Introduction

Although described in the original papers of Arnold Pick and other early reports of similar conditions, the recognition of primary progressive aphasia (PPA) as a relatively common syndrome followed the seminal work of Marsel Mesulam [45]. He defined PPA as a gradual isolated dissolution of language, contrasting with the preservation of other cognitive functions and behavior [46]. The structural basis of PPA has remained controversial [33].

Neuropathological examination of autopsied cases or of brain biopsies have shown that PPA is most commonly associated with different forms of frontotemporal lobar degeneration (FTLD) [5], including FTLD with Pick bodies [14, 20, 31, 66], corticobasal degeneration (CBD) [9, 22, 39], FTLD with motor neuron disease-type inclusions or ubiquitin-only inclusions (FTLD-U) [34, 60], and dementia lacking distinctive histopathology [7, 15, 35, 44, 57, 59, 61], although the latter diagnosis, always questionable, cannot even be considered in the absence of ubiquitin immunostains.

On the other hand, a number of reported cases of PPA have shown as the only recognized pathology neurofibrillary tangles and senile plaques in a density sufficient for the diagnosis of Alzheimer’s disease (AD) by current criteria. As many as one third of patients diagnosed clinically as either progressive non-fluent aphasia (PNFA) or semantic dementia showed this pattern [36]. Some reports have remarked the relative sparing of medial temporal lobe structures (hippocampus and entorhinal cortex) [16, 17, 56]. Galton et al. [10] have reported a large series of atypical presentations of pathologically confirmed AD, which included six cases of PPA. Non-fluent, mixed, and fluent aphasia were represented. Although episodic memory deficits were present in most patients, there were exceptions. Thus, neither a clinical profile of patients with PPA and AD pathology, nor a pathological difference between cases with AD pathology with and without PPA symptoms has been reported.

We first encountered an opportunity to make a contribution to the questions above when we classified 13 patients with PPA into a probable and a possible group, depending on whether non-language memory deficits were present or absent at onset. All ten patients with definite PPA showed FTLD pathology (3 FTLD with Pick bodies, 4 CBD, 3 FTLD-U). In contrast all three patients with memory deficits showed AD pathology [52]. However, there was a surprising lack of correspondence between the widespread AD pathology and the localized clinical manifestation. Later, two of us (DM and JW) observed a distinctive type of glial abnormality in the brains of these patients, and we investigated the presence and distribution of this lesion in all available brains with a reliable diagnosis of possible PPA and a group of AD patients without possible PPA.

Patients and methods

Among the patients examined by one of us (AK) at the Cognitive Neurology Clinic of the University of Western Ontario, 20 who presented with PPA came to autopsy prior to the end of 2002, and the brains were examined by one of the authors. 12 of these received a diagnosis of probable PPA, when aphasia, whether anomic, logopenic, or non-fluent was the only initial symptom. The pathological diagnosis was one of the varieties of FTLD: FTLD with Pick bodies, CBD, or FTLD-U [32, 53]. Eight patients received a diagnosis of possible PPA because the initial history was considered to have a component of memory impairment accompanying language deficits: they are the subject of the present report. All patients were assessed with the Western Aphasia Battery (WAB) [29]. Other tests used included the mini mental state examination (MMSE), and Mattis dementia rating scale (DRS). The frontal behavioral inventory (FBI) was obtained with the main caregiver.

For comparison, we used six patients with a clinical diagnosis of probable Alzheimer’s disease by NINCDS criteria [42] in whom review of the clinical history did not reveal any suggestion of PPA, and who received a pathological diagnosis of Alzheimer disease by CERAD criteria [47].

