ABI3 and PLCG2 missense variants as risk factors for neurodegenerative diseases in Caucasians and African Americans

Rare nonsynonymous variants in ABI3 (p.Ser209Phe; rs616338-T) and PLCG2 (p.Pro522Arg; rs72824905-G) have recently been implicated in conferring risk and protection, respectively, for Alzheimer’s disease (AD) [1]. Identified using a whole-exome microarray and genotype imputation in the largest AD case-control series to date, these variants are part of a growing number of rare variants now implicated in AD [1]. However, the association of rs616338-T and rs72824905-G has not yet been replicated in an independent cohort and their mechanisms of pathogenesis remain unknown.

ABI3 encodes the Abelson (Abl) interactor (Abi) family protein 3 (a.k.a. NESH), which is involved in actin cytoskeleton organization and functionally distinct from ABI1 or ABI2 [2]. ABI3 interacts with WASp-family verprolin homologous protein 2 (WAVE2), as part of the WAVE regulatory complex (WRC), a heterocomplex, which also includes NCK-associated proteins (NAP), Specifically Rac-associated 1 (SRA1), and Hematopoietic stem progenitor cell 300 (HSPC300) [3]. WRC activates the actin nucleator actin-related protein-2/3 (Arp2/3) to induce actin polymerization, which is necessary for cell motility in many functions including immune responses [3]. These findings and the identification of ABI3 expressing microglia clusters exclusively in AD brains and around amyloid beta (Aβ) plaques [4] may suggest a role for ABI3 in microglia motility.

PLCG2 encodes phosphoinositide-specific phospholipase C family protein PLCɣ2, which, upon extracellular ligand stimulation of receptor tyrosine kinase, is activated by recruitment to the cell membrane and phosphorylation [5, 6]. Consequently, PLCɣ2 hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP2) to inositol 1,4,5-trisphosphate (IP₂) and diacylglycerol (DAG), which increases calcium (Ca²⁺) influx and extracellular signal-regulated kinase (ERK) phosphorylation, thereby inducing cellular activation in settings including inflammation and innate immunity [5, 6]. The murine Plcγ2^Ali5 gain-of-function point mutation in the catalytic domain of this protein near its auto-inhibitory domain leads to enhanced Ca²⁺ influx in B-cells and expansion of innate inflammatory cells, resulting in autoimmunity and inflammation in this model [7]. Autosomal dominant PLCG2 in-frame deletion [8] or missense [p.Ser707Tyr] mutations [9] in the autoinhibitory domain, result in constitutively active or hyperactive phospholipase function, respectively, and lead to diseases characterized by autoimmunity and immunodeficiency, known as PLAID (PLCɣ2-associated antibody deficiency) [8] or APLAID (autoinflammation and PLAID) [9].

Given the above and the potential roles of ABI3 and PLCG2 in various arms of the immune system, it is likely that the rare missense AD-associated variants within these genes confer their effects through alterations in neuroimmunity, including neuroinflammation, a vital aspect of AD pathophysiology [1, 10]. It is well established that innate and/or adaptive immune changes are observed as features of multiple neurodegenerative diseases as well as in the aging brain [11]. It is therefore possible the ABI3 and PLCG2 variants may also influence risk of other neurodegenerative diseases including primary tauopathies and synucleinopathies [11‐13]. Identification of association with disease status in other neurodegenerative diseases may aid in understanding the mechanism by which ABI3 and PLCG2 contribute to disease pathophysiology. Therefore, in this study, we sought to replicate the association with AD previously observed with ABI3 rs616338-T and PLCG2 rs72824905-G in Caucasians, test if this association is also observed in African-Americans, and determine if these variants associate with risk of other neurodegenerative diseases, namely Parkinson’s disease (PD), dementia with Lewy bodies (DLB), progressive supranuclear palsy (PSP) and multiple system atrophy (MSA).

In this study we evaluated the association of the recently discovered [1] variants in the microglia-enriched genes ABI3 and PLCG2 in AD and four other neurodegenerative diseases comprised of three α-synucleinopathies (PD, DLB, MSA) and a primary tauopathy (PSP); investigated an African-American AD case-control cohort for presence and frequency of these variants; and studied the expression patterns of these genes in two brain regions (temporal cortex, cerebellum) in AD vs. control; and PSP vs. control samples. We validated the associations previously reported with ABI3_rs616338-T (p.Ser209Phe) and PLCG2_rs72824905-G (p.Pro522Arg) in a Caucasian AD case-control cohort, and observed a similar direction of effect in DLB. In contrast, the effect estimates observed for PSP and MSA were in the opposite direction to that in AD for both variants. Neither of the two variants appeared to associate with risk of PD and they both had exceedingly rare frequencies in the African-American AD case-control cohort.

