Alzheimer's disease risk alleles in TREM2 illuminate innate immunity in Alzheimer's disease

Genetic studies have provided the best evidence for cause and effect relationships in Alzheimer's disease (AD). Indeed, the identification of deterministic mutations in the APP, PSEN1 and PSEN2 genes and subsequent preclinical studies linking these mutations to alterations in Aβ production and aggregation have provided pivotal support for the amyloid cascade hypothesis. In addition, genetic, pathologic and biological studies of APOE have also indicated that the genetic risk for AD associated with APOE4 can be attributed, at least in part, to its pro-amyloidogenic effect on Aβ. In recent years a number of SNPs that show unequivocal genome-wide association with AD risk have implicated novel genetic loci as modifiers of AD risk. However, the functional implications of these genetic associations are largely unknown. For almost all of these associations, the functional variants have not been identified. Very recently, two large consortiums demonstrated that rare variants in the triggering receptor expressed on myeloid cells 2 (TREM2) gene confer significant risk for AD. TREM2 is a type 1 membrane receptor protein primarily expressed on microglia in the central nervous system that has been shown to regulate phagocytosis and activation of monocytes. Previously it had been shown that homozygous loss of function mutations in TREM2 cause polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy (PLOSL, Nasu Hakola disease) and also a pure form of early-onset dementia. The association of TREM2 variants with AD brings innate immune signaling into the light, affirming innate immunity's role as a significant factor in AD pathogenesis.


TREM2 variants are associated with Alzheimer's disease risk
Recently, the unequivocal associations of SNPs within genetic loci that encode genes that function in innate immunity have added genetic support to the notion that innate immunity may have a signifi cant role in AD. Variants in CR1 and CLU, which play roles in the complement system, repeatedly show signifi cant genetic associations with AD [9][10][11][12][13] whereas other genes (CD33, MS4A6A, MS4A4E, ABCA7, CD2AP) with either established or likely roles in innate immune function are also implicated as AD risk loci [12][13][14][15]. In addition, there appears to be a signifi cant overrepresentation of association within genetic loci that encode innate immune genes [16,17]. However, as with the majority of genetic associations with AD, the functional variants within these loci are unknown; thus, it is even premature to defi nitively conclude that such association refl ects a func tional variant that impacts function or expression of the encoded innate immune gene or alternatively alters a neighboring gene or non-coding RNA. In addition, the overall AD genetic risk or protection associated with these loci is small. Although if functional variants within these loci are defi nitively identifi ed, it is possible that the risk associated with such rare functional variants could be much more signifi cant.
Because of these issues, it has been challenging to experimentally assess the biological underpinnings of the potential genetic link between these novel loci and AD. However, recent studies have demonstrated that rare coding variants in triggering receptor expressed on myeloid cells 2 (TREM2), a known regulator of microglial activation and phagocytosis, confer substantial risk for AD [18,19]. TREM2 is highly expressed on microglia, as well as osteoclasts, dendritic cells and macrophages. It is a type 1 transmembrane glycoprotein that binds poorly characterized ligands (for example, bacteria, cell debris, and an astrocytoma cell-line), and, upon ligand binding, signals through DAP12 (TYROBP), an immunoreceptor tyrosine-based activation motif (ITAM)-containing transmembrane adaptor protein, and the SYK kinases that interact with the ITAM domain of DAP12 [20][21][22]. Since the cytoplasmic domain of TREM2 by itself has no intrinsic signaling capacity, it relies on DAP12 for signal transduction [23]. Homozygous, loss of function mutations in both TREM2 and DAP12 are known causes of polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy (PLOSL), which is also known as Nasu Hakola disease [24][25][26][27]. By inference, one would expect that variants that reduce DAP12 function (and possibly other downstream signaling molecules) might also confer risk for AD if they result in partial but not complete loss of function. Notably, other homozygous mutations in TREM2 (T66M, Y38C, Q33X) have been associated with dementia without bone cysts [19,28].
