Mutations in CHMP2B in Lower Motor Neuron Predominant Amyotrophic Lateral Sclerosis (ALS)

Background Amyotrophic lateral sclerosis (ALS), a common late-onset neurodegenerative disease, is associated with fronto-temporal dementia (FTD) in 3–10% of patients. A mutation in CHMP2B was recently identified in a Danish pedigree with autosomal dominant FTD. Subsequently, two unrelated patients with familial ALS, one of whom also showed features of FTD, were shown to carry missense mutations in CHMP2B. The initial aim of this study was to determine whether mutations in CHMP2B contribute more broadly to ALS pathogenesis. Methodology/Principal Findings Sequencing of CHMP2B in 433 ALS cases from the North of England identified 4 cases carrying 3 missense mutations, including one novel mutation, p.Thr104Asn, none of which were present in 500 neurologically normal controls. Analysis of clinical and neuropathological data of these 4 cases showed a phenotype consistent with the lower motor neuron predominant (progressive muscular atrophy (PMA)) variant of ALS. Only one had a recognised family history of ALS and none had clinically apparent dementia. Microarray analysis of motor neurons from CHMP2B cases, compared to controls, showed a distinct gene expression signature with significant differential expression predicting disassembly of cell structure; increased calcium concentration in the ER lumen; decrease in the availability of ATP; down-regulation of the classical and p38 MAPK signalling pathways, reduction in autophagy initiation and a global repression of translation. Transfection of mutant CHMP2B into HEK-293 and COS-7 cells resulted in the formation of large cytoplasmic vacuoles, aberrant lysosomal localisation demonstrated by CD63 staining and impairment of autophagy indicated by increased levels of LC3-II protein. These changes were absent in control cells transfected with wild-type CHMP2B. Conclusions/Significance We conclude that in a population drawn from North of England pathogenic CHMP2B mutations are found in approximately 1% of cases of ALS and 10% of those with lower motor neuron predominant ALS. We provide a body of evidence indicating the likely pathogenicity of the reported gene alterations. However, absolute confirmation of pathogenicity requires further evidence, including documentation of familial transmission in ALS pedigrees which might be most fruitfully explored in cases with a LMN predominant phenotype.


Introduction
Amyotrophic lateral sclerosis (ALS) is a late-onset, relentlessly progressive and eventually fatal neurodegenerative disorder characterised by the injury and death of upper motor neurons (UMN) in the cortex and lower motor neurons (LMN) in the brainstem and spinal cord [1]. The disorder comprises a range of clinical phenotypes depending on the pathoanatomical distribution of the motor system degeneration. Classical ALS is a combined UMN and LMN disorder. The pure LMN disorder of progressive muscular atrophy (PMA) and the pure UMN disorder of primary lateral sclerosis (PLS) share common molecular pathology hallmarks with ALS, and are considered syndromic variants. The majority of ALS cases are sporadic, although 5-10% of cases are familial. Fifteen loci are known to be associated with ALS, and eight causative genes have been identified, the most common of which is SOD1 (Cu/Zn superoxide dismutase 1) [2]. Recently, we identified missense mutations in CHMP2B (charged multivesicular protein 2B) in two individuals with familial ALS, one of whom had associated features of frontotemporal dementia (FTD) [3]. CHMP2B is expressed in all major areas of the human brain, as well as multiple other tissues outside the CNS. Although the exact function of CHMP2B is unknown, its yeast orthologue, vacuolar protein sorting 2 (VPS2), is a component of the ESCRTIII complex (endosomal secretory complex required for transport) [4]. ESCRTIII is an important component of the multivesicular bodies (MVBs) sorting pathway, which plays a critical role in the trafficking of proteins between the plasma membrane, trans-Golgi network and vacuoles/lysosome [5]. Alterations to VPS2 in yeast results in the formation of dysmorphic hybrid vacuole-endosome structures; additionally, disruption to other ESCRTIII components abolishes the ability of MVBs to internalise membrane-bound cargoes [5,6,7]. Interestingly, two other MND-causing genes, ALS2 and ALS8, are proposed to contribute to motor neuron injury by causing disruption to the processes of endocytosis and vesicle trafficking [8,9,10,11].
Defects in CHMP2B were originally reported in a Danish pedigree with autosomal dominant FTD [4]. The G.C single nucleotide change in the acceptor splice site of exon 6 of CHMP2B affected mRNA splicing, resulting in two aberrant transcripts: inclusion of the 201-bp intronic sequence between exons 5 and 6 (CHMP2B Intron5 ), or a short deletion of 10bp from the 59 end of exon 6 (CHMP2B D10 ). Expression of mutant CHMP2B protein in cells resulted in aberrant structures dispersed throughout the cytosol [4], and the ectopic expression of CHMP2B Intron5 in cortical neurons caused dendritic retraction prior to neurodegeneration [12]. In addition, autophagosome accumulation and the inhibition of autophagy have been seen in cells expressing the CHMP2B mutations found in FTD [12,13].