Brains removed at autopsy were weighed, fixed in formalin for 1–2 weeks, and coronally sectioned. Standard blocks were embedded in paraffin, and 8 μm sections stained using the H&E, Luxol Fast Blue, Bielschowsky, and Gallyas techniques [50]. Immunostains were performed on 4 μm sections by the avidin–biotin–peroxidase method, using Vector Elite kit and diaminobenzidine (DAB) as previously described [40]. Appropriate positive and negative controls were used in all cases. For detecting pathological tau proteins we used two mouse monoclonal antibodies, one conformation-dependent, tau-2 (Sigma, 1:200), directed against Ala95 through Ala108 of bovine tau, and specifically for the Ser conformation of the amino-terminal portion of human tau [65], and one phosphorylation-dependent, AT8 (Pierce Endogen, 1:100), that recognizes tau protein phosphorylated at both ser 202 and thr 205, [12] (Innogenetics, Ghent, Belgium). Other antibodies used include anti GFAP (Dako, 1:500), anti ubiquitin (Dako, 1:50), anti α-synuclein (Zymed, 1:50), and anti Chromogranin A (Dako, 1:2,000). Double labeling was carried out by the sequential method, using a brown chromagen (DAB) for the monoclonal tau antibodies and a purple chromagen (VIP substrate, Vector laboratories) for GFAP. Runs omitting each primary antibody were used as controls.

In the initial evaluation all the clinical information was available to the pathologists. In order to confirm the consistency of the finding reported here ten slides showing argyrophilic thorny astrocyte clusters (ATAC) and ten slides from corresponding areas in the AD controls were masked and given in random order to one of the pathologist (DM) who correctly identified the presence or absence of ATAC in all cases.

Results

Clinical

The common feature of the history was the salience of language impairment in contrast to the relative preservation of activities of daily living, navigation, calculation, and planning, even if episodic memory was definitely involved from the outset. For example, a 75-year-old woman was able to drive her car around town without difficulty, depositing or withdrawing money from the bank, maintaining her house, cooking and baking at a time when her language was reduced to garbled speech. The onset was insidious and progression gradual in all cases. On examination patients were attentive and cooperative. Word finding difficulty was more apparent on connected speech than on naming. Frequent paraphasias included mostly word substitutions. Impairment of both comprehension and repetition initially involved complex sentences and eventually short ones as well. The following written response to a question about the weather demonstrates the same severe semantic jargon and paragrammatism found in spontaneous speech: “They must dry sometimes wet in (town) in snow after in the monthly”. Spontaneous speech fluency varied among patients, depending on the stage they were examined, but characteristically fluency was preserved late into the illness in contrast to PNFA. Constructional apraxia was common, but limb apraxia was not a feature. Preservation of motor skills was exemplified in patients taking long bicycle rides. Many patients developed behavioral symptoms. Some of these such as the mirror and Capgras delusion are in our experience more characteristic of Alzheimer’s than Frontotemporal Disease (mirror = confusion of a mirror image as a stranger, Capgras = delusion of a close relative as an impostor). The FBI scores averaged 18 (below frontotemporal dementia cutoff, but compatible with PPA or AD).

Eventually patients progressed to a severe global dementia and some to mutism. The clinical features of the eight patients are summarized in Table 1. The WAB scores (for quantization of language features) in the table are from the initial examination. The pattern suggested fluent aphasia ranging from anomia to Wernicke’s aphasia, more characteristic of AD than PPA [1, 30]. Other cognitive test averages/possible maximum: MMSE 8.5/30, DRS 72.8/144 are compatible with other PPA or severe AD patients. The impaired language skills are the likely cause of the low MMSE scores.

Table 1 Clinical features
Full size table

Structural brain imaging with CT or MRI showed mild diffuse atrophy in four patients and focal atrophy accompanied by distinct asymmetry in the other four (Fig. 1). In most cases HMPAO-SPECT showed parietal hypoperfusion.

Fig. 1
figure 1

Magnetic resonance images of patients at presentation. a Patient 7, 74-year-old female, T1 weighted image, showing distinctly asymmetric atrophy preferentially involving the left cerebral hemisphere. b Patient 1, 75-year-old male, T2 weighted image, demonstrating atrophy primarily affecting the left Sylvian region

