Neuropathological fingerprints of survival, atrophy and language in primary progressive aphasia

Abstract

Primary progressive aphasia is a neurodegenerative disease that selectively impairs language without equivalent impairment of speech, memory or comportment. In 118 consecutive autopsies on patients with primary progressive aphasia, primary diagnosis was Alzheimer’s disease neuropathological changes (ADNC) in 42%, corticobasal degeneration or progressive supranuclear palsy neuropathology in 24%, Pick’s disease neuropathology in 10%, transactive response DNA binding proteinopathy type A [TDP(A)] in 10%, TDP(C) in 11% and infrequent entities in 3%. Survival was longest in TDP(C) (13.2 ± 2.6 years) and shortest in TDP(A) (7.1 ± 2.4 years). A subset of 68 right-handed participants entered longitudinal investigations. They were classified as logopenic, agrammatic/non-fluent or semantic by quantitative algorithms. Each variant had a preferred but not invariant neuropathological correlate.

Seventy-seven per cent of logopenics had ADNC, 56% of agrammatics had corticobasal degeneration/progressive supranuclear palsy or Pick’s disease and 89% of semantics had TDP(C). Word comprehension impairments had strong predictive power for determining underlying neuropathology positively for TDP(C) and negatively for ADNC. Cortical atrophy was smallest in corticobasal degeneration/progressive supranuclear palsy and largest in TDP(A). Atrophy encompassed posterior frontal but not temporoparietal cortex in corticobasal degeneration/progressive supranuclear palsy, anterior temporal but not frontoparietal cortex in TDP(C), temporofrontal but not parietal cortex in Pick’s disease and all three lobes with ADNC or TDP(A).

There were individual deviations from these group patterns, accounting for less frequent clinicopathologic associations. The one common denominator was progressive asymmetric atrophy overwhelmingly favouring the left hemisphere language network. Comparisons of ADNC in typical amnestic versus atypical aphasic dementia and of TDP in type A versus type C revealed fundamental biological and clinical differences, suggesting that members of each pair may constitute distinct clinicopathologic entities despite identical downstream proteinopathies. Individual TDP(C) participants with unilateral left temporal atrophy displayed word comprehension impairments without additional object recognition deficits, helping to dissociate semantic primary progressive aphasia from semantic dementia. When common and uncommon associations were considered in the set of 68 participants, one neuropathology was found to cause multiple clinical subtypes, and one subtype of primary progressive aphasia to be caused by multiple neuropathologies, but with different probabilities. Occasionally, expected clinical manifestations of atrophy sites were absent, probably reflecting individual peculiarities of language organization.

The hemispheric asymmetry of neurodegeneration and resultant language impairment in primary progressive aphasia reflect complex interactions among the cellular affinities of the degenerative disease, the constitutive biology of language cortex, familial or developmental vulnerabilities of this network and potential idiosyncrasies of functional anatomy in the affected individual.

asymmetry, dementia, Alzheimer’s disease, tauopathy, frontotemporal lobar degeneration

For the podcast associated with this article, please visit https://academic.oup.com/brain/pages/podcast

Introduction

The literature on aphasia was initially dominated by accounts of acute language impairments caused by stroke. Different forms of aphasia, progressive rather than acute, started to be reported during the late 19th century but did not attract much attention.^1-5 The 1980s witnessed a resurgence of interest in these syndromes, which were named ‘primary progressive aphasia’ (PPA).^6,7 By 1992, a review of the emerging literature found 63 cases of PPA, 13 with post-mortem information.⁸ The two cardinal features of PPA, namely the heterogeneity of neuropathology and the asymmetry of neurodegeneration, were identified even in this small cohort. These two characteristics have subsequently been confirmed by pivotal autopsy series of PPA.^9–17

Alzheimer’s disease neuropathological changes (ADNC) and the three repeat tauopathy of Pick’s disease were the first two entities linked to PPA. Frontotemporal lobar degenerations (FTLD) with the four-repeat tauopathies of corticobasal degeneration (CBD) and progressive supranuclear palsy (PSP) were added to the list. As abnormalities in transactive response DNA binding protein-43 (TDP-43) were found to underlie what used to be called dementia lacking distinctive histopathology¹⁸ and FTLD with ubiquitin,^19,20 PPA became associated with three distinct forms of FTLD-TDP, types A–C.^21–23 Additional but less frequent neuropathologic linkages were also reported, including diffuse Lewy body disease,^24,25 encephalopathy with axonal spheroids,^26,27 FTLD-TDP type B,²² globular glial tauopathy²⁸ and argyrophilic grain disease.²⁹

Parallel developments were also unfolding in clinical characterization. The initial subdivision of PPA into progressive non-fluent aphasia and semantic dementia in 1998 led in 2011 to the further classification into three types: non-fluent-agrammatic, logopenic and semantic.^30,31 A fourth variant, mixed PPA, was added to encompass some of the 30–40% of individuals that could not be classified by the 2011 system.³² Clinicopathologic investigations revealed that logopenic PPA was most frequently associated with Alzheimer’s disease; non-fluent-agrammatic PPA with the three- and four-repeat tauopathies of FTLD (Pick’s disease, CBD and PSP); and semantic PPA with FTLD-TDP type C.^11,23,33,34

It soon became clear that the linkage of PPA variants to individual neuropathologic entities was probabilistic rather than deterministic, and that it reflected common as well as uncommon associations.^23,35 For example, the assumption that logopenic PPA is invariably associated with ADNC was challenged;³⁶ Pick’s disease neuropathology and globular glial tauopathy were identified as additional correlates of semantic PPA;^23,28,37,38 FTLD-TDP type C was reported in a patient with non-fluent-agrammatic PPA rather than semantic PPA;³⁹ and the same GRN mutation was shown to trigger two different forms of aphasia in siblings.⁴⁰ This clinicopathologic heterogeneity indicated that aphasia types are not invariably bound to the cellular nature of the underlying neurodegeneration, and that they reflect common and infrequent interactions among its preferred anatomy, the intrinsic biology of the language network and peculiarities of this network in the affected individual.

