† Corresponding author. E-mail:
Project supported by the National Natural Science Foundation of China (Grant Nos. 61473208 and 61876132) and the Tianjin Research Program of Application Foundation and Advanced Technology, China (Grant No. 15JCYBJC47700).
Parkinson’s disease (PD) is characterized by pathological spontaneous beta oscillations (13 Hz–35 Hz) often observed in basal ganglia (BG) composed of subthalamic nucleus (STN) and globus pallidus (GPe) populations. From the viewpoint of dynamics, the spontaneous oscillations are related to limit cycle oscillations in a nonlinear system; here we employ the bifurcation analysis method to elucidate the generating mechanism of the pathological spontaneous beta oscillations underlined by coupling strengths and intrinsic properties of the STN–GPe circuit model. The results reveal that the increase of inter-coupling strength between STN and GPe populations induces the beta oscillations to be generated spontaneously, and causes the oscillation frequency to decrease. However, the increase of intra-coupling (self-feedback) strength of GPe can prevent the model from generating the oscillations, and dramatically increase the oscillation frequency. We further provide a theoretical explanation for the role played by the inter-coupling strength of GPe population in the generation and regulation of the oscillations. Furthermore, our study reveals that the intra-coupling strength of the GPe population provides a switching mechanism on the generation of the abnormal beta oscillations: for small value of the intra-coupling strength, STN population plays a dominant role in inducing the beta oscillations; while for its large value, the GPe population mainly determines the generation of this oscillation.
Parkinson’s disease (PD) is a common chronic neurodegenerative disorder that mainly occurs in the elderly and has an adverse effect on the quality of the patient’s life. The PD is attributed to the degeneration and death of dopaminergic neurons in the midbrain and the substantia nigra pars compacta.[1] Such a pathological degeneration results in the decrease of the dopamine (DA) content of the brain, which causes the abnormal discharges to occur in BG nuclei of both patients[2] and animal models[3] of PD. Hence, the decrease at the level of dopamine severely affects the neuronal activities in the cortico-basal ganglia (BG)-thalamocortical loops.[4,5] As is well known, the rigidity and bradykinesia[6] are the main Parkinsonian symptoms, which are closely correlated with oscillations in the beta band frequency (12 Hz–35 Hz) in the basal ganglia.[7]
The exact pathophysiological origin for the appearance of the synchronized beta oscillations remains controversial.[8–14] Some studies have shown that the pathological beta oscillations were generated in the striatum and then spread to the rest of the loop.[12] The striatal theory suggested that the STN–GPe model could generate beta frequency oscillations when the input from striatum to GPe increased.[9,11] Others suggested that beta oscillations originated from the cortico-thalamic circuit and spread to the BG nuclei,[8] and there were also studies suggesting the beta oscillations were generated internally within the BG network model.[10,13,14] Overall, most of previous studies have shown that beta oscillations can be observed in the dopamine-depleted BG nuclei composed of STN and GPe, and in the striatum and cortex.[15,16]
Computational research is becoming a promising method to elucidate the mechanism behind the various neurological diseases, such as Parkinson’s disease[5,17,18] and absence seizures.[19–21] The STN and GPe in the basal ganglia connect to each other to construct a negative feedback loop, a typical neural motif for the generation of limit cycle oscillations. Thus, the STN–GPe feedback circuit model describes the possible origin of the abnormal beta rhythms.[10,13,14,22] The study of Holgado et al.[23] indicated that the coupling strengths between STN and GPe need to be sufficiently strong for the generation of beta oscillations in the STN–GPe model. Fei Liu et al.[24] proposed a neuron mass model of the basal ganglia composed of STN and GPe networks to reproduce spontaneous beta oscillations related to Parkinson’s disease by changing the coupling strengths and intrinsic parameters of the model.
Spontaneous oscillations are a special behavior in nonlinear systems, and related to limit cycle oscillations, an intrinsic phenomena of nonlinear dynamical systems.[25,26] The generation of spontaneous oscillations in the STN–GPe circuit model is primarily attributed to the sigmoid function.[27] In the study of Holgado et al.,[23] for the simplicity of analysis, the sigmoid functions in both STN and GPe populations are linearized with a linear function. Thus, the linear stability theory is used to derive the conditions for the generation of beta oscillations in the subthalamic nucleus–globus pallidus network. In the study of Fei Liu et al.,[24] to use the describing function method to explain the generation of excessive beta oscillations, the sigmoid function in the GPe population is approximated by its equivalent gain. However, replacing the sigmoid function by its linear form[23] or equivalent gain[24] may result in analysis errors due to the model approximating error. Moreover, most of previous studies conducted simulations to determine the boundary between the regions generating normal and abnormal beta activities in the two-dimensional model parameter space.[23,24] Thus, it still remains to be addressed by using a nonlinear analysis method to probe the dynamical mechanism behind the spontaneous beta oscillations in the STN–GPe circuit model.
