Maternal choline supplementation rescues early endosome pathology in basal forebrain cholinergic neurons in the Ts65Dn mouse model of Down syndrome and Alzheimer’s disease

Abstract

Individuals with DS develop Alzheimer’s disease (AD) neuropathology, including endosomal-lysosomal system abnormalities and degeneration of basal forebrain cholinergic neurons (BFCNs). We investigated whether maternal choline supplementation (MCS) affects early endosome pathology within BFCNs using the Ts65Dn mouse model of DS/AD. Ts65Dn and disomic (2N) offspring from dams administered MCS were analyzed for endosomal pathology at 3-4 months or 10-12 months. Morphometric analysis of early endosome phenotype was performed on individual BFCNs using Imaris. The effects of MCS on the endosomal interactome were interrogated by relative co-expression (RCE) analysis. MCS effectively reduced age- and genotype-associated increases in early endosome number in Ts65Dn and 2N offspring, and prevented increases in early endosome size in Ts65Dn. RCE revealed a loss of interactome cooperativity among endosome genes in Ts65Dn that was restored by MCS. These findings demonstrate MCS rescues early endosome pathology, a driver of septohippocampal circuit dysfunction. The genotype-independent benefits of MCS on endosomal phenotype indicate translational applicability as an early-life therapy for DS as well as other neurodevelopmental/neurodegenerative disorders involving endosomal pathology.

1. Introduction

Down syndrome (DS) is the most prevalent non-lethal chromosomal disorder in humans, with an estimated population of ~5-6 million worldwide (Hartley et al., 2015). Caused by the triplication of human chromosome 21 (HSA21), individuals with DS typically present with a variety of systemic conditions and neurological deficits including impairments in language acquisition, attention, and hippocampal-dependent learning and memory (Lott and Dierssen, 2010; Millan Sanchez et al., 2012). Additionally, individuals with DS often display a premature aging phenotype, with the majority developing Alzheimer’s disease (AD) neuropathology by 30-40 years of age (Franceschi et al., 2019; Startin et al., 2019). Gene dosage imbalances in triplicated regions leading to global dysfunction of gene expression networks contribute to the neurodegenerative DS phenotype (Antonarakis et al., 2001; Martínez-Cué and Rueda, 2020; Reeves et al., 2001). DS/AD sequelae include, but are not limited to, the development of amyloid-beta peptide (Aβ)-containing senile plaques and tau bearing neurofibrillary tangles (NFTs), degeneration of basal forebrain cholinergic neurons (BFCNs), and endosomal-lysosomal system dysregulation (Cataldo et al., 2000; Hartley et al., 2015; Wisniewski et al., 1985).

Morphological abnormalities in early endosomes are an early pathological marker in AD and DS. Dysfunctional vesicles, often immunoreactive for early endocytic markers rab5 and EEA1, appear enlarged and accumulate in large quantities within the cell soma (Cataldo et al., 1997; 2003; 2008). These irregularities present decades prior to the deposition of amyloid plaques and NFTs in AD (Cataldo et al., 1996; 1997), and are observable in individuals with DS in utero (Cataldo et al., 2000). Expression profiling of vulnerable brain regions in AD patients demonstrates upregulation of positive mediators of endocytosis (Ginsberg et al., 2010a; 2010b; 2011). Additionally, pathologic phenotypes have been induced in vitro (Bucci et al., 1992; Grbovic et al., 2003; Laifenfeld et al., 2007; Rodriguez et al., 2017) and in vivo (Pensalfini et al., 2020; Xu et al., 2016) via the overexpression of Rab family genes, strengthening the hypothesis that endocytic dysfunction is a driver in neurodegenerative events.

Ts65Dn mice are trisomic for portions of mouse chromosome 16 (MMU16) and 17 (MMU17) that are orthologous to the long arm of HSA21 (Gardiner et al., 2003; Sturgeon and Gardiner, 2011). These mice survive until adulthood and recapitulate key aspects of human DS neuropathology including early endosomal abnormalities (Cataldo et al., 2003), altered amyloid-beta precursor protein (APP) expression (Choi et al., 2009; Seo and Isacson, 2005), and degeneration of BFCNs (Hamlett et al., 2016). Ts65Dn mice are born with intact BFCNs that begin to atrophy around 6 months of age (MO) (Granholm et al., 2000; Holtzman et al., 1996), in conjunction with deficiencies in retrograde transport and diminished neurotrophic signaling (Cooper et al., 2001; Hunter et al., 2003; Salehi et al., 2006; Xu et al., 2016). Enlarged early endosomes are detectable in a majority of BFCNs of the medial septal nucleus/vertical limb of the diagonal band (MSN/VDB) and nucleus basalis of Meynert/substantia innominata (NBM/SI) by 6 MO (Cataldo et al., 2003). Early endosome pathology increases throughout the septohippocampal circuit of trisomic mice with age, appearing in increasing numbers in hippocampal pyramidal cells by 12-18 MO (Cataldo et al., 2003; Jiang et al., 2016; Salehi et al., 2006).

One therapeutic under investigation for the treatment of DS/AD neuropathology is early choline delivery during the perinatal period (Strupp et al., 2016). Choline is an essential nutrient vital for fetal neurodevelopment and nervous system function (Blusztajn, 1998; IOM, 1998). When consumed via the maternal diet, choline passes through the placenta and breast milk to the developing infant (Caudill et al., 2020). Choline is used in the synthesis of acetylcholine (Abreu-Villaça et al., 2011; Lauder and Schambra, 1999) and is an important constituent of phospholipids, playing a role in lipid metabolism and membrane biogenesis (Klein, 2000; Lauwers et al., 2016). Choline is also a primary dietary source of methyl groups (Anderson et al., 2012; Zeisel, 2017). Through methyl group donation, choline derivatives influence gene expression and modify the fetal epigenome by altering DNA methylation patterns (Blusztajn and Mellott, 2012; Jiang et al., 2014; Pauwels et al., 2017).

