Endoplasmic Reticulum- and Plasma-Membrane-Driven Calcium Oscillations

Chapter 3
Endoplasmic Reticulum- and Plasma-Membrane-Driven Calcium Oscillations


Arthur Sherman


Laboratory of Biological Modeling, NIDDK, National Institutes of Health, USA


3.1 Introduction


Calcium (c03-math-0001) is of great importance throughout biology because it regulates many processes. In neuroendocrine cells, it plays a central role as the main trigger of hormone and peptide secretion. In the experimental literature, one often finds statements along the lines of “Agent X increases c03-math-0002 concentration (c03-math-0003), which increases secretion”. However, secretion is generally not controlled by a mere rise and fall of c03-math-0004 concentration, but rather by c03-math-0005 oscillations, and “Agent X” works by changing the pattern of those oscillations.


One of the key contributions of theorists has been to develop models that explain the mechanisms behind such oscillations and how they are regulated. Indeed, there are multiple mechanisms, so it is helpful to distinguish their origins and how they are regulated. In particular, we find oscillations that reflect variation in c03-math-0006 influx and arise from ion channels at the plasma membrane (PM), as well as oscillations that reflect release from internal stores and arise from ion channels in the membrane of the endoplasmic reticulum (ER). A particularly interesting scenario, where the analytical power of models really shines, is when both mechanisms are present in the same cell and interact with each other.


Our objectives are limited. We will consider simplified examples that illustrate the basic principles rather than aim to model in detail any particular system. An understanding of these principles will provide good preparation to tackle particular problems in the future. Our focus is cellular; other chapters will address systems level models that involve multiple cell types and organs. We also limit ourselves to temporal phenomena, modeling cells as spatially homogeneous. This is often adequate, sometimes not. Mathematically, this means that we only need to deal with ordinary differential equations (a single independent variable, time), rather than partial differential equations (time plus one or more spatial dimensions.) We also treat only deterministic systems, which leaves out the sometimes important effects of noise, but allows us to make good use of dynamical systems tools, such as phase planes and bifurcation diagrams (BDs).


3.2 Calcium balance equations


3.2.1 Derivation


The fundamental physical principle needed to model calcium dynamics in cells is the conservation of mass, which is determined by flux across boundaries. In physics, flux is the rate of flow of some quantity per unit area (http://en.wikipedia.org/wiki/Flux). As we will be dealing with small round cells with a fixed boundary, it is convenient to multiply implicitly by the cell surface area, giving just quantity per time, symbolized by c03-math-0007. Our goal is to calculate the changes in the cytoplasmic or intracellular c03-math-0008 concentration, c03-math-0009, denoted here more simply as c03-math-0010, which means that we must divide by the cytosol volume to convert the rate of flux to the rate of change of concentration. The cytosol volume can be taken to be fixed, so that we can further simplify by absorbing the volume into the flux; we denote this scaled flux by c03-math-0011.


Cells can often (but not always) be well approximated with two compartments: one for the cytosol and one for the ER, as shown in Figure 3.1. The cytosolic c03-math-0012, c03-math-0013, then satisfies the equation



where we are using the volume-scaled fluxes.

Image described by caption.

Figure 3.1 Calcium balance. c03-math-0023, influx through plasma membrane; c03-math-0024, efflux through plasma membrane; c03-math-0025, influx through ER membrane; c03-math-0026, efflux through ER membrane.


The factor c03-math-0015 accounts for the fact that most of the c03-math-0016 that enters a cell is buffered, that is, complexed with c03-math-0017-binding proteins. Typically, only about 1% of the c03-math-0018 ions that enter remain free, which means that the rate of increase of the free c03-math-0019 concentration is only 1% of what it would be in the absence of buffering. This keeps the free concentration low and prevents runaway activation of the many reactions that are regulated by c03-math-0020. This form of the equation is valid only because the buffering reactions are much faster than c03-math-0021 fluxes, so that it is safe to assume that free and bound c03-math-0022 are in the quasi-steady state. (A more complete derivation of this can be found in Sherman et al. (2002)).


The equation for ER c03-math-0027, c03-math-0028, can be similarly written as



Note that the ER gets its own fraction of free c03-math-0030, which may differ from that in the cytosol. Also, because we absorbed the cytosolic volume into the fluxes, we have to multiply by the cytosolic-ER volume ratio to get the correct effect of the flux on c03-math-0031. The ER volume is much smaller than that of the cytosol, so that the concentration change due to the flux will be much greater in the ER. However, the ER c03-math-0032 concentration is much greater, so the most typical case is small relative changes in the ER that result in large relative changes in the cytosol. We can make Equation (3.2) look much better, and reduce the number of parameters from 4 to 2, by multiplying the right-hand side by c03-math-0033:



where


equation

3.2.2 Applications


We will flesh out the c03-math-0036 equations (3.1) and (3.3) and incorporate them into more complex models shortly, but pause here to show that even these simple equations can be used to draw nontrivial conclusions.


