C9orf72-associated SMCR8 protein binds in the ubiquitin pathway and with proteins linked with neurological disease

A pathogenic GGGCCC hexanucleotide expansion in the first intron/promoter region of the C9orf72 gene is the most common mutation associated with amyotrophic lateral sclerosis (ALS). The C9orf72 gene product forms a complex with SMCR8 (Smith-Magenis Syndrome Chromosome Region, Candidate 8) and WDR41 (WD Repeat domain 41) proteins. Recent studies have indicated roles for the complex in autophagy regulation, vesicle trafficking, and immune response in transgenic mice, however a direct connection with ALS etiology remains unclear. With the aim of increasing understanding of the multi-functional C9orf72-SMCR8-WDR41 complex, we determined by mass spectrometry analysis the proteins that directly associate with SMCR8. SMCR8 protein binds many components of the ubiquitin-proteasome system, and we demonstrate its poly-ubiquitination without obvious degradation. Evidence is also presented for localization of endogenous SMCR8 protein to cytoplasmic stress granules. However, in several cell lines we failed to reproduce previous observations that C9orf72 protein enters these granules. SMCR8 protein associates with many products of genes associated with various Mendelian neurological disorders in addition to ALS, implicating SMCR8-containing complexes in a range of neuropathologies. We reinforce previous observations that SMCR8 and C9orf72 protein levels are positively linked, and now show in vivo that SMCR8 protein levels are greatly reduced in brain tissues of C9orf72 gene expansion carrier individuals. While further study is required, these data suggest that SMCR8 protein level might prove a useful biomarker for the C9orf72 expansion in ALS.


Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease that afflicts about 1 in 50,000 people each year and involves loss of upper and lower motor neurons [1]. Death typically follows 2 to 3 years after first onset. About 95% of cases are sporadic, while the rest have a family history of the disease. ALS also has overlapping clinical presentations with frontotemporal lobar degeneration (FTLD) and its most common subtype frontotemporal dementia (FTD), a neurological condition affecting the frontal and temporal lobes and marked by cognitive and behavioral impairment [2]. About 20% of ALS patients also exhibit FTLD, and ALS and FTLD have been considered to be part of a continuous disease spectrum [3].
A series of studies have shown that the long isoform of human C9orf72 protein forms a complex with SMCR8 (Smith-Magenis Syndrome Chromosome Region, Candidate 8) and WDR41 (WD Repeat domain 41) proteins [22,[26][27][28][29][30][31][32][33][34][35]. The SMCR8 gene is within the deleted region of chromosome 17 associated with Smith-Magenis Syndrome (SMS), a developmental disorder of children involving intellectual disability, distinctive facial features, and behavioral problems, but no reported motor defects [36,37]. WDR41 is a member of the WDrepeat family of proteins that act as protein-protein or protein-DNA interaction scaffolds for a variety of cellular functions [38]. SNPs within the WDR41 gene region have been associated with human caudate volume [39].
Bioinformatic analyses first identified both C9orf72 and SMCR8 proteins as having DENN (Differentially Expressed in Normal and Neoplastic cells) domains that are present in guanine nucleotide exchange factors (GEFs) for Rabs, multi-functional small GTPases involved in intracellular membrane trafficking and fusion, vesicle formation and transport, and autophagy [40][41][42]. The autolysomal-autophagy pathway involves generation of the autophagosome, an organelle surrounded by a double lipid bilayer. Autophagosomes engulf cytoplasmic components, such as protein aggregates, damaged organelles, and foreign pathogens, and fuse with lysosomes to generate autolysosomes that mediate degradation of the cargo. Autophagosomes also fuse with endosomes, forming an intermediate organelle called the amphisome, before fusion with lysosomes. Various studies have linked wild-type C9orf72 protein with proteostasis, showing that, in complex with SMCR8 and WDR41, it bind Rabs and plays roles in autophagy and initiation of autophagosome formation, as well as being linked by function and colocalization to endocytosis and lysosomal and endosomal trafficking ( [9, 21-23, 26, 28, 29, 31, 32, 34, 43-48], and Discussion for review). A role in the endolysosome pathway has also been shown for the C. elegans C9orf72 ortholog alfa-1 [49]. Aoki et al. [46] linked the interaction of C9orf72 and RAB7L1 with regulation of vesicle trafficking, and WDR41 is necessary for recruitment of the C9orf72 complex to lysosomes [35,50]. Thus, C9orf72 is a regulator of cellular proteostasis.
Additional roles for the C9orf72 complex have also been reported. C9orf72 alters phosporylation of cofilin and activates the small GTPase ADP-ribosylation factor-1/2 (ARF1/2) involved in actin dynamics [51]. Altered C9orf72 protein levels also causes changes in glutamatergic receptor levels, glutamate cycling and endothelin signaling, and excitotoxicity in response to glutamate, as well as widespread transcriptional changes [21,[52][53][54]. However, the consequences of loss of C9orf72 protein for motor neuron function remain unclear. In vivo, diminished motor function and axonal degeneration of motor neurons have been reported in zebrafish and C. elegans depleted of C9orf72 [55,56]. However, subsequent studies detected no or only mild motor function defects in mice deficient for murine C9orf72 ortholog 3110043O21Rik [45]. On the other hand, in a gain-offunction C9ALS/FTD mouse model, Shao et al. [57] found that 3110043O21Rik haploinsufficiency or loss was associated with increased motor behavior deficits in a dose-dependent manner, while Liang et al. [25] reported that Smcr8 knockout (KO) mice displayed motor behavior defects and axonal swelling. While effects on motor function are uncertain, immune system pathology, spleen and lymph node enlargement, defects in macrophage, myeloid and microglial cell function, altered lysosomal trafficking, and decreased body weight and survival have all been reported for C9orf72 or SMCR8 knockout mice [21,28,32,45,[58][59][60][61][62][63][64][65]. Despite these findings, so far no pathogenic loss-of-function coding mutation in C9orf72, SMCR8 or WDR41 genes has been found [66].
