Defining the brain control of physiological stability

The last few decades have seen major advances in neurobiology and uncovered novel genetic and cellular substrates involved in the control of physiological set points. In this Review, I discuss the limitations in the definition of homeostatic set points established by Walter B Canon and highlight evidence that two other physiological systems, namely rheostasis and allostasis provide distinct inputs to independently modify set-point levels. Using data collected over the past decade, the hypothalamic and genetic basis of regulated changes in set-point values by rheostatic mechanisms are described. Then, the role of higher-order brain regions, such as hippocampal circuits, for experience-dependent, allostatic induced changes in set-points are outlined. I propose that these systems provide a hierarchical organization of physiological stability that exists to maintain set-point values. The hierarchical organization of physiology has direct implications for basic and medical research, and clinical practice.


History of physiological stability
Homeostasis was described by Walter Cannon as a condition which may vary, but which is relatively constant (Cannon, 1963).He built the concept of homeostasis on the ideas of an internal milieu described by Claude Bernard (1878) and Eduard Pfluger (1877) which proposed that innate adjustments are used to maintain a steady state, and Fredericq (1885) proposition that any disturbance to the internal milieu induced a compensatory change to neutralise or repair the disturbance.Canon highlighted that homeostasis was not an equilibrium, but instead consisted of several coordinated changes across multiple systems involving central and peripheral nervous systems working cooperatively to maintain a normative state.The description of the components (e.g., hypothalamus and peripheral tissues) of homeostatic systems were heavily influenced by engineering and mathematical models used to describe dynamical systems, such as control theory (Maxwell, 1868).Therein, the self-regulation of physiological stability could be achieved using a series of connected components that required a mechanism focussed on feedback to readjust any physiological system to a deviation from a defended set-point (Rosenblueth et al., 1943).

Multi-level control of physiological stability
In a simple form, the components of homeostasis include a sensor, error detector, controller, and effector (Fig. 1A; Carpenter, 2004).For any physiological variable there is a set-point which is maintained to optimize an animal's fitness.Sensors monitor the levels of the physiological variable and transmit its ongoing value to an error detector.The cells which provide error detection monitor any deviation from the setpoint and initiate compensatory change(s) to maintain internal stability through a controller.The controller acts on effector components, which are usually peripheral tissues, to modify the level of the physiological variable by increasing or decreasing the signal output of the variable being monitored by a sensor.Core body temperature is a common example used to describe a homeostatic system (Fig. 1B) (reviewed by Tan and Knight, 2018).
Although Cannon stated that a level of variability was inherent in physiological systems (i.e., fluid matrix), he proposed that variations around set-points were kept within relatively narrow limits.Homeostasis, as originally described, did not consider that physiological stability may show regulated longer-term changes, beyond the natural variation typically observed around set-point values, for example daily or seasonal rhythms.However, such short-term changes apply to the regulation of core body temperature, where it is possible to see quick responses to cold adaptation and the activation of acute homeostatic 'emergency measures' to prevent hypothermia.But the genetic and cellular homeostatic mechanisms are distinct from the slow, gradual changes that are seen in core body temperature in animals that hibernate (Carey et al., 2003).A challenge to homeostasis about which Cannon stated, 'Unfortunately we know very little about the stimuli which cause these slow adjustments of the organism that protect against an unfavorable or harmful external environment ' (1963).

Limitations of homeostasis: short-versus long-term changes in set-points
Several physiological concepts have been proposed to account for long-term changes in physiological stability such as predictive homeostasis (Moore-Ede, 1986), and reactive scope model (Romero et al., 2009).These additional concepts were generally influenced by physiological responses to ecological conditions and based on the fundamental premise that homeostasis acts as a single regulatory system and functions to both 1) maintain stability around a set-point value and 2) control both the predictive and reactive changes in the set-point value.Basing physiological stability on a single control system has resulted in the common parlance that any adaptive change in the internal milieu is 'homeostasis'.However, there is a growing body of genetic, cellular, neural, and circuit-level evidence that indicates short-term deviations from physiological set-points (i.e., homeostasis) can be dissociated from long-term (e.g., daily) physiological changes; indicating that more than one mechanism maintains physiological stability.
Nicholas Mrosovsky (1990) proposed a second mechanism to achieve long-term physiological stability and coined the term rheostasis.Rheostasis is derived from the modern Latin rheo meaning 'stream' or 'flow' and used to refer to physiological states in which, at any phase of a rhythm, homeostatic set-points are maintained, but over a span of time (i.e.oscillation or single bout) there is a change in the regulated level.Across a range of animal models and physiological processes, Mrosovsky demonstrated these regulated changes in set-point values could be achieved, alongside the short-term homeostatic stability in set-point values.Rheostatic regulation of physiological set point are driven by intrinsic timing mechanisms that are independent of environmental input (e.g., light), such as the endogenous circadian and circannual clocks.Unfortunately, Mrosovsky's ideas were limited to measurements of homeostatic effector and sensor components and lacked the neuroanatomical precision and knowledge of genetic modifications needed to dissociate mechanisms involved in homeostasis and rheostasis (Stevenson, 2023).Over the past 30 years, these limitations have been overcome yielding novel insight into how homeostasis and rheostasis can be characterised.
