Porocytosis: A transient pore array secretes the neurotransmitter packet

Abstract

The porocytosis hypothesis is based on the arrayed nature of synaptic vesicles which forms the anatomical functional unit of secretion. The presynaptic array and the postsynaptic array of receptors form a synaptomere which is the unit of transmission. A transient increase in calcium ions, triggered by an action potential, activates all pores of the array to pulse transmitter. The array insures transmission while permitting a frequency dependent amount of secretion. Therefore the amount of secretion is variable which permits plasticity. Secretion from the array has the property of immediate synaptic plasticity whereas a change in array size would change synaptic strength. The robust nature of the array insures fidelity of transmission, a frequency dependent dynamic signature of transmission giving the property of immediate plasticity; and, a change in array size yields a change in synaptic strength for long term reliability. Anat Rec (Part B: New Anat) 282B:38–41, 2005. © 2005 Wiley-Liss, Inc.

POROCYTOSIS

We have developed the porocytosis hypothesis of secretion, which makes physiological data compatible with the structural organization of the neuromuscular junction and other synapses. In our hypothesis, the postsynaptic quantal response results from transmitter pulsed from an array of transient pores rather than from a single presynaptic vesicular exocytotic event (Fig. 1) (Kriebel et al., 2000, 2001). In the mechanism we propose, presynaptic vesicles are anchored to the active zone of the plasma membrane and juxtaposed to calcium ion-selective channels by proteins such as SNARE (Sudhof, 2000). Calcium ions resulting from the action potential bind with the lipid bilayers of vesicles and plasmalemma to form an array of transient pores to release some transmitter from each docked vesicle (Silver et al., 2003). The array pulses a standard packet of transmitter, which in turn generates the postsynaptic quantal response. By contrast, the quantal vesicular exocytosis (QVE) hypothesis states that the released packet of transmitter is stored within a single vesicle and that secretion is by the process of exocytosis; after exocytosis, the vesicle membrane is either incorporated into the presynaptic membrane to be recycled by endocytosis (Heuser and Reese, 1974; Zimmermann and Denston, 1977; Zimmermann, 1982) or the vesicle membrane fuses only momentarily (Ceccarelli et al., 1973; Neher, 1993; Bennett and Scheller, 1994; Stevens and Williams, 2000).

Figure 1. Diagrams of arrayed vesicles in a synaptic bouton. A: Presynaptic surface view of docked vesicles held in a hexagonal array by the paravesicular grid (shading between vesicles). The entire array is the functional unit of secretion in that all docked vesicles pulse transmitter when internal calcium is raised by an action potential. This array along with the postsynaptic receptors is termed a “synaptomere.” B: Cross-section of bouton showing a synaptomere during the moment of secretion. The density of stippling represents the concentration of neurotransmitter within vesicles and the amount of transmitter pulsed from each vesicle into the synaptic cleft. A large inward calcium current through voltage-gated calcium channels juxtaposed to the vesicles ensures that the entire array is flooded with calcium ions to activate the secretory pore at each vesicle. Even though the amount of pulsed transmitter from each vesicle varies, the summated amount is standardized by the array. Each flicker of the pore would release only part of the vesicle contents (roughly 8%) (Neher, 1993), so a docked vesicle would secrete repetitively. The immediately available store is represented within all docked vesicles. The transmitter transporter could maintain levels of transmitter in docked vesicles. In addition, a process that exchanges docked with cytoplasmic vesicles would ensure competent docked vesicles. Since the store of cytoplasmic vesicles is large, the array ensures a standard amount of secreted transmitter regardless of which vesicles are exchanged and how much transmitter is within each vesicle of the array.

We propose that the arrayed nature of components composing a secretory site provides an ultrastructural basis for the extensive electrophysiological and statistical data that is incompatible with the QVE hypothesis. We view the “secretory organelle” described by Harlow et al., (2001), which we have called the synaptomere (Kriebel et al., 2000, 2001), to be the unit of secretion. The scaffold of the synaptomere positions vesicles in juxtaposition to the neurolemma so that transient pores connect vesicular contents to the synaptic space, permitting a pulse of transmitter. Consequently, repetitive flickers would pulse a decreasing amount of transmitter. The processes of filling vesicles with transmitter and exchanging docked with the other cytoplasmic vesicles would maintain an immediately available store of transmitter. The transient calcium current into the secretory apparatus would synchronously activate the array of transient pores (Silver et al., 1994). Even though the distribution of the amounts of transmitter pulsed through single pores would be fit with an exponential curve and the amount within docked vesicles could vary from empty to full, the summated characteristics of an array ensure a packet of pulsed transmitter (Falk-Vairant et al., 1996; cf. Dunant and Israel, 1997). Pulsed secretion through an array of transient pores means that the packet of transmitter is not preformed within a single vesicle. The arrayed nature of both secretory sites and receptors ensures fidelity of transmission and guarantees a small variation in the postsynaptic response.