To increase understanding of the diverse functions of the C9orf72-SMCR8-WDR41 complex, we sought to determine by mass spectroscopy (MS) analyses the interactome composition of the SMCR8 component. Notably, we found that the SMCR8 complex includes numerous ubiquitin-related proteins and products of genes associated with numerous Mendelian neurological disorders. MS analyses, co-IP experiments, and association of SMCR8 with cytoplasmic stress granules (SGs) in cultured cells support a link between SMCR8 and the ubiquitin pathway. Furthermore, we reinforce previous observations that SMCR8 and C9orf72 protein levels are positively linked, now showing in vivo that SMCR8 might prove to be a useful biomarker for the C9orf72 expansion mutation in ALS patients.
Post-mortem ALS spinal cord and unaffected control brain motor cortex tissues were obtained from Drs. J. Ravits and R. Batra of the Department of Neurosciences, University of California San Diego School of Medicine [74], and C9ALS and unaffected control samples were from the Target ALS Multicenter Postmortem Tissue Core (Table S4). C9ALS samples had been confirmed for the C9orf72 expansion using repeat-primed PCR (RP-PCR) and Illumina Expansion Hunter (M. Harms, Columbia University, pers. comm.). Frozen spinal cord tissues were obtained from the University of Maryland Brain and Tissue Bank of the NIH NeuroBioBank. Frozen C9ALS and unaffected control cerebrospinal fluid (CSF) samples were from the Northeast ALS Consortium (NEALS). CSF was resuspended directly in 3X SDS loading buffer or first concentrated by tricholoracetic acid precipitation, and then analyzed by Western blotting using α-SMCR8 antibodies. Up to 20 μg of CSF total protein were loaded per well.

Protein isolation and immunoprecipitation
For MS sequence determination, HEK 293T cells in T 75 flasks were transfected using FuGENE HD (Promega) with 15 μg of FL-SMCR8, C9orf72-FL, or pcDNA6/myc-His B empty vector and expanded for approximately 45 h, followed by whole cell lysate preparation by sonication using a Diagenode Bioruptor. IP and sample recovery were as previously described [75,76]. Treatment of samples with 25 μg/ml DNase-free RNase (Roche) and 25 μg/ml RNaseA (Qiagen) was conducted in the absence of RNase inhibitors.
For other protein extracts, tissues or cells were lysed in RIPA buffer (Sigma) with Mammalian Protease Inhibitor Cocktail and phenylmethanesulfonyl fluoride (Sigma) and homogenized by Diagenode Bioruptor. For tissues, 2 mm ziconium silicate beads (Next Advance, Inc.) were added to the tubes. Supernatants were recovered by centrifugation at 11K rpm at 4°C for 15 min and resuspended in 3X SDS loading buffer.
Immunostained cells were examined using a Nikon Eclipse Ti-A1 confocal microscope with NIS-Elements AR software.

MS sequencing and data analyses
MS sequencing and database analyses was performed by the Johns Hopkins Mass Spectrometry and Proteomics Facility as previously described [75,76]. Peptide sequences were identified using Proteome Discoverer and Mascot software (Matrix Science) to search the NCBInr 167 database, including gly-gly modifiation on lysine as a variable protein modification. False discovery rate (FDR) was set at 1.0. Mascot search result *.dat files were processed in Scaffold (Proteome Software, Inc.) to validate protein and peptide identifications. Exclusion criteria for proteins are described in the Results section.
We next assessed the efficacy of several commercial antibodies against these proteins. Human C9orf72 expresses a 54-kilodalton (kD) long protein isoform (C9-L) and a 25-kD short isoform (C9-S). It has been noted that commercial C9orf72 antibodies often detect additional bands other than C9-L or fail to detect C9-S [17,19,23]. Consistent with this observation, the Santa Cruz S-14 antibody (α-C9orf72-SC) detected multiple bands in whole cell lysates as well as products of size consistent with both endogenous C9-L and C9-S; only C9-L co-IPed with tagged SMCR8 or WDR41 (Fig. 1a,c). Similarly, in both cultured cells and brain and spinal cord tissue lysates, the Proteintech 22,637-1-AP antibody (α-C9orf72-PT) marked a major band consistent with C9-L (arrow), plus additional products (Fig. S1A). A small number of non-commercial C9orf72-specific antibodies have also been described [17,23,30,79].
