Neuroligins (NLGNs) organize neuronal connectivity by engaging a diverse set of interaction partners, yet how extracellular recognition couples to intracellular growth programs remains unclear. Using affinity proteomics, we identify intercellular adhesion molecule-5 (ICAM5), a cell-surface protein localized to dendritic filopodia, as a novel neuroligin interactor. Surface plasmon resonance and cell-based assays demonstrate direct binding between the ICAM5 and NLGN3 extracellular domains and reveal that ICAM5 engages all neuroligin isoforms. ICAM5 is required for NLGN-induced dendritic outgrowth, but the NLGN3–ICAM5 complex does not contribute to synaptogenesis. Mechanistically, ICAM5 sustains PAK–Cofilin signaling and F-actin organization in growth cones, positioning it as a downstream effector that links neuroligin engagement to actin remodeling. Together, these findings define a neuroligin–ICAM5 axis that couples extracellular recognition to intracellular actin remodeling to control neuronal structural development.​

Decades of research have linked lysosomal failure to Alzheimer’s disease, but the role of ageing has remained unknown, hindered by the absence of in vitro models that combine both ageing and Alzheimer’s. A study now identifies how disrupted lysosome repair fuels dysproteostasis and inflammation in both aged and diseased neurons.

The subcellular localization of γ-secretase is to a large extent defined by the presenilin homologue: a unique sorting motif restricts PSEN2/γ-secretase to late endosomes/lysosomes (LE/Lys) whereas PSEN1 complexes reside at the cell surface and in endosomal compartments. While these distinct locations promote substrate specificity and different amyloid β pools, they also suggest additional roles in maintaining organellar homeostasis. Using new APP knockin models with altered PSEN2 expression, we recently correlated impairments in working memory and LTP to PSEN2's role in neuronal endolysosomal homeostasis (Perdok et al., 2024). Currently we are exploring in detail how altered PSEN2 expression molecularly affects LE/Lys functions and dynamics. We used nonneuronal, primary hippocampal and iPSC-derived human neurons that are either deficient in PSEN2 or gene-edited to express the FAD-linked PSEN2 N141I mutation. PSEN-deficient cells stably expressing APEX2-tagged PSEN2 were used for proximity-dependent biotinylation to identify the local PSEN2 interactome in LE/Lys. Hits were prioritized based on their relevance to organellar homeostasis and validated using biochemical assays and functional studies in neuronal models. Proximity labeling identified novel interactors connecting PSEN2 function to LE/Lys transport regulation and nutrient sensing. Furthermore, super-resolution microscopy was used to more precisely identify the nano-domain organization of PSEN2 at the limiting membrane of LE/Lys. In neurons, axonal LE/Lys motility is differently affected in PSEN2KO versus FAD-mutant PSEN2, suggesting a more complex toxic mechanism in disease context. We are currently exploiting microfluidic chambers combined with omics for a more comprehensive view on the impact of altered PSEN2 expression on the axonal compartment. Our data revealed new insights in how PSEN2/γ-secretase may regulate LE/Lys homeostasis, providing novel starting points to better understand the etiology of endolysosomal dysfunctions as observed at early, preclinical stages of AD. Notably, specific FAD-mutations also re-locate PSEN1/γ-secretase to LE/Lys, underscoring that knowledge gained here could aid to understand the more heterogeneous PSEN1 mutations.

Autosomal Dominant Alzheimer's Disease (ADAD), caused by mutations in Presenilins (PSEN1/2) and Amyloid Precursor Protein (APP) genes, typically manifests with early onset (< 65 years). Age at symptom onset (AAO) is relatively consistent among carriers of the same PSEN1 mutation, but more variable for PSEN2 and APP variants, with these mutations associated with later AAOs than PSEN1. Understanding this clinical variability is crucial for understanding disease mechanisms, developing predictive models and tailored interventions in ADAD, with potential implications for sporadic AD. We performed biochemical assessment of γ-secretase dysfunction on 28 PSEN2 and 19 APP mutations, including disease-associated, unclear and benign variants. This analysis has been valuable in the assessment of PSEN1 variant pathogenicity, disease onset and progression. Our analysis reveals linear correlations between the molecular composition of Aβ profiles and AAO for both PSEN2 (R2 = 0.52) and APP (R2 = 0.69) mutations. The integration of PSEN1, PSEN2 and APP correlation data shows parallel but shifted lines, suggesting a common pathogenic mechanism with gene-specific shifts in onset. We found overall “delays” in AAOs of 27 years for PSEN2 and 8 years for APP variants, compared to PSEN1. Notably, extremely inactivating PSEN1 variants delayed onset, suggesting that reduced contribution to brain APP processing underlies the later onset of PSEN2 variants. This study supports a unified model of ADAD pathogenesis wherein γ-secretase dysfunction and the resulting shifts in Aβ profiles are central to disease onset across all causal genes. While similar shifts in Aβ occur across causal genes, their impact on AAO varies in the function of their contribution to APP processing in the brain. This biochemical analysis establishes quantitative relationships that enable predictive AAO modelling with implications for clinical practice and genetic research. Our findings also support the development of therapeutic strategies modulating γ-secretase across different genetic ADAD forms and potentially more broadly in AD.

