Pay attention to the basal ganglia: a volumetric study in early dementia with Lewy bodies

Dementia with Lewy bodies (DLB) is the second most common form of neurodegenerative disease after Alzheimer’s disease (AD), with prevalence rates up to 5% in the elderly population and up to around 20% of all cases of dementia [1, 2]. According to the most recent consensus criteria [3], a diagnosis of probable DLB can be made if two or more core clinical features are present, among fluctuating cognition, recurrent visual hallucinations, rapid eye movement sleep behavior disorder (RBD), and spontaneous features of parkinsonism. But essential to the diagnosis of DLB remains the progressive cognitive decline, with deficits on tests of attention, executive functions, and visuo-perception in the foreground, which may occur early [4, 5]. Among these deficits, attentional dysfunction, which is typically assessed in clinical routine by measures of speed of processing, is a prominent and distinguishing neuropsychological feature of DLB as compared to AD at the prodromal stages [4, 6]. For instance, based upon a large prospective study comparing DLB with AD, the slowing of cognitive processing (i.e., deficits of cognitive reaction time) appears to be specific to DLB in the early stages of the disease [7]. Tasks of attention and processing speed requiring a graphomotor response seem to be particularly useful in discriminating DLB from normal status or versus AD [6]. Previous studies have also indicated a disruption of attention in the visual modality in DLB as compared to AD and Parkinson’s disease (PD) patients [8‐11].

Attention is a broad concept and, accordingly, many supporting brain regions are involved. Current neuroanatomical descriptions mostly come from the literature on fMRI in normal subjects and are drawn in a network perspective: in the context of attention-consuming tasks, the dorsal attention network (DAN) and the salience network (SN) both exhibit increased activity whereas the default network (DN), including regions of the cortical midline, is disengaged [12, 13]. The DAN encompasses the frontal eye field and inferior parietal sulcus and is involved in directed attention and working memory (Corbetta and Shulman [14]). The SN involves the anterior insula (AI), which plays a prominent role in the detection of salient stimuli, and the dorsal anterior cingulate cortex (dACC), more involved in task control [15]. The subcortical areas, including the thalamus, striatum, superior colliculus, and connecting white matter tracts, also participate in these networks [16].

Interestingly, structural neuroimaging studies using voxel-based morphometry (VBM) in patients with DLB, relative to AD patients, PD patients, and/or normal controls, have revealed gray matter (GM) atrophy in some of the abovementioned attentional regions. If relative to AD patients, DLB patients show a relative preservation of medial temporal lobe volumes [17‐20]; compiling VBM studies of DLB patients versus control subjects revealed consistent decreased GM volumes in the right lateral temporal/insular cortex and left lenticular nucleus (putamen and globus pallidus)/insular cortex [21]. These data are at least partially corroborated by studies conducted in our team with DLB patients versus controls, showing reduced GM volumes of the bilateral insula and anterior cingulate cortex in prodromal DLB (pro-DLB) using VBM [22], reduced GM volumes of the bilateral insula in pro-DLB using DARTEL-VBM, and diminished right insula and right orbito-frontal cortex volumes in pro-DLB using measures of cortical thinning [23]. Additionally, reduction of GM density has been shown in the occipito-parietal areas and bilateral lenticular nucleus in DLB versus PD [24], in the left caudate nucleus in DLB versus controls [25], in the bilateral thalamus in DLB versus controls [26], and in the bilateral putamen in DLB versus controls and AD [27].

So far, the question of the cerebral regions involved in attention deficits in DLB patients has been investigated in a few papers only. In an fMRI study, Firbank at al [28]. have shown in Lewy body disease (LBD) patients (DLB and PD) greater activation of the DAN for incongruent and more difficult trials as well as heightened deactivation of the default network, interpreted as an attempt to allocate resources to impaired attentional networks. Using functional connectivity analysis in the same cohort of patients, Koboleva et al. [29] found a decreased connectivity between the DAN and the ventral regions. Based on the same cohort again, Cromarty et al. [30], in a VBM study, did not show any significant correlations between attentional performance, despite deficits in the tasks, and GM volumes, suggesting, according to the authors, that these effects were unlikely due to region-specific structural deficits. On the other hand, Watson et al. [26] found in a pure group of DLB patients that impaired attentional function (measured by simple and choice reaction times) was correlated to the left thalamic regions (pulvinar and ventral lateral nucleus). Overall, these studies have not so far provided a clear picture of the regions involved in the attentional deficits of DLB patients.

The aim of the present study was to better understand the underlying structural mechanisms of the attentional deficit in DLB patients. Based on the fact that both the insular cortex and basal ganglia (caudate and thalamus) are important to attention function, that both areas are highly interconnected [31, 32], and given that these regions are atrophied at an early stage in DLB patients, we posit that attention disturbances in DLB might be related to diminished volume in these regions. To test this hypothesis, we took advantage of a unique cohort of 93 patients diagnosed with prodromal to moderate DLB, who underwent a full neuropsychological assessment, including attentional measures, as well as structural MRI.