Full size image

Pathology

Gross exam of the brain in all cases showed diffuse, moderately severe cortical atrophy lacking any obvious lobar accentuation, but often with perisylvian predominance (Fig. 2a), and on serial sections ventricular dilatation and severe hippocampal atrophy (Fig. 2b). In only one case was the atrophy considered grossly asymmetrical, left predominant. On microscopic examination using either Gallyas (Fig. 3a) or tau immunostains (Fig. 3b), all cases showed sufficient neurofibrillary tangles and neuritic plaques to be assigned to the “high likelihood that the dementia was due to AD” category by the NIA-Reagan criteria [21]. The regional distribution of lesions was consistent with Braak stages 5–6. In one case, a woman aged 70 at the time of death (patient #8), no additional histological abnormalities were present, other than those expected in AD, such as amyloid angiopathy. The other seven cases showed on tau immunostains (Fig. 4) and Gallyas stains (Fig. 5) large patchy compact formations in the cortex and subcortical white matter, made up of argyrophilic thorny astrocytes (ATA). We have named these structures ATAC (Figs. 4a, 5a). ATA had an eccentric nucleus and plump perikarya with a thorny outline, the latter labeled by both tau-2 and AT8 tau antibodies in a patchy pattern of coarse dense strands or clumps, usually on lighter stained background (Fig. 4b). Also stained were a variable proportion of ATA’s long varicose processes, often showing sucking feet on reaching the wall of blood vessels (Fig. 4c). ATA did not send processes to neighboring neuritic plaques (Fig. 4d, e). Tau-2 ATA staining was distinctly less intense than neurofibrillary tangles (Fig. 4f), but this was not obvious on AT8 staining. The Gallyas technique detected the same number of clusters and of astrocytes in the clusters as the tau immunostains (Fig. 5a). It stained either the entire perikaryon (Fig. 5b) or only some strands (Fig. 5c). The astrocytic nature of the argyrophilic cells was supported by the formation of vascular end-feet (Fig. 5d). When in the cortex ATA were not topographically related to tangle-bearing neurons (Fig. 5e). ATA were not stained by the Bielschowsky technique. Double labeling experiments with GFAP and tau (both tau-2 and AT8) showed some cells intensely expressing both markers, but in most tau-expressing cells GFAP staining is limited to bundles at the perikaryal periphery and the processes (Fig. 6). Astrocyte morphology and areal density was similar regardless of tau expression, and weak or intense GFAP labeling. No immunoreactivity to either ubiquitin or α-synuclein was found in ATAC.

Fig. 2
figure 2

Gross pathology. a Patient 3, 74-year-old male. Diffuse cortical atrophy lacking any obvious lobar pattern, but with perisylvian accentuation. b Coronal sections demonstrating cortical and hippocampal atrophy with no obvious asymmetry

Full size image
Fig. 3
figure 3

Alzheimer pathology. The middle frontal gyrus shows neurofibrillary tangles (black arrow) and neuritic plaques (white arrow), utilizing the Gallyas silver stain (a) or tau (AT8) immunostains (b). Two neurofibrillary tangle-bearing neurons are enlarged in the insets. Calibration bars a 200 μm, b 50 μm

Full size image
Fig. 4
figure 4

Argyrophilic thorny astrocyte clusters (ATAC) demonstrated with tau immunostains. a A large ATAC (white asterisk at its center) is present and the subcortical white matter near the grey–white junction. Some extensions form collars around blood vessels (black arrow). Inferior frontal cortex, AT8 immunostain. Bar 1 mm. b The cells in ATAC show a diffuse staining of patchy intensity of stout perikarya with a thorny outline (two labeled with black arrows). Tau-2 immunostain. Bar 20 μm. c The perivascular collars (asterisk on vessel) are formed by astrocytic processes and their end-feet. AT8 immunostain. Bar 20 μm. d Unlike the neighboring vessel (white arrow), neuritic plaques in the cortex (black arrow) are not associated with astrocytic process, even if immunolabeled astrocytes (red arrow) are present in the neighborhood. Tau-2 immunostain. Bar 100 μm. e Higher magnification of d, slightly rotated. The same plaque, with its tau positive neurites, is arrowed. Bar 20 μm. f Thorny astrocytes (black arrow) are less intensely stained by tau-2 than neurofibrillary tangles (red arrow). Bar 20 μm

Full size image
Fig. 5
figure 5

Argyrophilic thorny astrocyte clusters demonstrated with the Gallyas stain. a Two ATAC (arrows) in the subcortical white matter, frontal lobe. Bar 200 μm. b A thorny astrocyte with diffuse perikaryal staining, albeit of patchy intensity. Bar 10 μm. c Staining is limited to strands in the perikarya of these two thorny astrocytes. Bar 10 μm. d The processes of some thorny astrocytes (arrow) form end-feet on blood vessels. Bar 10 μm. e Thorny astrocytes in the cortex in the vicinity of a neuron (arrow) lacking neurofibrillary pathology. Bar 10 μm

