Exocytosis proceeds through prefusion stages such as hemifusion, but hemifusion is still an elusive intermediate of unknown duration. Using darkfield and fluorescence microscopy in ATII (alveolar type II) cells containing large secretory vesicles (LBs; lamellar bodies), we show that exocytotic fusion events were accompanied by a mostly biphasic SLID (scattered light intensity decrease) originating from the vesicle border. Correlation with the diffusional behaviour of fluorescence markers for either content or membrane mixing revealed that the onset of the fast second phase of SLID corresponded to fusion pore formation, which was followed by vesicle swelling. In contrast, a slow first phase of SLID preceded pore formation considerably but could still be accompanied by diffusion of farnesylated DsRed, an inner plasma membrane leaflet marker, or Nile Red. We conclude that hemifusion is an exocytotic intermediate that may last for several seconds. SLID is a new, non-invasive approach by which a prefusion phase, including hemifusion, can be continuously recorded and distinguished from fusion pore formation and postfusion vesicle swelling.
Regulated exocytosis denotes stimulated fusion between the limiting membrane of a storage vesicle and the plasma membrane, catalysed by SNARE (soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor) proteins under the influence of cytoplasmic factors such as Ca2+ [1–6]. The fusion pore is the initial aqueous connection between the vesicle lumen and the extracellular space . The initial fusion pore exhibits a channel-like behaviour that may flicker between closed and open states, allowing partial release of stored materials [8–11]. The diffusion of hydrophilic compounds through the pore can be preceded by the diffusion of amphiphilic and outer plasma membrane probes, suggesting that the initial lipidic pore is too narrow to enable passage of solutes [11,12]. Models of proteinaceous exocytotic pores have also been proposed .
In contrast to fusion pore formation, exocytotic pre- and hemi-fusion states are still much more refractory to experimental detection. Knowledge thereof is largely based on theoretical considerations, suggesting the ‘stalk model’ as a relatively low energy cost intermediate that initiates hemifusion, through which fusion occurs (‘fusion-through-hemifusion’, reviewed in detail in [14,15]). Hemifusion is defined as the fusion between two opposing membrane leaflets, enabling diffusion of lipids and probably lipid-anchored proteins from one membrane to the other. Several studies of content versus lipid mixing convincingly demonstrated the existence of hemifusion in viral and other models of fusion [16–18]. In pure phospholipid bilayers, hemifusion was also postulated from electrophysiological evidence . However, hemifusion has not yet been demonstrated as an intermediate stage of exocytosis in a physiological setting, i.e. in an intact, secreting cell. A FRAP study on sea urchin eggs concluded that docked vesicles may be kept stably in a hemifused state by immobile SNARE proteins, but a transition between these stages, i.e. fusion-through-hemifusion, was not directly demonstrated . Likewise, an electron tomography study on synaptic vesicles suggested docked vesicles to be in a hemifused state . Docking and other processes, such as priming, that make a vesicle fusion-competent, are considered to occur prior to hemifusion and may represent situations where a vesicle is spatially very close to the plasma membrane despite the lack of a lipid connection between both membranes.
Strictly speaking, hemifusion is a morphological description, denoting a certain arrangement of membrane leaflets. However, due to methodological impediments to prove this condition, recent studies have used a more functional description, meaning the exchange of lipid-anchored components between opposing membranes [16,20,22]. Although this functional definition is certainly less accurate in a structural sense, it probably reflects the physiological implications of hemifusion (e.g. exchange of signalling molecules) better than the morphological one. Therefore, we use the term ‘hemifusion’ here in its functional sense, bearing in mind the limitations to predict a clear structural state.
