GluT4: a central player in hippocampal memory and brain insulin resistance.

Abstract

Insulin is now well-established as playing multiple roles within the brain, and specifically as regulating hippocampal cognitive processes and metabolism. Impairments to insulin signaling, such as those seen in type 2 diabetes and Alzheimer’s disease, are associated with brain hypometabolism and cognitive impairment but the mechanisms of insulin’s central effects are not determined. Several lines of research converge to suggest that the insulin - responsive glucose transporter GluT4 plays a central role in hippocampal memory processes, and that reduced activation of this transporter may underpin the cognitive impairments seen as a consequence of insulin resistance.

Background: glucose, insulin, and hippocampal memory.

Over the past two decades or so, insulin signalling within the brain has moved from being seen as non-existent (or, at most, a minor player in hypothalamic fuel-sensing¹) to being recognised as an important scientific and clinical focus, central to the impact of both type 2 diabetes (T2DM) and Alzheimer’s disease (AD) - indeed, the latter has been characterised in some cases as ‘type 3 diabetes’ ^{2, 3}. The fact that insulin is a key component of hippocampal memory processes, and the concept of systemic insulin resistance extending to the brain such that it impairs both hippocampal metabolism and cognitive function,^4-7 are now well-established ^{6, 8-20}.

Study of the metabolic regulation of cognition, and specifically of hippocampal memory processing, began with investigation of glucose, the brain’s primary fuel source (which field in turn emerged from earlier work on the mechanisms modulating the impact of stress on memory²¹). Reviews both in this issue and elsewhere have covered this history in detail: administration of exogenous glucose, either systemically or directly into the brain, is well established to acutely improve performance on challenging cognitive tests, both in laboratory animals and in a wide range of human populations ^22-25. Importantly, glucose supply to the hippocampus was shown to be a constraint on performance of challenging memory tasks, so that cognitive demand depletes ECF glucose levels within brain regions involved in processing the specific task and to an extent that correlates with task difficulty ^{26, 27}. This task-associated depletion was later directly confirmed to be correlated with both increased local lactate production ²⁸ and caused by local glucose uptake and metabolism ²⁹. Depletion is local rather than brain-wide, occurring only in the brain regions in processing a particular task ^30-34. Glucose metabolism has become the primary marker for neural and cognitive activity in humans, forming the basis for techniques including fMRI and PET.

Conditions that impair brain glucose supply, such as aging, T2DM or AD, lead to more profound task-associated depletion of local brain glucose correlated with impaired cognitive performance ^35-40. Details of the exact causes of task-associated increases in glucose demand remain to be precisely described and may vary, but there are a large number of candidate processes downstream of increased local glycolytic metabolism: for example, metabolic support for the energetically costly Na⁺/K⁺ ATPase and promotion of enhanced cellular excitability and synaptic plasticity ^41-46.

Given the close relationship between local glucose metabolism and cognitive performance, interest in factors that might modulate brain glucose supply and metabolism increased: insulin was an early, and clinically-important, candidate and one whose receptors had already been shown to be altered, in the rat hippocampus, by water maze training ⁴⁷. Indeed, insulin was confirmed to be a key player in hippocampal cognitive processes: specific blockade of endogenous intrahippocampal insulin markedly impairs spatial working memory, while physiological doses delivered to the hippocampus enhance it ^4-6. Those studies also showed that diet-induced obesity and systemic insulin resistance (DIO; a rat model of lifestyle-induced type 2 diabetes) impairs both cognition and hippocampal metabolism in a manner resembling direct blockade of intrahippocampal insulin ⁵: the impairment seen resembles the cognitive and metabolic impairment seen in human patients with T2DM, who are especially prone to hippocampal dysfunction ^48-57. Enhancement of memory by insulin has also been shown in several human studies ^58-64, possibly in at least some cases by increasing insulin-mediated glucose transport through GluT4 ⁶⁵. In vivo studies using microdialysis directly confirmed that insulin acutely stimulates local hippocampal glycolysis ¹⁹, and other work has shown that insulin’s modulation of brain glucose metabolism is region-dependent, with cortex and hippocampus being most sensitive ^{18, 66-72}. Nonetheless, the mechanisms by which insulin regulates memory processes are less understood ^{6, 13, 14, 19, 20, 59, 65, 73-78}. Understanding the interplay of metabolic regulators is vital for understanding of both hippocampal function and the impact of e.g. T2DM and AD. The clinical fact that central insulin resistance causes cognitive impairment is now reasonably well-established, but not yet well understood at a cellular or molecular level.

