Introduction

Dementia is a growing problem in the aging population and has become a major challenge to societies. Alzheimer’s disease neuropathological change (ADNC), characterized by amyloid plaques and neurofibrillary tangles (NFT), is the most common pathology in the autopsied brains of older adults with cognitive impairment. However, ADNC is frequently accompanied by multiple neuropathologic comorbidities that contribute to cognitive impairment including transactive response DNA-binding protein of 43 kDa (TDP-43) proteinopathy [67, 77]. TDP-43 is a DNA and RNA binding protein involved in several neurodegenerative diseases, including amyotrophic lateral sclerosis (ALS), frontotemporal lobar degeneration (FTLD), and Alzheimer’s disease (AD) [1, 71]. Under normal circumstances, the presence of TDP-43 is largely limited to the nucleus of neurons, but in its proteinopathic state TDP-43 is mislocalized, forming neuronal cytoplasmic inclusions (NCI) and it is excluded from the nucleus (nuclear clearance). Several studies have demonstrated that TDP-43 proteinopathy with a predominant distribution in the mesial temporal lobe is frequent in the aging population, with a prevalence ranging from 10 to 70% [3, 25, 68, 86]. In 2019, a consensus group suggested the term “Limbic-predominant Age-related TDP-43 Encephalopathy neuropathologic change” (LATE-NC), the term we will use henceforth to refer to this TDP-43 proteinopathy and associated neuropathologic changes in the aging population [69]. Increasing evidence indicates that LATE-NC is associated with cognitive decline, hippocampal atrophy, and hippocampal sclerosis [17, 31, 41,42,43, 45, 64, 92]. Many studies have demonstrated that LATE-NC is associated with cognitive decline independent of AD [9, 25, 29, 33, 36, 42, 45, 51, 62, 63, 65, 67, 68, 78, 79, 92]. However, we still need large-scale community-based studies with longitudinal cognitive assessments to ascertain the clinical significance of the TDP-43 proteinopathy, its relationship with ADNC and its contribution to declines in different cognitive domains either associated or independent from AD. In terms of the mechanism of TDP-43 proteinopathy, some studies have demonstrated that loss of TDP-43 splicing repression is an important mechanism in neurodegenerative diseases including ALS and FTLD [50, 83]. However, the age of onset of TDP-43 loss-of-function in the human brain and its relationship with nuclear clearance and TDP-43 NCI remain unknown. Here, we use clinical and neuropathologic data from participants in the Baltimore Longitudinal Study of Aging (BLSA), a community-based cohort study, to explore LATE-NC and TDP-43 splicing repression in the aging population. All participants were cognitively normal at enrollment and had periodic longitudinal cognitive assessments. We also assessed for TDP-43 splicing repression in a younger forensic autopsy cohort. To examine the cryptic exon inclusion in cases with LATE-NC, we used RNA in situ hybridization (BaseScope) followed by immunofluorescent demonstration of a cryptic exon-encoded neoepitope within HDGFL2 using a recently characterized monoclonal antibody [35]. Our results showed that high levels of ADNC significantly increase the odds of LATE-NC, which is associated with accelerated decline in multiple cognitive domains. Notably, there are associations of LATE-NC with verbal memory and visuospatial ability that are independent of ADNC. Importantly, our data revealed that loss of TDP-43 splicing repression first occurs in the aging population during the 6th decade of life and a decade preceding the appearance of TDP-43 + NCIs.

Materials and methods

Subjects

Participants in this study were 309 individuals from the BLSA autopsy cohort (age range 57–109, mean age 88.5 years). The BLSA is a prospective study of community-dwelling volunteer participants begun in 1958 and conducted by the National Institute on Aging. Participants had normal cognition at the time of BLSA enrollment. Neuropsychological assessments, neurological, laboratory and clinical evaluations were conducted during each visit. The BLSA visit schedule has changed over time. Participants were typically seen every two years before 2003. Since then, participants were seen every 4 years for age < 60 years, every 2 years for 60–79 years, and annually for ≥ 80 years until they are untestable or expire [90]. The BLSA study was approved by the local Institutional Review Board and the National Institute on Aging. Written informed consent was obtained at each visit from all participants. An additional 24 donated autopsy brains from individuals 21–68 years of age (mean age 43 years) were included for examination of TDP-43 nuclear clearance in a younger age group. Individuals in this younger autopsy cohort did not have history of cognitive decline, but no cognitive evaluation was available. These autopsies were performed at the Office of The Chief Medical Examiner (OCME) of the State of Maryland in Baltimore and accessioned by the University of Maryland Brain and Tissue Bank with appropriate consent from the next-of-kin. Research proceedings were approved by institutional review boards. All brains were examined in the Division of Neuropathology at Johns Hopkins Medicine under a protocol approved by the JHU IRB.

BLSA cohort cognitive assessments

Subjects enrolled in the BLSA cohort had comprehensive longitudinal cognitive assessments. The Combined Clinical Dementia Rating (CDR) scale [61] was administered at 1- to 2-year intervals to all autopsy study participants [90]. Clinical and neuropsychological data from participants were reviewed at consensus conferences if their Combined Clinical Dementia Rating scale score was ≥ 0.5 or if they had > 3 errors on the Blessed Information-Memory-Concentration Test [27]. After death, subjects were adjudicated as having normal cognition, mild cognitive impairment (MCI), or dementia at a consensus diagnostic meeting by study examiners who were unaware of the neuropathologic diagnoses. Diagnoses of dementia and AD, respectively, were based on criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders, third edition, revised [2], and the National Institute of Neurological and Communication Disorders and Stroke–Alzheimer’s Disease and Related Disorders [56]. MCI was based on the Petersen criteria [74].

For the analysis of longitudinal cognitive trajectories, we included all longitudinal cognitive assessments conducted within 20 years prior to participants' death, from individuals who were ≥ 60 years old at the time of data collection, and whose assessments were conducted in or after 1993. The final analytic sample of cognitive function included 243 participants with 1405 cognitive assessments. The cognitive assessments include Mini Mental State Exam (MMSE) [26], California Verbal Learning Test (CVLT) immediate recall across five learning trials to assess verbal memory [18], Trail Making Tests A and B to measure attention, psychomotor speed and executive function [76], Card Rotations test to measure visuospatial ability [22, 91], Letter and Category Fluency tests to measure phonemic and semantic fluency, respectively [7, 72]. A detailed description of the cognitive measures and procedures have been published previously [54].

