Molecular mechanisms linking type 2 diabetes mellitus and late-onset Alzheimer's disease: A systematic review and qualitative meta-analysis

is to provide a comprehensive review of common mechanisms, which have hitherto been discussed in separate perspectives, and to assemble and evaluate candidate loci and epigenetic modifications contributing to polygenic risk linkages between T2DM and LOAD. For the systematic review on pathophysiological mechanisms, both human and animal studies up to December 2023 are included. For the qualitative meta-analysis of genomic bases, human association studies were examined; for epigenetic mechanisms, data from human studies and animal models were accepted. Papers describing pathophysiological studies were identified in databases, and further literature gathered from cited work. For genomic and epigenomic studies, literature mining was conducted by formalised search codes using Boolean operators in search engines, and augmented by GeneRif citations in Entrez Gene, and other sources (WikiGenes, etc.). For the systematic review of patho-physiological mechanisms, 923 publications were evaluated

metabolic dysregulation in T2DM and LOAD linkage.The results allow for more streamlined longitudinal studies of T2DM-LOAD risk linkages.

Epidemiological evidence and risk factors
The earliest descriptions of alterations of blood glucose and insulin metabolism in late-onset Alzheimer's disease (LOAD) patients date back to the 1980s (Bucht et al., 1983;Reubi and Palacios, 1986a).However, conclusive evidence supporting the notion that Type 2 Diabetes (T2DM) could be a contributory factor to LOAD was not available until 1 5 y ago (Mattson, 2004).Both diseases, T2DM (Khera et al., 2018), and LOAD (Escott-Price et al., 2017;Leonenko et al., 2019) are polygenic according to genome-wide studies, thus suggesting the importance of environmental risk factors are of key importance in disease onset and progression.
Given the increased prevalence of T2DM over recent decades, resolving the issue of a cerebrovascular vs neurodegenerative causation of cognitive impairments in LOAD has been regarded as essential (Luchsinger, 2012).Possible mechanisms discussed here are cerebral infarcts, white matter (WM) disintegrity, hyperinsulinaemia, insulin insensitivity, and advanced glycation end products (AGEs).An unresearched area remains the delineation of mechanisms for diabetic encephalopathy, including brain atrophy, which are, like the peripheral and central diabetic neuropathies, not yet sufficiently explained (Shinohara and Sato, 2017).However, progress in scientific understanding exist for diabetic retinopathy, with implications diabetic mechanisms possibly involved in LOAD.

Disease burden
The IDF estimated in 2019 that 463 million people have diabetes worldwide and that by 2030 this number is predicted to rise to 578 million and to 700 million by 204 5.According to 8th Diabetes Atlas (Federation, 2017), the prevalence rates for the over 60s was 12%-1 5% in 2017, and T2DM constitutes 90%-9 5% of all diabetes cases (Kähm et al., 2018;Koekkoek et al., 2015).T2DM is one of the most costintensive diseases globally (Scarlett et al., 2016): According to recent forecasts, the global economic burden caused by diabetes and its consequences will rise from US$1.3 trillion in 2015 to US$2.5 trillion if current trends continue (Bommer et al., 2018).The U.S. increase in mortality was age-adjusted +39% for 2000-2010 (Association, 2019), and further +33% are expected for 2020-202 5 (www.statista.com/statistics/452934/).

Epidemiological evidence for linkage of risks
Presence of (peripheral) insulin resistance as a risk factor for LOAD was first reported in the Rotterdam ageing study [1999] with RRs = 1.9-4.3(Lemche, 2017).In addition, 81% of autopsied LOAD cases were found to display elevated levels of fasting glucose (>110 mg/dl), or to have T2DM (>126 mg/dl) (Janson et al., 2004), whereas T2DM is typically present in only 30% of otherwise healthy elderly populations (Biessels et al., 2006).Elevated fasting glucose levels increased LOAD risk in nondiabetic elders at HR = 1.19 per 0.9 nmol/l difference (Crane et al., 2013).Other epidemiological studies also revealed that T2DM is associated with an 1.5-2.5-foldincreased risk of dementia (Strachan et al., 2011).
Fasting glucose levels in younger adults, as well as glycated haemoglobin, were determined as independent significant predictors for LOAD (Li et al., 2017), at HRs = 1.27 and 1.32 for their variability coefficients, respectively.Diabetes-induced cognitive deficits were identified as present in T2DM, possibly already present from adolescence, in the magnitude of 0.3-0.5 SD, but uncorrelated with age, and typically preceding amnestic MCI (Koekkoek et al., 2015).
During a 10-yr New York prospective study, the relationship between T2DM and LOAD was estimated at HR 1.6 (Cheng et al., 2011), but this figure was increased with vascular dementia (HR = 3.0 with mixed cases).Similar figures (RRs 1.67-1.77)were reported for prediabetes and LOAD (Xu et al., 2007).An early systematic review on the relationship between diabetes and Alzheimer (Biessels et al., 2006) concluded that the evidence base supporting a relationship between T2DM and LOAD had increased over time: Overall, the increase of T2DM risk for LOAD is at 50-100%, and for VD 100-150%, with ORs ranging 1.0-4.4(Biessels et al., 2006;Gudala et al., 2013).
When combining MetS-related risk factors, based on their world-wide prevalence in LOAD cases, the combined prevalence rates were 36.4%,compared to a general population risks of 22.7% (Norton et al., 2014).Sleep loss is associated with MetS and T2DM, due to alterations in glucose and insulin metabolism, and causes perturbations in neuroendocrine systems including the HPA axis (Schmid et al., 2015).HPA dysregulation underlying MetS (Lemche et al., 2016) is biased towards brain alterations.The few available studies on MetS in children and adolescents suggest (Yates et al., 2012) that even in these early life stages, brain alterations are present including (a) smaller hippocampi, (b) frontal lobe atrophy, and (c) alterations in the HPA axis.It was concluded that brain insulin resistance reduces capillary reactivity and causes impairments in endothelial-dependent vasodilatation resulting in increased BBB permeability at neurovascular units enabling glucose entry into brain parenchyma under inflammatory conditions (Section 5) (Yates et al., 2012).

Obesity and APOE ε4 allele interaction
Being overweight in midlife as a characteristic of MetS is related to LOAD at RR = 1.3 5-2.04 (Anstey et al., 2011), according to pooled meta-analytic evidence, when adjusted for T2DM and APOE ε4 status.
As our rationale for this systematic review, we considered mainly laboratory biomarker studies for T2DM and LOAD and to find mechanisms that were described for both of the conditions.The main objective herein is to identify key players of in-parallel switches and to extract their genomic loci.The qualitative meta-analysis following aims at testing for disease linkages and epigenetic modifications in each gene locus extracted.The detailed method and genomic result tabulation can be found as online supplemental files linked to the main text.

Histological, neuroimaging and biomarker evidence
Autopsy studies have been postulated to to be the gold-standard for the related pathogenesis of T2DM and LOAD.However, studies specifically addressing this issue have only become available in the past decade.The basic question herein is: which alteration is related to diabetic encephalopathy, and which are sequelae to LOAD-induced neurodegeneration.This issue has had few discussion in the existing literature to date.
Three major problems havehave been listed with the amyloid cascade hypothesis (Yamashima, 2013): (a) amyloidogenic processing in animal models has consistently failed to produce neuronal loss and brain atrophy as seen in human LOAD cases; (b) normal ageing may be accompanied by up to 30% amyloid deposition without clinical signs of dementia; (c) pharmacological removal of amyloid has not shown neuronal or cognitive recovery, and conversely, degree of amyloid burden hasis not been the best correlate of cognitive decline.These arguments may support T2DM as the major underlying pathology driving brain atrophy in LOAD.

Whole brain atrophy in T2DM
Evidence reviewed by La Ferla and colleagues (Baglietto-Vargas et al., 2016) suggests alterations in brain structure and function at various levels, in individuals with T2DM: (a) accelerated global cognitive decline due to whole-brain atrophy, (b) incident vascular impairment, (c) general neuron loss and pancreatic β-cell loss, (d) downregulation of O-GlcNAcylation, thus potentiating τ hyperphosphorylation, (e) abnormal fatty-acid metabolism and oxidative pathways.Aside from general structural losses (Roberts et al., 2014), volume reductions have been registered in the hippocampus and entorhinal cortex (Devanand et al., 2012;Moran et al., 2013), the DLPFC, medial frontal, and anterior cingulate (Moran et al., 2013), temporal and limbic cortices (Devanand et al., 2012), together with reductions in PCC and MTG connectivity (Chen et al., 2014) in T2DM.
In T2DM, large-scale alterations were found related to the insulin system: Intact brain myelination and larger brain size are dependent of the maintenance of IGF-1 signalling, whereas deficient IGF-1 functioning probably accounts for the whole-brain atrophy rate of 40% associated with T2DM (Messier and Teutenberg, 2005).In particular, ventricular enlargement in T2DM accounted for about 7%-20% of WM loss (Biessels et al., 2014).As WM fibre structures are believed to contribute to around 20% of IQ (Gordon et al., 2018;Ngandu et al., 2015), structural loss doubtless confers cognitive decline.In autopsy studies (but these are not whole-brain results), microinfarcts were found to be increased in T2DM (Ahtiluoto et al., 2010;Nelson et al., 2009) in the oldest age range (>8 5 yrs), however, Aβ and neurofibrillary tangle (NFT) burden decreased.These findings had thus been attributed to diabetic encephalopathy (Nelson et al., 2009).However, these studies had no in-vivo measurement of diabetes, but only post-hoc analyses.

Neuropathological impacts in T2DM
Recent evidence on brain atrophy in T2DM during ageing (Callisaya et al., 2018) suggests that reductions in total brain volume and ventricular enlargement begins early in midlife, resulting in reductions in verbal memory and fluency.Microvascular lesions are typical for LOAD: neurovascular dysfunction causes WM lesions (indicated also by T 2hyperintense MRI, WMHs) and cognitive impairment.Further, serum tau NFT correlates with these WM lesions.
In a number of studies reviewed by Pruzin, neither neuropathological-histological counts of Aβ neuritic plaques nor NFTs showed significant correlations with T2DM (Pruzin et al., 2018), but this lack could be explained by statistical weaknesses.However, most studies do support a relationship between (peripheral) insulin resistance (as assessed by HOMA1/2-IR) and increased levels of hyperphosphorylated τ in the CSF and decreased regional cerebral glucose uptake.Abnormally phosphorylated insulin receptor substrate IRS-1 (IRS1 2q36.1) may serve as a biomarker of cerebral insulin resistance (Pruzin et al., 2018), and hyperphosphorylated IRS-1 correlated positively with Aβ plaques, and negatively with cognition, irrespective of the presence or absence of T2DM status (Talbot et al., 2012).The negative histological findings require further examination.
Normal ageing-related decreases in hepatic IGF-1 production have an early onset in the third decade of life (Piriz et al., 2011), followed by a reduction in cerebral IGF-1 functioning in the brain, despite greater insulin-like growth factor-1 receptor (IGF-1R 1 5q26.3)expression, accompanied by decreased hippocampal levels of the vesicular glutamate transporter 1 (VGlut1, SLC17A7 19q13.33) in glutamatergic synapses (Muller et al., 2012).Abnormal acceleration of brain insulin deficits were found in early post-mortem LOAD histochemical investigations the form of reduced insulin and c-peptide levels (Frölich et al., 1998).Another post-mortem histochemical study demonstrated increased insulin receptor density and stronger staining of insulin itself in LOAD brains with neuropathological lesions.Reduced levels of IGF1R and more IGF-1-binding protein-2 (IGF1BP2 2q35) were found in the temporal cortices of LOAD sufferers (Moloney et al., 2010;Quesnel et al., 2023).This study reported increased IGF-1R embedded into and attached to amyloid plaques and decreased levels of the key insulin substrate adaptor proteins IRS-1 and IRS-2 (IRS2 13q34) in plaque affected regions (Moloney et al., 2010).