Mutations in Presenilin-2 (PSEN2), a catalytic subunit of the γ-secretase complex, cause familial Alzheimer’s disease (AD). Beyond its role in β-amyloid generation, we hypothesized that PSEN2 also supports key cellular functions whose disruption may underlie neuronal vulnerability. We now demonstrate that PSEN2 in late endosomes/lysosomes (LE/Lys) specifically localizes at contact sites with the endoplasmic reticulum (ER), implicating it in inter-organellar communication. Using proximity-dependent biotinylation of APEX2-tagged PSEN2, we defined its local interactome and identified a novel interaction with Lamtor1 –a regulator of vesicle trafficking and mTORC1 signaling– within LE/Lys-ER contacts. Mechanistically, altered PSEN2 expression disrupted Lamtor1-dependent signaling, promoting the recruitment of the BORC complex at LE/Lys, thereby driving their repositioning, while also affecting nutrient-dependent mTORC1 activity. These findings are recapitulated in mouse hippocampal neurons, uncovering a novel role for PSEN2 in maintaining organelle homeostasis with relevance to early events in AD pathogenesis.

Proteolytic processing of the amyloid precursor protein (APP) generates not only the well-known β-amyloid (Aβ) peptides but also APP C-terminal fragments (APP-CTFs). Recent evidence from studies in murine- or human-derived (neuronal) models suggests that APP-CTFs may independently contribute to Alzheimer's disease (AD) pathology by disrupting cellular homeostasis. This review highlights pathological effects unique to APP-CTFs that are independent of Aβ, shedding light on their distinct role in disease progression. We explore the mechanisms underlying APP-CTF-induced toxicity, with a focus on their contribution to endolysosomal dysfunction. APP-CTFs impair lysosomal function and disrupt calcium signaling between the endoplasmic reticulum and lysosomes, compounding organelle dysfunction. Understanding these mechanisms will aid the design of preventive therapeutic strategies that take into account the impact of APP-CTFs on AD pathology.

Background:Over 300 mutations in PSEN1 have been identified as causes of early-onset Alzheimer's disease (EOAD). While these include missense mutations and a few insertions, deletions, or duplications, none result in open reading frame shifts, and all alter γ-secretase function to increase the long/short Aβ ratio. Methods:We identified a novel heterozygous PSEN1 nonsense variant, c.325A > T, in a patient and his father, both presenting with EOAD, resulting in the substitution of lysine 109 with a premature stop codon at position (p.K109*). This produces a truncated 109 amino acid (aa) N-terminal PSEN1 fragment. Functional characterization was performed using overexpression models and a heterozygous mouse model (Psen1K109*/+). Results:In overexpression models, downstream ATGs serve as alternative starting codons, generating a > 37kDa and a > 27 kDa PSEN1 C-terminal fragment (PSEN1-CTFA and PSEN1-CTFB, respectively) that retain the two catalytic aspartates of γ-secretase. Heterozygous Psen1K109*/+ mice exhibited subtle phenotypic defects, including reduced Pen2 expression and mild APP-CTF accumulation. Notably, aged mice demonstrated significantly increased Psen2 protein expression, potentially contributing to an elevated Aβ42/Aβ38 ratio. Conclusions:These findings indicate that PSEN1 c.325A > T (p.K109*) is not a complete loss-of-function mutation. However, to what extent and by what mechanism it contributes to EOAD pathogenesis remains unclear.

