Neuronal Deletion of Tumor Susceptibility Gene 101 (Tsg101) Causes Rapid Apoptotic Loss of Hippocampal CA3 Neurons

1. Introduction

Alzheimer’s disease (AD) is the most common neurodegenerative disorder, with an estimated 55 million people worldwide currently living with AD. It is associated with the accumulation of extracellular amyloid β (Aβ) plaques and intracellular neurofibrillary tangles of hyperphosphorylated tau protein (p-tau), but one of the earliest cellular features is endosomal dysfunction in neurons [1,2,3,4]. Studies have detected abnormally enlarged early endosomes and elevated expression of the small guanosine triphosphatase (GTPase), ras-related proteins RAB5 (early endosome marker) and RAB7 (late endosome marker) in hippocampal neurons in individuals with mild cognitive impairment and AD [4,5,6]. In multiple cell types from AD patients or mouse models, endocytic activation associated with increased RAB5 activity led to late endosomal dysfunction [7,8], which suggests that increased RAB5 activity could in turn lead to elevated RAB7 expression. RAB5 is aberrantly activated by APP-βCTF (secretase-cleaved C-terminal fragment (CTF) of amyloid precursor protein, APP), and transgenic expression of activated RAB5 in mouse neurons led to enlargement and mistrafficking of early endosomes as well as synaptic plasticity deficits, tau hyperphosphorylation, and cognitive deficits [7,9,10]. Thus, endosomal defects may play a direct role in AD pathogenesis.

Tumor susceptibility gene 101 (TSG101) is a housekeeping gene that encodes a multidomain protein with many functions in different cellular compartments. TSG101 has been proposed to play roles in transcriptional regulation, cell proliferation, cell division, retroviral particle assembly, and viral budding, but it is best characterized as a component of the Endosomal Sorting Complex Required for Transport (ESCRT)-I [11,12,13,14,15,16,17,18,19,20,21]. The ESCRT machinery sorts ubiquitinated proteins onto intraluminal vesicles (ILVs) of multivesicular bodies (MVBs) for trafficking either to the lysosome, for degradation, or to the plasma membrane, whereupon ILVs are secreted as a subclass of nanosized extracellular vesicles (EVs) called exosomes [22,23,24]. The cellular mechanisms that determine whether MVBs traffic their ubiquitinated cargoes to the lysosome or the plasma membrane remain poorly understood. APP is ubiquitinated and sorted through the endosomal system in an ESCRT-dependent manner [25,26,27,28], and Tsg101 depletion in cultured cells inhibited the sorting of APP onto ILVs and to the lysosome, resulting in intracellular accumulation of Aβ and a concomitant decrease in Aβ secretion [26].

There are many additional links between the endo-lysosomal pathway and AD. APP cleavage to form amyloid-β (Aβ) and CTFs is believed to occur, at least in part, in early endosomes [29,30,31,32], and the acidic environment of late endosomes/lysosomes would promote the self-assembly of Aβ peptides into fibrils that seed plaque formation when released from the cell, perhaps via exosomes or other EVs [33,34,35]. Neuronal cells in the brains of human AD patients and mouse and rat AD models, as well as in cell culture, have been shown to accumulate aggregation-prone and oligomerized Aβ peptides in MVBs, and to secrete exosomes containing full-length APP as well as CTFs and Aβ [36,37,38,39,40,41]. Neurally-derived exosomes isolated from the blood of AD patients were reported to contain significantly higher levels of lysosomal proteins (cathepsin D and LAMP-1) and ubiquitinated proteins, indicative of endo-lysosomal dysfunction [42]. ESCRT components were associated with amyloid plaques in AD transgenic mice [35]. Knockdown of ESCRT proteins, including TSG101, promoted tau aggregation [43], and endosomally-derived exosomes contain and may propagate the spread of p-tau [39,44,45,46,47].