While in Cannons definition, homeostatic set-points were considered innate and could not be modified by experience, there are several examples (e.g., stress response) in which individual responses and the underlying physiological set-point value vary across individuals, even when the external stimuli presented is the same.The ability of physiological set-points to change in anticipation of a new environmental condition (e.g., territorial intrusion in birds, or a stressor in humans) is referred to as allostasis (McEwen and Wingfield, 2003).Allostasis is another physiological regulatory system that accounts for achieving physiological stability through change (Sterling and Eyer, 1988).The process of allostasis is thought to be predominantly controlled by higher order brain structures such as the limbic system (i.e., hippocampus), Fig. 1.Schematic representation of a homeostatic system.The components of homeostatic systems include error detectors, controllers, effectors and sensors that act to maintain the set-point for a physiological variable (A).The circuit for the homeostatic maintenance of core body temperature includes hypothalamic, endocrine, and vascular systems (b).Temperature sensitive receptors (TMSR) act as a sensor and monitors core body temperature.Cold-and warm-sensitive neurons in the preoptic area (POA) function as error detectors and signal any deviation in core body temperature beyond the set-point to the controller.The dorsomedial hypothalamus (DMH) acts as a controller to induce effectors to increase or decrease activity to readjust core body temperature.Key effectors that adjust core body temperature consist of skeletal muscle (via shivering), vascular (i.e., vasodilation/vasoconstriction) and skin (via water evaporation).
pallial structures in non-mammals and the cortex in mammals (i.e., prefrontal cortex) (McEwen, 2000).Allostatic stability functions to help an organism adapt to unpredictable perturbations in the environment, such as harsh environmental conditions (Wingfield, 2015) or responses to real or perceived stressors (McEwen, 2000).Facultative responses (Type 1) to environmental perturbations (e.g., emergency life history stages) are regulated differently and engage higher-order brain regions that are distinct from the predominantly hypothalamic controlled programmed rheostatic oscillations associated with timing transitions in life history stages.Moreover, it is recognised that continued activation of allostatic mechanisms over weeks, months or even years can result in severe pathophysiological consequences, such as allostatic load (Type 2) (McEwen and Wingfield, 2003).Changes in set-points driven by chronic stress, social status and parasite load have individual differences that are governed by complex interactions in higher order brain regions involved in personality (e.g., prefrontal cortex), the cognitive representation of the environment (e.g., association cortices/pallial structures) and sensory cortices/pallial structures (McEwen and Stellar, 1993;McEwen, 1998).Unlike the intrinsic control of homeostasis and rheostasis, allostasis permits a level of individual experience into adaptive or maladaptive changes in physiological set-points.Hamsters transferred to short photoperiod (SD; dark circles) lose approximately 30 % of body mass after 8-12 weeks.Short term food restriction (FR) to SD hamsters (triangles) induces a homeostatic adjustment and triggers a loss in body mass.A switch back to ad libitum food (AL) access returns body mass to the seasonally programmed rheostatic set-point.Hamsters held in LD and SD show approximately 1 g reduction in body mass (F) illustrating the same homeostatic response is recruited despite the 15 g reduction in body mass driven by the programmed seasonal rheostatic response.Neuropeptides in the arcuate nucleus dissociate homeostatic and rheostatic processes (G-J).Neuropeptide y (Npy) (G) and agouti-related protein (Agrp) (H) are sensitive to short-term homeostatic changes in energy stability (i.e., fasting).Other neuropeptides are genetic correlates of programmed rheostatic control of seasonal energy stability and include proopiomelanocortin (Pomc) (I) and somatostatin (Sst) (J) and insensitive to short-term homeostatic signals.Note bar plots in G-J are schematic diagrams to represent data used to reflect 'low' (i.e., 0.5) and 'high' (i.e., 2.0) expression levels.Fever response in homeotherms is associated with a 1-2 • C increase in core body temperature for several hours to days (K).In control conditions (blue labels) humans maintain a core body temperature of approximately 37.5 • C. Participants sitting in a pool will adjust the temperature in a gloved-hand (preference temperature) which is negatively correlated with core body temperature.A simulated infection induces a reactive rheostatic fever response (red labels) associated with an elevation in core body temperature, yet the homeostatic variation (preference temperature) around the transient increase in core body temperature remains.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Dissociating rheostatic and homeostatic mechanisms
Since Mrosovsky defined rheostasis (1990), a series of neuroanatomical and genetic substrates have been identified that provide direct, independent input to regulate the diverse range of physiological homeostatic set-point values (Fig. 2A).There are two different means in which rheostatic inputs can regulate changes in physiological set-points: programmed or reactive responses (Fig. 2B-C) (Stevenson, 2023;Mrosovsky, 1990).Programmed rheostasis consist of regular, repeated oscillations in set-points values (Fig. 2B).These programmed mechanisms are generally characterised by biological rhythms and serve to provide adaptive anticipatory adjustment in physiological states that enhance an organism's fitness for a future environment.Clear examples outlined by Nicolas Mrosovsky (1990) include core body temperature and body mass cycles in hibernators, pre-migratory fattening in birds, changes in body mass during human menstrual cycles, and cyclical antler growth (see Cornelius et al., 2013;Jastroch et al., 2016;Stevenson, 2023 for recent reviews).Conversely, reactive rheostasis is a regulated change in a physiological set-point but instead of repeated cycles, consist of a single change over a relatively defined period (Fig. 2C).Changes in circulating glucose concentrations during pregnancy, and core body temperature during fever are two robust examples of reactive rheostasis as the driver of the set-point change that is not predictive and is precipitated by an event (impregnation, infection).Both programmed and reactive rheostasis are associated with distinct genetic and/or cellular populations and hypothalamic regions to induce the regulated changes in set-point which would then be maintained by homeostatic error detectors.
