Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Iron loading is a prominent feature of activated microglia in Alzheimer’s disease patients

Iron loading is a prominent feature of activated microglia in Alzheimer’s disease patients Brain iron accumulation has been found to accelerate disease progression in amyloid‑β(Aβ) positive Alzheimer patients, though the mechanism is still unknown. Microglia have been identified as key players in the disease patho ‑ genesis, and are highly reactive cells responding to aberrations such as increased iron levels. Therefore, using histo‑ logical methods, multispectral immunofluorescence and an automated in‑house developed microglia segmentation and analysis pipeline, we studied the occurrence of iron‑accumulating microglia and the effect on its activation state in human Alzheimer brains. We identified a subset of microglia with increased expression of the iron storage pro ‑ tein ferritin light chain (FTL), together with increased Iba1 expression, decreased TMEM119 and P2RY12 expression. This activated microglia subset represented iron‑accumulating microglia and appeared morphologically dystrophic. + + Multispectral immunofluorescence allowed for spatial analysis of FTL Iba1 ‑microglia, which were found to be the + + predominant Aβ‑plaque infiltrating microglia. Finally, an increase of FTL Iba1 ‑microglia was seen in patients with high Aβ load and Tau load. These findings suggest iron to be taken up by microglia and to influence the functional phenotype of these cells, especially in conjunction with Aβ. Keywords: Alzheimer, Microglia, Iron, Ferritin, Human Introduction Not only can microglia modulate Alzheimer’s disease, Alzheimer’s disease is the most common cause of demen- but many transcriptomic studies showed microglia to undergo the most pronounced changes in response to tia, and is defined by the presence of amyloid-β (Aβ) pathology. In mice, a subset of responding microglia has plaques and tau tangles. In addition, the brain’s resident been found to lose their homeostatic molecular signature innate immune cells, microglia, have been found to be and transition into a so-called ‘disease-associated micro at the centre-stage of the disease, as most identified risk - genes are predominantly or even exclusively expressed in glia’ (DAM) state [3]. In humans, a comparable yet dispa- microglia [1, 2]. rate state coined the human Alzheimer microglia (HAM) has been identified [4]. Upregulated genes in these sub - sets do not only indicate loss of homeostatic function and increased pro-inflammatory activation, but also dysregu - *Correspondence: b.kenkhuis@lumc.nl lated iron-metabolism, manifested via upregulation of the Boyd Kenkhuis and Antonios Somarakis have contributed equally to this FTL-gene and downregulation of FTH1 and SLC2A11 [4, work 5]. FTL encodes the ferritin light chain (FTL) protein, the Department of Human Genetics, Leiden University Medical Center, Albinusdreef 2, 2333 ZA Leiden, The Netherlands component of the major iron-storage complex ferritin, Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 2 of 15 responsible for the long term storage of iron. These tran - Histology and immunohistochemistry scriptomic findings coincide with previously observed Formalin fixed paraffin embedded (FFPE) tissue was seri - ferritin microglia in Alzheimer’s disease [6, 7]. Though ally cut into ten 5-μm-thick and four 10-um-thick sec- increased iron concentration likely plays a role, the exact tions. Consecutive 10-μm-thick sections were used for link between the two has not yet been established. histological detection of iron using an enhanced Perl’s Iron accumulation, irrespective of microglial activa- stain and IHC detection of Ferritin Light Chain (FTL). tion, on the other hand, has been reported in disease- 5-μm-thick sections were used for staining of the micro- affected areas in Alzheimer’s disease, using both in-vivo glia multispectral immunofluorescence (mic-mIF) panel and post-mortem human MRI [8]. Several MRI and (Additional file  1: Table  S2) to verify expression of FTL histology studies found high correlations between iron in microglia/macrophages (Iba1), look at the activation accumulation and cortical Aβ and tau spreading [9–11]. state of these cells (P2RY12/TMEM119) and study the Clinically, increased iron concentrations were shown to interaction with Aβ-plaques. Finally, of three subjects, accelerate cognitive decline in Aβ-positive Alzheimer 20-μm-thick sections were obtained for 3D confocal patients, indicative of a disease-modifying role for iron imaging. Step-by-step histological and IHC optimiza- accumulation [12, 13]. Again, how iron accelerates cogni- tion protocols, together with the imaging parameters, are tive deterioration is poorly understood. reported in the Supplementary Methods. A step-by-step Therefore, in this study we aimed to research the possi - mIF protocol and further analysis of the described histo- ble link between iron accumulation and functionally acti- logical, IHC and mIF staining will be described in the fol- vated microglia, and finally, its relation with Aβ-plaques. lowing sections. We performed a comprehensive investigation of iron- accumulating microglia, and first identified that the iron- storage protein FTL, specifically reflected increased iron Microglia multispectral immunofluorescence (mic‑mIF) accumulation in microglia. Secondly, by using multispec- panel tral immunofluorescence and an in-house automated One 5-μm-thick section of each subject was stained with cell-analysis pipeline, we found F TL microglia to show the mic mIF panel with the following protocol, based significant activation, shown via both downregulation on a previously described protocol by IJsselsteijn et  al. of homeostatic markers TMEM119 and P2RY12 and [17]. Sections were deparaffinized with 3 × 5  min xylene, dystrophic morphology, and to predominantly infiltrate rinsed twice in 100% alcohol and subsequently washed Aβ-plaques. This provides evidence for iron dysregula - with 100% ethanol for 5  min. Endogenous peroxidases tion as a prominent feature of activated microglia in Alz- were blocked for 20  min in 0.3% H O /methanol, after 2 2 heimer’s disease in humans. which the slides were rinsed with 70% and 50% alcohol. Heat induced antigen-retrieval was performed by cook- ing the slides for 10  min in pre-heated citrate (10  mM, Methods pH = 6.0) buffer for 10  min. After cooking, excess buffer Tissue acquisition was removed and slides were cooled for 60  min. Non- Brain autopsy tissue of the middle temporal gyrus (MTG) specific antibody binding sites were blocked with block - of 12 Alzheimer patients and 9 age-matched controls ing buffer (0.1% BSA/PBS + 0.05% Tween) for 30  min. was collected at the Leiden University Medical Center Firstly, slides were incubated with anti-TMEM119 (1:250, (LUMC), Netherlands Brain Bank (NBB) and the Normal Sigma Aldrich) diluted in blocking buffer overnight at Aging Brain collection Amsterdam (NABCA). Patients RT. Slides were washed thrice with PBS and incubated were included based on clinical presentation and diag- with Poly-HRP secondary antibody for 30  min. Slides nosis was confirmed by a neuropathologist. The neuro - were washed again and incubated with the appropriate pathologists also evaluates Braak stage, based on Gallyas Opal tertiary antibody (1:100 in amplification diluent, and Tau immunohistochemistry (IHC), and Thal phase Perkin Elmer) for 60 min, which causes permanent bind- based on Congo Red and Amyloid Beta IHC, in eighteen ing of the fluorophore to the antigen site. All subsequent standard regions, according to the latest international steps are performed in the dark where possible. Finally, diagnostic criteria [14–16]. Patient demographics are the slides are placed back in citrate buffer and cooked in reported in Additional file  1: Table  S1. All material has the microwave for 15 min to wash the primary antibody been collected with written consent from the donors and off. The same steps are repeated for anti-P2RY12 (1:2500, the procedures have been approved by the Medical Ethi- Sigma Aldrich). After binding of the two antibodies cal committee of the LUMC and the Amsterdam UMC. amplified with Opal, slides are incubated with a primary antibody mix with anti-FTL (1:100, Abcam), anti-Aβ (17–24) (1:250, Biolegend) and anti-Iba1 (1:20, Millipore) Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 3 of 15 antibodies, diluted in blocking buffer, overnight at room Fig.  S2d). As a result, a novel segmentation algorithm temperature. The next day, after three washes with PBS, for this type of data was developed. slides are incubated with a secondary antibody mix of Identification of the entire cytoplasmic area of micro - G-a-rIgG A594, G-a-mIgG2b A647 and G-a-mIgG1 glia cells is error-prone, especially in regions close to CF680 (1:200, ThermoFisher), diluted in 0.1% BSA/BPS. Aβ-plaques, where microglia cells are densely packed. Finally, the slides are washed and incubated with 0.1 μg/ This problem was tackled by starting with the iden - mL DAPI (Sigma Aldrich) for 5 min, after which they are tification of the microglia’s soma. This part the of the mounted with 30 uL Prolong diamond (ThermoFisher). microglia cells should overlap with its nucleus and shows high intensity values, making it easily discern- ible. Segmentation of microglia nuclei and somas was Post‑mortem MRI acquisition and analysis performed using a customized level-set-based cell seg- MRI data and T2*-w severity scores were obtained from mentation method [19]. The main algorithm param - a previous study by Bulk et  al. [10], on the same tissue- eters are the weight for the energy terms minimizing blocks. In this study, tissue blocks were put in proton- the perimeter (ν) and the area (μ), which are empirically free fluid (Fomblin LC08, Solvay), and scanned at room selected for each segmentation task. Larger param- temperature on a 7  T horizontal-bore Bruker MRI sys- eter values correspond to smoother segmentation tem equipped with a 23 mm receiver coil and Paravision results. For the microglia nucleus segmentation, the 5.1 imaging software (Bruker Biospin, Ettlingen, Ger- DNA component image was used as input, and level- many). A gradient echo scan was acquired with repeti- set parameters were set to ν = 2 and μ = 3. Similarly, tion time = 75.0 ms, echo time = 33.9 ms, flip angle = 25° for the microglia soma segmentation the summation of at 100  μm isotropic resolution with 20 signal averages. the intensity values of the membrane (TMEM, PRY12, Subsequently, cortices were assessed for changes in MRI FTL, Iba) component images was utilized as input and contrast following a pre-defined scoring system. level-set parameters were set to ν = 2 and μ = 3. In both cases, level-sets were initialized with regions obtained Iron‑positive cell identification using the Otsu thresholding method [20] which is Whole slide scans of the histochemical iron staining were robust to intensity variation between images originat- exported from Philips Intellisite digital Pathology Solu- ing from the white and grey matter. Additionally, somas tion platform (Philips, the Netherlands) and imported and nuclei with a total area smaller than 50 and 30 pix- into ImageJ. RGB images were converted into 8-bit grey- els, respectively, were removed. scale images. Subsequently, while blinded for diagnosis, For the extension of the obtained segmentation to for each subject an optimal threshold was set to include the whole cytoplasmic area, the approach previously DAB-positive intracellular iron depositions, but exclude described for soma was repeated with less strong regu- extracellular background signal. The cortex of the MTG larization (v = 2, μ = 2). The result of this step was a finer was delineated and the number of positive cells was segmentation capturing microglia areas that are less determined using the ImageJ particle analyser, with a bright than the soma. Connected components overlap- size threshold of 4–100 pixels. Subject AD5 was excluded ping with the previously identified somas were regarded from this analysis, as iron-accumulating cells could not as microglia cells, whereas not overlapping components be distinguished due to high extracellular iron load. were considered as possible detached processes. At this step, in case a blood vessel was identified in an image, the Single cell segmentation Li thresholding method [21] was chosen over Otsu for Identification of the different microglia types was based the initialization of the level-sets algorithm, as it is less on the amount of expressed proteins over the seg- sensitive to the high intensity pixels representing the ves- mented area (Additional file  1: Fig.  