A wide range of tyrosine concentrations, 39-180 μM, have been measured in serum in infants and adults 4950, with means near 100 μM. In our model we take the serum concentration to be btyr = 97 μM. In the model experiments described in Results A, this concentration varies throughout the day due to meals but averages 97 μM. Tyrosine is transported from the serum across the blood-brain barrier (BBB) to the extracellular space and from there into the neuron. We simplify this two-step process into a single step from the serum into the neuron with velocity V_TYRin and assume that the kinetics are those of the neutral amino acid transporter across the BBB. The K_m of the transporter has been measured as 64 μM 51 and we take V_max = 400 μM/hr, so

If btyr has its average value of 97 μM, then V_TYRin = 244 μM/hr, which corresponds almost exactly to the 4 μM/min reported in 51 for the import of tyrosine into the brain.

Intracellular tyrosine is used in a large number of biochemical and molecular pathways and is produced by many pathways 52. Over 90% of the tyrosine that enters the intracellular pool of the brain is used in protein synthesis 535455 and even in the striatum a relatively small fraction is used for dopamine synthesis 55. To represent all of the other products and sources of tyrosine, we will use a single variable tyrpool, and assume that it exchanges linearly with the tyrosine pool:

We choose the rate constants k₁ = 6 μM/hr and k_-1 = 0.6 μM/hr so that tyrpool is approximately 10 time larger than tyr. As we will see below, with this choice, about 10% of the imported tyrosine goes to dopamine synthesis and the steady state tyrosine concentration is 126 μM in the model, well within the normal range of 100-150 μM 56. The importance of tyrpool is that, without it, all imported tyrosine would have to go to dopamine in the model. Not only would that be incorrect physiologically, but dopamine synthesis would be extremely sensitive to tyrosine import, which it is not 575856.

Tyrosine (tyr) and tetrahydrobiopterin (bh4) are converted by tyrosine hydroxylase (TH) into 3,4-dihyroxyphenylalanine (l-dopa) and dihyrobiopterin (bh2). The velocity of the reaction, V_TH, depends on tyr, bh4, cytosolic dopamine (cda), and extracellular dopamine (eda) via the autoreceptors:

The third term (on the right side of the equation) is simply Michaelis-Menten kinetics including the inhibition of TH by cda which competes with bh4 35923. Values for the rate constants and references are given in Table 2. The first term (on the right) is substrate inhibition of the enzyme by tyrosine itself 23. A range of values for K_i(tyr), 37-74 μM, was found in 60. We have computed K_i(tyr) = 160 μM directly from the data in figure 2 of 23. The number 0.56 in the numerator is chosen so that at steady state the overall value of this term is one. That means the the steady states with and without substrate inhibition will be the same and this will allow us to make comparisons of the the dynamic behaviors of the TH reaction in the two cases (Results A).

The second term (on the right) requires more discussion. It was first proposed in the 1970's 2627 that the binding of dopamine to presynaptic autoreceptors affects TH and therefore the synthesis of dopamine. Although the details of the mechanisms are not certain, research since that time has demonstrated clearly that the autoreceptors modulate the activity of TH as well as the neuronal firing rate and the release of dopamine2928616263306431. All three effects are consistent: higher eda means more stimulation of the autoreceptors and this decreases the activity of TH 2963, lowers firing rate 6162, and inhibits release 2829. The evidence in these papers suggests that dopamine agonists can inhibit TH by at least 50% 2829. The more difficult question is how much synthesis is increased if the normal inhibition by the autoreceptors is released? In 63 only a 40% increase was found, but the data in 30 and 31 suggest that synthesis can be increased by a factor of 4 to 5. This is consistent with the original data in 27, Table 1. The third factor in the formula for V_TH(tyr, bh4, cda, eda) has the following properties: at the normal steady state it equals one; as eda gets large it approaches 0.5; as eda gets smaller and smaller it approaches 5. The exponent 4 was chosen to approximate the data in 30, figure 2. Note that, in this first model, we are not including explicitly the effects of the autoreceptors on firing rate and dopamine release.