Both mouse and human SMCR8 have two predicted isoforms, the full length 105-kD protein and a Cterminal truncated 87.4-kD isoform generated by alternate splicing [36]. Our search of GenBank revealed additional human SMRC8 mRNA isoforms potentially encoding 35.9-kD (accession numbers BC001018, BC005067), 75.2-kD (AK296847.1), and 93-kD (BC101116, BC101117) protein products. Tissue-specific transcripts of various sizes have also been experimentally observed for human SMCR8 [36]. The Proteintech, Bethyl (A304-694A), and Abcam (ab186504 and ab202283) α-SMCR8 antibodies all marked a band consistent in size with full-length SMCR8 (i.e. 105 kD, arrows in Fig. S1B-E), plus additional bands of unknown specificicy, but which could in part relate to the above desribed SMCR8 protein isoforms. The Bethyl and Abcam ab202283 α-SMCR8 antibodies have been used in other studies, and our observations are similar [34,50,65,79]. Interestingly, although expression of full length SMCR8 protein was detected in human brain tissues, none was seen in spinal cord tissue lysates of multiple samples (Fig. S1B-E).
WDR41 has two predicted isoforms of 51.7 and 45.5 kD (Swiss Prot. Q9HAD4-1, Q9HAD4-2). For selected cancer cell lines, both the Santa Cruz (S-12) and Proteintech (26817-1-AP) polyclonal antibodies detected doublet bands consistent in size with these isoforms (Fig. S1F, G). These bands were very faint (Proteintech) or absent (Santa Cruz) from human brain and spinal cord tissue lysates, although bands of larger and smaller sizes were visible by Western blotting of cultured cells.

SMCR8 interactome contains many central nervous system (CNS) disease proteins
Because of the possible non-specific protein interactions described above, we considered commercial antibodies unsuitable for co-IP interactome studies. Therefore, we exploited a co-IP/MS protocol that we have successfully used in previous studies [75,76]. We transfected C9-L with C-terminal FLAG (FL)-tag, full-length SMCR8 with N-terminal FLAG-tag, or empty vector control in HEK 293T cells and performed α-FLAG IP from whole cell extracts in the presence or absence of RNase (Fig. 1d). Complex immunoprecipitated samples were analyzed by liquid chromatography tandem MS. After excluding ribosomal proteins and likely contaminants (such as keratins), 340 and 201 proteins having three or more spectra and not detected in vector only control cell lysates were associated with FL-SMCR8 and C9-L-FL, Fig. 1 Protein interaction analyses by Western blotting and co-IP of the SMCR8 complex in HEK 293T cells (see Fig. S1 for antibody analyses). a Endogenous C9-L (arrow) co-IPs with FLAG-tagged SMCR8. The thick arrowhead marks a band consistent in size with C9-S. b FLAG-tagged C9orf72 co-IPs both endogenous and co-transfected HA-tagged SMCR8. c FLAG-tagged WDR41 protein co-IPs both endogenous C9orf72 and SMCR8 proteins (indicated by arrows). Tagged C9orf72 and WDR41 proteins of (b) and (c) are not visible in whole cell lysates at the Western blot film exposure times shown. d C9orf72-FL, FL-SMCR8, and empty vector were immunoprecipitated on α-FLAG agarose from transfected 293T whole cell lysates, resolved on a polyacrylamide gel, and silver-stained. IP reactions were in the presence or absence of 50 μg/ml RNases. Complex immunoprecipitate samples were analyzed by MS sequencing. Arrows indicate full-length protein bands. Protein molecular weight markers are those of Novex Sharp Pre-stained Protein Standard (Thermo Fisher Scientific)     respectively (Tables 1, S1, S2). Furthermore, 71 proteins were found in both proteomes, although it should be noted that C9-L-FL was expressed at significantly lower levels than FL-SMCR8 (Fig. 1d), as previously reported [29]. Tables S1 and S2 also note interacting partners of C9orf72 or SMCR8 proteins reported in previous studies [19,28,29,31,32,34,51,87,88]. During the course of our investigations, another MS experiment was published that listed 1532 proteins that co-IPed with HA-tagged SMCR8 from 293T cells [34], and a total of 272 of these (80%) were also present in our dataset (Table S2).
To further confirm the effectiveness of our MS analyses, we next analyzed some of the interactors identified. To do that, a subset of cDNAs identified from the SMCR8 proteome were cloned with an N-terminal V5-TEV (tobacco etch virus)-epitope tag or were obtained as gifts. Notably, following cotransfection in 293T cells, 73% (22/30) of proteins tested directly co-IPed with FL-SMCR8 on α-FLAG agarose, further confirming the efficiency of our protocol (Fig. 2). In almost all cases, interactions were resistant to RNase digestion. Some proteins bound non-specifically to the agarose (BAG5, PPP2R1A, RUVBL2) or failed to bind FL-SMCR8 (G3BP1, GTF2I, RAB1A, RANGAP1, STIP1). It is possible some of these latter proteins are only able to bind SMCR8 when in complex with over-expressed C9orf72 and/or WDR41.
Several studies have proposed a role for C9orf72 in the regulation of autophagy by Rab GTPases, although with disagreement concerning which of the many Rab family members binds the C9orf72/SMCR8/WDR41 complex. Farg et al. [43] first reported C9orf72 to interact with RAB1, RAB5, RAB7 and RAB11. Webster et al. [22] confirmed that C9orf72 associates with GTP-bound RAB1A and the ULK1 complex, and it has been demonstrated that C9orf72 in complex with SMCR8 and WDR41 is a GEF for RAB8A, RAB11A, and RAB39B, and that its loss perturbs autophagy in neurons [27,29,31,89]. We detected only RAB1B in our SMCR8 and C9orf72 interactomes (Tables S1, S2), but failed to confirm binding of V5-tagged RAB1A, a paralog highly similar in sequence to RAB1A, with SMCR8 in direct co-IP experiments. However, we also tested and confirmed weak binding of V5-RAB7A with overexpressed SMCR8 (Fig. 2) and C9orf72 (not shown), but only in the presence of RNase.
