Abstract
Background:
Alzheimer’s disease (AD) is characterized by an involvement of brain dopamine (DA) circuitry, the presence of which has been associated with emergence of both neuropsychiatric symptoms and cognitive deficits.
Objective:
In order to investigate whether and how the DA pathways are involved in the pathophysiology of AD, we assessed by in vivo neuroimaging the structural and metabolic alterations of subcortical and cortical DA pathways and targets.
Methods:
We included 54 healthy control participants, 53 amyloid-positive subjects with mild cognitive impairment due to AD (MCI-AD), and 60 amyloid-positive patients with probable dementia due to AD (ADD), all with structural 3T MRI and 18F-FDG-PET scans. We assessed MRI-based gray matter reductions in the MCI-AD and ADD groups within an anatomical a priori-defined Nigrostriatal and Mesocorticolimbic DA pathways, followed by 18F-FDG-PET metabolic connectivity analyses to evaluate network-level metabolic connectivity changes.
Results:
We found significant tissue loss in the Mesocorticolimbic over the Nigrostriatal pathway. Atrophy was evident in the ventral striatum, orbitofrontal cortex, and medial temporal lobe structures, and already plateaued in the MCI-AD stage. Degree of atrophy in Mesocorticolimbic regions positively correlated with the severity of depression, anxiety, and apathy in MCI-AD and ADD subgroups. Additionally, we observed significant alterations of metabolic connectivity between the ventral striatum and fronto-cingulate regions in ADD, but not in MCI-AD. There were no metabolic connectivity changes within the Nigrostriatal pathway.
Conclusion:
Our cross-sectional data support a clinically-meaningful, yet stage-dependent, involvement of the Mesocorticolimbic system in AD. Longitudinal and clinical correlation studies are needed to further establish the relevance of DA system involvement in AD.
Keywords
INTRODUCTION
Alzheimer’s disease (AD) is pathologically characterized by brain accumulation of amyloid-β (Aβ) plaques, neurofibrillary tau tangles, and neurodegeneration [1]. AD is further characterized by the progressive involvement of subcortical neuronal population [2], which produces neurochemical alterations in cholinergic and monoaminergic circuitry [3]. The evidence of a dopaminergic neurotransmission deficit in AD dates back to pioneering postmortem studies, reporting mild histological abnormalities in dopaminergic brainstem nuclei [4], and decreased levels of dopamine (DA), levodopa, and DA metabolites in subcortical and cortical dopaminergic targets [5]. In the nineties, molecular in vivo imaging studies based on positron emission tomography (PET) reported decreased post-synaptic DA receptor expression, particularly D2-like receptor, as well as a reduced transporter activity and expression of tyrosine hydroxylase levels in the major target of the ventro-tegmental area (VTA), i.e., the ventral striatum (VST) of AD patients [6–9]. The role of DA systems degeneration in AD pathogenesis has recently gained a renewed interest [10]. Animal studies suggested that neurons of the dopaminergic VTA might be especially vulnerable in AD (for a review, see [11]). Notably, VTA dopaminergic neurons project diffusely to both subcortical and cortical regions [12], innervating; 1) the VST via the mesolimbic route [13]; 2) the medial fronto-cingulate cortex and orbitofrontal cortex, via the mesocortical route [14, 15]; 3) the medial temporal lobe structures, via the meso-hippocampal route [16]; 4) the amygdala, via the meso-amygdaloid route [10]. These widespread subcortical and cortical VTA projections jointly constitute the Mesocorticolimbic DA system, the dysregulation of which has been also associated with emergence of neuropsychiatric symptoms in AD [17, 18].
So far, few structural and functional magnetic resonance imaging (MRI) studies have assessed the alterations of the Mesocorticolimbic system in AD [19, 20]. These studies have shown significant associations between atrophy in VTA and medial temporal lobe, and cognitive performance in the AD clinical spectrum [19], as well as clinical stage-dependent disconnections within the VTA functional network, showing differences between mild cognitive impairment due to AD (MCI-AD) and dementia due to AD (ADD) [20]. These studies focused on the assessment of neurodegeneration and functional connectivity of the VTA, a small region, with limitations arising from the limited spatial resolution of the adopted neuroimaging techniques. It remains to be determined whether the dopaminergic routes projecting from the VTA are equally vulnerable, i.e., whether subcortical and cortical dopaminergic VTA targets are differentially affected in AD. Additionally, it is not yet fully understood whether VTA projections (i.e., Mesocorticolimbic pathway) are particularly vulnerable in AD compared to substantia nigra (SN) projections (i.e., Nigrostriatal pathway).