Most studies on the molecular, biochemical, and intracellular functions of TSG101 have emphasized the relevance of their specific findings to diseases like cancer and neurodegeneration, but the proposed mechanisms are rarely validated in genetically defined, in vivo disease models or primary human tissue samples. Consequently, the biological significance of the multifaceted functions of TSG101 in normal development, differentiation, and tissue homeostasis, as well as the potential roles of TSG101-mediated pathways in disease pathogenesis, are poorly defined [22]. In the present study, we used tamoxifen-induced conditional mutagenesis to delete Tsg101 from forebrain neurons in adult mice. Neuronal deletion of Tsg101 resulted in reactive astrogliosis in the dentate gyrus starting within 1-week of Tsg101 deletion, followed by apoptosis of hippocampal CA3 neurons and forebrain atrophy. No upregulation of autophagy was detected. Impaired endosomal trafficking was indicated by the accumulation of ubiquitinated proteins, RAB5, and RAB7, and there was an increase in intracellular APP and/or its proteolytic derivatives in hippocampal neurons. This study supports the hypothesis that disrupted ESCRT function affects the trafficking of APP and its proteolytic derivatives, such as Aβ, and demonstrates that impairing TSG101-dependent processes leads to rapid onset neurodegeneration in brain areas impacted in AD.

2. Materials and Methods

2.1. Mice

Animals were housed under standard conditions in the Animal Resource Center at the McLaughlin Research Institute. Tsg101^tm1-Kuw (Tsg101^fl) conditional knockout mice, which were created and maintained on a 129/SvJ genetic background, have been described previously [19]. As this allele has loxP sites 3kb upstream and 230 bp downstream of the first coding exon of Tsg101, Cre recombinase excises the proximal promoter region and first exon of Tsg101, resulting in a null allele. Tsg101^fl/fl mice were crossed to the tamoxifen-inducible Cre transgenic line, B6.FVB-Tg(Camk2a-cre/ERT2)2Gsc/Ieg (referred to as Camk2a-cre/ERT2), which expresses CreERT2 in forebrain neurons [48]. Camk2a-cre/ERT2 mice were obtained as frozen embryos from the European Mutant Mouse Archive, Munich, Germany (EMMA ID EM:02125) and rederived by the McLaughlin Research Institute Transgenic Facility (RRID:SCR_023108). Cre-positive F1’s were backcrossed to Tsg101^fl/fl animals to produce Tsg101^fl/fl and Tsg101^fl/+ mice segregating for the Cre transgene. Tamoxifen treated Tsg101^fl/+; Camk2a-cre/ERT2(+) or Tsg101^fl/fl without the Camk2a-cre/ERT2 transgene, or vehicle-treated Tsg101^fl/fl; Camk2a-cre/ERT2(+) were used as controls, and no significant differences in brain pathology were observed between these groups. Despite the mixed genetic background, no differences were observed in brain pathology between animals within the same genotype/treatment group. All animal procedures adhered to Association for Assessment and Accreditation of Laboratory Animal Care guidelines and were approved by the Institutional Animal Care and Use Committee of the McLaughlin Research Institute.

2.2. Genotyping

Animals were genotyped using PCR GoTaq Green 2X Master Mix (Promega). Tsg101 genotyping primers were: Tsg101^wt GTTCGCTGAAGTAGAGCAGCCAG and CATTTCTGGAGTCCGATGCGCAG; Tsg101^fl AGAGGCTATTCGGCTATGACTG and TTCGTCCAGATCATCCTGATC; Tsg101^null GATGGTCATACCTGGTTAGAAAGC and CATTTCTGGAGTCCGATGCGCAG. The Camk2a-cre/ERT2 transgene was detected using forward primer GGTTCTCCGTTTGCACTCAGGA and reverse primer GCTTGCAGGTACAGGAGGTAGT.

