© 2016. Published by The Company of Biologists Ltd | Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

RESEARCH ARTICLE SEPT8 modulates β-amyloidogenic processing of APP by affecting the sorting and accumulation of BACE1 Kaisa M. A. Kurkinen1,*, Mikael Marttinen1,*, Laura Turner2, Teemu Natunen1, Petra Mäkinen1, Fanni Haapalinna3, Timo Sarajärvi1, Sami Gabbouj1, Mitja Kurki4, Jussi Paananen4, Anne M. Koivisto3, Tuomas Rauramaa5, Ville Leinonen4, Heikki Tanila6, Hilkka Soininen3, Fiona R. Lucas2, Annakaisa Haapasalo3,6,‡ and Mikko Hiltunen1,3,‡

ABSTRACT a large number of neuritic plaques and neurofibrillary tangles (NFTs) in the neocortex of the brain are detected in Alzheimer’sdisease Dysfunction and loss of synapses are early pathogenic events in ’ patients (Cummings, 2004; Mandelkow and Mandelkow, 1998). Alzheimer s disease. A central step in the generation of toxic amyloid- These consist of amyloid-β (Aβ) peptide and hyperphosphorylated tau β (Aβ) peptides is the cleavage of amyloid precursor protein (APP) by β protein, respectively. It is well established that the increased -site APP-cleaving enzyme (BACE1). Here, we have elucidated β-amyloidogenic processing of amyloid precursor protein (APP), whether downregulation of septin (SEPT) protein family members, leading to the augmented production of Aβ is a key feature underlying which are implicated in synaptic plasticity and vesicular trafficking, the molecular pathogenesis of Alzheimer’s disease (Neve et al., affects APP processing and Aβ generation. SEPT8 was found to 2000). The cleavage of APP by β-site APP-cleaving enzyme 1 reduce soluble APPβ and Aβ levels in neuronal cells through a post- (BACE1) leads to the formation of the N-terminal soluble APPβ translational mechanism leading to decreased levels of BACE1 (sAPPβ) and a membrane-bound APP C-terminal fragment, called protein. In the human temporal cortex, we identified alterations in the C99, which is subsequently cleaved by γ-secretase producing Aβ.By expression of specific SEPT8 transcript variants in a manner that contrast, the majority of APP is cleaved by α-secretases, which leads correlated with Alzheimer’s-disease-related neurofibrillary pathology. to the release of the soluble ectodomain portion of APP (sAPPα)and These changes were associated with altered β-secretase activity. We prevents Aβ formation. Although the role of sAPPα is still under also discovered that the overexpression of a specific Alzheimer’s- debate, recent findings have suggested that the overexpression of disease-associated SEPT8 transcript variant increased the levels of α ’ β sAPP in the brain of Alzheimer s disease model mice overexpressing BACE1 and A peptides in neuronal cells. These changes were human APP with the Swedish mutation and presenilin-1 with deletion related to an increased half-life of BACE1 and the localization of of exon 9 (APP/PS1ΔE9) mitigates synaptic and cognitive defects BACE1 in recycling endosomes. These data suggest that SEPT8 β caused by pre-existing pathology (Fol et al., 2016). modulates -amyloidogenic processing of APP through a mechanism In neurons, BACE1 shows presynaptic localization and it is largely affecting the intracellular sorting and accumulation of BACE1. conveyed in the neuronal recycling endosomes (Cole and Vassar, KEY WORDS: Alzheimer’s disease, Amyloid precursor protein, 2007; Das et al., 2013). BACE1 and APP are segregated in neurons Amyloid-β, BACE1, SEPT8 during resting state conditions, whereas, upon activity-dependent induction, these proteins converge into acidic microdomains INTRODUCTION through an endocytosis-dependent pathway (Das et al., 2013). This Alzheimer’s disease is the most common neurodegenerative disorder is important because BACE1 is optimally active in the acidic in the world, and affects up to 50% of individuals above the age of 85. intracellular compartments, such as early and late endosomes (Kang Owing to the global increase in the aging population, early diagnosis et al., 2012). Based on the amyloid cascade hypothesis, increased and treatment of Alzheimer’s disease are becoming increasingly soluble Aβ peptide levels augment synaptic dysfunction, Ca2+ important from both a human and socioeconomic perspective. dyshomeostasis, inflammation and oxidative stress, as well as tau Alzheimer’s disease is clinically associated with a global cognitive hyperphosphorylation and the formation of NFTs in specific brain decline, and a progressive loss of memory and reasoning. At autopsy, regions in Alzheimer’s disease patients (DeKosky and Scheff, 1990; Hu et al., 2014; Kuperstein et al., 2010). Thus, if started early enough, β 1Institute of Biomedicine, School of Medicine, University of Eastern Finland, 70211 lowering A levels could be sufficient to slow down the disease Kuopio, Finland. 2Eisai Ltd., Bernard Katz Building, University College London, process in Alzheimer’s disease. This idea is further supported by the 3 London WC1E 6BT, UK. Institute of Clinical Medicine – Neurology, School of genetic finding indicating that a naturally occurring A673T rare Medicine, University of Eastern Finland and Department of Neurology, Kuopio University Hospital, 70211 Kuopio, Finland. 4Institute of Clinical Medicine – variant in APP protects from Alzheimer’s-disease- and age-related Neurosurgery, School of Medicine, University of Eastern Finland and Neurosurgery cognitive decline by reducing BACE1-mediated cleavage of APP of NeuroCenter, Kuopio University Hospital, 70211 Kuopio, Finland. 5Institute of β Clinical Medicine – Pathology, School of Medicine, University of Eastern Finland and/or producing less of the aggregation-prone A 40 containing the and Department of Pathology, Kuopio University Hospital, 70211 Kuopio, Finland. A→T variation at position 2 (Di Fede et al., 2009; Jonsson et al., 2012; 6Department of Neurobiology, A.I. Virtanen, Institute for Molecular Sciences, School Benilova et al., 2014; Das et al., 2016). Soluble Aβ42 dimers extracted of Medicine, University of Eastern Finland, 70211 Kuopio, Finland. ’ *These authors contributed equally to this work from the Alzheimer s disease brain have been shown to specifically increase the phosphorylation of tau and subsequently promote ‡ Author for correspondence ([email protected]; [email protected]) neurodegeneration, supporting the intimate link between Aβ and tau ’ M.H., 0000-0003-3566-4096 in Alzheimer s disease pathogenesis (Shankar et al., 2008). This observation also concurs with the notion that synaptic dysfunction is

Received 23 December 2015; Accepted 11 April 2016 the best correlate of the cognitive decline in Alzheimer’s disease, Journal of Cell Science

2224 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 1. See next page for legend. emphasizing the importance of understanding the early mechanisms predictive biomarkers in Alzheimer’s disease, it is necessary to focus leading to synaptotoxicity (Coleman and Yao, 2003). Collectively, in on the cellular processes involved in the early steps of Alzheimer’s order to identify factors that could be utilized as novel drug targets or disease pathogenesis. Journal of Cell Science

