and KLC1 expression in Alzheimer’s disease: relationship and genetic influences.

Early disturbances in axonal transport, before the onset of gross Background: neuropathology, occur in a spectrum of neurodegenerative diseases including Alzheimer’s disease. Kinesin superfamily motor proteins (KIFs) are responsible for anterograde protein transport within the axon of various cellular cargoes, including synaptic and structural proteins. Dysregulated KIF expression has been associated with AD pathology and genetic polymorphisms within kinesin-light chain-1 (KLC1) have been linked to AD susceptibility. We examined the expression of KLC1 in AD, in relation to that of the KLC1 motor complex (KIF5A) and to susceptibility genotypes. We analysed KLC1 and KIF5A gene and protein expression in Methods: midfrontal cortex from 47 AD and 39 control brains. We found that gene expression of both and increased Results: KIF5A KLC1 with Braak tangle stage (0-II vs III-IV and V-VI) but was not associated with significant change at the protein level. We found no effect of KLC1 SNPs on KIF5A or KLC1 expression but KIF5A SNPs that had previously been linked to susceptibility in multiple sclerosis were associated with reduced mRNA KIF5A expression in AD cortex. The findings raise the possibility that genetic polymorphisms Conclusions: within the gene locus could contribute to disturbances of axonal KIF5A transport, neuronal connectivity and function across a spectrum of neurological conditions, Hares et al. describe the expression of KLC1 and KIF5A gene and protein expression in the mid frontal cortex of Alzheimers Disease compared to healthy control patients. This study builds upon their body of work and observations of the kinesin expression in Multiple Sclerosis. The methodological approaches used are appropriate to address the questions being asked in the study and I would recommend this manuscript for indexing with the following addition to the discussion of the manuscript. Could the authors comment and perhaps discuss the relevance and potential merits of the use of live imaging of axonal transport in animal models of AD and how they may attempt to relate this to KIF5A and KLC protein expression in these models. Please see relevant references below. A substantial literature has implicated axonal transport deficits as a component of pathogenesis in AD and other adult-onset neurodegenerative diseases. Some evidence has been published from GWAS studies suggesting that altered expression of molecular motor proteins or genetic polymorphisms in KLC1 may be risk factors for AD, but relatively little has been done to document specific changes in patient materials. In this study, the investigators have evaluated expression of two genes that are candidates for a role in AD, taking advantage of access to midfrontal cortex tissue from 47 AD and 39 control human brains. The set of brains are relatively well matched in terms of mean age and postmortem delay for control and AD. Based on age at the time of death, the AD patients are likely to be mostly sporadic cases with 5-10% familial cases, which is consistent with the population levels. Both males and females are considered together. The authors do not stratify by familial vs sporadic or male vs female. The focus here is on expression of KIF5A and KLC1, both of which have been reported to have polymorphisms associated with neurological diseases. In addition, the authors cite literature claiming that KIF5A complexed to KLC1 is responsible for transport of vesicles containing APP as well as band g-secretases. Unfortunately, those studies have never been replicated despite concerted effort by multiple groups (see for example O. Lazarov, et al. (2007). Impairments in fast axonal transport and motor ). A neuron deficits in transgenic mice expressing familial Alzheimer's disease-linked mutant presenilin 1 . In the present manuscript Kelly Hares and co-workers aimed (as stated by the authors) “to determine whether KLC1 levels are altered in AD, the relationship between the levels of KLC1 and KIF5A, and whether expression of KLC1 in the cerebral cortex is influenced by KLC1 polymorphisms, which have been linked to AD susceptibility.”


