Lipid alterations in human frontal cortex in ALS‐FTLD‐TDP43 proteinopathy spectrum are partly related to peroxisome impairment

Abstract Aim Peroxisomes play a key role in lipid metabolism, and peroxisome defects have been associated with neurodegenerative diseases such as X‐adrenoleukodystrophy and Alzheimer's disease. This study aims to elucidate the contribution of peroxisomes in lipid alterations of area 8 of the frontal cortex in the spectrum of TDP43‐proteinopathies. Cases of frontotemporal lobar degeneration‐TDP43 (FTLD‐TDP), manifested as sporadic (sFTLD‐TDP) or linked to mutations in various genes including expansions of the non‐coding region of C9ORF72 (c9FTLD), and of sporadic amyotrophic lateral sclerosis (sALS) as the most common TDP43 proteinopathies, were analysed. Methods We used transcriptomics and lipidomics methods to define the steady‐state levels of gene expression and lipid profiles. Results Our results show alterations in gene expression of some components of peroxisomes and related lipid pathways in frontal cortex area 8 in sALS, sFTLD‐TDP and c9FTLD. Additionally, we identify a lipidomic pattern associated with the ALS‐FTLD‐TDP43 proteinopathy spectrum, notably characterised by down‐regulation of ether lipids and acylcarnitine among other lipid species, as well as alterations in the lipidome of each phenotype of TDP43 proteinopathy, which reveals commonalities and disease‐dependent differences in lipid composition. Conclusion Globally, lipid alterations in the human frontal cortex of the ALS‐FTLD‐TDP43 proteinopathy spectrum, which involve cell membrane composition and signalling, vulnerability against cellular stress and possible glucose metabolism, are partly related to peroxisome impairment.

ageing and redox balance, under the control of peroxisome/mitochondria function, which are altered in age-related diseases such as diabetes, hypertension, cancer and neurodegenerative diseases [32][33][34][35]. Impaired peroxisomal function occurs in Alzheimer's disease (AD) and related transgenic mouse models [36][37][38]. More precisely, accumulation of C22:0 and very long-chain fatty acids, and decreased levels of plasmalogens, together with increased volume density and loss of peroxisomes in neurons with neurofibrillary tangles, are all observed with AD progression [39]. These alterations have prompted the study of several specific therapeutic tools directed to curbing altered peroxisomal function in AD [40][41][42][43][44][45].
Amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) are two fatal neurodegenerative disorders with considerable clinical, pathological and genetic overlap. ALS is a fatal neurodegenerative disorder characterised by the progressive degeneration of both upper and lower motor neurons, resulting in a multitude of motor symptoms, including muscle weakness, fasciculations, spasticity, dysphagia and, eventually, respiratory dysfunction [46]. FTLD is a pathological diagnosis that manifests clinically in the form of frontotemporal dementia (FTD), characterised by cognitive, behavioural and linguistic dysfunction. The link between these disorders is made clear by the fact that almost 50% of ALS patients show cognitive impairment of the type observed in FTD, and also that 15% of ALS cases meet the diagnostic criteria for FTD at the time ALS is diagnosed [47]. In addition, 15% of FTLD cases have clinically detectable motor symptoms [48]. Both disorders are charac-

