Microglia in frontotemporal lobar degeneration with progranulin or C9ORF72 mutations

Abstract Objective To identify clinicopathological differences between frontotemporal lobar degeneration (FTLD) due to mutations in progranulin (FTLD‐GRN) and chromosome 9 open reading frame 72 (FTLD‐C9ORF72). Methods We performed quantitative neuropathologic comparison of 17 FTLD‐C9ORF72 and 15 FTLD‐GRN with a focus on microglia. For clinical comparisons, only cases with high quality medical documentation and concurring diagnoses by at least two neurologists were included (14 FTLD‐GRN and 13 FTLD‐C9ORF72). Neuropathological analyses were limited to TDP‐43 Type A to assure consistent assessment between the groups, acknowledging that Type A is a minority of C9ORF72 patients. Furthermore, only cases with sufficient tissue from all regions were studied (11 FTLD‐GRN and 11 FTLD‐C9ORF72). FTLD cases were also compared to age– and sex–matched normal controls. Immunohistochemistry was performed for pTDP‐43, IBA‐1, CD68, and GFAP. Morphological characterization of microglia was performed in sections of cortex blinded to clinical and genetic information. Results FTLD‐GRN patients had frequent asymmetric clinical features, including aphasia and apraxia, as well as more asymmetric cortical atrophy. Neuropathologically, FTLD‐C9ORF72 had greater hippocampal tau pathology and more TDP‐43 neuronal cytoplasmic inclusions. FTLD‐GRN had more neocortical microvacuolation, as well as more IBA‐1–positive ameboid microglia in superficial cortical layers and in subcortical white matter. FTLD‐GRN also had more microglia with nuclear condensation, possibly indicating apoptosis. Microglial morphology with CD68 immunohistochemistry in FTLD‐GRN and FTLD‐C9ORF72 differed from controls. Interpretation Our findings underscore differences in microglial response in FTLD‐C9ORF72 and FTLD‐GRN as shown by significant differences in ameboid microglia in gray and white matter. These results suggest the differential contribution of microglial dysfunction in FTLD‐GRN and FTLD‐C9ORF72 and suggest that clinical, neuroimaging and pathologic differences could in part be related to differences in microglia response.


Introduction
Frontotemporal lobar degeneration (FTLD) is clinically, neuropathologically and genetically heterogeneous. Among clinical presentations are changes in behavior, personality, and language. Some patients also have motor neuron disease. 1 The most common neuropathologic findings in FTLD are tauopathies or TDP-43 proteinopathies. 2,3 The most common genetic causes of FTLD-TDP are mutations in progranulin (GRN) 4,5 and chromosome 9 open reading frame 72 (C9ORF72). 6,7 Mutations in GRN account for about one fourth 8,9 and C9ORF72 for about one half 10 of familial FTLD-TDP. Clinically, almost all patients with mutations in GRN have frontotemporal clinical syndromes, with only rare reports of motor neuron disease. 11 In contrast, patients with C9ORF72 hexanucleotide repeat expansions often have motor neuron disease with or without clinical features of frontotemporal dementia. 12 The neuropathologic features of FTLD-TDP are focal cortical atrophy of frontal and temporal lobes with variable involvement of parietal lobe, as well as cytoplasmic inclusions in neurons and glia that are immunoreactive for TDP-43. 13 The relative density and distribution of neuronal and glial inclusions, as well as dystrophic neurites permits subtyping of FTLD-TDP into Types A, B, and C, as well as less common subtypes. 14 FTLD-GRN is almost always Type A, while FTLD-C9ORF72 can be Type B and less often Type A. 15 The pathogenesis of neurodegeneration in FTLD-TDP remains poorly understood. Several pathomechanisms have been hypothesized for FTLD-C9ORF72, including toxic gain of function of RNA, protein aggregation, and impairment of nucleocytoplasmic transport. 16,17 Microglial dysfunction has recently been suggested to play a role, based upon studies of C9orf72 knock-out mice. 18 Mechanisms thought to play a role in FTLD-GRN are based on the fact that progranulin has neurotrophic properties, while the proteolytic products of progranulin, the granulins, may be proinflammatory modulators. 19 Given that FTLD-C9ORF72 and FTLD-GRN may have different pathogenic mechanisms, we aimed to study clinical and neuropathological differences that might be related to this fact. Several previous studies have addressed clinicopathological characteristics of FTLD-TDP; however, few have stringently controlled for subtype. In this study, we hypothesize that different clinical features between GRN and C9orf72 could be driven by a differential role of microglia in neuroinflammation. To address this hypothesis, we measured microglial phenotypes based on density and morphology in cases of FTLD-GRN and FTLD-C9ORF72 with Type A TDP-43 pathology.