Full size image
Fig. 6
figure 6

Double labeling for GFAP (purple, black arrows) and tau-2 (brown, red arrows). a Most astrocytes within a cluster predominantly express either GFAP or tau. Bar 40 μm. b–d Co-expression of GFAP and tau is visible in a subset of astrocytes. GFAP may be condensed in bundles at the periphery of the perikaryon (b), in the processes (c), or occupy much of the core of the cell body (d). Bar 10 μm. e The morphology of processes forming vascular end-feet is similar in astrocytes chiefly expressing either tau or GFAP. Bar 20 μm. f–g Controls. After omitting the primary antibody for tau, only purple GFAP stained astrocytes are seen (f), whereas omitting the GFAP primary results in brown only staining for tau (g). Bars f 40 μm, g 20 μm

Full size image

The clusters as a whole (ATAC) had a compact oval shape and measured from a few hundred micrometers to 5–10 mm in diameter. When located in the subcortical white matter their margin was parallel to the grey–white junction (Fig. 4a). Blood vessels within an ATAC were surrounded by a halo of argyrophilic, tau immunoreactive processes, but ATAC showed no relationship to subpial or subependymal collections of thorn-shaped astrocytes. White matter areas containing ATAC demonstrated no discernible change on myelin stains, and ATAC could not be delineated on GFAP immunostains in the cortex. ATAC showed no consistent topographic correlation with the density of amyloid deposits or neurofibrillary tangles.

The areas involved were the white and grey matter of the temporal lobe, including the gyrus ambiens, parahippocampal gyrus, fusiform gyrus, inferior, middle, and superior temporal gyri, the frontal dorsolateral and orbitofrontal cortices, the cingulate gyri and occasionally the inferior parietal lobule. They were not found in the insula, occipital lobe, hippocampus, amygdala, thalamus, basal ganglia, brainstem, or cerebellum. There were no obvious focal accentuations of the severity of AD pathology, or differences between right and left cerebral hemispheres. Neither tufted astrocytes nor glial plaques were seen. Additional neurodegenerative features were the presence of argyrophilic grains and in the parahippocampal gyrus ballooned neurons in a 83-year-old female (patient #6), and of neurofibrillary tangles in the substantia nigra and periaqueductal grey matter of a 74-year-old male (patient #3). Severe subicular atrophy with relative sparing of the hippocampus was observed in patients #3 and #6. No infarcts were identified on gross examination in any patients, but microscopic infarcts (less than 0.5 mm in diameter) were observed in the cortex of patients #1 and #5, and in the left thalamus of patient #5.

The group of patients showing ATAC consisted of four men and three women. The average age at onset for men and women was 67.5 and 69.7 years, and at death 76.3 and 79.7, resulting in an average duration until death of 8.8 and 10 years, respectively. Brain weight averaged 1,145 g for men, and 1,008 g in women.

None of the six cases of AD without PPA used as the comparison group showed ATAC (Fig. 7).

Fig. 7
figure 7

Alzheimer’s disease patient without aphasia. This section, including the depth of the collateral sulcus (top left) and the CA1 sector of the hippocampus immunostained with AT8, shows abundant plaques and tangles, but complete absence of subpial, subependymal, or white matter tau deposits in astrocytes. Bar 1 mm

Full size image

Discussion

Although the literature does not describe this association, finding AD pathology as the pathological substrate of possible PPA is not surprising, providing a known basis for the memory impairment accompanying the aphasia. Recently, subtypes of PPA have been identified on clinical, imaging, and genetic bases: PNFA, semantic dementia, and logopenic progressive aphasia. The latter is characterized by anomia, slow speech, and impaired syntactic comprehension clinically, and left posterior temporal cortex and inferior parietal lobule atrophy on imaging. In contrast to the two other forms of PPA, there is over-representation of the ApoE ε4 allele, as high as 67%, suggesting a link to AD [13]. The relationship of this form to pathologically identified AD remains to be seen. In the present series the clinical characteristics in the first assessment were closer to the fluent aphasia seen in Alzheimer’s disease than the nonfluent aphasia associated with FTLD. Many patients at early stages of PPA and Possible PPA are fluent but the main distinguishing feature is the episodic memory loss in possible PPA. Semantic dementia patients remain fluent but maintain episodic memory while loosing semantic memory and comprehension. The speech at the last assessment was nonfluent in four patients with ATAC and still fluent in the patient without ATAC. Aphasia was more prominent in all the possible PPA patients than those diagnosed as AD clinically.

