U1‐70K and U1A and tau pathogenesis in demented and non‐demented individuals with Down syndrome

Abstract

INTRODUCTION

Splicing protein mislocalization is associated with tau pathogenesis, but its role in Down syndrome (DS) is under‐investigated.

METHODS

Spliceosome associations with tau and plaque pathology were examined in frontal cortex from DS with dementia (DSD+) and without dementia (DSD−) using quantitative immunoblotting and immunohistochemistry.

RESULTS

U1‐70K and U1A levels were downregulated, and hnRNPA2B1, 3Rtau, and 4Rtau were upregulated, whereas SRSF2 and CLK1 were unchanged in DSD+. The number of U1‐70K lightly labeled cells was only greater in layer III in DSD+. U1A intensely stained nuclei decreased significantly in layer III in DSD+, whereas those lightly labeled significantly increased in layers III and V–VI in DSD+. U1 mislocalization and tangles appeared in each laminae examined in both DS groups but were significantly greater in DSD+. Mislocalized U1s that co‐localized with AT8, and not TauC3, were significantly increased in DSD+.

DISCUSSION

U1 splicing proteins play a key role in tau pathogenesis in individuals with DS.

Highlights

• U1 mislocalization and tangle‐like profiles appeared in frontal cortex layer III and V–VI neurons in Down syndrome (DS).

• Frontal cortex hnRNPA2B1 nuclear morphometric values decreased in layer III in DS with dementia.

• Frontal cortex SRSF2 and CLK1 nuclear proteins were unchanged in DSD+ and without dementia (DSD−).

1. BACKGROUND

Although individuals with Down syndrome (DS) develop an Alzheimer's disease (AD) type dementia before age 50, its prevalence is variable, with approximately one third of older adults with DS lacking dementia at any age despite the onset of amyloid plaque and tau‐containing neurofibrillary tangles (NFTs) in their forties. ¹ , ² DS is estimated to affect 200,000 people in the United States, ³ and 5 to 8 million worldwide (www.globaldownsyndrome.org). Unfortunately, treatment approaches for dementia in DS are inadequate. ⁴ , ⁵ The discovery that abnormal splicing of exon 10 caused neurodegeneration linked to the formation of tau‐bearing NFTs in frontotemporal dementia–amyotrophic lateral sclerosis (FTD‐ALS), ⁶ , ⁷ suggested a molecular link between tau splicing/spliceosome defects and tauopathies, including DS and AD. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² Abnormal splicing factors that increase protein and expression levels of 4Rtau to 3Rtau isoforms ¹³ , ¹⁴ seem to precede NFT formation in tauopathies. ⁷ , ¹⁵ , ¹⁶ Recent evidence indicates that the nuclear splicing complex plays a crucial role in the pathogenesis of AD, but less is known about DS. ¹² , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹

Splicing is a highly regulated process that removes introns (constitutive) and selected exons (alternative) from a precursor transcript (pre‐mRNA [messenger RNA]) to generate a mature messenger (mRNA) in the spliceosome. The latter is a dynamic nuclear ribonucleoprotein (RNP) complex, comprising five small nuclear RNAs: snRNAs (U1, U2, U4, U6, and U5), specific for each RNA or snRNP, such as U1‐70K, U1A, and U1C for U1 RNA, which are regulators of early spliceosome assembly, ²² and RNPs, including the heterogeneous nuclear (hn) RNP and serine/arginine (SR) families. ²³ , ²⁴ Alternative splicing is an highly regulated process that creates an increase in transcriptome heterogeneity and proteome diversity, ²⁵ which carries a risk of splicing defects/mutations, linked to tauopathies. ²⁶ , ²⁷ In fact, disruption of alternative splicing of tau exon 10 via serine‐ and arginine‐rich splicing factor 2 (SRSF2) causes altered expression of 3R/4Rtau isoforms, leading to NFT formation and neurodegeneration, ¹⁵ , ²⁸ , ²⁹ , ³⁰ supporting a critical role of SRSF2 in tau pathogenesis. ³¹ Alterations in hnRNPA2B1 expression associated with FTD‐ALS ³² , ³³ , ³⁴ and AD ⁹ have been suggested to mediate tau pathology in AD. ¹¹ , ¹² Studies by our group demonstrated aberrant extranuclear mislocalization of snRNP U1‐70K and U1A proteins in the form of insoluble NFT‐like structures in the frontal cortex (FC) in preclinical and frank AD and DS. ¹⁸ , ¹⁹ These data suggest that splicing proteins play a role in abnormal RNA splicing and tau pathogenesis in the FC of individuals with AD and DS. The FC is a hub of the memory dorsal default mode network (DMN), which plays a key role in episodic memory and the retrieval of autobiographical memories during attentionally demanding tasks, displays spatiotemporal differences in the onset of amyloid ³⁵ , ³⁶ , ³⁷ , ³⁸ and hypometabolism that reflects tau neurodegeneration ³⁹ as well as a loss of connectivity early in AD and DS. ³⁸ , ⁴⁰ , ⁴¹ However, whether differential changes in spliceosome protein levels occur in DS with (DSD+) and without dementia (DSD−) and their association with 3Rtau and 4Rtau isoforms and NFT pathology remain unknown. ¹⁹ , ⁴² , ⁴³ Here, we examined changes in several splicing proteins including: U1‐70K, U1A, hnRNPA2B1, SRSF2, the tanscription coordinator RNA polymerase II (RNA pol II) C‐terminal domain (CTD), ⁴⁴ the alternative splicing regulator kinase CLK1, ⁴⁵ and their association with tau pathology in FC neurons using semi‐quantitative immunoblotting, immunohistochemistry and quantitative morphometry in DSD+ and DSD−.

RESEARCH IN CONTEXT

1. Systematic review: Authors reviewed the literature using traditional (e.g., PubMed) sources, meeting abstracts, and presentations. Although there are few publications describing spliceosome involvement in DS pathology, there are virtually no studies in people with Down syndrome (DS) with or without dementia. Relevant citations are appropriately cited.

2. Interpretation: Our findings led to an integrated hypothesis describing the pathophysiology of splicing proteins and tau pathology in the frontal cortex of people with DS with or without dementia. This hypothesis is consistent with the non‐clinical and clinical findings currently in the public domain.

3. Future directions: We propose a framework for the generation of new hypotheses and the conduct of additional studies. For example, understanding: (1) the role of splicing proteins in the onset of non‐AD pathologies; (2) spliceosome alterations in subjective memory and in the non‐demented oldest of the old; (3) potential development of novel drugs; and (4) spatial transcriptomics to measure gene expression and local transcripts, down to the single‐cell gene‐array level.

2. METHODS

2.1. Subjects and tissue preparation

Tissue containing the FC—frozen or immersion fixed in 4% paraformaldehyde or 10% neutral buffered formalin fixed—was obtained from a total of 37 DS cases clinically diagnosed with dementia (DSD+; n = 25) or without dementia (DSD−; n = 12), from the University of California, Irvine Alzheimer's Disease Research Center (UCI ADRC; n = 18), Institut d'Investigacions Biomediques August Pi i Sunyer (IDIBAPS) sodium  Biobanc (Barcelona, Spain; n = 11), and Barrow Neurological Institute (BNI; n = 2), all components of the Down Syndrome Biobank Consortium (DSBC), and Rush University Department of Pathology (n = 6). DS diagnosis was confirmed by the presence of an extra HSA21 using fluorescence in situ hybridization and/or chromosome karyotype. Case 12 from UCI was confirmed as partial trisomy without triplication of the APP gene. ⁴⁶ Clinical, demographic, and neuropathological details are presented in Tables S1 and S2.

Dementia status in UCI ADRC participants with DS was determined in accordance with the International Classification of Diseases, Tenth Revision (ICD‐10) and dementia questionnaire for people with intellectual disability (DSM‐IV‐TR) criteria. All UCI ADRC subjects were followed using longitudinal research protocols prior to death. Assessments included physical and neurological exams and a history obtained from both the participant and a reliable caregiver. Standardized direct and indirect cognitive and behavioral assessments were also completed. A diagnosis of dementia requires deficits in two or more areas of cognitive functioning domains and progressive worsening of cognitive performance compared to an individual's baseline performance. Cognitive decline due to confounding factors that may mimic dementia (e.g., depression, sensory deficits, hypothyroidism) was eliminated. Premorbid IQ (intelligence quotient) was also determined for the UCI cases. Dementia diagnosis for the Rush, BNI, and IDIBAPS cases was determined by a neurologist trained in gerontology and discussion with a caregiver. Human research committees of Rush University, BNI at St. Joseph's Hospital and Medical Center, and UCI approved this study.