Conversion biomarkers and neuroimaging
The prediabetic stage is characterised by biomarkers, of which impaired fasting glucose, glycated haemoglobin, and (peripheral) insulin resistance have been found predictive for dementia risk (Biessels E. Lemche et al. et al., 2014).Animal evidence also indicated that BDNF depletion in the brain causes an insulin resistant state, which reduced adult neurogenesis in the dentate gyrus proliferative zone (Pugazhenthi et al., 2017).Furthermore, malondialdehyde (MDA), a product of lipid peroxidation of polyunsaturated fatty acids (PUFAs) and a biomarker of oxidative stress, has been found to be predictive for MCI-LOAD conversion in concord with plasma insulin levels (Monacelli et al., 2015).
Increasingly is CBF hypoperfusion recognised as an early sign of MCI and LOAD (Sweeney et al., 2018): hypoperfusion by about 40% was found present in the PCC and praecuneus, parietotemporal, frontal and occipital cortices, hippocampus, parahippocampal and entorhinal gyri.Likewise, T2DM patients showed greater cortical gray matter (GM) atrophy, and subcortical GM hypoperfusion, as compared to controls.However, as yet it is unclear whether insulin-related cortical hypoperfusion or vascular restrictions are responsible for cognitive impairments in T2DM (Jansen et al., 2016).
Leukoaraiosis, believed to be based on gliosis and demyelination, are periventricular WM hypodense lesions usually found in brain scans of subjects over 80 years-old (Roriz-Filho et al., 2009).Myelin disintegrity and leukoaraiosis are also known to be significantly accelerated in T2DM (Habes et al., 2016a;Habes et al., 2016b;Maldjian et al., 2013), but different to lacunar infarcts.The T 2 -hyperintensity and reduced proton density are limited to WM, and exhibit no changes in GM, ventricles and gyrification.WM integrity was shown to be dependent on the intact Notch neuroproliferative signalling pathway (Madden et al., 2009), where the receptor complex NOTCH1-SOX9-SOX2 provides a positive-feedback loop via its ligand surface protein Jagged1 (CD29, JAG1 20p12.2).Notch1 (NOTCH1 9q34.3)signalling is also critically involved in LOAD structural abnormalities (Lemche, 2017).
In addition, BBB leakage corresponds to WM T 2 -hyperintensities prior to LOAD conversion; BBB integrity is compromised in T2DM, where microbleeds associate with hypertension and other MetS components (Biessels et al., 2014).Large-scale multimodal neuroimaging combining plasma and CSF biomarkers revealed that WMH and vascular dysregulations are predictive of LOAD conversion (Medina et al., 2016): Preceding endothelial dysfunction, vascular dysregulation and vascular stiffness impair astrocyte endfeet and oligodendrocytes even prior to Aβ deposition.

Brain insulin resistance concepts, and insulin/IGF signalling
An analogy of T2DM and LOAD metabolisms was first postulated by Hoyer (Hoyer, 2002), as both exhibit decline in insulin levels, insulin binding, and tyrosine kinase activity.In LOAD brains, glucose uptake and functioning of insulin degrading enzyme or insulysin (IDE 10q23.33)(Section 2.2.) were observed to be abnormally reduced (Hoyer, 2002), alongside diminished pyruvate dehydrogenase complex (PDHc, below) yielding reduced levels of acetyl-coenzyme A (CoA), and adenosine triphosphate (ATP).Furthermore, insulysin and ATP levels are not reduced in EOAD brains (Hoyer, 2002), in contrast, thus indicating their pathophysiological specificity for LOAD metabolism.

Peripheral metabolic condition and LOAD
Demetrius' postulation of LOAD as a metabolic condition (Demetrius and Driver, 2013;Demetrius et al., 2015) was proposed to augment the amyloid cascade hypothesis (Hardy and Higgins, 1992).This hypothesis has generated assumptions of a bioenergetic explanation for hypometabolic brain circuits (Gibas, 2017): Progressive peripheral insulin resistance leads to cerebral hyperinsulinaemia with lowering bioavailability of glucose crossing the BBB (Willette et al., 2015).In this state, starvation promotes reprogramming of the neural phenotype towards Fig. 1.Flowchart diagramme according to PRISMA 2020: http://www.prisma-statement.org/.
Brain insulin resistance has recently (Arnold et al., 2018) been defined as the inability of brain cells, either neurones or glia cells, to respond to insulin action due to deficient insulin receptor signalling.Brain insulin resistance is further specifically characterised by reduced and altered expression of insulin/IGF polypeptides (de la Monte, 2014).Also implicated in impaired brain insulin metabolism are (a) reduced neuronal glucose uptake, (b) reduced expression of GLUT4 (Williamson et al., 2023), (c) restricted homeostatic or inflammatory responses to insulin, (d) then impairing neuroplasticity and cognition (Arnold et al., 2018).

Results from animal models
Various rodent studies demonstrated that (a) τ protein was hyperphosphorylated in T2DM models, indicating that T2DM may be a precursor to LOAD; (b) brain insulin deficiency in T2DM rats correlated with their peripheral insulin resistance; (c) impairment of insulin/PI3K/ Akt signalling occurred in both T2DM and AD models (Yang et al., 2016).In high-fat diet (HFD) fed diabetic animals, widely used for hyperlipidaemia and hyperglycaemia (Li et al., 2020), BBB-infiltrating polarised macrophages induced brain insulin resistance (Jais et al., 2016), similar to that seen in adipose tissue (Olefsky and Glass, 2010), through nuclear factor κB (NFκB).High fat/high sugar (HSD) diet in mice induced a stress-type reaction in the brain (Kothari et al., 2017), indicated by pro-inflammatory pro-apoptotic (NFκB, JNK) and cellular stress signals (p38 MAPK, damage-inducible transcript 3 DDIT3 12q13.3).Intrahippocampal insulin injection in rats induced a neuroprotective effect (Ghasemi et al., 2014) in Aβ treated animals reversing the pro-apoptotic effect through caspase-3 activation via p38/MAPK, ERK, and JNK signalling pathways.

Insulin receptors, insulin receptor substrates, and insulin degrading enzyme (insulysin)
Insulin appears to impact on cognition at multiple levels including (a) fast and tonic neuromodulation, (b) increased adult neurogenesis, (c) enhanced neuroplasticity, (d) improved synaptic density, and (e) accelerated synaptic transmission (Fernandez and Torres-Aleman, 2012).Insulin receptor (IR) signalling is recognised as essential for dendritic outgrowth, neuronal survival, circuit development, and postsynaptic neurotransmitter receptor interaction (Derakhshan and Toth, 2013).There is also evidence that insulin signalling has links with (a) cholesterol metabolism, (b) acetylcholine production, (c) cerebral noradrenaline uptake, and (d) expression of NMDA receptors at synaptic sites (Williamson et al., 2012).A direct link of insulin action upon cognitive performance however, has not yet been demonstrated (Williamson et al., 2012) (but see Talbot et al., 2012 for correlations of hippocampal IRS-1 with working memory), whilst brain ageing generally is characterised by overall reduced insulin transport across the blood-brain barrier (Cholerton et al., 2011).

Impairment of insulin receptor functioning as cause of insulin resistance
Strong parallels exists between T2DM and LOAD in the blockade of IR functioning (Morgen and Frolich, 2015).It is currently believed that insulin crosses the BCSFB, and that cerebral ILPs are secreted through the choroid plexus into the CSF: Insulin transport into CSF follows a saturable mechanism, where the two receptor classes IGF-1R and IGF-2R are regulated by nutrition states (Begg, 2015), possibly subserved by chemosensing in the area postrema (AP).Epithelial cells in the choroid plexus and endothelial cells in brain vessels express the highest levels of both IRs (Fernandez and Torres-Aleman, 2012).Overall, it was found that the insulin CSF/serum ratio was significantly associated with peripheral insulin sensitivity, with reduced insulin transported into the CSF in insulin-resistant subjects (Heni et al., 2014).
Binding of insulin to IRs activates the intrinsic tyrosine kinase activity of the cytoplasmic domain of the IR (Lee and Pilch, 1994).This leads to autophosphorylation of tyrosine residues, triggering intracellular signalling cascades, and modifying release and reuptake of neurotransmitters.Downstream targets of cerebral insulin binding differ from the periphery and consist mainly of neuronal glutamate receptors (Roriz-Filho et al., 2009).In visceral adipose tissue, miR-15 5 derived from exosomes secreted by macrophages, targets PPARG, thus being inducive for insulin resistance and hyperglycaemia (Ying et al., 2017).In the CNS, Aβ oligomers tend to dock on IRs and push them from the outer membrane into the cytosol, impairing glutamatergic signalling.In T2DM, both CNS and peripheral IRs inactivation occurs through c-Jun N-terminal kinases (JNK) (Morgen and Frolich, 2015).

Insulin receptor malfunction and triglycerides
Cerebral glucose intolerance (Section 2.3.), termed "T3DM" is a resistance at the IR level (Banks et al., 2012;Mittal and Katare, 2016) induced by increased triglyceride levels.Only recently, it was established that triglycerides transgress the BBB and are active in CSF (Banks et al., 2018).Dyslipidaemia (triglycerides) related to obesity was observed to be directly linked to LOAD hippocampal atrophy, possibly based on deficient leptin signalling.Cerebral triglycerides blocked both IRs and leptin receptors (LEPRs) in the hypothalamus (Banks et al., 2018), and were inversely correlated with cognitive performance.
Previously, it had been shown that triglycerides can prevent leptin from crossing the BBB (Banks et al., 2004).Recently, it was shown that triglyceride droplets can also impede microglial functioning, a finding, however, specific to APOE ε homozygotes (Haney et al., 2023).Accumulated evidence suggests that aside from the ILP-system, effective glucose-regulation is possible also by means of leptin and LEPRs, presumably under the control of sympathetic and parasympathetic outflows (Fujikawa and Coppari, 2015).Recent electrophysiological evidence in rodents however, suggests that this effect is mediated by leptin levels and regulated by coherence of local field potentials (LFPs) in the lower β-band by the nucleus accumbens (NAcc) and ventral tegmental area (VTA) (Maurer et al., 2017).
A reverse mechanism has also been demonstrated (Marciniak et al., 2017), showing that deficient or missing τ specifically in hippocampal neurones induced an impaired hypothalamic insulin response via IRS-1 and the PTEN enzyme (PTEN 10q23.31).It therefore becomes possible to conceive that misfolded or otherwise pathophysiological τ becomes a cause of cerebral insulin resistance (Marciniak et al., 2017).In addition, GWAS found that τ (MAPT) haplotypes (i.e.combinations with other genotypes) are associated with hyperglycaemic traits in humans.

Role of insulysin (insulin degrading enzyme)
Diffusible oligomeric forms of Aβ impair synaptic functioning by blocking IRs, leading to a 70-fold increase in Aβ accumulation (Gong et al., 2003): The metalloprotease insulysin is able to degrade not only oligomeric Aβ, but also other proteins with a propensity to form amyloid fibrils, including glucagon, amylin, or calcitonin (Kurochkin, 2001).
Analysis of shared pathways (Hao et al., 2015;Karki et al., 2017) between T2DM and LOAD by meta-analysis of multiple GWAS studies in each domain yielded significant calls (in genomic analyses) for insulin, neurotrophin, PI3K/Akt, mTOR and MAPK signalling, gap junction channels, phagocytosis, calcium signalling (abnormal calcium central to LOAD (Mattson, 2004)), long-term potentiation, Wnt and chemokine signalling, and microglial mediated immune responses, amongst other Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.

Cerebral glucose intolerance and hypometabolism
Perturbations in peripheral glucose levels accompanied by hyperinsulinaemia in LOAD were observed in 1983 (Bucht et al., 1983), and were then interpreted as a consequence of the neuroendocrine imbalance of growth hormone, somatostatin and thyroid-stimulating hormone (Reubi and Palacios, 1986b).Reduction of glucose uptake is now seen as an early invariant biomarker of LOAD, and assumed being the cause of cognitive dysfunction by neuron loss: Reduced brain glucose levels are generally reported in LOAD brains (Lee et al., 2016), are accompanied by reductions in O-GlcNAcylation (Section 2.6.)(Liu et al., 2004), in turn increasing τ phosphorylation.