Introduction
Over 90% of Alzheimer's disease (AD) cases are sporadic rather than attributable to a single gene mutation. In these cases, as for many other sporadic adult-onset neurodegenerative diseases, the pathological changes are likely to be caused by a combination of environmental and genetic factors that activate variably independent pathogenic pathways that increase neuronal vulnerability to damage and death 1 . One major genetic risk factor for sporadic disease is the ε4 allele of APOE 2 . Aside from APOE, genome-wide association studies (GWAS), have linked single nucleotide polymorphisms (SNPs) in over 20 genetic loci to AD susceptibility 3 .
A common theme across neurodegenerative diseases is alterations in neuronal connectivity that can give rise to clinical symptoms before neuronal loss 4,5 . Targeting the processes responsible for the changes in neuronal connectivity is a key therapeutic aim, before consequential axonal loss and neuronal death cause irreversible disability in patients with neurodegenerative disease 6,7 . A crucial step in this transition from potentially reversible to irreversible disease is the disruption in axonal delivery of membrane-bound organelles from the neuronal soma to the synapse. The development of such disruption has been recognised pathologically in brain tissue, including in AD 8 , by the formation of axonal swellings at an early stage in the disease process.
Kinesin superfamily motor proteins (KIFs) are responsible for most anterograde protein transport within the axon of various cellular cargoes, including synaptic and structural proteins 9 . There are at least 38 different neuronally-enriched KIFs which contain a conserved microtubule-binding domain and a motor domain at their amino (NH 2 -) terminus, which hydrolyses ATP in order to generate motile forces to shift associated cargoes along the axon, via microtubule tracks 10 . The heterogeneous tail (COOH-terminal) domain of KIFs determines their cargobinding specificity 11 .
Early disturbances in axonal transport are a feature of several neurodegenerative diseases [12][13][14] , and studies have linked KIF dysregulation to AD 15,16 . Conventional kinesin member KIF5A transports several cargoes, including APP and βand γsecretases, by complexing with kinesin light chain-1 (KLC1) 17,18 . Deletion of the KLC1 subunit in mice leads to early selective axonal transport defects, resulting in a disorganised neuronal cytoskeleton with APP, neurofilament (NF) and hyperphosphorylated tau accumulation 19 . Mice transgenic for human APP with mutations that cause familial AD develop abnormal axonal swellings positive for KLC1 and structural axonal component phosphorylated NF-H, long before AD pathology can be detected 8,20 .
Single nucleotide polymorphisms (SNPs) in the region of the KLC1 gene have been linked to AD susceptibility; in particular, rs8702 has been shown to predict conversion of mild cognitive impairment (MCI) to AD 16,21,22 . Our previous studies have shown that SNPs within the KIF5A gene locus, linked to multiple sclerosis (MS) susceptibility, influence KIF5A expression in post-mortem brain 23,24 .
We and others have found KIF5A expression to be increased in post-mortem AD and PD tissue 25,26 . KIF members show a degree of functional redundancy, such that more than one KIF may transport a specific cargo 27 . We therefore suggested that KIF5A might be upregulated in AD as compensatory protective response to defects in other, interoperable, axonal transport proteins, facilitating the transport and clearance of aggregating cargoes in the cell body and axon. In support of this hypothesis, we showed that KIF5A levels in both MS and AD correlated inversely with the amount of cargo APP 24,25 . KIF5A shares many cargoes with anterograde motor protein KIF1B, such as mitochondria and synaptic vesicle proteins, and it can transport cargoes without complexing to KLC1 9,28 .
In this study we aimed to determine whether KLC1 levels are altered in AD, the relationship between the levels of KLC1 and KIF5A, and whether expression of KLC1 in the cerebral cortex is influenced by KLC1 polymorphisms, which have been linked to AD susceptibility.

Study cohort
Samples of midfrontal cortex (Brodmann area 8; 500mg) were obtained from the South West Dementia Brain Bank (Bristol, UK), under the terms of South West -Central Bristol Research Ethics Committee approval no 08/H0106/28+5. Our previous studies indicated an influence of post-mortem delay (PMD) on mRNA and protein expression in brain tissue 25 . Therefore, cases with a PMD >72 h were excluded from analysis. We studied 47 cases of AD (in which, AD neuropathological change was an adequate explanation of dementia, according to National Institute on Aging-Alzheimer's Association guidelines 29 ), and 39 age-matched controls with no history of cognitive impairment (Table 1).