It is well established that calcium ions serve to couple both excitation-secretion and excitation-contraction. In the striated muscle myofibril, the transverse tubule excites multiple triad structures to release calcium into the myoplasm. Calcium ions spread by diffusion into the array of myofilaments to initiate a synchronous twitch throughout all myofibrils composing the sarcomere. During the rising phase of the muscle twitch, calcium ions spread throughout a 50 micron diameter muscle fiber within milliseconds. By comparison, the nerve terminal is only 1 micron in diameter and therefore activation of the secretory apparatus would be expected to occur within a millisecond. The extent of the calcium microdomain is the synaptomere (Silver et al., 1994; Sugimori et al., 1994; Kriebel and Keller, 1999; Silver et al., 2003). In the presynaptic neuromuscular junction, the synaptomere extends 1 micron across the terminal diameter and is 200 nanometers wide. In the frog neuromuscular junction, the secretory unit is the ridge located opposite the postsynaptic junctional fold of the muscle fiber. On each side of the presynaptic fold are double row arrays of calcium channels. The rows are separated by 20 nanometers and the individual calcium channels are spaced 14 nanometers apart. The 200 calcium channels would ensure that the entire synaptomere would be bathed in calcium ions even if only a fraction were activated (cf. Llinás et al., 1992; Silver et al., 1994). The short distances between components of the array would accommodate a 200 microsecond latency between the action potential and secretion.

SYNAPTOMERES AND TRANSMITTER RELEASE SITES

A critical observation at the neuromuscular junction underlying our hypothesis is that the number of synaptomeres (secretory organelles) equals the number of release sites based on physiological analyzes of the end-plate potential (EPP) (del Castillo and Katz, 1954; Quastel, 1997; Kriebel and Keller, 1999). The small variation in EPP amplitudes demonstrates a narrow binomial distribution proving that each of the 200 synaptomeres secretes a given amount of transmitter with few failures. Pulsed secretion from each array would ensure a standard postsynaptic quantal response at each of the 200 sites of the neuromuscular junction. An alternative mechanism of secretion compatible with the QVE hypothesis would require that 1 vesicle from each of the 200 sites (each with 50 docked vesicles) is selected for secretion (Kriebel et al., 2001). However, a selection mechanism has not been defined (Schikorski and Stevens, 1997; Kriebel et al., 2004). Even though each secretory site normally generates one postsynaptic response, failures of secretion can be experimentally induced with a low Ca²⁺, high Mg²⁺ saline. In this experimental condition, Poisson statistics do fit EPP amplitude data when the numbers of secreted packets are reduced to fewer that 10 (del Castillo and Katz, 1954). The normal small observed coefficient of variation (< 3%) (Kriebel and Keller, 2000) of end-plate potentials is consistent with each of the 200 release sites secreting one packet as originally proposed by del Castillo and Katz (1954) and is the basis for our array model. In contrast, release of neurotransmitter via the QVE mechanism where 200 vesicles are selected for exocytosis from 10,000 docked vesicles would result in an end-plate potential coefficient of variation of 14–30%, which would require Poisson statistics (Kriebel and Keller, 1999). Without a defined selection process (Schikorski and Stevens, 1997; Kriebel et al., 2001, 2004), the notion that neurotransmitter release is mediated by a single-quantum, single-vesicle mechanism is precluded. An example of how an array ensures a standard response is found in amplitudes of action potentials recorded at an axon, where there is essentially no variation in amplitudes. In this case, the numbers of sodium ions passing through single pores fit an exponential curve but the summated number of ions through 2000 channels is essentially invariant because the array is so large.