Case materials
All cases were submitted for diagnostic studies and research to the brain bank for neurodegenerative disorders at Mayo Clinic in Jacksonville, Florida. We identified 17 cases of FTLD-C9ORF72 with Type A TDP-43 pathology and 15 cases of FTLD-GRN. All cases were neuropathologically classified by a single neuropathologist (DWD), and genotyping was performed on DNA extracted from frozen brain tissue. For clinicopathological studies, cases were included only if they had good quality medical documentation and there was diagnostic concurrence by a minimum of two neurologists. The final set of cases for clinical comparisons was 13 FTLD-C9ORF72 and 14 FTLD-GRN. For quantitative microglial morphological studies, tissue from all brain regions had to be available for assessment. The final set of cases for pathologic studies was 11 FTLD-C9ORF72 and 11 FTLD-GRN. A summary of cases included in clinical and pathologic analyses are listed in Table 1 and additional details are provided in Table S1.

Clinical assessment
Cognitive and psychiatric features were abstracted from medical records of neuropsychological and psychiatric evaluations. The clinical diagnosis of included cases fulfilled the criteria for behavioral variant frontotemporal dementia (bvFTD), 20 progressive nonfluent aphasia (PNFA), 21 or corticobasal syndrome (CBS). 22 Clinical asymmetry was considered present for cases with diagnosis of PNFA and CBS or by asymmetry in motor findings from neurologic examinations.

Genetic analyses
Frozen brain tissue was used for genotyping. All FTLD-C9ORF72 cases had hexanucleotide repeat expansions in C9ORF72 using a repeat-primed polymerase chain reaction method to detect expansions of the GGGGCC hexanucleotide, 6 and for most cases, the expansions were confirmed with Southern blotting. 23 FTLD cases with GRN mutations were confirmed using Sanger sequencing. 5,9 Microscopic pathology Samples from the middle frontal gyrus (mFCtx), superior temporal gyrus (sTCtx), inferior parietal cortex (iPCtx), primary motor cortex (MCtx), hippocampus, basal ganglia, and medulla were cut at 5 µm thickness, mounted on glass slides and stained with hematoxylin and eosin (H&E). Thioflavin-S fluorescent microscopy was performed to evaluate senile plaques and neurofibrillary tangles. A Braak neurofibrillary tangle stage and a Thal amyloid phase were derived from quantitative data, as described previously. 24 The neuropathologic assessment included immunohistochemistry for phospho-TDP-43 or an antibody to a neoepitope in the midregion of TDP-43. 25 The TDP-43 Type A was assigned to each case using features described in the harmonized classification. 14

Image analysis
Digital microscopy methods have been previously described. 28 Briefly, immunostained sections were scanned on an Aperio ScanScope XT slide scanner (Leica Biosystems Aperio, USA) producing high resolution digital images. Analyses were performed using Aperio ImageScope software in defined regions of interest (ROIs). A color deconvolution algorithm was used to count the number of pixels that had strong immunoreactivity. The output was percentage of strong positive pixels relative to the total area of the ROI. For digital analysis of the hippocampus, several ROIs were defined, including the entire hippocampus, the dentate fascia, CA4 sector, and the parahippocampal gyrus. For microvacuolation analyses, H&E sections of mFCtx, sTCtx, and parahippocampal gyrus were evaluated. The percentage of the area occupied by microvacuolation was assessed in cortical layer II as the number of weak positive pixels divided by the total number of positive pixels according to previously published methods. 29 Severity of microvacuolation was also assessed by semiquantitative scores (none, mild, moderate or severe), and these results were similar to those obtained by image analysis (Fig. S1).
For measurement of cortical thickness in mFCtx and sTCtx, ImageScope ruler tool was used by drawing a line perpendicular to the pial surface and the gray-white junction.

