Neuroinflammation in Preclinical Alzheimer's Disease: A Review of Current Evidence

The pathology of sporadic Alzheimer’s disease (AD) may be present at mid-life and precede the prodromal and clinical dementia syndromes associated with the disorder by decades. Few successful therapeutic treatments exist and, as a result, attention is turning to the preclinical stages of the disease for the development of future intervention strategies. The success of such strategies will rely on well-defined biomarkers of preclinical disease to identify and monitor changes earlier in the disease course. Here, we consider whether immune function changes are potentially useful markers of preclinical disease. We have selected studies spanning epidemiological, animal, clinical and imaging research pertaining to the earliest stages of AD pathogenesis, as well as studies of non-demented adults at high AD risk. We examine changes in inflammatory markers, alongside changes in established biomarkers, to highlight their suitability as disease indicators across preclinical and prodromal stages. We conclude that further work surrounding this topic is required, calling for larger prospective epidemiological studies of preclinical disease that incorporate serial assessment designs with a wider range of inflammatory mediators. We anticipate that future benefits of work in this area include improved disease detection and modification, as well as diagnostic accuracy of trial participants, leading to more cost-effective observation and intervention studies.


Introduction
Increasing evidence suggests sporadic Alzheimer's disease to be a clinically silent disease of mid-life, which remains undetected for decades until its terminal stage, characterized by dementia [1][2][3][4]. In light of multiple failures to develop an effective treatment for prodromal or clinical (dementia) AD, research efforts are now focusing on earlier stages of the disease process and clinical services are moving towards a similar model [5]. The development of therapeutic intervention strategies targeting the pre-clinical stages of the disorder currently relies, however, on our ability to identify biomarkers, which may be used to monitor change at this early stage. Within this context the focus is currently on neurological markers, notably neurodegeneration, amyloid and tau accumulation and pre-clinical cognitive markers [6,7]. Immune system changes on the other hand, although extensively studied in relation to prodromal AD and AD dementia, have to date been relatively neglected.
While considerable attention was given to early evidence of neuroinflammation in AD dementia, it fell out of favour as the amyloid cascade hypothesis [8] became the predominant aetiological model. In the last decade there has been a resurgence of interest in the potential role that neuroinflammation may play not only as pathological consequence of Aβ pathology, but also as a potential trigger and accelerator of amyloid plaque formation and tau-related neuronal injury [9][10][11]. Inflammatory reactions have been implicated in the pathogenesis of AD dementia in the form of glial cells, such as astrocytes and microglia, as well as inflammatory components, such as cytokines and chemokines [11,12]. Additional support comes from associations of AD dementia with genes implicated in immune responses, such as TREM2, CR1 and CD33 and chronic psychological stress [13][14][15]. An inflammatory contribution to disease pathogenies has also been observed in the prodromal stages of AD [16]. Given accumulating biological evidence that neuroinflammation may precede both Aβ and tau formation [11], it is of considerable interest as a potential pre-clinical marker. raised levels of cytokine TNF-α, have been found to correspond to an acceleration of cognitive decline in AD dementia patients [26], but the nature of inflammation's impact in the pre-clinical period has not been specifically examined within epidemiological studies.
A retrospective general population study of persons diagnosed with AD dementia observed that C-reactive Protein (CRP) levels were not significantly different ten years previously in individuals remaining cognitively healthy (i.e., controls) and those evolving towards AD dementia [4]. However, against expectation, levels of CRP in the AD dementia cases showed a significant downward slope closer to the time of diagnosis, while the control group showed a slightly raised curve towards the same time-point similar to the sigmoidal trajectories observed by Jack et al in relation to other biomarkers, notably amyloid and tau [27][28][29]. The study was limited by its restriction to a single inflammatory marker assessed at only one point in time and also limited by the older age of the population (over 65 at recruitment). The value of future epidemiological studies of inflammatory markers in the pre-clinical period will very much depend on their ability to examine a wider range of markers across time from mid-life, when exposure to many of the principal inflammation-related disorders also implicated in AD risk (notably obesity, diabetes and stroke) are currently postulated to have their principal impact.