RNA and protein extraction
Both RNA and protein were extracted from each sample by use of the Paris™ Kit (ThermoFisher Scientific; AM1921), according to the manufacturer's instructions. Samples were weighed, divided into 2x 50 mg portions and placed in 2 ml homogenisation tubes, pre-filled with 5x 2.3 mm silica beads (Stratech Scientific; 11079125z-BSP) and 300 µL of Paris™ kit lysis buffer, with the addition of Halt™ Protease and phosphatase inhibitor cocktail and 0.5M EDTA (1:100) (ThermoFisher Scientific; 78440). Samples were homogenised using a Precellys®24 automated homogeniser (Stretton Scientific). Samples were subsequently centrifuged (Sorvall ST 16R) at 10,000 xg for 2 min at 4°C and the soluble protein supernatant removed and stored at -80°C until required. The remaining homogenate was taken through the additional manufacturer's steps for extraction and elution of RNA, before storage at -80°C.

gDNA extraction
An approximately, 2 mm slice was taken from each brain sample and gDNA extracted using the TaqMan® Sample-to-SNP™ Kit (ThermoFisher Scientific; 4403313). Cortex was placed in lysis buffer, briefly centrifuged and incubated for 3 min at 95°C, before addition of DNA stabilising solution. Samples were then stored at -20°C until use. Membranes were blocked in 5% BSA/Tris-buffered saline-Tween 20 (TBS-T) or 5% milk/TBS-T, for 1 h at room temperature. Primary antibodies were reconstituted in membrane blocking solution (detailed in Table 2) and added overnight at 4°C. The primary antibodies used were: mouse anti-AβPP (ThermoFisher Scientific; 13-0200), mouse anti-GAPDH (Abcam; Ab9484); mouse anti-PHF-TAU (ThermoFisher Scientifc UK; MN1020), rabbit anti-KIF5A (Sigma-Aldrich Ltd; HPA004469), rabbit anti-KLC1 (Abcam; Ab174273) and rabbit anti-NeuN (Abcam; Ab177487). Bound primary antibody was detected by incubation with HRP-conjugated secondary antibodies for 1 h at room temperature (as detailed in Table 2). Protein expression was visualised using a chemiluminescence Clarity™ Western ECL Substrate (Biorad; 1705060). Antibodies displayed bands relative to their reported molecular weight, as described on manufacturer data sheets ( Figure 1).

Dot blot
Dot blot was performed using a Bio-Dot Microfiltration manifold (Biorad), according to the manufacturer's instructions. Nitrocellulose membrane was pre-soaked in 1x TBS and placed in the manifold, before addition of 100 µL of protein homogenate/well (diluted 1:200 with 1x TBS). After microfiltration for 90 min, the membrane was removed and placed in blocking antibody, before antibody incubation (Table 2), as per the western blot protocol. Protein expression was visualised on a Biorad Universal III Bioplex imager, using a chemiluminescence Clarity™ Western ECL Substrate (Biorad). Image Lab™ 5.0 software (Biorad) was used for densitometric protein analysis.
Integrated density values were expressed relative to the neuronal nuclear protein NeuN. In addition, NeuN protein levels were normalised to ubiquitous protein GAPDH to study any effect of PMD on protein expression.
Enzyme-linked immunosorbent assay (ELISA) AD frontal tissue (200 mg) was homogenised in TBS extraction buffer to obtain the soluble protein fraction and the protein pellet was homogenised in guanidine to obtain the insoluble fraction, as previously described 25,30 . Sandwich ELISA was used to measure total Aβ in the fractions, with mouse monoclonal anti-Aβ (4G8 clone) (Biolegend; 800712; RRID: AB_2734548), for the capture step and biotinylated anti-human Aβ monoclonal antibody (10H3 clone) (ThermoFisher Scientific; MN1150B; RRID: AB_223641), for the detection step, as previously described 31 .
Immunoperoxidase staining of paraffin sections 7 µm sections from frontal cortex were dewaxed, hydrated, and immersed in 3% hydrogen peroxide in methanol for 30 min to block endogenous peroxidase activity. Sections were subsequently rinsed in running tap water and pre-treated with sodium citrate buffer in the microwave (0.01 M, pH 6.0, 5 min) to unmask antigenic sites, then rinsed in PBS. Immunostaining was performed using VECTASTAIN® Universal Elite® ABC Kit (Vector Laboratories; PK-6200). Non-specific binding was blocked using horse serum for 30 min at room temperature, before incubation overnight at 4°C with anti-mouse monoclonal PHF-tau (1:2000) (ThermoFisher Scientific; MN1020; RRID: AB_223647), diluted in PBS. Sections were rinsed in PBS and incubated for 20 min with VECTASTAIN® biotinylated universal secondary antibody, followed by 20 min with VECTASTAIN® Elite ® ABC complex (both Vector Laboratories; PK-6200), before a final 7 min with 3,3'-diaminobenzidine (DAB) (Vector Laboratories; SK4100). Sections were washed in water, immersed in copper sulphate DAB enhancer for 4 min (0.16 M Copper (II) sulphate 5-hydrate/0.12 M sodium chloride) and counterstained with Gills haematoxylin II (Sigma-Aldrich Ltd; GHS216) for 15 s. Sections were subsequently dehydrated, cleared and mounted in Clearium® (Leica Biosystems; 3801100). Controls in each run included sections incubated overnight in PBS instead of the primary antibody.