2.3. Activation and Verification of Cre Recombinase Activity

Tsg101 deletion was induced at 12 weeks of age by twice-daily intraperitoneal injection of 0.025 mg tamoxifen (Sigma) per kg of body weight for three consecutive days, except for the oldest cohort (brains collected 140 days post-injection), which received twice-daily injections of 0.025 mg/kg for five consecutive days. Tamoxifen treated Tsg101^fl/fl; Camk2a-cre/ERT2(+) mice are referred to hereafter as Tsg101^ck2-null. Tamoxifen was dissolved to 5 μg/ml in a 1:9 mixture of absolute ethanol and pharmacological grade sunflower oil (Sigma). Control animals received injections of the 1:9 ethanol/oil vehicle mixture without tamoxifen. As Tsg101 deletion caused weight loss in Tsg101^fl/fl; Camk2a-cre/ERT2(+) mice, excessive weight loss (≥20% of initial body weight) was used as a humane end-point criterion. A liquid dietary supplement (Ensure Plus or Esbilac) was provided to treated animals as needed to support the maintenance of a healthy body weight.

Cre recombinase-mediated deletion of the floxed Tsg101 sequence was verified by PCR using the genotyping primers described above and a template of 10 ng genomic DNA isolated from dissected brain regions of 3 vehicle- and 3 tamoxifen-injected animals (2 weeks post-treatment), as well as from brains of 3 Camk2a-cre/ERT2 negative animals, as a further negative control. Reduction of TSG101 protein levels was confirmed by immunoblotting of brain lysates isolated two weeks after tamoxifen treatment (n=3 per genotype/treatment), using rabbit anti-TSG101 (Proteintech Cat# 14497-1-AP, RRID:AB_2208090). Brain lysates were prepared by homogenizing brain tissue in solubilization buffer (50 mM Tris-HCl pH 8.0, 1 mM EDTA, 1% Igepal CA-630) containing cOmplete Protease Inhibitor Cocktail (Roche). Blots were imaged using ECL Plus chemiluminescent substrate (Pierce).

2.4. Histology

Brains were fixed in formalin, processed for histology by standard methods and embedded in paraffin. Experimental and control brains were processed together in each batch, with care taken to avoid long exposure to 70% ethanol as this can cause artifactual vacuolation in rodent nervous tissue [49]. Paraffin-embedded brains were sectioned at 3 or 5 mm and stained with hematoxylin and eosin (H&E). For the 1- and 2-weeks-post-injection cohort, brains of at least 4 animals were examined for each genotype/treatment combination. Brains from at least 3 animals of each genotype/treatment combination were examined for later time points except 20 weeks, as only two Tsg101^ck2-null mice survived that long. Immunohistochemistry (IHC) was performed on paraffin sections, which were pretreated for antigen retrieval using 10mM sodium citrate (pH 6.0, 100 C, 10 minutes). Antibody information is provided in section 2.6. Immunolabelling was detected using fluorescent secondary antibodies (Vector Labs, CA), horseradish peroxidase (HRP)-conjugated secondary antibodies (anti-rabbit IgG-peroxidase, Sigma-Aldrich Cat# A0545, RRID:AB_257896), or HRP-conjugated avidin and biotinylated secondary antibodies of the appropriate isotype (Vector Labs, CA) and the peroxidase substrates diaminobenzidine (DAB) (Trevigen) and/or NovaRed (Vector Labs). Sections were counterstained with hematoxylin. For all studies, experimental and control samples were processed together to eliminate inter-experimental variability and at least 3 independent samples were examined for each genotype. Stained sections were imaged on a Zeiss AxioImagerM1 microscope using a PixieLink A623C color camera or on an Olympus Fluoview 1000 confocal microscope.