Neuropathological assessments

This study included 309 autopsy brains from BLSA participants and 24 brains from the younger autopsy cohort. After autopsy, all brains from the BLSA cohort were fixed in neutral buffered formalin for 2 weeks. Following external examination, brains were cut in coronal slabs and diagnostic tissue blocks were dissected from the left cerebral hemisphere in most instances. For microscopic examination, tissues from the middle frontal gyrus, superior and middle temporal gyri, inferior parietal cortex, precuneus, occipital cortex, cingulate gyrus, basal ganglion, thalamus, amygdala, hippocampus, midbrain, pons, medulla, and cerebellum were obtained. Brains from the younger autopsy cohort were examined and cut fresh in coronal slabs after autopsy. Tissues from the middle frontal gyrus, motor cortex, superior and middle temporal gyri, inferior parietal cortex, occipital cortex, basal ganglion, amygdala, hippocampus, midbrain, pons, medulla, and cerebellum were obtained. Tissues were fixed in 10% neutral buffered formalin, dehydrated in graded alcohols, cleared in xylene, and embedded in paraffin. Tissue sections were cut at 10 µm and stained with H&E. For BLSA brains, Hirano-silver stain was performed on middle frontal gyrus, superior and middle temporal gyri, inferior parietal cortex, precuneus, occipital cortex, amygdala, and hippocampus. Immunohistochemistry (IHC) of β-amyloid was performed on thalamus, midbrain, and cerebellum. IHC of phosphorylated tau was performed on middle frontal gyrus, superior and middle temporal gyri, inferior parietal cortex, occipital cortex, thalamus, amygdala, hippocampus, midbrain, and pons. IHC of α–synuclein was performed on cingulate gyrus, amygdala, and midbrain. Additional IHC was performed in brains with suspected histological abnormalities. For young autopsy cohort brains, IHC of phosphorylated tau was performed on middle frontal gyrus, amygdala, entorhinal cortex, and pons. IHC of β-amyloid was performed on middle frontal gyrus and amygdala. Additional IHCs including phosphorylated tau and β-amyloid were performed in brains with positive phosphorylated tau and β-amyloid staining or suspected histological abnormalities. IHC of α–synuclein was performed in brains with suspected Lewy body.

IHC of TDP-43 was performed using both phosphorylation-independent and phosphorylation dependent TDP-43 antibodies. The phosphorylation-independent antibody recognizes all forms of TDP-43 protein. The phosphorylation dependent TDP-43 antibody recognizes TDP-43 phosphorylation at Serine 409/410 and was used to evaluate TDP-43+ NCIs as well as other staining patterns. Phosphorylation independent TDP-43 antibody was used to assess the nuclear localization of TDP-43 and also TDP-43+ NCIs. In all brains, tissue sections of the amygdala and hippocampus were immunostained with both types of anti-TDP-43 antibodies. Sections of middle frontal gyrus were stained in cases with any positive staining in the amygdala, hippocampus, entorhinal cortex, or adjacent structures. Sections of precuneus and inferior parietal lobule were stained in randomly selected cases encompassing all LATE-NC stages (n = 22), including 7 cases of stage 0, 2 cases of stage 1, 10 cases of stage 2, and 3 cases of stage 3.

To examine the relationship of LATE-NC and tau pathology, we examined tau pathology in the dentate gyrus. Tau-positive granule cells were counted manually by a neuropathologist blinded to the pathologic diagnosis and TDP-43 status. The counts of tau-positive granule cells were adjusted by the lengths of the dentate gyrus granule cell layers.

Immunohistochemical (IHC) method

IHC stains were conducted on 10-µm or 40-µm thick sections from formalin fixed paraffin embedded tissues. Each section was deparaffinized with xylene and rehydrated through graded alcohols and water followed by heat-based antigen retrieval. Pretreatment with 88% formic acid for 5 min was performed for α-synuclein and β-amyloid immunostaining. Endogenous peroxidases were blocked using a 3% hydrogen peroxide solution. The slides were blocked with 3% normal goat serum in Tris-buffered saline (TBS) solution for 1 h at room temperature (RT). Primary antibodies (anti-β-Amyloid, 1:500, clone 6E10, mouse, 803023, BioLegend, San Diego, CA; anti- phosphorylated tau, 1:200, clone AT8, MN1020, Invitrogen, Waltham, MA; anti-α–synuclein, 1:500, mouse, clone 42, 610787, BD, Franklin Lakes, NJ; anti-TDP-43 (phosphorylation independent), 1:1000, rabbit polyclonal, 12782-1-AP, Proteintech, Rosemont, IL; anti-TDP-43 phosphorylation at Serine 409/410, 1:200, rat, clone 1D3, 829901, BioLegend, San Diego, CA) were applied for 1 h at RT. Slides were washed with TBS. The biotinylated horseradish peroxidase secondary antibodies (Abcam) were applied for 1-h at RT. After washing with TBS, ABC (Vector Laboratories) and DAB kits (Vector Laboratories) were used. After washing, slides were counterstained with hematoxylin, dehydrated by alcohol and xylene, and coverslipped using Permount mounting medium (Fisher Chemical).

Assessment of Alzheimer’s and other neurodegenerations

For assessment of neurodegeneration, we used Hirano-silver stained sections to score the density of the NP according to CERAD criteria [59] and stage NFT according to Braak [10]. β-amyloid distribution was assessed using anti-β-Amyloid antibody according to the Thal phase system [84]. Alzheimer’s disease neuropathologic change was staged according to NIA-AA criteria [34]. Lewy body pathology was examined using the anti-α–synuclein antibody and scored according to Third Report of the Dementia with Lewy Bodies Consortium [55]. The diagnosis of primary aging-related tauopathy (PART) was formulated according to Crary et al. [15].