Immunohistochemical analysis
The percentage field fraction immunopositive for PHF-tau (insoluble tau load) was assessed in the cortex of frontal sections from AD and controls by computer-assisted image analysis using Histometrix software (Kinetic Imaging, Nottingham, UK) driving a Leica DM microscope with a motorised stage, as previously described 32,33 . The area to be analysed was selected at low-magnification (x2.5) and the threshold for PHF-tau labelling density calibrated at high magnification (x20). The software was programmed to measure tau load in 30 random regions (x20 objective fields) within the selected area and determine the cumulative area fraction with a density exceeding the threshold value.

Statistical analysis
Univariate mRNA and protein analysis were performed using GraphPad Prism 5™ (GraphPad Software Inc.; San Diego, USA). Data normality was tested using the Shapiro-Wilk test.
One-way ANOVA with post-hoc Bonferroni, or Kruskal-Wallis with post-hoc Dunns test, as appropriate, was used to analyse differences in mRNA and protein expression according to Braak stages and SNP genotype. Parametric Pearson's or non-parametric Spearman's correlation was used to assess any relationship between proteins. A multiple regression model (STATA v12; StataCorp LLC; Texas, USA) was used to analyse mRNA expression in relation to Braak stage pathology, patient age of death and tissue post-mortem delay. Where necessary, data were transformed to normality before regression analysis. For all tests, values of p < 0.05 were considered statistically significant.