After dopamine is synthesized it is packaged into vesicles by the vesicular monoamine transporter, MAT. We take the K_m of the transporter in the literature range (see Table 2) and choose the V_max so that the concentration of cytosolic dopamine is in the range 2-3 μM under normal circumstances. The experiments in 65 and the calculations in 66 suggest strongly that there is transport from the vesicles back into the cytosol, either dependent or independent of the MAT. We assume this transport is linear with rate constant, k_out, chosen so that the vast majority (i.e., 97%) of the cellular dopamine is in the vesicular compartment. The vesicles take up a significant fraction of the volume terminal, perhaps 1/4 to 1/3 (reference). For simplicity we are assuming that the vesicular compartment is the same size as the non-vesicular cytosolic compartment. This assumption is unimportant since we take the cytosol to be well-mixed and we are not investigating vesicle creation, movement toward the synapic cleft, and recyling where geometry and volume considerations would be crucial.

Vesicular dopamine, vda, is put into the synaptic cleft, where it becomes eda, by the term fire(t)(vda) in the differential equations for vda and eda (see above). fire is a function of time in some of our in silico experiments, for example in Results G where we investigate individual spikes. However, for most of our experiments fire = 1 μM/hr, which means that vesicular dopamine is released at a constant rate such that the entire pool turns over once per hour. This is consistent with a variety of experimental results on turnover and we will see in Results C that this choice gives decay curves after α-methyl-p-tyrosine (α-MT) inhibition of TH that match well the findings of Caron and co-workers 2425.

Extracellular dopamine has three fates. It is pumped back into the cytosol by the DATs; it is catabolized; it is removed from the system. The parameters for the DATs are taken from the literature. The other two fates are discussed next.

Cytosolic dopamine is catabolized by monoamine oxidase (MAO) and aldehyde dehydrogenase to dihydrophenylacetic acid (dopac), which is exported from the neuron and methylated by catecholamine methyl transferase (COMT) to homovanillic acid (hva). In this simple model we are not investigating the details of catabolism, only how cda is removed from the system. Since the cytosolic dopamine concentration is low (2-3 μM) and the K_m for MAO is high (210-230 μM, 67), the removal of cda is basically a linear process that we model by the first order term (cda). We choose the rate constant = 10/hr so that the rate of cytosolic catabolism is somewhat less than the synthesis rate of cda at steady state. Extracellular dopamine is also catabolized, first by COMT and then by MAO. In this case, we use a Michaelis-Menten formula for this process because the K_m of dopamine for COMT is low enough (approximately 3 μM, 68) that the process saturates in some of our in silico experiments in which large amounts of DA are dumped into the extracellular space. The half-life of hva is the brain is approximately hr 6970, which determines = 3.45/hr for the removal of hva from the system.

In our model the extracellular space is a single compartment. One should think of it as the part of the entire extracellular space corresponding to this particular synapse. Of course, if we had many model synapses, the eda from one will diffuse into the extracellular compartment of another (volume transmission). We are assuming for simplicity that the extracellular space is well-mixed, that is, we are ignoring diffusion gradients between different parts of the extracellular space. In fact, Venton et al. 48 have shown using a combination of experiments and modeling that the extracellular space is well-mixed during tonic firing but that substantial gradients exists between "hot spots" of release and reuptake and the rest of the extracellular space during and just after episodes of burst firing. In addition, when SNc projections die, as in Parkinson's disease or in denervation experiments, the terminals will be further apart making it certain that diffusion gradients will play an important role (see the Discussion). The term k_rem(eda) in the differential equation for eda represents removal of eda through uptake by glial cells, uptake by the blood, and diffusion out of the striatum. After some experimentation we chose k_rem = 400/hr because it gave good fits to the experimental data in 33 discussed in Results B and the experimental data in 2425 discussed in Results D.

In all cases, steady states or curves showing the variables as functions of time were computed using the stiff ODE solver in MATLAB.

Figure 1 shows the concentrations and velocities at steady state in our model. Only about 10% of the cellular tyrosine input goes to dopamine synthesis with the remainder going to the tyrosine pool (80%) or being catabolized (10%) as seen experimentally 535455. Cellular tyrosine itself has a steady state concentration of 126 μM in the model consistent with a large number of experimental observations 58564.

It is known that the cytosolic concentration of dopamine is quite low and the concentration of l-dopa is extremely low 3. In the model, at steady state, cda = 2.65 μM and the concentration of l-dopa is 0.36 μM, consistent with these observations. It is instructive to look at the flux balance of cda in the steady state. 27.3 μM of cda are manufactured from tyrosine per hour. 81 μM/hr of dopamine are put into the vesicles by the monoamine transporter and 80.1 μM/hr are put back into the cytosol from the extracellular space by the DATs. Finally, 26.5 μM/hr of dopamine is catabolized in the cytosol.

