‘Slow’ and ‘fast’ endocytosis
We have found that in rat corticotrophs, endocytosis is strongly coupled to exocytosis. In this study, the first two or three voltage steps (0.5 s duration) of a train of depolarization typically elevated [Ca2+]i to a peak value of several micromolar. A similar [Ca2+]i elevation has been observed in corticotrophs during AVP or α-adrenergic receptor stimulation where [Ca2+]i rose to several micromolar within 1-2 s (Tse & Lee, 1998; Tse & Tse, 1998). Although the [Ca2+]i elevation induced by AVP or α-adrenergic receptor stimulation was due to intracellular Ca2+ release from IP3-sensitive calcium stores and depolarization-stimulated Ca2+ entry was used in this study, we have previously shown that voltage-gated Ca2+ entry and intracellular Ca2+ release generate a similar spatial Ca2+ gradient near the exocytic sites (Tse & Lee, 2000). Thus the [Ca2+]i near the exocytic sites during a train of depolarizations is comparable to that elicited during AVP or α-adrenergic receptor stimulation. At this moderate [Ca2+]i elevation, the endocytosis following stimulated exocytosis is predominantly the ‘slow’ endocytosis which has a time constant of ≈6 s. Endocytosis with a similar time constant has been described in melanotrophs (4 s; Thomas et al. 1994), chromaffin cells (5-15 s; Smith & Neher, 1997; Engisch & Nowycky, 1998), as well as hippocampal neurons (6 s; Klingauf et al. 1998) and retinal bipolar neurons (2 s; von Gersdorff & Matthews, 1994). Previous studies in pancreatic β cells (Eliasson et al. 1996) and chromaffin cells (Smith & Neher, 1997) have shown that endocytosis rapidly disappears within 20-100 s of whole-cell dialysis. However, we found that endocytosis in corticotrophs persisted even after 15 min of whole-cell dialysis. While there was significant rundown of exocytosis and Ca2+ entry during whole-cell dialysis, neither the efficiency nor the kinetics of endocytosis was affected (Fig. 1C and D). These results show that the maintenance of endocytosis is not dependent on small soluble proteins, and that endocytosis is closely coupled to exocytosis in rat corticotrophs.
For many exo/endocytotic events, ‘slow’ endocytosis retrieved more membrane than was added by exocytosis. This excess membrane retrieval is reflected in an average efficiency of endocytosis that is greater than one (1.2 at [Ca2+]i < 3.0 μm; Fig. 2). It occurred at all [Ca2+]i that stimulated exocytosis, and with low or high amplitudes of exocytosis (Fig. 2B and C). This excess membrane retrieval could be intrinsic to endocytosis. Previous studies with electron microscopy have shown that endocytic vesicles are often attached to the plasma membrane through a 25 nm wide neck (Willingham & Pastan, 1983; Takei et al. 1995). This neck is formed by dynamin, which is activated at high [Ca2+]i (Takei et al. 1995). If this neck consists of plasma membrane, it could contribute to the excess membrane that is retrieved. Assuming the diameter of the endocytic vesicle without the neck is the same as that of the exocytosed granule (200 nm) and the neck is 25 nm wide, the neck would have to be 320 nm long to account for the 20 % excess membrane. Necks as long as 700 nm has been reported (Willingham & Pastan, 1983), so the formation of such necks may account for some of the excess membrane retrieval. In a small percentage of exo/endocytotic events (9/91), an efficiency of endocytosis of 2 or more is observed. It seems unlikely that the formation of a neck alone can account for such large excess membrane retrieval, because the neck would have to be 1600 nm or longer. However, it should be noted that 6 of the 9 events with an efficiency of endocytosis > 2 were stimulated by [Ca2+]i transients larger than 3 μm. It is possible that at high [Ca2+]i, the endocytic machinery also retrieves non-granule membrane directly through the formation of endocytic vesicles. The observation that high [Ca2+]i promotes a larger efficiency of endocytosis may therefore reflect the formation of longer necks that connect the endocytic vesicle to the plasma membrane, as well as the formation of endocytic vesicles from non-granule membrane.