The focal accentuation of the atrophy and asymmetry observed in imaging in half of the cases at presentation was much less prevalent at the time of autopsy, once the disease had run its course, a pattern commonly seen in FTLD. Interestingly, the patient without ATAC also showed left predominant atrophy.

Our second finding concerns the presence of ATAC in the majority of possible PPA cases. The constituent thorny astrocytes are a subtype of glial fibrillary tangles (GFT), argyrophilic masses containing deposits of abnormal tau proteins located in the cytoplasm of glial cells [37]. Electron microscopy typically shows bundles of 15 nm straight tubules [22]. All forms lack tau isoforms corresponding to the expression of alternatively spliced exons 2 and 3 [55], but otherwise are labeled with all tau antibodies. In contrast to neurofibrillary tangles, GFT are poorly labeled by antibodies to ubiquitin, and not at all by antibodies to apolipoprotein E, alpha-1-antichymotrypsin, and heparan sulfate [22]. Oligodendrocytes present a single form, named coiled bodies, whose processes appear as argyrophilic threads in the white matter [4], whereas at least three varieties can be identified in astrocytes. Two of these appear in protoplasmic astrocytes, identified by their CD44 positive, GFAP negative immunohistochemical profiles, are predominantly located in the cerebral cortex and subcortical grey masses rather than the white matter, and involve the processes rather than the cell body. Tufted astrocytes are characterized by fine process radiating from the inconspicuous cell body, whereas glial plaques consist of a crown of processes devoid of an obvious connection to a cell body. The other variety, named thorny astrocytes or thorn-shaped astrocytes, show a conspicuous GFAP positive cytoplasm drawn into funnel-shaped thick process, reaching as end feet perivascular subpial and subependymal linings [69].

In terms of the significance of GFT, coiled bodies are present in many taupathies, including dementia with argyrophilic grains [4, 25, 41], PSP [68], CBD [63], and subacute sclerosing panencephalitis [24]. Tufted astrocytes are considered pathognomonic of PSP [19, 67], and glial plaques are characteristic of CBD [8, 37, 38, 2],

The nosological significance of thorny astrocytes is less clear. Although initially described in small numbers in a limited area in patients with several diseases characterized by neuronal tau deposits [26], tau + thorny astrocytes in subpial and subependymal regions are common in the medial temporal lobe of elderly subjects, albiet only a small subset demonstrate argyrophilia. There is no correlation between the abundance of thorny astrocytes and neurofibrillary tangle stage, amyloid deposition, or argyrophilic grain disease severity. The presence of ApoE ε4 allele is not associated with the presence of thorny astrocytes [58]. On the other hand, abundant ATAs have been restricted to postencephalitic parkinsonism of von-Economo type [23] and three forms of frontotemporal degeneration, FTLD with Pick bodies, PSP, and diffuse neurofibrillary tangles with calcification. In FTLD with Pick bodies they were described in the subcortical white matter of atrophic middle and inferior temporal gyri [70], or in affected cortical areas [28, 71]. In PSP thorny astrocytes appear in areas with relatively minor neuronal loss including the striatum, thalamus, superior colliculus and frontal and motor cortex, but not in the white matter [54, 69]. In dementia with diffuse neurofibrillary tangles with calcification, which can be considered a subset of neurofibrillary tangle dementia [43], thorny astrocytes form subcortical and cortical clusters, similar to our ATAC [18]. However the stated coincidence of the distribution of thorn-shaped astrocytes and gliosis [18] is at odds with our observations on ATAC.