2225 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 1. RNAi screening of septin family member proteins in HEK293-AP– members are known to play a role in the regulation of presynaptic APP cells identifies SEPT8 as a target affecting APP processing and activity and function (Amin et al., 2008; Ito et al., 2009), we β β A 40 and A 42 levels. (A) qRT-PCR was conducted to assess the studied the effects of downregulation of septin family members knockdown effect of 5 mM siRNAs for each target after 72 h of transfection in HEK293-AP–APP cells. GAPDH-normalized mRNA levels of SEPT5 (n=4), (SEPT5, SEPT7, SEPT8 and SEPT10) on APP processing by SEPT7 (n=4), SEPT8 (siSEPT8_2, n=3) and SEPT10 (n=4) in siSEPT- and assessing the levels of total soluble APP and Aβ in human siControl-transfected cells are presented. (B) The levels of total soluble HEK293 cells overexpressing alkaline-phosphatase-conjugated alkaline-phosphatase-linked APP (sAP–APP) in each target siRNA- APP (HEK293-AP–APP) (Lichtenthaler et al., 2003). Validated transfected cell line was normalized to the cell survival (white bars) in each small interfering RNAs (siRNAs) for BACE1 and ADAM17 were sample and is shown as the percentage of the level compared to that with used as positive controls in the screening of total soluble AP–APP siControl. Cell survival was not affected due to the downregulation of septins. (sAP–APP) and Aβ levels. Downregulation of SEPT5, SEPT7 and siControl, siSEPT5, siSEPT7, siSEPT10, siBACE1 and siADAM17 (n=4), and – siSEPT8 (siSEPT8_2, n=3). (C) Three different SEPT8 siRNAs significantly SEPT8 (Fig. 1A) significantly decreased sAP APP levels in the decreased the mRNA levels of SEPT8 in HEK293-AP–APP cells (left). The cell culture medium when normalized to the cell survival levels of Aβ40 and Aβ42 in the cell culture medium are normalized to APP total (Fig. 1B). In addition, a trend towards decreased Aβ40 and protein levels in the respective cell lysates and shown as a percentage of the Aβ42 levels was observed in culture medium samples from cells in level compared to that with siControl (right). n=4. (D) Three different SEPT8 which SEPT5 and SEPT8 was downregulated (Fig. S1A). siRNAs significantly decreased SEPT8 mRNA levels in SH-SY5Y cells stably Conversely, a trend towards increased Aβ40 and Aβ42 levels overexpressing the human APP751 isoform (SH-SY5Y-APP751) (left). The levels of Aβ40 and Aβ42 in the cell culture medium are normalized to the total was observed in cells with SEPT7 and SEPT10 downregulation protein levels of APP in the respective cell lysates and shown as a percentage (Fig. S1A). To further elucidate the SEPT8 findings, two other of the level compared to that with siControl (right). n=4. In A–D, n values are siRNAs targeting different parts of SEPT8 mRNA sequence were biological replicates and results are mean±s.d. *P<0.05, **P<0.01 also applied to the HEK293-AP–APP cells to exclude potential (independent sample t-test). off-target effects (Fig. 1C). Decreased Aβ40 and Aβ42 levels were observed with all three SEPT8 siRNAs when normalized to total Septins (denoted SEPT) are a conserved family of GTP-binding APP levels in the cell lysates (Fig. 1C). Total APP-normalized Aβ proteins, which are highly expressed in the brain (Marttinen et al., levels were assessed in order to account for the potential changes 2015). The septin protein family is composed of 13 genes, which in the total APP levels due to the knockdown of SEPT8 are divided into four subgroups based on their sequence homology (Fig. S1B). The same siRNAs were used to downregulate and domain composition: the SEPT2 group (SEPT1, SEPT2, SEPT8 in human SH-SY5Y neuroblastoma cells overexpressing SEPT4 and SEPT5), SEPT3 group (SEPT3, SEPT9 and SEPT12), the APP751 isoform (SH-SY5Y-APP751) (Sarajarvi et al., 2011). SEPT6 group (SEPT6, SEPT8, SEPT10, SEPT11 and SEPT14) Although the downregulation of SEPT8 mRNA levels was not as and SEPT7 group (SEPT7 only). Septins are implicated in the pronounced in SH-SY5Y-APP751 cells as in HEK293-AP–APP regulation of several cellular processes, including cytokinesis, cells, all three SEPT8 siRNAs resulted in a 20–40% decrease in membrane remodeling and compartmentalization, cytoskeleton both Aβ40 and Aβ42 levels (Fig. 1D) when normalized to total rearrangement, apoptosis and vesicle trafficking (Hall and Russell, APP levels in the cell lysates (Fig. S1C). Collectively, these data 2012). SEPT8 and SEPT5 have been ascribed a key role in the suggest that the downregulation of SEPT8 in non-neuronal and regulation of presynaptic functions, such as neurotransmitter neuronal cells affects APP processing by lowering total soluble release (Ito et al., 2009) and vesicle trafficking (Amin et al., 2008). APP and Aβ levels. In addition to these physiological functions, septins have been linked to different neurodegenerative diseases and psychiatric Downregulation of SEPT8 decreases endogenous BACE1 disorders; for example, Alzheimer’s disease, Parkinson’s disease, protein levels in SH-SY5Y neuroblastoma cells and primary Huntington’s disease, frontotemporal dementia, Down syndrome hippocampal neurons and schizophrenia (Barr et al., 2004; Kinoshita et al., 1998; The observation that the downregulation of SEPT8 decreased the Munoz-Soriano and Paricio, 2007). levels of both secreted Aβ and total soluble APP led us to assess Here, we have investigated the involvement of septin family the effects of SEPT8 downregulation on APP processing in the members SEPT5, SEPT7, SEPT8 and SEPT10 on Alzheimer’s total protein lysates of SH-SY5Y-APP751 cells. An average 60% disease molecular pathogenesis. We show that SEPT8 affects APP reduction in the γ-tubulin-normalized SEPT8 protein levels processing and Aβ peptide production through a molecular (Fig. 2A) did not significantly affect the total protein levels of mechanism that modulates the intracellular sorting and stability APP (APPtot). However, a trend towards reduced levels of APP of BACE1 in neuronal cells. We also observed an imbalanced C99, but not APP C83, the C-terminal fragment produced in the SEPT8 transcript variant expression profile that is already present non-amyloidogenic APP processing pathway, was observed at the early stages of the Alzheimer’s-disease-related (Fig. 2A). Furthermore, coinciding with the decreased Aβ levels neurofibrillary pathology in the human temporal cortex. These (Fig. 2B; Fig. S2A), protein levels of endogenous BACE1 were findings suggest that SEPT8 is centrally linked to cellular reduced (∼25%) upon the downregulation of SEPT8 (Fig. 2A). processes relevant to Alzheimer’s disease both in vitro and in the Similarly, lentiviral siRNA-mediated downregulation of SEPT8 in human brain. primary hippocampal neuronal cultures prepared from E18 mouse embryos resulted in significantly decreased (∼30%) endogenous RESULTS BACE1 protein levels (Fig. 2C). Consistent with this finding, the RNAi-based screening of septin protein family members in levels of Aβ40 and Aβ42 were significantly reduced in the culture non-neuronal and neuronal cells overexpressing APP medium of hippocampal neurons (Fig. 2D). The mRNA levels of identifies SEPT8 as a target reducing soluble APP and Aβ BACE1 in the primary hippocampal neurons were unchanged levels [control siRNA (siControl) 100.0±50.0%, siRNA against SEPT8 Given that Aβ-mediated synaptic dysfunction is one of the earliest (siSEPT8) 100.8±24.6%, n=4, mean±s.d., independent sample t- features in Alzheimer’s disease, and that septin protein family test, not significant], suggesting that the decrease in BACE1 Journal of Cell Science

2226 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 2. See next page for legend. protein levels takes place at the post-transcriptional level. even after normalization to the total APP levels in the primary Moreover, the levels of APPtot were significantly decreased in hippocampal neurons (Fig. S2B). Taken together, these data the primary hippocampal neurons after downregulation of SEPT8 suggest that downregulation of SEPT8 leads to a reduction of

(Fig. 2C). Importantly, Aβ40 levels were significantly reduced endogenous BACE1 and Aβ levels in neuronal cells. Journal of Cell Science

2227 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 2. Downregulation of SEPT8 leads to decreased levels of which does not involve prominent changes in the cell surface endogenous BACE1 in neuroblastoma cells and cultured mouse primary levels of BACE1. hippocampal neurons. (A) Western blot analysis to elucidate the effects on APP processing in the SH-SY5Y-APP751 neuroblastoma cells transfected SEPT8 with SEPT8 (siSEPT8_2) or control siRNAs. On average, a 60% decrease in transcript variant mRNAs are differentially expressed γ-tubulin-normalized SEPT8 protein levels was observed. Levels of APP C83 in the human temporal cortex in a manner that is correlated and C99 are normalized to both γ-tubulin and total APP levels. All samples with Alzheimer’s-disease-related neurofibrillary pathology were run on the same gel. (B) Downregulation of SEPT8 led to a reduction in Next, we wanted to address the relationship between the expression the Aβ levels. Aβ40 and Aβ42 levels in the cell culture medium are normalized of SEPT8 and the Alzheimer’s-disease-associated neurofibrillary to total protein levels in the respective cell lysates. (C) Downregulation of pathology in the human temporal cortex. The human SEPT8 gene SEPT8 protein levels using shRNA lentiviral particles (20 MOI) in mouse encodes four different transcript variants (denoted TV) (Fig. 4A): primary hippocampal neurons cultured for 20 days in vitro (DIV 20). Western ‘ ’ blot analysis shows a significant downregulation of mouse SEPT8 protein SEPT8 TV1 [483 amino acids (aa), referred to as the canonical 14 days after transduction. Downregulation of SEPT8 significantly decreased sequence, 56 kDa], SEPT8 TV2 (429 aa, differs from the canonical endogenously expressed BACE1 protein levels in mouse primary hippocampal sequence as follows: aa 430–483 missing, 50 kDa), SEPT8 TV3 neurons. Protein levels are normalized to GAPDH. The molecular mass in kDa (442 aa, differs from the canonical sequence as follows: 429–442, is indicated on the blots. (D) Aβ40 and Aβ42 levels were significantly NRSDIGAHQPGMSL→KASGWSSIYSVTIP, aa 443–483 decreased in mouse primary hippocampal neurons after siSEPT8 lentiviral β missing, 52 kDa), and SEPT8 TV4 (369 aa, differs from the transduction. A levels in cell culture medium are normalized to total protein – – levels in the respective cell lysates. In A–D, n=4 biological replicates and canonical sequence as follows: aa 1 60 missing, aa 430 483 results are mean±s.d. *P<0.05, **P<0.01 (independent sample t-test). missing, 44 kDa). To determine the expression status of SEPT8 transcript variants in our well-established human brain sample cohort (Martiskainen et al., 2015), we applied a microarray-based Downregulation of SEPT8 decreases BACE1 protein levels approach in which probes were designed to capture all the exons in through a post-translational mechanism the SEPT8 mRNA (Fig. 4B). This cohort consists of RNA samples As several lines of evidence suggested that the downregulation of from the post mortem temporal cortex of 60 subjects with varying SEPT8 decreases BACE1 protein levels and consequently degrees of Alzheimer’s-disease-related neurofibrillary pathology BACE1-mediated APP processing, we next wanted to assess (phosphorylated tau protein staining with AT8 antibody) (Table 1). whether the observed reduction in BACE1 protein levels is The subjects were grouped based on Braak staging into three regulated at the post-transcriptional or post-translational level. To subgroups: Braak stages 0–II, Braak stages III or IV, and Braak do this, we utilized two different BACE1 cDNA plasmids resulting stages V or VI (Braak et al., 2006; Martiskainen et al., 2015). We in either high or moderate expression levels of BACE1. A Myc- found that the expression profile at the 3′ end of SEPT8 was tagged BACE1 cDNA construct under the cytomegalovirus significantly different between the Braak groups in the temporal (CMV) promoter and without 5′ and 3′ untranslated regions cortex (Fig. 4B). More specifically, probes 317 and 807 targeted to (UTRs) of endogenous BACE1 (BACE1–Myc) yields high the exon 10b (TV1 variant) showed decreased binding, whereas BACE1 protein levels. Transfection of cells with a cDNA probes 314, 315 and 827 targeted to exon 10a and 10a* (TV2–TV4 construct encompassing the endogenous 5′ and 3′ UTRs of variants) showed increased binding. This alteration in the BACE1 (5′3′UTR-BACE1) results in a moderate overexpression expression of these variants led to a significantly reduced ratio of of BACE1 protein. SH-SY5Y-APP751 cells were co-transfected TV1 versus TV2–TV4 variants (TV1:TV2–TV4) according to the either with control or SEPT8 siRNAs together with 5′3′UTR- Alzheimer’s-disease-related neurofibrillary pathology (Fig. 4C). BACE1 (Fig. 3A,B) or BACE1–Myc plasmids (Fig. 3C). The Based on the results from our in vitro cell culture findings relating downregulation of SEPT8 significantly decreased BACE1 levels in to SEPT8 and BACE1, we next investigated whether the mRNA cells upon expression of either 5′3′UTR-BACE1 or BACE1–Myc. profiles of the different SEPT8 transcript variants showed a As the decrease was observed in cells transfected with BACE1– relationship with β-secretase activity, which was measured Myc, which has no regulatory UTRs, the observed reduction in previously from the same tissue samples (Laitera et al., 2014; BACE1 protein levels is likely due to post-translational Martiskainen et al., 2015). We observed a significant negative mechanisms. In line with the previous findings described above, correlation between exon 10b (probe 807) and β-secretase activity the downregulation of SEPT8 reduced BACE1-mediated APP (r=−0.46, P<0.001, Spearman correlation, n=55) in the temporal processing by significantly decreasing Aβ40, Aβ42 and sAPPβ, cortex. The opposite effect was detected with exon 10a (probe 314) but not sAPPα levels (Fig. 3B). and β-secretase activity (r=0.62, P<10−6, Spearman correlation, To further elucidate the post-translational mechanism n=55). Consequently, the TV1:TV2–TV4 ratio showed significant underlying reduced BACE1 protein levels after the knockdown negative correlation with β-secretase activity in the temporal cortex of SEPT8, we next assessed whether the trafficking of BACE1 (Fig. 4D). Taken together, these data suggest an imbalanced and/or APP onto the cell surface was affected by potentially transcript variant expression of SEPT8 in the early phase of altered intracellular trafficking within the secretory or endocytic Alzheimer’s disease pathogenesis (Braak stages 0–II versus III or recycling pathways. Western blot analysis performed after cell IV) and that the increased mRNA levels of the SEPT8 TV2–TV4 surface biotinylation in SH-SY5Y-APP751 cells co-transfected variants coincide with increased β-secretase activity in human with the 5′3′UTR-BACE1 plasmid and control siRNA or siSEPT8 temporal cortex. (Fig. S3A) demonstrated a moderate, but statistically insignificant, increase in the BACE1 protein levels at the cell surface when Overexpression of TV3 variant of SEPT8 stabilizes BACE1 normalized to total levels of BACE1 in the cell lysates (Fig. S3B). protein in SH-SY5Y neuroblastoma cells The levels of mature APP at the cell surface were not altered after Given the data that the levels of SEPT8 TV2, TV3 and TV4 normalization to the total APP levels in the cell lysates (Fig. S3B). are increased in Alzheimer’s disease brain and that the increase In summary, these results suggest that the knockdown of SEPT8 correlated with altered BACE1 activity, next we assessed whether decreases BACE1 levels through a post-translational mechanism, transient overexpression of the TV3 variant of SEPT8 affects Journal of Cell Science