Some of the best characterised programmed rheostatic processes are daily and seasonally regulated changes in physiological processes.Daily rhythms in physiology are driven by an endogenous clock that approximates a 24-h oscillation and is commonly referred to as a circadian rhythm.Circadian rhythms involve two evolutionary ancient components: a transcriptional-translational feedback loop (TTFL) (Takahashi, 2017) and a distinct hypothalamic nucleus: the suprachiasmatic nucleus (SCN) (Hastings et al., 2018).In all nucleated animal cells, a TTFL is provided by a select group of genes that provides an endogenous timing mechanism that acts to impart daily time in a cell autonomous manner.Basic helix-loop-helix ARNT-like protein 1 (Bmal1) and circadian locomotor output cycles kaput (Clock) form protein dimers that induce transcription of multiple proteins including clock control genes (CCG).The transcription of two other circadian genes, namely period (Per) and cryptochrome (Cry), form a protein dimer that inhibits the transcription of Bmal1, thus completing the feedback loop.The circadian clock genes establish an evolutionary conserved intrinsic timing system at a cellular level and are ubiquitously expressed in many cells that acts as sensors, error detectors, controllers, and effectors of physiological systems.Even enucleated cells such as red-blood cells contain intrinsic circadian oscillations that are driven by peroxiredoxins (Edgar et al., 2012;Hastings et al., 2008).
The fundamental interaction between homeostatic and circadian programmed rheostatic mechanisms is evidenced through observation of the effects of genetic mutations in the clock genes on physiological systems.Animals with Clock mutations display complete arrhythmia regarding locomotor activity (Vitaterna et al., 1994) yet maintain homeostatic core body temperature (Shostak et al., 2013).The loss of circadian programmed rheostatic regulation of core body temperature can be achieved by targeted mutations in Bmal1 within ventromedial hypothalamic Sf1-neurons via disrupted thermogenesis due to loss of brown adipose tissue activity (Orozco-Solis et al., 2016).Indeed, circadian clock genes establish programmed oscillations in physiological systems directly, and independently regulate homeostatic set-point values established by key error-detection components (e.g., corticotropin-releasing hormone; Jones et al., 2021).
The SCN in the hypothalamus, is an evolutionarily ancient structure that provides the predominant signal that governs circadian oscillations in physiological processes (Fig. 2D) (Patton and Hastings, 2018).Initial studies using neuroanatomical lesions of the SCN demonstrated that rodents become arrhythmic in locomotor activity and ingestive behaviours (Stephan and Zucker, 1972).It is now well established that the SCN provides the predictive, programmed change in physiological set-points.In the past thirty years, the molecular and cellular (Takahashi, 2017), and neural circuits (Paul et al., 2020) in which the SCN entrains programmed rheostasis in set-points values have been described and shown to comprise a distinct neuroanatomical substrate in which set-point values can be regulated independently from the circuits that control homeostasis (Herzog et al., 2017).For example, afferent projections from the SCN innervate multiple hypothalamic nuclei that govern error detection and controllers (Fig. 2D).Using core body temperature as an example, daily rhythms in this variable are common across animals (Refinetti and Menaker, 1992).Lesions that destroy the SCN abolish daily rhythms in body temperature in hamsters (Ruby et al., 1989) and rats (Eastman et al., 1984) yet animals do not experience hypo-or hyperthermia.In these instances, core body temperature is maintained by the temperature sensitive neurons in the POA but the amplitude in daily rhythms is eliminated.In situations in which the SCN or circadian clock genes in tissue/tissues are ablated, there are severe consequences of dysregulation in core body temperature, but homeostatic cells and circuits retain functionality to allow activation of 'emergency measures' (i.e., hypothermia).