Ts65Dn offspring that received perinatal maternal choline supplementation (MCS) perform better in spatial learning and memory, emotional regulation, and attentional focus tasks as assessed at old age timepoints (e.g., 12-17 MO) (Ash et al., 2014; Kelley et al., 2016; Moon et al., 2010; Powers et al., 2017; Velazquez et al., 2013). Cognitive improvements were accompanied by increases in the number and density of cholinotrophic neurons in the basal forebrain (Ash et al., 2014; Gautier et al., 2023; Kelley et al., 2014). MCS also normalized the expression of upregulated trisomic genes in both the hippocampus and basal forebrain (Alldred et al., 2018, 2019; 2023; Kelley et al., 2019). However, cellular mechanism(s) whereby MCS attenuates the neurodegenerative phenotype of trisomic BFCNs remains unknown.

In the present study we hypothesize MCS ameliorates endosomal pathology that contributes to trisomic septohippocampal circuit dysfunction. We assessed early endosome pathology within individual MSN/VDB BFCNs in Ts65Dn and disomic (2N) offspring via morphometric analysis of 3D reconstructed z-stacks (Gautier and Ginsberg, 2021). Additionally, we interrogated multiple early endosome genes using single population gene expression analysis to determine if MCS mitigates endolysosomal interactome defects.

2. Methods

2.1. Mice

Animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the Nathan Kline Institute and NYU Grossman School of Medicine, and were in accordance with NIH (Bethesda, MD) guidelines. Female Ts65Dn (strain #005252) and male C57Bl/6JEiJ x C3Sn.BliA-Pde6b⁺/DnJ F1 (strain #003647) were purchased from Jackson Laboratories (Bar Harbor, ME) and mated at the Nathan Kline Institute. Standard cages contained paper bedding and a variety of objects for enrichment (e.g., plastic igloo, t-tube, and cotton square nestlets). Mice were maintained on a 12-hour light/dark cycle under temperature- and humidity-controlled conditions.

2.2. Maternal choline supplementation (MCS)

Dams assigned to the choline-supplemented diet were given rodent chow supplemented with 5.0 g/kg of choline chloride (AIN-76A; Dyets, Bethlehem, PA), while dams assigned to the choline-normal control diet received chow containing 1.1 g/kg choline chloride (AIN-76A; Dyets) (Alldred et al., 2018, 2019; 2023). Choline-supplemented chow contained ~4.5 times the concentration of choline present in the control diet, although well within the normal dietary range in humans (Detopoulou et al., 2008). The control diet contained choline levels equivalent to standard rodent chow. Pups born to choline-supplemented (+ = MCS diet) and unsupplemented dams were weaned on postnatal day 21 and placed on the control diet with ad libitum access to food and water. Tail clips were taken at the time of weaning and genotyped as described by Duchon et al. (2011). After weaning, all pups were placed on the control diet with ad libitum access to food and water. Offspring were kept in mixed genotype housing and aged to approximately 3-4 MO (^Y = Young) or 10-12 MO (^A = Aged). The 3-4 MO mouse cohort contained: ^Y2N n = 4 (1F/3M, 119 cells), ^YTs65Dn n = 4 (2F/2M, 113 cells), ^Y2N+ n = 4 (3F/1M, 116 cells), ^YTs65Dn+ n = 4 (2F/2M, 113 cells). The 10-12 MO mouse cohort contained: ^A2N n = 6 (3F/3M, 173 cells), ^ATs65Dn n = 8 (5F/3M, 230 cells), ^A2N+ n = 7 (4F/3M, 222 cells), ^ATs65Dn+ n = 6 (1F/5M, 169 cells).

2.3. Tissue preparation

Mice were anesthetized via intraperitoneal injection of a ketamine (83 mg/kg) and xylazine (13 mg/kg) solution and perfused transcardially with ice-cold 0.15 M phosphate buffer (PB). Whole brains were drop-fixed in a solution of 4% paraformaldehyde buffered in PB and placed on an orbital shaker for 48 h at 4 °C. Brains were cryopreserved at 4 °C via a series of sucrose solutions in PB (12%, 18%, 30%), with a minimum of 24 h in each solution. Sectioning was performed at 40 μm on a cryostat (CM1860; Leica Biosystems, Buffalo Grove, IL) set to −25 °C. Sections were subsequently stored in a cryoprotectant solution (30% glycerol, 30% ethylene glycol, 40% PB) (Gautier and Ginsberg, 2021) at −20 °C until immunostaining.

2.4. Double label Rab5/ChAT immunohistochemistry

Rab5/ChAT immunolabeling was performed as previously described (Gautier and Ginsberg, 2021). Briefly, free-floating sections were rinsed in phosphate buffered saline (PBS; pH 7.4) and blocked in a dilution buffer solution containing 2% bovine serum albumin, 3% fetal bovine serum, and 0.8% Triton X-100 in PBS. Tissue sections were incubated overnight in a rabbit polyclonal antibody against Rab5 (1:500; ab18211, Abcam, Cambridge, MA) on an orbital shaker at 4 °C. Following incubation, sections were washed in dilution buffer then incubated in Alexa-Fluor donkey anti-rabbit 568 secondary antibody (1:500; Invitrogen, Eugene, OR) for 2 h. Tissue sections were washed in PBS and incubated for 36 h in a goat polyclonal antibody against choline acetyltransferase (ChAT; 1:250; AB144P, Millipore Sigma, Temecula, CA) on an orbital shaker at 4 °C. Sections were again washed in dilution buffer and incubated in Alexa-Fluor donkey anti-sheep 488 secondary antibody (1:500; Invitrogen) (Kaur et al., 2017; 2018) for 2 h. Sections were rinsed in PBS, mounted and coverslipped on gelatin-coated glass slides using VectaShield mounting medium (Vector Labs, Burlingame, CA), and stored in the dark at room temperature until imaging.

2.5. Image acquisition

Z-stacks of BFCNs within the MSN/VDB were imaged on a confocal microscope (LSM510, Zeiss, White Plains, NY) using LSM5 software (Zeiss). Z-stack images of individual BFCNs were collected using a Plan-Apochromat 100X/1.4 oil objective at a 0.5 μm slice interval (Gautier and Ginsberg, 2021). BFCNs were selected for imaging in a semi-random manner. Exclusion criteria for imaging included the presence of obfuscating structures such as blood vessels or adjacent non-cholinergic cells, staining artifacts, and tears in the surrounding tissue. A total of 25-35 BFCNs were scanned per brain, distributed throughout the rostrocaudal axis of the MSN/VDB.

