Introduction
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease of the corticomotoneuronal system associated with progressive loss of motor neurons, secondary muscle weakness, and death, typically from neuromuscular respiratory failure, within three years of first symptom onset [
1]. Frontotemporal dementia (FTD) is characterised by behavioural change or language problems, with a longer disease course when it occurs in isolation [
2]. Neither ALS nor FTD has highly effective disease-modifying therapy. ALS and FTD are related in clinical, histopathological and genetic domains. Up to 15% of people with ALS will fulfil the criteria for FTD, but a larger proportion (up to 50%) will have detectable cognitive or behavioural dysfunction on testing [
3]. Insoluble neuronal and glial cytoplasmic inclusions of the ubiquitinated protein TDP-43 are the pathological hallmarks of 97% of ALS and 50% of FTD cases [
4]. Variants in several genes have been implicated in both ALS and FTD. Of these, an intronic hexanucleotide repeat expansion (HRE) in
C9orf72 is the commonest cause of ALS and FTD, inherited in an autosomal dominant pattern [
5].
C9orf72 HRE can manifest as ALS, FTD or both in the same family [
6]. Rare variants in several other genes have been implicated in causing both ALS and FTD [
7]. An increasing number of relatives of people affected by monogenetic forms of ALS and FTD are aware of their potentially higher risk of developing ALS or FTD through asymptomatic gene variant carrier status although penetrance is variable and poorly understood. Therefore, the search for modifiable factors influencing the risk of ALS and FTD has become more relevant, with an urgent imperative to provide evidence-based guidance.
Several lifestyle and metabolic factors have been implicated in influencing ALS risk, including body mass index, strenuous exercise and smoking [
8,
9]. Large population-based cohort studies indicate that higher levels of low-density lipoprotein cholesterol (LDL-c) and its primary apolipoprotein, apolipoprotein B (ApoB) [
10] and lower levels of high-density lipoprotein cholesterol (HDL-c) and its primary apolipoprotein, apolipoprotein A1 (ApoA1) [
11] are associated with ALS. Genetic epidemiological techniques which can circumvent the confounding that limits the causal interpretation of observational studies [
12], specifically Mendelian randomisation (MR), suggest that high LDL-c and ApoB directly increase the risk of ALS [
13]. Observational studies exploring the effect of lipid biomarkers on survival in people with ALS have been somewhat inconsistent, variably indicating potential relationships between lower HDL-c [
14], higher total cholesterol (TC), triglyceride (TG), and LDL-c, and improved survival [
15‐
17]. A meta-analysis of observational studies did not support a relationship between biomarkers of lipid metabolism and survival in ALS [
18], but a large cohort study demonstrated that increased HDL-c is associated with worse survival [
18].
The evidence regarding the impact of lipid-lowering medication use on ALS risk is conflicted, with some studies indicating an increased risk of ALS following statin initiation and others indicating no association [
19,
20]. MR methods studying the effect of lipid-lowering drugs on ALS risk, using genetic proxies for the targets of lipid-lowering drugs, suggest a protective effect of HMG-CoA reductase and APOB inhibition. Similar research relating lipid biomarkers and lipid-lowering therapies to the risk of FTD, using either classical epidemiological approaches or genetic epidemiological techniques, is limited.
This study aimed to summarise the current literature surrounding the impact of lipids on ALS and FTD risk using a meta-analysis of observational studies and to study the effect of lipid biomarkers on the risk of ALS and FTD and survival in ALS using MR analysis. Additionally, the potential effect of lipid-lowering drugs on ALS and FTD risk and ALS survival was explored using a genetic proxy approach.