Significantly, when we queried the OMIM (Online Mendelian Inheritence in Man) database (https://omim. org/), we found that 65 (19%) of our putative SMCR8interacting proteins are associated with neurodegenerative and neurological genetic disorders (Table 2). These include 8 proteins linked with ALS and/or FTD, 14 with other neurodegenerative diseases (including 4 associated with spinocerebellar ataxias), 7 with Charcot-Marie Tooth disease, 5 with hypomyelinating leukodystrophy, and 13 with mental retardation. Thus, SMCR8 may recruit some of these proteins to its complex with C9orf72 and WDR41, predicting roles for the complex in central nervous system (CNS) disorders.
Our SMCR8 interactome also contained 9 ubiquitination pathway factors, including ubiquitin ligases and peptidases (Table 1). Therefore, we examined MSsequenced peptides deriving from immunoprecipitated FL-SMCR8 for ubiquitin modification (72% coverage of the total protein). A total of 9 high confidence modified lysine residues were predicted by at least 5 peptides in two independent experiments, suggesting that SMCR8 is highly ubiquitinated. Eight of these lysines were also identified by at least one of three ubiquitination prediction algorithms, including UbPred [86], BDM-PUB (http://bdmpub.biocuckoo.org), and UbiSite (http://csb.cse.yzu.edu.tw/UbiSite/) (Table S3). We then considered the phylogenic conservation of these lysines by aligning SMCR8 protein sequences from 8 vertebrate (human, chimpanzee, dog, mouse, rat, chicken, zebrafish, and frog) and two mollusc (freshwater snail and sea slug) species (Fig. S3). Eight of the 9 lysines detected by MS as modified were conserved among at least 8 species, including 2 residues (K232, K479) found in both molluscs, suggesting that these post-translational modifications (PTMs) might be functionally relevant.
Immunoprecipitating FLAG-tagged SMCR8 and probing with α-ubiquitin on Western blots reveals highmolecular weight (HMW) proteins consistent with polyubiquitinated SMCR8 and/or other large ubiquitinated proteins bound in the SMCR8 complex (Fig. 3a). In whole cell lysates, SMCR8-V5, in the presence of the proteasome inhibitor MG-132 and/or coexpressed ubiquitin, showed HMW products consistent with multiple PTMs (Fig. 3b). Furthermore, FLAG-tagged ubiquitin coimmunoprecipitates on α-FLAG agarose, and so by implication is conjugated to cotransfected HA-or V5tagged SMCR8 (Fig. 3c). Although treatment with MG132 caused accumulation of HMW SCMR8 protein species, suggesting their regulation by the ubiquitinproteasome system (UPS), full-length SMCR8 signal was little decreased in the presence of coexpressed ubiquitin (Fig. 3b,c).
Using confocal IF microsopy, we observed that overexpression of red fluorescent protein (RFP)-tagged ubiquitin induces formation of a large aggregate consistent with the aggresome and marked by colocalization with coexpressed and therefore likely UBB-bound FL-SMCR8 (Fig. 4a). Aggresomes appear mainly within an indentation of the nucleus at the microtubule-organizing center and form when the protein-degradation machinery of the cell is overwhelmed [90]. Misfolded and ubiquitinated proteins, including perhaps SMCR8, are transported to the aggresome along the microtubule network by means of the dynein motor complex (which includes cytoplasmic dyneins DYNC1H1 and DYNC1I2, both detected in FL-SMCR8 iimmunoprecipitates, Table S2). An Fig. 2 Confirmation of proteins in the SMCR8 complex. Selected proteins detected in the SMCR8 interactome by MS sequencing were tagged and coexpressed with SMCR8 in HEK 293T cells. Most were found to specifically co-IP on α-FLAG agarose with FL-SMCR8 but not empty vector. Approximately 1% of the input lysate (lanes 1, 2) and 30% of the immunoprecipitate (lanes 4-7) were loaded on gels. IP reactions were in the presence or absence of 50 μg/ml RNases. Also included is a panel representative of tagged FL-SMCR8 protein present in the input and IP fractions (detected by α-FLAG antibody) and showing that RNase treatment did not affect SMCR8 immunoprecipitation (lower right). Test proteins were detected by α-V5 antibody, except FL-UBR5, which was detected by α-FLAG antibody (bottom right). The molecular weight of each test protein, including its epitope tag, is shown in brackets. Protein molecular weight markers are those of Novex Sharp Pre-stained Protein Standard (Thermo Fisher) alternative ubiquitin-independent pathway involves interaction of STUB1 and BAG3, which transfer misfolded proteins to heat shock protein 70 (all proteins that co-IPed with FL-SMCR8, Table S2) and the dynein motor complex to promote formation of aggresomes [91,92]. Thus, SMCR8 protein is bound by ubiquitin and may recruit UPS complexes to the vicinity of its other associated cellular proteins, numbers of which have been linked with neuropathologies ( Table 2).