Several neurodegenerative diseases are now believed to contain an element of dysregulation of the lysosomal degradation of proteins, as reviewed by Martinez-Vicente et al. [14]. The results from these recent studies are part of a growing body of evidence that common pathways are involved in a spectrum of neurodegenerative diseases [15]. It was recently discovered that the ubiquitin positive inclusions seen in FTD and ALS contain the same protein, TAR DNA-binding protein-43 (TDP43), and that mutations in progranulin (PGRN) give rise to FTD with ubiquitinpositive inclusion bodies similar to those seen in some ALS patients [16,17]. It is proposed that mutations in CHMP2B may lead to FTD and other neurodegenerative diseases, including ALS, which is predicted to involve disruption to the cellular processes involved in the recycling and degradation of proteins. However, the status of CHMP2B mutations as a contributor to ALS has remained uncertain, due to the lack of described pedigrees where mutations segregate with disease in multiple affected individuals. The aims of this study were i) to identify whether mutations in CHMP2B contribute significantly to the pathogenesis of familial and apparently sporadic ALS, by screening for genetic alterations in a cohort of 433 ALS cases in whom the clinical phenotype had been documented serially throughout the disease course; ii) to investigate in CNS tissue changes in the gene expression profile of motor neurons from cases with CHMP2B mutations compared to neurologically normal controls, and iii) to investigate the functional effects of CHMP2B mutations in an in vitro system. We propose that CHMP2B missense mutations associate with ALS and demonstrate that these mutations give rise to a distinct clinical and neuropathological phenotype, cause a distinctive alteration in the motor neuron transcriptome and disrupt cellular pathology, compared to controls.

Patients and controls
DNA samples from 433 ALS cases were screened. Of these 37 had familial ALS (FALS) and were negative for mutations in SOD1, TDP43, FUS/TLS, ANG and VAPB; 356 had classical sporadic ALS with UMN and LMN clinical signs and 40 had ALS with a LMN phenotype throughout the disease course (PMA variant). Autopsy CNS tissue was available for 123 of the cases screened and for all the patients in whom a mutation in CHMP2B was found. DNA control samples (N = 500) were obtained from the Sheffield and Birmingham MND DNA banks and the Newcastle Brain Tissue Resource. All control individuals were neurologically normal and matched to the disease cohort by age and sex. Approval for the use of DNA samples was obtained from the South Sheffield Research Ethics Committee and written consent was obtained from the donors. Ethnicity of cases and controls was UK Caucasian.

PCR amplification and mutation screening of CHMP2B
DNA was extracted from snap frozen cerebellum or blood as described previously [18]. PCR was performed as described in the original CHMP2B publication [4]. Following clean-up with ExoSAP-IT (GE Healthcare, UK), PCR products were bidirectionally sequenced [18]. The CHMP2B nucleotide sequence was taken from ENST00000263780 on the Ensembl database, and was used to determine sites of known polymorphisms. Nucleotides were numbered in accordance with the nomenclature recommended by the Human Genome Variation Society (www.hgvs. org).
To determine the prevalence of the c.311C.A change in the control population, exon 3 PCR products were digested with the restriction enzyme AccI, generating 244, 104 & 51bp fragments from the wild type sequence. The C to A substitution abolishes a restriction site, resulting in two fragments (244 & 155bp). The c.618A.C change was screened by digestion of exon 6 PCR products with BanI, which generated 207 & 96bp in the presence of the substitution, whilst the wild type sequence of 303bp remained uncut. The presence of c.-151C.A and c.85A.G in the control population were determined by bidirectional sequencing of exons 1 and 2, respectively.

Neuropathology
Brains and spinal cords were dissected so that one cerebral hemisphere, the midbrain, left hemi-brainstem and left cerebellar hemisphere were sliced for snap freezing. Selected spinal cord segments were also snap frozen. The remaining tissues were fixed in formalin for processing to paraffin wax and used in routine staining and immunocytochemistry from all CNS levels. Standard immunocytochemical methods, including antigen retrieval where appropriate, were used to demonstrate localization of ubiquitin, p62/sequestosome 1, TDP43, CD68 (a marker of microglial activation), a-synuclein and AT8 (a tau marker) ( Table 1). The pathological survey reported here includes examination of cervical and lumbar limb enlargements of the spinal cord, thoracic spinal cord, multiple medulla oblongata and pontine levels to include lower cranial motor nerve nuclei, upper pons and midbrain, the hippocampus, motor cortex, frontal and temporal neocortex and cerebellum.
Microarray analysis of cervical motor neurons from CHMP2B cases vs. controls Snap frozen cervical spinal cord was available for cases 1-3, but not case 4, and the cervical cord from 7 control cases was used for comparison. Ten micron sections were prepared, and 500 motor neurons were isolated from each sample using laser-capture microdissection (LCM); RNA was extracted as previously described [19]. The quality (2100 bioanalyzer, RNA 6000 Pico LabChip; Agilent, CA, USA) and quantity (NanoDrop 1000 spectrophotometer) of the RNA from all of the samples were assessed. Each RNA sample was linearly amplified, following the Eberwine procedure [20] using the Two-cycle Amplification Method (Affymetrix), and again checked for quality and quantity. Fifteen micrograms of amplified cRNA from the 3 mutant CHMP2B cases and 7 controls was fragmented and each hybridised individually to 10 Human Genome U133 Plus 2.0 GeneChips (Affymetrix), as per manufacturer's protocols. Following stringency washes, chips were stained and scanned, and GeneChip Operating Software (GCOS) used to produce signal intensities for each transcript. ArrayAssist (Iobion Informatics, CA, USA) was used to determine genes that showed significant differential expression in the presence of CHMP2B mutations compared to control samples. Transcripts were considered differentially expressed if there was a twofold or greater difference in the mean signal intensity of the CHMP2B cases compared to the control group (p,0.05, two-tailed t test). Transcripts in which the signal intensities on all 10 GeneChips were below 35 (the average intensity of background noise level) were discarded as were those transcripts of unknown function. The differentially expressed transcripts on the resulting list were classified by biological process, as determined by GeneOntology terms. To identify specific pathways affected by CHMP2B mutations, PathwayArchitect (Stratagene) and the DAVID Functional Annotation Tool bioinformatics software packages were used [21].