The preferential targeting of the hemisphere dominant for language (usually left) emerged as the only universal correlate of PPA neuropathology.^10,23,35,41 The cellular components of this asymmetry were quantitated for several neuropathologic entities.^37,42–46In vivo asymmetries of atrophy, for example, were shown to reflect asymmetrical distributions of neurofibrillary tangles in PPA with ADNC,⁴⁷ and microglia and TDP-43 precipitates in PPA with FTLD-TDP(A).^44,45 The mechanisms that underlie this selective vulnerability of the language-dominant hemisphere remain mysterious. One potential candidate is a familial vulnerability of the language network that leads to developmental delays of language acquisition in some members and selective vulnerability to independently arising neurodegenerative diseases in others.^48–51

The spectrum of clinicopathologic correlation in PPA is an evolving field of research. The complexity of these interactions, the multiplicity of PPA variants and the heterogeneity of the underlying proteinopathies justify the need for comprehensive and comparative analyses that can shed light on general mechanisms of selective vulnerability. The current report addresses these themes in 118 consecutive autopsies on systematically diagnosed PPA participants with longevity and demographic data, a subset of which was enrolled in a longitudinal research program that included uniform cognitive assessment, quantitative neuroimaging and neuropathologic evaluation.

Materials and methods

The Northwestern PPA Research Program registered 118 autopsies. Diagnosis was based on the isolated emergence of a progressive language disorder caused by a neurodegenerative process.^8,52 Two cases of FTLD-TDP(B), one of globular glial tauopathy and one of leukodystrophy with axonal spheroids were excluded from quantitative analyses, because they did not constitute groups of meaningful size. The remaining 114 cases were classified by primary neuropathologic diagnosis for onset age and survival. Of these 114, 67 participated in a longitudinal investigation with biennial cognitive assessment and quantitative imaging. One living participant with a GRN mutation and presumptive FTLD-TDP(A) was added to this set. The resultant 68 participants provided the basis for clinicopathologic characterizations. All 68 had identical cognitive assessment at entry. Sixty-one participants had quantitative imaging at the initial visit and 23 at the second, 2 years later. The clinical dementia rating scale⁵³ was used to assess global functionality. All participants were Caucasian English speakers and right-handed. The study was approved by the Institutional Review Board at Northwestern University and informed consent was obtained from all participants.

Language assessment

The aphasia quotient (maximum score 100) of the revised Western aphasia battery measured overall aphasia severity.⁵⁴ Additional tests assessed grammar, word comprehension, naming, repetition, object recognition, sentence comprehension and word fluency. Performance was expressed as a percentage of maximum scores, as controls have nearly perfect performance in each task.⁵⁵ For word fluency, a control value of 132 words per minute was used.³²

• Grammar: Grammar in sentence production was assessed with the Sentence priming production test and the Northwestern anagram test.^56,57 In the Sentence priming production test, the participant is shown reversible action pictures and asked to produce 15 non-canonical sentences (passive voice, object-extracted Wh-questions, object relatives); in the Northwestern anagram test, the participant is tested on the same sentence types (n = 15) by ordering single word movable tiles to match action pictures. As no verbal output is required, the influence of speech production ability is eliminated. Performance on these two sets of 15 noncanonical sentences was averaged to derive a composite Northwestern anagram/Sentence priming score of grammar.

• Word comprehension: Tested with 36 moderately difficult items (157–192) of the Peabody picture vocabulary test-IV.⁵⁸ Each item requires the participant to match an auditory word representing an object, action, concept or attribute to one of four picture choices.

• Naming: The Boston naming test was used to assess object naming through a 60-item standardized test in which items are administered in order of decreasing frequency of occurrence in the English language.⁵⁹

• Repetition: We selected the six most difficult items of the revised Western aphasia battery repetition subtest to generate the repetition score (Rep66).

• Object recognition: Non-verbal object knowledge was assessed with the picture version of the Pyramids and palm trees test⁶⁰ where the participant is asked to decide which of two pictures is conceptually more closely associated with a target object.

• Sentence comprehension: Comprehension of syntactically complex sentences was assessed with the Sentence comprehension test, where one of two reversible action scenes needs to be matched to a stimulus sentence spoken by the examiner.⁵⁶ Fifteen sentences of the type chosen for sentence production were used to measure non-canonical sentence comprehension.

• Fluency: Participants viewed a wordless picture book of the Cinderella story and were asked to tell the story. The narrative was entered into the Systematic analysis of language transcripts to measure fluency in words per minute.^61–64

Primary progressive aphasia classification

The consensus guidelines of 2011 provided the frame of reference.³¹ They do not specify the tests to be used or cut-offs, leave 30–40% of the cases unclassified and the same patient may fulfill criteria for two variants.^15,65,66 Modifications have therefore been implemented, including the addition of a ‘mixed’ variant (mixed PPA).²³ We used standardized tests, normative baselines and quantitative performance cut-offs for mild (79–60%), moderate (59–40%) and severe (<39%) impairment at the initial visit to implement a heuristic algorithm for classification into four variants as follows: (i) non-fluent-agrammatic PPA = (grammar <80%) AND (fluency <60%) AND (word comprehension ≥80%); (ii) logopenic PPA = (grammar ≥80% OR fluency ≥60%) AND (word comprehension ≥80%) AND (naming <80% OR repetition <80%); (iii) semantic PPA = (word comprehension <60%) OR (word comprehension <80% AND naming ≤40%) AND (grammar ≥60% OR fluency ≥80%); and (iv) mixed PPA = (grammar <80%) AND (fluency <60%) AND (word comprehension <80%).

This algorithm worked well in this cohort but may be challenging to apply at incipient or severe disease stages. Relative rather than absolute cut-offs might also be helpful for classification.

Imaging methods

T₁-weighted 3D MP-RAGE sequences (repetition time = 2300 ms, echo time = 2.91 ms, inversion time = 900 ms, flip angle = 9°, field of view = 256 mm) were used to acquire 176 slices 1.0-mm thick on a 3 T Siemens TIM Trio’s 12-channel birdcage head coil. Reconstruction was done with the FreeSurfer image analysis suite, version 5.1.^67,68 Geometric inaccuracies and topological defects were corrected by validated guidelines.⁶⁹ Cortical thickness maps of the PPA participants were contrasted against 35 right-handed cognitively healthy volunteers with a similar range of age and education to identify peak patterns of atrophy.⁷⁰ Differences in cortical thickness between groups were calculated by conducting a general linear model on every vertex along the cortical surface. False discovery rate was applied at 0.05 in individual maps and at 0.001 in group maps to detect areas of peak cortical thinning (i.e. atrophy) after adjusting for multiple comparisons.⁷¹

Neuropathological assessment

Histology sections were taken from nine homologous cortical areas of both hemispheres. They were processed with the Gallyas stain, thioflavin-S and immunohistochemistry for phosphotau (AT8), beta amyloid (4GR), TDP-43, p62 and alpha-synuclein (p129). Consensus criteria were used for the diagnoses of Alzheimer’s disease neuropathological changes (ADNC), Lewy body disease, FTLD-TDP (types A, B and C) and FTLD-tau (Pick’s-, PSP- and CBD-type).^{21,33,72–74} The primary diagnosis of ADNC was always associated with the A3B3C3 stage of neuropathology. Cases initially diagnosed as FTLD with ubiquitin were re-examined, stained for TDP-43 and classified according to current nomenclature. Cases were stratified by primary neuropathologic diagnoses. The CBD and PSP groups were combined because they are both 4R tauopathies and overlap neuropathologically and clinically.

