How the Vertebrate Brain Regulates Behavior

by Donald Pfaff

  The next step was to record from individual neurons in the dorsal root ganglia at lumbar levels, which cover the appropriate cutaneous area effective for lordosis (Kow and Pfaff 1979). Mapping the receptive fields of neurons in all lumbar dorsal roots revealed an orderly pattern proceeding posteriorly from L1 (flanks) through L6 (tail base and perineum) (Kow and Pfaff 1975). The primary somatosensory neuron type responsible for triggering the onset of lordosis behavior was determined by a process of elimination. Noncutaneous units, unresponsive neurons, and neurons primarily responsive to hair deflection could not be crucial. Instead, dorsal root ganglion neurons responsive to very low pressure, especially sustained pressure (as opposed to punctuate skin deformation) should be involved, so-called type II neurons. These are associated with Ruffini endings in the skin. Thus, pressure in the L6 receptive field of the female deforms Ruffini endings leading to an activation of type II neurons, which yield the input to the lumbar spinal cord effective for the sensory stimulation of lordosis behavior.

  In the spinal cord as well, neurons in the dorsal horn at lumbar levels showed strong responses to the types of cutaneous stimuli on the flanks and perineum that in the unanesthetized female rat would elicit lordosis. The big difference with dorsal root ganglion neurons was that, while the latter were absolutely silent in the absence of stimulation, dorsal horn neurons often had spontaneous activity (Kow, Zemlan, and Pfaff 1980). Among the dorsal horn neurons excited by lordosis-relevant cutaneous stimulation, some would show a sudden, stepwise increase in firing as soon as pressure was applied to the skin; others increased their firing gradually, tracking a steady increase of pressure on the skin.

  We were very pleased to see effects of estrogen treatment on sensory responses in this system. At the level of the primary sensory neuron, Lee-Ming Kow measured the receptive field size of the pudendal nerve, the electrical activity of which would obviously be important for causing the vertebral dorsiflexion of lordosis; it covers the perineal and hind leg skin area contacted by the male (Kow and Pfaff 1973). A large enough number of animals tested and the precise methodology used allowed the 22 percent increase caused by estrogen treatment (compared to estrogen-free, ovariectomized controls) to be highly statistically significant. At the level of the dorsal horn of the spinal cord as well there potentially lies the cellular basis for an estrogen effect on sensory processing as Joan Morrell found small numbers of estrogen-binding cells there (Morrell et al. 1982).

  Ascending Pathways: How Behaviorally Relevant Sensory Information Gets Transmitted to the Brain

  Spinal circuitry is not adequate, not sufficient for lordosis behavior to be performed; however, sophisticated sensory-motor circuitry at each spinal cord level might be. We prepared female rats with complete transections of the spinal cord at low thoracic levels (Kow, Montgomery, and Pfaff 1977). Good postoperative care kept these animals in amazingly good physical condition (for example, they were voided, cleaned, and fed if necessary)—only eighteen of the ninety-two rats operated on showed a deteriorating condition. Even subtotal transections of the spinal cord could eliminate lordosis. For example, spinal operations that left the dorsal columns intact or the dorsolateral columns or the ventromedial columns intact did not permit lordosis behavior to occur. If the anterolateral columns, containing the ascending spinoreticular system, were left intact, then we could observe lordosis. In contrast, if much of cord was left intact but the anterolateral columns were badly damaged, then percentage of lordosis performance was low or zero.

  We concluded that supraspinal control is required for normal lordosis behavior, and that fibers necessary and sufficient for lordosis run in the anterolateral columns. For ascending sensory signaling, that would include spinoreticular fibers. (For descending systems that would mean vestibulospinal and reticulospinal systems.)

  The distribution of these fibers ascending from the spinal cord is quite clear. Anne Robbins, under the leadership of Susan Schwartz-Giblin in my laboratory, microinjected the retrograde neuroanatomical tracer Fluoro-Gold among the cells in the medullary reticular formation, the large cells called nucleus gigantocellularis (NGC). Significant numbers of spinal cord cell bodies were backfilled in the deep layers of the dorsal horn, near the central canal and even in the ventral horn. They were found both ipsilateral to the microinjection and contralateral. Control injections far lateral to NGC yielded only an extremely small number of backfilled cells in the spinal cord (Robbins, Schwartz-Giblin, and Pfaff 1990).

