Dopamine release dynamics in the nucleus accumbens are modulated by the timing of electrical stimulation pulses when applied to the medial forebrain bundle and medial prefrontal cortex

Abstract

Electrical brain stimulation has been used in vivo and in vitro to investigate neural circuitry. Historically, stimulation parameters such as amplitude, frequency, and pulse width were varied to investigate their effects on neurotransmitter release and behavior. These experiments have traditionally employed fixed-frequency stimulation patterns, but it has previously been found that neurons are more precisely tuned to variable input. Introducing variability into the inter-pulse interval of stimulation pulses will inform on how dopaminergic release can be modulated by variability in pulse timing. Here, dopaminergic release in rats is monitored in the nucleus accumbens (NAc), a key dopaminergic center which plays a role in learning and motivation, by fast-scan cyclic voltammetry. Dopaminergic release in the NAc could also be modulated by stimulation region due to differences in connectivity. We targeted two regions for stimulation—the medial forebrain bundle (MFB) and the medial prefrontal cortex (mPFC)—due to their involvement in reward processing and projections to the NAc. Our goal is to investigate how variable inter-pulse interval stimulation patterns delivered to these regions affect the time course of dopamine release in the NAc. We found that stimulating the MFB with these variable stimulation patterns saw a highly responsive, frequency-driven dopaminergic response. In contrast, variable stimulation patterns applied to the mPFC were not as sensitive to the variable frequency changes. This work will help inform on how stimulation patterns can be tuned specifically to the stimulation region to improve efficiency of electrical stimulation and control dopamine release.

Introduction

Electrical stimulation is often used to activate or suppress activity in neurons and neuronal circuits to modulate the release of neurotransmitters.^1,2 Deep brain stimulation (DBS), a form of electrical stimulation, has been used for decades for the treatment for severe Parkinson’s disease^3,4 and, more recently, as a potential treatment for depression.⁵ Clinical use of DBS typically involves applying a fixed-frequency electrical stimulation to a brain region.⁶ In patients with major depression, the subcallosal cingulate cortex,⁷ nucleus accumbens,⁸ and the medial forebrain bundle⁹ are often targeted by DBS because they are essential to regulating emotions and motivation.^6,10 However, most research into electrical brain stimulation, including DBS, has utilized fixed-frequency stimulation patterns in a single brain region which aim to modulate endogenous neuronal activity. Such fixed-frequency approaches do not accurately mimic the variation in neuronal activity or differentiate the sensitivity of anatomically distinct neuronal circuits to variable input. Because of the many roles of dopamine, it is possible that the different roles are encoded into the timing of the neuronal firing.¹¹

Endogenous dopamine neurons typically fire in either a burst or pacemaker firing pattern.^12–14 Bursting neurons fire in clusters with approximately four spikes per burst with each burst lasting less than 0.5 seconds.¹² In contrast, pacemaker neurons fire in a single-spike pattern in a low frequency, consistent spiking over long periods of time.¹³ Similarly, dopamine has two forms of release, phasic and tonic, andmodeling work indicates these forms of release are associated with burst and pacemaker firing, respectively.¹⁵ Phasic release occurs on the sub-second time scale and is the result of stimulation whether it is an external stimulation such as receiving a reward or a physical stimulation such as electrical stimulation.¹⁶ Tonic release contributes to the long-term, basal concentration level of neurotransmitter in the extracellular space.¹⁷ Typical electrical stimulation is performed using fixed inter-pulse intervals; however, the majority of dopamine neurons fire in bursts.^12,18 Some electrical stimulation parameters such as pulse frequency,^19,20 pulse amplitude,²⁰ and pulse width^19,20 have been investigated in context of DBS but not fully characterized; however, little is known about how inter-pulse interval variability affects dopaminergic release.

To understand how variation in the stimulation pattern and the stimulation target can affect dopaminergic release, we targeted pathways that evoke release in the nucleus accumbens (NAc). The NAc is involved in motivation,²¹ error-driven learning,²² and encoding unpredicted reward.²³ The NAc is also implicated in many diseases such as addiction and depression,²⁴ and a few DBS studies have targeted the NAc to treat obsessive compulsive disorder,^25,26 depression,^5,8,27 and addiction.^28,29 Furthermore, dopaminergic release in the NAc is modulated by the timing of the neural activity or stimulation.¹¹ The primary accumbal afferent dopaminergic connections are projections from the ventral tegmental area (VTA). The NAc also has many glutamatergic projections from cortical structures, including the prefrontal cortex,³⁰ basolateral amygdala,^30,31 ventral hippocampus,^30,32 and intralaminar thalamic nuclei.^33,34