In addition to ‘slow’ endocytosis, ‘fast’ endocytosis also followed exocytosis in some exo/endocytotic events. ‘Fast’ endocytosis is different from ‘slow’ endocytosis, because it is not smooth, but contains step-like decreases in Cm and is typically complete within 2 s (Fig. 7). The size of the step-like Cm decrease can be as large as 75 fF, corresponding to a spherical organelle with 1.5 μm diameter. ‘Fast’ endocytosis probably does not retrieve granule membrane that was exocytosed immediately preceding it, because ‘fast’ endocytosis is always followed by ‘slow’ endocytosis, and the amounts of ‘slow’ endocytosis and exocytosis correlate well with each other. Furthermore, ‘fast’ endocytosis has a higher [Ca2+] requirement (> 10 μm) than both exocytosis and ‘slow’ endocytosis. A similar type of ‘fast’ endocytosis has also been described in rat melanotrophs and bovine chromaffin cells (Neher & Zucker, 1993; Thomas et al. 1994; Smith & Neher, 1997). ‘Fast’ endocytosis has been proposed to retrieve membrane that was added by previous rounds of stimulated exocytosis or spontaneous activities (Thomas et al. 1994; Smith & Neher, 1997). In corticotrophs, there is only a weak correlation between the cell size and the amount of ‘fast’ endocytosis (r= 0.47). However, the amount of ‘fast’ endocytosis in these cells is typically small, so the total added membrane from previous exocytic events might not be very large. Therefore, the possibility that ‘fast’ endocytosis may retrieve membrane that was added by previous exocytic events cannot be ruled out here. Alternatively, ‘fast’ endocytosis may retrieve non-granule membrane that could be involved in the normal turnover of the plasma membrane. As in excess membrane retrieval by ‘slow’ endocytosis, high [Ca2+]i may also promote the retrieval of non-granule membrane through ‘fast’ endocytosis.
Exocytosis and endocytosis
When corticotrophs were stimulated with depolarization pulses, ‘slow’ endocytosis typically followed exocytosis and the amount of ‘slow’ endocytosis correlated well with the amount of exocytosis. Furthermore, no endocytosis occurred if no exocytosis was stimulated. These observations suggest that ‘slow’ endocytosis and exocytosis are closely coupled, such that each exocytosed vesicle leads to an endocytosed vesicle. Such close coupling implies that our Cm measurement is a net result of exocytosis and endocytosis. For our depolarization experiments, the temporal overlap of endocytosis and exocytosis is probably small. In corticotrophs, we have previously shown that exocytosis stimulated via voltage-gated Ca2+ entry was due to the generation of a locally high [Ca2+] near the exocytic sites (Tse & Lee, 2000). Thus, following the termination of the train of depolarizations, the spatial Ca2+ gradient would dissipate quickly and little exocytosis would be expected. Consistent with this, simultaneous measurements of amperometry and capacitance in another neuroendocrine cell type, chromaffin cells, have shown that most exocytosis occurred during the depolarization pulses (Engisch & Nowycky, 1998). From our flash photolysis experiments, we find that exocytosis is more steeply Ca2+ dependent than ‘slow’ endocytosis, and that exocytosis is typically a much faster process than ‘slow’ endocytosis (Fig. 4 and Fig. 5). Nevertheless, at low [Ca2+]i or low rates of exocytosis in our flash photolysis experiments, some exocytosis may overlap with endocytosis. Under this condition, we may have overestimated the time constant of ‘slow’ endocytosis and the Ca2+ dependence of ‘slow’ endocytosis may be less steep than that shown in Fig. 5B. However, this does not affect our conclusion that ‘slow’ endocytosis is mediated via a high-affinity Ca2+ sensor (discussed later).
The temporal overlap of exocytosis and endocytosis would affect most strongly the measurement of exocytosis during low [Ca2+]i elevation in our flash photolysis experiments. Figure 4 shows that at lower [Ca2+]i, exocytosis appears to stop, even though [Ca2+]i is still elevated and the readily releasable pool of granules has not been exhausted. This phenomenon can be explained by a close coupling of endocytosis to exocytosis, such that exocytosis and endocytosis of individual vesicles are occurring consecutively. Simulation of a kinetic model with coupled exocytosis and endocytosis shows that cumulative exocytosis can be expected to increase with faster rates of exocytosis and that the time to the peak of exocytosis will also shorten with faster rates of exocytosis (Fig. 6a and B). This model is similar to the ‘kiss-and-run’ model that has been proposed for synaptic vesicle recycling (Palfrey & Artalejo, 1998). Like the ‘kiss-and-run’ model, our model assumes that the same exocytosed vesicle is endocytosed. However, unlike the ‘kiss-and-run’ model, endocytosis in corticotrophs is slow (time constant of 3-6 s). The slow kinetics implies that the exocytosed vesicle in corticotrophs is exposed extracellularly for several seconds. Since the dissolution of the dense core in prolactin-containing granules requires ≈3 s (Angleson et al. 1999), this time interval may allow the dissolution of the dense core in the ACTH-containing granules and thus a more complete release of ACTH. Nevertheless, this time interval is still brief for the mixing of granule and plasma membrane proteins (Thomas et al. 1994). The close coupling between endocytosis and exocytosis therefore not only ensures efficient retrieval of vesicle membrane but also preserves the integrity of the vesicle and plasma membrane.
