Mutant Cyclin F Impedes COPII Vesicle-Mediated ER-Golgi Tracking and ER-Associated Degradation, Inducing ER Stress and Golgi Fragmentation in ALS/FTD

Mutations in the CCNF gene encoding cyclin F are associated with sporadic and familial amyotrophic lateral sclerosis (ALS) and frontotemporal dementia, but the underlying pathophysiological mechanisms are unknown. Proper functioning of the endoplasmic reticulum (ER) is essential for physiological cellular function. We used human neuroblastoma SH-SY5Y and human embryonic kidney HEK293T cell lines and mouse primary neurons-overexpressing two familial ALS cyclin F mutants to examine whether mutant ALS/FTD-associated cyclin F perturbs key functions of the ER and Golgi compartments. Specic cellular assays were used to examine ER-Golgi transport (VSVG ts045 ), the budding of vesicles from ER membranes and ER-associated degradation (ERAD). Immunocytochemistry was used to examine the morphology of the Golgi and ER-exit sites, and to detect ER stress and apoptosis. Western blotting was used to examine the content of vesicles budding from ER membranes and the interaction between Sec 31 and cyclin F. Flow cytometry was used to examine cell death. It


Abstract Background
Mutations in the CCNF gene encoding cyclin F are associated with sporadic and familial amyotrophic lateral sclerosis (ALS) and frontotemporal dementia, but the underlying pathophysiological mechanisms are unknown. Proper functioning of the endoplasmic reticulum (ER) is essential for physiological cellular function.

Methods
We used human neuroblastoma SH-SY5Y and human embryonic kidney HEK293T cell lines and mouse primary neurons-overexpressing two familial ALS cyclin F mutants to examine whether mutant ALS/FTDassociated cyclin F perturbs key functions of the ER and Golgi compartments. Speci c cellular assays were used to examine ER-Golgi transport (VSVG ts045 ), the budding of vesicles from ER membranes and ER-associated degradation (ERAD). Immunocytochemistry was used to examine the morphology of the Golgi and ER-exit sites, and to detect ER stress and apoptosis. Western blotting was used to examine the content of vesicles budding from ER membranes and the interaction between Sec 31 and cyclin F. Flow cytometry was used to examine cell death.

Results
We demonstrated that mutant cyclin F inhibited protein transport from the ER to Golgi apparatus by a mechanism involving aberrant vesicle sorting from the ER. It also impeded ER-associated degradation, whereby misfolded ER proteins are ubiquitinated and degraded by the proteasome. This was associated with induction of ER stress and Golgi fragmentation, leading to apoptosis.

Conclusion
Together, these results demonstrate that ER dysfunction is a pathogenic pathway associated with ALS/FTD-variant cyclin F.

Background
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by the progressive degeneration of both upper and lower motor neurons. ALS displays genetic and pathogenic overlap with the related condition frontotemporal dementia (FTD) in up to 15% of ALS patients [1,2]. Approximately 10% of ALS cases are familial (FALS), which are clinically indistinguishable from sporadic ALS (SALS).
Mutations in genes including SOD1, TARDBP, and FUS are present in FALS, and we recently identi ed missense mutations in the CCNF gene, encoding cyclin F, in patients with FALS/FTD [3]. Understanding early pathogenic mechanisms is important for designing effective therapeutic targets, but they remain virtually uncharacterised for cyclin F-associated ALS/FTD.
In this study, we demonstrate that expression of mutant ALS/FTD cyclin F in neuronal cells perturbs key functions of the ER. Protein transport between the ER and Golgi apparatus was inhibited in mutant cyclin F expressing cells by a mechanism involving the formation of aberrant coat protein complex II (COPII) vesicles and/or ER-exit sites. This was accompanied by activation of ER stress, Golgi fragmentation, impairment of ERAD and induction of apoptosis. This study therefore reveals novel insights into the pathogenic mechanisms induced by cyclin F mutations in ALS/FTD, involving perturbations to both the ER and Golgi compartments.

SH-SY5Y and HEK293T cell lines
Undifferentiated human neuroblastoma SH-SY5Y and human embryonic kidney HEK293T cell lines (ATCC) were cultured for 24h to 80% con uence at 37°C in a humidi ed atmosphere of 5% CO 2 in Dulbecco's modi ed Eagle's medium (DMEM; Gibco), supplemented with 10% (v/v) heat-inactivated foetal bovine serum (Gibco). Authentication of cell line identities was con rmed via short tandem repeat pro ling (Garvan Institute, Sydney). Cells were transfected for 48h or 72h with plasmids encoding pmCherry-C1 empty vector (EV), WT or mCherry-tagged mutant S621G or S195R CCNF, or co-transfected with plasmids encoding EV, mCherry-tagged WT or mutant CCNF with plasmids encoding GFP-tagged VSVG ts045 , or Venus/ddVenus using Lipofectamine™ 2000 (Invitrogen), following the manufacturer's instructions. Neuronal SH-SY5Y cells were used for immunocytochemistry studies, whereas HEK293T cells were used primarily for Western blotting studies because of their high transfection e ciency.