Defects in protein homeostasis can induce proteotoxic stress, affecting cellular fitness and, consequently, overall tissue health. In various growing tissues, cell competition based mechanisms facilitate detection and elimination of these compromised, often referred to as 'loser', cells by the healthier neighbors. The precise connection between proteotoxic stress and competitive cell survival remains largely elusive. Here, we reveal the function of an endoplasmic reticulum (ER) and Golgi localized protein Rer1 in the regulation of protein homeostasis in the developing Drosophila wing epithelium. Our results show that loss of Rer1 leads to proteotoxic stress and PERK-mediated phosphorylation of eukaryotic initiation factor 2α. Clonal analysis showed that rer1 mutant cells are identified as losers and eliminated through cell competition. Interestingly, we find that Rer1 levels are upregulated upon Myc-overexpression that causes overgrowth, albeit under high proteotoxic stress. Our results suggest that increased levels of Rer1 provide cytoprotection to Myc-overexpressing cells by alleviating the proteotoxic stress and thereby supporting Myc-driven overgrowth. In summary, these observations demonstrate that Rer1 acts as a novel regulator of proteostasis in Drosophila and reveal its role in competitive cell survival.

Presenilin proteins (PS1 and PS2) represent the catalytic subunit of γ-secretase and play a critical role in the generation of the amyloid β (Aβ) peptide and the pathogenesis of Alzheimer disease (AD). However, PS proteins also exert multiple functions beyond Aβ generation. In this study, we examine the individual roles of PS1 and PS2 in cellular cholesterol metabolism. Deletion of PS1 or PS2 in mouse models led to cholesterol accumulation in cerebral neurons. Cholesterol accumulation was also observed in the lysosomes of embryonic fibroblasts from Psen1-knockout (PS1-KO) and Psen2-KO (PS2-KO) mice and was associated with decreased expression of the Niemann-Pick type C1 (NPC1) protein involved in intracellular cholesterol transport in late endosomal/lysosomal compartments. Mass spectrometry and complementary biochemical analyses also revealed abnormal N-glycosylation of NPC1 and several other membrane proteins in PS1-KO and PS2-KO cells. Interestingly, pharmacological inhibition of N-glycosylation resulted in intracellular cholesterol accumulation prominently in lysosomes and decreased NPC1, thereby resembling the changes in PS1-KO and PS2-KO cells. In turn, treatment of PS1-KO and PS2-KO mouse embryonic fibroblasts (MEFs) with the chaperone inducer arimoclomol partially normalized NPC1 expression and rescued lysosomal cholesterol accumulation. Additionally, the intracellular cholesterol accumulation in PS1-KO and PS2-KO MEFs was prevented by overexpression of NPC1. Collectively, these data indicate that a loss of PS function results in impaired protein N-glycosylation, which eventually causes decreased expression of NPC1 and intracellular cholesterol accumulation. This mechanism could contribute to the neurodegeneration observed in PS KO mice and potentially to the pathogenesis of AD.

Rare mutations in the gene encoding presenilin2 (PSEN2) are known to cause familial Alzheimer's disease (FAD). Here, we explored how altered PSEN2 expression impacts on the amyloidosis, endolysosomal abnormalities, and synaptic dysfunction observed in female APP knock-in mice. We demonstrate that PSEN2 knockout (KO) as well as the FAD-associated N141IKI mutant accelerate AD-related pathologies in female mice. Both models showed significant deficits in working memory that linked to elevated PSEN2 expression in the hippocampal CA3 region. The mossy fiber circuit of APPxPSEN2KO and APPxFADPSEN2 mice had smaller pre-synaptic compartments, distinct changes in synaptic vesicle populations and significantly impaired long term potentiation compared to APPKI mice. At the cellular level, altered PSEN2 expression resulted in endolysosomal defects and lowered surface expression of synaptic proteins. As PSEN2/gamma-secretase is restricted to late endosomes/lysosomes, we propose PSEN2 impacts endolysosomal homeostasis, affecting synaptic signaling in AD-relevant vulnerable brain circuits; which could explain how mutant PSEN2 accelerates AD pathogenesis. This latest research reveals that changes in PSEN2 expression accelerate amyloid build-up, disrupt lysosomal function, and weaken synaptic connections. These findings spotlight PSEN2 as a powerful target for early Alzheimer's disease intervention.