Immunofluorescent (IF) staining method for evaluation of TDP-43 nuclear clearance

TDP-43 nuclear clearance was evaluated in granule cells of the hippocampus dentate gyrus using IF staining with the phosphorylation independent anti-TDP-43 antibody. We counted only cells with granule cell morphology. IF staining was performed by triple labeling with phosphorylation independent TDP-43 (1:200), phosphorylation dependent TDP-43 (1:200), and NeuN (1:1000, mouse, clone A60, MAB377, MilliporeSigma, Burlington, MA) antibodies on 10-µm thick sections from formalin fixed paraffin embedded tissue blocks. Each section was deparaffinized with xylene and rehydrated with graded alcohols, and water. For antigen retrieval, tissue sections were microwaved for 10 min in citrate buffer (ab93678, Abcam). All sections were blocked with 3% normal goat serum in Tris-buffered saline (TBS) solution with 0.2% Triton X-100 for 1 h at RT. The primary antibodies were applied and incubated overnight at 4 °C. After washing with TBS, the secondary antibodies were applied for 1 h at RT: Alexa Fluor 488 anti-Rabbit IgG (1:400, ab150077, Abcam), Alexa Fluor 568 anti-Mouse IgG (1:400, ab175701, Abcam), Alexa Fluor 647 anti-Rat IgG (1:400, ab150159, Abcam), and DNA stain with DAPI (1:1000, Cat#10236276001, Roche). After washing with TBS, sections were coverslipped using ProLong Gold Antifade Mountant (P36930, Invitrogen). Slides were examined on a Leica Mica confocal microscope.

Cryptic exon-encoded neoepitope: IF staining

For IF staining of a TDP-43 dependent cryptic peptide, a mouse monoclonal antibody recognizing the cryptic exon-encoded neoepitope in Hepatoma-Derived Growth Factor-Like protein 2 (HDGFL2) was used [35] (Supplementary Information). Cryptic HDGFL2 was co-labeled with phosphorylation independent TDP-43 (1:200) and phosphorylation dependent TDP-43 (1:200). Antigen retrieval using HistoVT One (Nacalai Tesque, Kyoto, Japan) 95 °C 30 min was performed. IF staining method is as described above.

A peptide competition test was performed to evaluate the specificity of the cryptic HDGFL2 antibody IF staining. Three different peptides with overlapping amino acid sequences from the cryptic HDGFL2 sequence were tested: peptide 1: EPTIWFGKGHSGMLASEGREA; peptide 2: SEGREAVLTRLHESERVRKQ; peptide 3: HESERVRKQERERDTEERRE. Five times peptide to antibody by weight were added in the primary antibody mixture containing cryptic HDGFL2 antibody, phosphorylation independent TDP-43 antibody, and phosphorylation dependent TDP-43 antibody. IF staining was performed as described above. Slides were examined on a Leica Mica confocal microscope.

TDP-43 dependent cryptic exon: RNA in situ hybridization

For examination of cryptic exon containing RNA transcripts, BaseScope™ probes (Advanced Cell Diagnostics, Newark, CA) were designed for TDP-43 dependent cryptic exon containing transcripts of HDGFL2. Assays were conducted using a BaseScope™ RED Reagent Kit according to the manufacturer’s protocol in 5 brains with TDP-43+ NCIs, 5 brains with TDP-43 nuclear clearance only, and 5 brains without TDP-43 pathology. Sections were deparaffinized, rehydrated, and blocked for endogenous peroxidases with H2O2, pre-treated with ACD RNAScope Target Retrieval Reagents. The slides were incubated in Protease IV solution for 30 min at 40 °C, and then in probe solution for 2 h at 40 °C in the HybEZ II Oven. Amp reagents 1–8 were used according to the protocol. Slides were incubated in Fast Red for 10 min, counterstained with hematoxylin, dried, and coverslipped with VectaMount. Number of cryptic HDGFL2 transcripts (red dots) in 1000 granule cells in the dentate gyrus per brain were counted for quantification of cryptic HDGFL2 transcripts.

Statistical methods

Demographic, clinical, and neuropathologic characteristics were compared between the LATE-NC positive and negative groups using t-test and Chi-square (χ2) test, depending on the assessed variable. The number of cryptic HDGFL2 transcripts between pairs of different groups were compared using the Mann–Whitney test. For assessment of tau pathology in the dentate gyrus in different groups, the Kruskal–Wallis test was used for comparing all three groups and the Mann–Whitney test was used for comparing pairs.

To investigate the effects of age, sex and APOE ε4 status on the probability of LATE-NC presence, we used logistic regression analysis with LATE-NC as the outcome. The initial full model included the predictors of sex, APOE ε4 (carrier vs. non-carrier), age at death, and all 2-way interactions among these three factors. Backwards elimination was used to reduce the model for all the interactions at the p = 0.1 level.

We used separate logistic regression to model the relationship between CERAD scores and Braak stages and probability of LATE-NC presence. CERAD scores and Braak stages were used separately as continuous predictors, assuming linearity between logit of LATE-NC presence with CERAD scores and Braak stages. Models were conducted without and with adjustments of age of death, sex and APOE.

We used separate linear mixed effects models with each longitudinal cognitive measure as the outcome to estimate the effect of LATE-NC on the longitudinal cognitive trajectories. Time to death (years) was used as the time variable in linear mixed effects models. Two sets of models were tested for each cognitive measure. The first sets of models included covariates of sex and age of death. The fixed effects included LATE-NC (presence vs absence), age at death, sex, time, time square, and the interactions of LATE-NC, age at death and sex with time and time square. Since the cognitive trajectories were modeled to be quadratic and nonlinear, the effects of LATE-NC on cognitive trajectories were tested using F test with 2 degrees of freedom so that interactions of LATE-NC with time and time square were tested simultaneously. If the effect of LATE-NC on prior cognitive trajectories was significant, we estimated the cognitive performance difference between LATE-NC groups at 1 year, 5 years and 10 years prior to death based on the model for illustration purposes. For the second set of models, we additionally included CERAD scores and Braak stages as covariates. Random effects of the model included intercept, time and time square with unstructured covariance. The statistical tests were two-sided with a type I error of 0.05 unless stated otherwise.

All the analyses were conducted in SAS 9.4 (Cary, NC).