Mouse primary cortical neurons
Primary neurons were harvested from the cortex of C57BL/6 mouse embryos at embryonic day 16-18. The culture of primary neurons was performed as described previously [27]. Brie y, cortical tissue was dissected, cut into pieces under sterile conditions in Hanks' Balanced Salt solution (Gibco) and digested in 10 units/ml papain (Sigma) in 0.2 mg/ml L-cysteine, 1 mM CaCl 2 and 0.5 mM EDTA (pH 8) in DMEM (Gibco) for 10 min at 37ºC. Cells were subsequently dissociated, resuspended in platting medium (Neurobasal medium [Gibco] supplemented with 10% (v/v) heat-inactivated foetal bovine serum [Gibco], and 100 µg/ml penicillin-streptomycin) and seeded for 12h on 15 mm glass coverslips previously coated overnight with 0.1 mg/ml poly-D-lysine (Sigma). Cells were then incubated in neuronal medium (Neurobasal medium [Gibco] supplemented with and 100 µg/ml penicillin-streptomycin) at 37°C in a humidi ed atmosphere of 5% CO 2 . Half of the medium was changed every three days. After 5 days in vitro, neurons were transfected with constructs encoding EV, WT or mutant S621G or S195R CCNF using Lipofectamine™ 2000 (Invitrogen) following the manufacturer's instructions. Primary neurons were then incubated for 48h before xation in 4% paraformaldehyde (PFA) in 0.1M PBS. Immunocytochemistry SH-SY5Y and HEK293T cells grown on 13 mm coverslips were washed in 0.1 M PBS (pH 7.2) and xed in 4% PFA in PBS for 10 min. After 3 washes in PBS, cells were permeabilised in 0.1% (v/v) Triton X-100 in PBS for 5 min and the non-speci c background staining was blocked, using 3% (w/v) bovine serum albumin (BSA) in PBS for 45 min at room temperature with gentle rocking. Cells were then incubated overnight at 4°C with primary antibodies diluted in 1% (w/v) BSA in PBS: polyclonal rabbit anti-calnexin  Technology). Cells were then washed as above and treated with 0.5 µg/ml Hoechst 33342 reagent (Sigma). After 3 washes in PBS, coverslips were mounted onto slides in uorescent mounting medium (Dako) and cells were photographed with 20x/na = 0.8, 40x/na = 1.3, 63x/na = 1.4 or 100x/na = 1.46 objectives on a Zeiss LSM 880 inverted confocal laser-scanning microscope, equipped with a LSM-TPMT camera (Zeiss). Additional low-resolution images were acquired with 20x/na = 0.5, 40x/na = 0.75, 63x/na = 1.4 or 100x/na = 1.46 objectives on an AxioImager Z2 uorescent microscope (Zeiss) equipped with a monochrome AxioCamHRm digital CCD camera (Zeiss). VSVG ts045 transport assay SH-SY5Y cells co-transfected with constructs encoding cyclin F or VSVG ts045 were incubated overnight at 40 º C to accumulate VSVG ts045 in the ER. Cycloheximide (Sigma) diluted 20 µg/ml in DMEM was then added and the cells were incubated at 32°C for 30 min to allow VSVG ts045 to tra c to the Golgi. After one wash in PBS, samples were xed in 4% PFA in 0.1M PBS (pH 7.2) for 10 min and processed for immunocytochemistry as described above. At least 20 cells expressing both cyclin F and VSVG ts045 were photographed in each group. Mander's coe cient was calculated for each cell to determine the degree of colocalisation (where 0 indicated no colocalisation and 1 indicated total colocalisation) of VSVG ts045 or cyclin F with either calnexin or GM130, using the JaCoP plugin [28] in ImageJ (http://rsbweb.nih.gov/ij/index.html). All experiments were performed in triplicate.

