Activation in Neural Networks Controlling Ingestive Behaviors: What Does It Mean, and How Do We Map and Measure It? ⁎ Alan G

Physiology & Behavior 89 (2006) 501–510

Activation in neural networks controlling ingestive behaviors: What does it mean, and how do we map and measure it? ⁎ Alan G. Watts , Arshad M. Khan, Graciela Sanchez-Watts, Dawna Salter, Christina M. Neuner

Neuroscience Research Institute and Neuroscience Graduate Program, University of Southern California, Los Angeles, CA 90089-2520, United States Received 27 February 2006; received in revised form 5 May 2006; accepted 25 May 2006

Abstract

Over the past thirty years many of different methods have been developed that use markers to track or image the activity of the neurons within the central networks that control ingestive behaviors. The ultimate goal of these experiments is to identify the location of neurons that participate in the response to an identified stimulus, and more widely to define the structure and function of the networks that control specific aspects of ingestive behavior. Some of these markers depend upon the rapid accumulation of proteins, while others reflect altered energy metabolism as neurons change their firing rates. These methods are widely used in behavioral neuroscience, but the way results are interpreted within the context of defining neural networks is constrained by how we answer the following questions. How well can the structure of the behavior be documented? What do we know about the processes that lead to the accumulation of the marker? What is the function of the marker within the neuron? How closely in time does the marker accumulation track the stimulus? How long does the marker persist after the stimulus is removed? We will review how these questions can be addressed with regard to ingestive and related behaviors. We will also discuss the importance of plotting the location of labeled cells using standardized atlases to facilitate the presentation and comparison of data between experiments and laboratories. Finally, we emphasize the importance of comprehensive and accurate mapping for using newly emerging technologies in neuroinfomatics. © 2006 Elsevier Inc. All rights reserved.

Keywords: Fos; Neuroinfomatics; MAPK; Hypothalamus; Atlas; Rat; Anorexia; Brain maps

1. Introduction considering the organization of central control networks in behavior, and then consider the meaning of the term ‘activation’ A major focus for behavioral neuroscience is to understand from the viewpoint of neural physiology. We then describe the use the complete functional organization of the neural networks that and limitations of various cellular markers that are currently used initiate, maintain, and terminate specific behaviors. Clarifying to map neural activation. Finally, we will discuss strategies that neural network organization for ingestive behaviors would we can use to incorporate this type of data into standardized reveal which neurons contribute to the behavioral sequence, atlases and databases in a way that will enable investigators to when they contribute, and would provide the basis for further share and compare these complex datasets in a meaningful way, functional investigations into how collectively they can control and to construct testable models of behavioral control networks. eating and drinking. How can we clarify the organization of these networks in a way 2. Behavioral control networks and changes in neural that will help us understand their function? In this review, we will activation address this question from the perspective of mapping patterns of neural activation. We will first discuss a general model for 2.1. Behavioral control networks

The sequence of motor actions that make up a behavior ⁎ Corresponding author. Neuroscience Research Institute, Hedco Neurosci- ence Building, MC 2520, University of Southern California, Los Angeles, CA ultimately derives from the changing signaling patterns with- 90089-2520, United States. Tel.: +1 213 740 1497; fax: +1 213 741 0561. in networks that control sensory transduction, central integra- E-mail address: [email protected] (A.G. Watts). tion, and motor selection and execution (Fig. 1; [1,2]). Our

constituent neurons are the building blocks of network function. Thus, if we can identify the location, chemical phenotype, and projections of the neurons that change their firing rates–or as is often more generally stated, their ‘activity’–at critical times of the behavioral sequence then it should be possible to begin defining the nature of the networks responsible for behavioral initiation, maintenance, and termination. Before we discuss tools and strategies, we will briefly consider the meaning of the term neural activation. An accurate definition becomes important for considering how the most commonly used cellular markers are related to neural function.

2.2. Neural activation

The term ‘activation’ when applied to neurons (or indeed to any cell type), implies that some aspect of their function–but usually Fig. 1. A schematic representation of the neural systems and their interactions output–increases over a defined time period. Although ‘activation’ involved with controlling reflex (A) and motivated behaviors (B). Central neural is useful for general discussion, it is often used synonymously with connections are shown in black, and hormonal and feedback signals as dashed lines. The regions that constitute the central behavioral control networks in the increased firing rate, which is considered the neuron's primary brain are enclosed within the gray box in (B). Adapted from [49]. functional output. Closely examining the function of the most commonly used cellular markers together with their associated mechanisms shows that different physiological processes are understanding is clearest at the sensory and motor parts of the being tracked by each of these markers. Moreover, it is generally network because the nature of the sensory mechanisms in the accepted that the function of many commonly used markers is not brain and periphery is fairly easily probed using specific necessarily related directly to changes in firing rate. homeostatic and metabolic signals. For example, we know that Whether a neuron changes its activation state is determined angiotensin II acts on neurons in the subfornical organ (SFO) to by how it integrates its afferent inputs. Part of this integration stimulate drinking, and that leptin, insulin, and ghrelin acting on process involves changing the state of signal transduction neurons in the arcuate nucleus (ARH) have profound effects on pathways, which in turn can alter two fundamental processes feeding behavior [3,4]. Similarly, the neural bases of the simple (Fig. 2): neuronal excitability, by way of changes in membrane motor actions in the consummatory phase are relatively easily potential; and the biosynthesis of a variety of proteins and tracked in the hindbrain, and we know with some degree of peptides, which include transcription factors, proteins involved detail how several of the motor networks in the hindbrain that with transmitter release, receptor proteins, etc. A third cellular control chewing and swallowing are organized [5–9]. But process that is intimately involved with changes in neural ac- located between sensory transduction and motor output are the tivity is the neuron's energy metabolism (Fig. 2). Increases in control networks that add the adaptive value to behavior and firing rate require significant energy expenditure, particularly allow animals to survive successfully in their environment for maintaining Na+/K+-ATPase activity [10]. (Fig. 1). Despite their importance to the organization of behavior, the constituency and function of these critical circuits 2.3. Tracking neural activation are very poorly understood, for the most part because of their great complexity and the lack of suitable tools for investigation. How do we track changes in neural activity relative to the Fig. 1 posits that the exterosensory and interosensory signals development of a behavior? And if we can do this, what will the that initiate motivated–as opposed to reflex–behaviors require results tell us about the structure and function of the networks? sophisticated processing within central control networks. In There are three main strategies that experimentalists use to track terms of network function at the absolute simplest level, a specific changes in neural activity relative to the development of a behavior is initiated when input signal processing causes some behavior. One is direct electrophysiological recording to correlate neurons within a central control network to increase their firing changes in neural firing patterns in particular brain regions with rate while others are inhibited. These firing rate changes are key the expression of a behavior. A disadvantage with regard to parts of the integration process within the network that ultimately overall network structure is that we need to know the constituents initiates a specific motor sequence. of a given neural control network in advance and with some The way that a behavior emerges from the activity of these degree of certainty; i.e. where to place the electrodes. Electro- control networks is very poorly understood. In fact for all physiology can produce complex and sophisticated sets of data ingestive behaviors our knowledge of the constituent neurons that are extremely useful for determining the signaling properties within central networks together with their structure/function of a network [11,12]. These types of data are critical to under- relations, is rudimentary at best. However, what is clear is that standing how behaviors are controlled because they provide a changes in the nature and release rates of chemical signals (fast- way to analyze network function in real time. Furthermore, multi- acting transmitters, peptides, and neuromodulators) from electrode assemblies are now able to simultaneously gather data A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 503