2.6. Early endosome quantification

High-definition analysis of vesicular compartments was performed on 100x z-stacks of individual BFCNs reconstructed using Imaris Start version 9.1.2 (Bitplane, Concord, MA), as previously described (Gautier and Ginsberg, 2021). Briefly, individual ChAT-labeled BFCNs were designated as regions of interest (ROIs) within the reconstruction using the Surface creation tool. Immunolabeled early endosomes outside the BFCN of interest were then excluded from analysis by masking red signal external to the ROI. The number of Rab5-labeled early endosomes per BFCN, as well as the individual cross-sectional areas of each vesicle, were quantified using the “Spots” tool. Early endosome cross-sectional area was calculated by Imaris using the region growing algorithm (Bitplane, 2017). Quantification was performed blinded to age, sex, genotype, and maternal diet.

2.7. Early endosome statistical analysis

Data analyses of early endosome phenotype in 2N and Ts65Dn offspring are presented as mean ± standard error of the mean (SEM), as well as fold change from unsupplemented groups. Statistical comparisons between groups were performed in R by contrasts based on a linear model with 3 factors (equivalent to 3-way ANOVA) with false discovery rate (FDR) control (Miller, 1981). Factors evaluated included age, genotype, and maternal diet. Sex differences within cohorts were compared via t-test. For comparisons involving early endosome size, individual endosomes were pooled by mouse and sorted into 1 of 4 groups based upon endosomal cross-sectional area: small (“S” = 0.00-0.50 μm²), medium (“M” = 0.51-1.00 μm²), large (“L” = 1.01-1.50 μm²), and extra-large (“EL” ≥1.51 μm²). Bin sizes (S, M, L, and EL) were adapted from Cataldo et al. (2008) and Jiang et al. (2010; 2016) with minor modifications for z-stack quantification (Gautier and Ginsberg, 2021). Size data were evaluated in R using a linear mixed model with 3 factors (equivalent to 3-way ANOVA with multiple measures) to account for multiple measures per mouse, followed by FDR control (McCulloch et al., 2008). Statistical significance was set at p < 0.05.

2.8. Relative co-expression (RCE) analysis of early endosome genes obtained from BFCNs via laser capture microdissection (LCM)

RCE was conducted on early endosomal gene expression data mined from a cohort of Ts65Dn and 2N offspring on either MCS or a normal choline regimen (4-8 MO; n=40 (33F/7M), 10 per condition) independent from the mouse cohort used for the quantitative morphometry studies (Kelley et al., 2019). Mice were anesthetized with ketamine/xylazine, perfused transcardially with saline and 4% paraformaldehyde, and postfixed in 4% paraformaldehyde for 24 h at 4 °C before being placed in a 30% sucrose solution. Tissue sections were cut at 40 μm in the coronal plane on a freezing sliding microtome and stored in cryoprotectant at 4 °C until immunostaining.

ChAT immunolabeling was performed as described previously (Kelley et al., 2019). Briefly, free-floating sections were incubated in sodium metaperiodate to inhibit endogenous peroxidase activity, washed in Tris-buffered saline (TBS) solution containing 0.25% Triton X-100, and incubated in a blocking solution of 3% horse serum in TBS/Triton X-100. Sections were then incubated in ChAT antibody (1:1,000) in a solution of 1% horse serum in TBS/Triton X-100 overnight on a shaker at room temperature. After incubation in primary antibody solution, tissue sections were processed using the Vectastain Elite kit (Vector Laboratories, Burlingame, CA, USA) and visualized with a solution containing 0.05% 3,3’-diaminobenzadine tetrahydrochloride (Sigma-Aldrich, St. Louis, MO), 1% nickel ammonium sulfate hexahydrate (Sigma), and 0.0015% hydrogen peroxide (Sigma) as described previously (Kelley et al., 2019). Sections were rinsed, mounted on glass slides, and dried overnight at room temperature under RNAse free conditions.

Approximately 100 ChAT-positive MSN/VDB neurons were microisolated per subject via LCM (Arcturus PixCEll Ile, Molecular Devices, Sunnyvale, CA) and samples prepared for custom-designed microarray analysis using the terminal continuation (TC) RNA amplification procedure (Alldred et al., 2008, 2009; Che and Ginsberg, 2004). P³³ labeled TC RNA probes were hybridized to custom-designed microarrays and mounted on a phosphor screen (Amersham Biosciences, Little Chalfont, UK) for 24 h and developed on a Kodak Digital Science Imaging Station (440CF, Rochester, NY). Array images were standardized to the same brightness and contrast levels. Hybridization signal intensity for each gene was assayed as a ratio of the total hybridization signal intensity (global normalization) of the array utilizing ImageQuant software (Amersham Biosciences) (Ginsberg, 2008, 2010; Kelley et al., 2019).

Relative expression levels were compared within groups using a cross-gene correlation for early endosome genes Rab4a, Rab5a, and Eea1, creating a comparative relative co-expression (RCE) matrix to assess interactome connectivity of the three genes based upon genotype and maternal diet. This bioinformatic interrogation demonstrates the health of gene expression networks through robustness, or connectedness, of network plots using a bipartite hub-spoke format with designated hubs. Each gene was treated in isolation using a z-score prior to the calculation of the correlation matrix to prevent bias based on probe binding strength. RCE matrices were calculated in Python using base modules and imported into Gephi (0.9.2) to generate undirected force-generated network plots (Bastian et al., 2009) using the correlation values for edge weight. OpenOrd with default parameters, 750 iterations on 15 threads (cores - 1), and random seed was used to generate final RCE plots (Martin et al., 2011).

3. Results

3.1. Rab5-positive early endosomes increase with age in MSN/VDB BFCNs

To quantify early endosome changes within MSN/VDB BFCNs, brain tissue sections were immunofluorescently double labeled with antibodies directed against ChAT and Rab5 (Figure 1A). Quantification of the number of Rab5-positive early endosomes per BFCN was performed using the “Spots” analysis tool in Imaris (Figure 1B) (Gautier and Ginsberg, 2021).

Figure 1. Quantitative analysis of early endosomes within MSN/VDB BFCNs.

A. Representative confocal image (100x magnification) double-labeled with antibodies against ChAT (green) to identify cholinergic neurons and Rab5 (red) to identify early endosomes. B. A z-stack (100x magnification) of the same BFCN reconstructed as a 3D model using Imaris (left panel). Early endosomes were quantified using Spots analysis and color-coded by endosomal cross-sectional area (right panel). Scale bar: 10 μm.

