Cholinergic muscarinic M₁/M₄ receptor networks in dementia with Lewy bodies

Abstract

Cholinergic dysfunction is central in dementia with Lewy bodies, possibly contributing to the cognitive and psychiatric phenotypes of this condition. We investigated baseline muscarinic M₁/M₄ receptor spatial covariance patterns in dementia with Lewy bodies and their association with changes in cognition and neuropsychiatric symptoms after 12 weeks of treatment with the cholinesterase inhibitor donepezil. Thirty-eight participants (14 cholinesterase inhibitor naive patients, 24 healthy older individuals) underwent ¹²³I-iodo-quinuclidinyl-benzilate (M₁/M₄ receptor assessment) and ^99mTc-exametazime (perfusion) single-photon emission computed tomography scanning. We implemented voxel principal components analysis, producing a series of images representing patterns of inter-correlated voxels across individuals. Linear regression analyses derived specific M₁/M₄ and perfusion spatial covariance patterns associated with patients. A discreet M₁/M₄ pattern that distinguished patients from controls (W_1,19.7 = 16.7, P = 0.001), showed relative decreased binding in right lateral temporal and insula, as well as relative preserved/increased binding in frontal, precuneus, lingual and cuneal regions, implicating nodes within attention and dorsal visual networks. We then derived from patients an M₁/M₄ pattern that correlated with a positive change in mini-mental state examination (r = 0.52, P = 0.05), showing relative preserved/increased uptake in prefrontal, temporal pole and anterior cingulate, elements of attention-related networks. We also generated from patients an M₁/M₄ pattern that correlated with a positive change in neuropsychiatric inventory score (r = 0.77, P = 0.002), revealing relative preserved/increased uptake within a bilateral temporal-precuneal-striatal system. Although in a small sample and therefore tentative, we posit that optimal response of donepezil on cognitive and neuropsychiatric signs in patients with dementia with Lewy bodies were associated with a maintenance of muscarinic M₁/M₄ receptor expression within attentional/executive and ventral visual network hubs, respectively.

We investigated covariance patterns of muscarinic M1/M4 receptors in Lewy body dementia and their correlation with changes in cognition and neuropsychiatric signs after 12 weeks of donepezil treatment. Positive response in cognitive and neuropsychiatric symptoms could be associated with preserved/increased receptors within elements of attention/executive and ventral visual networks, respectively.

Graphical Abstract

Introduction

Dementia with Lewy bodies (DLB) is a common form of dementia in older age, where marked cholinergic dysfunction including reduced choline acetyltransferase is a key neurochemical feature and major contributor to the cognitive, sleep and psychiatric symptoms (Aarsland et al., 2003). In Parkinson’s disease dementia which overlaps with DLB in aspects of clinical phenotype, neuropathology and neurochemistry (Outeiro et al., 2019), several dysfunctional cholinergic networks have been proposed to affect the attention, visuoperceptual and memory domains (Gratwicke et al., 2015).

Acetylcholine neurotransmission is facilitated by both muscarinic and nicotinic receptors. Of the former, there are five muscarinic subtypes (M₁–M₅), which are widely distributed throughout the central nervous system and brain and are considered to play an important role in learning and memory with M₁ implicated, in particular, in regulating cognition (Anagnostaras et al., 2003; Erskine et al., 2019). Neuropathological assessment of the various muscarinic subtypes (M₁-M₄) has been examined in several brain regions in DLB. Studies have shown decreased M₁ expression in temporal, hippocampus and parietal areas relative to controls (Ballard et al., 2000; Shiozaki et al., 1999, 2001), and in the striatum compared to Alzheimer’s disease and healthy cases (Piggott et al., 2003). Increased M₁ in temporal cortex has also been reported (Shiozaki et al., 1999). Elevation of M₂ in anterior cingulate (Teaktong et al., 2005) and M₃ in frontal structures (Shiozaki et al., 1999) as well as decreased M₄ in temporal cortex (Shiozaki et al., 1999) has also been observed in DLB relative to controls. Furthermore, using (R, R) ¹²⁵I-iodo-quinuclidinyl-benzilate (QNB), a muscarinic antagonist whose binding is consistent with the distribution of M₁/M₄ receptors found increased binding levels within the insula, cingulate and claustrum in DLB relative to controls (Piggott et al., 2002, 2003).