2.5. Immunoblotting

For western blotting, brain tissue was homogenized in RIPA lysis buffer (50 mM Tris, 150 mM NaCl, 1% NP40, 0.1% sodium deoxycholate) supplemented with cOmplete Protease Inhibitor Cocktail (Roche). Cellular debris was pelleted by centrifugation and the supernatant diluted in 2X SDS loading buffer. Proteins were electrophoresed through 8% or 12% SDS-polyacrylamide gels and transferred to Immobilon P membrane (Millipore). Immunoblotting was performed following standard protocols. Briefly, membranes were blocked with 5% nonfat dry milk in tris-buffered saline with Tween 20 (TTBS). Antibodies were diluted in TTBS and HRP-conjugated secondary antibodies (goat anti-rabbit IgG-peroxidase, Sigma-Aldrich Cat# A0545, RRID:AB_257896 or goat anti-mouse IgG, BD Biosciences Cat# 554002, RRID:AB_395198) were imaged using Clarity ECL Western Blotting Substrate (BioRad) and either a Bio-Rad ChemiDoc imaging system or the rapid capture setting on an Azure 300 Imager (Azure Biosystems). For LC3 western, the Azure software to verify the signal was not saturated, then bands were quantified using AzureSpot Pro (Azure Biosystems). Antibody information is provided in section 2.6.

2.6. Antibodies

Primary antibodies used were: rabbit anti-cleaved caspase-3 (Trevigen Cat# 2305-PC-100, RRID:AB_2665453), mouse anti-NeuN (Millipore Cat# MAB377, RRID:AB_2298772), rabbit anti-GFAP (Proteintech Cat# 16825-1-AP, RRID:AB_2109646), mouse anti-mono and polyubiuquitinated conjugates (Enzo Life Sciences Cat# BML-PW8810, RRID:AB_10541840), rabbit anti-RAB5 (LifeSpan Biosciences LS-C138527, now sold by Abcam, Cat# ab109534, RRID:AB_10865740), guinea pig anti-p62 (ARP American Research Products Cat# 03-GP62-C, RRID:AB_1542690), rabbit anti-LC3A (Novus Cat# NB100-2331, RRID:AB_10001955), rabbit anti-LC3 (RRID:AB_2716621, a gift from Dr. Masahiro Shibata), mouse anti-RAB7 (Abcam Cat# ab58029, RRID:AB_945127), and mouse monoclonal 4G8 (BioLegend Cat#800701, RRID:AB_2564633) which is reactive to amino acids 17-24 of Aβ and to APP. For immunoblotting, mouse anti-beta-tubulin-III (3F3-G2) (Santa Cruz Biotechnology Cat# sc-53140, RRID:AB_793543) was used as a loading control.

2.6. Statistics

Brain weight, body weight, food consumption, and protein levels were assessed for statistically significant differences using Student’s t-test. Minimum sample sizes were selected to provide at least 90% power of detecting a significant difference (p = 0.05) if at least 90% of animals within a genotype/treatment group gave a consistent phenotype or result. As some of the Tsg101^ck2-null mice were lost within 14 days of Tsg101 deletion, an excess number of Tsg101^fl/fl; Camk2a-cre/ERT2(+) mice were treated with tamoxifen to collect the desired number of mice at later time-points.