The largest portion of cellular dopamine is in the vesicles; in our model vda = 81 μM at steady state. We assume that at a "normal" firing rate the vesicular contents would be emptied in an hour; that is, vda is released into the synaptic cleft at 81 μM/hr. The DATs put most of this eda back into the cytosol (80.1 μM/hr), with the remainder being removed (0.81 μM/hr) or being catabolized (.02 μM/hr). We will see below that these velocities are consistent with the half-life measurements of Caron and co-workers 2425.

Tyrosine hydroxylase (TH) converts the amino acid tyrosine into l-dopa and bh4 into bh2; l-dopa is then converted by aromatic amino acid decarboxylase into dopamine. Given the dynamic nature of neurons and the importance of dopamine, it is not surprising that TH is regulated by many different mechanisms. TH is inhibited by dopamine itself and is also inhibited by the D2 autoceptors that are stimulated by extracellular dopamine. The effects of these regulations will be discussed below. Here we focus on a third regulation, substrate inhibition of tyrosine hydroxylase by tyrosine 23. Substrate inhibition means that tyrosine can bind non-enzymatically to TH preventing TH from performing its function of converting tyrosine to l-dopa. Substrate inhibition can be competitive (one tyrosine binding to TH makes the catalytic site unavailable to another tyrosine) or uncompetitive (the catalytic site is available to another tyrosine but the enzyme does not perform its catalytic function). Substrate inhibition is not widely recognized as an important regulatory mechanism, though it was proposed by Haldane in the 1930s 71, and it known to have an important homeostatic function in the folate cycle 72. Figure 2 shows normal Michaelis-Menten kinetics, competitive substrate inhibition, and uncompetitive substrate inhibition. In uncompetitive substrate inhibition the velocity curves rises, reaches a maximum, and then descends to zero because at higher and higher tyrosine concentrations more and more enzyme is bound non-enzymatically to tyrosine.

The velocity curve, figure 2 of 23, shows clearly that the substrate inhibition of TH by tyrosine is uncompetitive and we have chosen our kinetic parameters to match the shape of that curve. The question that we wish to address here is what is the purpose of this substrate inhibition? We will see that it stabilizes vesicular dopamine in the face of variations in tyrosine availability.

It is known 57 that brain tyrosine levels can double after meals, and this implies that tyrosine levels in the blood vary even more dramatically. In our model the average tyrosine level in the blood is 97 μM. We assume that for 3 hours after breakfast and lunch this concentration is multiplied by 1.75 and for three hours after dinner by 3.25. At other times the concentration of blood tyrosine is .25 × 97 = 24.2 μM, which gives a daily average of 97 μM. The blood tyrosine concentrations are shown in Figure 3 along with the cellular tyrosine levels (computed from the model) over a 48 hour period. As found in 57 the intracellular tyrosine levels (roughly the brain levels) vary considerably.

To see the effect of substrate inhibition on the synthesis of L-Dopa by TH, we computed the time courses of the velocity of the TH reaction both with and without substrate inhibition, Panel B of Figure 3. Without substrate inhibition the velocity of the TH reaction varies from 23.5 to 28 μM/hr while in the presence of substrate inhibition the variation ranges only from 27 to 28 μM/hr.

This naturally raises the question of how much the levels of vesicular dopamine vary throughout the day in the two cases. Panel C of Figure 3 shows that substrate inhibition greatly reduces the variation.

We conclude that one important purpose of substrate inhibition is to stabilize the velocity of the TH reaction, and thus the vesicular stores of dopamine, in the face of large variations in tyrosine availability because of meals. The stabilization is a result of the relatively flat velocity curve in a large neighborhood (say 75 μM to 175 μM -see Figure 2) of the normal tyrosine concentration of 126 μM. We note that the non-monotone shape of the velocity curve helps explain some of the unusual relationships between tyrosine levels and dopamine synthesis and release reported in the literature 735856.

In a series of studies and one modeling paper, Justice and co-workers studied the dynamics of extracellular dopamine in dopaminergic neurons in rat brain 34353633. In one experiment they stimulated the ascending projections of SN neurons in the medial forebrain bundle for ten seconds and measured the time course of extracellular dopamine in the striatum. The results of a similar stimulation in our model are shown in Figure 4, which also shows the data in the original experiment. Note that the curve starts to descend before the end of stimulation because of depletion of the reservoir of vda. The close match between our model curve and the data suggests that our V_max for the DATs (the primary clearance mechanism) is in the right range.