Major impediments to the study of exocytotic hemifusion are the extremely small diffusion distances that represent a methodological challenge to detect small time lapses between content and membrane mixing. In the present study, we overcome these impediments by studying a cell model [the ATII (alveolar type II) pneumocyte] which is particularly suitable for the following reasons. (1) LBs (lamellar bodies) are large secretory vesicles of more than 1 µm diameter. (2) LBs contain a surface active complex of lipids and proteins (surfactant), which can be selectively stained by FM 1-43, a membrane-impermeable amphiphilic dye [23,24]. Thereby, content mixing, i.e. the mixing of the extracellular fluid with vesicle contents, elicits a bright localized signal. This readout enables a high temporal resolution of fusion pore formation, as revealed by combined patch clamp (capacitance) and fluorescence experiments . (3) In a previous report we showed that single LB fusion events were followed by actin coating of the fused LB, which was found to be a prerequisite for release . Since the fusion pores in these cells open slowly  and remain a mechanical barrier for release , the shape of a fused LB is maintained for several seconds until the actin coat contracts, facilitating analysis of post-fusion membrane mixing. In addition, surfactant is essentially insoluble in water, remaining in a compact form within the fused vesicle before active extrusion; and (4) LB exocytosis is slow, sequential, and can be stimulated, facilitating analysis of single fusion events .
In the present study, we show the first dynamic resolution of exocytotic prefusion stages, including hemifusion, and introduce darkfield microscopy as a direct and highly sensitive optical approach to distinguish between pre- and post-fusion stages of exocytosis.
Cell isolation and treatment
ATII cells were isolated from male Sprague-Dawley rats (150–200 g) as described . Animal experiments were approved by the Regierungspräsidium Tübingen, application number 833. In short, rats were anaesthetized with ketamin (10%) and xylazil (2%). Lungs were perfused, removed from the body, washed and incubated two times with elastase and trypsin at 37°C for 15 min. Lungs were immersed in DNase-containing solution and minced with scissors into portions of approx. 1 mm size. Enzyme reaction was stopped by incubation with FCS (fetal calf serum) and the tissue was filtered 3 times through gauze and nylon meshes (mesh width: 150, 20 and 10 μm). The final filtrate was centrifuged for 8 min at 130 g, the cells were suspended in DMEM (Dulbecco's modified Eagle's medium), and put on IgG-coated plastic dishes for 15 min at 37°C. Non-adherent cells were centrifuged for 8 min at 130 g, suspended in DMEM with 10% FCS and 1% penicillin/streptomycin (10000 units/ml of penicillin, 10 mg/ml of streptomycin) and seeded on to glass coverslips. ATII cells were cultured at 37°C, 5% CO2 and 95% humidity, and used for experiments 1–2 days after the isolation. In LAMP3–GFP (LAMP, lysosome-associated membrane protein; GFP, green fluorescent protein) and DsRed experiments, cells were transfected with adenovirus constructs after isolation. The composition of the experimental solution used for live cell imaging was (in mmol/l): 140 NaCl, 5 KCl, 1 MgCl2, 2 CaCl2, 5 glucose, 10 Hepes, pH 7.4. LB secretion was stimulated with ATP (100 μmol/l). LB fusions were detected with FM 1-43 (300 nmol/l) added to the bath, as described earlier . In the Nile Red experiments, the cells were stained with 200 nmol/l Nile Red before the experiment, in DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate) experiments the cells were incubated with 1 μmol/l DiI for 10 min, and in LTR (LysoTracker Red) experiments the cells were incubated with 100 nmol/l LTR for 10 min. All dyes were purchased from Molecular Probes. The experiments were performed at room temperature (21–25°C).
Farnesylated DsRed and LAMP3–GFP adenovirus vectors
LAMP3–GFP adenovirus vector, cloning and cell transfection were recently described in . Vectors pDsRed-Monomer-F and pEGFP1-N1 were obtained from Clontech. Vectors pDONR221 and pAd/CMV/V5-DEST were purchased from Invitrogen.
cDNA encoding DsRed-Farn was amplified from the vector pDsRed-Monomer-F using the primers DsRed-FarnF2 (ggggacaagtttgtacaaaaaagcaggcttcGTCAGATCCGCTAGCGCTACCGG) and DsRed-Farn-R2 (ggggaccactttgtacaagaaagctgggtcAGCTTGAGCTCGAGATCTGGATCC) using standard PCR methods. After isolation of PCR products, amplified cDNA was introduced into pDONR221 by in vitro recombination using BP-clonase II enzyme mix (Invitrogen) and cloned into Esherichia coli OmniMax™2-T1R (Invitrogen). Thereafter, DsRed-Farn cDNA was transferred into adenoviral expression vector pAd/CMV/V5-DEST by in vitro recombination using LR-clonase II enzyme mix (Invitrogen) and recombination products were cloned into E. coli OmniMax™2-T1R (Invitrogen).