Understanding how cholinergic receptors are altered at the network level may have implications for treatment. Cholinesterase inhibitors (ChEIs) are the mainstay of symptomatic treatment in DLB (Lam et al., 2009). However, little effort has been made into the impact and optimization of the use of these drugs, for example in relation to target symptoms and dose. At present, there is still a wide range of response heterogeneity to these drugs in DLB with improvements only occurring in about a third of patients in terms of cognition and activities of daily living (Matsunaga et al., 2015), where the picture is even less clear for behavioural and psychiatric symptoms (Matsunaga et al., 2015). While ChEIs show efficacy in improving neuropsychiatric symptoms, it remains to be established if changes in muscarinic receptor networks play a part in neuropsychiatric symptom response to treatment.

One way to examine function at the systems network level is by multivariate approaches such as spatial covariance, a form of principal components (PC) analysis, which overcomes the concept of functional segregation and provides connectivity information between brain regions. In the present study, our aim was to apply spatial covariance to ¹²³I-QNB SPECT scans, a ligand with high-binding affinity to M₁ and M₄ receptors (Piggott et al., 2002), to investigate muscarinic M₁/M₄ connectivity/networks in ChEI naïve patients with DLB. Specifically, we investigated and derived baseline spatial covariance patterns of M₁/M₄ that firstly, distinguished DLB from healthy individuals; and secondly in patients, were associated with positive changes in global cognition and neuropsychiatric symptoms after 12 weeks of open-label treatment with the ChEI, donepezil.

Materials and methods

Standard protocol approvals, registrations and patient consents

Approval was from the UK Department of Health’s administration of radioactive substances advisory committee (ARSAC) and Newcastle, North Tyneside and Northumberland research ethics committee. Participants gave written informed consent for the study unless they lacked capacity in which case an appropriate consultee (the nearest relative) provided written assent for their participation.

Participants

We included 38 individuals (14 DLB, 24 elderly controls). DLB participants were recruited from a community-dwelling population of patients referred to old age psychiatry, geriatric or neurology services. Cognitively normal controls were recruited from patient spouses, friends and volunteers. Diagnosis was carried out consensually between two consultant old age psychiatrists using the third report of the DLB consortium 2005 diagnostic criteria (McKeith et al., 2005). Of the 14 patients with DLB, 11 were ‘probable’ and 3 ‘possible’. Autopsy was performed on four cases and received clinico-pathological diagnoses using standardized criteria (Table 1) (Braak et al., 2003; McKeith et al., 2005; Braak et al., 2006). At the time of QNB imaging, DLB patients were naïve to ChEI or anti-Parkinsonian treatment. Participants on any of the following medications were excluded from the study: antipsychotic, cholinergic, anticholinergic and antidepressant medications.

Table 1.

Participant characteristics

	Controls	DLB				Statistic, P-value
N	24	14
Sex (m:f)	15:9	7:7				χ2 = 0.6, 0.5
Age	74.1 ± 5.1	74.1 ± 7.1				F(1,36) = 0.0003, 0.9
MMSE	28.3 ± 1.5	15.7 ± 6.2				U = 2.0, <0.001
CAMCOG (max 107)	95.0 ± 3.9	54.7 ± 17.5				W1,13.8 = 72.0, <0.001
CAMCOGmemory (max 27)	22.1 ± 1.9	11.9 ± 5.0				W1,15.2 = 53.6, <0.001
CAMCOGexec (max 28)	20.8 ± 4.2	9.1 ± 6.6				F(1,36) = 44.6, <0.001
NPI	Na	20.5 ± 19.5
NPIhall	Na	4.2 ± 3.7
UPDRS III	0.9 ± 1.5	27.4 ± 14.0				U = 332.5, <0.001
CAF	Na	4.2 ± 3.3
Neuropathologically diagnosed DLB cases (n=4)
Case		1	2	3	4
Braak Tau staging		1	3	3	3
(1 − 6)
Braak LB staging		6	6	6	6
(1 − 6)
McKeith		3	3	3	3
(0 − 3)

Values denote mean ± 1 SD.

MMSE = mini-mental state examination; CAMCOG = Cambridge cognition examination; NPI_hall = neuropsychiatric inventory hallucinations subscale; UPDRS = Unified Parkinson’s disease rating scale; CAF = Cognitive fluctuation scale; Na = Not applicable; LB = Lewy body.

