Post-mortem analysis of neuroinflammatory changes in human Alzheimer’s disease

Since the genome-wide association studies in Alzheimer’s disease have highlighted inflammation as a driver of the disease rather than a consequence of the ongoing neurodegeneration, numerous studies have been performed to identify specific immune profiles associated with healthy, ageing, or diseased brain. However, these studies have been performed mainly in in vitro or animal models, which recapitulate only some aspects of the pathophysiology of human Alzheimer’s disease. In this review, we discuss the availability of human post-mortem tissue through brain banks, the limitations associated with its use, the technical tools available, and the neuroimmune aspects to explore in order to validate in the human brain the experimental observations arising from animal models.


Introduction
The concept of research on human samples has been pioneered by neuropathologists who had an interest in research and who started to archive brain specimens referred for diagnosis, such as the one developed by John Corsellis in 1950 [1]. Development of novel technologies to detect biological molecules and increased collaboration with scientists highlights the need for the use of human samples, especially in the neurosciences. Indeed, one of the major limitations to our knowledge of human neurological diseases resides partly in the limits inherent to animal models, which mimic some aspects of the human neurological disorder without reproducing its complexity arising from both genetic and environmental factors. For example, more than 50 different animal models have been generated to explore Alzheimer's disease (AD) [2] and more than 20 models are available for the study of schizophrenia [3] without clear consensus about the similarities with human disease. The underuse of postmortem human brain tissue also impedes the deeper understanding of the pathophysiological processes ongoing in the diseased brain [3].
Therefore, in the mid-20th century, the notion of brain banking to archive, collect, and use human brain samples became essential with the aim to facilitate access to the tissue, to simplify the administrative burden for the researcher, and to improve their quality for cutting-edge research on neurological diseases. In this review, we discuss the pros and cons related to the use of human tissue, the parameters susceptible to influence the neuroinflammatory changes, and how to analyse them in AD.
Requirements and limitations to the use of postmortem brain tissue Ethics Networks of brain tissue banks have been created to allow request of tissue through a unique portal such as the consortium BrainNet Europe in 2001 under the European Commission or the UK Network of Brain Tissue Banks in 2009 by the Medical Research Council. In the UK, the use of human tissue is regulated by the Human Tissue Authority (HTA) and brain banks are licensed to operate as research tissue banks by the HTA under ethical approval provided by an ethics committee. This implies that the use of human tissue for a specific project is subject to approval by the brain bank committee. This is a compulsory step that could hinder the researcher and thus might appear as a limitation. However, under the approval of the brain bank, the study is ethically covered by the bank, saving administrative burden to the researcher and thus optimising the time spent on the project. This procedure is not restricted to the UK; the same principle applies worldwide [4]. Information about US brain banks is found under the platform National Institutes of Health Neuro-BioBank. The Australian Brain Bank Network, in addition to provide tissue for research, offers a neuropathology diagnostic service and education and training opportunities. Asia has started to set up brain banks in Japan, India, and China. Two brain banks are also present in South America: one in Brazil (the Brain Bank of the Brazilian Aging Brain Study Group, Sao Paulo) and one in Argentina.

Quality of the tissue
Quality of the tissue during its acquisition and long-term preservation is the principal objective of the bank. Different parameters may impact on the integrity of the tissue and thus on any biological molecules. Therefore, it is important to take into account these factors in the analysis of inflammatory events. These include age (the immune profile is known to evolve with ageing [5]), gender, genetic heterogeneity, agonal status (defined as the deep infrequent respiration in the final moments before death), preterminal medication, cause of death, concomitant disease, post-mortem interval [6], and time in the fixative. In addition, the potential role of systemic inflammatory diseases and infections may influence the cerebral inflammatory status [7]. Clinical studies have reported an accelerated deterioration of the cognition in the AD patients affected by systemic infections [8,9], and experimental studies demonstrated that systemic infection can switch the brain inflammation to a more aggressive phenotype, resulting in increased neurodegeneration [10,11].
