FormalPara Key Points

Mild cognitive impairment and Parkinson’s disease (PD) dementia (PDD) are common disorders and are associated with severe morbidity and increased mortality.

PDD is treated with cholinesterase inhibitors (ChIs). Ongoing trials against PDD target glutamate, monoamines, kinases, glucocerebrosidase and glucagon-like peptide-1 alone or together with ChIs.

Type 2 diabetes mellitus shares pathophysiological mechanisms with PD and neurodegenerative dementias, such as central and peripheral insulin resistance that in turn results in altered autophagy, cell proliferation and increased inflammation.

Antidiabetic drugs have shown positive effects on cognitive outcomes in PD animal models, as well as in early-phase clinical trials.

1 Introduction

Parkinson’s disease (PD) has traditionally been considered a movement disorder and its diagnosis relies on the presence of bradykinesia along with rigidity and/or rest tremor [1]. Non-motor symptoms have recently gained attention and quality-of-life studies have shown that PD patients are severely affected by its non-motor aspects. Indeed, increasing recognition has been given to non-motor manifestations; these are incorporated into both the latest Movement Disorder Society (MDS) criteria for PD [1] and particularly into separate MDS criteria for prodromal PD [2]. Among the non-motor symptoms, cognitive impairment and dementia are especially debilitating and they are the strongest component leading to nursing home placement of PD patients.

2 Prevalence of Cognitive Impairments in Parkinson’s Disease (PD)

Cognitive deficits are common in PD, but vary in both quality and severity in different stages of the disease. The spectrum ranges from subjective cognitive decline [3] to mild cognitive impairment (MCI) [4] and PD dementia (PDD) [5]. Subjective cognitive decline does not have established criteria, but it is generally defined as a mild impairment noted by the patient, family or caregivers, although performance in formal cognitive testing is within normal range. Little is known about the occurrence of subjective cognitive decline in PD, but it is assumed to precede future progression to MCI and dementia [6, 7]. The main difference between PD–MCI and PDD is the degree to which the cognitive decline affects everyday life functioning, which is minimal in the case of PD–MCI but substantial in PDD [4, 5]. The cognitive decline often hides under more profound motor deficits and only detailed neurocognitive evaluation reveals the grade to which these deficits extend [8]. Among patients with PD, 25–30% have PD–MCI [4, 9], while an additional 30% have a PDD diagnosis [10]. An 8-year cumulative prevalence of 78% for getting a diagnosis of PDD has been reported in a Norwegian cohort [11] and a cumulative incidence of 83% 20 years after initial diagnosis was reported in the Sydney multicentre study of PD [12].

2.1 PD–Mild Cognitive Impairment (MCI)

MCI can affect one or more cognitive domains, and is further subdivided into amnestic MCI, when memory is the predominantly affected domain, and non-amnestic MCI, when domains other than memory are more impaired [13]. Previously, MCI diagnosis was based on subjective complaints from the patient and/or observations from a caregiver or clinician about cognitive difficulties that cannot be explained by age alone and do not interfere with everyday living [14]. However, recently, the MDS Task Force has redefined PD–MCI and established criteria to facilitate the use of a common definition, additionally stating that the deficits should be attributed to PD, excluding other conditions, and that detailed neuropsychological testing should confirm any subjective observation [4].

Older age, longer disease duration, concomitant depression and a higher motor burden are positively correlated with PD–MCI [9]. While MCI most often develops into PDD, amelioration and return to normal cognitive functioning is not uncommon [15]. Non-amnestic, single-domain PD–MCI is the most common according to some studies [9], while others report that multi-domain MCI is the most common [16]. Executive functioning, visuospatial abilities, verbal fluency, memory and language have all been reported to be affected [9, 15]. Accumulating evidence suggest that there are two distinct categories of cognitive dysfunction in PD with different prognoses. The first one is a frontal executive dysfunction that correlates with dopaminergic loss, which is affected by dopaminergic therapy and has a better prognosis. The second type is characterised by visuospatial and semantic fluency deficits attributed to posterior and temporal dysfunction related to cholinergic loss. Prognosis in the latter is more severe, as the patients commonly progress to PDD [17, 18].