Survival analysis and statistics

Onset time was determined by harmonizing three sources of information: the patient’s report, reliable informants among family members or friends and medical records. Survival rates were illustrated by Kaplan-Meier curves. The Cox proportional hazards model, where proportional assumption has been checked, controlled for the onset ages in each group. Benjamini-Hochberg adjusted P-values were used to account for multiple comparisons. We also estimated the sensitivity and specificity with which specific PPA variants or individual test scores of <60% (moderate-to-severe impairment) could predict underlying neuropathology (Supplementary Tables 1 and 2).

Data availability

Data is available pending collaboration requests that abide by Northwestern University policies.

Results

Clinicopathological frequency, onset age and survival

The onset age and survival data in Table 1 and Fig. 1 include the information from 114 autopsied cases (118 minus four with rare neuropathologies), whereas the imaging and test performance data in Tables 2 and 3 and Figs 2–5 represent the 68 participants (67 autopsied cases plus one living patient with a GRN mutation) who were recruited into the longitudinal investigation. The numbers of participants that contributed data to the language tests, initial MRIs and 2-year follow-up MRIs are indicated in Table 3 and Figs 2–5. Of the 118 autopsied PPA patients, those with ADNC comprised 42% of the cohort; CBD/PSP, 24%; Pick’s disease neuropathology, 10%; TDP(A), 10%; TDP(C), 11%; and rare diagnoses [TDP(B), globular glial tauopathy, leukodystrophy with axonal spheroids], the remaining 3%. The mean age of onset clustered around 60 years, with CBD/PSP cases being the oldest at 65.2 years and TDP(C) patients the youngest at 54.4 years (Table 1). The TDP(A) group had the shortest mean survival period (7.1 years), while the TDP-C group (13.2 years) had the longest (Fig. 1). Even when the results were controlled for age of onset in each group, the period of survival for patients with TDP(A) was shorter than that in the other four groups and that for cases with CBD/PSP was shorter than in the TDP(C) and ADNC groups. Gender distribution was uneven in PPA-ADNC (31 males, 18 females) and the two PPA-TDP groups (seven males, 18 females). All P-values in Table 1 were significant before adjustment for age of onset, and either significant or trending afterwards.

Figure 1

Kaplan-Meier curves of survival probability.

Figure 2

PPA with ADNC, CBD and PSP neuropathology. Yellow and red areas indicate regions of significant atrophy at a false discovery rate of 0.001. (A and B) Atrophy maps at initial visit and 2 years later in PPA-ADNC. (C) Scatter plot of performance in cognitive tests in PPA-ADNC at the initial visit. The horizontal lines indicate boundaries of mild, moderate and severe impairment. Scores on the y-axis indicate per cent of maximum score. (D and E) Atrophy maps at initial visit and 2 years later in PPA with CBD and PSP (PPA-CBD/PSP). (F) Scatter plot of performance in cognitive tests in PPA-CBD/PSP at the initial visit. The horizontal lines indicate boundaries of mild, moderate and severe impairment. Scores on the y-axis indicate per cent of maximum score. The asterisks in A and E mark the temporal pole. IFG = inferior frontal gyrus (Broca’s area); STG = superior temporal gyrus; TPJ = temporoparietal junction (Wernicke’s area).

Figure 3

PPA with Pick’s disease and TDP(A) neuropathology. Yellow and red areas indicate regions of significant atrophy at the false discovery rate of 0.001. (A and B) Atrophy maps at initial visit and 2 years later in PPA with Pick’s disease neuropathology (PPA-Pick’s disease neuropathology). (C) Scatter plot of performance in cognitive tests in PPA-Pick’s at the initial visit. The horizontal lines indicate boundaries of mild, moderate and severe impairment. Scores on the y-axis indicate per cent of maximum score. (D) Atrophy maps at initial visit in PPA-TDP(A). (E) Scatter plot of performance in cognitive tests in PPA-TDP(A) at the initial visit. The horizontal lines indicate boundaries of mild, moderate and severe impairment. Scores on the y-axis indicate per cent of maximum score. IFG = inferior frontal gyrus (Broca’s area); TPJ = temporoparietal junction (Wernicke’s area).

Figure 4

PPA with TDP(C), TDP(B) and glial globular tauopathy. Yellow and red areas indicate regions of significant atrophy at the false discovery rate of 0.001 for group maps in A and B and 0.05 for individual maps in D–G. (A and B) Atrophy maps at initial visit and 2 years later in PPA-TDP(C). (C) Scatter plot of performance in cognitive tests in PPA-TDP(C) at the initial visit. The horizontal lines indicate boundaries of mild, moderate and severe impairment. Scores on the y-axis indicate per cent of maximum score. (D) Atrophy map and test scores at initial visit in a participant with PPA-TDP(B). (E) Atrophy map and test scores at initial visit in a participant with glial globular tauopathy pathology (PPA-GGT). (F and G) Atrophy maps and test scores at initial visit and 4 years later in a participant with PPA-TDP(C). IFG = inferior frontal gyrus (Broca’s area); TPJ = temporoparietal junction (Wernicke’s area).

Figure 5

Individual cases with agrammatic and semantic PPA variants. Yellow and red areas indicate regions of significant atrophy at the false discovery rate of 0.05. IFG = inferior frontal gyrus (Broca’s area); TPJ = temporoparietal junction (Wernicke’s area).

Table 1

Age of onset and survival

Pathology	Age at onset (years)	Onset differs from	Survival (years)	Survival differs froma
ADNC (n = 49) M31, F18	60.8 ± 8.0 [47–80]	CBD/PSP+, TDP(C)+	10.8 ± 4.5 [3–27]	CBD/PSP+, TDP(A)**
CBD/PSP (n = 28) M13, F15	65.2 ± 8.0 [41–80]	ADNC+, Pick’s+, TDP(C)**	9.0 ± 2.5 [5–16]	ADNC+, TDP(A)+, TDP(C)+
Pick’s (n = 12) M5, F7	59.1 ± 6.6 [45–70]	CBD/PSP+	10.9 ± 3.7 [5–17]	TDP(A)*
TDP(A) (n = 12) M3, F9	60.2 ± 6.2 [50–68]		7.1 ± 2.4 [4–11]	ADNC**, CBD/PSP+, Pick’s*, TDP(C)*
TDP(C) (n = 13) M4, F9	54.4 ± 4.8 [47–63]	ADNC+, CBD/PSP**	13.2 ± 2.6 [8–17]	CBD/PSP+, TDP(A)*

Values are presented as mean ± SD [range]. Age of onset and survival based on the cohort of 114 cases (118 minus four with rare neuropathologies). CBD/PSP = corticobasal- and progressive supranuclear-type neuropathology; F = female; M = male; n = number of cases per neuropathology group. Pick’s = Pick’s disease neuropathology.