Animal Studies
The occurrence of plaque-dependent inflammation has been robustly observed in animal models of AD [9,30], with microglia present in neutritic plaques [31,32] that are spatially related to dendritic spine loss [33]. Furthermore there is some evidence to suggest that a pro-inflammatory process may be initiated before plaque deposition. In transgenic mice overexpressing amyloid precursor protein, a microglial activation processes involving Il-1β and Il-6 was detected early at 3 months of age when amyloid plaques were not yet present [34]. Similarly, using a triple transgenic mice model (3xTg), Janelsins et al. [35] observed increased expressions of TNF-α and monocyte chemoattractant protein-1 (MCP-1) in the entorhinal cortex of three-month-old mice which coincided with the production and accumulation of intracellular amyloid but preceded extracellular plaque deposition, the latter of which occurred only at 12 months of age [35]. Another study found that exposure of a viral mimic (polyriboinosinicpolyribocytidilic acid) to wild-type and transgenic mice prenatally and again in adulthood corresponded to an emergence of AD-like pathology (amyloid and tau aggregates), microglia activation and reactive gliosis over the course of animal ageing [36].
Taken together, these studies suggest that pro-inflammatory processes may act as very early indicators of preclinical disease. However, their relevance to pre-clinical AD is limited by the difficulties inherent in defining 'pre-clinical' in animal models. This can only currently be estimated by reference to progression of amyloid and tau deposition. Although human and rodent amyloid plaques share morphological similarities, they are distinct in biochemical composition [37] and might interact differently with other pathological features, such as inflammatory agents. Moreover, animal models typically involve the manipulation of genetic markers associated with familial forms of AD and such models might not correspond to the relatively more prevalent sporadic form associated with humans. Therefore, while animal studies provide limited evidence of inflammation as an upstream marker of disease, prospective studies are required to monitor changes before and during the evolution of neurological changes within the human brain.