In the periphery, insulin is thought of as a regulator of glucose. However, the converse is also true: release of insulin from the pancreas occurs in response to elevations in blood glucose. The molecular machinery that regulates such release is present in the brain, and has been shown to modulate hippocampal memory ^{41, 42, 79}; moreover, there is increasing evidence for local synthesis and release of insulin in the brain ^{6, 80-86}. This raises the possibility that one role of hippocampal insulin might be an effector of glucose’s procognitive actions, including via increased GluT4 translocation that permits increased glucose flux; this has been suggested ²⁵ but remains untested. One potentially complicating factor both here and in consideration of the links between insulin resistance and cognitive impairment is that insulin transport across the blood-brain barrier may decrease with age⁸⁷, so that reduced central insulin signaling might be seen even without diminished central insulin sensitivity.

Mechanisms by which insulin enhances memory

The focus of this paper is on GluT4 and insulin resistance - it is explicitly not intended to be a comprehensive review of mechanisms of brain aging - but it is important to note that regulation of GluT4 is far from the only mechanism by which insulin likely regulates cognitive processes: the brief discussion below is incomplete and omits major lines of research such as regulation of neurotransmission and neurotransmitter receptors (including NMDA, AMPA, and GABA receptors ^78-88-90). However, it is worth discussing a couple of mechanisms that appear to regulate hippocampal cognition, especially in the context of this issue’s focus on central insulin resistance.

Equally important and conversely, GluT4 in the context of memory processes is not solely regulated by insulin: several other receptor tyrosine kinases (e.g. insulin-like growth factor-1 receptors and TrkB receptors) and other receptors involved in memory formation may regulate GluT4 ^91-94. Downstream of receptor activation, GluT4 translocation is regulated by several post-receptor signaling molecules critical for maintaining long term memory including insulin-like growth factor 2, brain-derived neurotrophic factor, Ca²⁺/calmodulin-dependent protein kinase II, phosphoinositide 3-kinase (PI3K), protein kinase A, MAPK, protein kinase-λ, and -protein kinase-ζ ^91-101. Some non-receptor activation-dependent conditions are associated with GluT4 trafficking and activity, such as cellular depolarization via elevated intracellular [Ca²⁺] and elevated extracellular [K⁺] ^102-106; as noted below, this includes recruitment of GluT4 to the synapse as a consequence of neuronal activity ¹⁰⁷, a seminal recent finding that is consistent with prior work showing increased cerebellar GluT4 after prolonged exercise¹⁰⁸. In general, although insulin’s CNS actions are likely to be mediated in significant partvia GluT4, the two are not synonymous. This is consistent with the fact that although GluT4 and insulin receptors (IR) show some degree of brain co-localization ^{17, 109}, there is not a one-to-one correlation, in contrast to peripheral tissues where the two showtight co-localization.

Antagonism of beta-amyloid

One insight into likely mechanisms of insulin action in the brain comes from clinical studies which show a clear link between T2DM and development of AD, and that many AD patients have reduced hippocampal insulin signalling along with brain hypometabolism and accumulation of beta-amyloid (Aβ)^{54, 74, 110-117}. Several studies including in vitro work suggest that insulin and Aβ oppose each other at a molecular level ^118-126; recent work showed that acute administration of oligomeric Aβ to the hippocampus causes rapid cognitive impairment, reduced local glucose metabolism, and impaired translocation of the insulin-regulated glucose transporter GluT4 ⁷, a pattern that closely resembles hippocampal insulin resistance. Related studies, discussed further below, showed that in DIO rats, specific antagonism of intrahippocampal oligomeric Aβ reversed cognitive impairment, suggesting a key interaction between insulin and Aβ in modulation of hippocampal cognitive processing. Insulin regulates Aβ processing and removal from the brain ^{127, 128}, a large literature shows that the mutual antagonism between insulin and beta-amyloid mutually extends to e.g. glucose metabolism, glutamate signalling, and other pathways ^{60, 129-137}, such that preventing impairment from accumulation of oligomeric Aβ is an important procognitive role of hippocampal insulin.