Cognitive function was evaluated with the mini-mental state examination (MMSE) and Cambridge Cognitive examination (CAMCOG) (Roth et al., 1986) tests. We also utilized the memory and executive subscales (CAMCOG_memory, CAMCOG_exec). Parkinsonism was assessed using part III (motor examination) of the unified Parkinson’s disease rating scale (UPDRS) (Fahn et al., 1987). Neuropsychiatric features and cognitive fluctuations were measured with the neuropsychiatric inventory (NPI) (Cummings et al., 1994) and Clinician Assessment of Fluctuation scale (CAF) (Walker et al., 2000). For analyses (see below), we also considered the NPI hallucinations subscale (NPI_{hallucinations}), which was specifically focussed on assessing visual hallucination severity and frequency.

Following their ¹²³I-QNB scan, most patients (n = 12) were then treated with the ChEI donepezil titrated up to the standard daily clinical dose of 10 mg. After a period of 12 weeks, patients underwent repeated MMSE and NPI assessments.

Radiochemistry

Employing the technique of Lee et al. (1996) (R, R) ¹²³I-QNB radiosynthesis was performed, the details of which have been previously described (Pakrasi et al., 2007).

Acquisition

Individuals were scanned with a triple-head gamma camera (Picker 3000XP), 5 h post-injection of (R, R) ¹²³I-QNB using a previously reported imaging protocol (Colloby et al., 2006). Within 4 weeks of the (R, R) ¹²³I-QNB scan, participants also underwent ^99mTc-exametazime regional cerebral blood flow (rCBF) SPECT scanning (Colloby et al., 2006).

Spatial pre-processing

Scans were registered to match their specific ¹²³I-QNB and ^99mTc-exametazime SPECT templates in standard MNI space using the image registration tool FLIRT (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FLIRT, accessed 20 July 2020). Information regarding the template images have been reported elsewhere (Pakrasi et al., 2007; Colloby et al., 2008). The registered images were then smoothed with a 10 mm FWHM 3D Gaussian filter.

Spatial covariance

Figure 1 depicts a simplified schematic of the spatial covariance analysis pipeline. In detail, spatial covariance analysis was simultaneously applied to ‘n’ pre-processed (registered and smoothed) ¹²³I-QNB SPECT scans using covariance software (http://www.nitrc.org/projects/gcva_pca/) (Habeck et al., 2005), capturing the major sources of between- and within-group variation, producing (n – 1) PC images organized in a descending order of attributed variance. For each PC image, voxels had positive and negative weights representing the sign and strength of voxel covariance that remained fixed across subjects. Specifically, positive and negative voxels were interpreted as concomitant increased and decreased M₁/M₄ binding, respectively. The degree to which a subject expressed a PC image (PC₁, PC₂,……., PC_n−1) was by means of the subject-scaling factor (SSF₁, SSF₂,…, SSF_n−1), obtained by multiplying each voxel value of a PC image by the corresponding voxel value in the subject’s QNB scan followed by calculating the total sum of these products to yield a score. Therefore, a high SSF score for a PC image indicates a greater increased binding in voxels with positive weights and a greater decreased binding in voxels with negative weights.

Figure 1.

Analysis pipeline. Simplified schematic of the spatial covariance analysis pipeline.

To derive the QNB spatial covariance pattern (SCP_QNB) that best discriminated DLB from controls; all subject scale factor scores (SSF₁, SSF₂,…, SSF_n−1) were entered into a linear regression model as independent variables with group membership as the dependent measure. Akaike’s information criteria then determined how many SSFs (PCs) should be included in the regression model to achieve optimum trade-off between goodness of fit and model simplicity (Burnham and Anderson, 2002). The set of SSFs (PCs) generating the lowest Akaike’s information criteria value was then chosen as predictor variables for the model that differentiated DLB from controls, where the resulting linear combination and their coefficients formed the composite PC image pattern (SCP_QNB). The extent to which each subject then expressed the SCP_QNB was by the SSF_QNB, calculated similarly by multiplying each voxel value of the SCP_QNB image by the corresponding voxel value in the subject’s QNB scan followed by computing the total sum of these products to produce a score.

Spatial covariance analysis was then conducted on the perfusion ^99mTc-exametazime SPECT scans, primarily to assess whether the muscarinic M₁/M₄ disease pattern (SCP_QNB) differed from the perfusion pattern. Therefore, positive and negative weights of these images were viewed as concurrent relative increased and decreased rCBF, respectively. The analysis revealed an SCP_rCBF that best separated DLB from controls, with subject expression scores (SSF_rCBF).