Validation of microarray results by Q-PCR
Primers were designed and optimised to validate significant changes in expression of genes with key roles in the pathways affected by CHMP2B mutations (Table 2). Q-PCR was perfomed using 25ng of cDNA, 16 Brilliant II SYBR Green QPCR Master Mix (Agilent, CA, USA), the optimised concentrations of forward and reverse primers, and nuclease free water was used to make a final reaction volume of 20ml. Samples were run on an Mx3000P Real-Time PCR system (Stratagene) using previously published parameters [19]. Gene expression values, normalised to actin expression, were determined using the ddCt calculation [22]. Actin was selected as a housekeeping gene as its expression was consistent across the 10 GeneChips. An unpaired two-tailed t test was used to analyse the data and to determine the statistical significance of any differences in gene expression (GraphPad Prism 5, Hearne Scientific Software).
In vitro models to investigate the functional effects of CHMP2B mutations HEK-293 cells (ECACC) were plated on 13mm coverslips in Dulbecco's Modified Eagle's Medium ((DMEM) w Glucose + Lglutamine, w/o Na pyruvate; Lonza) plus 10% fetal calf serum (FCS, Biosera), in the absence of antibiotic (termed +/2). Cells were transfected with 500ng myc-tagged CHMP2B with either the wild-type, c.85A.G (p.I29V), c.311C.A (p.T104N) or c.618A.C (p.Q206H) sequence, under a constitutive cytomegalovirus promoter, using Lipofectamine 2000 (Invitrogen) as per the manufacturer's protocol. After transfection, cells were grown in DMEM plus 10% FCS for 24 hours. To investigate changes in the lysosomal degradation pathway, autophagy was induced in selected cells by serum withdrawal for two hours, 22 hours posttransfection. Cells are subsequently referred to as +/2 (10% FCS, no antibiotic) or 2/2 (no FCS for the last 2 hours of growth posttransfection, no antibiotic). Cells were fixed, (4% paraformaldehyde (Sigma)), permeabilised (0.1% triton (Sigma)), and blocked in 5% goat serum (Sigma) in PBS for one hour. Primary antibodies were mouse a-CD63 (in-house), a marker of late-endosomes and lysosomes, and either rabbit or mouse a-myc (AbCam). Secondary antibodies were anti-mouse or a-rabbit AlexaFluor488 and amouse or a-rabbit AlexaFluor555 (Molecular Probes). Cells were visualised using the Zeiss Axioplan2 microscope and the Zeiss LSM 510 confocal microscope. Transfected cells were scored for the presence of vacuoles and halos. To measure vacuole area, images were captured from a minimum of 25 fields of view per transfection round. ImageJ was used to convert the images to greyscale, subtract the background and adjust the threshold. The 'analyse particles' plug-in was used to measure the area of the vacuoles in mm 2 . COS-7 cells (ECACC) were plated in 6 well tissue culture plates in DMEM, plus 10% FCS, in the absence of antibiotic 24 hours prior to transfection (DMEM+/2). Cells were transfected with 2mg of plasmid DNA, as described above, using Lipofectamine 2000 as per the manufacturer's protocol. To induce autophagy, 22 hours post-transfection cells were serum starved for 2 hours, before being washed in PBS and collected in 500ml 16 Trypsin-EDTA (TE). An equal volume of DMEM+/2 was added to the cells to quench the TE, and cells were pelleted by spinning at 3,0006g for 2 minutes. Cells were washed with 500ml PBS and spun again to re-pellet. The supernatant was discarded and cells were resuspended in 40ml of lysis buffer (25mM Tris pH7.4, 0.5% (v/v) Triton, 50mM NaCl, 2mM EDTA, plus protease inhibitor cocktail) and lysed at 4uC. Protein concentration was estimated using a Bradford assay, and 20mg of total protein per sample was separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (12% acrylamide gels) and transferred to PVDF membranes. Blots were blocked in TBS-T (20mM Tris-HCl pH7.6, 137mM NaCl, plus 0.1% (v/v) Tween-20) and 5% (w/v) dried skimmed milk. They were then probed with rabbit polyclonal anti-LC3 (Stratech) diluted 1:1000 and rabbit polyclonal anti-actin diluted 1:1000 (used as a protein loading control) in TBS-T plus milk for one hour at room temperature, followed by peroxidase-conjugated secondary antibody (1:4000, one hour at room temperature). Antibody binding was revealed using enhanced chemiluminescence, as per the manufacturer's instructions.

Mutation screening of CHMP2B in ALS patients
Sequence analysis of the entire coding region and intron/exon boundaries of CHMP2B from 433 ALS cases identified point mutations in four cases (0.9%) ( Figure 1). One patient (Case 1) was heterozygous for a previously undescribed mutation, a single nucleotide substitution, c.311C.A, which results in the substitution of threonine by asparagine (p.T104N). Two subjects (Cases 2 & 3), were heterozygous for a c.85A.G substitution, resulting in a previously described isoleucine to valine substitution (p.I29V). The fourth patient (Case 4) was the previously published glutamine to histidine (c.618A.C, p.Q206H) case [3]. These mutations are all highly conserved in mammals, with the p.T104N and p.Q206H also conserved in chicken and zebrafish (see Figure S1). The c.311C.A, c.85A.G and c.618A.C changes were absent in 1000 control chromosomes from 500 neurologically normal individuals. The c.618A.C mutation has previously been shown to be absent in 1280 control chromosomes [3]. A novel SNP, c.-151C.A, was also identified in the 59UTR of CHMP2B in four cases (0.9%), however, this change was also present in 2% of controls.

Clinical phenotypes of patients carrying mutations in CHMP2B
In all cases neurological investigations including haematological and biochemical blood tests, CSF analysis, neuroimaging and neurophysiological evaluation were compatible with a diagnosis of ALS. None of the cases presented with or developed signs of frontotemporal dementia (FTD).