The MCS (multi-cloning site) and coding region of EGFP (enhanced GFP) were amplified from plasmid pEGFP-N1 using primers MCSgfp-F1 (GGGGACAAGTTTGTACAAAAAAGCAGGCTATGCTAGCGCTACCGGACTCAG) and MCSgfp-R1 (GGGGACCACTTTGTACAAGAAAGCTGGGTATTTACTTGTACAGCTCGTCC) by standard PCR techniques. PCR products were introduced into pDONR221 by in vitro recombination using BP-clonase II enzyme mix and cloned into E. coli OmniMax™2-T1R.
Recombinant adenoviral expression vectors were isolated from E. coli cells using the Qiagen plasmid midi kit (Qiagen). Plasmids were linearized using restriction endo-nucleases PacI (New England Biolabs). Linearized plasmid was transfected into 293A cells (Invitrogen) using Lipofectamine™ 2000 (Invitrogen) without further purification directly after heat inactivation of PacI. Transfected cells were cultivated for 5–8 days and virus particles were isolated using ViraBind Adenovirus purification kit (Cell Biolabs) according to the manufacturers' protocol. Virus particles were eluted in elution buffer (Cell Biolabs), and stored in elution buffer containing 10% glycerol in aliquots at −80°C until use.
Imaging with the RTM 2.5 microscope
Darkfield and fluorescence imaging were performed on an upright Richardson RTM 2.5 microscope (Improvision). For darkfield illumination, we used the red or the blue filter in the condenser; therefore the light-reflecting structures appeared red or blue on a dark background. Darkfield and fluorescence light was detected with a 3CCD colour video camera (ExwaveHAD DXC-C33, Sony). Images were acquired with the 63× water immersion objective (Leica) and the RTM 3.0 blue filter cube (490 nm dichroic, 400–500 nm excitation filter, 510 LP emission filter) for FM 1-43, LAMP3–GFP and CTG (CellTracker Green; l,5-chloromethylfluorescein diacetate) imaging, and with the RTM 3.0 green filter cube (560 nm dichroic, 540–550 nm excitation filter, 570 LP emission filter) for farnesylated DsRed, DiI, LTR and Nile Red imaging. Volocity software (Improvision) was used for image acquisition and subsequent measurements. At the image acquisition rate of 5–10/s, the darkfield and fluorescence signals were acquired simultaneously and were subsequently digitally separated from an RGB image based on their different colours. In combined darkfield and FM 1-43 fluorescence experiments, we used the red channel for darkfield and green channel for FM 1-43 fluorescence signals, and in combined darkfield and farnesylated DsRed, DiI and Nile Red experiments, we used the blue channel for darkfield and red channel for fluorescence signals.
Imaging with the fluorescence microscope
The combined LTG (LysoTracker Green) and Nile Red measurements were performed with a 100× Neofluar oil, numerical aperture 1.3 on a Zeiss Axiovert equipped for epifluorescence (Polychrome II, TILL Photonics), emission beam splitting (DV-CC, Optical Insights LLC) and synchronous image acquisition (Sensicam, PCO) at a rate of 1/s. Wavelength and filters used were: 485 nm excitaton, 520 nm dichroic 1, 565 nm dichroic 2, 525 nm BP (LTG) and 630 nm LP (Nile Red). Cells were pre-incubated 30 min with 200 nm LTG, Nile Red was added at 4 μmol/l 5 min prior to ATP (50 μmol/l).
Image and statistical analysis
A square region of interest was defined around the fusing LB, on which darkfield light and fluorescence intensity were measured. In darkfield and fluorescence measurements, the lowest value was subtracted and the change in light intensity was expressed in arbitrary units. Unless otherwise stated, data are presented as means±S.D. Vesicle size was measured as the surface of the vesicle intersection in one focal plane. Vesicle volume change after fusion was calculated from the measured vesicle intersection in one focal plane, and median and interquartile ranges were calculated. Tau values were obtained by fitting the curve with a single exponential in Igor Pro 5.04B. Excel and SigmaPlot 10.0 were used for statistics and graphs. Some images were corrected for brightness and contrast with Adobe Photoshop CS2.