To evoke release in the NAc, we targeted two afferent NAc connections: the medial forebrain bundle (MFB), a well-studied dopaminergic pathway, and the medial prefrontal cortex (mPFC), a key cortical center capable of modulating dopamine release by glutamatergic inputs.^32,35 The MFB contains dopaminergic axons extending from the VTA to the NAc.³⁶ Applying an electrical stimulus to the MFB allows direct activation of unmyelinated dopaminergic axons and other myelinated neurons to evoke dopaminergic release in the NAc.³⁷ On the other hand, electrically stimulating the mPFC primarily activates glutamatergic neurons, which project to the NAc^30,32,35,38 and VTA.³⁹ Subsequently, glutamate is released in the VTA and NAc which causes the downstream release of dopamine in the NAc.^37,40,41 Glutamatergic neurons projecting to the VTA synapse with GABAergic neurons and dopaminergic neurons, which can evoke dopamine release in the NAc.^40,42,43 Glutamatergic projections to the NAc synapse with both D1- and D2-expressing medium spiny neurons, which use NMDA-receptor modulation.^33,44 NMDA receptors are not colocalized with tyrosine hydroxylase on accumbal dopaminergic axons, but NMDA receptors can modulate presynaptic and postsynaptic dopaminergic release.⁴⁵ Therefore, dopaminergic release in the NAc could be directly controlled by adjacent glutamatergic neurons. However, stimulating D2-expressing neurons in the mPFC inhibit dopamine release in the NAc,^11,42 so it is feasible that the mPFC-VTA-NAc pathway is primary driver of accumbal dopaminergic release. Because of these differences in release pathways, we consider MFB stimulation primarily as a direct axonal stimulation while mPFC stimulation primarily causes dopaminergic release in the NAc by multi-synaptic activation through the glutamatergic projections to the NAc and VTA.

Previous studies have shown that spike activity in neurons respond reliably and dynamically to variable input.^46,47 To investigate how variability in the time-course of activation alters downstream release, we varied the inter-pulse interval distribution of stimulation patterns from fixed interval stimulation patterns to burst-like stimulation patterns (e.g., series of short inter-pulse intervals with intermittent long intervals). This was accomplished by generating pulse sequences where we modified the ‘local variance,’ a measure of variation in the inter-pulse interval.⁴⁸ In this study, average stimulation rate and local variance were systematically varied while simultaneously measuring dopaminergic release in the NAc using fast-scan cyclic voltammetry (FSCV). Two different stimulation pathways were investigated to understand how differences in connectivity to the NAc respond differently to variable inter-pulse intervals. Stimulating the MFB allows for investigation of direct activation of axons while stimulating the mPFC triggers dopamine release through a multi-synaptic pathway. We observed that when stimulation of the MFB and mPFC is compared, increased variability of the stimulation pattern caused different effects. Altering the overall frequency had a larger effect on the MFB than the mPFC, but altering the local variance, or the inter-pulse interval, influenced the release in the NAc when stimulating the mPFC but not the MFB.

Results and Discussion

The primary aim of this study is to investigate how the time-course of dopamine release is impacted by variance in the inter-pulse interval and the target of the stimulation. Previous studies have investigated stimulation timing parameters such as frequency, duration, and number of pulses; however, these studies used fixed inter-pulse intervals applied to either the MFB⁴⁹ or mPFC.⁵⁰ Neuronal circuits are highly responsive to the variability and predictability of their inputs and may utilize such variation to improve information processing;¹¹ therefore, it is possible that variation in the inter-pulse interval distribution could significantly affect the level and time-course of dopamine release. Further investigation into varying the inter-pulse interval allows for better understanding of the neuronal and neurochemical responses to stimulation. A study from Montague et al. investigated the effect of irregular stimulation patterns on dopamine release in the striatum.⁵¹ Their stimulation patterns consisted of a fixed-frequency 24-pulse stimulation burst that occurred at irregular intervals over a 70-second time period.⁵¹ In contrast, this study investigates variation of the inter-pulse intervals and the subsequent effect on dopamine release. We used local variance (LV) to quantify the variation in the stimulation sequence.^52–54 Local variance measures variability by comparing adjacent inter-pulse intervals in a pulse sequence (Equation 1).⁴⁸

$$LV=\frac{3}{n-1}\sum _{i=1}^{n-1}{\left(\frac{I_i-I_{i+1}}{I_i+I_{i+1}}\right)}^2$$ (1)

In Equation 1, n is the number of inter-pulse intervals, I_i is the i-th inter-pulse interval, I_i+1 is the i+1^st inter-pulse interval.⁴⁸

Typical neurons fire with local variance values in the range of 0–1.25.⁵⁵ The local variance patterns used in this study were randomly generated and selected with spontaneous local variance measures in mind. Figure 1D – K contain histograms of the inter-pulse interval for different local variance values to exemplify how the distributions of inter-pulse intervals are affected. Stimulation patterns with fixed inter-pulse intervals, as past electrical stimulation experiments implemented, have a local variance of zero (Figure 1D & E). Stimulation patterns with low local variance values mimic tonically active neurons due to their regular inter-pulse intervals (Figure 1D – G). As the local variance increases, the variance of inter-pulse intervals increases, which better mimics burst-firing neurons (Figure 1F – K). Stimulation patterns with intervals derived from a Poisson series have local variance values approaching one (Figure 1H & I). See Methods for more details on local variance. In addition to exploring local variance, we also investigated the effects of average stimulation frequency. Due to the variable inter-pulse intervals, stimulation frequencies were calculated as averages over the full stimulation pattern. In this study, we used lower average frequencies, 10 Hz and 20 Hz, for a duration of 10 seconds, based on a previous study investigating the interplay between stimulation frequency and duration.⁵⁰ Hill et al. found that lower stimulation frequencies in the mPFC (i.e. 10 Hz) evoked larger peak dopamine release with longer stimulation durations.⁵⁰ They showed that longer, specifically 10- and 20-second, mPFC stimulation evokes more dopamine in the NAc for frequencies of 4, 10, 20, and 60 Hz.⁵⁰ Longer stimulation patterns also help to better understand dopamine dynamics when measuring with techniques detecting the synaptic overflow such as fast-scan cyclic voltammetry (FSCV).¹ In addition, past studies investigating electrical stimulation using microdialysis often stimulated for 10–20 minutes.^56–58 Similarly, DBS studies apply stimulation for long periods of time, often hours or days. The stimulation parameters investigated in this study are in Table 1.