^a

Corrected for age at onset in individual path groups.

⁺

P = 0.053;

*

P < 0.01;

**

P < 0.001. All P-values were significant before adjustment for age of onset, and either significant or trending afterwards.

Table 2

Demographics and clinical characteristics of the cohort of 68 cases

	ADNC	CBD/PSP	Pick’s disease	TDP(A)	TDP(C)
Age at onset (years)	60.9 ± 7.4 (51–74)	66.2 ± 9.4 (41–80)	59.1 ± 4.6 (53–66)	60 ± 6.9 (50–68)	54.2 ± 5.5 (47–63)
Disease duration (years)	9.9 ± 4 (3–23)	8.8 ± 2.9 (5–16)	10.1 ± 3.9 (5–17)	6.8 ± 3 (4–11)	12.7 ± 2.7 (8–17)
Age at death (years)	70.8 ± 7.1 (60–85)	74.9 ± 9.4 (53–88)	69.3 ± 4.3 (64–76)	67.3 ± 8.5 (56–77)	66.9 ± 6.5 (55–77)
Symptom duration, initial visit (years)	4.8 ± 2.4 (1.5–10)	3.1 ± 1.7 (1–7)	4.1 ± 1.9 (2–8)	2.7 ± 1.3 (1.5–5)	4 ± 1.5 (2.5–6.5)
CDR global, initial visit	0:9, 0.5:17, 1:0	0:10, 0.5:5, 1:0	0:0, 0.5:7, 1:1	0 : 2, 0.5:3, 1:1	0:3, 0.5:4, 1:2
WAB AQ, initial visit	79.7 ± 15.1 (35.1–96.8)	83.8 ± 9.8 (61.1–96.8)	75.7 ± 9.4 (57–85.2)	65.9 ± 26.9 (25.2–96.8)	80.4 ± 10.7 (65.9–95.8)
WAB AQ, second visit	67.3 ± 22.4 (16.8–94.3)	54.5 ± 22 (9.3–84.1)	57.8 ± 22 (42.2–73.3)	–	65.6 + 14.4 (45–88.2)
Sex (male/female)	18/9	7/9	4/4	1/6	4/6
Education (years)	16.2 ± 2.4 (12–20)	16 ± 2.1 (14–20)	15 ± 2.6 (12–20)	14.7 ± 2.8 (11–18)	15.3 ± 2.9 (12–20)
Subtype	G: 10, L: 10, S: 0, M: 5, U: 2	G: 13, L: 1, S: 0, M: 1, U: 1	G: 4, L: 1, S: 1, M: 2, U: 0	G: 3, L: 0, S: 0, M: 3, U: 1	G: 0, L: 1, S: 8, M: 0, U: 1
n	27	16 (CBD: 9, PSP: 7)	8	7	10

Values are presented as mean ± SD (range). CBD/PSP = corticobasal- and progressive supranuclear-type neuropathology; CDR = clinical dementia rating scale; G = agrammatic PPA variant; L = logopenic PPA variant; M = mixed PPA variant; n = number of cases per neuropathology group; Pick’s = Pick’s disease neuropathology; S = semantic PPA variant; U = unclassifiable PPA cases; WAB AQ = aphasia quotient of the Western aphasia battery, revised.

Table 3

Language scores

	ADNC (N = 27)			CBD/PSP (N = 16)			Pick’s (N = 8)			TDP-A (N = 7)			TDP-C (N = 10)
	Mean ± SD	Range	n	Mean ± SD	Range	n	Mean ± SD	Range	n	Mean ± SD	Range	n	Mean ± SD	Range	n
Grammar	59.7 ± 25.5	10–96.7	22	52.6 ± 26.2	0–90	14	57.9 ± 24.4	20–93.3	8	42.8 ± 28.9	13.3–86.7	6	88.5 ± 18.1	50–100	9
Word comprehension	92.1 ± 8.7	66.7–100	23	89.8 ± 8.9	66.7–100	16	75.7 ± 24.1	36.1–100	8	73.8 ± 25.4	44.4–97.2	7	43.3 ± 23.4	22.2–97.2	10
Naming	61.7 ± 26.9	0–98.3	26	84.9 ± 15.1	45–98.3	16	55.6 ± 40.7	0–98.3	8	46 ± 37.3	10–95	7	16.8 ± 19.2	5–61.7	10
Repetition	62.4 ± 19.7	25.8–93.9	26	71.7 ± 19.8	34.8–97	16	80.5 ± 11.6	60.6–97	8	56.6 ± 37	0–97	6	82.7 ± 13.4	60.6–100	10
Object recognition	95.6 ± 3.1	90.4–100	24	94.2 ± 4	84.6–98.1	16	83.5 ± 19.9	51.9–100	7	92.3 ± 7.3	82.7–100	7	76.7 ± 13.6	53.8–98.1	10
Sentence comprehension	89.2 ± 10.8	30–100	21	77.1 ± 23.8	20–100	16	76.2 ± 18.4	53.3–100	7	66.7 ± 24.9	40–100	6	96.3 ± 6.8	80–100	9
Fluency	61.7 ± 20.3	20.5–114	22	39.4 ± 11.2	20.8–60.8	14	42.1 ± 35.3	9.9–100.3	5	41.1 ± 21.2	6.8–59.3	6	70.3 ± 33.8	15.6–103.1	5

Language and object recognition scores in the cohort of 68. The specific tests used to derive the performance scores are described in the methods section. Numbers refer to percentage of total possible scores in a given test. CBD/PSP = corticobasal- and progressive supranuclear-type neuropathology; CDR = clinical dementia rating scale; G = agrammatic PPA variant; L = logopenic PPA variant; M = mixed PPA variant; N = number of cases in each neuropathology group; n = number of participants that contributed data to that measure; Pick’s = Pick’s disease neuropathology; S = semantic PPA variant; U = unclassifiable PPA cases; WAB AQ = aphasia quotient of the Western aphasia battery, revised.

