What metabolic abnormalities would you expect when the energy inflow to the blood, brain or muscle/fat tissue is disrupted? This question divides the beliefs. To shed light on this issue, we have completed a three-part series of systematic reviews, the first two of which have recently been published (Sprengell et al., 2021a,b) and the third one we present here with this article (Figure 1).

Disrupted energy inflow to the blood, as occurs with caloric restriction, resulted in minor mass (energy) changes in the brain as opposed to major changes in the body. Disrupted energy inflow into the brain, as occurs with cerebral artery occlusion, led to an increase in systemic blood glucose. Disrupted energy inflow to the muscle/fatty tissue, as it occurs in type 1 diabetes mellitus (T1DM), is subject of the current systematic review.

What is the theoretical background that can lead to controversial predictions? A set of long-held theories on energy metabolism regards the brain as only passively supplied. This is the gluco-lipostatic theory (Kennedy, 1953; Mayer, 1953) and its modern variants (Chaput and Tremblay, 2009; Schwartz et al., 2017). In contrast, the Selfish-Brain theory describes the brain as a self-regulating compartment that primarily covers its own high energy need (Peters et al., 2004). In this respect, the brain behaves “selfishly.” The current systematic review aims to test whether the gluco-lipostatic theory or the Selfish-Brain theory makes more accurate predictions.

The gluco-lipostatic model consists of two components, the lipostatic and the glucostatic. The lipostatic theory was founded by Gordon C. Kennedy (Kennedy, 1953). He believed that body mass was the regulated quantity in energy metabolism. He postulated that a signal from the body energy stores controls food intake. Yet he could not name this signal. Kennedy's lipostatic theory gained momentum in the 1990s when leptin was identified as the signal from adipose tissue that he had suspected (Zhang et al., 1994).

Glucostatic theories include Jean Mayer's original “glucostatic theory” and the “blood-glucose-insulin models” pioneered by Bergman and Cobelli (1980). Jean Mayer was the first to formulate a glucostatic theory (Mayer, 1953). In his model of energy metabolism, blood glucose is the most important physiological variable to be regulated and should be kept as constant as possible. That is why he called his model “glucostatic theory.” The “blood-glucose-insulin models” also see blood glucose as the main regulated variable. While Mayer's model assumed blood glucose to be controlled by adjusting food intake, Cobelli's and Bergman's blood-glucose-insulin models assumed blood glucose to be controlled by adjusting insulin-dependent storage in muscle and fat tissue. Against this background, Mayer's glucostatic model is only one representative of a class of glucostatic theories that consider keeping blood glucose levels constant as the main goal in energy metabolism.

The gluco-lipostatic theory and its modern variants, which combine Kennedy's and Mayer's model with a blood glucose-insulin model, have in common that they see the brain as passively supplied. This view on the brain becomes particularly clear in the blood glucose-insulin models, as they are formulated in mathematically explicit form. Cobelli's blood glucose-insulin model postulated that insulin-independent utilisation is constant and represents glucose uptake by the brain and erythrocytes (Dalla Man et al., 2007; Li et al., 2016). Therefore, in this model, the brain is supplied independently of the cerebral energy content, i.e., in a passive manner. Bergman's so-called “minimal model” postulated that insulin and glucose per se influence net glucose utilisation (i.e., production by the liver minus uptake by brain and muscle; Bergman, 2005). The latter assumption means that Bergman also saw the brain as only passively supplied.

The Selfish-Brain theory is formalised as a supply chain model (Peters and Langemann, 2009). Herein, the brain occupies the position of the final consumer. The brain is the organ that consumes the largest share of glucose in the organism (Reinmuth et al., 1965). In the cerebral supply chain, energy from the environment is taken up by the body (into the blood stream), and from there ~2/3 of the circulating glucose enters the brain. In the field of logistics such supply chains have been extensively studied (Slack et al., 2004). The logistic push-principle operates according to the following rule: the supplier delivers material and in so doing determines the activity of a production step. In contrast, the pull-principle works in the following manner: the material required for a production step is provided only when the receiver needs it (just-in-time). In comparison with the push-principle the pull-principle offers clear economic advantages; with the latter there are short set-up times and only small (economically optimised) storage sites (Figure 2, upper panel).