The aim of the present study was to investigate which brain regions are involved in the attentional difficulties of DLB patients. For this purpose, we used VBM correlational analysis between behavioral performances on two typical clinical measures of attention and the degree of GM cerebral atrophy. As expected, we found an attentional deficit by means of tasks such as the TMTA and the DSST that are used in clinical routine. In accordance with our hypothesis, volumetric analyses highlighted the correlations between altered attentional scores and decreased volumes in the basal ganglia: in the striatum (mainly the caudate nucleus) and subthalamic nucleus for the TMTA, in the left thalamus for both tasks, and in the right inferior frontal gyrus (BAs 44 and 45) and the left cerebellum for the DSST. The results of our additional analysis, including MMSE scores as a nuisance covariate, confirmed that correlations between TMTA mean RT and caudate nucleus and between DSST scores and left cerebellum were not explained by the severity of the disease, reinforcing the role of these regions independently of the progression of the disease.

By comparing our group of DLB patients to healthy controls, we were able to confirm the existence of attentional deficits, even at the early stage of the disease, which is consistent with the literature [4, 6, 45]. Indeed, the majority of the patients included in the present study were at the prodromal or mild stages of the disease (only 6 out of the 93 patients were at the moderate stage at the time of the evaluations). The 2 tasks we used have in common that they measure the speed of processing as they require a series of operations to be performed under time pressure, and they both involve visual analysis, focused attention, response selection, and motor execution, yet each task has its own specificity. The DSST is a widely used, standardized psychometric test that also targets the maintenance of stimulus-response associations, and has high re-test reliability [42, 46]. Hence, this measure involves more operations than the TMTA, which might explain, based on the variability of patterns of atrophy in DLB patients, why correlations between DSST scores and the degree of atrophy are less specific of one particular brain region. Conversely, the TMTA requires more basic processes involving the correct sequencing of simple elements. The present results are in accordance with the literature suggesting that this measure constitutes a key discriminator between DLB and AD [11].

Negative correlations were found between mean reaction times in the TMTA and clusters that were very delimited to the regions of the basal ganglia, including primarily the caudate nucleus, the putamen, the globus pallidus bilaterally, the left thalamus, and the subthalamic nuclei, meaning that the slower the processing speed, the more these regions were atrophied. The basal ganglia consist of an array of the subcortical nuclei, including the caudate nucleus and the putamen (collectively referred to as the striatum in humans), the globus pallidus, the subthalamic nuclei, and the substantia nigra [47].

The striatum is involved in loops interacting with the prefrontal cortex (PFC). There are at least five of these circuits anatomically described: motor, oculomotor, dorsolateral PFC, orbital PFC, and anterior cingulate circuits [48], supported by both neuroanatomical and neuroimaging studies [47]. They all originate in the PFC and form a loop passing through the striatum (caudate or putamen), globus pallidus, and substantia nigra and finally through the thalamus. In the present study, part of the striatum, the caudate nucleus, was more involved in the dorsal part of its head, and our results also indicate that the most significant correlation was for the left caudate nucleus, for which the cluster peak reached significance at FWE-corrected level. Previous physiological, disease, and lesion studies have demonstrated that the caudate nucleus is associated with processing speed. Patients with traumatic brain injury and initial severe concussion presented a correlation between the volumes of caudate and the speed of processing, assessed using the Wechsler Adult Intelligence Scale-IV (WAIS IV) processing speed index [49]. In rhesus monkeys, a spatial visual working memory task activated bilaterally the head of the caudate nuclei [50] (see also Derauf et al. for patients with reduced caudate volumes due to prenatal methamphetamine exposure correlated to reaction times in an attention task [51] and Spies et al. for HIV patients with previous childhood psychological trauma presenting with diminished volume of the left caudate nucleus, together with slower processing speed [52]). However, in prodromal Huntington’s disease, where the caudate nucleus is atrophied, no correlation was found between the caudate nucleus and speed of processing, using DSST or TMTA [50, 53]. The strong correlation we found between the TMTA and the caudate nucleus could reflect the different processes involved in this task, with a more particular involvement of the dorsal caudate/dorsolateral PFC for the cognitive aspect of the task. Such a correlation has never previously been reported in DLB. However, our results confirm the robust link between processing speed and caudate lesion recently demonstrated in traumatic brain injury by Tate et al. [49]. In line with the idea that the caudate atrophy is involved in non-motor impairment in DLB patients, the volumetric study by Barber et al. [25] highlighted a reduction of the volume of the left caudate nucleus that was not itself correlated with parkinsonism symptoms. These data are consistent with the dorsal caudate nucleus being involved in cognitive rather than motor aspects of the TMTA [47, 54]. Based on all these data, we would suggest that in this pathological condition (DLB), the basal ganglia are the point of disruption of the cognitive loops, even if VBM-GM methods cannot inform us whether afferent pathways, efferent pathways, or both are disturbed.

We also found correlations between scores on the TMTA and atrophy in the subthalamic nuclei. This region was initially described as sustaining motor functions, because of its involvement in some motor symptoms, for example, in PD; interestingly, however, this initial view of its role has been revised in the past decades to incorporate wider cognitive functions [55].