2.2. Western blotting

Of the seven splicing proteins examined, only U1‐70K and U1A displayed insoluble extranuclear tangle‐like structures; quantitative immunoblotting was carried out for only the soluble fraction of these splicing proteins in 5 DSD− and 13 DSD+ cases due to limited frozen tissue availability (see Table S1). FC gray matter was dissected free of white matter on dry ice and frozen at −80°C until the time of biochemical assay. Prior to use, samples were homogenized (150 mg/mL) in a phosphate buffer containing protease inhibitors (Sigma; St. Louis, MO, USA). ⁴⁷ , ⁴⁸ , ⁴⁹ Proteins were denatured in sodium dodecyl sulfate (SDS) loading buffer to a final concentration of 5 mg/mL and separated (50 µg/sample) by 4%–20% gradient sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS‐PAGE; Lonza; Rockland, ME, USA) and electrophoretically transferred to polyvinylidene fluoride membranes (Immobilon P, Millipore). Membranes were blocked in Tris‐buffered saline (TBS)/0.05% Tween‐20/5% milk (1 h) at room temperature (RT). After adding each primary antibody to the blocking buffer (Table S3 for dilutions and manufacturer's information), membranes were incubated overnight (4°C), and then washed, incubated at RT for 1 h with horseradish peroxidase‐conjugated goat anti‐mouse immunoglobulin G (IgG secondary antibody (1:8000, Bio‐Rad; Hercules, CA, USA) or goat‐anti‐rabbit IgG secondary antibody (1:5000, Bio‐Rad; Hercules, CA, USA), visualized by chemiluminescence on a Kodak Image Station 440CF (Perkin‐Elmer; Wellesley, MA, USA), and quantified with Kodak 1. Several proteins were developed in the same membrane per case due to the restricted frozen tissue availability, and signals were normalized to β‐tubulin and analyzed in three independent experiments across clinical groups. ⁵⁰ Immunoblot controls consisted of the deletion of the primary antibody that resulted in the absence of immunolabeled membrane bands. Regarding western blotting for hnRNPA2B1, we observed a strong 36‐37 kDa (A2) and a weaker 38 kDa (B1) band ⁵¹ in the FC in both groups. Based on these observations, we evaluated only the more intensely reactive 36–37 kDa band. In addition, we evaluated protein levels for different tau isoforms using a tau protein ladder containing the six tau isoforms (Peptide #T‐1007‐1; Watkinsville, GA, USA). Because 1N3Rtau and 1N4Rtau isoforms are the most abundant forms found in adult human brain, ⁵² , ⁵³ these were analyzed in FC tissue from the DS cases (see Figure S1). Figure 1 shows images of non‐spliced western blots, and Figures S2 and S3 show full immunoblot images.

FIGURE 1.

Representative immunoblots and scatter blots with bars showing significant downregulation of frontal cortex (A) U1‐70K and (B) U1A and upregulation of (C) pS5,2‐RNA pol II, (D) hnRNPA2B1, (E) 1N3Rtau, and (F) 1N4Rtau protein levels in DSD+ compared to DSD–. Immunoreactive signals were normalized to β‐tubulin levels. Light gray rectangular bands below the tubulin bands are known as shadows or ghost bands, caused by intense, localized signals that exhaust the enhanced chemiluminescence reaction. Mann–Whitney test, *p < 0.05, **p < 0.01,***p < 0.001. β‐tub, beta tubulin; DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia.

2.3. Immunohistochemistry

FC fixed tissue was cut frozen at 40 microns on a sliding microtome, and paraffin‐embedded cases were sectioned on a rotatory microtome at 8‐micron thickness. Free‐floating sections were stored in cryoprotectant at −20°C, washed in phosphate buffer pH 7.4 solution, and mounted on slides prior to processing. These sections were used for better visualization and quantification of U1‐70K and U1A tangle‐like structures in six DSD− and nine DSD+ cases (see Table S1). While paraffin‐embedded sections were used to measure U1A, U1‐70K, hnRNPA2B1, SRSF2 optical density, nuclear size, nuclear and cytoplasmic counts, phosphorylated AT8 and truncated TauC3 NFTs, neuropil threads (NTs), tau‐labeled plaques, 3Rtau and 4Rtau NFTs, and APP/Aβ (amyloid beta) plaque counts as well as the presence of APP/Aβ blood vessels in the FC of 11 DSD− and 22 DSD+ cases (see Table S1).

Both free‐floating and paraffin‐embedded sections were processed for single‐label avidin‐biotin–based immunohistochemistry. After tissue was deparaffinized, both types of sections were hydrated and pretreated either with a solution of citric acid (pH 6) for 10 min in a microwave (U1A, U1‐70K, hnRNPA2B1, AT8, and TauC3), 20 min with 1x Dako target retrieval pH 9 (Dako, Denmark) in a steamer (SRSF2), or with 88% formic acid for 10 min to aid in the visualization of APP/Aβ plaques. To detect 3Rtau and 4Rtau isoforms, sections were pretreated with a citric acid solution for 10 min (pH 6) and washed in TBS followed by 88% formic acid for 15 min. Immunostaining used antibodies directed against U1A (1:250, Sigma Aldrich), U1‐70K (1:150, Gift from Dr. Hales), SRSF2 (1:300, Abcam), hnRNPA2B1 (1:500, Invitrogen), APP/Aβ (6E10; 1:1000, BioLegend), 3Rtau (1:500, Millipore), 4Rtau (1:500, Millipore) isoforms, and phosphorylated (AT8; 1:800, Invitrogen) and truncated tau (TauC3; 1:200, Invitrogen) epitopes (Table S3). Sections were incubated with the appropriate primary antibody concentration in a TBS Triton X‐100/1% goat serum (GS) solution overnight at RT. After three 1% GS TBS washes, sections were incubated with a secondary goat biotinylated anti‐mouse antibody (1:200; 1 h) (Vector Labs; Burlingame, CA, USA) followed by a 1‐h incubation using a Vectastain ABC kit (Vector Labs; Burlingame, CA, USA). Subsequently, sections were developed in acetate‐imidazole buffer solution containing 0.05% 3,3′‐diaminobenzidine tetrahydrochloride (Thermo Fisher Scientific; Waltham, MA, USA) and 0.005% H₂O₂. Slides were dehydrated in an ascending series of ethanol concentrations (50%, 70%, 95%, and 100%), cleared in xylenes, and cover‐slipped using DPX (Electron Microscopy Sciences; Hatfield, PA, USA). To control batch‐to‐batch variation in chemicals, sections from each case were processed simultaneously for each antibody. APP/Aβ, 3Rtau, 4Rtau, AT8, and TauC3 immuno‐stained sections were counterstained with 0.1% Mayer's hematoxylin for 2 min (Electron Microscopy Sciences; Hatfield, PA, USA). Controls consisted of the omission of primary antibodies, resulting in an absence of immunoreactivity.