A total of 461 and 794 BFCNs were reconstructed and quantified in the 3-4 MO and 10-12 MO cohorts, respectively. The average number of early endosomes per BFCN for each group analyzed is presented in Table 1. Comparisons of the average number of early endosomes per cell within the young cohort showed no significant differences by genotype between choline-normal diet ^YTs65Dn mice and ^Y2N littermates (Figures 2A, C; 3A, B). Quantitative analysis of the 10-12 MO cohort showed significant differences in the average number of endosomes per BFCN by both age and genotype (Figure 2B, E, G). ^ATs65Dn offspring had a ~1.2-fold increase compared to ^A2N littermates (p = 1.15 x 10⁻⁶) and a ~1.9-fold increase compared to ^YTs65Dn counterparts (p = 8.97 x 10⁻¹⁶) (Figures 2E, G). Early endosomes in ^ATs65Dn mice often formed large globular clusters within the soma (Figure 2I; 3D). In contrast, early endosomes in ^A2N littermates were spread uniformly throughout the cytoplasm. Unexpectedly, ^A2N mice also had a significant ~1.6-fold increase (p = 1.26 x 10⁻⁹) in the average number of early endosomes when compared to ^Y2N counterparts (Figure 2G). Despite an age-associated increase in the number of Rab5-immunoreactive early endosomes in 2N mice, endosomes did not aggregate to form clusters observed in ^ATs littermates (Figure 2I).

Table 1.

The average number of early endosomes per BFCN in the MSN/VDB.

Group	Total BFCNs	Total Endosomes	Average Endosomes per BFCN
Y2N	119	4,125	34.8 ± 1.3
YTs65Dn	113	3,911	34.6 ± 1.3
Y2N+	116	3,871	33.3 ± 0.3
YTs65Dn+	113	3,776	33.4 ± 1.2
A2N	173	9,415	54.6 ± 1.6
ATs65Dn	230	15,353	66.5 ± 1.8
A2N+	222	10,039	45 ± 1.6
ATs65Dn+	169	7,861	46.5 ± 1.1

Total Rab5-immunoreactive early endosome numbers and average (mean ± SEM) early endosome numbers per ChAT-immunoreactive BFCN in the MSN/VDB by experimental group.

Figure 2. MCS significantly reduces age- and genotype-associated increases in BFCN early endosome number in aged offspring.

A-B. Total early endosomes for each BFCN analyzed in the 3-4 MO (A) and 10-12 MO (B) cohorts. C-D. MCS had no effect on average early endosome numbers per BFCN across the 3-4 MO cohort normalized to ^Y2N (C), or when ^YTs65Dn+ mice were compared to ^YTs65Dn (D). E. ^ATs65Dn mice have increased early endosomes per BFCN (p < 0.001) normalized to ^A2N. MCS resulted in a decrease in early endosome number in ^A2N+ (p < 0.001) and ^ATs65Dn+ (p < 0.001) offspring. F. MCS decreased average early endosome counts in 10-12 MO ^ATs65Dn+ offspring (p < 0.001) normalized to ^ATs65Dn. G. Comparing age groups, early endosome counts increased in both 2N (p < 0.001) and Ts65Dn (p < 0.001) mice, with a greater increase in Ts65Dn. Age-associated increases were lessened in MCS ^A2N+ (p < 0.001) and ^ATs65Dn+ (p < 0.001) offspring. H. Comparing only trisomic mice, early endosome counts increased with age in ^ATs65Dn mice (p < 0.001) normalized to ^YTs65Dn. MCS attenuated this increase (p < 0.001). I. Representative 3D reconstructions of a BFCN from ^A2N (left panel) and ^ATs65Dn (right panel) littermates. Increased Rab5-ir early endosomes are visible in the ^ATs65Dn BFCN. Error bars on scatter plots indicate SD. Error bars on bar graphs indicate SEM. Statistical analysis was performed via contrasts based on linear model with 3 factors (age, genotype, MCS) with FDR control. A total of (n = 461) BFCNs in the 3-4 MO cohort and (n=794) BFCNs in the 10-12 MO cohort were analyzed, with 4-8 mice per group. Scale bar: 5 μm. Key: ***p < 0.001.

Figure 3. MCS reduces Rab5-immunoreactive early endosomes within MSN/VDB BFCNs in aged offspring.

Representative 3D reconstructed confocal z-stacks (100x magnification) depict littermate pairs (A and B, C and D, E and F, G and H). 3-4 MO ^Y2N (A) and ^YTs65Dn (B) littermates have similar endosomal phenotypes. C. No significant early endosome pathology is visible in 10-12 MO ^A2N mice, although early endosome counts are increased compared to ^Y2N mice. D. 10-12 MO ^ATs65Dn mice have increased Rab5-ir early endosomes in MSN/VDB BFCNs. 3-4 MO ^Y2N+ (E) and ^YTs65Dn+ (F) MCS offspring appear phenotypically similar, with no apparent MCS effects on early endosome number. MCS significantly decreases Rab5-ir early endosome counts in 10-12 MO ^A2N+ (G) and ^ATs65Dn+ (H) offspring. Scale bar: 4 μm.

3.2. MCS decreases early endosome number per BFCN in aged mice independent of genotype

In 3-4 MO MCS offspring, no significant changes in the average number of early endosomes were observed between ^YTs65Dn+ and ^Y2N+ offspring (Figure 2A, C; 3E, F). At 10-12 MO, MCS induced a significant ~1.3-fold decrease (p = 1.24 x 10⁻¹¹) in early endosome number per BFCN in ^ATs65Dn+ offspring compared to unsupplemented ^ATs65Dn mice (Figure 2F; 3D, H). Comparisons between ^ATs65Dn+ and ^A2N offspring showed a significant ~1.2-fold decrease (p = 0.0006) (Figure 2E). ^A2N+ offspring also had a significant, ~1.2-fold decrease (p = 5.45 x 10⁻⁵) in the number of early endosomes per BFCN compared to ^A2N mice (Figures 2E; 3C, G). Strikingly, no significant differences in early endosomes per BFCN were found between ^A2N+ and ^ATs+ littermates (Figures 2E; 3G, H), indicating rescue of the early endosome phenotype.

