Resting‐state frontostriatal functional connectivity in Parkinson's disease–related apathy

One of the most common neuropsychiatric symptoms in Parkinson's disease (PD) is apathy, affecting between 23% and 70% of patients and thought to be related to frontostriatal dopamine deficits. In the current study, we assessed functional resting‐state frontostriatal connectivity and structural changes associated with the presence of apathy in a large sample of PD subjects and healthy controls, while controlling for the presence of comorbid depression and cognitive decline. Thirty‐one healthy controls (HC) and 62 age‐, sex‐, and education‐matched PD patients underwent resting‐state functional magnetic resonance imaging (MRI). Apathy symptoms were evaluated with the Apathy Scale (AS). The 11 Beck Depression Inventory‐II items that measure dysphoric mood symptoms as well as relevant neuropsychological scores were used as nuisance factors in connectivity analyses. Voxel‐wise analyses of functional connectivity between frontal lobes (limbic, executive, rostral motor, and caudal motor regions), striata (limbic, executive, sensorimotor regions), and thalami were performed. Subcortical volumetry/shape analysis and fronto‐subcortical voxel‐based morphometry were performed to assess associated structural changes. Twenty‐five PD patients were classified as apathetic (AS > 13). Apathetic PD patients showed functional connectivity reductions compared with HC and with non‐apathetic patients, mainly in left‐sided circuits, and predominantly involving limbic striatal and frontal territories. Similarly, severity of apathy negatively correlated with connectivity in these circuits. No significant effects were found in structural analyses. Our results indicate that the presence of apathy in PD is associated with functional connectivity reductions in frontostriatal circuits, predominating in the left hemisphere and mainly involving its limbic components. © 2015 International Parkinson and Movement Disorder Society

apathy has been found to correlate with frontal gray matter (GM) 5 and ventral striatal volume reductions. 6 Apathy after subthalamic nucleus stimulation surgery was shown to be associated with mesolimbic dopaminergic denervation 7 and is amenable to dopaminergic therapy. 8,9 These findings give support to the relevance of frontostriatal circuits in PD-related apathy; to our knowledge, nonetheless, no published studies have evaluated the presence of associated functional connectivity (FC) changes.
The presence of apathy in PD is associated with cognitive deficits [10][11][12] and with a higher risk of dementia. 13 Moreover, it often coexists with depression, and the symptomology of both syndromes overlaps 2 ; specifically, symptoms of apathy occur as part of the depressive syndrome. 14 Although evidence supports the existence of apathy as a distinct entity in PD, 2,10,[15][16][17] the study of its neural substrates, as well as its clinical detection, are complicated by such overlap.
Our aim in the current study was to evaluate resting-state FC and structural changes affecting the frontostriatal pathways in a large sample of PD patients and matched controls, while controlling for the associated effects of cognitive decline and depressive symptoms. We hypothesized that apathy in PD patients would be associated with disrupted FC in frontostriatal circuits, especially affecting the ventral striatum and the ventromedial prefrontal cortex. Considering that dopamine modulates frontostriatal FC, and that dopamine deficiency due to nigral degeneration precedes forebrain GM pathology in PD, 18 we also expected that FC changes would be more marked than structural degeneration.

Methods
Eighty-four non-demented PD patients and 38 healthy controls (HC) matched for age, sex, and years of education were included. Patients were recruited from the Parkinson's Disease and Movement Disorders Unit, Hospital Cl ınic de Barcelona. Healthy controls were recruited from individuals who volunteered to participate in scientific studies at the Institut de l'Envelliment, Universitat Aut onoma de Barcelona. The inclusion criterion for patients was the fulfillment of the UK PD Society Brain Bank diagnostic criteria for PD. 19 Exclusion criteria were Mini-Mental State Examination scores less than 25 or dementia according to Movement Disorder Society criteria 20 ; Hoehn and Yahr (H&Y) score greater than III; significant neurological, systemic, or psychiatric (except depressive symptoms) comorbidity; pathological magnetic resonance imaging (MRI) findings other than mild white matter (WM) hyperintensities; root-mean-square head motion greater than 0.3 mm translation or 0.6 rotation.
Four patients were excluded because of macroscopic movement, 14 because of head motion greater than 0.3 mm translation or greater than 0.6 rotation, and one for being an outlier in connectivity analyses. Eight HCs were excluded (2 because of microvascular WM changes, 5 because of incomplete filling of the AS), leaving a final sample of 31 HC and 65 PD patients. This subject sample was used in a recent FC study, 21 and all except one HC and one PD patient were part of the sample used in a cortical thickness study. 22 All patients except one were taking antiparkinsonian drugs, consisting of different combinations of levodopa (L-dopa), cathecol-O-methyl transferase inhibitors, monoamine oxidase inhibitors, dopamine agonists, and amantadine. All assessments were done while patients were in the on state. Levodopa equivalent daily dose (LEDD) was calculated as suggested by Tomlinson et al. 23 Motor disease severity was evaluated using H&Y and Unified Parkinson's Disease Rating Scale motor section (UPDRS) scores.
The study was approved by the institutional ethics committee, and all subjects provided written informed consent to participate.