In addition, Liu et al. found that the upper beta frequency originates from the high frequency cortical input, and there is a transition mechanism between the upper and lower beta oscillatory activities underlined by inhibitory self-feedback within the GPe.[27] Furthermore, the study of Liu et al. indicated that the STN primarily influenced the generation of the beta oscillation while the GPe mainly determined its frequency.[24] It is still unknown why the two populations play different roles in generating the abnormal beta oscillations, and what is responsible for the transition mechanism between the upper and lower beta oscillations.
A bifurcation analysis method, as an effective means to research the dynamic behavior of nonlinear systems, has been extensively used to study the neurological diseases,[25,28–30] such as epileptic seizures.[29,31] Here we employ the bifurcation analysis method to elucidate the mechanism for generating the pathological spontaneous beta oscillations underlined by coupling strengths and intrinsic properties of the STN–GPe circuit model. Moreover, by plotting frequency distribution in the two-dimensional model parameter space, we further reveal the regulating mechanism of the spontaneous oscillations frequency.
The rest of this paper is organized as follows. In Section
The STN–GPe feedback neural circuit model is proposed in previous studies to investigate the generation of pathological beta oscillations related to Parkinson’s disease (PD).[23,24,27] In the present study, this model is employed to further investigate the model parameters underlined generation mechanisms of beta oscillations associated with Parkinson’s disease. The architecture of the STN–GPe model is given schematically in Fig.
Both STN and GPe neuronal populations are composed of the linear convolution operator TSTN(t) or TGPe(t) modeling excitatory or inhibitory synapse, and the sigmoid nonlinear function modeling spike generation close to the populations. The mathematical formulation at the population level is extensively used in some previously proposed models,[24–27,32–34] which were originally used to describe the neural activities of cortex.
The mean firing rate of every population is computed from its average membrane potential by the sigmoid function. Thus, for STN and GPe, the two functions of sigmoid shape are given by
For both STN and GPe populations, the synapses transfer mean firing rates of every population into their average membrane potential modeled by the two linear convolution operators TSTN(t) and TGPe(t), in the form of second-order linear functions as follows:
All the parameters of the STN–GPe circuit model are given in Table
![]() | Table 1. Parameter interpretation and values in STN–GPe feedback circuit model. . |
Bifurcation analysis is an efficient nonlinear method to elucidate the regulating mechanisms of the dynamics of a nonlinear dynamical system, and extensively employed to study the effects of model parameters on the dynamics of various neural models.[28–30] In the present study, to detect the beta pathological oscillations in the STN–GPe model, we conducted bifurcation analysis with respect to parameters of the model by using XPPAUT, a valid tool for simulating and analyzing dynamical systems.[41] To explore the influence of model parameters on the oscillation frequency, frequency curves with respect to model parameters and frequency distribution diagrams in two-dimensional parameter plane were plotted by using the MATLAB software.[42] The Runge–Kutta numerical integration method could be used to solve the differential equations at a zero initial condition.
To probe the role played by STN and GPe populations in generating the limit cycle oscillation, we define a measure named effect factor K to quantitatively evaluate the effect of the model parameters on the generation of spontaneous oscillations. The effect factor is defined as K = Δ2/Δ1, where Δ1 represents the change of the model parameter value, and Δ2 the change of the spontaneous oscillation region caused by Δ1. In Fig.
We first conduct codimension-one bifurcation analysis to study the influence of coupling strengths between STN and GPe populations on the spontaneous beta oscillations. In addition, through codimension-two bifurcation analysis, we explore how the model parameters interact to exert an influence on the generation of abnormal spontaneous beta oscillations, including inter-coupling strengths (for example, C1 and C2) between STN and GPe populations, and the properties of STN and GPe populations itself, such as synaptic gains (i.e., Hs and Hg) and sigmoid elements (for example, es and eg, and rs and rg). Especially, we focus on exploring how the self-feedback of the GPe (i.e., C3) exercises an influence on the spontaneous beta oscillations.