Discussion
Meta-analysis of three cohort studies indicated that elevations in LDL-c and HDL-c are associated with increased and decreased risk of ALS, respectively. No observational studies examining lipid traits and FTD risk were found. In support of a causal role for the observed association between LDL-c and ALS risk, two-sample MR analysis provided evidence of a potential causal association between genetically predicted higher levels of LDL-c, TC and ApoB and risk of ALS, and of higher ApoB levels and FTD risk. However, no association was identified for HDL-c or ApoA1 in relation to ALS or FTD. Using genetic proxies for lipid-lowering therapies targeting four individual genes, this study suggests that reducing LDL-c levels through targeting of APOB (a genetic proxy for Mipomersen treatment) could reduce the risk of ALS and FTD. All findings were robust to extensive sensitivity analyses. In relation to the effects of lipid biomarkers on the aggressiveness of ALS, no evidence of a causal association between lipid biomarkers or cholesterol-targeting therapies and survival in ALS was found.
The extant literature examining the role of lipids and apolipoproteins influencing the development of ALS and FTD is limited. One finding of our meta-analysis, that elevated levels of LDL-c are associated with increased risk of ALS, is supported by causal evidence from our MR study. Previous MR studies have reported similar results, highlighting the negative effect of LDL-c [
13,
50]. We did not find a statistically significant association between ApoB and ALS in the meta-analysis of two cohort studies, which contrasts with our MR analysis and previous genetic epidemiological studies [
50]. This might be explained by the high heterogeneity in the meta-analysis (86%), the relationship between ApoB and LDL-c—since ApoB is the major apolipoprotein constituent of a range of circulating lipid particles beyond LDL-c—or potential alterations in lipid biomarkers that occur before the onset of symptomatic ALS, influencing the associations in observational studies [
51].
The second finding of our meta-analysis, that lower levels of HDL-c are associated with a higher risk of ALS, was not supported by causal evidence from MR. One case–control study [
49] also identified an inverse association, where higher HDL-c increased the risk of ALS, which could be explained by reverse causation, as case–control studies are more susceptible to this, along with variations in population characteristics, measurement methods, and confounding factors that may contribute to the conflicting findings. This is in keeping with prior genetic epidemiological studies, which have also failed to show evidence of a causal role for HDL-c in ALS risk [
13]. This parallels findings in cardiovascular disease prevention, in which the robust association between lower HDL-c and higher risk of cardiovascular disease using classical epidemiological methods has not been supported by genetic epidemiological analysis or randomised control trials of treatments aiming to increase HDL-c levels [
52,
53]. There is some evidence for very high levels of HDL-c increasing the risk of cardiovascular disease [
54], but it is not clear that this explains the discrepancy between established observational and genetic epidemiological findings, the cause of which remains enigmatic. It may, therefore, indicate that HDL-c levels have a role in the prediction of ALS risk but do not themselves mediate that association.
Though no observational studies were identified that examine FTD risk in relation to lipid biomarkers, our MR analysis provides evidence that elevated levels of ApoB increase the risk of FTD. Despite the high correlation between ApoB and LDL-c levels, no causal association was identified between LDL-c and FTD risk. Although all MR analyses of FTD risk were of lower power than analyses of ALS due to the much smaller sample size and SNP coverage in the FTD GWAS, the near-zero association of LDL-c and FTD risk and higher estimated power to detect an effect for LDL-c suggest that insufficient power does not explain this discrepancy.
Multiple observational studies have tied metabolic factors during the course of symptomatic ALS with differences in disease progression and survival, including an association between lower levels of HDL-c and shorter survival, higher levels of TG, TC and LDL-c and longer survival [
14‐
18]. Our MR analysis did not find evidence of a causal relationship between any of the lipid species investigated and survival in ALS. This might be attributable to the lower power to detect a small effect as a consequence of the small sample size of the outcome GWAS of survival in people with ALS, as well as that ALS survival may not be very heritable and that the genetic determinants of lipid metabolism in health may differ from those in disease. However, our findings are in accordance with previous genetic epidemiological work in which polygenic scores for lipid levels were not associated with survival in ALS patients [
18]. Observational associations between lipids and survival might be explicable due to systemic metabolic alterations occurring in more aggressive diseases or relate to dietary changes occurring because of the disease process, as opposed to directly influencing survival. Studies of the effect of lipids on the risk of FTD are limited. One previous genetic epidemiological study exploring this relationship did not find an association between lipids and risk of FTD (specifically FTD with TDP-43 pathology), perhaps owing to the smaller sample size and lower genomic coverage of the outcome GWAS [
55].
