Pathologic Correlates of Apraxia in Alzheimer Disease

Objective  To examine the neuroanatomical correlates of apraxia in Alzheimer disease.

Patients  Twenty-three patients with clinically overt Alzheimer disease.

Design  Anterograde study and neuropathologic case series. Clinical severity was assessed using the Global Deterioration Scale. Ideomotor praxis was examined on transitive and intransitive movements and meaningless gestures, and dressing ability was evaluated clinically. Constructive praxis was tested using a 3-dimensional figure copying task. Correlations between neurofibrillary tangle and senile plaque densities and praxis test performance were studied using stepwise logistic regression models.

Setting  Studies were conducted at the Psychiatric and Geriatric Hospitals of the University of Geneva School of Medicine, Geneva, Switzerland.

Main Outcome Measures  Odds ratios to estimate the associations between neurofibrillary tangle and senile plaque densities in each neocortical area and the presence of ideomotor, dressing, and constructional apraxia.

Results  Statistically significant relationships were found between neurofibrillary tangle densities in the anterior cingulate cortex and ideomotor and dressing apraxia and between neurofibrillary tangle densities in the superior parietal, posterior cingulate, and occipital cortex and constructional apraxia. Senile plaque counts did not correlate with praxic performance.

Conclusions  These results suggest that ideomotor and dressing apraxia are associated with mild damage of the anterior cingulate cortex, whereas constructional apraxia is related to the disruption of cortical pathways mediating visuospatial cognition in Alzheimer disease. Senile plaque densities do not represent a valuable pathologic correlate of apraxia in this disorder.

ALTHOUGH THERE have been many studies^1-9 on clinicopathologic correlations in Alzheimer disease (AD), few attempts have been made to determine whether the distribution of AD lesions is related to cognitive deficits that can be localized to specific brain regions.^10,11 Apraxia is a good candidate for these types of studies since investigations of focal pathology have suggested that this condition is associated with both structural and functional abnormalities in restricted cortical areas.^12-23 Apraxia is defined as a disorder of skilled movements that cannot be explained by palsy, paresis, ataxia, akinesia, posture, tone or movement disorders, comprehension impairment, or deafferentation. Three of the main forms of apraxia are ideomotor, dressing, and constructional apraxia.^11,20,21 Ideomotor apraxia represents a difficulty in making gestures caused by an inability to translate the concept of a motor sequence into the corresponding motor action. In the artificial context of testing, patients are unable to select, sequence, and orient spatially meaningful transitive and intransitive and meaningless movements. Bilateral injuries of the parietal lobes as well as the anterior segment of the corpus callosum have been related to ideomotor apraxia.^{14,16,18-20,22-25} Dressing apraxia represents a particular form of apraxia confined to clothing use and is often associated with focal lesions in the right parietal lobe.^15,17 Constructional apraxia refers to a visuospatial disorder characterized by an impairment in the spatial organization required when parts of objects are assembled to form a single entity. It is thought to be related to the parietal-occipital cortex pathology^15,17,26-31; however, this view has been challenged.^11,32

Apraxia usally appears in late stages of AD,³³ although patients with early constructional and ideomotor praxis disability have been described.^10,16 Since the neuroanatomical correlates of apraxia have been drawn from patients with discrete lesions, it is not yet clear whether this condition is related to the damage of specific cortical circuits or whether it is the consequence of diffuse neuropathologic changes in AD. To address this issue, we performed an anterograde clinicopathologic analysis of 23 patients with AD including detailed examination of ideomotor, dressing, and constructional praxis and quantitative assessment of neurofibrillary tangles (NFTs) and senile plaques (SPs) in several neocortical areas.

The sample included 23 right-handed patients (18 women aged 88.1 ± 1.3 years and 5 men aged 88.8 ± 2.4 years [mean ± SD]) who died and underwent autopsy in the Psychiatric and Geriatric Hospitals of the University of Geneva School of Medicine, Geneva, Switzerland (Table 1). All patients presented with considerable intellectual deterioration at the beginning of the study and were classified clinically as having probable AD according to the criteria of the National Institute of Neurological Disorders and Stroke and Alzheimer's Disease and Related Disorders Association.³⁴ Detailed neuropsychological testing, performed at least once during the 3 months prior to death, revealed severe memory deficits, temporal and spatial disorientation, and mild language deficits in all the patients. Gnosia was preserved in all the patients. The clinical severity of dementia was assessed using the Global Deterioration Scale (GDS).³⁵ The most common causes of death were global heart failure (10 patients), pneumonia (8 patients), cancer (3 patients), and pulmonary embolism (2 patients). There was no patient with a cerebral cause of death, such as tumor or stroke. All causes of death were pathologically confirmed.