Longitudinal investigation participants and association of variants with neuropathology

All 68 participants were right-handed and therefore had a >90% likelihood of left hemisphere language dominance. Mean age of onset, survival length and gender distribution paralleled that of the larger cohort (Tables 1 and 2). Mean symptom duration at the time of the initial visit varied from around 4 years in the ADNC, Pick’s disease neuropathology and TDP(C) groups to 2–3 years in the others (range: 1–10 years; Table 2). Mean aphasia severity at initial visit was in the aphasia quotient range 83.8–75.7 in all groups except TDP(A), which exhibited the lowest aphasia quotient (65.9). Selectivity of language impairment was reflected in the preservation of daily living activities as shown by the CDR scores that ranged almost entirely from 0 to 0.5 (Table 2). As previously reported,²³ the logopenic variant was most prevalent in the ADNC group, the agrammatic variant most prevalent in the three- and four-repeat tauopathies of Pick’s disease and CBD/PSP and the semantic variant most prevalent in TDP(C) (Table 2).

Primary progressive aphasia-Alzheimer’s disease neuropathological changes

The ADNC group had leftward asymmetric atrophy at initial encounter and 2 years later (Fig. 2A–C). All major language network components were involved, including the inferior frontal gyrus (Broca’s area), the posterior dorsolateral frontal cortex (DFC), the confluence of the inferior parietal lobule with the posterior superior and middle temporal gyri at the temporoparietal junction (Wernicke’s area), the fusiform gyrus, the middle sectors of the temporal lobe (MTL) and the anterior temporal lobe (ATL), with the exception of the temporal pole. There was additional but much lesser right hemisphere atrophy within the temporoparietal junction and MTL. Two years later, atrophy had expanded, concentrically, within both hemispheres but without loss of leftward asymmetry. The aphasia quotient dropped from 80 at initial visit to 67.3 at the second (Table 2). At initial examination, severe impairments (performance <40%) were found in four domains: grammar, fluency, naming and repetition (Fig. 2C). Even in these domains, performance ranged broadly from intact to severely impaired (Table 3). Some participants had normal repetition and would therefore not fit the 2011 criteria for logopenic PPA, whereas others had impaired fluency and grammar, characteristic of agrammatic PPA (Fig. 2C). Neither age of onset nor aphasia severity measured by the mean aphasia quotient at initial encounter differentiated PPA-ADNC with agrammatic (aphasia quotient = 82 ± 7.5) versus logopenic PPA (aphasia quotient = 87.8 ± 8.5) variants. Conceivably, co-morbidities accompanying ADNC could account for the clinical heterogeneity, although this could not be ascertained in this cohort. None of the 27 cases in this group had word comprehension scores under 60% at initial testing.

Primary progressive aphasia-corticobasal degeneration/progressive supranuclear palsy

The CBD/PSP group displayed complete asymmetry of neurodegeneration at both initial and return visits (Fig. 2D–F). The cortical atrophy at the initial visit involved only small patches of the left DFC. Two years later, the atrophy had spread further into the DFC, inferior frontal gyrus and entire superior temporal gyrus but spared the temporal pole. The aphasia quotient dropped from 85.6 to 54.5. Moderate to severe impairment was most frequent in fluency, grammar, repetition and the comprehension of grammatically complex sentences, functionalities associated with the dorsal route of language processing. Word comprehension and object knowledge were largely preserved. In keeping with this pattern of impairments, 80% of the cases had an agrammatic form of PPA. There was no significant difference in mean age of symptom onset for the CBD versus PSP groups (63.9 years versus 67.5 years).

Primary progressive aphasia-Pick’s disease neuropathology

Pronounced leftward asymmetry of atrophy was detected at initial and return visits. Significant cortical thinning in the left hemisphere at the initial visit extended into DFC, inferior frontal gyrus, orbitofrontal cortex, ATL, MTL, anterior fusiform gyrus and parahippocampal gyrus, including the temporal pole (Fig. 3A–C). The second visit displayed slight concentric extension of atrophy into adjacent regions and emergence of restricted atrophy in some homologous parts of the right hemisphere, especially in the frontal lobe. The aphasia quotient dropped from 75.7 to 57.8. The involvement of both the dorsal and ventral components of the language network was reflected in the heterogeneity of clinical presentation, which included moderate to severe impairments in all domains except repetition. Similar to the four-repeat tauopathy in PPA-CBD/PSP, the three-repeat tauopathy of PPA-Pick’s disease neuropathology was also associated mostly with agrammatic PPA. However, one participant in this group had semantic PPA with severe word comprehension impairment without significant impairment of grammar or fluency. This was the only neuropathology group where both semantic and agrammatic variants were seen.

Primary progressive aphasia-TDP(A)

The TDP(A) group showed one of the most extensive distributions of left hemisphere neurodegeneration (Fig. 3D and E). There was no detectable right hemisphere atrophy. Grammar, naming, repetition and fluency displayed severe impairments in individual participants even at the initial visit. Agrammatic and mixed PPA variants were the most common. There were too few cases with repeat scans to yield meaningful progression data.

Primary progressive aphasia-TDP(C)

This group displayed the most uniform and distinctive neurodegeneration patterns and clinical manifestations (Fig. 4A–C). The signature atrophy site was located within the ATL and MTL, including the pole, with extension into adjacent orbitofrontal cortex, insula, fusiform gyrus and parahippocampal gyrus. Asymmetry was pronounced but there was a small patch of right hemisphere atrophy at the tip of ATL. Progression over 2 years led to modest extension of atrophy to adjacent areas with no evidence of spread to dorsal components of the language network. The aphasia quotient dropped from 80.4 to 65.6. The uniformity of the atrophy was reflected in the uniformity of the clinical picture where eight of the 10 cases had semantic PPA and where, with the exception of one case, the only severe abnormalities were found in word comprehension and object naming.

Diagnostic predictions by variant and test

Given the greatest concentration of logopenic PPA in ADNC, non-fluent-agrammatic PPA in CBD/PSP and semantic PPA in TDP(C), we calculated the ability of variants to predict pathology. Logopenic PPA predicted ADNC with 37% sensitivity and 93% specificity; non-fluent-agrammatic PPA predicted CBD/PSP with 81% sensitivity and 67% specificity; semantic PPA predicted TDP-C with 80% sensitivity and 98% specificity. Moderate-to-severe impairment (i.e. score under 60%) in word comprehension predicted underlying TDP-C with sensitivity of 80% and specificity of 91%. No other language test had similar diagnostic sensitivity and specificity for any neuropathology (see Supplementary Tables).