Immunoprecipitation
HEK293T cells transfected for 48h with constructs encoding cyclin F were removed by scraping in 400 µl of non-denaturing lysis buffer (20 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, 1% (v/v) NP40 pH 7.4, 1% protease and 1% phosphatase inhibitor cocktail [Roche]) for 30 min, followed by sonication for 10s on ice and centrifugation at 14,000 g for 10 min at 4°C. The concentration of protein in each lysate was calculated using a Pierce BCA Protein Assay Kit (Thermo Fisher Scienti c) following the manufacturer's instructions. Cell lysate (500 µg of proteins) was incubated with either 2 µg of rabbit polyclonal anti-FLAG antibody (Sigma F7425) or 2 µg of mouse monoclonal anti-ubiquitin antibody (SantaCruz sc-8017) for 1h at 4°C in a rotary shaker. Protein G/A Dynabeads (Thermo Fisher Scienti c 10003D) were washed 3 times in lysis buffer and then incubated with each sample for 2h at 4°C in a rotary shaker. The beads along with protein complexes were separated by placing the tubes in a magnetic rack. After rinsing in lysis buffer, the beads were mixed with 30 µl of Laemmli buffer (BioRad 161-0747) and boiled at 95°C for 5 min before Western blotting was performed. Supernatants (10 µg of both input and ow-through) were mixed with Laemmli Buffer and NuPage sample reducing agent (Novex NP0009) and boiled at 95°C for 5 min.
In vitro budding assay A modi ed in vitro assay was used to analyse ER vesicle budding [23,29]. Brie y, HEK293T cells cotransfected with constructs encoding cyclin F and VSVG ts045 were incubated overnight at 40 º C to accumulate VSVG ts045 in the ER. Cells were then washed in PBS, resuspended in 5/90 buffer (50 mM HEPES and 90 mM potassium acetate in demineralized water) and incubated with rat liver cytosol (Thermo sher) and an energy regenerating system (50 mM creatine phosphate, 0.2 mg/ml creatine phosphokinase and 1 mM ATP) at 32°C for 30 min. Identical samples were incubated at 4°C to monitor non-speci c ER fragmentation. The cells were removed by low-speed centrifugation at 4,000 g for 1 min at 4°C, followed by 15,000 g for 1 min, and budded vesicles were recovered by centrifugation at 100,000 g for 1h at 4°C from the resulting supernatant. The levels of VSVG ts045 cargo in the budded vesicle fractions were quanti ed by Western blotting using anti-VSVG (1:1000; Sigma V4888) and anti-COPII (Sec23; 1:500; Pierce, Rockford, IL, PA1-069A) antibodies. The relative intensity of VSVG and COPII/Sec23 to β-actin was normalised to untreated cells.
ERAD assay SH-SY5Y cells co-transfected with constructs encoding cyclin F and either NHK-Venus, NHK-ddVenus or SS-ddVenus for 48h, were xed in 4% PFA in 0.1M PBS, pH 7.2, and mounted as above. Images were acquired using an Axio Imager Z2 uorescent microscope at 20x/na = 0.8 magni cation. At least 100 cells expressing cyclin F and Venus/ddVenus were scored as the percentage of NHK-Venus, NHK-ddVenus or SS-Venus uorescent cells from three different experiments.

Quantitative analysis of cells
The percentage of cells expressing cyclin F with Golgi fragmentation was quanti ed from 10-30 primary neurons per group and from at least 50 SH-SY5Y or 100 HEK293T cells per group from n = 3 independent experiments. Only cells where the Golgi structure was clearly visible were analysed. The Golgi was considered fragmented when at least 5 fragments were clearly visible. The area covered by the Golgi fragments was calculated using ImageJ.
The percentage of cells displaying nuclear immunoreactivity to CHOP or XBP1 was quanti ed from 30 + primary neurons per group and at least 100 + HEK293T cells per group expressing cyclin F from n ≥ 3 independent experiments. All analyses were performed blind.

Quantitative analysis of apoptotic nuclei
Apoptotic nuclei were de ned as condensed when they were under 5 µm in diameter or fragmented (multiple condensed Hoechst-positive structures in one cell) [30]. The percentage of apoptotic cells was quanti ed from 10-30 primary neurons per group or from at least 100 HEK293T cells per group expressing cyclin F from n ≥ 3 independent experiments. Cells undergoing cell division were excluded from analysis. Sytox Blue staining HEK293T transfected with cyclin F constructs for 72h were harvested by adding trypsin for 1 min at room temperature. The cells were then collected in PBS, centrifuged at 1,200 rpm for 5 min and resuspended in 200 µl of buffer containing 10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 , pH 7.4. The cell suspension was treated with 1 µM Sytox Blue nucleic acid stain (Invitrogen) for 10 min at room temperature in the dark. The cell suspension was then analysed for SYTOX blue positive cells after gating for cells positive for mCherry uorescence, using a BD FACS Canto™ II ow cytometer (BD Biosciences).

Statistics
Data are presented as mean value ± standard error of the mean (SEM). Statistical comparisons between group means were performed using GraphPad Prism 6 software (Graph Pad software, Inc.). One-way or two-way ANOVA followed by post hoc Tukey test for multiple comparisons was used when justi ed. The signi cance threshold was set at p = 0.05.