It is usual practise in most of the prospective brain banks to fix one hemisphere and freeze the other hemisphere when collecting tissue. The preparation of the fixed tissue is a routine procedure used for diagnosis in any pathology department and thus a standardised protocol among the brain banks. Fixed tissue is used for histological staining and protein detection by immunohistochemistry, and the main difficulty is to achieve optimal specificity of the staining. Indeed, optimisation of the staining might require testing antibodies from different companies, using several pre-treatments (that is, antigen retrieval), and importantly ensuring that the detected staining is specific. This could be tested by using samples from another organ that expresses the protein of interest (such as the lymph node for any macrophage/ microglia proteins) or by a blocking experiment to test the specificity of the primary antibody or by doing both [12]. The preparation of fresh frozen tissue is more challenging as genetic material is susceptible to degradation under the factors described above and sometimes is not available from the archives of tissue banks. In addition, the development of genomic and proteomic technologies has highlighted the importance of obtaining high-quality RNA. If available, brain tissue pH and RIN (RNA integrity number) value are proxy indices of agonal status [13] and thus good indicators of the quality of the frozen tissue as far as the preservation of RNA and proteins is concerned [14]. Proteins are known to be more resistant as they can still be detected in tissue even when RNA is degraded [14]. Overall, standardized protocols have been developed and optimized by brain banks to minimize the impacts of preand post-mortem conditions.
Information required for the analysis of the immune response in Alzheimer's disease To ensure that the detection of the inflammatory profile is optimal in the human samples, the study should be adequately designed to overcome or minimize the impact of the factors discussed above. This can be achieved by the following: Age and gender: to match selected cases. Genetic heterogeneity: to determine the sample size necessary to reach statistical power [15]. Agonal status: to obtain pH and RIN values. The RIN algorithm has been designed to provide unambiguous assessment of RNA integrity based on a numbering system from 1 to 10, with 1 being the most degraded profile and 10 being the most intact [13]. The choice of the RIN value will depend on the technique to be used, as microarray analysis will need a higher RIN value than RT-PCR [14]. Pre-terminal medication or concomitant disease or both: to get access to the post-mortem report and to control the quantitative analysis for these confounding factors. Cause of death: to refer to the death certificate. Post-mortem interval: to select cases with the shortest interval and usually no more than 72 hours. Time in the fixative: to ensure that sections are provided from blocks taken at the post-mortem examination after a standardised time in the fixative.
Overall, the experimental group should be matched as closely as possible with the controls and the case selection based on the question investigated. The quantitative findings should be controlled for the influences of any preor post-mortem factors or both. Access to detailed clinical records is becoming an important point for the analysis of any quantitative assessment, information which might not be available or which might be incomplete because of the age of the case.
Other factors to account for the analysis are the known genetic risk factors. The polymorphism of the apolipoprotein E (APOE) gene is the major risk factor for sporadic AD [16], a fact recently reiterated in a number of genomewide association studies (GWASs) [17,18]. In the context of inflammation in AD, this is an important point to consider in the constitution of the cohort to study as APOE genotype has been associated with microglial activation both in the degree of microglial activation in AD brains [19] and in the microglial expression of inflammatory molecules [20].
How to analyse the neuroinflammatory response As mentioned above, different types of tissue are available from brain banks, and the most common are fixed paraffinembedded and frozen tissue. Fixed paraffin-embedded tissue will be useful to perform immunohistochemical detection of markers of interest. Brain sections could be used at a range of thickness (4 to 30 μm), depending on the experimental needs. The use of thick sections (20 to 30 μm) combined with free-floating immunohistochemistry is highly encouraged (sometimes available if fixed tissue is not paraffin-embedded), as it facilitates the permeation of the antibodies and provides a more effective removal of background staining. Glass-mounted thin sections (4 to 10 μm) can also give good results provided that appropriate protocol optimization is performed. Neuroinflammation can be analyzed in fixed tissue in different but complementary approaches: the qualitative or quantitative assessment or both. Qualitative assessment relies on description according to set criteria and thus can be interpreted as a subjective assessment. Qualitative assessment to be obtained on tissue is (i) the presence or absence of the marker of interest, (ii) the type of cell or feature recognized by the marker, and (iii) the cell morphology (for example, amoeboid, ramified, or dystrophic). However, it is now recognized that morphology is not sufficient to reflect the multitude of functions or activation states expressed by microglia [21]. Therefore, the quantitative approach is essential to obtain an objective measurement of the different markers studied. Quantification relies on sampling and statistical analysis based on numerical data collected. A semi-quantitative analysis can be performed on the basis of a rating system such as a scale of severity/intensity of the marker of interest, defined according to the pattern of immunostaining (for example, 0, 1+, 2+, and 3+) and usually assessed blindly by at least two researchers. Quantitative assessment can be obtained as (i) the number of positive cells per field or per area/ volume unit, (ii) the protein load defined as the percentage of the immunostained area of region sampled, and thus (iii) the protein load per cell. For example, Iba1 (ionized calcium-binding adaptor molecule 1) is currently acknowledged as being expressed independently of microglial functional state [22], and its expression is increased during neuroinflammation. Detection of Iba1 is widely used in animal studies, and Iba1 is the reference marker for microglial assessment in the human brain [23]. The numerical data collected are important for statistical power, and collection can be achieved in different ways: (i) by having sufficient cases in each group, (ii) by assessing several brain areas if there are not enough cases, (iii) by collecting enough individual data within each case, or (iv) by doing a combination of these.