S2a). Hence, accu- sel. Vessels were defined as components larger than 4000 rate segmentation of the whole microglia cell area is pixels, after Otsu thresholding of the autofluorescent of paramount importance for our method. Solutions component image. currently available for microglia cell segmentation For correct identification of microglia cells in the (Abdolhoseini et  al. 2019 [18], Inform, PerkinElmer) proximity of Aβ-plaques, the watershed segmentation typically fall short of capturing the whole microglia was applied specifically to those cells whose cytoplas - area (Additional file  1: Fig.  S2b). These are focused on mic area is shared among multiple microglia somas [22]. either capturing the skeleton of the cells, without prop- Aβ-plaque identification was performed employing a erly identifying the cell boundaries (Additional file  1: semi-supervised approach using Ilastik [23]. Fig.  S2c), or segmenting the microglia’s soma exclud- Finally, branches identified within a 10 pixel radius ing their processes, which in the acquired 2D images from the region corresponding to each identified are typically detached from the soma (Additional file  1: Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 4 of 15 Statistical analysis microglia soma were identified as detached processes Firstly, variables were inspected for being gaussian dis- and assigned to the microglia cell. tributed. If normally-distributed, data plots represent A sample resulting segmentation of entire microglial the mean and the standard deviation. For not normally- cells is illustrated in Additional file  1: Fig.  S2e. For the distributed data, data plots show the median with the evaluation of our algorithm, 186 cells were manually seg- corresponding interquartile range. Comparison of two mented, in 7 images from different subjects and regions. continuous variables was performed using a two-tailed Our proposed segmentation framework outperformed unpaired Student’s independent t-test (normally-distrib- the available segmentation solutions correctly captur- uted) or a Mann–Whitney U test (not normally-distrib- ing 153 cells (Inform: 12 cells, Abdolhoseini et  al., 2019: uted). Paired normally distributed data were analysed 49 cells compare Additional file  1: Fig.  S2g), with false using a two-tailed paired Student’s t-test. Bonferroni positive 33 cells (Additional file  1: Fig. S2h) and false neg- post-hoc analysis was performed, and a significance atives 40 cells (Additional file  1: Fig. S2i). Among the cor- level of P < 0.05 was used. The linear correlation between rectly identified cells, median Dice’s similarity index [24] identified number of cells and different pathological hall - of 0.8 was achieved (Additional file 1: Fig. S2f ). marks was assessed using the Pearson correlation coeffi - cient. All statistical tests were performed using GraphPad Cell phenotype identification Prism (Version 8.00, La Jolla, San Diego, CA, USA). Superimposing the segmentation masks onto the compo- nent image of all membrane markers, four mean intensity Results values were extracted for each cell. Afterwards, inten- FTL ‑microglia reflect iron accumulating microglia sity values were normalized imposing Z-score trans- in Alzheimer’s disease formation. For the definition of the different microglia An enhanced Perl’s staining for iron revealed an abun- cell types Phenograph [25], an unsupervised clustering dance of iron-positive cells in the cortex of the MTG in method, was utilized. For Phenograph 100 nearest neigh- Alzheimer’s patients. On further inspection, iron-pos- bours along with the default parameters were selected, in itive cells showed characteristic microglia morphology order to avoid overclustering due to the limited amount with a small soma and many thin processes (Fig.  1a) and of markers. Subsequently, for each Phenograph identified quantification indicated a significant increase of iron- cluster the variability of the single-cell marker expression positive cells in Alzheimer patients compared to con- values was examined (Additional file  1: Fig. S3) through a trols (P = 0.0024; Fig.  1b). Additionally, iron-positive violin plot [26] indicating the variation in each cluster, in cells appeared to cluster in groups, something that was parallel with their expression patterns as illustrated in the not observed in control patients (Fig.  1a). All MTG tis- composite images. sue blocks have also previously been scanned using T2*-w MRI, sensitive for paramagnetic substances such as iron. MRI images were scored based on alterations in Analysis of cellular phenotypes signal intensity reflecting overall parenchymal iron accu - The median expression value of each marker for each mulation and focal iron depositions, and were published phenotype was illustrated with a heatmap. The similari - by Bulk et  al. [10]. An increase of iron-positive micro- ties among the identified phenotypes were observed from glia appeared to be only present in cases with the high- a t-SNE [27] embedding using the same input as in Phe- est MRI severity score, indicating a significant increase nograph and the default parameters. The t-SNE embed - of iron-positive microglia only to occur in subjects with a ding was coloured according to the cluster of each cell, pronounced macroscopic iron-phenotype (Fig.  1c). Sub- its cohort or its individual marker expression values [28]. sequently we studied the correspondence of iron accu- To explore the differences between the Alzheimer mulation with altered expression of the main iron-storage patients and controls regarding their phenotypes and protein ferritin light chain (FTL), as FTL is known to be their spatial relationship with the Aβ-plaques, an interac- expressed in microglia and oligodendrocytes, whereas tive, data-driven pipeline described by [29] was utilized. heavy chain ferritin is primarily expressed by neurons First, using a version of raincloud plots [29] the pheno- in Alzheimer tissue [31]. The Perl’s staining and the FTL types that exist predominantly in each cohort are iden- staining showed a highly similar staining pattern, with tified and consequently, their relative position regarding focal clusters of cells representing microglia morphology the Aβ-plaques using a visual query system are explored. (Fig.  1d). Thus, increased expression of the main iron- For the exploration of the variability in each subject and storage protein FTL appears to reflect iron accumulation the validation of our findings, a customized version of in microglial cells. the motif glyphs described in our previous work [30] was employed. Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 5 of 15 Fig. 1 Increased iron‑positive and corresponding FTL ‑microglia in Alzheimer’s disease a MTG cortex of Alzheimer patients shows increased positivity for iron inside cells with microglial morphology. b Significant increase of iron‑positive cells in Alzheimer patients (n = 11) compared to controls (n = 9)(Mean, Student’s t‑test). c Iron‑positive microglia number only increased in cases with severe signal alterations on iron‑sensitive T2*‑ w MRI, reflected by MRI severity score. d FTL expression reflects intracellular iron accumulation. Scale overview images, 200 μm. Scale zooms, 30 μm Quantitative analysis enables microglia phenotyping (Fig.  3b; Additional file  1: Fig.  S2). After segmentation, To confirm the microglial origin of FTL cells, study unsupervised clustering using Phenograph assigned sin- their activation state and potential interaction with Aβ, gle segmented cells to 20 separate clusters. Following we designed the microglia multispectral immunofluores - manual evaluation of the unsupervised clusters, 6 clus- cence (mic-mIF) panel that can simultaneously detect 6 ters were excluded based on non-microglial morphol- different markers (Additional file  1: Table S2). The MTG ogy and/or sub-threshold expression of all microglial of 12 Alzheimer patients, both of early- and late onset, markers (TMEM119/P2RY12/Iba1). In addition, three and 9 control subjects (Additional file  1: Table  S1) was times two clusters were merged based on similarity in stained and imaged. After image acquisition and mul- protein expression levels and their visual appearance tispectral unmixing of the data, images were exported (Additional file  1: Fig.  S3). Exclusion of the non-micro- for automated segmentation, phenotyping and spatial glial cells resulted in identification of 69,227 cells, with analysis (Fig.  2). In total, 3149 images (110–236 per sub- no significant differences in the number of microglia ject) were obtained. Multispectral unmixing allowed for per mm between control and Alzheimer patients in simultaneous detection of FTL with the nuclear marker either grey matter (GM) or white matter (WM) (Fig. 3c). DAPI, TMEM119, P2RY12, Iba1 and Aβ at 0.5 × 0.5  μm The remaining 11 clusters (C1–C11) were identified as resolution (Fig.  3a). TMEM119 and P2RY12 are gener- major microglia phenotype clusters (Fig.  3d). Though ally considered homeostatic microglia-specific mark - the 11 different phenotypes clustered on the t-SNE plot, ers, based on transcriptomic[3, 32], in vitro [33–35] and the low degree of separation suggests a rather continu- post-mortem IHC studies [35–37], whose expression ous spectrum of expression of the microglia markers decreases when activated. Iba1, on the other hand, is a (Fig.  3e). The control and Alzheimer patients did clus - pan microglia/macrophage marker, which is upregulated ter together, and the marker-based t-SNE plots already upon activation. Finally Aβ stains the characteristic path- revealed more cells with high TMEM119 and P2RY12 ological Aβ-plaques that form in the parenchyma of Alz- expression in controls, but increased FTL expression in heimer patients. Images were segmented using a targeted Alzheimer patients (Fig.  3f). With regard to anatomi- in-house segmentation pipeline allowing segmentation cal region, only C1 and C2 appeared to be more present of cells with processes (like microglia) in 2D images in the grey matter (GM), whereas C5 and C6 appeared Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 6 of 15 Fig. 2 Schematic of mic‑mIF acquisition and analysis pipeline to be proportionally more present in the white mat- Alzheimer-associated cluster shows increased expression ter (WM) (Fig.  3g). Four FTL clusters (C1–C3, C5) of a combination of FTL and Iba1. were identified, with differing expression levels and co-expression levels of P2RY12, TMEM119 and Iba1 Spatial analysis of  FTL microglia clusters + + (Fig.  3d). Cluster C1 (FTL Iba1 ) appeared significantly After cell phenotype identification, all microglia were more present in Alzheimer patients (P = 0.0264), while assessed for proximity to parenchymal Aβ-plaques. For + + + + C2 (P2RY12 TMEM119 FTL Iba1 ) was more pre- visualization purposes, a second image was created, + + sent in controls (P = 0.0055; Fig.  3h). FTL Iba1 clus- where infiltrated Aβ-plaques were plotted onto the ters lacking either P2RY12 (C3) or TMEM119 (C5) did original image as a ‘glyph’ (Fig.  4a) [30], with the dif- not differ significantly in prevalence between control ferent colours corresponding to the respective cluster and Alzheimer patients. Cluster C4 showed solely Iba1 of the infiltrating microglia, to analyse which clusters expression, meaning that this cluster likely also consists predominantly infiltrated Aβ-plaques. Subsequently, all of non-resident infiltrating macrophages. Additionally, individual cells represented as cluster-colored dots or three P2RY12 clusters (C6–C8) were identified, with the cluster-colored glyphs were plotted back onto the the highest expressing cluster (C8) being more present in original whole slide image (Fig. 4a), to assess differences controls. The same applied for the TMEM119 clusters in cluster composition of microglial Aβ infiltration (C9–C11), with C10 and C11 having higher expression on a whole-section scale. As expected, quantifica - and being more present in control patients. These results tion showed significantly more identified Aβ-plaques indicate a small shift of homeostatic microglia positive in Alzheimer patients, although some were found in for P2RY12 and TMEM119 in controls towards activated controls as well (P = 0.0002; Fig .  