Results

Characteristics of participants in the BLSA cohort

This study included 309 deceased BLSA participants. Males outnumber females because the BLSA cohort was initially limited to men with enrollment of women beginning in 1978. Mean age at death was 88.5 years (ages range 57–109 years). The participants were predominantly white (93.5%), 33.0% had normal cognition, 14.6% had MCI, and 52.4% had dementia before death. Other demographic and clinico-pathological characteristics including CERAD neuritic plaque (NP) score and Braak neurofibrillary tangle (NFT) stage for ADNC are summarized in Table 1.

Table 1 Demographics and clinicopathologic characteristics of the BLSA autopsy participants (n = 309)
Full size table

Clinical and pathological features of TDP-43 proteinopathy

Among the 309 BLSA participants, 118 (38.2%) had LATE-NC comprised of TDP-43+ neuronal cytoplasmic inclusions (NCIs) accompanied by nuclear clearance of TDP-43 (Fig. 1a–c). According to the recently updated LATE-NC staging system [70], among the 309 BLSA participants, 27 (8.74%) are stage 1a, 9 (2.91%) are stage 1b, 71 (23.0%) are stage 2, and 11 (3.56%) are stage 3. Other LATE-NC, including dense networks of TDP-43+ neurites (Fig. 1d), TDP-43+ glial staining (Fig. 1e–g), subpial and subependymal TDP-43+ processes (Fig. 1h–j), and perivascular TDP-43+ pattern (Lin bodies) (Fig. 1k, l) [49] were also observed. TDP-43+ glial staining or subpial and subependymal TDP-43+ cell processes in the absence of NCIs were present in 12 (3.88%) brains, corresponding to LATE-NC stage 1c, which is not included in the LATE-NC positive group in all the following analyses [70] (Table 1). Perivascular TDP-43 pathology is mostly found in brains with TDP-43+ NCIs. From a total of 118 brains with TDP-43+ NCIs, 62 (52.5%) had perivascular TDP-43 pathology, while of 44 TDP-43 with nuclear clearance only, none had perivascular TDP-43 pathology. For total of 122 brains without TDP-43 pathology, only 2 (1.6%) had perivascular TDP-43 pathology. Sections of precuneus and inferior parietal lobule were stained in randomly selected cases (n = 22) encompassing all LATE-NC stages. Only 4 of the 22 brains revealed TDP-43 lesions. One LATE-NC stage 2 case had TDP-43+ NCIs and neurites in both precuneus and inferior parietal lobule. Another LATE-NC stage 2 case had one TDP-43+ NCI and few neurites in the inferior parietal lobule, while two additional LATE-NC stage 2 cases had only rare TDP-43+ neurites in the inferior parietal lobule.

Fig. 1
figure 1

Pathologic features of TDP-43 immunohistochemistry in BLSA cohort. a TDP-43+ neuronal cytoplasmic inclusions (NCIs) (phosphorylation dependent TDP-43 antibody). b TDP-43+ NCIs and neurites (arrow) (phosphorylation dependent TDP-43 antibody). c TDP-43 nuclear clearance (arrow) and cytoplasmic inclusion by phosphorylation-independent TDP-43 antibody. d Dense networks of TDP-43+ neurites (phosphorylation dependent TDP-43 antibody). e–g TDP-43+ glial cells, including oligodendrocytes (e), astrocytes (f), and astrocytes with ARTAG-like pattern (g) (phosphorylation dependent TDP-43 antibody). h–j TDP-43+ cell processes in subpial region and subependymal region (phosphorylation dependent TDP-43 antibody), usually associated with abundant corpora amylacea. k, l perivascular TDP-43+ pattern (phosphorylation dependent TDP-43 antibody). Scale bars, 10 µm

Full size image

The participants with LATE-NC have older mean age of death than the TDP-43 negative group (91.3 vs. 86.8 years, p < 0.001) (Table 1). The frequency of LATE-NC increases with age (Fig. 2a), is higher in females than males (52.1% vs. 29.5%, p < 0.001) (Fig. 2b), but not significantly different in APOE ε4 carriers as compared to non-APOE ε4 carriers (45.3% vs. 34.4%, p = 0.126) (Fig. 2c). To investigate the effects of age, sex, and APOE ε4 status on the probability of LATE-NC presence, we used logistic regression analysis with LATE-NC as the outcome. None of the interaction terms is significant in the logistic regression model relating age, sex, and APOE ε4 with the probability of LATE-NC. The final model included the main effects of age, sex and APOE ε4, and this model has c statistics (AUC) of 0.695. Age at death and female sex are significantly predictive of LATE-NC (p = 0.0002 and p = 0.0014, respectively; Fig. 2b, Table 2). Every year increment increased the odds of being positive by 6.7%. Being female increased the odds of LATE-NC by 2.33-fold. Although APOE ε4 increased the odds of LATE-NC positivity by 60.7%, it did not reach statistical significance at 0.05 level. Results are not adjusted for other covariates. (Fig. 2c, Table 2).

Fig. 2
figure 2

Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC) positivity with age, sex, APOE ε4 status. a The frequency of LATE-NC increases with age at death. b Predicted probability of LATE-NC positivity increases 6.7% per year and is higher in females. Being female increases the odds of LATE-NC by 2.33-fold (p = 0.0014). c The APOE ε4 allele increases the probability of LATE-NC by 60.7% but does not reach statistical significance at 0.05 level (p = 0.1052). Results are not adjusted for covariates (also see Table 2)

Full size image
Table 2 Relation of age, sex, and APOE ε4 with LATE-NC by logistic regression
Full size table

Association of LATE-NC with Alzheimer’s disease neuropathologic changes

The frequency of LATE-NC increases with higher CERAD NP scores and Braak NFT stages (Fig. 3) [10, 59]. After adjusting for age at death, sex, and APOE ε4 status, both CERAD NP score and Braak NFT stage are significantly associated with probability of LATE-NC from logistic regression models (p = 0.0008 and < 0.0001, respectively; Table 3); and each unit increase in CERAD NP score and Braak NFT stage is associated with a 57% and 58% increase in odds of LATE-NC, respectively (Table 3, Supplementary Table 1 and 2). Due to higher frequency of LATE-NC in females compared to males, we further examined the relationship between sex and ADNC by a regression model. After adjusting for age at death and APOE ε4 status, males had lower CERAD NP score (beta = − 0.296, standard error = 0.131, p = 0.024) and Braak score (beta = − 0.540, standard error = 0.157, p = 0.0007) compared with females.

