Secretion without membrane fusion: Porocytosis

Abstract

We have recently proposed a mechanism to describe secretion, a fundamental process in all cells. That hypothesis, called porocytosis, embodies all available data and encompasses both forms of secretion, i.e., vesicular and constitutive. The current accepted view of exocytotic secretion involves the physical fusion of vesicle and plasma membranes; however, that hypothesized mechanism does not fit all available physiological data. Energetics of apposed lipid bilayers do not favor unfacilitated fusion. We consider that calcium ions (e.g., 10⁻⁴ to 10⁻³ M calcium in microdomains when elevated for 1 ms or less), whose mobility is restricted in space and time, establish salt bridges among adjacent lipid molecules. This establishes transient pores that span both the vesicle and plasma membrane lipid bilayers; the diameter of this transient pore would be ∼ 1 nm (the diameter of a single lipid molecule). The lifetime of the transient pore is completely dependent on the duration of sufficient calcium ion levels. This places the porocytosis hypothesis for secretion squarely in the realm of the physical and physical chemical interactions of calcium and phospholipids and places mass action as the driving force for release of secretory material. The porocytosis hypothesis that we propose satisfies all of the observations and provides a framework to integrate our combined knowledge of vesicular and constitutive secretion. Anat Rec (Part B: New Anat) 282B:18–37, 2005. © 2005 Wiley-Liss, Inc.

INTRODUCTION

Secretion is a fundamental process common to all cells. Two general modes of secretion have been identified: vesicular and constitutive. Vesicular secretion encompasses release of cellular products contained within membrane-bound vesicles to the extracellular space. Fusion of the vesicle with the plasma membrane, with requisite involvement of an elevation in calcium levels near or at the vesicle-plasma membrane juncture, is widely considered a feature common to vesicular secretion (Llinás et al., 1992). Constitutive secretion encompasses release of cellular products contained free within the cytosol that are not otherwise packaged within the lumen of vesicles or other membranous organelles. In this process, cell product is passed through the confining phospholipid bilayer from the cytosol to the extracellular space (Dan and Poo, 1994; Peaker and Wilde, 1996; Atlas, 2001; Cavalli et al., 2001). While suggestions are made that leader sequences are important for threading cell products via facilitated diffusion across the bilayer membrane (Blobel and Dobberstein, 1975; Walter et al., 1984), the mechanism by which constitutive secretion is accomplished has yet to be identified. Later in this article, we will proffer a hypothesis that encompasses a means by which a cell can accomplish regulated vesicular and constitutive secretion.

Recently, we hypothesized an alternative mechanism, named porocytosis, to describe secretion from vesicular compartments without fusion of the vesicular and plasma membranes (Kriebel et al., 2000, 2001; Silver et al., 2001, 2003). The dimensions of the transient pore would be about 1 nm in diameter, or twice the radius of a lipid molecule, the change in spacing expected for the freezing of lipid with calcium in a bilayer. In porocytosis, cell products contained in a membrane-bound vesicle lumen pass to the extracellular space through a calcium-dependent transient passageway or pore that spans both the lipid bilayers of the juxtaposed vesicle and plasma membrane. The porocytosis mechanism is consistent with all observations in synapses, including the so-called quantal vesicular release, with some morphological evidence (Kriebel, et al., 2001; Silver et al., 2001, 2003). While our earlier reports focused on the porocytosis hypothesis as applied to release of neurotransmitter from the presynaptic terminal of the neuromuscular junction (NMJ) (Kriebel et al., 2000, 2001; Silver et al., 2001, 2003), the hypothesis is applicable to other forms of vesicular secretion, which accounts for the often observed secretions in which only part of the vesicular contents (at presynaptic endings) are released, or by granular inclusions such as chromaffin cells, pancreatic islet cells, mast cells, basophils, and eosinophils, and a wide variety of cells central to cancer and inflammatory diseases (Sagen and Pappas, 1987; Pappas and Kriho, 1988; Dvorak, 1991, 2000; Plattner et al., 1997).

BACKGROUND CRITIQUE ON VESICULAR AND CONSTITUTIVE SECRETION

Much of our knowledge of secretion has been obtained from morphological studies of isolated monosynaptic preparations of peripheral cholinergic synapses such as the neuromuscular junction and from other isolated or cultured neurons. Most studies of neurotransmitter secretion from the presynaptic terminal describe secreted transmitter in quantal packets, and the process is thus within the framework of the quantal-vesicular exocytotic (QVE) hypothesis. This quantal hypothesis is based on the tenet that quanta are unitary, independent of stimulation frequency, always the same, and that packet size does not change (Katz and Miledi, 1979). The so-called quantum is thus thought to represent a discrete amount of neurotransmitter that can just be recorded postsynaptically, and that is the entire amount of neurotransmitter contained in a single presynaptic vesicle.

In secretion of neurotransmitter from presynaptic vesicles and secretion from chromaffin and other nonneuronal secretory cells, the amount of cellular product released may vary, and variations in amount released are typically observed in physiological preparations. In addition, morphological variations in the secretory vesicles, consistent with the tenet of constancy, are most typically observed, i.e., coexistence of partially filled vesicles or granules with both filled and empty vesicles and granules (Rose et al., 1978; Sagen, and Pappas, 1987; Pappas and Kriho, 1988; Crivellato et al., 2004), variations in vesicular diameter and volume (Fox, 1988, 1996; Dvorak, 2000; Koval et al., 2001), variations in amounts of cellular product released as a function of the temporal frequency of stimulation (Florey and Kriebel, 1983; Lowen et al., 1997), and variations in vesicle pools with respect to age and content (Duncan et al., 2003).

Several forms of indirect measurement of secretion provide evidence supposedly consistent with outright fusion of secretory vesicle membrane with plasma membrane, i.e., measurements of changes in the electrical capacitance of plasma membranes (Neher and Marty, 1982). Increased membrane capacitance has been interpreted as an increase in the amount of plasma membrane through the incorporation of exocytosed vesicular membrane into the plasma membrane (Haimann et al., 1985; Heidelberger et al., 1994). Other reports involving indirect methods, such as deconvolution of inward currents, attributed one quantum to each bouton, and the incorporation of a secretory vesicle into the plasma membrane would increase the area of that membrane and thus prevent exocytosis of the remaining vesicles (Triller and Korn, 1985; Korn and Faber, 1987). In addition, it has been suggested that the fusion of a single vesicle with the plasma membrane established a hypothesized “energy barrier for other vesicles enough that a second exocytotic event hardly ever takes place” (Schikorski and Stevens, 1997). It seems likely that the energy barrier noted by Schikorski and Stevens (1997) represents the van der Waals and other intermolecular forces that serve to repel apposed membranes. Capacitance changes may also be interpreted in a manner quite different from exocytosis.

While indirect measurements of the amount of cellular product needed to elicit a postsecretory cell response at a neuromuscular junction have been made (Kuffler and Yoshikami, 1975), and a direct relationship between the inward calcium current and release of neurotransmitter is well established (Llinás and Heuser, 1977), there are no direct measurements of the amount of cellular product contained within a single secretory vesicle. Similarly, there are no measurements of the statistical distribution of cellular secretory product contained within one or more populations of such vesicles (Garofalo and Satir, 1984; Matthiesen et al., 2001). Thus, we are constrained to interpret vesicular secretion and to understand its mechanism from the standpoint of our strongest sources of evidence: morphology, electrophysiology, and physical chemistry.

The most prevalent model for vesicular secretion is the QVE hypothesis, which holds that release of cellular product (e.g., chemical neurotransmitter) occurs through the release of the entire content of a single vesicle that is extruded to extracellular space, and that the vesicle membrane becomes incorporated into the presynaptic or plasma membrane. Indeed, the QVE hypothesis has gained such acceptance that the terms “vesicle” and “quantum” have become interchangeable and synonymous, i.e., a single vesicle represents a single quantum (cf. Schneggenburger et al., 2002). In a quest to suggest a unifying tenet, this terminology has been applied to other, nonneuronal, secretory processes in which there is no evidence for unity or quanta of released cellular product (Uvnas and Aborg, 1980; Bloc et al., 1999). That being said, the QVE hypothesis has well served those interested in secretion in that it has provided a model to test and against which new findings could be judged. Several variations on the central theme of QVE are currently being tested with fluorescence microscopy, capacitance measurements, and even atomic force microscopy, including kiss-and-run and necking (Fesce et al., 1994; Ales et al., 1999; Fesce and Meldolesi, 1999). Gathering all the evidence, plus new understanding of the physical limits of the methods of observation used now, challenges us to reinterpret the cell's mechanism for secretion.

The QVE hypothesis, while widely cited, does not adequately address properties described in the past decades. Two such considerations are as follows. One, the early work of some researchers (Fatt and Katz, 1952; del Castillo and Katz, 1954) has shown that there is greater molecular leakage of neurotransmitter (constitutive secretion) than that measured as spontaneous miniature end-plate potentials (mEPPs) at the frog neuromuscular junctions, i.e., vesicular release (c.f., Erxleben and Kriebel, 1988a, 1988b). This has also been reported at the squid giant axon (Miledi, 1967) and at calyx synapses (Schneggenburger et al., 1999).