4b). Furthermore, a microglia, with downregulated expression of P2RY12 and higher percentage of the plaques showed microglia TMEM119 in Alzheimer patients. In addition, a specific infiltration in Alzheimer patients (P = 0.013; Fig .  4c). Looking at the whole slide distribution, Aβ-plaques (See figure on next page.) Fig. 3 Identification of homeostatic and activated Alzheimer ‑associated microglia clusters a Example of mIF image of an Alzheimer patient. b Exemplary images of segmented microglia in a control and an Alzheimer patient. c Number of identified cells in the GM and WM of controls (blue; n = 9) and Alzheimer patients (red; n = 12)(Mean, Student’s t‑test). d Heatmap showing the expression of the four different markers (P2RY12, TMEM119, FTL and Iba1), in the 11 identified microglia clusters. e t ‑SNE plot of all individual cells showing the distinct colour ‑ coded clusters and of control‑ vs. Alzheimer ‑patient ‑ derived cells. f t‑SNE plots colour ‑ coded for intensity of the four individual markers. g Distribution of clusters in GM and WM. h Prevalence of identified clusters (C1–C11) in individual control (blue; n = 9) and Alzheimer patients (red; n = 12) (Median, Mann–Whitney U test). Scale bar, 100 µm. Scale bar zooms, 20 µm. GM Grey matter, WM White matter Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 7 of 15 Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 8 of 15 Fig. 4 FTL ‑microglia show significant Aβ ‑plaque infiltration a Schematic of how microglial Aβ‑plaque infiltration is studied. Both cells and Aβ‑plaques are identified and an interaction map showing ‘glyphs’ in the colour of the cluster of the infiltrating microglia is created. Subsequently glyphs are plotted back onto the whole slide image to also enable studying the spatial distribution pattern. Number of identified Aβ‑plaques (Mean, Student’s t‑test) (b) and the percentage of microglia infiltrated Aβ‑plaques (Mean, Student’s t ‑test) (c) are increased in Alzheimer’s disease (n = 12) compared to controls (n = 9). d Microglia clusters differ spatially, depending on the presence of Aβ‑plaques in their proximity. e Distribution of all‑mic clusters compared to Aβ‑mic clusters of controls and Alzheimer patients. Comparison of prevalence of all‑mic compared to Aβ‑mic of C1‑ (f) and C3‑microglia (g) of all individual Alzheimer patients (n = 12) (paired Student’s t‑test). h Percentage of all identified clusters infiltrating Aβ‑plaques. i Representative images of C1 and C3‑microglia infiltrating an Aβ‑plaque. Scale bar, 50 µm. All mic = all microglia, Aβ‑mic = Aβ‑plaque infiltrating microglia were found to be more present in the coronal sulcus To quantify the influence of Aβ-plaques on micro - rather than the gyrus. This also appeared to be associ - glia phenotype, we compared all phenotyped micro- ated with the regional microglia phenotype, as can be glia (all-mic) with the subset of microglia infiltrating seen for the predominantly purple (C6–C8) microglia Aβ-plaques (Aβ-mic). Controls showed a slight per- populating the Aβ-plaque deplete regions (Fig.  4d). cental increase of C1 and C5 in Aβ-mic compared to Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 9 of 15 + + Fig. 5 C1‑microglia (FTL Iba1 ) reflect iron‑positive microglia a Number of identified C1‑microglia correlates well with number of identified iron‑positive microglia (n = 20, Pearson coefficient). Increased number of C1‑microglia are associated with higher overall Aβ load ( Thal) (b) and Tau Load (Braak) (c). Comparison between APOE4 (n = 6) vs. APOE3 (n = 4) carriers shows increased number of identified Aβ‑plaques (d) and similar microglia infiltration (e). Increased prevalence of C1‑all‑mic (f), no increased proportion of Aβ‑plaques infiltrated with C1‑mic (Aβ‑mic) (g), and significantly increased proportion of C1‑microglia infiltrating Aβ‑plaques (h). d and e Median, Mann–Whitney U test. Patients AD8 and AD12 were excluded from the APOE comparison analysis as they harbour a familial mutation in the APP and PSEN1 gene, respectively, which could be of more influence than the APOE‑ genotype + + all-mic, and less Aβ-plaque infiltration of TMEM119 Correlation of  FTL ‑microglia with pathology clusters C9–C11 (Fig.  4e), though this was based on a As already shown in Fig.  1d, FTL staining closely fol- limited total number of Aβ-plaques. Alzheimer patients lowed the enhanced Perl’s staining showing microglial on the other hand, showed a large percental increase iron loading. Therefore, we also checked the correlation of FTL -clusters C1 and C3 in the Aβ-mic population of the number of iron-positive microglia with the num- (Fig.  4e), which was also statistically significant when ber of identified microglia of different FTL clusters. The + + looking at subject-specific proportional increases (C1: number of identified C1 (FTL Iba1 ) microglia corre- P < 0.0001, C3: P = 0.0004; Fig .  4f, g). While C1 and lated well with number of iron–positive-cells (R = 0.7601, C3-microglia together make up less than 20% of all- p = 0.0004; Fig. 5a), while other FTL clusters with lower mic, they constitute almost 50% of the Aβ-mic popu- expression (C2, C3) did not show correlation with num- lation (Fig.  4e). P2RY12 clusters C6–C8, on the other ber of iron-positive cells (Additional file  1: Fig. S4a, d). hand, showed a small contribution to Aβ-mic com- This suggests that it is especially the marked increase of pared to all-mic (Fig.  4e). Finally, not only did C1 and FTL expression found in C1-microglia that reflects sub - C3 make up the majority of Aβ-mic, but also when stantial iron loading, while moderate FTL expression is examining the proportions of these individual clusters also found in non-iron accumulating cells in controls. that directly infiltrated Aβ-plaques, they showed much Although we already found C1-microglia to significantly higher proportion of infiltration than all the other clus - infiltrate Aβ-plaques, we also checked for its correlation ters (Fig. 4h). A visual example of the C1 and C3-micro- with overall Aβ and Tau load, as assessed by a neuro- glia infiltrating an Aβ-plaque on the original mic-mIF pathologist using Thal stage and Braak stage, respectively. images can be found in Fig.  4i. All in all, these results A marked increase of the number of C1-microglia was suggest Aβ-plaques to be predominantly infiltrated solely found in high-pathology load subjects with Thal by a specific subset of microglia, characterized by phase V, and Braak stage V/VI (Fig. 5b, c), though not all increased FTL and Iba1 expression and loss of expres- high-pathology load subjects show increase of C1-micro- sion of homeostatic markers P2RY12 or TMEM119 and glia. C2-microglia were primarily found in controls with P2RY12. low Braak stage I/II and Thal I-II, whereas C3-microglia were present in both controls and Alzheimer patients Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 10 of 15 Fig. 6 C1 and C3‑microglia show distinct dystrophic morphology compared to homeostatic control microglia a Representative images of the five different morphological subtypes of microglia: homeostatic, activated, dystrophic, phagocytic and macrophage ‑like. b Controls show predominantly homeostatic and activated microglia, while Alzheimer patients show a variety of homeostatic, activated, phagocytic and dystrophic microglia. c Representative images of C1 and C3‑microglia surrounding Aβ‑plaques showing dystrophic morphology. d 3D confocal imaging + + confirms cytorrhexic appearance of FTL Iba1 ‑microglia. Scale bar represents 20 μm unless otherwise stated. Colorcoding for IF‑images in 6A‑ C are according to the box in the top right corner. Colorcoding of 3D confocal images are according to the legend adjacent to the images with varying pathological burdens (Additional file  1: showed no differences in Aβ load, microglia prevalence, Fig. S4b, c, e, f ). This is in line with the finding that iron- or Aβ-infiltration of C1, C2 nor C3 (Additional file  1: Fig. positive microglia were particularly present in Alzhei- S4g–q). In addition, we looked at differences between mer patients with advanced iron loading. However, there APOE3 and APOE4 carriers, as the latter have been is lack of Alzheimer patients with intermediate Thal- found to have elevated ferritin levels in the CSF [38]. As and Braak-scores, making it impossible to state that an expected, APOE4 carriers had more Aβ-plaques (Fig. 5d), increase of C1-microglia is exclusive to advanced stage but did not show overall increased microglia infiltra - disease, and C3 represents an intermediate state between tion (Fig.  5e). Though sample sizes for both groups were C2 in controls and C1 in advanced disease. Further inves- small (n = 4–6), a trend indicating higher prevalence of tigation into the differences between early-onset Alzhei - C1-microglia in the GM could be observed (P = 0.0667; mer’s disease (EOAD, onset < 65y) patients and late-onset Fig. 5f ), which was not the case for C2 and C3-microglia Alzheimer’s disease (LOAD, onset > 65y) patients, (Additional file  1: Fig. S4r, s) However, no difference was Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 11 of 15 observed when looking at the proportion of Aβ-plaques accompanied the pronounced FTL expression. They also infiltrated by C1-microglia (Aβ-mic) (P = 0.5096; Fig. 5g). reflected the morphological appearance of the iron-posi - This suggests that even though a higher percentage of tive microglia identified on the Perl’s staining (Fig. 1a). C1-microglia infiltrate Aβ-plaques (P = 0.0381; Fig.  5h), this is likely due to the increased number of Aβ-plaques Discussion present in the APOE4 carriers. In this manuscript, we confirmed that increased FTL expression reflects an increase in iron accumulation in + + FTL Iba1 ‑microglia have a dystrophic morphological microglia in the cortex of Alzheimer patients. Microglia appearance with increased FTL expression also showed higher Iba1 Finally, we visually evaluated the morphological appear- expression, but loss of homeostatic markers TMEM119 ance of all phenotyped microglia in the same dataset, as and P2RY12, indicative of an activated phenotype. On + + this provides additional information about the activation further investigation this F TL Iba1 phenotype appeared stage of the microglia. Two authors (BK and LdH), evalu- to be increasingly present in Alzheimer patients and the ated the cells according to five distinctive morphological predominant Aβ-plaque infiltrating microglia phenotype. clusters: homeostatic, activated, dystrophic, phagocytic Morphologically they appeared to be in a dystrophic acti- and perivascular macrophages (Fig.  6a), based on pre- vation stage. viously described morphological phenotypes [39]. The Firstly, in this study we confirmed that previously parenchyma of controls was predominantly populated identified iron-positive cells in Alzheimer patients [40, by C6–C11-microglia, which consistently expressed 41] are of microglial rather than astrocytic origin, and TMEM119 and/or P2RY12. These cells presented with show high FTL expression. Subsequently, using mul- homeostatic morphology, showing small circular or tispectral fluorescence and unsupervised clustering, oval cell bodies, with thin highly ramified processes and we identified several FTL -clusters, which were vari- extensive branches (Fig.  6b). Morphological appearance ably present in controls and Alzheimer disease stages. therefore appeared to be in line with the homeostatic