Comparing 3-4 MO and 10-12 MO MCS offspring, significant age-associated increases in the average number of endosomes per neuron were still observed, albeit to a lesser extent than in unsupplemented mice. Both ^A2N+ and ^ATs65Dn+ mice displayed ~1.3-fold increases in the number of early endosomes per BFCN relative to ^Y2N+ (p = 1.86 x 10⁻⁵) and ^YTs65Dn+ (p = 4.67 x 10⁻⁶) mice, respectively (Figure 2G). Although the cohorts were not optimally sex balanced, no significant sex differences in early endosome number were observed by age, genotype, or maternal diet in MSN/VDB BFCNs (Supplementary Figure 1).

3.3. MCS decreases early endosomes of all sizes in aged Ts65Dn mice

In addition to quantifying the number of Rab5-immunoreactive early endosomes per BFCN, the size of endosomes within identified BFCNs was measured. Endosomes were pooled by mouse and sorted into one of four bins based on the cross-sectional area of the individual vesicle. Size bins were designated as either small (“S” = 0.00-0.50 μm²), medium (“M” = 0.51-1.00 μm²), large (“L” = 1.01-1.50 μm²), or extra-large (“EL” ≥1.51 μm²). A total of 15,683 early endosomes were analyzed in the 3-4 MO cohort, and 42,668 in the 10-12 MO cohort. No significant genotype-dependent changes were observed in early endosome number per bin in the 3-4 MO cohort. Comparisons of pooled endosome counts between ^ATs65Dn and ^A2N littermates revealed genotype-associated increases in the number of Rab5-identified early endosomes across all bins in trisomic mice (Table 2). ^ATs65Dn mice displayed a ~1.2-fold increase in the average number of early endosomes for each bin size compared to ^A2N littermates, with significance attained in the L (p = 0.04) and EL (p = 0.05) bins (Figure 4A).

Table 2.

The average number of early endosomes per size bin within BFCN in the MSN/VDB.

		Bin Size
		S		M		L		EL
		(0.00-0.50 μm2)		(0.51-1.00 μm2)		(1.01-1.50 μm2)		(1.51+ μm2)
Group	n	Total #	Average #	Total #	Average #	Total #	Average #	Total #	Average #
Y2N	4	199	49.7 ± 9.4	955	238.7 ± 31.2	1,192	298 ± 23.9	1,779	444.7 ± 49.6
YTs65Dn	4	207	51.7 ± 12.4	975	243.7 ± 45.5	1,038	259.5 ± 23	1,691	422.7 ± 48.3
Y2N+	4	333	83.2 ± 8.2	1,354	338.5 ± 31.1	1,060	265 ± 11.8	1,124	281 ± 39.6
YTs65Dn+	4	253	63.2 ± 14	1,127	281.7 ± 34.1	946	236.5 ± 20.2	1,450	362.5 ± 64.8
A2N	6	574	95.6 ± 11.7	2,620	436.6 ± 49.8	2,101	350.1 ± 18.7	4,120	686.6 ± 60.6
ATs65Dn	8	968	121 ± 13.9	4,382	547.7 ± 59	3,322	415.2 ± 30.1	6,681	835.1 ± 44.8
A2N+	7	511	73 ± 7.5	2,700	385.7 ± 38.6	2,431	347.2 ± 18.9	4,397	628.1 ± 66.8
ATs65Dn+	6	432	72 ± 12	1,990	331.6 ± 53.2	1,768	294.6 ± 25.1	3,671	611.8 ± 73

The average (mean ± SEM) number of Rab5-immunoreactive early endosomes in each size bin within identified BFCNs.

Figure 4. MCS reduces age- and genotype-associated increases in endosome number in early endosomes of multiple sizes.

A. The average number of Rab5-ir early endosomes significantly increased in the L (p < 0.05) and EL (p < 0.05) size bins in 10-12 MO ^ATs65Dn mice normalized to ^A2N. B. Comparing aged trisomic mice, MCS significantly decreased the average number of early endosomes in all size bins (S: p < 0.01, M: p < 0.001, L: p < 0.001, EL: p < 0.01). C-D. The average number of early endosomes increases in S, M, and EL size bins (all p < 0.01) in 2N mice with age (C) and in all size bins (all p < 0.001) in Ts65Dn mice with age (D). E-F. MCS attenuates age-associated increases in S, M, and L early endosome bins in 2N mice (E) and Ts65Dn mice (F), but the number of early endosomes ≥ 1.51 μm² still significantly increased compared to counterparts (p < 0.001 and p < 0.01, respectively). Error bars indicate SEM. Statistical analysis was performed via contrasts based on linear mixed model with 3 factors (age, genotype, and MCS) with FDR control. A total of (n = 15,683) endosomes in the 3-4 MO cohort and (n = 42,668) endosomes in the 10-12 MO cohort were analyzed, with 4-8 mice per group. Key: *p < 0.05, **p < 0.01, ***p < 0.001.

In the 10-12 MO cohort, MCS significantly lowered the average number of early endosomes across all bins in ^ATs65Dn+ compared to ^ATs65Dn offspring (Table 2). Endosomes in S (p = 0.002) and M (p = 0.001) bins were reduced by ~1.4-fold in ^ATs65Dn+ compared to ^ATs65Dn offspring. Endosomes in the L (p = 0.0004) and EL (p = 0.005) bins decreased by ~1.3-fold compared to ^ATs65Dn (Figure 4B). No significant differences were observed between ^A2N+ and ^ATs65Dn+ littermates in any of the endosome size bins (Figure 4A).

3.4. MCS reduces age-associated increases in early endosomes of multiple sizes independent of genotype

Comparisons between 3-4 MO and 10-12 MO cohorts revealed significant age-associated effects in both genotypes (Table 2). ^A2N mice had significant increases in the average number of early endosomes in S (~1.9-fold; p = 0.01), M (~1.8-fold; p = 0.01), and EL (~1.5-fold; p = 0.01) bins compared to ^Y2N counterparts (Figure 4C). ^ATs65Dn mice had significant increases in the average number of endosomes in S (~2.3-fold; p = 0.0003), M (~2.2-fold; p = 0.0001), L (~1.6-fold; p = 8.84 x 10⁻⁵), and EL (~2-fold; p = 3.2 x 10⁻⁵) bins compared to ^YTs65Dn mice bins (Figure 4D).