2.2 PD with Dementia

Dementia is the most severe cognitive syndrome, defined as acquired objective cognitive impairment affecting multiple cognitive domains that is severe enough to affect activities of daily life (ADL) [19]. PDD affects multiple cognitive domains, including attention, memory, executive and/or visuospatial ability [5]. In addition to this, PDD is characterised by neuropsychiatric features such as hallucinations and delusions, apathy, depression and anxiety [5]. PDD shares a lot of similarities with dementia with Lewy bodies (DLB), the latter presenting in close chronological proximity to the initiation of motor symptoms as opposed to PDD where parkinsonian semiology must precede cognitive deficits by at least 1 year [20]. Progression from PD or PD–MCI to PDD cannot be predicted but several risk factors have been identified [15]. These includes PD diagnosis at an older age, akinetic–rigid motor subtype, psychiatric manifestations, loss of postural control, dysautonomia, and rapid eye movement (REM) behaviour disorder (RBD) [5, 21].

Multiple mechanisms seem to be involved in the genesis of cognitive decline in PD. Limbic and cortical Lewy body pathology, as well as amyloid plaque and tau pathologies, are the most common contributors to PDD [22]. Additionally, the role of vascular disease [23] and neuroinflammation [24] has been emphasised. In order to explain the clinical heterogeneity between PDD patients, other mechanisms have also been examined such as the role of synaptic proteins [25], neurotransmitters [26] and genetic factors [26, 27].

Studies on glucosylceramidase beta (GBA) mutations have indeed highlighted a high risk for carriers to develop PD with a more rapid cognitive decline [27]. Moreover, several autosomal dominant α-synuclein mutations cause PD with a worse cognitive prognosis [28]. On the other hand, leucine-rich repeat kinase 2 (LRRK2) and parkin mutations have been described to lead to a milder PD phenotype with less cognitive decline [15, 28]. The role of polymorphisms in genes such as microtubule-associated protein tau (MAPT), apolipoprotein E (APOE) and catechol-O-methyltransferase (COMT) on cognitive outcomes has been examined, but results are not yet conclusive [15, 18, 28].

With regards to wet biomarkers, reduced cerebrospinal fluid (CSF) amyloid β 1–42 (Aβ1–42) concentration is consistently associated with cognitive impairment and predicts dementia development [29], while high levels of neurofilament (NFL) [30], inflammatory glycoprotein YKL-40 [31] and neurogranin [25] have been associated with cognitive decline in PD. Furthermore, it has been reported that PDD patients have increased CSF levels of total and phosphorylated tau and α-synuclein; however, there are still inconsistencies between different studies [32].

In the field of neuroimaging, several modalities have been used to investigate PD–MCI and PDD. Magnetic resonance imaging (MRI) with various structural and functional techniques along with positron emission tomography (PET) and single-photon emission computerised tomography (SPECT) of various tracers (glucose metabolism, cholinergic, amyloid and tau PET) have mainly been used [15, 33, 34]. Structural MRI studies have shown that there is an early cortical thinning in PDD [35] and PD–MCI patients have parietal cortical thinning as well as lower left orbitofrontal cortex grey matter volume as compared to non-demented PD patients [36]. It has also been reported that PD patients have reduced glucose metabolism compared to controls in a range of regions including the parietal and prefrontal cortices [37,38,39]. Moreover, PD patients with cognitive impairment have reduced metabolism in the temporal, parietal and premotor cortices as compared to PD patients without cognitive decline [38, 39]. Additionally, electroencephalography and magnetoencephalography studies have shown promising results in terms of predicting cognitive deterioration in PD [40].

3 Role of Diabetes-Related Pathways and Antidiabetic Drugs in PD Dementia (PDD)

Various mechanisms are involved in the underlying pathology of cognitive decline in PD. These include limbic and cortical Lewy body pathology, amyloid plaque and tau pathology, neuroinflammation, synaptic plasticity dysregulation, neurotransmitter alterations as well as genetic factors [15, 18, 22,23,24,25,26, 28]. The complex roles of these in the underlying pathology of PDD has been previously reviewed in depth [41, 42]. There is accumulating evidence that diabetes-related pathways are involved in cognitive decline in PD and here we focus on these mechanisms.