ALS/FTD mutant Cyclin F inhibits ER-Golgi transport
To probe cellular disease mechanisms induced by mutant cyclin F, we rst expressed familial ALS/FTDassociated mutant cyclin F (Fig. 1A) [3], tagged with mCherry, in HEK293T cells. Western blotting of cellular lysates prepared from cells expressing mutant cyclin F demonstrated no signi cant difference (p > 0.05) between the expression levels of wild-type (WT) and mutant cyclin F (Fig. 1B, C).
Protein tra cking defects have been previously detected in ALS but have not been described before in association with mutant cyclin F [23,31]. Hence, we rst examined if mutant cyclin F inhibits ER-Golgi tra cking. We used a temperature sensitive mutant of vesicular stomatitis viral glycoprotein (VSVG ts045 ) [32,33,23,34], to examine tra cking of VSVG ts045 from the ER to Golgi in cells co-expressing cyclin F and VSVG ts045 (Fig. 1D). VSVG ts045 has been widely used as a reporter of membrane tra cking for canonical ER-to-Golgi transport [35,36]. The transport of VSVG ts045 was examined by immunocytochemistry using anti-calnexin (Fig. 1E) or anti-GM130 (Fig. 1G) antibodies to label the ER and Golgi compartments, respectively. Analysis using Mander's coe cient, where 0 indicated no colocalisation and 1, high co-localisation, demonstrated that VSVG ts045 co-localised signi cantly more with calnexin in mutant cyclin F S621G expressing cells (0.6 ± 0.03), after incubation at the permissive temperature (32°C), compared to control cells expressing either cyclin F WT (0.3 ± 0.03) or empty vector (EV) (expressing mCherry alone) (0.06 ± 0.009) (Fig. 1F). These data therefore indicate that more VSVG ts045 was retained in the ER in mutant cyclin F S621G expressing cells compared to controls. Additional control cells from each group were incubated only at 40°C, con rming that VSVG misfolds and is retained within the ER at 40°C (Fig. 1F). Similarly, in cells expressing mutant cyclin F S621G , signi cantly less VSVG ts045 colocalised with GM130 (0.2 ± 0.02) compared to control cells expressing EV (0.8 ± 0.02) or cyclin F WT (0.6 ± 0.03) (Fig. 1H), where VSVG ts045 extensively colocalised with GM130, indicating e cient ER-Golgi transport. Hence, VSVG ts045 is retained in the ER and less is transported to the Golgi apparatus in cells expressing mutant cyclin F compared to control cells, indicating that mutant cyclin F perturbs ER-Golgi transport.
ALS/FTD mutant cyclin F interferes with the budding of COPII vesicles from the ER ER-Golgi tra cking includes multiple steps, consisting of COPII-dependent vesicle budding from the ER, tethering and docking to target membranes, and nally SNARE protein-dependent vesicle fusion [18,19]. Newly synthesized proteins are exported from the ER via coat protein complex II (COPII)-coated vesicles, which form in specialized zones within the ER, termed ER exit sites (ERES) [16,18,19]. The rst stage of ER-Golgi transport is therefore the formation of COPII vesicles from the ER membranes.
First, to provide evidence that VSVG ts045 is transported from ER to Golgi in COPII vesicles, we repeated the experiment above (Fig. 1), but performed immunocytochemistry for Sec31A (rather than calnexin or GM130), a marker of the newly budded ER-derived vesicles at ER exit sites ( Fig. 2A). Neuronal SH-SY5Y cells expressing GFP-tagged VSVG ts045 were incubated overnight at 40 o C to accumulate misfolded VSVG ts045 followed by 30 min incubation at 32 o C. Following immunocytochemistry, VSVG ts045 was found to be associated with Sec31-positive budding vesicles (Fig. 2B), consistent with its packaging into COPII vesicles to be transported from the ER.
We next used a modi ed in vitro reconstituted assay (Xu and Hay, 2004) to directly assess whether VSVG ts045 -GFP was associated with membranes that form following ER retention, which includes those from COPII vesicles, the ER-Golgi intermediate compartment (ERGIC) and intra/post Golgi carriers (Fig. 2B). This assay therefore aims to reconstitute ER vesicle budding using rat liver cytosol (as a source of soluble COPII) and an ATP regeneration system. ER-derived COPII vesicles are released into the extracellular buffer during incubation [29,37] and semi-intact cells are used as a source of ER, given that the integrity of the ER is preserved under these conditions [29,38,39]. Semi-intact HEK293T cells coexpressing cyclin F and VSVG ts045 for 24h were incubated at 40 o C overnight to retain VSVG ts045 in the ER, then incubated at 32 o C to allow VSVG ts045 to be incorporated into COPII vesicles. The light membranes, including ER-derived vesicles, were then recovered by cellular fractionation (Fig. 2A). VSVG ts045 and Sec23 expression were rst examined in control cell lysates sampled before the budding reaction. No signi cant difference in VSVG and Sec23 levels was detected between all groups (untreated cells or cells expressing EV, WT or mutant cyclin F) (Fig. 2C-E). This indicates that differences in protein expression do not account for the results of the budding assay.
VSVG ts045 was then quantitated in the budded vesicular preparations by Western blotting using an anti-VSVG antibody (Fig. 2F). Signi cantly less VSVG ts045 was present in the preparations derived from cells expressing mutant cyclin F S621G compared to UT cells (4.9-fold) and cells expressing mCherry only (EV) (3.8-fold, Fig. 2G), consistent with impairment in protein transport from the ER to Golgi (Fig. 1). There was a trend towards less VSVG ts045 in preparations derived from cells expressing mutant cyclin F S195R , but this was not statistically signi cant. The presence of COPII was also assessed in the vesicular fractions using anti-Sec23 antibodies (Fig. 2F). Interestingly, signi cantly less Sec23 was also associated with vesicles obtained from cells expressing mutant cyclin F S621G or mutant cyclin F S195R compared to UT (cyclin F S621G : 2.3-fold; cyclin F S195R : 2.3-fold) and EV cells (cyclin F S621G : 1.9-fold; cyclin F S195R : 1.9-fold) (Fig. 2H). These results reveal that less COPII and vesicular cargo (VSVG ts045 ) are present in ER-derived vesicular preparations from cells expressing mutant cyclin F compared to controls. This further implies that vesicular budding between the ER and Golgi is defective in cells expressing mutant cyclin F. ER-derived exit sites are perturbed in cells expressing ALS/FTD mutant cyclin F We next examined the rst step of ER-Golgi tra cking by analysing Sec31-positive clusters of ER exit sites in HEK 293T cells expressing mutant cyclin F by airyscan microscopy (Fig. 3A-C). ER-exit site clusters were identi ed as hollow and spherical structures staining positive for Sec31, with a diameter ranging from 60 to 350 nm. The mean diameter of these clusters was signi cantly decreased (1.2-fold, p < 0.05) in cells expressing ALS/FTD mutant cyclin F compared to UT cells or those expressing either EV or WT cyclin F (Fig. 3B). Signi cantly more small ER-exit site clusters and conversely, signi cantly fewer large ER-exit site clusters were present in cells expressing mutant cyclin F compared to UT, EV or WT cyclin F cells (Fig. 3C). These results suggest that the rst step of ER-Golgi tra cking, vesicle budding from the ER, was perturbed in cells expressing ALS/FTD-mutant cyclin F.
It has been previously established that ubiquitination of Sec31 regulates the size of COPII coats, allowing for the tra cking of large cargo [40], and we previously demonstrated that mutant cyclin F impairs ubiquitination [12]. This raises the possibility that aberrant ubiquitination of Sec31 in cells expressing ALS/FTD mutant cyclin F impacts the size of COPII vesicles, perturbing the transport of large cargoes. The ubiquitination of Sec31 was therefore examined by immunoprecipitation (IP) in cells co-expressing FLAG-tagged Sec31 with either EV or cyclin F proteins (Fig. 3D). IP of cell lysates was performed using an anti-FLAG antibody, and Western blotting for ubiquitin was performed. The ubiquitination of Sec31 was signi cantly decreased in cells expressing mutant cyclin F S621G compared to those expressing mCherry alone (EV, 5-fold) or cyclin F WT (8-fold) (Fig. 3E). These ndings were not due to differences in the expression levels of Sec31 between populations (Fig. 3F, G). Hence, the ubiquitination of Sec31 was signi cantly decreased in cells expressing mutant cyclin F.
To further examine the transport of large cargoes, bulk protein secretion was examined in the conditioned medium from cells expressing cyclin F (Fig. 3E). Silver staining revealed no signi cant difference in the levels of secreted proteins overall in the media. However, when proteins over 100 kDa were speci cally examined, signi cantly less total protein of this size was detected in the conditioned medium from mutant cyclin F expressing cells compared to the other groups (Fig. 3G). These results therefore imply that bulk secretion of large proteins is impaired in cells expressing mutant cyclin F, providing further evidence that mutant cyclin F perturbs protein tra cking between the ER and Golgi compartments.