Here, with a multimodal neuroimaging approach, i.e. combining structural MRI and 18F-FDG-PET brain metabolic data, we aim to provide a thorough assessment of the alterations in morphology and network topology characterizing the DA Mesocorticolimbic and Nigrostriatal pathway projections, addressing cross-sectionally the prodromal MCI-AD and ADD phases. We hypothesized that alterations of the DA Mesocorticolimbic pathway are present in AD since the earliest disease phases, and possibly associate with neuropsychiatric symptoms. Network measures were estimated with 18F-FDG-PET brain glucose utilization data, employing an emerging methodological approach estimating metabolic connectivity [21]. 18F-FDG-PET data provide crucial information about directionality of brain signaling, since increases in local metabolism reflect an increase in afferent effective connectivity [22]. This approach is based on the assumption that brain energy consumption is influenced by multiple pathological events, including altered neurotransmission [21, 22], and on the evidence of a significant coupling between neurotransmission impairment and integrity of metabolic networks [24].
Our group has previously shown that this approach successfully allows detection of metabolic connectivity alterations within the DA Mesocorticolimbic and Nigrostriatal systems, in both Parkinson’s disease [25] and dementia with Lewy bodies [26].
METHODS
Participants
We retrospectively selected N = 60 patients with ADD, N = 53 subjects with MCI-AD, and N = 54 cognitively-normal healthy control subjects (HC) from the ADNI database (http://adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial MRI, PET, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early ADD. For up-to-date information, see http://www.adni-info.org. All procedures performed in studies involving the ADNI participants were carried out in accordance with the Helsinki declaration. The ADNI study was approved by the Institutional Review Boards of all of the participating institutions. Informed written consent was obtained from all participants and their legal representatives at each site prior to the collection of clinical, genetic, and imaging data.
We included only subjects with both MRI scan performed on a 3T scanner and 18F-FDG-PET scan, cognitive and neuropsychiatric evaluations available and performed within 6 months. As in the ADNI1 study MRI scans were acquired on a 1.5T scanner, subjects/patients enrolled in the ADNI1 study were excluded, hence resulting in a sample limited to ADNIGO and ADNI2 cohorts. As for clinical protocol, ADD patients were enrolled based on NINCDS/ADRDA criteria for probable AD [27]; MCI subjects were enrolled based on evidence of memory impairment and in the absence of significant impairment in other cognitive domains, thus representing amnestic single domain MCI cases (see clinical protocols on http://adni.loni.usc.edu/). Since ADNI enrollment criteria are based on the 1984 NINCDS/ADRDA clinical criteria for AD [27], known to yield some mismatch against definite neuropathological diagnosis [28], we set additional and more stringent inclusion criteria to select our clinical sample. Specifically, an additional inclusion criterion was the availability of an amyloid 18F-Florbetapir-PET scan to establish presence of cerebral fibrillary amyloid-β plaques [29]. We thus selected only MCI-AD subjects and ADD patients with available neuroimaging biomarkers indicating a high likelihood of underlying AD etiology [30, 31], i.e., 1) amyloid positive at 18F-Florbetapir-PET [30, 31], also in accordance with the current research framework for AD [1], and 2) presenting evidence of neuronal injury, as determined by 18F-FDG-PET (i.e., temporal-parietal/precuneus hypometabolism) [30, 31]. Amyloid positivity was determined based on semi-quantification of 18F-Florbetapir-PET images, using a validated cut-off of SUVRCEREBELLUM>1.11, as detailed elsewhere [32, 33]. Presence of an AD-like hypometabolism pattern at 18F-FDG-PET was determined based on an optimized SPM-based procedure, extensively described elsewhere [34], allowing accurate detection of disease-specific hypometabolism patterns not only in cases with overt dementia [34–38] but also in prodromal disease phases [34, 37–40]. As biomarker of neuronal injury, we selected 18F-FDG-PET hypometabolism over MRI atrophy due to the higher accuracy reported for 18F-FDG-PET hypometabolism in identifying ADD patients [average 18F-FDG-PET sensitivity: 0.84, specificity 0.86; average MRI sensitivity: 0.81, specificity 0.75] and predicting progression to ADD in MCI subjects [average 18F-FDG-PET sensitivity: 0.74, specificity 0.76; average MRI sensitivity: 0.73, specificity 0.62], as compared to MRI [41]. This was also done in light of the recent evidence suggesting that MRI hippocampal atrophy might not be a specific marker for AD, as it is also observed in non-AD syndromes with amnestic presentations [42], differently from 18F-FDG-PET AD-like hypometabolism [43]. As for controls, we selected only healthy individuals lacking evidence of underlying neurodegenerative processes at 18F-FDG-PET, based on the well-established association between hypometabolism topographies and neurodegenerative conditions [34, 44] and following from the evidence that an 18F-FDG-PET scan negative for neurodegeneration significantly predicts clinical stability in healthy controls [40]. A demographic summary of the study groups is available in Table 1.