Case 1 was a 54 year old man who presented with respiratory failure and bulbar dysfunction. His first problem was dyspnoea whilst swimming. He was treated with non-invasive ventilation (NIV) from the time of diagnosis throughout his illness. There was no reported family history of neurological disease. On examination at the time of presentation, he had a moderately severe dysarthria and bilateral wasting, fasciculation and weakness of the tongue. He had wasting and proximal fasciculation in the upper and lower limbs, weakness of the ulnar and median innervated intrinsic hand muscles and the hip flexors. The deep tendon reflexes were normal and plantar responses were flexor, consistent with a pure lower motor neuron phenotype. His symptoms progressed rapidly over the course of 15 months, at which stage he died of respiratory failure with relatively preserved limb function.
Case 2 was a 64 year old female who presented with back pain and progressive weakness of the right leg. She underwent L4/5 spinal decompression to no avail and her symptoms continued to progress to affect both legs, with later development of upper limb, bulbar and respiratory muscle weakness. There was no significant family history. On examination she had wasting and predominantly distal weakness in the lower limbs with upper limb fasciculations. The reflexes were normal throughout, and the plantar reflexes were flexor. She died 66 months after symptom onset.
Case 3 was a 49 year old man who initially noticed a deterioration in his ability to play football, due to weakness affecting his left leg. His symptoms progressed steadily. Fifteen months after onset he had wasting and fasciculations in the upper limbs with preserved power. The lower limbs were wasted globally, most severely in the left anterior tibial compartment. Power in hip flexion and ankle dorsiflexion was reduced on the left and the ankle jerks were absent bilaterally. He developed bulbar symptoms 24 months after disease onset. No UMN signs were detectable clinically throughout the disease course. There was no family history of ALS, with healthy parents and 5 siblings, including one with developmental delay described as possible autism. He died of respiratory failure 29 months after his initial symptom onset.
Case 4 was a 73 year old male who presented with dysarthria, dysphagia and clumsiness of the hands. His symptoms progressed rapidly and 12 months after onset he had a wasted, fasciculating tongue, LMN weakness in the upper and lower limbs, depressed reflexes and flexor plantar responses. He died 14 months after disease onset. He had a cousin who had died from ALS several years earlier.

Neuropathological findings in the four cases with CHMP2B mutations
The anatomical and molecular pathology of all four cases was similar and corresponded to a rather stereotypical pattern of LMN predominant degeneration (Figure 2). Involvement of UMNs was not detectable in conventional stains and the Betz cell somata were readily identified and appeared normal in all cases. There was no evidence of myelin pallor in the corticospinal tracts ( Figure 2a) other than equivocal changes in the medullary pyramid of one case (Case 1), which was not present in the spinal cord at any level. Microglial activation demonstrated by CD68 staining is a more sensitive indication of early white matter tract degeneration [23]. Three of the cases showed no evidence for corticospinal tract degeneration at any level. One case (Case 1) showed marked subcortical microglial activation centred on the precentral gyrus, mild changes in the medulla and only equivocal changes in the spinal cord white matter. Glial pathology was previously reported in the motor cortex in case 4, comprising oligodendroglial coiled bodies immunoreactive for p62 and TDP43, which were not readily demonstrated by conventional ubiquitin immunocytochemistry [3]. These lesions were present in the motor cortex and the ventral horns of all the cases. These lesions are not confined to CHMP2B-related ALS as previously reported [3], but are consistently present to a variable degree in sporadic and non-SOD1-related familial variants of ALS [24].
The LMN pathology in all cases was typical of the primary muscular atrophy (PMA) variant of ALS. There was variable severe loss of motor neurons from all spinal levels. Bulbar motor nuclei were generally less severely affected. Surviving motor neurons contained a population of inclusion bodies demonstrated by ubiquitin, p62 and TDP43 staining. In three cases these were exclusively of 'compact' morphology with no demonstrable 'skeins' (Figures 2d,g,h, i). All the glia and neurones containing TDP43 intracytoplasmic inclusions showed decreased staining in the cell nucleus in comparison to retention of normal nuclear staining in cells negative for inclusions. One case showed very few intraneuronal lesions, in the context of massive motor neuronal loss, of rather indeterminate morphology. Bunina bodies were absent from all the cases. Extramotor involvement of the CNS assessed by p62/TDP43 immunoreactive lesions was absent including evaluation of the hippocampal dentate granule cells and the frontal and temporal neocortex. Staining for related neurodegenerative pathologies (tau and alpha-synuclein) showed minimal expression of Alzheimer's disease pathology (low Braak stage and minimal cortical amyloid deposition) and absence of synucleinopathy.
Gene expression profiling of motor neurons from CHMP2B cases compared to controls The percentage of genes described as present on the 10 GeneChips was an average of 22.7% for the three available CHMP2B cases (26 p.I29V, 16 p.T104N), and 28.2% for the seven neurologically normal control cases. (CEL files for each of the 10 GeneChips have been submitted to the Gene Expression Omnibus Repository, Accession GSE19332). ArrayAssist (Affymetrix) was used to analyse expression data. Duplicate probes were removed, as were probes for unknown genes, and those for which the signal intensity in all 10 GeneChips was below 35 (the average background level). After filtering, 890 genes were found to be downregulated, and 555 upregulated, all with a fold change $2, and p value #0.05. These 1,445 differentially expressed genes were then categorised according to their biological process, as defined by GeneOntology terms (Table 3) (Full list available in Table S1). An additional 398 genes were downregulated and 358 upregulated at a fold change of 1.5-1.99, p#0.05. These were only considered if they were involved in the pathways of interest.