Cerebral glucose transport across BBB
Cerebral glucose transportation depends on the physiological function of astrocytes participating in the outer layer of vascular capillaries at the BBB, and on various glucose transporters distributed in the brain, particularly the essential GLUT1 and GLUT3 (Chen and Zhong, 2013).GLUT2, expressed in mainly pancreatic β-cells and liver, is in CNS known also as a glucose-sensing mechanism, and point mutations and transcriptomic signatures have been implicated in insulin resistance in T2DM and LOAD (Matone et al., 2017;Mueckler et al., 1994).Reduced glucose metabolism in LOAD is attributable to significant reductions in glucose transporters GLUT1 and GLUT2, which are responsible for neuronal glucose uptake, and to down-regulation of hypoxia-inducible factor 1α subunit (HIF1A 14q23.2), the regulator of GLUT1 and GLUT3, coinciding with decreased O-GlcNAcylation, hippocampal atrophy, hyper-phosphorylation of τ protein, and density of NFTs (Abbott, 2004;Liu et al., 2004).Glucose transporters GLUT1, GLUT4, and GLUT8 are themselves insulin sensitive (Woods et al., 2003), and regulated by IGF1 and IGF1R (Ostrowski et al., 2016;Verdile et al., 2015).GLUT4 is expressed in the cerebellum, sensorimotor cortex, hippocampus, pituitary and hypothalamus; GLUT8 has been observed in the hippocampus and hypothalamus, but not found responsible for glucose transport (Cholerton et al., 2011;Verdile et al., 2015).These five GLUTs were also identified as critical in LOAD core pathophysiology (Lemche, 2017).

Human histology
In the Baltimore ageing study (An et al., 2018), post-mortem dissection of LOAD tissue indicated higher cerebral glucose concentrations and ratios of the glycolytic amino acids, serine, glycine, and alanine to glucose decline of glycolytic flux in LOAD.Protein levels of the neuronal (GLUT3) and astrocytic (GLUT1) glucose transporters correlated with LOAD severity.Abnormalities in brain glucose homeostasis were associated with elevated fasting glucose levels, and preceded overt LOAD pathology onset.

Human neuroimaging
Glucose hypo-metabolism as an invariant biomarker in LOAD (Chen et al., 2013) has become evident from 18 F-FDG-PET neuroimaging, and because in this depletion it is different from all other dementias, LOAD has been characterised as "T3DM" (Steen et al., 2005) (Section 2.2.).Cerebral metabolism as measured by radiolabelled glucose and ketone body acetoacetate (hepatic glucose surrogate) by PET neuroimaging (Chiotis et al., 2018;Croteau et al., 2018) showed the most pronounced reductions (11-15%) in LOAD cortices and subcortical regions, as compared to MCI and healthy elders. 18F-FDG PET neuroimaging of glucose uptake in prediabetic subjects identified with HOMA-IR (Baker et al., 2011) revealed that greater insulin resistance was associated with LOAD-like patterning of glucose uptake in frontal, parieto-temporal, and cingulate regions in adults with prediabetes and T2DM (Baker et al., 2011).The occurrence of the LOAD pattern of reduced cerebral glucose uptake was independent of ApoE4 allelic status (Baker et al., 2011).

Animal research
Rodent research suggests that in T2DM (a) neurocircuits involved in glucose homeostasis are dysfunctional; that (b) this dysfunction may contribute to hyperglycaemia; and that (c) acidic fibroblast growth factor 1 (FGF1 5q31.3) is involved in cerebral glucoregulation.FGF1 showed association with hypothalamic content of synaptophysin (a synaptic marker protein) (SYP Xp11.23) involved in synaptogenesis (Scarlett et al., 2016).
Long-term HFD induced the aggregation of Aβ in the dentate gyrus of wild-type MetS-prone mice.This was accompanied by neuroinflammatory states decreasing neuronal progenitor cells and causing impairment of normal autophagy (Busquets et al., 2017).In the T2DM rat model with hyperglycaemia and hypoinsulinaemia (Wistar Bonn Kobi, WBN/Kob), radiolabelling investigation confirmed a HPA imbalance (Tojo et al., 1996) towards chronic sympathicotony, an ANS bias inducing a prediabetic state (Lemche et al., 2016).
Observations in ageing murine LOAD models suggest that perturbations in cerebral glucose precede neurotoxic Aβ 1-42 appearance (Chua et al., 2012): The IR-β subunit, IGF-1R, IRS-1, and IRS-2, accompanied E. Lemche et al. by reduced glucose and insulin content, were present in the brains of AβPPsw/PS1δE9 mice before Aβ accumulation, but not present in wildtype animals.This was accompanied by increased GLUT3 and GLUT4 levels and alterations in PI3K signalling (Chua et al., 2012).However, animal studies in several APOExAPP murine models suggest that the presence of the ApoE4 allele can induce insulin dysfunction (Chan et al., 2016), independently of Aβ 1-42 deposition, resulting in memory decline as a result of impaired AMPA-receptor functioning and attenuation of GluR1 activity (Chan et al., 2016).

Calpain-cathepsin hypothesis
The calpain-cathepsin hypothesis for LOAD was formulated in 1996 (Oikawa et al., 2009;Siklos et al., 2015;Yamashima, 2016), drawing upon parallelisms in apoptosis following ischaemic strokes and LOAD and explaining the strong comorbidity between TBI and neurodegeneration.Cathepsins are lysosomal cysteine proteases mediating terminal protein degradation, cell regulation, apoptosis, lipid metabolism and immune response (Siklos et al., 2015).Calpains are a class of intracellular proteolytic enzymes in the calcium/calmodulin metabolism, several of which are relevant to both T2DM and LOAD.Little is yet known about in-vivo regulatory mechanisms of calpains, however, the two enzymes co-operate in lysosomal rupture under oxidative stress.

Cathepsins and insulin resistance
In 198 5, it was discovered that cathepsins B and H were decreased in diabetic muscle tissue (Stauber and Fritz, 1985).In 1988 lysosomal cathepsin D was found in pharmacologically induced insulin resistance covarying with insulin signalling (Nerurkar et al., 1988).In two Uppsala community cohorts (Nowak et al., 2016), cathepsin D was found to be predictive for the transition between insulin resistance and T2DM, which was mediated by predictors IL-1 receptor antagonist (IL1RN 2q14.1)(HR = 1.28) and tissue plasminogen activator (PLAT 8p11.21)(HR = 1.30), that were associated with incident T2DM.

Glucagon
Glucagon (GCG 2q24.2) itself forms a complementary regulatory system by opposing insulin function; glucagon converts glycogen into glucose, which insulin action then enables to be taken up by glucosedependent tissues.The proprotein convertase 2 produces glucagon by cleavage of proglucagon in the α-cells of pancreatic Langerhans islets, inhibited by insulin production via GABA signals of neighbouring β-cells, and downregulated by amylin (Section 3.1.).In glycogenolysis, glucagon activates the glucagon receptor (GCGR 17q25.3),which ultimately releases glucose through cAMP action, catalysed by pyruvate kinase.Further to glycogenolysis, glucagon promotes gluconeogenesis and also lipolysis.Glucagon is inactivated by neprilysin (MME 3q25.2), which also degrades Aβ (Kanekiyo et al., 2013;Kim et al., 2023;Quinto et al., 2013).Inhibition of neprilysin is known to ameliorate glucagon receptor functioning (below), and therefore improves insulin sensitivity (Esser and Zraika, 2019).

GLP-1
Insulin secretion is potentiated by intestinal nuclear farnesoid X receptor (FXR 12q23.1) in the bile acids, which trigger GLP-1 production by enteroendocrine L-cells (Trabelsi et al., 2015) from pre-proglucagon expressed by GCG.GLP-1 also drives insulin secretion, which may be relevant to this hypothesis.Peripheral GLP-1 can cross the BBB, but appears binding preferentially in the subfornical organ and the area postrema (AP) (Orskov et al., 1996), both of which have projections into the hypothalamic areas involved in water homeostasis and appetite regulation.In the AP, GLP-1 acts directly on neurones, where c-Fos expression triggers ANS homeostasis functions such as regulation of emesis, blood pressure, and food and water intake (Kawatani et al., 2018;Zhang et al., 2020).In contrast to AP neurones, neurones in the nucleus tractus solitarius (NTS) did not exhibit the respective depolarisation bursts (Kawatani et al., 2018).

GLP-1 receptors in pancreas and subcortical areas
GLP-1Rs are present in the pancreas and in vascular endothelium, but also in neuron somata, dendrites, and astrocytes (Drucker, 2018).Aside from pancreatic β-cells, GLP-1 receptors were found in the brain mainly in the circumventricular organs, the amygdala, hippocampus, hypothalamic nuclei, ventrolateral medulla, and olfactory bulb (Calsolaro and Edison, 2015;Cork et al., 2015).Within the hypothalamus, GLP-1R have been located in ARC, PVN, supraoptic nucleus, and supraoptic decussation (Drucker, 2018) in rodents.Being highly expressed also in the gut-brain-axis, GLP-1Rs were found mediating antiinflammatory signals, although they are not found on immune cells (Drucker, 2018;During et al., 2003).Particularly in the hippocampus, overexpression of GLP1Rs was related to neuroprotection and enhancement of learning and memory (During et al., 2003).It has been shown recently (Hsu et al., 2018) that GLP-1Rs participate in an executive higher-order regulatory loop between hippocampal CA1 glutamatergic neurones and mPFC for food recognition, possibly participating in feeding behaviours.GLP-1Rs activate signal transduction through related pathways Raf, ERK1/2, PI3K/Akt, MAPK, but only in the presence of glucose.
Fast axonal transport (FAT) is dependent on insulin signalling and requires GLP-1R activation (Takach et al., 2015).Aβ oligomers impair insulin signalling, with the consequence of suppressing the protein kinase B/Akt responsible for BDNF transport in murine LOAD models (Takach et al., 2015).Stimulating the GLP-1R to activate the insulin pathway restores normal BDNF axonal transport.In the LOAD brain, GLP-1R activation appears to reduce chronic neuroinflammatory activation of microglia and prevents synapse loss (McClean and Hölscher, 2014).Studies in cultured cell lines (Chen et al., 2016a) have shown that activation of the GLP-1R reduced oxidative stress, the production of AGEs, and promoted upregulation of RAGEs through suppression of NFκB.The neuroprotective effect was evident by reduction of apoptosis factors caspase-3, − 9, Bax and beclin-2 (Chen et al., 2016a).

GLP-2 and GLP-2 receptors
Together with GLP-1, GLP-2 is generated and secreted in posttranslational proteolytic cleavage of proglucagon, produced by enteroendocrine L-cells upon nutrient ingestion and in CNS neurones.GLP-2Rs, alike GLP-2, are believed to serve intestinal growth and enteric apoptosis.All are in close relation with glucagon receptors (GCGR 17q2 5.3), which are expressed in most internal organs, the gastrointestinal tract, and in the cerebral cortex; a mutation of GCGR has been linked to T2DM (Hager et al., 1995), but most known mutations are related to cardiometabolic traits (Drucker, 2018).