In the neuroenergetic supply chain, the “brain-pull” refers to the force with which the brain actively demands energy from the body. This brain-pull is the very feature that distinguishes the Selfish-Brain theory from the gluco-lipostatic theory and its variants. The “body-pull” is the force with which the body demands energy from the environment – a pull component that we usually refer to as hunger and which also occurs in the gluco-lipostatic model.

The existence of the postulated brain-pull mechanisms was supported by evidence from the 1990s showing that when a neuron fires and needs more energy to do so, it pulls glucose from the blood via the astrocytes (Pellerin and Magistretti, 1994). Against this background, the brain turns into an active part in energy metabolism according to the principle of “energy on demand” (Magistretti et al., 1999). Other brain-pull mechanisms have been discovered that are activated by sensors in amygdala (Zhou et al., 2010) and ventromedial hypothalamus (VMH) (Spanswick et al., 1997; Routh et al., 2014; Toda et al., 2016) when intraneuronal adenosine triphosphate (ATP) falls. These neuroendocrine mechanisms involve the sympathetic nervous system and the hypothalamic-pituitary-adrenal axis: first, cerebral insulin suppression prevents the body stores to take up glucose so that more glucose is available to the brain (Woods and Porte, 1974; Ahren, 2000; Hitze et al., 2010); second, visceral lipolysis and hepatic ketogenesis increase the availability of ketones to the brain as alternative energy substrates (Kubera et al., 2014); and third, acceleration of the heart rate increases the blood volume that reaches the brain, procuring it with additional energy (Jones et al., 2011).

A brain that works according to the energy-on-demand principle will under certain circumstances produce different metabolic outcomes than its passive counterpart. For a given model of energy metabolism, the expected outcomes will influence individual treatment decisions, guidelines, and practical recommendations, for example, for obesity, type 2 diabetes mellitus, and cardiovascular disease. Since such decisions can have far-reaching consequences for the individual and society, clarification should be brought about by testing the theories at issue.

So far, two out of three systematic reviews have put the theory of the Selfish-Brain to the test. In our first systematic review, the predictions of the Selfish-Brain theory were met in terms of what happens when the energy flow to the blood is disrupted (calorie restriction), while the predictions of the gluco-lipostatic theory and its variants were violated (Sprengell et al., 2021a). As mentioned above, during caloric restriction, the brain mass (energy) was much less affected than that of the body (Sprengell et al., 2021a). From the perspective of the gluco-lipostatic theory and its variants, brain and body should be equally affected by the lack of energy – which was not the case.

In our second systematic review, the Selfish-Brain theory continued to demonstrate the accuracy of its predictions by passing the second test, which dealt with disrupted energy inflow to the brain (occlusion of the cerebral artery) (Sprengell et al., 2021b). As already noted, the occlusion of the cerebral arteries increased systemic blood glucose (Sprengell et al., 2021b). The gluco-lipostatic theory and its variants failed with their prediction that states that during cerebral ischaemia the blood glucose level would remain “static” (i.e., kept constant). After all, this was not the case either.

The aim of our third systematic review, the current paper, is to again test the predictions of the Selfish-Brain theory against those of the gluco-lipostatic theory in experimental T1DM. Therefore, we have formulated a hypothesis on this issue that, if confirmed, would match the predictions of the Selfish-Brain theory – but not those of the gluco-lipostatic theory:

To this end, we systematically searched the literature to test whether the T1DM studies we found actually confirm this hypothesis or not.

We have pre-registered the protocol for this systematic review on PROSPERO on 30th of January 2020; updated versions were published on 28th of September 2020, 14th of December 2020, 24th of February 2021, and 9th of April 2021 (International prospective register of systematic reviews; CRD42020156816). The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines for systematic reviews of interventions have been followed (Moher et al., 2009) and the Cochrane Handbook for systematic reviews of interventions was used (Higgins and Thomas, 2019).

The systematic search of the literature generated 706 articles, which were processed as summarised in Figure 3. Six hundred and eight works were screened by title or abstract, and 64 articles were analysed by full text. We identified 18 studies that met all inclusion criteria, had no exclusion criteria, and focused on how experimental T1DM affects brain mass or energy vs. blood glucose concentrations.