Fig. 3
figure 3

Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC) positivity with CERAD neuritic plaques (NP) score and Braak neurofibrillary tangles (NFT) stage. a The frequency of LATE-NC positivity increases with higher CERAD NP scores. b The frequency of LATE-NC positivity increases with higher Braak NFT stages. By logistic regression models, there are significant association between the LATE-NC positivity with both CERAD NP score (see Table 3 and Supplementary Table 1) and Braak NFT stage (see Table 3 and Supplementary Table 2)

Full size image
Table 3 Relation of CERAD NP score and Braak NFT stage with LATE-NC
Full size table

LATE-NC and domain-specific cognitive decline

Participants of the BLSA cohort underwent comprehensive longitudinal cognitive assessments. The characteristics of the longitudinal cognitive sample are in Supplementary Table 3. Out of the total 309 participants, 243 participants who had complete longitudinal cognitive assessments were included in the analysis of cognitive decline. We used separate linear mixed effects models with each longitudinal cognitive measure as the outcome, and sex and age at death as covariates in the first set of models. The effects of LATE-NC on longitudinal trajectories of CVLT immediate recall (verbal memory) and Card Rotations Test (visuospatial ability) are highly significant in the first models (CVLT immediate recall p = 0.0001, Card Rotations Test p < 0.0001; Fig. 4a, b, Table 4). Both tests showed significant differences in scores between LATE-NC positive and negative groups. The difference estimates for CVLT immediate recall scores are − 13.40 at 1 year prior to death (p < 0.0001), − 8.94 at 5 years prior to death (p < 0.0001), and − 4.41 at 10 years prior to death (p = 0.0124). The difference estimates for Card Rotations Test scores are − 13.62 at 1 year prior to death (p = 0.0135), − 12.28 at 5 years prior to death (p = 0.0018), and − 7.68 at 10 years prior to death (p = 0.0709). The effects of LATE-NC positivity on longitudinal slopes of MMSE and Category Fluency Test are also significant (p = 0.039 and p = 0.026, respectively) in the first model (Fig. 4c, d, Table 4). The differences in MMSE and Category Fluency Test scores between LATE-NC positive and negative groups are also significant. The difference estimates of MMSE and Category Fluency Test scores are − 1.87 and − 2.05 at 1 year prior to death (p = 0.0054 and 0.0044, respectively), − 1.42 and − 1.45 at 5 years prior to death (p = 0.0016 and 0.0048, respectively), and − 0.86 and − 0.81 at 10 years prior to death (p = 0.0023 and 0.0092, respectively). After adding CERAD NP score and Braak NFT stage as covariates, the effects of LATE-NC on longitudinal trajectories of CVLT immediate recall and Card Rotations Test remain significant (CVLT immediate recall p = 0.0048, Card Rotations Test p = 0.0007). However, the effects on the longitudinal slopes of MMSE and Category Fluency Test are no longer significant (p = 0.286 and p = 0.183, Table 4). The effects of LATE-NC on the longitudinal trajectories and levels of cognitive measures of Trail Making Test A, Trail Making Test B, Trail Making Test B-A, and Letter fluency test are not significant. In summary, LATE-NC has significant effects on verbal memory (CVLT immediate recall), visuo-spatial ability (Card Rotations Test), mental status (MMSE) and semantic fluency (Category Fluency Test). Notably, the effects of LATE-NC on verbal memory and visuospatial ability remain significant after adjusting for ADNC. However, LATE-NC does not significantly affect attention, executive function (Trail Making Tests) or phonemic fluency (Letter Fluency Test).

Fig. 4
figure 4

Longitudinal trajectories of cognitive function in different domains. The effects of Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC) on longitudinal trajectories of CVLT immediate recall (verbal memory) (a) and Card Rotations Test (visuospatial ability) (b) are highly significant. The differences in MMSE (c) and Category Fluency Test (d) trajectories between LATE-NC positive and negative groups are also significant

Full size image
Table 4 LATE-NC effects on cognitive domains
Full size table

TDP-43 nuclear clearance: an early marker of TDP-43 proteinopathy

Under normal circumstances, TDP-43 is localized in the nucleus of neurons, a feature demonstrated using a phosphorylation independent TDP-43 antibody (Figs. 1c and 5d). By contrast, TDP-43 proteinopathy is characterized by a triad of TDP-43+ cytoplasmic and neuritic inclusions (NCI) plus nuclear clearance. There is growing evidence suggesting that TDP-43 nuclear clearance, in the absence of NCI, is sufficient to cause loss of function of TDP-43 in neurons, and it is likely an early event of TDP-43 proteinopathy [66, 81, 87]. We examined this notion by evaluating TDP-43 immunohistochemistry (IHC) in the dentate gyrus of the hippocampus with a phosphorylation independent TDP-43 antibody in all subjects in the BLSA cohort. Only cells with neuronal morphology were evaluated. Triple immunofluorescent labeling with phosphorylation independent and phosphorylation dependent TDP-43 antibodies, and neuronal marker NeuN was performed to assess nuclear clearance of TDP-43 as well as NCI (Fig. 5a–c). Results defined three distinct groups: (1) no TDP-43 pathology (n = 147, 47.6%), (2) presence of both TDP-43+ NCIs with nuclear clearance of TDP-43 (n = 118, 38.2%), and (3) presence of TDP-43 nuclear clearance only without TDP-43+ NCIs (n = 44, 14.2%). Remarkably, the mean age of participants in the TDP-43 nuclear clearance only group is younger than the TDP-43+ NCIs group (86.7 vs. 91.3, p = 0.002). Comparison of TDP-43 nuclear clearance only group and no TDP-43 pathology group showed that the TDP-43 nuclear clearance only group has higher rate of APOE ε4 (36.6% vs. 18.8%, p = 0.033; Table 5). The effects of TDP-43 nuclear clearance in absence of NCI on the longitudinal trajectories and levels of cognitive measures are not significant (Table 6, Supplementary Table 4).