Figure 1.

Experimental design of electrode placement and stimulation patterns. (A) Schematic of electrode placement and stimulation regions. Stimulation of the MFB directly activates dopaminergic axons. Stimulating the mPFC causes multi-synaptic activation because the signal must be propagated through multiple neurons to ultimately cause a release of dopamine in the nucleus accumbens. (B) The triangular waveform from −0.4 to +1.3 V vs. Ag/AgCl is applied to the carbon-fiber microelectrode (CFME) at a frequency of 5.0 Hz. (C) A 2-second window of an example stimulation pattern. This example shows a stimulation pattern (biphasic) with a 20 Hz average frequency and a local variance of 1.01 over the full 10-second stimulation. In the MFB, each pulse is 4.00 ms wide with an amplitude of ± 300 μA while in the mPFC each pulse is 2.00 ms with an amplitude of ± 600 μA so that the charge passed at the stimulating electrode by both regions is equal. (D-K) Histograms of inter-stimulation intervals (ISI) in each of the stimulation patterns. The stimulation patterns in (H) and (J) each have 3 data points that are greater than 400 ms.

Table 1.

Stimulation parameters of experimental patterns. Frequency and local variance were calculated over the 10-second stimulation pattern. Each frequency and local variance combination was repeated 5 times per rat.

Frequency	Local Variance
10 Hz	0.00	0.39	1.00	1.28
20 Hz	0.00	0.38	1.01	1.24

Downstream dopamine release was detected in the nucleus accumbens (NAc) after MFB or mPFC stimulation. Activating mPFC targets glutamatergic projections that are involved in goal-directed behaviors.^59–61 In contrast, MFB stimulation directly activates dopamine axons, causing release in the NAc (Figure 1A).³⁷ Dopamine release was monitored in the NAc (Figure 1A) with a carbon-fiber microelectrode (CFME). We used FSCV which allows dopamine concentration to be monitored at a high temporal resolution, to investigate the effect of inter-pulse interval variability on dopamine release. The triangular FSCV waveform was applied from −0.4 V to +1.3 V vs. Ag/AgCl at 400 V/s (Figure 1B).⁶² A 5.0 Hz application frequency was used in contrast to a typical 10 Hz application frequency to minimize artifacts on electrophysiological recordings (see Methods).⁶² The same stimulation patterns were applied to each stimulation region, and an example of a stimulation pattern with an average frequency of 20 Hz and a local variance of 1.01 is shown (Figure 1C). The stimulation pattern was designed so that the stimulation pulses did not overlap with the application of the FSCV waveform (Figure 1B).

Data from a representative stimulation pattern is shown in Figure 2. Results in Figure 2 exemplify the dopaminergic release after stimulating the MFB with a 20 Hz, 1.01 LV stimulation pattern. This stimulation pattern was used as an example due to the high variability in the pulse timing (Figure 2C). To analyze the effect of local variance on dopamine release, an instantaneous stimulation frequency (ISF) is determined by counting the number of pulses within a given bin interval and dividing by the interval duration to determine the frequency in each bin. The bin size was chosen to be 200 ms to align with the 5.0 Hz voltammetry data collection. The ISF for the 20 Hz, 1.01 LV stimulation pattern is variable ranging from 0 Hz to 55 Hz (Figure 2C). Dopamine levels before, during, and after the stimulation were monitored by continuously collected cyclic voltammograms, which are visualized using false color plots (Figure 2A). The current of the oxidation peak at approximately 0.6 V vs. Ag/AgCl (Figure 2B, inset) is measured on each voltammogram and plotted over time to measure the relative change of dopamine concentration (Figure 2B).

Figure 2.

Representative experimental data of a stimulation pattern with a local variance of 1.01 and a 20 Hz average frequency. The stimulation is applied from 0–10 seconds, denoted by the dashed lines. (A) Representative color plot of dopamine release in response to an MFB stimulation. The y-axis is applied voltage and current is encoded in false color. (B) The current is monitored at the oxidation peak around 0.6 V vs. Ag/AgCl (inset) over time to construct a concentration vs. time plot. Concentration vs. time results for MFB stimulation are in black (Mean ± SEM, n = 5 rats) and the results for mPFC stimulation are in blue (Mean ± SEM, n = 5 rats). (C) The biphasic stimulation pattern overlayed with the instantaneous stimulation frequency (ISF) in black. ISF values range from 0 to 55 Hz. (D) Derivative of (B) to represent the change of dopamine in response to the stimulation. Change of dopamine over time results for MFB stimulation are in black (Mean ± SEM, n = 5 rats) and the results for mPFC stimulation are in blue (Mean ± SEM, n = 5 rats).