Over the last 15 years Caron and co-workers have conducted numerous experiments with dominergic neurons. We focus here on the experiments reported in 24, 25 and 74 that compare the behavior of extracellular dopamine and striatal tissue dopamine in wild type mice (WT) and mice that express no DATs at all (DAT^-/-), the heterozygote (DAT^+/-), and mice that overexpress the DATs (DAT-tg). The experiments of Caron and co-workers provide an exceptional opportunity to analyze the effects and importance of the DATs.

When we turn off the DATs in our model (by setting the V_max to zero), we see changes in steady state values that are qualitatively similar to those seen in 24 and 25 but the magnitudes differ somewhat. The steady state value of eda rises by a factor of 10 in the model when the DATs are turned off, while it rises by only a factor of 5 in the DAT^-/- mouse. In the model, vesicular dopamine declines from 81 μM to 11 μM when the DATs are turned off, while 24 and 25 report that striatal tissue dopamine in DAT^-/- mice is only 1/20 of the value in WT. We modeled the heterozygote (DAT^+/-) by reducing the V_max of the DATs to 1/2 the normal value. The model eda increases by 50% compared to WT and vda declines by 27%, which is almost exactly the decline in striatal tissue DA reported in DAT^+/- mice in (24, figure 3). In general, one would not expect the model and experimental results to correspond exactly because the DAT^-/- and DAT^+/- mice have not had their DATs suddenly turned off as we are doing in the model. These mice have lived their whole lives with no or reduced DATs, respectively, so their dopaminergic neurons may differ in other ways from those of the WT mice.

The studies 24, 25 and 74 report on various experiments that highlight the physiological difference between the WT, DAT^-/-, and DAT^+/- mice. We conducted similar experiments with the model and compared our results to theirs. Figure 1(E,F) of 25 shows the time courses of eda for WT and DAT^-/- mice after treatment with α-methyl-p-tyrosine (α-MT), a potent TH blocker. They find half-lives of approximately 2.5 hours for WT and 15-20 minutes for DAT^-/- mice. In the model, the half-life of eda is 2 hours and 40 minutes for WT mice and 37 minutes for DAT^-/- mice; see Figure 5.

One important focus of the experiments in 24, 25 and 74 is the clearance of eda after stimulation. In the model we can test this directly by raising the concentration of eda at time T = 0 to 10 times normal (for either WT or DAT^-/-) and measuring the half-life of eda as the system relaxes back to equilibrium. See Figure 6.

We find in the model that the half-life of eda for WT mice is .067 seconds and the half-life for DAT^-/- mice is approximately 6 seconds, giving a ratio of DAT^-/- half-life to WT half-life of about 90. Caron and coworkers find similar numbers experimentally except that their ratio is about 300. We note that the ratio is sensitive, of course, to small changes in the small WT half-life that is determined both in experiments and in the model from a very steep curve.

Some of the most interesting experiments in 2425 and 74 measure the actual time courses of eda in WT, DAT^-/-, DAT^-/-, and DAT-tg mice in response to pulse stimulation. Panel A of Figure 7, below, shows a composite of the experimental data taken from 24 and 74. We were intrigued by the non-monotone character of the peaks. Either more transporters (DAT-tg) or fewer transporters (DAT^+/-) lowers the eda peak compared to wild-type. Our investigations show that this behavior is due to two competing effects.

The first effect is that the amount of dopamine available to be released in response to an external pulse is not strictly proportional to the number of DATs. In the model "the number of DATs" is represented by the V_max of the transporter. We increased the V_max by 50% for DAT-tg mice compared to wild type, decrease it by 50% for DAT^+/- mice, and decrease it to zero for DAT^-/- mice. However, vda, the pool available for release, has the following model values: vda = 98.9 μM (DAT-tg), vda = 81 μM (WT), vda = 59 μM (DAT^+/-), vda = 11.4 μM (DAT^-/-). In the model, the amount of dopamine released in response to a pulse small enough not to deplete the vda pool very much is proportional to vda. Thus more dopamine is released in DAT-tg compared to WT compared to DAT^+/- compared to DAT^-/-. If this were the only effect then one would expect the eda peak to be highest for DAT-tg, lower for WT, still lower for DAT^+/-, and lowest for DAT^-/-.