Like circadian rhythms, circannual rhythms are an innate, endogenous mechanism of timing seasonal changes in multiple physiological processes (Helm and Stevenson, 2014).As with circadian rhythms the homeostatic and rheostatic control of the seasonal changes in energy balance, thermoregulation, and reproduction can also be clearly dissociated.In this regard, the Siberian hamster (Phodopous sungorus) has yielded significant insights into the genetic and neural mechanisms in which rheostasis provides long-term input to regulate seasonal changes in set-point values which control energy balance (Lewis and Ebling, 2018).Hamsters breed and maintain heavy body mass during the summer periods and in the laboratory reproduce and engage in elevated food intake when day lengths (also referred to as photoperiod) are longer than 12 h (typically 16 h light; 8 h darkness) (Bao et al., 2019).A simple shift to short winter-like photoperiods will induce a cessation in reproduction and involution of gonads; reduced body and adipose mass; and hamsters experience bouts of torpor (Bao et al., 2019;Haugg et al., 2022).A simple change from long t short photoperiods induces a programmed rheostatic change in body mass in the SD animals reflected in an approximately 30 % loss in body mass (Fig. 2E).
Challenges to homeostatic energy stability in the Siberian hamster such as neuroendocrine and physiological responses to simple food restriction manipulations allows differentiation of the systems which are engaged during short-term; versus long-term programmed rheostatic processes.For example, when hamsters housed in short photoperiod are food restricted there is a significant reduction in body mass (approximately 2 g) and once returned to ad libitum feeding, body mass returns to the programmed seasonal body mass value (Fig. 2E).Here hamster body mass is readjusted to the seasonally appropriate rheostatic setpoint.The same energy homeostatic and energy rheostatic changes in body mass has been reported in Svalbard ptarmigans (Mortensen and Blix, 1985) indicating a conserved response in chordates.Siberian hamsters that are kept in either long-or short-photoperiod and are subjected to acute food restriction both enter negative energy balance and typically lose approximate 1-2 g in body mass (Fig. 2F) (Bao et al., 2019).As the change in body mass is equivalent between the two seasonal physiological states, this indicates an acute physiological regulation whereby the animals are attempting to homeostatically maintain body mass.Acute food restriction in hamsters provides a powerful experimental design to interrogate the cellular, and genetic differences between homeostatic and rheostatic processes.
A series of discrete neuropeptides located in the arcuate nucleus are well established to provide orexigenic and anorexigenic control over T.J. Stevenson energy stability (Yeo and Heisler, 2012) and potential substrates for either energy homeostasis or rheostasis.Neuropeptide Y (Npy) and Agouti-related protein (Agrp) integrate nutritional signals from adipose, gut and pancreatic tissues and stimulate feeding intake via melanocortin-4 receptors.Interestingly, Npy and Agrp expression remain comparable between short and long photoperiod conditions despite the 30 % reduction in hamster body mass (Fig. 2G-H).However, in food restricted hamsters, the levels of Npy and Agrp are increased in both long-and short-photoperiod conditions demonstrating that the orexigenic neuropeptides are activated in response to the short-term homeostatic demands (Reddy et al., 1999;Bao et al., 2019).Short-term homeostatic changes in these neuropeptides in response to fasting and feeding are observed across mammalian and avian species (reviewed by Helfer and Stevenson, 2020) and reliably induce controllers to maintain energy homeostasis.This contrasts with Pomc (Reddy et al., 1999;Bao et al., 2019) and somatostatin (Sst; Dumbell et al., 2015;Bank, J.H.H, 2017) the expression of which exhibit robust changes between the longand short-photoperiod physiological states but crucially, do not change in response to short-term homeostatic challenges (i.e., fasting) (Fig. 2I-J).These data establish that anorexigenic neuropeptides Npy and Agrp are genetic correlates of energy homeostasis whereas other neuropeptides (e.g., Pomc and Sst) govern programmed energy rheostasis.