ALS/FTD mutant cyclin F impedes ER-associated degradation
The results described above demonstrate that one important function of the ER, the transport of secretory and transmembrane proteins from the ER to the Golgi apparatus, is perturbed by mutant cyclin F. ERAD is another important quality control function of the ER that monitors the delity of protein folding, and those proteins that fail to fold or assemble properly are degraded [41,42]. Furthermore, ERAD is a complex, multistep process, that is regulated by ubiquitination, which is defective in cells expressing mutant cyclin F. Hence, we next examined whether ERAD is defective in cyclin F-associated ALS/FTD. ERAD begins with the recognition and targeting of substrates, followed by ubiquitination, retro-translocation and proteasomal degradation. Here we used a substrate with an ER-targeted signal sequence (K b -SS) fused to a mutant version of Venus, ddVenus (deglycosylation-dependant Venus), in which Asp is substituted to Asn at position 82 [26]. This mutation results in glycosylation and a sharp reduction in uorescence, which is restored to wildtype levels when Asn is converted back to Asp. Removal of oligosaccharides by endogenous peptide:N′glycanase (PNGase) in the cytosol results in deamidation of glycosylated Asn, converting it to Asp which restores uorescence [26]. Given that ERAD involves two stages, entry of substrate into the ER followed by its retro-translocation to the cytoplasm, both glycosylation of ddVenus in the ER and its deglycosylation in the cytosol are required for uorescence (Fig. 4A, B). Hence the accumulation of ddVenus uorescence indicates speci c impairment of ERAD. In addition, a second ERAD substrate was used, uorescent ddVenus fused to the null Hong Kong genetic variant of α1antitrypsin (NHK-ddVenus), which misfolds terminally in the ER and is also degraded speci cally by ERAD [25,26,43]. In cells expressing empty vector, cyclin F WT or mutant cyclin F S621G , ERAD was probed using substrates NHK-Venus (a control for transfection e ciency), NHK-ddVenus or SS-ddVenus (Fig. 4C-E).