Two, a membrane fusion event (i.e., omega figures) is rarely observed in small- and medium-diameter vesicles except in cases in which the presynaptic transmitter secretion becomes either irreversible (e.g., with 4-aminopyridine, black widow spider venom, calcium ionophore), or there is excessive amplitude of electrical stimulation, i.e., beyond the bounds of normal physiology. In these special nonphysiological cases, the ending does increase in size, presumably due to incorporation of vesicle membrane into the plasma membrane. While high rates or excessive influence by applied toxins will permit visualization of a desirable phenomenon, it is not physiological and thus must be viewed with that reservation.

Other investigators, having recognized some or all of these limitations inherent with the QVE hypothesis, have proposed thematic variations that include necessarily rapid liaisons and mixings of the phospholipids of vesicle and plasma membranes. Indeed, suggestions of different pools of heterogeneous vesicle populations have been proposed to explain short-term synaptic depression, an interpretation based on electrophysiology without supportive morphological evidence (Schneggenburger et al., 2002). While evidence is accruing that there are distinct populations of vesicles, based on vesicle age and content, a rapid selective recruitment of a distinct subpopulation of vesicles is morphologically inconsistent with existing observations (Horrigan and Bookman, 1994; Duncan et al., 2003). The notion of one single vesicle being selected for release, from a population of 50–100 docked presynaptic vesicles that are otherwise identical, is statistically untenable. A mechanism that relies on multiple pools or subset populations of vesicles to provide different amounts of neurotransmitter (which must be exchanged with a different population in a wholesale fashion) makes the logistics of trafficking separate sets of 50–100 presynaptic vesicles, each set containing distinctly different amounts of neurotransmitter, difficult to support in light of the complete absence of supportive morphological evidence (Harlow et al., 2001). As satisfying as such metaphors may be when viewing a subset of the available information, these metaphors become obfuscating in light of the whole body of information already available.

MEMBRANES DO NOT UNDERGO SPONTANEOUS FUSION

It is established that stable membranes do not undergo spontaneous fusion; among the impediments to such fusion are the repulsive forces of the polar head groups of the membrane lipids, and the lateral tension and other forces within the bilayer and that are influenced by membrane waveform and curvature. Similarly, the precise mechanism by which pores form in lipid bilayers and govern their size, structure, and stability is poorly understood. To date, both “wet” experiments and their empirical data and molecular dynamics simulations are of a coarser grain than that needed to address the questions at hand adequately. It is also clear from the literature that the term “pore” has different meanings in different areas of study. Some results can, however, provide clues to the mechanism.

The physical fusion of vesicles with diameters at or below 70 nm is not readily achieved in physical chemical studies of lipid membranes (Wilschut et al., 1981; Lentz et al., 1992; Weikl and Lipowsky, 2001). This is especially true for vesicles studied at physiological levels of monovalent cation concentrations (i.e., 100–150 mM) (Duzgunes and Ohki, 1981; Sundler and Papahadjopoulos, 1981; Okhi, 1984; Wilschut et al., 1985; Rupert et al., 1987; Lentz et al., 1992; Ortiz et al., 1999; Averbakh and Lobyshev, 2000; Chanturiya et al., 2000; Weikl and Lipowsky, 2001; Binder and Zschornig, 2002). Under those conditions, calcium-dependent effects on the associations of vesicles with planar lipid membranes (i.e., the plasma membrane) require calcium levels to reach 10⁻⁴ M or higher, levels that are in fact observed in vivo (Llinás et al., 1992). That elevated level of calcium is short-lived, existing for not more than 1 ms (Silver et al., 1994; Sugimori et al., 1994). Such elevations in local levels of calcium can also be expected to result in concomitant elevations in the protons released, i.e., from lipids to which the calcium would bind (Schnetkamp and Kaupp, 1985). While vesicles may be induced to fuse with planar membranes, the diameters of such fusion-competent vesicles typically must exceed 100 nm, which is larger than diameters observed in vivo. Vesicles smaller than 100 nm have a very small contact site with the planar surface, while vesicles of diameter large than 500 nm have a contact site large enough to support the interbilayer mixing, hemifusion, and ultimate fusion of the vesicle and planar membrane. Such an interpretation is also consistent with the energy barrier to fusion of supernumerary vesicles described by Schikorski and Stevens (1997); thermodynamics tells us that there can be no exception to this energy barrier, such as the suggestion that an initial fusion of a vesicle is somehow permitted, while all subsequent vesicle-plasma membrane interactions are subject to thermodynamic constraints.

In addition to the small contact site, a key limitation to fusion of the smaller diameter vesicles appears to reside in overcoming the high lateral tension forces among the acyl chains that compose the bilayer lipids (Wilschut et al., 1981; Lentz et al., 1992; Weikl and Lipowsky, 2001). Such high lateral forces, balanced as part of a surface tension, would tend to favor a curved, not flat, surface, and thus necessarily pose a strict limit on the number of lipids and area of interaction between the vesicle and planar membrane. In addition, physical chemical studies show that the existence of even transiently stable tubular connections that unite the vesicle lumen and its contents with the extracellular space is energetically unfavorable under physiological conditions, especially on the known millisecond time scale of inward calcium currents and transmitter release.

In many instances, specific fusion proteins regulate protein-protein and protein-lipid interactions required for fusion (Jahn et al., 2003). Fusion of two membranes requires both local perturbations in phospholipid packing, reduction in repulsion forces at the intermembrane hydrophilic surfaces, and local changes in organization and thermal movements of lipids (Cevc and Richardson, 1999; Jahn and Grubmuller, 2002). Genetic deletions of specific regions of such binding proteins show that calcium appears to be required for fusion in which proteins such as SNARE-syntaxin complexes are involved (Reim et al., 2001), i.e., that these proteins do not provide all of the mechanism to drive fusion of vesicles with planar membranes and thus exocytotic release of vesicle contents. It is well established that SNARE-synaptotagmin complexes stabilize the interface and serve to anchor vesicles and the plasma membrane (Sugita and Sudhof, 2000). It is also well established that membrane-encapsulated viruses, such as influenza virus (Skehel and Wiley, 2000), Dengue virus (Modis et al., 2004), and Semliki Forrest virus (Gibbons et al., 2004), utilize viral envelope proteins to bind to cell surface receptors and, following a conformational change in the viral proteins, induce fusion of the viral and cell membranes, thereby affecting viral entry into host cells. While calcium appears to play a role in this viral protein-cell protein association, and in the necessary resulting conformational change in the viral capsid protein, proton levels appear to play a more significant role especially in the hemifusion and fusion of virus and plasma membranes (Forzan et al., 2004; Melikyan et al., 2004). Details of the membrane fusion of the viral and cell lipid membranes, and the involvement of calcium, have yet to be described in full detail. Thus, a picture emerges of a hierarchy of serial steps involved in secretion. One of the early steps for large-diameter bodies such as viruses and large vesicles (e.g., dense-core vesicles in chromaffin cells and insulin-containing vesicles in β-cells) is the use of specific fusion proteins to facilitate the recognition and close association of apposed curved and planar membrane surfaces, i.e., vesicle and plasma membranes, viral and plasma membranes. Calcium's role appears to emerge after the fusion proteins have brought the two membranes into close physical proximity, and that effect is less pronounced than the effect of local pH.

Melikov et al. (1999, 2001) and others have described the calcium-free conditions for formation of pores in black lipid membranes by mechanical or electrical perturbations. Such pores, whose head groups line the lumen of the pore, have a mean diameter of 2 nm with conductance values between 250 and 500 pS (Wilhelm et al., 1993; Melikov et al., 2001); these pores are similar to those described in earlier modeling experiments (Nanavanti et al., 1992). The pores are also similar to the openings (pores) formed and observed during electroporation of excitable cells (eggs) with large secretory vesicles (cortical granules) and lipid vesicle bilayer membranes (Hibino et al., 1991, 1993; Tekle et al., 1994, 2001). Results from these sets of experiments performed with different and complementary approaches are fully consistent. Melikov et al. (1999, 2001) also describe detecting ∼ 1 nm diameter prepores from which pores must arise. From these values, these prepores appear to be the same as the ∼ 1 nm pores that we hypothesize are formed from the interaction of calcium and lipids that are needed for release of chemical neurotransmitter, while the larger pores observed in black lipid membranes and in modeling studies are more applicable to membrane fusions and cell disruption. Recent synthetic studies demonstrate that channel-like pores can be formed in lipid bilayers at reduced pH via an as yet undefined pH-dependent mechanism (Chanturiya et al., 2003) and that a slow rate of membrane fusion of liposomes can occur in such preparations in a pH-dependent manner (Tomohiro et al., 2003), all in calcium-free milieu. Thus, interlipid pores of about 1 nm diameter can be formed in lipid bilayers, induced by calcium, with larger pores being induced by high transbilayer electrical potential, mechanical stretch, and reduction of pH to 6.5 or lower. The calcium levels needed for such lipid-lipid associations are in the order of 10⁻⁴ M under physiological monovalent cation concentration, i.e., the level of calcium measured at secretory microdomains (Llinás et al., 1992).