Ca2+ dependence of endocytosis
Endocytosis has been reported to be both [Ca2+]i independent (Ryan et al. 1996; Wu & Betz, 1996) and [Ca2+]i dependent (Artalejo et al. 1995; Smith & Neher, 1997; Engisch & Nowycky, 1998). In nerve terminals, high [Ca2+]i has been reported to inhibit endocytosis (von Gersdorff & Matthews, 1994; Hsu & Jackson, 1996; Cousin & Robinson, 2000). In corticotrophs, high [Ca2+]i clearly does not inhibit endocytosis, as endocytosis occurs even when [Ca2+]i > 10 μm (Fig. 4a and Fig. 5a). In corticotrophs, endocytosis can also occur at resting [Ca2+]i (< 300 nm). In depolarization experiments, the recovery time course of the stimulated [Ca2+]i transient is three times faster than that of endocytosis, so significant endocytosis occurs after [Ca2+]i has returned to resting levels. Endocytosis therefore does not require high [Ca2+]. Nevertheless, we found that τendo decreased by ≈3-fold when [Ca2+]i increased from 2 to 14 μm (Fig. 3C and Fig. 5B), suggesting that ‘slow’ endocytosis is dependent on Ca2+. Furthermore, the Ca2+ dependence of endocytosis is also suggested by the Ba2+ experiment which shows that substitution of Ba2+ for Ca2+ slows down the kinetics of ‘slow’ endocytosis ≈2-fold (Fig. 8a). Our observation that high [Ca2+]i promotes a larger efficiency of endocytosis (Fig. 2B) also supports the suggestion that endocytosis is Ca2+ dependent. The above observations are consistent with the notion that endocytosis may involve a Ca2+ sensor with a high Ca2+ affinity (Cousin & Robinson, 2000).
It has been proposed that the Ca2+ sensor for endocytosis has a high Ca2+ affinity - a few hundred nanomolar (Lai et al. 1999; Cousin & Robinson, 2000). For endocytosis with such a high Ca2+ affinity, endocytosis would appear to be relatively [Ca2+]i independent at the high level of [Ca2+]i (in the micromolar range) that triggers exocytosis, because the Ca2+ sensor for endocytosis would be saturated at this high [Ca2+]i. A high Ca2+ affinity for endocytosis will also allow endocytosis to occur at resting [Ca2+]i (< 300 nm). Thus our experimental results would be consistent with a high Ca2+ affinity for endocytosis. Direct testing of this proposal will involve examining exo/endocytosis at lower [Ca2+]i (< 1 μm). Unfortunately, in corticotrophs, significant exocytosis could be detected only when depolarization raised [Ca2+]i to > 1 μm (Tse & Lee, 2000). Therefore, in this study, we numerically simulated endocytosis (see Methods) with an EC50 for [Ca2+]i set at 500 nm for a [Ca2+]i transient with a decay time constant of 2.3 s and peak [Ca2+]i of 2 μm. With these parameters, our simulation generated an endocytosis with a time constant of 6.0 s, which is similar to the measured value of 6.4 s at 1.9 μm[Ca2+]i (Fig. 3C) in our depolarization experiments. For a peak [Ca2+]i of 5 μm, the simulation generated an endocytosis with a time constant of 4.2 s which is also close to the measured value of 5.2 s at 5.0 μm (Fig. 3C) in our depolarization experiments. We have suggested earlier in the Discussion that there is little temporal overlap between exocytosis and endocytosis in our depolarization experiments. If there is indeed some temporal overlap, the time constant of endocytosis will be smaller and a Ca2+ sensor of higher affinity (< 500 nm) will be needed to simulate the time course of endocytosis. Nevertheless, endocytosis with a Ca2+ sensor of high Ca2+ affinity could account for the small decrease in the time constant of endocytosis with increasing [Ca2+]i that we measured in our depolarization experiments (Fig. 3C). As for our flash photolysis experiments, the range of [Ca2+]i (8-15 μm) in those experiments would almost completely saturate the Ca2+ sensor for endocytosis, so the time constant of endocytosis should only decrease slightly with increasing [Ca2+]i (Fig. 5B). Therefore, endocytosis in corticotrophs is not [Ca2+]i independent, but instead involves a Ca2+ sensor with high Ca2+ affinity.
