The Genetics of Dementia with Lewy Bodies: Current Understanding and Future Directions

Dementia with Lewy bodies (DLB) is a neurodegenerative disease that shares clinical, pathological and genetic features with Parkinson’s disease (PD) and Alzheimer’s disease (AD). Lewy body dementia is a term that encompasses both DLB and Parkinson’s disease dementia (PDD), and the ‘one year rule’ is used to assess the temporal onset of dementia versus parkinsonism, in order to differentiate the two diseases. If parkinsonism occurs at the same time or within 1 year of dementia, a diagnosis of DLB is made, whereas if parkinsonism precedes the onset of dementia by a year or more, PDD is diagnosed. The fundamental feature of DLB is dementia, which often occurs with parkinsonism, early fluctuations in cognition and attention, visual hallucinations and REM sleep behaviour disorder (RBD) [1••, 2]. Autonomic dysfunction, sleep irregularities, depression and anxiety are also common. DLB is a devastating disease because of its symptoms, as well as the fact that there is currently no cure or disease modifying therapy available. Furthermore, DLB is thought to have a shorter disease duration [3] and decreased survival rate compared to AD [4]. The need to understand the disease pathobiology is imperative for the development of disease-specific therapies.

It is now clear that DLB has a strong genetic component. Although most cases are sporadic, a number of reports have demonstrated the occurrence of the disorder in families [5‐13], in addition to the identification of genetic loci that modulate risk for the development of DLB [14, 15, 16••]. As such, using common genetic variants, the proportion of the phenotype that can be attributed to genetic factors was estimated to be 36% [16••].

Despite the fact that we now know that genetics plays a role in the disease, genes that cause DLB are still to be identified. In comparison with AD and PD, we know far less about the genetic basis of DLB. There are multiple reasons for this. Firstly, DLB is difficult to diagnose, as phenotypic overlaps with other neurodegenerative diseases can diminish chances of an accurate diagnosis. The lines between different neurodegenerative diseases can often be blurred, and a patient may show features of one or several other diseases, both ante and post mortem. Additionally, DLB is not as frequently recognized as a disease compared to other well-known disorders, such as AD. Both factors result in a substantial rate of mis- and under-diagnosis. This has hindered the collection of large cohorts of cases whose diagnosis is certain, and as a consequence, has limited large-scale genetic analyses. Furthermore, as the disease is age-related, and typically occurs in those aged 65 or older, the likelihood of gathering biomaterials from multiple family members for genetic testing of familial DLB is small. This, coupled with the fact that families with DLB are rare, has limited the understanding of Mendelian DLB genetics. Although still a prevalent cause of disease in the elderly, DLB is less common than diseases such as AD and PD, and hence cohorts will generally be smaller for this disease.

Moreover, some previous genetic studies have focused on Lewy body disease (PD, DLB and PDD) as a whole, and not specifically to genetic alterations that are unique to DLB (Table 1). Ultimately, there is less research conducted in DLB compared to AD and PD (Fig. 1a). DLB was only described as a separate disease entity in 1984, comparatively later than AD or PD (Fig. 1b). Thus, genetic research in DLB is only just gathering momentum; for example, the first genome-wide association study (GWAS) in DLB was conducted in 2017 [16••], at a time when we have reached the post-GWAS era in many diseases. Genetic studies have already shown that DLB shares genetic risk factors with AD and PD; however, recent findings suggest that DLB may also have a unique genetic architecture [16••].

In order to adequately implicate a gene in disease, detailed assessment of the gene in question is required, with a view to reproduce findings in independent studies. Replication studies are fundamental to distinguish true associations from false positive findings, and thus in the validation of the results of the initial study [34]. To date, only three genes have been convincingly established to be involved in DLB: APOE, GBA and SNCA. Variation in the SNCA gene can modulate risk for or cause DLB phenotypes. The established risk factors for DLB are also known to impart risk to either AD (APOE) or PD (GBA, SNCA). Of note, AD and PD do not share common genetic risk factors between them [35]. This is in accordance with a recent study, which showed no evidence of correlation between PD and AD, but showed that the genetic correlation between DLB and AD was 0.578 (SE ± 0.075), and between DLB and PD was 0.362 (SE ± 0.107). The APOE locus is highly associated with AD and DLB. When removing this locus from the analysis, the correlation between AD and DLB was reduced to 0.332, which does not differ significantly from the correlation between DLB and PD [36]. It should be noted that this study used a genotyping array enriched for neurological disease variants and so was not entirely unbiased.