2.4. Immunofluorescence

Double immunofluorescence was used to evaluate the mislocalization of the splicing proteins U1‐70K and U1A within NFTs labeled with antibodies against AT8 and TauC3 in the FC tissue from five DSD− and nine DSD+ cases (Table S1). Of the various splicing antibodies used, only the U1s revealed mislocalization. After pretreatment with a solution of citric acid (pH 6) for 10 min in a microwave, a mounted free‐floating section per case was incubated overnight at room temperature with either U1‐70K or U1A and AT8 and TauC3 (see Table S3 for dilutions and manufacturer's information). The appropriate secondary antibodies were applied for 1 h at RT as follows: first Cy5‐conjugated donkey anti‐rabbit IgG for U1‐70K and U1A (1:200, Jackson Immuno‐research; West Grove, PA, USA) and second, after several washes, tissue was placed in Cy2‐conjugated donkey anti‐mouse IgG for AT8 and TauC3 (1:200, Jackson Immuno‐research). Sections were also stained with the nuclear fluorescence marker DAPI (1:2000, Thermo Fisher Scientific; Waltham, MA, USA) at RT for 10 min. Auto‐fluorescence was blocked with Auto‐fluorescence Eliminator Reagent (Millipore; Burlington, MA, USA) according to manufacturer's instructions. Sections were cover‐slipped with an aqueous mounting media (Thermo Fisher Scientific). Fluorescence was visualized with the aid of a Revolve Fluorescent microscope (Echo laboratories; San Diego, CA, USA) with excitation filters at wavelengths 405, 489, and 649 nm for 4'6'‐diamidino‐2‐phenylindole (DAPI; emission blue), Cy2 (emission green; AT8 and TauC3), and Cy5 (emission far‐red; U1‐70K and U1A), respectively, as described previously. ⁵⁴

2.5. Optical density and counts

Optical density (or OD) measurements and counts of FC layers III and V–VI SRSF2, U1A, U1‐70K, and hnRNPA2B1 immunoreactive (‐ir) nuclei were performed in 10 unbiasedly selected regions of interest (ROIs) per slide (n = 2) with a 40x objective across clinical groups using the Nikon elements software program as described previously. ⁵⁵ SRSF2‐, U1A‐, U1‐70K‐and hnRNPA2B1‐labeled nuclei were outlined manually, and OD and nuclear area measurements (µm²) were automatically analyzed in gray‐scale images in the same focal plane with the aid of Nikon Elements software. The average of background OD values from non‐labeled cortical areas was subtracted from the average measurements of positive nuclei. Counts of splicing protein positive nuclei that were semi‐qualitatively determined to display either strong or light immunoreactivity and neurons displaying mislocalized U1A and U1‐70K labeling with or without a stained nucleus were performed across clinical groups using the same image evaluated for OD. The number of AT8‐, 3Rtau‐, 4Rtau‐positive NFTs and APP/Aβ plaques in layers III and V–VI was determined in 10 unbiased ROIs at 200× in one section per case, whereas neuropile thread (NT) counts were undertaken in the same number of ROIs at 400×. Because paraffin‐embedded tissue limited the visualization of U1A‐ and U1‐70K‐positive tangle‐like structures, these profiles were counted in free‐floating slide‐mounted sections either single immunostained or double‐labeled immunofluorescently with AT8 and TauC3 in 10 unbiased ROIs with a 40x objective in one section per case. All measurements were performed blinded to demographics and clinical and neuropathological diagnosis and averaged per case.

2.6. Statistical analysis

FC protein levels, densitometry, cell counts, nuclear size, clinical‐neuropathological, and demographic values were compared between DS groups using a non‐parametric Mann–Whitney test, and a chi‐square test (sex and apolipoprotein E [APOE] ε4). The Wilcoxon signed‐rank test was used to evaluate differences in OD, cell counts, and nuclear size between cortical layers III and V–VI within each clinical group. Associations between splicing markers, counts, tau and plaque pathology, and neuropathological scores were evaluated using the Spearman test. Statistical significance was set at 0.05 (two‐sided). False discovery rate (FDR) was used to correct for multiple comparisons. Data were presented as scatter plots with bar‐plots (GraphPad Prism 10, GraphPad Software Inc.; Boston, MA, USA), expressed as mean ± standard deviation (SD) and linear regression graphs (SigmaPlot v15‐Inpixon, Graffiti; Palo Alto, CA, USA).

3. RESULTS

3.1. Demographics

Study subjects included cases with an antemortem clinical diagnosis of DSD− (n = 12) and DSD+ (n = 25), with ages ranging from 36 to 72 years. There were no differences in age at death (Mann–Whitney test, p = 0.14) and post‐mortem interval (PMI; Mann–Whitney test, p = 0.54) between DSD+ and DSD−. Sex and APOE e4 frequency were comparable between groups (chi‐square test, p > 0.05). However, brain weight was significantly larger in the DSD− compared to the DSD+ group (Mann–Whitney test, p < 0.001). Virtually all DSD+ cases displayed Braak stage VI, whereas 54% of DSD− were Braak stage III, 9% Braak stage IV, and 36% Braak stage V, and greater NFT pathology was seen in the demented DS cases (Mann–Whitney test, p < 0.001). When available, neuropathological ABC criteria were higher in DSD+ than DSD− (see Table S1). Clinical, demographic, and neuropathological details are summarized in Table S2.

3.2. Immunoblotting

FC quantitative immunoblotting for all splicing proteins and tau isoforms was carried out for the soluble fraction. We found a significant downregulation of U1‐70K (Figure 1A; Mann–Whitney test, p = 0.01) and U1A protein levels (Figure 1B; Mann–Whitney test, p = 0.03) in DSD+ compared to DSD−, whereas pS5,2‐RNA pol II (Figure 1C; Mann–Whitney test, p = 0.02), hnRNPA2B1 (Figure 1D; Mann–Whitney test, p = 0.02), 1N3Rtau (Figure 1E; Mann–Whitney test, p = 0.001), and 1N4Rtau (Figure 1F; Mann–Whitney test, p = 0.001) were upregulated in DSD+ compared to DSD−. There were no significant differences in SRSF2 and CLK1 protein levels between groups. Comparatively, FC U1‐70K levels were significantly higher than U1A in DSD+ (Table S4; Mann–Whitney test, p = 0.005), but not in DSD−. Although the ratio of pS5,2‐RNA pol II over pS5‐RNA pol II levels was comparable between the groups (Mann–Whitney test, p > 0.05), DSD+ displayed significantly higher levels pS5‐RNA pol II than pS5,2‐RNA pol II in both DS with (Mann–Whitney test, p ≤ 0.001) and without dementia (Mann–Whitney test, p = 0.01). The levels of 1N3Rtau were significantly higher than 1N4Rtau in DSD+ (Mann–Whitney test, p < 0.0001) and DSD− (Mann–Whitney test, p = 0.008). There was a significant upregulation of 4Rtau/3Rtau ratio (Mann–Whitney test, p = 0.001) and 4Rtau+3Rtau (Mann–Whitney test, p = 0.001) protein levels in DSD+ compared to DSD−. Correlational analyses revealed positive associations between Braak staging and FC 1N3Rtau, 1N4Rtau, pS5,2‐RNA pol II (Spearman test: 1N3Rtau r = 0.72, 0.001, 1N4Rtau r = 0.8, p = 0.0000002, pS3,2‐RNA pol II r = 0.56, p = 0.01), but negative association with U1‐70K (Spearman test, r = −0.59, p = 0.01) across groups. FC pS5,2‐RNA pol II levels correlated positively with 1N3Rtau (Spearman test, r = 0.57, p = 0.01) and 1N4Rtau (Spearman test, r = 0.63, p = 0.01) values and hnRNPA2B1 positively correlated with 1N3Rtau (Spearman test, r = 0.65, p = 0.005). FDR significance level is p = 0.02. Western blot results are summarized in Table S4.

3.3. Characteristics of frontal cortex U1‐70K, U1A, SRSF2, and hnRNPA2B1 immunostaining in DS

Bright‐field microscopy revealed U1A, U1‐70K, SRSF2, and hnRNPA2B1‐ir nuclei in the FC throughout all cortical layers in DS with and without dementia (Figure 2). Strong and light U1‐70K and U1A immunostained nuclei were observed in the FC of DSD− and DSD+ cases (Figure 2A–L). In addition to U1‐70K (Figure 2E,F) and U1A (Figure 3K,L) nuclear labeling, U1A and U1‐70K immunoreactivity was also observed in the cytoplasm, displaying a fine granular (Figure 3B,P) or tangle‐like appearance (Figure 3, Figure S4), mainly in layers III and V–VI in both DS groups. However, only stronger SRSF2 immunostaining was associated with nuclear speckles compared to the nucleoplasm (Figure 2 M–R). Robust hnRNPA2B1 labeling appeared within the cell nucleus and weakly in the cytoplasm in both demented and non‐demented individuals with DS (Figure 2S–X).

FIGURE 2.