Fig. 2. Three processes within neurons have components that are amenable for use as markers of neural activation: biosynthesis, excitability, and energy metabolism. Within these processes, markers vary in the mechanisms that lead to their accumulation, the time frames over which they are detectable, and their downstream functions. about neural firing rates in a way that is providing real insights dimensional approach using a combination of behavioral analysis, about how neural networks, particularly in the cortex, are put a variety of neuroanatomical and biochemical techniques, along together and operate to develop complex behaviors [13,14].But with newly emerging techniques in neuroinfomatics [19–21].Ifwe even with this level of technical sophistication our ability to can identify which neurons change their activity as a behavior determine the actual function of the group of neurons from which develops, we might be in a position to clarify the organization of recordings are taken currently remains relatively rudimentary the networks responsible for different aspects of behavioral control. within the context of behavioral control [12]. Currently our ability to do this for any behavior is very poor. Asecondstrategythatprovidesresolutiononabroaderscale than electrophysiology is to track changes in energy utilization 3. Markers within defined regions of the brain. There are a number of markers associated with energy metabolism that have been used in this Fig. 2 illustrates the cellular events that follow transmitter manner, e.g. cytochrome oxidase [15,16]. But the most well- binding to receptors on the neural membrane. It highlights the fact 14 known is to measure changes in the uptake of [ C]2-deoxy-D- that many components in the biosynthetic cascades initiated by glucose (2DG) in response to a stimulus. 2DG is a glucose analog afferent integration might serve as indices of neural activation. that is not isomerized to fructose-6-phosphate after its phosphor- Changes in the levels of these components are going to reflect ylation by hexokinase. Upon administration into the circulation changes in the state of signal transduction pathways regulated by [14C]2DG is taken up by all cells in proportion to their glucose ligand–receptor binding. However, their utility as markers utilization. Measuring the accumulation of [14C]2DG after a fixed depends upon whether these levels change in a manner that can time interval reveals brain areas that alter their energy expenditure track the stimulus in a meaningful way, and upon the availability following an experimental manipulation. [14C]2DG autoradiog- of sufficiently sensitive detection methods. raphy was pioneered by Sokoloff in the early 1970s and has been Three sets of molecules have been used as markers to determine applied to a number of neuroscientific problems [17,18]. For neural activation response patterns to a multitude of stimuli: many years it provided the only way to image neural activation phosphorylated proteins; primary and mature RNA transcripts; patterns throughout the brain in a manner that is now more and the peptides/proteins that are the end products of biosynthesis commonly tracked using c-Fos in small animals, and fMRI in (Fig. 2). The nature of these molecules means that in situ hy- humans. Of course tracking changes in metabolic activity bridization and/or immunocytochemistry are the most common provides indices of neural function that are inherently different methods for identifying neurons where the levels of a particular to those monitored by electrophysiology or immunocytochemis- marker change following a stimulus. Each of these molecules has try (Fig. 2), and these methods do not have the cellular resolution characteristics that make them more or less useful for tracking of immunocytochemical techniques. specific aspects of neural activation. For example, some protein The third strategy for tracking changing patterns of neural kinases are widely distributed in the brain and are rapidly phos- activity–and the one upon which we will focus our attention for the phorylated following a stimulus making them potentially very rest of this article–is to map the location, chemical phenotype, and useful for identifying neurons that quickly become involved connections of all the neurons whose activity changes before and in organizing behavioral responses. We will describe some of during the initiation of a behavioral episode. This requires a multi- our recent results showing the utility of one such marker–the 504 A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 phosphorylated (p) forms of p44/p42 mitogen-activated protein [33,34], we can use this information to consider the type of cellular (MAP) kinases (also known as ERKs 1/2)–in tracking rapid neural process engaged by a particular stimulus if levels of pERK1/2 responses. increase. The function of other cellular markers in neurons is much The discovery in the mid-1980s that certain proteins less clear. In this regard c-Fos acts as a transcription factor for a (immediate–early gene products [22]) quickly accumulate in large number of genes that regulate cell growth, proliferation, and neurons following an extracellular stimulus provided the break- differentiation [35]. However, the role of these types of long-term through that led to their use as activation markers [23]. The proto- adaptive processes in neurons is unknown. oncogene c-Fos [24] is by far the most commonly used cellular marker, although others are also used [25–27]. 5. Resolution Cellular markers, by their nature, have been used almost exclusively to track how neurons and circuits respond to a The utility of the data generated from experiments using single stimulus. Because endpoints are detected histologically, cellular markers depends upon at least three analytical resolution experimental design is generally linear within a set of animals: issues. First, how well can we temporally correlate the patterns of stimulus application, tissue processing, data gathering and marker expression with the identified components of a behavior? interpretation. However, we should note that it is possible to This is determined largely by the kinetics of marker accumulation track the convergence of two stimuli given at discrete intervals and degradation, and can be considered as the marker's temporal onto neural networks of the same animal. The clearest examples resolution. Second, how accurately can we describe and resolve of this approach are the use of Arc and Homer 1a gene products. the sequence of motor events that make up the behavior being These immediate early genes have rapid (Arc) or slower (Homer investigated? Third, how accurately can we represent (i.e. map) 1a) activation and decay patterns, meaning that detecting their the location of activated neurons within the brain? mRNAs can separately track two stimuli that are presented 20– 30 min apart [26−28]. Further developing this type of technology 5.1. Temporal resolution will dramatically improve our ability to understand how behavioral networks are organized. Having identified the categories If the distribution of a particular marker within the brain is to of molecules that can act as cellular markers, we should consider provide a meaningful picture of network structure during a both the cellular mechanisms that lead to marker accumulation, behavioral sequence then we need to understand its response and the functions that the markers have in neurons. kinetics (Fig. 2). How long does it take for a marker to accumulate in a neuron, and what is its half-life? 4. Mechanisms Some markers respond to a stimulus very rapidly and are only transiently detectable. Changes in membrane potential occur Two considerations are central to understanding what changes within milliseconds of an afferent stimulus, and can be equally in the concentrations of cellular markers can reveal about neural brief in duration once the stimulus is removed. Changes in the network organization. First, what are the cellular events that activity of some biosynthetic pathway components also occur culminate in the accumulation of the marker? And second, what rapidly—protein phosphorylation, for example. Other events, function does the marker itself have within the neuron? such as the accumulation of newly synthesized bioactive proteins The downstream products of the signal transduction pathways and peptides, occur much more slowly and are more persistent in engaged by a neuron's afferent inputs are critical cellular events neurons. c-Fos and the fos-related antigens (FRA) are examples of regulating the accumulation of the marker. Understanding these markers of this type. Their accumulation can take 30 min or mechanisms instructs us about the processes to which a marker longer, and in the case of FosB, can persist for many hours [36]. may or may not respond, and they constrain the way we can Each type of marker thus provides a different picture of interpret what the neuron's overall function might be in relation to network structure and function at a given time of the behavioral the stimulus. A good example is the fact that c-Fos biosynthesis sequence. For example, different expression patterns are seen if we tracks stimuli that increase intracellular calcium following the examine the c-Fos in the brain 30 min and 2 h following a stimulus occupation of a variety of receptors in a way that is related to that triggers feeding [37]. If the behavior is made up of a complex frequency of excitatory neural inputs [22,30,29].Furthermore, sequence of motor events, then the complete map obtained at a this property means that increased c-Fos expression is not time well into the sequence will be a useful composite of the neural invariably coupled to neuronal firing rates (see also [22,31,32]); a networks involved with the events expressed up to that point [38]. feature that also applies to all currently used cellular activation The pattern of c-Fos distribution seen within the brain 2 h markers. It is also important to note that these mechanistic issues after a challenge that leads to feeding will likely be different mean that the absence of a cellular marker does not exclude the again to what is revealed by a marker that accumulates within neuron's potential contribution to behavioral control. 5 min of the challenge (see below). In the latter case, because of The physiological role that a particular marker plays within a the rapidity with which the marker appears, it should be possible neuron provides the context for recognizing how changes in its to correlate the resulting expression patterns much more tightly levels relate to neural activity. This is an important consider- to a specific aspect of the behavior than with a marker such as c- ation when trying to place the neuron within the overall context Fos. This point leads us now to consider that the utility of the of network function. For example, because a great deal is maps derived from different markers is determined by how well known about the signaling properties of p44/42 MAP kinases we can resolve the structure of the behavior. A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 505