Evidence that endogenous SMCR8 accumulates in cytoplasmic stress granules
The accumulation of neuronal RNA and protein aggregates, including cytoplasmic stress granules, is a pathogenic hallmark of a number of neurodegenerative diseases, among them FTD and ALS [93][94][95]. SGs assemble rapidly under cellular stress and include the small, but not large, ribosomal subunits bound to translation initiation factors such as eIF2 and eIF3 (reviewed in [96]). Processing-bodies (PBs) and SGs are dynamic cytoplasmic aggregates that participate in mRNA decay, and SGs in mammalian cells are heavily ubiquitinated [97]. Because previous publications implicated C9orf72 protein expression in the metabolism of SGs [88,98], we wished to determine if the C9orf72 binding partner SMCR8 associates with SGs in various tumor cell lines.
As reported by others [28], we observed epitope-tagged SMCR8 and C9orf72 proteins to both have a diffuse cytoplasmic distribution with protein also observed in nuclei, although nuclear localization was more evident for C9orf72 (Fig. 4b,c, S4A-C). However, although Maharjan et al. [98] reported that SGs were induced in a majority of unstressed mouse Neuro2A (N2A) neuroblastoma cells when transfected with myc-tagged C9-L, we failed to observe this phenomenon for tagged C9-L or SMCR8 proteins transfected alone or in combination (not shown) in unstressed human osterosarcoma U2OS, HEK 293T, or neuroblastoma cell lines (Fig. 4b, S4B). Furthermore, when cells were treated with 250 μM of the oxidative stressor sodium arsenite (NaAsO 2 ) for 80 min, tagged C9orf72, SMCR8, or WDR41 protein very rarely colocalized in aggregates with endogenous canonical SG marker protein TIA1 in multiple cell lines (Figs. 4c,d, S4C). Fig. 3 Evidence that SMCR8 protein is poly-ubiquitinated. a The FL-SMCR8 construct was transfected in 293T cells and immunoprecipitated with α-FLAG antibody-bound agarose. A Western blot of whole cell lysates probed with α-FLAG antibody shows expression of full-length FL-SMCR8 protein plus HMW products consistent with PTMs (left). Probing with α-UBB antibody marks HMW products in immunoprecipitates consistent with either poly-ubiquitinated FL-SMCR8 protein or the presence of other HMW ubiquitinated proteins that co-IP with the SMCR8 complex (right). IP reactions were in the presence or absence of 50 μg/ml RNases. b C-terminal V5-tagged SMCR8 and empty vector or HA-tagged ubiquitin were coexpressed in 293T cells and treated or not treated with the proteasome inhibitor MG132. Expression of SMCR8-V5 protein and empty vector, in the presence but not absence of MG132, produces HMW bands on Western blots that are consistent with post-translational modification of SMCR8 at multiple sites. SMCR8-V5 protein coexpressed with HA-UBB and without MG132 shows the same HMW bands, which increase in signal intensity upon incubation with MG132. c V5-or HA-epitope-tagged SMCR8 was coexpressed with empty vector or FLAG-tagged UBB in 293T cells and incubated overnight in the presence or absence of MG132. Cell lysates were subjected to immunoprecipitation with α-FLAG agarose, followed by Western blotting and probing with α-HA (top left panel), α-V5 (top right) or α-FLAG (bottom left) antibodies. A HMW smear seen in immunoprecipitates is consistent with poly-ubiquitination of tagged SMCR8 proteins. In general, overexpression of ubiquitin does not lead to a significant decrease in full-length SMCR8 protein levels We next examined localization of endogenous C9orf72 and SMCR8 proteins in cells. The α-C9orf72-SC and α-C9orf72-PT antibodies both detected nuclear and cytoplasmic distribution for C9orf72 protein, with fine cytoplasmic granululazion visible in unstressed cells that was more evident for the latter antibody (Fig. S4D). Fig. 4 Immunofluorescence microscopy shows evidence for association of endogenous SMCR8 protein with cytoplasmic aggregates. a FLAGtagged SMCR8 and RFP-tagged ubiquitin transfected in 2102Ep cells colocalize in a structure consistent with the aggresome. b Overexpression of V5-tagged C9orf72 does not induce stress granule formation in unstressed U2OS cells. c Exogenously expressed HA-SMCR8 protein is not observed in SGs of U2OS cells stressed with NaAsO 2 . d WDR41-FL protein does not colocalize with SG marker protein TIA1 in U2OS cells stressed with NaAsO 2 . e Endogenous C9orf72 protein detercted by the α-SMCR8-SC antibody does not colocalize with SGs in NaAsO 2 -stressed U2OS cells (see also Fig. S4E). f Endogenous C9orf72 protein detected by the C9-L antibody [54] does not colocalize with SGs in DTT-stressed U2OS cells. g,h Endogenous SMCR8 detected by the α-SMCR8-ab202283 antibody localizes to SGs of stressed (h), but not unstressed (g) U2OS cells (see also Fig.  S4G-I). i The α-WDR41-SC antibody does not detect endogenous protein in SGs of NaAsO 2 -stressed 2102Ep cells. NT: no treatment. Cell nuclei were stained with Hoechst 33342 (right-most panels). Size bars are 10 μm However, contrary to previous studies that used these antibodies to report SG localization [88,98], we failed to detect endogenous C9orf72 in stress-induced U2OS (Figs. 4e, S4E) or 2102Ep (not shown) cells, although C9orf72 infrequently justaposed or overlapped with SGs and/or PBs in N2A cells (Fig. S4F.G). To confirm further these observations, two polyclonal antibodies developed by the Robertson lab [19,79], and specific for C9-L (Fig.  4f) and C9-S (not shown) isoforms, were also tested but failed to show obvious C9orf72 protein presence in TIA1-marked SGs in DTT-or NaAsO 2 -stressed cells of multiple lines, including U2OS and 2012Ep cells. Thus, detection of C9orf72 in SGs appears to be cell line and possibly antibody dependent.