Low‐power photomicrographs showing frontal cortex laminar distribution of nuclear U1‐70K (A, D), U1A (G–J), SRSF2 (M, P), and hnRNPA2B1 (S, V) immunoreactivity in DSD− (left side panels) and DSD+ (right side panels). High‐power images show similar nuclear immunostaining for U1‐70K (B, C, E, F), U1A (H, I, K, L), SRSF2 (N, O, Q, R), and hnRNPA2B1 (T, U, W, X) in the same cases. There were many more lightly stained U1‐70K‐ and U1A‐positive nuclei in layers III (H, K) and V–VI (I, L) in the DSD+ case. Scale bars: S = 200 µm applies to all low magnification panels, X = 25 µm applies to all high‐magnification panels. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia.

FIGURE 3.

High‐power photomicrographs showing mislocalized cytoplasmatic U1‐70K‐ and U1A‐ immunolabeled profiles with a granular (B, P) or tangle‐like (A, C–H, I–O) appearance in frontal cortex layers III and V–VI in DSD− and DSD+ cases. U1‐70K and U1A tangle‐like structures exhibit an annular (J), flame (A, D, L), or a skein of yarn (N, O) shape, whereas some (F, L) lacked nuclear U1A and U1‐70K immunostaining and others displayed pale nuclei (A, I, K, M) compared to those without a tangle‐like structure (B, P, H). Scale bar in P = 10 µm; applies to all panels. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia.

3.4. OD, counts, and area values of U1‐70K‐, U1A‐, SRSF2‐, and hnRNPA2B1‐labeled nuclei

3.4.1. U1‐70K

We observed intense nuclear U1‐70K immunostaining in layers III, V, and VI in all DSD− and DSD+ cases, whereas pale nuclear immunostaining was observed in 90% (20/22) of the DSD+ versus 50% (6/11) in DSD− cases. Quantitation revealed that the percentage of nuclei displaying intense U1‐70K immunostaining compared to the total number of U1‐70K‐positive profiles examined (i.e., number of nuclear intensely + weakly immunostained + cytoplasmic U1‐70K‐positive neurons) was significantly greater in DSD− than in DSD+ in layers III (Figure S5A; Mann–Whitney test, p = 0.0005) and V–VI (Figure S5A; Mann–Whitney test, p = 0.006), whereas the percent of weakly stained nuclei was significantly decreased in layers III (Figure S5B; Mann–Whitney test, p = 0.001) and V–VI (Figure S5B; Mann–Whitney test, p = 0.04) in DSD−. Moreover, a greater percentage of cytoplasmic U1‐70K‐positive cells was found in layers III (Mann–Whitney test, p = 0.0002) and V–VI (Mann–Whitney test, p = 0.01) in DSD+ compared to DSD− (Figure S5C).

The number of intensely reactive U1‐70K nuclei, nuclear area, and OD measurements were comparable in the cortical layers examined between demented and non‐demented cases with DS (Mann–Whitney test, p > 0.05). By contrast, the number of lightly U1‐70K‐positive nuclei in layer III were higher in DSD+ than in DSD− (Figure 4A; Mann–Whitney test, p = 0.02), whereas there were no differences in layers V–VI between groups (Figure 4A; Mann–Whitney test, p > 0.05).

FIGURE 4.

(A) Scatter plots with bars show a significant increase in the number of lightly labeled U1‐70K‐ (p = 0.02) and a reduction in intensely reactive U1A‐ (p = 0.01) (B) positive nuclei in layer III. (C) The number of light U1A‐positive nuclei in layers III (p = 0.007) and V–VI (p = 0.003) was increased in DSD+ compared to DSD−. (D) A significantly decreased number of SRSF2‐stained profiles was seen in layer III (p = 0.01) in DSD+. (E) Optical density (OD) values for hnRNPA2B1 nuclei and (F) nuclear area were significantly reduced in layers III (OD p = 0.001, area p = 0.0003) and V–VI (OD p = 0.004, area p = 0.04) in DSD+ compared to DSD−, whereas hnRNPA2B1 nuclear number (p = 0.01) was decreased in layer III only in DSD+ (G). A within‐group analysis revealed that intensely stained (B) U1A and (D) SRSF2 nuclear numbers as well (F) hnRNPA2B1 positive nuclear size, and (G) hnRNPA2B1 numbers were significantly greater in layer V–VI than III in DSD+, but not in DSD−. Mann–Whitney test and Wilcoxon signed‐rank test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.00001. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia; ir‐, immunoreactive.

A within‐group analysis revealed that the number of intensely stained U1‐70K nuclei (Figure S6A; Wilcoxon signed‐rank test: DSD− p = 0.02 and DSD+ p < 0.0001) and nuclear area (Figure S6B; Wilcoxon signed‐rank test: DSD− p = 0.001 and DSD+ p = 0.03) were greater in layers V–VI than layer III in both groups. Comparison of OD values between layers with intensely and lightly stained U1‐70K nuclei revealed significant greater measurements in layers V–VI than in layer III in DSD+ (Figure S6C,D; Wilcoxon signed‐rank test, intense p = 0.01 and light p = 0.04). The nuclear area of weakly labeled U1‐70K nuclei was significantly larger in layers V–VI than in layer III in DSD+ (Figure S6E; Wilcoxon signed‐rank test, p = 0.02).

OD (Figure S7A–D), area (Figure S7E–H), and counts (Figure S7I–L) of the intensely compared to weakly labeled U1‐70K nuclei were significantly greater in layer III (Wilcoxon signed‐rank test, DSD+: OD p < 0.0001, area p = 0.003 and counts p < 0.0001, and DSD−: OD p = 0.01, area p = 0.01, and counts p = 0.0005) and layers V–VI (Wilcoxon signed‐rank test, DSD+: OD p < 0.0001, area p = 0.0001, and counts p < 0.0001 and DSD−: OD p = 0.03, area p = 0.03 and counts p = 0.0005) in both groups.

3.4.2. U1A

Like U1‐70K immunostaining, strong and weak U1A‐positive nuclei were found in FC layers III and V–VI in demented and non‐demented DS cases (Figure 2G–L). Strongly stained U1A nuclei were observed in all DSD+ and DSD− cases, whereas pale nuclear U1A reactivity was detected mostly in DSD+ cases. Specifically, in DSD+, weakly stained nuclei were found in layers III in 80% and V–VI in 64% of cases, whereas only 16% of DSD− cases exhibited pale nuclear U1A immunostaining in layers III and V–VI. Analysis revealed no differences in the total number of U1A‐positive profiles (i.e., intense + pale‐stained nucleus + cytoplasmic staining) between groups in layers III and V–VI (Figure S8A; Mann–Whitney test, p > 0.05). However, the percentage of intensely U1A‐labeled nuclei was significantly greater in layers III (Figure S8B; Mann–Whitney test, p = 0.001) and V–VI (Figure S8B; Mann–Whitney test, p = 0.001) in DSD− than in DSD+, whereas the percentage of weak nuclear staining was significantly reduced in layers III (Figure S8C; Mann–Whitney test, p = 0.004) and V–VI (Figure S8C; Mann–Whitney test, p = 0.003) in DSD− compared to DSD+. Moreover, a greater percentage of cytoplasmic U1A‐positive cell numbers was found in layers III (Figure S8D; Mann–Whitney test, p = 0.005) and V–VI in DSD+ (Figure S8D; Mann–Whitney test, p = 0.04) compared to DSD−.

Quantitative analysis revealed a significantly greater number of intensely stained U1A nuclei in layer III, but not layers V–VI, in DSD− compared to DSD+ (Figure 4B; Mann–Whitney test, p = 0.01). By contrast, the number of nuclei displaying light U1A labeling was significantly reduced in layers III (Figure 4C; Mann–Whitney test, p = 0.007) and V–VI (Figure 4C; Mann–Whitney test, p = 0.003) in DSD− compared to DSD+. There were no differences in OD measurements and nuclear area in the intensely and lightly positive U1A nuclei in layers III and V–VI between groups (Table S5; Mann–Whitney test, p > 0.05). Comparison between layers in each clinical group revealed a significant increase in the number of intensely positive U1A nuclei (Figure 4B; Wilcoxon signed‐rank test, p < 0.0001) and OD values (Table S5; Wilcoxon signed‐rank test, p = 0.02) in layers V–VI than III in DSD+, but not in DSD−.