5.2. Behavioral resolution structure of these complex behaviors means that it is more difficult to correlate expression patterns of cellular markers with a The structure of the behavior chosen for study will determine particular aspect of the behavior. However, the analysis of ensuing how well data can be defined within the context of network c-Fos expression patterns is beginning to provide real insight into organization. All behaviors consist of a temporally arranged the different networks that contribute to these complex behaviors series of sensory-motor events [2]. For some behaviors this [38]. sequence will be relatively fixed and easily documented, while A further analytical complexity is that some experimental for others it will vary considerably each time they are expressed, manipulations that stimulate ingestive behaviors are accompanied particularly in the appetitive phase. Those behaviors that are by a wide range of autonomic and/or neuroendocrine activities. more easily dissected will be those that have simple, well For example, consider the feeding that results from an injection of defined, and easily quantified stimulus-response characteristics. norepinephrine (NE) into the region of the paraventricular nucleus An example of an ostensibly simple ingestive behavior is the set of the hypothalamus (PVH). Feeding begins within a few minutes of motor actions utilized when a rat approaches a water source of injection [43] meaning that changes in the relevant feeding to drink. Despite its overt simplicity a quite complex sequence control networks are initiated quickly. However, NE applied to the of motor actions is revealed when this behavior is dissected into PVH has other significant physiological effects that are not its component parts (Table 1 [39–41]). Another example of a directly related to the overt motor actions of feeding. First, NE simple ingestive behavioral sequence is the compensatory injections increase the release of ACTH secretogogues from CRH feeding that occurs when dehydration–anorexia is reversed neuroendocrine neuronal terminals in the median eminence, and [42]. Within any of these behavioral sequences, each compo- ultimately glucocorticoid from the adrenal cortex [44];second, nent set of sensory-motor events will involve its own pattern of NE injections have significant cardiovascular effects [45].Eachof neural activation. Other ingestive behaviors are more varied and these motor events will have their own sensory-motor outcomes complex in their initiation and expression; the predatory behavior and associated patterns of neural activation. To probe the circuits expressed as an animal searches of food, for example [38].The involved with NE-associated feeding, a typical experiment might