Insulin-regulated aminopeptidase (IRAP)

GluT4 translocates to the cell surface via complex machinery ^{138, 139} downstream of insulin, primarily through phosphatidyl inositol 3-kinase (PI3K), and Akt¹⁴⁰. Activation of Akt inactivates glycogen synthase kinase 3 beta (GSK3β) and activates AS160 ^141-144. Activation of AS160, along with non-PI3K mechanisms, permits release of GluT4-containing glucose storage vesicles (GSVs) from their intracellular tethers ^{138, 139, 143, 145-147}, allowing GluT4 to move to the cell surface. However, the GSVs contain proteins other than GluT4, including some involved in regulation of memory processes such as insulin-like growth factor 2 (IGF2) receptors, insulin-regulated aminopeptidase (IRAP), and the Ras GTPase-activating-like protein IQGAP1; ^{106, 148, 149-151}. Hence, recruitment of GluT4 to the cell surface will concurrently recruit these proteins (and vice versa), so that increased plasma membrane GluT4 might in some cases be a correlate, rather than a cause, of insulin’s cognitive effects.).

IRAP, in particular, colocalises with GluT4 in the hippocampus ¹⁰⁶ and is the receptor for angiotensin IV (AngIV), which enhances hippocampal memory ^{152, 153}. The presence of IRAP in GSVs may both promote stability of the GLUT4 protein and regulate compartmentalization and recycling of GLUT4 from endosomes to GSVs following translocation and subsequent endocytosis ¹⁵⁴. Interestingly, though, we have shown that glucose flux through GluT4 is required for cognitive enhancement by hippocampal AngIV ¹⁵⁵, suggesting that IRAP is not an alternative to GluT4 as the transducer of insulin’s cognitive effects.

GluT4 and glucose metabolism

In the periphery, the primary role of insulin is to remove glucose from the blood via GluT4, which moves to the cell surface to permit increased glucose entry into cells when needed. GluT4 is heavily expressed in the hippocampus ¹⁵⁶. Taken together with the facts that (i) hippocampal cognitive processes are limited by glucose supply^32-34, (ii) administration of glucose to the hippocampus causes an increase in hippocampal metabolism that correlates with improved memory ^{5, 7, 26, 27, 39, 157-159}, and (iii) intrahippocampal insulin acutely increases local hippocampal metabolism⁵⁵, an attractive hypothesis is that cognitive enhancement by insulin requires GluT4 translocation as a mediator of on-demand glucose supply during memory processing ^{156, 160, 161}. Researchers have speculated that because GluT4 exists in neuronal populations with high-energy demands, such as the hippocampus, it is likely that GluT4 supports GluT3, the principal neuronal glucose transporter, in meeting the demand for glucose supply to neurons during times of enhanced energy demand ^161-163. This role of GluT4 is likely to be specific to neurons: in the hippocampus GluT4 appears to be abundant wherever there are insulin receptors, but the converse is not true ^{156, 164-166}. Insulin receptors have been identified in many cell types (e.g. neurons, astrocytes, endothelia, microglia), whereas GluT4 shows neuron-specific localization ^{156, 167-169}.