The presence of ideomotor apraxia was assessed by requesting the patient to perform 5 transitive movements on commands given in 5 different modalities: verbal, visual, pantomime imitation, tactile, and actual use.³⁶ The test required demonstration of the use of the following 5 tools: comb, toothbrush, soup spoon, hammer, and saw. The following 5 intransitive movements were tested using verbal commands and imitation: wave good-bye, salute like a soldier, come here, stop, and go away.³⁶ In addition, 4 meaningless movement sequences were evaluated: making a curl with the thumb and the index finger of both hands, making "butterfly wings" with both hands (the hands are held with the thumbs crossed, the palms facing upward and the left wrist resting on the right one), joining both hands by the fingertips, and putting the index finger of the right hand on the middle finger of the left hand. Patients who were unable to imitate more than 1 of 5 transitive or intransitive movements and 1 of 4 meaningless gestures were classified as having ideomotor apraxia. Dressing praxis ability was evaluated clinically by asking the patients to put on and button a piece of their clothing. Apraxia was diagnosed when the patients failed to button this piece of clothing. To examine the presence of constructional apraxia, a Necker cube figure¹⁰ was placed in front of the patients and the patients were asked to reproduce the figure.^10,37 In case of failure, the examiner drew the cube for patients who were then immediately asked to reproduce it. Performances were scored according to the constructional praxis subtest of the Consortium to Establish a Registry for Alzheimer's Disease³⁷ neuropsychological battery. Patients who scored less than 3 were considered to have constructional apraxia.^10,37 All patients included in this study showed no visual impairment or any other physical limitation that may have influenced praxis test performance. All tests were administered by 2 trained neuropsychologists with high interrater reliability (r=0.97).

The brains obtained during the autopsy (postmortem delay, 3-41 hours) were fixed in 10% formalin for at least 6 weeks and cut into 1-cm-thick coronal slabs. Following macroscopic examination, tissue blocks were taken from the hippocampal formation, entorhinal cortex, superior and middle frontal cortex (Brodmann areas 9 and 46), superior, middle, and inferior temporal cortex (Brodmann areas 22, 21, and 20), superior parietal cortex (Brodmann area 7), angular cortex (Brodmann area 39), primary visual cortex (Brodmann area 17), occipital visual association cortex (Brodmann area 19), and anterior and posterior cingulate cortex (Brodmann areas 24 and 23) in the left hemisphere and the midbrain including substantia nigra. For microscopic purposes, paraffin-embedded blocks were cut into 6-µm-thick sections. For routine neuropathologic evaluation, tissues were stained with hematoxylin-eosin, cresyl violet acetate, Globus silver impregnation, and modified thioflavine S stains.³⁸ Subsequently, the clinical diagnosis of AD was confirmed in all the brains using the criteria of the Consortium to Establish a Registry for Alzheimer's Disease.³⁹ Brains with coexistent pathologic changes, such as cerebral infarcts or Lewy bodies, were not included in the present series to ensure a pure AD sample. The quantitative assessment of the localization and the distribution of NFTs and SPs was made on sections stained with highly specific and fully characterized antibodies to the microtubule-associated protein tau^40,41 and the amyloid Aβ protein.⁴² The anti-tau antibody used in the present study was a polyclonal antibody (961-S28T) raised against a synthetic peptide corresponding to a sequence located in the C-terminus of tau proteins (serine 400–threonine 429). This sequence contains 2 putative sites of phosphorylation serine-proline at serine residues 404 and 422 that are found in paired helical filaments. This antibody detects both intracellular and extracellular NFT.^40,41 The tissues were incubated overnight at a dilution of 1:4000 for both the anti-tau and anti-Aβ antibodies and subsequently processed by the peroxidase-antiperoxidase method using 3,3′-diaminobenzidine as a chromogen.