Glycosylation and glycation including AGEs
O-GlcNAcylation is a common nutrient-responsive post-translational modification (PTM) by glycosylation (Wang et al., 2017b;Yang and Qian, 2017).It can only occur in the Golgi apparatus and ER (Wu et al., 2023).It is currently regarded as a highly dynamic process, called GlcNAc cycling, balanced by the counteracting enzymes protein O-GlcNAc transferase (OGT Xq13.1) and protein O-GlcNAcase (OGA; MGEA5 10q24.32 adjacent to IDE) (three isoforms: fOGA, sOGA, vOGA) (see below).O-GlcNAcylation herein protects its target proteins as a buffering coat against protein aggregation by modulating protein-protein interactions (Yang and Qian, 2017).This is accomplished by dynamic alternation between O-GlcNAcylation and phosphorylation on nuclear and cytosolic proteins: in (a) competition between OGT and OGA for the serine or threonine functional group of proteins; and by (b) occupancy, where O-GlcNAc and O-phosphatase reciprocally influence functioning of proteins (Hart et al., 2007).
Maintenance of O-GlcNAc homeostasis is a precondition for multiplex cellular functions, whereas an imbalance is documented for T2DM and LOAD.OGT is highly expressed in the pancreas, and further in brain, heart, muscles, and liver (Lubas et al., 1997).In the glucose-sensitive pathway in T2DM, OGT is effective as nutrient and stress sensor (Hart et al., 2007;Lubas et al., 1997).OGA is extensively expressed, in in brain, lymph nodes, and immune-system relevant organs such as spleen and thyroid.Both enzymes involved in O-GlcNAc cycling, OGT and OGA, also affect transcriptional and epigenetic structures and regulators including RNA polymerase II, histones, and histone deacetylase complexes (Hanover et al., 2012;Yang et al., 2002).It has therefore been conjectured by Hanover and others (Dehennaut et al., 2014;Hanover et al., 2012) that O-GlcNAc cycling may serve as a homeostatic mechanism linking nutrient availability with chromatin remodelling.
Alteration of nucleocytoplasmic proteins occurs with β-N-acetylglucosamine induced by the glucose metabolism, which is increasingly decreased in both T2DM and LOAD cerebra.OGT is induced by insulin on membrane cholesterol domains dependent on PI3K signalling, here reducing insulin receptor expression and promoting PI3K and MAPK activation (Perez-Cervera et al., 2013).Furthermore, hepatic overexpression of OGT hampered insulin-responsive genes thus inducing insulin resistance and dyslipidaemia (Yang et al., 2008).In rodent experimentation, it was found that hyperglycaemic states promote apoptosis of pancreatic β cells though accumulation of O-GlcNAc (Liu et al., 2000).
OGT has been identified a nutrient sensing mechanism for cellular insulin signal transduction (Akimoto et al., 1999).Phosphatidylinositol 3,4,5-trisphosphate (PIP 3 , Akt pathway) recruits OGT from the nucleus E. Lemche et al. to the plasma membrane, where it hampers insulin signalling and thus creates the basis for insulin resistance and T2DM (Yang et al., 2008).Further analogous mechanisms have been identified with oxidative stress already at adipocyte levels, by reducing the phosphoinositide 3-kinase (PI3K/h-Akt) interactions with both IRS-1 and IRS-2 (Ma et al., 2018;Whelan et al., 2010).Nutrient sensing is exerted by OGT on αCaMKII-positive neurones in the PVN of the hypothalamus, and its loss precludes satiety signals, thus causing hyperphagia (Lagerlöf et al., 2016).OGT was furthermore found promoting autophagy in nutrientdependent manner (Guo et al., 2014).

OGT and OGA in protein degradation
Together with its counterpart enzyme O-GlcNAcase, is the dynamic cycling PTM (Hart et al., 2007) controlled at serine and threonine residues (Yang and Qian, 2017).Herein, OGT catalyses the O-GlcNAcylation of proteins, and OGA their hydrolytic cleavage from O-GlcNAc, dependent on nutrients and stresses, reviewed in Ref (Yang and Qian, 2017).Dynamic and transient O-GlcNAcylation is induced by insulin, glucagon and ghrelin secretion in adaptation to metabolic demands, in order to modulate tissue-specific signalling pathways; this has been considered a specific link between T2DM and LOAD (Hart et al., 2007).Both human analyses and metazoan models suggested that both OGT and OGA are involved in protein degradation (Lazarus et al., 2012;Wang et al., 2012).Furthermore, alterations in O-GlcNAcylation of transcription factors in macrophages and lymphocytes affect inflammation and cytokine production with increasing age (Banerjee et al., 2016).

OGT and neural transmission
In the brain, O-GlcNAc glycosylation of proteins was found occurring with neuronal excitation and believed essential for communication of neurones, similar to phosphorylation (Khidekel et al., 2007).In the CNS, loss of OGT leads either to postnatal lethality or progressive neuroinflammation, tauopathy and Aβ deposition (Wang et al., 2016) (see below).OGT ko studies indicated that AMPA receptor subunits GluA2 and GluA3 were less expressed on excitatory hypothalamic neurones, with fewer presynaptic terminals and lesser dendritic spines (Lagerlöf et al., 2017).In GluA2 subunits, a specific LTD was discovered in hippocampal CA1-CA3 subfields dependent on O-GlcNAcylation, suggesting its importance for synaptic efficacy in memory and cognition (Taylor et al., 2014).
Ko murine models suggested that OGT enrichment in the PNS is a prerequisite of axonal outgrowth and maintenance of sensory fibre functioning (Su and Schwarz, 2017).O-GlcNAcylation of the myelin protein periaxin (PRX 19q13.1)was found critical for maintenance of peripheral myelin and axon integrity (Kim et al., 2016;Kim et al., 2018); whereas its loss is associated with progressive peripheral sensory neuropathy mediated by the AP-1/JUN transcription factors (Kim et al., 2018).To date, however, periaxin was not observed in oligodendrocytes, yet in cerebral endothelial cells (Wang et al., 2018a).However, it was demonstrated in cultured cells that neuronal stress induced by glucose deprivation increases OGT expression, which in turn activated p38 MAPK interaction (Cheung and Hart, 2008); it is believed that this mechanism could affect axonal structures.

Tauopathy and LOAD
Glycosylation by protein O-GlcNAcylation (Wani et al., 2017), a common PTM (Wang et al., 2017b;Yang and Qian, 2017) of nucleocytoplasmic, cytoskeletal, mitochondrial, synaptic and memory-related proteins with β-N-acetyl-glucosamine, which is regulated by glucose metabolism, was found markedly decreased in LOAD (Liu et al., 2009a).Proteomic analysis of post-mortem LOAD brains indicated that ~2 5% of tested proteins are significantly altered in LOAD (Wang et al., 2017b).
Herein, τ protein deposition is known to be covaried with episodic memory encoding (Chiotis et al., 2018).The posttranslational modification of τ protein by O-GlcNAcylation reciprocal to τ hyperphosphorylation, was in 1996 first observed to modifying all splice variants of microtubule-associated protein τ (Arnold et al., 1996).More importantly, the decrease in O-GlcNAcylation correlated negatively with phosphorylation at most phosphorylation sites of τ protein (Liu et al., 2009a), suggesting their inverse association.In addition, also reductions of Aβ neurotoxicity have been observed (Zhu et al., 2014), suggesting it has a neuroprotective function (Wani et al., 2017), but which is diminished in LOAD.It has also been found to be directly related to neuroplasticity through protein synapsin-1 (SYN1 Xp11.3-p11.23)(Cole and Hart, 1999).
This downregulation of both O-GlcNAcylation and protein phosphatase 2A in LOAD suggests that impaired brain glucose metabolism (Section 2.3.) may lead to abnormal τ hyperphosphorylation, oligomerisation, and degeneration by NFTs via depletion of τ O-GlcNAcylation in LOAD (Liu et al., 2009a;Yuzwa et al., 2012).In addition, increasing levels of OGT for O-GlcNAcylation or controlling its extent with OGA could be therapeutic for LOAD patients.Bisecting N-acetylglucosamine (GlcNAc) is the enzymatic product of a glycosyltransferase, N-acetylglucosaminyl-transferase III (GnT-III; MGAT3 22q13.1).Bisecting GlcNAc levels are increased in the LOAD patients CSF, and levels of GnT-III, the glycosyltransferase responsible for synthesising a bisecting GlcNAc residue, were found elevated in LOAD brains (Wang et al., 2018b).

AGEs and AGE receptors
Structural brain damage is jointly induced by oxidative stress, Olinked glycoprotein, and advanced glycation endproducts (AGEs) (Irie et al., 2008).Glycation and the glycolytic pathway produce α-ketoaldehyde methylglyoxal (MG), formed nonenzymatically by degradation as a by-product by sugar fragmentation reactions (Angeloni et al., 2014), which is the main precursor of AGE formation.In T2DM and LOAD, neurotoxic AGEs in hippocampal neurones (Krautwald and Munch, 2010;Srikanth et al., 2011) and glial cells indicate Maillard reactions due to accelerated ageing.AGEs and their receptor RAGE resulting from chronic hyperglycaemia provide critical links between T2DM and LOAD (Pugazhenthi et al., 2017).AGER produces several splice variants resulting in protein isoforms of RAGE, including an extracellular form and a N-truncated form (Srikanth et al., 2011).AGEs and RAGEs become critically co-involved in LOAD core pathophysiology (Mushtaq et al., 2014;Yang and Song, 2013), as AGE albumin is synthesised and secreted by microglia, triggering the expression of RAGE, and high mobility group protein B1 (HMGB1 13q12.3).Furthermore, RAGE is also a receptor for Aβ 1-42 (Pugazhenthi et al., 2017), which triggers downstream interaction of the JNK/SAPK pathway (Bogoyevitch et al., 2010;Tamagno et al., 2012) in both MetS and LOAD.

Mitochondrial metabolic functions: Irisin, AMPK, and mitochondrial uncoupling proteins
The myokine irisin (FNDC5 1p35.1) is a transmembrane protein (Jedrychowski et al., 2015), which has been shown to be significant for homeostatic functions of energy, fat and sugar metabolism, and cognition (Boström et al., 2012;Chen and Gan, 2019).Neurotrophic functioning of irisin implicates the regulation of neural differentiation and proliferation, cognitive functioning, regulation of energy and cardiac vagal tone.Vagotomy experiments indicated that irisin has effects in the hypothalamic nuclei, where it exerts metabolic functions through oxygen consumption, thermoregulation, and basal metabolic turnover (Zhang and Zhang, 2016).Its presence in various organ system suggests that irisin could have cross-organ messaging functions linking brain with skeletal muscular, adipose tissue and the cardiovascular system (Zhang and Zhang, 2016).Neuronal FNDC5 expression was found regulated by proliferator-activated receptor γ-coactivator 1α (PGC-1α) (Boström et al., 2012), and forced expression of FNDC5 in primary cortical neurones covaried with BDNF expression (Wrann et al., 2013).In addition, it was demonstrated that elevated peripheral irisin induced also neuroprotective factors c-Fos, ARC, and ZIF268 expression in the E. Lemche et al. hippocampus (Wrann et al., 2013).

Neuroprotective functions of irisin
It was discovered that FNDC5 covaries with transition of neural stem cells into neuronal precursor cells and astrocytes (Forouzanfar et al., 2015;Ostadsharif et al., 2011), implicating a role of FNDC5 in controlling neurodifferentiation in CNS maturation (Hashemi et al., 2013;Zhang and Zhang, 2016).CSF irisin levels directly correlated with BDNF levels (Lourenco et al., 2020).At two stages of neurogenesis, reduction of FNDC5 expression resulted in differential effects, (a) inhibition of neural precursors, and (b) reduction of mature neuronal markers (Hashemi et al., 2013).Reverse effects were found with FNDC5 overexpression (Forouzanfar et al., 2015), and were attributed to BDNF (Forouzanfar et al., 2015;Wrann et al., 2013).Furthermore, there was an association identified of the ERK1/2 signalling pathway with FNDC5 expression (Kim et al., 2023), and a significant controlling effect of basic fibroblast growth factor (FGF2 4q28.1) on the FNDC5 promoter (Hosseini Farahabadi et al., 2015).There is recent confirmation that irisin can be expressed in the CNS in vivo, however, as well as evidence supporting active transport of peripheral irisin across the BBB (Zhang and Zhang, 2016).FNDC5 release by neurones in the hippocampus was demonstrated (Wrann et al., 2013), and also expression in liver (Zhang and Zhang, 2016) and in adipocytes.