2228 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 3. Downregulation of SEPT8 reduces exogenous BACE1 protein levels in SH-SY5Y-APP751 cells overexpressing 5′3′UTR-BACE1 or BACE1–Myc. SH- SY5Y-APP751 cells were co-transfected with BACE1 plasmid containing both the 5′- and 3′-untranslated regulatory regions of human BACE1 mRNA (5′3′UTR-BACE1) or Myc-tagged BACE1 (BACE1–Myc) and siControl or siSEPT8. (A) Western blot analysis showing significantly decreased BACE1 levels in 5′3′UTR-BACE1- and siSEPT8-transfected (siSEPT8_2) cells. Protein levels were normalized to GAPDH and are shown as a percentage of that with siControl. UNT, untreated SH-SY5Y-APP751 cells. (B) Western blots showing levels of soluble APPβ (sAPPβ), APPα (sAPPα) and soluble total APP (sAPPtot) in the cell culture medium from 5′3′UTR-BACE1- and siSEPT8-transfected cells. Levels of Aβ40 and Aβ42 are normalized to total protein levels in respective cell lysates, whereas sAPPβ and sAPPα levels were normalized to sAPPtot levels. The data are shown as a percentage of that with siControl. The levels of Aβ40 and Aβ42 as well as sAPPβ showed a significant decrease upon the downregulation of SEPT8 as compared to control samples. (C) Western blot analysis showing that downregulation of SEPT8 decreased BACE1–Myc protein levels in SH-SY5Y-APP751 cells. Protein levels are normalized to GAPDH and shown as a percentage of the value with siControl. The molecular mass in kDa is indicated on the blots. In A–C, n=4 biological replicates replicates and results are mean±s.d. *P<0.05, **P<0.01, ***P<0.001 (independent sample t-test).

endogenous BACE1 protein levels in SH-SY5Y-APP751 cells. experiment was performed in SH-SY5Y-APP751 cells, which Importantly, a transient overexpression of SEPT8 TV3 tagged with were analyzed at the 3, 6 and 12 h time-points to delineate the V5 (SEPT8–V5) significantly increased endogenous BACE1 effects of SEPT8 on the stability of BACE1 protein. Cycloheximide protein levels (Fig. 5A). When SEPT8–V5 was co-overexpressed inhibits de novo protein synthesis and, thus, enables the follow-up of with 5′3′UTR-BACE1 plasmid, the total protein-normalized Aβ40 the degradation of existing proteins over time. BACE1 protein levels levels were also significantly increased (Fig. 5B). BACE1 protein were significantly increased at the time-point zero in SH-SY5Y- levels were also increased when untagged SEPT8 TV3 was APP751 cells co-transfected with SEPT8 and 5′3′UTR-BACE1 co-overexpressed with 5′3′UTR-BACE1 plasmid in the SH- plasmids, as was also detected in the previous experiment

SY5Y-APP751 cells (Fig. 5C). Next, a cycloheximide timecourse (Fig. 6A). Although not statistically significant, the half-life (t½) Journal of Cell Science

2229 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 4. The expression profile of SEPT8 transcript variants is altered in correlation with the Alzheimer’s-disease-related neurofibrillary pathology in human temporal cortex. (A) Exon and transcript variant (TV) structure of the SEPT8 gene, as well as the positions of microarray probes used for the expression and splicing screening. Untranslated regions are shown as a light gray boxes and protein-coding exons as black boxes. (B) Exon array analysis of mRNA samples extracted from the temporal cortex unveils a statistically significant increase in SEPT8 exon 10a expression (probe 314) in the Braak stages III and IV (n=13) and V and VI (n=19) as compared to the Braak stages 0–II (n=28). Conversely, the expression of SEPT8 exon 10b (probes 317 and 807, located at the far 3′ end), is significantly decreased in correlation with the Alzheimer’s-disease-related neurofibrillary pathology. Results are mean±s.e.m., n=60 subjects. *P<0.05, **P<0.01 [one-way ANOVA with least significant difference (LSD) post-hoc analysis]. (C) The SEPT8 TV1:TV2–TV4 ratio (TV1/TV2-4) is significantly decreased in correlation with the Alzheimer’s-disease-related neurofibrillary pathology. Results are mean±s.e.m., n=60 subjects. **P<0.01 [one-way ANOVA with least significant difference (LSD) post-hoc analysis]. (D) The altered transcript variant ratio of SEPT8 (TV1:TV2–TV4) correlates with increased β-secretase activity in the human temporal cortex in correlation with the Alzheimer’s-disease-related neurofibrillary pathology. n=55 subjects. ***P<0.001 (Spearman’s correlation). of BACE1 was increased in the SEPT8 and 5′3′UTR-BACE1 co- data suggest that transient overexpression of SEPT8 TV3 stabilizes transfected cells (t½=7.3±0.4 h) as compared to the control- BACE1 protein levels and augments its activity in the neuronal transfected cells (t½=5.7±1.3 h) (Fig. 6B). Taken together, these cells. Journal of Cell Science

2230 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Table 1. Demographic and clinical characteristics of the subjects with Alzheimer’s-disease-related neurofibrillary pathologya Subjects with clinical Females/Males Age at death Braak stage Number of subjects per stage Alzheimer’s disease b (% women) (years)c Post-mortem delay (hours)d 0 6 5 3/3 (50) 79.2±10.9 18±19.6 I 11 1 5/6 (45) II 11 5 6/5 (55) III 6 6 5/1 (83) 82.8±6.8 15.8±16.1 IV 7 5 6/1 (86) V 12 12 10/2 (83) 81.3±6.95 7.1±5.0 VI 7 7 7/0 (100) aClassification of Alzheimer’s disease-related neurofibrillary pathology according to Braak staging. bClinically diagnosed Alzheimer’s disease according to NINCDS-ADRDA criteria. cThere were no significant differences in the age of death between Braak 0–II, Braak III or IV, and Braak V or VI groups in the Alzheimer’s-disease-related sample cohort. Data are mean±s.d. dThere was a significant difference in the post-mortem delay between Braak 0–II, and Braak V and VI groups (P=0.008). Data are mean±s.d.

SEPT8 modulates intracellular trafficking of BACE1 in SH- 5′3′UTR-BACE1 with the V5-tagged SEPT8 TV3 was detected by SY5Y neuroblastoma cells immunofluorescence microscopy (Fig. 6C), a co-immunoprecipitation The observation that SEPT8 TV3 stabilized BACE1 protein analysis revealed no direct interaction between SEPT8 TV3 and prompted us to investigate the subcellular localization of BACE1 BACE1 in the SH-SY5Y-APP751 cells (data not shown). SEPT8- in SH-SY5Y-APP751 cells. Although partial colocalization of TV3–V5 localized throughout the SH-SY5Y-APP751 cells in

Fig. 5. Overexpression of the SEPT8 TV3 variant increases both endogenous and exogenous BACE1 protein levels in SH-SY5Y-APP751 cells. (A) Western blot analysis showing that transient overexpression of C-terminally V5-tagged SEPT8 TV3 (SEPT8–V5) significantly increased endogenous BACE1 protein levels. Protein levels were normalized to GAPDH and are shown as a percentage of control (pcDNA). (B) The levels of Aβ40 and Aβ42 in the cell culture medium from pcDNA or SEPT8–V5 and 5′3′UTR-BACE1 co-transfected cells were normalized to total protein levels in respective cell lysates and are shown as a percentage of control (pcDNA). (C) Western blot analysis showing that the transient co-overexpression of untagged SEPT8 TV3 with 5′3′UTR-BACE1 plasmid significantly increased BACE1 levels as compared to control transfected (pcDNA) samples. Protein levels were normalized to TfR levels and are shown as a percentage of control (pcDNA). Molecular masses in kDa are indicated on the blots. In A–C, n=4 biological replicates and results are mean±s.d. *P<0.05, **P<0.01, ***P<0.001 (independent sample t-test). Journal of Cell Science