The functional annotation tool of the DAVID Bioinformatics Resource was used to identify pathways with a significant number of differentially expressed genes, namely: axon guidance, regula-tion of actin cytoskeleton and SNARE interactions in vesicular transport, mammalian target of rapamycin (mTOR) signalling and regulation of autophagy, mitogen activated kinase (MAPK) signalling, calcium signalling, and cell cycle and apoptosis (Table 4). We focused our analysis on pathways that were of most biological interest in relation to the predicted function of CHMP2B and related proteins.
Fusion of ER-derived vesicles with the Golgi requires the pairing of the v-SNARE, blocked early in transport 1 (BET1), with its t-SNARE complex comprising syntaxin 5 (STX5), Golgi SNAP receptor complex member 2 (GOSR2;X2.76) and SEC22A, B (X1.6), C. SEC23 and SEC24, whose function is to coat the vesicles travelling along microtubules to the cis-Golgi, are also upregulated (X1.5 and X1.8 respectively). In contrast, USE1 (unconventional SNARE in the ER homologue 1), which is involved in retrograde transport from the Golgi to the ER [26], is downregulated (Y2.66). Furthermore, two components of the coatomer protein complex I, COPE and COPZ1 are downregulated (Y3.8 and Y3 respectively). This complex is responsible for coating of the vesicles travelling between ER and Golgi. Also downregulated are the t-SNAREs, VTI1 (Y2.9) and STX4 (Y2.2), which encode proteins involved in vesicle docking and transport from the GA to the endosome. In addition, many of the adaptor-related protein complex transcripts are strongly downregulated, AP1B1 (Y2.29), AP2A2 (Y2.25), AP2S1 (Y3.35), AP3D1 (Y2.67) and AP4B1 (Y2), and these also play a key role in the transport of vesicles from the GA to the endosomal sorting pathway. KIF1A, TUBB and MAP4 were selected for verification by Q-PCR and were found to be significantly downregulated by 2.96 (p = 0.0009), 3.07 (p,0.0001) and 3.52 fold (p = 0.0003), respectively (Table 5).
MAPK signalling (Figure 4). There are two main pathways through which MAP kinases signal: the classical MAPK pathway and the p38 MAP kinase pathway [35]. The first step in the activation of classical MAPK signalling is ligand binding to a receptor tyrosine kinase. The phosphorylated tyrosine of the target protein is bound by the SH2 domain of GRB2, which also contains an SH3 domain that binds to the proline-rich region of SOS. Once bound by GRB2, SOS catalyses the substitution of GDP to GTP on RAS (Y4.96). GTP-bound RAS activates the MAPKKK, RAF1 (Y2.10), by phosphorylation. RAF1 subsequently phosphorylates the MAPKKs, MEK1/2, which in turn activate the MAP kinases ERK1 and ERK2 (Y1.71). MAPK interacting serine/threonine 2 (MKNK2) is directly activated by ERK, and is itself downregulated (Y3.24). MKNK2 contributes to the basal phosphorylation of eIF4A, which is essential for translation initiation [36]. Downregulation of these core pathway components suggests repression of the basal response to growth factors, an effect that appears to be amplified by upregulation of the Ras inhibitor, NF1 (X1.81) and the ERK1/2 inhibitor, PTPN5, (X1.59).
Activation of p38 is mediated by multiple upstream kinases and this pathway plays a key role in the cell's stress response as well as the phosphorylation of multiple target proteins including phospholipase A 2 and MAP tau. TAK1 (Y2.42) is an upstream kinase that phosphorylates MAPK kinases 3 and 6 (MEK3, 6), which subsequently phosphorylate p38 (Y5.34). There is strong down- Table 3. Grouping of 891 downregulated and 556 upregulated genes, with a fold change (FC) $2, and p value #0.05, into their biological process, as defined by GeneOntology terms.   regulation of the p38 MAPK signalling pathway, which is strengthened by upregulation of the p38 (and ERK1/2) inhibitor PTPN5 (X1.59). Additionally, the p38 target, MK2 is downregulated (Y1.60), which plays a role in the regulation of mRNA stability and the reorganisation of actin [35]. MK2 activates Hsp27 (Y12.85), which binds to and inactivates the pro-apoptotic molecules caspase-3, caspase-9 and cytochrome c [37]. Calcium signalling (Figure 5). The G protein coupled receptor, cholinergic receptor, muscarinic 3 (CHRM3), is upregulated (X2.92) in CHMP2B mutant motor neurons. Muscarinic 3 (M 3 ) receptors couple to phospholipase Cb (PLCb) via the G q class a-subunits [38]. However, as both G q and G q11 are downregulated (Y3.08 and Y2.51, repectively), this would predict a decrease in PLCb activation, despite the increase in CHRM3 transcription. PLCb functions by catalysing the hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) to inositol 1,4,5triphosphate (IP 3 ) and diacylglycerol (DAG). IP 3 binds to the IP 3 receptor (IP 3 R, Y1.93), which is located on the ER membrane, resulting in the opening of a IP 3 -gated Ca 2+ -release channel, thus allowing calcium to exit the ER lumen and enter the cytoplasm, where it can propagate the signal by activating protein kinase C (PKC). Following depletion of ER stores, calcium is reaccumulated through sarco-endoplasmic Ca 2+ ATPase (SERCA, X2.12), which is directly controlled by the ATP supplied by mitochondria [39]. ANT3 catalyses the exchange of ATP for ADP from the mitochondrial matrix through the inner mitochondrial membrane into the inter-membrane space, and is downregulated (Y1.82). Solute carrier family 8, sodium/calcium exchanger 1  (2). This is enhanced by upregulation of STMN1, a MT-destabilising protein, whose over-expression has also been demonstrated to result in Golgi