Cohort variables
With exclusion criteria, the study cohort comprised 47 AD cases and 39 controls, as detailed in Table 1. The age of AD cases ranged from 54-98 y (mean 81, SD 9). Control cases ranged from 43-95 y (mean 80, SD 11). There were more females (n = 29) than males (n = 18) in the AD group, and fewer females (n = 15) than males (n = 24) in the controls. The PMD did not differ significantly between AD (37 h, SD 19) and controls (34 y, SD 20; two-tailed Mann-Whitney p = 0.34).
Significant increase in kinesin gene expression linked to AD pathology Expression of kinesin relative to RBFOX3 mRNA was measured by qPCR and analysed in relation to Braak tangle stage. KIF5A mRNA was significantly elevated at Braak stages III-IV (p < 0.01) and V-VI (p < 0.001), compared to stages I-II, (Figure 2A). KLC1 mRNA was significantly elevated in late Braak stage disease (V-VI, p < 0.001) compared to stages I-II ( Figure 2B). In a previous study we found an association between PMD and the level of KIF5A relative to RBFOX3 (NeuN) mRNA 25 but in the present cohort we found no effect of either patient age at death or PMD on expression of KIF5A or KLC1 relative to RBFOX3 (Table 3).
Single nucleotide polymorphisms within the KIF5A gene locus influence KIF5A mRNA expression We previously showed that KIF5A protein level was reduced in post-mortem brain tissue from MS patients homozygous for SNPs within the KIF5A gene locus which have been linked to MS susceptibility 24 . We explored this in AD brain tissue using pre-designed TaqMan® SNP assays for rs12368653 and rs4646536, the latter in linkage disequilibrium (r 2 = 1) with rs703842, making it suitable as a proxy 34,35 . The accuracy of the SNP was verified using 18 MS positive control genotypes from our previous studies 24 .
KIF5A mRNA expression was significantly lower in AD patients homozygous for the rs12368653 SNP (AA) than in AD patients with no copies (GG) (p<0.05; Figure 3A). Similarly, AD patients homozygous for the rs4646536 SNP (AA) (proxy for rs703842), had significantly lower KIF5A mRNA than did heterozygous patients (AG) (p<0.05; Figure 3B). There was no significant difference in KLC1 mRNA expression in AD patients stratified according to KIF5A SNPs (rs12368653 and rs4646536; Figure 3C and 3D). We subsequently investigated whether SNPs in the KLC1 gene that were previously linked to AD   susceptibility 21,22 influence kinesin mRNA expression. There was no significant difference in KLC1 mRNA between AD patients heterozygous for the rs8702 SNP (GC) and those with no copies (GG; Figure 4A). Similarly, there was no significant difference in KLC1 mRNA between AD patients heterozygous for the rs8007903 SNP (AG) and those with no copies (AA; Figure 4B). Due to low cases numbers (n=2), homozygous expression could not be analysed. There was no significant effect of KLC1 SNPs on KIF5A mRNA expression ( Figure 4C and 4D).
Kinesin protein level does not vary significantly with AD pathology Kinesin protein expression was measured by dot blot. In keeping with previous studies 25 , KIF5A expression did not differ significantly between Braak stages 0-II, III-IV and V-VI ( Figure 5A). In addition, there was no significant variation in KLC1 protein level with Braak stage ( Figure 5B). As expected, the levels of KIF5A and KLC1 protein were positively correlated (p < 0.001, Figure 5C). There was no influence of SNPs within the KIF5A gene locus (rs12368653 and rs4646536) on KIF5A ( Figure 6A and 6B) or KLC1 protein level ( Figure 6C and 6D).
Kinesin levels inversely correlate with amyloid precursor protein in AD As in previous studies 25,36 , APP level did not differ significantly between AD and control cases ( Figure 7A). APP levels correlated inversely with both KIF5A and KLC1 protein in AD (p < 0.05; Figure 7B and 7C). Neither KIF5A or KLC1 correlated with the level of soluble or insoluble Aβ or with the ratio of soluble: insoluble Aβ (Table 3), although a trend was seen for lower levels of soluble Aβ to be associated with higher levels of KIF5A (p=0.10; Table 3), as in our previous study 25 . As expected, soluble and insoluble Aβ levels were significantly higher in AD than controls ( Figure 8A and 8B).

KIF5A protein level correlates inversely with soluble paired helical filament tau
The level of hyperphosphorylated tau in the soluble protein fraction was significantly increased in AD compared with control cases ( Figure 9A). The level correlated inversely with KIF5A protein level (p<0.05) but not with KLC1 protein in AD ( Figure 9B and 9C). Western blot analysis showed additional bands above the predicted molecular weight of tau (45-65kDa),  most likely attributed to phosphorylated epitopes 37 . In addition to immunoblotting, hyperphosphorylated tau levels were measured by quantitative immunohistochemistry and showed a significant positive correlation with hyperphosphorylated tau protein levels measured by dot blot (p<0.001; Figure 9D). There was no correlation between KIF5A or KLC1 protein level and insoluble tau load (Table 4).