Previous experimental and simulation studies have suggested that inter-coupling connection between STN and GPe and inhibitory intra-coupling connection in GPe play important roles in inducing beta frequency band behavior in the STN–GPe model.[23,24] In this subsection, to explore the influence of the coupling strengths on the generation of pathological beta oscillations, we conducted bifurcation analysis and plotted the curves of the frequency versus the coupling strength, while the other parameters had default values as listed in Table
We first plotted the codimension-one bifurcation diagram versus coupling strength C, letting C2 = C, C1 = 0.5C, and C3 = 0.5C, which means that the inter-coupling strength between STN and GPe populations and intra-coupling strength of GPe population change in the same way. The bifurcation result is illustrated in Fig.
![]() | Fig. 3. (a) Codimension-one bifurcation versus C, where HB is Hopf bifurcation point. (b) Frequency versus C. |
The bifurcation result in Fig.
Furthermore, the curve of frequency versus coupling strength C as illustrated in Fig.
Next, we draw the diagrams for codimension-one bifurcation versus coupling strengths C1, C2, and C3, respectively, (and other parameters remain the default values as listed in Table
The effects of coupling strengths C1, C2, and C3 on the frequency are shown in Fig.
We further conduct theoretical analysis to explain why and how the increasing of C3 can enhance the generation of spontaneous neural oscillations and causes the oscillation frequency to increase.
According to Fig.
![](cpb_29_5_058701/cpb_29_5_058701_ieqn1.gif)
Next, letting ω = 0 in Eq. (
In this subsubsection, we explore the influence of the interaction of coupling strengths on pathological beta oscillations. Codimension-two bifurcation results versus coupling strength C1 and C2 are illustrated in Figs.
In Figs.
To probe how the coupling strengths cooperate with each other to regulate the frequency of spontaneous oscillation, the frequency distribution diagrams of the spontaneous oscillations in the two-dimensional parameter space of C1 and C2 are obtained as shown in Figs.
It should be noted that the above results are in agreement with the results of Holgado et al.,[23] but inconsistent with those of Liu et al.,[24] where the increase of C1 results in the fact that the spontaneous neural oscillations in the STN–GPe model disappear (for the details see Fig.
The main idea of describing function method is to split a nonlinear system into two parts: nonlinear part N(A) and linear part L(jω). Solving the equation 1 + N(A)L(jω) = 0, i.e., L(jω) = −1/N(A), one can determine the limit cycle oscillations. For convenience, plotting the line of −1/N(A) and Nyquist diagram of L(j ω), if the two sides are crossed, there exists limit cycle oscillations in the nonlinear system.[43] As an effective method to analyze the limit cycle oscillations of nonlinear system, the describing function method is extensively used to gain an insight into the mechanism of neural spontaneous oscillations.[24,26]
In line with the study of Liu et al., we transform the sigmoid function in STN population into N(A), and the others in the STN–GPe model into linear part
![]() | Fig. 6. Describing function analysis results of spontaneous oscillations with different values of C1 and C2, showing (a) line (yellow) of −1/N(a) and (b) Nyquist diagrams of L(jω). |
Moreover, according to the principle of the describing function method, we can obtain the following function defining the stability boundary in the two-parameter plane (C1,C2):
In summary, the increasing of inter-coupling strengths C1 and C2, and the decreasing of intra-coupling strength C3 can cause the spontaneous beta oscillations to be generated in the STN–GPe model. It can be concluded that the STN–GPe circuit model can be shifted into a Parkinson-related oscillation state by changing the coupling strengths.
Previous studies demonstrated that increase of the coupling strength induces the the beta oscillation to be generated, as well as the oscillation frequency to be enhanced.[24] Our study reveals that the role of inter-coupling strength between STN and GPe populations played in generating the abnormal oscillations is different from that of intra-coupling strength of GPe, thus their influences on the oscillations should be separately studied. The present results reveal that the increase of inter-coupling strength between STN and GPe populations induces the spontaneous beta oscillations to be generated, and causes the oscillation frequency to decrease, while the increase of intra-coupling strength of GPe can prevent the model from generating the oscillations, and dramatically reduce the oscillation frequency. The present conclusions have some differences from those in previous study,[24] the main reason lies in the fact that Liu et al. combined inter-coupling strength and intra-coupling strength into one parameter (represented by C), thus drawing a confusing conclusion. Furthermore, we conduct a theoretical analysis to probe the influence of the inter-coupling strength of GPe population on the oscillations, revealing that the increase of inter-coupling strength leads the equivalent gain to decrease and the natural frequency to increase. The theoretical analysis results provide an insight into the generating mechanism and regulating mechanism of the oscillations induced by the inter-coupling strength of GPe population.