We also used the α-SMCR8-PT and α-SMCR8-ab202283 antibodies to examine endogenous SMCR8 protein localization. In unstressed U2OS cells, endogenous SMCR8 was nuclear and more prominently cytoplasmic with speckled staining (Fig. 4g). However, when cells were stressed with NaAsO 2 , SMCR8 redistributed to large intensely staining foci that colocalized with TIA1 (Fig. 4h). Fig. S4H shows SMCR8 protein in large cytoplasmic aggregates of NaAsO 2 -stressed HEK 293T cells that costain with a different endogenous SG marker, the LINE-1 retrotransposon-encoded ORF1 protein [81], while Fig. S4I shows costaining with SGmarker eIF3η in human neuroblastoma SK-N-SH cells. In N2A cells treated with the endoplasmic reticulum stressor thapsigargin, SMCR8 granules were marked by a p70 S6 kinase antibody known to recognize HEDLS/ EDC4, a PB component (Fig. S4J, [77]): PBs frequently overlap or juxtapose with SGs in stressed cells [99]. Endogenous SMCR8 granules in unstressed N2A cells also partially colocalized with GW182 autoantigen, which marks PBs (Fig. S4K) [80]. However, as noted above, SMCR8 commercial antibodies detect multiple protein species (Fig. S1B-E), some possibly non-specific, and we cannot be certain that canonical full-length endogenous SMCR8 proteins are what we see in SGs. Nevertheless, our data suggest that in stressed cells a fraction of endogenous SMCR8 protein is directed to cytoplasmic SGs.
Our analyses showed that TAR DNA binding protein 43 (TDP-43, product of the TARDBP gene) binds SMCR8 (Fig. 2; Table 2). Mutations in TARDBP are involved in about 4% of familial and 1% of sporadic ALS (sALS) cases. However, even wild-type TDP-43, while mostly nuclear in healthy cells, is cleaved and hyperphosphorylated and accumulates in ubiquitinated cytoplasmic aggregates in neurons of almost all ALS and about half of FTLD patients (reviewed in [100]). We tested if endogenous or overexpressed SMCR8 protein colocalizes with TDP-43 protein in cytoplasmic granules but found this not to be the case in unstressed or stressed U2OS or 2102Ep cells (Fig. S4L).
Hexanucleotide expansions within transcripts of the C9orf72 ALS gene may undergo non-conventional repeat-associated non-ATG (RAN) translation and generate dipeptide repeats that aggregate in the cytoplasm of neuronal cells of C9ALS patients (reviewed in [101]). To see if such aggregates might colocalize with SMCR8, we coexpressed in 293T cells FL-SMCR8 and a C9orf72 RAN translation product of 50 GA-dipeptide repeats tagged with EGFP [68]. Overexpressed dipeptide proteins formed one to three large cytoplasmic aggrgates in each cell that were were ringed by, but mostly excluded SMCR8 (Fig. S4M).
Finally, the α-WDR41-SC antibody marks WDR41 protein as predominantly nuclear but also with faint cytoplasmic granules that fail to colocalize with SGs in unstressed or stressed U2OS, 2102Ep, 293T, or N2A cells ( Fig. 4i and not shown). On the other hand, the α-WDR41-PT antibody colocalizes with a minor subset of granules positive for 4-ET, a marker of PBs (Fig. S4N). However, while the α-WDR41-SC antibody recognizes only bands consistent in size with WDR41 isoforms in HEK 293T, 2102Ep, and SK-N-SH cells (Fig. S1F), the α-WDR41-PT antibody detects other non-canonical protein species (Fig. S1G), and the specificity of its SG staining is thus uncertain.
Searching the Mammalian Stress Granules Proteome Database (https://msgp.pt) [102], we found that 18% of the SMCR8 proiein interactome (61/340) and 26% (35/ 201) of the C9orf72 interactome are known SGassociated proteins. It is thus possible that SG components bind endogenous SMCR8-C9orf72 complexes and shepherd them to SGs, although why this would not also be the case for overexpressed exogenous SMCR8 or C9orf72 proteins is unclear.

SMCR8 expression in ALS patient brain tissues
Despite its strong association with protein-degradation factors, SMCR8 overexpression does not stimulate degradation of C9orf72 protein with which it is in complex. Contrarily, multiple studies in cells and knockout mice have shown that protein but not RNA levels of SMCR8 and C9orf72 are positively correlated, suggesting that in complex the two proteins stabilize and protect each other from degradation [26,28,29,32,47,54,65,103]. On the other hand, increased SMCR8 protein reportedly has little effect on WDR41 levels in KO mice or cells [32,35]. We confirmed in 293T cells that overexpression of SMCR8 with various tags strongly increased levels of cotransfected FL-C9-L protein, while cotransfection of empty vectors or an unrelated protein (RO60) did not (Fig. 5a). Considering the interplay between SMCR8 and C9orf72 proteins, and the fact that C9orf72 RNA expression is reduced in some C9ALS patient cohorts, we asked if SMCR8 expression levels are altered in the brains of C9ALS patients compared with non-affected controls.
