serves as control for membrane integrity. For the experiments in the bottom row, the C+4B mixture monitors the insertion of bilayer-spanning gA channels into type-C LUVs, and the C+4A mixture served as a negative control. Mean ± SD (n = 3–5) except 40 h D+4B, which is mean ± range/2 (n = 2). To see this figure in color, go online. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 5 Summary of results on gA exchange between DC18:1PC vesicles. The fluorescence quench rates, determined in experiments similar to those in Fig. 3 but with DC18:1PC LUVs, were normalized to the rates determined for the standard vesicle (type-D) after 10 min incubation, where the average quench rate was 89 ± 15 s–1, mean ± SD (n = 8). For the experiments in the top row, the D+4A mixture monitors the removal of gA subunits (bilayer-spanning channels) from type-D LUVs. The D+4B mixture served as positive control, and type-D by itself serves as control for membrane integrity. For the experiments in the bottom row, the C+4B mixture monitors the insertion of bilayer-spanning gA channels into type-C LUVs, and the C+4A mixture served as a negative control. Mean ≥ SD (n = 3–6) except for the 2 h experiments, which is mean ± range/2 (n = 2). To see this figure in color, go online. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 6 Extraction of gA subunits from planar lipid bilayers using gA-free vesicles. Current traces were recorded from a planar DC18:1PC/n-decane bilayer that had been doped with 100 fM gA, added through the aqueous phase, which had been stirred for 60 s, and gA channel activity was monitored. Over the first 15 min (after gA addition), channel activity increased to reach an apparent equilibrium, where the average number of conducting channels was 2.4. The aqueous phases then were stirred for one min, which caused some membrane instability (periods of membrane instability are denoted by the gray box; they were not included in the analysis) but no further increase in channel activity. After 25 min, 40 μL 1.0 M CsCl was added to both aqueous phases, as control for the later addition of lipid vesicles, and the solutions were stirred for 60 s (indicated by solid box), which led to minimal change in channel activity; the average number of conducting channels after 5 min was 2.2. Then 40 μL of gA-free DC22:1PC lipid vesicles (type-A) were added to both sides of the bilayer, to a final lipid concentration of 80 μM in the aqueous phases, which were stirred for 60 s (again indicated by solid box), and channel activity was monitored over the next 70 min (only the first 30 min was used for analysis). The current trace fragments below the trace show segments at higher time resolution, sampled at the locations indicated by (∗) below the time axis for the main trace. 1.0 M CsCl, 100 mV. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 7 gA single-channel activity in planar DC18:1PC/n-decane bilayers before and after addition of gA-free DC22:1PC vesicles. The average number of conducting channels over consecutive observation periods was determined from single-channel current traces, like Fig. 6. Each data point denotes the average number of conducting gA channels observed over 3 min, normalized to the maximum number (nmax) observed in the particular current trace (10–15 min after gA addition). The three stages in the experiments were: Control, after gA addition but before any manipulation; Salt Control, where first the aqueous phases were stirred for 60 s to test whether the system had reached apparent equilibrium, followed by the addition of 40 μL electrolyte to the aqueous phase on each side of the bilayer, then stirred for 60 s and channel activity was determined; and Vesicle Addition, where gA-free DC22:1PC vesicles were added to both sides of the bilayer to a final lipid concentration (in the aqueous phases) of 80 μM and the solutions were stirred for 60 s and channel activity was monitored. 1.0 M CsCl, 100 mV. Mean ± SD (n = 3). Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 8 Insertion of gA subunits into planar lipid bilayers from gA-containing DC22:1PC vesicles. In the first half of the trace, lasting 27 min, DC22:1PC vesicles (80 μM lipid, 40 nM gA) were added only to the cis chamber, the aqueous phases were stirred for 60 s (indicated by a solid box), and channel activity was recorded. There is only one identifiable channel event (marked by (∗) below the time axis for the main trace) after 24 min of recording, which is shown at higher resolution in the first of the three currents in the bottom of the figure. Then the same amount of gA-containing DC22:1PC vesicles were added to the trans chamber, the solutions were stirred for 60 s (again indicated by a solid box), and gA channel activity was observed. The second and third current trace fragments below the trace show segments at higher time resolution, sampled at the locations indicated by (∗) below the time axis for the main trace. 1.0 M CsCl, 100 mV. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions

Figure 9 Evaluating models of gA channel formation using Poisson and binomial distribution fits to the data. The three panels denote the results from three different experiments where we were able to observe single-channel activity for >50 min (the current trace in Fig. 6 is from Experiment 2). For each experiment, the average number of conducting channels (〈n〉) and the maximal number of conducting channels (nmax) were determined from the current traces (e.g., Fig. 6), and the distribution of the number of conducting channels (0, 1, 2, … nmax) in each 3-min test period (the different conditions are defined in the legend to Fig. 7) was compared to predictions based on the Poisson and binomial distributions. For the binomial fits, the number of gA dimers (conducting and nonconducting) was estimated as the maximal number of observed channels, nmax and as 1.5⋅nmax (∗) or 2⋅nmax (∗∗), to account for a possible underestimate of nmax. To see this figure in color, go online. Biophysical Journal 2017 113, 1757-1767DOI: (10.1016/j.bpj.2017.08.049) Copyright © 2017 Biophysical Society Terms and Conditions