The second effect is that if the cell has more DATs, then it should be able to pump the released eda back into the cell faster. Thus, if the amount released were the same in each case, we would expect the eda peak to be lowest for DATtg, somewhat higher for WT, still higher for DAT^+/-, and highest for DAT^-/-. However, as we have seen, the amount released is not the same but decreases as one progresses from DAT-tg to WT to DAT^+/- to DAT^-/-. These are the two competing effects that determine the heights of the peaks. The situation is even more complicated and interesting, however. The K_m for the DAT has been measured in a number of experiments. A reasonable range of possible values is 0.2 μM to 2 μM; see 75 and 76. In the experiments in 24 and 74 the maximal eda concentrations were in the range 2-3 μM. This means that if K_m = 0.2 then the height of the peak will be highly affected by how much dopamine is released because the DATs are saturated.

In Panel B of Figure 7, we show the time courses of extracellular dopamine in model experiments where K_m = 0.2 μM. Notice that the peaks decrease as one goes from DATtg to WT to DAT^+/- to DAT^-/-. In each case the amount of stimulation was the same. This is what one would expect if the first effect, the amount of dopamine released, dominates. To test whether the saturation of the DATs causes this effect we did a model experiment in which the amount of stimulation was reduced to 1/10 of what it was before. The results can be seen in Panel C of Figure 7. At this lower stimulation level, the second effect dominates because the DATs are no longer saturated and the peaks increase as one goes from DAT-tg to WT to DAT^+/-. Note also how much narrower the peaks are because the DATs are operating on the linear parts of their response curves rather than the saturation part. Panels B and C show that the peaks can be either decreasing or increasing as one proceeds from DAT-tg to WT to DAT^+/-, depending only on the amount of stimulation. The DAT^-/- peak remains lowest because it is not affected by saturation effects on the DATs, since there aren't any.

If we increase the K_m of the DATs we will decrease the saturation effects seen in Panel B of Figure 7 and thus the two effects (amount released and rapidity of uptake) should be more evenly balanced. That is indeed the case as shown by Panel D of Figure 7. In Panel D the K_m of the DATs has been raised to 1.6 μM and everything else remains the same as in the model experiments shown in Panel B. Now the WT peak is higher than both the DAT-tg peak and the DAT^+/- peak. The differences are not great, but the non-monotone effect is clear. As before, the DAT^-/- peak is the lowest since so little dopamine is released.

These model experiments show that the relative heights of the peaks depend on the size of the vesicular stores, the number of DATs, their K_ms, and the amount of stimulation.

Bergstrom and Garris 46 measured the time course of extracellular dopamine after two seconds of stimulation in rat striatum after partial denervation. We will let f denote the fraction of striatal terminals still alive. With 20 Hz stimulation, the peaks of the resulting eda curves are almost independent of f until f = .15 (46, figure 3(a), and Panel A of Figure 8). This homeostasis, coined "passive stabilzation" in 77, is the main focus of 46. By contrast, with 60 Hz stimulation for 2 seconds, the resulting peaks decrease almost linearly as f decreases from one towards zero (46, figure 3(b), and Panel C of Figure 8). Our model shows very similar behavior in both cases (Panels B and D of Figure 8).

Bergstrom and Garris do not explain the reasons why the two cases (20 Hz and 60 Hz) are so different, although the reasons are implicit in their discussions. We give an explanation here using a simple model for eda introduced in 44. As we will see, the difference depends on the K_m of the DATs. We denote the (well-mixed) extracellular concentration of dopamine in the striatum by E(t). We'll ignore removal from the system and catabolism because we are interested in events on a very short time scale. Compared to the concentrations we get after stimulation, E starts very small, so we'll assume E(0) = 0. Assume that dopamine is released from the cells at a total rate of C/sec for ^to seconds. In 46, t_o = 2, so the concentration E(t) will satisfy the differential equation:

where we take K_m = 0.2 μM as in our large model. For t > 2 the C isn't there any more but that doesn't affect the maximum concentration of E(t), which occurs at t = 2. We want to calculate E(2) and then introduce our scale factor f and see how the value scales with f.