fragmentation. Downregulation of transcripts maintaining Golgi structure (GLG1, GORASP1, COG2 and COG7) support the hypothesis of Golgi fragmentation (3). COPI is used to coat empty vesicles exiting the Golgi for recycling back to the ER, however, two main constituents, COPE and COPZ1, are downregulated, as is USE1, which is required for vesicle fusion with the ER. These findings predict an eventual deficit of material available to ER for the packaging of newly synthesised proteins (4). There is dysregulation of multiple SNARE transcripts, which are required for vesicle fusion. BOS1 and SEC22 are upregulated, which may be the Golgi's response to the reduction in vesicles being transported along destabilised microtubules. Multiple SNAREs and adaptor proteins (VTI1, STX4, AP1B1, AP2A2, AP2S1, AP3D1 and AP4B1), which are required for fusion between vesicles carrying mature proteins and the cell surface, endosomes and lysosomes, are downregulated, predicting impairment in the delivery of proteins throughout the cell (5). Finally, inhibition of autophagy by the mTORC1 complex and downregulation of ATG1, which forms the phagophore assembly site (PAS) with ATG17 and ATG13 to initiate autophagy, indicates a decrease in the clearance of cellular debris which may result in cytosolic accumulations and contribute to motor neuron injury (6). doi:10.1371/journal.pone.0009872.g003 (NCX1, X3.21) and voltage-dependent anion channel 3 (VDAC3, X2.93), are located on the inner mitochondrial membrane and outer mitochondrial membrane, respectively, and pump calcium out of the mitochondrial matrix and into the cytoplasm, where it is transported to the ER lumen via SERCA [39]. NCX1 and SERCA were selected for verification by Q-PCR, and were found to be upregulated by 25.75 fold (p,0.0001) and 1.66 fold (p = 0.0044), respectively (Table 5).

Cellular pathology of CHMP2B mutations
To identify pathogenic effects of CHMP2B mutations, HEK-293 cells were transfected with constructs containing myc-tagged fulllength CHMP2B with either wild-type or p.I29V, p.T104N or p.Q206H mutant CHMP2B. 24 hours after transfection, cells expressing wild-type CHMP2B showed a typically diffuse pattern of staining throughout the cell ( Figure 6A). Cells expressing mutant CHMP2B isoforms had large cytoplasmic vacuoles that were devoid of lumenal staining for the recombinant protein, such vacuoles were rarely seen in cells expressing wild-type protein  ( Figure 6B-D). Image analysis showed there were significantly more vacuoles with an area greater than 1mm 2 in cells expressing mutant CHMP2B protein compared to wild-type (one way ANOVA: p,0.0001, Figure 7). Additionally, the presence of mutant CHMP2B increased the number of cells with vacuoles greater than 1mm 2 in area (Bonferroni post-test, p,0.001), as did serum withdrawal (Bonferroni post-test, p,0.0001). A second phenotype observed in the mutant cells was the presence of vacuoles with an accumulation of CHMP2B mutant protein on the outer membrane, termed halos ( Figure 6E). The number of cells with halos in cells expressing the p.T104N isoform of CHMP2B was significantly increased (Bonferroni post-test, p,0.05). When cells were deprived of serum the presence of halos was even more striking in mutant cells. Co-staining with CD63, a tetraspanin that is abundant in late endosomes and lysosomes, revealed an interesting change in staining pattern in cells expressing mutant CHMP2B. In cells expressing WT CHMP2B, CD63 co-localises with small vacuoles within the cytoplasm ( Figure 6F-H). In cells expressing mutant CHMP2B, CD63 does not co-localise with the large cytoplasmic vacuoles caused by mutant CHMP2B transfection, but instead is found on the membrane of these aberrant structures ( Figure 6 I-K). Immunoblotting of LC3 typically reveals two bands, LC3-I (18kDa) and LC3-II (16kDa). During the formation of autophagosomes, the cytoplasmic form of LC3 (LC3-I) is recruited, where it undergoes site-specific proteolysis and lipidation, generating LC3-II which sequesters to the membrane of the autophagosomes [40]. The level of LC3-II can be used to monitor autophagic activity, as it correlates with the number of autophagosomes. Western blotting for LC3 in COS-7 cells showed a significant increase in LC3-II levels in cells expressing mutant CHMP2B compared to those expressing wild-type protein (Figure 8).

Discussion
In our cohort, we have identified mutations in CHMP2B in four out of 433 individuals with ALS, giving a frequency of just less than 1%. This is approximately half of the frequency of SOD1 mutations, which have been reported to account for approximately 20% of familial ALS cases [41], and 2% of all ALS cases [42,43,44]. Of note, only 1 of our 4 cases of CHMP2B associated ALS/MND had a discernible family history compatible with ALS in a second-degree relative. This supports the hypothesis that some apparently sporadic ALS cases have a genetic component. Evidence from a UK twin study, which examined concordance in both mono-and dizygotic twins (having first excluded probands from families in which dominant inheritance of MND had already been identified), estimated the heritability of ALS to be between 0.38 and 0.85; indicating that genetic factors are likely to make a substantial contribution to the sporadic form of the disease [45]. Although the present study has not been able to document the segregation of CHMP2B in multiple affected members of specific pedigrees in ALS, we believe our results support the body of evidence for the contribution of genetic factors to apparently sporadic ALS.