Quantitative analysis of microglial phenotypes
Digital image analysis was used to assess microglial density based on signal intensity in cell bodies and processes. To further investigate the role of microglia in FTLD-TDP, we performed qualitative morphologic characterization of different microglia phenotypes using the Aperio counting tool for manual labeling of IBA-1-positive or CD68-positive microglia in fixed-sized ROI. Digital image analysis and manual counting of IBA-stained sections gave comparable results (Fig. S2). Individual microglial subpopulations were assessed in mFCtx, a severely affected region, and compared to MCtx, a minimally affected region and in the subcortical white matter. Microglia were categorized as ramified, elongated, rodshaped or ameboid (Fig. 1). Dystrophic microglia had evidence of fragmentation or beading of cell processes according to the criteria of Streit and Braak. 30 Elongated cells had a polarized appearance with minimal branching, with rodshaped nuclei, and process lengths between 75-150 µm. The number of IBA-1-positive ameboid, ramified, elongated, and dystrophic microglia were counted in each ROI. Cells were divided into CD68-low, which lacked immunostaining of cell processes, but had perinuclear granules (Fig. 1) and CD68high, which had immunostaining of cell processes. Microglial counts in defined ROIs were made by two observers, one blinded (SFR) to all clinical, genetic and pathologic information. Five neurologically normal individuals, matched for age and sex, were used as controls for this analysis (Table S2).

Statistical analyses
SigmaPlot (Systat Software, San Jose, CA) was used for statistical analyses. Due to small sample sizes, non-parametric Kruskal-Wallis analysis of variance (ANOVA) on ranks was performed on quantitative measures and differences in median values were assessed. Post hoc pairwise comparisons were performed between groups using Mann-Whitney rank sum test. For categorical data (sex, presence or absence of extrapyramidal signs, and symmetry/asymmetry of neuroimaging findings), a Chi-squared test was used to compare group differences. Fisher's exact test was used for pairwise categorical data when the count was less than 5. Correlative analysis was performed using Spearman rank order correlation. A P-value of < 0.05 was considered statistically significant. Given that FTLD-GRN tended to have more women and FTLD-C9ORF72 to be older, analyses were also adjusted for age and sex in a multiple logistic regression models assessing clinical differences, such as memory impairment similar adjustments were made in multiple linear regression analyses for Braak stage.   There were no statistically significant differences between the two groups with respect to median disease duration and male-to-female ratio; but there were more women in FTLD-GRN (57%) than FTLD-C9ORF72 (27%). This observation fits with a recent meta-analyses showing that FTLD-GRN is more common in women than men. 31

Asymmetric clinical syndromes and neuroimaging findings
The asymmetrical clinical syndrome of CBS was noted in three of 14 patients (21%) with FTLD-GRN, but in none of the FTLD-C9ORF72 patients. Available MRI or CT images (14 for FTLD-GRN and 11 for FTLD-C9ORF72) showed asymmetrical cortical atrophy in only one patient with FTLD-C9ORF72, but in seven (50%) patients with FTLD-GRN (Table 2).

Language disorders
Progressive aphasia (four PNFA and two progressive aphasia not otherwise specified) was noted in six of 14 patients with FTLD-GRN, but in only three of 13 FTLD-C9ORF72 patients. One patient appeared to have logopenic aphasia, and one patient appeared to have semantic components. A receptive component could not be excluded confidently. The remainder clearly had progressive aphasia, but could not be further subtyped.

Amnestic clinical syndromes and extrapyramidal signs
Amnestic dementia was based on neuropsychological evaluations and considered for patients where memory complaints were the predominant feature. Alzhiemer dementia was in the clinical differential diagnosis of some of these patients. Memory problems overshadowed other clinical features of FTLD, such as behavioral and language problems. Amnestic dementia (AD) was prominent in two patients (14%) with FTLD-GRN, but in eight patients (62%) with FTLD-C9ORF72. AD was the clinical diagnosis of four of eight FTLD-C9ORF72 patients (Table 1). In FTLD-C9ORF72, prominent memory problems were noted in four patients with dementia with Lewy bodies (DLB) (31%), and these patients also had extrapyramidal signs and visual hallucinations. The frequency of extrapyramidal signs in the entire cohort was similar in FTLD-GRN and FTLD-C9ORF72 ( Table 2).