A within‐group analysis revealed that OD measurements (Figure S9A,B) and number (Figure S9C,D) of light U1A‐ir nuclei were significantly reduced compared to more intensely reactive U1A‐labeled nuclei in layers III and V–VI in DSD+ (Wilcoxon signed‐rank test, p < 0.0001). The area of lightly stained U1A compared to more intensely immunostained nuclei was smaller in layers III (Figure S9E; Wilcoxon signed‐rank test, p = 0.0003) and V–VI (Figure S9F; Wilcoxon signed‐rank test, p = 0.0003) in DSD+. Due to the limited number of non‐demented DS cases displaying weak U1A nuclear immunostaining versus strong U1A nuclei, a statistical evaluation was not performed.

3.4.3. SRSF2

FC nuclear SRSF2 immunostaining was observed in all cortical laminae in DS with and without dementia. The intensity of SRSF2 nuclear immunoreactivity was not different between groups. There were no statistically significant differences in SRSF2 nuclear OD and area values in layers III and V–VI between groups (Table S5). The number of stained SRSF2 nuclei was significantly decreased in layer III, but not in layers V–VI, in DSD+ compared to DSD− (Figure 4D; Mann–Whitney test, p = 0.01). Within‐group analysis found that the numbers of the SRSF2‐positive nuclei (Figure 4D; Wilcoxon signed‐rank test, p = 0.0007) and nuclear area (Table S5; Wilcoxon signed‐rank test, p = 0.0002) in layers V–VI were significantly increased compared to layer III in DSD+.

3.4.4. hnRNPA2B1

OD measures and area of hnRNPA2B1‐positive nuclei were significantly higher in layers III (Figure 4E,F; Mann–Whitney test: OD p = 0.001, area p = 0.0003) and V–VI (Figure 4E,F; Mann–Whitney test: OD p = 0.004, area p = 0.04) in DSD− than in DSD+. There was a significantly increased number of hnRNPA2B1‐labeled nuclei in layer III, but not in layers V–VI, in DSD− compared to DSD+ (Figure 4G; Mann–Whitney test, p = 0.01). A within‐group analysis revealed that the area of hnRNPA2B1‐positive nuclei was greater in layers V–VI than in layer III in DSD− (Figure 4F; Wilcoxon signed‐rank test, p = 0.02), whereas numbers of hnRNPA2B1‐positive nuclei in layers VI–VI were higher than in layer III only in DSD+ (Figure 4G; Wilcoxon signed‐rank test, p = 0.005). By contrast, the OD measurements of nuclear hnRNPA2B1 immunostaining were comparable between layers in both DSD− and DSD+ (Table S5).

3.5. Cytoplasmic U1‐70K‐ and U1A‐positive cell counts

Cytoplasmic mislocalized U1‐70K and U1A displayed a pale granular or tangle‐like appearance, predominantly in layers III and V–VI neurons in DSD+ (Figures 2, 3, and S4). Some lacked nuclear immunostaining for both U1s (Figure 3). Of interest, mislocalized U1‐70K and U1A immunostaining was not observed in NTs (Figure S4). Cytoplasmic U1‐70K immunostaining was observed in 16% of DSD−, whereas mislocalized U1‐70K immunostaining was seen in layer III in 100% compared to 72% in layers V–VI of DSD+ cases.

We quantified cells containing cytoplasmic U1‐70K and U1A, independent of morphology or nuclear labeling in paraffin‐embedded sections in both DSD− (n = 11) and DSD+ (n = 22). Statistical analysis revealed greater numbers of cells displaying mislocalized U1‐70K (Figure 5A; Mann–Whitney test, p < 0.0001) and U1A (Figure 5B; Mann–Whitney test p = 0.008) immunoreactivity in layer III in DSD+ compared to DSD−, whereas higher numbers of U1‐70K cells were seen in layers V–VI (Figure 5A; Mann–Whitney test, p = 0.009), but not for U1A (Figure 5B; Mann–Whitney test, p = 0.06) in DSD+. A within‐group analysis found that cytoplasmic U1‐70K‐positive cell counts were similar between layers in DS with and without dementia (Figure 5A; Wilcoxon signed‐rank test, p > 0.05), whereas the number of cells exhibiting cytoplasmic U1A was significantly higher in layer III than in layers V–VI in DSD+ (Figure 5B; Wilcoxon signed‐rank test, p = 0.005), but not in DSD−.

FIGURE 5.

Scatter plots with bars showing significantly higher numbers of neurons containing cytoplasmic mislocalized U1‐70K (A; p < 0.000) and U1A (B; p = 0.008) in frontal cortex layer III in DSD+ compared to DSD−, as well higher numbers of U1‐70K (A; p = 0.009), but not U1A positive cells in layers V–VI in DSD+. A within‐group analysis found that (B) cytoplasmic U1A positive cell numbers were significantly increased in layers V–VI than in layer III in DSD+. The number of U1‐70K positive was significantly higher than U1A tangle‐like structures in layers III (D; p = 0.01) and V–VI (D; p = 0.03) DSD+, but not in (C) DSD−. Mann–Whitney test and Wilcoxon signed‐rank test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.00001. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia; ‐ir, immunoreactive.

Counts of tangle‐like structures positive for U1A and U1‐70K in layers III and V–VI, which were performed on floating sections mounted on slides (see Table S1), were similar between demented (n = 9) and non‐demented (n = 5) people with DS. However, the number of U1‐70K‐positive tangle‐like structures was significantly higher than U1A in layer III in DSD+ (Figure 5D; Mann–Whitney test; p = 0.01) and layers V–VI (Figure 5D; Mann–Whitney test, p = 0.03), but not in DSD− (Figure 5C).

3.6. Phosphorylated AT8 tau, truncated TauC3, and 3Rtau and 4Rtau isoform profiles in DS with and without dementia

AT8‐ and TauC3‐positive NFTs, NTs, and clusters of swollen dystrophic neurites (tau neuritic plaques) were observed in layers III and V–VI in both DSD− and DSD+ (Figures S10A–D, S11A–D). Comparatively, many more AT8‐positive NTs compared to TauC3 were observed in DSD+ (Figures S10C,D and S11C,D). Quantitively, AT8‐ir NFTs and NT counts were increased in layers III (Figure S10E,F; Mann–Whitney test: NFTs p = 0.0002 and NT p = 0.004) and V–VI (Figure S10E,F; Mann–Whitney test: NFTs p < 0.0001 and NT p = 0.003) in DSD+ compared to DSD−, but no statistically significant differences in number of AT8‐positive plaques were observed in the cortical layers between DS groups (Figure S10G; Mann–Whitney test, p > 0.05). Within‐group analysis demonstrated that the number of AT8‐stained NFTs in layer III was significantly higher compared to layers V–VI (Figure S10E; Wilcoxon signed‐rank test, p = 0.0001). The number of AT8‐positive plaques was significantly greater in layer III than in layers V–VI in DS with (Figure S10G; Wilcoxon signed‐rank test p = 0.01) and without dementia (Figure S10G; Wilcoxon signed‐rank test, p = 0.03).

The number of the TauC3‐positive NFTs and NTs was significantly increased in layers III (Figure S11A–F; Mann–Whitney test: NFTs p = 0.004 and NT p = 0.02) and layers V–VI (Figure S11A–F; Mann–Whitney test: NFTs p = 0.003 and NT p = 0.01) in DSD+ compared to DSD−. The FC displayed very few TauC3‐positive plaques, and there were no differences between groups (Figure S11A–D,G). Although TauC3‐positive NFTs were greater in layers V–VI than layer III in DSD+ (Figure S11E, Wilcoxon signed‐rank test, p < 0.0001), this was not observed for the TauC3 plaques and NTs in either group (Figure S11F,G; Wilcoxon signed‐rank test, p > 0.05).