Table 1 A summary of some of the motor actions, motor systems, associated sensory inputs, and time frames engaged as a rat approaches a drinking spout to drink fluid (see [39–41] for more details) Behavior phase Motor action Motor system Sensory input Time frame Initiation Potential Initiators Increased plasma osmolality – Osmoreceptors Decreased plasma volume – Angiotensin II/baroreceptors Circadian Suprachiasmatic nucleus Incentive (cognitive) Visual/olfactory/gustatory Dry mouth (eg. from eating) ANS Trigeminal Inhibition of salivation ANS Trigeminal

Appetitive Foraging/exploration/purposive locomotion Whole body Visual/olfactory etc. min/s Body and head orientation to fluid source Whole body/head/neck Visual/olfactory etc. s

Consummatory Approach Body orientation to source Whole body Visual/olfactory etc. s Head/snout orientation to source Head/neck Visual/olfactory etc. s/ms Preparation Contact of perioral vibrissae with fluid source Head/neck Trigeminal s/ms Forward movement of snout Head/neck/trigeminal Trigeminal ms Contact of upper lip with fluid source Head/neck/trigeminal Trigeminal ms Contact of lower lip with fluid source Head/neck/trigeminal Trigeminal ms Identification & confirmation Opening of mouth Trigeminal Trigeminal ms Extension of tongue Hypoglossal Trigeminal/gustatory ms Initial exploratory licking Hypoglossal Trigeminal/gustatory s/ms (tongue extension–retraction) Performance Licking Hypoglossal Trigeminal/gustatory s/ms Swallowing

Termination Retraction of tongue Hypoglossal Trigeminal ms Closure of mouth Trigeminal Trigeminal ms Withdrawal from fluid source Head/whole body Visual/olfactory etc. s 506 A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 use immunocytochemistry to reveal c-Fos expression patterns 90–120 min after the NE injection. As we analyze these data, the question will arise as to which c-Fos labeled neurons are directly related to feeding and which ones derive from the initiation and consequences of the neuroendocrine and autonomic events. Clearly, understanding the complete motor sequence–behavior, autonomic and neuroendocrine–for the selected behavior will significantly improve the resolution of the analysis. Ultimately, how accurately a particular behavior is described will determine how well the resulting patterns of neural activation can be correlated with the behavioral sequence. A clear and detailed understanding of the behavioral sequence is essential for interpreting the ensuing expression patterns in the brain; a well- defined behavior is going to be easier to correlate with neural activation patterns than one that is poorly defined or documented.

5.3. Examples

To illustrate the utility of these various markers, we now discuss two examples where we have used cellular markers to explore the structure of central control networks.

5.3.1. Neural activation following an intravenous injection of 2-deoxyglucose Systemic injections of 2DG in rats generate a triad of motor responses aimed at normalizing energy metabolism [46]: increased sympathoadrenal output that elevates adrenaline secretion from the adrenal medulla; increased release of CRH from neurons in the hypothalamic paraventricular nucleus (PVH) that stimulates ACTH and then glucocorticoid secretion from the adrenal cortex; and feeding. We have recently used immunocytochemistry to detect the Fig. 3. Photomicrographs of phospho(p)ERK1/2 immunoreactivity in the oval phosphorylated form of the ERK1/2 MAP kinases (pERK) as nucleus of the bed nuclei of the stria terminalis (A, A′), lateral part of the central part of a study aimed at mapping the central responses to reduced nucleus of the amygdala (B, B′), and paraventricular nucleus of the hypothalamus ′ glucose availability. As discussed earlier, phosphorylation of (C, C ) from animals killed 5 min after being injected intravenously with either saline (A, B, C) or 200 mg/kg of 2-deoxy-D-glucose (2-DG; A′,B′,C′). these widely distributed kinases rapidly follows a range of ligand–receptor interactions. We have shown previously that both iv 2DG and halothane anesthesia elevate pERK1/2 levels in nocturnal food intake in the hour following the return of water. PVH CRH neurons within 10 min [47]. However, we have now The behavioral sequence is activated by a simple stimulus, is gone on to show that within 5 min of an iv 2DG pERK1/2 is seen highly predictable, and prominently features eating during its in the PVH, the lateral part of the central nucleus of the amygdala initial phases, making it an excellent candidate for investigating (CEAl), the oval nucleus (ov) in the dorsolateral part of the bed the organization of the central circuits that initiate feeding. nuclei of the stria terminalis (BST), and part of the insular cortex To this end, we have used c-Fos and pERK1/2 immunocyto- (IC). These results (Fig. 3 [48]) emphasize the utility of pERK1/ chemistry to determine the location and phenotype of neurons that 2 for identifying neurons that are rapidly activated by a stimulus. are activated after water is returned to dehydrated (DE)–anorexic rats. Our first study [50] used c-Fos immunocytochemistry to 5.3.2. Neural activation during the reversal of clarify the role of a population of CRH neurons in the LHA [51] dehydration–anorexia during both the development and reversal of DE–anorexia. Be- Rats become anorexic during the consumption of hypertonic cause CRH neurons are found in the same LHA regions as orexin saline as part of a behavioral reflex aimed at protecting their and MCH neurons [52], we also examined the behavior of these fluid balance [1,49]. Anorexia becomes particularly pronounced neurons as DE–anorexia develops and is then reversed. We found after 4–5 days, when nocturnal food intake drops to about 25– that 5 days of DE increase c-Fos-immunoreactivity (ir) in large 35% of baseline levels. Returning drinking water stimulates a numbers of neurons in the LHA, some of which also show stereotypic behavioral sequence that begins with drinking, increased CRH, but not orexin or MCH gene expression. We also followed after about 8 or 9 min by a robust bout of eating that showed that the behavioral sequence exhibited by DE animals in lasts about 20–30 min [42]. Animals that have been drinking the minutes following water drinking is accompanied by a further hypertonic saline for 5 days will eat about 25% of their normal increase in the number of c-Fos-ir nuclei, which was independent of A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 507