ALS/FTD cyclin F mutants induce ER stress
The impairment of two important ER functions, protein tra cking to the Golgi and ERAD, implies that mutant cyclin F perturbs the overall homeostasis of the ER. To investigate this possibility further, we next examined whether cyclin F induces ER stress using immunocytochemistry following established methods [44][45][46] (Fig. 5). XBP-1 and CHOP are both transcription factors that translocate to the nucleus when activated during ER stress, and CHOP becomes activated only when the UPR becomes pro-apoptotic [47]. Nuclear immunoreactivity to XBP-1 and pro-apoptotic CHOP therefore indicate activation of the UPR (Fig. 5A). Expression of mutant cyclin F signi cantly increased the proportion of cells with nuclear XBP1 immunoreactivity, compared to cells expressing WT (cyclin F S621G : 2-fold; cyclin F S195R : 2.2-fold) or mCherry only (EV) (cyclin F S621G : 5.9-fold; cyclin F S195R : 6.3-fold) (Fig. 5B, C). Similarly, signi cantly more cells with nuclear immunoreactivity to CHOP were observed in populations expressing mutant cyclin F compared to cells expressing WT (S621G: 1.7-fold; S195R: 1.9-fold) or EV (cyclin F S621G : 5.9-fold; cyclin F S195R : 6.6-fold; Fig. 5D, E). To note, compared to mCherry, expression of WT cyclin F resulted in signi cantly more cells with nuclear immunoreactivity to CHOP, but not to XBP1. These results indicate that ALS/FTD-associated mutations in cyclin F induce ER stress in neuronal cells.
To con rm the ndings obtained in cell lines, ER stress was next examined in primary cortical neurons expressing cyclin F by immunocytochemistry, where nuclear reactivity to CHOP indicated activation of the UPR (Fig. 5F). Quanti cation demonstrated that signi cantly more primary neurons expressing mutant cyclin F displayed nuclear immunoreactivity to CHOP (S621G: 1.7-fold; S195R: 2.2-fold) compared to WT or mCherry (EV) cells (Fig. 5G). Hence, ALS/FTD-associated mutations in cyclin F activate ER stress in primary cortical neurons, con rming the results obtained in cell lines.

ALS/FTD mutant cyclin F induces fragmentation of the Golgi
Fragmentation of the Golgi apparatus results following induction of cellular stress, including ER and oxidative stress [48,49], and when ER-Golgi tra cking is impaired [23,[50][51][52]. Hence, perturbation of ER-Golgi transport and induction of ER stress by mutant cyclin F implies that Golgi fragmentation is also induced. Therefore, we next assessed whether the Golgi apparatus was fragmented in cells expressing mutant cyclin F (Fig. 6). The morphology of the Golgi was examined by immunocytochemistry using an anti-GM130 antibody. In SH-SY5Y (Fig. 6A) and HEK293T (Supplementary Fig. 2A)