C2-microglia, which displayed positivity for all included protein phenotype. Occasionally activated microglia microglia markers, were almost exclusively present in + + were identified, which have larger cell bodies and notice - control patients. Conversely, C1-microglia (F TL Iba1 ) ably fewer branches and ramifications (especially sec - were significantly more present in AD patients, and + + + ond degree) (Fig.  6a). Activated cells generally showed C3-microglia (TMEM119 FTL Iba1 ) were margin- higher Iba1 and FTL expression and were often pheno- ally present in either group. Interestingly, both C1 and typed as C2-microglia (Fig.  6a). Microglia in Alzheimer C3-microglia showed a strong tendency to infiltrate patients, on the other hand, had a much more hetero- Aβ-plaques. C1-microglia were almost exclusively pre- geneous appearance; homeostatic, activated, dystrophic sent in advanced stage Alzheimer patients, whereas and phagocytic microglia could all be observed within C2-microglia were primarily detected in controls (with the coronal sulcus of a single patient (Fig.  6b). Though low Thal/Braak stages), and C3-microglia were variably almost all phenotype clusters and morphological clus- present across controls and Alzheimer patients of all ters could be observed, we focussed on the C1-microglia, stages. Regarding the temporal dynamics of these clus- as they reflected iron-positive  microglia. We found the ters, one could therefore hypothesize that in Alzheimer’s striking majority of C1-microglia to have a dystrophic disease microglia surround Aβ-plaques and lose P2RY12 morphological appearance. The dystrophic cells show expression, as has been observed previously by others a very distinct phenotype, often with a cloudy or cytor- (transition from C2 to C3) [36, 42]. As of yet we do not rhexic (fragmentation of the cytoplasm) appearance know what the relevance is of the preserved TMEM119- which results in ill-defined processes (Fig.  6a). There is expression. Over time, these microglia take up iron, caus- often deramification and the remaining branches show ing a pronounced increase of FTL expression and loss spheroids and fragmentation. Especially microglia (both of TMEM119. This corresponds to the fact that only C1 and C3) infiltrating Aβ plaques showed highly dys - C1-microglia appeared to correlate with iron-accumulat- trophic morphological characteristics, indicative of an ing microglia. However, our study population is not ideal advanced activated/neurodegenerative state (Fig. 6c). The to dissect the temporal dynamics of these clusters, since dystrophic morphology was also verified using 3D confo - the majority of Alzheimer patients showed advanced dis- cal microscopy, which also showed the same cytorrhexic ease (Braak V/VI) and only two patients showed mild to appearance of microglia surrounding the Aβ-plaques moderate (Braak III/IV). Future work studying these phe- (Fig. 6d). All in all, the finding of a dystrophic phenotype notypes in a larger cohort with a larger range of disease in C1-microglia was in line with the increased Iba1 and stages would be highly relevant to accurately determine decreased TMEM119 and P2RY12 expression, which Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 12 of 15 at what stage of the disease C2-microglia prevalence exhaustion, but also increased iron levels. This is in line decreases and C1 and C3-microglia prevalence increases. with a previous study, which found ferritin levels in the Several qualitative studies had previously identified CSF to not be associated with an inflammatory response increased presence of dystrophic ferritin microglia in in Alzheimer patients and hypothesized ferritin levels to brain tissue of Alzheimer patients [6, 40, 43, 44]. The rather reflect changes in iron associated with tangle and dystrophic morphological appearance was also con- plaque pathology [47]. firmed in this study, though the functional insights of Why iron increases with age and even more profoundly these morphologically defined states remains debat - in neurodegenerative diseases is still largely unknown able. Our spatial analysis revealed a strong tendency of [8, 48]. It is hypothesized to be caused by several fac- + + FTL Iba1 to infiltrate Aβ-plaques; significantly more tors including increased blood–brain barrier perme- than can be expected based on prevalence of the clus- ability and disorganization of the iron-dense myelin ter itself, and more than any other identified microglia sheaths [49–51]. Alongside a general increase of iron cluster. Although some other studies had already looked in the parenchyma, iron was also shown to accumulate into the association of dystrophic ferritin microglia inside Aβ-plaques [51, 52]. Therefore, a possible hypoth - with Aβ-plaques [6, 7, 31, 40, 45], results were inconsist- esis for why iron is sequestered in microglia surrounding ent, as none of these studies so far looked into the rela- Aβ-plaques, could be that the iron is taken up as byprod- tive proportion of these microglia in the total population. uct while attempting to phagocytose the Aβ aggregates. The importance of this is also stressed in a recent study Conversely, considering we only found approximately by Nguyen et  al. [46], in which they found an amyloid- 25% of iron-accumulating C1-microglia to infiltrate responsive microglia (ARM) subset, characterized by Aβ-plaques, iron is more likely sequestered using either CD163, but did not pick up on the Aβ-plaque-infiltrating DMT1 or Transferrin-receptors and stored inside FTL, properties of their identified ferritin microglia. Finally, in an attempt to mitigate the potentially toxic effects we were able to further characterize iron-positive/FTL - of free iron, which in its free form is suggested to par- microglia by analyzing co-expression of several other take in Fenton’s reaction to form hydroxyl radicals and microglia markers on a single cell level. This revealed that cause toxic oxidate stress [50]. When iron is taken up by C1-microglia, with the highest FTL protein expression microglia, it first becomes part of the labile iron pool, and increased Iba1 expression, showed complete loss where it can produce reactive oxygen species damaging of expression of homeostatic markers TMEM119 and the mitochondria and other cell organelles [53]. Studies + + P2RY12. Although we acknowledge that our F TL Iba1 ( performed using peripheral tissue cells showed the non- − − P2RY12 TMEM119 )-microglia were only characterized CNS equivalent of microglia, macrophages, to respond using four protein-markers, which is only a fraction com- to intracellular iron accumulation by also activating the pared to the total amount of genes used to define specific NLRP3 inflammasome [54]. Accordingly, in  vitro and transcriptomic states such as the DAM/HAM-states, we in  vivo studies have shown that exposure to a combina- do want to highlight the similarities. The DAM/HAM- tion of iron and Aβ induces the production of cytokine subsets showed FTL among the highest upregulated IL-1β and a switch to glycolytic metabolism in microglia, genes, with coinciding downregulation of TMEM119 both of which can be interpreted as NLRP3-inflammas - and P2RY12 [3, 4]. Additionally clustering around ome activation [55, 56]. NLRP3-inflammasome activa - Aβ-plaques was also reported as a characteristic feature tion in microglia was shown to be able to modify disease of DAM  microglia [3], as is observed for the identified progression in two different Alzheimer mouse models + + FTL Iba1 -microglia. [57, 58]. Our data support the in vitro and mouse model To date, the reason for the observed increase of FTL- evidence that iron and Aβ can act together to accelerate expression remains disputed. With FTL being the long- disease progression via microglial inflammasome activa - term storage component of ferritin, its expression is tion, by showing that in human brain tissue of Alzhei- likely to be increased in response to increased intracel- mer patients, microglia are exposed to a combination lular labile iron concentrations. Yet, ferritin is also widely iron and Aβ. Finally, these findings are also in line with recognized as an acute phase reactant and it has also recent clinical studies, in which iron was found to act as been suggested that microglia upregulate ferritin as a a potential disease modifier by accelerating deterioration response to exhaustion, caused by the attempt to phago- in Alzheimer patients with high Aβ load [12, 13]. cytose aggregated Aβ [45]. However, our findings show Thanks to the possibility to visualize up to six protein + + that the identified FTL Iba1 -microglia closely reflected markers on the same section using mIF, we could bet- microglia with high levels of the metal iron, and therefore ter study the great heterogeneity in microglia phenotype suggest that the observed increased FTL-expression at and its spatial relationship with pathology. A limitation least does not merely reflect inflammatory activation or of mIF compared to other high-dimensional techniques Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 13 of 15 such as single-cell or imaging mass cytometry is the microglia functioning and consequentially accelerate dis- limited number of markers available to characterize the ease progression. complex microglial activation states. However, single- cell mass cytometry lacks the spatial component, which is essential when studying the relation with Aβ. Imaging Conclusion mass cytometry, on the other hand, does capture the spa- In summary, we showed that our multispectral immu- tial distribution, however to date does not enable high- nofluorescence pipeline allowed for accurate identifica - throughput analysis and offers limited resolution. Since tion of specific microglia clusters, and more importantly microglia have very complicated and variable morphol- for the spatial analysis with respect to pathological hall- ogy, solely evaluating protein expression directly sur- marks. In this specific study we identified dystrophic + + − − rounding the nucleus is insufficient, and high-resolution FTL Iba1 TMEM119 P2RY12 -microglia to be signifi - images are required for proper segmentation and pheno- cantly more present in Alzheimer’s disease patient, and typing. Secondly, as we are studying relatively rare acti- to be the predominant Aβ-plaque infiltrating microglia vated microglia subtypes that will not be present in every cluster. Finally, in correspondence with the increase of + + ROI or even subject, we required high-throughput quan-FTL-expression, FTL Iba1 -microglia showed massive titative analysis methods. The mIF-mic panel, together iron-loading. with our optimized microglia segmentation pipeline for 2D-images, enabled accurate segmentation and analysis Supplementary Information of > 60,000 cells to carefully identify the FTL -microglia The online version contains supplementary material available at https ://doi. org/10.1186/s4047 8‑021‑01126 ‑5. in an unbiased fashion. In this study, we adopted an unsupervised learning Additional file 1. approach to generate distinct clusters in our dataset, and avoid bias in the identification in clusters, as can be pre - sent in more classical IHC studies. However, as already Acknowledgements We would like to thank all patients who donated their brain to the Leiden indicated in the results section, even though distinct University Medical Center (LUMC), Netherlands Brain Bank (NBB) or the Normal clusters were identified, the low degree of separation on Aging Brain collection Amsterdam (NABCA), and prof. A.J.M. Rozemuller for the t-SNE mapping and similarity on the associated heat- neuropathological evaluation of the brains. We would also like to thank I.M. Hegeman‑Klein for technical assistance with histological and immunohisto ‑ map, suggest these clusters may be more of a continuum chemical techniques. rather that distinct subsets. This is in line with other transcriptomic and proteomic studies, in which they also Author contributions B.K. and L.v.d.W. conceived and designed the project. B.K., M.I. and N.F.C.C.d.M showed the microglia clusters to be more of a continuum, designed the antibody