PORE FORMATION VIA MECHANICAL STRETCHING OF LIPID BILAYERS

Recent molecular dynamics studies have shown, in agreement with “wet” empirical observations, that the tension forces needed to induce rupture is higher in equilibrated membranes than in those with an existing pore. That is, rupture of a membrane due to application of an increase in mechanical tension in the bilayer essentially requires the presence of a preexisting pore (Wilhelm et al., 1993; Olbrich et al., 2000; Heinrich et al., 2001; Evans et al., 2003; Tieleman et al., 2003; Leontiadou et al., 2004). In a similar fashion, the electrical fields used to form these pores in such preparations is 50-fold higher than that of the resting membrane potential. Thus, while illustrative of behavior of lipids in black membrane studies, these pores are likely to be much larger than those needed for flow of vesicle contents from a vesicle lumen to the extracellular space. These experiments are also performed in the absence of appreciable levels of calcium known to be required for calcium-lipid associations at the physiological levels of monovalent cations used (Duzgunes and Ohki, 1981; Sundler and Papahadjopoulos, 1981; Okhi, 1984; Wilschut et al., 1984; Rupert et al., 1987; Lentz et al., 1992; Ortiz et al., 1999; Averbakh and Lobyshev, 2000; Chanturiya et al., 2000; Weikl and Lipowsky, 2001; Binder and Zschornig, 2002), as well as for secretory release of neurotransmitter (Dodge and Rahamimoff, 1967; Llinás et al., 1991, 1992; Heidelberger et al., 1994; Silver et al., 1994; Sugimori et al., 1994).

RESULTS OF MORPHOLOGICAL STUDIES AND STATISTICS OF VESICULAR SECRETION: SINGLE-VESICLE FUSION VERSUS VESICLE ARRAY SECRETION AND ENTRY OF MEMBRANE-ENCAPSULATED VIRUSES

Fusion of vesicles with the plasma membrane, which is a hallmark of QVE and thus should be readily observed in thorough morphological studies, apparently does not occur in healthy terminals. Several studies report morphological evidence showing vesicle fusion and claim that such evidence is proof enough to support the claim of vesicle fusion being required for QVE (Heuser and Reese, 1981; Torri-Tarelli et al., 1985). Such published images are necessary to support the case for vesicle fusion and QVE. However, are such images truly representative of the normal physiology of secretion? In actuality, vesicles-plasma membrane fusion is only observed in compromised terminals, i.e., those stimulated to amplitudes or frequencies that are well beyond those normally observed, and which do not recover after rest (Rose et al., 1978). The use of reagents that result in exaggerated physiological responses or excessive electrical depolarization may be justifiable to reveal events whose natural occurrence is of low probability, but their results cannot be considered reflective of normal physiological functioning. While high rates of stimulation or excessive influence by applied toxins will permit visualization of a desirable phenomenon, it is not physiological per se and thus must be viewed with that reservation. Thus, the assumption that data obtained in morphological studies showing vesicle membranes fusing with plasma membranes following exposure to either 4-aminopyridine (Heuser et al., 1979; Tokunaga et al., 1979; Heuser and Reese 1981; Forsman and Elfvin, 1983; Torri-Tarelli et al., 1985), calcium ionophore (Kita and Van Der Klotz, 1976; Llados et al., 1988; Hunt and Silinsky, 1993), or excessive depolarization must be carefully weighed.

Another view of secretion of quanta of neurotransmitter (e.g., acetylcholine) through complex protein particles called mediatophores located in the presynaptic membrane of Torpedo has been advanced (Dunant and Israel, 1995). In this scheme, a calcium-dependent protein pore forms and opens to release neurotransmitter from the presynaptic cytosol. Further, the body of work of Israel and Dunant also challenges the notion of a functional linkage between exocytosis and endocytosis, which would account for the observed transient increase in plasma membrane capacitance. In a divergence of the notion of “one vesicle = one quanta,” they view quantal release instead as representing pulsatile release of some amount of neurotransmitter, but a mechanism by which pulsatile becomes quantal has yet to be described and tested. However, while their idea about release of neurotransmitter through protein pores is intriguing, several groups report no such particle appearance/formation related to secretion, and especially not in relationship to local changes in calcium levels, i.e., studies by Harlow et al. (2001) with muscle, Gilligan and Satir (1983) with Tetrahymena, and Garofalo and Satir (1984) with Paramecium, who find no evidence of the formation of such particles, especially in response to local elevations of calcium levels. When viewed in the light of the physical chemical constraints or the lipid-calcium interactions in membranes discussed above, the recent findings (Taraska et al., 2003; Taraska and Almers, 2004) that promote the kiss-and-run scheme for secretion are more consistent with the salt bridge transient pore formation of the porocytosis hypothesis.

Interestingly, electron microscopic studies using thin sections, semithin sections, frozen thin sections, freeze fracture/freeze etch, etc., have not revealed a pore traversing the lipid bilayer. This is not surprising when one considers the inherent physical constraints of the methods. Typical thin (i.e., silver interference color) sections are between 50 and 70 nm thick as used by Harlow et al. (2001), while semithin sections (e.g., 70–100 nm thick) were used in the study of Torri-Tarelli et al. (1985) and by Harlow et al. (1998). If one considers the lipid-free volume of the pore itself to represent a signal and the surrounding membrane to represent background, then the background would be about 50–70 times that of the signal for typical thin sections (epoxy-embedded or frozen), and about 70–100 times that of the signal in semithin sections. Such unfavorable ratios of signal to background are not likely to be detected in any current imaging system. Similarly, a 1–2 nm diameter pore in an en face planar membrane would be obscured by layer of metal ions (e.g., platinum) several atoms thick used to shadow and provide contrast in either freeze fracture or freeze etch, and detail would be further obscured by the multiatom-thick overlain carbon layer used to stabilize the metal shadow (Gilligan and Satir, 1983; Garofalo and Satir, 1984; Torri-Tarelli et al., 1985; Harlow et al., 2001). Thus, it is easy to understand why no one has yet to report observations of pores (whose dimensions are consistent with biophysical and physical chemical studies on lipids and membranes per se) in the plasma membrane.

Another imaging method recently applied to granular secretion is atomic force microscopy (AFM) (Beuers, 1997; Jena, 2003; Cho et al., 2004), in which successive lines are scanned with a tip moved laterally across a field of view. While AFM tips traverse a sample along a single line in a field of view at a rate of 2–4 Hz (500 to 250 ms per scan line), acquisition of a typical two-dimensional image requires one to several minutes. This is an enormous duration relative to a single open time for a synaptic vesicle, i.e., 1 ms. However, there is no way to know with certainty the location of the site of secretion from a single vesicle or granule. Thus, a typical open time would at best represent 1:60,000 or about 0.2% of the duration of a single typical scan. Therefore, the time needed to acquire a single scan with an AFM significantly exceeds the 1-ms duration of the single vesicle release at the neuromuscular junction observed with electrophysiological methods or the open time of a transient pore. Furthermore, the small dimensions of the passageway or transient pores (1–2 nm diameter), residing in a semirigid lipid bilayer, make detection of such pores problematic at best. Given the characteristic physical dimensions involved, it would seem very difficult to ensure that the opening for any one random synaptic vesicle could be achieved, which would be required if the QVE hypothesis were operant. Therefore, given the inherent methodological limitations of AFM, it is exceedingly difficult to understand how a method such as AFM can be used to resolve the structure or dynamics of a transient pore of 1 or 2 nm diameter whose lifetime is on the order of 1 ms. Therefore, the inherent methodological limitations of AFM preclude a statistically significant observation of the structure and dynamics of vesicular secretion via either QVE or the transient pores of porocytosis.

VESICLE DIAMETER AND VOLUME

Morphological analyses of secretory vesicle diameters indicate that volumes vary by more than 20% (Table 1) (Fox, 1988, 1996; Koval et al., 2001). Such variation in volume, with the reasonable expectation that vesicle contents should be at similar concentrations (amount of cellular product molecules per unit volume), stands in contrast to the requirement for constancy of the amount of cellular product released in a vesicle secretion event, i.e., the quantum required by the QVE hypothesis, or its variations. In addition, unlike the case of intercellular junctional complexes or membrane-localized ion channels, no secretion channel particles are observed at either the plasma or the vesicle membranes, especially at their neighboring surfaces. This point is elegantly illustrated by the electron microscopy of Harlow et al. (2001), in which the density of vesicle membranes is consistent throughout their circumference, and the optical density range and pattern of plasma membrane segments shown are the same as those of the vesicles, i.e., there are no prominent protein particles within these membranes that could serve as pores (Garofalo and Satir, 1984; Dunant and Israel, 1995, 2000; Seagar et al., 1999; Harlow et al., 2001). Furthermore, the micrographs shown by Harlow et al. (2001) are clearly consistent with the notion of vesicle arrays that can partially release their contents in unison.

