Using Neurons to Realize Control Hierarchies
So far, I have illustrated the idea of top-down control without invoking the nervous system. This is appropriate since there is a great deal of control in single-cell organisms. Such control is required to integrate the activities of multiple mechanisms so that each performs as needed to maintain the existence of the organism. As we have seen, this control is typically exerted by altering the constraints in the mechanisms that channel and distribute energy into the performance of work. Moreover, yet higher levels of control can modulate lower levels. Within bacterial colonies there is differentiation of tasks between individual bacteria, and signaling systems exist that enable the colony to alter the operation of control mechanisms in individual bacteria. This differentiation of function and accompanying hierarchical control becomes even more manifest in multi-cell organisms. One of the central modes of control is achieved through the emergence of specialized cells, neurons, with long projections (axons and dendrites) from the cell body. Neurons conduct electrical charges along these projections until a synapse is reached. There they release transmitters that can excite other cells. (In some cases, electrical signals are directly communicated to other cells through what are known as gap junctions.)
Much thinking about neural control systems has adopted Charles Scott Sherrington’s (1923) view of the nervous systems as largely a reactive system in which sensory stimulation initiates a sequence of neural activity culminating in a motor response. On this view, the importance of the nervous system is to enable organisms to respond appropriately to conditions in their internal or external environments. In simple reflex cases, nerves from sensors control motor outputs, enabling them to respond appropriately to stimulus condition. When determining the needed response is more difficult, a network of neurons intervenes. If appropriately configured, such networks can learn to respond differentially to the encountered circumstances. To then exercise control, some of the neurons must connect directly to other tissues that perform physiological processes or motor actions. On this view, which Keijzer (2015) characterizes this as an input-output conception of the nervous system, brains are hierarchies of complex networks. Networks higher in the hierarchy control those lower and the network at the lowest level controls the motor outputs.
On this reactive input-output view of the nervous system, one would expect an organism to remain passive until it received input. But observing any animal confounds this assumption—animals are endogenously active. This is true not just of animals; even single-celled organisms are characteristically active both in carrying out basic life functions and in moving through space. Reversing the usual perspective, activity might be viewed as the default state with special arrangements required in order to stop activity. From this perspective, what the nervous system must do is constrain endogenous activity so as to enable coordinated action. (Keijzer thus contrasts the input- output view with what he terms the coordination view. For him, the first neurons to evolve served to coordinate contractile tissues so as to generate locomotion. Even if coordinating motility was the original role of neurons, they provided as well a basis for coordination of other activities, including more basic physiological functions.)
Fundamental to the coordination view is the contention that the systems that need coordination are endogenously active. A similar assumption is appropriate for the neurons that specialize in coordination. Within Sherrington’s laboratory, Thomas Graham Brown (1914) offered just such a view of the nervous system. Although ostensibly investigating reflexes, he began to attend to the endogenous rhythmic activity that persisted even in deafferented legs in rabbits and other mammals. This research received little uptake at the time. It was revived, however, in research on central pattern generators—networks of neurons that are active in generating cycles of motor activity without external stimulation (Wilson (1961)). More recently, central pattern generators have been found to control a great variety of other neural activity including visual and olfactory processing and cognitive activities including memory formation.
Neural pattern generators require ongoing physiological activity within neurons (resulting from constraining the release of free energy within them) and a mode of organization (itself either within or between neurons) through which the products of these activities constrain others activities. Neurons and the nervous system are endogenously active systems (Bechtel 2013) that can then control other mechanisms. The mechanism for generating circadian rhythms discussed above is one example of an endogenously active control system, but there are many others found in the nervous system. For these endogenously active neural mechanisms to control other biological mechanisms, they must affect constraints in these mechanisms. Sometimes a complex set of operations intervenes between the neural controller and the controlled organs. In the case of muscles, for example, those neurons whose axons synapse onto muscles release neurotransmitters that bind to receptors on the muscle. This generates an electrical current within the muscle cell that leads to a release of calcium from the sarcoplasmic reticulum into the cytoplasm. There the calcium reacts with troponin, causing it to bind to tropomyosin, which was blocking the binding sites between actin and myosin. This then permits the cycling of cross-bridges that cause actin and myosin filaments to pull each other in. This continues until the electrical current ceases, stopping the release of calcium. In this scenario, different constraints are modified in sequence resulting in releasing the endogenous interaction of actin and myosin filaments.