A total of 608 works was screened by title and abstract, and 64 were analysed in full text. According to strict selection criteria defined in our PROSPERO pre-announcement and complying with PRISMA guidelines, 18 studies met all inclusion criteria. Thirteen studies provided sufficient data to test our hypothesis. These 13 evaluable studies (15 experiments) showed that in experimental T1DM, major glycaemic disturbance does not translate into major disturbance of brain energy homeostasis. The use of STZ or alloxan to destroy pancreatic beta cells was found to be very effective in producing hyperglycaemia (+294% change compared to controls), but much less effective in affecting brain mass or energy (−4% change compared to controls), leaving the blood compartment much more susceptible to perturbations than the brain (p < 0.001) (Figure 4). Thus, all 13 evaluable studies confirmed and none contradicted our hypothesis that experimental induction of T1DM causes minor changes in brain mass (energy) as opposed to major changes in blood glucose levels.

The insight that brain mass (energy) is resistant to hyperglycaemic oversupply is new. Although data on this topic have been available for many years, it is only the synopsis of data and the theoretical background that make the insight possible. Given that the Selfish-Brain theory makes accurate predictions in cases where those of conventional theories have failed (Sprengell et al., 2021a,b), there is reason to believe that the astrocytic-neuronal and systemic-neuroendocrine mechanisms as assigned by the Selfish-Brain theory provide a reliable explanation for brain mass (energy) being resistant to hyperglycaemic oversupply (Peters et al., 2004, 2007, 2017; Peters and McEwen, 2015).

In terms of mechanisms, the Selfish-Brain theory refers to the principle of supply chains stating that “when push increases, pull relaxes” (Figure 2, lower panel). The cerebral-supply-chain model predicts that an increase in the blood-push component (e.g., hyperglycaemia) is compensated by down-regulation of brain-pull components.

Brain-pull function is carried out at the astrocytic level (glutamatergic neuron demands energy from blood) and at the systemic level (brain demands energy from body). The astrocytic brain-pull describes the force with which glutamatergic neurons, when they fire, demand the required energy via the astrocytes from the blood (Peters and Langemann, 2009). Pierre Magistretti and Luc Pellerin were the first to describe a brain-pull mechanism and called it the Astrocyte-Neuron Lactate Shuttle (ANLS) (Pellerin and Magistretti, 1994). Since its introduction, the ANLS has received a great deal of attention, with approval as well as criticism; however, substantial evidence has accumulated from independent sources that supports the model (Mason, 2017). The ANLS hypothesis states that firing glutamatergic neurons release glutamate into the synaptic cleft, from where it is taken up by astrocytes; the incoming glutamate is a trigger for astrocytes to take up more glucose, which in turn is converted to lactate to be released via monocarboxylate transporters for neuronal use. Thus, according to the ANLS hypothesis, brain-pull activity depends on neuronal glutamatergic activity and subsequent uptake of glutamate in astrocytes. In contrast, GABA does not couple inhibitory neuronal activity with glucose utilisation, as does glutamate for excitatory neurotransmission (Chatton et al., 2003).

The demand and need of glutamatergic and GABAergic neuron populations can be fundamentally changed during a lack or excess of brain energy concentrations. In this respect, the Selfish-Brain theory made detailed predictions many years ago (Peters et al., 2004): These forecasts were based on experimental findings showing that in different brain regions, e.g., cortex, hippocampus, there is a typical constellation where low-affinity ATP-sensitive potassium (K_ATP) channels are localised on presynaptic GABAergic neurons and high-affinity K_ATP channels on post-synaptic glutamatergic neurons (Luhmann and Heinemann, 1992; Ohno-Shosaku et al., 1993; Lee et al., 1996; Jiang and Haddad, 1997) (Figure 5). Latest work also shows the key role of high affinity K_ATP channels on pyramidal cells (Karagiannis et al., 2021). The predictions were as follows: When neuronal ATP concentrations fall, this type of multi-site neuronal ensemble leads to biphasic responses. In mild ATP deficiency, low-affinity K_ATP channels hyperpolarise GABAergic neurons, thereby disinhibiting post-synaptic glutamatergic neurons and increasing glutamatergic activity (which serves a pull function), while in severe ATP deficiency, high-affinity K_ATP channels hyperpolarise glutamatergic neurons, causing glutamatergic activity to cease (which serves a neuroprotective function). Such a theory-predicted biphasic response to falling ATP levels could indeed be confirmed experimentally (Steinkamp et al., 2007). However, in the case we are interested in here, namely the case of an intraneuronal ATP excess, the presynaptic K_ATP channels of the GABAergic neurons close completely, so that these neurons are more likely to depolarise. As the presynaptic GABAergic neurons inhibit the postsynaptic glutamatergic neurons, an ATP surplus will shift the glutamatergic/GABAergic balance towards the GABAergic neuron population.