3.1 PD and Diabetes

Although PD and diabetes are clinically different, an association between the two disorders was described in the 1960s, with the assessment of glucose tolerance in PD patients [43]. It was later suggested that alterations in similar pathways and common underlying mechanisms exist between the two disorders [44, 45].

Several epidemiological studies have demonstrated an association between type 2 diabetes mellitus (T2DM) and an increased risk of develo** PD [46,47,48,49,50]. More specifically, a large proportion of PD and PDD patients have impaired insulin signalling and insulin resistance [44, 51]. A prospective epidemiological study in a Finnish population found an association between T2DM and the risk of develo** PD [46] and, furthermore, T2DM has been associated with a younger onset of PD and more severe symptoms [47, 48]. However, in contrast, some studies did not find diabetes to be a risk factor for PD and even reported a lower prevalence of diabetes in PD patients than in controls [45, 52,53,54]. Despite these conflicting reports, it has been suggested that the diseases may have common underlying biological mechanisms [44, 45]. Mitochondrial dysfunction, autophagy, inflammation and impaired insulin signalling deficits and resistance have all been associated with PD, PDD and diabetes [44]. In addition to shared biological pathways, whole-genome transcriptome profiling of the substantia nigra of PD patients has provided evidence of genetic links between PD and diabetes [55]. It was found that 892 dysregulated priority genes were altered, and various ‘hub’ genes with multiple interactions with other genes were identified, including those encoding glycogen synthase kinase-3β (GSK-3β) and insulin-like growth factor-1 (IGF-1) receptor (IGF-1R). Furthermore, diabetes was among the top three diseases showing the strongest probable relationship to the top upregulated priority genes [55]. Network-based approaches provide further evidence for a molecular link between diabetes and PD [56]. Interactome map** revealed more than 400 genes linking both T2DM and PD and identified insulin receptor and lipid signalling, activation of the immune response, mitogen-activated protein (MAP) kinase (MAPK) cascade and protein serine–threonine kinase activity as convergent pathways [57].

3.2 PD, Diabetes and Dementia

The link between diabetes and PDD was noted back in 1992 when Sandyk and Awerbuch [58] reported an association between the two conditions in a small study of 12 PD patients. In their study, all five PD patients with diabetes also had dementia while the seven diabetes-free PD patients did not exhibit dementia at the time [58]. Since then the field investigating diabetes and PDD has grown. Bosco et al. [51] reported that PD patients with dementia are two times more likely to be insulin resistant than PD patients without dementia [51].

3.3 Diabetes and Dementia

In addition to lifestyle factors, such as obesity and physical inactivity, increased longevity and population aging are crucial factors contributing to the increasing prevalence of T2DM worldwide [59]. Similar trends are observed in the prevalence of dementia, resulting in co-occurrence of the two diseases [60]. It is now established that there is an increased risk of dementia in patients with T2DM [61, 62]. Diabetes is also linked to less severe forms of cognitive dysfunction, including MCI [63, 64] and subtle cognitive changes reported under the term diabetes-associated cognitive decrements [65]. Diabetes-associated cognitive decrements in adults with T2DM are commonly subtle changes in cognitive function that may be bothersome, but do not affect ADL [65]. These cognitive decrements affect one or several domains, including processing speed, executive function, visual and verbal memory, as well as motor function [66]. It has been suggested that diabetes-associated cognitive decline develops during the pre-diabetic stages, and advances very slowly, over many years [67]. Cognitive decrements do not represent a pre-dementia stage in patients younger than 60 years, whereas older patients may progress to MCI and dementia [65].