Individual cases

Individual cases illustrated rare entities associated with PPA and details of clinicopathological correlations. Figure 4D, from one of the two cases with FTLD-TDP(B), shows neurodegeneration limited to the left ATL. At initial imaging, the only language abnormality was in naming. There was also speech apraxia reflected in the mildly low words per minutes. Fig. 4E shows the atrophy pattern in the globular glial tauopathy case. Word comprehension was marginally low at 83% but naming was severely impaired (22%) and led to a diagnosis of semantic PPA. In contrast to the case in Fig. 4D, left hemisphere atrophy also encompassed MTL, inferior frontal gyrus and DFC. The latter two atrophy sites explain the mildly low words per minute (71%). In comparison to Fig. 4D, the presence of atrophy not only in ATL but also in MTL is likely to underlie the presence of word comprehension in Fig. 4E. Figure 4F and G, taken 4 years apart in a case of PPA-TDP(C), illustrate an analogous clinico-anatomical progression. Atrophy was initially similar to that of Fig. 4D, and the only substantial language impairment was in naming, at a level of severity that placed the participant at the boundary between logopenic PPA and semantic PPA, despite intact word comprehension. Figure 4G shows the extension of atrophy 4 years later into the MTL, fusiform gyrus, parahippocampal gyrus and insula. As in Fig. 4E, the posterior extension of temporal neurodegeneration into mid-temporal cortex (but not into Wernicke’s area) was associated with the onset of word comprehension impairment. Figure 5A–D illustrate left hemisphere atrophy patterns in four different neuropathologies underlying non-fluent-agrammatic PPA. The common denominator is the atrophy in inferior frontal gyrus and DFC. Of the four, only Fig. 5D had abnormal object naming, presumably reflecting the lateral temporal atrophy that was absent in the others. Figure 5E shows a case with semantic PPA and Pick’s disease. In contrast to Fig. 5C, also with Pick’s disease neuropathology, the case in Fig. 5E had more extensive lateral temporal atrophy in the ATL and MTL, with extension into the temporal pole, a pattern that explains the impairment of word comprehension. Figure 5E also shows that verbal semantics can be severely undermined by strictly unilateral temporal neurodegeneration.

Discussion

Although PPA is a relatively rare syndrome, it has had a major impact on highlighting the heterogeneity of clinicopathologic interactions. Starting from its initial characterization in 1982, PPA has shown that the same syndrome (e.g. progressive aphasia) can be caused by multiple neuropathological entities, that the same neuropathological entity (e.g. ADNC) can cause multiple syndromes (amnestic, aphasic), that the relationship between disease and symptom is probabilistic and that the clinical syndrome is determined by the network-level neuroanatomy of the neurodegeneration rather than the nature of the underlying cellular pathology. The current report reinforces these principles and adds new information based on a uniformly tested and imaged cohort.

Relative frequencies, genetics, age of onset, survival and clinical predictors of neuropathology

The 42% incidence of ADNC, 34% FTLD-tau (CBD/PSP and Pick’s disease neuropathology), 21% FTLD-TDP and 3% of rare entities is in line with some but not all previous autopsy series in PPA.^15,17,23 Genetic mutations were seen only in GRN and were detected in 4 of the 12 TDP(A) autopsies in the larger cohort of 118, reinforcing the conclusion that such mutations constitute the most common genetic cause of PPA.^75–77 In GRN families, some members may have PPA and others behavioural variant FTD.^78,79 Rarely, all affected members will have PPA.⁸⁰ Even then, the type of aphasia may differ between siblings and there is considerable heterogeneity of PPA subtypes associated with GRN mutations.^40,81 The literature also contains rare associations of PPA with mutations in the presenilin (PSEN1), MAPT, TARDBP and C9orf72 genes.^82–85 The cellular neuropathology is FTLD-TDP(A) in GRN mutations and mostly FTLD-TDP(B) in C9orf72 mutations. The most common clinical variants associated with dominantly-inherited diseases are non-fluent-agrammatic and logopenic PPA, but rare cases of semantic PPA have been reported.^{19,38,81,85,86}

Regardless of the underlying disease process, PPA is a ‘presenile’ dementia with mean age of onset at or under 65 years. Age of onset is lowest in PPA-TDP(C) and highest in PPA-CBD/PSP. PPA-TDP(A) has the shortest survival period, whereas FTLD-TDP(C) has the longest. The Kaplan-Meier curves show that the probability of being alive after 7 years of disease is less than 50% in TDP(A), almost 100% in TDP(C) and around 75% in ADNC, Pick’s disease neuropathology and CBD/PSP. The slow disease progression in TDP(C) is consistent with results obtained in other cohorts.^19,87,88 In cellular models, TDP(A) aggregates exert a much stronger toxicity than TDP(C) aggregates, a feature that may account for the more malignant disease course.⁸⁹ The distinction of TDP(A) from TDP(C) is further underlined by their differential associations with ribonuclear proteins.⁹⁰

The usefulness of clinical subtyping for predicting the underlying neuropathology is modest at best. The one exception is semantic PPA, which predicts TDP(C) with a sensitivity of 80% and specificity of 98%. However, these numbers would need to be revised downward if rare associations of semantic PPA with globular glial tauopathy (Fig. 4E) and ADNC⁹¹ are taken into account. Logopenic PPA had a specificity of 93% but sensitivity of 37% in detecting ADNC, a relationship that is unlikely to be of much clinical value, considering the much higher specificity and sensitivity offered by PET, CSF and plasma biomarkers. As Fig. 2C shows, word comprehension scores of <60% at initial visit can be used as a ‘negative biomarker’ that makes ADNC very unlikely, whereas the same score appears to have sensitivity and specificity at or above 80% for predicting TDP(C), but with the caveats noted above concerning rare causes of semantic PPA. We did not quantitate phonemic paraphasias or apraxia of speech, factors that might have predictive value for ADNC and CBD/PSP, respectively. To be useful, any clinical ‘surrogate biomarker’ for underlying neuropathology would need to have high sensitivity and specificity at the individual rather than group level.

Aphasic variant of Alzheimer’s disease neuropathological changes

Typical ADNC leads to an amnestic dementia; advancing age and presence of apolipoprotein E4 are its most important risk factors; the neurofibrillary tangles follow the Braak & Braak stages of hippocampofugal progression; limbic TDP-43 deposits are frequent co-morbidities; and neurodegeneration tends to display symmetrical distributions. In contrast, neither age nor ApoE4 is a risk factor for the aphasic variant of ADNC (PPA-ADNC)^92–94; TDP-43 co-morbidity is much less frequent^95,96; the Braak and Braak staging can be violated and replaced by higher neocortical/limbic ratios of neurofibrillary tangles and hippocampal sparing^{42,46,47,97,98}; functional connectivity perturbations can spread along the language rather than memory networks^99,100; and the neurodegeneration is always asymmetrical, being greater within the language-dominant hemisphere.^{23,36,101–103} In further contrast to typical dementias of ADNC where there is a predominance of females, we also found a distinct preponderance of males in the PPA-ADNC group, perhaps reflecting the association of PPA with familial dyslexia^49,51 and the higher susceptibility of males to dyslexia.¹⁰⁴ The ADNC that causes PPA is therefore so divergent from the typical pattern that it might qualify as a true biological variant.