Similar direction of effect for both variants in AD and DLB, despite lack of or opposite trends of association in other α-synucleinopathies (PD, MSA) or a tauopathy (PSP) may be due to high rates of existing AD pathology in DLB, observed in about 2/3 of autopsy-proven DLB patients [32, 33]. It is also possible that these variants represent a shared genetic component between AD and DLB, similar to APOE [34‐36], the effects of which have also been demonstrated in pure DLB [36]. AD and DLB share similarities with respect to neuroinflammation, such as microglial activation and cytokine induction, which are observed in both neuropathology studies of these diseases, as well as in vitro studies of effects of Aß or α-synuclein on microglia [33]. Thus, it remains a possibility that the potential effects of ABI3_rs616338-T and PLCG2_rs72824905-G on microglial function may influence AD and DLB risk, similarly. It should be noted that in a recent study of a larger DLB cohort, including 829 pathologic Lewy body diseases (LBD) patients, another microglial gene variant TREM2 p.R47H was found to associate with disease risk only in those LBD patients with predominant AD pathology, but not in those with low AD pathology [37], indicating lack of a role for this variant in DLB pathophysiology. In our study, only 67 of 306 DLB patients were autopsy confirmed, hence we were not powered to test for genetic associations while adjusting for the presence of concomitant AD pathology. Future studies of pathologically confirmed DLB cohorts with low or no AD pathology are needed to distinguish the effects of ABI3_rs616338-T and PLCG2_rs72824905-G on pure DLB.

One surprising finding from our study is the identification of opposite trends of association for these variants with both PSP and MSA patients, which reached nominal significance for PLCG2_rs72824905-G in the MSA cohort but were not statistically significant in the other comparisons. These results could merely represent false positive findings given the rarity of the tested variants and the relatively small PSP and MSA cohorts, although all patients for both diseases were autopsy proven, which ensures diagnostic accuracy. Nevertheless, despite its rarity, the higher MAF of PLCG2_rs72824905-G in PSP and MSA which is ~ 2–2.4 times as that of controls and 3–4 times as that of AD patients, raises the possibility of opposite effects of this variant in AD vs. these two neurodegenerative diseases. To hypothesize on the biological basis of such opposing effects, the potential effect of the PLCG2_rs72824905-G variant on immune function needs to be considered.

PLCG2_rs72824905-G causes a proline to arginine substitution (p.Pro522Arg), located between the nspPH and nSH2 domains, which comprise part of the autoinhibitory region of PLCγ2 [38]. In vitro studies have demonstrated that deletion of amino acids in the nspPH-nSH2 linker domain cause an increase in PLCγ2 activity, both basal and in response to stimulation [38]. Deletions [8] and a missense [9] variant within the nSH2 domain in humans have been shown to cause PLAID and APLAID (p.Ser707Tyr), respectively [39]. These diseases are characterized by immune dysregulation, antibody deficiency and autoinflammation due to a complex mix of loss and gain of function of PLCγ2 [39]. Additionally, point mutations in the nspPH domain in a murine model have also been shown to increase PLCγ2 activity [40] and in vivo studies have demonstrated that PLCG2 gain of function mutations can lead to autoinflammation [7]. The PLCG2_rs72824905-G variant causes an amino change from a non-charged to a positively charged residue, likely altering the structure of the protein. This can affect the interaction of the autoinhibitory and catalytic regions, consequently reducing the autoinhibition of PLCγ2. Thus, PLCG2_rs72824905-G is expected to increase the PLCγ2 signaling activity, which may induce activation of inflammation and innate immunity [5, 6].

Identification of AD candidate genes and risk variants in innate immunity pathways [41], enriched expression of many of these genes in microglia [28], gene expression network [42] and expression quantitative trait loci (eQTL) studies [43‐46] implicating regulatory changes of these genes and variants in brain tissue collectively provide strong evidence for role of innate immunity in AD. Rare, coding TREM2 variants that increase risk of AD appear to have loss of function effects and are associated with reduced amyloid plaque-associated microgliosis [47]. A common potentially regulatory variant that is associated with modestly higher brain levels of TREM2, also associates with a protective effect in AD [46]. Consistent with this, elevated TREM2 gene dosage reduced amyloid pathology and improved memory in a mouse model of AD [48]. These studies suggest that enhanced function or increased levels of microglial, innate immunity genes may confer protection in AD, especially through their effects on Aß clearance.