In all brains, analyses were performed in layers II and III and V and VI of Brodmann areas 9, 7, 39, 19, 24, and 23. These areas were selected to study possible correlations between different types of apraxia and the AD pathologic changes in the posterior cortical areas and the cingulate cortex. Area 9 was included as a control area in which no correlation with apraxia was expected. The number of NFTs and SPs was determined in 10 different serial sections for each area (10 randomly selected fields of 0.1 mm² per section) and the mean density of lesions per square millimeters in cortical layers II and III and V and VI separately was calculated using a computer-assisted image analysis system consisting of a microscope (Axioplan, Zeiss A. G., Zurich, Switzerland), a high-sensitivity camera (CP-3000, Tokina Optical Co Ltd, Tokyo, Japan), a microcomputer (COMPAQ Deskpro 386/20, COMPAQ, San Jose, Calif) and a software system (SAMBA 2005, ALCATEL, Grenoble, France).

Relationships between neuropathologic measures and GDS score, brain weight, years of education, and postmortem delay to autopsy were evaluated by univariate correlation analysis using Spearman coefficient (rs). Since praxis test performance was expressed in nominal values (presence or absence of apraxia), forward stepwise logistic regression models were built to examine correlations between NFT and SP densities and praxis. The forward stepwise logistic regression model is advantageous when dealing with relatively small samples since it allows variable reduction while maintaining appropriate statistical power. Total NFT and SP densities in each of the 6 studied neocortical areas as well as covariates, which were significantly correlated with neuropathologic data in the univariate correlation analysis, were used as predictors in this model with praxis test performance as the dependent variable. For each predictor that was retained into the regression models, an odds ratio was calculated to determine whether this independent variable significantly predicts the dependent variable. The odds ratio represents a direct measure of association between an independent variable and the dependent variable estimated from a logistic model. In addition, sensitivity and specifity values were calculated for each logistic regression model using the clinical diagnosis of apraxia as the criterion standard. All statistical analyses were performed using software (Stata software, Release 5, Stata Corporation, College Station, Tex).

Nine patients presented with mild dementia (GDS score, 4), 4 patients presented with moderate dementia (GDS score, 5), 6 patients presented with severe dementia (GDS score, 6), and 4 patients presented with very severe dementia (GDS score, 7). At least 1 type of apraxia was present in all patients with a GDS score of 5 or higher and 3 of the 9 patients with a GDS score of 4 (Table 1). Of the 23 patients, ideomotor apraxia was observed in 11 (51%), dressing apraxia in 7 (26%), and constructional apraxia in 15 (69%). Four patients with a GDS score of 7 and 2 patients with a GDS score of 6 displayed the 3 forms of apraxia. In this series, 13 (56%) of the patients had an age-related SP score of B according to the criteria of the Consortium to Establish a Registry for Alzheimer's Disease³⁷ in the most severely affected neocortical area and were classified as having neuropathologically probable AD (Table 1). The remaining patients met the criteria of the Consortium to Establish a Registry for Alzheimer's Disease of neuropathologically definite AD (age-related SP score of C and clinical history of AD). In addition, the hippocampal formation, entorhinal cortex, and area 20 exhibited massive NFT formation in all the patients.

Quantitatively, substantial NFT formation (≥10/mm²) was observed in 13 (55%) of the brains in layers II and III of area 23, while areas 9, 7, and 39 displayed lower NFT densities (<5/mm²) in all the brains. The lowest NFT densities were found in area 19. Senile plaques were distributed throughout the cerebral cortex in all the brains. High SP densities (>20/mm²) were seen in 17 (75%) of the brains in layers II and III of area 9, 9 (40%) of the brains in layers II and III of area 39, and 15 (65%) of the brains in layers II and III of areas 24 and 23. More than 10 SP/mm² were observed in 12 (50%) of the brains in layers II and III of area 7 and 15 (67%) of the brains in layers II and III of area 19 (Table 2).

There was a positive relationship between the GDS score and NFT densities only in areas 9 (rs=0.17; P<.05) and 23 (rs=0.23; P<.05). Conversely, no significant correlation was found between the GDS score and SP densities in any of the areas studied. Brain weight was correlated negatively with NFT densities in areas 9 (rs=−0.18; P<.05), 24 (rs=−0.17; P<.05), and 23 (rs=−0.30; P<.01) and with SP densities in all the areas studied (ranges: rs=−0.21 to −0.68; P<.05-.001). The number of years of education and postmortem delay to autopsy did not correlate significantly with neuropathologic data in any of the areas studied.