Irisin as a biochemical link between T2DM and LOAD
As mentioned, is irisin the effector protein of PGC-1α (Boström et al., 2012;Chen et al., 2024) and thus counteracting ageing-related obesity and T2DM (Chen et al., 2024;Wenz et al., 2009).The fibronectin ectodomain of FNDC5 is cleaved from a prohormone into the soluble peptide hormone, which is secreted from muscle, in analogous mechanisms as EGF and TGF-α (Boström et al., 2012).Its dependence of aerobic exercise has been reliably confirmed by mass spectrometry by covariation of its circulating levels in human plasma (Jedrychowski et al., 2015) with levels of incretins and insulin.Irisin acts on white adipose cells in culture and in vivo to stimulate mitochondrial thermogenin (UCP1 4q31.1)(Wu et al., 2012) expression (see below), and beige (Wu et al., 2012) and brown adipocyte development (Boström et al., 2012;Chen et al., 2024;Zhang et al., 2016).
Muscular irisin release was shown to reduce cerebral oligomeric Aβ burden (Lourenco et al., 2019): Peripheral FNDC5 expression into circulation was found being transported into the brain, where it rescued memory loss and improved synaptic plasticity.Furthermore, expression of the prohormone FNDC5 was also seen in human hippocampus, whereas it was found reduced in LOAD and aged CSF.In the brain, FNDC5/irisin triggered cAMP-PKA-CREB signalling and prevented dendritic spine loss (Lourenco et al., 2019).Murine AD models suggested irisin also being essential for LTP (Lourenco et al., 2019).Unclear though remains the issue, through which receptor irisin exerts its action (Chen and Gan, 2019), although this was considered being a G-coupled receptor.Alternatively, the existence of a specific irisin receptor has also been suggested albeit it remains yet unidentified (Zhang and Zhang, 2016).Recently, it was demonstrated that integrin aV/b5 can function as an irisin receptor (although not generally (Wang et al., 2023)) on astrocytes required for irisin-induced release of astrocytic neprilysin, leading to clearance of Aβ (Kim et al., 2023).

AMPK and insulin secretion
5′ adenosine monophosphate-activated protein kinase (AMPK) is expressed in the liver, brain, pancreas, and peripheral muscles.In muscles, it is considered to interacting with (a) irisin and (b) cathepsin B, and (c) BDNF, by the latter improving cognitive functions (Townsend and Steinberg, 2023).It is considered crucial for inter-organ communication and cell growth (Mihaylova and Shaw, 2011).AMPK activation stimulates hepatic and muscular fatty acid oxidation, glucose uptake, triglyceride synthesis, inhibition of adipocyte lipogenesis, activation of adipocyte lipolysis, and exerts modulation of insulin secretion by pancreatic β-cells.AMPK signalling and hypothalamic food regulation (Lopez, 2018b) have been shown being important for visceroception, integration of hormone networks with neural circuits (Lopez et al., 2016), and as the interface of vagal and sympathetic ANS branches.Further to the hypothalamus, AMPK is also expressed in the pontine hindbrain (Lopez et al., 2016): the medial NTS and AP.In the hypothalamic ventromedial nucleus, AMPK regulates energy expenditure, glucose, and lipid metabolism; whereas AMPK in the hypothalamic paraventricular nucleus modulates dietary preference for carbohydrate over fat (Lopez, 2018a).AMPK in ventromedial hypothalamus regulates glucagon and GLP-1, and in the NTS GLP-1 (Lopez et al., 2016).
In the pancreas, both insulin secretion and β-cell mass dynamics are regulated by the liver kinase B1-AMP-activated kinase (LKB1-AMPK) pathway (subunits α1 and α2, β1 and β2) (Rourke et al., 2018).AMPK is inhibited by insulin, leptin, and triglycerides by inducing various phosphorylations.AMPK is believed to partly mediate exercise-induced glucose uptake by GLUT-4 and subsequent stimulation of glycolysis.Hereby, an epigenetic mechanism is at hand: histone deacetylase 5 (HDAC5) is a substrate of AMPK (McGee et al., 2008), where it causes phosphorylation, and HDAC5 acetylation (reviewed by (Zhang et al., 2009).HDAC5 thereby loses its function as transcriptional repressor, which allows AMPK-activated GLUT-4 transcription (Zhang et al., 2009).Histone modifications by both phosphorylation and acetylation induced by AMPK were documented in the metabolic pathway flux (McGee and Hargreaves, 2019).

Energy homeostasis and synaptic functions
AMPK is considered responsible for cellular energy homeostasis and glucose and energy sensing (Carling et al., 2011), and triggered by a drop in ATP levels (Mihaylova and Shaw, 2011).It becomes inactivated by phosphorylation of acetyl-CoA carboxylase (ACC) and HMG-CoA reductase, the enzymes restricting cholesterol synthesis (Carling et al., 2011).Because of this cell energy homeostatic function, AMPK was demonstrated critically involved in establishing long-term memory (LTM) formation in primary neurones (Didier et al., 2018).Energetic requirements of synaptic transmission are regulated by AMPK by activation of transcription of Intermediate Early Genes via the PKA/CREB pathway through soluble adenyl cyclase (Didier et al., 2018).

AMPK and glucose-sensing neurones
There is a specific population of GABAergic GLUT-2 expressing neurones in the NTS (Lamy et al., 2014), which are sensitive to hypoglycaemia.Hypoglycaemic levels consistent with reduced intracellular glucose metabolism increased AMPK activity, inducing closure of leaking K+ channels (Lamy et al., 2014).NTS GLUT-2 neurones that project to dorsal vagal motor root nucleus increased parasympathetic nerve outflow and glucagon secretion (Lamy et al., 2014).Furthermore, amylin-activated projections converge in the NTS: the amylin-activated neurones in the lateral parabrachial nucleus (LPB) project from the NTS to the primary orexigenic dorsal lateral hypothalamic area (Potes et al., 2012).Amylin-related structures are AP, NTS, LPB, the central amygdaloid nucleus (CeA), and the lateral bed nucleus of the stria terminalis (BNST) (Potes et al., 2012).Further glucose-excited and glucoseinhibited neurones have been described in the ventrolateral portion of the ventromedial hypothalamic nucleus (VL-VMN) responsive to leptin and oestrogen (Routh, 2010;Santiago et al., 2016).

Loss of AMPK glucose sensing and T2DM
In T2DM, PI3K/Akt, MAPK and AMPK signalling pathways fail to interact, which impairs glucose homeostasis (Rowart et al., 2018;Schultze et al., 2012).It is established that loss of AMPK prevents glucose sensing, however the precise mechanisms remain unexplained.

Restoring insulin sensitivity
AMPK is dysregulated in MetS and T2DM in states of insulin resistance, and its reactivation improves insulin sensitivity (Coughlan et al., 2014).Evidence from transgenic murine models suggests that both AMPK and SIRT1 orchestrate PCG-1α (Canto and Auwerx, 2009), the master regulator of mitochondrial biogenesis, presumably through ageing-related posttranslational modification of PPARCG1A.AMPK has been shown to trigger PPARγ expression (Lyons and Roche, 2018), but with divergent results on different subunits α1 and α2.AMPK has furthermore been found specifically active in cultured endothelial cells (Strembitska et al., 2018), where it enhances insulin-stimulated Akt activation.AMPK is also known to interact in tight junction (TJ) ontogenesis with claudin, occludin, and protein zonula occludens (ZO)-1 (Rowart et al., 2018), and therefore decisive for BBB integrity.

AMPK and hyperphosphorylation of tau
There is converging evidence that AMPK and the AMPK-related microtubule affinity regulating kinase (MARK) family are decisive for hyperphosphorylation of microtubule-associated proteins including protein tau (Drewes et al., 1997;Nishimura et al., 2004;Thornton et al., 2011).This capability of AMPK for targeting serine sites on protein tau has been shown to be also related to Aβ presence (Thornton et al., 2011).In neuronal plasticity, AMPK was found exerting a limiting function for polarisation of neurones (conversion of dendrites into axons and directing these) through control of phosphatidylinositol 3-kinase (PI3K) localisation (Amato et al., 2011).Here, AMPK was shown to induce phosphorylation of kinesin light chain 2 (KLC2 11q13.2),thus inhibiting axonal outgrowth through prevention of PI3K localisation to the dendrite tip.Previously, a study reported the related protein kinesin light chain 1 (KLC1 14q32.33)as a target of AMPK in pancreatic β-cells, where AMPK activation blocks glucose-stimulated insulin secretion.Whilst both kinesin-1/Kif5B and kinesin light chain-1 (KLC1) contain consensus AMPK phosphorylation sites, however, only in-vitro peptides were found phosphorylated by AMPK (McDonald et al., 2009).

Mitochondrial uncoupling proteins
The family of membrane bound uncoupling proteins (UCP1-5) regulate in mitochondrial OXPHOS respiration the pumping of protons during oxidative phosphorylation of ADP into ATP (Ramsden et al., 2012), and because of parallel heat release, are hence crucial for cellular thermogenesis.UCPs function dependent on thyroid hormones, leptin, and catecholamines as metabolic regulators; the exact mechanisms yet remain unresearched.By counteracting release of reactive oxygen species in a process called "mild uncoupling" (Echtay et al., 2002), UCPs promote proton leakage into the cell matrix to prevent excessive accumulation of ROS (Ramsden et al., 2012).This uncoupling of ROS formation is dependent of fatty acids present (Echtay et al., 2002).UCPs have greater presence in neurones with glycolytic metabolism (Rupprecht et al., 2014), and are therefore T2DM-associated.In the LOAD rat model OXYS, early mitochondrial dysfunction was confirmed by reduction of UPC expression (Tyumentsev et al., 2018) and subsequent decrease in mitochondrion counts.
The universal thermogenin UCP1 (4q31.1)has tissue specificity for brown adipocytes, and is stimulated by irisin (see above).UCP2 (11q13.4) is preponderant in the immune system (bone marrow, lymph nodes, spleen, appendix); SNPs are associated with T2DM and BMI, its -866G/A polymorphism in the UCP2 promoter (Gomathi et al., 2019) is related to insulin release and pancreatic β-cell function.It was also found active in early neuronal differentiation and proliferation (Rupprecht et al., 2014).UCP2 is required for POMC synaptic plasticity (Kim et al., 2019), and therefore, UPC2 deletion prevents diet-induced obesity (DIO).HFD induces functional and morphological changes into microglial activation in arcuate microglial mitochondria through UCP2, thereby producing neuroinflammation (Kim et al., 2019).UCP3 (11q13.4,two splice variants) is believed to protecting mitochondria against lipid-induced oxidative stress, if fatty acids exceed mitochondrial oxidation capacity, by enabling their export.Specifically, its SNP rs1800849 is related to prediabetes and T2DM.UCP4 (6p12.3),with its multiple splice variants, is mainly expressed in brain tissue.The UCP4 promoter is the target effector of the NFκB-prosurvival pathway in reducing the effects of oxidative stress (Ho et al., 2012;Ramsden et al., 2012).Its rs9472817 is associated with individual susceptibility to LOAD, and found modulating ApoE4 impact (Montesanto et al., 2016).UCP4 is present in highly differentiated neurones (Rupprecht et al., 2014).UCP5 (Xq26.1)is, with multiple splice isoforms, as UCP4, primarily expressed in neurones, related to lipid peroxidation and metabolic rate (Ramsden et al., 2012), and performing neuroprotective functions.In reverse, under exacerbation of oxidative stress, a participation of UCP5 in neurodegenerative processes has been assumed (Ho E. Lemche et al. et al., 2006).The neuronal UCPs 2, 4, 5 are, as reviewed in ref. (Andrews et al., 2005), neuronal regulators of mitochondrial biogenesis, calcium flux, free radical production, and thus influence neurotransmission and synaptic plasticity (Andrews et al., 2005;You et al., 2024).This hypothesis is supported by greater presence near axon terminals and ion transport features in neuronal UCPs (Hoang et al., 2012): Recent expression studies (Hoang et al., 2012) indicated that neuronal UCPs 2, 4, 5 exhibit transmembrane chloride transport activity modulated by the mitochondrial lipid cardiolipin.

Further established links: cystatin, Klotho
Cystatin C (CST3 20p11.21)or neuroendocrine basic polypeptide, is considered a housekeeping gene associated with the basal metabolic rate (Delanaye et al., 2008).Three SNPs in the CST3 promoter region express two common variants, alternative splicing results in multiple transcript variants encoding a single protein.CST3 may be a further link between T2DM and LOAD, since polymorphisms have been found associated with both conditions.Cystatin C levels were found correlating with microalbuminuria in MetS and T2DM (Vijay et al., 2018), cognitive impairment (Kono et al., 2017), ageing with accelerated frailty (Hart et al., 2017), white matter hyperintensities (Lee et al., 2017), and myelin instability (due to its regulation of myelin-associated glycoprotein MAG proteolysis) (Stebbins et al., 1998).Myelin instability was recently identified altering APP processing and activating microglia (Depp et al., 2023).