2231 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 6. See next page for legend. Journal of Cell Science

2232 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 6. Overexpression of SEPT8 TV3 stabilizes BACE1 protein levels by formation, growth and stability of axons and dendrites, synaptic increasing the half-life of BACE1 and enhances the sorting of BACE1 to plasticity and vesicular trafficking (Marttinen et al., 2015). Beyond the recycling and early endosomal compartments in SH-SY5Y-APP751 these physiological functions, septins have been linked to different cells. (A) Western blot analysis of SH-SY5Y-APP751 cells co-transfected with ’ untagged SEPT8 TV3 and 5′3′UTR-BACE1 and treated with 30 µg/ml neurodegenerative and psychiatric disorders, such as Alzheimer s cycloheximide for 0, 3, 6 and 12 h. The levels of BACE1 were already disease, Parkinson’s disease, Huntington’s disease, frontotemporal significantly increased at the 0 h time-point in SEPT8-TV3-overexpressing dementia, Down syndrome and schizophrenia (Ageta-Ishihara et al., cells as compared to control cells (pcDNA). Total APP (APPtot) was used as a 2013; Barr et al., 2004; Dong et al., 2003; Gozal et al., 2011; Ihara control. UNT, untreated SH-SY5Y-APP751 cells. Molecular masses in kDa are et al., 2007; Kinoshita et al., 1998; Pissuti Damalio et al., 2012; indicated on the blots. (B) Quantification of BACE1 protein levels at different Zhang et al., 2000). time-points after cycloheximide treatment suggested that the overexpression of Based on our initial RNAi screening of SEPT5, SEPT7, SEPT8 SEPT8 TV3 reduced the degradation of BACE1. BACE1 protein levels are β normalized to the BACE1:GAPDH ratio at the 0 h time-point. In A and B, n=4 and SEPT10, we found that SEPT8 in particular affected the - biological replicates and results are mean±s.d. (C) BACE1 (green) displays amyloidogenic processing of APP. More specifically, we found that partial colocalization (yellow) with SEPT8-TV3–V5 (red). SEPT8-TV3–V5 the downregulation of SEPT8 decreased the levels of sAPPβ and localizes in different cellular compartments throughout the cell, except in the Aβ in conjunction with reduced protein levels of BACE1 through a nucleus. (D) Overexpression of SEPT8-TV3–V5 (n=69 individual cells) potential molecular mechanism involving regulation of sorting and decreases BACE1 (green) colocalization with Rab7 (red), a marker of late accumulation of BACE1 protein in neuroblastoma cells. Consistent endosomes and lysosomes, as compared to control cells (pcDNA, n=45 individual cells). Results are mean±s.e.m. (E) Overexpression of SEPT8-TV3– with this, endogenous BACE1 protein, but not mRNA levels, as V5 (n=13 individual cells) significantly increases BACE1 (green) colocalization well as Aβ levels, were significantly downregulated in cultured with the recycling and early endosomal marker TfR (red) as compared to primary hippocampal neurons upon lentivirus-mediated control cells (pcDNA, n=21 individual cells). Results are mean±s.e.m. Scale downregulation of SEPT8. However, it should be emphasized bars: 5 μm. *P<0.05, ***P<0.001 (Mann–Whitney U-test). that the total APP levels were also significantly decreased in the hippocampal neurons, possibly also contributing to the reduced Aβ different intracellular compartments, except in the nucleus (Fig. 6C). levels. Although the total APP-normalized Aβ40 levels were still To validate that the experimental settings in the quantitative significantly reduced in the primary hippocampal neurons, we immunofluorescence microscopy allowed us to detect differences cannot completely rule out the possibility that some other in the colocalization of two different proteins, we first mechanism beyond BACE1 might underlie the altered APP- and downregulated the well-known BACE1-interacting and sorting Aβ-related changes. In general, the effects of SEPT8 protein Golgi-localized γ-ear-containing ADP-ribosylation factor- downregulation on the endogenous levels of BACE1 were more binding protein 3 (GGA3) (Tesco et al., 2007) by using siRNA in the substantial in the primary hippocampal neurons as compared to SH-SY5Y-APP751 cells transiently overexpressing 5′3′UTR- neuroblastoma cells. In contrast, overexpression of the BACE1 (Fig. S4A). We detected a significantly decreased Alzheimer’s-disease-associated SEPT8 transcript variant 3 colocalization of BACE1 with GGA3 in these SH-SY5Y-APP751 (SEPT8 TV3) resulted in an opposite effect to that of cells. The GGA3 protein levels were decreased by ∼40% and the downregulation, indicated by augmented levels of Aβ and colocalization of GGA3 with BACE1 was similarly decreased by BACE1 protein in neuroblastoma cells. These changes coincided ∼40% (Fig. S4B). Co-immunostaining of BACE1 and markers with a moderately increased half-life and the alteration of the specific for different subcellular compartments was next used to subcellular localization of BACE1 from late-endosomal or examine BACE1 subcellular localization. BACE1 localization was lysosomal compartments to early and/or recycling endosomes. found to be significantly decreased in the Rab7-positive late- Previously, a similar change of BACE1 trafficking from late- endosomal or lysosomal compartments in cells overexpressing endosomal or lysosomal compartments to early and recycling SEPT8–V5 and 5′3′UTR-BACE1 (Fig. 6D). In line with this, endosomes has been shown to coincide with the stabilization of colocalization of BACE1 with transferrin receptor (TfR; a marker for BACE1 after depletion of the BACE1-sorting protein GGA3 (Tesco recycling endosomes) (Fig. 6E) was significantly increased in cells et al., 2007). Thus, our biochemical and immunohistochemical co-overexpressing SEPT8–V5 and BACE1 (Das and Pellett, 2011). findings together suggest that overexpression of SEPT8 TV3 These findings led us to further study the protein levels of well- promotes sorting and accumulation of BACE1 to the recycling or established factors known to regulate endosomal and intracellular early endosomal compartments, which are optimal sites for its trafficking of BACE1, such as ADP-ribosylation factor 6 (Arf6), enzymatic activity (Tesco et al., 2007). These changes were not GGA1 and GGA3 (He et al., 2005; Sannerud et al., 2011; Tesco associated with altered protein levels of factors affecting the et al., 2007). However, no significant changes in the protein levels endosomal or intracellular trafficking of BACE1, such as Arf6, of Arf6, GGA1 or GGA3 were observed in SH-SY5Y-APP751 cells GGA1 or GGA3, or with alterations in the Lys48- or Lys63-linked co-overexpressing SEPT8 and 5′3′UTR-BACE1 as compared to ubiquitylation known to regulate the lysosomal and proteosomal cells co-transfected with control plasmid and 5′3′UTR-BACE1 trafficking of BACE1 (Kang et al., 2012; Walker et al., 2012). (data not shown). In addition, overexpression of SEPT8 in SH- Importantly, our results are in line with recent studies that have SY5Y-APP751 cells did not affect the Lys48- or Lys63-linked shown that BACE1 mainly accumulates in the TfR-positive ubiquitylation of BACE1 (data not shown). Collectively, endosomes and is degraded in the lysosomes (Kandalepas et al., cycloheximide timecourse and co-immunostaining data suggest 2013). Collectively, these results suggest SEPT8 is a potential new that the overexpression of SEPT8 TV3 promotes the sorting and sorting protein for BACE1 and thus underline its potential role in accumulation of BACE1 to the recycling or early endosomal the early pathogenesis of Alzheimer’s disease (Fig. 7). However, it compartments in SH-SY5Y-APP751 neuroblastoma cells. should be noted that the other identified septin family members beyond SEPT8 are also potential targets that might affect APP DISCUSSION processing and the generation of Aβ. It is important to perform Septins are a conserved family of GTP-binding proteins, which are similar detailed molecular characterizations of these septin family highly expressed in the brain and known to take part in the members as with SEPT8 in the future. Journal of Cell Science

2233 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

Fig. 7. The potential mechanism of the effects of SEPT8 in the sorting and accumulation of BACE1. In non-neuronal and neuronal cells, downregulation of SEPT8 levels results in the decreased levels of BACE1 and subsequently diminished production of sAPPβ and Aβ. In contrast, increased expression of the SEPT8 TV3 variant facilitates sorting of BACE1 to early endocytic compartments, where BACE1 protein is stabilized leading to increased BACE1 levels and activity, and in the increased generation of Aβ. Similarly, in Alzheimer’s disease brain, an imbalance in SEPT8 transcript variant expression [reduced TV1:TV2–TV4 variant ratio (TV1/TV2-4)], which occurs during the progression of the brain pathology, might lead to increased BACE1 levels and activity, and consequently increased Aβ generation. Full details are in the Discussion. EE, early endosome; RE, recycling endosome; L, lysosome; PM, plasma membrane.