Subsequently, GDS score and brain weight in the areas studied were included as covariates in the stepwise logistic regression models with each type of apraxia as the dependent variable and NFT and SP densities in the 6 studied neocortical areas as independent variables. In this statistical analysis, test performance on both ideomotor praxis and dressing praxis was correlated negatively only with NFT densities in area 24 (P<.05); test performance on constructional praxis was correlated negatively with NFT densities in areas 7, 19, and 23 (P<.05) (Table 3). The GDS score and brain weight did not significantly predict praxis. The specificity of regression models was 86.7% for ideomotor apraxia, 50.0% for dressing apraxia, and 86.1% for constructional apraxia. The best sensitivity was obtained for dressing apraxia (97.8%), followed by ideomotor apraxia (95.3%) and constructional apraxia (86.7%). Senile plaque densities did not correlate with the different types of praxis performance in any of the areas studied.

The results presented herein indicate that ideomotor and dressing apraxias are associated with mild NFT formation in the anterior cingulate cortex, whereas constructional apraxia may be a good clinical indicator of NFT development in posterior cortical areas in AD. In addition, the absence of correlation between SP densities and praxis test performance in this series suggests that amyloid deposits are not causally related to these specific deficits in AD. The lack of correlation between SP densities and dementia severity observed herein is consistent with the findings reported in several series of elderly patients^1-9 and indicates that amyloid deposits do not represent a valuable pathologic hallmark of the progression of dementia in AD.

However, 2 limitations should be considered to interpret our data. First, this study only includes neuropathologic analysis of the left hemisphere. The hemispheric localization of apraxia is still unclear. Although ideomotor apraxia commonly follows damage to the left hemisphere, patients with damage to the right hemisphere may present with impairment in the production of intransitive movements to verbal commands.³⁶ In addition, although lesions in the posterior right hemisphere are likely to cause dressing apraxia and drawing disability,^{15,17,26,27,29-31} it has been proposed that these deficits are also related to damage to the posterior left hemisphere in AD.¹⁶ Second, the definition of apraxia in patients with AD may vary depending on the neuropsychological tests selected. No standardized test modalities are available for ideomotor and dressing apraxias, and the performance of 2-dimensional constructional tasks does not correlate well with the 3-dimensional figure copying task used in this study, suggesting that distinct cortical circuits may subserve these abilities.²⁶ Therefore, we cannot exclude that the relationships between the AD pathologic changes and different forms of apraxia may vary depending on the cerebral hemisphere studied and test design. Furthermore, it should be kept in mind that the correlations between NFT formation and different types of apraxia observed herein do not prove a causal relationship.

There have been several studies^14-24 on the relationship between focal cortical pathology and ideomotor and dressing apraxias; however, the pathologic correlates of these cognitive disabilities in AD are still unknown. The current study reveals a significant correlation between ideomotor and dressing apraxias and NFT densities in area 24, yet mild NFT formation was present in this area in all patients with AD.Three previous studies^18,19,25 showed that lesions restricted to the rostral part of the corpus callosum may cause ideomotor apraxia. Giroud and Dumas¹⁸ and Kazui and Sawada¹⁹ reported left ideomotor apraxia in conjunction with ischemic lesions in the anterior part of the corpus callosum. In addition, Goldenberg and collaborators²⁵ described the development of ideomotor apraxia with both verbal command and imitation in a patient with destruction of the anterior two thirds of the cingulate gyrus. Our observations amend these findings to show that the association between the damage of area 24 and ideomotor apraxia is also observed in AD. In contrast, the link between the damage of the anterior cingulate cortex and dressing apraxia remains poorly understood. It is well documented that the anterior cingulate cortex participates in the acquisition and performance of spatial memory tasks in experimental animals, implying that this area plays a key role in selection and recruitment of processing centers appropriate for complex task execution.^43-48 Furthermore, the anterior cingulate cortex possesses a motor area somatotopically connected to the primary motor cortex and is involved in the integration of somatic and visceral activities.^49-51 These observations suggest that even mild NFT formation in the anterior cingulate cortex may cause dressing apraxia by disturbing both programming and execution of complex motor behaviors. However, such a conclusion should be drawn with caution from the data presented herein since the specificity of the logistic regression model for dressing apraxia was quite low. Further clinicopathologic studies are needed to determine whether neocortical areas not included herein are better correlated with dressing apraxia than area 24.