  This neuroanatomical result matched well the electrophysiological results from extracellular single neuron recordings by Lee-Ming Kow (Kow and Pfaff 1982). In anesthetized female rats we were able to find a large number of NGC neurons (among the 742 cells studied) that responded to the type of cutaneous stimulation that elicits lordosis behavior—pressure on the flanks, rump, and tail base. Recording in unanesthetized animals was harder—the rats had to be surgically prepared with implanted “floating” wire electrodes. Nevertheless, of the hundred and eighteen cells we could study, forty-two responded. Estrogen pretreatment of the females studied under anesthesia tended to facilitate the excitability of these medullary reticular neurons and to increase the ratio of the number of cells excited by lordosis-relevant cutaneous stimulation to the number of cells inhibited.

  In turn, hindbrain reticular neurons project to the midbrain. Joan Morrell used the retrograde neuroanatomical horseradish peroxidase technique with initial injections in the dorsal midbrain, which covered the dorsal mesencephalic central grey and adjacent midbrain reticular formation including the neuronal areas contacted by ventromedial hypothalamic outflow (mentioned above and covered below) (Morrell and Pfaff 1983). Considerable numbers of cells in the medullary reticular formation were backfilled, and even some of the axons of some neurons in the lumbar spinal cord made it up to the dorsal midbrain. This illustrates is the ladder-like character of the ascending side of the lordosis behavior neural circuit: spinal cord projects to the region of NGC, and NGC projects to dorsal MCG and reticular formation; some spinal neurons project all the way to the dorsal midbrain.

  Joan Morrell’s neuroanatomical results are complemented by the electrophysiological data of Yasuo Sakuma, the premier Japanese neuroendocrine physiologist (Sakuma and Pfaff 1980a,b). Yasuo recorded neurons in the MCG and in the mesencephalic reticular formation just lateral to central grey. Large numbers of such neurons responded to lordosis-eliciting cutaneous stimulation (fifteen times as many excitatory responses as inhibitory responses) and also responded to electrical stimulation of medullary reticular NGC (seven times as many excitatory responses as inhibitory responses). What was most exciting was that many also responded to electrical stimulation of the ventromedial hypothalamus (four times as many excitatory responses as inhibitory responses). Thus, in terms of lordosis behavior-circuit building, we got convincing evidence of convergent effects of lordosis-relevant somatosensory and ventromedial hypothalamic influences on central grey cells in the female rat mesencephalon.

  A side point: at several points in the circuitry for lordosis behavior it is clear that not all sensory information is transmitted. As receptive fields grow larger, for example, from primary receptors to higher brain regions, spatial information is reduced. Apparently, for this type of neural circuitry, if a certain amount and form of sensory information is required for a behavior’s regulation, the nerve cell in question receives and responds to it. If not required, the sensory information is less precise or the nerve cell in question gets none at all. Clearly, this arrangement reflects an economy in the use of nerve cell signaling capacities.

  Thus, to meet the estrogen-dependent signals emerging from the hypothalamus, lordosis-relevant sensory information reaches neurons in and just lateral to the MCG (and its adjoining mesencephalic reticular formation). The central grey integrates hypothalamic neuroendocrine signals with brainstem sensory-motor physiology.

  Descending Pathways

  How do central grey neurons exert their behavioral influence? Yasuo Sakuma
used antidromic stimulation techniques while recording from cells of origin of descending projections (Sakuma and Pfaff 1980b,c). The stimulating electrodes were placed in the nucleus gigantocellularis (NGC) of the medullary reticular formation, while recording microelectrodes sampled the electrical activity of cell bodies in the MCG. Antidromic spike latencies ranged from 2 to 20 milliseconds, indicating a range of nerve fiber diameters. Central grey neuronal cell bodies thus identified in these experiments (n = 82) were found near the lateral borders of the central grey in positions that appeared clearly to be in contact with axons descending from the hypothalamus. Most exciting, Yasuo discovered that the electrical activity of some of these central grey neurons is influenced by estrogen treatment and by VMH stimulation (Sakuma and Pfaff 1980b,c).

  Electrical responsiveness was measured by the ability of the antidromic action potential (these neurons project to the medullary reticular formation and thus are likely of a high degree of relevance to lordosis behavior) to propagate from the descending axon into the somatodendritic complex. This spike invasion happened almost twice as often in the estrogen-treated female rats compared with the ovariectomized, estrogen-free controls. Further, there were more neurons with high rates of resting discharge in the estrogen-treated group. Important for the regulation of excitability in the lordosis circuit, we studied the ability of low-intensity VMH stimulation to influence central grey cells that project to the NGC. Routinely, VMH stimulation markedly facilitated antidromic spike invasion to the cell body of central grey neurons, thus facilitating the excitability of those neurons that project down to the NGC (Sakuma and Pfaff 1980b).