panel for microglia multispectral immunofluorescence even when studying substantially more genes or proteins (mic‑mIF). A.S., O.D. and B.K. created the microglia segmentation pipeline. A.S. [5, 59, 60]. However, employment of distinct clusters created the spatial analysis tools for mic‑mIF data under supervision of B.P.F.L, J.D. and T.H.. B.K and L.d.H. performed morphological evaluation of microglia. allows for studying the extreme ends of the continuum B.K. and A.S., analysed and interpreted the mic‑mIF data. B.K., A.S., W.M.C.v.R‑M, of the clusters to find meaningful changes in activation T.H., and L.v.d.W. wrote the manuscript. All authors read and approved the final state. Finally, to verify that we were not looking at arbi- manuscript. trary differences in expression levels, we visually checked Funding distinguishability of all independent clusters on the asso- B.K. is supported by an MD/PhD‑ grant from the Leiden University Medical ciated immunohistochemical images and merged clusters Center. In addition, he has received funding from an early career fellowship from Alzheimer Nederland ( WE.15‑2018‑13) and a Eurolife Scholarship for Early where this was not possible, as illustrated in Additional Career researcher. A.S. has received funding through Leiden University Data file 1: Fig. S3. Science Research Programme. LvdW received funding from The Netherlands Future studies looking into the effect of iron and Aβ in Organization for Scientific Research (NWO) Innovational Research Incentives Scheme ( VIDI 864.13.014). humanized models such as iPSC-derived microglia would be extremely valuable to decipher the functional effect of Availability of data and materials this combination, and the influence of Alzheimer-asso - The data that support the findings of this study are available from the cor ‑ responding author upon reasonable request. ciated genetic risk variants such as APOE. In addition, since microglia, as well as iron accumulation, are shown Competing interests to be involved in many different neurodegenerative and The authors have no conflicts of interest to declare. All co ‑authors have seen and agree with the contents of the manuscript and there is no financial inter ‑ neuro-immunological disease such as Parkinson’s disease est to report. and multiple sclerosis, it would be worthwhile looking into this interaction as a common pathway in neurode- Consent for publication Not applicable. generation. Like for Alzheimer disease, iron could inter- act with the accumulating protein of interest to affect Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 14 of 15 Ethics approval and consent to participate multispectral imaging without tyramide signal amplification. J Pathol Clin All material has been collected with written consent from the donors and Res 5:3–11 the procedures have been approved by the Medical Ethical committee of the 18. Abdolhoseini M, Kluge MG, Walker FR, Johnson SJ (2019) Segmentation, LUMC and the Amsterdam UMC. tracing, and quantification of microglial cells from 3D image stacks. Sci Rep 9:8557 Author details 19. Dzyubachyk O, Van Cappellen WA, Essers J, Niessen WJ, Meijering E (2010) Department of Human Genetics, Leiden University Medical Center, Albinus‑ Advanced level‑set ‑based cell tracking in time ‑lapse fluorescence micros‑ dreef 2, 2333 ZA Leiden, The Netherlands. Department of Radiology, Leiden copy. IEEE Trans Med Imaging 29:852–867 University Medical Center, Leiden, The Netherlands. Department of Pathol‑ 20. Otsu N (1979) Threshold selection method from gray‑level histograms. ogy, Leiden University Medical Center, Leiden, The Netherlands. Department IEEE Trans Syst Man Cybern SMC‑9:62–66 of Intelligent Systems, Delft University of Technology, Delft, The Netherlands. 21. Li CH, Lee CK (1993) Minimum cross entropy thresholding. Pattern Recog‑ Department of Cell and Chemical Biology, Leiden University Medical Center, nit Pergamon 26:617–625 Leiden, The Netherlands. 22. Beucher S, Lantuejoul C (1979) Use of watersheds in contour detec‑ tion. Int Work Image Process Real‑ Time Edge Motion Detect Rennes Received: 21 December 2020 Accepted: 30 January 2021 132:2.1–2.12 23. Sommer C, Straehle C, Kothe U, Hamprecht FA (2011) Ilastik: interactive learning and segmentation toolkit. In: Proceedings of the international symposium on biomedical imaging, pp 230–233 24. Dice LR (1945) Measures of the amount of ecologic association between species. Ecology 26:297–302 References 25. Levine JH, Simonds EF, Bendall SC, Davis KL, Amir EAD, Tadmor MD et al 1. Jansen IE, Savage JE, Watanabe K, Bryois J, Williams DM, Steinberg S et al (2015) Data‑ driven phenotypic dissection of AML reveals progenitor‑like (2019) Genome‑ wide meta‑analysis identifies new loci and functional cells that correlate with prognosis. Cell 162:184–197 pathways influencing Alzheimer’s disease risk. Nat Genet 51:404–413 26. Hintze JL, Nelson RD (1998) Violin plots: a box plot‑ density trace syner‑ 2. Lambert JC, Ibrahim‑ Verbaas CA, Harold D, Naj AC, Sims R, Bellenguez C gism. Am Stat 52:181–184 et al (2013) Meta‑analysis of 74,046 individuals identifies 11 new suscepti‑ 27. Van Der Maaten L, Hinton G (2008) Visualizing data using t‑SNE. J Mach bility loci for Alzheimer’s disease. Nat Genet 45:1452–1458 Learn Res 9:2579–2605 3. Keren‑Shaul H, Spinrad A, Weiner A, Matcovitch‑Natan O, Dvir ‑Szternfeld 28. Höllt T, Pezzotti N, van Unen V, Koning F, Eisemann E, Lelieveldt B et al R, Ulland TK et al (2017) A unique microglia type associated with restrict‑ (2016) Cytosplore: interactive immune cell phenotyping for large single‑ ing development of Alzheimer’s disease. Cell 169(1276–1290):e17 cell datasets. Comput Graph Forum 35:171–180 4. Mathys H, Davila‑ Velderrain J, Peng Z, Gao F, Mohammadi S, Young JZ 29. Somarakis A, IJsselsteijn ME, Luk SJ, Kenkhuis B, de Miranda NFCC, et al (2019) Single‑ cell transcriptomic analysis of Alzheimer’s disease. Lelieveldt BPF et al (2021) Visual cohort comparison for spatial single‑ cell Nature 570:332–337 omics‑ data. IEEE Trans Vis Comput Graph 27:733–743 5. Zhou Y, Song WM, Andhey PS, Swain A, Levy T, Miller KR et al (2020) 30. Somarakis A, Van Unen V, Koning F, Lelieveldt BPF, Hollt T (2019) ImaCytE: Human and mouse single‑nucleus transcriptomics reveal TREM2‑ visual exploration of cellular microenvironments for imaging mass dependent and TREM2‑independent cellular responses in Alzheimer’s cytometry data. IEEE Trans Vis Comput Graph 1:1 disease. Nat Med 26:131–142 31. Meadowcroft MD, Connor JR, Yang QX (2015) Cortical iron regulation and 6. Lopes KO, Sparks DL, Streit WJ (2008) Microglial dystrophy in the aged inflammatory response in Alzheimer’s disease and APPSWE/PS1ΔE9 mice: and Alzheimer’s disease brain is associated with ferritin immunoreactivity. a histological perspective. Front Neurosci 9:255 Glia 56:1048–1060 32. Krasemann S, Madore C, Cialic R, Baufeld C, Calcagno N, El Fatimy R et al 7. Grundke‑Iqbal I, Fleming J, Tung YC, Lassmann H, Iqbal K, Joshi JG (1990) (2017) The TREM2‑APOE pathway drives the transcriptional phenotype Ferritin is a component of the neuritic (senile) plaque in Alzheimer of dysfunctional microglia in neurodegenerative diseases. Immunity dementia. Acta Neuropathol 81:105–110 47:566–581 8. Damulina A, Pirpamer L, Soellradl M, Sackl M, Tinauer C, Hofer E et al 33. Banerjee P, Paza E, Perkins EM, James OG, Kenkhuis B, Lloyd AF et al (2020) (2020) Cross‑sectional and longitudinal assessment of brain iron level in Generation of pure monocultures of human microglia‑like cells from Alzheimer disease using 3‑ T MRI. Radiology 296:192541 induced pluripotent stem cells. Stem Cell Res 49:102046 9. Spotorno N, Acosta‑ Cabronero J, Stomrud E, Lampinen B, Strandberg 34. Van Wageningen TA, Vlaar E, Kooij G, Jongenelen CAM, Geurts JJG, Van OT, van Westen D et al (2020) Relationship between cortical iron and tau Dam AM (2019) Regulation of microglial TMEM119 and P2RY12 immuno‑ aggregation in Alzheimer’s disease. Brain 143:1341–1349 reactivity in multiple sclerosis white and grey matter lesions is dependent 10. Bulk M, Kenkhuis B, van der Graaf LM, Goeman JJ, Natté R, van der Weerd on their inflammatory environment. Acta Neuropathol Commun 7:66 L (2018) Postmortem T2*‑ weighted MRI imaging of cortical iron reflects 35. Bennett ML, Bennett FC, Liddelow SA, Ajami B, Zamanian JL, Fernhoff NB severity of Alzheimer’s disease. J Alzheimer’s Dis 65:1125–1137 et al (2016) New tools for studying microglia in the mouse and human 11. van Duijn S, Bulk M, van Duinen SG, Nabuurs RJA, van Buchem MA, van CNS. Proc Natl Acad Sci USA 113:E1738–E1746 der Weerd L et al (2017) Cortical iron reflects severity of Alzheimer’s 36. Mildner A, Huang H, Radke J, Stenzel W, Priller J (2017) P2Y 12 receptor is disease. J Alzheimer’s Dis 60:1533–1545 expressed on human microglia under physiological conditions through‑ 12. Ayton S, Fazlollahi A, Bourgeat P, Raniga P, Ng A, Lim YY et al (2017) out development and is sensitive to neuroinflammatory diseases. Glia Cerebral quantitative susceptibility mapping predicts amyloid‑β‑related 65:375–387 cognitive decline. Brain 140:2112–2119 37. Zrzavy T, Hametner S, Wimmer I, Butovsky O, Weiner HL, Lassmann H 13. Ayton S, Wang Y, Diouf I, Schneider JA, Brockman J, Morris MC et al (2019) (2017) Loss of “homeostatic” microglia and patterns of their activation in Brain iron is associated with accelerated cognitive decline in people with active multiple sclerosis. Brain 140:1900–1913 Alzheimer pathology. Mol Psychiatry 66:1–10 38. Ayton S, Faux NG, Bush AI, Initiative ADN (2015) Ferritin levels in the cer‑ 14. Braak H, Braak E (1991) Neuropathological stageing of Alzheimer‑related ebrospinal fluid predict Alzheimer’s disease outcomes and are regulated changes. Acta Neuropathol 66:239–259 by APOE. Nat Commun 6:6760 15. Braak H, Braak E (1995) Staging of alzheimer’s disease‑related neurofibril‑ 39. Streit WJ, Xue QS, Tischer J, Bechmann I (2014) Microglial pathology. Acta lary changes. Neurobiol Aging Neurobiol Aging 16:271–278 Neuropathol Commun 2:1–17 16. Hyman BT, Phelps CH, Beach TG, Bigio EH, Cairns NJ, Carrillo MC et al 40. Connor JR, Menzies SL, St. Martin SM, Mufson EJ, (1992) A histochemi‑ (2012) National Institute on Aging‑Alzheimer’s Association guidelines cal study of iron, transferrin, and ferritin in Alzheimer’s diseased brains. J for the neuropathologic assessment of Alzheimer’s disease. Alzheimer’s Neurosci Res. 31:75–83 Dement NIH Public Access 8:1–13 41. Zeineh MM, Chen Y, Kitzler HH, Hammond R, Vogel H, Rutt BK (2015) Acti‑ 17. IJsselsteijn ME, Brouwer TP, Abdulrahman Z, Reidy E, Ramalheiro A, vated iron‑ containing microglia in the human hippocampus identified Heeren AM et al (2019) Cancer immunophenotyping by seven‑ colour Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 15 of 15 by magnetic resonance imaging in Alzheimer disease. Neurobiol Aging 52. Nabuurs RJA, Hegeman I, Natté R, van Duinen SG, van Buchem MA, van 36:2483–2500 der Weerd L et al (2011) High‑field MRI of single histological slices using 42. Walker DG, Tang TM, Mendsaikhan A, Tooyama I, Serrano GE, Sue LI et al an inductively coupled, self‑resonant microcoil: application to ex vivo (2020) Patterns of expression of purinergic receptor P2RY12, a putative samples of patients with Alzheimer’s disease. NMR Biomed 24:251–357 marker for non‑activated microglia, in aged and Alzheimer’s disease 53. Dixon SJ, Stockwell BR (2014) The role of iron and reactive oxygen species brains. Int J Mol Sci 21:66 in cell death. Nat Chem Biol 25:9–17 43. Kaneko Y, Kitamoto T, Tateishi J, Yamaguchi K (1989) Ferritin immunohis‑ 54. Nakamura K, Kawakami T, Yamamoto N, Tomizawa M, Fujiwara T, Ishii T tochemistry as a marker for microglia. Acta Neuropathol 79:129–136 et al (2016) Activation of the NLRP3 inflammasome by cellular labile iron. 44. Jellinger K, Paulus W, Grundke‑Iqbal I, Riederer P, Youdim MB (1990) Brain Exp Hematol 44:116–124 iron and ferritin in Parkinson’s and Alzheimer’s diseases. J Neural Transm 55. McIntosh A, Mela V, Harty C, Minogue AM, Costello DA, Kerskens C et al Park Dis Dement Sect 2:327–340 (2019) Iron accumulation in microglia triggers a cascade of events that 45. Streit WJ, Braak H, Del Tredici K, Leyh J, Lier J, Khoshbouei H et al (2018) leads to altered metabolism and compromised function in APP/PS1 mice. Microglial activation occurs late during preclinical Alzheimer’s disease. Brain Pathol 29:606–621 Glia 66:2550–2562 56. Nnah IC, Lee C, Wessling‑Resnick M (2020) Iron potentiates micro ‑ 46. Nguyen AT, Wang K, Hu G, Wang X, Miao Z, Azevedo JA et al (2020) APOE glial interleukin‐1β secretion induced by amyloid‐β. J Neurochem and TREM2 regulate amyloid‑responsive microglia in Alzheimer’s disease. 