We choose the following expressions for the fluxes:






where PMCA is the “plasma-membrane calcium ATPase,” which pumps c03-math-0058 out of the cell, and SERCA is the “sarco-endoplasmic reticulum calcium ATPase,” which pumps c03-math-0059 into the ER. We assume that flux out of the ER is a passive, diffusive leak. Equations (3.4) and (3.5) correspond to Equation (2.15), Chapter 2, this volume, for bulk cytosolic c03-math-0060, except that the plasma-membrane calcium current is for the moment constant, not voltage dependent. In general, the pump expressions are not linear, but rather saturating Hill functions, but these are not needed for the moment. There may also be multiple plasma-membrane c03-math-0061 currents (e.g., L-type and T-type currents, as described in Chapter 2, Equations (2.6) and (2.7)), multiple ER c03-math-0062 channels (e.g., IP3 and Ryanodine receptors), and more than one c03-math-0063 efflux component (e.g., the c03-math-0064c03-math-0065 exchanger in addition to PMCA).


In Figure 3.2, the black curves show the effect of partial (50%) block of the SERCA pump, such as with thapsigargin, by cutting c03-math-0066 in half. For simplicity, we assume that the block becomes fully effective instantaneously at c03-math-0067 min. Then, c03-math-0068 rises to a peak and then recovers to baseline. One often hears or reads something like “thapsigargin was used to increase c03-math-0069,” but the simulation shows that such an increase is only transient because the c03-math-0070 released from the ER to the cytosol is pumped out of the cell, leaving the ER depleted (in proportion to the inhibition of SERCA) but the cytosol at its original level once equilibrium is re-established. It is simple to show that the rise in c03-math-0071 must be transient: setting Equation (3.3) to steady state gives


equation

and substituting this into Equation (3.1) and setting to the steady state implies that


equation

Thus, the ER terms drop out of the c03-math-0074 equation, and the steady-state value of c03-math-0075 depends only on the parameters of c03-math-0076 and c03-math-0077. We conclude that c03-math-0078 cannot be affected by the change in c03-math-0079. The same would be true if we had emptied the ER by increasing efflux, which could correspond to the activation of IP3 or Ryanodine receptors. This is in general enough that it is worth stating as a theorem that holds independent of our choices for the fluxes:

Image described by caption and surrounding text.

Figure 3.2 Illustration of the steady-state calcium theorem (Theorem 3.2.1) using the passive model, Equations (3.1) and (3.3), with fluxes as in Equations (3.4)–(3.7). c03-math-0037, cytosolic c03-math-0038; c03-math-0039, ER c03-math-0040. Parameters: c03-math-0041M/pC; c03-math-0042 pA, c03-math-0043 pA (red), 0 (black); c03-math-0044 sc03-math-0045; c03-math-0046 sc03-math-0047, reduced to 0.2 sc03-math-0048 at c03-math-0049 min; c03-math-0050 sc03-math-0051; c03-math-0052; c03-math-0053.


A maintained change in c03-math-0083 is illustrated by the red curve in Figure 3.2. In that simulation, c03-math-0084 is increased in magnitude at the same time that the PMCA is inhibited; this could correspond, for example, to a store-operated current (SOC) activated by the reduction in c03-math-0085 (Hogan et al., 2010). The combination of the two effects results in an increase in c03-math-0086 at the steady state. Conversely, if one observes a steady state increase in c03-math-0087 when SERCA is inhibited in an experiment, one can be sure that a plasma-membrane ion current was affected (though not necessarily SOC).


The inability of c03-math-0095 to affect steady state c03-math-0096 is not due to the depletion of the ER (indeed, we deliberately only depleted it partially in Figure 3.2 to emphasize this point), but to the fact that c03-math-0097 is constant. The reader may check this explicitly by solving for c03-math-0098 in terms of c03-math-0099 in Equations (3.1) and (3.3) using the expressions in Equations (3.4)–(3.7).


We now illustrate that Theorem 3.2.1 does not hold when the system is not in steady state. In Figure 3.3, the black curves show the response to a train of square pulses of c03-math-0100; this represents roughly the effect of a train of action potentials. The red curves show the response when SERCA is partially inhibited. The pulses of c03-math-0101 are larger in amplitude. This happens because more of the c03-math-0102 that enters with each pulse stays in the cytosol instead of going into the ER. Theorem 3.2.1 does not apply because c03-math-0103 fluctuates too rapidly for the system to reach steady state during each pulse.