ough the clinical presentation of these subjects is consistent with PLOSL without bone involvement, postmortem brain pathology has not been reported.
Association of TREM2 with AD was initially shown using whole exome sequencing and whole genome sequencing [18,19]. In these studies, the most defi nitive risk for AD was associated with the heterozygous R47H variant (rs75932628-T) of TREM2. Notably, the risk associated with this allele was strong with odds ratios in the initial two studies of 2.9 (95% confi dence interval 2.16 to 3.91) [18] and 4.5 (95% confi dence interval 1.7 to 11.9) [19]; thus, roughly equivalent to the risk associated with one APOE4 allele [29]. In addition, a number of other variants in TREM2 were present in AD patients but not controls, with one variant (D87N) showing signifi cant asso ciation with disease [30]. Since these initial publications, two other publications have confi rmed the risk associated with the R47H variant to other populations and even early onset AD [31,32].
us, TREM2 represents the fi rst gene within the innate immune signaling pathway for which functional variants show association with AD risk.
On the surface these exciting genetic fi ndings appear to provide the fi rst rapidly tractable genetic association between a gene that is known to regulate innate immunity and AD. However, some caution is warranted, as inferring both functional eff ects of these variants on TREM2 and the relationship to the AD pathological cascade is, at this point, largely speculative. e TREM2 R47H and other variants more tentatively associated with AD risk are all located within the extracellular immunoglobulin-like domain of TREM2.
us, similar to the mutations in the same region that cause PLOSL, it is hypo thesized that the variants in TREM2 associated with AD cause loss of function or partial loss of function, reduc ing ligand binding and downstream signaling [18,30]. It is also possible that they result in nonsensemediated RNA decay and reduce TREM2 levels. A soluble form of TREM2 and other variants have also been identifi ed that could infl uence the function, level, or both of TREM2 [33,34].
us, TREM2 could have non-cell autonomous actions. In addition, as the TREM2 ectodomain can be shed [34], it has been speculated that TREM2 may undergo regulated intramembrane proteolysis with the membrane stub being further processed by γ-secretase, and it is possible that these cleavages could be altered by these mutations. Finally, although the initial focus of TREM2 in AD will likely focus on its role in microglial activation, it is important to consider other possible functional roles of TREM2 on cells other than microglia within the central nervous system and also in regulating peripheral immune cell entry and activity in the brain.
Given that Trem2 activation has been shown to enhance phagocytosis and suppress cytokine production in mice [22,35], TREM2 could have very complex biological eff ects relevant to regulation of Aβ deposition as well as innate immune responses triggered by Aβ accumulation, such as regulation of cell-to-cell trans mission of tau, induction of other intracellular proteinopathies and neurodegeneration. Trem2 has been shown to be present in plaque-associated microglia in young and aged APP transgenic mice, presumably as a response to Aβ pathology [19,36,37]. us, TREM2 variants could infl uence age of onset, progression of disease, or both. However, at this point much more data are needed to understand how AD-associated variants infl uence TREM2 function, and how that variation in function alters factors relevant to AD pathogenesis. Notably, postmortem pathological phenotypes of brains from D87N and R47H carriers were well within the normal spectrum of pathologies noted in typical AD [19]. To date, there have been no reports of distinguishing clinical phenotypes in R47H carriers.