154:177–189 Acta Neuropathol 1:3 57. Ising C, Venegas C, Zhang S, Scheiblich H, Schmidt SV, Vieira‑Saecker A 47. Ayton S, Janelidze S, Roberts B, Palmqvist S, Kalinowski P, Diouf I et al et al (2019) NLRP3 inflammasome activation drives tau pathology. Nature (2020) Acute phase markers in CSF reveal inflammatory changes in 575:669–673 Alzheimer’s disease that intersect with pathology, APOE ε4, sex and age. 58. Heneka MT, Kummer MP, Stutz A, Delekate A, Schwartz S, Vieira‑Saecker A Prog Neurobiol 47:101904 et al (2013) NLRP3 is activated in Alzheimer’s disease and contributes to 48. Acosta‑ Cabronero J, Betts MJ, Cardenas‑Blanco A, Yang S, Nestor PJ (2016) pathology in APP/PS1 mice. Nature 493:674–678 In vivo MRI mapping of brain iron deposition across the adult lifespan. J 59. Sankowski R, Böttcher C, Masuda T, Geirsdottir L, Sagar, Sindram E et al Neurosci 21:36 (2019) Mapping microglia states in the human brain through the integra‑ 49. Farrall AJ, Wardlaw JM (2009) Blood–brain barrier: ageing and microvas‑ tion of high‑ dimensional techniques. Nat Neurosci 22:2098–2110 cular disease—systematic review and meta‑analysis. Neurobiol Aging 60. Böttcher C, Schlickeiser S, Sneeboer MAM, Kunkel D, Knop A, Paza E et al 25:337–352 (2019) Human microglia regional heterogeneity and phenotypes deter‑ 50. Ward RJ, Zucca FA, Duyn JH, Crichton RR, Zecca L (2014) The role of mined by multiplexed single‑ cell mass cytometry. Nat Neurosci 22:78–90 iron in brain ageing and neurodegenerative disorders. Lancet Neurol 13:1045–1060 Publisher’s Note 51. Bulk M, Abdelmoula WM, Nabuurs RJA, van der Graaf LM, Mulders CWH, Springer Nature remains neutral with regard to jurisdictional claims in pub‑ Mulder AA et al (2018) Postmortem MRI and histology demonstrate dif‑ lished maps and institutional affiliations. ferential iron accumulation and cortical myelin organization in early‑ and late‑ onset Alzheimer’s disease. Neurobiol Aging 62:231–242 Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Acta Neuropathologica Communications Springer Journals

Iron loading is a prominent feature of activated microglia in Alzheimer’s disease patients

Download PDF
15 pages

Loading next page...
 
/lp/springer-journals/iron-loading-is-a-prominent-feature-of-activated-microglia-in-XFc9LeraMs

References (79)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
eISSN
2051-5960
DOI
10.1186/s40478-021-01126-5
Publisher site
See Article on Publisher Site

Abstract

Brain iron accumulation has been found to accelerate disease progression in amyloid‑β(Aβ) positive Alzheimer patients, though the mechanism is still unknown. Microglia have been identified as key players in the disease patho ‑ genesis, and are highly reactive cells responding to aberrations such as increased iron levels. Therefore, using histo‑ logical methods, multispectral immunofluorescence and an automated in‑house developed microglia segmentation and analysis pipeline, we studied the occurrence of iron‑accumulating microglia and the effect on its activation state in human Alzheimer brains. We identified a subset of microglia with increased expression of the iron storage pro ‑ tein ferritin light chain (FTL), together with increased Iba1 expression, decreased TMEM119 and P2RY12 expression. This activated microglia subset represented iron‑accumulating microglia and appeared morphologically dystrophic. + + Multispectral immunofluorescence allowed for spatial analysis of FTL Iba1 ‑microglia, which were found to be the + + predominant Aβ‑plaque infiltrating microglia. Finally, an increase of FTL Iba1 ‑microglia was seen in patients with high Aβ load and Tau load. These findings suggest iron to be taken up by microglia and to influence the functional phenotype of these cells, especially in conjunction with Aβ. Keywords: Alzheimer, Microglia, Iron, Ferritin, Human Introduction Not only can microglia modulate Alzheimer’s disease, Alzheimer’s disease is the most common cause of demen- but many transcriptomic studies showed microglia to undergo the most pronounced changes in response to tia, and is defined by the presence of amyloid-β (Aβ) pathology. In mice, a subset of responding microglia has plaques and tau tangles. In addition, the brain’s resident been found to lose their homeostatic molecular signature innate immune cells, microglia, have been found to be and transition into a so-called ‘disease-associated micro at the centre-stage of the disease, as most identified risk - genes are predominantly or even exclusively expressed in glia’ (DAM) state [3]. In humans, a comparable yet dispa- microglia [1, 2]. rate state coined the human Alzheimer microglia (HAM) has been identified [4]. Upregulated genes in these sub - sets do not only indicate loss of homeostatic function and increased pro-inflammatory activation, but also dysregu - *Correspondence: b.kenkhuis@lumc.nl lated iron-metabolism, manifested via upregulation of the Boyd Kenkhuis and Antonios Somarakis have contributed equally to this FTL-gene and downregulation of FTH1 and SLC2A11 [4, work 5]. FTL encodes the ferritin light chain (FTL) protein, the Department of Human Genetics, Leiden University Medical Center, Albinusdreef 2, 2333 ZA Leiden, The Netherlands component of the major iron-storage complex ferritin, Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 2 of 15 responsible for the long term storage of iron. These tran - Histology and immunohistochemistry scriptomic findings coincide with previously observed Formalin fixed paraffin embedded (FFPE) tissue was seri - ferritin microglia in Alzheimer’s disease [6, 7]. Though ally cut into ten 5-μm-thick and four 10-um-thick sec- increased iron concentration likely plays a role, the exact tions. Consecutive 10-μm-thick sections were used for link between the two has not yet been established. histological detection of iron using an enhanced Perl’s Iron accumulation, irrespective of microglial activa- stain and IHC detection of Ferritin Light Chain (FTL). tion, on the other hand, has been reported in disease- 5-μm-thick sections were used for staining of the micro- affected areas in Alzheimer’s disease, using both in-vivo glia multispectral immunofluorescence (mic-mIF) panel and post-mortem human MRI [8]. Several MRI and (Additional file  1: Table  S2) to verify expression of FTL histology studies found high correlations between iron in microglia/macrophages (Iba1), look at the activation accumulation and cortical Aβ and tau spreading [9–11]. state of these cells (P2RY12/TMEM119) and study the Clinically, increased iron concentrations were shown to interaction with Aβ-plaques. Finally, of three subjects, accelerate cognitive decline in Aβ-positive Alzheimer 20-μm-thick sections were obtained for 3D confocal patients, indicative of a disease-modifying role for iron imaging. Step-by-step histological and IHC optimiza- accumulation [12, 13]. Again, how iron accelerates cogni- tion protocols, together with the imaging parameters, are tive deterioration is poorly understood. reported in the Supplementary Methods. A step-by-step Therefore, in this study we aimed to research the possi - mIF protocol and further analysis of the described histo- ble link between iron accumulation and functionally acti- logical, IHC and mIF staining will be described in the fol- vated microglia, and finally, its relation with Aβ-plaques. lowing sections. We performed a comprehensive investigation of iron- accumulating microglia, and first identified that the iron- storage protein FTL, specifically reflected increased iron Microglia multispectral immunofluorescence (mic‑mIF) accumulation in microglia. Secondly, by using multispec- panel tral immunofluorescence and an in-house automated One 5-μm-thick section of each subject was stained with cell-analysis pipeline, we found F TL microglia to show the mic mIF panel with the following protocol, based significant activation, shown via both downregulation on a previously described protocol by IJsselsteijn et  al. of homeostatic markers TMEM119 and P2RY12 and [17]. Sections were deparaffinized with 3 × 5  min xylene, dystrophic morphology, and to predominantly infiltrate rinsed twice in 100% alcohol and subsequently washed Aβ-plaques. This provides evidence for iron dysregula - with 100% ethanol for 5  min. Endogenous peroxidases tion as a prominent feature of activated microglia in Alz- were blocked for 20  min in 0.3% H O /methanol, after 2 2 heimer’s disease in humans. which the slides were rinsed with 70% and 50% alcohol. Heat induced antigen-retrieval was performed by cook- ing the slides for 10  min in pre-heated citrate (10  mM, Methods pH = 6.0) buffer for 10  min. After cooking, excess buffer Tissue acquisition was removed and slides were cooled for 60  min. Non- Brain autopsy tissue of the middle temporal gyrus (MTG) specific antibody binding sites were blocked with block - of 12 Alzheimer patients and 9 age-matched controls ing buffer (0.1% BSA/PBS + 0.05% Tween) for 30  min. was collected at the Leiden University Medical Center Firstly, slides were incubated with anti-TMEM119 (1:250, (LUMC), Netherlands Brain Bank (NBB) and the Normal Sigma Aldrich) diluted in blocking buffer overnight at Aging Brain collection Amsterdam (NABCA). Patients RT. Slides were washed thrice with PBS and incubated were included based on clinical presentation and diag- with Poly-HRP secondary antibody for 30  min. Slides nosis was confirmed by a neuropathologist. The neuro - were washed again and incubated with the appropriate pathologists also evaluates Braak stage, based on Gallyas Opal tertiary antibody (1:100 in amplification diluent, and Tau immunohistochemistry (IHC), and Thal phase Perkin Elmer) for 60 min, which causes permanent bind- based on Congo Red and Amyloid Beta IHC, in eighteen ing of the fluorophore to the antigen site. All subsequent standard regions, according to the latest international steps are performed in the dark where possible. Finally, diagnostic criteria [14–16]. Patient demographics are the slides are placed back in citrate buffer and cooked in reported in Additional file  1: Table  S1. All material has the microwave for 15 min to wash the primary antibody been collected with written consent from the donors and off. The same steps are repeated for anti-P2RY12 (1:2500, the procedures have been approved by the Medical Ethi- Sigma Aldrich). After binding of the two antibodies cal committee of the LUMC and the Amsterdam UMC. amplified with Opal, slides are incubated with a primary antibody mix with anti-FTL (1:100, Abcam), anti-Aβ (17–24) (1:250, Biolegend) and anti-Iba1 (1:20, Millipore) Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 3 of 15 antibodies, diluted in blocking buffer, overnight at room Fig.  