Image described by caption and surrounding text.

Figure 3.3 Depleting ER c03-math-0088 (c03-math-0089) increases amplitude of cytosolic c03-math-0090 (c03-math-0091) response to pulses. Black: control, Red: SERCA (sarco-endoplasmic reticulum calcium ATPase) pump inhibited 50%. Equations and parameters as in Figure 3.2 but with c03-math-0092 pulsed between c03-math-0093 and c03-math-0094 pA every 30 s.


Note that there are two time constants evident in Figure 3.3: with each pulse, c03-math-0104 jumps up rapidly and then rises slowly. The rapid jumps reflect the fast intrinsic kinetics of c03-math-0105, whereas the slow rise reflects the slow kinetics of c03-math-0106, which pushes c03-math-0107 up slightly with each pulse as the ER fills. The slow kinetics of the ER are also manifest in the slow rise (black) and slow fall (red) of the envelope of c03-math-0108 as the ER fills or empties. The mean trend of c03-math-0109, however, averaged over the rapid fluctuations, does approach approximately the same level irrespective of the fact whether the pump is inhibited or not. That is, Theorem 3.2.1 applies approximately to the averaged system.


3.3 ER-driven calcium oscillations


3.3.1 Closed cell


The model we have considered so far has a very limited repertoire because it is linear. There is a unique steady state, and the most the system can do is relax to that steady state exponentially or as a sum of exponentials. Put in other terms, the model has only negative feedback, embodied in the SERCA and PMCA pumps, which suppresses any fluctuation of c03-math-0110 away from rest. There is a strongimperative to keep c03-math-0111 stable, but some cells have evolved the ability to generate and profit from controlled large excursions away from rest. In order to obtain more dynamic responses, we need to make the model nonlinear, specifically to add positive feedback. We do this by postulating that c03-math-0112 increases the flux out of the ER, often referred to as “calcium-induced calcium release” (CICR). Moreover, the negative feedback, which is needed to limit and terminate the excursions, has to develop slowly enough that it does not cancel out the rise in c03-math-0113 too soon. We have already encountered examples of oscillations mediated by fast positive feedback combined with slow negative feedback in Chapter 1, McCobb and Zeeman, and Chapter 2, Bertram et al., where they were the bases of action potentials. Indeed, the model we introduce here has much in common with those electrical oscillators, though it works via c03-math-0114 rather than the membrane potential.


IP3 regulation of ER c03-math-0115


A simple example of cytosolic c03-math-0116 oscillations is the model of Li and Rinzel (1994), which is a simplification of the model of DeYoung and Keizer (1992). This will serve as an exemplar of a wide class of models based on the IP3 receptor (IP3R), an internal receptor located on the ER membrane, which opens in response to the second messenger IP3. IP3 in turn is produced by the activation of a G-protein-coupled receptor (GPCR) on the plasma membrane that responds to an external signal, such as acetylcholine or GnRH; the GPCR is coupled to phospholipase C via Gc03-math-0117, which forms IP3 from phosphatidylinositol 4,5-bisphosphate (PIP2). The IP3R is moreover a ligand-gated c03-math-0118 channel; IP3 opens the channel and allows c03-math-0119 to flow down its concentration gradient from the ER to the cytosol. Thus, one way to introduce positive feedback is to assume that IP3 production is stimulated by rises in c03-math-0120. This was the basis of the earliest models for c03-math-0121 oscillations (Meyer and Stryer, 1988; Swillens and Mercan, 1990). DeYoung and Keizer, in contrast, motivated by reports that c03-math-0122 could oscillate even when IP3 is held fixed, proposed that c03-math-0123 enhanced the activity of the IP3R directly. Both effects of c03-math-0124 are known to exist, and there has been a long, yet unresolved debate about which is more important, or whether they each occur in different types of c03-math-0125 oscillations. We will not enter that debate here, but simply use the hypothesis of DeYoung and Keizer as a learning tool.


Modeling the role of IP3


Bezprozvanny et al. (1991) showed that the steady state open probability of the IP3R was a bell-shaped function of c03-math-0126, which suggested that c03-math-0127 in high concentrations inactivates the channel. DeYoung and Keizer proposed that this inactivation provides negative feedback beyond that of SERCA to terminate the c03-math-0128 spike, and that its time scale controls the period of the oscillation. They developed a model with eight states, for all the combinations of activating c03-math-0129, activating IP3 and inactivating c03-math-0130, bound or unbound.