Peaks of high ISF values seem to correlate to larger releases of dopamine. The peak dopaminergic release occurs at about 8.2 seconds (Figure 2B), which follows a period of high ISF values from 7.0–8.2 seconds (Figure 2C). Another way of determining the effect of ISF is to compare it to the change of concentration of dopamine by taking a derivative of Figure 2B to better visualize differences in dopamine release and rate of release (Figure 2D). ISF (Figure 2C) and change of concentration (Figure 2D) are well-correlated here (Pearson’s correlation, r = 0.60); therefore, variability in pulse timing of stimulation patterns causes variable dopamine release.

A range of local variance values was used to investigate the effect of different levels of “burstiness” in different stimulation patterns (Table 1). The resulting dopaminergic release profiles of each of these stimulations are in Figure 3 (n = 5 rats). Each release profile was normalized to the 20 Hz, 0.00 LV stimulation for its respective region because the release was most consistent during application of this stimulation pattern. For example, each result in Figure 3A and 3C is normalized to the 20 Hz, 0.00 LV stimulation, or the black trace, in Figure 3C. The 10 Hz average frequency stimulations are in Figure 3A & B while the 20 Hz average frequency stimulations are in Figure 3C & D. The MFB stimulations are in 3A & C and the mPFC stimulations are in 3B & D. These release profiles serve as a basis for Figure 4 and Figure 5.

Figure 3.

Concentration profiles during stimulation of the MFB (A and C) and the mPFC (B and D). Each trace in A and C is normalized to the 20 Hz, 0.00 LV (the black trace) stimulation in C. Each trace in B and D is normalized to the 20 Hz, 0.00 LV stimulation (the black trace) in D. The average global frequencies are 10 Hz (A-B) and 20 Hz (C-D) (Mean + SEM, n = 5 rats). Stimulation patterns were the same for each region but were different for each frequency due to the change in the number of pulses. For 10 Hz stimulations, local variance values were 0.00, 0.39, 1.00, and 1.28 (A-B). For 20 Hz stimulations, local variance values were 0.00, 0.38, 1.01, and 1.24 (C-D). The scale for each figure is different due to varying amounts of release.

Figure 4.

Area under the curve during the stimulation from 0 to 10 seconds. (A) MFB stimulation (Mean + SEM, n = 5 rats). Two-way repeated measures ANOVA shows frequency is significant (F_(1,4) = 37.02, **p = 0.0037) but local variance is not (F_(3,12) = 1.19, p = 0.355). Holm post-hoc tests: **p < 0.01, ***p < 0.001. The interaction is not significant (F_(3,12) = 1.013, p = 0.421). (B) mPFC stimulation (Mean + SEM, n = 5 rats). Two-way repeated measures ANOVA shows frequency is not significant (F_(1,4) = 6.15, p = 0.0682) and local variance is significant (F_(3,12) = 18.42, ****p = 0.000087). Holm post-hoc tests: *p < 0.05. The interaction is not significant (F_(3,12) = 0.119, p = 0.947).

Figure 5.

Derivative analysis. (A) Change of dopamine in response to a 20 Hz, 1.01 LV stimulation is in black. This is compared to the instantaneous stimulation frequency (ISF) in blue in the MFB (n = 5 rats, mean ± SEM). Both the rate of change of dopamine and the ISF were binned into 400 ms bins. (B) The same 20 Hz, 1.01 LV stimulation dopaminergic response is compared to the ISF in the mPFC (n = 5 rats, mean ± SEM). (C) Dopamine release in response to varying levels of ISF in 200 ms bins. The release in the MFB is represented in filled, black circles while the mPFC response is represented by empty, red squares. Responses from the two stimulation regions are plotted on separate axes. The MFB (slope = 4.0 ± 0.5 nM, linear regression, R² = 0.84) is more sensitive to ISF changes than the mPFC (slope = 0.4 ± 0.2 nM, linear regression, R² = 0.19). (D) Linear regression values from (C) compare the ISF to the subsequent dopaminergic release for each animal. The dopaminergic release in the MFB is more correlated to the ISF than the release in the mPFC (unpaired t-test, ****p < 0.0001). (E) Representative correlogram correlating change of dopamine to the ISF in the MFB and mPFC (n = 1). (F) The max correlation values from the correlograms are plotted (n = 5 rats, mean + SEM). Two-way repeated measures ANOVA shows that both brain region (F_(5,40) = 5.48, ***p = 0.0006) and stimulation pattern (F_(1,8) = 9.46, *p = 0.0152) are significant. Bonferroni post-hoc test: ***p < 0.001.