  Studies I will describe show that the most important descending systems for regulating and stimulating lordosis are the reticulospinal system and the vestibulospinal system. Rockefeller doctoral student Sandra Cottingham discovered that central grey stimulation facilitates both systems. The most important features of both systems have to do with their elevation of excitability of the motor neurons that cause contraction of the deep back muscles that cause lordosis behavior. Cottingham placed monopolar electrodes in the lateral vestibular nucleus, the origin of a major vestibulospinal tract, and showed that stimulation there could cause electromyographic signals in the deep back muscles that run lordosis behavior (Cottingham and Pfaff 1987). Then, preparing for the critical experiment, she decreased the amount of current going into the lateral vestibular nucleus so that it could not activate deep back muscles. Superimposed upon that lower amount of current, stimulation of the MCG reinstated the ability of vestibulospinal to activate electrical responses in the deep back muscles. Effective stimulation points were in the lateral central grey; if we missed that midbrain target, no potentiation of the vestibulospinal system was possible.

  Ann Robbins, working with a long-time leader in the laboratory, Professor Susan Schwartz-Giblin, followed up Cottingham’s results with neuroanatomical studies (Robbins, Schwartz-Giblin, and Pfaff 1990). The retrograde fluorescent tracer Fluoro-Gold was microinjected into NGC; after a survival time of four days, Ann identified numerous cells in the central grey—both ipsilateral and contralateral to the NGC site of application—as projecting to NGC. These presumably supply the neuroanatomical basis of Cottingham’s findings.

  Thus, in a hierarchical fashion, the cell group in the MCG that receives estrogen-dependent input from the hypothalamus revs up the ability of a lower hindbrain center, the lateral vestibular nucleus, to activate the motor neurons that run lordosis behavior. I further inferred that our reproductive behavioral system had co-opted a regular postural control system—the lateral vestibulospinal tract—in order to produce the female’s behavior required for fertilization and reproduction.

  Then Cottingham used similar strategies for the medullary reticulospinal system, including those descending influences emanating from the NGC (Cottingham, Femano, and Pfaff 1987). In the medullary reticular formation, her best sites for stimulating the motor neurons of lordosis-relevant axial muscles were in and lateral to the NGC. Ineffective control sites in her work tended to be far lateral or at the very bottom of the brain. As earlier, she set up her experiments by calibrating the stimulus current strength. For example, in one experiment the mesencephalic central grey stimulation itself gave a very small activation of axial muscle motor neurons as measured by axial muscle electromyography, the key to lordosis behavior. An NGC stimulus train of 50 microamps amplitude at a frequency of two hundred pulses per second yielded a larger axial muscle activation. The crucial experiments took place when she superimposed mesencephalic central grey stimulation on NGC stimulation and recorded a vast amplification of electromyographic activity. A key example would be when this combination central grey-on-NGC phenomenon worked on motor neurons for the axial muscle lateral longissimus, a key muscle in executing lordosis behavior.

  Thus, the central grey of the midbrain plays an important role in relaying hormone-dependent hypothalamic information to the medullary reticular formation, which in turn has direct access to the motor neurons essential for lordosis behavior.

  Lower Brainstem

  The reticulospinal link in the lordosis circuit was characterized extensively by Philip Femano, who had come to our laboratory from Rutgers and introduced us to computer analyses of neurophysiological data (Femano, Schwartz-Giblin, and Pfaff 1984a,b). Femano worked under the leadership of Susan Schwartz-Giblin who later became dean of the State University of New York (SUNY) medical school. Fermano anesthetized rats with urethane and used bipolar stimulation methods that were designed to minimize the spread of current, thus to define most precisely the NGC source of the effect on the deep back muscles that execute lordosis. Currents in the range of 15 to 40 microamps were typical. Under these conditions he discovered that all axial, deep back muscles could be activated by NGC stimulation: muscles named transversospinalis, medial longissimus, and lateral longissimus.