Table 1. Summary of the probabilities for secretion of the contents of a single vesicle for different mechanisms of vesicular fusion

Proposed Mechanisms for Ssecretion by Vesicle Fusion	Equation (Above)	Probability
Secretion of the contents from 1 vesicle of a group of 50 vesicles	PA-50	8.88 × 10^−16
Secretion of the contents from 1 vesicle of a group of 100 vesicles	PA-100	5.77 × 10^−30
Wholesale exchange of two different sets of vesicles for a ridge of 50 docked vesicles and 50 undocked vesicles, from which the contents of 1 of those exchanged vesicles is secreted	PB	3.17 × 10^−29
Release of contents from 1 vesicle from a set of 50 vesicles that is exchanged for a different set of 50 vesicles	PC	1.78 × 10^−15
Release of contents of 1 vesicle from a set of 50 vesicles that was exchanged for another vesicle from a different set of 50 vesicle	PD	3.15 × 10^−30
Only 10 vesicles out of 100 have the full complement of accessories and are thus able to release their contents	PE	7.89 × 10^−30

Morphological studies of granule secretion from chromaffin cells show that numerous secretory vesicles lie apposed to the plasma membrane (Rose et al., 1978; Sagen and Pappas, 1987; Pappas and Kriho, 1988). The temporally coincident presence of empty, partially filled, and filled vesicles without observation of omega figures offers significant cause to question a model that requires vesicle fusion per se (Rose et al., 1978; Sagen and Pappas, 1987; Fox, 1988, 1996; Pappas and Kriho, 1988; Dvorak, 2000; Koval et al., 2001). A similar story emerges with secretory vesicles and granules in Paramecium (Gilligan and Satir, 1983), Tetrahymena (Garofalo and Satir, 1984), and hagfish (Downing et al., 1981), in which release of large amounts of cellular material (e.g., strands of mucus) are released from vesicles, each surrounded by a rosette of particles (Satir and Satir, 1974; Satir and Oberg, 1978; Gilligan and Satir, 1983). The particles, which are observed in both resting and active states (and not formed as a consequence of changes in calcium levels), are believed to be calcium channels that would provide localized elevation of calcium levels required for secretion (Satir and Oberg, 1978; Gilligan and Satir, 1983). These cases are morphologically consistent with the observations of particles (voltage-dependent calcium channels) adjoining secretory vesicles anchored in the active zones in the presynaptic terminal (Gulley et al., 1978; Pumplin et al., 1981; Schaeffer et al., 1982; Pawson et al., 1998). In the cases of the relatively large granules of Paramecium, Tetrahymena, hagfish, and chromaffin cells, fusion of the granule and plasma membranes is readily observed, consistent with the physical chemical studies discussed below. In these cases, fusion of membranes is readily observed when the granule diameter exceeds 100 nm.

Table 1 shows that the volumes in secretory vesicles of presynaptic terminals and secretory granules in chromaffin cells, and thus presumably the vesicle or granule contents, vary by more than 20%. If the notion of “one vesicle = one quantum” was accurate, then observed coefficients of variance for amounts of cellular products would have to exceed 10%, when they are observed to be 3% in cases of the so-called quantal release. This is an additional complication for the notion that the content of a single vesicle represents the amount of cell product released at a single secretory event.

Electrical Capacitance

It is argued that the changes in electrical capacitance that accompany secretion are necessary proof of a vesicle fusion event and thus imply occurrence of morphological change that is not observed (Schneggenburger et al., 2002). Simply dismissing the morphological argument, and claiming that morphologists have not looked hard enough (Schneggenburger et al., 2002), is not a viable argument. Capacitance changes are electrical events and indicate the nature of surfaces (volumes) accessible to electrons. Outright fusion of the vesicle membrane with the plasma membrane is not the only means of increasing that electron-accessible surface (volume). Capacitance is a measure of the number of electrons that can be held by a body; thus, that capacity is a function of the surface area available for electrons. Placing vesicles in sufficiently close proximity to the plasma membrane should suffice to increase membrane capacitance in a manner analogous to the production of capacitors in integrated circuits, i.e., placing dots or small objects sufficiently close to a capacitance plate will increase plate capacitance. As capacitance is a measure of electron-accessible (surface) space, allowing electrons to bridge from the planar plasma membrane to spherical plasma membrane (i.e., via a calcium-lipid salt bridge) would be seen as increased plasma membrane capacitance. Indeed, arrays of such dots are typically used in combination with planar surfaces to ensure the reliability and uniformity of spatial and temporal response in integrated circuits. Such dots and plates are a direct analogy to the placement of anchored spheroidal secretory vesicles next to the planar plasma membrane that is observed in cells (Harlow et al., 2001).

The single-vesicle hypothesis would require that the increased electron carrying capacity of the plasma membrane be concentrated on the surface area corresponding to that single vesicle, with a corresponding increase in net negative charge on that vesicle's cytosolic surface. This is consistent with the interpretation of single-vesicle fusion and reuptake suggested by Neher and Marty (1992), which in turn is consistent with the single-vesicle (quantum) interpretation of Katz. However, such an increase in charge could be expected to result in a very large increase in interbilayer repulsion by existing polar head groups of membrane lipids rather than the expected attraction at the contact site.

Alternatively, the many electrons represented by the capacitance change could be distributed on an array of many small vesicles. The total increase in capacitance of the array would be equal for the array and the single vesicle, but distribution of the total charge across an array would greatly reduce the changes in interbilayer repulsion forces at each arrayed vesicle's contact site compared with that for a single vesicle, and without a requisite need for either fusion of spherical vesicles with planar membranes or energetically costly budding of small diameter spherical vesicles from planar surfaces. The second case, in which small numbers of electrons are added to each of many arrayed vesicles (e.g., the ridge of the neuromuscular junction), would be analogous to the capacitance properties of spatially ordered quantum dots whose diameters are less than 300 nm (Fig. 1) (Macucci et al., 1993; Zhitenev et al., 1999; Brodsky et al., 2000).

Figure 1. A graph of the vesicle or granule volume for each of the examples listed by line number in Table 2, sorted by increasing particle diameter. In addition, the 100 and 300 nm boundaries identified for vesicle fusion with planar membranes are also indicated. Physical chemical studies and numerical analysis and modeling indicate that no fusion with the plasma membrane will occur for vesicles of 100 nm diameter or less (e.g., the dashed red line), and that fusion is possible but problematic for vesicles and granules whose diameters are between 100 and 300 nm, e.g., between the dotted green (upper) line and dashed red (lower) horizontal lines. Physical chemical studies and numerical analysis and modeling indicate that no fusion with the plasma membrane will occur for vesicles of 100 nm diameter of less (e.g., the dashed red line), and that fusion is possible but problematic for vesicles and granules whose diameters are between 100 and 300 nm (e.g., between the dashed red and dotted green lines). Vesicle membrane fusion with plasma membranes is readily observed for particles (vesicles and granules) whose diameter is greater than 300 nm (e.g., dotted green line).

Table 2. Source of secretor/vesicles*

Line	Source of Secretory Vesicles	Mean Diameter (nm)	SD (nm)	Volume nm^3
* Summary of the available morphological analyses of mean secretory vesicle diameters, the standard deviation of the mean (SD) in those diameters, and the volume of those vesicles and granules from various neurons and chromaffin cells that indicate that vesicle volumes vary by more than −20% (Fox, 1988, 1996; Koval et al., 2001), and that those volumes cluster according to vesicle contents and function. In several cases, SD values were not provided in the cited publications and are thus omitted. The table is sorted by mean diameter of a particle (vesicle or granule) type. Line numbers (leftmost column) for each entry are used to identify ordinate values for Figure 1. Chemical neurotransmitters are typically found in small vesicles whose diameters are less than 100 nm, while vesicles and granules with larger diameters are seen to fuse with the plasma membrane under physiological conditions. Such variation in volume, with the reasonable expectation that vesicle contents should be at similar concentrations (amount of cellular product molecules per unit volume), stands in contrast to the requirement for constancy of the amount of cellular product released in a vesicle secretion event, i.e., the quantum required by the QVE hypothesis or its variations. a Fox (1988). b Koval et al. (2001). c Heijnen et al. (1999). d Rorsman and Renstrom (2003). e Amano et al. (2001). f Ornberg and Resse (1981). g Martin et al. (1999).
1	Malapterurus presynaptic vesiclesa	45.00	1.10	3.609E + 04
2	Torpedo eye muscle 1 presynaptic vesiclesa	62.00	2.10	1.022E + 05
3	Chromaffin cell, small light vesiclesb	63.71	9.18	1.115E + 05
4	Rat diaphragm presynaptic vesiclesa	65.00	1.00	1.188E + 05
5	Torpedo eye muscle 2 presynaptic vesiclesa	68.00	3.20	1.373E + 05
6	Chromaffin cell, small coarse vesiclesb	70.99	16.98	1.574E + 05
7	Chromaffin cell, small dense coarse vesiclesb	71.61	11.13	1.618E + 05
8	Skate presynaptic vesiclesa	73.00	3.10	1.720E + 05
9	74 mm Torpedo presynaptic vesiclesa	78.00	4.40	2.122E + 05
10	Neonate Torpedo presynaptic vesiclesa	79.00	3.60	2.209E + 05
11	Adult Torpedo presynaptic vesiclesa	83.00	3.90	2.582E + 05
12	Chromaffin cell, extra coarse vesiclesb	91.36	15.19	3.491E + 05
13	Activated platelets, small secretory vesiclesc	100.00		4.632E + 05
14	Chromaffin cell, low-density coarse vesiclesb	133.99	32.88	1.150E + 06
15	Chromaffin cell, coarse vesiclesb	161.92	31.69	2.062E + 06
16	Chromaffin cell, large coarse vesiclesb	249.32	55.30	7.730E + 06
17	Chromaffin cell, large light corase vesiclesb	267.97	73.35	9.631E + 06
18	Beta cells, insulin granulesd	300.00		1.358E + 07
19	Chromaffin cell, large dense coarse granulesb	309.12	60.10	1.487E + 07
20	Rat basophilic leukemia cells, small vesiclese	500.00		6.389E + 07
21	Activated platelets, large secretory vesiclesc	1,000.00		5.173E + 08
22	Rat basophilic leukemia cells, large vesiclese	1,500.00		1.753E + 09
23	Limulus amebocyte secretory granulesf	1,650.00		2.335E + 09
25	Brine shrimp hemocyte granulesg	6,000.00		1.131E + 11
	Summary
	Maximum	6,000.00	73.35	1.13E + 11
	Minimum	45.00	1.00	3.61E + 04
	Median	91.36	6.79	3.49E + 05
	Mean	524.12	18.42	4.71E + 09
	Standard deviation	1,221.29	22.81	2.26E + 10
	Coefficient of variation	2.33	1.24	4.79E + 02