Age-associated increases in early endosome number were significantly attenuated in several bins within MCS offspring. Unlike unsupplemented counterparts, no significant differences were observed between ^YTs65Dn+ and ^ATs65Dn+ offspring in S, M, and L bins (Figure 4F). Similarly, no significant increases were documented between ^Y2N+ and ^A2N+ offspring in S, M, and L bins (Figure 4E). However, MCS did not impact age-associated increases in the average number of extra-large sized early endosomes in offspring of either genotype. ^ATs65Dn+ offspring had a significant ~1.7-fold increase in extra-large early endosomes (p = 0.009) compared to ^YTs65Dn+ mice, and ^A2N+ mice had a significant ~2.2-fold increase in extra-large early endosomes (p = 0.0003) compared to ^Y2N+ counterparts (Figure 4E, F).

3.5. MCS reduces enlargement of extra-large early endosomes in aged trisomic mice

The average cross-sectional area of individual Rab5-positive early endosomes within the S, M, L, and EL size bins were evaluated for alterations based on age, genotype, and maternal diet. The results of this analysis are presented in Table 3. No significant cross-sectional area changes were found between littermates in any size bin within the 3-4 MO and 10-12 MO cohorts. Comparison between 3-4 MO and 10-12 MO cohorts revealed a significant ~1.1-fold increase (p = 0.002) in the average cross-sectional area of extra-large early endosomes in ^ATs65Dn mice compared to ^YTs65Dn (Figure 5B). This significant age-associated increase was not observed in supplemented ^ATs65Dn+ offspring compared to ^YTs65Dn+ (Figure 5D). No significant age- or maternal diet-associated changes were observed between ^Y2N and ^A2N or ^Y2N+ and ^A2N+ mice (Figure 5A, C).

Table 3.

The average size of early endosomes per size bin within BFCN in the MSN/VDB.

		Bin Size
		S		M		L		EL
		(0.00-0.50 μm2)		(0.51-1.00 μm2)		(1.01-1.50 μm2)		(1.51+ μm2)
Group	n	Total #	Average Size	Total #	Average Size	Total #	Average Size	Total #	Average Size
Y2N	4	199	0.34 ± 0.02	955	0.75 ± 0.02	1,192	1.25 ± 0.01	1,779	2.13 ± 0.05
YTs65Dn	4	207	0.34 ± 0.02	975	0.75 ± 0.02	1,038	1.25 ± 0.01	1,691	2.14 ± 0.05
Y2N+	4	333	0.34 ± 0.02	1,354	0.75 ± 0.02	1,060	1.25 ± 0.01	1,124	2.07 ± 0.05
YTs65Dn+	4	253	0.34 ± 0.02	1,127	0.75 ± 0.02	946	1.25 ± 0.01	1,450	2.11 ± 0.05
A2N	6	574	0.34 ± 0.01	2,620	0.76 ± 0.01	2,101	1.25 ± 0.01	4,120	2.23 ± 0.04
ATs65Dn	8	968	0.35 ± 0.01	4,382	0.75 ± 0.01	3,322	1.25 ± 0.01	6,681	2.34 ± 0.03
A2N+	7	511	0.33 ± 0.01	2,700	0.75 ± 0.01	2,431	1.26 ± 0.01	4,397	2.13 ± 0.04
ATs65Dn+	6	432	0.34 ± 0.01	1,990	0.75 ± 0.01	1,768	1.26 ± 0.01	3,671	2.23 ± 0.04

The average (mean ± SEM) early endosome cross-sectional area in each size bin within identified BFCNs.

Figure 5. MCS rescues age-associated increases in early endosome cross-sectional area in trisomic offspring.

A. No significant changes in average early endosome cross-sectional area are observed in 2N mice with age. B. Cross-sectional area of Rab5-ir early endosomes ≥ 1.51 μm² significantly increased (p < 0.01) in 10-12 MO ^ATs65Dn mice compared to 3-4 MO counterparts. C. No significant changes in average early endosome cross-sectional area are observed in 2N+ mice with age. D. Ts65Dn+ offspring do not have significant changes in EL early endosome cross-sectional area with age, in contrast to unsupplemented Ts65Dn counterparts. Error bars indicate SEM. Statistical analysis was performed via contrasts based on linear mixed model with 3 factors (age, genotype, and MCS) with FDR control. A total of (n = 15,683) endosomes in the 3-4 MO cohort and (n=42,668) endosomes in the 10-12 MO cohort were analyzed, with 4-8 mice per group. Key: **p < 0.01.

3.6. MCS attenuates loss of interactome connectivity in select early endosome genes in trisomic mice

To interrogate the effects of MCS on interactome cooperativity of early endosome genes, single population gene expression analysis was performed on BFCNs from an independent cohort of mice (Kelley et al., 2019). In RCE networks within the three isolated hubs (Rab4a, Rab5a, and Eea1), the clustering coefficient approximated 0, with modularity indicative of imbalance across hubs. Node and edge numbers were approximately equal across all groups (2N = 112:113; 2N+ = 70:69; Ts65Dn = 58:57; Ts65Dn+ = 322:325), as expected with RCE networks paired for hub-spoke modeling. Disomic offspring had similar modularity scores (0.60 for 2N and 0.61 for 2N+), suggesting little difference across the three early endosome genes and their degree of relative co-expression as a function of maternal diet (Figures 6A, B). Comparison of interactome connectivity maps between Ts65Dn and 2N littermates indicated trisomic mice had a lower RCE score (0.50), suggesting decreased interactome connectivity (Figure 6A, C). This was most apparent for Rab5a, where an interactome of 25 co-expressed genes in 2N mice was reduced to 3 in Ts65Dn littermates (Supplementary Figure 2A, C).

Figure 6. Comparative RCE networks of early endosome genes reveal loss of interactome connectivity in trisomic mice.

A comparison of relative expression network connectedness for Rab4a, Rab5a, and Eea1 by genotype and maternal diet using bipartite hub-spoke modeling. Early endosome genes are represented as isolated hubs, with interacting genes connected via individual spokes. RCE plots for 2N mice (A) and 2N+ mice (B) are similar (modularity score = 0.60 for 2N and 0.61 for 2N+), indicating little change in gene connectivity networks as a function of MCS. Ts65Dn mice (C) have decreased interactome connectivity in all 3 genes (modularity score = 0.50) compared to 2N counterparts. MCS alters interactome connectivity in Ts65Dn+ mice (D) in all 3 genes (modularity score = 0.25). Matrices were calculated in Python, undirected force-generated network plots were generated using Gephi, and final RCE plots were created using OpenOrd. A total of n = 40 mice (10 mice per group) were analyzed.