Biological basis of reactive rheostasis
Not all regulated changes in physiological set-points oscillate and instead function to induce a transient adjustment.Fever, or pyrexia, is an evolutionary conserved physiological response to increase core body temperature above a normal set-point (~40 • C) which occurs during the activation of the innate immune system (Evans et al., 2015).Here reactive rheostatic systems that involve the induction of fever act on distinct cellular populations that are different from homeostatic systems to regulate a transient change in core body temperature.Increased core body temperature occurs due to the immune induction of elevation in systemic interleukin-1β signalling which acts on brain endothelial cells to increase prostaglandin E (PGE2) synthesis and secretion into the preoptic area (Park et al., 2016).PGE2 binds to EP3 prostaglandin receptors (EP3R) on temperature-sensitive neurons in the preoptic area to drive an increase in the set-point of core body temperature (Lazarus et al., 2007).PGE2 is an evolutionary conserved hormone that acts to induce a temporary increase in core body temperature established by temperature-sensitive neurons in the preoptic area.In this case, brain endothelial cells are the substrates for reactive rheostatic regulation of core body temperature during fever providing direct input to the control of the temperature set-point which is then maintained by the homeostatic systems.These homeostatic and reactive rheostatic response systems can also be differentiated at a behavioural level.For example, humans have a linear preference between surface temperature (i.e.hand) and core body temperature (Fig. 2K).In response to an intramuscular injection of Propidon (attenuated Streptococcus, Staphylococcus and Bacillus pyocyaneus), core body temperature significantly increased and induced a fever state, reflecting a regulated change in set-point.Despite the elevated core body temperature during the fever rheostatic response, the homeostatic variation in core body temperature, measured by temperature preference in the gloved hand was nearly identical compared to control conditions.Here, a linear shift to a higher temperature preference is a key indicator of the temporary change in core body temperature set-point value which is associated with elevated PGE2 stimulation.

Modelling physiological dynamics
Homeostasis, rheostasis and allostasis can be modelled by distinct mathematical formulations.The dynamics of homeostasis are best captured using models of control theory (Maxwell, 1868).A simple formulation to homeostasis is to simply obtain the difference between the set-point (Sp) and error value (e) and divide the value by the regulated value (y).Control theory requires feedback to correct deviations in a physiological variable which are either elevated or reduced compared to a set-point value.The addition of a closed-loop controller permits feedback input from the effector to be integrated by sensors and any deviation from a compared value (i.e., set-point) is transmitted to a controller to initiate a change in the effector.In a Control theory mathematical equation, the homeostatic components include a controller (C), the system or physiological effector (E) and the sensor (S).If we can take the functional values of C, E, and S and plug these values into the equation for the closed-loop transfer function (feedback) and calculate homeostatic stability as: In this equation, the error value consists of the functional value of the controller and effector divided by the functional value of the controller, effector and sensor.The numerator is the feed-forward gain from the setpoint and the denominator is the closed loop transfer function.Together, the numerator and denominator form the overall error signal (e) and provides the closed-loop component for the control of the regulated variable (y).Alternatively, mathematical models of rheostasis commonly used are Fourier analyses or Lomb-Scargle periodogram for programmed and reactive regulated changes, respectively.One of the most common analyses of programmed circadian rheostasis uses time series analyses that include cosinor rhythmometry (Refinetti et al., 2007).Thus, programmed rheostasis can be described as: In this equation, M is the MESOR (Midline statistic of rhythms, or the amplitude mean), A is the amplitude (a measure of half variation in the cycle), ϕ is the acrophase (the time of overall high values recurring in each cycle) and τ is the period (duration of one cycle), and e is the error term.
Finally, a leading mathematical model of neural processing, especially in hippocampal circuits, are Hopfield networks (Hopfield, 1982), and can be used to characterize allostatic systems.At the simplest level, Hopfield networks describe the relation between two neurons or 'nodes' and using Hebb's law that neurons wired together, fire together, thus allostatic input can be formulized as: Here, w ij is the strength of the relation between neurons i and j, N is the number of nodes in the network, and p i is the value of the i-th node.
Hopfield networks have expanded beyond Eq. 3 to include measurements of 'energy' which determines the stability of discrete connections.When stability is maintained, whenever the state of any 'neuron' changes, the energy function will decrease.With sufficient data that incorporates short-term homeostatic, long-term regulated rheostatic and experience-induced changes in a set-point value, a single mathematical model using differential equations which incorporates control theory, time-series analyses (i.e., Fourier analyses) and Hopfield networks could, in theory, model any physiological process.Fig. 4 describes a theoretical equation that captures homeostatic closed feedback control (error function + set point value) in which a rheostatic Fourier analysis component is incorporated along with allostatic input in the form of Hopfield networks.From the homeostatic closed-loop function in Eq. 1, the main error value forms the inside bracket (i.e., left) and each component (i.e., sensor, controller and effector).Rheostatic input (r [Eq.2]) is provided directly into each control theory component and directly modifies the set-point.Allostasis (a [or w ij in Eq. 3]) only impacts the set-point value and is dependent on rheostatic mechanisms.The maintenance of the set point Sp is a product of the homeostatic value, the programmed (and reactive) rheostatic factor and the allostatic value.

