Interaction of systemic oxidative stress and mesial temporal network degeneration in Parkinson’s disease with and without cognitive impairment

Parkinson’s disease (PD) is the second most common neurodegenerative movement disorder among the elderly and also presents as a spectrum of cognitive dysfunction, ranging from PD with normal cognition (PDN) to PD with mild cognitive impairment (PDMCI) to PD with dementia (PDD) [1]. PDMCI is an early form of neurodegeneration that carries a risk of further degeneration into dementia. It is now known that early cognitive deficits constitute an important issue for diagnostic, therapeutic, and prognostic factors in PD [1]. In previous studies, dementia in PD has been demonstrated to be associated with widespread gray matter (GM) atrophy [2], especially in the mesial temporal lobe (MTL) [3]. Although the brain is less affected in cases of PDMCI than dementia, the initial atrophy patterns seen in PDMCI can be used for dementia prediction [3, 4].

In addition to structural changes, alterations of both dopaminergic and non-dopaminergic transmitter systems in the PD brain have also been found to modify several distinct functional networks underlying the different cognitive impairments [3]. The functional network disruptions assessed with “resting state” or intrinsic connectivity fMRI has shown distinct patterns of global network disruption or disruption within specific networks in PD in recent studies. Different network connectivity changes affect different types of cognitive impairment in PD Different network connectivity changes result in different types of cognitive impairment in PD [5‐8]. Progressive cognitive decline in PD is associated with altered resting status functional connectivity in multiple brain regions, including the MTL [9]. Several studies have also discussed dopamine-dependent functional connectivity disruptions and other network alternations in PD [10]. Furthermore, a previous review study has shown worsening memory storage deficits associated with worsening MTL atrophy and further damage to the MTL network with the progression from PD to PDD. Another network implicated in memory is the basal nucleus of Meynert (BNM) cholinergic network. The volume of the BNM cholinergic network degenerates significantly in PDD, which correlates with progressive electrocortical depression in magnetoencephalography [3].

In the large-scale study of human brain networks by Seeley et al., it was reported that spatial disease atrophy patterns reflect the healthy brain’s intrinsic functional network architecture. These findings support the network neurodegeneration hypothesis of Seeley et al. [11] and suggest that human neural networks can be defined by synchronous baseline activity and selective vulnerability to neurodegenerative illness. However, while these functional networks are associated with cognitive dysfunctions, the specific factors that affect the vulnerable areas, and their locations and contributions to the degeneration from mild cognitive impairment to dementia in PD are still undetermined.

Systemic oxidative stress is an important etiology of the neuroinflammation seen in PD and is associated with further progression of the disease via various pathways, such as blood-brain barrier (BBB) dysfunction and the infiltration of peripheral immune cells and circulating cytokines [12, 13]. The infiltration of peripheral immune cells, especially lymphocytes and monocytes, which can mediate apoptosis [14], is one of the interactive pathways of systemic oxidative stress and neuroinflammation. The migration of lymphocytes to the central nervous system (CNS) through the BBB is dependent on lymphocyte function-associated antigen 1 (LFA-1) [15]. This association has been further been demonstrated by the increased peripheral leukocyte apoptosis with striatal dopamine neuron loss [16] and white matter alteration observed through the use of diffusion tensor imaging [17]. In addition, increased oxidative stress has been found to be associated with cognitive function status even in the early stage of the disease in young PD patients [18]. The accumulating evidence has shown an association between inflammation and cognitive impairment not only in PD [19] but also in other neurodegenerative diseases such as AD [20] and even in changes due to normal aging [21]. Peripheral immune cells, which reflect the level of systemic oxidative stress, might play important roles both in disease progression and cognitive function deterioration in PD.

The main focus with respect to peripheral immune cells was on their role in dopamine neuron loss in the substantia nigra (SN) or striatum in previous studies. Although the peripheral immune cells are associated with the initial pathogenesis of PD [22], it is still not known whether this association is widely spread in a non-specific manner to all brain tissues or if it is only focused on particular regions, i.e., the so-called vulnerable areas. Some recent studies have demonstrated the correlational relationships between structural alterations of various areas and different peripheral oxidative stress markers [17, 18, 23]. Moreover, two mediation studies by our team revealed the interactions between altered brain structures and elevated serum oxidative stress [24, 25]. A related important question that has yet to be clearly answered by existing research is whether peripheral inflammation affects these vulnerable areas and, if so, whether it disrupts cognition-related networks, thus causing subsequent downstream degradation and functional impairment. In any case, the identification of any highly sensitive areas might be helpful in the early detection of cognitive deficits, in treatment monitoring, and in dementia progression prevention.