Neurodegenerative diseases encompass a spectrum of conditions characterized by the gradual deterioration of neurons in the central and peripheral nervous system. While their origins are multifaceted, emerging data underscore the pivotal role of impaired mitochondrial functions and endolysosomal homeostasis to the onset and progression of pathology. This article explores whether mitochondrial dysfunctions act as causal factors or are intricately linked to the decline in endolysosomal function. As research delves deeper into the genetics of neurodegenerative diseases, an increasing number of risk loci and genes associated with the regulation of endolysosomal and autophagy functions are being identified, arguing for a downstream impact on mitochondrial health. Our hypothesis centers on the notion that disturbances in endolysosomal processes may propagate to other organelles, including mitochondria, through disrupted inter-organellar communication. We discuss these views in the context of major neurodegenerative diseases including Alzheimer's and Parkinson's diseases, and their relevance to potential therapeutic avenues.

Neuronal endosomal and lysosomal abnormalities are among the early changes observed in Alzheimer’s disease (AD) before plaques appear. However, it is unclear whether distinct endolysosomal defects are temporally organized and how altered γ-secretase function or amyloid precursor protein (APP) metabolism contribute to these changes. Inhibiting γ-secretase chronically, in mouse embryonic fibroblast and hippocampal neurons, led to a gradual endolysosomal collapse initiated by decreased lysosomal calcium and increased cholesterol, causing downstream defects in endosomal recycling and maturation. This endolysosomal demise is γ-secretase dependent, requires membrane-tethered APP cytoplasmic domains, and is rescued by APP depletion. APP C-terminal fragments (CTFs) localized to late endosome/lysosome-endoplasmic reticulum contacts; an excess of APP-CTFs herein reduced lysosomal Ca2+ refilling from the endoplasmic reticulum, promoting cholesterol accretion. Tonic regulation by APP-CTFs provides a mechanistic explanation for their cellular toxicity: failure to timely degrade APP-CTFs sustains downstream signaling, instigating lysosomal dyshomeostasis, as observed in prodromal AD. This is the opposite of substrates such as Notch, which require intramembrane proteolysis to initiate signaling.

Phospholipase D3 (PLD3) polymorphisms are linked to late-onset Alzheimer's disease (LOAD). Being a lysosomal 5'-3' exonuclease, its neuronal substrates remained unknown as well as how a defective lysosomal nucleotide catabolism connects to AD-proteinopathy. We identified mitochondrial DNA (mtDNA) as a major physiological substrate and show its manifest build-up in lysosomes of PLD3-defective cells. mtDNA accretion creates a degradative (proteolytic) bottleneck that presents at the ultrastructural level as a marked abundance of multilamellar bodies, often containing mitochondrial remnants, which correlates with increased PINK1-dependent mitophagy. Lysosomal leakage of mtDNA to the cytosol activates cGAS-STING signaling that upregulates autophagy and induces amyloid precursor C-terminal fragment (APP-CTF) and cholesterol accumulation. STING inhibition largely normalizes APP-CTF levels, whereas an APP knockout in PLD3-deficient backgrounds lowers STING activation and normalizes cholesterol biosynthesis. Collectively, we demonstrate molecular cross-talks through feedforward loops between lysosomal nucleotide turnover, cGAS-STING and APP metabolism that, when dysregulated, result in neuronal endolysosomal demise as observed in LOAD.

γ-Secretase is an intramembrane protease which plays a pivotal role in the onset and progression of Alzheimer’s disease (AD). Presenilin provides the catalytic activity of which two homologues exist, Presenilin-1 and Presenilin-2 (Psen1/2). Psen2/γ-secretase complexes are restricted in their localization to the late endosomes and lysosomes, as opposed to the broader distribution of PSEN1/ γ-secretase, making it the main generator of intracellular abeta (Aβ). Although the focus has been on the extracellular pool, this toxic intracellular pool has been shown to precede plaque and tangle formation and correlates well with synaptic dysfunction highlighting its importance. This project aims to unravel the effects of altered Psen2 expression and this intracellular pool on AD pathogenesis. We generated novel mouse models by crossing the well characterized APP NL-G-F model with a Psen2 knock out (KO) mouse and an in-house generated Psen2 knock in (KI) model carrying the familial AD-linked N141I mutation (FAD-Psen2). Curiously, opposed to expectations, we found an identical accelerated plaque pathology coinciding with the presence of dystrophic neurites and amyloid precursor protein c-terminal fragment (APP-CTF) accumulation in both genotypes. These alterations in pathology equally translated to an earlier deficit in working memory for both APPxPsen2KO and APPxFADPsen2. In contrast, gliosis was markedly different with increased and earlier microglia recruitment in the case of APPxPsen2KO whereas APPxFAD-Psen2 displayed a delayed recruitment. These surprising differential effects on distinct pathological features suggest both a protective role for Psen2 as well as specific, yet unexplored, roles in neurons as well as glial cells. We are currently examining primary neurons and microglia from the different genotypes to explore underlying molecular mechanisms.