Human cases
Post-mortem samples of fresh-frozen FC area 8 were obtained from the Institute of Neuropathology HUB-ICO-IDIBELL Biobank and the Hospital Clinic-IDIBAPS Biobank following the guidelines of Spanish legislation on this matter and approval of the local ethics committees and in accordance with criteria of sample quality This post-mortem delay does not compromise the quality of the sample [49][50][51]. One hemisphere was immediately cut in coronal sections, 1 cm thick, and selected areas of the encephalon were rapidly dissected, frozen on metal plates over dry ice, placed in individual air-tight plastic bags and stored at −80°C until the use for biochemical studies. The other hemisphere was fixed by immersion in 4% buffered formalin for 3 weeks for morphological studies.
The neuropathological study was carried out on 20 selected 4-μm-thick de-waxed paraffin sections of representative regions of the brain. Sections were stained with haematoxylin and eosin, Klüver-Barrera, or processed for immunohistochemistry with anti-β-amyloid, phospho-tau (clone AT8), α-synuclein, αB-crystallin, TDP-43, ubiquitin, p62, glial fibrillary acidic protein, CD68 and Iba1 antibodies [52]. Sporadic FTLD-TDP (sFTLD-TDP) cases were diagnosed following well-established criteria: frontotemporal atrophy, loss of neurons and variable spongiosis in the upper cortical layers, astrocytic gliosis and presence of TDP-43-immunoreactive inclusions in the cytoplasm or in the nucleus of neurons, and in dendrites (NCIs, NIIs and DNs respectively), and were then categorised as type A, B or C. [53,54] Cases with familial frontotemporal lobar degeneration linked to C9ORF72 expansion (henceforth referred to as c9FTLD for practical purposes), all of them carrying more than 30 intronic hexanucleotide repeats, were classified as type A or B.
All these cases showed a sequential pattern II or III [55]. The frontal cortex of sporadic ALS (sALS) cases showed variable alterations; TDP-43-immunoreactive small dystrophic neurites and/or TDP-43positive granules and/or small cytoplasmic globules in neurons were observed in 11 of 18 cases, but they were abundant in only three cases (cases 56, 57 and 58) ( Table 1); spongiosis in the upper cortical layers was found in only one case (case 55). The whole series included 16 sFTLD-TDP (71.6 ± 9.6 years; 11 men and 3 women), 19 c9FTLD (mean age 70 years; 10 men and 9 women), 15 sALS (mean age 54 years; 11 men and 4 women) and 17 control cases (64.7 ± 8.9 years; 11 men and 6 women), as summarised in Table 1.
Although there are significant differences in the average age of the groups analysed, the age range of the study subjects is between 60 and 70 years. Previous studies on brain (and particularly in frontal cortex) lipid composition showed that lipids remain stable in adulthood; minimal changes appear in older ages than those analysed here [56,57].
Patients with additional associated pathologies of the nervous system, excepting early stages of neurofibrillary tangle pathology in the entorhinal cortex and hippocampus, and those with the presence of mild small blood vessel disease, were excluded, as were those cases with infectious, inflammatory or autoimmune diseases.  Table   S1. The values for β-glucuronidase (GUS-β) were used as internal controls for normalisation purposes [58]. The parameters of the reactions were 50°C for 2 min, 95°C for 10 min and 40 cycles of 95°C for 15 s and 60°C for 1 min. Finally, Sequence Detection Software (SDS version 2.2.2, Applied Biosystems) was used to capture TaqMan PCR data. The double-delta cycle threshold (ΔΔCT) method was utilised to analyse the data. The statistical study was performed using the T-student test or ANOVA-one way when necessary. The significance level was set at *p < 0.05, **p < 0.01 and ***p < 0.001 versus control group; # p < 0.05, ## p < 0.01 and ### p < 0.001 versus sALS; and $ p < 0.05, $$ p < 0.01 and $$$ p < 0.001 versus sFTLD-TDP.

Fatty acid profiling
Briefly, samples were incubated for lipid extraction and FAs transesterification in 2 ml of 5% methanolic HCL at 75 °C for 90 min. FAs methyl esters were extracted by adding 2 ml of n-pentane and 1 ml of saturated NaCl solution. Samples were separated and evaporated under N2 gas n-pentane phase and finally dissolved in 80 µl of carbon disulphide. Gas chromatography (GC) analysis was then performed.
The GC method was used for separation with a DBWAX capillary column (30 m × 0.25 mm ×0.20 μm) in a GC System 7890 A with a Series Injector 7683B and an FID detector (Agilent Technologies, Barcelona, Spain). The temperature of the injector was 220 °C using the splitless mode. A constant rate (1.8 ml/min) of helium (99.99%) was maintained. The column temperature was held at 145°C for 5 min; subsequently, the column temperature was increased by 2°C/ min to 245°C for 50 min, and held at 245°C for 10 min, with a postrun of 250°C for 10 min as previously described [59][60][61]. Based on FA composition, different indexes were calculated, and elongase and desaturase activity was estimated from specific product/substrate ratios [61,62].