MCS rescued interactome connectivity in Ts65Dn+ offspring compared to unsupplemented counterparts (Figure 6C, D). Ts65Dn+ offspring had a lower RCE score (0.25) than Ts65Dn mice, which can be attributed to the interactome hyperconnectivity of Eea1 (Figure 6D). Interactome connectivity of Rab4a and Rab5a approached levels seen in 2N littermates (Figure 6B, D; Supplementary Figure 2B, D). Although individual genes varied from hub to hub, similar pathways were represented by genotype. AD-amyloid pathway genes were found in trisomic mice within the Rab5a hub {e.g., amyloid beta precursor protein (App), amyloid beta precursor like protein 1 (Aplp1), and amyloid beta precursor protein binding family B member 1 (Apbb1)}, whereas neurotrophin and neurotransmitter pathway members were found in the Rab5a hub in 2N littermates {e.g., brain derived neurotrophic factor (Bdnf), insulin like growth factor 1 (Igf1), cholinergic receptor muscarinic 2 (Chrm2), dopamine receptors Drd2 and Drd4, adrenoceptor alpha 1B (Adra1b), and galanin receptor 2 (Galr2)}. A complete list of Rab5a co-expressed genes for each group can be found in Supplementary Figure 2.

4. Discussion

We evaluated Rab5-immunoreactive early endosomes within ChAT-immunoreactive BFCNs in the MSN/VDB of young and aged Ts65Dn and 2N offspring following perinatal MCS. In addition to rescuing age- and genotype-associated changes in the pathologic early endosome phenotype in Ts65Dn offspring, MCS decreased the number of early endosomes in aged 2N littermates. RCE analysis of key early endosomal genes, including Rab5a, indicates loss of interactome connectivity found in trisomic offspring is attenuated by MCS. These results suggest a potential mechanism by which MCS is able to offset alterations in endosomal function in Ts65Dn mice, rescuing vulnerable BFCNs from degeneration (Gautier et al., 2023).

Early endosome number and size is directly proportional to the rate of endocytic intake and fusion in vitro (Bucci et al., 1992; Gorvel et al., 1991; Stenmark et al., 1994). Accordingly, increased quantities of very large Rab5-positive early endosomes in human DS and AD, as well as relevant cellular/animal models, are a marker of pathologically increased endocytosis (Cataldo et al., 1996; 1997; 2008; Grbovic et al., 2003; Jiang et al., 2010; 2016; Pensalfini et al., 2021). No significant differences were found in average early endosome number or size between 3-4 MO offspring. By 10-12 MO, ^ATs65Dn offspring had significant increases in the number of early endosomes per BFCN of all sizes compared to ^A2N littermates. Age-associated increases correspond with the progressive impairment of neurotrophic signaling (Granholm et al., 2000; Hunter et al., 2003; Seo and Isacson, 2005). Neurotrophin trafficking studies demonstrate retrograde transport of NGF-TrkA complexes (Marlin and Li, 2015) between the hippocampus and medial septum is nearly abolished in trisomic mice by 12 MO (Cooper et al., 2001; Salehi et al., 2006). The fate of endocytosed NGF-TrkA in the soma is complex, and its relation to alterations in endosomal structure and function in trisomic mice requires further investigation.

Age-associated increases in some Rab5-immunoreactive early endosomes were also found in 2N mice. While the exact cause of increased early endosomes in aged 2N mice is unknown, it may be due in part to upregulation within endosomal pathways that accompany normal neuronal senescence. Support for this hypothesis comes from the observation that clathrin-mediated endocytosis and clathrin-independent endocytosis are upregulated in the brains of elderly individuals with no cognitive impairment (Alsaqati et al., 2018). Notably, accelerated APP endocytosis, trafficking, and production of APP metabolites (Burrinha and Claudia, 2022) has been documented in aged primary neurons (Burrinha et al., 2021) and normal human fibroblasts (Zou et al., 2010), even in the absence of pathologic Aβ accumulation and amyloid plaque deposition. Taken together, these data suggest an overall increase in the number of early endosomes within vulnerable BFCNs may be a normal biproduct of the aging process. The extent of this increase, however, and whether this becomes pathological over time, is dependent on genotype. Global increases in early endosomes of all sizes in ^ATs65Dn mice may be reflective of the neurodegenerative process as the endosomal-lysosomal system becomes increasingly dysregulated (Colacurcio et al., 2018), leading to downstream consequences including neurotrophin signaling failure.

MCS had a profound impact on endosomal phenotype in MSN/VDB BFCNs in 10-12 MO supplemented offspring, reducing age-associated increases in early endosome number in S, M, and L size categories. However, MCS did not mitigate increases in endosomes greater than 1.51 μm² in either genotype, suggesting biogenesis of extra-large early endosomes may occur through a distinct mechanism. Broad-spectrum decreases in endosomal number, rather than precision reduction of enlarged vesicles, suggests MCS may be ameliorating early endosomal phenotype by rescuing endocytic homeostasis. In vitro and in vivo experiments exploring both Rab5 overexpression and silencing demonstrated that even moderate alterations in Rab5 expression can significantly increase receptor-mediated uptake or prevent it entirely (Bucci et al., 1992; Grbovic et al., 2003; Laifenfeld et al., 2007; Stenmark et al., 1994; Pensalfini et al., 2020). To this end, MCS may help rebalance endocytic flux and correct the reciprocal anomalies documented in recycling endosomes and late endosomes (Cataldo et al., 2000; 2003; 2008; Xu et al., 2018). The impact of MCS on lysosomal function and autophagy await further study.