The first DLB GWAS was conducted in late 2017, and incorporated a total of 1743 DLB patients (1324 of which had autopsy supported diagnosis) and 4454 controls [16••]. Genotyping data was enriched with SNPs imputed from the haplotype reference consortium, to allow detection of lower frequency variants. SNCA, GBA and APOE loci were significantly associated with DLB in the discovery and replication phases, as well as the meta-analysis of both stages. In the discovery phase, two other loci were genome-wide significant: BCL7C/STX1B (OR 0.74, 0.67–0.82; p = 3.30  ×  10⁻⁹) and GABRB3 (OR 1.34, 1.21–1.48; p = 2.62  ×  10⁻⁸). However, the GABRB3 signal did not maintain significant association when restricting analysis to pathologically diagnosed samples (p = 1.21  ×  10⁻⁷). Loci with p < 5  ×  10⁻⁶ were genotyped in the replication phase, and CNTN1, a locus that reached suggestive association in the discovery phase (OR 1.58, 1.32–1.88; p = 4.32  ×  10⁻⁷, pathologically diagnosed cases only), was significantly associated in the replication phase (p < 0.05). As CNTN1 lies < 500,000 bp away from the LRRK2 locus, samples that harboured the p.Gly2019Ser LRRK2 variant were removed from analysis, without significant effect on the association at CNTN1. However, whether other LRRK2 variants may mediate the association cannot be ruled out. BCL7C/STX1B was genome-wide significant in the meta-analysis; however, this was mainly driven by the discovery phase, and requires further replication. Despite the identification of interesting candidates, replication of novel associations is required, and further GWA studies with larger cohorts are warranted. Heritability was shown to be higher than expected given chromosome size on chromosomes 19, 5, 6, 7 and 13 but apart from chromosome 19, no genome-wide significant variants were identified, suggesting that, perhaps, other DLB genes may be present on these chromosomes.

Although not particularly common, several DLB families have been studied, but these have failed to shed much light on the genetics of the disease. Only a small proportion of patients within these families underwent genetic analysis, and if performed, it was usually minimal with very few cases assessed at the exome or genome levels. As mentioned above, variation in SNCA sequence or dosage has been identified in families as the cause of mixed PD and DLB phenotypes. Some DLB patients have a family history of neurodegenerative disease, but not necessarily of DLB, with family members being diagnosed with AD or PD. Alzheimer families with mutations in APP, PSEN1 and PSEN2 have been described with phenotypes of mixed parkinsonism and dementia suggestive of DLB [91‐93], and extensive Lewy body pathology has also been found in Alzheimer’s families with these mutations [94, 95], suggestive of a possible mechanistic link. Indeed, DLB cases frequently present Aβ pathology at autopsy [96], and it has been suggested that Aβ accumulation can trigger Lewy body disease [97]. A recent review by Vergouw and colleagues [98] provides a summary of familial studies in DLB. The most promising study to identify a novel DLB gene relied on linkage analysis in a DLB family which identified a locus on chromosome 2q35-q36 [11]; however, no gene was subsequently identified [99]. This could be because the variant was not detectable with current technology or, perhaps, the linkage results were misleading.

As our knowledge of the disease progresses, the diagnostic criteria for DLB are updated (Fig. 1b) to provide increasing sensitivity and specificity for accurate diagnoses [1••]. It is also worth considering whether previous reports in DLB families, which occurred when the first consensus criteria were in use, have patients that would meet current diagnostic criteria for DLB. The lack of autopsy data in some of these studies makes a reliable diagnosis difficult. Moving forward, genetic analysis of multiple family members with well characterized clinical and autopsy data that meet more recent diagnostic criteria for DLB, will be paramount in identifying novel disease-causing genes.