Using genetic proxies for drug targets, we identified a potential effect of targeting APOB as a means to reduce the LDL-c levels and the risk of ALS and FTD. APOB is the major protein constituent of LDL-c, and levels of the two are highly correlated, but it is also the main protein constituent of other lipoprotein particles, including very low-density lipoprotein cholesterol and chylomicrons. APOB acts as a ligand for the activation of the LDL receptor [
56]. The discrepancy between causal effects for LDL-c and ApoB levels on FTD risk is, therefore, somewhat unexpected, particularly given that estimates of statistical power (Figure S4) indicate higher power to detect a given effect for LDL-c compared with ApoB. A potential explanation for this discrepancy is that direct measurement of ApoB represents a more accurate measure of the concentration of lipoprotein particles, independent of the amount of cholesterol or other lipids per particle [
57,
58]; again, this parallels cardiovascular disease in which the LDL particle number and ApoB levels more accurately reflect disease risk [
58].
As a major constituent of the central nervous system with crucial roles in its normal functioning, there is great interest in defining the role of lipids in neurodegenerative disease [
59]. Beyond ALS and FTD, dysfunctional cholesterol metabolism has been identified in Alzheimer’s, Parkinson’s and Huntington’s disease, implicating lipid pathways as a broad aetiological factor in neurodegenerative disease [
60]. Since brain cholesterol is largely synthesised in situ without significant translocation of cholesterol from blood [
61], it remains unclear how circulating lipids reflect neurodegenerative disease risk. Hypotheses that might explain this relationship include a role for oxidised cholesterol species [
62], a means of transporting cholesterol between the central nervous system and circulation that are toxic to neurons, or cholesteryl esters [
63], a means of sequestering excess cholesterol but which exert oxidative stress on neurons. Lipid levels may also reflect far more complex changes evolving at the synaptic level [
64].
The small sample size of the GWAS of FTD and of survival in people with ALS are significant limitations. This has inevitably impacted the power of the analysis. ALS is a more pathologically homogeneous disease, with 97% of cases being associated with TDP-43 pathology, and only 50% of FTD cases are associated with TDP-43 pathology. Although exposures such as alterations in blood lipids might have an effect on FTD risk that is not pathology-specific, any pathology-specific effect would negatively impact the power of this analysis. No GWAS of sufficient size examining pathological subtypes of FTD exists to probe this question. Limitations relating to the genetic proxies of therapeutic targets include that genetic variants reflect the effect of lifelong changes in lipid levels on ALS or FTD risk, and the magnitude of the effect may not be comparable with the short-term effects of lipid-lowering drugs. Our study only predicts the on-target effects of specific drug targets, and these models do not estimate potential off-target effects. Horizontal pleiotropy cannot be completely excluded, although various sensitivity analyses were performed to test the assumptions of MR analyses. Our analyses also assume no gene–environment or gene–gene interactions and linear and time-dependent effects of drug targets on ALS or FTD risk. Furthermore, since our findings were limited to GWAS of individuals of European ancestry, these findings are not necessarily valid for other genetic ancestries.
In conclusion, these data support a causal role for higher LDL-c and total cholesterol increasing the risk of ALS and higher APOB increasing the risk of both ALS and FTD. The findings reveal the potential for APOB inhibitors to reduce the risk of sporadic ALS and FTD. Further work in monogenic forms of ALS and FTD is necessary to determine whether reducing blood lipids influences risk in those at high risk. Understanding the mechanisms by which LDL-c and ApoB mediate ALS and FTD risk may help identify additional approaches to the prevention of these diseases.