Among the major neuropathologic correlates of PPA, the PPA-ADNC group was associated with the most widespread right hemisphere atrophy at initial imaging. Nonetheless, leftward asymmetry was maintained throughout the course of the disease.^36,101 In individual cases of PPA-ADNC, detectable neurodegeneration in the first MRI could be confined entirely to the left hemisphere. The language domains where severe impairments were detected (grammar, naming, repetition and fluency) are consistent with the presence of peak atrophy sites within the temporoparietal junction, MTL, inferior frontal gyrus and DFC.^105–107 Word comprehension was almost always preserved, probably reflecting the lack of prominent neurodegeneration within the temporal pole. A remarkable feature of PPA-ADNC is the sparing of memory even when the hippocampus and entorhinal cortex have dense accumulations of neurofibrillary tangles bilaterally.⁹⁵ Several factors may contribute to this resilience including less TDP-43 comorbidity than in typical amnestic ADNC, lesser frequency of ApoE4, and apparent absence of prominent medial temporal atrophy (parahippocampal gyrus, fusiform gyrus) as shown in Fig. 2A.⁹⁵ PPA-ADNC was most frequently associated with the logopenic and agrammatic forms of PPA, each accounting for 37% of the cases. Only 77% of logopenic PPA cases in the cohort of 68 had ADNC. This finding is consistent with some reports¹⁵ but not others^9,17 where the reported correlation of ADNC to logopenic PPA is much higher.

Three- and four-repeat tauopathies in primary progressive aphasia: PPA-CBD/PSP, PPA-Pick’s disease neuropathology

The PPA-CBD/PSP group had the most limited cortical atrophy at initial assessment, mostly confined to posterior DFC. Language tests showed severe impairments in grammar, fluency and non-canonical sentence comprehension, whereas verbal and non-verbal semantics were preserved. This pattern is in keeping with the predominantly agrammatic clinical presentation of PPA in the CBD/PSP group.^37,105 No detectable atrophy was seen in the right hemisphere. Digital histopathology had demonstrated asymmetric neuronal degeneration consistent with this distribution of atrophy in PPA-CBD/PSP.³⁷ White matter degeneration is likely to play a more substantial role in CBD/PSP than in the other neuropathologic entities associated with PPA and may account for the emergence of symptoms in the absence of substantial cortical atrophy.^108,109 Some participants had difficulties with pronunciation of multisyllabic words and displayed symptoms of speech apraxia. However, and as required for the PPA diagnosis, these components of motor speech impairment were distinctly less prominent than the language impairment. At initial testing, none of the participants had ophthalmoplegia, neck rigidity, dystonia, limb apraxia or cortical somatosensory signs of classic CBD or PSP syndromes.^110–112 Such atypical manifestations of CBD have been categorized as ‘cognitive predominant’ forms of the disease.¹¹³ The seven participants with PSP in the current cohort and additional case reports in the literature^114,115 show that ‘cognitive predominant’ forms also occur in PSP.

Compared with the astrocytic plaques and tufted astrocytes of four-repeat tauopathies (CBD and PSP), the three repeat tauopathy of Pick’s disease is characterized by round cytoplasmic inclusions. Electron cryo-microscopy has shown that three- and four-repeat tauopathies have different filament structures.¹¹⁶ The cortical atrophy pattern in PPA-Pick’s disease neuropathology was much more extensive than in PPA-CBD/PSP. Leftward asymmetry, while still very pronounced, was not as profound as in PPA-CBD/PSP. Peak atrophy encompassed frontotemporal swaths of cortex that included DFC and inferior frontal gyrus in the frontal lobe as well as insula, orbitofrontal cortex, MTL and ATL, including the temporal pole. In contrast to PPA-ADNC, the temporoparietal junction initially had no major atrophy, a feature that is consistent with the relative sparing of repetition. This was the only group where the clinical picture had the greatest variations, ranging from severe agrammatism to severe semantic PPA. The differences were in part attributable to individual differences in the distribution of atrophy. Comparisons of Fig. 5C and E illustrate two cases of PPA-Pick’s disease neuropathology, one with non-fluent-agrammatic PPA and the other with semantic PPA. A comparison of the two atrophy maps shows that word comprehension and severe naming deficits arose in the participant with the more extensive ATL and MTL atrophy that included the temporal pole. However, while both participants had inferior frontal gyrus as well as DFC atrophy, areas critical for grammar and fluency, the participant in Fig. 5E had preserved fluency and only mildly impaired grammar. It appears therefore, that the determinant of the clinical presentation may reflect an interaction between the neuroanatomy of degeneration and peculiarities of brain organization at the individual level. As in CBD/PSP, however, the most common clinical presentation of Pick’s disease neuropathology was that of non-fluent-agrammatic PPA. In rare instances PPA can also be associated with other tauopathies such as argyrophilic grain disease and globular glial tauopathy as shown in Fig. 4F.^28,29

TDP-43 proteinopathy: PPA-TDP(A), -TDP(B) and -TDP(C)

Survival was the shortest in PPA-TDP(A). Moderate to severe impairment was distributed across all language parameters, and there was no single characteristic clinical presentation. Of the two participants with GRN mutations and TDP(A) neuropathology in this group, one had non-fluent-agrammatic PPA and the other agrammatism combined with selective auditory word comprehension impairment.¹¹⁷ Another study, based on a much larger group of GRN mutations, found that logopenic PPA was the most frequent clinical correlate.⁷⁷ The atrophy in PPA-TDP(A) was one of the most extensive in the left hemisphere, while the right hemisphere was entirely spared, at least according to MRI imaging. Quantitative microscopy in sporadic as well as GRN mutation cases had shown that TDP-43 inclusions, neuronal loss, neuronal shrinkage and activated microglia were more prominent within the language-dominant hemisphere and mirrored the pattern of atrophy.^43–45,118 Further clinico-anatomical concordance was demonstrated by showing that the TDP-43 inclusions were more numerous in DFC in a participant with agrammatism, whereas they were more numerous in temporoparietal junction in a participant with anomia.⁴⁴

Two cases, excluded from the quantitative analyses, had TDP(B) as the primary neuropathologic diagnosis. They both presented with features of logopenic PPA but with mildly decreased fluency and variable components of speech impairment (Fig. 4D). No other signs of motor neuron disease characteristic of TDP(B) were noted but the autopsy showed sparse to moderate motor neuron disease-type changes in upper and lower motor neurons.²² Age of onset was in line with the other TDP types (59 years, 61 years) and survival from onset to death was short (5 years, 7 years). The neurodegeneration was sharply confined to the anterior tip of the left temporal lobe and the patients had an isolated moderately severe anomia without word comprehension impairment. These 2 TDP(B) cases have survival patterns similar to PPA-TDP(A) and anatomy of neurodegeneration similar to TDP(C).