CST3/7 and DAMs in neuroinflammation
CST3 has been associated with LOAD risk (Kaeser et al., 2007): It is established that cystatin C binds Aβ, and contributes to its proteolysis, however, it has also been shown that it fosters aggregation of Aβ 1-40 and Aβ 1-42 at certain concentrations (Kaeser et al., 2007).The homozygous CST3A/A genotype specifically conferred risk for LOAD associated with reduced level of cystatin C in the peripheral circulation (Chuo et al., 2007).The dimerising L68Q mutant of human cystatin C causes massive amyloidosis due to its tendency for aggregation in cerebral arteries during ageing (Janowski et al., 2001).Two LOAD-specific microglia types (DAMs, disease-associated microglia) have been identified (Keren-Shaul et al., 2017) constituting 6% of classified cells, with microglial gene expression of the amyloid neuroendocrine basic polypeptide cystatin C and lysosomal enzyme β-hexosaminidase (HEXB 5q13.3) in combination with an additional unique signature of lipid metabolism and phagocytic genes, including apolipoprotein E (APOE 19q13.32),lipoprotein lipase (LPL 8p21.3), and cystatin F (CST7 20p11.21).CST3 and APOE have been identified as parts of the plaque-induced gene network in App NL-G-F and C57BL/6 murine models (Chen et al., 2020).
CST7 was identified an advanced DAM marker related to τ burden and cognitive decline (Pereira et al., 2022).

Klotho as a metabolic gene regulator
The enzyme α-Klotho (KL 13q11.1), a β-glucuronidase responsible for cleavage of complex carbohydrates, is a gene regulator of ageing and longevity (Castner et al., 2023;Masso et al., 2018), but has implications for insulin sensitivity, vascular integrity, and cognition.KL variant VS (SNP rs9536314) elevating Klotho in mice enhanced synaptic long-term potentiation (LTP), synaptic plasticity, and enriched synaptic GluN2B, an N-methyl-D-aspartate receptor (NMDAR) subunit with key functions in learning and memory (Castner et al., 2023;Dubal et al., 2014;Dubal et al., 2015).In humans, this SNP is associated with greater GM volume, specifically in the DLPFC (Yokoyama et al., 2015).KL-VS heterozygosity was also found related to unspecific dementia risk and cognitive deficits in schizophrenia (Almeida et al., 2017b;Morar et al., 2018).Recently, an inverse correlation between higher levels of CSF Klotho was demonstrated with lower CSF Aβ42 and τ burdens (Grøntvedt et al., 2022).Such neuroprotective effects were observed ageingindependently (Castner et al., 2023;Dubal et al., 2014;Masso et al., 2018).There is an ageing-related decline of Klotho in serum and CSF (Semba et al., 2014), which is caused by loss of epigenetic modification in the Klotho promoter, and which induces mitochondrial dysfunction (Sahu et al., 2018;Xiao et al., 2004).Klotho-deficient rodents exhibit a progeria syndrome, which is characterised by endothelial dysfunction through vasodilation and impairment of angiogenesis, suggesting a role in microvascular stability by its interaction with transcription regulator early growth response protein 1 (EGR1 5q31.1) and the fibroblast growth factor family (Choi et al., 2010).
Through regulatory roles in cleavage of carbohydrates, LDL metabolism, and cholecalciferol production, it is a potential link between T2DM and LOAD.Klotho expression is deficient in T2DM (Masso et al., 2015), which causes microalbuminuria and disease progression towards diabetic complications (Nie et al., 2017).In return, enhanced expression of Klotho provided multiple β-cell protective aspects, including insulin secretion and storage, and reduction of oxidative ER stress (Lin and Sun, 2015).Likewise, were neuroprotective mechanisms for Klotho in cultured hippocampal neurones described through its target, the erythropoietin receptor (EPOR 19p13.2),and erythroid transcription factor (GATA1 Xp11.23), acting against oxidative injury and Jak2/ STAT5 signalling (Cheng et al., 2015).The α-Klotho transmembrane isoform m-KL was found most abundant in brain tissue (Masso et al.,Fig. 2. Summary of coaction of metabolic mechanisms 1 (Section 2).This summary diagramme depicts how various elements in the disturbed glucose metabolism are interconnected with switches in the LOAD core pathophysiology.On the left side, there is a relation of advanced glycation endproducts with APP.There is a modulatory effect of GLP-1R on BDNF.GSK-3β, which is responsible for phosphorylation of protein tau, is modulated by IGF-1, on the one hand, and by phosphatidylinositol 3,4,5-triphosphate (PIP 3 ), on the other; the latter downregulating insulin signalling.There are reciprocal effects of O-GlcNAcylation on brain glucose metabolism.The mechanisms displayed were independently verified by separate literature searches.Acronyms are listed.2015), and associated with exercise preventing decay in AD murine models.In humans, chronic psychosocial stress lead to Klotho depletion (Prather et al., 2015).In mice transgenic for human islet amyloid precursor protein (hIAPP; Section 3.1), Klotho expression diminished premature mortality, network dysfunction, and enhanced spatial learning and memory (Dubal et al., 2015).During neurogenesis, Klotho was shown to be involved in oligodendrocyte differentiation and myelination (Chen et al., 2015), whereas in cultured oligodendrocyte precursor cell line it was observed that Klotho exerted control in gene expression, cell differentiation, and ERK/Akt signalling pathways.In cultured hippocampal neurones was Klotho found being protective against ROS mediators, glutamate excitotoxicity, and toxicity of oligomeric Aβ (Zeldich et al., 2014).With respect to LOAD core pathophysiology, it was found that the transmembrane isoform of Klotho may undergo analogous shedding processes by sheddases α-, β-secretase, and a final cleavage by γ-secretase as is the amyloid-β precursor protein (Bloch et al., 2009).However, there were no interactions found between KL-VL allelic and ApoE4 allelic status in neuroimaging of Aβ burden (Porter et al., 2019).

Amylin deposition and manifestation of T2DM
Human islet amyloid polypeptide (hIAPP; also known as amylin) is a 37-amino acid peptide and incretin hormone co-secreted with insulin at a molar rate 15:1 from β-cells in the pancreas, but in antagonism to glucagon (Masters et al., 2010;Pruzin et al., 2018).Human IAPP is produced by the β-cells as a prohormone, and proIAPP is processed into IAPP by three endoproteases: prohormone convertase 2 (PC2), prohormone convertase 1/3 (PC1/3), and by carboxypeptidase E (CPE) in the secretory granules; proIAPP is also known being expressed by neurones (Schultz et al., 2011).Cleavage of preproIAPP takes place in the endoplasmic reticulum, and processing of both proIAPP and proinsulin occurs in the Golgi (Westermark et al., 2011).

Amylin physiological functions
Major physiological functions of amylin are activation of insulin targets (Adler et al., 2014), and reduction of food intake and glycaemic control (Baase and Hayes, 2014;Fu et al., 2017a).Accordingly, amylin is mainly a downregulatory peptide by inhibiting insulin and glucagon secretion locally in the pancreatic Langerhans islets (Stienstra et al., 2012).In healthy humans, fasting plasma amylin concentrations are in the range of 4-2 5 pmol/l, and amylin is distributed equally to insulin in plasma and interstitial fluids.Unlike insulin, however, amylin is not eliminated in the liver, but mainly through renal metabolism (Qiu and Zhu, 2014).Amylin therefore accumulates in states of insulin resistance and T2DM (Pruzin et al., 2018) leading into hyperamylinaemia.It is possible that IAPP may also have extra-pancreatic sites of action, since expression of the peptide has been detected in the gastrointestinal tract and in sensory neurones (Jürgens et al., 2011;Schultz et al., 2011.In addition, amylin has cerebral binding sites, probably contributing to satiety regulation by inhibiting gastric emptying.Amylin has known receptors (Section 3.1.)in the area postrema and mesolimbic and hypothalamic regions, and alike insulin, has the capability to transgressing the BBB.Recent studies have shown preferential accumulation in the CNS and fatty tissues {Schultz et al., 2011;Westermark et al., 2011).
Indeed, amylin was found exerting neurotrophic properties in the AP (Liberini et al., 2016), where it stimulated neurogenesis, neuron differentiation and fostered neuronal projections in neonatal and adult rodents.This is mediated by the transcription factor neurodifferentiation 1 (NEUROD1 2q31.3),known to being associated with T2DM risk (Malecki et al., 1999).The neurogenic differentiation factor 6 (NEUROD6 7p14.31.3) (Chen et al., 2020) has been identified as a neuronal factor of a gene signature module expressed with increasing amyloid plaque burden in App NL-G-F and C57BL/6 murine AD models and confirmed in humans.
The S20G (AGCSer to GGCGlY) polymorphism of IAPP (12p12.1)was found disease-linked with incidence and severity of T2DM in some populations (Seino and Study Group of Comprehensive Analysis of Fig. 3. Summary of coaction of metabolic mechanisms 2 (Section 2).This summary diagramme depicts how various elements in brain insulin resistance interact at the extracellular level, the cell membrane and the intracellular level.The activation of the NLRP3 inflammasome, caused by amylin amyloidosis promotes blockade of the insulin receptor in brain insulin resistance.Reactive oxygen species trigger calpain 1, which contributes to Aβ42 production.Disease-associated microglia (DAMs) express cystatins C, F, lipoprotein lipase and phagocytic genes.Cystatin F is also a DAM marker related to protein tau.The master energy homeostasis regulator peroxisome proliferator-activated receptor gamma (PPARγ) coactivator 1 alpha (PGC-1α) interacts with thermogenin, irisin, and sirtuin 1. AMPK triggers phosphorylation of PGC-1α via p38/ERK signal transduction.The mechanisms displayed were independently verified by separate literature searches.Acronyms are listed.
Genetic Factors in Diabetes, 2001): a GWAS in the ADNI-1 and ADNI-GO/2 cohorts for middle temporal cortical thickness and cognitive function (Roostaei et al., 2016) identified the SNP rs73069071 within the amylin locus IAPP and overlapping solute carrier organic anion transporter family member (SLCO1A2 12p12.1)genes.Using post-mortem Aβ immunohistochemistry in the Religious Orders Study and Memory and Ageing Project, the SNP rs73069071 also showed an allelic Aβ-deposition interaction effect for global cognitive function (Roostaei et al., 2016).

NLRP3 inflammasome activation by amylin
Islet amyloid polypeptide in diabetic pancreas can activate the autoinhibited NLRP3 inflammasome, and thus induce in parallel IL-1β secretion and inflammation of the islets, leading to the development of pancreatitis (Fu et al., 2017a;Masters et al., 2010) (Section 4).Human NLRP3 is expressed by alternative splicing as six isoform variants, of which only the full length variant is yet fully understood (Yao et al., 2023).Human amylin is capable to trigger caspase-1 release and ASC specks (the adaptor protein apoptosis-associated speck-like protein containing a pyrin domain and a caspase recruitment domain, an inflammasome read-out) also known as (PYCARD 16p11.2),thus initiating IL-1β production from the inflammasome within the pancreas (Masters et al., 2010).Herein, hIAPP oligomers energise the autoinhibited NRLP3 inflammasome, which produces phagolysosomes upon its activation.Critical roles of cathepsin B and ROS were confirmed in amylin amyloidosis just as with Aβ (see Section 2.4.).In contrast could acetylation of the NLRP3 in concord with sirtuin 2 reverse chronic inflammation in aged mice (He et al., 2020) .