Physiological Stability
In this equation, both the rheostasis and allostasis value range from − 1.0 to 1.0.The R denotes the natural range for the regulated variable.In this case, the numerator is simply the sum of rheostatic and allostatic input.The denominator is the product of rheostatic and allostatic input plus one.In a simple test of the equation, when programmed circadian rheostatic input to the set point is at the peak (i.e., 1) and there is zero allostatic input, the computation is simply r/1 (or 1) and half the naturally occurring range of the regulated variable.In this scenario, the formulation for stability will be the product of closed loop transfer function and the daily peak for the regulated variable.Conversely, a regulated variable at the mesor (midpoint) will have a value of 0, and only allostatic input will modify the set point.If there is no rheostatic, nor allostatic input, then the regulated value will only be driven by the error value and the homeostatic set point.In theory, the model can be expanded to include all physiological processes or regulated variables to provide a comprehensive metric.The Greek uppercase letter sigma ( ∑ ) represents the summation of all regulated variables and range from one (or i) to the n th physiological variable (n = 5 if investigating five different variables).

Physiological stability is a hierarchical organization
Physiological systems are controlled at multiple cellular levels and neural circuits indicating a hierarchical structure.The physiological regulation of circulating cortisol is an excellent example of how homeostasis, rheostasis and allostasis independently regulate adaptive, stable changes in a physiological process.Cortisol exhibits a programmed circadian rheostatic oscillation but is also stimulated at time when an individual encounters an allostatic stressor.Circulating cortisol concentration set-point value is established by parvocellular cells in the paraventricular nucleus (PVN) that express corticotropin-releasing hormone (CRH) (Fig. 3A).CRH released into the hypophyseal portal system stimulates the controller (i.e., corticotropes) to release adrenocorticotropin (ACTH).ACTH travels in the systemic circulation to the effector (i.e., adrenal cortex) and acts in the zona fasciculata to initiate the synthesis and secretion of the physiological variable (i.e., glucocorticoid; Lightman et al., 2020).Cortisol has rapid and dose-dependent negative feedback via the glucocorticoid receptor to maintain a stable state (Andrews et al., 2012).Here, the glucocorticoid receptor in the PVN acts as the sensor to detect any deviation from the cortisol set-point established by the CRH neurons.During the midpoint of the light phase, ultradian pulsatile activity of CRH neurons maintain serum cortisol concentrations around 150 nmol/l (Fig. 3B).Yet, serum cortisol levels are highly dependent on the time of day.Here rheostatic input in the form of direct innervation of CRH cells by vasoactive intestinal peptide (VIP) expressing neurons in the suprachiasmatic nucleus (SCN) establish programmed, regulated changes in serum cortisol concentrations (Fig. 3B).These VIP neurons project from the SCN to the CRH neurons in the PVN to entrain circadian clock gene rhythms and in turn establish the daily cycle in the hypothalamo-pituitary-adrenal axis (Jones et al., 2021).Allostatic control of set-points established by CRH is communicated via glutamatergic cells located in the ventral hippocampus.These hippocampal cells provide inhibitory input to CRH neurons in the PVN likely via the bed nucleus of the stria terminalis (BNST) (Cole et al., 2022).Individual, experience dependent control of CRH derived from allostatic input can be illustrated by findings showing that early experience to environmental stimuli (i.e., developmental programming) can alter neural pathways that govern how an individual will respond the stimuli in adulthood.A good example is maternal licking and grooming pups early in life.Here high levels of maternal licking and grooming behaviour reduces DNA methylation in the glucocorticoid receptor which permits protein synthesis in the hippocampus (Weaver et al., 2004).Higher glucocorticoid receptor expression in the hippocampus is associated with reduced stress responses in rats compared to conspecifics raised by low licking and grooming mothers (Weaver et al., 2004).
Evidence such as the level of glucocorticoid receptor in key brain regions involved in emotion and cognition illustrate how individual experience can shape the control of physiological set-points and establish the variation in physiological and behavioural responses to the same stimulus (i.e., stressor).
A hierarchical structure is also observed in ecologically relevant conditions.Unpredictable environmental perturbations are common across habitats and are increasing in frequency and intensity due to global climate changes.Seasonal variation in the stress response is common (Borniger et al., 2017).Birds, such as the European starling (Sturnus vulgaris) have robust programmed daily and seasonal rheostatic rhythms in glucocorticoid concentrations in laboratory conditions.However, the glucocorticoid response to a stressor (i.e., restraint) vary seasonally with lower corticosterone levels in non-breeding physiological states compared to pre-breeding and breeding (Romero and Remage-Healey, 2000).Lapland longspurs (Calcarius lapponicus) breed in the high arctic and exhibit robust programmed seasonal rheostatic rhythms in corticosterone; furthermore, the allostatic response to capture stress is modulated with a dampened response during the nonbreeding state (Astheimer et al., 1995).Unpredictable natural environmental stressors (i.e.storm) also induce robust increases in plasma corticosterone analogous to capture stress (e.g., Astheimer et al., 1995).Similar seasonally adjusted allostatic responses to capture stress are observed in desert habitats at a lower latitude (Wingfield et al., 1992).These patterns of circulating glucocorticoid levels support the notion that seasonally programmed rheostatic processes control predictive, regulated adjustments in glucocorticoids by CRH neurons and that allostatic input via stress responses are modified during unpredictable environmental perturbations.

