Magnetic Resonance Imaging Signatures of Tissue Pathology in Frontotemporal Dementia

Background  The pathologic substrates of frontotemporal dementia (FTD) are difficult to predict in vivo.

Objective  To determine whether different pathologic substrates of FTD have distinct patterns of regional atrophy on magnetic resonance imaging (MRI).

Design  Retrospective case study.

Setting  The Institute of Neurology, University College London, and the Institute of Psychiatry, King’s College London.

Patients  Twenty-one cases of FTD selected on pathologic grounds (9 with ubiquitin-positive [tau- and α-synuclein–negative] inclusions [FTD-U], 7 with Pick disease [PiD], and 5 with familial FTD with tau exon 10+16 mutations [tau exon 10+16]) and 20 healthy controls were studied.

Main Outcome Measures  Patterns of gray matter atrophy in each group as assessed by voxel-based morphometry (VBM) and a blinded visual assessment of each MRI study.

Results  All pathologic substrates were associated with atrophy that involved the inferior and medial temporal and inferior frontal lobes. Additionally, specific VBM signatures were identified for each subgroup: FTD-U was associated with asymmetric (left > right) temporal lobe atrophy, PiD was associated with severe dorsolateral bifrontal atrophy, and tau exon 10+16 was associated with asymmetric (right > left) medial temporal lobe atrophy. The VBM findings were supported by blinded visual assessment.

Conclusion  These findings suggest that MRI patterns of regional gray matter atrophy constitute signatures of tissue pathology in FTD.

Frontotemporal dementia (FTD) is a group of neurodegenerative disorders with relatively focal involvement of the frontal and temporal lobes. These diseases together constitute a common cause of dementia.¹ Three main syndromic variants are recognized: frontal variant FTD (fvFTD), which presents with predominant behavioral problems; semantic dementia (SD), which presents as an impairment of semantic memory; and progressive nonfluent aphasia (PNFA), which presents with predominant speech production difficulties.² These clinical phenotypes are associated with characteristic patterns of brain atrophy on magnetic resonance imaging (MRI).³ Typically, fvFTD is associated with bilateral frontal atrophy,³ SD is associated with predominantly left anterior temporal lobe atrophy,⁴ and PNFA is associated with left perisylvian atrophy.⁵ However, definitive diagnosis of FTD continues to require pathologic examination.

Frontotemporal dementia is pathologically heterogeneous.² One common subtype is characterized by the presence of ubiquitin-positive, tau- and α-synuclein–negative, intracellular inclusions that involve the frontal and temporal cortices and the dentate gyrus of the hippocampus.⁶ This disease is herein designated FTD with ubiquitin-positive, tau- and α-synuclein–negative inclusions (FTD-U); it has also been termed FTD with motor neuron disease–like inclusions.⁷ Another pathologic subtype is Pick disease (PiD), which is characterized by the presence of ballooned neurons with argyrophilic intraneuronal rounded inclusions (Pick bodies), immunoreactive for the microtubule-associated protein tau, in the frontal and temporal cortices, dentate granule cells, and subcortical nuclei.⁸ Approximately 10% to 50% of FTD cases are genetically determined, and some autosomal dominant pedigrees have mutations in the tau gene on chromosome 17⁹; in such cases, tau-positive inclusions are characteristically found in the neurons and glial cells. Other pathologic substrates of FTD include dementia that lacks distinctive histologic features and corticobasal degeneration.¹⁰

The identification of reliable in vivo predictors of tissue pathology is essential for the development, evaluation, and monitoring of disease-modifying therapies in FTD. In a recent clinicopathologic study of FTD,⁷ the clinical features of corticobasal degeneration, FTD with motor neuron disease, and PNFA were found reliably to predict tissue pathology at autopsy. However, tissue pathology was generally unpredictable in fvFTD and SD, suggesting that clinical classification alone is not sufficiently reliable to guide specific disease-modifying therapies across the FTD spectrum. Volumetric brain MRI can reliably distinguish Alzheimer and non-Alzheimer dementias based on patterns of regional atrophy⁴; however, the role of MRI techniques in the identification of specific tissue pathology in FTD remains contentious.^11-13

We undertook this study to evaluate retrospectively the potential of brain MRI to predict specific pathologic substrates of FTD. Different pathologic subgroups were compared using 2 independent MRI analyses: the unbiased and automated technique of voxel-based morphometry (VBM) and blinded visual assessment of individual MRI studies by an experienced neuroradiologist (J.M.S.).