S2d). As a result, a novel segmentation algorithm temperature. The next day, after three washes with PBS, for this type of data was developed. slides are incubated with a secondary antibody mix of Identification of the entire cytoplasmic area of micro - G-a-rIgG A594, G-a-mIgG2b A647 and G-a-mIgG1 glia cells is error-prone, especially in regions close to CF680 (1:200, ThermoFisher), diluted in 0.1% BSA/BPS. Aβ-plaques, where microglia cells are densely packed. Finally, the slides are washed and incubated with 0.1 μg/ This problem was tackled by starting with the iden - mL DAPI (Sigma Aldrich) for 5 min, after which they are tification of the microglia’s soma. This part the of the mounted with 30 uL Prolong diamond (ThermoFisher). microglia cells should overlap with its nucleus and shows high intensity values, making it easily discern- ible. Segmentation of microglia nuclei and somas was Post‑mortem MRI acquisition and analysis performed using a customized level-set-based cell seg- MRI data and T2*-w severity scores were obtained from mentation method [19]. The main algorithm param - a previous study by Bulk et  al. [10], on the same tissue- eters are the weight for the energy terms minimizing blocks. In this study, tissue blocks were put in proton- the perimeter (ν) and the area (μ), which are empirically free fluid (Fomblin LC08, Solvay), and scanned at room selected for each segmentation task. Larger param- temperature on a 7  T horizontal-bore Bruker MRI sys- eter values correspond to smoother segmentation tem equipped with a 23 mm receiver coil and Paravision results. For the microglia nucleus segmentation, the 5.1 imaging software (Bruker Biospin, Ettlingen, Ger- DNA component image was used as input, and level- many). A gradient echo scan was acquired with repeti- set parameters were set to ν = 2 and μ = 3. Similarly, tion time = 75.0 ms, echo time = 33.9 ms, flip angle = 25° for the microglia soma segmentation the summation of at 100  μm isotropic resolution with 20 signal averages. the intensity values of the membrane (TMEM, PRY12, Subsequently, cortices were assessed for changes in MRI FTL, Iba) component images was utilized as input and contrast following a pre-defined scoring system. level-set parameters were set to ν = 2 and μ = 3. In both cases, level-sets were initialized with regions obtained Iron‑positive cell identification using the Otsu thresholding method [20] which is Whole slide scans of the histochemical iron staining were robust to intensity variation between images originat- exported from Philips Intellisite digital Pathology Solu- ing from the white and grey matter. Additionally, somas tion platform (Philips, the Netherlands) and imported and nuclei with a total area smaller than 50 and 30 pix- into ImageJ. RGB images were converted into 8-bit grey- els, respectively, were removed. scale images. Subsequently, while blinded for diagnosis, For the extension of the obtained segmentation to for each subject an optimal threshold was set to include the whole cytoplasmic area, the approach previously DAB-positive intracellular iron depositions, but exclude described for soma was repeated with less strong regu- extracellular background signal. The cortex of the MTG larization (v = 2, μ = 2). The result of this step was a finer was delineated and the number of positive cells was segmentation capturing microglia areas that are less determined using the ImageJ particle analyser, with a bright than the soma. Connected components overlap- size threshold of 4–100 pixels. Subject AD5 was excluded ping with the previously identified somas were regarded from this analysis, as iron-accumulating cells could not as microglia cells, whereas not overlapping components be distinguished due to high extracellular iron load. were considered as possible detached processes. At this step, in case a blood vessel was identified in an image, the Single cell segmentation Li thresholding method [21] was chosen over Otsu for Identification of the different microglia types was based the initialization of the level-sets algorithm, as it is less on the amount of expressed proteins over the seg- sensitive to the high intensity pixels representing the ves- mented area (Additional file  1: Fig.  S2a). Hence, accu- sel. Vessels were defined as components larger than 4000 rate segmentation of the whole microglia cell area is pixels, after Otsu thresholding of the autofluorescent of paramount importance for our method. Solutions component image. currently available for microglia cell segmentation For correct identification of microglia cells in the (Abdolhoseini et  al. 2019 [18], Inform, PerkinElmer) proximity of Aβ-plaques, the watershed segmentation typically fall short of capturing the whole microglia was applied specifically to those cells whose cytoplas - area (Additional file  1: Fig.  S2b). These are focused on mic area is shared among multiple microglia somas [22]. either capturing the skeleton of the cells, without prop- Aβ-plaque identification was performed employing a erly identifying the cell boundaries (Additional file  1: semi-supervised approach using Ilastik [23]. Fig.  S2c), or segmenting the microglia’s soma exclud- Finally, branches identified within a 10 pixel radius ing their processes, which in the acquired 2D images from the region corresponding to each identified are typically detached from the soma (Additional file  1: Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 4 of 15 Statistical analysis microglia soma were identified as detached processes Firstly, variables were inspected for being gaussian dis- and assigned to the microglia cell. tributed. If normally-distributed, data plots represent A sample resulting segmentation of entire microglial the mean and the standard deviation. For not normally- cells is illustrated in Additional file  1: Fig.  S2e. For the distributed data, data plots show the median with the evaluation of our algorithm, 186 cells were manually seg- corresponding interquartile range. Comparison of two mented, in 7 images from different subjects and regions. continuous variables was performed using a two-tailed Our proposed segmentation framework outperformed unpaired Student’s independent t-test (normally-distrib- the available segmentation solutions correctly captur- uted) or a Mann–Whitney U test (not normally-distrib- ing 153 cells (Inform: 12 cells, Abdolhoseini et  al., 2019: uted). Paired normally distributed data were analysed 49 cells compare Additional file  1: Fig.  S2g), with false using a two-tailed paired Student’s t-test. Bonferroni positive 33 cells (Additional file  1: Fig. S2h) and false neg- post-hoc analysis was performed, and a significance atives 40 cells (Additional file  1: Fig. S2i). Among the cor- level of P < 0.05 was used. The linear correlation between rectly identified cells, median Dice’s similarity index [24] identified number of cells and different pathological hall - of 0.8 was achieved (Additional file 1: Fig. S2f ). marks was assessed using the Pearson correlation coeffi - cient. All statistical tests were performed using GraphPad Cell phenotype identification Prism (Version 8.00, La Jolla, San Diego, CA, USA). Superimposing the segmentation masks onto the compo- nent image of all membrane markers, four mean intensity Results values were extracted for each cell. Afterwards, inten- FTL ‑microglia reflect iron accumulating microglia sity values were normalized imposing Z-score trans- in Alzheimer’s disease formation. For the definition of the different microglia An enhanced Perl’s staining for iron revealed an abun- cell types Phenograph [25], an unsupervised clustering dance of iron-positive cells in the cortex of the MTG in method, was utilized. For Phenograph 100 nearest neigh- Alzheimer’s patients. On further inspection, iron-pos- bours along with the default parameters were selected, in itive cells showed characteristic microglia morphology order to avoid overclustering due to the limited amount with a small soma and many thin processes (Fig.  1a) and of markers. Subsequently, for each Phenograph identified quantification indicated a significant increase of iron- cluster the variability of the single-cell marker expression positive cells in Alzheimer patients compared to con- values was examined (Additional file  1: Fig. S3) through a trols (P = 0.0024; Fig.  1b). Additionally, iron-positive violin plot [26] indicating the variation in each cluster, in cells appeared to cluster in groups, something that was parallel with their expression patterns as illustrated in the not observed in control patients (Fig.  1a). All MTG tis- composite images. sue blocks have also previously been scanned using T2*-w MRI, sensitive for paramagnetic substances such as iron. MRI images were scored based on alterations in Analysis of cellular phenotypes signal intensity reflecting overall parenchymal iron accu - The median expression value of each marker for each mulation and focal iron depositions, and were published phenotype was illustrated with a heatmap. The similari - by Bulk et  al. [10]. An increase of iron-positive micro- ties among the identified phenotypes were observed from glia appeared to be only present in cases with the high- a t-SNE [27] embedding using the same input as in Phe- est MRI severity score, indicating a significant increase nograph and the default parameters. The t-SNE embed - of iron-positive microglia only to occur in subjects with a ding was coloured according to the cluster of each cell, pronounced macroscopic iron-phenotype (Fig.  1c). Sub- its cohort or its individual marker expression values [28]. sequently we studied the correspondence of iron accu- To explore the differences between the Alzheimer mulation with altered expression of the main iron-storage patients and controls regarding their phenotypes and protein ferritin light chain (FTL), as FTL is known to be their spatial relationship with the Aβ-plaques, an interac- expressed in microglia and oligodendrocytes, whereas tive, data-driven pipeline described by [29] was utilized. heavy chain ferritin is primarily expressed by neurons First, using a version of raincloud plots [29] the pheno- in Alzheimer tissue [31]. The Perl’s staining and the FTL types that exist predominantly in each cohort are iden- staining showed a highly similar staining pattern, with tified and consequently, their relative position regarding focal clusters of cells representing microglia morphology the Aβ-plaques using a visual query system are explored. (Fig.  1d). Thus, increased expression of the main iron- For the exploration of the variability in each subject and storage protein FTL appears to reflect iron accumulation the validation of our findings, a customized version of in microglial cells. the motif glyphs described in our previous work [30] was employed. Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 5 of 15 Fig. 1 Increased iron‑positive and corresponding FTL ‑microglia in Alzheimer’s disease a MTG cortex of Alzheimer patients shows increased positivity for iron inside cells with microglial morphology. b Significant increase of iron‑positive cells in Alzheimer patients (n = 11) compared to controls (n = 9)(Mean, Student’s t‑test). c Iron‑positive microglia number only increased in cases with severe signal alterations on iron‑sensitive T2*‑ w MRI, reflected by MRI severity score. d FTL expression reflects intracellular iron accumulation. Scale overview images, 200 μm. Scale zooms, 30 μm Quantitative analysis enables microglia phenotyping (Fig.  3b; Additional file  1: Fig.  S2). After segmentation, To confirm the microglial origin of FTL cells, study unsupervised clustering using Phenograph assigned sin- their activation state and potential interaction with Aβ, gle segmented cells to 20 separate clusters. Following we designed the microglia multispectral immunofluores - manual evaluation of the unsupervised clusters, 6 clus- cence (mic-mIF) panel that can simultaneously detect 6 ters were excluded based on non-microglial morphol- different markers (Additional file  1: Table S2). The MTG ogy and/or sub-threshold expression of all microglial of 12 Alzheimer patients, both of early- and late onset, markers (TMEM119/P2RY12/Iba1). In addition, three and 9 control subjects (Additional file  1: Table  S1) was times two clusters were merged based on similarity in stained and imaged. After image acquisition and mul- protein expression levels and their visual appearance tispectral unmixing of the data, images were exported (Additional file  1: Fig.  