Demographic and clinical features
Table showing demographic summary plus baseline biomarker and clinical data. ADD, dementia due to Alzheimer’s disease; CDR, Clinical Dementia Rating; HC, healthy controls; MCI-AD, mild cognitive impairment due to Alzheimer’s disease; MMSE, Mini-Mental State Examination; NPI, Neuropsychiatric Inventory. §ADD>HC, MCI-AD>HC; #ADD<HC, MCI-AD<HC, ADD<MCI-AD; °MCI-AD>HC; *AD>HC; ¥ADD>MCI-AD, AD>HC. All results were corrected for multiple comparisons with the Bonferroni correction.
Global cognitive and neuropsychiatric evaluations
We relied on Mini-Mental State Examination (MMSE) [45] and Clinical Dementia Rating (CDR) [46] questionnaires to provide a measure of global cognitive status also supporting MCI and dementia clinical diagnosis in the included clinical groups.
Given that we were specifically interested in testing whether alterations of the DA Mesocorticolimbic pathway were associated with presence of neuropsychiatric symptoms, we also considered the Neuropsychiatric Inventory (NPI) [47], so as to obtain a standardized measure of severity/frequency of a wide range of neuropsychiatric symptoms in our clinical groups. The NPI questionnaire is composed of 12 sub-items addressing delusions, hallucinations, agitation/aggression, dysphoria/depression, anxiety, euphoria/elation, apathy/indifference, disinhibition, irritability, aberrant motor behavior, night-time sleep disturbances, and appetite/eating changes. The maximum total score (a×b) obtainable from each neuropsychiatric symptoms sub-item is 12 and the NPI total cumulative score is 144 [47].
Differences across the three groups in demographic, cognitive, and neuropsychiatric data were assessed with an Analysis of Variance (ANOVA), by applying a Bonferroni-type adjustment for multiple comparisons.
MRI pre-processing
Structural MRI (3 Tesla T1-weighted MPRAGE sequence) were downloaded from the ADNI database. Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) was used to generate a study-specific template by aligning the gray and white matter images nonlinearly to a common space. DARTEL procedure consisted in three steps: 1) tissue segmentation of the T1 images for each subject using the segmentation tool of SPM12; 2) construction of a study-specific template of each tissue type in MNI space; 3) non-linear transformation estimation necessary to warp grey and white matter images to match the study-specific template.
An explicit mask for subsequent voxel-level analysis was created by averaging the smoothed, warped, and modulate gray matter segmented maps of the healthy controls, to be then thresholded at >0.1.
Voxel-level atrophy W-score maps estimation
To create maps of statistical deviation of brain volume from the reference control group, we estimated voxel-level atrophy W-score maps [48, 49]. Briefly, w-scores are computed by subtracting, for each individual and each voxel, the measured value with the expected value considering the mean of the healthy controls and nuisance factors, then dividing by the standard deviation of the residuals, according to the following formula:
A detailed description of the W-score approach in neuroimaging analysis is available in [48, 49]. W-scores in the control group have a normal z-like distribution (mean = 0; standard deviation = 1) and thus patients’ W-score maps at the voxel-level show where each voxel value per patient would fall and whether the deviation would be significant or not. W-score atrophy maps were calculated considering the effect of age, sex, and total brain volume (TBV; gray matter + white matter volume), in order to investigate regional effects relative to global atrophy/shrinkage.