As further support for the suggested role of SEPT8 in the early TV3, which could consequently facilitate the observed sorting of stages of Alzheimer’s-disease-associated pathophysiology, we BACE1 to the recycling or early endosomal compartments. identified transcript-variant-specific alterations in the expression Although immunofluorescence microscopy in the neuroblastoma profile of SEPT8 mRNA that could be related to the Alzheimer’s- cells suggested that BACE1 and SEPT8 are closely located, a direct disease-related neurofibrillary pathology in the human temporal interaction between the two proteins was not observed in the co- cortex. In these assessments, we used a well-established brain tissue immunoprecipitation analysis. According to the STRING database sample set, which has been previously used to assess the expression (http://string-db.org/), which allows the assessment of the known and splicing status of Alzheimer’s-disease-associated risk genes as and predicted protein–protein interactions, SEPT8 interacts with well as α-, β- and γ-secretase activity in the human temporal cortex other septin protein family members, such as SEPT5. Apart from (Laitera et al., 2014; Martiskainen et al., 2015). The analysis of this, interaction of SEPT8 with well-established synaptic proteins SEPT8-exon-specific microarray probes revealed that although total (CDK5, PARK2 and STX4) linked to neurodegenerative diseases SEPT8 mRNA levels were unaltered, the levels of TV1 were was suggested, indicating that SEPT8 is a central component in the significantly reduced and those of TV2–TV4 were increased interactome regulating presynaptic functions. according to the Alzheimer’s-disease-related neurofibrillary One of the earliest hallmarks in Alzheimer’s disease is the pathology. This indicates that the ratio of SEPT8 TV1:TV2–TV4 dysfunction and loss of synapses in the hippocampus and association mRNAs is significantly reduced even by the early stages of the cortices, but the detailed molecular mechanisms underlying these Alzheimer’s-disease-related neurofibrillary pathology (Braak stage processes still remain partially elusive (DeKosky and Scheff, 1990; III). This is a central discovery given that the transcription of the Scheff et al., 2006). Importantly, the degree of synaptic loss shows a SEPT8 gene is controlled by multiple promoters (Hall et al., 2005), close correlation with the degree of dementia (Coleman and Yao, and it underscores the possibility that the expression of different 2003). It has been previously shown that the protein levels of transcript variants of SEPT8 might be adapted upon varying different septins are altered in the temporal neocortex of Alzheimer’s physiological and pathophysiological circumstances (Fig. 7). The disease patients as compared to non-Alzheimer’s disease subjects reduced transcript variant ratio (TV1 versus TV2–TV4) negatively (Hanai et al., 2004; Musunuri et al., 2014), suggesting that septins correlated with increased activity of BACE1 in the temporal cortex. could represent early markers linked to synaptic dysfunction and This observation is in line with the in vitro cell-based data showing synaptotoxicity. Consistent with this idea, presynaptically enriched that the overexpression of the SEPT8 TV3 variant increased BACE1 SEPT8 controls the formation of the soluble N-ethylmaleimide- levels and the production of Aβ, whereas the opposite effect was sensitive factor attachment protein receptor (SNARE) complex and evident upon downregulation of SEPT8, particularly in mouse the subsequent docking of synaptic vesicles to the presynaptic primary hippocampal neurons. Nevertheless, it is important to note membrane (Ito et al., 2009). Based on this, it is conceivable that that the observed effects of SEPT8 TV3 on BACE1 might differ SEPT8 could play a central role at the synaptic level in the early from those of SEPT8 TV2 and TV4 due to structural differences. stages of Alzheimer’s-disease-related synaptotoxicity. It has been Thus, the effects of the SEPT8 TV2 and TV4 variants should also be recently shown that protein affinity purification combined with mass elucidated in subsequent studies. The structural differences at the spectrometry can be utilized for the identification of fragments of the protein level between SEPT8 transcript variants are primarily presynaptic SNARE complex component synaptosomal-associated located at the α-helical coiled-coil domain at the C-terminus, which protein 25 (SNAP25) in the cerebrospinal fluid (Brinkmalm et al., is a site that enables isoform-specific contacts with other interaction 2014). Indeed, significantly increased levels of SNAP25 fragments partners (Hall et al., 2005; Souza and Barbosa, 2010). Structural have been observed in the cerebrospinal fluid of Alzheimer’s disease heterogeneity at the C-terminus raises the question of whether patients even by the very early stages of the disease. This pinpoints BACE1 is able to specifically interact directly or through other SNAP25 as a potential biomarker for synaptic integrity even in the scaffold partner(s) with certain SEPT8 transcript variants, such as early stages of Alzheimer’s disease and further suggests that Journal of Cell Science

2234 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215 presynaptic markers beyond SNAP25 might be plausible tools for (10 mg/ml, Sigma) and 5 µl DNAse I (10 mg/ml, Sigma)]. The hippocampi the early diagnosis and the assessment of disease progression. were dissociated in 5 ml papain-dissociation solution at +37°C for 5 min. Therefore, it would be important to evaluate in the future whether the Papain-treated tissue was pelleted by centrifugation at 800 g for 5 min. After careful removal of papain-dissociation solution, 2 ml of trituration medium imbalanced transcript variant expression of SEPT8 in the 2+ 2+ cerebrospinal fluid could serve as a potential biomarker in subjects [in ml: 9.8 Ca - and Mg -free HBSS, 0.1 100 mM sodium pyruvate, 0.1 ’ 1 M HEPES (pH 7.2) and 10 µl DNase I (10 mg/ml)] was added on the with prodromal Alzheimer s disease. Finally, in order to validate the hippocampal tissue. The tissue was triturated using a 1-ml pipet by pipetting biomarker potential and specificity of the SEPT8-related transcript up and down 15 times. The indissoluble tissue was left to sink to the bottom variant changes, the comprehensive elucidation of any role for of the tube, and the supernatant containing the suspension of singular cells SEPT8 in other neurodegenerative disorders beyond Alzheimer’s was moved into a clean tube. Then, 2 ml of trituration medium was added on disease is essential. top of the tissue pellet and trituration was repeated. Cells were harvested by centrifugation at 800 g for 5 min and resuspended in 10 ml of feeding MATERIALS AND METHODS medium containing Neurobasal medium supplemented with 1× B27, siRNA oligonucleotides 0.5 mM L-glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin. The following siRNAs (Life Technologies) were used to knockdown Cells were plated on 12-well-plates at a density of 300,000 cells per well, the target sequences: human SEPT8 siRNA_1 ID s230020, sense and incubated at +37°C in 5% CO2 cell culture incubator. One-half of the 5′-AGAUUAACGCAGUCAUGAATT-3′ and antisense 5′-UUCAUGA- feeding medium was changed after 5 days in vitro. CUGCGUUAAUCUCT-3′; human SEPT8 siRNA_2 ID s230021, sense 5′-AGGUGAACAUUAUUCCCAUTT-3′ and antisense 5′-AUGGGAA- Lentiviral transduction shRNA particles UAAUGUUCACCUTG-3′; human SEPT8 siRNA_3 ID s230022, For validation of the knockdown efficiency of mouse SEPT8 siRNA in sense 5′-GGAUCCACGUUUGCCUCUATT-3′ and antisense mouse primary hippocampal neurons, short hairpin RNA (shRNA) targeting 5′-UAGAGGCAAACGUGGAUCCTT-3′; human SEPT5 siRNA ID s22- the open reading frame of mouse SEPT8 mRNA from Sigma (Mission 4294s, sense 5′-CGAGUGCACUGCUGCCUAUTT-3′ and antisense 5′- shRNA self-inactivating replication incompetent clone), TRCN0000112992 AUAGGCAGCAGUGCACUCGGT-3′; human SEPT7 siRNA ID s2741, sequence 5′-CCGGCGCCGAGATTAATGCAGTCATCTCGAGATG- ® sense 5′-GAAGGGAGCAUGUAGCUAATT-3′ and antisense 5′-UUAG- ACTGCATTAATCTCGGCGTTTTTG-3′, was used. MISSION lenti- CUACAUGCUCCCUUCTT-3′; human SEPT10 siRNA ID s45506 viral control short hairpin transduction particles, clone ID SHC002H, 5′-CCAUGAUUCUCGCAUCCAUTT-3′ and antisense 5′-AUGGAUG- non-mammalian short hairpin targeting 5′-CCGGCAACAAGATGAAG- CGAGAAUCAUGGTA-3′; ON-TARGETplus Human GGA3 siRNA 5′- AGCACCAACTCGAGTTGGTGCTCTTCATCTTGTTGTTTTT-3′, GAGAACAAGAGGCGGACUU-3′ (J-012881-09-005, Dharmacon); hu- was used as a control. The bold letters represent the target sequence of the man BACE1 siRNA ID s24219 (Ambion Life Technologies); and human hairpin structure. The lentiviral particles were introduced to the hippocampal ADAM17 siRNA ID s13718 (Ambion Life Technologies). All target siR- neuron cultures at 20 multiplicity of infection (MOI) at 5 days in vitro (DIV). NAs were compared to siControl Silencer® Select Negative Control #1 The samples were analyzed at 20 days after transduction. siRNA (catalog #4390843, Ambion, Life Technologies). RNA extraction and cDNA synthesis cDNA constructs Total RNA was extracted from siRNA-transfected HEK293-AP-APP or SH- The following plasmids were used to overexpress target proteins: SEPT8 SY5Y-APP751 cells or lentivirus-transduced mouse primary hippocampal ® transcript variant 3 (TV3) in pCMV6-XL vector (Origene), and a plasmid in neurons using TRI Reagent (Molecular Research Center Inc.) according to ’ which SEPT8 TV3 cDNA was C-terminally tagged with V5 (SEPT8-TV3- the manufacturer s protocol. Equal amounts (500 ng) of total RNA samples V5) and subcloned into the pcDNA3.1 vector. BACE1 was overexpressed were subjected to cDNA synthesis utilizing SuperScript III Reverse from a plasmid with a C-terminally Myc-tagged BACE1 protein under the Transcriptase (Invitrogen). CMV promoter, and a 5′3′UTR-BACE1 plasmid containing 5′-and3′- untranslated regulatory regions. pcDNA3.1 served as a control. Real-time quantitative PCR analysis Target-specific primers for human SEPT8 (5′-GAATCACTGCGACTTC- Cell culture, transfections and measurement of cell survival GTGA-3′ and 5′-CTTCCTCTTGGCCTCGTATG-3′), SEPT5 (5′-ACAA- Human embryonic kidney cells stably overexpressing alkaline-phosphatase- GCAGTACGTGGGCTTC-3′ and 5′-TCAGCACTGAGCAGCTTCC-3′), linked APP (HEK293-AP–APP) were cultured in Dulbecco’s modified SEPT7 (5′-TGGGAAAGTCGACATTAATCAAC-3′ and 5′-CCACTGC- Eagle’s medium (DMEM) containing 10% FBS, 2 mM L-glutamine, ATCTCCAAATCCT-3′), SEPT10 (5′-CCGTTTGCTGTTGTGGGAAG- 100 U/ml penicillin and 100 mg/ml streptomycin and supplemented with 3′ and 5′-GCCCATTTCCTCCAGTTTGC-3′), BACE1 (5′-ATGGGTGA- ′ ′ ′ 50 mg/ml hygromycin B and 0.3 µg/ml puromycin at +37°C in a 5% CO2 GGTTACCAACCA-3 and 5 -GACAACGTAGAAGCCCTCCA-3 ), GA- cell culture incubator (Lichtenthaler et al., 2003; Viswanathan et al., 2011). PDH (5-GGTCTCCTCTGACTTCAACA-3′ and 5′-GTGAGGGTCTCT- A human neuroblastoma SH-SY5Y cell line stably overexpressing the CTCTTCCT-3′), and for mouse SEPT8 (5′-ACTCCCTGAAGTCCCTG- human APP751 isoform (SH-SY5Y-APP751) was cultured as previously GAT-3′ and 5′ATCTGGACTCCGTTGCTGAC-3′), mouse BACE1 (5′- described (Sarajarvi et al., 2009; Viswanathan et al., 2011, 2013). Cells were CAACCAGACCGAGGCACT-3′ and 5′-ATGCTCTTGTCGTAGTTGT- transfected with 5 nM of each target-specific siRNA, using Lipofectamine ACTCC-3′) and mouse GAPDH (5′-AACTTTGGCATTGTGGAAGG-3′ (Thermo Fisher Scientific) transfection reagent according to the and 5′-ACACATTGGGGGTAGGAACA-3′) were designed to amplify the manufacturer’s protocol. Cell survival levels were determined with region flanking at least two different exons of the target gene. The cDNA propidium iodide and digitonin from HEK293-AP–APP cells plated on a samples were amplified using SYBR Green Master Mix (Applied 48-well plate and treated as described previously (Loikkanen et al., 1998; Biosystems) or FastStart Universal SYBR Green Master (ROX, Roche) in Vepsalainen et al., 2013). a real-time quantitative PCR (qRT-PCR) machine (7500 Fast Real Time PCR System, Applied Biosystems). The standard curve method and Primary hippocampal cell cultures comparative Ct-method were used for analyzing mRNA levels, and each Primary hippocampal neuronal cultures were prepared from embryonic day sample was normalized to the GAPDH level from the same sample. 18 JAXC57BL/6J mouse brains. All animal experiments were performed according to approved guidelines. Briefly, hippocampi were dissected in Soluble APP and Aβ measurements ice-cold HC-buffer containing 1 mg/ml glucose and 10 mM PBS. Tissue Soluble sAPPα and sAPPβ were detected using a mouse monoclonal was pelleted by centrifugation at 800 g for 5 min. During the centrifugation antibody (1:1000; 6E10, Biosite) and a specific in-house rat monoclonal pre-warm 5 ml papain-dissociation buffer [in mg: 10 DL-Cysteine-HCl, 10 antibody (BAWT; 1:10) generated against amino acids 11–16 of Aβ