In the present study, we used voxel-based morphometry (VBM) and resting state functional MRI (rs-fMRI) to evaluate the effects of systemic oxidative stress on brain morphology and functional network alterations, respectively, in PD patients with different cognitive statuses. More specifically, the study’s aims were (1) to evaluate the differences in systemic oxidative stress and GM volume changes in PD patients and healthy controls, (2) to investigate the vulnerable areas and associated functional networks in the brain that are highly sensitive to systemic oxidative parameters, and (3) to evaluate the relationships between cognitive impairment and the structural and functional integrity of the networks of the vulnerable areas.

This prospective study enrolled 41 patients who were diagnosed with idiopathic PD from December 2010 to May 2015 according to the United Kingdom Brain Bank criteria [26] by an experienced neurologist from our hospital. For comparison, 29 sex- and age-matched healthy volunteers were recruited as a normal control (NC) group. All the participants in both groups had no other history of neurologic or psychiatric illness. All the evaluations for the PD patients, including the evaluations of clinical disease status, the MRI studies, and the neuropsychological tests, were initially assessed in the OFF-medication state achieved by the withdrawal of dopaminergic medications 12 to 18 h before testing. The Chang Gung Memorial Hospital Ethics Committee approved the study, and all of the participants provided written informed consent.

Each patient’s disease severity and cognitive functional status were evaluated using the Unified Parkinson’s Disease Rating Scale (UPDRS), the modified Hoehn and Yahr Staging Scale, the Schwab and England Activities of Daily Living Scale, and the MiniMental State Examination (MMSE). Neuropsychological evaluations of five cognitive domains (i.e., attention and working memory, executive, language, memory, and visuospatial) were conducted using subtests from the Cognitive Ability Screening Instrument [27] and the Wechsler Adult Intelligence Scale-III [28].

All the patients were classified as having PDN, PDMCI, or PDD. PDMCI was defined according to the Movement Disorder Society Task Force Guidelines [1]. Cognitive impairment was defined as a score 1.5 SD below the normative mean in each of the domains [29]. PDN was defined as less than two domains of cognitive impairment. PDMCI was defined as one score at − 1.5 SD in each of two or more domains but without dementia. PDD was defined as impairment in more than one cognitive domain with an MMSE score of less than 26 [30]. The analyses of the five cognitive function domains were used with the average Z-score of all the subtest scores in each domain.

For each subject, an MRI scan was performed using a 3.0 Tesla whole-body GE Signa MRI system equipped with an eight-channel head coil. A 3D high-resolution T1-weighted anatomical image was acquired using an inversion recovery fast spoiled gradient-recalled echo pulse sequence (TR/TE/inversion time = 9.5/3.9/450 ms; flip angle = 20°; FOV = 256 mm; matrix size = 512 × 512). A resting state functional image was also acquired using 300 contiguous echo planar imaging whole-brain functional scans (TR  =  2 s, TE  =  30 ms, FOV  =  240 mm, flip angle 80°, matrix size 64 × 64, thickness  =  4 mm). During the resting experiment, the scanner room was darkened and the participants were instructed to relax, with their eyes closed, without falling asleep. All patients were in the OFF-dopaminergic-medication state during the MRI scans.

T1-weighted structural MRI data were analyzed using VBM and the Statistical Parametric Mapping software program (SPM12, Wellcome Institute of Neurology, University College London, UK, http://​www.​fil.​ion.​ucl.​ac.​uk/​spm/​) with default settings. All native space T1-weighted structural MRI scans were segmented into GM, white matter (WM), and cerebrospinal fluid (CSF) components. During normalization, the GM and WM images were affine-registered to the tissue probability maps in the Montreal Neurological Institute standard space. The resulting GM tissue segment images were transformed by the DARTEL registration procedure and were smoothed with a Gaussian kernel of full width at half maximum of 8  mm. The final GM images were analyzed within the framework of a full factorial design, whereas ANCOVA was performed with age, sex, and total intracranial volume (TIV) as covariates to investigate the gray matter volume (GMV) differences between the PD and control groups. The resulting statistical inferences were considered significant if the cluster-level family-wise error corrected P value was < 0.05. For the correlation analyses, whole-brain voxel-based multiple regression analysis with age and gender controlled for was used to identify the GM vulnerable areas associated with significant oxidative stress markers in PD.