Regions-of-interest definition
In order to reconstruct the anatomical neurotransmitter pathways, regions of interest (ROIs) delineating the Mesocorticolimbic (VTA-based) and Nigrostriatal (SN-based) DA targets were selected according to previous literature [15, 50], following a similar methodological approach adopted in previous studies [25, 26]. The Nigrostriatal DA pathway was included as a “control” pathway, for comparison only. More specifically, we selected for the Nigrostriatal pathway: the middle/superior frontal gyri, the supplementary motor cortex, the medial segment of the postcentral gyrus, the medial segment of the precentral gyrus, and the medial segment of the superior frontal gyri. For the Mesocorticolimbic pathway, we selected: the entorhinal area, the hippocampus, the amygdala, the parahippocampal gyrus, the medial orbitofrontal cortex, the lateral orbitofrontal cortex, the subcallosal area, the orbital part of the inferior frontal gyrus, the medial frontal cortex, and the anterior/middle cingulate gyri.
Additionally, the dorsal striatum and VST ROIs were created by splitting caudate and putamen ROIs on the axial plane taking the anterior commissure/posterior commissure line as reference. All the ROIs were bilateral and provided by Neuromorphometrics Inc. (http://www.neuromorphometrics.com/), under academic subscription. The selection resulted in a total of N = 19 ROIs, of which N = 12 ROIs for the Mesocorticolimbic DA pathway and N = 7 ROIs for the Nigrostriatal DA pathway.
Gray matter reductions at the regional level
Regional gray matter reductions were estimated by first extracting regional atrophy W-scores from the above-mentioned ROIs and for all the groups, with the REX toolbox for SPM (http://web.mit.edu/swg/software.htm). Lower W-scores indicated more severe regional atrophy compared to HC. W-scores, corrected for age, sex, and TBV, were then compared across the three groups with a multivariate analysis of covariance (MANCOVA). The resulting statistical models were deemed significant at p < 0.05 Bonferroni corrected for multiple comparisons. Bonferroni correction was applied for post-hoc analysis.
In order to better characterize the clinical significance of these findings, we investigated the relationship between NPI measures (when clinically relevant, i.e., significantly affected in our patients’ cohorts) and grey matter reductions in the pathological cohort (ADD and MCI-AD) by means of non-parametric Spearman correlations. We selected a non-parametric test as NPI scores follow a non-continuous distribution [51].
18F-FDG-PET pre-processing
18F-FDG-PET images were downloaded from the ADNI database and were co-registered to their respective structural MRI scan in the native space with SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Then, warping parameters obtained through the DARTEL spatial normalization were applied to the co-registered 18F-FDG-PET images, which were subsequently smoothed with an 8 mm3 Gaussian filter. This step limits statistical noise and bias due to inter-subject anatomical differences, thus increasing statistical power [52, 53].
After warping and smoothing 18F-FDG-PET images were respectively scaled to their cerebral global mean (CGM), in order to account for between-subject and between-scan uptake variability [54]. This method has been shown to yield higher signal-to-noise ratio compared to the choice of other references regions [55], that might be affected by a number of factors, including cerebrovascular disease (e.g., [56]), brain trauma (e.g., [57]), or aging (e.g., [58]). This procedure provided warped, smoothed SUVRCGM 18F-FDG-PET images.
Metabolic connectivity analysis in Mesocorticolimbic and Nigrostriatal pathways
To evaluate the association among regional metabolism within the DA pathways, we tested the correlation between each possible pairwise ROIs combination considering the effect of all the remaining Mesocorticolimbic network regions. Regional metabolism was estimated by first extracting the average regional 18F-FDG-PET SUVRCGM from the above-mentioned ROIs and for all the groups, with REX toolbox for SPM. We then ran a partial correlation model, separately for HC and symptomatic groups (MCI-AD and ADD). Additionally, correlation coefficients resulting for each pairwise combination were tested for difference among groups by means of a z-test, considering age and gender as nuisance variables. Results are reported at a relatively liberal threshold (p < 0.01) as well as following correction for multiple comparisons using Bonferroni correction (p < 0.05).
