Multi-Scale Models of Neural Mechanisms Controlling Breathing in Mammals

研究涉及进一步开发神经元和网络的新型神经动力学模型，包括在啮齿动物大脑中并行实验研究的呼吸神经控制系统。开发的基于数据的模型包括：（1）脑干呼吸神经元的生物物理学现实的细胞水平计算模型，结合了细胞结构和生物物理特性的当前信息，如神经元活动的离子电导机制;（2）脑干呼吸神经网络的大规模模型，结合了网络功能和结构结构的可用信息。这些建模研究的总体目标是获得机制的见解的方式，其中细胞和电路级的性能集成到微电路以及大规模的呼吸网络的动态操作的哺乳动物呼吸神经控制系统。进一步发展了啮齿动物脑干中呼吸中枢模式生成（CPG）网络的新模型，该模型由分布在连续排列的脑干结构隔室中的相互作用的兴奋性和抑制性子网络组成，每个子网络在呼吸神经活动模式的生成和控制中具有不同的功能作用，这些呼吸神经活动模式在吸气和呼气的正常呼吸周期中演变。该CPG模型中使用的基本网络结构和细胞特性来自于在大鼠脑干-脊髓原位中进行的电生理学和神经解剖学重建研究，以及在具有活性电路的体外活体脑干切片制备物中分离的子网络。这些模型还纳入了第一次调节不同的电路组件的建模传入输入信号，包括节奏活跃的输入，从关键的神经调节控制系统，已知参与呼吸模式的产生。对于CPG网络操作的动态分析，还应用来自动力系统理论的方法来识别电路操作的关键动态变量和参数，这些变量和参数是呼吸节律和模式生成的基础，并控制吸气和呼气神经活动的功能不同阶段之间的有序过渡。微电路和大规模模型的计算机模拟模拟了在不同的体外和原位条件下实验发现的单细胞和神经元群体活动模式的许多特征。一个主要的新的假设来自实验研究和进一步测试这些模型是，产生振荡活动的能力存在于呼吸CPG在多个层次的细胞和网络组织，形成一个动力系统的耦合振荡机制。因此，呼吸节律产生的不同机制可以以脑状态依赖的方式在功能上表达，并且是多种呼吸运动行为的基础，其中一些发生在正常生理条件下，而其他的出现在病理生理传导下，例如在严重脑缺氧（异常低氧的条件）期间。不同层次的蜂窝和网络复杂性模型的模拟进一步证实了这一新概念的可行性，并提供了深入了解所涉及的基本蜂窝和网络机制。我们还发起了模拟方法的实施，涉及集群计算的大型分布式并行处理系统，包括美国国立卫生研究院的Biowulf集群以及桌面超级计算系统，利用图形处理单元（GPU），允许实时模拟大规模的网络模型。在系统水平上，呼吸神经控制系统的模型已经被进一步开发，其将基本神经回路动力学与外周氧和二氧化碳交换、血气运输以及通过诸如氧和二氧化碳的血液/脑水平的信号的中枢呼吸回路的生理反馈调节相耦合。这些模型代表了第一代系统级控制模型，其集成了神经系统结构功能特性的基本要素和呼吸气体交换和运输系统的现实特征。所有这些模型目前都被应用于进一步探索脑干呼吸回路的操作原理和呼吸活动的控制，包括在与脑和身体氧/二氧化碳稳态紊乱相关的各种（病理）生理条件下。

Research involved the further development of novel neurodynamical models of neurons and networks comprising the respiratory neural control system as studied experimentally in parallel in the rodent brain. Data-based models developed included: (1) biophysically realistic cellular-level computational models of brainstem respiratory neurons incorporating current information on cellular architecture and biophysical properties such as ionic conductance mechanisms underlying neuronal activity; and (2) large-scale models of brainstem respiratory neural networks incorporating available information on network functional and structural architecture. The overall objective of these modeling studies was to gain mechanistic insights into the manner in which cellular- and circuit-level properties are integrated into microcircuits as well as large-scale respiratory networks for dynamical operation of the mammalian respiratory neural control system. A new model of respiratory central pattern generation (CPG) networks in the rodent brainstem was further developed consisting of interacting excitatory and inhibitory subnetworks distributed in serially arranged brainstem structural compartments, each with distinct functional roles in generation and control of the respiratory neural activity patterns that evolve during the normal breathing cycle of inspiration followed by expiration. The basic network architecture and cellular properties used in this CPG model were derived from electrophysiological and neuroanatomical reconstruction studies conducted in the rat brainstem-spinal cord in situ and on subnetworks isolated in living brainstem slice preparations in vitro with active circuits. These models also incorporated for the first time regulation of different circuit components by modeled afferent input signals, including rhythmically active inputs from critical neuromodulatory control systems that are known to be involved in regulation of respiratory pattern generation. For dynamical analysis of CPG network operation, methods from dynamical systems theory were also applied to identify critical dynamical variables and parameters of circuit operation that underlie respiratory rhythm and pattern generation and control the orderly transitions between the functionally distinct phases of inspiratory and expiratory neural activity. Computer simulations with the microcircuit and large-scale models mimicked many features of the single-cell and neuron population activity patterns found experimentally under different in vitro and in situ conditions. A major new hypothesis derived from experimental studies and further tested with these models was that the capability to generate oscillatory activity exists within the respiratory CPG at multiple levels of cellular and network organization, forming a dynamical system of coupled oscillatory mechanisms. Thus different mechanisms of respiratory rhythm generation can be functionally expressed in a brain state-dependent manner and underlie multiple respiratory motor behaviors, some of which occur under normal physiological conditions and others of which emerge under pathophysiological conductions such as during severe brain hypoxia (conditions of abnormally low oxygen). Simulations with models of different levels of cellular and network complexity further confirmed the plausibility of this new concept and have provided insights into the essential cellular and network mechanisms involved. We have also initated implementation of simulation approaches involving cluster computing on large distributed parallel processing systems including the NIH Biowulf cluster as well as desktop supercomputing systems utilizing graphics processing units (GPUs) that allow real-time simulation of large-scale network models. At the system level, models of the respiratory neural control system have been further developed that couple essential neural circuit dynamics with peripheral oxygen and carbon dioxide exchange, blood gas transport, and physiological feedback regulation off central respiratory circuits by signals such as blood/brain levels of oxygen and carbon dioxide. These latter models represent the first generation of system-level control models that integrate essential elements of nervous system structural-functional properties and realistic features of the respiratory gas exchange and transport system. All of these models are currently being applied to further explore principles of operation of brainstem respiratory circuits and control of respiratory activity including under various (patho)physiological conditions associated with disturbances of brain and body oxygen/carbon dioxide homeostasis.