An increasing body of evidence suggests that neurodegenerative and vascular dementias share some common underlying pathologies, such as insulin dysregulation, that act both through disease-specific and general mechanisms [68]. In a meta-analysis of 14 studies including over 2 million people, T2DM was associated with a 60% increased risk for dementia of all types, and for vascular dementia the additional risk was 19% higher in women than in men [69]. Radiological signs of vascular brain injury such as lacunes and white matter hyperintensities are common findings in patients with T2DM [70]; however, neuropathologic studies do not confirm increased occurrence of large artery infarcts or microinfarcts in these patients [71]. Also, although the risk for Alzheimer’s dementia is increased in patients with T2DM, neither Alzheimer’s pathology in the brain [72] nor in vivo biomarkers of amyloid β deposition and tau pathology [73, 74] are more common in patients with diabetes than in non-diabetic individuals. These findings may indicate that T2DM accelerates neurodegeneration by other, non-Alzheimer’s-specific mechanisms, which may also contribute in other dementia syndromes, such as PDD. One such mechanism is insulin resistance, which appears in T2DM and neurodegenerative dementias including PD [75, 76]. Insulin resistance is a core sign of T2DM that, besides hyperglycaemia, also contributes in oxidative stress, inflammation, atherosclerosis and hyperlipidaemia. Brain insulin resistance has been defined as the failure of nerve cells to respond normally to insulin, with subsequent disturbances in synaptic, metabolic and immune response functions [77]. Systemic insulin resistance and T2DM are associated with brain insulin resistance and with neurodegeneration; however, it is unclear whether the inter-relation of these conditions depends on mechanistic links or co-occurrence of processes of aging [75].

3.4 Insulin, Insulin-Like Growth Factor (IGF)-1 and Insulin Resistance in Relation to PD

Insulin is thought to have neuroprotective functions in the central nervous system (CNS) [78]. Animal studies have shown that downregulating insulin receptor expression in the rat hippocampus impairs hippocampal plasticity, spatial learning and long-term potentiation [79]. Conversely, insulin treatment has been shown to improve memory and cognition in rats [80] and promote protective effects in rat models of PD [81]. IGF-1 is a neuronal survival factor and is closely related to insulin (Fig. 1). IGF-1 can bind to both the insulin receptor and IGF-1R with similar outcomes, and animal studies have demonstrated that IGF-1 can exert a neuroprotective effect on neurons in in vitro and in vivo models of PD and reduce α-synuclein aggregation [82,83,84]. IGF-1 levels are increased in the serum and CSF of PD patients [85,86,87], possibly as an endogenous neuroprotective response. Lower serum IGF-1 levels in PD patients have been associated with poor executive function [88] and found to predict poor verbal memory performance [89].

Fig. 1
figure 1

Potential neuroprotective mechanisms of antidiabetic drugs in the context of Parkinson’s disease dementia. ChIs block ACh degradation by ChE allowing ACh to bind to either the mAChR or nicotinic receptors resulting in MAP kinase signalling. IGF-1 or insulin bind to the IR activating MAP kinase signalling, promoting cell growth and synaptic plasticity via BDNF, or the PI3 K/AKT signalling pathway, resulting in cell proliferation and survival via modulating multiple downstream targets including mTOR, GSK-3β, FoxO and NFκB. DPP-IV inhibitors block the degradation of GLP-1 by DDP-IV, promoting signalling, while GLP-1 analogues directly promote GLP-1R signalling, which increases cAMP activating of the aforementioned signalling pathways and also activates the PKA signalling pathway resulting in the activation of CREB and resultant gene transcription. Proinflammatory cytokines, which are elevated in Parkinson’s disease, such as TNFα and IL-1 promote pro-inflammatory cytokine release via NFκB signalling and can also inhibit insulin and IGF-1 signalling, resulting in the disruption of MAP kinase and PI3 K signalling. In contrast to this, glitazones can activate PPARγ, resulting in gene transcription of neuroprotective mediators. PGC-1α can bind to various transcription factors including PPARγ and exerts a neuroprotective effect. The insulin degrading enzyme can target α-syn, binding to α-syn oligomers inhibiting the formation of fibrils. α-syn alpha-synuclein, ACh acetylcholine, AKT protein kinase B, ChE cholinesterase, ChI cholinesterase inhibitor, BDNF brain-derived neurotrophic factor, cAMP cyclic adenosine monophosphate, CREB cyclic adenosine monophosphate response element-binding protein, DPP-IV dipeptidyl peptidase IV, FoxO forkhead box O, GLP-1 glucagon-like peptide-1, GLP1R glucagon-like peptide-1 receptor, GSK3β glycogen synthase kinase-3β, IDE insulin degrading enzyme, IGF-1 insulin growth factor-1, IL-1 interleukin-1, IR insulin receptor, mAChR muscarinic acetylcholine receptors, MAP mitogen-activated protein, mTOR mammalian target of rapamycin, NFκB nuclear factor-κB, PI3 K phosphatidylinositol 3 kinase pathway, PGC-1α PPARγ coactivator 1α, PKA protein kinase A, PPARγ proliferator-activated receptor-γ, TNFα tumour necrosis factor-α