We will consider two cases. Suppose the release is relatively low as it is in the 20 Hz case where the peak concentrations are (on average) between 0.1 and 0.2 μM. Since the concentrations are below K_m we can approximate (1) by:

We easily solve this differential equation and find

Suppose, now, that only the fraction f of cells are still alive. Then C is replaced by fC and V_max is replaced by fV_max, so the maximum is:

The shape of the graph of E(2) as a function of f depends on the size of . Bergstrom and Garris measured a V_max close to 4 μM/sec and K_m = 0.2 μM, so ≈ 40. This means that the graph of E(2) will be almost constant as f gets smaller from 1. Only when f gets very very small will E(2) plunge down to zero. This is the constant behavior of the peaks seen in 46, figure 3(a), for 20 Hz stimulation.

In the case of 60 Hz stimulation the peaks are quite high, as much as 3 μM for the intact striatum, far above the K_m of the transporters. As long as the concentrations are well above the K_m the transporters are saturated and we can approximate (1) by

so the solution at t = 2 is:

If only the fraction f of cells are left, then,

Thus the maxima should decrease linearly with f and that's what Bergstrom and Garris see in their figure 3(b).

This calculation is correct as long as the concentration of E stays well above K_m. The complete behavior as f goes from 1 to 0 in the 60 Hz case should be: first, this linear decrease, then a middle range around the K_m value where the rate of decrease is more modest, then a flat plateau as in the first case we considered, until, finally, E(2) should plunge to 0 for almost complete denervation. As the ability to take accurate voltametric measurements improves, it will be interesting to see if this prediction is correct.

It is well established that the expression levels of proteins vary substantially, even in genetically identical cells, and often vary substantially in time in individual cells 78. Some of this variation is due to control functions in the cell but other variation is due to the stochastic nature of gene expression when small numbers of molecules are involved. The D2 autoreceptors stabilize the velocity of the TH reaction (and therefore vesicular stores) against variation in gene expression level. The mechanism is easy to understand. If the expression of TH drops then the dopaminergic neuron will have less cytosolic and vesicular dopamine and less will be released into the extracellular space. When the concentration of extracellular dopamine drops, the inhibition of TH via the autoreceptors is released, partially compensating for the drop in TH expression. Similarly, if TH is overexpressed, extracellular dopamine rises and the inhibition of TH by the autoreceptors is increased. Panel A of Figure 9 shows how the velocity of the TH reaction depends on the expression level of TH both with and without the autoreceptors. In the presence of autoreceptors the effect of expression level is much milder.

Autoreceptors on the presynaptic membrane upregulate tyrosine hydroxylase(TH) when eda drops and downregulate TH when eda rises as it will do if the firing rate of the neuron increases 6261. Thus, the autoreceptors provide a mechanism whereby the eda concentration provides feedback inhibition to TH. The strength of the effect can be seen in Panel B of Figure 9 where eda is graphed as a function of firing rate. The eda curve is much flatter in the presence of the autoreceptors. We note that the eda also affects firing rate directly via somatic autoreceptors 62, but this is not included in our model.

It is known 34 that there are two typical firing patterns seen in the dopaminergic neurons of the SNc, tonic firing at about 5 Hz and bursts of action potentials with an intraburst frequency of about 15-30 Hz. Dopaminergic neurons respond to reward-related stimuli with increased burst firing 879 and burst firing is more effective at raising dopamine levels than tonic firing 8045. Recently bursts have been measured in awake, freely-moving animals 814782 in response to rewards and in response to cues for the rewards when the cues have been learned; for a review, see 83.

We see in our model responses that quite similar to those observed experimentally. Panel A of Figure 10 shows the time course of extracellular dopamine in response to steady firing at 5 Hz. All extra dopamine is cleared from the extracellular space before the next action potential arrives as reported in 76. Note that on this short time scale cytosolic dopamine and vesicular dopamine remain approximately constant. However, Panel B shows that a burst of action potentials at 15 Hz causes a substantial rise in average eda 76. The model results shown in Figure 10 are similar to the model and experimental results reported in 82, figure 2. Thus, even a very short term shift from tonic firing at 5 Hz to burst firing at 15 Hz produces a large dopamine signal. This shows how sensitive the system is to a brief short-term change in frequency of firing.

However, if firing continues for a long time at 15 Hz, the feedback on TH via the autoreceptors will cause eda to decline to an intermediate level, higher than normal but not as high as the short term response. The inhibition of TH by increased binding to the autoreceptors happens quickly, but the resulting decrease in cda and vda happens slowly over a nine hour period (Figure 11), and this causes a gradual decrease in eda even though the firing rate remains elevated. Thus, over the long term, the eda concentration gradually habituates to the increased firing rate. It would be interesting to test this prediction of the model experimentally.