Fatt and Katz (1952), using electrical amplifiers of the time, were able to record postsynaptically in the muscle fiber at a neuromuscular junction preparation of frog. They termed the smallest signal they were able to record the miniature end-plate potential, which was designated at the basic unit of postsynaptic response. This mEPP, of course, does not cause muscle contraction. It is well known that it takes 50 or more of these unit responses to cause contraction, depending on the muscle (Fatt and Katz, 1952; del Castillo and Katz, 1954, 1957; Miledi, 1967; Kuffler and Yoshikami, 1975; Katz and Miledi, 1979). The extension of the unit response as a quantum to equal a single vesicle awaited electron microscopic observations of synaptic vesicles; these were first published in the early 1950s.

When Katz became aware of the publications of Palay and Palade showing discrete membrane-bound vesicles clustered at the presynaptic plasma membrane (Palay and Palade, 1954, 1955; Palay, 1956), he suggested that each mEPP resulted from exocytosis of the contents of a single vesicle—thus was born the quantal vesicular exocytosis hypothesis. There was no morphological proof to Katz's suggestion, only an interpolation that the observed postsynaptic signal (mEPP) was directly related to ultrastructural features at the synapse (synaptic vesicles), i.e., the hypothesized fusion of single vesicles with the presynaptic plasma membrane. Therefore, borrowing the then-popular jargon from physics, Katz called the released contents from a single vesicle a quantum.

Some years later, using more sensitive amplifiers having lower intrinsic noise levels and very much improved electrodes, Kriebel et al. recorded bona fide signals, which were smaller than mEPPs, from much smaller-diameter edge muscle cells in frog sartorius muscle (Kriebel and Gross, 1974; Kriebel and Stopler, 1975; Kriebel et al., 1976). While Katz's studies used muscle cells that were 30–50 micrometers in diameter, the muscle cells studied by Kriebel were only 6–10 micrometers in diameter. These findings were subsequently confirmed by others (Vautrin and Mambrini, 1981; Muniak et al., 1982; Wiegand et al., 1984; Erxleben and Kriebel 1988a, 1988b; Vautrin and Kriebel, 1991, 1992). Kriebel named each of the subunit signals of mEPPs as a subminiature end-plate potentials (sub-mEPP). The demonstration of the sub-mEPP, a singular subunit, presented an insurmountable conundrum for Katz's interpolation: how could there be secretion of less than a quantum of neurotransmitter?

Thus, Katz's quantum may be more appropriately considered as a recorded unit signal that had subcomponents not resolved in measurements and from larger muscle fibers made before those of Kriebel and others (Kriebel and Gross, 1974; Kriebel and Stopler, 1975; Kriebel et al., 1976). Furthermore, the hypothesized quantal release, which requires the wholesale release of the entirety of vesicle contents, was known to vary as a function of stimulation frequency and amplitude, and which depended on an inward flow of extracellular calcium (del Castillo and Katz, 1957; Dodge and Rahamimoff, 1967; Kuffler and Yoshikami, 1975; Rose et al., 1978; Florey and Kriebel, 1983; Garofalo and Satir, 1984; Erxleben and Kriebel, 1988a; Llinás et al., 1991, 1992; Fox and Kriebel, 1994; Heidelberger et al., 1994; Silver et al., 1994; Sugimori et al., 1994; Travis and Wightman, 1998), could not represent the contents of a single vesicle of a finite volume culled from a population of identical vesicles, as has been suggested (Schneggenburger et al., 2002). Nor could the quantum come from vesicles that are selectively trafficked at close quarters, then perfectly chosen from one or another set of different-sized vesicles, to the absolute exclusion of all other vesicles.

Dynamics of Neurotransmitter Release

A biochemical study of synaptic vesicles from cholinergic neurons shows that they contain an integral acetylcholine transporter that moves acetylcholine from the cytoplasm into the vesicle lumen (Eder-Colli and Amato, 1985). That transporter acts to maintain vesicles at a particular level of neurotransmitter. As such, this transporter will replenish the contents of anchored discharged vesicles. Thus, a partially discharged vesicle could be refilled with the amount of neurotransmitter released upon depolarization. Since it has been established that in identified neuromuscular synapse preparations following tetanic stimulation causing synaptic fatigue is not correlated with changes in the synaptic vesicle population (Rose et al., 1978; Pappas et al., 1988), the synapses are not completely depleted of their vesicles (Pappas et al., 1988). This then places the dynamics of neurotransmitter release under the control of five parameters: voltage dependency, extracellular calcium levels, stimulation frequency, mass action driven by levels of neurotransmitter in the vesicle lumen versus the level in the cleft, and interaction of the cytoplasmic faces of the vesicle and plasma membranes.

Mechanistic Schemes Based on Morphologic Studies

Morphological studies of neuromuscular junctions with conventional and cryoelectron microscopy have established that there are between 50 and 100 docked presynaptic vesicles at each ridge (Palay and Palade, 1954, 1955; Palay, 1956; Birks et al., 1960; Heuser and Reese, 1981; Torri-Tarelli et al., 1985; Harlow et al., 2001). In such a highly organized system, there are four possible mechanistic schemes that we consider here: one, 1 vesicle from a group of 50 or 100 vesicles secretes its entire contents across the plasma membrane; two, the wholesale exchange of two different sets of vesicles for a ridge of 50 docked vesicles and 50 undocked vesicles, from which the contents of 1 of those exchanged vesicles is secreted; three, the release of the contents of 1 vesicle from a set of 50 vesicles that was exchanged for another vesicle from a different set of 50 vesicles; and, four, the release of contents from 10 competent vesicles having the full complement of accessory proteins out of a ridge population of 100 that have the full complement of accessories and are thus able to release their contents. Let us look at each of these four possible cases.

Secretion of contents from 1 vesicle and from a group of 50 vesicles

In this case, the probability of the release of contents of a single secretory vesicle in a set of 50 vesicles in a ridge or bouton would be

• 

Similarly, the probability for the release of material from a single vesicle from a population of 100 vesicles would be

• 

In such a highly organized system in which all vesicles were considered equally competent, the probability of the release of contents of a single secretory vesicle in a set of 50–100 vesicles would range between 8.88 × 10⁻¹⁶ and 7.889 × 10⁻³¹, i.e., extremely improbable (if not prohibitive) odds. Such odds make it impossible for a single cell to release a quantum at a single synapse by the QVE hypothesis, and some cells have more than 20,000 synapses, e.g., Purkinjie cells.

From these two examples, we can see the difficulty in relying on purported mechanisms that require release of cell material from the complete fusion of one vesicle from a relatively large population of vesicles in chromaffin cells (Sagen and Pappas, 1987; Fox, 1996; Plattner et al., 1997), PC-12 cells (Saegusa et al., 2002), salivary gland (Segawa et al., 1998), or the presynaptic ridge at the neuromuscular junction (Birks et al., 1960; Rose et al., 1978; Pawson et al., 1998; Harlow et al., 2001).