Non-targeted lipidomic analysis
A previously validated method was used for lipid extraction [63].
Briefly, 5 μl of miliQ water and 20 μl of methanol were added to 10 μl of homogenised tissue. Samples were then shaken vigorously for 2 min. Following this, methyl tert-butyl ether (MTBE) containing isotopically labelled lipid standards was added. Samples were then immersed in a water bath (ATU Ultrasonidos, Valencia, Spain) with an ultrasound frequency of 40 kHz and power of 100 W, at 10°C for 30 min. After this, 25 μl of miliQ water was added to the mixture, which was centrifuged at 300 rpm at 4°C for 10 min to separate the organic phase. Finally, the upper phase was collected and stored for mass-spectrometry analysis. A pool (20 µl of each sample) of all lipid extracts was prepared and used as quality control [64].
Lipid extracts were analysed by LC-MS according to the method described [59]. An Agilent UPLC 1290 system coupled to an ESI-Q-TOF MS/MS 6545 (Agilent Technologies, Barcelona, Spain) was used. Two runs were performed to collect positive and negative electrospray ionised lipid species. Data pre-processing was done as published before [65][66][67]. Finally, identities were confirmed by searching experimental MS/MS spectra against in silico libraries, using HMDB and LipidMatch, an R-based tool for lipid identification [68,69]. Multivariate and univariate statistics were calculated using Metaboanalyst [70].

Peroxisome biogenesis
When compared with controls, only PPARG expression was significantly increased in sALS (p < 0.01) and significantly decreased in c9FTLD cases (p = 0.018), whereas PPARGC1A was significantly increased in c9FTLD (p = 0.033). However, differences were also identified when comparing expression levels among the three disease groups. Thus, PEX14 was significantly increased in c9FTLD when compared with sALS (p < 0.001) and sFTLD-TDP (p < 0.001); PPARD was decreased in sFTLD-TDP when compared with sALS (p < 0.001); PPARG was decreased in sFTLD-TDP and c9FTLD when compared with sALS (p = 0.006 and p < 0.001 respectively); PPARGC1A was down-regulated in sFLTD-TDP when compared with sALS (p < 0.005) and up-regulated in c9FTLD when compared with sFTLD-TDP (p < 0.001); and finally, DNM1L mRNA expression was  were observed in sALS, sFTLD-TDP or c9FTLD when compared with controls. Nor were significant differences observed among the three disease groups ( Figure 1D).

β-oxidation components
ACAA1 gene expression was decreased in sALS and sFTLD-TDP when compared with controls (p < 0.001 and p < 0.001 respectively), but it was increased in c9FTLD, not only when compared with controls (p < 0.001), but also with respect to sALS (p < 0.001) and sFTLD-TDP cases (p < 0.001). Following a similar trend, levels of ACOX3 were increased in c9FTLD when compared with controls (p < 0.001), sALS (p < 0.001) and sFTLD-TDP (p < 0.001). EHHADH transcript levels were down-regulated in sFTLD-TDP when compared with controls and sALS (p = 0.004 and p = 0.049 respectively), and up-regulated in c9FTLD when compared with sFTLD-TDP (p < 0.001). No differences in the expression of ACOX1 and ACOX2 were found between controls and disease cases, nor among the three pathological groups ( Figure 1E).

Plasmalogen biosynthesis
Expression levels of genes coding for components of the plasmalogen biosynthesis pathway, AGPS, DHAP-AT and FAR1, were evaluated, but no differences in the expression of these genes were found between controls and disease cases, or among the three pathological groups ( Figure 1F).

Acylcarnitine biosynthesis
Acylcarnitine biosynthesis components revealed few differences in the expression of ACOT, CRAT and CROT transcripts. ACOT gene expression was increased in c9FTLD when compared with sALS (p = 0.003), whereas CRAT and CROT were significantly increased in c9FTLD when compared with sFTLD cases (p = 0.034 and p = 0.029 respectively). Regarding control cases, only CROT mRNA expression levels were significantly decreased in sFTLD cases when compared to controls (p = 0.021) ( Figure 1G).