Finally, two cases of PPA in which astrocytic tau deposits constituted the dominant pathology have been reported. The patient described by Molina et al. [49] was an 80-year-old woman with PPA who in a left temporal lobe biopsy showed astrocytes stained by Gallyas and containing tau and ubiquitin positive inclusions located in cortex and white matter, predominantly at the grey–white junction, as in our cases. In addition, coiled bodies were present. There was no hint of Alzheimer disease, nor were any tau deposits found in neurons other than rare diffuse perikaryal staining. Ballooned neurons and superficial spongiosis were not present. The second case, reported by Wakabayashi et al. [64] in a 72-year-old man with a 6-year typical aphemic history, showed tufted astrocytes in layers II–V in the lower third of the left primary motor area, and to a lesser degree in perirolandic cortex bilaterally. The astrocytic inclusions were stained by phosphorylation dependent (AT8) and independent (tau-2) antibodies, but not by antibodies to alpha synuclein, or ubiquitin. There were numerous oligodendroglial inclusions (coiled bodies) with the same immunoreactive pattern in deep cortical layers and white matter. Superficial linear spongiosis was observed in the affected regions. Western blot analysis of insoluble tau proteins after dephosphorylation was reported as demonstrating three major bands corresponding to 4R and 3R isoforms, with a preponderance of 4R. Other findings included argyrophilic grains and a few neurofibrillary tangles in the hippocampus, but no neurofibrillary tangles in cortical or subcortical neurons. Thus, the authors did not consider that the patient’s disease could be classified as PSP, even if the astrocytic inclusions were similar to those seen in this disease [64]. Abundant cortical tufted astrocytes are a characteristic feature in PSP cases presenting with PNFA [3, 27, 48].

It should be mentioned that mutations in exons 1 and 10 and the intron following exon 10 of the gene MAPT are associated with robust astrocytic and oligodendroglial, in addition to neuronal, tau deposition [11].

Argyrophilic thorny astrocyte cluster represents a lesion not fully characterized in the literature. The basic process of deposition of tau in GFAP immunoreactive fibrous astrocytes is one shared with aging, but in contrast to the periventricular, subependymal, and subpial location of the thorny astrocytes described in normal aged subjects [58], ATAC are located in the cerebral cortex and subcortical white matter, are not restricted to the medial temporal lobe but involve the frontal, temporal, and parietal lobes, and display intense argyrophilia. Similar points of distinction in relation to normal aging were advanced in the report of the clusters of GFTs associated with diffuse neurofibrillary tangles with calcification [18], lesions closely related to ATAC. More research will be necessary to establish whether ATAC represent an intensification of the astrocytic changes seen in aging, or else a different phenomenon, possibly related to FTLD pathologies [51] This distinction is well known in neuronal tau deposits, where neurofibrillary tangles in AD replicate the biochemical composition and extend the anatomical distribution of those in normal aged brains, whereas neurofibrillary tangles in CBD or progressive supranuclear palsy do neither [6]. Other issue for future research is the range of conditions in which ATAC appear. Our preliminary observations indicate that ATAC are not found in FTLD-U with either PPA or behavioral presentations.

We hypothesize that ATAC represent a marker of a process, co-exiting but probably independent of AD, and responsible for the prominent focal clinical manifestations. The argument for the independence of ATAC and AD is grounded on cases in which subcortical thorny astrocytes are either the only reported pathology [49] or accompany FTLD with Pick bodies rather than AD [70]. In addition, the brains reported here showed no topographic relationship between the severity of plaques and tangles on the one hand and ATAC on the other. ATAC could be the result of pathogenetic processes related to those subjacent to FTLD. Alternatively, they could be secondary to alterations in the blood-brain barrier, a possibility raised in relation to the thorny astrocytes of normal aged brains [58], and compatible with ATAC’s patchy, nonsystematic distribution. Although our findings are consistent with either hypothesis, it should be noted that a related lesion, tufted astrocytes in PSP, are considered a degenerative rather than reactive process because their distribution is unrelated to reactive astrocytes in the vicinity of infarcts or neurofibrillary tangles [62]. The consistent development in aged baboons, but not other primate species, of tau + thorny astrocytes in a stereotyped distribution suggests that genetic variations contribute to the formation of ATAC. An important focus for future research is the relationship between the marker (ATAC) and the clinical manifestation (aphasia).