As DLB may clinically and pathologically resemble Alzheimer’s or Parkinson’s diseases, speculation as to whether AD- or PD-causing genes may also be involved in the pathogenesis of DLB prompted the study of these genes in small cohorts of mainly sporadic DLB cases. However, due to the phenotypic similarities between diseases, it is still unclear whether the mutations identified play a role in DLB or simply occur in misdiagnosed cases. This issue is complicated further by the heterogeneity of phenotype that can be associated with some of these mutations [41, 46‐48, 91, 93, 100].

Mutations that are established to be pathogenic in APP and PSEN1 were found in either a clinical DLB case [29]; or in pathologically confirmed DLB cases, of which 69% of the cohort also met pathological criteria for Alzheimer’s disease [27]. Furthermore, studies of genes that cause other neurodegenerative diseases have identified variants in DLB patients that have been previously reported, but are of unknown pathogenic consequence in genes such as CHMP2B, SQSTM1, PSEN2 [30] and GRN [29]. Novel variants in MAPT [29] and multiple variants of uncertain significance have also been reported. A compound heterozygous mutation in PARK2 was identified in a DLB patient [30]; however it is unclear whether the data was phased. Rare variants in GRN were hypothesized to be associated with DLB, albeit in a very small study of 58 DLB cases and 380 controls [32]. Nevertheless, if the reported mutation carriers do in fact have DLB, mutations in genes known to cause Mendelian forms of other neurodegenerative diseases only occur in a small proportion of DLB cases.

The MAPT H1G sub-haplotype was associated with clinical DLB; however, the association was attenuated when pathologically diagnosed samples were included in analysis [24], weakening evidence for a role specific to DLB. The H1 haplotype may be associated with more severe alpha-synuclein deposition, as suggested in a small study of 22 DLB brains [101]. MAPT p.Ala152Thr has been proposed to be associated with DLB in a clinical DLB cohort [25]. This variant is also associated with AD and FTD-spectrum disorders [102], but not PD [25, 102]. However, the MAPT locus showed no evidence of association in the DLB GWAS [16••], and this is of interest as the MAPT locus is a highly significant result in PD, reaching genome-wide significance even in smaller studies of approximately the same number of cases as the DLB GWAS [103]. Therefore, there is limited convincing evidence for a role of MAPT variation in DLB, and further studies of risk associated haplotypes in larger cohorts are needed.

Variants in the TREM2 gene confer a risk for the development of AD similar to that associated to one ε4 allele of APOE, an association mediated primarily by the p.Arg47His variant, but also by others, such as p.Arg62His. The p.Arg47His variant was not found to be associated with DLB [16••, 22]; however, in a small study of 58 DLB cases, p.Arg62His was nominally associated with DLB (uncorrected p value = 0.0024, OR = 3.2 [95% CI 1.7–27] [32], although this would not survive multiple-test correction. Again, further study is required to identify the role, if any, of TREM2 in DLB.

Pathogenic hexanucleotide repeat expansions in the C9ORF72 gene are the most common cause of FTD and/or ALS. Analysis of the expansion in neuropathologically diagnosed DLB patients found three cases with either 32 or 33 repeats, in a combined total of 1562 patients [17, 19, 104]; whereas studies of clinically diagnosed DLB patients have identified two patients with > 30 repeats [18, 19, 105]. This suggests that repeat expansions in C9ORF72 are not a common cause of DLB.

An increased frequency of the LRRK2 p.Gly2019Ser variant was seen in DLB compared to controls (0.0021 versus 0.0003, respectively) [16••]; however, this was not statistically assessed given its low frequency. This variant has previously been seen in a clinical DLB case [26], and in two Ashkenazi Jewish individuals with DLB [80], demonstrating a low frequency in this disease. Where studied, no other pathogenic LRRK2 mutations were found in DLB cases [27, 29, 30], and no LRRK2 variants were significantly associated with DLB [26]. Whilst LRRK2 genetic variation does not seem to occur often in DLB, it is difficult to distinguish between PD, PDD and DLB and thus assess the contribution of LRRK2 specifically to DLB [106]. LRRK2 mutation carriers in PD have a lower rate of dementia [107], perhaps providing further evidence against a role in DLB.

The apparent lack of association for other strong AD and PD risk loci, such as TREM2, MAPT, CLU, PICALM, and BIN1 may hint at distinct genetic features for DLB, or may be attributed to insufficient power to detect associations.