Discussion
Previous studies in mouse models implicated defective functioning of KLC1 in disruption of axonal transport in AD 19,38 .
In the current study of human tissue from AD and control cases, we have found KLC1 and KIF5A gene expression to be elevated in AD and associated with Braak tangle stage. We have also demonstrated significant inverse correlations between kinesin levels and AD-associated proteins, emphasising the importance of kinesins in AD pathology. We did not find any association between KLC1 gene expression and KLC1 SNPs reportedly linked to AD.
Four KLC genes are expressed in mammals but those involved in axonal transport are KLC1 and KLC2. KLC1 is neuronally enriched whereas KLC2 is ubiquitously expressed 39 . A coiledcoil region at the amino terminus of the KLCs binds to the stalk domain (close to the carboxyl terminus) of conventional kinesins (KIF5A, KIF5B and KIF5C) and acts as an adaptor complex for indirect binding and transport of cellular cargoes, such as APP 27 . Expression of both KLC1 and KIF5A mRNA was positively associated with Braak tangle stage, a widely used marker of AD progression, defined by the distribution of hyperphosphorylated tau pathology distribution within the brain 40 . KIF5A mRNA was upregulated at relatively early stages of disease (Braak stages III-IV), in keeping with previously reported evidence of early-stage axonal abnormalities in AD and other neurodegenerative diseases 13,19,38 . Upregulation of both KIF5A and KLC1 mRNA may be a compensatory response to other pathological processes that interfere with axonal transport of protein cargoes to the synapse. Elevated kinesin gene expression in human tissue has been demonstrated in several neurodegenerative diseases 15,26,41 .
Genome-wide association studies (GWAS) have shown linkage of multiple genes to AD susceptibility, including some genes encoding proteins with roles in axonal transport and cytoskeletal function (NME8, CELF1, and CASS4) 3 . However, GWAS are unable to detect rare variants linked to susceptibility. Such variants include SNPs within the KLC1 gene, which have been linked to early AD pathogenesis. In particular, rs8702 was reported to predict conversion from mild cognitive impairment to AD and was associated with levels of hyperphosphorylated tau in patient cerebrospinal fluid 16 . It is hypothesised that this SNP could influence the splicing of c-terminal epitopes in KLC1 gene transcripts 16 , which could affect cargo binding affinity, axonal transport and synaptic function, causing structural instability and eventual neuronal degeneration. The effective 'strain' on the neuron caused by transport defects is likely to be exacerbated by a reduction in mitochondrial ATP supply associated with aging 42 . We found no effect on KLC1 expression of heterozygosity of the rs8702 'C' risk allele or rs8007903 'G' allele but did not have sufficient cases to analyse the influence of homozygosity of the SNPs. Larger genetic studies found no effect of rs8702 on AD susceptibility but a significant interaction with APOE ε4 carrier status in patients 16 . The continuing uncertainty as to the influence of KLC1 polymorphisms on the risk of AD highlights the difficulties of linkage studies for rare variants, which are often hampered by small sample size, locus heterogeneity and false-positive results 3 .
We previously found a significant reduction in KIF5A expression in MS tissue in homozygous carriers of the rs703842 'A' allele and rs12368653 'A' allele 23 Irrespective of overall kinesin protein levels, evidence points to disruption of convention kinesin transport in AD, through   As kinesin phosphorylation is a potent modulator of axonal transport in neurodegenerative diseases, this pathway presents a potential modifiable target for treatment 1,55 . However, evidence suggests cargo transport is bi-directional in axons with both anterograde motor kinesin and retrograde motor dynein bound simultaneously in a 'tug-of-war' model, such that the loss or post-translation modification of one motor protein will affect motility in both directions 56,57 . Indeed, studies have reported reduced dynein intermediate chain in AD cortex 48 . In addition, protein kinases regulate a wide range of cellular activities, and for any attempted therapeutic interventions it will be important to establish the specificity of kinase inhibitors and phosphatases on kinesin expression.
Overall, the findings from this study do not indicate that upregulation of KIF5A is a result of functional redundancy between KIF5A and KLC1 in protein transport, or that the expression of KLC1 in cerebral cortex is influenced by KLC1 polymorphisms, which have been linked to AD susceptibility. However, the findings highlight the importance of KIF5A in maintaining axonal transport and raise the possibility that patients with higher 'reserve' levels of KIF5A are less susceptible to axonal transport disruption and pathology. The findings also suggest genetic polymorphisms within the KIF5A gene locus could represent a common neurodegenerative pathway across a spectrum of neurological conditions.

Data availability
The raw data files used for figure generation are available at the University of Bristol data repository, data.bris In the present manuscript Kelly Hares and co-workers aimed (as stated by the authors) "to determine whether KLC1 levels are altered in AD, the relationship between the levels of KLC1 and KIF5A, and whether expression of KLC1 in the cerebral cortex is influenced by KLC1 polymorphisms, which have been linked to AD susceptibility."