Gene expression linked to fatty acid metabolism
The

Fatty acid profiling
Since the biosynthesis of highly unsaturated fatty acids is dependent on peroxisomal beta-oxidation activity, fatty acid composition of total lipids from frontal cortex area 8 was analysed (Table 2)

Lipidomic profiling
In order to gain an overview of whole lipidome, an untargeted lipidomic approach was applied. Baseline correction, peak picking and peak alignment were performed on acquired data, resulting in a total of 7951 molecules from both ionisation modes (negative and positive). After quality control assessment, filtering and signal correction, 1119 features remained, which were log-transformed and auto-scaled (mean centering/standard deviation) and used for multivariate and univariate statistical analysis.

A lipidomic pattern is associated with ALS-FTLD-TDP43 proteinopathy spectrum
To investigate whether there was a common pattern for TDP-43 proteinopathies, an untargeted lipidomic analysis was performed in tissue samples from frontal cortex region 8 with the three neurological disorders grouped together. Unsupervised methods such as Principal Component Analysis (PCA) and Hierarchical Clustering visualised as a heatmap were used to find patterns in the samples.
A PCA analysis was performed using the whole detected lipidome; no differences were found between the diseased (DIS) group and healthy controls (CTL) ( Figure 3A). However, when the 25 lipid species with the lowest p -values were represented using hierarchical clustering analyses, as shown in a heatmap ( Figure 3B), a clear separation between groups was observed revealing a specific shared trend in patients with neurodegenerative diseases within the ALS-FTLD-TDP43 proteinopathy spectrum. Finally, the Wilcoxon test on all acquired data was performed to determine whether there were any significant lipid feature differences between healthy and diseased groups (p < 0.05). Dunn's test was used to correct for multiple comparisons. The statistical test resulted in 63 differential molecules with p < 0.05 (Table 3)

Differences between neurological diseases within the ALS-FTLD-TDP43 proteinopathy spectrum
To determine whether there was a real difference between the distinct TDP43 proteinopathy phenotypes characterised as sALS, sFTLD-TDP and c9FTLD, the three diseases were compared with each other. Frontal cortex region 8 samples were used to uncover characteristic lipidomic trends and features for each disorder. The heatmap representing the hierarchical clustering of the individual samples ( Figure 3D) showed perfectly arranged samples in disease groups when the top differential metabolites obtained with the Kruskal-Wallis test were used. Interestingly, FTLD-TDP patients (both sporadic and c9) clustered together, indicating that these groups are more similar to each other than they are to the ALS group.
The Kruskal-Wallis test revealed changes in glycerolipids, glycerophospholipids, sphingolipids and sterol lipids listed in Table 4.
Most of the compounds identified in this analysis ( Figure 3E) of frontal cortex region 8 were TGs and most contained 18:1 oleic acid; these TGs were higher in the ALS group and lower in the sFTLD-TDP group. Interestingly, CE (20:1) was increased in c9FTLD compared with the other groups.
TA B L E 2 Fatty acid compositional profiles of total lipids from frontal cortex area 8 in controls, sALS, sFTLD-TDP and c9FTLD cases assessed with gas chromatography