Fig. 4. Photomicrographs of phospho(p)ERK1/2 immunoreactivity in the insular cortex (A, B), paraventricular nucleus of the hypothalamus (C), and lateral hypothalamic area (D, E, F) of animals given 2.5% saline to drink for 5 days, and then either given water to drink or maintained on saline and killed 10 min later. Note theincreasein pERK1/2 immunoreactivity in the insular cortex and lateral hypothalamic area but not the paraventricular nucleus 10 min after the return of water. whether animals ate or were denied access to food. c-Fos-ir signifi- each positively labeled neuron on a drawing or a map taken from a cantly increased in orexin neurons but not CRH or MCH neurons. standardized brain atlas, and use photomicrographs as adjuncts Together these data implicate CRH but not orexin or MCH neurons for data representation (e.g. [50]). Ideally, maps should accurately in the motor events accompanying the development of DE– represent the area of interest in terms of nuclear boundaries and anorexia; and orexin but not CRH or MCH neurons in controlling the position of major fiber tracts etc. The other point to consider is behavioral sequence that is stimulated by drinking water. In a broader study (A.G. Watts, C. Neuner and D. Salter, unpublished observations) we have recently used pERK1/2 immunocytochemistry to identify neurons that are activated during the first 10 min after water is returned. We posit that rather than waiting 45 min or longer after water is returned (as is required for c-Fos expression), a tighter correlation between behavioral onset and the identity of central circuits can be achieved if we identify neurons at the same time as the animal is in the process of initiating feeding. With this technique we find that there is increased pERK1/2-ir in the insular cortex and lateral hypothalamic area, but not the PVH or arcuate nucleus 10 min after the return of water (Fig. 4). These data show that pERK immunocytochemistry can be used to track changes of neural activation with a temporal resolution that is much closer to the time of stimulus onset than that achieved using c-Fos as a cellular marker.

5.4. Mapping activation patterns and the spatial resolution of data analysis

The third important aspect of resolution concerns how data from neuroanatomical experiments are represented, analyzed, and compared across experiments and laboratories. Considered at the simplest level, after sections containing the positively labeled neurons are processed then it is a relatively straightforward task to Fig. 5. Plate 18 of the Swanson brain atlas [54] with the pERK immunoreactivity photograph and publish them. Although this technique provides from Fig. 3 placed in approximately the correct alignment (left plate). The border on the right delineates the equivalent BST regions. Abbreviations: al, an exact representation of the data, it often offers little utility to anterolateral area of the BST; am, anteromedial area of the BST; BST, bed nuclei other investigators for comparison with their datasets. A more of the stria terminalis; fu, fusiform nucleus of the BST; ju, juxtacapsular nucleus sophisticated and useful representation is to plot the position of of the BST; ov, oval nucleus of the BST; PS, parastrial nucleus. 508 A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 that the entire brain should be mapped with a particular marker, phenotype. This is particularly important when deriving data from not just favored regions. Only if we know the location of all the complex heterogeneously organized brain regions. As an cells that change their activity will we have a complete picture of example, consider the neurons showing pERK1/2 labeling in the regions engaged in the processing of information required to the dorsolateral part of the BST 5 min after a 2DG injection control an ingestive behavior. (Fig. 5). These neurons are found in the oval (ov) nucleus of the But however the data are represented, how can they be most BST, and because of the complexity of this part of the BST it is easily interpreted and used by others in the field? Neuroana- critical to be able to locate these neurons with a precise descriptor, tomical experiments are no different to those that generate rather than a more general term such as the ‘dorsolateral BST’. parametric data in that maps showing the locations of labeled Fig. 6 shows the projection array of all the subnuclei in the neurons must also be replicated and independently verified to dorsolateral BST (see also [55]). With this in mind, if we compare become usefully established in the literature. To do this, data the projection patterns of the BSTam, BSTal, and BSTju with the from independent experiments must be represented in a manner BSTov, we see that there is less likelihood of pERK1/2-labeled that other investigators can accurately compare. Simply put, two neurons in the BSTov projecting to the PVH, than if we had seen independent sets of data showing the distribution of labeled labeling slightly more medially in the BSTam, or more ventrally cells can only be compared if they share common reference in the BSTal (Figs. 5 and 6). Fig. 6 shows other examples of points and fiducials. Currently the best way to do this is to plot differential projection patterns of these BST nuclei to the lateral data on maps from a standardized brain atlas, the most widely hypothalamic area, parabrachial, and dorsomedial nuclei. These used of which are Paxinos and Watson [53] and Swanson [54]. differences in projections are crucial if we wish to construct the A second advantage of accurately plotting the positions of detailed neural network that is involved with generating responses labeled neurons on standardized atlas maps is that the investigator to 2DG. However, determining the final structure of these control can access published information about projections and chemical networks will require data derived from experiments that combine