Genetic research in DLB is only just beginning to come together, providing hope for future characterization of the genetic architecture of the disease. In order to identify additional genes implicated in DLB, it will be imperative to study more individuals with the disease. This will require collaborative approaches in order to increase cohort numbers and more studies that are focused on replicating results. As well as generating genetic data, it is important to collect detailed clinical and pathological data on patients studied.

There is a clear need for more unbiased genetic studies (genome- or exome-wide). The majority of genetic studies in DLB thus far have been hypothesis based, largely trying to identify candidate genes (Table 1). Table 1 also highlights the fact that some DLB samples have been used in multiple genetic studies, an event that should be made clear and that can severely bias results, certainly for a disease where sample collections are small.

Large-scale genetics studies in neurodegenerative diseases have been dominated by European and American cohorts. The study of patients from other populations could allow the discovery of population-specific, predisposing variants, which in turn may provide novel insights into the biological processes that occur in disease. DLB research in Japan has been invaluable for furthering our understanding of the disease. For instance, the earliest identification of DLB patients [108], and the study of [(123)I]MIBG myocardial scintigraphy to distinguish DLB [109], were first proposed by Japanese scientists. Detailed, large-scale genetic analyses of these patients could be transformative for the field.

Understanding the genetic bases of a disease can allow us to identify pathways and mechanisms involved in the disease pathobiology. An example of this is in AD, where the identification of APP, PSEN1 and PSEN2 mutations were crucial for the development of the amyloid cascade hypothesis. Furthermore, genetics may be able to help us tease apart PD, AD, DLB and PDD using molecular data. As diagnostic criteria [1••] and understanding of the disease improves [110‐112], we will be able to have a better understanding of the biological processes underlying DLB, and this may lead to the identification of disease-specific therapeutic targets.

Next-generation sequencing technology has revolutionized genetic analysis, and in combination with large-scale databases of genetic variation such as gnomAD or ExAC [113], has allowed us to have a better understanding of genetic variation in humans. This knowledge is going to improve as datasets increase—the 1000 Genomes Project, which began in 2008, included whole genome data from 2504 individuals [114], yet current databases such as ExAC and gnomAD [113], now provide data from 60,706 exomes or 123,136 exomes and 15,496 genomes, respectively. Although these types of datasets will include information from individuals who will suffer from neurodegenerative disease, they have been invaluable in informing the evaluation of potentially pathogenic variants [115]. Furthermore, improved interpretation of GWAS results will enable genetic variants to be linked to a functional role in disease risk [116]. In addition, polygenic risk scores can be used to calculate an individual’s lifetime risk for disease based on their genetic profile, a process that has already been implemented in Alzheimer’s and Parkinson’s diseases [117, 118], and that may allow for identification of presymptomatic individuals.

The role for genetics in human biology is multifaceted and complex, and thus we need to integrate DNA sequencing, with RNA and protein expression, tissue and cell-specific expression and epigenetic analyses in order to reveal as complete a picture as possible. The overall objective is to understand the pathways that are perturbed in disease and that should be therapeutically targeted. Therapies will likely need to be administered prior to clinical disease onset, and therefore must target the prodrome. Genetic studies may allow us to identify those particularly susceptible for the development of the disorder, by utilizing polygenic risk scores or causative mutations.

Moreover, genetics has informed the use of biomarker: by associating genes with disease, there is the potential to analyse the proteins encoded by those genes in patients in order to aid diagnosis. For example, it has been suggested that Gcase [90] or alpha-synuclein [54] levels may be altered in the CSF of DLB patients. A recent study identified phosphorylated alpha-synuclein in skin biopsies of DLB patients, which was absent in controls, or those with an alternative diagnosis of dementia [119]. Although requiring replication, phosphorylated alpha-synuclein in skin was suggested to contribute to autonomic dysfunction and may provide an easy and relatively inexpensive biomarker for DLB [120].

Although DLB genetic research is still only in its beginning, interest in this area has increased, resulting in the identification of several genes that are involved in the disease. This is the first step to obtain a complete picture of the genetic architecture of DLB. As we get closer to this stage, we will be able to better understand disease pathogenesis and to nominate candidate disease-specific therapeutic targets, which will enable us to slow, or even halt, this devastating disease.