Fig. 5
figure 5

Nuclear clearance of TDP-43 in dentate gyrus of the hippocampus. a–c Three representative cases of TDP-43 nuclear clearance without TDP-43+ neuronal cytoplasmic inclusions (NCIs). Immunofluorescent stains with neuronal marker (NeuN), phosphorylation-independent TDP-43 (TDP-43), phosphorylation-dependent TDP-43 (pTDP-43), and DNA stain with DAPI revealed some neurons (white arrow) have loss of nuclear TDP-43 but no TDP-43+ NCIs on either by phosphorylation-dependent TDP-43 and phosphorylation-independent TDP-43 stains. d The immunoperoxidase stain with phosphorylation-independent TDP-43 antibody showed loss of TDP-43 staining in the nuclei of the granule cells (arrows) but no TDP-43+ NCIs. e Age distribution of TDP-43 nuclear clearance only and TDP-43+ NCIs in the BLSA and younger autopsy cohorts. Scale bars, 10 µm

Full size image
Table 5 Clinico-pathologic characteristics of cases with TDP-43 nuclear clearance only vs. no TDP-43 pathology in the BLSA cohort
Full size table
Table 6 TDP-43 nuclear clearance effects on cognitive domains
Full size table

Because the age of death of participants in BLSA cohort was ≥ 57 years of age (mean 88.5 years), we examined the autopsy brain of 24 younger individuals aged 20 to 60 years (mean age 43 years). Among these 24 brains, 4 (16.7%) had low-level ADNC, 6 (25%) had primary aging related tauopathy (PART), and 14 (58.3%) were normal. Their demographic and clinico-pathological characteristics are summarized in Table 7. TDP-43 nuclear clearance without TDP-43+ NCIs was first noted in the 50 s, while the TDP-43+ NCIs appeared in the 60 s (Fig. 5e, Supplementary Table 5). We did not find any brains with TDP-43 nuclear clearance in brains younger than 50 years old. In summary, TDP-43 nuclear clearance precedes the onset of TDP-43+ NCIs by approximately a decade.

Table 7 Demographics and clinico-pathologic characteristics of individuals from the younger autopsy cohort (n = 24)
Full size table

Cryptic exon RNA and protein detection

TDP-43 represses the splicing of cryptic exons and loss of this function has been proposed to be an important mechanism of TDP-43 associated proteinopathy and neurodegeneration [13, 46, 50, 52, 57, 75]. To assess for TDP-43 splicing repression, we used BaseScope RNA in situ hybridization with an RNA probe targeting a cryptic exon in Hepatoma-Derived Growth Factor-Like 2 (HDGFL2) transcript in 5 brains with TDP-43+ NCIs, 5 brains with TDP-43 nuclear clearance only, and 5 brains without TDP-43 pathology. All 5 cases (100%) with TDP-43+ NCIs and all 5 cases (100%) with TDP-43 nuclear clearance only revealed cryptic HDGFL2 RNA transcripts (Fig. 6a, b, 7b, d). All five brains without TDP-43 pathology were negative for cryptic HDGFL2 RNA transcript (Fig. 6c). The quantification of cryptic HDGFL2 RNA transcripts revealed that brains with TDP-43+ NCIs had an average of 260.5 per 1000 cells (median: 197 per 1000 cells) (Fig. 6d). Brains with TDP-43 nuclear clearance only had an average of 174.7 per 1000 cells (median: 166 per 1000 cells) (Fig. 6d). Brains without TDP-43 pathology had an average of 12.2 per 1000 cells (median: 13 per 1000 cells), which was considered background signal. Both brains with TDP-43+ NCIs and brains with TDP-43 nuclear clearance only showed significantly higher numbers of cryptic HDGFL2 transcripts compared to brains without TDP-43 pathology (Mann–Whitney test, p < 0.01 for both comparisons) (Fig. 6d). No significant difference in levels of cryptic HDGFL2 transcripts was observed between brains with TDP-43+ NCIs compared to those with TDP-43 nuclear clearance only (Mann–Whitney test, p = 0.15).

Fig. 6
figure 6

RNA in situ hybridization using a probe targeting cryptic HDGFL2 transcripts. a Cryptic HDGFL2 RNA in situ hybridization shows positive cryptic HDGFL2 labeling in a case with TDP-43+ NCIs. b Cryptic HDGFL2 RNA in situ hybridization shows positive cryptic HDGFL2 labeling in a case with TDP-43 nuclear clearance without TDP-43+ NCIs. c Cryptic HDGFL2 RNA in situ hybridization is negative in a case with no TDP-43 pathology. In all cases, the positive control probe (ubiquitin C) shows prominent labeling, and the negative control shows no signal (a–c). Scale bar, 25 μm. d Quantification of cryptic HDGFL2 transcripts showed higher numbers of cryptic HDGFL2 transcripts in TDP-43+ NCIs group and TDP-43 nuclear clearance only group compared to no TDP-43 pathology group. *p < 0.05, **p < 0.01

Full size image
Fig. 7
figure 7

Cryptic HDGFL2 protein and RNA transcript detection in cases with TDP-43+ NCIs (a, b) and TDP-43 nuclear clearance only (c, d). Immunofluorescent stains with antibody against cryptic HDGFL2 peptide, phosphorylation-independent TDP-43 (TDP-43), phosphorylation-dependent TDP-43 (pTDP-43), and DAPI for DNA stain. a presence of cryptic peptide in neuronal nucleus and perikaryon (arrows) and TDP-43+ NCIs (arrowheads). b positive cryptic HDGFL2 RNA in situ hybridization (arrows) in the same brain as in a. c presence of cryptic peptide (arrow) and absence of nuclear TDP-43. d positive cryptic HDGFL2 RNA in situ hybridization (arrows) in the same brain as in c. Scale bars, 25 µm

Full size image

We then used a monoclonal antibody recognizing the cryptic exon-encoded neoepitope in HDGFL2 to corroborate its cryptic exon inclusion (Supplementary Information) [35]. Cryptic HDGFL2 IF staining was performed in 6 brains with TDP-43+ NCIs, 10 brains with TDP-43 nuclear clearance only, and 6 brains without TDP-43 pathology. All 6 (100%) brains with TDP-43+ NCIs were positive for cryptic HDGFL2 accumulation (Fig. 7a). Among brains with TDP-43 nuclear clearance only, 4 out of 10 (40%) had positive cryptic HDGFL2 accumulation (Fig. 7c). All 6 brains without TDP-43 pathology were negative for cryptic HDGFL2 staining. A peptide competition study was conducted to ascertain the specificity of the cryptic HDGFL2 antibody. Three different peptides, each possessing amino acid sequences derived from the cryptic peptide within HDGFL2, were tested. The peptide bearing the C-terminal of the cryptic HDGFL2 peptide effectively blocked the staining of the cryptic HDGFL2 antibody, whereas the other two peptides failed to block the cryptic HDGFL2 antibody staining (Fig. 8). Together, these data suggest that during aging loss of TDP-43 splicing repression occurs initially in neurons depleted of nuclear TDP-43 lacking NCI, supporting the notion that such loss contributes to the progression of disease or exacerbates neurodegeneration and cognitive decline in mixed etiology dementia harboring TDP-43 pathology.