Important features associated with different phenotypes
To identify the differential lipid molecules in each disorder, we performed a Kruskal-Wallis test with a post hoc comparison using Dunn's test on frontal cortex region 8 samples from sALS, sFTLD-TDP, c9FTLD and controls. The differential molecules are listed in Table 5. Globally, phospholipid species (mainly PC) presented the greatest differences between ALS (lower) and CTL (higher) groups.
A sustained fatty acid profile does not exclude, however, potential changes in the content of lipid species that can be uncovered with a lipidomics approach. An adult human brain contains the largest amount and diversity of lipids (in terms of classes and molecular species) including glycerolipids, glycerophospholipids, sphingolipids and cholesterol. Glycerophospholipids are the major phospholipid components ubiquitously found in neural cell membranes [59,79,80].
In the human brain, phospholipids constitute 4.2% of the wet weight of the grey matter [79,81,82]. Phosphatidic acid occurs in low concentrations in brain (about 2% of total phospholipids). The and inositol phosphoglycerides account for about 2.6% of the total phospholipids in the human brain [84,85]. The brain contains the highest concentrations of phosphoinositides among animal tissues. Abbreviations: AcCar, acylcarnitine; DG, diacylglycerol; FAHFA, fatty acid ester of hydroxyl fatty acid; FDR, false discovery rate; mz value, mass-tocharge ratio; PC, phosphocholine; PE, phophoethanolamine; SM, sphingomyelin. a Represents confirmation by data-dependent MS2.
b Represents confirmation by data-dependent MS2 by class. Glycerophospholipids are important building blocks of cell membranes that provide an optimal environment for protein interactions, trafficking and function. In the ageing process and in the pathological context of neurodegeneration, decreased brain phospholipid levels and alterations in brain phospholipid metabolism appear, as observed in brain post-mortem tissue, CSF and blood [86]. This study demonstrates a down-regulation in PC and PE levels in the ALS-FTLD spectrum, suggesting alterations in the architecture of the neural cells [87,88]. In addition, PC is an important source for the formation of second messengers and lipid mediators [89,90].
Disturbance of its production interferes with cell proliferation and differentiation, and membrane movement throughout the cell [91].
Furthermore, PE plays essential roles in autophagy, cell division and protein folding, representing a precursor for the synthesis of several protein modifications [88]. In addition, both PC and PE are intermediates in the synthesis of other glycerophospholipid classes [92,93]. In line with our results, a recent study in cells [94] revealed that TDP-43-mediated toxicity induced lower levels of glycerophospholipids (especially glycerophosphocholines) and sphingolipids.
Decrease levels of glycerophospholipids were also described in ALS animal models [95]. Globally, our results suggest a minor but cru- Ether lipids are a subclass of glycerophospholipids that have two chemical forms: as 'plasmanyl' (also termed alkyl ethers and represented by the 'O-' prefix), and as 'plasmenyl' (also termed alkenyl ethers or plasmalogens, and represented by the 'P-' prefix) [96,97]. Ether lipids are mostly present as PC and PE species [96]. At the cellular level, ether lipid biosynthesis begins in the peroxisome and is completed in the endoplasmic reticulum [96][97][98]. The physiological role of ether lipids is linked to their function as membrane components (fluidity, formation of lipid rafts and a source of second messengers). Other functions in which ether lipids are involved are cholesterol transport, G-protein-mediated signal transduction, membrane fusion events, transmembrane protein function and vesicular function [96][97][98]. Interestingly, an antioxidant effect has also been assigned to plasmalogens [99]. Available evidence reveals that ether lipids are inversely associated with genetic peroxisomal disorders, and also suggests that they are negatively associated with prevalent disease states such as cancer, cardiovascular diseases and Alzheimer disease, among others [100]. Notably, these pathological conditions share as a common trait, increased oxidative stress, and a potential mechanistic role for plasmalogens. Our study clearly demonstrates down-regulation of the ether lipid content in frontal cortex area 8 of TDP-43 proteinopathies. Thus, a reduction in its levels may confer vulnerability against oxidative stress insults potentially contributing to neurodegeneration in these disorders, in addition to acting as a marker of impaired peroxisomal function. Reinforcing this idea, the lack of support at the transcriptional level (no changes were observed in the present work) suggests the existence of alterations at the translational level or, more probably, functional defects mediated by potential oxidative stress conditions. Further studies are needed to obtain a more detailed mechanistic view.
A major category of lipids is the sphingolipids [101] that play a key role in the formation of lipid rafts in cell membranes [102].
The metabolism of sphingolipids is a complex network with ceramide at the core [103]. The result is a wide diversity of lipid species with structural (e.g. sphingomyelins) and bioactive/messenger (e.g. sphingosines, dihydroceramides and ceramides) functions [101]. down-regulation of acylcarnitines does not seem to be mediated by defects at the transcriptional level. In fact, the contrary is observed, with an increase in mRNA content of the main components of the biosynthesis pathway. Therefore, it is plausible to hypothesise that this apparent contradiction between phenotype and genotype is caused by translational alterations or, analogously to the postulated for plasmalogens biosynthesis, by functional defects at the protein level leading to a reduced peroxisomal biosynthestic capacity. Further studies are, however, also needed to consolidate these new ideas.
Finally, a special mention should be made of the detection of two

CO N FLI C T O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/nan.12681.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon request. Represents confirmation by data-dependent MS2 by class.
c Represents confirmation by MS1 exact mass and retention time.