We then derived a ChEI naïve M₁/M₄ baseline SCP that correlated with ΔMMSE_{rel_b}, describing the percentage change in MMSE between baseline and 12 weeks of post-treatment. Furthermore, we generated ChEI naïve M₁/M₄ baseline SCP’s that correlated with ΔNPI_total and ΔNPI_{hallucinations}, measures that describe changes in global neuropsychiatric and hallucinatory symptoms between baseline and 12 weeks of post-treatment, respectively. This involved performing a separate spatial covariance analysis of the donepezil-treated DLB sample, generating a series of PC images expressed by the SSFs, which in turn were entered into regression models as predictor variables with either ΔMMSE_{rel_b}, ΔNPI_total or ΔNPI_{hallucinations} as the dependent parameter. The resulting linear combinations with the lowest Akaike’s information criteria values defined the SCP_{ΔMMSE_rel_b}, SCP_{ΔNPI_total} and SCP_{ΔNPI_hallucinations}, with individual pattern expressions scores of SSF_{ΔMMSE_rel_b}, SSF_{ΔNPI_total} and SSF_{ΔNPI_hallucinations}, respectively.

Stability and reliability of all SCPs were assessed by bootstrap resampling (1000 iterations), to identify areas that contributed to the patterns with high confidence. This transforms the voxel weights of each SCP into Z maps, computed as the ratio of voxel weight and bootstrap standard deviation. The Z-statistic follows roughly a standard normal distribution where a one-tailed P ≤ 0.05 infers a threshold of |Z| ≥ 1.64 (Habeck et al., 2010).

Statistical analyses

Demographic, cognitive and behavioural variables were tested for normality and variance homogeneity using Shapiro–Wilk and Levene’s tests, respectively. Where applicable, the data were examined using parametric (ANOVA F, Welch’s ANOVA W) and non-parametric (Mann–Whitney U, χ²) tests, and interpreted as significant if P ≤ 0.05. Correlations were examined, where appropriate, with Pearson’s r and Spearman’s ρ coefficients. Analysis was conducted using IBM SPSS v.23.0.0.3.

Data availability

The anonymized data are available to qualified investigators upon request to the corresponding and senior authors.

Results

Demographics and clinical characteristics

Table 1 shows demographic and clinical characteristics of study participants. Groups were similarly matched for gender and age, while as expected, cognitive and clinical variables differed (P < 0.001).

Spatial covariance

Thirty-seven QNB PC images were initially generated from which PC₁, PC₃ and PC₆ formed the baseline SCP_QNB that distinguished DLB from controls (Fig. 2A and B). SSF_QNB scores were standardized so that the control mean/SD were 0/1, respectively, and were significantly higher in DLB than controls (mean ± SD; controls = 0 ± 1.0, DLB = 1.9 ± 1.5; W_{1,19.7 =} 16.7, P = 0.001, Fig. 2C). The baseline disease pattern was characterized by concomitant bilateral preserved/increased M₁/M₄ binding (red regions) in medial/middle frontal gyrus, precuneus, cuneus and lingual gyrus with concomitant decreased binding (blue regions) in right middle/superior temporal gyrus, insula and left caudate regions. Supplementary Table 1 details specific regions contributing to the M₁/M₄ disease pattern. Thirty-seven rCBF PC images were then generated from which PC_1,2,3,4, PC₁₀ and PC₁₁ derived the associated baseline SCP_rCBF (Fig. 2D and E). SSF_rCBF scores were similarly standardized and greater in DLB than controls (controls: 0 ± 1.0, DLB: 6.0 ± 1.6; F(1,36) = 184.6, P < 0.001, Fig. 2F). The rCBF covariance pattern involved bilateral relative increased rCBF (red) in cerebellum, central, lentiform-thalamic, anterior cingulate, insula, amygdala and inferior frontal regions with bilateral relative decreased rCBF (blue) in occipital, temporal and parietal areas. Supplementary Table 2 details specific regions of the rCBF disease pattern.

Figure 2.

Spatial covariance patterns in DLB. Disease M₁/M₄ (A, B) and rCBF (D, E) baseline covariance patterns in DLB projected onto axial and rendered displays of an MRI brain template. Distribution of subject scores (SSF_QNB, SSF_rCBF) across groups, respectively (C, F). Images displayed neurologically (left is left).