Effects of age and Alzheimer type pathology on clinical features
To determine potential confounding variables accounting for observed differences in clinical syndromes and neuroimaging findings, we addressed possible contributing pathologic features, such as brain weight, Thal amyloid phase, Braak neurofibrillary tangle (NFT) stage, and presence of hippocampal sclerosis (HpScl). There were no differences between FTLD-GRN and FTLD-C9ORF72 for age at death, disease duration, brain weight, frequency of HpScl or Thal amyloid phase. On the other hand, the median Braak NFT stage was greater (P < 0.001) in FTLD-C9ORF72 than in FTLD-GRN. To adjust for the effects of age at death and sex on differences in Braak NFT stage, a multiple linear regression model showed higher (P = 0.011) Braak NFT stage in FTLD-C9ORF72 compared with FTLD-GRN.

TDP-43 pathology
To assess differences in TDP-43 pathology histological sections from the mFCtx, sTCtx, iPCtx and hippocampus were assessed and scored semiquantitatively as either present or absent. The density of phospho-TDP-43 pathology was also assessed with image analysis. There were no significant differences in TDP-43 pathology in MCtx (not shown); however, densities of TDP-43 pathology tended to be greater in FTLD-GRN than FTLD-C9ORF72 in mFCtx, sTCtx and iPCtx, (Fig. 2). On the other hand, FTLD-C9ORF72 had greater TDP-43 density in the hippocampus than FTLD-GRN (P = 0.002). In subregions of the hippocampus, TDP-43 density was significantly greater in dentate gyrus (P < 0.001) and CA4 sector (P < 0.001) in FTLD-C9ORF72 compared with FTLD-GRN.

Cortical thickness
We assessed cortical thickness in mFCtx and sTCtx. FTLD-GRN had significant atrophy of the mFCtx (P = 0.036) ( Table 3). The cortical thickness in the sTCtx was not significantly different.

Neocortical microvacuolation
The degree of superficial microvacuolation was assessed using digital image analysis in the mFCtx and sTCtx. In most cases, at least some degree of microvacuolation was detected in both ROIs (Fig. 3). Quantitative analysis revealed that microvacuolation was more severe in FTLD-GRN than FTLD-C9ORF72, reaching statistical significance in mFCtx (P = 0.036).

Neuroinflammation
There were no significant differences in overall density of IBA-1, CD68 and GFAP-positive glial cells between FTLD-C9ORF72 and FTLD-GRN using digital image analysis in mFCTx, sTCTx, iPCTx, MCTx. The only exception was more CD68-positive microglia in the hippocampus in FTLD-C9ORF72 (data not shown).