cells expressing EV or
Cyclin F WT , the Golgi displayed its typical morphology of continuous stacked membranous vesicles.
Hence Golgi fragmentation was present in cells in which VSVG ts045 secretion was impaired, con rming that inhibition of ER-Golgi secretion correlates with fragmentation of the Golgi.
To further con rm these results and to provide a more unbiased quanti cation method, the surface area covered by the Golgi fragments was next assessed. Golgi stacks can be dispersed (mini-stacks) or completely disassembled, hence the surface area covered by the Golgi also indicates fragmentation [48].
Quanti cation of this area revealed a signi cant increase in the area covered by fragmented Golgi in cells expressing cyclin F S621G (1.9-fold) and cyclin F S195R (2.2-fold) mutants compared to cyclin F WT expressing cells and other controls (Fig. 6C). Similarly, we also examined the area covered by Golgi fragments in the GM130-immunostained HEK293 cells prepared for the VSVG ts045 assay (Fig. 1).
Quanti cation revealed that a signi cant increase in the area covered by fragmented Golgi was present in cells expressing cyclin F S621G compared to WT (2.6-fold) or EV cells (13.5-fold, Supplementary Fig. 3b) again to a similar degree (57.4%, compared to 64.7% in Fig. 6). Hence inhibition of ER-Golgi secretion correlates with fragmentation of the Golgi. Together these results demonstrate that mutant cyclin F induces Golgi fragmentation in neuronal cells.
To further con rm the above results, we next examined mouse cortical primary neurons expressing mutant cyclin F and controls for Golgi fragmentation. Immunocytochemistry of primary neurons expressing cyclin F was rst performed using an anti-GM130 antibody to examine the morphology of the Golgi apparatus (Fig. 6D). Signi cantly more neurons expressing mutant cyclin F displayed Golgi fragmentation compared to those expressing WT (cyclin F S621G : 2.2-fold; cyclin F S195R : 2.9-fold) and mCherry (EV) (cyclin F S621G : 6.3-fold; cyclin F S195R : 8.1-fold) (Fig. 6E). Hence, ALS/FTD-associated mutations in cyclin F induce Golgi fragmentation in primary cortical neurons, con rming the results obtained in cell lines.

ALS/FTD mutant cyclin F induces cell death
ER stress induces apoptosis when prolonged or severe, and Golgi fragmentation is also associated with apoptosis [54,55]. Hence cellular death was next analysed by ow cytometry following Sytox Blue staining, in SH-SY5Y cells expressing cyclin F (Fig. 7A, B). Quantitative analysis demonstrated signi cantly more dead cells in populations expressing mutant cyclin F S621G compared to UT cells (3.7fold) and those expressing cyclin F WT (1.4-fold) or mCherry (3.8-fold), indicating that mutant cyclin F S621G expression induces cell death.
This was next examined in primary neurons, where the presence of condensed nuclear morphology indicated the induction of apoptosis, as previous [44,56]. Examination of the nuclei and quantitative analysis of primary neurons expressing cyclin F demonstrated that the percentage of neurons undergoing apoptosis was signi cantly increased in populations expressing mutant cyclin F compared to those expressing cyclin F WT (cyclin F S621G : 1.9-fold; cyclin F S195R : 1.8-fold) or mCherry (cyclin F S621G : 7.9-fold; cyclin F S195R : 7.3-fold) (Fig. 7C, D). Hence, these data con rm that ALS/FTD-cyclin F mutants induce apoptosis.