Neuropsychiatric Evaluation
Apathy symptoms were evaluated with the selfadministered Apathy Scale (AS), recommended for use in PD. 24 Subjects were classified as apathetic if they scored greater than 13. 15 We also administered the Beck Depression Inventory-II (BDI) to all subjects. Kirsch-Darrow et al. 25 dissociated BDI items into four factors (apathy, dysphoric mood, loss of interest/pleasure, somatic factor); the loss of interest/pleasure factor was shown to be sensitive to symptoms of both depression and apathy, whereas the somatic factor can be influenced by other PD-related symptoms. The 11 items constituting the dysphoric mood factor loaded on negativity/ sadness-symptoms not related to apathy-and showed the lowest correlation with apathetic symptoms. We used the score in these 11 items (henceforth referred to as dysphoric mood score) as covariates of no interest in FC analyses to control for associated depression.

Neuropsychological Assessment
Subjects underwent a thorough neuropsychological battery assessing cognitive functions frequently impaired in PD, using the following tests: Attention/executive functions: backward minus forward digit spans; Trail-Making Test part A minus part B scores; phonemic fluency scores (words beginning with "P" produced in 60 seconds), and Stroop Color-Word Test interference scores). Visuospatial/ visuoperceptual functions: Benton's Visual Form Discrimination and Judgment of Line Orientation tests. Memory: Rey's Auditory Verbal Learning Test total learning and 20-minute free recall scores. Composite z-scores for each cognitive function (referred to as attention/executive [A/E] scores, memory scores, and VS/VP scores) were calculated as the mean of the z-scores of all tests within that function.
We have also investigated the presence of mild cognitive impairment (MCI) following the Movement Disorder Society Task Force criteria, as described previously. 22

MRI Acquisition
Structural T1-weighted images, functional restingstate images, and FLAIR images were acquired on a 3T Siemens MRI scanner as previously described. 26

Processing of fMRI
The preprocessing of resting-state images was performed with FSL (release 5.0.4, http://fsl.fmrib.ox.ac. uk/fsl/fslwiki/FSL) and AFNI (http://afni.nimh.nih.gov/ afni). Briefly, it included removal of the first 5 volumes to allow for T1 saturation effects, skull stripping, grandmean scaling, and temporal filtering (0.01-0.1 Hz). To control for the effect of subject head movement, physiological artifacts, and other nonneural sources of signal variation on the estimation of connectivity, motion correction and regression of nuisance signals (six motion parameters, cerebrospinal fluid, and WM) were performed. To remove the effects of images corrupted by motion, a scrubbing procedure as suggested by Power et al. 27 was applied. Images were then smoothed with a 6-mm full-width at half maximum Gaussian kernel.
Additionally, head motion was calculated as the average Euclidean displacement between consecutive timepoints for rotatory and translatory motion. 28

Definition of Regions of Interest
The frontal cortices were parcellated into limbic (anterior, posterior, and medial orbital gyri, gyrus rectus, and subcallosal gyrus/ventral anterior cingulate), executive (rostral superior and middle frontal gyri and dorsal prefrontal cortex), rostral motor (caudal portions of lateral and medial superior frontal gyrus, caudal middle and inferior frontal gyrus), and caudal motor (precentral gyrus and caudal premotor area), as described by Tziortzi et al. 29 In this study, these frontal divisions were used as seeds for probabilistic tractography analyses that defined the functional striatal subregions included in the Oxford-GSK-Imanova Striatal Connectivity Atlas, which we used to parcellate the striata into limbic, executive, and sensorimotor regions. We also included the thalami, defined using the Harvard-Oxford subcortical structural atlas. Supplemental Data Figure 1 displays the frontal and striatal segmentation scheme used.
To obtain each seed region's resting-state functional MRI (fMRI) time series, the mask for each structure was non-linearly registered to each subject's T1weighted image using FSL FNIRT, and subsequently linearly registered to native functional space.