Relationship between the disease pattern expressions and age, MMSE, CAMCOG, CAMCOG_memory, CAMCOG_exec, total NPI, NPI_hall, UPDRS III and CAF scores were explored in DLB. Age (r = 0.49, P = 0.04), CAMCOG (r = −0.48, P = 0.04) and CAMCOG_memory (r = −0.52, P = 0.03) correlated with the M₁/M₄ disease pattern expression, however, after controlling for multiple testing (Bonferroni correction), these associations did not remain significant. Trends between the rCBF pattern (SSF_rCBF) and these measures did not yield any significant correlations (|r| ≤ 0.45, P ≥ 0.07).

Data for the donepezil-treated group are presented (Table 2). Eleven QNB PC images were produced from the DLB_donepezil sample. Differences in MMSE were found between baseline and 12-week scores (P = 0.02). The resultant baseline M₁/M₄ pattern (PC₅) that correlated with ΔMMSE_rel__b (r = 0.52, P = 0.05) was generated and termed the ‘cognitive’ response pattern (Fig. 3A). Figure 3B depicts the pattern expression scores (SSF_PC5) plotted as a function of ΔMMSE_rel__b. The covariance pattern consisted of bilateral concurrent preserved/increased M₁/M₄ binding in superior/middle/orbitofrontal gyri, inferior temporal, temporal pole, fusiform and anterior cingulate with bilateral concurrent decreased uptake in superior/inferior parietal, precuneus and basal forebrain, superior and middle temporal gyri. Supplementary Table 3 details specific regions contributing to the pattern. Differences in NPI_total were also observed between baseline and 12-week post-treatment scores (P = 0.008). The M₁/M₄ pattern (PC₄) that was found to be associated with ΔNPI (r = 0.77, P = 0.002) was derived (‘neuropsychiatric’ response pattern, Fig. 3C). Figure 3D presents the pattern expression scores (SSF_PC4) plotted as a function of ΔNPI. The covariance pattern comprised of bilateral concomitant preserved/increased M₁/M₄ binding in fusiform, parahippocampal gyrus, inferior/middle/superior temporal gyri, precuneus, striatum and superior central regions with bilateral concomitant decreased uptake in medial orbitofrontal, superior/middle frontal gyri, inferior central, insula, cuneus, basal forebrain and posterior cingulate areas (Supplementary Table 4). Lastly, differences in NPI_{hallucinations} were observed between baseline and 12-week post-treatment scores (P = 0.04), where the M₁/M₄ pattern (PC_3,5) that correlated with ΔNPI_{hallucinations} was also obtained (r = 0.75, P = 0.003) (‘hallucinatory’ response pattern, Fig. 3E). Figure 3F shows pattern expression scores (SSF_PC3,5) plotted as a function of ΔNPI_{hallucinations}. The covariance pattern involved bilateral concomitant preserved/increased M₁/M₄ binding in fusiform, inferior temporal, precuneus, caudate and left cuneus with bilateral concomitant decreased uptake in anterior and mid-cingulate, medial frontal and superior postcentral gyri (Supplementary Table 5).

Table 2.

Summary data of DLB patients treated with donepezil

	DLBdonepezil	Statistic, P-value
N	12
Sex (m:f)	6:6
Age	73.5 ± 6.5
MMSEba	18.2 ± 4.9
MMSE12wa	22.2 ± 5.3
MMSE12w – MMSEba	4.0 ± 4.0	W = 42.5 (Z = 2.4), 0.02b
ΔMMSErel_b (%)a	25.1 ± 29.0
NPI_10b	20.9 ± 18.5
NPI_1012w	6.1 ± 7.6
NPI12w – NPIb	−14.8 ± 16.8	W = 5.0 (Z = 2.7), 0.008b
NPI_hallucinations_b	4.0 ± 3.8
NPI_hallucinations_12w	1.2 ± 2.5
NPIhall12w – NPIhallb	−2.8 ± 4.1	W = 5.0 (Z = 2.1), 0.04b

Values denote mean ± 1 SD.

_b = at Baseline, _12w = at 12 weeks.

ΔMMSE_{rel_b} (%) = [(MMSE_12w – MMSE_b)/MMSE_b] × 100%.

^a n = 11.

^bRelated samples Wilcoxon signed-rank test.

Figure 3.

Responder patterns in DLB. Baseline cholinergic M₁/M₄ cognitive (A), neuropsychiatric (C) and hallucinatory (E) response patterns in DLB.