Amylin-amyloidosis
Amylin possesses, like Aβ, cytotoxic properties, but only at its premature aggregation stage with up to 6000 molecules, termed intermediate-sized toxic amyloid particles (ISTAPs) (Janson et al., 1999): this smaller oligomeric aggregation state caused membrane interruptions.However, the precise cerebral roles of amylin are still under dispute (Baase, 2017): it is yet unclear whether misfolded or aggregated amylin contributes to Aβ burden in LOAD, or whether amylin uptake could also reduce overall Aβ burden.
Amylin amyloidosis is toxic both to pancreatic β-cells (Pruzin et al., 2018), and to cerebral parenchyma.Histological studies in post-mortem tissues confirm amylin deposition co-varying with T2DM, but not in controls nor in VD, in close co-localisation with Aβ in temporal lobar sites in LOAD (Jackson et al., 2013).Pancreatic oligomerised amylin enriched in cerebral leads independently of Aβ deposition to amyloid formation.Brain parenchyma infiltrated by amylin exhibited vacuolations, enlarged interstitial spaces, spongiform alterations, and capillaries bent at amylin accumulation sites (Jackson et al., 2013).These results suggest in-parallel β-cell islet amylin deposition, in the brain colocalised with Aβ 1-42 in temporal cortices, BBB breakdown related to pericyte degeneration, and disturbance of brain lymphatic and the glial lymphatic systems (Pruzin et al., 2018).

Amylin and retinol binding protein
Amylin agent pramlintide is able to distinguish LOAD patients from MCI and healthy elders (Zhu et al., 2017a) by decrease of retinol-binding protein 4 (RBP4 10q23.33).Diverging results were found in post-mortem cerebral and CSF levels of RBP4 in LOAD patients (Maury and Teppo, 1987;Puchades et al., 2003).Yet other studies (Ishii et al., 2019), however, found no relations of peripheral RBP4 levels with dementia rating nor circulating Aβ 1-42 and p-τ.Retinol-binding protein 4 has been found an index of brain insulin resistance in the Swedish APP murine AD model (Mody et al., 2011).Bovine studies indicated moderate intercorrelations between plasma RBP4, insulin and fat mass (Liu et al., 2019).RBP4 expression was observed in adipocytes and hepatic stellate cells, but not in hepatocytes (Liu et al., 2019).RBP4 overexpressing murine models are insulin resistant, glucose intolerant, and have increased macrophage and CD4+ T-cell infiltration in adipose tissue (Moraes-Vieira et al., 2014).Thereby, CD206+ macrophages express proinflammatory markers triggered by RBP4 through toll-like receptor 4 (TLR4 9q33.1) and JNK-dependent pathways, thus inducing CD4+ Tcell T h 1 polarisation and adipose tissue inflammation (Moraes-Vieira et al., 2014).

Amylin receptor binding and cerebral targets
Amylin receptors (AmyRs) are G-coupled receptors subserving a multiplicity of functions, involved in ameliorating glucose metabolism, relaxing cerebrovascular structure, modulating inflammatory reactions, and potentially enhancing neural regeneration (Qiu, 2017).Neuroprotective effects are mediated by AmyRs, but misfolded amylin cannot activate AmyRs (Qiu, 2017).AmyRs, however, possess a capability for Aβ binding, and thereby support its drainage and efflux into the blood stream (Mohamed et al., 2017;Zhu et al., 2015).This drainage effect induced by amylin upon brain Aβ to blood clearance was found exerted through amylin receptor 3 (AMY 3 R) by subcellular translocation via LRP1 towards the plasma membrane of the BBB endothelium (Mohamed et al., 2017).
Amylin can further dock to monomeric CALCR or CALCRL without RAMPs (Baase, 2017).All calcitonin and AMYRs contain different allosteric binding sites providing differential activation of downstream signalling pathways (Fu et al., 2017a).Amylin at higher micromolar E. Lemche et al. doses does activate the mechano-osmo-sensitive vanilloid receptor 4 (TRPV4 12q24.11)(Zhang et al., 2017b).This TRPV4 activation was identified as the key molecular mediator for the cytotoxic effects of hIAPP on hippocampal neurones.CALCR has known interactions with apolipoprotein B and apolipoprotein E receptor LRP1, which is involved in LOAD core pathophysiology (Lemche, 2017).The interactions of CALCR with LRP1 are likely responsible for beneficial effects of amylin agonists versus detrimental effects of misfolded amylin aggregates in LOAD (Baase, 2017;Mohamed et al., 2017).

Amylin receptor binding
Amylin, alike insulin, crosses the BBB and has neuronal receptors.However, differential uptake for amylin is but greater than for insulin, indicated by influx and transport constants (Banks and Kastin, 1998).While insulin is mainly resorbed in pons-medulla and hypothalamus, but not in midbrain and occipital cortex, the main uptake regions of amylin were in cerebellum, and frontal, parietal, occipital cortices (Banks and Kastin, 1998).Amylin binding sites in primate brains, as determined with autoradiography, were in the arcuate nucleus, parts of the ventromedial hypothalamic nuclei, and the NTS.High density binding was present in dorsal raphe and AP (Paxinos et al., 2004).Furthermore, high density binding was also detected in parts of the preoptic area, BNST, amygdala and NAcc.
AmyR expression was shown in VM hypothalamic neurones, astrocytes, and microglia, and the latter released IL-6 and IL-10 upon amylin infusion (Foll et al., 2015), as well as leptin-induced phosphorylation of transcription factor STAT3 (STAT3 17q21.2),which triggers T H 17 differentiation, in wild-type rodents.Studies in murine models revealed transcription of IAPP also in neurones in the lateral hypothalamus, ARC, and the medial preoptic area (MPO) (Li et al., 2015), regulated by leptin (Li et al., 2015), where it appeared to co-regulate hunger-satiety jointly with leptin.In astrocytes were significant differences for IL-1β and TNFα observed (Foll et al., 2015), suggesting pro-inflammatory actions of amylin in the ARC and VM hypothalamic nuclei.

Aβ uptake by AMY 3 R
A number of histochemical studies revealed that pericytes in the BBB, hippocampus, and parahippocampal gyrus (Schultz et al., 2017) particularly suffer from hyperamylaemia, where intraneuronal amylin inclusion induced nuclear changes, Ca ++ imbalance, and subsequent neuron death.It is therefore likely that hyperamylaemia contributes to BBB breakdown as a consequence of vascular wall disintegration in patients with obesity, T2DM and LOAD (Jackson et al., 2013;Pruzin et al., 2018).In addition, parabiosis (i.e.blood exchange) used for transgenic murine AD models indicated that peripherally circulated human Aβ entered WT brain parenchyma and induced AD pathologylike changes (Bu et al., 2018).

Amylin cross seeding and co-deposition in cerebral plaque formation
The physiological functions of the amyloid-β precursor protein (APP 21q21.3)and APP-like proteins consist in the CNS of neurotrophic, neuroprotective and synaptic-stabilising contributions (brain growth, dendritic branching, spine density, axonal targeting, neurogenesis), which are absent in ko animals (Müller et al., 2017).Experiments in the LOAD-model SAMP8 mouse with the amylin analogue pramlintide suggest that also amylin exerts such neuroprotective properties (Adler et al., 2014): Neuronal AmyRs are capable of amylin uptake, which restores haem oxidase HO-1 (HSP32; HMOX1 22q12.3),and triggers expression of synapsin-1, CDK-5, and p-ERK1/2 (Adler et al., 2014).A current view thus is that oligomeric amylin may have beneficial effects in LOAD, whereas multimeric and misfolded aggregates are involved in the LOAD Aβ cascade and plaques.Examination of diabetes duration in humans suggests that amylin secretion is attenuated with diabetes manifestation (Qiu et al., 2014b), and hence neuroprotective effects are then abolished.There are currently few reliable life-span data available in humans for ageing-related changes in plasma amylin levels (Baase, 2017;Li et al., 2016).

Physiological and neuroprotective effects
In the mammalian CNS, soluble APPα (sAPPα, the α-secretase cleaved APP extracellular domain fragment) exerts LTP-facilitative and memory rescuing functions inhibitory to those processes involved with the LOAD-specific amyloidogenic pathway, and counteracting GSK-3βdependent hyperphosphorylation of tau (Müller et al., 2017).Analogous physiological functions of amylin are conceivable (Bharadwaj et al., 2017).Indeed demonstrated several studies recuperation of cognitive functions by amylin supply (Zhu et al., 2015;Zhu et al., 2017b).In contrast, a reverse neurodegenerative effect for amylin is likely in presence of Aβ oligomers.In human plasma samples, positive association between amylin levels (but not plasma insulin levels) and Aβ 1-42 as well as Aβ 1-40 is found only in patients with LOAD or amnestic MCI, irrespective of APOE allelic status (Qiu et al., 2014c;Zhu et al., 2015).It was suggested that this may be a result of amylin action in cerebral Aβ clearance (Zhu et al., 2015).

Amylin and cognitive functioning
Plasma amylin levels, but not plasma Aβ levels, were found covarying with several cognitive domains such as logical memory delayed recall, executive function, and visuospatial cognition (Qiu et al., 2014a;Zhu et al., 2015;Zhu et al., 2017b).Cognitive impairment and hyperinsulinaemia are associated with higher fasting amylin (Morris et al., 2016), although plasma levels of amylin in MCI and LOAD are strongly E. Lemche et al. reduced as compared to healthy elderly (Adler et al., 2014).Amylin levels were also associated with ApoE4 status and T2DM, but were substantiated independent risk factors for LOAD (Adler et al., 2014).Post-mortem histology comparing T2DM, LOAD with and without manifest diabetes (Jackson et al., 2013) detected amylin oligomers and plaques in the GM of T2DM sufferers, but not controls.Extensive amylin deposition was measured in vascular walls and perivascular spaces in LOAD patients, even without peripheral T2DM (Jackson et al., 2013).

Protein misfolding and pathological protein cross seeding
Different amyloid classes have the tendency for crosspolymerisation, their amyloidoses will mutually potentiate (Bharadwaj et al., 2017;Gonzalez et al., 2017;Oskarsson et al., 2015): despite different sequences share amylin and Aβ similar folding properties, and hence may complement mechanistically.Simulations showed in this respect that hIAPP in its monomeric form is more likely to co-aggregate with helicoidal Aβ structures (Ge et al., 2018).Likewise, due to shared epitopes, generic amyloid oligomer vaccination proved equally effective for hIAPP and Aβ in the murine AD model TG2576 (Rasool et al., 2012).
The most straightforward explanation to amylin as a link between T2DM and LOAD, is hyperamylaemia in insulin resistance (Fu et al., 2017a), which then causes amylin deposition not only in the pancreas, but also in the cerebrum.Animal experimentation suggests that amylin seeding into brain parenchymata promotes also Aβ deposition (Gonzalez et al., 2017).Multiple evidence for cross-seeding of Aβ and hIAPP suggests that this mechanism could drive Aβ-amylin spreading and colocalisation (Fu et al., 2017a;Martinez-Valbuena et al., 2019;Oskarsson et al., 2015).This is supported by positive association between amylin and Aβ 1-42 and Aβ 1-40 found in LOAD and aMCI (Zhu et al., 2015).However, there is an ageing-related decline in insulin secretion, whilst amylin levels were not found age-correlated (Li et al., 2016), irrespective of T2DM or CVD events.At the same time, there was a significant age-related increase in Aβ 1-40 (Li et al., 2016).
Protein-protein interactions revealed by proximity-ligation assays (Oskarsson et al., 2015), in both neuropathological post-mortem and transgenic murine analyses, confirmed that there is co-localisation of hIAPP and Aβ in cerebral tissue (Martinez-Valbuena et al., 2019), and Aβ precipitation in amyloid seeds in pancreata of T2DM patients and mice (Martinez-Valbuena et al., 2019;Miklossy et al., 2010;Wijesekara et al., 2017) (but see opposing findings in (Oskarsson et al., 2015)).Solely IAPP, but not proIAPP immunoreactivity, was detected in human senile plaques, highest in LOAD, but lacking in other dementias (Oskarsson et al., 2015).These findings are supportive of the notion of peripheral metabolic causation of LOAD.