Wholesale exchange of two different sets of vesicles for a ridge of 50 docked vesicles and 50 undocked vesicles from which the contents of 1 of those exchanged vesicles is secreted

The amount of material released per secretory event can vary depending on the nature or frequency of the stimulation (del Castillo and Katz, 1957; Kuffler and Yoshikami, 1975; Rose et al., 1978; Florey and Kriebel, 1983; Garofalo and Satir, 1984; Erxleben and Kriebel, 1988a; Fox and Kriebel, 1994; Travis and Wightman, 1998); these quanta are not immutable, i.e., at the neuromuscular junction, and granular secretion in protozoa, chromaffin cells, and pancreatic β-cells. A recent explanation for this phenomenon, which is consistent with notions of quantal release through single granules or vesicles, is that there exist within cells distinct populations of vesicles whose contents differ by proscribed amounts and concentrations (Schneggenburger et al., 2002). There is no published morphological or biochemical evidence to support such a notion. Neither is there evidence of how such populations would be distinguished within cells, in millisecond time scales (Llinás et al., 1992; Silver et al., 1994; Sugimori et al., 1994), i.e., the submillisecond time frames needed to change response patterns as observed in published records. In light of an absence of morphological and biochemical support for such a mechanism, let us consider the statistical footing for such a model. The probability for the wholesale exchange of two different sets of vesicles for a ridge of 50 docked vesicles and 50 undocked vesicles (P_B) would be

• 

Release of contents of 1 vesicle from a set of 50 vesicles that was exchanged for another vesicle from a different set of 50 vesicles

Alternatively, if only 1 vesicle among a set of 50 docked vesicles was exchanged for another vesicle of different content, then the probability for that exchange (P_C) would be

• 

However, such a mechanism would have to ensure that only that distinct new vesicle from among the total of 50 docked vesicles would be selectively and specifically released. That would change the picture considerably, i.e., the combined probability of the fusion/release of the contents of the specific single exchanged vesicle (P_D) would be

• 

These probabilities would prove even more unfavorable for ridges or synapses with more than 50 docked vesicles. Therefore, without a demonstrated mechanism to support the assertion of wholesale vesicle exchange, these are prohibitive constraints on any demonstrated mechanism and thus the quantum vesicular exocytosis hypothesis.

Release of contents from 10 competent vesicles having the full complement of accessory proteins out of a ridge population of 100 that have the full complement of accessories and are thus able to release their contents

Let us also consider the case in which 10 vesicles out of 100 have the full complement of accessories and are thus able to release their contents. In this case, the probability for that release (P_E) would be

• 

These findings are summarized in Table 1. In each of these cases, a mechanism would have to be invoked to ensure that one and only one vesicle was released and that all of the other vesicles (i.e., number of vesicles − 1) were prevented from releasing their contents. No such exclusionary mechanism has yet been demonstrated (Triller and Korn, 1985; Korn and Faber, 1987).

Thus, the observed constancy of the amount of neurotransmitter secreted with nerve stimulation, attested to by the small value of the coefficient of variance of end-plate potentials, can only be explained by release of small amounts of neurotransmitter molecules from arrays of anchored vesicles at each release site of the neuromuscular junction. Since the coefficient of variation of the quantal packet is a function of 1 over the square root of the number of contributing vesicles, and there are 30–50 in an array, a standard amount of secretion is guaranteed by the array with each action potential. The array notion, especially at the neuromuscular junction, is viable in maintaining a standard packet size so that anchored vesicles need not become empty as transporters on the vesicle membrane can continuously replenish them since 60% of the acetylcholine in the synapse is present in the cytosol (Zimmerman and Denston, 1977; Zimmermann, 1982). In addition, acetylcholine is readily available for transporters because synthesis, mediated by acetyl-O-transferase, is known to occur on the cytoplasmic surface of the vesicle membrane (Eder-Colli and Amato, 1985). In fact, it has been reported that newly synthesized transmitter is preferentially released (Zimmerman, 1982). Complimenting this perspective, recent numerical simulations indicate that arrays of receptor molecules increase the likelihood of accurate detection of ligand, i.e., released chemical neurotransmitter (Duke and Bray, 1999).

It is also likely that, concurrently, there are mechanisms for docking and undocking vesicles that are independent of secretion, each process having its own identifiable rate constants. Small variations in amounts of neurotransmitter released are readily accommodated by modulating (e.g., through small changes in calcium dynamics) the amount released from many vesicles whose diameters are observed to vary by 3–10% (Fox and Kriebel, 1994; Fox, 1996). Most importantly, the array concept permits quantal size to be frequency-dependent. Thus, to achieve the observed characteristic constancy of quantal release (i.e., miniature end-plate potential − mEPP size), the synapse must rely on simultaneous secretion from many anchored vesicles within an array of anchored vesicles. The porocytosis mechanism we have proposed uniquely meets these requirements. We believe further that the porocytosis mechanism extends to secretion in other nonsynaptic systems.

Lipid Bilayer Physical Chemistry

Evidence from the physical chemistry of lipid bilayers must also be taken into account when considering the process and mechanisms of secretion. Current hypotheses, other than porocytosis (Kriebel et al., 2001; Silver et al., 2001, 2003), view exocytotic secretion as involving/requiring the physical fusion of vesicle and plasma membranes. That hypothesized mechanism does not fit all available physiological (Kriebel et al., 2001; Silver et al., 2001, 2003) or physical chemical data (Cheng et al., 1999; Averbakh and Lobyshev, 2000; Huster et al., 2000; Binder and Zschornig, 2002). Furthermore, energetics of apposed lipid bilayers does not favor unfacilitated fusion. Calcium ion levels are elevated in microdomains for 1 ms and to levels of 10⁻⁴ to 10⁻³ M and are of limited lateral mobility at the site of secretion (Llinás et al., 1992; Silver et al., 1994; Sugimori et al., 1994).

Physical chemical studies of membrane-membrane interactions among bilayer membranes offer some important insights into the cell physiology of vesicular secretion. It has been observed that under conditions of physiological monovalent cation levels (e.g., 100–150 10⁻³ M Na⁺), the physical fusion of bilayers requires calcium at levels of more than 10⁻⁴ M. This is the level of calcium we observed at the presynaptic terminal (Llinás et al., 1991, 1992; Silver et al., 1994; Sugimori et al., 1994). At these levels, phospholipids undergo a phase transition and form a complex with calcium (Galla and Sackmann, 1975; Duzgunes et al., 1981; Okhi, 1984; Rupert et al., 1987; Ortiz et al., 1999; Averbakh and Lobyshev, 2000; Chanturiya et al., 2000; Binder and Zschornig, 2002). In these and other model systems, no fusion or release of contents from the lumen of a bilayer vesicle occurs at physiological monovalent cation levels (e.g., 150 mM) and calcium levels below 10⁻⁴ M, especially for vesicles whose diameters are less than 100 micrometers (cf. Fig. 1). This restriction is due largely to the resistance to changes in vesicle curvature imparted by lateral tension forces between the acyl chains of bilayer phospholipids in vesicles whose diameters are less than 100 nm. The energetics for those systems is only slightly more favorable for vesicles whose diameters are between 100 and 300 nm. Complete fusion is more readily observed for vesicles whose diameters are more than 500 nm (Duzgunes and Ohki, 1981; Sundler and Papahadjopoulos, 1981; Wilschut et al., 1985; Rupert et al., 1987; Lentz et al., 1992; Weikl and Lipowsky, 2001; Binder and Zschornig, 2002).

Such studies that consider artificial lipid bilayers as a model system for the plasma membrane have a mechanistic focus at the sites of calcium-lipid and lipid-lipid interactions. Many of these contemporary studies seek to extend these to more macroscopic model systems, including planar lipid bilayers, or the so-called black membranes. These powerful tools have yielded much important information (Schindler, 1980). However, they may not be suitable for studies of the release of materials through either vesicular or constitutive secretion due to the fact that they typically rely on global changes in calcium levels on one or both leaflets of the lipid bilayer membrane, i.e., calcium changes occurring along membrane surfaces whose surface dimensions are large relative to the diameter and surface area of the secretory vesicle, which is contrary to the well-accepted observation of calcium microdomains at the site of vesicle-membrane interaction and secretory release (Llinás et al., 1991, 1992; Silver et al., 1994; Sugimori et al., 1994). Such studies of artificial membranes treat the plasma membrane as a continuum of large area, while the changes in calcium and thus calcium-induced alterations in lipid-lipid interactions visible on a macroscopic scale (cf. Silvius and Gagne, 1984; Averbakh and Lobyshev, 2000) are more limited and focused in vivo. Thus, each localized elevation in calcium levels, sufficient to cause the freezing of lipids and thus induce local perturbations in membrane permeability (i.e., within microdomains), would represent a discontinuum within the lipid bilayer at a contact site between the vesicle membrane and plasma membrane and break symmetry within the plane of the membrane. It is at such breaks that activities are typically found (Winfree, 1977; Peacocke, 1983). A similar mechanism could be expected to permit constitutive secretion, i.e., without the vesicle, and transmembrane bridging by calcium.

We envision a mechanism in which release of cellular materials stored in anchored vesicles is achieved across the lipid bilayers of vesicle and plasma membranes, i.e., through a transient pore established through the interaction of calcium and membrane lipid. Thus, we consider that a calcium ion, whose lateral mobility is restricted in space and time (Silver et al., 1994; Sugimori et al., 1994), establishes salt bridges among adjacent lipid molecules (Ortiz et al., 1999; Ravoo et al., 1999; Averbakh and Lobyshev, 2000), as well as transient pores that span the anchored vesicle and plasma membrane lipid bilayers (Figs. 2 and 3). The lifetime of that pore will be dependent on the duration of sufficiently high calcium levels. The anchored vesicles are positioned by being docked next to the plasma membrane with SNARE-syntaxin, SNARE-synaptotagmin, and other anchoring and fusion proteins (Sugita and Sudhof, 2000; Jahn et al., 2003). Calcium enters this site through juxtaposed calcium channels, i.e., voltage-gated channels in the plasma membrane (Llinás et al., 1992; Silver et al., 1994; Sugimori et al., 1994). The calcium levels are known to reach 10⁻⁴ to 10⁻³ M within the microdomain for a brief period of time, i.e., 1 ms (Llinás et al., 1992; Silver et al., 1994; Sugimori et al., 1994). Then calcium associates with the phosphate on a phospholipid molecule and bridges to an adjacent phospholipid (Averbakh and Lobyshev, 2000). The phospholipids freeze as single phospholipid molecule entities become dimers (Galla and Sackmann, 1975; Duzgunes et al., 1981; Ravoo et al., 1999). The resulting phase transition decreases spacing between complexed lipids and increases the spacing between frozen and free phospholipids, i.e., forming a transient tunnel or pore through which small entities may traverse. The bridging by calcium overcomes the ionic and van der Waals charge-based repulsion among opposing phospholipid head groups (Galla and Sackmann, 1975; Ortiz et al., 1999; Ravoo et al., 1999; Averbakh and Lobyshev, 2000).