The different release profiles of the MFB and mPFC are likely due to their NAc activation pathways. Stimulation of the MFB directly activates dopaminergic neurons that innervate the NAc from the VTA.³⁶ Due to this direct activation, the MFB is highly responsive to changes in frequency resulting in varied dopaminergic release over the 10-second stimulation (Figure 3A & C). This responsiveness is emphasized when comparing a fixed-frequency stimulation (i.e., a stimulation with local variance of zero) to stimulation patterns with variation in pulse timing. The release caused by each LV 0.00 stimulation shows a steady rise in dopamine while the release from non-zero local variances have multiple maxima and minima in the MFB.

In contrast, mPFC-stimulated release in the NAc is not as highly responsive to frequency changes, likely because mPFC stimulation activates two different pathways (Figure 1A),⁶³ indicating that multiple circuits could be affecting dopaminergic release simultaneously. However, previous studies show dopaminergic release in the NAc is primarily controlled by dopaminergic neurons in the VTA.⁶⁴ Furthermore, mPFC-stimulated dopamine release seems to plateau around 5 seconds. This may have to do with the rate of dopaminergic and glutamatergic release and clearance from the mPFC to the NAc. You et al. investigated the effect of stimulation frequencies on the mPFC to NAc pathway. When examining stimulation frequencies from 6 Hz to 400 Hz, they found glutamate release did not increase significantly with frequencies greater than 25 Hz.⁶⁵ This indicates that higher rates of mPFC stimulation may not be associated with more release in the NAc and perhaps may have inhibitory effects. High-frequency stimulation could preferentially activate D2 autoreceptors which inhibit calcium channels therefore attenuating dopamine release.⁶⁶ Other studies show that some neurons in cortical pathways have decreased activity after continual, high-frequency stimulation.⁶⁷ Though, these studies used higher frequency stimulations than investigated here. Our results indicate that there may be interplay between the duration and magnitude of high frequency bursts that affect the release of dopamine. Future studies using stimulation patterns with periods of high frequency stimulation (i.e., 100 Hz or greater) will help further elucidate the dynamics of dopamine in the MFB and mPFC.

To analyze the overall effects of local variance and frequency on dopamine release, we calculated area under the curve (AUC) of dopaminergic release profiles for the stimulation duration of each stimulation pattern (Figure 4). Analyzing the AUC allows us to understand the effects of local variance on the overall degree of dopaminergic release. As in Figure 3, each result is normalized to the 20 Hz, 0.00 LV stimulation. In the MFB (Figure 4A), frequency has a large effect on the release (Two-way ANOVA, F_(1,4) = 37.02, **p = 0.0037). This is consistent with previous reports that indicate that MFB stimulation is frequency dependent and dopamine release will increase with frequency until approximately 100 Hz.⁶⁸ Local variance in the MFB stimulation did not have a statistically significant effect (Two-way ANOVA, F_(3,12) = 1.19, p = 0.355) on the total AUC. On the other hand, increasing local variance attenuated the dopaminergic release during mPFC stimulation (Figure 4B, Two-way ANOVA, F_(3,12) = 18.42, ****p = 0.000087). Changing the frequency did not significantly affect the AUC (Two-way ANOVA, F_(1,4) = 6.15, p = 0.0682). This is consistent with other studies that show that mPFC stimulation elicits more dopamine with lower frequency stimulations than high frequency stimulations for 10 second stimulations.⁵⁰ The higher local variance stimulations experience periods of high frequency stimulation and thus may not elicit as much dopamine as a consistent, low-frequency stimulation in the mPFC.

To more closely investigate differences in rates of release, derivatives of each trace were taken. Average traces of the derivative for the dopaminergic response to the 20 Hz, 1.01 LV stimulation of both the MFB (Figure 5A) and mPFC (Figure 5B) are compared to the ISF in blue. The rate of change of dopamine under these stimulation parameters is more highly correlated with the ISF during MFB stimulation and not as well correlated during mPFC stimulation (Pearson’s correlation, r = 0.60 in the MFB and r = 0.46 in the mPFC). Plots for the other stimulation patterns show similar correlations (Figure S2). To further support this statement, the change of dopamine was plotted in response to the ISF for both MFB and mPFC stimulation (Figure 5C). The dopaminergic response to higher ISF values applied to the MFB resulted in a linear increase of dopamine while the dopaminergic response of mPFC stimulation had a linear increase until it reached a plateau around 20 Hz (Figure 5C). These results are consistent with previous findings that high frequency mPFC stimulations are not as effective at releasing dopamine.^50,65 Furthermore, the dopaminergic response to MFB stimulation (linear regression, slope = 4.0 ± 0.5 nM, R² = 0.84) is more sensitive to ISF changes than mPFC stimulation (linear regression, slope = 0.4 ± 0.2 nM, R² = 0.19). Similarly, a comparison of the R² values of the linear relationship between ISF and the rate of dopaminergic release for each animal showed that the relationship to MFB stimulation was stronger than mPFC stimulation (Figure 5D, t-test, ****p < 0.0001). The R² values in Figure 5C are a result of pooling all the animal data as one data set while the results in Figure 5D are a result of averaging individual animal results. This reveals that the dopaminergic response to MFB stimulation is more tightly linked to the frequency and timing changes of the stimulation than mPFC stimulation.