Li and Rinzel (1994) simplified this in the following formula:



where


equation

c03-math-0133 is the concentration of IP3, treated as a parameter; and c03-math-0134 is the fraction of available IP3 receptors, that is, the fraction not inactivated by binding c03-math-0135. c03-math-0136 satisfies the differential equation



where


equation

Equations (3.8) and (3.9) together say that the IP3R open probability is for fixed c03-math-0139 an increasing function of [IP3] and c03-math-0140, but decreases over time as receptors inactivate. This was deliberately formulated to highlight its similarity to the conductance of the voltage-dependent c03-math-0141 channel, described in Chapter 2, Equation (2.1), which increases with the membrane potential c03-math-0142 but inactivates over time.


We can now assemble the pieces to specify the Li–Rinzel model, taking Equations (3.1) and (3.3) with Equation (3.8) in place of Equation (3.7) and


equation

in place of Equation (3.6) (not strictly necessary but customary and more accurate). The equations for c03-math-0144 and c03-math-0145 then become




Finally, we add Equation (3.9) for receptor inactivation.


This system with three dependent variables (c03-math-0148, c03-math-0149, and c03-math-0150) can be simplified by assuming that the cell is closed, that is, that the fluxes across the plasma membrane, c03-math-0151 and c03-math-0152 are 0. This serves two purposes. First, it emphasizes that the oscillations only require flux in and out of the ER. Second, it reduces the number of equations to 2. The total c03-math-0153 content of the cell is



where c03-math-0155 takes into account the differences in ER and cytosolic volume and buffering. This quantity is constant when the cell is closed: its time derivative is 0, as can be verified by multiplying Equation (3.3) by c03-math-0156 and adding to Equation (3.1). This reduction to two equations now allows us to use phase planes, shown in Figure 3.4, to understand the behaviors. (The concept and usage of phase planes are explained in Chapter 1, McCobb and Zeeman.)

Image described by caption and surrounding text.

Figure 3.4 Phase planes (a) and timecourses (b) for the closed-cell Li–Rinzel model, Equations (3.9)–(3.11). c03-math-0157, fraction of IP3 receptors that are available to be opened (i.e. are not inactivated). Oscillations exist for an intermediate range of IP3 concentrations and, within that range, frequency increases with IP3. Parameters: c03-math-0158 as specified in the panels; c03-math-0159; c03-math-0160M/s; c03-math-0161 1/s; c03-math-0162M; c03-math-0163M; c03-math-0164M/s; c03-math-0165M; c03-math-0166 1/c03-math-0167Ms; c03-math-0168M; c03-math-0169.


We solve the model equations for c03-math-0170 and c03-math-0171 (but can recover c03-math-0172 from Equation (3.12) should the need arise; see Figure 3.6). The left column shows the phase planes for three values of [IP3], and the right column shows the corresponding timecourses. The c03-math-0173 nullcline (blue) is obtained by setting the right-hand side of Equation (3.9) to 0 and solving for c03-math-0174 as a function of c03-math-0175. It decreases monotonically, reflecting the increase of inactivation with c03-math-0176 (see Equation (3.9)).


The c03-math-0177 nullcline (red) is obtained by setting the right-hand side of Equation (3.1) to 0. Conceptually, we want to solve for c03-math-0178 as a function of c03-math-0179, but that equation can have up to three solutions for some values of c03-math-0180, so it is easier to solve for c03-math-0181 as a function of c03-math-0182. The c03-math-0183 nullcline is, thus, defined by


equation

where we have neglected the small terms c03-math-0185 and c03-math-0186 in the second line (the latter is valid as long as c03-math-0187 is not too small). The resulting curve has a distorted N shape, increasing for small c03-math-0188, then decreasing, and then increasing again. The left branch of the c03-math-0189 nullcline has a steep slope because c03-math-0190 as c03-math-0191. The N shape results from the competition between activation, c03-math-0192, of the receptor, which increases c03-math-0193, and SERCA, c03-math-0194, which refills the ER. The general trend is for c03-math-0195 to increase due to SERCA, but in the range where the IP3R is strongly activated, c03-math-0196 turns down. This is typical for activator–inhibitor systems, where the positive feedback mechanism creates a kink in the activator variable nullcline and thereby creating instability and oscillations. In Hodgkin–Huxley-type systems, the activation of the c03-math-0197 current plays the same role. In that context, it creates a dip in the current–voltage relation, which translates to a negative resistance and gives away its role as the source of dynamism.