Clinical and Imaging Studies
Research on the role of brain inflammation in the evolution of the AD began in the mid-1980s [38]. Subsequently, the inflammatory profiles of prodromal AD and AD dementia are relatively well described but present a heterogenous picture; possibly as a result of discrepant methodological practices and patient characterisation between studies [39]. Inflammation has primarily been studied in the context of advanced AD pathology involving amyloid plaques. Therefore, the involvement of elevated inflammatory agents in evolution from prodromal changes to dementia remains unclear, with inconsistent evidence for a specific risk for AD dementia [40]. Relatively few studies have assessed inflammatory profiles in preclinical AD, so it is even less clear if or how neuroinflammation promotes AD-related cognitive and brain changes in the direction of prodromal and clinical AD. In addition, the initiation of the acute inflammatory response is counterbalanced by an active resolution and specialized pro-resolving mediators (SPMs) have been identified that drive resolution by diminishing inflammatory molecules, such as cytokines [41]. Thus it has been suggested that agerelated deficits in resolution of inflammatory responses may contribute to the development of late onset AD dementia [42], including in humans [43]. Finally, while, preclinical evidence suggests that inflammation can induce tau hyperphosphorylation, a better understanding of whether this increases neurofibrillary tangles is needed [44].
Given that the preclinical phase of AD is, by definition, not observable in terms of everyday functioning, research participants with preclinical AD are likely to be included in studies of "healthy" or "normal" cohorts. Elevated levels of pro-inflammatory cytokines and markers, such as IL-1β, IL-6, TNF-α and CRP have been observed in some but not all persons over time and attributed to age [45]. It is, however, possible that such persons are manifesting pre-clinical AD, a proposition supported by studies of midlife and older normal adults showing elevated levels of inflammatory markers in association with cognitive deficits and brain changes ( Table 1).
In healthy adults, elevated YKL-40 concentrations have been associated with poorer Mini-Mental State Exam (MMSE) performance [46,47] and PAI-1 has been negatively correlated with motor speed and coordination [48]. Serum CRP levels have been negatively associated with performance on executive function tasks [49], while composites of CRP with other inflammatory markers have correlated with visuospatial function (CRP+TNF-α [50]; CRP+IL-6 [51]), verbal proficiency and short-term memory (CRP+IL-6 [51]). Midlife inflammation levels may contribute to later-life cognitive performance, as one study found that inflammation composite scores created from immunoassays ascertained at mid-life (45-65 years old) were associated with reduced episodic performance 24 years later [52].
These relationships between raised inflammatory levels and cognitive scores are not always demonstrated [53,54], possibly due to heterogeneous cognitive ability within "healthy" aged samples. For example, in one study, participants' levels of inflammatory markers were classified into tertiles of high, middle and low values. Cognitive scores were adjusted for age, education-level and gender. At baseline, the highest tertiles of ACT and lowest tertiles of albumin were associated with delayed memory recall and MMSE performances, respectively. However, excluding participants with MMSE scores<21 at baseline, revealed a further linear relationship between ACT and information processing speed, while the association between MMSE performance and albumin disappeared. At follow-up both ACT and albumin tertiles were associated with decline in MMSE performance. Only the association between ACT and MMSE performance remained when participants with MMSE<21 at baseline were excluded from the longitudinal analyses [54]. These findings imply the possible sensitivity of different inflammatory markers to cognitive impairment at different stages along dementia continuums and underscore the importance of sub-group analyses in cognitively heterogeneous samples.
Where these relationships between cognitive performance and inflammation are present, they may be partially mediated by inflammation's influence on brain morphology [50,51]. Smaller hippocampal volumes have been found for individuals with higher levels of STNFR-1, STNFR-2; IL-6 and CRP [51,55,56]; these effects being strongest for individuals between 60-70 years [55]. In both middle and older healthy aged adults, higher CRP has been related to reduced global and regional fractional anisotropy in the frontal and temporal lobes, the cortico-subcortical tracts, and corpus callosum [48,49]. Higher IL-6 has been associated with smaller total brain volume, total and regional grey and white matter volumes as well as with reduced cortical thickness of the inferior occipital and temporal gyri [56][57][58]. The influence of prolonged inflammation on brain morphology is not yet clear, with one recent study finding that the presence of baseline inflammatory biomarkers did not modify the change of brain measures over time [59], but earlier reports indicating greater cortical thinning and white matter volume reduction at follow-up [56,57]. Age and race may mediate the relationship between inflammation and later-life brain integrity as younger and white participants with higher levels of systemic inflammation during midlife were more likely to show reduced brain volumes 24 years later [52].
Some studies of inflammation in healthy adult samples incorporate the assessment of established AD risk factors alongside immunoassay analyses but few studies have compared inflammatory mediators across the stages of preclinical dementia as defined by accepted research criteria [6]. These criteria propose a spectrum of stages pertaining to biomarker load within preclinical AD: stage 1 individuals show evidence of amyloid deposition only; stage 2 individuals possess both amyloid and neurodegeneration; while stage 3 individuals show subtle cognitive symptoms in addition to these biomarkers. In one study of cognitively normal individuals [60], those participants qualifying for stages 2 and 3 as well as those with suspected non-Alzheimer's pathophysiology (SNAP) showed higher levels of CSF YKL-40 than those in stages 1 and those without amyloid evidence. At the cohort level, YKL-40 concentrations correlated with t-tau and p-tau deposition, in keeping with other research [46,61,62]. This correlation remained significant when amyloid positive and amyloid negative participants were analysed separately, emphasising a possible role for YKL-40 in tau hyperphosphorylation and neuronal injury, which may be at least partially independent from amyloid plaque deposition. The overlap of elevated YKL-40 levels across preclinical and SNAP participants further suggests that neuroinflammation may emerge independent of amyloidios. More recently, an assessment with a broader range of inflammatory markers again found no correlation of any markers with AB41-42 levels. Instead, six CSF markers (IL-15, MCP-1, sFLT-1, siCAM-1 and sVCAM-1) were associated with tau pathology independent of age, gender, cognitive status and APOE-E4 status [63]. Alternatively, certain inflammatory agents may interact with amyloid and tau pathologies to alter brain structure in preclinical AD. In a cohort of mid and later life adults at risk for AD, higher MCP-1 in combination with lower CSF AB42 levels was associated with measures of neuronal injury in the bilateral frontal cortex and lateral temporal lobe, while higher MCP-1 in combination with higher CSF p-tau was related to altered microstructure in the precuneus. Elevated YKL-40 was directly associated with CSF levels of neurofilament light chain protein and t-tau, markers of neuronal injury [62]. Similarly, YKL-40 was associated with a change in global cognition over the course of two years in older adults, but only for those who were positive for amyloid pathology [64]. Previously, YKL40 had been associated with reduced cortical thickness in the middle and inferior temporal areas in a cross-sectional research, but this association was again only observed in participants with low CSF Aß42 [65]. In both studies, CSF YKL-40 and AB42 levels did not correlate in the positive AB42 group, suggesting that amyloid build-up and neuroinflammation may underlie distinct processes but may have additive effects on cognition and brain structure.

Marker
Main Findings

YKL-40
Levels are significantly increased in preclinical and prodromal AD groups [46]. Levels are higher in preclinical stages 2-3 and SNAP stage compared to Stage 0, 1 [60]. Levels significantly correlate with t-tau and p-tau levels in preclinical AD [46], middle and older adults [62]. Levels correlated with MMSE and MAT performance for combined preclinical and prodromal AD groups [46] as well as change scores for global cognition in older adults [64]. Levels correlates with cortical thinning in middle and inferior temporal areas but only in individuals with low CSF Aß42 [65].