Functional Connectivity Analysis
Functional connectivity analyses were performed with FSL and AFNI. Initially, a mean time series was obtained from each seed region (4 frontal, 3 striatal, 1 thalamic per hemisphere) by averaging the time series of every voxel contained in it before smoothing, in native functional space. Subsequently, these time series were correlated with the time series of every voxel inside the regions of interest, thus producing a Pearson's r coefficient correlation map. These were then converted to z maps, using Fisher's r-to-z transformation.

Cortical and Subcortical Gray Matter Volume Analysis-Voxel-Based Morphometry
Structural data was analyzed with FSL-VBM, 30 a voxelbased morphometry (VBM)-style analysis. First, nonbrain tissue from structural images was extracted. After segmentation, GM images were aligned to MNI152 standard space using affine registration. Resulting images were averaged to create a study-specific template, to which native GM images were then non-linearly re-registered. The registered partial volume images were then modulated by dividing by the Jacobian of the warp field. The modulated segmented images were then smoothed with an isotropic Gaussian kernel (sigma 5 3 mm). A mask with the structures of interest (frontal cortices, striatal divisions, and thalamus) was created to define a search volume for subsequent statistical testing.

Subcortical Volumetry and Shape Analysis
Subcortical structures of interest (accumbens, pallidum, caudate, putamen, and thalamus) were segmented using FIRST as implemented in FSL. Because subcortical structural volumes scale with head size, we used FSL SIENAX to calculate intracranial volumes, to be entered as nuisance factors in volume analyses. For shape analysis, surface meshes were fitted for each region of interest as a vertex distribution model, modeling its shape. Multivariate testing was then performed on the three-dimensional coordinates of each vertex, followed by multiple-comparison correction, thus allowing the detection of localized structural changes. 31

White-Matter Hyperintensity Load Analysis
To detect and quantify global (periventricular 1 deep) WM hyperintensity load, we used an automated segmentation procedure. 32 Results are given in normalized volumes, taking brain volume into account.

Statistical Analysis
Voxelwise general linear model was applied using non-parametric testing (5,000 permutations) for connectivity and VBM, testing all voxels inside the regions of interest (frontal lobes, striata, thalami), as well as for shape analyses. To evaluate the association between the presence of apathy and changes in FC, GM volume or shape, we performed both intergroup comparisons and correlations with AS scores. Sex and dysphoric mood scores were entered as nuisance variables in all analyses. Because group differences were found for A/E and memory scores, these variables were entered in intergroup analyses; for correlation analyses, memory scores were entered as a nuisance factor as they correlated with AS scores. Significance level was set at P < 0.05, corrected for multiple comparisons using false-discovery rate (FDR) control, applied to P-value maps obtained from threshold-free cluster enhancement. 33 Pearson's chi-square test was used to compare categorical variables (hand dominance, sex, HY stage). Student's t test was used to compare clinical data means between patient subgroups (PD-NA, PD-A). Three-level one-way analyses of variance were used to compare clinical and sociodemographic data between HC and patient subgroups. Three-level one-way analyses of covariance were used to compare subcortical volumes between HC and patient subgroups, controlling for variables that showed intergroup differences in the previous step. Pearson's correlation was used to evaluate the relationship between demographic, clinical, and neuropsychological measures. Partial correlations were performed separately in the HC group and in the collapsed PD group to assess the relationship between subcortical volumes and AS scores, controlling for the variables that correlated significantly with them. Statistical significance threshold was set at P < 0.05.