Amylin and τ hexapeptides
Amylin could also be the searched link between protein τ and Aβ (Kurnellas et al., 2013;Martinez-Valbuena et al., 2019), as there are structural similarities between protein τ, Aβ and amylin in the form of hexameric amyloid peptides.Given the interactions of Aβ, hyperphosphorylated τ, and amylin (Martinez-Valbuena et al., 2019), it was recently argued that LOAD pathophysiological mechanisms could drive T2DM manifestation as well (Bharadwaj et al., 2017).Specifically, tauopathy mechanisms present early in LOAD, could promote amylin amyloidosis in pancreatic β-cells, thus accelerating T2DM by peripheral insulin resistance (Martinez-Valbuena et al., 2019).Unlike hIAPP, however, rodent IAPP has no fibrillogenic and neurotoxic properties, and does not accumulate as aggregates.Therefore, animal model evidence is limited on this issue.There is a larger species of protein τ expressed in skeletal muscles, whose tauopathy in term of hyperphosphorylations could contribute to insulin resistance (Bharadwaj et al., 2017).Post-mortem studies detected aggregated Aβ colocalised with amylin in islet amyloid deposits, hyperphosphorylated τ, ubiquitin, apolipoprotein E, apolipoprotein A, IB1/JIP-1 and JNK-1 in Langerhans islets in T2DM and LOAD patients (Martinez-Valbuena et al., 2019;Miklossy et al., 2010).

Neuroinflammation and amylin
Also, peripheral activation of AmyRs (Wang et al., 2017a) was associated with activated-microglial markers scavenger receptor CD68, allograft inflammatory factor 1 AIF-1, and mitochondrial enzyme ATP synthase F1 subunit β (ATP5F1B 12q13.3),which were correlated with NFT burden in human probes.In APP transgenic mice, the respective genes were identified as hubs in a co-expression analysis (Wang et al., 2017a), where CD68 was linked with IGF-1.Peripheral lymphatic vessel changes in T2DM (Haemmerle et al., 2013) as apparent from ex-vivo endothelial gene expression analysis indicates that insulin resistance activates gene clusters related to inflammation and immune responses.Specifically, the transmembrane glycoprotein CD68 expressed both in macrophages and microglia, was upregulated on macrophages.Monocytes expressing the scavenger receptor CD68, which is identified binding oxidised LDL from blood vessel walls (Doi et al., 1990;Fig. 4. Summary of coaction of amylin with other factors (Section 3).This summary diagramme depicts how protein misfolding triggers neuroinflammation under conditions of hyperglycaemia, LDL excess, hyperamylaemia.Triggered by cathepsin B, phagolysosomes play a role in agglomeration of amylin and Aβ42.Blockade of amylin receptor 3 is correlated with a number of inflammation and immune responses involving the scavenger receptor CD68.Amylin/Aβ42 oligomers play a pivotal role under conditions of oxidative stress in the activation of the NLRP3 inflammasome.The release of interleukin 1β then ultimately triggers the JNK pathway and co-activator beclin-3 expression.The mechanisms displayed were independently verified by separate literature searches.Acronyms are listed.Matsumoto et al., 1990), secrete proinflammatory cytokines.
Double transgenic AD mice expressing hIAPP revealed that peripheral insulin resistance, hyperglycaemia, and glucose intolerance are amyloid-inducible (Wijesekara et al., 2017).Whilst co-deposition of Aβ and hIAPP covaried with hyperphosphorylation of τ, was β-cell depletion observed in pancreatic islets, resulting in deficient insulin signalling (Wijesekara et al., 2017).Aβ immunisation rescued insulin production in this murine model, thus suggesting a role of Aβ in β-cell loss in T2DM (Wijesekara et al., 2017).Transgenic mice producing hIAPP exhibited microglia/macrophage activation, inflammatory response profiles and cognitive impairments (Srodulski et al., 2014) attributable to heightened amylin oligomerisation.As mentioned, amylin induced endoplasmic reticulum (ER) stress, which triggered macroautophagy: If formation of autophagolysosomes is hampered, accumulation of amyloid in autophagic vacuoles leads to disturbance of the lysosomal system (Westermark et al., 2011).It was presumed that beclin-1 deficiency could interfere with the formation of autophagosomes, a further parallelism between LOAD and T2DM (Westermark et al., 2011).

Pancreas-inflammation and islet-associated immune cells
Insulin resistance is a decrease in insulin-stimulated glucose uptake in peripheral tissues, and impairment of insulin-mediated suppression of hepatic glucose production (Donath and Shoelson, 2011).This decrease will be initially countered by islet compensation, characterised by increased β-cell mass and function and prediabetic persons may remain on this compensatory level.Overt T2DM only occurs when such islet compensation fails, due to progressive loss of functional β-cell mass.The onset of an autoinflammatory process during progression to T2DM is indicated by IL-1β, IL-6, IL-18, TNF-α and CRP levels (Bae et al., 2019;Donath and Shoelson, 2011), and the development of insulitis.Inducive factors for insulitis are glucotoxicity, lipotoxicity, ER stress, and oxidative stresses.Several epidemiological and experimental studies indicated that IL-1β receptor antagonist (IL1RN 2q14.1)increases precede T2DM manifestation and its elevation is therefore characteristic for prediabetic stages (Donath and Shoelson, 2011;Hui et al., 2017).In contrast, established T2DM is further characterised by (a) amylin deposition, (b) β-cell apoptosis, and (c) possibly β-cell dedifferentiation leading to loss of β-cell identity.Soluble oligomeric human IAPP is herein likely to contribute early to NLRP3 inflammasome-related IL-1β secretion in islets (Martinez-Valbuena et al., 2019;Masters et al., 2010;Morikawa et al., 2018), whilst hIAPP deposition in the pancreas is a more prolonged process (Ehses et al., 2007).

Microglial chemotaxis of ASC inflammation specks
Macrophage chemotactic protein or chemokine (C -C motif) ligand 2 (MCP1; CCL2 17q12) attracts monocytes and basophils towards inflammation sites.The CCL2 monomeric polypeptide circulates peripherally, but mainly resides in endothelial cell plasma membranes, induced by platelet-derived growth factor A (PDGFA 7q22.3), and through NFκB signalling.Hypomethylation of CpG sites within the CCL2 promoter region correlated with high levels of blood glucose and triglycerides, which were associated with serum CCL2 levels and vascular damage in T2DM (Liu et al., 2012).

Interaction of diet, oxidative stress and inflammation
Dietary components, such as overload of free fatty acids have been hypothesised to influence the conditions, under which the autoinhibited NLRP3 inflammasome becomes activated (Legrand-Poels et al., 2014); specifically, saturated fatty acids exhibiting toxicity, triggering ROS production, ER stress, apoptosis, and inflammation.In this mechanism, it is conceived that e.g., palmitate decreases AMPK phosphorylation, thus reducing AMPK-dependent autophagy (Choi et al., 2023), and leading to mitochondrial ROS damage.Furthermore, palmitate induces lysosome destabilisation and cathepsin B release.Studies of phospholipids containing polyunsaturated fatty acids and triglycerides in nonobese T2DM Goto-Kakizaki-rats (Malaisse et al., 2006;Portois et al., 2007), with comparisons of brain tissue with spleen, liver, and peripheral tissue, revealed substantial differences in support of organ-specific regulation (Malaisse et al., 2006;Portois et al., 2007), with most pronounced difference between spleen and liver (Portois et al., 2007).Through increased ROS production, oxidation of the phospholipid cardiolipin, essential for mitochondrial membrane integrity, is promoted: a process which NLRP3 is able to recognise.ER integrity is compromised by palmitate being incorporated into ER membrane phospholipids.The resulting leakage causes Ca ++ release from ER into mitochondria through mitochondria-associated membranes (MAMs), the connectors between ER and mitochondrion, thus inducing mitochondrial disruption (Legrand-Poels et al., 2014).This mechanism is present both in LOAD (Halle et al., 2008) and T2DM (Donath and Shoelson, 2011;Masters et al., 2010), as well as in other conditions such as atherosclerosis.

Genes and epigenomics: qualitative meta-analysis
The previous sections of the systematic review summarised in a first step molecular links and biochemical mechanisms connecting the hitherto known pathophysiologies of T2DM and LOAD.During this step, gene loci of the identified key players were collected and extracted (Fig. 1, PRISMA 2020 flowchart).In a second step, a qualitative meta-analysis was conducted on the extracted gene loci and epigenetic mechanisms, by which genome databases and other online resources were mined to identify respective genomic and epigenomic studies.The genomic variations associated with T2DM or LOAD were listed, alongst with accumulated findings on possible epigenetic and/or gene regulatory mechanisms, in Supplemental Table 1.
Further strong candidates are gene loci, for which associations with either T2DM and LOAD were documented in one of the categories: CCL2, SNP suspected to associate with T2DM and LOAD; EGR1, SNP in T2DM, epigenetic mechanism relevant to LOAD haplotype; GSK3B, SNP in LOAD, gene regulation in T2DM; HMOX, SNP in T2DM, epigenetic mechanism in LOAD; IGF1, SNP in T2DM, epigenetic LOAD suspect; IGFBP3, SNP in T2DM, epigenetic LOAD suspect; LDLR, SNP in LOAD, gene regulation in T2DM; MGAT3, SNP in T2DM, gene regulation in LOAD; OGA, SNP in T2DM, epigenetic PTM in LOAD suspect; NEUROD1, SNP in T2DM, epigenetic mechanism LOAD suspect; PARP1, SNP in LOAD, epigenetic LOAD suspect.

Discussion of results
The state of research may of course introduce a bias and limitation of the results, as certain key players received more attention and research efforts than others.Therefore, any lack of finding is not truly a negative result in this listing, as this may change with future positive findings and discoveries.On the other hand, is up-to-date replicated genomic association evidence and respective epigenetic findings a strong weight for a critical role in paralleling pathophysiological mechanisms.In summary, the present evaluation of results highlights mainly insulin signalling, inflammation and inflammasome pathways, proteolysis, calpain mechanism, gluconeogenesis and glycolysis, glycosylation, lipoprotein metabolism and oxidation, intracellular calcium or sodium, cell cycle regulation or survival, apoptosis, autophagic-lysosomal pathways, and energy homeostasis.For a larger part of the genomic loci, interactions with amyloid and/or τ protein processing have been documented.

Summary and conclusion
The present systematic review and qualitative meta-analysis examined biochemical parallelisms conferring pathophysiological and genomic risks between T2DM and LOAD.We described findings in histology and neuropathology, metabolic and endocrine mechanisms, amyloid formation, and inflammasome activation and immunology.Identified were molecular regulators or keys, recognised in both conditions.In a further level, extracted gene loci were screened for known disease associations and epigenomic modifications implicated in both or either of the conditions.
In summarising the main findings, present-state genomic reports support influence of insulin signalling system, inflammation and inflammasome pathways, calpain mechanisms, glucagon metabolism, lipoprotein metabolism, glycosylation mechanisms.Caveats must be formulated as greater efforts in established or attention-gaining loci are likely to bias such findings, whereas rarely investigated loci are less likely to yield positive findings.This limitation, however, does not restrict replicated positive findings.
Nevertheless, a strength of this systematic review is its comprehensiveness due to combining several levels of investigation.In particular, research reviews typically discuss isolated aspects of pathological mechanisms, yet few aim to connect various levels of hitherto accumulated knowledge.In the present systematic review, this was achieved by relating metabolic, protein misfolding, and inflammation-related mechanisms, which may be acting in parallel to both conditions.
For further steps, the results of the present systematic review will enable researchers to make use of NGS technologies with the capacity to target larger arrays of genomic loci, and also assess respective RNA expression levels in populations at risk.This will provide a broader basis for prevention and early intervention in insulin resistance states conducive to T2DM and LOAD.

Fig. 5 .
Fig. 5. Summary of coaction of inflammation and immune mechanisms (Section 4).This summary diagramme depicts how calcium influx and potassium efflux in oxidative stress conditions lead to M1 macrophage polarisation.Triggered by caspase 1, the release of interleukin 1β stimulates chemokine CCL2 and ultimately triggers the pro-apoptotic JNK pathway initiating neurodegeneration.This process again reinforces amylin amyloidosis in brain insulin resistance conditions.The mechanisms displayed were independently verified by separate literature searches.Acronyms are listed.