The intersection of the two nullclines is the steady state, and in each row there is only one. Increasing IP3 pulls down the c03-math-0198 nullcline because c03-math-0199 is an increasing function of c03-math-0200 and moves the steady state along the c03-math-0201 nullcline. In the top row, c03-math-0202M, the steady state is on the rising left branch of the c03-math-0203 nullcline, and is therefore stable. The trajectory (black) swings around from the upper right to end at the steady state, producing a single transient spike and then steady c03-math-0204, as shown in the right panel. In this state, the system is not oscillatory but is said to be excitable, meaning that a suitable perturbation away from the rest state is amplified into a spike rather than returning immediately to rest.


For c03-math-0205M, the steady state is on the rising right branch of the c03-math-0206 nullcline and is again stable, but the approach to rest is a damped oscillation. In order to get sustained oscillations, the steady state must be destabilized, which is guaranteed if it lies on the falling middle branch of the c03-math-0207 nullcline. This is the case for c03-math-0208M, shown in the middle panels. The trajectory now takes the form of a closed curve (a limit cycle, see Chapter 1, McCobb and Zeeman) that orbits around the steady state.


The phase plane also explains a characteristic feature of the oscillation, namely that the spikes are narrow and the interspike intervals are long. This happens because the trajectory is closer to the unstable steady state during the interspike portion, and the flow is therefore slow. The period decreases markedly as IP3 is increased within the oscillatory range because the steady state moves to the right, away from the low c03-math-0212 values traversed during the interspike period (compare c03-math-0213M to c03-math-0214M in Figure 3.4). Thus, it is not difficult for models like this to capture the frequency encoding observed experimentally as the IP3-generating stimulus is varied (Goldbeter et al., 1990).


If we stack a series of phase planes like those in Figure 3.4 and plot the values of c03-math-0215 (the steady states as well as the minima and maxima of the periodic orbits when those are present), we get Figure 3.5. A similar picture would be obtained if c03-math-0216 were plotted against IP3, so we omit the third dimension. The line shows the steady states, solid and black for stable, dashed and red for unstable. The green dotted line shows the average c03-math-0217 during spiking, and the filled circles the amplitude of the oscillations, which exist for a range of values of IP3, or between lower and upper thresholds. Those thresholds are labeled “HB” for Hopf bifurcation, which indicates the particular way in which the steady state goes unstable (see Chapter 1, McCobb and Zeeman). The existence of upper and lower thresholds, dividing the system behaviors into off, oscillating and fully on, is a ubiquitous feature of activator–inhibitor systems.

Image described by caption and surrounding text.

Figure 3.5 Bifurcation diagram for the closed-cell Li–Rinzel model. Oscillations in cytosolic c03-math-0209 (c03-math-0210) exist for an intermediate range of IP3, demarcated by Hopf bifurcations (HB). Black: stable; lines: steady states; filled circles: min and max of oscillation; red: unstable steady states; green: average cytosolic c03-math-0211 during oscillations. Equations and parameters as in Figure 3.4

Image described by caption and surrounding text.

Figure 3.6 In the open-cell Li–Rinzel model (Equations (3.9), (3.13), (3.14)), c03-math-0222 influx is needed to sustain oscillations. Parameters as in Figure 3.4 except: c03-math-0223M; c03-math-0224M/s; c03-math-0225M; c03-math-0226M; c03-math-0227. We also add the plasma-membrane parameters: c03-math-0228/s; c03-math-0229M; and c03-math-0230, which is initially 2.5 c03-math-0231M/s and at c03-math-0232 s (red arrow) reduced to 0.


The peak c03-math-0218 of the oscillation is in most cases higher than the steady c03-math-0219 level when c03-math-0220 is above the upper threshold, but the average is lower. The model thus suggests that if one could measure c03-math-0221 experimentally but not IP3, then the average would be a better indicator of the IP3 level (or “activity”) than the peak.


3.3.2 Open cell


Suppose now that the cell is open. Now during each c03-math-0233 spike, some of the c03-math-0234 will be pumped out of the cell rather than back into the ER. Unless there is some c03-math-0235 influx to balance the efflux, the ER would eventually empty out and oscillations would cease. We restore the plasma-membrane fluxes, setting c03-math-0236 to a constant and introducing a saturating PMCA pump:


equation

Stay updated, free articles. Join our Telegram channel

Oct 27, 2017 | Posted by in ENDOCRINOLOGY | Comments Off on Endoplasmic Reticulum- and Plasma-Membrane-Driven Calcium Oscillations

Full access? Get Clinical Tree

Get Clinical Tree app for offline access