All 4 cases in this report were negative for changes in SOD1, ANG, TDP43, VAPB and FUS/TLS. Clinically, all four cases presented with a phenotype consistent with a lower motor neuron phenotype of ALS. In this cohort only 40 cases are recorded clinically as having a lower motor neuron predominant phenotype and CHMP2B mutations were found in 4 (10%) of these cases. In our autopsy cohort, 15/123 cases had a LMN phenotype pathologically. Of these 15, three lack any evidence of ubiquitin/TDP43 neuronal inclusion pathology, one in the presence of a mitochondrial transfer RNA gene mutation [46], and another as one of a pair of brothers with MND and colonic neoplasia recently found to have a mutation in FUS/TLS [47]. As such CHMP2Brelated ALS is highly over-represented in this LMN predominant clinical subgroup. It is interesting to note that although mutations in CHMP2B were originally identified in patients with FTD, none of the four cases we identified had clinically apparent cognitive changes and there were no noteworthy pathological changes in the hippocampus.
A previous report of 166 familial and 372 sporadic ''classical'' ALS cases of predominantly Anglo-Celtic origin from Australia and London found no evidence of mutations in CHMP2B [48]. However, population differences in other ALS pathogenic mutations have been previously reported as exemplified by the lack of TARDBP mutations in some populations and [49,50] the Combining the decrease in IP3, with downregulation of its receptor, IP3R (Y), would reduce the opening of a Ca 2+ -release channel on the ER membrane, which allows Ca 2+ to exit the ER lumen into the cytosol. SERCA, which pumps Ca 2+ out of the cytoplasm into the ER lumen is upregulated, so these findings predict an increased Ca 2+ concentration in the ER lumen. Upregulation of NCX1(X) and VDAC3(X), and downregulation of ANT3(Y), predict that mitochondria are increasing the amount of calcium pumped out of the matrix, but decreasing the amount of ATP leaving the organelle. Amplifying the aberrant intracellular calcium levels, in addition to reducing the amount of energy available for cellular processes. (PM = plasma membrane; OMM = outer mitochondrial matrix; IMM = inner mitochondrial matrix). doi:10.1371/journal.pone.0009872.g005 very low frequency of SOD1 mutations reported in some countries such as the Netherlands [51].
In silico analysis of the identified amino acid substitutions predicts that the p.I29V and p.Q206H mutations decrease protein stability (http://gpcr2.biocomp.unibo.it/cgi/predictors/ I-Mutant2.0/I-Mutant2.0.cgi). The native threonine of the p.T104N change is predicted to be a site of phosphorylation (www.cbs.dtu.dk/services/NetPhos), thus substitution of threonine with asparagine is predicted to affect protein activity. We did not identify any of the described codon changes in our 500 controls, whilst previous studies have sequenced 1495 samples for exon 3 and 2035 samples for exon 6, without identifying the p.T104N and p.Q206H substitutions [3,4,48,52]. We therefore propose that these changes are not rare benign polymorphisms, but are associated with disease. Although the p.I29V substitution has been reported in a single control sample, as well as in a familial FTLD case [53], screening of our 500 controls, 640 previously screened controls, 708 ALS cases and 546 FTD cases [3,4,48,52] has failed to detect this change. Therefore, whilst it is recognised that not all missense mutations are pathogenic [54], the clinical, neuropathological and cellular phenotype common to all 3 CHMP2B mutations, and which is distinct from controls, supports the proposal that all 3 nucleotide substitutions described in this report are associated with a lower motor neuron dominant-ALS. The c.-151C.A polymorphism is predicted to alter the binding site for an unknown transcription factor (www-bimas.cit.nih.gov/ molbio/index.shtml). However, there is no significant difference in frequency between subjects and controls (Chi Square p = 0.9), and therefore this change is likely to represent a non-functional polymorphism.
The pathology is rather stereotypical and, whilst firmly within the ALS/MND spectrum, appears to represent a rather distinctive lower motor neuron variant. The inclusion morphology does not correspond to the predominant pattern seen in classical ALS/ MND, where skein inclusions predominate in ,90% of cases, so that the absence of classical skeins in this group is distinctive. Bunina bodies are found in up to 75-80% of ALS/MND cases, and their absence in these 4 patients is again suggestive of an atypical group. While these features are not sufficiently distinctive to allow a morphological prediction of CHMP2B-related MND, our results indicate that a predominance of compact inclusions in a PMA case may warrant examination of the CHMP2B gene.