Conversely, since the putamen is connected with the PFC supplementary motor area (SMA), and hence is implicated in motor circuits, this region might rather sustain the motor aspects of the TMTA. Accordingly, it has been proposed that the putamen contributes to movement preparation during self-initiated behavior [47].

The last subcortical region that we found to be correlated with attentional scores and common to both tasks was the left thalamus in its ventral portion. More precisely, the degree of atrophy of the ventral anterior nucleus was correlated to scores on the TMTA and those of the ventral lateral nucleus to scores on the DSST. It is interesting to note that distinct regions of the thalamus are also integrated into the cortico-subcortical circuits described before, receiving afferent projections from the striatum [48]. It has recently been suggested that two of the DLB core features, hallucinations and fluctuations, might be related to the pathological mechanisms at the level of the thalamus, by disturbing the thalamocortical circuits [56]. In DLB patients, attentional impairment is associated with cognitive fluctuations, for which thalamic dysfunction has been shown (Chabran E, Noblet V, Loureiro de Sousa P et al: Changes in gray matter volume and functional connectivity in dementia with Lewy Bodies compared to Alzheimer’s disease: implications for fluctuations, submitted). More importantly, our result corroborates the recent finding by Watson et al. [26], one of the two studies that, to date and to our knowledge, have looked for correlations between attention scores and GM volumes: using linear regression analysis centered on the thalamus, and attentional tasks of simple reaction time, choice reaction time, and vigilance, they found that atrophy in the ventral lateral nucleus was associated with impaired attentional function in DLB. Taken together, these data point towards a dysfunction of the thalamic ventral lateral nucleus underlying attentional impairment in DLB patients.

Overall, the implication of the subcortical regions is consistent with the subcortical syndrome clinical picture that is characterized by a general slowing of cognitive and motor functions, occurring as a consequence of strokes, vascular lesions, and metabolic diseases [57].

Turning now to non-subcortical regions, neither the insula nor other regions of the attentional networks described in the introduction were found to be correlated to attentional scores. Given the fact that the insula, which is part of the SN, has been shown to be involved in the detection of salient stimuli and might not be involved in the tasks we studied. As for the absence of involvement of the DAN [12, 13], the tasks we used might not have required either a redirection of attention or the recruitment of strong working memory processes.

On the contrary, we found positive correlations between DSST scores and atrophy in the right inferior frontal gyrus (BA 45) and in the left cerebellum. Regarding the former, if its left counterpart corresponds to the motor speech area of Broca, the right inferior frontal region has been associated with attentional control [14] and also with reasoning about indeterminate relations [58]. This appears consistent with our result since the specificity of the DSST is to associate unknown and insignificant symbols with numbers. We could also speculate that BA 45 in the right hemisphere serves as a premotor area for non-language codes.

Similar to other brain regions we have already discussed in the present paper, there is cumulative evidence, based on neuroimaging findings and neuropsychological research, suggesting that the cerebellum is involved in cognition in a way that is independent of motor functions [59]. Relevant to the combined atrophy we displayed in the cerebellum, thalamus, and PFC correlated to impaired performance on the DSST is the concept of “cognitive dysmetria” proposed by Andreasen et al. [60], observed in schizophrenic patients, which implicates a disruption of connectivity between these three nodes—cerebellum, thalamus, and prefrontal cortex—and produces, behaviorally, “difficulties in prioritizing, processing, coordinating, and responding to information” [60, 61]. Disturbed connectivity between these three brain regions, as reflected by our correlational results, might explain the impaired performances on the DSST in our group of DLB patients. As these results were reported at an uncorrected level of p < .001, these interpretations need to be treated with caution.

Overall, the positive correlations we found between raw scores on the DSST and GM volumes were less clearly delimited to one brain structure and less strong than with the TMTA, which might indicate a lower sensitivity of this test to gray matter damage of the basal ganglia. Conversely, DSST performance seems to be sensitive to white matter damage due to injury or disease [62, 63] and aging [64]. As a perspective, we will now examine the correlations between white matter morphometry as well as diffusion tensor imaging and attention scores on the same set of data. To disentangle the aspects of motor and/or cognitive speed impairment, we will explore motor speed independently, for instance, by focusing on walking speed in DLB patients.

This is the first study to have examined the neural correlates of attentional dysfunction in DLB patients using voxel-based morphometry and attention tasks typically used in clinical routine. While the existing literature on the neurobiological correlates of attentional dysfunction in DLB is still unclear, we were able to demonstrate, using correlation analysis between attentional scores and GM volumes, the involvement of the striatum—particularly the left caudate nucleus—and the subthalamic nucleus, associated with increasing time on the TMTA, the involvement of the right inferior PFC and the left cerebellum, associated with diminished performances on the DSST, and the involvement of the left thalamus in both tasks. Partially confirming our hypotheses, these results reveal the involvement of the basal ganglia in processing speed and attention in DLB.