RESULTS
Demographics and clinical/neuropsychiatric evaluation
The included participants did not differ in terms of sex (p = 0.190), age (p = 0.601), and education (p = 0.253) (Table 1). As expected, MMSE and CDR scores showed that global cognitive status was significantly, and progressively, deteriorated in MCI-AD and ADD (Table 1). Both symptomatic groups showed higher NPI total and depression scores compared to HC (NPI –Total, MCI-AD p = 0.006, ADD p < 0.001; NPI - Dysphoria/Depression, MCI-AD (p = 0.006, ADD p = 0.02). At Bonferroni post-hoc comparison, MCI-AD group showed significantly higher NPI scores than HC for one item, i.e., anxiety (p = 0.043), whereas ADD group showed higher scores than HC in Apathy/Indifference (p < 0.001), Irritability (p = 0.042), Aberrant Motor Behavior (p = 0.022), and Appetite/Eating Changes (p = 0.011) (Table 1). At last follow-up visit, 33 out of 53 (62.3%) MCI-AD converted to ADD.
Gray matter reductions at the regional level
Nine out of the total 19 selected regions of interest showed significant gray matter volume reductions in MCI-AD and ADD patients, with the net majority (77.7%) involving structures belonging to the Mesocorticolimbic DA pathway. Regions in the Mesocorticolimbic DA pathway indeed presented the most significant differences between patients and HC, with 58.3% of VTA targets presenting significant gray matter reductions (see Table 2 and detailed effect sizes therein). Ventral striatum, orbitofrontal cortex, and medial temporal lobe structures, namely hippocampus, parahippocampal gyrus, entorhinal area, and amygdala, showed the peaking gray matter reductions, with most of the effect already detectable at the MCI-AD stage (see Fig. 1).
ROI-level gray matter reductions
Table showing regions with significant gray matter reductions across the three clinical groups. The statistical threshold was set at p < 0.05 Bonferroni corrected for multiple comparisons. Regions are sorted according to decreasing effect size, as measured by Partial-Eta squared. The MCI-AD<CN, ADD<HC and ADD<MCI-AD columns show the p-values of post-hoc tests, Bonferroni corrected (see Methods). ADD, Alzheimer’s disease dementia; MCI-AD, mild cognitive impairment due to Alzheimer’s disease; HC, healthy controls; M, mesocorticolimbic; N, nigrostriatal.
Box plot showing atrophic regions as expressed by W-score values, obtained from the comparison of patients with healthy control group in the Mesocorticolimbic and Nigrostriatal DA pathways. The horizontal dashed line represents a threshold for significance (W < –1.64, p < 0.05). ACgG, anterior cingulate gyrus; ADD, Alzheimer’s disease dementia; AMY, amygdala; DST, dorsal striatum; EntA, entorhinal area; HP, the hippocampus; LOFC, the lateral orbitofrontal cortex; MCgG, the middle cingulate gyri; MCI-AD, mild cognitive impairment due to Alzheimer’s disease; MFC, medial frontal cortex; MFG, the middle frontal gyrus; MOFC, medial orbitofrontal cortex; MPoG, the medial segment of the postcentral gyrus; MPrG, the medial segment of the precentral gyrus; MSFG, the medial segment of the superior frontal gyrus; OrIFG, the orbital part of the inferior frontal gyrus; SCA, the subcallosal area; PHG, parahippocampal gyrus; SFG, the superior frontal gyri; SMC, the supplementary motor cortex; VST, ventral striatum.
Considering the Nigrostriatal pathway regions, grey matter reductions involved only 28.57% of the SN targets, and limited to the middle frontal gyrus and precentral gyrus medial segment, in both MCI-AD and ADD in comparison with HC, whereas subcortical structures showed no differences (Fig. 1). Severity of grey matter reductions in these regions was relatively mild, as suggested by the small effect sizes (0.08–0.13) (Table 2).
When testing the correlation between NPI measures and grey matter reductions in the Mesocorticolimbic regions, we found that NPI depression sub-scale score correlated with atrophy in the medial orbitofrontal cortex (Rs = –0.22, p = 0.004); NPI anxiety sub-scale score with atrophy in the ventral striatum (Rs = –0.22; p = 0.004); NPI apathy sub-scale score with atrophy in the hippocampus (Rs = –0.26; p = 0.001), parahippocampal gyrus (Rs = –0.24; p = 0.002), entorhinal area (Rs = –0.25; p = 0.001), and amygdala (Rs = –0.25; p = 0.001). In all aforementioned Mesocorticolimbic regions, higher NPI scores were associated with more severe atrophy in both the symptomatic groups (MCI-AD and ADD). All reported correlations survived Bonferroni-correction for multiple comparisons.