A cross-correlation analysis was performed to determine the maximal correlation between changes in dopamine concentration and ISF. Representative correlograms demonstrate the differences in the MFB- and mPFC-stimulated release in terms of peak correlation values (Figure 5E). The max correlation values for each stimulation pattern can be extracted from each correlogram (Figure 5F). MFB stimulation results in higher correlation values (Two-way ANOVA, ***p = 0.0006, F_5,40 = 5.48) than the mPFC stimulation indicating again that the MFB is more responsive to the stimulation pattern than mPFC (Figure 5F).

The lag time at the point of maximal correlation can also be extracted from the cross-correlation analysis (Figure 5E); however, this is a challenging analysis due to high variability in lag time. Because the mPFC utilizes a multi-synaptic circuit, the process is likely to be slower than the direct activation of the dopaminergic neurons in the MFB. Synaptic release,⁶⁹ concentration and clearance,^70,71 and receptor kinetics⁷² all contribute to the dopaminergic release caused by mPFC activation. Due to these constraints for dopamine release in the NAc, we expected the release caused by mPFC activation to be slower than in the MFB. However, while MFB-stimulated release typically had a lag time around zero for each stimulation pattern, mPFC-stimulated release had very variable responses to the stimulation patterns (Figure S3). The most variable stimulation patterns, 10 Hz 1.28 LV and 20 Hz 1.24 LV, caused the MFB-stimulated release to be tightly correlated with the stimulation pattern while the mPFC-stimulated release resulted in variable lag times (Figure S4; F-test, 10 Hz 1.28 LV F = 95.6 and 20 Hz 1.24 LV F = 35.0).

The differences between dopaminergic release in the NAc caused by mPFC and MFB stimulation could be attributed to the circuit architecture of the stimulation pathways. When directly targeting MFB dopaminergic axons, the release in the NAc is frequency-driven and closely resembles the instantaneous frequency of the stimulation pattern (Figure 5A). On the other hand, stimulation of the mPFC resulted in a dopaminergic response in the NAc that was not as tightly correlated as the MFB-stimulated release (Figure 5B). Efferent glutamatergic mPFC neurons synapse with GABAergic neurons in the NAc and VTA, which could inhibit dopaminergic release in the NAc.^40,43 It is possible that the stimulation frequencies larger than 20 Hz cause GABAergic neurons to be preferentially activated and dopamine release decreased (Figure 5C). In this case, the primary driver of mPFC-stimulated dopamine release is the glutamatergic connections to dopaminergic neurons in the VTA. Additionally, the dopaminergic response to MFB and mPFC stimulation may be contingent on the type of firing in those regions. For example, VTA neurons with axons in the MFB may exhibit more bursting firing behavior and are therefore better equipped to respond to variable stimulation patterns. To better understand the mPFC evoked release, lidocaine can be used to inactivate the VTA and isolate the projections from the mPFC to the NAc to determine the different activities of the different pathways.⁷³.⁷³ The effect of stimulation parameters is crucial to understand in the context of electrical stimulation and DBS, and while many experiments have investigated amplitude, frequency, and pulse width, they have used fixed inter-pulse intervals. By understanding the dynamics of dopamine release due to variable inter-pulse interval, we may develop novel stimulation patterns to modulate neural activity and release. In addition, due to the complexity of neuronal signaling, it is possible that different frequencies could activate different neuronal pathways. Electrical stimulation non-specifically activates or inhibits neuronal inputs which could modulate different downstream release pathways in different ways. Monitoring downstream release in multiple release sites would provide more insight into the overall effect of varying the inter-pulse interval. Exposing the system to variable inter-pulse intervals may create a more responsive stimulation depending on the region being stimulated. Tuning the stimulation pattern to the region being stimulated will be crucial to understanding the complexities in neuronal signaling. Developing computational models to predict dopaminergic release using a variety of stimulation patterns could elucidate the most effective stimulation parameters. Our work shows that dopaminergic release in response to variable inter-pulse intervals is different depending on the targeted region, so models must be able to incorporate differences in direct activation and multi-synaptic activation. It is important to create these computational models so that stimulation parameters can be tuned to the targeted stimulation region for the precise control of dopaminergic release.

Limitations

A limitation of the current study is the use of male-only rodents, and, in the future, it is important to study effects on both males and females to investigate possible sex-specific effects. Furthermore, this study was performed on anesthetized rats which respond to stimulation differently than awake and behaving rats. Finally, this study investigated only a few stimulation patterns and levels of local variance while there are many stimulation patterns that could be investigated.