Fig. 8
figure 8

Peptide competition study for cryptic HDGFL2 antibody demonstrating block of staining by peptide 3. a Immunofluorescent stains of cryptic HDGFL2, phosphorylation-independent TDP-43 (TDP-43), phosphorylation-dependent TDP-43 (pTDP-43), and DNA stain with DAPI reveals cryptic HDGFL2 expression in cells with TDP-43+ NCIs. b, c the cryptic HDGFL2 antibody staining is not blocked by peptide 1 or peptide 2. d peptide 3 completely blocks the cryptic HDGFL2 antibody staining. Scale bars, 20 µm

Full size image

Association of TDP-43 nuclear clearance and tau pathology

TDP-43 proteinopathy is associated with increased tau pathology in patients with AD, even in cases lacking high Braak NFT stages [58, 80, 85]. We examined the tau pathology in dentate gyrus of the hippocampus in brains from three TDP-43 pathology groups: brains with TDP-43+ NCIs, brains with TDP-43 nuclear clearance only, and brains without TDP-43 pathology. Subsets of brains from Braak NFT stage 0-III or Braak NFT stage V/VI with similar CERAD NP scores and Braak NFT stages within each group were selected. There were no significant differences in the average CERAD NP scores and Braak NFT stages between three TDP-43 pathology groups (Supplementary Table 6 and Supplementary Table 7), confirmed by Kruskal–Wallis tests (p = 0.452 for CERAD NP scores and p = 0.437 for Braak NFT stages for Braak NFT stage 0-III; p = 0.764 for CERAD NP scores and p = 0.670 for Braak NFT stages for Braak NFT stage V/VI). We identified increased tau-positive granule cells in the dentate gyrus from brains with TDP-43 NCIs and TDP-43 nuclear clearance only (Fig. 9a). Brains with TDP-43+ NCIs had a significant higher number of tau-positive granule cells compared to brains with no TDP-43 pathology in both Braak NFT stage 0-III group and Braak NFT stage V/VI group (p < 0.0001 for Braak NFT stage 0-III, and p = 0.0147 for Braak NFT stage V/VI, Fig. 9b, Supplementary Table 8–10). Importantly, brains with TDP-43 nuclear clearance only also showed a significant higher number of tau-positive granule cells compared to brains with no TDP-43 pathology in both Braak NFT stage 0-III group and Braak NFT stage V/VI group (p = 0.0243 for Braak NFT stage 0-III, and p = 0.028 for Braak NFT stage V/VI, Fig. 9b, Supplementary Table 8–10). Taken together, these results support the idea that loss of TDP-43 splicing repression exacerbates tau pathology to accelerate neurodegeneration and cognitive decline in cases of ADNC.

Fig. 9
figure 9

Increased tau pathology in dentate gyrus in brains with TDP-43+ NCIs and brains with TDP-43 nuclear clearance. a Immunohistochemical stains of phospho-tau (AT8) showed increased tau pathology in the dentate gyrus in brains with TDP-43+ NCIs and TDP-43 nuclear clearance. b Quantification of number of tau-positive granule cells in dentate gyrus showed significant higher number in brains with TDP-43 NCIs and TDP-43 nuclear clearance compared to brains with no TDP-43 pathology in both Braak NFT stage 0-III and Braak NFT stage V/VI groups. Scale bar, 25 µm; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Full size image

Discussion

This clinico-pathological study of more than 300 BLSA participants shows that LATE-NC is present in 38.2% of the cohort, it frequently coexists with ADNC, and is associated with domain-specific cognitive decline in the aging population. Although LATE-NC is associated with higher levels of ADNC, our results reveal that TDP-43 proteinopathy has clinical impacts on verbal memory and visuospatial function which are independent from the ADNC. This observation suggests that TDP-43 proteinopathy and canonical pathology of Alzheimer’s disease impacts different pathogenic processes in the aging brain. Thus, TDP-43 proteinopathy should be considered as one of several key independent factors that contributes and interacts with other pathogenic pathways underlying cognitive decline in the aging population.

Our study shows that the prevalence of TDP-43 proteinopathy and LATE-NC is higher in cases with higher CERAD NP score and Braak NFT stage, similar to a recent study that collected data from 13 different cohorts [68]. Our study also shows a higher frequency of LATE-NC in females independent of age or APOE status. This higher frequency could be associated with higher CERAD NP scores and Braak NFT stages in females in our cohort. Some previous studies have also shown higher ADNC pathology scores in females [24, 32], but the presence of a greater frequency of LATE-NC in females has not been previously addressed. Some previous studies have shown an association between APOE ε4 allele and LATE-NC [16, 21, 39, 40, 94], while other studies have failed to demonstrate this association [14, 28, 29, 78, 88]. In our study, we observed a statistically significant relationship between the APOE ε4 allele and TDP-43 nuclear clearance. Although the APOE ε4 allele increased the probability of TDP-43 NCIs, this relationship did not reach statistical significance. This may be attributed to an insufficient sample size.