The PPA-TDP(C) group had the most distinctive clinicopathologic fingerprints. Peak atrophy was most extensive within the left temporal lobe and extended into the adjacent insula and orbitofrontal cortex. Severe language abnormalities were seen only in word comprehension and object naming. Eight of the 10 participants had the semantic PPA variant. One had severe word comprehension impairment but was unclassifiable as semantic PPA because of abnormal grammar scores. Another, shown in Fig. 4F and G, had distinct anomia but no impairment of word comprehension at a time when the atrophy was confined to the anterior tip of ATL including the pole. As the atrophy spread posteriorly to MTL during the subsequent 4 years, word comprehension deficits emerged and became consistent with the clinical picture of semantic PPA. In the group of PPA-TDP(C), non-verbal object recognition showed mild to moderate impairment in only 3 of the participants, further supporting the distinction of semantic PPA from semantic dementia, the latter commonly reflecting bilateral ATL neurodegeneration leading to disruption of verbal as well as non-verbal semantics.^23,119Figure 4 strengthens the conclusion that left temporal pole degeneration is pivotal but not sufficient to undermine word comprehension and that caudal extension into MTL and anterior components of the parahippocampal gyrus and fusiform gyrus may be necessary. As shown in Figs 4E and 5E, globular glial tauopathy and Pick’s disease neuropathology can also be associated with the semantic PPA syndrome.^11,28,38 In our group of cases, however, these latter two entities had atrophy patterns that go beyond the temporal lobes and into the frontal lobes. Atrophy confined to the ATL in a PPA patient with severe anomia or word comprehension impairment can therefore predict TDP type C atrophy with considerable certainty.

The classic language network of Wernicke-Lichtheim-Geschwind makes no mention of the ATL¹²⁰ probably because this area is not vulnerable to focal cerebrovascular lesions. The identification of prominent semantic impairments in FTLD-TDP(C) patients with selective ATL neurodegeneration led to major revisions of the classic language network so that the left ATL rather than Wernicke’s area was found to be critical for word comprehension.^{16,21,38,88,105,121–123} FTLD-TDP type C can also cause predominantly right sided or bilateral ATL neurodegeneration, in which case it becomes associated with behavioral variant FTD, prosopagnosia or a combination of impaired word comprehension with associative agnosia that collectively define semantic dementia.^{23,55,119,124–128}

No mechanistic explanation exists for the remarkable affinity of TDP-C for the ATL. As shown in this report and in the literature quoted above, TDP(A) and TDP(C) have substantially different rates of progression, anatomy of neurodegeneration, morphology of precipitates, genetic linkages, cellular toxicity, ribonuclear protein associations, and clinical manifestations. Furthermore, neurodegeneration in TDP(A) is associated with increased densities of abnormal TDP precipitates whereas the inverse relationship in seen in TDP(C).¹²⁹ Both entities entail translocation and truncation of the same protein, which is a normal constituent of the nucleus and involved in RNA processing. If the cause of neurodegeneration were the loss of normal TDP-43 functionality, types A and C should not be that different in clinical and neuropathologic manifestation. The striking differences suggest that TDP-43 inclusions in types A and C are likely to reflect downstream manifestations of fundamentally different upstream causes that happen to converge on the same proteomic network.

Conclusions

Each neuropathologic entity underlying PPA has preferred anatomical targets and corresponding clinical patterns. Less favored associations also arise so that the linkage of clinical features to neuropathology becomes probabilistic rather than absolute. The highest consistencies occur in CBD/PSP, which targets the dorsal components of the language network with resultant impairment of grammar and fluency, and TDP(C), which targets the ventral components of the network with resultant impairment of word comprehension and naming. The left-sided predominance of cortical neurodegeneration is the fundamental and defining feature of all PPA. This asymmetry cannot be attributed to a chance occurrence at onset because it is frequently maintained until death and the aphasia remains the principal deficit, without equivalent impairments of memory or comportment, for up to 10–15 years.^11,130 Remarkably, each neuropathologic entity that causes PPA can cause other syndromes where the degeneration is bilateral or predominantly right sided. The biological underpinnings of the asymmetric neurodegeneration underlying PPA remain unknown. In some cases of PPA, it may reflect genetic or developmental vulnerabilities that make the left hemisphere a locus of least resistance for independently arising neurodegenerations.^49,51 For example, increased familial incidence of learning disability, including dyslexia, has been reported in persons with PPA; and in at least one kindred with a PPA proband, six of nine unaffected siblings had dyslexia and abnormal functional connectivity of the language network.⁴⁹ In other cases, constitutive or individual molecular differences may underlie the differential vulnerability of the left hemisphere.¹³¹ In Parkinson’s disease with unilateral symptoms, for example, hemispheric differences in neuronal DNA methylation, transcriptomics, and proteomics have been shown to correspond to the lateralization of symptoms.¹³² Similar findings in PPA, assuming they can be linked to the cause rather than result of the neurodegeneration, would provide pivotal insights into the biology of selective vulnerability and the evolution of language in the human brain.

Acknowledgements

We are grateful to our participants and their families for their remarkable dedication to research on PPA.

Funding

Funding was provided by R01DC008552 from the National Institute on Deafness and Communication Disorders, R01NS075075 and R01NS085770 by the National Institute on Neurological Disorders and Stroke, R01AG056258, R01AG62566, K08AG065463 and P30AG013854 from the National Institute on Aging, the Davee Foundation and the Jeanine Jones Fund.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.

 
• ADNC

  Alzheimer’s disease neuropathological changes

 
• ATL

  anterior temporal lobe

 
• CBD

  corticobasal degeneration

 
• DFC

  dorsolateral frontal cortex

 
• FTLD

  frontotemporal lobar degeneration

 
• MTL

  middle sectors of the temporal lobe

 
• PPA

  primary progressive aphasia

 
• PPA-ADNC

  aphasic variant of Alzheimer’s disease neuropathological changes

 
• TDP-43

  transactive response DNA-binding protein-43