  The temporal aspects of the deep back muscle activation were interesting. With the low currents used, a single pulse in NGC was not effective, but a train of pulses always was. NGC circuitry could handle stimulation frequencies as high as two hundred pulses per second. The deep back muscles (and thus their motor neurons) could respond steadily to train after train after train of pulses; in fact, latencies to the onset of muscular activation tended to go down as train numbers increased. The minimal onset latency to the electromyographic activation of lateral longissimus was 12 milliseconds.

  The real-life physiological mode of this effect was analyzed by Rockefeller graduate student Mark Cohen (Cohen, Schwartz-Giblin, and Pfaff 1987a), working with Schwartz-Giblin. We knew that short trains of electrical stimuli to the pudendal nerve could evoke clear-cut responses in the axial muscle motor neurons essential for lordosis. The magnitude of the motor neuron response is drastically reduced by spinal transaction, and the fibers responsible for the supraspinal influence travel in the lateral columns, precisely as do the fibers responsible for facilitating lordosis behavior as a whole. Therefore, Mark asked the question, might NGC stimulation have its effect on the behavior primarily by enhancing the throughput from a lordosis-relevant sensory stimulation to the motor neurons essential for lordosis?

  Mark set up the experiment this way. Pudendal nerve stimulation intensity was adjusted so that, by itself, it gave only a small increase in the electrical response recorded in the lumbar-level nerves leading to the deep back muscle lateral longissimus. Likewise, NGC stimulation was calibrated such that hardly any electrophysiological response was seen in the nerve for this deep back muscle. The crucial experiment came when the two types of stimuli were combined (Figure 2.3). The answer was that concurrent stimulation of NGC combined with the pudendal nerve led to a very high motor nerve response, about ten times the number of spikes of either stimulus site alone. Thus, we understand how, in the lordosis circuit, NGC neurons (which themselves had received estrogen-dependent inputs from the ventromedial hypothalamus via the MCG) can greatly amplify lordosis-relevant motor neuron responses to lordosis-r
elevant stimulus inputs.

  Lateral vestibular nucleus stimulation (Modianos and Pfaff 1977) also facilitates lordosis. Low-intensity (10 microamps) stimulation was delivered bilaterally at two hundred pulses per second. The magnitude of the lordosis increase varied from female to female, but we knew we had a phenomenon; the increase during bilateral stimulation usually was a doubling of lordosis strength and in some cases much more. Unilateral stimulation also worked and yielded just as large a behavioral facilitation as bilateral stimulation. Follow-up tests showed that the magnitude of lordosis facilitation was an orderly increasing function of microamps per pulse (beginning at 0.5 microamps) and pulses per second (studied from ten through four hundred). When the lateral vestibular nucleus stimulation was turned off, lordosis measures returned to control levels.

  Figure 2.3. Electrical stimulation of fibers descending from the medullary reticular formation (horizontal black bars) greatly potentiates deep back muscle (for lordosis behavior) responses to segmental (sensory) inputs from the pudendal nerve (diagonally striped bar). (Adapted from Cohen, Schwartz-Giblin, and Pfaff 1987a.)

  Conversely, lesions of the lateral vestibular nucleus led to marked loss of lordosis (Modianos and Pfaff 1976). The degree of behavioral decrement was an orderly (declining) function of the number of giant cell loss in the nucleus, and the effect was specific because lesions of the superior or the medial vestibular nucleus did not harm lordosis performance. Likewise, cerebellar lesions did not cause lordosis decrements.

  In turn, vestibulospinal and reticulospinal pathways (which we just covered here individually) interact with each other (Cottingham, Femano, and Pfaff 1988). They synergize. During these experiments, Cottingham had either monopolar or bipolar electrodes in the lateral vestibular nucleus as well as bipolar electrodes (twisted 75-micron tungsten wire) in the medullary reticular formation. Recording electrodes were in the deep back muscles essential for lordosis performance reflecting, obviously, the excitation of their respective motor neurons. As in previous experiments, Cottingham set up the experiments by using currents that were effective but not overwhelming. In a typical experiment, therefore, lateral vestibular nucleus stimulation would cause some motor neuron and muscle activation, but at a low level. Reticular formation stimulus, by itself, was the same. The combination of simultaneous vestibular and medullary (primarily NGC) stimulation vastly increased deep back muscle activity, for example, by more than ten times either of them by themselves. Thus, two major brainstem neuron groups, each of which receives descending inputs from central grey in the lordosis circuit, massively amplify each other’s actions to stimulate the motor neurons and thus the muscles involved in lordosis.