Discussion

This was the first study examining the association between muscarinic M₁/M₄ patterns of spatial covariance and improvements in cognitive and neuropsychiatric symptoms with cholinergic therapy in DLB. We undertook a spatial covariance approach of (R, R)¹²³I-QNB SPECT scans, an M₁/M₄ receptor ligand, in ChEI naïve patients with DLB. This methodology offers connectivity information across brain voxels and therefore functions at the network level, with advantages over traditional univariate techniques where brain voxels are treated as independent. A covariance pattern that discriminated DLB from healthy older individuals was identified and implied M₁/M₄ network dysfunction in DLB. We also observed distinct baseline M₁/M₄ covariance patterns that were associated with positive changes in MMSE, NPI_total and NPI_{hallucinations} following 12 weeks of donepezil treatment. These patterns/networks had clear spatial foci suggesting certain cortical regions and their cholinergic innervation may be selectively relevant for cognitive and neuropsychiatric amelioration through cholinergic therapy.

A muscarinic M₁/M₄ covariance pattern was found that distinguished DLB from controls. This disease pattern consisted of concomitant decreased and preserved/increased M₁/M₄ binding in several brain regions. The decreased M₁/M₄ uptake pattern largely converged on the lateral temporal cortex and insula, where studies have shown their roles in language comprehension and episodic memory (Xie et al., 2012; Goldstein et al., 2017). Thus, the pattern suggests a lateral temporal and insula network of M₁/M₄ receptor dysfunction. The disease pattern also showed regions of preserved/increased M₁/M₄ activity within frontal, parietal and occipital lobes that implicate two key brain systems important in the symptomatology of DLB, i.e. the dorsal visual stream (Foley et al., 2015) and fronto-parietal (FP) (Markett et al., 2014) networks. This may highlight an M₁/M₄ receptor role in both vision and attention in DLB, and chimes with the clinical phenotype of visual hallucinations, visuo-perceptual and attentional deficits (Bradshaw et al., 2006), all of which show response to ChEI (McKeith et al., 2004).

The rCBF disease pattern, that was distinct from the associated M₁/M₄ pattern, comprised bilateral relative decreases in occipital, temporal and parietal areas with relative increases in cerebellum, striato-thalamic, limbic, inferior frontal and motor regions in DLB. Regions of relative reduction appear to mainly involve elements of the default mode network (DMN) (precuneus, inferior parietal, lateral temporal cortex) (van den Heuvel and Hulshoff Pol, 2010; Buckner, 2013), where theories infer their contribution to cognitive decline (Seeley et al., 2009). In addition, the relative decreased occipitoparietal perfusion pattern is a characteristic functional deficit that distinguishes DLB from Alzheimer’s disease and healthy individuals (Watson and Colloby, 2016). Our previous fMRI study in DLB showed alterations to FP rather than DMN resting-state networks (Peraza et al., 2015), with FP connectivity strongly linked to cognitive fluctuations (Franciotti et al., 2013; Peraza et al., 2014). Similarly, we also found relative sparing of the DMN in a more recent fMRI study of DLB (Schumacher et al., 2018). However, these studies were in mild disease whereas in the present study, patients were in the moderate/severe stage of illness (mean MMSE: previous 23.6, 22.0; present 15.7). Inconsistency between results could be methodological, or that FP network modulation is perhaps an early feature of this condition, while changes to the DMN emerge later in the course of the disease due to possible increased beta-amyloid and tau co-pathology.

After controlling for multiple testing, no significant correlations were found between the disease pattern expressions (M₁/M₄, rCBF) and cognitive/neuropsychiatric variables in DLB. These results were consistent with our previous findings in Parkinson’s disease dementia where we similarly observed no correlations between the disease M₁/M₄ pattern expressions with cognitive and neuropsychiatric symptoms (Colloby et al., 2016). The lack of such correlations could be the small sample size, or that covariance disease patterns derived from combined control-patient cohorts characterize a number of composite networks that are less sensitive to specific cognitive and clinical symptoms.