Fig. 6. A comparison of the projections from the anteromedial group (amg) anterior lateral group (alg) of the bed nucleus of the stria terminalis (BST) derived from anterograde and retrograde tracing studies. The horizontal bars highlight the projections of the juxtacapsular (ju), the oval nucleus (ov), anterolateral area (al), and anteromedial area (am) of the BST. The size of the dots represents the density of the projection. (See text for more details, and [55] for references and a full description of the connections.) Abbreviations: ACBc, nucleus accumbens, core; ACBp, nucleus accumbens, posterior part; ACBsd, nucleus accumbens, dorsal shell; ACBsv, nucleus accumbens, ventral shell; ADP, anterodorsal preoptic nucleus; AHN, anterior hypothalamic nucleus; AHNc, anterior hypothalamic nucleus, central part; AHNp, anterior hypothalamic nucleus, posterior part; AMBd, nucleus ambiguus, dorsal division; AVP, anteroventral preoptic nucleus; AVPV, anteroventral periventricular nucleus hypothalamus; AVPV, anteroventral periventricular nucleus hypothalamus; B, Barrington's nucleus; CEAc, central amygdalar nucleus, capsular part; CEAl, central amygdalar nucleus, lateral part; CEAm, central amygdalar nucleus, medial part; CP, caudoputamen; DMHa, dorsomedial hypothalamic nucleus, anterior part; DMHp, dorsomedial hypothalamic nucleus, posterior part; DMHv, dorsomedial hypothalamic nucleus, ventral part; DMX, dorsal motor nucleus vagus nerve; EW, Edinger-Westphal nucleus; I, internuclear area, hypothalamic periventricular region; ISN, inferior salivatory nucleus; LCN, lateral cervical nucleus; LHAjd, lateral hypothalamic area, juxtadorsomedial region; LHAjvd, lateral hypothalamic area, juxtaventromedial region, dorsal zone; LHAjvv, lateral hypothalamic area, juxtaventromedial region, ventral zone; LHAs, lateral hypothalamic area, suprafornical region; LHAsfa, lateral hypothalamic area, subfornical region, rostral zone; LHAsfp, lateral hypothalamic area, subfornical region, posterior zone; LSc, lateral septal nucleus, caudal (caudodorsal) part; LSr, lateral septal nucleus, rostral (rostroventral) part; LSv, lateral septal nucleus, ventral part; MEPO, median preoptic nucleus; MEV, midbrain trigeminal nucleus; MM, medial mammillary nucleus, body; MPNc, medial preoptic nucleus, central part; MPNl, medial preoptic nucleus, lateral part; MPNm, medial preoptic nucleus, medial part; MPO, medial preoptic area; MRNm, midbrain reticular nucleus, magnocellular part; NTSco, nucleus of the solitary tract, commissural part; NTSl, nucleus of the solitary tract, lateral part; NTSm, nucleus of the solitary tract, medial part; PAGvl, periaqueductal gray, ventrolateral division; PARN, parvicellular reticular nucleus; PBl, parabrachial nucleus, lateral division; PBm, parabrachial nucleus, medial division; PD, posterodorsal preoptic nucleus; PGRNl, paragigantocellular reticular nucleus, lateral part; PMd, dorsal premammillary nucleus; PMv, ventral premammillary nucleus; PS, parastrial nucleus; PSCH, suprachiasmatic preoptic nucleus; PSTN, parasubthalamic nucleus; PVHap, paraventricular hypothalamic nucleus, anterior parvicellular part; PVHd, paraventricular hypothalamic nucleus, descending division; PVHdp, paraventricular hypothalamic nucleus, dorsal parvicellular part; PVHlp, paraventricular hypothalamic nucleus, lateral parvicellular part; PVHmpv, paraventricular hypothalamic nucleus, medial parvicellular part, ventral zone; PVi, periventricular hypothalamic nucleus, intermediate part; PVp, periventricular hypothalamic nucleus, posterior part; PVpo, preoptic periventricular nucleus; RR, midbrain reticular nucleus, retrorubral area; SFO, subfornical organ [Pines]; SNr, substantia nigra, reticular part; SSN, superior salivatory nucleus; SUT, supratrigeminal nucleus; V, motor nucleus of the trigeminal nerve; VMHdm, ventromedial nucleus hypothalamus, dorsomedial part; VMHvl, ventromedial nucleus hypothalamus, ventrolateral part; VTA, ventral tegmental area. A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510 509 tract-tracing methodologies with the identification of markers for [5] Fay RA, Norgren R. Identification of rat brainstem multisynaptic neuronal activation (eg. [56]). connections to the oral motor nuclei using pseudorabies virus. III. Lingual muscle motor systems. Brain Res Rev 1997;25:291–311. Finally, accurate data representation using standardized [6] Fay RA, Norgren R. Identification of rat brainstem multisynaptic atlases opens up the possibility of using newly emerging connections to the oral motor nuclei in the rat using pseudorabies virus. neuroinfomatics tools to generate and compare distribution II. Facial muscle motor systems. Brain Res Rev 1997;25:276–90. patterns of activated neurons during behavior [19–21,57,58]. [7] Fay RA, Norgren R. Identification of rat brainstem multisynaptic Given the fact that there are an estimated 500–1000 identified connections to the oral motor nuclei using pseudorabies virus. I. Masticatory muscle motor systems. Brain Res Rev 1997;25:255–75. brain regions and 25,000 connections in the brain [19], it goes [8] Travers JB, Dinardo LA, Karimnamazi H. Motor and premotor mechan- without saying that analyzing and correlating data from cellular isms of licking. Neurosci Biobehav Rev 1997;21:631–47. markers with components in a behavioral sequence are going to [9] Travers JB, Travers SP, Norgren R. Gustatory neural processing in the require informatics tools that rank in complexity and utility hindbrain. Annu Rev Neurosci 1987;10:595–632. along side those used for probing and mapping mammalian [10] Sokoloff L. Energetics of functional activation in neural tissues. Neurochem Res 1999;24:321–9. genomes. Two neuroinfomatics systems that our lab is currently [11] Rizzolatti G, Luppino G. The cortical motor system. Neuron 2001;31:889–901. using are Neuroscholar (http://chasseur.usc.edu/website/index. [12] Rolls ET. Taste, olfactory, and food texture processing in the brain, and the html) and the Brain Architecture Management System (http:// control of food intake. Physiol Behav 2005;85:45–56. brancusi.usc.edu/bkms/). [13] Mulder AB, Nordquist RE, Orgut O, Pennartz CM. Learning-related changes in response patterns of prefrontal neurons during instrumental conditioning. Behav Brain Res 2003;146:77–88. 