Once neurons evolved as cells that could control the operation of other cells by altering constraints in them, the path was open for creating a hierarchy of such constraints. Constraints in individual neurons could be modified by activity in networks of neurons, and yet higher-level networks could operate on neurons in these networks. I will illustrate this potential by returning to the example of the circadian feedback mechanism operative in individual cells, including individual neurons. In animals, either collections of neurons (e.g., in fruit flies) or whole nuclei (in mammals) assume a regulatory role with respect to the oscillators in individual cells. In mammals, a structure known as the suprachiasmatic nucleus (SCN) performs this function. If the SCN is surgically removed, the animal ceases to exhibit circadian rhythms in behavior or in physiological function (Moore and Eichler 1972). If slices from the removed SCN are maintained in an appropriate medium, the neurons continue to generate circadian rhythms (Herzog et al. 2004), indicating that slices of the SCN can function autonomously. If, however, SCN neurons are dispersed so that many of the connections between them are lost, individual cells still oscillate, but with substantially varying periods, ranging from 21.25 to 26.25 hours with a SD of 1.2 hours (Welsh et al. 1995). Since individual oscillations are out of phase with each other, there is no detectable rhythm in the overall populations. Given that regular rhythms are found in normal SCN tissue in which cells communicate, the communication must synchronize the endogenous oscillations. Thus, collectively the cells of the SCN regulate each other’s behavior, resulting in far more reliable timekeeping than individual neurons can produce. This top-down effect from the population to the individual results from many individual
SCN neurons sending signals to which others can respond by advancing or delaying their own oscillation.
As I noted above, individual cells in mammals possess the requisite mechanism for generating circadian rhythms. What they lack is the ability to synchronize the rhythms in individual cells. This requires the SCN, which functions as a controller on their rhythms. How a signal is communicated from the SCN to other cells of the organism is not understood. When Ralph et al. (1990) removed the native SCN in a hamster and inserted the SCN from a mutant strain that exhibited short periods into a ventricle, they succeeded in restoring some circadian behavior but with a short period. Since the inserted SCN did not make neural projections, its effects on other tissues must have been through hormones. But the fact that not all behavioral or physiological rhythms could be restored suggests that the effect of the SCN on other mechanisms may require neural transmission.
Since circadian rhythms, as the name implies, have a period of only approximately 24 hours, it is important that SCN cells also be entrained to the external environment by sensory information. Otherwise, after a few days an organism will be out of phase with the light-dark cycle in its environment. In fact, one of the initial clues that the SCN was the central clock was that it receives projections from the retina. After the details of the circadian mechanism were discovered, researchers identified the pathway by which the signal from the retina serves to enhance the concentration of PER within a population of SCN cells. If the signal is received around expected dawn, when PER levels are beginning to increase, the signal serves to advance the phase of the oscillation. If, on the other hand, it is received around expected dusk, it serves to delay the phase. The retina thus provides higher-level control over the SCN, which in turn regulates individual cells throughout the body that directly affect the transcription of many proteins which figure in basic activities of organisms. Moreover, one can even view the retina as part of a higher-level control circuit that includes the locomotor system and decision-making operations since exposure to light is also affected by the behavior of the organism. This is particularly true of nocturnal organisms, which must exit their burrows to receive light input. Such higher-level intervention is also a factor in us: when humans expose themselves to light at night (e.g., in performing shift work), they cause their circadian rhythms to be desynchronized from the light cycle in their environment. This in turn frequently results in obesity, diabetes and various cancers.
The circadian system is just one example of a hierarchical control system realized through neurons. There is not space to describe others in detail, but the basic pattern is the same. As research with decorticated animals makes clear, basic motor activity is retained, but less coordinated, when neural control is removed. Sub-cortical brain regions provide a great deal of the needed control. Cortex serves as a higher-order control system that is linked to subcortical ones through numerous loops involving projections both up to cortex and back down to sub-cortical areas. At each level, researchers are identifying complex mechanisms that maintain their own dynamical behavior while modulating constraints in ones lower in the constraint hierarchy.