Several studies, from our lab and others, support this ‘on-demand energy supply’ hypothesis for hippocampal GluT4. In the hippocampus, insulin enhances GluT4 translocation in the brain in a time-course and kinase-dependent (i.e. PI3K) manner that is very consistent with insulin’s effects on peripheral GluT4 ¹⁷⁰. Physiologically-relevant increases in intrahippocampal insulin that enhance memory upregulate GluT4 translocation, and intrahippocampal Aβ at a dose that impairs cognition and hippocampal metabolism reduces GluT4 translocation ^{5, 7}. More recently, we were able to show (using indinavir, see below) that GluT4 is a key player in hippocampal memory processes and specifically the impact of insulin. Increased glucose metabolism via GluT4 was required for enhancement of spatial memory by exogenous insulin; moreover, cognitive challenge using either of two hippocampally-mediated tasks in the absence of any treatment increased hippocampal GluT4 translocation after spatial working memory or fear-encoding (but not retrieval, interestingly; it appears that the role of GluT4 in hippocampal memory processes may be specific to encoding) ^{29, 171}. Consistent with the ‘on-demand’ hypothesis, blockade of hippocampal GluT4 did not affect hippocampal metabolism at baseline, but prevented any increase in metabolism by insulin treatment²⁹. Moreover, the role of GluT4 in memory appears to be at least somewhat region-specific: no effect of indinavir was seen on an amygdala-dependent task, in contrast to the marked effect on a very similar hippocampally-dependent task ¹⁷¹. These findings strongly support a key role for increased glucose metabolism, via GluT4, at times of increased glucose demand during hippocampally-mediated memory encoding.

In our 2016 paper ¹⁷¹, we noted as a potential inconsistency in this hypothesis that “Others have used high-resolution [¹⁴C]-2DG imaging to show that increased glucose utilization during upregulated neuronal activity occurs primarily in synaptic areas (the neuropil) rather than perikarya ^172-174. Because GluT4 is expressed in perikarya rather than neuropil^{163, 175}, which shows limited 2DG phosphorylation during hippocampally-dependent memory acquisition, increased neuronal glucose utilization seen during SA testing may not have been mediated entirely, or even primarily, through GluT4.” One possibility to resolve this apparent conflict was that insulin might undergo relatively bulk release in the hippocampus ¹⁷⁶, and hence could conceiveably act simultaneously on both synapses to control glutamatergic neurotransmission and the cell body to enhance GluT4 translocation and glucose utilization. The apparent dissociation between the location of plasma membrane GluT4 and the site of energy demand was, though, potentially resolved in a seminal subsequent study which confirmed our finding of activity-dependent neuronal GluT4 translocation and showed that this occurs in large part at the synapse, rather than the cell body ¹⁰⁷. This result expanded our understanding of the role of hippocampal GluT4, as potentially including both an insulin-dependent role (primarily at the neuronal cell body) and also a role in providing on-demand glycolytic support at the synapse (which is the location of the majority of activity-associated energy demand ^{45, 177, 178} and where glycolytic enzymes localize when energy demand exceeds supply¹⁷⁹). Ashrafi et al.¹⁰⁷ suggested that activation of AMP-kinase might mediate the synaptic recruitment of GluT4 during hippocampal cognitive processing, consistent with the discussion, above, of GluT4 being a target of multiple cognitive signalling pathways.

While acute blockade of GluT4 impaired memory encoding, chronic direct inhibition of hippocampal GluT4 impaired long-term memory, accompanied by e.g. reduced hippocampal BDNF, and alterations in task-associated hippocampal metabolism, but spared short-term memory ¹⁷¹. Further investigation of this dissociation is needed.