The rs-fMRI data were preprocessed using the SPM software program and the Data Processing Assistant for Resting-State fMRI software program. After normalization and smoothing by SPM, the waveform of each voxel was finally used for removal of the linear trends of time courses and for temporal band-pass filtering (0.01 to 0.08 Hz) [31]. For network construction of the vulnerable areas associated with oxidative stress, the GM regions previously identified as vulnerable to any significant levels of oxidative stress markers were set as seeds. The voxel-wise FC analyses of the four groups were performed by computing the temporal correlations between the mean time series of each seed and the time series of each voxel within the whole brain. The correlation coefficients of each voxel were normalized to z-scores with Fisher’s r-to-z transformation. A head movement parameter, the mean frame-wise displacement (FD), which indexes the volume-to-volume changes in head position, was also assessed for each participant. There was no significant difference between the NC and PD groups or among the PD subgroups.

To test whether the changes in FC patterns were associated with the underlying structural atrophy, the GMV values of the four groups were calculated under the intrinsic connectivity networks of the healthy controls connecting to the vulnerable areas. ANCOVAs with age, sex, and TIV (only for GMV) as the covariates were then used to compare the groups in terms of the mean GMV and functional connectivity correlation coefficient (fc-CC) values of the network area. The intergroup differences in the seed-based functional connectivity to the vulnerable areas were tested using the voxel-wise two-sample t test embedded in SPM. Analyses of the correlations between the fc-CC and GMV of the network and the correlations among the fc-CC, disease severity, and cognitive impairment were also performed, with the former analyses controlling for TIV and the latter analyses controlling for age and sex. The cluster size was determined over 1000 Monte Carlo simulations using the AlphaSim program distributed with the REST software tool (http://​resting-fmri.​sourceforge.​net/). A corrected significance level of P = 0.01 was obtained by a combined threshold of P = 0.01 for each voxel and an extent threshold of 43 voxels (cluster size > 1161 mm³). The threshold for statistical significance was set at P < 0.05.

In the comparison between the PD patients and healthy controls, the PD patients presented with higher levels of systemic oxidative stress, including increased apoptosis percentages and increased LFA-1 values. Compared with the NC group, the PD patients had significantly lower GMVs with diffuse atrophy patterns, especially the PDD patients. The MTL atrophy was used to identify the MTLs as the vulnerable areas associated with lymphocyte apoptosis, indicating their relationship with systemic oxidative stress. In addition, we also found deterioration of the structural and functional network integrity of the MTL network, a finding which further associated the network with the progression of disease severity and cognitive deficits. These associations may reflect the possible key role of lymphocyte apoptosis in the GMV atrophy and functional connectivity alterations seen in the development of PD.

Increased apoptosis of lymphocytes and monocytes and increased expression of LFA-1 might support the view that there is only one primary pathophysiology of PD: abnormal systemic oxidative stress and the acceleration of peripheral immune cell infiltration into the brain. Neurodegeneration in PD associated with systemic inflammation via various pathways has been thoroughly discussed [12, 13, 16‐18, 24, 25]. Several previous studies showed that higher percentages of peripheral apoptotic lymphocytes [32] and the CNS infiltration of lymphocytes and monocytes contribute to neurodegeneration [22, 33, 34]. The activation of LFA-1 has been shown to mediate the recruitment of peripheral lymphocytes and monocytes into the PD brain [22]. Although researchers have faced some difficulty in verifying the direct interactions between systemic oxidative stress and neuroinflammation in PD subjects, the infiltration of peripheral immune cells into the SN of animal PD models [13] and the association between leukocyte apoptosis and striatal dopamine neuron loss [16] constitute indirect forms of evidence provided by previous studies. Current study may further provide an indirect evidence of peripheral inflammation communicating with the brain with immune cells in PD.