Measurement of single LB fusion with the plasma membrane by darkfield microscopy and FM 1-43 fluorescence was recently described in detail . It exhibits the following key feature: the darkfield light intensity originating from the border of a fusing LB decreases [SLID (scattered light intensity decrease)], whereas vesicle contents appear dark (Figure 1). When FM 1-43 is present in the bath, this is accompanied by an increase in FFM1-43 (FM 1-43 fluorescence intensity) within the fused LB, starting at the site of the fusion pore (Figure 1). FFM1-43 originating from surfactant exceeds outer plasma membrane staining so much that the plasma membrane appears virtually dark (Figure 1).
ATII cell imaged with darkfield illumination (left) and FM 1-43 fluorescence (right)
To investigate further the obvious correlation between SLID and the fusion pore opening, we used higher image acquisition rates (10/s). These measurements revealed that SLID consists of two clearly discernible stages (Figure 2), a slow component at the beginning (mean duration 2.24±2.19 s, n=22) and a subsequent fast one (mean duration 0.92±0.49 s). The onsets of the two phases of SLID were measured as shown in Figure 2. The linear parts of the darkfield data were fitted and the crossing points of the linear regressions were considered as onsets of SLID phases 1 and 2. The onset of the FFM1-43 increase (indicating fusion pore opening ) corresponded to the start of the fast component of SLID (Figure 3). The highest observed discrepancy between the two events was 0.5 s. Therefore, it is reasonable to assume that the light intensity decrease gets faster when the fusion pore opens (either transiently or permanently). This assumption leads to the question of what the initial slow component of SLID might reflect. Despite its variable duration, it was detectable in the majority (84.6%) of the observed fusion events. Theoretically, several explanations for the biphasic nature of SLID are possible, which will be discussed below. The most likely one is that the first phase of SLID reports the hemifusion stage (‘membrane mixing’), where the proximal leaflets of the plasma and vesicle membranes merge but the fusion pore is not yet open. During this stage, a non-membrane-permeable dye such as FM 1-43 would not be able to enter the LB. The second phase of SLID corresponds to an open fusion pore and ‘content mixing’.
Determination of the times of onset of SLID phases 1 and 2
The time courses of darkfield light intensity change and
To test further the hypotheses that hemifusion intermediates of considerable duration indeed exist in type II cells, we sought several independent experimental approaches, applying markers for either content or membrane mixing, as illustrated in Figure 4(A); DiI was used as a marker of the outer plasma membrane leaflet . In contrast with FM 1-43, it was absent in the bath solution. Therefore, LB staining was exclusively due to lateral membrane diffusion, but not to bulk fluid entry. Conversely, LTG or LTR were used as fluid phase content markers. They are trapped in LBs by non-ionic diffusion and protonation, and rapidly discharged into the extracellular space upon the formation of an aqueous pore . The onset of LTG discharge coincides (<1 s) with uptake of FM 1-43  and also with the dissipation of the proton gradient, probably the earliest diffusional event within the lumen of the pore (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/424/bj4240007add.htm). To probe for diffusion of components between the leaflets of the inner plasma membrane and outer LB membrane (hemifusion), we used two different dyes: the first being farnesylated DsRed, which selectively incorporates into the inner leaflet of the plasma membrane by post-translational farnesylation of the protein through membrane-bound enzymes . Under the assumption that the hemifusion stalk does not contain diffusion barriers for this dye, it should diffuse from the inner leaflet of the plasma membrane to the outer leaflet of the LB membrane during hemifusion. The second dye was the unspecific lipid marker Nile Red, which is trapped within phospholipids , resulting in stable staining of LBs (Figures 4 and 6). In the course of hemifusion, the LB should lose Nile Red fluorescence intensity, possibly due to dye redistribution along the plasma membrane.
SLID and its relation with fluorescence of membrane/content markers
Figure 4(A) exhibits the fluorescence stainings of individual LBs before and after both phases of SLID, i.e. the initial pre- and final post-fusion stages. It illustrates that all tested dyes behaved according to the theoretical predictions outlined above, with certain limitations regarding DsRed, which exhibited a wide scatter in its diffusional behaviour (discussed below; Figures 5A and 5B). In the absence of SLID, the fluorescence intensities of these markers were stable or subject to variable extents of photobleaching.