A shift of the glutamatergic/GABAergic balance has two major consequences, one for the cerebral energy budget, the other for cognitive function. The cerebral energy budget is affected in that the glutamatergic neurons, which are the ones involved in the energy demand process, become less active in the presence of surplus energy. As a result, the ensemble of astrocytes, glutamatergic neurons and GABAergic neurons takes up less energy during a cerebral energy surplus. The ATP surplus is thus reduced over time and energy homeostasis is restored. This prediction, based on the Selfish-brain theory, is consistent with the finding from our current systematic review that despite hyperglycaemia brain energy homeostasis can be maintained.

Cognitive function is also affected by a predominance of cerebral GABA tone. Since prevailing GABA-inhibition of glutamatergic neurotransmission limits information processing, we would expect cognitive dysfunction to occur. In fact, cognitive dysfunction has been shown to occur with hyperglycaemia, no matter whether type 1 diabetes or type 2 diabetes (Moheet et al., 2015). In type 2 diabetes, hyperglycaemia was found associated with increased cerebral GABA levels, which went along with cognitive dysfunction (van Bussel et al., 2016; Thielen et al., 2019). STZ diabetes was found to exhibit decreased cerebral glutamate levels and increased cerebral GABA levels (Datusalia and Sharma, 2016), and the animals displayed cognitive dysfunction (Lin et al., 2018). The protection of the cerebral energy budget against hyperglycaemic oversupply thus has its price, namely the impairment of cognitive performance.

Systemic brain-pull function is mainly carried out by the sympathetic nervous system (SNS). The SNS supplies the brain with additional energy when needed. Systemic brain activity may result in the blood containing more substrates (e.g., glucose, ketones, or lactate) for brain use (Hitze et al., 2010; Kubera et al., 2012, 2014) or in more blood flowing to the brain (e.g., due to increased heart rate, which increases cardiac output) (Jones et al., 2011).

More substrates for brain use can be provided by different mechanisms. In the VMH, energy sensors are positioned to detect a decrease in cerebral ATP concentrations. Specifically, VMH neurons are targeted by GABAergic presynaptic nerve terminals, the latter being equipped with K_ATP channels (Chan et al., 2007). Upon the decline in ATP concentrations in the VMH, the K_ATP channels can no longer bind ATP, so that the presynaptic GABAergic nerve terminals become hyperpolarised, thereby disinhibiting the postsynaptic glutamatergic VMH neurons. The postsynaptic VMH neurons, via their glutamatergic output, activate the autonomic centres in the brainstem, which in turn increase the tone of the SNS (Tong et al., 2007; Lindberg et al., 2013). SNS activation in turn leads to the release of noradrenaline from the sympathetic nerve endings and adrenaline from the adrenal medulla. In one of the sympathetic target organs, the endocrine pancreas, SNS nerves stimulate the release of glucagon from alpha cells (Havel et al., 1996) and suppress the release of insulin from beta cells (Ahren, 2000).

Some researchers studied the behaviour of the VMH-pancreas axis as a whole. Chan et al. (2007) showed that a decrease in GABA release in the VMH increased plasma adrenaline and glucagon levels. Stanley showed that acute activation of specific VMH neurons increases plasma glucose and glucagon, but suppresses insulin levels (Stanley et al., 2016). And Meek and colleagues showed that the activation of certain VMH neurons increases blood glucose without increasing insulin concentrations (Meek et al., 2016).

According to the supply chain principle “when push increases, pull relaxes,” the brain-pull is expected to be turned down. Indeed, Chan et al. showed in another paper that STZ diabetic rats exhibited increased GABAergic inhibition in the VMH resulting in blunted glucagon and adrenalin responses (Chan et al., 2011). Furthermore, the researchers showed that when GABA tone in the VMH of STZ-induced diabetic animals was lowered either pharmacologically or genetically, they were able to restore the absent glucagon response and also normalise the impaired adrenalin response (Chan et al., 2011). This study in STZ rats clearly showed that glucagon is in principle present in pancreatic alpha cells, but is only suppressed by output from the VMH. The brain, by down-regulating SNS activity, can suppress glucagon secretion, thus limiting further hepatic glucose production.