3. Results

Tsg101 was ablated in forebrain neurons of adult Tsg101 conditional knockout mice carrying the Camk2a-cre/ERT2 transgene by treating them with tamoxifen at 12 weeks of age (Figure 1A). This cre transgene has previously been shown to have low activity in the hippocampus in the absence of tamoxifen and robust activity, especially in neurons of the cortex, dentate gyrus, and hippocampus upon tamoxifen treatment [48]. Initially, Tsg101^fl/fl; Camk2a-cre/ERT2+ and control (Tsg101^fl/fl; Camk2a-cre/ERT2- and Tsg101^fl/-; Camk2a-cre/ERT2+) mice were given twice daily IP injections of tamoxifen or vehicle control (1:9 ethanol/sunflower oil) for 5 days. Deletion of Tsg101 in tamoxifen-treated mice was verified by PCR on DNA isolated from dissected regions of the CNS and the sciatic nerve using primers that distinguish the endogenous wild-type allele, the floxed allele, and the recombined (deleted) allele (Figure 1B). Immunoblotting confirmed a reduction in TSG101 protein levels in the brains of tamoxifen-treated Tsg101^fl/fl; Camk2a-cre/ERT2+ mice, referred to hereafter as Tsg101^ck2-null (Figure 1C). In initial studies, these mice showed significant, progressive weight loss and reduced food intake relative to controls, starting 9 days after the initiation of tamoxifen treatment (Figure 2A,B). By 13-14 days post-tamoxifen, Tsg101^ck2-null mice lost ≥ 20% of their body weight and were humanely euthanized. Home cage observation indicated that reduced food consumption coincided with onset of a head “shaking” or bobbing behavior, which may have made it difficult for the mice to feed from the overhead hopper or on pieces of food on the cage floor, but they also appeared to generally lose interest in even trying to eat. The tamoxifen treatment regime was reduced to 2 injections per day for 3 consecutive days and Tsg101^ck2-null mice were offered a pipette of liquid Ensure meal replacement or Esbilac supplement every 3-4 hours for ~2 weeks, starting 8 days after the initiation of tamoxifen injections. Under these conditions, the anorexia and weight loss experienced by Tsg101^ck2-null mice appeared to be transient as, after liquid oral supplementation, most of the mice resumed appropriate feeding behavior and weight gain, although some still appeared frail and had reduced lifespan. As brain histology did not discern any differences between Tsg101^ck2-null mice generated using either tamoxifen treatment regime, samples were used interchangeably.

Twenty weeks after tamoxifen treatment was initiated, Tsg101^ck2-null mice showed a substantial reduction in forebrain size, apparent even upon gross inspection (Figure 3A,B). Hematoxylin and eosin (H&E)-stained coronal sections revealed marked reductions in the volume of the hippocampal formation (Figure 3C–H) and cortex layer 1 (Figure 3I–K). The granule cell layers of the dentate gyrus and hippocampus were much thinner in Tsg101^ck2-null brains (Figure 3C–H), consistent with loss of neurons. To investigate the time course of neurodegeneration, Tsg101^ck2-null and control brain sections were examined 1, 2, 4, and 6 weeks following tamoxifen treatment, using H&E staining to assess overall morphology and IHC for glial fibrillary acidic protein (GFAP) to detect reactive astrogliosis. One week after tamoxifen treatment, Tsg101^ck2-null brains showed a small number of neurons with pyknotic nuclei in the CA3 subfield of the hippocampus and elevated GFAP immunoreactivity surrounding the granule cell layer of the dentate gyrus, indicative of reactive astrogliosis. The latter suggests neuronal homeostasis was disrupted in the dentate gyrus, despite no overt neuron loss (Figure 4A–D). Within 2 weeks of tamoxifen treatment, Tsg101^ck2-null brains showed extensive loss of NeuN-positive neurons in the CA3 subfield of the hippocampus, associated with the presence of pyknotic nuclei (Figure 4E–J). Astrogliosis was still concentrated in the dentate gyrus, with weaker GFAP staining present more broadly throughout the hippocampal formation (Figure 4K,L). By 4 weeks after tamoxifen treatment, the pyramidal layers of all regions of the hippocampus as well as the dentate gyrus were thin and showed the presence of pyknotic nuclei (Figure 4M,N), and GFAP staining remained strongest in the dentate gyrus (Figure 4O,P). By the 6-week time point, only a thin layer of neurons remained in the granule cell layers of the hippocampus and dentate gyrus (Figure 4Q,R). At this time-point, astrogliosis was observed throughout the hippocampal formation and the dentate gyrus continued to show stronger GFAP staining than other regions (Figure 4S,T). Although cortex layer 1 was significantly thinner in the brains of Tsg101^ck2-null mice 20 weeks after tamoxifen treatment, no difference was observed up to the 6-week time point (Figure 4U,V). Consistent with a significant loss of neurons, brains of Tsg101^ck2-null mice weighed significantly less than those of control mice by the 2-week time point (Figure 4W) and atrophy was progressive, as indicated by the significant reduction in the brain weight of Tsg101^ck2-null mice between the 2 and 6 week time points (Figure 4W). By 6-weeks post-tamoxifen, there was a ~20% reduction in total brain mass in Tsg101^ck2-null mice relative to controls.