The use of frozen tissue will allow the study of gene expression (for example, RT-PCR and microarrays) and protein expression (such as multiplex assays and Western blotting). The RNA or protein isolation methods should match the requirements of the techniques to be applied, and sample size, RNA species, and purity are the main variables to take into account. Owing to the intrinsic value of the human samples, it is highly encouraged when analyzing RNA expression to use isolation kits allowing the purification of all species of RNA, including microRNA, which will allow the multiple analysis. As mentioned previously, quality of frozen tissue is one challenge of post-mortem brain; thus, in addition to the parameters described above, adequate experimental controls are essential for the data analysis.
Microglia, as the brain macrophages, have the property to express a range of inflammatory markers [21,[24][25][26][27][28]. If the considerations described above are taken into account, functional immune changes can be assessed in the human brain by using specific markers as detailed below.

Inflammatory profile
Despite a broad interest in the inflammatory response in AD and the extensive research in this disease, the scientific community has failed to shed clear and uniform light on the contribution of local inflammation to the disease [29,30]. The neuropathology of AD shows a robust innate immune response characterized by the presence of activated microglia, with increased or de novo expression of diverse macrophage antigens [21], and at least in some cases production of inflammatory cytokines [29,31]. It has been suggested that non-steroidal anti-inflammatory drugs protect from the onset or progression of AD [32], suggesting that inflammation is a causal component of the disease rather than a consequence of the neurodegeneration. Recent GWASs have highlighted several genes involved in innate immunity, indicating also a causal role for inflammation in the disease [33]. Additionally, a solid body of evidence shows that systemic inflammation may interact with the innate immune response in the brain to act as a 'driver' of disease progression and exacerbate symptoms [7]. The impact of systemic inflammation on the progression of AD means that any neuropathology study on the inflammatory response in the AD brain must take into account systemic co-morbidities that may influence the microglia phenotype (see 'Information required for the analysis of the immune response in Alzheimer's disease' section).
The definition of the brain inflammatory profile of AD shows conflicting ideas in the literature, probably arising from the heterogeneity of the post-mortem samples and the difficult application of the detection methods [21]. AD has been associated with a pro-inflammatory phenotype, characterized by expression of interleukin-1 beta (IL-1β) and complement proteins [34,35]. The upregulation of genes linked to an anti-inflammatory phenotype, arginase 1, or the transforming growth factor-beta (TGF-β) has also been reported in association with AD [36,37]. The consensus defines that, in the human AD brain, the inflammatory response cannot be classified as strictly M1-like or M2-like [38] and that the changes in expression level are compounded by the various detection methods (for review, see [29]).
Although the precise inflammatory phenotype of microglia in AD seems elusive, the link of AD with inflammation seems clear, as highlighted by a recent study using microarray technology on the gene signature of ageing and AD [39]. These ideas support the model of an activation of the innate inflammatory response in microglia as a prelude to the development of AD [39]. Furthermore, studies on incipient AD samples show a strong correlation of genes associated with the microglial response and the progression into AD [40]. The concept of the interconnection of AD and the innate immune response is supported by evidence from a GWAS implicating genes involved in innate immunity [41]. These promising studies are opening new avenues into the understanding of the impact of the innate immune response in AD while supporting the need for future exploration.
Characterising the inflammatory response in human post-mortem AD samples by using reliable and consistent methods will provide valuable information in the field. It could be agreed that analysing the expression of inflammatory mediators at the protein level, rather than the analysis of the mRNA expression, is highly desirable. To accomplish this task, the market offers a number of multiplex systems to analyse several molecules simultaneously, accelerating research and minimising costs. It is highly encouraged to analyse a broad range of inflammatory mediators instead of using a limited number of molecules as a proxy. New technical progress aimed at increasing the panel of molecules to be analysed, as well as the detection levels, will provide a valuable approach to be able to trace comparisons like those recently used to define the microglial gene signature in mice [42,43].