In DLB, we derived an M₁/M₄ covariance pattern that positively correlated with improvements in MMSE following 12 weeks of ChEI. This cognitive response pattern showed relative decreases in lateral temporal and parietal regions along with relative preserved/increased binding in prefrontal, temporal pole, fusiform and anterior cingulate. From a network perspective, there were covariant preservation/upregulation in areas converging on key components of the FP (lateral/medial prefrontal), semantic (Tsapkini et al., 2011) (temporal pole), visual (Grill-Spector et al., 2017) (fusiform) and salience (anterior cingulate) networks. Notably, the FP network underpins attention and executive function and these are key cognitive domains which demonstrate improvements in DLB with ChEI treatment (Meng et al., 2019). Therefore, the results may imply that a relative M₁/M₄ receptor maintenance within some elements of attention/executive/visual/salience networks are pre-requisite for ChEI treatment response, and could point towards the potential relevance of these networks and their cholinergic innervation as a predictor of positive treatment response to cognition in DLB.

We also identified an M₁/M₄ covariance pattern that correlated with a decrease in the total NPI score. This ‘neuropsychiatric’ response pattern showed relative decreases in lateral/medial prefrontal, cuneus, insula, basal forebrain and posterior cingulate regions with relative preserved/increased uptake in lateral/medial temporal, precuneus and striatum. In terms of networks, the covariant preserved/increased uptake appears to involve a bilateral temporal-precuneal-striatal network that might implicate elements/hubs of the ventral visual stream, DMN and frontostriatal circuits. As an adjunct to the neuropsychiatric pattern and that visual hallucinations are one of the core features of DLB, we generated an M₁/M₄ covariance pattern that correlated with changes in NPI_{hallucinations}, revealing a pattern which was associated with decreased visual signs. This more specific pattern (contained in entirety within the broader, distributed neuropsychiatric pattern described above) showed bilateral preserved/increased binding in fusiform, inferior temporal, cuneus/precuneus and caudate with bilateral decreased uptake in anterior/mid-cingulate, frontal and postcentral areas. The concomitant preserved/increased pattern suggests contributions from primary visual, visual association and DMN regions, i.e. an inferior temporal-occipito-parietal network. Notably, a recent study showed an association between visual hallucinations and decreased metabolic connectivity in the right lateral temporal and bilateral fusiform regions, outlining their potential impact on visual symptoms in DLB (Iaccarino et al., 2018).

The present findings support the conjecture that a number of brain regions are central to the neuropsychiatric profile of DLB, i.e. fusiform, inferior temporal gyrus, precuneus and striatum, and that cholinergic maintenance of these structures is a precondition for ChEI treatment response of neuropsychiatric symptoms, particularly visual hallucinations in DLB. The underlying neuropathology of neuropsychiatric symptoms in DLB and neurodegenerative disorders in general is poorly understood. While ChEIs do improve neuropsychiatric symptoms, they have modest effects on psychotic symptoms including delusions and hallucinations (Armstrong and Weintraub, 2017). Results from the current study suggest changes in M₁/M₄ receptor networks may play a role in neuropsychiatric symptom response to treatment. In line with symptoms of hallucinations and delusions occurring in psychotic disorders without dementia, there is increasing evidence that dysregulated M₁/M₄-mediated modulation of glutamate and dopamine in frontal-temporal-striatal regions contribute to the development of these symptoms (Erskine et al., 2019). Treatment options beyond ChEIs for neuropsychiatric symptoms are very limited, with antipsychotics being associated with significant risks (Armstrong and Weintraub, 2017). Our findings support ongoing development of M₁/M₄ agonist drugs to treat these symptoms in DLB (Erskine et al., 2019).

Finally, we noted significant overlap between regions associated with both cognitive and neuropsychiatric improvement, i.e. inferior temporal gyrus, temporal pole, fusiform and putamen, and aligns with the known multi-symptom benefits of cholinergic treatment in DLB (McKeith et al., 2000). Overall, this observation suggests that these regions, in particular, may have a core role (from an M₁/M₄ perspective) in the clinical phenotype of DLB and relevance as important therapeutic targets for designing outcome measures in trials and for clinical assessment in practice.