Summary of SLID and fluorescence markers
A direct comparison of the earliest onsets of fluorescence changes between these dyes is shown in Figure 4(B): DiI and LTR behaved like FM 1-43, exhibiting fluorescence change during phase 2 of SLID exclusively. The onset of DsRed and Nile Red fluorescence change, however, could start as early as the beginning of phase 1 of SLID (Figure 4B). Hence, the earliest observed onsets of fluorescence changes by markers of lipid versus content mixing coincided almost precisely with the onsets of phase 1 and phase 2 of SLID respectively.
Assuming the onset of the second, fast, phase of SLID as a marker for the instance of fusion pore formation, Figure 5(A) exhibits the temporal relation (negative values represent prefusion events) between all fluorescence and SLID phase 1 changes with reference to this event (=zero time). It is again apparent that the diffusion of content/outer leaflet markers (LTR, DiI and FM 1-43) followed the fusion pore formation. In contrast, Nile Red and DsRed fluorescence changes could start several seconds before fusion. There was, on average, a delay between the onset of SLID phase 1 and particularly that of DsRed, which could extend into the postfusion phase (Figure 5A). Hence, analysis of hemifusion by specific membrane diffusion probes is considerably limited by highly variable diffusion times and/or limitations of detection, as will be discussed below.
To analyse further a possible relation between SLID phase 1 and membrane merger (hemifusion), we plotted individual fluorescence changes as relative delays following the onset of SLID phase 1, shown in Figure 5(B). The data reveal that, although there was a clear tendency of Nile Red and DsRed to lag behind the onset of SLID phase 1, some events closely coincided (Figure 4B). Importantly, however, a fluorescence change was never observed before SLID phase 1. Hence, hemifusion cannot precede SLID phase 1.
To obtain a direct comparison between lipid and content mixing within a single vesicle, we used a combination of Nile Red and LTG. As indicated above, both dyes accumulate within LBs (Figure 6A). The spectral emission properties of both dyes allowed their synchronous recording with a beam-splitting device in the emission path. Figure 6(B) demonstrates that the decline in Nile Red intensity precedes that of LTG (1.36±0.8 s, n=7). Also here, the most reasonable explanation is a situation where vesicle and plasma membrane are physically conjunct, seconds before a water-permeable pore is formed.
Comparison between content and membrane mixing
The concept of hemifusion versus fusion was further examined by analysing vesicle swelling (Figure 7A). Postfusion swelling is a common finding, driven by fluid uptake through the fusion pore due to the physico-chemical properties of the vesicle matrix [33,34]. Postfusion swelling can therefore be taken as another parameter of ‘content mixing’, independent of content marker diffusion. Fusion pore formation (measured as the start of FM 1-43 staining) was followed by swelling of LBs (Figure 7B), measured as the surface change of the scattered light signal. Prefusion swelling, as reported in another cell type , was never observed here. The median increase of the LB volume estimated from darkfield images was 17.5% (interquartile range 5.8–24.3%, n=50). Postfusion LB swelling occurred within seconds and was therefore several-fold faster than the initiation of release [24,26], (time constant of postfusion swelling: τ=2.1±1.4 s, n=6). To exclude the possibility that the diameter increase of the scattered light signal was just the result of a change in light diffraction properties, we transfected the ATII cells with LAMP3–GFP fusion protein. LAMP3 is localized at the membrane of lysosome-associated organelles, and in ATII cells the LAMP3–GFP fluorescence and the darkfield signal overlap . As in the darkfield illumination, the swelling was also detectable by measuring the LAMP3–GFP signal (Figure 7C; median increase in volume 30.9%, interquartile range 18.5–38.4%, n=5). Additional evidence for LB swelling is presented in Supplementary Figure S2 (at http://www.BiochemJ.org/bj/424/bj4240007add.htm).