Phagocytosis
The phagocytic ability of microglia is a feature shared with peripheral macrophages, helping to eliminate bacterial, necrotic, or apoptotic cells during development or disease. In AD, the amyloid plaque burden increases with age in both mouse models [44] and human patients [45], indicating the rather ineffective phagocytic potential of microglia. Amyloid-beta (Aβ) deposits have been shown to have a potent chemoattractant activity on microglia, although their removal by phagocytosis has not been clearly evidenced in vivo [46]. However, it has been shown that the removal of Aβ can be improved by further challenge of microglia with high doses of lipopolysaccharide [47] or the induction of IL-1β [48]. In human AD, active immunotherapy directed against Aβ has been successful in Aβ removal, partly by redirecting the microglia toward Aβ [23] and by increasing their phagocytic activity [49]. Recent evidence supports a differential contribution of perivascular macrophages and parenchymal microglia, not bone marrow-derived cells, to the clearance of Aβ [50]. In this study, the authors used mice deficient in CCR2, a molecule expressed by monocytes defining their migration, to rule out the contribution of circulating monocytes, further evidencing a prominent role of the perivascular macrophage population to Aβ clearance [50]. Interestingly, as microglia do not express CCR2 in healthy and diseased conditions [51,52], the analysis of this molecule in comparison with other markers expressed by microglia (that is, CX3CR1) can help to potentially differentiate the infiltrated monocytes/macrophages from the resident microglia. This comparison has not been performed yet in the human brain and would provide valuable insights for the understanding of the balance of microglia/macrophages.
The regulation of the phagocytic activity of microglia appears as a key genetic determinant of AD pathology. Recent studies link genetic variants of TREM2, a protein regulating the activation and phagocytic functions of myeloid cells, with the risk of developing AD [53,54]. TREM2 has a balancing role between phagocytic and pro-inflammatory microglial activities and is expressed in microglia around plaques in an experimental model of AD [55]. Similarly, dysregulation of the complement system in humans has been associated with AD [18]. However, no clear consensus defines the overall level of microglial phagocytosis in the human AD brain. The use of refined experimental approaches to directly study microglial phagocytosis [56], together with the analysis of immunological markers such as CD68 (related with phagocytic activity), will shed light on the understanding of the phagocytic activity of microglia and other macrophage populations in the AD brain.

Proliferation
Microglial activation in neurodegeneration is accompanied by an increase in their numbers. The contribution of circulating progenitors to the microglial population is minor, or even absent, as shown in a mouse model of AD [50], pointing to in situ microglial proliferation as the mechanism regulating microglial turnover [57]. In mice, microglia are maintained and function largely independently of circulating progenitors in health [58,59] and disease [50,52,60]. Therefore, the analysis of microglial proliferation in AD is necessary for understanding how the innate inflammatory response contributes to disease onset or progression or both.
Proliferation was assumed to be responsible for the increased number of microglial cells observed in AD samples, although direct evidence of proliferating microglial cells (that is, Ki67 expression in Iba1 + cells) was reported only recently [60]. The expansion of the microglial population has been consistently documented in transgenic mouse models of AD, mainly accumulating around plaques [61]. However, direct evidence of microglial proliferation (incorporation of 5-bromo-2-deoxyuridine in Iba1 + cells) was only recently reported, suggesting a direct effect of the plaque microenvironment over the regulation of microglial proliferation [62]. These studies pinpoint the importance of the control of microglial proliferation during AD. Establishing reproducible and consistent methods to monitor microglial proliferation in post-mortem AD brains will provide the scientific community with valuable tools to better compare results across cohorts of patients, contributing to our better understanding of the pathophysiology of AD [63].