Discussion
In this study, we compared FTLD-GRN and FTLD-C9ORF72 matched for TDP-43 subtype (Type A) to exclude for differences due to TDP-43 subtype. Although there are several clinical, neuropsychological, and radiological studies of FTLD-GRN and C9ORF72, to our knowledge, this is the first direct clinicopathological comparison of FTLD-GRN and FTLD-C9ORF72 matched for TDP-43 type. In our clinical comparison, we found that FTLD-C9ORF72 more frequently had amnestic dementia, including antemortem diagnoses of AD or DLB, as well as more symmetrical and milder cortical atrophy compared with FTLD-GRN. Although amnestic deficits in FTD can be affected by executive dysfunction, 32 our observations are in line with previous reports, including Mahoney and coworkers and Simon-Sanchez and coworkers, who reported about half of FTLD-C9ORF72 patients presented with memory impairment. 33,34 Our previous study showed FTLD-C9ORF72 had greater tau pathology than FTLD-GRN. A previous study showed less tau pathology in FTLD-GRN compared to FTLD-C9ORF72. 27 In this study, we similarly found decreased tau densities in FTLD-GRN in the mFCtx and hippocampus compared to FTLD-C9ORF72. In line with this, Papegaey and coworkers biochemically reported reduction of tau protein expression in frontal cortex of FTLD-GRN. 35 We also found significant greater TDP-43 in the hippocampus of FTLD-C9ORF72 compared with FTLD-GRN in addition to tau pathology. It is likely that hippocampal TDP-43 pathology and tau pathology both contribute to the amnestic phenotype in FTLD-C9ORF72.
FTLD-GRN patients more often had asymmetrical clinical presentations, such as PNFA and CBS. This is in line with previous reports, 36,37 including a study by Pickering-Brown and coworkers who reported that PNFA was more common in FTLD-GRN (36%). 38 FTLD-GRN also had more severe cortical atrophy, microvacuolation, and more ameboid microglia in cortical regions (but not in the hippocampus) compared with FTLD-C9ORF72. More severe and asymmetrical cortical atrophy, as well as greater TDP-43 pathology in, sTCtx and iPCtx may underlie a pathological substrate of aphasia.
In the brain, expression studies based upon RNA sequencing show that it is extremely high in microglia and sparse in neurons. 39 Loss of function mutations in GRN, leading to nonsense mediated decay of mRNA, 9 are thought to be associated with microglial dysfunction. 40,41 There are only a few quantitative neuropathologic studies on neuroinflammation in FTLD-GRN and FTLD-C9ORF72. Lant and coworkers performed a semiquantitative analysis of CD68-positive microglia in FTLD-GRN (Type A), FTLD-C9ORF72 (Types A and B) and FTLD-MAPT in frontal and temporal cortices and found significantly more CD68-positive microglia in FTLD-MAPT than in FTLD-GRN and FTLD-C9ORF72, but no difference between the two genetic forms of FTLD-TDP. 42 In the present study, however, we assessed critically affected areas such as cortical layer II and also subtyped microglia based on morphology. We did not find any significant differences in rod-shaped or dystrophic microglia between FTLD-GRN and FTLD-C9ORF72, but showed that FTLD had significantly more dystrophic microglia compared to control. Several studies highlight a role for rod-shaped and dystrophic microglia in neurodegenerative diseases. 30,43,44 The exact significance of these findings remain to be identified, but may be related to microglial dysfunction. On the other hand, FTLD-GRN had more microglia with condensed nuclei, which may suggest either increased vulnerability of microglia deficient in GRN or an increased turnover of macrophages.
Furthermore, FTLD-GRN had significantly more IBA-1-positive ameboid cells in layer II of mFCtx and deep white matter. This is consistent with a study by Woollacott and coworkers reporting high density of amoeboid microglia in the cortex and white matter of a single FTLD-GRN case. 45 These findings suggest a role for ameboid microglia in the pathogenesis of both white matter and gray matter damage in FTLD-GRN. FTLD-GRN also showed more significant neocortical microvacuolation. The underlying molecular mechanism for microvacuolation is unknown, but synaptic damage has long been presumed to play a role. 46 Lui and coworkers showed that progranulin knockout mice have synaptic loss through defective synaptic pruning by microglia. 40 They also found increased IBA-1positive microglia in the frontal cortex of FTLD-GRN compared with neurological controls. 40 In conclusion, we found distinct clinicopathological differences between in FTLD-GRN and FTLD-C9ORF72 with respect to type A TDP-43. Clinical asymmetric phenotypes, such as PNFA or CBS, as well as asymmetric brain atrophy, were more common in FTLD-GRN. Early memory deficits and symmetric brain atrophy were more common in FTLD-C9ORF72. Our neuropathological analyses highlight differential neocortical microvacuolation and phagocytic microglial phenotype between FTLD-GRN and FTLD-C9ORF72. Microglial dysfunction may be implicated in both mutations, but our data suggest different roles for neuroinflammation between FTLD-GRN and FTLD-C9ORF72. Additional experimental studies are warranted to better determine shared or distinct downstream      Table S2. Demographic and pathologic features of FTLD cases and normal controls used in microglial morphologic studies. Figure S1. Comparison of manual microglial counts to image analysis of IBA-1 density. Figure S2. Comparison of manual microvacuolation scores and vacuolation burden from image analysis.