The number of 3Rtau‐positive NFTs in layers III (Figure S12A–E; Mann–Whitney test, p < 0.0001) and V–VI (Figure S12A–E; Mann–Whitney test, p < 0.0001) was significantly higher in DSD+ compared to DSD−, whereas counts of 3Rtau‐immunostained non−tangle‐bearing cells were greater only in layer III (Figure S12A–D,F; Mann–Whitney test, p = 0.02) in DSD−. The numbers of 3Rtau‐ir NTs in layers III (Figure S12A–D,G; Mann–Whitney test, p = 0.0009) and V–VI (Figure S12A–D,G; Mann–Whitney test, p = 0.0006) were significantly increased in both layers in DSD+ compared to DSD−. However, 3Rtau‐positive plaque numbers in layers V–VI, but not layer III, were significantly higher in DSD+ compared to DSD− (Figure S12A–D,H; Mann–Whitney test, p = 0.04). Within‐group examinations found that the numbers of 3Rtau‐positive NFTs were significantly increased in layers III compared to layers V–VI in DSD+ (Figure S12E; Wilcoxon signed‐rank test, p < 0.0001). Likewise, the number of 3Rtau‐positive non−tangle‐bearing cells was higher in layer III compared to V–VI in DSD− (Figure S12F; Wilcoxon signed‐rank test, p = 0.04), but decreased in DSD+ (Figure S12F; Wilcoxon signed‐rank test, p = 0.003). No differences in the number of 3Rtau‐positive NTs and plaques were observed within the layers examined in each group (Figure S12G,H).

Although there were no differences in 4Rtau‐positive NFTs and NT counts in layers III and V–VI between groups (Figure S13A‐E,G; Mann–Whitney test, p > 0.05), only 4Rtau‐positive non−tangle‐laden cell numbers in layers V–VI were greater in DSD+ than in DSD− (Figure S13A–D, F; Mann–Whitney test, p = 0.02). Comparatively, the numbers of 4Rtau‐positive NFTs (Figure S13E; Wilcoxon signed‐rank test, p = 0.003), non−tangle‐bearing cells (Figure S13F; Wilcoxon signed‐rank test, p = 0.006), and NTs (Figure S13G; Wilcoxon signed‐rank test, p = 0.003) were significantly increased in layers V–VI compared to layer III in DSD+, but not in DSD−.

Correlational analysis revealed positive associations between the number of AT8‐, TauC3‐, and 3Rtau‐positive NTs and NFT counts and Braak scores in layers III and V–VI across groups (Table S6; Spearman test, r > 0.59, p ≤ 0.001).

3.7. Frontal cortex APP/Aβ plaque and cerebral amyloid angiopathy (CAA) in DS with and without dementia

Qualitative microscopic examination revealed that most APP/Aβ‐labeled plaques displayed a dense‐core morphology in layers I to III in DSD− (Figure S14A,B) compared to laminae I to VI in the FC in DSD+ (Figure S14C,D). Diffuse plaques were observed mainly in layers V and VI in DSD− (Figure S14A,C). APP/Aβ‐reactive plaques were more abundant in layer III than in layers V and VI in all DSD+ cases (Figure S14A–C,F–H). Moreover, DSD+ displayed gray and white matter APP/Aβ‐labeled plaques (Figure S14D,E,I) and leptomeningeal arteries or CAA in 90% of DSD+ (20/22) cases (Figure S14J–L, Table S1). Qualitatively, the number of APP/Aβ‐reactive plaques in layers III was significantly higher in DSD+ compared to DSD− (Figure S14M; Mann–Whitney test, p = 0.004), whereas in layers V–VI, APP/Aβ plaque counts were similar between groups (Figure S14M; Mann–Whitney test, p > 0.05). Conversely, there were no differences in the total number of APP/Aβ‐positive plaques when counts from layers III and V–VI were combined in DSD+ compared to DSD− (Figure S14N; Mann–Whitney test, p > 0.05). A within‐group analysis found that the number of APP/Aβ‐positive plaques in layer III was significantly greater than in layers V–VI in DS with (Figure S14M; Wilcoxon signed‐rank test, p < 0.0001) and without dementia (Figure S14M; Wilcoxon signed‐rank test, p = 0.01). Moreover, plaque load in layer III, but not in layers V–VI, correlated with Braak stage across groups (Table S6; Spearman test, r = 0.59, p < 0.01).

3.8. Frontal cortex extranuclear U1s and tau pathology in DS with and without dementia

To confirm whether U1‐70K‐ and U1A NFT‐like structures co‐occurred with the early phospho‐AT8 and late caspase truncation‐mediated TauC3 NFT pathology, double immunofluorescence was performed using mounted free‐floating sections. Double immunofluorescence showed that extranuclear U1‐70K (Figure 6) and U1A (Figure S15) tangle‐like structures that co‐labeled with AT8 and TauC3, displayed a morphological appearance of a skein of yarns (Figure 6C,G), flame (Figures 6E,I and S15K,O), or annular shape (Figures 6A,M and S15C,D). Quantitation revealed a greater number of U1‐70K and U1A tangle‐like structures that co‐stained with AT8 in FC layers III (Figure 7A,B; Mann–Whitney test: U1‐70K p = 0.01, U1A p = 0.01) and V–VI (Figure 7A,B; Mann–Whitney test: U1‐70K p = 0.003, U1A p = 0.02) in DSD+ than in DSD−. By contrast, FC neurons co‐labeled for U1‐70K and TauC3 exhibited a trend toward an increase in DSD+ compared to DSD−, which was not statistically significant (Figure 7C; Mann–Whitney test, p > 0.05), whereas counts of dual‐labeled U1A‐ and TauC3‐positive profiles were higher in layers V–VI in DSD+ than in DSD− cases (Figure 7D; Mann–Whitney test, p = 0.02).

FIGURE 6.

Immunofluorescent single‐labeled images showing nuclear and/or mislocalized U1‐70K‐ (red) positive tangle‐like structures (A, C, E, G, I, K, M, O). Dual fluorescent labeling revealed neurofibrillary tangles (NFTs) that were positive for U1‐70K (red), AT8 (green), and TauC3 (green), and merged images appeared orange; these are shown (B, D, F, H, J, L, N, P) in individuals with DS with and without dementia. Cells showing extranuclear U1‐70K immunostaining displayed a skein of yarn (C, G), flame (E, I), or annular (A, M, O) tangle‐like shape within AT8‐positive NFTs. NFTs were found that were single labeled for U1‐70K (B, D) and TauC3 (K, L). Note also the absence of U1‐70K‐positive NTs, whereas AT8‐ (B, D, F, H) and TauC3‐ (J, L, N) positive NTs were observed in both groups. DAPI nuclear counterstain appears blue and/or pink in merged images. 10 µm scale bar in L applies to A–K and in P applies to M–O panels. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia.

FIGURE 7.

Scatter plots with bars showing a greater number of U1‐70K (A) and U1A (B) tangle‐like structures co‐stained with AT8 in frontal cortex layers III (U1‐70K p = 0.01, U1A p = 0.01) and V–VI (U1‐70K p = 0.003, U1A p = 0.02) in DSD+ compared to DSD−, whereas the numbers of neurons dual labeled for U1A‐ and TauC3‐positive profiles were higher in layers V–VI (D; p = 0.02) in DSD+ than DSD−. Cells co‐labeled for U1‐70K and TauC3 exhibited a trend toward an increase in DSD+ compared to DSD− (C). Mann–Whitney test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.00001. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia; NFT, neurofibrillary tangle.