The analysis of microglial proliferation is best achieved by double/triple immunohistochemical analysis by using either fluorescence or bright-field microscopy [63] (Figure 1). The use of fluorescence-based techniques needs to be supplemented by the use of a fluorescence-quenching step (for example, Sudan Black). This step is particularly important in the case of AD human tissue, as the occurrence of autofluorescent artefacts (for example, lipofuscin granules) is very frequent and can compound the interpretation of results. Double bright-field immunohistochemistry can be achieved by combining DAB and alkaline phosphatase reactions, labelling two individual antibodies with a brown or blue precipitate, respectively. Both fluorescent and bright-field microscopy methods need to implement a membrane or cytoplasmic microglial marker (Iba1, CD68, and CD11b) and a nuclear proliferation marker (Ki67, phospho-histone H3, and PCNA), together with nuclear counterstaining to decipher the subcellular localization of the proliferation markers. The analysis of double-or triple-staining techniques needs to be coupled to colour deconvolution methods.

Cell infiltration
Cell infiltration in the brain during AD is an important question related to the potential roles for recruited monocytes/macrophages and T cells within the brain parenchyma. Infiltration of peripheral leukocytes in the human AD brain is very limited when compared with classic autoimmune diseases like multiple sclerosis [64]. However, the rare coincidence of stroke and AD leads to an increase in infiltrating macrophages in the brain, which contained Aβ fibrils suggesting an effective plaque clearance response [65]. Although limited evidence is available in the literature regarding the existence and role of infiltrated leukocytes in human AD, these studies suggest that systemic co-morbidities could determine the degree of infiltration of circulating leukocytes. It should be noted that the findings on analysis of post-mortem tissue several years following any event which might impact on neuroinflammation (insult, trauma, disease, treatment) may not Figure 1 Microglial proliferation in post-mortem human Alzheimer's disease brain. Representative images of the detection of Ki67 in microglial cells (Iba1 + ) by bright-field double immunohistochemistry (A) (DAB, brown, Iba1; AP, blue, Ki67) or double immunofluorescence (B) (Alexa 568, red, Iba1; Alexa 488, green, Ki67) from the temporal cortex of a patient with Alzheimer's disease. Images adapted from [63]. Iba1, ionized calcium-binding adaptor molecule 1.
necessarily reflect those immediately after the event, and instead represent the later effects.
As explained before (in the 'Proliferation' section), experiments in mouse models of AD suggest that the infiltration of circulating monocytes is scarce and does not contribute to the pool of parenchymal microglia [52]. Translating these findings to the human situation is a challenging task because of the lack of specific markers to distinguish microglia from peripheral leukocytes, although the analysis of the levels of molecules like CD45, Ly6C, or CD11c could help to split the contribution from both populations [66].
The study of the adaptive immune response in AD has provided valuable information in the last few years. T-cell number-both the CD4 + (T helper) and the CD8 + (T cytotoxic/suppressor) populations-increases in patients with AD [67]. Although the number of T cells is higher in AD brains, they do not express markers of proliferation, indicating an absence of antigen-triggered clonal expansion [67]. However, there is evidence indicating the orchestration of a systemic T-cell response, as suggested by the presence of the RO isoform of CD45 in peripheral T cells in patients with AD, linked to T-cell memory [68], and by the increase in the CD4 + and CD25 + regulatory subsets in patients with AD [69]. However, the role of the T cells infiltrated in the brains of patients with AD is unclear. Major histocompatibility complex class II is found upregulated in microglia surrounding Aβ plaques in the AD brain, indicating possible antigen presentation [30]. However, the co-stimulatory factors CD80 and CD86 are required for the induction of primary adaptive immune responses and their description in microglia in AD remains elusive [70]. Infiltration of T lymphocytes has been associated with the development of side effects in a number of AD patients immunised against Aβ [71,72]. However, analysis of T cells in a cohort of immunised AD cases compared with unimmunised AD cases did not show a difference [23]. Therefore, a detailed analysis of the T cell-dependent responses in the brain parenchyma is required to fully understand the role of these cells in AD pathology. The use of fresh samples, allowing flow cytometry analysis and cell culture [73], would provide the optimal tool to overcome the limitations of using post-mortem tissue for this purpose.

Conclusions
A number of recent publications have highlighted specific inflammatory profiles associated with healthy or diseased brain [38,39]. However, our current knowledge of the neuroinflammatory response in AD is based mainly on in vitro and animal studies. Therefore, it is essential to confirm or contradict the experimental findings in the human brain in order to increase our knowledge of the pathogenic mechanisms of AD. This strategy would lead to the identification of potential therapeutic targets without undermining the benefit of animal models. The recent development of brain banks with the aim of providing good-quality tissue for research, in association with the tools now available to identify genes and proteins (Table 1), should soon increase our understanding of the role of immunity in neurodegenerative diseases.