6. Conclusion [14] Buzsaki G. Theta rhythm of navigation: link between path integration and landmark navigation, episodic and semantic memory. Hippocampus Brain imaging techniques have come a long way in the 50 years 2005;15:827–40. since the idea that monitoring local blood flow and glucose [15] Hevner RF, Liu S, Wong-Riley MT. A metabolic map of cytochrome utilization could give us a dynamic picture of brain activation. oxidase in the rat brain: histochemical, densitometric and biochemical studies. Neuroscience 1995;65:313–42. Today, to define the organization of the neural networks that [16] O'Keefe LP, Levitt JB, Kiper DC, Shapley RM, Movshon JA. Functional control ingestive behavior we are steadily improving the resolution organization of owl monkey lateral geniculate nucleus and visual cortex. of our methods in regard to cellular markers and behavioral J Neurophysiol 1998;80:594–609. analysis. Newly emerging techniques are providing fascinating [17] Kennedy C, Des Rosiers MH, Jehle JW, Reivich M, Sharpe F, Sokoloff L. pointers towards what can be achieved with cellular markers that Mapping of functional neural pathways by autoradiographic survey of local metabolic rate with (14C)deoxyglucose. Science 1975;187:850–3. respond in the same time domain as the behavior. But if we want to [18] Sokoloff L, Reivich M, Kennedy C, Des Rosiers MH, Patlak CS, Pettigrew delineate the structure and function of entire neural networks KD, et al. The [14C]deoxyglucose method for the measurement of local perhaps the biggest challenge for behavioral neuroscientists and cerebral glucose utilization: theory, procedure, and normal values in the neuroanatomists is to analyze, compare, and share the large quan- conscious and anesthetized albino rat. J Neurochem 1977;28:897–916. tities of data that are generated by these methods. The complexity [19] Bota M, Dong HW, Swanson LW. From gene networks to brain networks. Nat Neurosci 2003;6:795–9. of the approaches needed to achieve this goal is of the same [20] Burns, G.A., Khan, A.M., Cheng, W.-C., Watts, A.G. A computational magnitude as problems in genomics, proteomics, metabolomics knowledge representation for physiological experiments with direct links etc., and will require the same sophisticated computational approa- to the primary literature and raw data. Program No. 571.9. 2005, Abstract ches to data analysis and presentation. Viewer/Itinerary Planner. Washington, DC: Society for Neuroscience, 2005. Online. [21] Khan AM, Hahn J, Cheng W-C, Watts AG. Burns GAPC . NeuroScholar's Acknowledgements electronic laboratory notebook and its application to neuroendocrinology. Neuroinformatics 2006;4:139–62. The work from the authors laboratory was supported by PHS [22] Curran T, Morgan JI. Fos: an immediate–early transcription factor in grants NS29728, MH66168, MH67392, and MH71108. We neurons. J Neurobiol 1995;26:403–12. would like to thank Larry Swanson and Hong-Wei Dong for [23] Morgan JI, Cohen DR, Hempstead JL, Curran T. Mapping patterns of c-fos expression in the central nervous system after seizure. Science permission to use part of their original figure and the unpublished 1987;237:192–7. data shown in Fig. 6. [24] Van Beveren C, van Straaten F, Curran T, Muller R, Verma IM. Analysis of FBJ-MuSV provirus and c-fos (mouse) gene reveals that viral and cellular fos gene products have different carboxy termini. Cell 1983;32:1241–55. References [25] Chaudhuri A, Nissanov J, Larocque S, Rioux L. Dual activity maps in primate visual cortex produced by different temporal patterns of zif268 mRNA and [1] Watts AG. Neuropeptides and the integration of motor responses to protein expression. Proc Natl Acad Sci U S A 1997;94:2671–5. dehydration. Annu Rev Neurosci 2001;24:357–84. [26] Petrovich GD, Holland PC, Gallagher M. Amygdalar and prefrontal [2] Watts AG, Swanson LW. Anatomy of motivational systems. In: Pashler pathways to the lateral hypothalamus are activated by a learned cue that Hal, Gallistell Randy, editors. ‘Stevens’ handbook of experimental stimulates eating. J Neurosci 2005;25:8295–302. psychology, volume 3, learning, motivation, and emotion. Third edition. [27] Vazdarjanova A, McNaughton BL, Barnes CA, Worley PF, Guzowski JF. John Wiley & Sons; 2002. p. 563–632. Experience-dependent coincident expression of the effector immediate– [3] Cowley MA. Hypothalamic melanocortin neurons integrate signals of early genes arc and Homer 1a in hippocampal and neocortical neuronal energy state. Eur J Pharmacol 2003;480:3–11. networks. J Neurosci 2002;22:10067–71. [4] Niswender KD, Baskin DG, Schwartz MW. Insulin and its evolving [28] Rosen JB, Adamec RE, Thompson BL. Expression of egr-1 (zif268) partnership with leptin in the hypothalamic control of energy homeostasis. mRNA in select fear-related brain regions following exposure to a Trends Endocrinol Metab 2004;15:362–9. predator. Behav Brain Res 2005;162:279–88. 510 A.G. Watts et al. / Physiology & Behavior 89 (2006) 501–510