Discussion
This study demonstrates that ALS/FTD-associated mutations in cyclin F perturb ER homeostasis by inhibiting two key functions; ERAD and transport of proteins from the ER to the Golgi apparatus. Consistent with these observations, Golgi fragmentation, ER stress, and induction of apoptosis was also detected in cell lines and mouse cortical primary neurons expressing mutant cyclin F. Together these data imply that dysfunction to the ER is an important process in neuronal degeneration induced by mutant cyclin F, thus providing novel insights into CCNF-associated ALS/FTD. ER-Golgi transport is a vital gateway to the endomembrane system because one third of all proteins transit through the ER-Golgi compartments before reaching their nal cellular locations [18,57]. Our results imply that ALS/FTD-associated p.S621G and p.S195R mutations in cyclin F perturb the rst stage of ER-Golgi transport: the budding of COPII vesicles from the ER. The COPII coat is composed of ve separate proteins, including Sec23 and Sec31, and its assembly is essential for the formation of transport vesicles on the cytosolic face of the ER membrane. Curvature of the ER membrane, concentration of cargo and vesicular release then results. Defective COPII vesicles are known to inhibit secretion [40] and here we detected several abnormalities in COPII in mutant cyclin F cells. Less COPII and less vesicular cargo were present in ER-derived membranes obtained from mutant cyclin F expressing cells compared to controls. Furthermore, the Sec31-clusters (representing one or more ER-exit sites) were smaller in these cells, implying that either fewer COPII vesicles were formed overall or that the same number of COPII vesicles were present, but each contained less Sec31. Typical COPII vesicles are 60-70 nm in diameter, and it was not possible to resolve individual COPII vesicles or ER-exit sites using the methods used here. Furthermore, whilst Sec23 levels correlate with the load of protein cargo [40], here the levels of Sec23 were comparable to cargo load, suggesting that the defect is related to aberrant COPII vesicles themselves rather than the incorporation of protein cargo. Hence, these results imply that there is a defect in the formation of COPII vesicles and/or ER-exit sites in mutant cyclin F expressing cells.
COPII vesicles are normally too small to transport some secreted macromolecules, including procollagen brils, pre-chylomicrons, and pre-very low-density lipoproteins (VLDLs). However, ubiquitination of Sec31 by the CUL3-KLHL12 complex allows for the generation of large vesicles to facilitate the transport of these bulky cargo, thus controlling the tra c of COPII vesicles [40,58]. Calcium-dependant control of CUL3-KLHL12 also ubiquitinates lysine residues on Sec31 [40,59]. Similarly, deubiquitylation of Sec31 by USP8 antagonizes the formation of large procollagen-containing carriers [60]. The exact site of Sec31 ubiquitination may not be crucial for regulation of COPII vesicle tra cking, but it is a signal to recruit other effectors to assemble the coat or to regulate its catalytic activity. We previously demonstrated that mutant cyclin F S621G dysregulates ubiquitination at Lys48, resulting in disruption of biological networks responsive for cellular survival and maintenance [11,13]. Here, we demonstrate that in cells expressing mutant cyclin F, ubiquitination of Sec31 was reduced and bulk secretion of large proteins (> 100kDa) was inhibited, although the secretion of proteins smaller than this was unchanged. This together with inhibition of ER-Golgi tra cking in these cells implies that aberrant ubiquitination of Sec31 impairs the formation of COPII vesicles, forming smaller ER-exit site clusters and inhibiting the transport of large cargoes, which would ultimately impact neuronal functions. Future studies are therefore warranted to con rm this possibility and to identify the speci c large cargoes whose secretion is inhibited by mutant Impairment of ER-Golgi and other forms of transport has been previously described in ALS [23,[70][71][72][73], in cells expressing mutant SOD1 [23], mutant TDP-43 [23], mutant FUS [23] or mutant ubiquilin2 [31], and these events were previously linked to ER stress [23]. ER stress has now been widely implicated in ALS and mutations in SOD1 [44], TARDBP [74], FUS [75] [70,71,83], or from the cytoplasmic face of the ER [23] respectively, by inhibiting ER-Golgi transport [23]. Preliminary experiments examining whether cyclin F is present in the ER were inconclusive, raising the question of how ER stress is induced in ALS/FTD. As ALS mutant cyclin F mislocalizes to the cytoplasm where it promotes cytoplasmic aggregation of TDP- 43 [84], it is possible that mutant cyclin F induces ER stress from the cytoplasm by inhibiting ER-Golgi transport.
However, we cannot rule out the possibility that mutant cyclin F may directly induce ER stress from the ER itself, although mutant cyclin F aberrantly ubiquitinates Sec 31, which is also localised in the cytoplasm.
Golgi fragmentation is also a well-described event in ALS [24,53,85] 90] and both membrane ow and cargo load in uence its structure and function.
Hence, blocking the export of cargo-containing ER carriers [91,92] or depleting cargo receptors [93] results in Golgi fragmentation and some of the tubulo-vesicular Golgi clusters can further fuse with the ER, increasing ER stress [94]. In addition, inhibition of intra-Golgi tra cking or vesicle transport from the Golgi to the plasma membrane can result in Golgi fragmentation, if prolonged [50][51][52]. Hence, in this study, it is possible that the inhibition of ER-Golgi transport by mutant cyclin F triggers Golgi fragmentation and ER stress. However, there are many possible cellular stressors than induce Golgi fragmentation so the directionality of these links cannot be conclusively established.

Conclusion
In summary, this study identi es novel cellular mechanisms triggered by ALS/FTD-associated mutant cyclin F. In Fig. 8, we provide one hypothetical model to illustrate the cellular events triggered by mutant cyclin F, based on the ndings of this study. However, it is also possible that ER stress or Golgi fragmentation is the upstream trigger, which would subsequently trigger impairment of ER-Golgi transport and impede ERAD given that these events are all closely related. Alternatively, it is possible that the cellular events detected in this study result from a combination of these defects. Further studies are therefore required to probe the directionality of these links.    Misfolded proteins in the ER accumulate, induce ER stress and activate the UPR. UPR modulators PERK and IRE1 are activated by dimerization followed by phosphorylation, inducing activation of the UPR transcription factor XBP1, resulting in production of its spliced form (sXBP1). Under long-term ER stress, the UPR pro-apoptotic pathway is triggered, leading to the activation of the transcription factor CHOP and expression of apoptotic UPR genes45. PERK, protein kinase R-like endoplasmic reticulum kinase; CHOP, C/EBP-homologous protein; IRE1α, inositol-requiring protein 1α; XBP1, X-box binding protein 1.