From a series of patients with dementia who had undergone postmortem examination or in vivo brain biopsy at the Institute of Neurology, University College London, or the Institute of Psychiatry, King’s College London, between 1990 and 2003, we identified all cases of pathologically confirmed FTD with at least 1 volumetric brain MRI study. Two cases with a recognized mutation in exon 10 of the tau gene (exon 10+16) and pathologic confirmation in a sibling were also included. We evaluated 21 FTD cases and 20 age- and sex-matched healthy controls. Cases were classified as FTD-U (n = 9), PiD (n = 7), or familial FTD with tau exon 10+16 mutations (tau exon 10+16) (n = 5); other pathologic diagnoses were not included because numbers were insufficient to support group analyses. A family history of dementia was present in all tau exon 10+16 cases and 3 FTD-U cases. Clinical phenotype,² disease duration, Mini-Mental State Examination (MMSE) score,¹⁴ and Clinical Dementia Rating¹⁵ at the time of MRI were ascertained from records. Informed consent was obtained for all patients, and the study was conducted with the approval of the local ethics committee.

In the autopsy cases, half of each brain was fixed in 10% buffered formalin for at least 4 weeks. Each fixed half-brain was then cut into 0.5-cm-thick slices in the coronal plane, and tissue blocks were selected from representative areas. After processing and paraffin wax embedding, sections were examined using routine techniques including hematoxylin-eosin, Luxol fast blue, cresyl violet, and modified Bielschowsky silver. Immunohistochemical staining was performed with the following antibodies: neurofilament proteins (RT97 and BF10) (B. H. Anderton, PhD, Department of Neuroscience, Institute of Psychiatry, King’s College London); β-amyloid (Dako, Carpinteria, Calif); ubiquitin (Dako); tau 12E8 (Elan Corporation, San Francisco, Calif); paired helical filament tau (AT8; Autogen Bioclear UK Ltd, Wiltshire, England); αβ-crystallin (Novocastra, Newcastle upon Tyne, England); α-synuclein (Diane Hanger, PhD, Department of Neuroscience, Institute of Psychiatry, King’s College London); and prion protein (12F10; G. Hunsmann, MD, German Primate Center, Göttingen, Germany, or KG9, Institute of Animal Health, Compton, England, and 3F4, Senetek PLC, Napa, Calif). In 2 cases, tissue was obtained in vivo from a full-thickness right frontal biopsy specimen fixed in 10% formalin and processed in paraffin wax using standard procedures. Sections were prepared as described herein.

Pathologic diagnosis was based on consensus criteria.² The diagnosis of FTD-U was based on the presence of ubiquitin-positive, tau- and α-synuclein–negative, abnormal neurites and neuronal, intracytoplasmic inclusions in the frontotemporal cortices or the hippocampal dentate fascia. None of the cases classified as FTD-U in this series had clinical or pathologic characteristics of motor neuron disease. The diagnosis of PiD was based on the presence of silver-, tau-, and ubiquitin-positive, intraneuronal Pick bodies in the frontal and temporal cortices and subcortical gray matter structures, including the amygdala, basal ganglia, and brainstem nuclei. Cases in the tau exon 10+16 group all had a family history of FTD and tau exon 10+16 C-to-T (cytosine-to-thymine) splice site mutations (OMIM 157140.0006 [microtubule-associated protein tau, IVS10, C-to-U, +16]). These cases were included in a previous clinicopathologic series.¹⁶

T1-weighted volumetric MRI studies were acquired using a 1.5-T Signa Unit (General Electric Medical Systems, Milwaukee, Wis) with a spoiled gradient echo technique (time to echo, 5 milliseconds; time to repeat, 35 milliseconds; flip angle, 35°; field of view, 24 × 24 × 19.2 cm; 124 contiguous 1.5-mm-thick coronal slices).