The array of components in the synaptomere—secretory organelle named by Harlow et al. (2001)—fits the physiological data of one site per micron of terminal length at the frog neuromuscular junction. Similarly, in a synaptic bouton, one paravesicular apparatus anchoring docked vesicles would be the functional array. The frog neuromuscular junction (NMJ) has 200 secretory arrays, which secrete 200 packets. By comparison, synaptic boutons secrete one packet of transmitter from each array with few, if any, failures. This observation could be explained within the QVE hypothesis with a selection process as discussed above but the decreasing packet size of about one-third with each action potential during a tetanus can only be explained with the entire array of docked vesicles secreting a percentage of the remaining transmitter with each action potential of the tetanus. Within the confines of the QVE hypothesis with exocytosis of one vesicle per action potential, the 20–40 docked vesicles and hundreds in reserve would require at least 100 action potentials before a decrease in packet size could occur.

The array of transient secretory pores ensures a standard packet with each action potential even though the amount of transmitter pulsed through individual transient pores would vary as a function of transmitter remaining within the given vesicles. Thus, the packet size of transmitter pulsed from the array would be frequency-dependent (Kriebel, 1988). In the NMJ, 60% of the aceylcholine is cytoplasmic. Thus, the transmitter transporter would continuously add aceylcholine to docked vesicles. A vesicular recycling mechanism (Fig. 1) would ensure competency of docked vesicles.

CATION-EXCHANGE HYPOTHESIS

Our model also accommodates the cation-exchange hypothesis proposed by Rahamimoff and Fernandez (1997), who suggested a single pore generates the subminiature end-plate potential [The subminiature end-plate potential is a postsynaptic response 1/10th of the size of a miniature end-plate potential (Kriebel, 1988).] cf. Dunant and Israel, 1997). We simply extend this notion with the porocytosis array model where secretion from the array releases a packet that generates the postsynaptic quantal response (cf. Dunant and Israel, 1997). Most importantly, the dynamic nature of frequency-dependent changes in packet size readily accommodates the wide range of frequency-dependent facilitation processes.

The array offers an explanation to the heretofore unexplained physiological observation that quantal size decreases during a tetanus at both NMJs and other synapses. At the NMJ, the number of secretory packets is sufficiently large that the end-plate potential exceeds threshold in normal physiological frequencies. The NMJ is a one-to-one synapse such that there is no integration. However, at most neuronal synapses, the size of the excitatory postsynaptic potential (EPSP) is relatively small such that spatial and temporal summation of currents are required to reach threshold (exception is the one-to-one synapse of Held). Small EPSPs and inhibitory postsynaptic potentials permit ionic integration in the postsynaptic neuron. Therefore, the array confers a frequency-dependent packet size that yields a signature amount of secretion during different neuronal frequencies. Thus, a given synapse would code different behavioral responses with different frequencies, each having a signature secretory packet size as well as a rate-dependent decrease in packet size reaching a stable size reflecting the dynamic state where the rate of transmitter transported into docked vesicles is stable. Thus, the array permits transmission across the synapse to accommodate different behavioral responses. This property would confer a mechanism for short-term learning because the secreted packet size is mutable.

IMPLICATIONS FOR LEARNING AND MEMORY

The robust nature of the array provides a simple mechanism to increase the strength of a synapse for a basis in learning and memory. The detailed theories of short- and long-term learning and memory provided by Kandel et al. (2000) can be reinterpreted using array theory. The retrograde messenger NO, for example, might cause an increase in the size of the synaptomere to increase the strength of the synapse. An increase in array size would signal not only an increase in the number of docked vesicles but an increase in the titer of proteins forming the paravesicular grid and cytoplasmic enzymes.

In conclusion, application of array theory to secretion at the synapse confers many properties of transmission not achieved with the concept that secretion is immutable. The QVE hypothesis implies that packet size would not change until the docked vesicles and most of the reserve vesicles are depleted. The QVE hypothesis also requires a selection process to ensure that only one vesicle suffers exocytosis without failures and this mechanism has not been defined. By contrast, the array secures transmission across the synapse without failures while conferring plasticity of the secretory response. With different frequencies, the synaptomere secretes signature packet sizes (and thus signature postsynaptic responses) with unique steady states, conferring an analogue aspect of transmission. Therefore, given synapses can be employed in different behavioral responses. Slight changes in afferent synaptic frequencies would produce immediate changes in the size of the packet of secretion. The array thus provides the basis of immediate plasticity, which is necessary for immediate learning. An increase in presynaptic calcium levels and the NO retrograde messenger, which accompany synaptic activity, would increase enzymatic titers, leading to an increase in the array size to form the basis of memory.

Acknowledgements

Supported in part by NIH, DA015511 and SYNTOS-04a.