Degenerating pyramidal cells in the CA3 region of the hippocampus of Tsg101^ck2-null mice showed morphological and immunohistochemical evidence of apoptosis, such as an abundance of pyknotic nuclei (Figure 4H,J) and expression of the cleaved (active) form of the apoptosis effector caspase-3 (CC3) (Figure 5A–C). CA3 pyramidal cells also showed increased staining for mono- and poly-ubiquitinated proteins (Figure 5D–F), consistent with impaired sorting of ubiquitinated proteins into MVBs, a process that includes substrate deubiquitination. As loss of TSG101 function has been reported to up-regulate autophagy and dysregulated autophagy contributes to caspase-dependent neuronal apoptosis in cultured cells [51,52], we examined expression of the autophagy adapter protein, p62/SQSTM1 (Sequestosome-1), as well as LC3 (microtubule-associated protein 1A/1B-light chain 3), which undergoes conversion from LC3-I to LC3-II via the addition of a phosphatidylethanolamine group to the C terminus prior to being recruited to autophagosomes. In Tsg101^ck2-null brains collected 1 week following the initiation of tamoxifen treatment, SQSTM1-positive punctae were observed in some pyramidal cells in the dentate gyrus, and more frequently in the hippocampal CA3 region (Figure 5G–K). As SQSTM1 is degraded through autophagy, elevated levels could indicate impairment of this pathway. IHC for LC3B did not reveal punctae of staining in Tsg101^ck2-null brains (data not shown). Immunoblotting indicated reduced expression of LC3-I in Tsg101^ck2-null hippocampi but a band consistent in size with LC3-II was not detected, even on long exposures or on blots of subcellular fractions expected to contain autophagosomes, prepared as described by Lai et al. [53] (Figure 5L,M and data not shown). This result was replicated using two different antibodies against LC3. The reduced expression of LC3-I was unlikely to be due to a loss of neurons as expression was normalized against a neuron-specific marker, TUJ1. Taken together, these data suggest that either autophagy was not upregulated in Tsg101^ck2-null neurons, or that it was a very early and transient effect since no LC3-II could be detected within one week of induction of Tsg101 deletion.

Our mouse model provided the opportunity to examine, in vivo, whether neuronal depletion of Tsg101 leads to endosomal defects similar to those observed in AD brains. IHC on coronal sections from Tsg101^ck2-null and control brains collected 2-weeks post-tamoxifen revealed accumulation of RAB5 and RAB7 in hippocampal CA3 neurons (Figure 6A–D). Staining using an antibody reactive to APP amino acids 17-24 (clone 4G8), thus recognizing APP and Aβ, revealed more abundant intraneuronal staining in Tsg101^ck2-null brains relative to controls (Figure 6E–H), consistent with impaired trafficking of APP/Aβ.

4. Discussion

The results of this study demonstrate that deleting Tsg101 from Camk2a-expressing neurons led to severe, progressive neurodegeneration. Within 1 week of Tsg101 deletion, significant GFAP staining was observed around the granule cell layer of the dentate gyrus and remained most pronounced in this region through all stages examined. Astrogliosis in the dentate gyrus would disrupt its ability to filter information passing through to the CA3 region of the hippocampus from the entorhinal cortex, which can lead to increased neuronal activity in the CA3 due to reduced recycling of glutamate and potassium from the synaptic cleft by activated astrocytes [54,55,56,57,58]. Hippocampal CA3 neurons showed caspase-dependent apoptosis within two-weeks of deleting Tsg101, consistent with astrogliosis in the dentate gyrus leading to excitotoxicity and apoptosis in the CA3. Although embryonic fibroblasts showed a transient upregulation of autophagy prior to apoptosis following the conditional deletion of Tsg101 [59], we did not detect lipidated LC3 (LC3-II) in Tsg101^ck2-null brains 1 week after tamoxifen treatment was initiated. This suggests that, in vivo, either autophagy is not upregulated in neurons as a result of loss of Tsg101, or it occurs early and is very brief. The presence of reactive astrocytes in the dentate gyrus may explain the difference in cell death pathways triggered in neurons in culture and neurons in vivo. Reactive astrogliosis in the dentate gyrus could reflect either local effects of loss of Tsg101 from neurons in this region or disruption of neuronal signaling in the entorhinal cortex, which projects to and can induce astrocyte activation in the dentate gyrus. The entorhinal cortex is believed to play a role in the early stages of AD and often shows the earliest histopathological alterations, including the presence of neurofibrillary tangles and cell death [60]. We hypothesize that CA neurons may be the earliest to die in Tsg101^ck2-null mice because the dentate gyrus provides excitatory inputs to the CA3, and impaired glutamate uptake by activated astrocytes in the dentate gyrus would be predicted to lead to excitotoxicity and apoptosis in CA3 neurons.