CRP
An association with cognitive performance is not always demonstrated [53,54]. Nonetheless, higher levels are associated with worse performance in executive function [49], while individuals within the highest tertiles of CRP show significantly lower MMSE scores overtime [47]. CRP levels inversely related to white matter integrity in corticosubcortical pathways and association fibres of frontal and temporal lobes [48]; global and regional FA scores of the frontal lobes, the corona radiata and the corpus callosum [49] as well as regional gray matter volume in the posterior and lateral aspects of left temporal cortex [58].
PAI-1 Levels correlate with lower processing speed and motor coordination [48]. Higher levels associated with white matter integrity loss in corticosubcortical pathways and association fibres of frontal and temporal lobes [48].
ACT Highest tertile of ACT was associated with lower delayed recall scores cross-sectionally and greater decline on MMSE longitudinally [54].

IL-6
As part of a composite with CRP, has been negatively associated with tests of spatial processing, short-term memory, verbal proficiency, verbal learning and memory, executive function [51] as well as episodic performance [52]. Higher levels associated with reduced cortical thickness in inferior occipital gyrus and sulcus, inferior temporal gyrus (both hemispheres) crossectionally. Higher levels associated with cortical thinning in: frontopolar gyri and sulci; right subcentral gyrus and sulci; inferior temporal poles (bilatreal) and left occipital pole and right calcarine sulcus cortex overtime [57].
TNF-a As part of composite with CRP, has been associated with correlated with visual spatial ability. DA in the body and isthmus of the corpus callosum was shown to mediate this association [50].

Composites
Composite (fibrinogen, albumin, white blood cell count, von Willebrand factor, Factor VIII,) was negatively associated with episodic memory and was associated with greater ventricular degeneration, smaller hippocampal and occipital volumes [52].

Discussion and Conclusion
The evidence considered in this review largely suggests that inflammation is present in preclinical AD and is associated with AD pathogenesis. Some findings from animal and clinical studies propose that inflammatory processes might precede or be independent from amyloid deposition, suggesting that these markers constitute the earliest indicators of preclinical disease. Some studies demonstrate direct associations with tau pathology, cognitive performance and brain changes; others identified interaction effects, where the presence of inflammation in combination with amyloid and/or tau pathology influenced neurodegeneration.
It is unclear whether inflammatory markers are associated with particular clinico-pathological characteristics that may underlie heterogeneous outcomes within aging and preclinical cohorts; for example, individuals who subsequently develop prodromal and clinical dementia compared to those who do not. Nonetheless, there is some evidence that inflammatory markers may possess diagnostic or prognostic value for future AD symptomology. Additional studies focusing on impaired resolution of inflammation, especially on tau related neurofibrillary pathology are needed, as are studies that examine whether such responses are dose-dependent and relevant to AD in humans.
Large epidemiological studies of preclinical AD are notably absent. The use of animal models, while undoubtedly facilitate knowledge of the various molecular mechanisms of AD, may not be the optional paradigms to simulate the preclinical stages of the disease. While it appears that elevated levels of inflammation are associated with cognitive processing deficits and morphometric changes in "healthy" adults, there is a distinct lack of preclinical AD samples available to investigate similar relationships in these individuals. The range of inflammatory markers investigated in preclinical AD studies is narrow and it is not yet possible to infer which agents, acting alone or in combination, are relevant in the proposed inflammatory process, and if they may support or prevent AD development at different stages of disease. The majority of these studies are cross-sectional, limiting inferences regarding causality, and temporal relationships between inflammation, cognitive performance and neurodegeneration. Where longitudinal designs have been adopted, few studies have captured inflammatory measures serially, using only baseline data to predict future cognitive and brain changes at followup. Single clinical assessments may over-or under-estimate the level of inflammation chronicity for the individual at baseline and precludes investigations of the impact of changes in inflammation markers, or combinations thereof, on neurodegenerative and cognitive outcomes. Further clinical and imaging studies with longitudinal data and the ascertainment of inflammatory agents alongside neuropathological and cognitive markers earlier in the disease course and at predefined times throughout disease duration is needed to the determine the contribution of inflammation within the cascade of pathological and cognitive changes associated with AD.
Interest in inflammation and AD is building and there are numerous benefits of developing this area of research within preclinical AD. Delineating the role and timing of neuroinflammation in this population will help to reconcile conceptual debates surrounding AD aetiology. Relatedly, including inflammatory markers alongside established biomarkers in future research studies could improve diagnostic accuracy of participant samples and help stratify individuals according to greatest risk for cognitive impairment, leading to smaller and more cost-effective observational and clinical trials that promote earlier disease detection and targeted therapeutic intervention. At present, the immediate clinical application of these markers for diagnostic purposes remains elusive since research studies are only beginning to consider their use for extending current definitions of preclinical AD. Future work will be able to assess the impact of utilising inflammatory markers on diagnostic decision-making and patient management in clinical contexts.