BSA, 250 glucose and 50 ml PBS (pH 7.4) containing 250 µl papain (a kind gift from Prof. Stefan Lichtenthaler, DZNE, Munich, Germany), Journal of Cell Science

2235 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215 respectively. Total sAPP (sAPPtot=sAPPα+sAPPβ) levels were detected PBS-Ca-Mg supplemented with 0.1 mM glycine for 20 min. Cells were using a mouse monoclonal antibody recognizing the N-terminus of APP washed twice with PBS-Ca-Mg. The cells were scraped in T-PER (1:1000; MAB348, clone 22C11; Millipore). Aβ40 and Aβ42 levels were containing protease inhibitors and centrifuged at 10,000 g for 10 min at measured from HEK293-AP-APP, SH-SY5Y-APP751, and wild-type 4°C. Protein concentration was measured from the supernatant and 450 µg mouse primary hippocampal neuron cell culture medium using the human of proteins were mixed with binding buffer (PBS and 1% Nonident P40) and or rat Aβ40 (294-64701) ELISA Kit (Wako) and the Aβ42 (High-Sensitive; incubated with agarose beads cross-linked with streptavidin (Pierce) 292-64501) ELISA Kit (Wako) according to manufacturer’s protocol as overnight. The samples were centrifuged at 5000 g for 1 min to obtain a described previously (Natunen et al., 2013b). pellet containing the biotinylated protein fraction and supernatant containing the unbiotinylated protein fraction for western blotting. Western blot analysis Total proteins were extracted from HEK293-AP–APP, SH-SY5Y-APP751 Immunofluorescence microscopy and quantitative protein and mouse primary hippocampal neurons using transmembrane protein colocalization analysis extraction reagent (TPER) buffer (Pierce) containing diluted protease and Cells were washed twice in PBS, fixed in 4% paraformaldehyde (PFA) in phosphatase inhibitors (1:100; Thermo Scientific). Protein concentrations PBS at room temperature for 15 min, and permeabilized in PBS containing were measured by using a bicinchoninic acid (BCA) assay (Pierce). A total 0.1% Triton X-100 for 10 min. Unspecific antibody binding was prevented of 10–50 µg of total protein lysates were separated using 4–12% Bis-Tris by incubation in blocking solution containing 1.5% (w/v) goat IgG (Zymed) polyacrylamide gel electrophoresis (PAGE; Invitrogen) and subsequently in PBS at room temperature for 30 min. Cells were incubated with the transferred to Hybond-P polyvinylidene fluoride (PVDF) membrane (GE following primary antibodies for 1.5 h at room temperature: anti-V5 mouse Healthcare). Unspecific antibody binding was prevented by incubating monoclonal antibody (R960-25, Life Technologies, 1:200; recognizing the PVDF membrane in blocking solution containing 5% non-fat milk in 1× V5 tag in SEPT8), rabbit anti-BACE1 (D10E5, 5606, Cell Signaling, Tris-buffered saline containing 0.05% Tween-20 (TBST) for 1 h at room 1:250), mouse anti-Rab7 (B-3, sc-376362, Santa Cruz Biotechnology, 1:50; temperature. Primary antibodies were diluted in appropriate dilution in 1× a marker for late endosomes and lysosomes), anti-TfR monoclonal mouse TBST and incubated overnight at +4°C. SEPT8 was recognized by anti- antibody (13-6800, Zymed, 1:200; a marker of early and recycling SEPT8 polyclonal antibody (1:2000, 11769-1-AP; ProteinTech Group), endosomes), and mouse monoclonal anti-GGA3 antibody (612310; BD BACE1 protein was detected by rabbit monoclonal anti-BACE1 D10E5 Transduction Laboratories, 1:50). This was followed by a 1-h incubation at antibody (1:1000, #5606; Cell Signaling Technology). The mouse room temperature with secondary Alexa-Fluor-488-conjugated goat anti- monoclonal antibody against the N-terminal of APP (1:1000, MAB348, rabbit-IgG and Alexa-Fluor-594-conjugated goat anti-mouse-IgG (both at clone 22C11; Millipore) recognized total sAPP (sAPPtot). The soluble 1:500 dilution; Molecular Probes) antibodies. Cell imaging was performed APPα (sAPPα) N terminus was recognized by a mouse monoclonal using a Zeiss Axio Imager microscope connected to ApoTome.2 at 63× antibody (1:1000, 6E10; Biosite). Soluble APPβ (sAPPβ) was detected by a magnification and images were prepared using the Zeiss ZEN 2012 specific rat monoclonal antibody (BAWT, see above). Rabbit polyclonal program. Colocalization of 5′3′UTR-BACE1 and subcellular marker antibody recognizing the APP C-terminus (1:2000, A8717; Sigma) proteins in pcDNA- (control) or SEPT8-V5-transfected cells was identified immature APP (APPim), mature APP (APPm) and APP C- quantified by assessing the number of colocalizing pixels using Zeiss terminal fragments (CTFs; C83 and C99). Anti-Golgi-associated, γ-adaptin ZEN Module Colocalization analysis hardware according to instructions on ear containing, ARF binding protein 3 (GGA3) was detected using a mouse the Zeiss web page (http://www.ecu.edu/csdhs/bsomresearchgradstudies/ monoclonal antibody (1:2500, 612310; BD Transduction Laboratories). CoreImagingCenter/upload/Colocalization_AIM_ZEN.pdf). Briefly, a Glyceraldehyde-3-phosphate dehydrogenase (GAPDH; 6C5) was region of interest (ROI) was manually drawn to include each individual recognized by a mouse monoclonal antibody (1:15,000, ab8245; Abcam). analyzed cell. Every pixel within the ROI was plotted in a scatter diagram Anti-γ-tubulin rabbit polyclonal antibody (1:1000, ab11317; Abcam) and based on its intensity level from each channel. From the scatter diagram, the anti-transferrin receptor mouse monoclonal antibody (1:1000, 13-6800; number of colocalizing pixels was quantified in the ZEN Module Zymed) were used for immunoblotting. Secondary anti-mouse-IgG or anti- Colocalization analysis tool. The values from individual cells were rabbit-IgG antibodies, conjugated to horseradish peroxidase (HRP), were averaged and the data are shown as mean±s.e.m. number of colocalizing diluted in 1× TBST. Enhanced chemiluminescence (ECL) substrate (GE pixels. Numbers of analyzed cells for each transfection and staining are Healthcare) was evenly applied on the top of the membrane and shown in the figure legends. subsequently protein bands were detected with an ImageQuant RT ECL camera (GE Healthcare) and intensities quantified using Quantity One (Bio- Microarray-based global exon expression and splicing screening Rad) software. of human SEPT8 gene in human brain Neuropathological sample cohort Cycloheximide timecourse Human post-mortem brain samples from the temporal lobe were obtained SH-SY5Y-APP751 cells were reverse-transfected using Lipofectamine from Kuopio University Hospital, Kuopio, Finland. The study was approved reagent (Invitrogen) with control plasmid and 5′3′UTR-BACE1, or SEPT8 by the Ethical Committee of the Kuopio University Hospital. Informed and 5′3′UTR-BACE1. Transfected cells were split into four wells in four consent was obtained for all tissue donors and all clinical investigations were six-well-plates and were assessed at the 0, 3, 6, 12 h time-points for the conducted according to the principles expressed in the Declaration of cycloheximide timecourse. Cycloheximide was added to transfected cells at Helsinki. All subjects (n=60) were investigated at the memory clinic and 30 µg/ml. A total of 30 µg protein from each time-point was used for western later autopsied and evaluated for Alzheimer’s disease-related neurofibrillary blotting. The BACE1 levels at each time-point were normalized to the pathology (18 males and 42 females; mean±s.d. age 81±9.0 years). BACE1 level at 0 h and blotted against the appropriate cycloheximide Altogether, 41 subjects out of 60 were diagnosed as having clinical treatment to obtain the half-life of BACE1. The half-life of BACE1 was Alzheimer’s disease according to the National Institute of Neurological and calculated by plotting intensity values (normalized to the BACE1:GAPDH Communicative Disorders and Stroke and the Alzheimer’s disease and value at 0 h) with respect to chase time. Polynomial fitting was subsequently Related Disorders Association (NINCDS-ADRDA) criteria (Table 1) executed to obtain half-life calculations. (McKhann et al., 1984). Neurofibrillary pathology of human post-mortem brain temporal cortex samples were assessed with immunostaining of Biotinylation of cell surface proteins paraffin sections with AT8 antibody, which detects hyperphosphorylated SH-SY5Y-APP751 cells were rinsed twice with PBS supplemented with tau (Braak and Braak, 1991). The brain samples were subdivided into three 0.01 mM CaCl2 and 1 mM MgCl2 (PBS-Ca-Mg), and pre-incubated with groups based on the degree of the Alzheimer’s-disease-related fresh PBS-Ca-Mg for 15 min at 4°C. Sulfo-NHS-LC-Biotin (EZ Link™, neurofibrillary pathology [Braak stages 0–II (n=28), Braak stages III or Pierce) in PBS-Ca-Mg was added to the cells and incubated on a shaker IV (n=13), and Braak stages V or VI (n=19)], in which increased for 30 min at 4°C. Excess biotin was quenched by incubating cells in immunostaining of hyperphosphorylated tau by AT8 antibody denotes more Journal of Cell Science