Endosomal abnormalities are one of the earliest pathological features of AD, characterized by enlarged RAB5-positive early endosomes and RAB7-positive late endosomes and progressive accumulation of MVBs, lysosomes and autophagosomes [4,8,61]. APP is known to traffic through the endosomal system, and the accumulation of APP and/or its N-terminal proteolytic derivatives in Tsg101^ck2-null hippocampal neurons supports a role for the ESCRT pathway in normal APP trafficking. Our in vivo results are consistent with in vitro studies that demonstrated inhibited sorting of APP onto ILVs and to the lysosome, with intracellular accumulation of Aβ and a concomitant decrease in Aβ secretion when Tsg101 was depleted [26]. It is not clear whether accumulation of APP in Tsg101^ck2-null neurons was due to reduced lysosomal degradation or impaired exosomal release, as both can be MVB-dependent. Exosomes isolated from mouse brain are enriched with toxic APP fragments [36], suggesting exosomal secretion may clear them from cells, but also deposit them into the extracellular space where they may contribute to Aβ plaques. Mice homozygous for the human apolipoprotein E4 (ApoE4) variant, which is the greatest genetic risk factor for AD in people, showed an age-dependent decrease in exosomes associated with reduced levels of TSG101 and RAB35 [62], both of which are involved in exosome biosynthesis and release [24,63]. The role of exosomes and other EVs in AD pathogenesis remains to be fully elucidated.

Generally, ESCRT dysfunction is associated with neurodegeneration. Mutations in the gene encoding the ESCRT-III component Charged Multivesicular Body Protein 2B (CHMP2B) are associated with frontotemporal dementia (FTD), FTD with motor neuron disease, and Amyotrophic Lateral Sclerosis (ALS) [64,65,66]. Studies using transgenic mice suggest Chmp2b mutations act through a gain-of-function mechanism [67,68]. Deleting the gene encoding the ESCRT-0 protein, HGF-regulated tyrosine kinase substrate, HGS (formerly referred to as HRS), from Synapsin-1-expressing neurons in mice resulted in apoptotic loss of hippocampal CA3 neurons and reduced locomotor activity and learning ability [69], but their phenotype was less severe than that observed in Tsg101^ck2-null mice. Deleting Hgs from mouse forebrain neurons using a constitutive Camk2a-cre resulted in a phenotype more similar to what we report here for Tsg101^ck2-null mice, including a marked accumulation of ubiquitinated proteins by 5 weeks of age and apoptotic and necrotic loss of hippocampal neurons by 7 weeks of age, with the CA3 region being most severely affected [70]. We have also demonstrated that deleting Tsg101 from Schwann cells in the mouse peripheral nervous system resulted in severe, rapid onset peripheral neuropathy associated with “onion bulb” formations and de/dysmyelination [71]. Mice lacking Hgs in Schwann cells displayed a much milder phenotype, displaying mild motor and sensory defects, fewer myelinated axons, and thinner and aberrantly folded myelin sheaths in the sciatic nerve [72]. Our group also previously demonstrated that conditional ablation of Tsg101 in oligodendroglia of the central nervous system (CNS) resulted in severe and rapid onset spongiform neurodegeneration and de/dys-myelination [73]. In the same study, we showed that oligodendroglial deletion of the gene encoding RAB7 did not produce any histopathological abnormalities, suggesting that the severe phenotypes associated with TSG101 deficiency were not due to defects in late endosomal, lysosomal, or autophagosomal functions, which are RAB7-dependent [74,75,76]. Taken together, these studies demonstrate critical roles for TSG101 in a variety of nervous system cell types.