Full size image

Peripheral insulin and IGF-1 can cross the blood–brain barrier (BBB) and animal studies have shown that they are also produced in regions of the CNS associated with cognition and dementia, including the hippocampus, cortex and olfactory bulb [76, 90–92]. The insulin receptor is widely expressed in the brain—in the olfactory bulb, cerebral cortex, hippocampus, hypothalamus, amygdala [93] and substantia nigra [76, 94]. It is more concentrated in neurons than in glial cells and is present in post-synaptic neuron terminals [93]. To induce signalling, IGF-1 or insulin bind to the insulin receptor, causing autophosphorylation of the intracellular unit. This causes a signalling cascade resulting in the activation of either the mammalian MAPK/extracellular signal-regulated kinases (ERK) signalling pathway, resulting in cell growth, or the phosphoinositide 3-kinase (PI3 K)/protein kinase B (AKT) signalling pathway for metabolic functions, resulting in cell proliferation and survival, and the synthesis of lipids and proteins [95] (Fig. 1). The activation of these pathways can modulate multiple downstream targets [76, 96] (Fig. 1). This results in the modulation of pathways in PD and PDD, such as apoptosis (mammalian target of rapamycin [mTOR], GSK-3β and forkhead box O [FoxO]), autophagy (mTOR), inflammation (GSK-3β and nuclear factor-κB [NFκB]) and synaptic plasticity (GSK-3β, and cyclic adenosine monophosphate [cAMP] response element-binding protein [CREB]). GSK-3β has been widely researched in neurogenerative disorders, inflammation and cognition as well as diabetes, playing a role in each. Of note, α-synuclein, a key player in both PD and PDD can activate GSK-3β directly and indirectly by inhibiting insulin receptor substrate 1, thereby altering insulin signalling [76, 97–99].

Dysfunctional insulin signalling due to insulin resistance, diabetes or excessive inflammation [100] may result in activated GSK-3β and NFκΒ and inactivated mTOR and CREB signalling, among others, resulting in altered autophagy, cell proliferation and increased inflammation (Fig. 1). Indeed, loss of insulin sensitivity centrally in animal models of Alzheimer’s disease (AD) has resulted in increased formation of AD-related pathology [1). A press release has reported positive data on cognition and apathy, but no results have been published.

5.2 Drugs Primarily Acting on Intracellular Enzymes

5.2.1 Protein Kinase Inhibitors

Masitinib is a non-selective tyrosine kinase receptor (TKR) inhibitor currently being tested against cognitive impairment in PD in a phase II study assessing safety and efficacy (EudraCT: 2010-024424-26; Table 1). The primary target of masitinib is the TKR c-Kit, but it also exerts weak inhibitory activity in fibroblast growth factor receptor 3, lymphocyte-specific kinase (Lck), Lck/Yes-related protein, focal adhesion kinase and the Fyn TKR. It also targets platelet-derived growth factor receptor and is currently being used in the treatment of mast cell (MC) tumours in dogs, and in oncology for the treatment of unresectable or metastatic gastrointestinal stromal tumours [173]. Clinical use of masitinib in non-oncological applications seems reasonable, based on the involvement of c-Kit receptors in MC-mediated inflammatory pathogenesis.

Another biochemical pathway of increasing interest in PD is that of the C-Abelson (C-Abl) tyrosine kinase enzyme. C-Abl is phosphorylated and activated upon mitochondrial dysfunction, resulting in increased oxidative stress and dopamine neuron degeneration supposedly through parkin inactivation, α-synuclein aggregation and impaired autophagy of toxic elements [174]. The C-Abl inhibitor nilotinib, which is used in much higher doses in leukaemia treatment, has been investigated in a small, open-label, proof-of-concept trial on 12 patients with advanced PD or DLB with good results on safety and tolerability during 24 weeks of treatment; however, one patient had a myocardial infarction and two had a corrected QT (QTc) interval prolongation on electrocardiogram [175]. Secondary objectives included the determination of the ability of nilotinib to cross the BBB, and to determine target engagement through measurements of phosphorylation of Abl in CSF. The study had very high coverage by the media when initially presented and a great impact on decision-making among patients and physicians. This led to a great escalation of off-label prescription of nilotinib, which in turn led to call for further, larger studies to include a control arm [176]. A larger phase II study is now underway (NCT02954978; Table 1) and is estimated to be completed by 2020.