GluT4 and hippocampal insulin resistance

Peripheral vs. central regulation of GluT4

Outside the CNS, GluT4 translocation mediates the anabolic functions of insulin by increasing conversion of glucose into glycogen, protein, and fatty acids ^180-183. Some of these effects are mechanistically impossible in the hippocampus, though; for instance, GluT4 is expressed in neurons ¹⁸⁴, and glycogen conversely made only in astrocytes ^{185, 186}, and thus insulin is unlikely to effect glucose conversion into glycogen through GluT4-mediated glucose uptake in the brain. In general, brain insulin signalling and regulation of GluT4 may be different from that in peripheral tissues. For example, brain IRs are structurally distinct from peripheral IRs ¹⁸⁷. The neuronal IR is largely the -α isoform (IRα), which differs from the dominant -β (IRβ) isoform present in astrocytes, muscle, fat and liver due to alternative splicing of exon 11 ^188-191. Insulin and IGF2 bind IRα with high affinity ^{189, 192}, although it is unknown whether IGF2-activation of IR promotes GluT4 translocation in the hippocampus. IRα is linked to mitogenic actions of insulin, and is somewhat less effective at increasing glucose metabolism than IRβ ¹⁸⁹: in general, insulin’s effects in the hippocampus may not precisely match those seen elsewhere in the body.

As mentioned above, IR and GluT4 may have distinct spatial expression patterns within neurons, and these may alter at times of increased neuronal activity. Moreover, although rodent studies have identified GluT4 as primarily or exclusively neuronal, there is some evidence in the human brain that GluT4 may also be found on non-neuronal cells, such as microglia and endothelial cells ¹⁹³, so that insulin’s effects in human brain may be more diverse than those observed in rodent studies. Several molecules and events necessary for hippocampal memory (e.g. CaMKII, DHA, PKC, PKA, BDNF, cellular depolarization, etc.) regulate GluT4 in adipocytes and muscle ^91-106. It will be important to determine which of these effects also occur in neurons.

Impaired hippocampal GluT4 translocation in insulin resistance

In peripheral tissues, deficits in GluT4 trafficking and intrinsic activity are linked to the adverse effects of insulin resistance ^{101, 183, 194-199}. Because insulin resistance impairs hippocampal cognitive and metabolic processes ^{19, 134, 176, 200-207}, identifying the molecular consequences of brain insulin resistance is important DIO animals show markedly reduced cognitive responsiveness to intrahippocampal insulin, accompanied by an absence of the increased local glucose metabolism seen in response to insulin in control animals ⁵⁵. Perhaps surprisingly, although factors linked to T2DM such as elevated corticosterone impair hippocampal GluT4 translocation ¹⁶⁸, there appear to be no data reported on hippocampal GluT4 in an animal T2DM model. As would be predicted, we found translocation of GluT4 to be impaired (Figure 1): after 12 weeks of a high-fat diet the proportion of hippocampal GluT4 found at the plasma membrane was less than half of that of control animals, consistent with impaired GluT4 translocation being a key factor in cognitive and metabolic impairments seen with hippocampal insulin resistance.

Figure 1.

DIO rats have reduced hippocampal plasma membrane (active) GluT4 compared to rats fed a control chow diet Y-axis shows the ratio of plasma membrane to total GluT4 in the hippocampus, normalised to the group mean for Control animals set at 100%. Data are mean + SEM. * = p<.05.

Interestingly, a second glucose transporter, GluT8, may be regulated by insulin^{208, 209} (although reports differ ²¹⁰) and has been identified in the hippocampus, its primary location ^{184, 211}. The impact of insulin resistance on hippocampal GluT8 is unknown.

Hippocampal beta-amyloid and GluT4.

T2DM and AD are linked mechanistically by insulin resistance: patients with AD have reduced brain insulin signalling ^{54, 74, 100}, accompanied by hypometabolism, and both insulin and insulin sensitizers have proven effective at ameliorating AD ^{60, 74, 212, 213}. Conversely, T2DM is a major risk factor for AD, and leads to both elevated brain Aβ and impaired clearance of Aβ ^{54, 74, 110-117}. Abnormal accumulation of brain Aβ has been reported in transgenic rodent models, and we see elevated hippocampal Aβ in our DIO rat model also (unpublished data). In vivo, after as little as 10 min, oligomeric Aβ causes impaired hippocampal insulin signaling and reduced GluT4 translocation, accompanied by cognitive impairment and hippocampal hypometabolism but no effect on either GluT1 or GluT3 ¹²⁹. Moreover, cognitive impairment seen in the DIO rat model of diet-induced insulin resistance is completely reversed by blockade of intrahippocampal oligomeric Aβ ²¹⁴, along with restoration to baseline of dysregulated hippocampal glutamate signalling that is a hallmark of amyloid’s hippocampal effects ^{215, 216}, showing that Aβ is a key downstream effector of the cognitive impairment caused by insulin resistance. Further, GluT4 immunoreactivity colocalises with cholinergic markers ²¹⁷, suggesting a role for GluT4 dysregulation in AD which is characterised by damage to cholinergic neurons ^{218, 219}. Taken together, these findings strongly support the potential for (dys)regulation of GluT4 to be central to the pathological links between T2DM and AD. One possible caveat to this conclusion is offered by Nelson et al.²²⁰, who suggest that a direct T2DM-AD link is not certain and that cardiovascular pathology may be a better explanation for the link between T2DM and impaired cognition during aging.