Conclusions
This review aimed to build a clear and comprehensive model that dissociates homeostasis, rheostasis and allostasis systems at multiple biological levels of analyses (i.e., genetic and neural).We can now start to develop a framework for outlining the three distinct physiological control systems that serve to maintain set-points (Fig. 4).Homeostasis is evolutionarily conserved and provides the primary input for the maintenance of set-points and could be classified as primary stasis.Homeostatic stability predominantly requires sympathetic and parasympathetic inputs, hypothalamo-pituitary-tissue axes and induce rapid (i.e., seconds to minutes) modifications to controllers, effectors and sensors to achieve the desired set-point.Any damage to homeostatic control often leads to the activation of emergency responses and usually lethal outcomes (e.g., hypothermia).Rheostatic mechanisms are also evolutionary conserved, intrinsic processes which provide regulated input to set-points that serve to increase an individual's fitness by anticipatory changes in physiological variables.Substrates for rheostatic input are predominantly genetic and function in specific hypothalamic nuclei/cellular to provide direct input to homeostatic error detectors.This additional layer of control could be classified as secondary stasis.Unlike homeostasis, damage to rheostatic inputs (e.g., SCN) does not necessarily cause life-threatening physiological changes, but instead causes pathological conditions (Wulff et al., 2010).Finally, allostasis includes inputs from a series of higher order brain structures (e.g., cortical/pallial) which integrate individual, species-specific environmental cues and cognitive states to induce a transient change in a physiological set-point.Like rheostasis, damage to allostatic input is not lethal, but induces long-term pathological conditions and could be considered as tertiary stasis.The stress response and dieting are two examples in which allostatic inputs can drive a change in physiological set-points (Fig. 4).This theoretical framework provides a structure in which homeostasis acts to establish the fundamental level for any given set-point and rheostatic and allostatic systems provide direct, and independent input to modify set-point levels.Our understanding the neural connections and coordinated interactions between hierarchical levels (i.e., rheostatic and allostatic) is limited, further scientific attention to the hodology of physiological stability will be a fruitful endeavour.
The present paper described the control of physiological set-points within well-defined physiological processes but did not include how a regulated variable can be adjusted by multiple neural circuits in parallel.Physiological set-points are maintained by homeostatic, rheostatic and allostatic systems.Homeostasis is a primary stasis that functions to maintain evolutionary conserved, innate physiological processes that are essential for survival (e.g., blood chemistry).Rheostasis provides higher order, secondary stasis to optimize physiological set-points by establishing a regulated change that functions to anticipate a change in the environment.Regulated changes in physiological stability are common across the animal kingdom and serve to enhance animal fitness.Allostasis provides a level of top-down tertiary stasis which is experience-dependent and provides species-specific, and individual variation in adaptive changes in physiological set-point (e.g., reactions to perceived stressor).Examples of physiological processes are listed below homeostasis, rheostasis and allostasis.
The concept of poikilostasis which is defined as various states of equilibrium was developed by Wayne Kuenzel et al. (1999) to account for how different neural systems can modify the homeostatic set point of a physiological variable.For example, adjustments in either sympathetic or parasympathetic neural systems can independently change the homeostatic set-point for body mass.Undoubtedly, homeostatic set-points can be modified from multiple neural systems in parallel and consequently, the cellular and neural basis of 'lateral processing' of programmed and reactive rheostasis, and allostasis.The implications of poikilostasis and parallel processing of physiological set-points requires further attention and is beyond the scope of the current manuscript.
A hierarchical organization of internal stability has significant implications for understanding most physiological, immunological, behavioural and cognitive processes.Several homeostatic systems serve to ensure survival of the organism such as maintaining core body temperature or blood pH values within narrow limits (Fig. 4).The underlying causes of many acute and chronic illnesses are not necessarily due to changes in homeostatic systems and instead should be consider as disruption to rheostatic (e.g., shift-work; social jet-lag; climate change) or allostatic (e.g., stress response) systems.Alterations in local climates driven by global change is causing increased frequency, duration and intensity of environmental perturbations for wild animals which has negative impacts for programmed seasonal rheostasis (Stevenson et al., 2015).Significant advances have been achieved in understanding how these factors impact allostatic processes at a physiological level (Wingfield, 2005).The framework outlined in this paper will help improve our understanding of how global change has different impacts on the neural control of rheostatic regulation of physiological set-points (i.e., seasonal timing) versus allostatic modification of facultative setpoints (i.e., emergency life history responses).It will be important to shift the current focus from a single homeostatic mechanism of internal stability and adopting terminology that accurately reflects the genetic and neural basis of how physiological set-point values are adjusted.This Review is particularly relevant for the field of circadian biology where the use of 'homeostasis' to describe daily changes in genetic, cellular, and hormonal signalling needs to be reconsidered as distinct changes in programmed rheostasis.