Some limitations of the study warrant discussion. First, the study had a small sample size and therefore results should be interpreted as tentative. Small samples could also explain the lack of correlations observed with the baseline behavioural and clinical symptoms. Second, given the historical nature of the dataset, MoCA assessments were not widely in use at that time of the study. However, MMSE has shown consistent benefits in ChEI trials whereas MoCA remains unproven (Birks and Harvey, 2018). Third, the SPECT ligand used (¹²³I-QNB) is non-selective for muscarinic receptor subtypes (i.e. M₁ vs. M₄) and hence there is uncertainty regarding which receptor subtype contributed to the pattern of changes observed in specific areas of these networks. Both receptor subtypes have been shown to be associated with neuropsychiatric symptoms in DLB, albeit with non-selective ligands (Ballard et al., 2000; Teaktong et al., 2005). While a recent in vivo PET imaging study using a new M₄ selective ligand, ¹¹C-MK-6884, showed 20–50% reduction in receptor binding in frontal and temporal cortices in patients with moderate to severe Alzheimer’s disease (Wang et al., 2019). This suggests there is reduced acetylcholine release and potentially selective changes in cortical M₄-mediated signalling in dementia, consistent with similar cortical muscarinic receptor reductions observed in Alzheimer’s disease with the non-selective ligand ¹²³I-QNB (Pakrasi et al., 2007). These findings suggest the pattern of cortical and subcortical changes that we observed in DLB may be driven by changes in both M₁ and M₄ receptor subtypes. Fourth, given the inclusion of severe DLB cases in this study, we cannot exclude the possibility that some patients will have significant mixed pathology, which in turn could affect the results. Although reassuringly, of the four cases that did have autopsy confirmation of their diagnoses, none appeared to have significant Alzheimer’s disease co-pathology. Finally, the investigation on the treatment-related changes had no placebo comparison group and replication of this study with newer selective M₁ and M₄ receptor PET ligands along with neuropsychological assessments, which align more with the cognitive deficit profile of DLB, may provide more specific cholinergic receptor response patterns. The study also had strengths including clinical assessment of DLB patients, muscarinic SPECT scans conducted prior to ChEI treatment and follow-up assessments from the effects of 12 weeks of ChEI treatment. Autopsy confirmation of diagnoses was available for a minority of cases confirming clinical diagnoses, though not for the entire patient sample. It is also worth noting that since responder M₁/M₄ covariance patterns correlated with changes in MMSE/NPI scores, and because multiple factors may affect both cognition and neuropsychiatric symptoms at baseline in DLB, the observed changes were assumed specific to cholinergic modification.

In summary, we infer a number of dysfunctional M₁/M₄ receptor and perfusion networks associated with DLB. The relevance of these networks may be important in terms of their contribution to clinical features of DLB, in particular, the attentional deficits and hallucinations. The use of ChEIs could improve the symptoms; but there is marked heterogeneity between DLB patients in response to these agents and impractical to reliably predict on clinical grounds who might respond to these drugs. Although speculative and in a small sample, we identified patterns that suggest those with a pre-existing M₁/M₄ receptor maintenance of ‘attentional/executive’ and ‘visual’ networks may experience cognitive and neuropsychiatric symptomatic improvement with ChEI treatment, respectively. These findings provide further insights into therapies targeted at improving cholinergic neurotransmission (mediated via muscarinic M₁ and M₄ receptor agonists) and treatment outcomes in DLB.

Funding

Medical Research Council UK [grant number G9817682]. National Institute for Health Research Dementia Biomedical Research Unit at Cambridge University Hospitals National Health Service Foundation Trust and the University of Cambridge. The National Institute for Health Research Newcastle Biomedical Research Centre in Ageing and Chronic Disease and Biomedical Research Unit in Lewy Body Dementia based at Newcastle upon Tyne Hospitals National Health Service Foundation Trust and Newcastle University. Industry grant from Sosei Heptares, Cambridge, UK.

Competing interests

S.J.C. reports no disclosures. P.J.N. is an employee of Sosei Heptares and holds shares in the company. I.G.M. has been a consultant for GE and Bayer Healthcare. G.B. is an employee of Sosei Heptares. J.T.O’B. has been a consultant for GE Healthcare, Lilly, Bayer Healthcare, TauRx, Axon, Eisai and Roche. J.-P.T. has been a consultant of Kyowa Kirin and Sosei Heptares, received grant funding from Sosei Heptares and honoraria for talks from GE Healthcare.

Glossary

CAMCOG

Cambridge Cognitive examination

ChEI

cholinesterase inhibitor

DLB

dementia with Lewy bodies

DMN

default mode network

FP

fronto-parietal

MMSE

mini-mental state examination

MNI

Montreal neurological institute

NPI

neuropsychiatric inventory

PC

principal components

QNB

iodo-quinuclidinyl-benzilate

rCBF

regional cerebral blood flow

SCP

spatial covariance pattern

SPECT

single-photon emission computed tomography

SSF

subject scale factor