Mechanistically, the effect of MCS on early endosome pathology in Ts65Dn mice may be a downstream effect of the downregulation of APP (Alldred et al., 2018). Aberrant APP expression and processing in DS/AD is linked to endosomal abnormalities through the activity of APPL1 (Kim et al., 2016). Under normative conditions, hydrolysis of GTP to GDP releases APPL1 from Rab5 (Chial et al., 2008; Miaczynska et al., 2004; Zhu et al., 2007). In DS/AD, simultaneous binding of APPL1 to Rab5 and β-CTF locks Rab5 in its active GTP-bound state, inhibiting GTP hydrolysis and APPL1 release (Kim et al., 2016). This leads to an accumulation of APPL1 on the surface of early endosomes and a pathological overactivation of Rab5 (Nixon, 2017; Pensalfini et al., 2020). Reducing the beta C-terminal fragments of APP (β-CTF) (Jiang et al., 2010; 2016) or preventing interactions between β-CTF and GTP-Rab5 (Xu et al., 2016) rescues endosomal pathology and BFCN degeneration. Choline supplementation decreases APP levels in the hippocampus of aged Ts65Dn mice (Alldred et al., 2018) and reduces amyloidosis in APP/PS1 mice (Mellott et al., 2017; Velazquez et al., 2019). Resolution of the early endosome phenotype in ^ATs65Dn+ offspring may therefore be correlated with a reduction in the intracellular β-CTF that induces the pathologic positive feedback loop.

To generate a nuanced view of the impact of genotype and maternal diet on relevant early endosome genes in BFCNs, we employed a focused bioinformatic investigation that enables the generation of network maps using RCE. Previous collaborative expression profiling studies of trisomic mice in the context of MCS revealed a correction in trisomic gene expression to 2N levels that closely mirror histological changes and have been validated by both qPCR and Nanostring nCounter (Alldred et al., 2018, 2019; 2021, 2023; Kelley et al., 2019). These profiling studies revealed expression level alterations across many functional groups, with a mosaic of upregulated and downregulated genes that made pattern recognition difficult to ascertain, both globally and in select pathways. RCE network analysis revealed a loss of cooperativity between gene expression levels in trisomic mice, rather than unilateral increase or decrease. MCS rebalanced early endosome interactome connectivity for Rab5a and Rab4a, indicating the ability to positively affect network integrity of key hub genes. Whether MCS induced connectivity changes in Eea1 are beneficial has yet to be determined.

A caveat of this approach is that no specific pathways can be analyzed or extracted from these networks in terms of overall expression levels. The RCE method does not enable an assessment as to whether individual genes are more or less co-expressed with other genes in the interactome. Rather, the advantage of this approach is that it allows for an unbiased view of cooperativity across a network of genes driven through a single hub. The more connected a hub is through its interactome network, the greater the degree of molecular cohesion. Accordingly, this approach resulted in elucidation of gene dysfunction in Ts65Dn mice, particularly in Rab5a, that is not present in any of the other groups examined. Although this analysis was performed on an independent cohort of mice, these results reflect what we observed pathologically through quantitative morphometry of endosomes within BFCNs in the MSN/VDB, providing molecular validation of our cellular-based analysis.

The connection between enlarged endosomal phenotypes and neurotrophic deficits in trisomic BFCNs is well established (Cooper et al., 2001; Hunter et al., 2003; Salehi et al., 2006; Xu et al., 2016). Based on our endosomal findings and RCE results, we posit MCS plays a role in correcting connectivity-based septohippocampal communication through preservation of neurotrophic signaling, likely downstream of endosomal trafficking. In light of the age-related changes we observed, MCS may not be as beneficial for neurotrophin signaling in young, 3-4 MO trisomic mice compared to older, 10-12 MO counterparts that have a profoundly abnormal endosomal phenotype. Although we currently have supportive immunohistochemical and transcriptomic evidence in trisomic BFCNs (Alldred et al., 2023; Gautier et al., 2023; Kelley et al., 2014; 2016; 2019), this hypothesis awaits formal NGF-TrkA septohippocampal transport studies (Choi et al., 2009) in trisomic MCS offspring compared to trisomic offspring exposed to a normal choline diet.

5. Conclusions

Through a combination of morphometric analysis of early endosomal phenotype and RCE interrogation of early endosome associated genes, we demonstrate perinatal MCS reduces age- and genotype-associated changes in endosome pathology in supplemented offspring and restores interactome integrity. Pathologic alterations within the endosomal-lysosomal and autophagy pathways are found in a number of neurodegenerative disorders, including Parkinson’s disease (Nixon, 2013), Huntington’s disease (Kegel et al., 2000; Pal et al., 2006), Neimann Pick Type C (Cabeza et al., 2012; Jin et al., 2004), and amyotrophic lateral sclerosis (Mitra et al., 2019). The present results derived from an established murine model of DS support continued translational investigation into the use of choline supplementation in the treatment of neurological conditions that involve deleterious vesicular phenotypes.

Highlights.

• Evaluated maternal choline supplementation in a murine model of DS/AD

• MCS decreases early endosome pathology in trisomic offspring

• MCS attenuates age-associated increases in early endosome number in BFCNs

• MCS restores early endosome interactome connectivity

Abbreviations

2N

disomic offspring

^A2N

aged disomic offspring

^A2N+

aged maternal choline supplemented disomic offspring

^Y2N

young disomic offspring

^Y2N+

young maternal choline supplemented disomic offspring

Aβ

amyloid-beta peptide

Adra1b

adrenoceptor alpha 1B receptor

AD

Alzheimer’s disease

App

amyloid-beta precursor protein

Aplp1

amyloid-beta precursor like protein 1

Apbb1

amyloid-beta precursor protein binding family B member 1

β-CTF

beta C-terminal fragments of APP

BDNF

brain derived neurotrophic factor

BFCNs

basal forebrain cholinergic neurons

ChAT

choline acetyltransferase

Chrm2

cholinergic receptor muscarinic 2

Drd2 & Drd4

D2 & D4 dopamine receptors

DS

Down syndrome

FDR

false discovery rate

Galr2

galanin receptor 2

HSA21

human chromosome 21

Igf1

insulin like growth factor 1

LCM

laser capture microdissection

MCS

maternal choline supplementation

MMU16

mouse chromosome 16

MMU17

mouse chromosome 17

MO

months of age

MSN/VDB

medial septal nucleus/vertical limb of the diagonal band

NBM/SI

nucleus basalis of Meynert/substantia innominata

NFTs

neurofibrillary tangles

PB

phosphate buffer

PBS

phosphate buffered saline

RCE

relative co-expression

ROI

region of interest

SEM

standard error of the mean

TBS

Tris-buffered saline

TC

terminal continuation

Ts65Dn

trisomic offspring

^ATs65Dn

aged trisomic offspring

^ATs65Dn+

aged maternal choline supplemented trisomic offspring

^YTs65Dn

young trisomic offspring

^YTs65Dn+

young maternal choline supplemented trisomic offspring