In contrast, enhanced innate immunity may have detrimental effects in non-AD neurodegenerative diseases or in non-Aß components of AD pathology. Complement and microglial activation increased tau pathology in mouse models, whereas their inhibition and depletion, respectively, reduced synapse/neuron loss and tau pathology (reviewed [10]). Microglia were found to induce neuron-to-neuron spread of tau in an adeno-associated virus–based mouse model [49]. Although prion-like spread has been suspected for tau in PSP and α-synuclein in MSA [50], the role of microglia in propagation of proteinopathy in these diseases has not been demonstrated. Neuroinflammation and microglial activation is observed in both PSP [51, 52] and MSA [53] in disease affected brain regions, although whether this is beneficial or detrimental to the disease progress remains to be established.

In light of these collective data, one model, which may reconcile the opposing trends in the effects of PLCG2 and ABI3 variants in AD vs. PSP and MSA in our study is that activation of innate immunity relatively early in the neurodegenerative process may be beneficial, especially in the context of extracellular Aß pathology. However, persistently activated innate immunity and microglia may be detrimental for propagation of intracellular tau or α-synuclein, which characterizes diseases such as PSP and MSA, as well as for late-stage neurodegeneration. It should be emphasized that the genetic association findings in our study are statistically marginal, and until replicated in other cohorts, the above model remains speculative.

Our study also investigated an African-American AD case-control cohort for associations with ABI3_rs616338-T and PLCG2_rs72824905-G. While this cohort is of modest size, it enabled the observation that MAF for both variants are smaller than those of both Caucasians AD and elderly control patients. This suggests that these variants are unlikely to influence AD risk in African-Americans to the extent observed in Caucasians. This does not, however, rule out ABI3 or PLCG2 as potential AD risk genes in African-Americans. We have identified TREM2 coding variants that confer AD risk in African-Americans [54], but these were different than the AD risk variants identified in Caucasian subjects [55, 56]. Deep sequencing efforts on African-Americans and other non-Caucasian races are necessary to identify the full spectrum of AD risk variants in these and other genes.

We also characterized the gene expression patterns and co-expression networks of ABI3 and PLCG2 in two brain regions, namely temporal cortex that is affected with AD neuropathology and cerebellum that is relatively spared, in AD, PSP and control samples. PSP was included in the expression analysis as a neurodegenerative disease, which like AD, has tau pathology, but unlike AD, lacks Aß pathology, and as such may help distinguish expression changes in the context of these different neuropathologies. We previously showed that comparative transcriptomics utilizing different neurodegenerative diseases may identify pathways that are commonly vs. distinctively perturbed in these conditions [26]. We determined that temporal cortex, but not cerebellum, levels of these genes and their co-expression networks are increased in AD, but not in PSP, compared to control samples. These elevations are abolished after adjusting for levels of cell type markers, including CD68, which is highly expressed by monocytes and macrophages and upregulated in actively phagocytic cells [57]. Our findings suggest that ABI3 and PLCG2 are members of a network of co-expressed microglial genes, the levels of which are upregulated in brain regions affected by AD pathology, where this upregulation is either the result of enhanced numbers or activation state of microglia or both. In this study, we utilized our gene expression data from the AMP-AD Consortium [19], which is primarily focused on brain regions affected with or relatively spared in AD. The fact that we did not observe microglial transcript elevations in PSP suggests that either such microgliosis is not a key aspect of this disease or that PSP-affected brain regions should be examined to observe such changes. Future studies focusing on brain regions affected in PSP are needed to address these possibilities. Although the gene expression changes we observed in AD may be driven by the microglial response to AD pathology, the presence of many AD candidate risk genes in these immune/microglial networks suggests that perturbed function or levels of these genes and networks are likely to play a causal role in AD pathophysiology.

In summary, our study provides effect size estimates for the recently discovered ABI3 and PLCG2 variants [1] in five different neurodegenerative diseases and an African-American AD case-control cohort and also characterizes the expression patterns of these genes in two brain regions in AD, PSP and controls. The strengths of our study are the sizable AD case-control cohort, evaluation of a variety of neurodegenerative diseases and an African-American cohort, neuropathologic diagnosis of all PSP and MSA and > 40% of AD cases and thorough examination of gene and co-expression networks for these genes. Despite these strengths, the sizes for the non-AD neurodegenerative cohorts remain modest, therefore assessment of larger cohorts for replication of the findings is necessary.

Our findings strengthen the evidence for the role of ABI3_rs616338-T and PLCG2_rs72824905-G as risk and protective factors, respectively, for AD and suggest a similar role for these variants in DLB, but suggest opposite effects, especially for PLCG2_rs72824905-G in PSP and MSA. These and our gene expression data provide a rationale to investigate the roles of these and other innate immunity genes for their distinct roles in different neurodegenerative diseases.