Figure 2. A sketch showing the effect of adding calcium ions to a layer of phospholipids (e.g., dipalmitoylphosphatidylcholine, DPPC). The sketch shows the effect of calcium addition to a bilayer of lipids; the left side of the panel shows the calcium-free configuration, while the right side of the panel shows the effect of addition of a calcium ion, which can form a salt bridge between adjacent lipids. The salt bridge has the effect of decreasing the intermolecular distance between the lipids complexed with calcium, thereby increasing their distance from neighboring lipid molecules not complexed with calcium. Extending this to a plane would reveal a collar of closely spaced lipid-calcium complexes around a gap or pore that would exist for only as long as the localized level of calcium was high enough to form calcium-lipid complexes. This salt bridge effect on interlipid spacing is comparable to the phase change observed in lipid monolayers and bilayers locally exposed to calcium solutions. Calcium ions are shown as large red dots. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com].

Figure 3. A montage showing greater detail of the relationships between secretory vesicles, the plasma membrane, membrane phospholipids, and calcium ions according to the porocytosis hypothesis. A: Electron micrograph shows the presynaptic vesicles and the plasma membranes of a portion of a neuromuscular junction (Pappas et al., 1988). The box outlines several vesicles immediately adjacent to the plasma membrane and points to B, which shows the juxtaposition of lipid bilayers, i.e., the curved (upper) lipid membrane of a secretory vesicle and the planar (lower) lipid bilayer membrane of the plasma membrane. C shows this region in detail, with calcium ions (large red dots) bridging adjacent lipid molecules within (cis) and between (trans) membranes. The path through this transient pore is indicated with the vertical blue double-ended arrow.

In this way, the amount of material released per stimulation would then depend on the following six constraining factors: close physical spacing between vesicle and plasma membrane bilayers; restricted lateral motion of vesicles by SNAREs or similar or related anchoring proteins; localized elevation of calcium levels to a sufficiently high level, i.e., 10⁻⁴ to 10⁻³ M; finite duration of elevation of calcium levels, i.e., 1 ms or slightly longer; volume of vesicle lumen; and internal concentration or level of cell product to be released from a vesicle would occur along a concentration gradient, i.e., mass action.

Such constraints would ensure that the amount of cell product to be released would be limited by physical chemistry, not some imaginative construct. Small amounts of material would be released per pore opening. Increases in amounts to be released would be due to the flow from more than one vesicle. Concerted release of considerable amounts of material (e.g., quanta at the synapse) would require the simultaneous action of arrays of anchored vesicles. No proteinaceous pore molecule need be invoked or conjured (which is especially important as none is observed). The porocytosis mechanism would also ensure that while release of vesicular product was an all-or-none deterministic event and significantly decreased the likelihood of spontaneous release, not all of the vesicle contents would necessarily be discharged. Deviations in the factors noted above would contribute to aberrant physiology or pathology. This interpretation is consistent with the published observations cited above.

A variation of this mechanism could readily achieve constitutive secretion. Creation of the transient pore (i.e., via a localized elevation of calcium level to 10⁻⁴ to 10⁻³ M) would provide a portal across which cytosolic cell product could flow, driven by mass action, during the period of the transient pore being open. In vesicular secretion, one uses a localized significant elevation of calcium level in the immediate vicinity of two juxtaposed membranes. However, formation of a transient pore does not require two juxtaposed bilayers; only one bilayer affected by a localized elevation of calcium level is needed (Figs. 2 and 3). Such an elevation has been demonstrated in the form of spontaneous entry of calcium to the cytosol. It has been demonstrated that such an entry and localized calcium elevation occurs within discernable microdomains at the presynaptic terminal (Llinás et al., 1991, 1992; Silver et al., 1994; Sugimori et al., 1994). Thus, a single mechanism utilizing localized elevation in calcium level at the cytosolic surface of the plasma membrane would result in formation of a transient pore, whose open time, and thus amount of cell product released, would be determined by the period during which the localized calcium level was elevated to the level near 10⁻⁴ M. Reduction in localized calcium level, which is observed to be a rapid process (Silver et al., 1994; Sugimori et al., 1994), would result in a rapid disassociation of calcium from the bilayer and a consequent melting and resealing of the plasma membrane bilayer. In this way, the use of localized interaction of calcium and phospholipids in the bilayer membrane could provide a common mechanism for regulated release of cell product by either vesicular or constitutive secretion.

CRITIQUE AND DISCUSSION

Secretion is a fundamental cell process that provides for transfer of material from inside the cell to its surroundings. Classical studies of neurotransmitter release in motor neurons, chromaffin cells, and other cell types have focused our attention on the vital importance of the spatial relationship between the plasma membrane and the vesicles immediately adjacent to it. Electrophysiological studies have indicated the quantal nature of neurotransmitter release and highlighted the relationship of the amount of neurotransmitter released and the frequency and amplitude of the stimulation. Morphometric studies reveal the high degree of variability in the diameters, and thus the volume of the vesicles and the amount of releasable material that they contain, within populations of secretory vesicles and granules—variability that greatly exceeds the statistical constancy of the end-plate potentials. Recent morphological studies reveal an absence of protein particles that could serve as pores, junctions, or conduits through which neurotransmitters could travel from a vesicle's lumen to the synaptic cleft. Studies of the calcium dynamics that accompany and are required for vesicular secretion show that they occur during 2 ms, and within microdomains, consistent with the dynamics of vesicular secretion. Physical chemical studies and numerical modeling of membrane-membrane interactions all point to the requirement for the high observed calcium levels that accompany vesicular secretion (e.g., 10⁻⁴ to 10⁻³ M calcium). In essence, the reliance of secretion on the dynamics of the interaction of calcium and phospholipids that is the hallmark of porocytosis hypothesis squarely places the secretory process in the realm of the physical and physical chemical interactions of calcium and phospholipids, with mass action as the driving force for release of secretory material. This model mechanism is readily extended to constitutive secretion. It is also clear that methods used to study secretion must satisfy the Nyquist sampling criteria for empirical studies, i.e., that sampling frequency must be three- to fivefold greater than signal frequency (Nyquist, 1928; Shannon, 1949), and theoretical/modeling studies must be data-driven and of adequate granularity to provide the necessary new information and perspectives.

Without meeting these criteria, we risk remaining static in our understanding. The porocytosis hypothesis that we propose satisfies all of the observations, including the latest findings from wet and modeling studies, and provides a framework to integrate our combined knowledge of both vesicular and constitutive secretion.

MODEL MECHANISM

The porocytosis hypothesis considered in light of the information discussed above may be seen in two general variations: a single bilayer (e.g., constitutive secretion, entry of nonenveloped viruses, cell lysis) and two apposed bilayers (e.g., vesicular secretion). We review each of these variations below.

Model Mechanism for a Single Bilayer: Constitutive Secretion

Let us begin by considering of a single phospholipid bilayer and progress through rupture of the membrane and consequent cell lysis (Fig. 4). Under resting conditions, the membrane is in a steady state, during which it exists essentially as a planar bilayer, perhaps with some small degree of undulation. The spacing between the lipids is relatively constant, and individual molecules are otherwise free to spin about their long axis where they exhibit diffusional spreading and can mix freely within the confines of the bilayer (resting membrane, top panel of Fig. 4). The membrane will stay in that apparent equilibrium state so long as no excessive strain is imparted among the lipids.

Figure 4. A sketch showing a model for the four hierarchical steps in the formation of pores in membranes induced by increasing intramembrane strain resulting from the interaction of calcium and lipids. The mechanism would address constitutive secretion, part of vesicular secretion, cell disruption, and for other pore-forming processes in which accessory proteins were not required. The steps progress from (1) resting membrane, (2) transient pore, (3) 2 nm pore, and (4) rupture of membrane. The progress from each step to the next is depicted as a decision based on the level of strain in the bilayer; if strain is increased, then the membrane progresses to the next greater step; if the strain is decreased, then the membrane returns to the previous step. This model is consistent with observations that formation of the 2 nm pore is a stepwise process, having an intermediate between itself and the resting state. It is also consistent with the observations that membrane pores involved in cell lysis are of dimensions greater than 2 nm, typically between 5 and 20 nm in diameter. Calcium ions are shown as red dots.