Analysis was performed using SPM99 statistical software (http://www.fil.ion.ucl.ac.uk/spm) with MATLAB 6 (http://www.mathworks.com). An optimized method of VBM was used.¹⁷ Brain MRI studies from 10 healthy controls and 10 patients with a neurodegenerative disease (5 with FTD and 5 with Alzheimer disease), age and sex matched to the study cohort and acquired on the same scanner, were normalized to the Montreal Neurologic Institute standard brain.¹⁸ The mean template image was smoothed with an isotropic gaussian kernel of 8 mm full width at half maximum. All study images were normalized to this customized template, and each normalized image was segmented into gray matter, white matter, and cerebrospinal fluid. Each gray matter image was masked with a brain region to exclude all nonbrain voxels; brain segmentations were performed using a semiautomated technique based on MIDAS image analysis software.¹⁹ Gray matter images were modulated and smoothed with an isotropic gaussian kernel of 8 mm full width at half maximum. The smoothed images were analyzed using a single subject condition and covariate model that incorporated age, sex, and MMSE score (disease severity) as covariates.

Regional gray matter density was compared among pathologic subgroups and between each pathologic subgroup and controls. Gray matter differences were assessed both at an uncorrected statistical threshold (P<.0001) and after correction for multiple comparisons for the whole brain volume (P<.05). Masking procedures were used to identify regions of gray matter atrophy that were common to all pathologic subgroups and regions that were unique to each subgroup; these procedures applied Boolean criteria to every voxel in the thresholded statistical parametric map for each contrast. Inclusive masking was used to identify voxels at which gray matter atrophy occurred in every FTD subgroup relative to controls; exclusive masking was used to identify voxels at which atrophy occurred only in a particular FTD subgroup relative to controls.

Each MRI study was graded for regional atrophy by an experienced neuroradiologist (J.M.S.) blinded to the pathologic diagnosis and clinical information other than age. Brain regions assessed included the entire temporal lobe, the hippocampus and amygdala, and the frontal, parietal, and occipital lobes. Regional brain atrophy was graded numerically as absent (0), mild (1), moderate (2), or severe (3) and normalized for the mean atrophy score (corrected score equals rating for that region minus mean rating across all regions for that patient). Pathologic subgroups were compared using the Mann-Whitney test, and corrected atrophy scores were compared between hemispheres using the Wilcoxon signed rank test.

Characteristics of controls and patients with FTD are given in Table 1. Groups were well matched for age, handedness, sex, and disease duration. Most patients had a clinical diagnosis of fvFTD, approximately equally distributed among the pathologic subgroups; 3 of the 4 patients with SD had FTD-U, and the single patient with PNFA had PiD.

Compared with controls, the FTD-U group showed leftward asymmetric frontal and temporal lobe atrophy; the PiD group showed severe and extensive bifrontal atrophy, with milder atrophy of the temporal lobes; and the tau exon 10+16 group showed rightward asymmetric atrophy that involved the anterior and medial temporal lobes and orbitofrontal cortex (Figure 1 and Table 2). All 3 pathologic subgroups showed atrophy that involved the middle and inferior temporal gyri, medial temporal lobes, insula, orbitofrontal cortex, and subfornical regions (Figure 1, yellow). Atrophy unique to the FTD-U group involved the orbitofrontal cortex, posterior superior temporal lobe, and posterior fusiform gyri bilaterally; atrophy was more severe in the left hemisphere (Figure 1, green). Atrophy unique to the PiD group predominantly involved the dorsolateral frontal regions bilaterally (Figure 1, blue). Atrophy unique to the tau exon 10+16 group predominantly involved the right medial temporal lobe (Figure 1, red). On direct statistical comparison among pathologic subgroups (threshold P<.0001 uncorrected), the PiD group showed significantly more atrophy in the bilateral frontal regions than the FTD-U and tau exon 10+16 groups, and the tau exon 10+16 group showed significantly more atrophy in the right medial temporal lobe than the FTD-U and PiD groups; differences between the FTD-U group and the other pathologic subgroups did not reach statistical significance.