Fusion pore formation was followed by LB swelling
These experiments indicate that content markers or water can enter the LB during the second phase of SLID exclusively, which makes this phase a sensitive postfusion indicator. The DsRed and Nile Red experiments indicate the existence of hemifusion intermediates that may last for seconds. SLID phase 1 is a newly described prefusion indicator that may represent the hemifusion state even more reliably and precisely than any diffusion marker, as discussed below.
Two opposing theories describe initial exocytotic fusion pores as being either proteinaceous (direct fusion pathway), or lipidic, emerging through transient hemifusion states (stalk model) . Our study provides the first experimental evidence in favour of the ‘fusion-through-hemifusion’ model in a living, intact cell. It demonstrates prefusion and membrane merger seconds before content mixing in single secretory vesicles. Therefore it confirms predictions from a multitude of non-exocytotic or artificial fusion systems, reviewed in detail elsewhere [14,15,36]. It also supports evidence in favour of ‘stable’ hemifusion states of ‘docked’ vesicles [20,21].
Decades of intensive research, including breath-taking advances in molecular and live cell imaging techniques, have so far failed to develop a method to continuously record exocytotic hemifusion states in living cells. SLID (phase 1) is a new approach in this direction. We are aware that at present, SLID is no more than a new phenomenon, but it yields a continuous recording of two phases, of which phase 1 can clearly be assigned to a prefusion stage. Its validation as a method that accurately reflects the morphological postulate of the hemifusion state will only be possible when hemifusion becomes accessible to independent measurement. To this end, electron microscopy would be the method of choice, but this is as yet far beyond the temporal resolution and technical feasibility of the method, providing static information of the process only.
SLID (phase 1) is a prefusion indicator that coincides fairly well, but not precisely, with diffusional membrane merger (Figure 5). In addition to hemifusion, it could reflect any other conceivable pre-fusion condition, such as protein rearrangement and development of a tighter contact between the vesicle and the plasma membrane. Hence the key question remains: what does SLID actually report?
Darkfield microscopy has been extensively used to investigate the fusion and transformational properties of giant liposomes [37,38]. These studies suggest that the light scattering intensity from a liposomal membrane reflects its thickness, and that the changes in scattered light intensity of liposomal membranes observed with the darkfield microscope might also be based on the co-operative movement of lipid molecules in a membrane [37,38]. We also made the observation that extruded surfactant, in contrast with surfactant within LBs, generated a darkfield signal, probably because the highly organized lipid layers in LBs  became disordered. The origin of the darkfield signal from the limiting membrane of an LB is more difficult to interpret: the membrane of a secretory vesicle is not merely an even surface of free-floating phospholipids, but is probably endowed with a highly complex three-dimensional topography due to abundant membrane-associated proteins . Although we do not know the exact composition, distribution and density of these proteins in LBs , they may affect the optical properties of the incident light. In any case, it can be assumed that SLID reports a change of LB surface properties rather than of vesicle content, since the first phase of SLID precedes measurable content mixing and LB swelling by seconds.
We have the following reasons to favour the hypothesis that SLID phase 1 indeed reports the hemifusion state. Lipid flow explains the biphasic nature of SLID much better than other prefusion processes, since the two phases of SLID would correspond with the two leaflets of the membrane, merging in sequence. In this case, the first phase of SLID would reflect a change of lipid composition in the outer leaflet of the limiting LB membrane. Although there is no direct way to test this hypothesis, it is consistent with other observations. First, darkfield microscopy gives the best high-contrast images of giant liposomes compared with other methods, indicating that darkfield images do not require the presence of membrane proteins . Furthermore, detergents such as Triton X, which disrupt lipid domains, sequentially cause SLID in all LBs . The lipid composition of a membrane is fundamental in determining its curvature and tension . In this way, specific lipids not only determine the readiness of a secretory vesicle to hemifuse and fuse, but also create a force for lipids to flow when membrane leaflets merge . Monck et al.  demonstrated lipid flow from the plasma to the vesicle membrane during exocytosis, resulting in up to 40% loss of plasma