In T1DM, insulin secretion can no longer be controlled by the brain, as the irreversible damage to the beta cells leaves no insulin at all. Thus, from the onset of type 1 diabetes, an important brain-pull mechanism for controlling glucose energy flows within the organism gets lost, namely cerebral insulin suppression (CIS) (Woods and Porte, 1974; Ahren, 2000; Hitze et al., 2010). Normally, as mentioned earlier, the brain can suppress insulin secretion via sympathetic efferents when it needs energy (Ahren, 2000). This suppression prevents muscle and adipose tissues from taking up glucose via insulin-dependent pathways (Shepherd and Kahn, 1999) and instead makes the circulating glucose available to the brain via insulin-independent pathways (Deng et al., 2014). For the brain needs virtually no insulin to take up glucose (Hom et al., 1984; Hasselbalch et al., 1999; Seaquist et al., 2001). But if insulin is completely absent, the neuroendocrine mechanisms can no longer let insulin levels rise, so they cannot make surplus blood glucose be stored in muscle and fat tissue. The impact of this inability is reflected in the massive body mass change of −30.7 ± 13.3% (compared to controls) that was evident in this systematic review (Table 1). Such an ineffectiveness of CIS in T1DM means that the other brain-pull mechanisms that act through the sympathetic nervous system have to be turned down in order to protect the brain against oversupply. For example, a downregulated sympathetic nervous system could contribute to the reductions in heart rate (Howarth et al., 2005), circadian variation in heart rate and pulse pressure (Hicks et al., 1998) observed in type 1 diabetes. In all, downregulation of brain-pull mechanisms (shifted glutamatergic/GABAergic balance, reduced SNS activity and glucagon secretion) can help maintain cerebral energy homeostasis despite increased blood-push (hyperglycaemia).

The dataset found in our current systematic review violates the predictions of the gluco-lipostatic theory and its variants, which view the brain as purely passively supplied. From the perspective of these theories, experimental T1DM is expected to affect the energy content of the blood and brain in the same order of magnitude – which was not the case. This is the third time that the gluco-lipostatic theory and its variants have failed the test:

These tests systematically examined the three possible disruptions within the cerebral supply chain (Figure 1). We pre-registered all three tests at PROSPERO before we started any of our systematic literature searches. The evidence obtained from this testing may give reason to update the basic tenets of energy metabolism. Updating the long-held theory could replace a brain that is only passively supplied with a brain that independently regulates its energy concentrations and takes a primary position in a hierarchically organised energy metabolism. The latter proposition represents the axiom of the Selfish-Brain theory (Peters et al., 2002).

What could be the clinical significance of updating the basic beliefs about energy metabolism? Of course, the reviewed cases of caloric restriction, cerebral artery occlusion, and type 1 diabetes represent only a very limited group of metabolic states. In contrast, stress, sleep, obesity, and type 2 diabetes mellitus are much more common and have greater overall clinical relevance. Mental stress increases global brain glucose uptake and deep sleep decreases it, indicating that brain energy metabolism is not a constant (Maquet et al., 1990; Madsen et al., 1995). Most research on obesity and type 2 diabetes still relies on the gluco-lipostatic theory and its variants (Schwartz et al., 2017). Given that these gluco-lipostatic theories exhibit explanatory abnormalities, as demonstrated by the three-part series of systematic reviews completed here, the question arises whether these shortcomings also affect mainstream explanatory models for obesity and type 2 diabetes. Do these models overlook an important influence of the brain's energy demand and need on peripheral energy metabolism? Therefore, in the next research step, it should be questioned whether the long-held explanatory models for obesity and type 2 diabetes still offer the best explanation.

The data analysed in this study is subject to the following licences/restrictions: systematic review. Requests to access these datasets should be directed to achim.peters@uksh.de.

MS developed the search strategies that BK and AP approved. MS screened the article titles or abstracts against the inclusion and exclusion criteria. BK checked this step, and disagreements were resolved where necessary by consulting AP. MS and BK independently analysed the full text, and disagreements were resolved where necessary by consultation with AP. MS extracted the data, which BK and AP independently checked. MS assessed the risk of bias, which BK and AP independently checked and approved. BK and AP wrote the manuscript. All authors read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.