Putting TREM2 in context
Innate immune signaling in the brain is highly complex and may refl ect varying states of immune activation and suppression in both health and disease. A potentially useful framework to classify innate immune activation states that was adopted from studies of peripheral macrophages has been to describe microglial phenotype as a classic activation (M1) or alternative activation (M2) state [38][39][40]. However, though this classifi cation system is a useful framework, there is growing recognition that a rigid application of these dichotomous microglial phenotypes may be too simplistic [6]. In general, M1 microglial phenotypes have been associated with neurotoxicity and M2 with a neuroprotective/neuro-remodeling role. Based on what we know about TREM2 functions in myeloid cells and the presumed loss of function eff ects of TREM2 variants associated with AD, one could suggest that the genetic association of TREM2 with AD indicates that suppression of an M2-like neuroprotective microglial response with decreased phagocytosis and increased cytokine production might promote AD pathologies. However, a survey of published studies in the fi eld suggests that a unifi ed view of the role of innate immunity, microglial activation states and AD pathology is not feasible at this time [5,6] . For example, our data and that from other groups show that an M1-like pro-infl ammatory microglial activation state protects from Aβ pathology, whereas an M2-like alternative activation state can promote Aβ pathology [41][42][43][44][45][46][47][48][49]. In contrast, others have reported that an M2-like microglial activation state protects from both Aβ pathology and cognitive and synaptic dysfunction [50][51][52][53][54][55][56].
ere is also evidence, albeit much more limited, that factors that might protect from Aβ pathology might promote tau pathology and vice versa [57][58][59][60]. Like the preclinical data generated to date, epidemiologic and clinical data reveal a fairly confl icted literature on observational and clinical trial fi ndings on non-steroidal anti-infl ammatory drug (NSAID) use and AD [61][62][63][64][65][66][67][68][69][70][71][72]. Long-term NSAID use has been repeatedly shown to confer protection in epidemiologic studies, but clinical trials in AD patients with celecoxib and naproxen have not shown any benefi t [73]. e Alzheimer's Disease Anti-infl ammatory Prevention Trial (ADAPT) testing AD 'prevention' with naproxen and celecoxib was halted due to cardiovascular side-eff ects induced by naproxen; however, follow-up studies along with some post hoc analysis of this trial suggests that treatment eff ects could vary depending on underlying, clinically silent AD pathology at the time of trial enrollment [68][69][70][71][72]. Another intriguing observation relating to human NSAID use and AD was a report showing that naproxen use was associated with increased post-mortem brain Aβ pathology [74].
A fi nal factor that must be taken into account when trying to understand the role of the innate immune system in the AD brain is the recent data that indicate widespread upregulation of innate immune gene expression in the aging human brain [4]. At this point, it is unclear how this underlying age-related 'skewing' of the innate immune response towards an activated state aff ects development and progression of AD. By inference from other chronic infl ammatory conditions, one could argue that a chronic pro-infl ammatory environment would be harmful, and could either directly promote ageassociated decline in function, sensitize the brain to a second insult or even trigger a proteinopathy. Alternatively, one could argue that a proinfl ammatory environment might actually help to remodel the brain and protect it from proteinopathies by promoting the removal of misfolded proteins and danger signals released by degenerating cells. In actuality, the eff ects may be quite complex, with the balance between positive and negative eff ects of innate immune activation on proteostasis and neuro degeneration in AD contextually dependent on the nature, timing, duration, and strength of the specifi c signals. Furthermore, immune 'manipulations' probably have more complex eff ects on innate immune activation and other factors that could infl uence AD pathologies than what are currently being surveyed, and thus a broader approach may be needed to understand how manipulations of a given innate immune pathway impacts AD.
While there has been previous interest in therapeutic strategies targeting infl ammation and innate immunity in AD, probably because of the confl icted preclinical and clinical data, there has been limited activity relating to development of novel innate immune targeting therapies in AD. If future studies can tie the action of ADassociated TREM2 variants to changes in microglial function that infl uence AD-relevant phenotypes, there will be direct genetic evidence that alterations in innate immune responses confer risk for AD. Such data will likely spur renewed interest in development of innate immune modulatory strategies for AD. Indeed, as opposed to therapies targeting the protein aggregates that are most likely to be eff ective as prophylactics, therapies modulating innate immune targets could be predicted to have effi cacy during later disease stages. Given our current understanding of TREM2 function, as well as the large body of data showing that innate immunity can alter proteostasis and neurodegeneration, it is likely that many diff erent novel therapeutic approaches may arise from these studies. Some of these approaches may directly target TREM2 whereas others might target parallel pathways that also regulate microglial activation.
Another intriguing aspect of these new data is that they represent another example of how variants within a genetic locus can confer risk for or cause one type of neurodegenerative disorder when present in a hetero zygous state and cause a distinct disorder in the compound heterozygous or homozygous state. Other known examples are i) heterozygous progranulin (PRGN) mutations resulting in frontal temporal lobar degeneration and homozygous or compound heterozygous PRGN mutants causing neuronal ceroid lipofuscinosis [75,76], and ii) heterozygous glucocerebrosidase (GBA) mutations associated with risk for Parkinson's disease and homozygous or compound heterozygous GBA mutations causing Gaucher's disease [77,78]. Why variants or mutations produce these diff erent neurological phenotypes in the heterozygous versus the compound heterozygous or homozygous states is quite enigmatic, but certainly an area worthy of further study.

Conclusions
With the spotlight fi rmly placed on TREM2's role in AD, research advances will likely be quite rapid, and the emerging data will likely enable a more unifi ed understanding of the function of innate immune signaling in AD. Although the possibility of harnessing TREM2 for therapeutic benefi t is tempting, until we learn more about the functionality and regulation of this protein in the brain, it is challenging to envision how one would target TREM2 in AD. More generally, future genomic studies aimed at identifying rare functional variants in i) innate immune loci already associated with AD through genome-wide association studies or ii) innate immune genes shown to modulate relevant diseaseassociated pathology could provide a steady stream of new rare functional variants in innate immune genes that might impact AD risk. Indeed, it is important to note that TREM2 never reached genome-wide signifi cance in the published genome-wide association studies.
us, combining biological inference with whole exome or whole genome sequencing strategies is likely to yield a treasure chest of novel genetic variants in innate immune signaling factors that infl uence risk for AD. Hopefully, these will not only tell us more about AD pathogenesis, but will also reveal tractable therapeutic targets.

Competing interests
The authors declare that they have no competing interests.