[29] Thomas KL, Everitt BJ. Limbic–cortical–ventral striatal activation during [46] Salter D, Watts AG. Differential suppression of hyperglycemic, feeding, retrieval of a discrete cocaine-associated stimulus: a cellular imaging study and neuroendocrine responses in anorexia. Am J Physiol Regul Integr with gamma protein kinase C expression. J Neurosci 2001;21:2526–35. Comp Physiol 2003;284:R174–82. [30] Luckman SM, Dyball RE, Leng G. Induction of c-fos expression in [47] Khan AM, Watts AG. Intravenous 2-deoxy-D-glucose injection rapidly hypothalamic magnocellular neurons requires synaptic activation and not elevates levels of the phosphorylated forms of p44/42 mitogen-activated simply increased spike activity. J Neurosci 1994;14:4825–30. protein kinases (extracellularly regulated kinases 1/2) in rat hypothalamic [31] Fields RD, Eshete F, Stevens B, Itoh K. Action potential-dependent parvicellular paraventricular neurons. Endocrinology 2004;145:351–9. regulation of gene expression: temporal specificity in Ca2+, cAMP- [48] Khan AM, Watts AG. Glucoprivic challenge is associated with elevated responsive element binding proteins, and mitogen-activated protein kinase levels of phosphorylated signaling intermediates in discrete forebrain signaling. J Neurosci 1997;17:7252–66. regions: immunofluorescence studies of ERK1/2 and CREB [Program [32] Hoffman GE, Lyo D. Anatomical markers of activity in neuroendocrine #634.15] 2005, Abstract Viewer/Itinerary Planner. Washington, DC: systems: are we all ‘fos-ed out’? J Neuroendocrinol 2002;14:259–68. Society for Neuroscience. Online. [33] Grewal SS, York RD, Stork PJ. Extracellular-signal-regulated kinase [49] Watts AG, Salter DS. Neural mechanisms of anorexia. In: Stricker Edward signalling in neurons. Curr Opin Neurobiol 1999;9:544–53. M, Woods Steven C, editors. The handbook of behavioral neurobiology. [34] Sweatt JD. The neuronal MAP kinase cascade: a biochemical signal inte- Food and fluid intake. 2nd Ed. Plenum Press; 2004. p. 383–420. gration system subserving synaptic plasticity and memory. J Neurochem [50] Watts AG, Sanchez-Watts G. Rapid and preferential activation of c-fos 2001;76:1–10. protein in hypocretin/orexin neurons following the reversal of dehydra- [35] Ordway JM, Fenster SD, Ruan H, Curran T. A transcriptome map of tion–anorexia. J Comp Neurol in press. cellular transformation by the fos oncogene. Mol Cancer 2005;4:19. [51] Watts AG, Sanchez-Watts G, Kelly AB. Distinct patterns of neuropeptide [36] McClung CA, Ulery PG, Perrotti LI, Zachariou V, Berton O, Nestler EJ. gene expression in the lateral hypothalamic area and arcuate nucleus are DeltaFosB: a molecular switch for long-term adaptation in the brain. Brain associated with dehydration induced anorexia. J Neurosci 1999;19:6111–21. Res Mol Brain Res 2004;132:146–54. [52] Swanson LW, Sanchez-Watts G, Watts AG. Comparison of melanin- [37] Valdes JL, Farias P, Ocampo-Garces A, Cortes N, Seron-Ferre M, concentrating hormone and hypocretin/orexin mRNA expression patterns Torrealba F. Arousal and differential Fos expression in histaminergic in a new parcelating scheme of the lateral hypothalamic zone. Neurosci neurons of the ascending arousal system during a feeding-related Lett 2005;387:80–4. motivated behaviour. Eur J Neurosci 2005;21:1931–42. [53] Paxinos G, Watson C. The rat brain in stereotaxic coordinates — the new [38] Comoli E, Ribeiro-Barbosa ER, Negrao N, Goto M, Canteras NS. coronal set. Fifth edition; 2005. Functional mapping of the prosencephalic systems involved in organizing [54] Swanson LW. Brain maps: structure of the rat brain. A laboratory guide predatory behavior in rats. Neuroscience 2005;130:1055–67. with printed and electronic templates for data, models and schematics. 3rd [39] Jacquin MF, Zeigler HP. Trigeminal orosensation and ingestive behavior in ed. Amsterdam: Elsevier; 2004. the rat. Behav Neurosci 1983;97:62–97. [55] Dong HW, Swanson LW. Projections from bed nuclei of the stria [40] Zeigler HP, Jacquin MF, Miller MG. Trigeminal sensorimotor mechanisms terminalis, anteromedial area: cerebral hemisphere integration of neuro- and ingestive behavior. Neurosci Biobehav Rev 1984;8:415–23. endocrine, autonomic, and behavioral aspects of energy balance. J Comp [41] Zeigler HP, Semba K, Jacquin MF. Trigeminal reflexes and ingestive Neurol 2006;494:142–78. behavior in the rat. Behav Neurosci 1984;98:1023–38. [56] Kelly AB, Watts AG. The mediation of dehydration-induced peptidergic [42] Watts AG. Dehydration-associated anorexia: development and rapid gene expression in the rat lateral hypothalamic area by forebrain afferent reversal. Physiol Behav 1999;65:871–8. projections. J Comp Neurol 1996;370:231–46. [43] Leibowitz SF. Pattern of drinking and feeding produced by hypothalamic [57] Dashti AE, Ghandeharizadeh S, Stone J, Swanson LW, Thompson RH. norepinephrine injection in the satiated rat. Physiol Behav 1975;14:731–42. Database challenges and solutions in neuroscientific applications. Neuro- [44] Itoi K, Suda T, Tozawa F, Dobashi I, Ohmori N, Sakai Y, et al. image 1997;5:97–115. Microinjection of norepinephrine into the paraventricular nucleus of the [58] Bota M, Dong HW, Swanson LW. Brain architecture management system. hypothalamus stimulates corticotropin-releasing factor gene expression in Neuroinformatics 2005;3:15–48. conscious rats. Endocrinology 1994;135:2177–82. [45] Harland D, Gardiner SM, Bennett T. Paraventricular nucleus injections of noradrenaline: cardiovascular effects in conscious Long-Evans and Brattleboro rats. Brain Res 1989;496:14–24.