Cell competition is a developmental phenomenon that allows the selection of healthier cells in a developing tissue. In this process, cells with reduced fitness, conceivably due to harmful mutations, acquire the ‘loser’ status and are eliminated by the fitter (winner) neighboring cells via juxtacrine cell-cell interactions. How various mutations trigger cell competition is an extensively studied question. However, the mechanism of cell competition remains largely elusive. In this study, we reveal previously unknown functions of an ER and Golgi localized protein Rer1 in the regulation of cell competition in the developing Drosophila wing epithelium. Our data show that loss of Rer1 leads to the proteotoxic stress marked by the increased phosphorylation of eIF2α. The increased proteotoxic stress in the rer1 mutant cells led to their elimination via cell competition. Interestingly, we find that Rer1 levels are upregulated upon Myc overexpression, which generates super-competitive cells that overgrow at the expense of the normal neighboring cells. Loss of Rer1 also restricts the growth of Myc-induced super-competitive cells. Moreover, consistent with its known function as a negative regulator of the Notch pathway, our results show that the increased levels of Rer1 in Myc-overexpression led to the downregulation of Notch activity. In summary, these observations provide the first characterization of Rer1 in Drosophila and reveal its role in triggering cell competition via the regulation of proteotoxic stress.

Phospholipase D3 (PLD3) is a single-pass type II membrane protein that is majorly localized to late endosomes/lysosomes (LE/Lys) where it functions as a 5'-3' exonuclease that degrades ssDNA. Whereas PLD3 has been identified as risk factor for late-onset Alzheimer’s disease (LOAD), knockout studies have resulted in conflicting data regarding its impact on the amyloid precursor protein metabolism. Moreover, the substrates for PLD3 in LE/Lys remain unexplored CRISPR/Cas9 gene editing was used to generate PLD3 knockout (KO) SH-SY5Y cells that were subsequently stably rescued with wild-type PLD3 and coding-variants (M6R & V232M). All cell lines were evaluated for morphological and functional alterations of the LE/Lys compartment, including lipid profiling and analysis of the nucleotide content of LE/Lys magnetically isolated from the different cell lines. We optimized a PLD3 activity assay and analyzed the LE/Lys nucleotide content as a means to identify PLD3 substrates. We report that a PLD3 KO and LOAD-associated risk variants lead to a significant lysosomal pathology, comprising an increase in the size of LE/Lys, more LAMP1 immunoreactivity, a reduced catabolic activity, an increased propensity for leakage and a marked increase in cholesterol content and in storage lipids. At the ultrastructural level, this is accompanied with the appearance of multilamellar bodies, sometimes including mitochondrial remnants. This is suggestive for a defective mitophagy, which we confirm on the light microscopy level using mKeima assays. Furthermore, we provide support for a role of PLD3 in the homeostatic degradation of mtDNA. We argue that when PLD3’s role is compromised, this leads to a build-up of the lysosomal nucleotide content, impacting on the normal functioning of LE/Lys. We are currently analysing alterations in downstream cellular signaling pathways, including the TLR9 cascade. Collectively, our findings subscribe a key role for PLD3 exonuclease activity in lysosomal homeostasis. Its dysfunction, as occurs in LOAD, leads to the co-occurrence of lysosomal and mitochondrial defects.