We first examined transcription levels of C9orf72, SMCR8, and WDR41 genes in RNA-Seq datasets from several sequence read archives that contain C9ALS sample data. GEO dataset GSE67196 includes cerebellum and frontal cortex samples of 9 healthy, 8 C9ALS, and 10 sALS individuals. Using TEtranscripts [84] to analyze C9orf72 gene expression levels, we found a significant log 2 0.96-fold decrease (padj 4.6E-5) in the frontal cortex of C9ALS vs sALS individuals and a 1.1-fold decrease (padj 1.6-E4) in the cerebellum of C9ALS vs control individuals; however, in neither case was decrease in SMCR8 expression significant. The Neuro-LINCS dbGaP Study phs001231 (SRP098831) consists of poly(A) + non-stranded mRNA of iPSC-derived motor neurons from 4 C9ALS, 3 spinal muscular atrophy (SMA), and 3 unaffected individuals (2 or 3 replicates each). No significant changes in C9orf72 or SMCR8 transcript levels were seen in this dataset, although WDR41 sequence read numbers were reduced about 0.35-fold in both C9ALS vs control and SMA vs control samples (padj< 0.01). Finally, a recent RNA-Seq study comparing C9 FTLD and FTLD/ motor neuron disease patients with unaffected control individuals reported a highly significant decrease in C9orf72 RNA levels in C9 FTLD samples; however, this data showed no significant change in SMCR8 or WDR41 RNA expression [24]. We next assayed endogenous SMCR8 protein expression levels in the context of the C9orf72 hexanucleotide expansion. Motor cortex brain tissue lysate samples of 11 C9ALS and 10 unaffected control individuals were analyzed by Western blotting with α-SMCR8 antibodies (Fig. 5b, Table S4). Multiple film exposures were made to optimize signal to noise. Individual band intensities were quantitated with ImageJ software [104] and normalized against the summed exposures of all equivalent bands on the same gel. SMCR8 signal was then normalized to endogenous HSP90 protein signal detected on the same gel after reprobing with α-HSP90 antibody. Remarkably, an average 5-fold reduction in SMCR8 protein signal was seen in C9ALS vs control tissues (Fig. 5b). We also tested by Western blotting cerebrospinal fluid samples from 5 C9ALS patients and 5 unaffected controls, but were unable to detect full-length SMCR8 protein signal with either the α-SMCR8-PT or α-SMCR8-ab202283 antibodies (not shown). We also plotted normalized SMCR8 protein signal against ALS disease duration in months (Table S4), finding a weak negative but non-significant correlation (r = 0.34). Nevertheless, altogether our data recommend further investigation of SMCR8 protein level as a potential biomarker of the C9orf72 expansion disease mutation.

Discussion
In this study we characterized the SMCR8 protein interactome and found it to include numerous components Fig. 5 Expression of C9orf72 and SMCR8 proteins are positively correlated in cell lines and human brain tissues. a C9orf72-FL was coexpressed in HEK 293T cells with 3 different epitope-tagged SMCR8 constructs, FLAG-tagged RO60 protein, or empty vectors (pcDNA3 and pcDNA6 myc/his B). A Western blot of whole cell lysates was probed sequentially with rb α-FLAG, ms α-HA, ms α-V5, and rb α-HSP90 antibodies, the latter as a loading control. At the exposure time for the film shown, expression of C9orf72-FL was not seen in the presence of empty vector or RO60-FL, but signal was robust in the presence of SMCR8. b Western blot of brain motor cortex tissue lysates of C9ALS patients (lanes 1-5) and unaffected control individuals (lanes 6-9) probed with α-SMCR8 and α-HSP90 antibodies. Sample names are shown above the panels (see Table S4). Numbers below the middle panel are normalized ratios of SMCR8 to HSP90 expression determined by ImageJ analysis of band intensities and calculated as described in the text. The lower panel shows the approximately 150-kD unspecified band detected by α-WDR41-SC antibody in human brain tissue lysates (see Fig. S1F): this panel is included only as an additional loading control and is not intended to show expression of canonical WDR41 protein. Approximtely 50 μg of protein was loaded in each lane. c Dot plot of ratios of SMCR8 to HSP90 protein band intensities determined by ImageJ analyses of brain tissues lysates from 11 C9ALS and 10 control individuals. Each sample point is the average of 2 to 4 independent Western blot analyses. A short horizontal line indicates mean values. The presence of a C9orf72 hexanucleotide expansion in each C9ALS carrier individual was confirmed by Columbia University and Target ALS using RP-PCR and Illumina Expansion Hunter, but expansion copy numbers are not known of the ubiquitin-proteasome system, including ubiquitin ligases and peptidases. Of note, the IP method used here exploited FLAG-tagged proteins and so overcame limitations imposed by differences in isoform expression and non-specific protein species recognized by C9orf72 and SMCR8 antibodies. Despite evidence that SMCR8 itself is ubiquitinated at multiple residues, its degradation is not significantly induced in the presence of overexpressed ubiquitin suggesting other roles linking it with the UPS. Recruitment of UPS components to autophagy complexes could be one such role, and our SMCR8 interactome contains 24 autophagy pathway-associated proteins ( Table 1). Ubiquitin plays a fundamental role not only in proteasome-mediated protein degradation but also in the targeting of proteins for degradation by autophagic complexes. Protein ubiquitination also regulates multiple steps of the autophagy pathway (reviewed in [105,106]). For example, E3 ubiquitin ligase STUB1, a protein that co-IPs with SMCR8 (Table 1, Fig. 2), regulates autophagy by targeting TFEB for degradation by the UPS [107]. Also, E3 ligase HUWE1 (Table 1) mediates the ubiquitination and proteasomal degradation of WIPI2, a protein involved in autophagosome formation [108].