3.9. Associations between splicing proteins, tau, and plaque pathology

Correlational analyses demonstrated numerous associations between U1‐70K and U1A morphometry and tau pathology. In layer III, weak U1‐70K‐positive nuclei number correlated with AT8, 3Rtau NT (Figure S16A,B; Spearman test: AT8 r = 0.64, p = 0.00004, 3Rtau r = 0.61, p = 0.0001), and NFTs (Figure S16C,D; Spearman test: AT8 r = 0.50, p = 0.003, 3Rtau 0.54, p = 0.001), and cytoplasmic U1‐70K neurons correlated with AT8, TauC3, and 3Rtau NFTs (Spearman test: AT8 r = 0.79, p = 0.0000002, TauC3 r = 0.68, p = 0.00002, 3Rtau r = 0.81, p = 0.0000002) and negatively with numbers of non−tangle‐positive 3Rtau neurons (Spearman test, r = −061, 0.0001). In layers V–VI, weak U1‐70K‐positive nuclei number was associated with AT8 NTs (Figure S17A; Spearman test, r = 0.63, p = 0.00008), AT8 NFTs (Figure S17B; Spearman test, r = 0.51, p = 0.002), TauC3 NFTs (Spearman test, r = 0.58, p = 0.0008), 3Rtau NTs (Figure S17C; Spearman test, r = 0.63, p = 0.00008), NFTs (Figure S17D; Spearman test, r = 0.57, p = 0.0005), and 4Rtau NTs (Spearman test, r = 0.60, p = 0.0001) and NFTs (Spearman test, r = 0.54, p = 0.001). Counts of cytoplasmic U1‐70K neurons in layer III correlated strongly with AT8 NFTs (Figure 8A; Spearman test, r = 0.79, p = 0.0000002) and 3Rtau NFTs (Figure 8B; Spearman test, r = 0.81, p = 0.0000002). Number of U1‐70K neurons in layers V–VI were associated with AT8 NTs (Figure 8C; Spearman test, r = 0.68, p = 0.000005) and AT8 NFTs (Spearman test, r = 0.61, p = 0.0001), 3Rtau NTs (Spearman test r = 0.71, p = 0.0000002), and 3Rtau NFTs (Figure 8D; Spearman test, r = 0.70, p = 0.0000002), and negatively with numbers of non−tangle‐positive 3Rtau neurons (Spearman test, r = −0.59, p = 0.0002). Intensely U1A‐immunoreactive nuclei correlated positively with 3Rtau non−tangle‐positive cells (Spearman test, r = 0.70, p = 0.0000002) and negatively with 3Rtau NFTs (Spearman test r = −0.57, p = 0.0005) in layer III. Lightly immunostained U1A nuclei correlated with AT8 NT (Figure S18A; Spearman test, r = 0.57, p = 0.0005) and AT8 NFT (Figure S18B; Spearman test, r = 0.61, p = 0.0001) and 3Rtau NT (Figure S18C; Spearman test, r = 0.57, p = 0.0005) and 3Rtau NFT counts (Figure S18D; Spearman test, r = 0.57, p = 0.0004) in layer III. Cytoplasmic U1A‐positive neurons were associated with AT8 NT (Figure 8E; Spearman test r = 0.68, p = 0.000004), 3Rtau NT (Figure 8F; Spearman test r = 0.78, p = 0.0000002), and 3Rtau NFT counts (Spearman test, r = 0.62, p = 0.0001) in layer III. In layers V–VI, the numbers of lightly immunostained U1A nuclei correlated with AT8 NTs (Figure S19A; Spearman test, r = 0.62, p = 0.0001), 3Rtau NTs (Figure 8G; Spearman test, r = 0.68, p = 0.000006), 3Rtau NFTs (Spearman test, r = 0.64, p = 0.00004), and 4Rtau NFTs (Spearman test, r = 0.64, p = 0.00004) across groups. Light U1A‐positive nuclei were associated with AT8 NFT counts in layers V–VI (Figure S19B; Spearman test r = 0.52, p = 0.002). Counts of cytoplasmic U1A neurons in layers V–VI were also associated with AT8 NTs (Figure S19C; Spearman test, r = 0.61, p = 0.0001), 3Rtau NFTs (Figure 8H; Spearman test, r = 0.69, p = 0.000001) and 4Rtau NFTs (Spearman test, cytoplasm r = 0.68, p = 0.000003). Moreover, the number of hnRNPA2B1‐stained nuclei in layer III was positively associated with 3Rtau non−tangle‐positive cells (Spearman test, r = 0.57, p = 0.0005), and the area of the hnRNPA2B1 nucleus was negatively associated with AT8 NTFs (Spearman test, r = −0.55, p = 0.0008) in layer III. By contrast, none of the SRSF2 data revealed associations with tau pathology. Finally, there were no significant correlations of hnRNPA2B1 and SRSF2 OD, size, and nuclear numbers with plaque numbers. The lightly positive U1A nuclei (Spearman test, r = 0.68, p = 0.000006) and nuclear area (Spearman test, r = 0.62, p = 0.000009) correlated with plaque load in layer III, but not layers V–VI, across the groups.

FIGURE 8.

Linear regression graphs illustrate strong positive associations between numbers of frontal cortex cytoplasmic U1‐70K‐positive cells with AT8 NFTs (A; r = 0.79, p = 0.0000002) and 3Rtau NFTs (B; r = 0.81, p = 0.0000002) in layer III, as well as with AT8 NTs (C; r = 0.68, p = 0.000005) and 3Rtau NFTs (D; r = 0.70, p = 0.0000002) in layers V–VI across groups. Moreover, cytoplasmic U1A‐positive cell numbers correlated with 3Rtau (E; r = 0.78, p = 0.0000002) and AT8 (F; r = 0.68, p = 0.000004) NTs in layer III, and with 3Rtau NFTs (H; r = 0.69, p = 0.000001) in layers V–VI. Lightly stained U1A nuclei numbers were associated with 3Rtau NTs (G; r = 0.68, p = 0.000006) in layers V–VI across DSD− and DSD+ cases. Spearman test, FDR < 0.001. DSD+, Down syndrome with dementia; DSD‐, Down syndrome without dementia; ir‐, immunoreactive; NFT, neurofibrillary tangle; NT, neuropil thread.

4. DISCUSSION

4.1. Spliceosome proteins in the frontal cortex in DS

We found differences in the CTD domain of phosphorylated RNA polymerase II (aka RNAP II) between DS groups. Immunoblotting revealed that phosphorylated pS5,2‐RNA pol II, but not pS5‐RNA pol II, levels were significantly upregulated in DSD+ compared to DSD−, similar to the AD temporal cortex. ⁵⁶ Although the ratio of pS5,2‐RNA pol II/pS5‐RNA pol II was unchanged across groups, pS5‐RNA poll II levels were significantly higher than pS5,2‐RNA pol II in both DS groups. The increase of pS5,2‐RNA poll II suggests aberrant RNA transcription in response to neuronal dysfunction in DS. ⁵⁷ We observed positive correlations between pS5,2‐RNA pol II, Braak scores, and 1N4Rtau, suggesting an association with tau pathogenesis. U1‐70K downregulation negatively correlated with Braak staging, suggesting a link between U1‐70K dysregulation and tau pathology in DS. The presence of U1‐70K‐ and U1A‐positive tangle‐like structures supports the observation of insoluble aggregates of both splicing proteins in these DS groups as well as in the FC of familial and sporadic AD. ¹⁹

Immunohistochemical studies revealed intensely and weakly stained U1‐70K and U1A nuclei, granular cytoplasmic, and tangle‐like structures in both DS cohorts in FC layers III and V–VI. There were no differences in OD, size, or counts of intensely stained U1‐70K nuclei between layers III and V–VI across groups, whereas the number of the lightly stained U1 nuclei was significantly higher in layer III in DSD+, suggesting greater neuronal vulnerability in layer III. Cytoplasmic U1 immunoreactivity appeared as granules or tangle‐like structures in each layer in 70% of DSD+ compared to 16% of DSD− cases. Counts revealed greater numbers of neurons displaying mislocalized U1‐70K in layers III and V–VI in DSD+, compared to U1A upregulation only in layer III in DSD+ cases. U1‐positive tangles were reported in cortical and hippocampal neurons in DS cases that lacked a clinical diagnosis. ¹⁹ Although there were no group differences in number of U1 tangle‐like structures, U1‐70K tangle‐like profiles were increased compared to U1A in layers III and V–VI in DSD+. Total counts revealed a significant increase in the percentage of weak nuclear and cytoplasmic U1‐70K‐positive profiles in layers III and V–VI in DSD+. These observations suggest that mislocalized U1‐70K contributes to alterations in RNA splicing to a greater extent in DS with dementia.

Dual immunofluorescence revealed significantly greater numbers of AT8, but not TauC3‐positive NFTs displaying U1 in layers III and V–VI in DSD+ compared to DSD−. Cytoplasmic U1‐70K‐ and U1A‐positive layer III neurons were positively associated with Braak stage and AT8 and 3Rtau‐labeled profiles across groups. Bai and colleagues ¹⁷ reported the spread of cytoplasmic U1‐70K pathology followed Braak tau staging in the medial temporal lobe in AD, suggesting a mechanistic interaction between these two pathologies that contributes to transcriptome disruption and neurodegeneration in DS and AD. ²¹ Whether tau drives the mislocalization of U1s or other mechanisms that lead to tangle formation ¹⁷ , ⁵⁸ remains unknown.