2236 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215 severe Alzheimer’s disease pathology according to Braak staging (Table 1) Braak, H. and Braak, E. (1991). Neuropathological stageing of alzheimer-related (Braak et al., 2006; Natunen et al., 2013a). changes. Acta Neuropathol. 82, 239-259. The variation of SEPT8 gene exon expression in the inferior temporal Braak, H., Alafuzoff, I., Arzberger, T., Kretzschmar, H. and Del Tredici, K. (2006). Staging of alzheimer disease-associated neurofibrillary pathology using paraffin cortex was investigated using microarray-based global expression and sections and immunocytochemistry. Acta Neuropathol. 112, 389-404. splicing screening (Affymetrix Exon Array; Agilent). Briefly, 100 ng of Brinkmalm, A., Brinkmalm, G., Honer, W. G., Frölich, L., Hausner, L., Minthon, total RNA was amplified and labeled using a Low Input Quick Amp WT L., Hansson, O., Wallin, A., Zetterberg, H., Blennow, K. et al. (2014). SNAP-25 Labeling kit (5190-2943, Agilent). Samples were processed with the RNA is a promising novel cerebrospinal fluid biomarker for synapse degeneration in Spike-in kit (5188-5282, Agilent). Concentration of RNA, and amplified Alzheimer’s disease. Mol. Neurodegener. 9, 53. ’ and Cy3-labeled cDNA were measured and quality confirmed before Cole, S. L. and Vassar, R. (2007). The Alzheimer s disease beta-secretase enzyme, BACE1. Mol. Neurodegener 2, 22. hybridization of 600 ng of the Cy3-labeled samples onto the Agilent 8×60K Coleman, P. D. and Yao, P. J. (2003). Synaptic slaughter in Alzheimer’s disease. custom exon chip overnight at +65°C using the Gene Expression Neurobiol. Aging 24, 1023-1027. Hybridization Kit (5188-5242, Agilent). Next, samples were washed Cummings, J. L. (2004). Alzheimer’s disease. N. Engl. J. Med. 351, 56-67. using the Gene Expression Wash Pack (5188-5327, Agilent) and the chips Das, S. and Pellett, P. E. (2011). Spatial relationships between markers for were scanned using an Agilent Technologies Scanner, model G2565CA. secretory and endosomal machinery in human cytomegalovirus-infected cells versus those in uninfected cells. J. Virol. 85, 5864-5879. Numeric data was produced by Feature Extraction software (version 10.7.3, Das, U., Scott, D. A., Ganguly, A., Koo, E. H., Tang, Y. and Roy, S. (2013). Agilent). Activity-induced convergence of APP and BACE-1 in acidic microdomains via an endocytosis-dependent pathway. Neuron 79, 447-460. Statistical analyses Das, U., Wang, L., Ganguly, A., Saikia, J. M., Wagner, S. L., Koo, E. H. and Roy, IBM SPSS version 21 was used to analyze the data. Statistical comparisons S. (2016). Visualizing APP and BACE-1 approximation in neurons yields insight into the amyloidogenic pathway. Nat. Neurosci. 19, 55-64. of results obtained in experiments were performed by independent sample DeKosky, S. T. and Scheff, S. W. (1990). Synapse loss in frontal cortex biopsies in t-test or Mann–Whitney U-test for two-group comparisons. A comparison Alzheimer’s disease: correlation with cognitive severity. Ann. Neurol. 27, 457-464. of three or more groups was performed using one-way ANOVA followed Di Fede, G., Catania, M., Morbin, M., Rossi, G., Suardi, S., Mazzoleni, G., Merlin, by a Fisher’s least significant difference (LSD) post-hoc test. Results are M., Giovagnoli, A. R., Prioni, S., Erbetta, A. et al. (2009). A recessive mutation in expressed as mean±s.d. when the number of biological replicates was the APP gene with dominant-negative effect on amyloidogenesis. Science 323, n≤6 and otherwise as mean±s.e.m. P<0.05 was considered statistically 1473-1477. Dong, Z., Ferger, B., Paterna, J.-C., Vogel, D., Furler, S., Osinde, M., Feldon, J. significant. and Bueler, H. (2003). Dopamine-dependent neurodegeneration in rats induced by viral vector-mediated overexpression of the parkin target protein, CDCrel-1. Competing interests Proc. Natl. Acad. Sci. USA 100, 12438-12443. The authors declare no competing or financial interests. Fol, R., Braudeau, J., Ludewig, S., Abel, T., Weyer, S. W., Roederer, J.-P., Brod, F., Audrain, M., Bemelmans, A.-P., Buchholz, C. J. et al. (2016). Viral gene ’ Author contributions transfer of APPsalpha rescues synaptic failure in an Alzheimer s disease mouse K.M.A.K. and M.M. carried out the study designing and performed the majority of the model. Acta Neuropathol. 131, 247-266. laboratory experiments, participated in the data analysis and interpretation of data Gozal, Y. M., Seyfried, N. T., Gearing, M., Glass, J. D., Heilman, C. J., Wuu, J., Duong, D. M., Cheng, D., Xia, Q., Rees, H. D. et al. (2011). Aberrant septin 11 is and wrote the manuscript. L.T. and F.R.L. analyzed data. T.N., P.M., F.H., T.S. and associated with sporadic frontotemporal lobar degeneration. Mol. Neurodegener. S.G. participated in the biochemical experiments and data analysis. M.M., M.H., 6, 82. M.K. and J.P. performed bioinformatics data analysis. A.M.K., T.R., V.L. and Hall, P. A. and Russell, S. E. H. (2012). Mammalian septins: dynamic heteromers H.S. collected and analyzed human brain samples. A.H. and M.H. contributed with roles in cellular morphogenesis and compartmentalization. J. Pathol. 226, equally to the work, conceived of the study, participated in study design and analysis 287-299. of the data, and wrote the manuscript. All authors read and approved the final Hall, P. A., Jung, K., Hillan, K. J. and Russell, S. E. H. (2005). Expression profiling manuscript. the human septin gene family. J. Pathol. 206, 269-278. Hanai, N., Nagata, K.-I., Kawajiri, A., Shiromizu, T., Saitoh, N., Hasegawa, Y., Funding Murakami, S. and Inagaki, M. (2004). Biochemical and cell biological This study was funded by the Suomen Akatemia (Academy of Finland); Kuopion characterization of a mammalian septin, Sept11. FEBS Lett. 568, 83-88. He, X., Li, F., Chang, W.-P. and Tang, J. (2005). GGA proteins mediate the Yliopistollinen Sairaala (Kuopio University Hospital) [VTR grant V16001]; Sigrid recycling pathway of memapsin 2 (BACE). J. Biol. Chem. 280, 11696-11703. Juséliuksen Säätiö(Sigrid Jusélius Foundation); the Itä-Suomen Yliopisto (Strategic Hu, X., Li, X., Zhao, M., Gottesdiener, A., Luo, W. and Paul, S. (2014). Tau Funding of the University of Eastern Finland) [UEF-Brain]; VPH Dementia Research pathogenesis is promoted by Abeta1-42 but not Abeta1-40. Mol. Neurodegener. Enabled by IT VPH-DARE@IT of the Seventh Framework Programme [grant 9, 52. number 601055]; and the EADB project in the JPND-CO-FUND program of the Ihara, M., Yamasaki, N., Hagiwara, A., Tanigaki, A., Kitano, A., Hikawa, R., European Commission [grant number 301220]. Tomimoto, H., Noda, M., Takanashi, M., Mori, H. et al. (2007). Sept4, a component of presynaptic scaffold and lewy bodies, is required for the Supplementary information suppression of alpha-synuclein neurotoxicity. Neuron 53, 519-533. Supplementary information available online at Ito, H., Atsuzawa, K., Morishita, R., Usuda, N., Sudo, K., Iwamoto, I., Mizutani, http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.185215/-/DC1 K., Katoh-Semba, R., Nozawa, Y., Asano, T. et al. (2009). Sept8 controls the binding of vesicle-associated membrane protein 2 to synaptophysin. J. Neurochem. 108, 867-880. References Jonsson, T., Atwal, J. K., Steinberg, S., Snaedal, J., Jonsson, P. V., Bjornsson, Ageta-Ishihara, N., Yamakado, H., Morita, T., Hattori, S., Takao, K., Miyakawa, S., Stefansson, H., Sulem, P., Gudbjartsson, D., Maloney, J. et al. (2012). A T., Takahashi, R. and Kinoshita, M. (2013). Chronic overload of SEPT4, a parkin mutation in APP protects against Alzheimer’s disease and age-related cognitive substrate that aggregates in Parkinson’s disease, causes behavioral alterations decline. Nature 488, 96-99. but not neurodegeneration in mice. Mol. Brain 6, 35. Kandalepas, P. C., Sadleir, K. R., Eimer, W. A., Zhao, J., Nicholson, D. A. and Amin, N. D., Zheng, Y.-L., Kesavapany, S., Kanungo, J., Guszczynski, T., Sihag, Vassar, R. (2013). The Alzheimer’s beta-secretase BACE1 localizes to normal R. K., Rudrabhatla, P., Albers, W., Grant, P. and Pant, H. C. (2008). Cyclin- presynaptic terminals and to dystrophic presynaptic terminals surrounding dependent kinase 5 phosphorylation of human septin SEPT5 (hCDCrel-1) amyloid plaques. Acta Neuropathol. 126, 329-352. modulates exocytosis. J. Neurosci. 28, 3631-3643. Kang, E. L., Biscaro, B., Piazza, F. and Tesco, G. (2012). BACE1 protein Barr, A. M., Young, C. E., Sawada, K., Trimble, W. S., Phillips, A. G. and Honer, endocytosis and trafficking are differentially regulated by ubiquitination at lysine W. G. (2004). Abnormalities of presynaptic protein CDCrel-1 in striatum of rats 501 and the di-leucine motif in the carboxyl terminus. J. Biol. Chem. 287, reared in social isolation: relevance to neural connectivity in schizophrenia. 42867-42880. Eur. J. Neurosci. 20, 303-307. Kinoshita, A., Kinoshita, M., Akiyama, H., Tomimoto, H., Akiguchi, I., Kumar, S., Benilova, I., Gallardo, R., Ungureanu, A.-A., Castillo Cano, V., Snellinx, A., Noda, M. and Kimura, J. (1998). Identification of septins in neurofibrillary tangles Ramakers, M., Bartic, C., Rousseau, F., Schymkowitz, J. and De Strooper, B. in Alzheimer’s disease. Am. J. Pathol. 153, 1551-1560. (2014). The Alzheimer disease protective mutation A2T modulates kinetic and Kuperstein, I., Broersen, K., Benilova, I., Rozenski, J., Jonckheere, W., thermodynamic properties of amyloid-beta (abeta) aggregation. J. Biol. Chem. Debulpaep, M., Vandersteen, A., Segers-Nolten, I., Van Der Werf, K.,