Association of the SMCR8-C9orf72 complex with the UPS and autophagy would also be consistent with stress granule localization, since protein ubiquitination regulates SG dynamics. Components of the UPS, including ubiquitin, co-localize with SGs, while proteasome inhibition, and consequent increase in ubiquitinated proteins, induces SG formation [111][112][113]. Recent evidence also suggests SGs are regulated by autophagy [114,115], and it has been proposed that improper metabolism of SGs could be involved in ALS pathology [93,94]. Interestingly, Chitiproulu et al. [88] proposed that C9orf72 protein associates with autophagy cargo receptor p62 (encoded by the SQSTM1 gene) to control SG elimination rather than assembly by forming a complex that eliminates by autophagy SG proteins dimethylated on arginines (of note, we found p62 in the SMCR8 but not C9orf72 interactomes; Fig. 2, Table 2S).
However, our data disagree in some aspects with previously published results concerning C9orf72 colocalization with SGs. While, Maharjan et al. [98] reported that overexpression of myc-tagged C9-L led to the spontaneous appearance of SGs in a majority of N2A cells and cortical neurons in the absence of cellular stress, we failed to reproduce these observations in either U2OS or N2A cells for tagged C9orf72 or SMCR8 proteins, overexpressed together or separately. Furthermore, using the α-C9orf72-PT antibody (Fig. S1A), Maharjan et al. [98] noted that endogenous C9orf72 protein colocalized with a fraction of SGs in neuronal cell lines and cortical neurons in response to DTT and heat shock-induced cell stress, and that C9orf72 depletion inhibited SG assembly, impaired expression of proteins required for their formation, and increased cell sensitivity to stress. However, despite testing several antibodies, cell lines and conditions, we could not detect endogenous C9orf72 in SGs of selected non-neuronal cancer-derived cell lines, and we saw only minor colocalization of C9orf72 with SGs and PBs of N2A cells. Thus, association of C9orf72 protein to SGs appears to be cell line-dependent.
On the other hand, we observed endogenous, but not exogenously expressed, SMCR8 protein localization to SGs of all chemically stressed cell lines tested. Interestingly, about one-fifth of the putative interacting proteins we identified as members of our C9orf72 and SMCR8 interactomes are known SG proteins, which themselves might play a role in targeting of SMCR8 complexes to granules. It is conceivable that SMCR8-C9orf72 SG association is sensitive to cell type, cellular conditions, and levels of interacting proteins as determinants of entry into SGs, and perhaps these factors explain discrepancies between our data and previously published observations.
As reported in other studies, we also presented supporting evidence that C9orf72 protein levels are positively correlated with those of SMCR8 in cultured cells [26,28,29,32,47,54,65,103]. Furthermore, we now show that SMCR8 protein expression is reduced in the brains of C9ALS patients compared with unaffected controls (and as also recently noted by [25]. To date, it has been reported that a small number of proteins, including neurofilament proteins, are differentially expressed in the CSF of ALS and FTD proteins and have been proposed as candidate biomarkers for the C9orf72 mutation [116,117]. Whether or not SMCR8 protein can also be an effective CSF or plasma biomarker for C9 expansion patients remains to be determined and is likely contingent upon the development of better α-SMCR8 antibodies.

Conclusions
In this study we characterized the protein interactome of SMCR8, which binds the protein product of C9orf72, the major susceptibility gene for ALS. Using a robust and highly specific protocol, we demonstrated ubiquitination without significant degradation of SMCR8 protein and its association with many components of the ubiquitinproteasome system. Evidence was presented for localization of endogenous SMCR8 protein to cytoplasmic stress granules, although in several cell lines we failed to reproduce previous observations that C9orf72 protein enters these granules. SMCR8 protein levels were downregulated in whole tissue brain lysates of C9ALS patients compared with unaffected controls, suggesting the potential usefulness for SMCR8 as a biomarker of the disease state.
In addition to ALS and FTD, the C9orf72 gene expansion mutation has been linked with other neurodegenerative and psychiatric disorders, although etiological roles remain unknown [118][119][120][121][122][123]. We have shown that SMCR8, whose cellular levels positively correlate with C9orf72 protein expression, associates not only with many factors of protein metabolism and stress granule dynamics, but also with numerous products of genes linked with a range CNS disorders (65/340 in total, Table 2). It is therefore reasonable in future studies to consider a role for SMCR8 in these diverse neuropathologies, perhaps relating to recruitment of the UPS with consequent effects on protein homeostasis.