The longitudinal trajectories of cognitive function in our study show that LATE-NC was associated with faster rates of cognitive decline in verbal episodic memory (California Verbal Learning Test (CVLT) immediate recall), visuospatial ability (Card Rotations Test), mental status (MMSE) and semantic fluency (Category Fluency Test). The CVLT immediate recall and Card Rotations Test scores show large differences between participants with and without LATE-NC. Both test results have more than 13-point differences at 1 year prior to death. By contrast, the differences in MMSE and Category Fluency Test are smaller. Several lines of evidence indicate that NPs in AD are associated with rate of cognitive decline, and that NFTs are also associated with cognitive impairment in both AD patients and non-AD patients with PART [5, 6, 19, 37, 53, 60]. Therefore, we used both CERAD NP score and Braak NFT stage as covariates to adjust for effects of ADNC severity in our analysis of cognitive decline, demonstrating that effects of TDP-43 loss on decline of verbal memory and visuospatial ability were independent of ADNC. By contrast, most previous studies of LATE-NC and cognitive decline have not included CERAD NP score or Braak NFT stage as covariates, limiting the assessment of the separate contributions of TDP-43 proteinopathy and ADNC to cognitive decline. We confirmed the impact of TDP-43 proteinopathy on verbal episodic memory decline, already shown by some previous studies [44, 92]. In addition, our study is the first to show that cases with TDP-43 proteinopathy have significant decline in visuo-spatial ability independent of ADNC. The impairment of visuo-spatial ability is frequently seen in patients with Alzheimer’s dementia and often occurs in early stages [38, 89]. Many brain cortical regions are involved in visuo-spatial ability, including occipital lobe, parietal lobe, prefrontal cortex, premotor cortex, inferior temporal cortex, and mesial temporal lobe [47]. The parietal lobe, especially posterior parietal cortex which includes superior parietal lobule, inferior parietal lobule and intraparietal sulcus, is important for visuospatial processing [30, 47, 93]. The precuneus has strong functional connectivity with the angular gyrus and is important for visuospatial processing in humans [8, 47]. In Alzheimer’s disease, several studies have shown that early amyloid deposition occurs in the precuneus during preclinical stages [4, 20], but no prior study has examined the TDP-43 pathology in this region. To examine LATE-NC in the parietal lobe and precuneus, we randomly selected 22 brains spanning the whole range of LATE-NC stages. We found 4 of 22 cases had TDP-43 pathology in inferior parietal lobule or precuneus, and the burden of TDP-43 pathology in these cases is quite low. By contrast, the mesial temporal structures uniformly bear LATE-NC. These structures are part of the parieto-medial temporal pathway that subserves visuo-spatial processing in humans and is important in navigation, topographic and spatial learning [47] and the hippocampus plays an important role in allocentric spatial memory formation in human [73, 82]. This pathway is also important in the decline of visuo-spatial ability in patients with Alzheimer’s disease [47, 48, 95]. Therefore, loss of TDP-43 splicing repression occurring in the mesial temporal lobe may underlie the visuo-spatial impairment observed in patients with TDP-43 pathology. However, more research is needed to investigate how TDP-43 pathology affects the visuo-spatial abilities.

TDP-43 nuclear clearance: an early marker of TDP-43 loss of function and proteinopathy

TDP-43 pathology is characterized by TDP-43+ NCIs and nuclear clearance of TDP-43. Until now, few studies have focused on neurons with nuclear clearance of TDP-43 but without cytoplasmic inclusions [11, 12, 66, 81, 87]. An important function of TDP-43 is repressing the splicing of cryptic exons [13, 46, 50, 52, 57, 75]. Our previous study revealed that brains with nuclear clearance of TDP-43 without TDP-43+ NCIs had cryptic exon incorporation, indicating a loss of TDP-43 function when nuclear clearance is present in the absence of TDP-43 NCIs [81]. Others have confirmed that presence of TDP-43 regulated cryptic RNAs in AD brains [23]. In this study, we showed that 14.2% BLSA brains already had nuclear clearance of TDP-43 in the absence of TDP-43+ NCIs. Importantly, this phenomenon was first observed in individuals in their 50 s, about 10 years earlier than the earliest appearance of TDP-43+ NCIs. However, we did not find any nuclear clearance of TDP-43 in individuals younger than 50 years. We further used cryptic HDGFL2 antibody and RNA in situ hybridization to show that brains with nuclear clearance of TDP-43 exhibit cryptic exon inclusion and accumulation of cryptic peptide found in HDGFL2, an important pathogenic mechanism of TDP-43 proteinopathy. Recent studies have showed an association between TDP-43 pathology and increased tau pathology in the hippocampus, implying that TDP-43 pathology may exacerbate tau aggregation [58, 80, 85]. Our study demonstrated that not only brains with TDP-43+ NCIs but also brains with TDP-43 nuclear clearance without NCIs have higher burden of tau pathology in dentate gyrus of the hippocampus. This suggests that loss of TDP-43 splicing repression in neurons lacking NCI would initiate a pathogenic cascade that contributes to or accelerates tau aggregation. These findings indicate that the nuclear clearance of TDP-43 is an early pathogenic event of TDP-43 proteinopathy, upstream to the TDP-43 NCIs, and that it precedes cognitive decline. Furthermore, our observations suggest that TDP-43 proteinopathy is more widespread in the population and appears at a younger age than currently believed. Further studies on the clinical and biological significance of neurons with nuclear clearance of TDP-43 but lacking TDP-43+ NCIs are necessary.

In summary, this study shows that loss of TDP-43 splicing repression is a seminal and early event in TDP-43 proteinopathy which is first manifested by neurons that exhibit nuclear clearance of TDP-43 lacking NCI as early as 50 s years of age, and then followed by TDP-43+ NCIs a decade later. Loss of TDP-43 splicing repression is associated with increased tau pathology, suggesting that such loss contributes or exacerbates the pathological conversion of tau in ADNC. LATE is common in brains with advanced ADNC, and has specific effects on the decline of memory and visuo-spatial ability in the aging population that are independent of AD.

Limitations

Our observations have some limitations. As all histological sections for immunohistochemistry and in situ hybridization were performed on tissues from the left hemisphere, therefore, possible asymmetries between the cerebral hemispheres cannot be detected. Another unavoidable limitation is the lack of cognitive evaluation of subjects in the young cohort. It is conceivable that the presence of TDP-43 abnormalities in the subjects from the young cohort could represent the early stage of FTLD. Some patients with FTLD exhibit anosognosia, which could not be ascertained without a cognitive evaluation.