Metabolic connectivity analysis
Regions belonging to the Mesocorticolimbic pathway showed significant metabolic connectivity alterations in ADD (threshold, p < 0.01 uncorrected for multiple comparisons). These changes specifically involved 1) multiple VST functional connections, including those to frontal and cingulate regions, i.e., middle cingulate gyrus, lateral orbitofrontal cortex, and orbital part of the inferior frontal gyrus; 2) the parahippocampal gyri connections with medial temporal lobe structures, including the amygdala, the entorhinal cortex, and the hippocampus; and 3) the lateral orbitofrontal connections with the anterior cingulate gyrus. The finding of significant metabolic connectivity alterations between the VST and middle cingulate gyrus and between the VST and lateral orbitofrontal cortex, was also confirmed after Bonferroni correction for multiple comparisons (p = 0.04 and p = 0.05, respectively, see Fig. 2). The MCI-AD group showed a more limited connectivity reconfiguration, with the only significant (liberal threshold, p < 0.01 uncorrected for multiple comparisons) alteration between orbital part of the inferior frontal gyrus and the middle cingulate cortex, not surviving, however, Bonferroni correction for multiple comparisons at p < 0.05 (Fig. 2).
Metabolic connectivity analysis of the Mesocorticolimbic pathway. Figure represents metabolic connectivity changes obtained when comparing partial correlation coefficients between MCI-AD versus HC and ADD versus HC. Yellow edges represent significant metabolic connectivity changes, with ochre nodes for p-values thresholded at p < 0.01, uncorrected for multiple comparison and red nodes for connectivity changes surviving Bonferroni correction for multiple comparisons (p < 0.05). Metabolic connectivity changes are listed below each brain render; only significant metabolic connectivity changes at p < 0.01, uncorrected for multiple comparisons (yellow dotted lines) and at p < 0.05, Bonferroni-corrected for multiple comparisons (red dotted lines) are reported. Brain renderings were obtained using BrainNet [99]. ADD, Alzheimer’s disease dementia; HC, healthy controls, MCI-AD, mild cognitive impairment due to Alzheimer’s disease.
In the ADD group, the Nigrostriatal pathway showed no significant metabolic connectivity alterations. The dorsal striatum, main hub of the Nigrostriatal system, was well connected with the other Nigrostriatal nodes, in both MCI-AD and ADD groups.
DISCUSSION
This cross-sectional study explores the structural alterations and metabolic connectivity changes in the DA Mesocorticolimbic and Nigrostriatal systems in subjects with MCI and dementia due to AD. Within the Mesocorticolimbic DA system, we showed that both MCI-AD and ADD patients had significant tissue loss not only in memory-related medial temporal lobe regions, as expected, but also in other major VTA targets, notably the VST and the orbito-frontal cortices. Regions belonging to the Nigrostriatal pathway exhibited overall less pronounced alterations, with significant gray matter reductions limited to the middle frontal gyri and precentral gyrus medial segment, but no changes in the basal ganglia structures. Metabolic connectivity analyses revealed significant functional changes in the VST connections to the fronto-cingulate regions, but only in the dementia stage. Our data overall provide evidence for a significant, yet phase-dependent, structural and metabolic involvement of regions belonging to the Mesocorticolimbic pathway in AD [48]. Our finding of a more widespread loss of gray matter in regions innervated by the VTA, as compared to the SN, is consistent with recent evidence in animal models of AD, where selective degeneration of VTA neurons and impaired DA outflow to VTA targets, i.e., nucleus accumbens and hippocampus, were reported [11]. The factors underlying the more pronounced vulnerability of regions belonging to the Mesocorticolimbic over the Nigrostriatal pathway remain unclear. Recently, it has been suggested that the selective vulnerability of the VTA and its Mesocorticolimbic targets might be explained in the context of the cholinergic hypothesis for AD [10]. It has long been recognized that cholinergic projections from latero-dorsal and posterior peduncolo-pontine tegmental nuclei make selective synaptic contact with DA tyrosine hydroxylase-positive terminals in VTA [59], where they are critical modulators of dopaminergic activity [60, 61]. Building on this evidence, it is possible that selective vulnerability of the VTA DA neurons described in the literature [11], and of regions belonging to the Mesocorticolimbic pathway described here, might be at least partially ascribed to the progressive loss of cholinergic neurons characterizing AD [10].