Neuropsychiatric Assessment
Twenty-six PD patients (41.3%) were classified as apathetic (AS > 13). Table 1 shows sociodemographic, clinical, and head motion data, with intergroup comparisons. No intergroup differences in age or years of education were seen, although significant sex differences occurred (males tended to be overrepresented in the apathetic patient group). As expected, dysphoric mood scores were significantly higher in apathetic patients (PD-A) than in non-apathetic patients (PD-NA). The LEDD and measures of disease severity (H&Y, UPDRS scores) or duration were not significantly different between PD-A and PD-NA and did not correlate significantly with AS, BDI, or dysphoric mood scores.
The proportion of PD patients taking antidepressant medication was similar in both groups (6 PD-NA, 5 PD-A, P 5 0.702, Pearson's chi-square 5 0.146); AS and neuropsychological scores were not significantly different between medicated and unmedicated patients.
AS scores correlated significantly with memory scores (r 5 -0.36, P 5 0.004), whereas no significant effects were found for A/E (P 5 0.503) or Mini Mental State Examination (P 5 0.694) scores. Dysphoric mood scores did not correlate significantly with cognitive scores.
Although PD-A patients had lower A/E scores than HC and lower memory scores than both HC and PD-NA (Table 1)

Fronto-Striatal-Thalamic Functional Connectivity Analysis
PD-A patients showed FC reductions (P < 0.05, FDR-corrected) compared with HC and with PD-NA, mainly in left-sided circuits and involving limbic regions (see Table 2 and Figure 1). Compared with HC, PD-A displayed reduced FC between the limbic striatal division and the rest of the left striatum. Moreover, FC between the limbic division of the left striatum and the left frontal lobe was significantly reduced in PD-A compared with PD-NA; in PD-A compared with HC, evidence (P 5 0.06, FDR-corrected) suggested reduced connectivity in the left orbitofrontal cortex and inferior frontal gyrus.
Correlation analyses in the collapsed PD sample showed that AS scores correlated negatively with the FC between both limbic and executive divisions of the left striatum and the left frontal lobe; between the limbic region of the left frontal lobe and the left striatum; and between the caudal and rostral frontal lobe and right striatum. Additionally, AS scores correlated negatively with the FC between the different subdivisions of the left frontal lobe (see Fig. 2 and Supplemental Data Table). These regions of significant correlation with AS scores displayed a marked overlap with the clusters of significant intergroup differences described previously in the left frontal lobe and in the striata.
Significant intergroup/correlation analysis results are summarized schematically in Figure 3.
FC results were maintained after adding head movement parameters as nuisance factors.
No significant correlations were found in the HC group. To confirm that the connectivity differences observed were not attributable to the unequal sex distribution among groups, we performed additional connectivity analyses comparing male and female HC; no significant differences were observed. Additionally, we performed post hoc correlation analyses separately in male and female PD patients (Supplemental Data Fig.  2), revealing that the observed significant negative correlations between AS scores and frontostriatal functional connectivity were present in both groups.

VBM, Subcortical Volume, and Shape Analysis
No significant group differences or correlations with AS scores were observed for GM volume or subcortical volume/shape.  Description of clusters (>10 voxels) of significant intergroup connectivity differences, controlling for dysphoric mood, memory, and attentional/executive scores, as well as sex. HC, healthy controls; PD-A, Parkinson's disease patients with apathy; PD-NA, Parkinson's disease patients without apathy; SM, sensorimotor; OFC, orbitofrontal cortex; PCG, precentral gyrus.

White-Matter Hyperintensity Load
There were no significant intergroup differences or correlations between connectivity values in the clusters of intergroup differences and WM hyperintensity load.