On visual assessment, significant intergroup differences in the distribution of atrophy were found despite individual variation (Figure 2). The FTD-U group had significantly greater atrophy of the left temporal lobe than the PiD group (P<.04), the PiD group had significantly greater atrophy of the left frontal lobe than the FTD-U group (P<.03), and the tau exon 10+16 group had significantly greater atrophy of the right amygdala and right hippocampus than both the FTD-U and PiD groups (all P<.04). The FTD-U group had significantly greater left than right amygdala atrophy (P<.02). The positive predictive value, sensitivity, and specificity of the VBM signatures of pathology were estimated from the individual visual assessment scores for each pathologic subgroup with respect to the other subgroups (Table 3). The criteria for radiologic signatures were FTD-U, leftward asymmetric selective temporal lobe atrophy; PiD, bifrontal atrophy; and tau exon 10+16, asymmetric right medial temporal lobe atrophy.

This study analyzed MRI patterns of brain atrophy in pathologically defined FTD subgroups using the parallel, unbiased techniques of VBM and blinded visual assessment. Our findings suggest that MRI signatures of tissue pathology in FTD can be identified at the time of clinical presentation. We found that FTD-U was associated with predominantly left temporal lobe atrophy, PiD was associated with more severe bifrontal atrophy, and tau exon 10+16 was associated with relatively focal right medial temporal lobe atrophy. Certain brain regions (including the right inferior temporal lobe, orbitofrontal cortex, and insula) were involved across FTD subgroups and may have less discriminative value.

Macroscopic examination at autopsy of the brains of patients with FTD has shown predominantly frontal and temporal and, less commonly, parietal atrophy.^8,16 Although patterns of regional synaptic loss may correlate with neuropsychological and behavioral deficits in FTD,²⁰ the macroanatomical correlates of histopathologically or immunohistochemically defined FTD subgroups are heterogeneous.^7,11,21 Autopsy descriptions of asymmetric temporal lobe atrophy in FTD-U,¹³ severe frontal atrophy in PiD,²¹ and severe medial temporal lobe atrophy in tau exon 10+16¹⁶ are consistent with the present MRI findings. However, histopathologic studies have demonstrated severe involvement of the amygdala in PiD⁸ and diffuse involvement (without medial temporal lobe emphasis) in tau exon 10+16 cases.¹⁶ These apparent discrepancies from the present study may indicate that the microscopic distribution of tissue pathology does not translate simply to macroscopic atrophy or might arise from the comparison of different disease stages between studies. By the time of autopsy, atrophy is likely to be more widespread and group differences that may have existed at an earlier stage of disease may have been obscured. Ideally, imaging measures of regional atrophy in FTD should be directly correlated with quantitative pathologic measures of regional tissue loss^11,20 and other indexes of regional disease involvement (for example, density of inclusions) at autopsy. Such quantitative pathologic correlation was not available for the cases included in the present study but represents a clear direction for future work.

Several previous MRI studies have found distinctive patterns of atrophy in clinically defined subgroups of FTD.^3-5 In the present study, cases were selected on pathologic rather than clinical grounds; the MRI signatures we have identified are therefore contingent on specific tissue diseases rather than their clinical associations per se. Clinicopathologic correlations in FTD are imprecise,⁷ whereas the correlation between clinical and radiologic findings is often difficult. The present study provides complementary information that could be used to predict tissue pathology when this is not possible on clinical grounds alone (for example, in patients who present with fvFTD).