Conclusions

Understanding the effect of stimulation parameters is vital to the development of techniques such as deep brain stimulation. This work aimed to understand parameters of electrical brain stimulation, especially the effect of variable inter-pulse intervals. We created stimulation patterns with increasing degrees of variability in the inter-pulse intervals, which was quantified by local variance. These stimulation patterns were applied to two brain regions that ultimately evoke dopamine release in the NAc. When stimulating the MFB, the response is more frequency-dependent because the 20 Hz average frequency released more dopamine than the 10 Hz average frequency and the dopaminergic response tracked fast changes in the instantaneous stimulation frequency. When applying the same stimulation patterns to the mPFC, the signal was not as responsive to high-frequency bursts and responded slowly to changes in the instantaneous stimulation frequency. These data suggest that altering the frequency and inter-pulse intervals of stimulation sequences has region-dependent effects. These differences in release can be further investigated using computational methods and more stimulation patterns to better understand how stimulation affects the time-course of dopaminergic release.

Methods

Chemicals.

Artificial cerebrospinal fluid (aCSF) is prepared with 15 mM Tris HCl, 10 mM Tris base, 126 mM NaCl, 2.5 mM KCl, 1.2 mM NaH₂PO₄∙H₂O, 2.0 mM anhydrous Na₂SO₄, 2.4 mM CaCl₂∙2H₂O, and 1.2 mM MgCl₂∙6H₂O. Before adding CaCl₂ and MgCl₂, the pH is adjusted to 7.40 with HCl. The water was purified to 18.2 MΩ∙cm with a Milli-Q Gradient A10 water purification system (EMD Millipore, Burlington, MA). Tris HCl (CAS 1185-53-1), Tris base (CAS 77-86-1), magnesium chloride hexahydrate (CAS 7791-18-6), calcium chloride dihydrate (CAS 10035-04-8), and dopamine hydrochloride (CAS 62-31-7) were purchased from Sigma Aldrich (St. Louis, MO). Sodium sulfate anhydrous (CAS 7757-82-6) and sodium phosphate monobasic monohydrate (CAS 10049-21-5) were purchased from Mallinckrodt Baker (Phillipsburg, NJ). Sodium chloride (CAS 7647-14-5), potassium chloride (CAS 7447-40-7), and hydrochloric acid (CAS7647-01-0) were purchased from EMD Millipore (Burlington, MA).

Electrode Fabrication and Placement.

Carbon-fiber microelectrodes (CFMEs) were fabricated as previously described.⁷⁴ Briefly, a single carbon fiber (AS4, Hexcel, Stamford, CT) was inserted into a glass capillary. The capillary was then pulled into a point with a vertical micropipette puller (PE-2, Narishige, Tokyo, Japan). The exposed carbon fiber was trimmed to approximately 60 μm. Electrical connection was achieved with a silver wire (Kauffman Engineering, Cornelius, OR) coated in alcohol-based graphite conductive adhesive (Alfa Aesar, Ward Hill, MA) fixed in place with epoxy (Loctite, Henkel, Rocky Hill, CT). A PEDOT:Nafion coating was electrodeposited onto the carbon fiber surface to enhance dopamine selectivity and sensitivity in vivo.⁷⁵ Electrodes (n=10) were calibrated in vitro using a aCSF and known concentrations of dopamine. A calibration factor of 20 nA/μM was used to convert current measurements to dopamine concentration. The electrode was placed in the nucleus accumbens (AP: 1.5 mm, ML: 1.4 mm, DV: −6.4 to −7.9 mm from bregma). The Ag/AgCl reference electrodes were prepared by soaking a 0.25 mm diameter silver wire (CAS 7440-22-4, Thermo Fisher, Ward Hill, MA) in 8.25% sodium hypochlorite (The Clorox Company, Oakland, CA) for 24 hours. Reference electrodes were placed approximately 5 mm deep on the contralateral hemisphere.

Bipolar stimulating electrodes were purchased from Plastics One (MS303/1, Roanoke, VA). The electrode was either lowered into the medial forebrain bundle (MFB, n = 5 rats) or the medial prefrontal cortex (mPFC, n = 5 rats). Only one stimulation region was targeted in each rat. When targeting the MFB, the electrode was placed on the medio-lateral plane at the coordinates (AP: −2.5 mm, ML: 1.7 mm, DV: −8.3 to −9.4 mm from bregma). The stimulation patterns were applied as a square waveform with amplitude and pulse width depending on region using a Digitimer DS4 Bi-phasic Stimulus Isolator. The biphasic stimulation patterns in the MFB had current amplitudes of ±300 μA with a 4 ms overall pulse width. In the mPFC, the electrode was aligned with the anterior-posterior plane and lowered to the appropriate coordinates (AP: 3.2 mm, ML: 0.8 mm, DV: −3.1 to −4.1 mm from bregma). The biphasic stimulation pattern in the mPFC had a current amplitude of ±600 μA with a 2 ms overall pulse width. Pulse width was optimized to use longer durations to activate the unmyelinated dopaminergic axons in the MFB and shorter pulse widths to target the cell bodies in the mPFC. The amplitude of the stimulation pulses was adjusted so that the overall charge passed remained constant when comparing MFB and mPFC stimulation.

In some experiments, an electrophysiological array was implanted near the working electrode. Electrophysiological arrays were implanted in 4 of 10 of the animals used in this study. We previously developed a methodology to perform simultaneous electrochemical and electrophysiological measurements.⁶² This methodology limits interference, allowing the data from these measurements to provide a comprehensive understanding of neural activity. The data from these electrophysiological measurements will be used to understand the relationship of neural activity to dopamine dynamics in future publications.