Serine/arginine‐rich splicing factor 2 (or SRSF2) ⁵⁹ is linked to cellular apoptosis, ⁶⁰ cholinergic signaling, ⁹ , ⁶¹ acetylation, ⁶² and amyloid protein formation. ⁶³ SRSF2 acts on a SRSF2‐like enhancer, promoting the inclusion of exon 10 ⁶⁴ and generation of 4Rtau isoforms linking SRSF2 to tau pathogenesis. SRSF2 levels did not correlate with other splicing proteins, tau isoforms, nuclear OD, area, counts, Braak staging, patient demographics, or tau and plaque pathology across groups. Increased 4Rtau expression, plus an elevation of exon 10, enhances SRSF2 activity, tau expression and aggregation. ⁶⁴ , ⁶⁵ SRSF2 cytoplasmic mislocalization was not seen in either group, as reported in vivo. ⁶⁴ There were no significant differences in nuclear SRSF2 measures in layers III and V–VI between groups. There was a significant decrease in the number of SRSF2‐labeled nuclei in layer III in DSD+ compared to DSD−, indicative of a greater vulnerability of cortical neurons that innervate thalamic relay nuclei in DSD+. Our statistical analysis failed to show differences in CLK1 protein levels between DS groups or a correlation with Braak stages despite a CLK1 association with alternative splicing of exon 10. ⁶⁶ Correlations between nuclear SRSF2 OD values and 4Rtau pathology in layers V–VI suggest a relationship between SRSF2 and generation of tau isoforms in cortico‐cortical and corticothalamic projection neurons in DS.

hnRNPs mediate tau pathology in AD. ⁶⁷ , ⁶⁸ hnRNPA2B1 levels in the FC were upregulated, but did not correlate with Braak stage across groups. Conversely, morphometric analyses revealed a significant reduction in nuclear hnRNPA2B1 OD and area in layers III and V–VI in DSD+ than in DSD−, but significantly fewer hnRNPA2B1‐positive nuclei were seen in layer III compared to V–VI in DSD+. This discrepancy between hnRNPA2B1 protein levels and nuclear morphometrics is likely due to different methodologies, where the former includes non‐neuronal cell types. A reduction of nuclear hnRNPA2B1 immunostaining occurs in the AD entorhinal cortex ⁹ and hippocampus in prodromal and frank AD, ⁶⁹ suggesting a widespread downregulation throughout the brain. However, mislocalized hnRNPA2B1 did not co‐localize with phosphorylated AT8 aggregates in the FC and hippocampus in various tauopathies. ⁶⁹ hnRNPA2B1 dysregulation may be detrimental to neuron survival by binding with N⁶‐methyladenosine, a nuclear epigenetic modifier of RNA, ¹² , ⁷⁰ , ⁷¹ , ⁷² and oligomeric tau. ¹² Whether N⁶‐methyladenosine/oligomeric tau/hnRNPA2B1 complexes are increased in the FC requires investigation.

4.2. U1 splicing proteins, tau, and Aβ pathology in DS

We found greater numbers of mislocalized U1 neurons in DSD+. Correlations demonstrated associations between mislocalized U1s and NFTs and NTs containing AT8 and 3Rtau across groups, like other tauopathies. ¹⁷ , ¹⁹ Quantitative immunofluorescence revealed significantly greater numbers of AT8, but not TauC3 NFTs that displayed U1 tangle‐like structures in layers III and V–VI in DSD+ than in DSD−. Of note, many AT8 and TauC3 NFTs lack mislocalized U1 tangle‐like structures and vice versa in both groups, suggesting that U1 mislocalization is not a necessary precondition for NFT formation. Because AT8 and TauC3 epitopes are associated with more advanced tau pathology and severe cognitive decline in AD and DS, ⁵⁵ , ⁷³ it is likely that the co‐occurrence of extranuclear U1s and NFTs synergistically affect the acceleration of mRNA metabolism disturbances, axonal transport, synaptic deficits, and ultimately neuronal loss in DS and AD.

DS is also characterized by extracellular accumulations of Aβ plaques, generated by the triplication of the APP gene. ⁷⁴ Insoluble U1A and U1‐70K correlate with Aβ levels and abnormal APP transcription in AD, ¹⁷ alternate splicing plays a role APP processing, ⁷⁵ and SRSF2 and hnRNPA2B1 maintain a balance between APP isoforms. ⁶³ Although there was no correlation between SRSF2 and hnRNPA2B1 morphometric data and amyloid plaque pathology, densities of APP/Aβ‐containing plaques correlated with the number of lightly positive U1A nuclei and area in layer III, but not layers V–VI, across groups. Of interest, layer III exhibited a higher number of dense cored/neuritic plaques in DSD+ cases, ⁵⁵ which correlate with AD severity. Knockdown of U1‐70K in HEK293 cells increases endogenous APP and Aβ40 ¹⁷ and alters the balance of APP‐spliced isoforms, ¹⁷ indicative of a mechanistic interaction between the spliceosome and APP processing. Whether splicing protein dysfunction contributes to the formation of amyloid pathology or vice versa in DS requires further investigation.

Present findings indicate that U1‐70K, U1A, hnRNPA2B1, and SRSF2 display complex localization and morphometric patterns in the FC of DSD+ and DSD− cases. Nuclear U1A and U1‐70K are mislocalized to a greater extent in DSD+ and co‐occur with AT8 and TauC3 aggregates, confirming that U1s form a new type of cytoplasmic tangle‐like structure that occurs more frequently in FC neurons in trisomy cases ¹⁷ , ¹⁹ with a clinical diagnosis of dementia like that reported in sporadic and familial AD, but not in other tauopathies. ¹⁹ The functional consequences of mislocalized U1s leading to the formation of a tangle‐like structure in DS remain unclear. Because U1 plays a role in the suppression of polyadenylation (polyA), nuclear cytoplasmic transport of U1s, and formation of tangle‐like fibrils, dysfunction could affect its ability to protect against mRNA degradation. ¹⁷ Trisomy alters polyA, ⁷⁶ leading to increased or decreased mRNA stability, which in turn can affect protein production. Disruption of RNA splicing function can result in alterations in APP expression. ¹⁷ The differential effect of trisomy on the mislocalization and formation of U1 tangle‐like elements in those individuals with dementia suggests that triplication of chromosome 21 does not uniformly affect RNA splicing machinery in people with DS. By contrast, we found that other RNA splicing factors, including hnRNPA2B1, pS5,2‐RNA pol II, and pS5‐RNA pol II, did not demonstrate tangle‐like structures in DS like those reported in AD. ⁵⁶ , ⁶⁹ Therefore, dysfunction of U1 proteins provides insight into the molecular mechanisms beyond the standard AD‐related lesions that contribute to the dementia associated with DS. Our finding suggests that mislocalization of U1 nuclear proteins not only affects tau pathology, but also other cellular mechanisms related to mRNA stability and cytoplasmic transport, which may be more prone to dysfunction in trisomy with dementia. Therefore, understanding the contribution of U1 pathology to RNA transcription and AD pathogenesis in DS may be critical to drug development for dementia in trisomy 21. Future research aimed at defining the genetic phenotype that drives U1 dysfunction in demented individuals with DS could help the development of novel therapies and biomarkers.

AUTHOR CONTRIBUTIONS

Conceptualization: Elliott J. Mufson and Sylvia E. Perez. Methodology: Muhammad Nadeem, Jennifer C. Miguel, and Sylvia E. Perez. Investigation: Muhammad Nadeem, Jennifer C. Miguel, Sylvia E. Perez, and Michael Malek‐Ahmadi. Visualization: Sylvia E. Perez. Funding acquisition: Elliott J. Mufson, Sylvia E. Perez, and Chadwick M. Hales. Supervision: Elliott J. Mufson and Sylvia E. Perez. Writing‐original draft: Elliott J. Mufson and Sylvia E. Perez. Writing–review and editing: Elliott J. Mufson and Sylvia E. Perez.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Any author disclosures are available in the supporting information.

ETHICS STATEMENT

Tissue sample use was approved by a family member and the ethical committees of each brain repository in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Supporting information

Supporting Information

Supporting Information

Perez SE, Miguel JC, Nadeem M, Malek‐Ahmadi M, Hales CM, Mufson EJ. U1‐70K and U1A and tau pathogenesis in demented and non‐demented individuals with Down syndrome. Alzheimer's Dement. 2025;21:e70474. 10.1002/alz.70474

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available upon request from the corresponding authors.