The present findings were obtained from a relatively small group of well-characterized cases and may not generalize to more heterogeneous cohorts; the findings should be validated prospectively in larger populations. It is unclear whether the patterns of atrophy found in the tau exon 10+16 group are also associated with other tau mutations, given the substantial clinical and pathologic heterogeneity of familial FTD.^12,16 It is difficult to assess disease severity and duration reliably in FTD; the MMSE is particularly sensitive to language deficits and the Clinical Dementia Rating to memory impairment, whereas behavioral change is difficult to quantify. Moreover, the technique of VBM has potential limitations when applied to atrophic brains.²²

Despite these caveats, independent blinded visual assessment of atrophy supported the findings on VBM, although the convergence between the 2 methods of evaluation varied depending on the pathologic subgroup (Table 3). On visual rating, the VBM criterion of leftward asymmetric temporal lobe atrophy reliably predicted FTD-U; bifrontal atrophy had relatively high sensitivity and specificity (but poor positive predictive value) for PiD; and the VBM criterion of rightward asymmetric medial temporal atrophy had poor sensitivity and positive predictive value for tau exon 10 + 16. Medial temporal atrophy was clearly bilateral on visual rating in tau exon 10 + 16 (Figure 2), with only a weak trend for rightward asymmetry. This apparent discrepancy between VBM and visual rating methods may reflect a wider individual variation in left than right medial temporal lobe atrophy. Because of the small numbers in each pathologic subgroup, caution is needed in extending these findings to larger populations. A further issue, not addressed directly herein, concerns the usefulness of these findings in distinguishing FTD from other causes of dementia, in particular, Alzheimer disease. Nevertheless, the present findings suggest that certain radiologic patterns may assist in predicting pathologic diagnosis in patients with FTD. The variation between individuals and overlap between pathologic FTD subgroups suggest that these findings should be interpreted in conjunction with clinical and other indexes⁷ in the individual case. However, such population-level signatures could be incorporated into the design and analysis of clinical trials.

Correspondence: Jason D. Warren, FRACP, Dementia Research Centre, Institute of Neurology, University College London, 8-11 Queen Square, London WC1N 3BG, England (jwarren@dementia.ion.ucl.ac.uk).

Accepted for Publication: January 26, 2005.

Author Contributions:Study concept and design: Whitwell, Rossor, Fox, and Warren. Acquisition of data: Whitwell, Josephs, Stevens, Revesz, Holton, and Al-Sarraj. Analysis and interpretation of data: Whitwell, Rossor, Godbolt, Fox, and Warren. Drafting of the manuscript: Whitwell, Fox, and Warren. Critical revision of the manuscript for important intellectual content: Whitwell, Josephs, Rossor, Stevens, Revesz, Holton, Al-Sarraj, Godbolt, Fox, and Warren. Obtained funding: Whitwell and Fox. Administrative, technical, and material support: Josephs, Rossor, and Warren. Study supervision: Fox and Warren.

Funding/Support: Ms Whitwell received financial support from the Special Trustees of University College London Hospitals Trust and National Hospital for Neurology and Neurosurgery. Dr Fox holds a Medical Research Council Senior Clinical Fellowship. Dr Josephs was a Mayo Foundation scholar to the Institute of Neurology, Dementia Research Centre, and the Queen Square Brain Bank, Department of Molecular Neurosciences, Institute of Neurology, University College London, London, England. Dr Warren is supported by European Union Grant LSHM-CT-2003-503330 to the Abnormal Proteins in the Pathogenesis of Neurogenerative Disorders (APOPIS) Consortium.

Acknowledgment: We thank the patients and their families for participation; Hilary Watt, MSc (Medical Statistics), and Chris Frost, MA, DipStat (Cantab), for statistical advice; and Peter Lantos, MD, for conducting initial pathologic examinations in some cases.