membrane area. It is likely that the inner plasma membrane leaflet and outer LB membrane leaflet have different lipid compositions , which would be sufficient to explain substantial lipid flow during hemifusion. Secondly, the delay between the onset of SLID phase 1 and the change of DsRed or Nile Red fluorescence intensity does not exclude that membrane merger is the common underlying mechanism: free lipid flow through the hemifusion stalk may be limited [15,36]. This could explain the observation that phase 1 of SLID does not occur abruptly (as a step), but rather linearly and slowly. If true, the slope of phase 1 would actually reflect diffusional limitations. Accordingly, hemifusion can present a barrier for the diffusion of proteins with the size of farnesylated DsRed . In addition, we have to consider photobleaching (rapid decrease of DsRed fluorescence before and after fusion, Figure 4B) which can also explain the low detectability of a small DsRed increase before the fusion pore opening. Thirdly, the existence of fusion events, where diffusion of farnesylated DsRed into the vesicle almost precisely coincided with the onset of SLID phase 1, indicates that membrane merger and SLID phase 1 can occur simultaneously. Hence, SLID phase 1 does not require any other or additional explanation than hemifusion alone. Finally, in theory, both phases of SLID could occur after fusion pore formation, but the initial pore is so small that no ‘content mixing’ (entry of extracellular fluid) would take place. This possibility, however, is very unlikely for the following reasons. First, we showed in simultaneous patch-clamp and FM 1-43 measurements that the fusion pore opening (measured as capacitance steps) and the increase in FFM1-43 are apart by no more than 1 s , which is considerably shorter than the duration of the first phase of SLID (2.24 s on average, 9 s maximum). Secondly, even if such ‘narrow pores’ existed (small enough to escape detection by capacitance measurements), FM 1-43 should be able to enter the LB along the wall of the fusion pore, by lateral diffusion from the outer leaflet of the plasma membrane .
Our experiments revealed that there is no vesicle swelling prior to fusion pore formation. Hence, the prefusion phase was not accompanied by volume uptake. It does not appear, therefore, that vesicle swelling in ATII cells is a driving force for fusion pore formation [33,35].
Independent of the prefusion marker (SLID phase 1, Nile Red or DsRed), the hemifusion could last longer than 4 s (Figure 5A). Hence, considering the large variability in the duration of individual hemifusion states (not detectable to 9 s), these different methods yielded fairly consistent results. They suggest that hemifusion lifetimes cannot be solely explained on the basis of a different vesicle size. Rather, it appears that hemifusion does not proceed in a stereotypic way, but that specific lipid or protein domains in opposing membranes determine the transition to fusion. This prompts the question of whether hemifusion is an actively regulated process and which compounds are involved therein. It also prompts the question of whether transient hemifusion and LB retrieval exist in addition to kiss-and-run events . In any case, the duration of the hemifusion state should influence the amount of membrane mixing. This may induce translocation of plasma membrane signalling molecules to secretory vesicles, triggering signalling cascades that induce– among others – actin assembly and influence the future mobility of the fused vesicle [26,46,47]. In the lung, these processes may determine polarized secretion, compound exocytosis and membrane recycling .
The potential of darkfield microscopy will be to disclose the regulatory principles of hemifusion and possibly other prefusion events that are inaccessible by fluorescence probes, which are all subject to limitations caused by diffusion properties, photobleaching, staining specificity or spectral features.
Technical assistance by Melanie Timmler, Tatiana Felder, Gerlinde Siber and Irina Öttl is gratefully acknowledged.
alveolar type II
Dulbecco's modified Eagle's medium
fetal calf serum
FM 1-43 fluorescence intensity
green fluorescent protein
lysosome-associated membrane protein
scattered light intensity decrease
soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor
Pika Miklavc and Paul Dietl designed the study. Oliver H. Wittekindt and Susanne Albrecht provided DsRed and LAMP3–GFP adenovirus vectors. Pika Miklavc performed the darkfield experiments, Susanne Albrecht performed experiments with LAMP3–GFP, and Thomas Haller and Peter Schullian performed the experiments with the beam-splitting device. Paul Dietl, Thomas Haller and Pika Miklavc wrote the manuscript.
This work was supported by the Fonds zur Förderung der Wissenschaftlichen Forschung [grant numbers P15742, P15743 and P20472], the Deutsche Forschungsgemeinschaft [grant number D1402], and the 6th framework of the European Union, Pulmo-Net.