If increased strain is imparted to the membrane, and specifically within the bilayer with an orientation parallel to the plane of the membrane, then the membrane will deform. If that strain is focused at the sites of interactions between two adjacent lipid molecules, then the spacing between those two lipids can be expected to change. If the force is an attractive force, then the adjacent lipids will be drawn closer; consequently, these lipids will exhibit a decrease in their rotational freedom. If the force is repulsive, then the lipids will be displaced away from spacing between one or both of the attracted lipid molecules, causing an increase from the resting mean spacing.

Calicum and similar multivalent cationic ions can induce such an increase in intramembrane strain through participating in a lipid-calcium-lipid complex, e.g., Gd₃₊ can impart similar changes in erythrocyte membranes with a resulting increase in permeability to ascorbate (Cheng et al., 1998). This would result in formation of a gap in the bilayer, i.e., a pore through which materials could traverse the bilayer, i.e., driven by mass action (transient pore, second from the top panel of Fig. 4).

The transient pore would remain open while calcium levels exceeded 10⁻⁴ M (i.e., 1 ms or less, as is observed). With the rapid reduction of local calcium levels, as is normally observed, the lipids would be freed of their association with calcium and freed to resume their independent rotations, and tension within the bilayer would be reduced to the level of the resting membrane. This would result in a closure of the transient pore and consequent cessation of transmembrane flow of material.

If the local calcium levels remain at 10⁻⁴ M for more than 1 ms, or the calcium levels rise to higher levels for a longer time, then more lipids would associate with more calcium ions, resulting in the freezing of larger patches of membrane and consequently decrease in spacing of more lipids and an increase in the intramembrane tension. Such an increase would result in formation of the observed 2 nm pore, i.e., a pore whose width was equivalent to two lipid molecules.

Local calcium levels would play an essential role in the life of the 2 nm pore as well. Reduction of the locally elevated calcium levels would result in a decrease in intramembrane tension and a decrease in diameter of the pore from 2 to about 1 nm, i.e., to the transient pore state. In this way, the transient pore would serve as an intermediate step in the formation of the 2 nm pore when calcium is plentiful, and to the resting membrane state when local calcium levels are decreased, which is consistent with the observations described above. Similarly, if there is elevation of the calcium levels above 10⁻⁴ for more than 1 ms, or the calcium levels rise to higher levels for a longer time, then more lipids would associate with more calcium ions, resulting in the freezing of larger patches of membrane and consequently decrease in spacing of more lipids and an increase in the intramembrane tension. Such frozen patches would resemble lipid rafts (bottom of Fig. 4).

A further increase in strain would stress the membrane to a point of failure and a consequent rupture of membrane. This is observed in cases of cell rupture by massive influx of calcium into cells in T-cell-mediated nonsecretory cell lysis (Berke, 1994), osmotic shock, apoptosis, etc. Owing to the large diameter of the pores, this could likely be considered an irreversible process.

Model Mechanism for Two Opposed Bilayers: Vesicular Secretion

Let us now consider the case of vesicular secretion across two apposed phospholipid bilayers: one bilayer that is essentially planar in configuration, and a second bilayer being a membrane vesicle of less than 100 nm diameter that is held in very close spatial register with the planar membrane, i.e., a secretory synaptic vesicle held in place with SNARE-syntaxin or SNARE-synaptotagmin protein complexes (Fig. 5). The small diameter of the neurotransmitter-containing vesicle would provide a small contact area between the flat cell membrane and the spherical vesicle, i.e., several square nanometers. The vesicle is anchored in close proximity to the cell membrane by SNARE-containing complexes and functionally similar proteins, thereby ensuring that the vesicle cannot drift away and thus it is best able to convey its contents to the extracellular space.

Figure 5. A sketch showing a model for the five hierarchical steps in the formation of pores of two apposed membranes induced by increasing intramembrane strain resulting from the interaction of calcium and lipids. Calcium ions are shown as red dots. The mechanism would apply to vesicular secretion, i.e., in which accessory proteins were not required. The steps progress from (1) resting membrane, (2) juxtaposed membranes, (3) transient pore (1 nm diameter), (4) 2 nm pore, and (5) hemifusion, fusion, or rupture of membrane. The progress from each step to the next is depicted as a decision based on the level of strain in the bilayer; if strain is increased, then the membrane progresses to the next greater step; if the strain is decreased, then the membrane returns to the previous step. This model is consistent with observations that formation of the 2 nm pore is a stepwise process, having an intermediate between itself and the resting state. It is also consistent with the observations that membrane pores involved in cell lysis are of dimensions greater than 2 nm, typically between 5 and 20 nm in diameter. Accessory anchoring proteins such as SNAREs are not shown.

In this model mechanism, the resting state is one in which the anchored vesicle and cell membranes are near one another (Fig. 5, top panel). The anchored vesicle may then be anchored in register with the plasma membrane through the anchoring action of SNARE, synaptotagnin, syntaxin, or similar proteins. These bilayers would then be close to contact as juxtaposed membranes (Fig. 5, second topmost panel). The membranes may be considered to be in contact except for repulsive forces between the polar head groups of the two membranes (e.g., van der Waals).

Arrival of the action potential at the synaptic membrane results in an opening of voltage-gated calcium channels and an influx of calcium from outside the cell, resulting in an elevation of local calcium levels between 10⁻⁴ M and 10⁻³ M (Fig. 5, middle panel; calcium seen as red dots). Due to the close apposition of the spherical vesicle and planar plasma membranes, the following events would simultaneously occur in both membranes, but be restricted to the limited area of the contact site between the two membranes.

At that calcium level, and in the presence of physiological monovalent cation levels, the calcium is able to bind to the phosphates of the phospholipid and thereby decrease the interlipid distance of the bound lipids that are normally held apart by van der Waals and other forces. This will consequently produce a local increase in tension between the complexed lipids and increase the spacing of complexed and uncomplexed lipids, resulting in formation of a transient pore.

The eventual extent of the size and duration of the pore or passageway, and the subsequent state of the vesicle and plasma membranes, will depend on the duration of the time during which calcium levels were at or above 10⁻⁴ M.

If the levels of calcium were elevated for 1 ms or less, then the vesicle membrane-plasma membrane complex would be transient, forming a transient passageway or pore through which neurotransmitter and other small molecules could flow, driven by mass action, and the vesicle and plasma membranes would retain their separate compositions and identities. Capacitance measurements would show this event as a brief increase in capacitance of the plasma membrane, similar to what is observed experimentally.

If, on the other hand, the levels of calcium were sustained for a period lasting sufficiently more than 1 ms, then the lipids of the two membranes could either rearrange to exhibit hemifusion and eventual full fusion of the membranes, with the consequent expulsion of vesicle contents to the extracellular volume, or incorporate the vesicle membrane into the plasma membrane, with a consequent increase in membrane capacitance. This would not, however, be the only means of increasing membrane capacitance, as we have seen above. A subsequent reduction in membrane capacitance would require the budding of a portion of the plasma membrane or the same amount as the incorporated vesicle lipid, the energetics of which appear to be prohibitive, and which is not observed experimentally with planar membranes or soap bubbles. (Following extensive literature reviews, discussions with colleagues, and direct experimental observations, the authors know of no instance in which small bubbles spontaneously bud off from large bubbles. We therefore welcome any demonstration of such an occurrence.)

The fusion of the two membranes should be a rapid event, driven largely by the asymmetry in intramembrane tension. The amount of extracellular calcium flowing into the vesicle lumen (by mass action) and thus available to form a complex with E-face lipids will result in a nearly uniform increase in intramembrane tension. This would work against the less extensive complexing of calcium and lipid on the P-face due to the lower number of calcium ions available from the increase in intracellular calcium levels; hence, the intramembrane tension on the P-face leaflet would be lower than that in the E-face leaflet. This asymmetry in tension would have the effect of pulling the E-face toward the plane of the plasma membrane, much like a spring, and the vesicle membrane would rapidly become coplanar with the plasma membrane. Reversal of that process, which would be seen as budding, would require a localized reversal of asymmetric intraleaflet leaflet tension. To date, there are no experimental or data-driven theoretical observations that support such a notion, despite the popularity of the results of such a mechanism among contemporary workers.

CONCLUSION

Given the discussion above, we conclude that the porocytosis and vesicle array hypothesis described above best fits all available published data from living cells, excised patches of cells, biomolecular models, and computational models that pertain to secretion of chemical neurotransmitter at the synapse. We believe that porocytosis is also integrally involved in secretion from other vesicle-requiring systems, including chromaffin cells and β-cells.

The porocytosis mechanism would also be integrally involved in, and is fully compatible with, published data regarding the use of fusion proteins needed for enveloped viruses to enter or exit from cells and for cell-mediated killing of foe cells by the immune system. Lastly, as outlined above, the porocytosis mechanism is equally applicable to the process of constitutive secretion.

Thus, porocytosis provides a mechanism for understanding secretion and calcium-dependent membrane-membrane interactions, as well as secretion and other transmembrane transit processes, and brings our attention to the interactions of calcium and lipids, and the first principles of such interactions that must be integral to the membranes and physiology of secretion by cells.

Acknowledgements

Support by NSF MCB 9602056 (to R.B.S.), the U.S. Department of Veterans Affairs (to R.B.S.), and the National Institutes of Health grant DA 01551 (to G.D.P.).