The main result of the current study is the specific correlation between constructional apraxia and NFT densities in areas 7, 19, and 23. This result agrees with previous reports of focal pathology that showed that lesions confined to parietal and occipital areas may result in constructional apraxia.^15,17,26-31 Two studies^10,11 have investigated neuropathologic correlates of constructional apraxia in AD. Using silver stains, Förstl and collaborators¹¹ examined 23 patients with AD and found no specific relationship between constructional apraxia and parietal lobe pathology, although a significant positive correlation between brain weight and drawing performance was present. Based on these observations, the authors¹¹ postulated that constructional apraxia is indicative of diffuse rather than regional pathology in AD. This has been disputed by Nielson and collaborators,¹⁰ who performed detailed immunocytochemistry analysis of 16 patients with AD. Their results revealed a specific negative relationship between early neuritic pathology in area 18 (secondary visual cortex) and to a lesser degree in area 7 and figure copying performance, suggesting that this constructional praxis test may be valuable in assessing the underlying occipital cortex pathologic characteristics in AD. The data presented herein amend this latter report by showing that 3-dimensional figure copying disability in AD may be a clinical indicator of NFT formation in areas 7, 19, and 23. Although the function of these areas is still a matter of debate, several lines of evidence suggest that they participate in visuospatial processing.^45,52-60 Several studies^57-60 of the visual system of the monkey demonstrated the presence of a pathway between the visual and parietal cortex that subserves visuospatial tasks and motion analysis. Recently, positron emission tomographic activation studies have reported the involvement of the superior parietal and occipital cortex in visual learning and recognition of complex visual geometric patterns as well as in visuospatial attention processes.^52,53 Moreover, area 23 is strongly connected with areas 7 and 19, and studies of both humans and animals point to a role for this area in spatial orientation and in triggering appropriate motor responses to specific visual stimuli.^45,54-56,61 All together these observations imply that constructional praxis in AD may depend on the integrity of posterior cortical networks mediating visuospatial cognition.

In contrast to the hippocampal formation and temporal neocortex, which represent early sites of NFT formation, posterior cortical areas are thought to be affected only at a late stage of AD.^1-9,58 However, it has been suggested that disturbances of visual synthesis and spatial localization may be observed in most patients with AD independent of the clinical severity.⁶² The pattern of NFT distribution in the patients described herein is comparable to that reported usually in AD, in that there was a massive NFT formation in the hippocampus, entorhinal cortex, and area 20, and relatively low NFT densities in the other neocortical areas. While constructional apraxia correlates with moderate NFT involvement of areas 7, 19, and 23, it is important to note that patients with AD with simultanagnosia, ocular motor apraxia, and optic ataxia, the triad of Balint syndrome, display prominent NFT formation in the same areas.⁶³ This suggests the presence of 2 forms of posterior cortical pathology in AD: typical AD with drawing deficits associated with mild NFT formation in areas 7, 19, and 23 and AD with more complex visuospatial difficulties, such as Balint syndrome characterized by a preferential involvement of posterior parietal and occipital areas and relative preservation of the prefrontal and temporal neocortex.^62,63 Further clinicopathologic studies, including quantification of NFTs and SPs, estimation of neuronal and synaptic loss in additional posterior cortical areas, and standardized evaluation of complex visual task performance, may lead to a better understanding of the spectrum of posterior cortical pathology in AD.

Accepted for publication August 29, 1997.

This study was supported by grant AG05138 from the National Institutes of Health, Bethesda, Md (P.R.H.) and grant 31-52997-97 from the Fonds National Suisse de la Recherche Scientifique, Bern, Switzerland (P.G. and C. B.).

We thank F. Herrmann, MD, and E. Kövari, MD, for active participation in this study, N. K. Robakis, PhD, and A. Delacourte, PhD, for generously providing antibodies to tau and Aβ amyloid proteins, E. Giacobini, MD, for helpful discussion, and M. Surini and P. Y. Vallon for expert technical assistance. We also wish to thank J. Richard, MD, for the autopsy material.

Reprints: Panteleimon Giannakopoulos, MD, HUG-Belle-Idée-Psychiatry, Pavillon les Tilleuls, 2, chemin du Petit Bel Air, CH 1225 Chene-Bourg, Geneva, Switzerland.