Animals and Surgery.

Male Sprague Dawley rats (310–400 g, Envigo, Indianapolis, IN) were housed two per cage in a reversed 12-hour light/dark cycle room. Food and water were provided ad libitum. On the day of surgery, rats were anesthetized with 3.5% isoflurane gas (VetOne, Boise, ID) and a 1.5 L/min flow of oxygen. Ophthalmic ointment (Dechra Veterinary Products, Overland Park, KS) was applied to the eyes. During the surgery, isoflurane was lowered until reaching 1.5% and then maintained until the end of the surgery. The temperature of the rat was maintained with a water-circulating heating pad (T/Pump, Stryker, Portage, MI). The breath rate was monitored with a RespiRat, a device that utilizes a pressure sensor to monitor breath rate during surgery. Breathing rate was reviewed by surgical personnel every 15 minutes for the duration of the experiment. All procedures were in accordance with the NIH Guidelines for the Care and Use of Laboratory Animals and approved by the University of Arizona Institutional Animal Care and Use Committee standards.

Histology.

After surgery, the brain was harvested and kept in 4% paraformaldehyde (MP Biomedicals, Irvine, CA) for at least 48 hours. The brain was then moved to a 30% sucrose solution. Coronal sections of brain tissue at 40 μm thickness were prepared and electrode placement was verified. The brain slices were Nissl stained as previously described⁷⁶ to determine stimulation probe locations (Figure S1B & C). In some experiments, a fluorescent dye (DiI’; DiIC18(3), Invitrogen, Thermo Fischer Scientific, Waltham, MA) was used to coat the glass casing of the carbon-fiber microelectrodes to better visualize the fiber location. These slices were counterstained with DAPI (Vectashield H-1500–10, Vector laboratories, Newark, CA) and imaged (Figure S1A).

Fast-Scan Cyclic Voltammetry (FSCV).

A Dopamine And Neural Activity (DANA) system⁶² was used to acquire the signal. The DANA system was controlled using custom LabVIEW software (National Instruments, Austin, TX), WCCV 4.0. For FSCV measurement, a 60-Hz, triangular waveform was applied from −0.4 V to +1.3 V vs. Ag/AgCl at 400 V/s to cycle the carbon-fiber electrodes until the background current stabilized. The collection frequency for the experiment was lowered to 5.0 Hz so that interference with electrophysiological measurements was minimized and the stimulation pattern pulses did not overlap with the voltametric waveform.⁶² The background was allowed to stabilize before starting the experiment.

Stimulation patterns.

Stimulations were delivered every 5 minutes and lasted 10 seconds. Each stimulation has a mean stimulation frequency of 10 Hz or 20 Hz and a quantified inter-pulse interval variability which was set along a scale from fixed inter-pulse interval to “bursty” (See Quantifying Variability below). Stimulation sequences were generated using custom code in MATLAB. Individual pulses that overlapped in time with the FSCV triangular pulses were shifted ahead or behind the 8-ms pulses to limit electrical interference. Stimulation sequences with a desired level of variability were created by first generating a large corpus of over 1000 stimulation patterns. This corpus was created by randomly pulling sequences of inter-pulse intervals from exponential and gamma distributions (exprnd and gamrnd in MATLAB). Stimulation patterns with a desired level of variance were then selected from this corpus and constrained such that the inter-pulse intervals had to be at least 5 ms in duration and individual pulses could not overlap in time with a scan pulse. For each mean stimulation frequency, 10 Hz and 20 Hz, four stimulation patterns with different degrees of variance of the inter-pulse interval were selected for experimentation (Table 1). Each of these patterns were repeated 5 times in a random order throughout the experiment for a total of 40 stimulations applied.

Quantifying Variability.

Local variance (LV) is a measure of the variability of a time-series of events. Local variance quantifies the extent of the regularity of the inter-pulse intervals so that regular patterns have a local variance of approximately zero, Poisson series of events have an local variance of approximately one, and “bursty” patterns have an local variance greater than one.⁴⁸ The local variance measure differs from other measures of inter-pulse interval variance as it is robust to changes in stimulation or firing rate. The local variance of a given stimulation pattern (Table 1) was determined by Equation 1.⁴⁸

Data Analysis.

Data analysis was performed using WCCV 4.0 (Knowmad Technologies, LLC). To remove non-dopaminergic signals, such as pH and ionic changes, a differential measurement between the dopaminergic signal and interferents was performed. Data were smoothed by gaussian smoothing with a window of 0.6 seconds. Statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA) and R (RStudio, Boston, MA, USA). Analyses performed were linear regression, t-test, and F-test, 1-way ANOVA and 2-way ANOVA with post-hoc tests.

Abbreviations

DBS

Deep brain stimulation

MFB

Medial forebrain bundle

mPFC

Medial prefrontal cortex

NAc

Nucleus Accumbens

LV

Local variance

ISF

Instantaneous stimulation frequency

FSCV

Fast-scan cyclic voltammetry

CFME

Carbon-fiber microelectrode