The calmodulin-calcineurin-dynamin cascade has been proposed as the mediator between Ca2+ and endocytosis (Liu et al. 1994; Artalejo et al. 1996; Cousin & Robinson 2000). Ca2+ activates calmodulin which stimulates calcineurin. Calcineurin then dephosphorylates dynamin which leads to the formation of endocytic vesicles. We investigated the role of calmodulin in this cascade by substituting Ba2+, a poor activator of calmodulin, for Ca2+ and by intracellular dialysis of calmidazolium, a calmodulin inhibitor. Neither of these manipulations inhibited endocytosis. Ba2+ could clearly support endocytosis, although endocytosis became slower, while calmidazolium had no significant effects on endocytosis (Fig. 8a and B). An alternative to calmodulin as the Ca2+ sensor is the Ca2+-dependent interaction between calcineurin and dynamin (Lai et al. 1999). The EC50 of 0.1-0.4 μm of this interaction is consistent with a high Ca2+ affinity. This interaction is not dependent on the catalytic activity of calcineurin, which is also consistent with our observation that endocytosis is not affected by the calcineurin inhibitors cyclosporin A or calcineurin autoinhibitory peptide. However, this interaction is thought to bring calcineurin into proximity with dynamin and other proteins in the endocytic complex, where it and possibly other phosphatases can dephosphorylate these proteins and thus maintain a functional endocytic complex.
The focus of the Ca2+ activation cascade is dynamin, because dynamin is necessary for endocytosis in neuronal and non-neuronal cells (van der Bliek, 1999). Dynamin is a GTPase that can self-assemble into a ring-like structure which forms the narrow neck that connects an endocytic vesicle to the plasma membrane (Hinshaw & Schmid, 1995; Takei et al. 1995). Dynamin has also been proposed to directly detach the endocytic vesicle from the plasma membrane or to coordinate another molecule for the same function (Sever et al. 1999). Dynamin is activated by dephosphorylation which might belie the reason why substitution of ATP with ATP-γ-S in our pipette solution did not have a significant effect on endocytosis (van der Bliek, 1999). However, substitution of GTP with GDP-β-S in our pipette solution also failed to disrupt endocytosis in corticotrophs. In contrast, under similar experimental conditions, GDP-β-S completely abolished the activation of AVP or the α-adrenergic receptor-coupled G-protein (Tse & Lee, 1998; Tse & Tse, 1998). Since dynamin-GTP is the active form of dynamin, it is possible that GDP-β-S cannot easily displace GTP that is already bound to dynamin. Furthermore, the dynamin GTPase activity is only enhanced during endocytosis, so additional GTP may not be required until endocytosis has exhausted the pool of dynamin-GTP (Sever et al. 1999).
In summary, we have found that endocytosis in corticotrophs is closely coupled temporally and probably spatially to exocytosis. This close coupling is reflected in the good correlation between endocytosis and exocytosis and the persistence of this correlation even after prolonged whole-cell dialysis. Our experimental results are compatible with a slow ‘kiss-and-run’-like model wherein each exocytosed vesicle is eventually endocytosed. We also found that endocytosis involves a Ca2+ sensor with high Ca2+ affinity (EC50≈500 nm), which explains the weak [Ca2+]i dependence at [Ca2+]i levels that stimulate exocytosis, and the continuation of endocytosis at resting [Ca2+]i. With these characteristics for endocytosis, a small [Ca2+]i elevation in rat corticotrophs such as that occurring during CRH stimulation would not only stimulate exocytosis but also ensure efficient endocytosis and thus the sustained release of ACTH.