Although HGS interacts with TSG101 to recruit ESCRT-I to endosomes [77,78,79], in multiple mammalian cell lines, knockdown of TSG101 or HGS had significantly different effects on epidermal growth factor receptor (EGFR) trafficking and degradation and endosomal and MVB morphology [59,77,78,80,81]. Depleting TSG101 resulted in the formation of multicisternal early endosomes and defects in protein sorting, as well as endoplasmic reticulum (ER) stress, ER membrane remodeling, and upregulation of autophagy [59,80,81,82,83]. Down-regulation of TSG101 has also been reported to result in reduced secretion of exosomes [24,84]. Depleting HGS did not induce early endosome tubulation and led to the production of enlarged MVBs containing fewer ILVs [17,80,82]. Thus, TSG101 appears to be required for the formation of stable vacuolar domains within the early endosome that subsequently develop into MVBs, while HGS is more important in the formation and/or accumulation of ILVs within MVBs. TSG101 deletion also disrupts the recycling of membrane proteins by decoupling early endosomes from recycling endosomes [22]. Given their roles in MVB production, it is not surprising that HGS and TSG101 are both involved in exosome production, with HGS depletion resulting in smaller exosomes and TSG101 depletion leading to fewer exosomes and a concomitant increase in other EV subtypes [24,80].

5. Conclusions

The phenotype we observed in Tsg101^ck2-null mice validated some findings from cell culture studies, such as the intracellular accumulation of APP and/or its proteolytic derivatives in Tsg101-depleted neurons. Other findings in the whole brain contrasted with observations in cell culture models. For example, we did not observe lipidated LC3 in vivo within 1 week of Tsg101 deletion, which was prior to detecting autophagy in hippocampal CA3 neurons, suggesting that if autophagy was upregulated in vivo, it was not sustained. Assessing neuronal depletion of Tsg101 in the context of the whole brain also provided important insight into the influence of neuronal connectivity in disease pathogenesis and progression. Although all Camk2a-expressing forebrains neurons lost Tsg101, hippocampal CA3 pyramidal neurons were the first to undergo apoptosis and showed a more robust accumulation of ubiquitinated proteins and p62/SQTM1 by the 2-week timepoint, whereas the dentate gyrus showed the earliest and consistently most pronounced reactive astrocytosis. We hypothesize that depleting Tsg101 from neurons in the dentate gyrus disrupted neurotrophic signaling, triggering reactive astrogliosis to support neuronal survival and repair. The ability of activated astrocytes to execute normal functions, such as uptake of excess glutamate from the synaptic cleft, would be impaired and we propose that this leads to excitotoxicity and apoptosis in CA3 neurons. Our results stress the importance of in vivo studies to fully assess gene/protein function, especially for cells that contribute to organs like the brain, where communication between cells and different cell types is crucial.

These studies provide important proof-of-concept data toward the elucidation of complex neurodegenerative pathways in diseases such as AD. Endolysosomal abnormalities have been noted in AD-affected brain tissue and are hypothesized to contribute to the pathology of AD. Our results lend support to this hypothesis by demonstrating that impairment of the endo-lysosomal sorting machinery at TSG101 is sufficient to induce neurodegeneration in the hippocampal neurons that are most susceptible in AD, as well as to recapitulate a subset of features of AD pathology, including the accumulation of endosomal markers and Aβ. Finally, these studies demonstrate that TSG 101-dependent trafficking in neurons is critical for maintaining neuronal homeostasis.