Discussion
The main finding of our study is that the presence of apathy in PD was associated with resting-state FC reductions affecting frontostriatal circuits, predominantly in the left hemisphere. These changes were observed while controlling for the presence of associated depressive symptoms and cognitive impairment, and were not accompanied by significant structural changes, suggesting that frontostriatal FC disruption plays a relevant role in PD-related apathy.
Our results indicate that the occurrence of apathy in PD is accompanied by reduced resting-state FC mainly affecting the limbic divisions of the striatum and prefrontal cortex. These structures are central components of the brain's reward and motivation systems, 34 recently shown to be involved in PD-related apathy. 35 Moreover, the limbic division of the left striatum showed reduced connectivity with the ipsilateral frontal cortex and with the rest of the left striatum. The limbic striatum is hypothesized to influence motor activity as the limbic-motor interface 36 ; Haber et al. 37 found evidence that the mechanism through which this region influences other striatal regions-and, consequently, prefrontal and motor cortices-involves a "striatonigrostriatal spiral" through connections with mesencephalic dopaminergic nuclei. Dopamine deficits have been shown to result in reduced frontostriatal FC. 38 The loss of mesencephalic dopaminergic neurons might thereby lead to the intrastriatal, frontostriatal, and, ultimately, frontofrontal connectivity disruptions found in our study. Finally, the worsening of apathetic symptoms during the off state 39 and after dopaminergic treatment reduction after deep-brain stimulation surgery, 8 as well as its improvement with dopaminergic treatment, 7,40 suggest that dopamine deficiency is involved in PD-related apathy.
Taken together with our results, the abovementioned findings provide evidence that reduced striatofrontal resting-state FC is associated with apathy in PD, and is probably mediated by dopamine deficits. Furthermore, the "striatonigrostriatal spiral" model links the connections between the limbic striatum and other striatal/frontal regions not only to ventral tegmental area but also to substantia nigra pars compacta dopaminergic neurons. 37 The FC correlates found in our study had a clear leftsided predominance. This finding, alongside recently published studies, indicates that laterality of PD neuropathological changes influences the risk of developing apathy. Cubo et al., 3 assessing PD patients up to 2 years from diagnosis, found that subjects with leftpredominant motor symptoms (ie, with lesser left-sided dopamine deficits) were less likely to be apathetic. In line with these observations, the striatal volume correlates for apathy described by Carriere et al. 6 were more pronounced in left-hemisphere structures. Furthermore, in a recent study, Porat et al. 41 found that offmedication PD patients with greater left-sided dopamine deficits had impaired approach motivation-a symptom of apathy-whereas right-sided predominance was associated with impaired loss avoidance (possibly related to impulse control disorders). 41 No such relationship was found in our sample, which may be related to the fact that most patients had bilateral disease at assessment. Gray matter atrophy, which was not found in our study, has been inconsistently reported in association with apathy in PD. Reijnders et al. 5 described that the degree of apathy correlated with frontal, insular, and parietal reductions in GM volume, whereas Isella et al. 17 failed to find structural correlates. Carriere et al., 6 studying dopamine-resistant apathy through shape analysis, found striatal volume reductions, mainly in the nucleus accumbens, but no cortical thinning. These variable results may be a consequence of the distinct techniques used, different sample characteristics, and different con-trol over potential confounds such as cognitive deficits and motor disability. 42 Taking into consideration the known patterns of PD evolution, conceivably apathy in early PD is mediated by disrupted frontostriatal FC secondary to mesencephalic dopaminergic neuron degeneration; as the degenerative process progresses to forebrain structures, frontal and striatal atrophy may acquire a more important pathophysiological role. Longitudinal and radionuclide imaging studies are necessary to confirm this hypothesis and help disentangle the neurochemical and anatomical substrates of apathy in PD. Studies assessing PD patients in the on and off states also may shed light on the interaction between dopaminergic medication and the underlying pathological changes on FC and may provide a better characterization of the pathophysiology of apathy in PD. Moreover, future structural connectivity studies may be useful in Finally, apathy in our study was associated with worse cognitive performance, giving support to the association between apathy and cognitive deficits in PD. [10][11][12] Apathy was also more common in male patients, a finding that has been inconsistently reported. 1,10,11 Some limitations must be considered when interpreting our results. Groups were not matched for depressive or cognitive status. Although we did correct all connectivity analyses for these variables, we cannot exclude that they influenced the results obtained. Nonetheless, considering the high coexistence of apathy and depression or cognitive decline, our study sample is probably representative of the general population of apathetic PD patients. Additionally, with the aim of using a neuroanatomically valid scheme for parcellating the frontostriatal circuits, we used data from a carefully performed structural connectivity study. 29 The resulting regions of interest, however, are large, and their averaged functional time series may fail to reflect the activity of functionally distinct subregions, possibly reducing the sensitivity to localized connectivity disruptions. Finally, despite rigorous exclusion criteria, our study groups were not matched for head movement, which can affect FC parameters. 43 We have nonetheless applied several preprocessing steps to minimize these effects. The fact that results were maintained after adding movement parameters as covariates, and that the main group differences were found between HC and PD-A (which did not differ in head motion), suggests that the observed effects are not artifactual.
In conclusion, our findings suggest that apathy in PD is associated with reduced resting-state frontostria-tal FC, mainly affecting left-hemispheric limbic/ ventromedial regions but also extending to premotor and primary motor regions, even in the absence of significant structural degeneration and while controlling for associated depression and cognitive decline. These findings are compatible with the purported involvement of dopamine deficits in frontostriatal pathways in the genesis of apathy symptoms in PD.