The study of Liu et al. claimed that the upper beta frequency in STN–GPe model originates from a high frequency cortical input, and the intra-coupling strength of GPe plays a transition role in determining low and upper frequency of the beta oscillations,[27] however the triggered mechanism is unclear. Our study reveal that the upper beta frequency can be generated by the model itself as long as the intra-coupling strength is large enough. Furthermore, the theoretical and simulation analysis further show that the intra-coupling strength can dramatically regulate the oscillation frequency by changing the natural frequency of the STN–GPe model. These results can shed light on the switching mechanism of the spontaneous beta oscillations triggered by the intra-coupling strength.
In addition to the coupling strengths in the STN–GPe model, nonstructural changes in the STN–GPe circuit model, such as alterations of synaptic and intrinsic properties, are also critical for affecting dynamics behavior of the model.[27,44] In this subsection, we further explore how the STN and GPe populations exert influence on the spontaneous beta oscillations.
First, we explore the influences of synaptic properties on the generation of pathological beta oscillations in the model. There are two parameters describing the synaptic properties of the STN–GPe model, i.e., excitatory and inhibitory average synaptic gains Hs and Hg. To explore how the two synaptic parameters interact with each other to induce abnormal Parkinsonian oscillatory behaviors, we keep the other parameter values listed in Table
The bifurcation results are shown in Figs.
As shown in Figs.
To detect how Hs and Hg exert the influences on the frequency of beta oscillations, the frequency distribution diagrams in the (Hs, Hg) plane are given in Figs.
Finally, we argue that intrinsic properties, such as sigmoid parameters (i.e., es, eg, rs, and rg), are also capable of affecting the dynamic behaviors of the model, and induce pathological rhythmic activity related to Parkinson’s disease.
To verify our hypothesis, the diagrams of bifurcation and frequency distribution in both (es, eg) and (rs, rg) planes are derived to explore the influence of sigmoid elements on the generation of the spontaneous oscillations. Codimension-two bifurcation results with respect to es and eg, and with respect to rs and rg are illustrated in Figs.
To quantitatively evaluate the role of intra-coupling C3, we further investigate the effect factor of sigmoid parameters on the generation of spontaneous oscillation as shown in Figs.
The frequency distribution diagrams in the plane of (es, eg) and (rs, rg) are shown in Figs.
Previous studies showed that the STN primarily influences the generation of the beta oscillation while the GPe mainly determines its frequency. However, our study reveals that the intra-coupling strength of the GPe population plays a switching role in inducing the spontaneous beta oscillations through determining which population exerts a more important influence on the generation of the oscillations. For small value of the intra-coupling strength, the STN plays a dominant role in generating the beta oscillations; for large value of the intra-coupling strength, the GPe mainly determines the generation of the beta oscillations.
In the present study, we mainly employ the bifurcation analysis method to elucidate the mechanism of generating the pathological spontaneous beta oscillations triggered by coupling strengths and intrinsic properties of the STN and GPe populations. The results reveal that the increase of inter-coupling strength between STN and GPe populations enhances the generation of spontaneous beta oscillations, and results in the decrease of oscillation frequency. However, the increasing of intra-coupling strength of GPe can prevent the oscillations from being generated, and thus dramatically increasing the oscillation frequency. We further conduct theoretical analysis to provide an insight into the role of the inter-coupling strength of GPe population played in the generating and regulating of the oscillations. Furthermore, our study reveals that the intra-coupling strength of the GPe population can reshape the interacting relationship between STN and GPe populations, and determine which population exerts a more important influence on the generation of the spontaneous beta oscillations. The present results provide testable hypotheses for future experimental work.
In the present study, we focus on exploring the generation of the spontaneous beta oscillations caused by the excessive changes of parameters in the STN–GPe model. However, the inputs of the STN–GPe model from the cortex and striatum should also exert influence on the beta oscillations. In future study, we will further probe the roles of the inputs from cortex and striatum played in inducing physiological beta oscillations.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] | |
[22] | |
[23] | |
[24] | |
[25] | |
[26] | |
[27] | |
[28] | |
[29] | |
[30] | |
[31] | |
[32] | |
[33] | |
[34] | |
[35] | |
[36] | |
[37] | |
[38] | |
[39] | |
[40] | |
[41] | |
[42] | |
[43] | |
[44] |