Microarray analysis of the gene expression profile of motor neurons with CHMP2B mutations compared to neurologically normal control samples reveals some interesting changes to genes involved in key cellular processes, many of which are distinct to those shown by motor neurons isolated from SOD1-related ALS cases (Kirby et al, manuscript under review). CHMP2B cases show downregulation in multiple transcripts encoding proteins involved in the transport of cargoes along microtubules, suggesting impairments in axonal transport (KIF1A, KIF1C, KIF5C DYNLRB1, DYNLL2 and DYNC1H1), a phenomenon that has been well documented in multiple neurodegenerative diseases, including a SOD1-mediated model of ALS [55]. Interestingly, the dysregulation of microtubule proteins MAP1S, MAP4 [25] and stathmin, predicts loss of cell structure and transport network. It has been shown that stathmin overexpression in HeLa cells leads to Golgi fragmentation and microtubule disassembly, which are the same events observed in transgenic G93A SOD1 mice [56]. Stathmin overexpression and Golgi fragmentation seem to be early events in the neurodegenerative cascade characteristic of ALS and other neurological diseases [57], and has been confirmed by transcriptional analysis in pre-symptomatic G93A SOD1 mice [19]. ATG1 (3.48 fold) is strongly downregulated, and is part of a complex with ATG17 and ATG13. This complex is required for the formation of vesicles at the phagophore assembly site, which is a crucial step for  autophagy initiation [30]. A decrease in autophagy initiation may result in the accumulation of cytosolic aggregates, which could potentially contribute to motor neuron injury. It is of great interest that a recent study in a fly model of Huntington's disease combined the treatment of rapamycin and lithium [58]. Inhibition of mTOR by rapamycin and the induction of mTOR-independent autophagy by lithium resulted in an increase in autophagy in both an mTOR-dependent and mTOR-independent manner, enhanced the clearance of protein aggregates and provided greater protection than single treatment with either of these molecules [58]. In addition, our results predict a global repression of protein translation within the cell due to the downregulation of multiple translation initiation factors. Translation initiation is a critical checkpoint and step of the translation process, and is the stage at which control of mRNA translation is most strongly exerted [59]. This may be an effect of end-stage disease, where the cell is attempting to survive at the expense of general protein synthesis, or it may be in response to a compromised ER trying to prevent accumulation of misfolded proteins. The classical MAPK cascade and the p38 signalling pathway both show downregulation of multiple components, alongside upregulation of inhibitors of signal amplification. This would predict that the cell is unable to respond normally to stimulation from growth factors, cytokines or environmental stresses. Downregulation of HSP27, which plays a role in neuronal survival, in conjunction with the altered transcriptional levels of MKNK2 and AKT suggests that the motor neurons are in a state that predisposes to apoptosis. Interestingly, pharmacological abrogation of p38 by SB203580, resulted in the accumulation of large autolysosomes [60]. Downregulation of p38 is therefore potentially very important as CHMP2B functions in degradation and recycling pathways, and mutations result in autophagosome accumulation [12]. The downregulation of Ga q and Ga 11 suggests a reduction in IP 3 levels as a result of decreased PLCb activity, and this in conjunction with the downregulation of IP 3 R, predicts a reduction in calcium release from the ER. In addition, the upregulation of SERCA and the mitochondrial calcium transporters NCX1 and VDAC3, may result in excess calcium being transported out of the mitochondria and into the ER lumen, having the net effect of an increased [Ca 2+ ] in the ER lumen and reduced [Ca 2+ ] in the mitochondria. Tight control of ER Ca 2+ stores is crucial for cell survival, and it has been previously demonstrated that increased ER calcium levels trigger ceramide-induced apoptosis [61].
We have demonstrated in a HEK-293 cellular model that transfection of constructs carrying p.I29V, p.T104N or p.Q206H CHMP2B mutations results in the formation of large cytoplasmic vacuoles (Figure 6 B-D). Induction of autophagy resulted in an increase in the number of vacuoles, providing further evidence that functional CHMP2B is required for the degradation of cytoplasmic components. These mutations caused aberrant CD63 accumulation within the cytosol (p.T104N: Figure 6 I-J; p.I29V & p.Q206H: data not shown). CD63 is found abundantly in lysosomes, and its trafficking and cellular distribution is tightly regulated [62]. The small vacuoles found in cells expressing WT CHMP2B co-localised with large areas of CD63 stainingdisplaying the normal physiological fusion of late endosomes with lysosomes ( Figure 6 F-H). In contrast, cells expressing mutant CHMP2B often showed CD63 immunostaining on the membranes of large vacuoles ( Figure 6 I-K). These data indicate that the missense CHMP2B mutations identified in our patient cohort disrupt the fusion of the lysosome with the large cytoplasmic vacuoles, suggesting that mutant CHMP2B is altering the dynamics of this fundamental process. This hypothesis is further supported by the transfection of FTD-related CHMP2B truncating mutations in neuroblastoma cell lines, which led to the formation of large vesicles, with a morphology in keeping with aberrant endosomes [52].
The MVB pathway has an essential role in the delivery of ubiquitinated proteins to the lysosome for degradation, and is highly conserved from yeast to humans. It has been demonstrated that mice deficient in genes essential for autophagy are unable to effectively clear cellular debris, leading to the accumulation of polyubiquitinated proteins within inclusion bodies and subsequent neurodegeneration [63,64]. Additionally, there is strong evidence that CHMP2B mutations result in neuronal death; dendritic retraction and cell death seen in cortical neuron cultures transfected with CHMP2B Intron5 is a result of the failure of the aberrant protein to dissociate from the other subunits of the ESCRTIII prior to the next round of sorting, a step critical for the formation of MVBs [12]. Transfection of HeLa cells with CHMP2B Intron5 and CHMP2B D10 results in increased levels of ubiquitin and p62 and inhibition of protein degradation through the process of autophagy [13]. Inhibiting autophagy induction in cells transfected with mutant CHMP2B resulted in delayed neuronal loss, further implicating autophagy in the process of neurodegeneration [65,66]. In the present study, we have demonstrated that LC3-II levels are significantly increased in cells expressing mutant CHMP2B. Autophagosomes sequester the target organelle/protein and fuse with lysosomes, where the contents -and the LC3-II present on the luminal side of the autophagosome -are degraded [67]. The half-life of LC3-II is short due to the balance of autophagosome synthesis and degradation; thus the increased levels of LC3-II in mutantexpressing cells is indicative of an alteration in the degradation of cargo by the autophagic pathway. We therefore propose that the CHMP2B mutations we have identified may contribute to motor neuron injury through dysfunction of the autophagic clearance of cellular proteins, and the formation of ubiquitin positive inclusions with compact morphology (Figure 2). Further mutation screening now needs to take place in cohorts of patients, particularly those with a lower motor neuron disease phenotype, and from different geographical areas. Identification of CHMP2B mutations showing segregation with familial disease would establish conclusively the role of CHMP2B as a causative gene in ALS.

Supporting Information
Table S1 The 891 genes downregulated and 561 genes upregulated in motor neurons with CHMP2B mutations, grouped into biological processes.