289, 30977-30989. Subramaniam, V. et al. (2010). Neurotoxicity of Alzheimer’s disease abeta Journal of Cell Science

2237 RESEARCH ARTICLE Journal of Cell Science (2016) 129, 2224-2238 doi:10.1242/jcs.185215

peptides is induced by small changes in the Abeta42 to Abeta40 ratio. EMBO J. Sannerud, R., Declerck, I., Peric, A., Raemaekers, T., Menendez, G., Zhou, L., 29, 3408-3420. Veerle, B., Coen, K., Munck, S., De Strooper, B. et al. (2011). ADP ribosylation Laitera, T., Sarajarvi, T., Haapasalo, A., Puli, L., Kauppinen, T., Makinen, P., factor 6 (ARF6) controls amyloid precursor protein (APP) processing by mediating Rauramaa, T., Tanila, H., Jaaskelainen, J. E., Alafuzoff, I. et al. (2014). the endosomal sorting of BACE1. Proc. Natl. Acad. Sci. USA 108, E559-E568. Increased gamma-secretase activity in idiopathic normal pressure hydrocephalus Sarajarvi, T., Haapasalo, A., Viswanathan, J., Makinen, P., Laitinen, M., patients with beta-amyloid pathology. PLoS ONE 9, e93717. Soininen, H. and Hiltunen, M. (2009). Down-regulation of seladin-1 increases Lichtenthaler, S. F., Dominguez, D.-I., Westmeyer, G. G., Reiss, K., Haass, C., BACE1 levels and activity through enhanced GGA3 depletion during apoptosis. Saftig, P., De Strooper, B. and Seed, B. (2003). The cell adhesion protein P- J. Biol. Chem. 284, 34433-34443. selectin glycoprotein ligand-1 is a substrate for the aspartyl protease BACE1. Sarajarvi, T., Tuusa, J. T., Haapasalo, A., Lackman, J. J., Sormunen, R., J. Biol. Chem. 278, 48713-48719. Helisalmi, S., Roehr, J. T., Parrado, A. R., Makinen, P., Bertram, L. et al. Loikkanen, J. J., Naarala, J. and Savolainen, K. M. (1998). Modification of (2011). Cysteine 27 variant of the delta-opioid receptor affects amyloid precursor glutamate-induced oxidative stress by lead: the role of extracellular calcium. Free protein processing through altered endocytic trafficking. Mol. Cell. Biol. 31, Radic. Biol. Med. 24, 377-384. 2326-2340. ’ Mandelkow, E.-M. and Mandelkow, E. (1998). Tau in Alzheimer s disease. Trends Scheff, S. W., Price, D. A., Schmitt, F. A. and Mufson, E. J. (2006). Hippocampal Cell Biol. 8, 425-427. synaptic loss in early Alzheimer’s disease and mild cognitive impairment. Martiskainen, H., Helisalmi, S., Viswanathan, J., Kurki, M., Hall, A., Herukka, Neurobiol. Aging 27, 1372-1384. S. K., Sarajarvi, T., Natunen, T., Kurkinen, K. M., Huovinen, J. et al. (2015). Shankar, G. M., Li, S., Mehta, T. H., Garcia-Munoz, A., Shepardson, N. E., Smith, ’ Effects of Alzheimer s disease-associated risk loci on cerebrospinal fluid I., Brett, F. M., Farrell, M. A., Rowan, M. J., Lemere, C. A. et al. (2008). Amyloid- biomarkers and disease progression: a polygenic risk score approach. beta protein dimers isolated directly from Alzheimer’s brains impair synaptic J. Alzheimers Dis. 43, 565-573. plasticity and memory. Nat. Med. 14, 837-842. Marttinen, M., Kurkinen, K. M. A., Soininen, H., Haapasalo, A. and Hiltunen, M. Souza, T. A. C. B. and Barbosa, J. A. R. G. (2010). Cloning, overexpression, (2015). Synaptic dysfunction and septin protein family members in purification and preliminary characterization of human septin 8. Protein J. 29, neurodegenerative diseases. Mol. Neurodegener. 10, 16. 328-335. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D. and Stadlan, Tesco, G., Koh, Y. H., Kang, E. L., Cameron, A. N., Das, S., Sena-Esteves, M., E. M. (1984). Clinical diagnosis of Alzheimer’s disease: report of the NINCDS- Hiltunen, M., Yang, S.-H., Zhong, Z., Shen, Y. et al. (2007). Depletion of GGA3 ADRDA work group under the auspices of department of health and human stabilizes BACE and enhances beta-secretase activity. Neuron 54, 721-737. services task force on Alzheimer’s disease. Neurology 34, 939-944. Vepsalainen, S., Koivisto, H., Pekkarinen, E., Makinen, P., Dobson, G., Munoz-Soriano, V. and Paricio, N. (2007). Overexpression of septin 4, the McDougall, G. J., Stewart, D., Haapasalo, A., Karjalainen, R. O., Tanila, H. drosophila homologue of human CDCrel-1, is toxic for dopaminergic neurons. Eur. J. Neurosci. 26, 3150-3158. et al. (2013). Anthocyanin-enriched bilberry and blackcurrant extracts modulate Musunuri, S., Wetterhall, M., Ingelsson, M., Lannfelt, L., Artemenko, K., amyloid precursor protein processing and alleviate behavioral abnormalities in the ’ Bergquist, J., Kultima, K. and Shevchenko, G. (2014). Quantification of the APP/PS1 mouse model of Alzheimer s disease. J. Nutr. Biochem. 24, 360-370. ̈ brain proteome in Alzheimer’s disease using multiplexed mass spectrometry. Viswanathan, J., Haapasalo, A., Bottcher, C., Miettinen, R., Kurkinen, K. M. A., J. Proteome Res. 13, 2056-2068. Lu, A., Thomas, A., Maynard, C. J., Romano, D., Hyman, B. T. et al. (2011). ’ Natunen, T., Martiskainen, H., Sarajärvi, T., Helisalmi, S., Pursiheimo, J.-P., Alzheimer s disease-associated ubiquilin-1 regulates presenilin-1 accumulation Viswanathan, J., Laitinen, M., Mäkinen, P., Kauppinen, T., Rauramaa, T. et al. and aggresome formation. Traffic 12, 330-348. (2013a). Effects of NR1H3 genetic variation on the expression of liver X receptor Viswanathan, J., Haapasalo, A., Kurkinen, K. M. A., Natunen, T., Mäkinen, P., alpha and the progression of Alzheimer’s disease. PLoS ONE 8, e80700. Bertram, L., Soininen, H., Tanzi, R. E. and Hiltunen, M. (2013). Ubiquilin-1 Natunen, T., Parrado, A. R., Helisalmi, S., Pursiheimo, J. P., Sarajarvi, T., modulates gamma-secretase-mediated epsilon-site cleavage in neuronal cells. Makinen, P., Kurkinen, K. M., Mullin, K., Alafuzoff, I., Haapasalo, A. et al. Biochemistry 52, 3899-3912. (2013b). Elucidation of the BACE1 regulating factor GGA3 in Alzheimer’s disease. Walker, K. R., Kang, E. L., Whalen, M. J., Shen, Y. and Tesco, G. (2012). J. Alzheimers Dis. 37, 217-232. Depletion of GGA1 and GGA3 mediates postinjury elevation of BACE1. Neve, R. L., McPhie, D. L. and Chen, Y. (2000). Alzheimer’s disease: a dysfunction J. Neurosci. 32, 10423-10437. of the amyloid precursor protein(1). Brain Res. 886, 54-66. Zhang, Y., Gao, J., Chung, K. K. K., Huang, H., Dawson, V. L. and Dawson, T. M. Pissuti Damalio, J. C., Garcia, W., Alves Macêdo, J. N., de Almeida Marques, I., (2000). Parkin functions as an E2-dependent ubiquitin- protein ligase and Andreu, J. M., Giraldo, R., Garratt, R. C. and Ulian Araújo, A. P. (2012). Self promotes the degradation of the synaptic vesicle-associated protein, CDCrel-1. assembly of human septin 2 into amyloid filaments. Biochimie 94, 628-636. Proc. Natl. Acad. Sci. USA 97, 13354-13359. Journal of Cell Science

2238