Fig. 2 .
Fig. 2. Schematic representation of rheostatic processes.Rheostasis provides regulated changes in homeostatic set-point values directly from key hypothalamic cell populations (A).Regulated changes in set-point values consist of regular, repeated oscillations governed by programmed rheostasis (e.g., daily rhythms) (B).Reactive rheostatic responses are a regulated change in setpoint values that occur over a single period (e.g., fever) (C).The downward arrow indicates the initiation of a reactive rheostatic response (e.g., infection).The suprachiasmatic nucleus (SCN) is a distinct neural region that provide programmed circadian rheostatic changes in homeostatic set-points maintained by other hypothalamic nuclei (D).Blue lines denote rheostatic input and red arrows indicate homeostatic outputs.Photoperiod induced changes in Siberian hamster body mass can be used to dissociate homeostatic and programmed rheostatic mechanisms (E).Hamsters in long photoperiod (LD; white circles) maintain large body mass.Hamsters transferred to short photoperiod (SD; dark circles) lose approximately 30 % of body mass after 8-12 weeks.Short term food restriction (FR) to SD hamsters (triangles) induces a homeostatic adjustment and triggers a loss in body mass.A switch back to ad libitum food (AL) access returns body mass to the seasonally programmed rheostatic set-point.Hamsters held in LD and SD show approximately 1 g reduction in body mass (F) illustrating the same homeostatic response is recruited despite the 15 g reduction in body mass driven by the programmed seasonal rheostatic response.Neuropeptides in the arcuate nucleus dissociate homeostatic and rheostatic processes (G-J).Neuropeptide y (Npy) (G) and agouti-related protein (Agrp) (H) are sensitive to short-term homeostatic changes in energy stability (i.e., fasting).Other neuropeptides are genetic correlates of programmed rheostatic control of seasonal energy stability and include proopiomelanocortin (Pomc) (I) and somatostatin (Sst) (J) and insensitive to short-term homeostatic signals.Note bar plots in G-J are schematic diagrams to represent data used to reflect 'low' (i.e., 0.5) and 'high' (i.e., 2.0) expression levels.Fever response in homeotherms is associated with a 1-2 • C increase in core body temperature for several hours to days (K).In control conditions (blue labels) humans maintain a core body temperature of approximately 37.5 • C. Participants sitting in a pool will adjust the temperature in a gloved-hand (preference temperature) which is negatively correlated with core body temperature.A simulated infection induces a reactive rheostatic fever response (red labels) associated with an elevation in core body temperature, yet the homeostatic variation (preference temperature) around the transient increase in core body temperature remains.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3 .
Fig. 3. Multiple levels of control for stability in circulating glucocorticoid concentrations.The hierarchical organization of physiological stability involves homeostatic, rheostatic and allostatic processes (A).Corticotropin-releasing hormone (CRF) in parvocellular neurons in the paraventricular nucleus (PVN) stimulate the release of adrenocorticotropin (ACTH) from corticotropes in the anterior pituitary gland (Pt).ACTH passes through the circulatory system to the adrenal gland to stimulate cells in the zona fasciculata to synthesize and release glucocorticoids (i.e.cortisol).Glucocorticoids provide negative feedback to CRH and ACTH expressing cells for homeostatic maintenance of set-points.(B) serum cortisol in humans in the early afternoon is maintained around 150 nmol/l.The grey box indicates the homeostatic range for the daily phase of serum cortisol.Negative feedback is essential to maintain physiological constancy in the regulated variable (i.e., cortisol) (C) Programmed circadian rheostatic control of daily cycles in serum cortisol are established by vasoactive intestinal peptide innervation from the suprachiasmatic nucleus (SCN).Light and dark box represent the daily cycle in light and dark phases, respectively.(D) Allostatic mechanisms are activated in responses to many real and perceived stressors, and act to rapidly increase serum cortisol for approximately 60-120 min before returning to the daily phase set-point.Individual variation or pathophysiological conditions lead to increased allostatic load reflected in a prolonged increase in serum cortisol, or an inadequate response.

Fig. 4 .
Fig. 4. Hierarchical organization of physiological stability.Physiological set-points are maintained by homeostatic, rheostatic and allostatic systems.Homeostasis is a primary stasis that functions to maintain evolutionary conserved, innate physiological processes that are essential for survival (e.g., blood chemistry).Rheostasis provides higher order, secondary stasis to optimize physiological set-points by establishing a regulated change that functions to anticipate a change in the environment.Regulated changes in physiological stability are common across the animal kingdom and serve to enhance animal fitness.Allostasis provides a level of top-down tertiary stasis which is experience-dependent and provides species-specific, and individual variation in adaptive changes in physiological set-point (e.g., reactions to perceived stressor).Examples of physiological processes are listed below homeostasis, rheostasis and allostasis.