This is the first study to adopt a dopaminergic framework to study atrophy and metabolic connectivity in AD, reporting that vulnerability of regions innervated by the VTA is not limited to targets of the mesolimbic/hippocampal route but extends to the orbitofrontal cortices (mesocortical route) and amygdala (meso-amygdaloid route) as well, suggesting a widespread vulnerability of the Mesocorticolimbic pathway. With the exception of the entorhinal area, atrophy of all the aforementioned brain regions already reached a peak in the MCI-AD group, with no further neurodegeneration in ADD (Table 2), pointing at an involvement of these regions in early disease phase.
Among the targets of VTA dopaminergic projections, the hippocampus showed the most pronounced tissue loss in both MCI-AD and ADD cohorts. The hippocampal structures are known to be the locus of severe neurodegeneration in AD and MCI-AD [62], by accumulation of neurofibrillary tangles and diffuse amyloid load, and are the earliest and most severely affected brain regions in typical, late-onset AD [63]. Of note, while the ventral hippocampus receives massive dopaminergic projections from the VTA, the dense dopaminergic receptors in the dorsal hippocampus receive not only by VTA projections, but also by tyrosine hydroxylase-rich afferents from the locus coeruleus [64, 65]. The latter represents an early region of aggregation for AD pathology [66, 67], and thus the relevant role of the alterations of this dopaminergic input to the hippocampus and other cortical structures [68], particularly for the cognitive status [62], should be considered.
Additionally, our data show significant tissue loss in the other major mesolimbic target, i.e., ventral striatum, already in prodromal disease stages. Neurofibrillary tangles in the basal portions of the VST have been observed relatively early in the course of AD, already in Stage IV of Braak’s classification, when deposits are mostly restricted to limbic brain regions only [69]. The VST has a higher density of neurofibrillary tangles than throughout the striatum in AD [70], which is consistent with our observation of a ventral greater-than dorsal gradient of atrophy and with other evidence showing a prevalent involvement of VST in AD pathology [71]. As mentioned in the introduction, both hippocampal formation and VST are innervated by VTA dopaminergic projections, the degeneration of which has been shown to produce lower DA outflow, equally affecting the hippocampus and VST in AD transgenic animal models [11]. We also found that amygdala, the major target of the meso-amygdaloid route, suffered significant gray matter loss in both our MCI-AD and ADD cohorts. Notably, DA is recognized a neuromodulator of both hippocampal and amygdala synaptic plasticity; the DA projections arising from the VTA to the dorsal hippocampus are a major determinant of memory encoding, via an enhancement of long-term potentiation and learning performance [72–74], whereas VTA DA input to the amygdala modulates consolidation of long-term memories [75]. Accordingly, restoration of DA transmission ameliorates memory and learning performances in mouse models of AD [11, 77]. Last, we also found evidence for gray matter loss in cortical targets belonging to the mesocortical route, with significant atrophy in the medial orbitofrontal cortices and pars orbitalis of the inferior frontal gyrus, in both MCI-AD and ADD cases.
We found that severity of gray matter loss within regions belonging to the Mesocorticolimbic pathway was associated with greater severity of neuropsychiatric symptoms. Specifically, depression was linked to gray matter loss within the meso-cortical route, anxiety to gray matter loss within the mesolimbic route, and apathy to gray matter loss within the meso-hippocampal/meso-amygdaloid routes, suggesting that vulnerability of different Mesocorticolimbic targets is associated with specific neuropsychiatric signatures [78]. In this direction, multiple lines of evidence point at the clinical relevance of VTA degeneration (and of its targets) in both AD animal models [10] and humans [19, 80]. In humans, VTA and VST degeneration was associated with both neuropsychiatric symptoms [20, 80] and memory/cognitive dysfunction [19] in MCI-AD and ADD patients. Furthermore, the severity of cognitive impairment in AD has been previously associated with shape abnormalities of the VST [81]. MRI atrophy in the VST was found in both early and more advanced AD stages [82, 83], and with an associated increased risk of clinical progression from MCI-AD to ADD during 2 years of follow-up (Hazard Ratio = 1.59, 95% Confidence Interval: 1.16 to 2.18) [83]. Although the cardinal symptom of typical, late-onset AD pathology is memory impairment, a high percentage of AD patients actually develops neuropsychiatric symptoms during the disease course [84]. Among the neuropsychiatric symptoms, depression and apathy are the most frequently reported symptoms in both MCI-AD subjects and ADD patients [84], and our cohort showed a similar neuropsychiatric profile (Table 1). Presence of neuropsychiatric symptoms is also associated, in cognitively healthy subjects, with an higher risk of developing MCI-AD, and, in MCI-AD subjects, with a higher likelihood of progression to dementia [85–87].