Alzheimer’s disease (AD) is the most common form of dementia of which the incidence increases with age. AD has a strong genetic component, for which further study is essential to identify pathways that can be targeted therapeutically, as no cure is available yet. While most studies have focused over the years on the early onset risk factors (i.e. APP, PSEN1 and PSEN2), those represent only a minor fraction of all cases. GWAS studies have identified ≈30 late-onset AD (LOAD) susceptibility loci that, when functionally clustered, are directly or indirectly linked to cellular trafficking routes and functioning. These include genes linked to cholesterol metabolism, endocytic transport regulation and lysosomal processes. The genetic findings concur with the appearance of endolysosomal abnormalities at preclinical stages, before the appearance of senile plaques and tangles. Using a CRISPR-Cas9 based approach, knock-out (KO) of various risk factors from early (PSEN1) or late (PICALM, PLD3) onset risk factors were realized in fibroblasts and neuronal SH-SY5Y cells. Analysis of the endolysosomal pathway and lysosomal functions were carried out using biochemical assays, lipidomic analysis of isolated organelles, confocal imaging as well as electron microscopy. Both the deficiency of presenilins and late risk factors led to endolysosomal dysfunctions, as characterized by an enlargement of late endosomes/lysosomes that accumulate diverse types of lipids such as cholesterol or glycerophospholipids. Further focusing on PSENs, their depletion led to a delay in the endosomal maturation, resulting in an accumulation of endosomal cargoes and adapters such as the retromer subunit VPS35. Re-expression of either catalytic or inactive PSEN1 was sufficient to correct these aberrant accumulations. Whereas the endolysosomal abnormalities could be connected to the accumulation of APP-CTFs, the nature of γ-secretase-independent contribution remains enigmatic. Activation of recycling significantly alleviated endolysosomal defects observed in PSENdKO cells. Altogether, our results underscore that different upstream mechanisms could lead to similar endolysosomal dysfunctions. Our observations are also shedding light on the potential of modulating/restoring endosomal recycling regulation as a strategy to alleviate observed early endolysosomal dysfunctions at preclinical stages of AD pathogenesis.

Neurodegenerative diseases (NDs) are generally considered proteinopathies but whereas this may initiate disease in familial cases, onset in sporadic diseases may originate from a gradually disrupted organellar homeostasis. Herein, endolysosomal abnormalities, mitochondrial dysfunction, endoplasmic reticulum (ER) stress, and altered lipid metabolism are commonly observed in early preclinical stages of major NDs, including Parkinson's disease (PD) and Alzheimer's disease (AD). Among the multitude of underlying defective molecular mechanisms that have been suggested in the past decades, dysregulation of inter-organellar communication through the so-called membrane contact sites (MCSs) is becoming increasingly apparent. Although MCSs exist between almost every other type of subcellular organelle, to date, most focus has been put on defective communication between the ER and mitochondria in NDs, given these compartments are critical in neuronal survival. Contributions of other MCSs, notably those with endolysosomes and lipid droplets are emerging, supported as well by genetic studies, identifying genes functionally involved in lysosomal homeostasis. In this review, we summarize the molecular identity of the organelle interactome in yeast and mammalian cells, and critically evaluate the evidence supporting the contribution of disturbed MCSs to the general disrupted inter-organellar homeostasis in NDs, taking PD and AD as major examples.

Synapse structures, including neuronal and immunological synapses, can be seen as the plasma membrane contact sites between two individual cells where information is transmitted from one cell to the other. The distance between the two plasma membranes is only a few tens of nanometers, but these areas are densely populated with functionally different proteins, including adhesion proteins, receptors, and transporters. The narrow space between the two plasma membranes has been a barrier for resolving the synaptic architecture due to the diffraction limit in conventional microscopy (~250 nm). Various advanced super-resolution microscopy techniques, such as stimulated emission depletion (STED), structured illumination microscopy (SIM), and single-molecule localization microscopy (SMLM), bypass the diffraction limit and provide a sub-diffraction-limit resolving power, ranging from 10 to 100 nm. The studies using super-resolution microscopy have revealed unprecedented details of the nanoscopic organization and dynamics of synaptic molecules. In general, most synaptic proteins appear to be heterogeneously distributed and form nanodomains at the membranes. These nanodomains are dynamic functional units, playing important roles in mediating signal transmission through synapses. Herein, we discuss our current knowledge on the super-resolution nanoscopic architecture of synapses and their functional implications, with a particular focus on the neuronal synapses and immune synapses.

γ-Secretase affects many physiological processes through targeting >100 substrates; malfunctioning links γ-secretase to cancer and Alzheimer's disease. The spatiotemporal regulation of its stoichiometric assembly remains unresolved. Fractionation, biochemical assays, and imaging support prior formation of stable dimers in the ER, which, after ER exit, assemble into full complexes. In vitro ER budding shows that none of the subunits is required for the exit of others. However, knockout of any subunit leads to the accumulation of incomplete subcomplexes in COPII vesicles. Mutating a DPE motif in presenilin 1 (PSEN1) abrogates ER exit of PSEN1 and PEN-2 but not nicastrin. We explain this by the preferential sorting of PSEN1 and nicastrin through Sec24A and Sec24C/D, respectively, arguing against full assembly before ER exit. Thus, dimeric subcomplexes aided by Sec24 paralog selectivity support a stepwise assembly of γ-secretase, controlling final levels in post-Golgi compartments.