Indinavir, HIV, and GluT4-related cognitive impairment

The GluT4 blocker indinavir was developed as an inhibitor of the HIV retroviral protease, and has no effect on mammalian proteases in either rats or humans ^{221, 222}. Other than inhibition of HIV protease, the primary action of indinavir is to selectively inhibit GluT4 via competition at the cytoplasmic domain without affecting translocation of GSVs ^{223, 224}, causing rapid and reversible peripheral insulin resistance ^225-228. This dual activity offers a potential human model in which to assess the specific impact of GluT4 blockade: patients treated with GluT4-impairing protease inhibitors (PI; indinavir, nelfinavir) vs. those treated with the newer atazanavir, which has a low affinity for GluT4 ²²⁷.

A common consequence of HIV infection is cognitive impairment including brain pathology, commonly referred to as HIV-associated neurocognitive disorder (HAND) ²²⁹; the prevalence of HAND has not decreased concomitantly with e.g. decreased viral load ^230-235 and is roughly 30-50% among HIV-infected persons ²³⁶. Long-term PI use has been linked to HAND, with symptoms including hippocampal cognitive impairment and Aβ accumulation, resembling central insulin resistance ²³⁷. HIV impairs blood-brain barrier integrity, so that systemic drugs such as PI may have increased access to the brain ²³⁸; note that this may also conceivably alter transport of insulin and/or glucose from the periphery. This suggests the possibility that impaired hippocampal GluT4 activity may be a cause of HAND. As an initial test of this hypothesis, we analysed data from approximately 1100 patients in the CNS HIV Anti-Retroviral Therapy Effects Research (CHARTER) dataset²³⁶. Consistent with this suggestion, we found that HIV patients with a history of GluT4-targeting PI use had impaired hippocampally-dependent recall memory relative to those who never took GluT4-targeting PI (Figure 2). This result is further support for GluT4 as a central regulator of hippocampal memory processes, and for impairment of glucose flux through GluT4 as a causal factor in cognitive impairments associated with central insulin resistance.

Figure 2.

CHARTER data showing that HIV patients on GluT4-inhibiting PI such as indinavir have impaired memory recall compared to patients taking PI that do not affect GluT4-mediated glucose uptake. Y-axis is demographically adjusted domain T-score for the recall domain, as defined in the CHARTER study²³⁶. Data are mean + SEM. * = p< .05.

Conclusion

Central, and more specifically hippocampal, insulin resistance appears to be a central element in the cognitive impairment seen in both T2DM and AD. Impairment of glucose supply to neurons subsequent to dysregulation of the insulin-sensitive glucose transporter GluT4 may be a unifying mechanism that explains, at least in part, the comorbidity of these two diseases. Additional disease states, such as HAND, where central GluT4 is impaired are also associated with cognitive impairment on hippocampal tasks. GluT4 is thus an attractive target for future therapeutic intervention: indeed, several treatments that increase the activity of GluT4 such as alpha lipoic acid, AICAR, insulin sensitizing drugs, and histone deacetylase inhibitors also increase memory ^239-242. Moreover, GluT4 in the hippocampus of healthy subjects is a key component of memory processing and regulation, likely transducing on-demand glucose supply to neurons when needed to meet the metabolic needs of increased neuronal activity.