We investigated also metabolic connectivity changes between regions belonging to the Mesocorticolimbic pathway, using an emerging connectivity approach based on metabolic 18F-FDG-PET data. So far, metabolic connectivity studies in AD have focused mainly on the investigation of metabolic connectivity alterations in resting-state brain networks [88–90], and on the modulation of metabolic connectivity by modifiable life factors [91–94]. Altogether, these studies have consistently shown that the default mode network is especially vulnerable in AD [88–90, 95], with involvement of other large-scale networks in atypical clinical presentation [88]. Broadly, these results suggest a multi-factorial modulation of brain connectivity in AD (see [21, 96] for reviews). Here we adopted a different approach specifically aimed at assessing metabolic connectivity alterations in neurotransmission pathways. This novel approach has also been adopted in two previous studies, reporting metabolic connectivity alterations of the DA pathways in both Parkinson’s disease and dementia with Lewy bodies [25, 26], at consistence with the well-known neurochemical imbalance reported in these conditions [21]. Recently, it has been demonstrated that the investigation of metabolic connectivity of the dopaminergic pathways, based on 18F-FDG-PET, yields reliable information on the molecular architecture of dopaminergic pathways, and can provide comparable results to those based on 123I-FP-CIT imaging [97]. Here, we found that most regions within the Mesocorticolimbic pathway showed a preserved functional connectivity in AD. Only in the ADD group, the VST showed metabolic connectivity alterations with the lateral part of orbito-frontal cortex and the middle cingulate gyrus. Notably, the VST represents a crucial hub within the Mesocorticolimbic system, with tight interconnections towards medial and lateral orbitofrontal cortices [50]. On the contrary, we did not observe alterations of the ventral striatal interconnections in the MCI-AD group, suggesting that connectivity changes within the Mesocorticolimbic network are a later phenomenon, likely following ventral striatal neural loss (already detectable in the MCI-AD group) in the AD pathologic cascade. The discrepancy between detected atrophy and metabolic connectivity observed in the MCI-AD phase requires further investigation. Still, it can be noted that the degree of gray matter loss was significant but not particularly severe in Mesocorticolimbic structures, ranging approximately from 0.5 to 1.5 standard deviations below the mean (depending on the ROIs; Fig. 1); additionally, it must be underlined that structural MRI and 18F-FDG-PET measure related but partially independent aspects of neurodegeneration; as a prototypical example, although it is well-established that hippocampus shows significant gray matter loss in AD, it is also well-known that it does not show significant glucose metabolism reductions in AD maps [48]. Although the nature of the discrepancy between MRI and 18F-FDG-PET findings is poorly understood, it is possible that synaptic compensatory mechanisms play a role [98]. Possibly, similar compensatory mechanisms might be at play also in the atrophied regions of the Mesocorticolimbic pathway that do not show alterations in metabolic connectivity in the MCI phase.
In conclusion, this study points at an involvement of the Mesocorticolimbic DA neuromodulatory system in AD, that seems stage-dependent, as emerging from both structural and molecular connectivity data from two groups in the AD continuum at different cognitive stages. Structural and metabolic connectivity alterations in the Mesocorticolimbic pathway in ADD is not limited to VST and hippocampus but encompasses other relevant dopaminergic targets, including the amygdala and the orbitofrontal cortex, with a comparable volume loss along symptomatic disease phases. The lack of metabolic connectivity alterations in MCI-AD suggests that functional changes within the Mesocorticolimbic pathway degeneration are likely to appear in more advanced disease stages, likely following more severe pathology spreading. Molecular PET studies based on radiotracers specific for DA neurotransmission are necessary to definitely demonstrate that dopaminergic neurotransmission impairment is at the basis of the structural and metabolic changes here observed in crucial DA Mesocorticolimbic targets. Longitudinal prospective studies, at difference with the present cross-sectional evidence, are also needed to establish the diagnostic and prognostic relevance of DA system involvement in AD pathogenesis.
Footnotes
ACKNOWLEDGMENTS
This work was supported by the Italian Ministry of Health (Ricerca Finalizzata Progetto Reti Nazionale AD NET-2011-02346784).
Data collection and sharing for this project was also funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (
). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.