5.2.2 Glucocerebrosidase Targeting

There are currently two clinical trials investigating drugs targeting glucocerebrosidase (GCase). Targeting the function of the GBA-encoded enzyme, GCase is a relatively new therapeutic approach in PDD. NCT02914366 is a currently recruiting study on the safety, tolerability and efficacy of ambroxol, a GCase chaperone, in PDD patients (Table 1). It is thought that as ambroxol will increase lysosomal function and has been shown to improve GCase functions it may reduce α-synuclein levels [177]. The study will investigate cognitive and motor scales as well as changes in GCase in lymphocytes and the pharmacokinetics of ambroxol. Another study on PD patients carrying a GBA mutation is also currently ongoing and aims to assess the drug dynamics, efficacy and safety of the substrate inhibitor GZ/SAR402671 (MOVES-PD, NCT02906020; Table 1).

5.2.3 GLP-1 Receptor Agonists

A phase II clinical trial is currently underway evaluating the GLP-1R agonist lixisenatide (NCT03439943; Table 1) as an add-on therapy for early-stage PD patients. However, the primary endpoint for this study is motor improvement with no specific interest in the cognitive aspects of the disease. In addition to this a phase II clinical trial with the GLP-1 analogue semaglutide and newly diagnosed PD patients is planned for early 2019 (NCT03659682; Table 1). In this trial cognitive assessment is also going to be examined among other parameters. Both of these drugs have demonstrated protective effects in preclinical studies [178, 179]. In addition to this the GLP-1 receptor agonist exenatide has been investigated in an open-label, randomised controlled trial of 45 patients with moderate PD [180]. In this study, exenatide was well-tolerated, although weight loss was a common concern. The single-blinded ratings of motor and cognitive performance suggested clinically relevant improvements in the treatment group compared with control. Patients who had previously been exposed to exenatide showed persistent advantage in motor and cognitive performance 1 year after study completion [181]. Following these encouraging results, a phase II randomised, double-blind, placebo-controlled study was conducted including 60 patients with PD who were allocated to exenatide or placebo for 48 weeks, followed by 12 weeks of washout period [182]. The study reported positive results on practically defined off-medication motor scores, which were sustained after the exposure period. However, no significant differences were observed in cognitive performance between exenatide and placebo. This may be due to the short observation period, combined with the mixed population including both early and more advanced PD patients. The observation period may not have been long enough to show significant changes in cognitive outcome [134]. Interestingly, in a post hoc analysis of the Exenatide-PD trial, the authors report improvement in cognition in the subgroups of patients with insulin resistance and those with obesity [183]. Another open-label study of exenatide in PD is currently ongoing in the USA and aims to investigate how the brain and motor behaviour changes in response to treatment (NCT03456687; Table 1). It is also important to note that clinical trials are also evaluating GLP-1 analogues in AD, such as a randomised, placebo-controlled phase II trial that is assessing the safety and efficacy of liraglutide in 206 patients with early AD (NCT018430755). Another trial that aimed to evaluate exenatide in patients with AD or MCI (NCT01255163) was terminated due to insufficient recruitment.

6 Conclusions and Future Perspective

ChIs are being used for the treatment of PDD. However, there is a strong medical need for additional procognitive therapies. Several distinct pharmacological mechanisms are being targeted alone or together with ChIs to counteract MCI and PDD. T2DM appears to be associated with PD and neurodegenerative dementias, possibly through peripheral and cerebral insulin resistance that in turn results in altered autophagy, mitochondrial function, cell proliferation and increased inflammation. Drugs used in diabetes treatment have shown positive effects on neurodegenerative processes and on clinical outcome, regarding memory and cognition, and could, hopefully, be developed into novel therapies against PDD and related conditions. Since these drugs have primarily been developed for the treatment of T2DM, it is likely that a next generation of compounds with more BBB penetrance may be better neuroprotective agents. The fact that the structure of GLP-1R has been solved will likely facilitate the development of non-peptidergic GLP-1 compounds that may turn out to be favourable against neurodegeneration.