In terms of comparisons among the PD subgroups, our data revealed that the PDN group had a higher level oxidative stress than the other PD subgroups. At the same time, the trajectory of oxidative stress presented double peaks in the PDMCI and PDD subgroups (Table 1 and Additional file 1: Figure S1). To the best of our knowledge, there have been no previous studies discussing the changes in oxidative stress in PD in terms of different cognitive functions. The trajectory of oxidative stress seen in the present study, however, was similar to that of the hypothetical model of a dual peak of microglial activation in Alzheimer’s disease proposed by Fan et al. [35], wherein ramified microglia transform into anti-inflammatory and pro-inflammatory phenotypes. Dynamic changes to the peripheral immune system in mild (early) and severe (late) cognitive status might also reflect the progression of neurodegeneration in a non-cell-autonomous fashion with respect to neurons [35]. Although we could not verify the phenotypes of leukocytes in the present study, the higher immune response seen in PD without dementia might suggest a stage with higher levels of CNS pro-inflammatory phenotypes, such as increased apoptosis percentage and LFA-1 values.

Although GMV was diffusely and significantly reduced in all the PD patients relative to the normal controls in the present study, only the GMV reductions in the bilateral MTLs were significantly associated with peripheral lymphocyte apoptosis accompanied with cognitive impairment. The high susceptibility of the MTL suggests its role as an important area in maintaining cognitive ability in PD. The cell-autonomous factors of neuron govern both for and against the influence of peripheral immune cells [36], might determine the different sensitivity over different brain regions. The MTL is known as a very sensitive area for chronic stress and hypoxia conditions [37‐39] and is among the brain areas most easily affected in Alzheimer’s disease [40]. The neurons in the MTL have large cell surfaces and high-energy requirements, which makes these neurons more prone to receiving multiple forms of oxidative stress [40]. Our whole-brain VBM-based correlation analysis with lower presumptions further highlights the reliability of the observed findings.

We found that the topographic distribution of the MTL functional network derived from the healthy subjects was partially overlaid with the key components of the cholinergic pathways from the BNM network to the frontal and temporal cortices [3]. The BNM network is responsible for maintaining fronto-executive and memory functions [41]. Relatedly, decreased cholinergic neurons in the frontal and temporal lobes are believed to be associated with the cognitive impairments seen in PDMCI and PDD [3]. Alterations of MTL intra-network functional connectivity between the MTL and MOL were also found. Meanwhile, attention and memory impairments were observed to be associated with decreased functional connectivity between the MTL and MOL. Interestingly, the MTL-pons connectivity in the PDMCI subgroup was lower than that for the other PD subgroups. The cause of this presentation is still uncertain and requires further investigation and research.

We also found progressive underlying structural deficits of the MTL network in PDN, PDMCI, and PDD. Those temporo-spatial degradation results reflect the functional and anatomical proximity of the MTL network to presumed disease vulnerable areas, and highlight the value of systemic lymphocyte levels in the evaluation of the network degenerative mechanisms in PD. According to the network degeneration hypothesis [11], the different pathological molecules, such as mis-folded proteins, target different specific brain vulnerable areas and their networks. These selective forms of neuronal vulnerability may be related to the weakening of synaptic convergence zones, the loss of retrograde growth factor supplement, the transsynaptic spread of mis-folded protein, and other factors related to network destabilization [11]. The concordance between disease-related atrophy and healthy intrinsic functional connectivity reflects the network-driven neuronal vulnerability. However, the accumulation of mis-folded alpha-synuclein in Lewy bodies and Lewy neurites, a major hallmark of PD [42], was not evaluated in the present study. Its effects and interactions with systemic inflammation in the MTL were thus not determined, and a further validation study should therefore be conducted.

This study has some limitations. First, it was a small-sized and cross-sectional study, which limits any interpretations that can be made from it regarding the trajectory of oxidative stress markers, the structural or functional alterations as the disease progresses, and any causal relationships among the above parameters. Second, multiple comparison correction was not applied for the correlational analyses due to the small sample sizes after subgrouping, the exploratory nature of the study, and the possibility of overcorrection owing to the high collinearity of the studied clinical and cognitive parameters. As such, the correlational analyses might be underpowered, and the significance of the associated results should be interpreted cautiously. Third, we did not verify the phenotypes of peripheral immune cells. As such, only successful therapeutic trials targeting those oxidative stress makers, their upstream causes, or their downstream effects will confirm that PD are indeed caused by processes related to immune cell infiltration. Manipulations of the systemic immune system and mis-folded proteins and evaluations of their effects in PD animal models might help to verify the role of the MTL in cognition in the future.

In terms of cognitive impairment, this study demonstrates that the neuronal vulnerable areas associated with lymphocyte infiltration are located in the bilateral MTLs, with such infiltration resulting in the disruption of both architectural and functional connectivity. Lymphocyte-associated MTL atrophy may thus represent the possible etiology of PD.