S3). Exclusion of the non-micro- for automated segmentation, phenotyping and spatial glial cells resulted in identification of 69,227 cells, with analysis (Fig.  2). In total, 3149 images (110–236 per sub- no significant differences in the number of microglia ject) were obtained. Multispectral unmixing allowed for per mm between control and Alzheimer patients in simultaneous detection of FTL with the nuclear marker either grey matter (GM) or white matter (WM) (Fig. 3c). DAPI, TMEM119, P2RY12, Iba1 and Aβ at 0.5 × 0.5  μm The remaining 11 clusters (C1–C11) were identified as resolution (Fig.  3a). TMEM119 and P2RY12 are gener- major microglia phenotype clusters (Fig.  3d). Though ally considered homeostatic microglia-specific mark - the 11 different phenotypes clustered on the t-SNE plot, ers, based on transcriptomic[3, 32], in vitro [33–35] and the low degree of separation suggests a rather continu- post-mortem IHC studies [35–37], whose expression ous spectrum of expression of the microglia markers decreases when activated. Iba1, on the other hand, is a (Fig.  3e). The control and Alzheimer patients did clus - pan microglia/macrophage marker, which is upregulated ter together, and the marker-based t-SNE plots already upon activation. Finally Aβ stains the characteristic path- revealed more cells with high TMEM119 and P2RY12 ological Aβ-plaques that form in the parenchyma of Alz- expression in controls, but increased FTL expression in heimer patients. Images were segmented using a targeted Alzheimer patients (Fig.  3f). With regard to anatomi- in-house segmentation pipeline allowing segmentation cal region, only C1 and C2 appeared to be more present of cells with processes (like microglia) in 2D images in the grey matter (GM), whereas C5 and C6 appeared Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 6 of 15 Fig. 2 Schematic of mic‑mIF acquisition and analysis pipeline to be proportionally more present in the white mat- Alzheimer-associated cluster shows increased expression ter (WM) (Fig.  3g). Four FTL clusters (C1–C3, C5) of a combination of FTL and Iba1. were identified, with differing expression levels and co-expression levels of P2RY12, TMEM119 and Iba1 Spatial analysis of  FTL microglia clusters + + (Fig.  3d). Cluster C1 (FTL Iba1 ) appeared significantly After cell phenotype identification, all microglia were more present in Alzheimer patients (P = 0.0264), while assessed for proximity to parenchymal Aβ-plaques. For + + + + C2 (P2RY12 TMEM119 FTL Iba1 ) was more pre- visualization purposes, a second image was created, + + sent in controls (P = 0.0055; Fig.  3h). FTL Iba1 clus- where infiltrated Aβ-plaques were plotted onto the ters lacking either P2RY12 (C3) or TMEM119 (C5) did original image as a ‘glyph’ (Fig.  4a) [30], with the dif- not differ significantly in prevalence between control ferent colours corresponding to the respective cluster and Alzheimer patients. Cluster C4 showed solely Iba1 of the infiltrating microglia, to analyse which clusters expression, meaning that this cluster likely also consists predominantly infiltrated Aβ-plaques. Subsequently, all of non-resident infiltrating macrophages. Additionally, individual cells represented as cluster-colored dots or three P2RY12 clusters (C6–C8) were identified, with the cluster-colored glyphs were plotted back onto the the highest expressing cluster (C8) being more present in original whole slide image (Fig. 4a), to assess differences controls. The same applied for the TMEM119 clusters in cluster composition of microglial Aβ infiltration (C9–C11), with C10 and C11 having higher expression on a whole-section scale. As expected, quantifica - and being more present in control patients. These results tion showed significantly more identified Aβ-plaques indicate a small shift of homeostatic microglia positive in Alzheimer patients, although some were found in for P2RY12 and TMEM119 in controls towards activated controls as well (P = 0.0002; Fig .  4b). Furthermore, a microglia, with downregulated expression of P2RY12 and higher percentage of the plaques showed microglia TMEM119 in Alzheimer patients. In addition, a specific infiltration in Alzheimer patients (P = 0.013; Fig .  4c). Looking at the whole slide distribution, Aβ-plaques (See figure on next page.) Fig. 3 Identification of homeostatic and activated Alzheimer ‑associated microglia clusters a Example of mIF image of an Alzheimer patient. b Exemplary images of segmented microglia in a control and an Alzheimer patient. c Number of identified cells in the GM and WM of controls (blue; n = 9) and Alzheimer patients (red; n = 12)(Mean, Student’s t‑test). d Heatmap showing the expression of the four different markers (P2RY12, TMEM119, FTL and Iba1), in the 11 identified microglia clusters. e t ‑SNE plot of all individual cells showing the distinct colour ‑ coded clusters and of control‑ vs. Alzheimer ‑patient ‑ derived cells. f t‑SNE plots colour ‑ coded for intensity of the four individual markers. g Distribution of clusters in GM and WM. h Prevalence of identified clusters (C1–C11) in individual control (blue; n = 9) and Alzheimer patients (red; n = 12) (Median, Mann–Whitney U test). Scale bar, 100 µm. Scale bar zooms, 20 µm. GM Grey matter, WM White matter Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 7 of 15 Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 8 of 15 Fig. 4 FTL ‑microglia show significant Aβ ‑plaque infiltration a Schematic of how microglial Aβ‑plaque infiltration is studied. Both cells and Aβ‑plaques are identified and an interaction map showing ‘glyphs’ in the colour of the cluster of the infiltrating microglia is created. Subsequently glyphs are plotted back onto the whole slide image to also enable studying the spatial distribution pattern. Number of identified Aβ‑plaques (Mean, Student’s t‑test) (b) and the percentage of microglia infiltrated Aβ‑plaques (Mean, Student’s t ‑test) (c) are increased in Alzheimer’s disease (n = 12) compared to controls (n = 9). d Microglia clusters differ spatially, depending on the presence of Aβ‑plaques in their proximity. e Distribution of all‑mic clusters compared to Aβ‑mic clusters of controls and Alzheimer patients. Comparison of prevalence of all‑mic compared to Aβ‑mic of C1‑ (f) and C3‑microglia (g) of all individual Alzheimer patients (n = 12) (paired Student’s t‑test). h Percentage of all identified clusters infiltrating Aβ‑plaques. i Representative images of C1 and C3‑microglia infiltrating an Aβ‑plaque. Scale bar, 50 µm. All mic = all microglia, Aβ‑mic = Aβ‑plaque infiltrating microglia were found to be more present in the coronal sulcus To quantify the influence of Aβ-plaques on micro - rather than the gyrus. This also appeared to be associ - glia phenotype, we compared all phenotyped micro- ated with the regional microglia phenotype, as can be glia (all-mic) with the subset of microglia infiltrating seen for the predominantly purple (C6–C8) microglia Aβ-plaques (Aβ-mic). Controls showed a slight per- populating the Aβ-plaque deplete regions (Fig.  4d). cental increase of C1 and C5 in Aβ-mic compared to Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 9 of 15 + + Fig. 5 C1‑microglia (FTL Iba1 ) reflect iron‑positive microglia a Number of identified C1‑microglia correlates well with number of identified iron‑positive microglia (n = 20, Pearson coefficient). Increased number of C1‑microglia are associated with higher overall Aβ load ( Thal) (b) and Tau Load (Braak) (c). Comparison between APOE4 (n = 6) vs. APOE3 (n = 4) carriers shows increased number of identified Aβ‑plaques (d) and similar microglia infiltration (e). Increased prevalence of C1‑all‑mic (f), no increased proportion of Aβ‑plaques infiltrated with C1‑mic (Aβ‑mic) (g), and significantly increased proportion of C1‑microglia infiltrating Aβ‑plaques (h). d and e Median, Mann–Whitney U test. Patients AD8 and AD12 were excluded from the APOE comparison analysis as they harbour a familial mutation in the APP and PSEN1 gene, respectively, which could be of more influence than the APOE‑ genotype + + all-mic, and less Aβ-plaque infiltration of TMEM119 Correlation of  FTL ‑microglia with pathology clusters C9–C11 (Fig.  4e), though this was based on a As already shown in Fig.  1d, FTL staining closely fol- limited total number of Aβ-plaques. Alzheimer patients lowed the enhanced Perl’s staining showing microglial on the other hand, showed a large percental increase iron loading. Therefore, we also checked the correlation of FTL -clusters C1 and C3 in the Aβ-mic population of the number of iron-positive microglia with the num- (Fig.  4e), which was also statistically significant when ber of identified microglia of different FTL clusters. The + + looking at subject-specific proportional increases (C1: number of identified C1 (FTL Iba1 ) microglia corre- P < 0.0001, C3: P = 0.0004; Fig .  4f, g). While C1 and lated well with number of iron–positive-cells (R = 0.7601, C3-microglia together make up less than 20% of all- p = 0.0004; Fig. 5a), while other FTL clusters with lower mic, they constitute almost 50% of the Aβ-mic popu- expression (C2, C3) did not show correlation with num- lation (Fig.  4e). P2RY12 clusters C6–C8, on the other ber of iron-positive cells (Additional file  1: Fig. S4a, d). hand, showed a small contribution to Aβ-mic com- This suggests that it is especially the marked increase of pared to all-mic (Fig.  4e). Finally, not only did C1 and FTL expression found in C1-microglia that reflects sub - C3 make up the majority of Aβ-mic, but also when stantial iron loading, while moderate FTL expression is examining the proportions of these individual clusters also found in non-iron accumulating cells in controls. that directly infiltrated Aβ-plaques, they showed much Although we already found C1-microglia to significantly higher proportion of infiltration than all the other clus - infiltrate Aβ-plaques, we also checked for its correlation ters (Fig. 4h). A visual example of the C1 and C3-micro- with overall Aβ and Tau load, as assessed by a neuro- glia infiltrating an Aβ-plaque on the original mic-mIF pathologist using Thal stage and Braak stage, respectively. images can be found in Fig.  4i. All in all, these results A marked increase of the number of C1-microglia was suggest Aβ-plaques to be predominantly infiltrated solely found in high-pathology load subjects with Thal by a specific subset of microglia, characterized by phase V, and Braak stage V/VI (Fig. 5b, c), though not all increased FTL and Iba1 expression and loss of expres- high-pathology load subjects show increase of C1-micro- sion of homeostatic markers P2RY12 or TMEM119 and glia. C2-microglia were primarily found in controls with P2RY12. low Braak stage I/II and Thal I-II, whereas C3-microglia were present in both controls and Alzheimer patients Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 10 of 15 Fig. 6 C1 and C3‑microglia show distinct dystrophic morphology compared to homeostatic control microglia a Representative images of the five different morphological subtypes of microglia: homeostatic, activated, dystrophic, phagocytic and macrophage ‑like. b Controls show predominantly homeostatic and activated microglia, while Alzheimer patients show a variety of homeostatic, activated, phagocytic and dystrophic microglia. c Representative images of C1 and C3‑microglia surrounding Aβ‑plaques showing dystrophic morphology. d 3D confocal imaging + + confirms cytorrhexic appearance of FTL Iba1 ‑microglia. Scale bar represents 20 μm unless otherwise stated. Colorcoding for IF‑images in 6A‑ C are according to the box in the top right corner. Colorcoding of 3D confocal images are according to the legend adjacent to the images with varying pathological burdens (Additional file  1: showed no differences in Aβ load, microglia prevalence, Fig. S4b, c, e, f ). This is in line with the finding that iron- or Aβ-infiltration of C1, C2 nor C3 (Additional file  1: Fig. positive microglia were particularly present in Alzhei- S4g–q). In addition, we looked at differences between mer patients with advanced iron loading. However, there APOE3 and APOE4 carriers, as the latter have been is lack of Alzheimer patients with intermediate Thal- found to have elevated ferritin levels in the CSF [38]. As and Braak-scores, making it impossible to state that an expected, APOE4 carriers had more Aβ-plaques (Fig. 5d), increase of C1-microglia is exclusive to advanced stage but did not show overall increased microglia infiltra - disease, and C3 represents an intermediate state between tion (Fig.  5e). Though sample sizes for both groups were C2 in controls and C1 in advanced disease. Further inves- small (n = 4–6), a trend indicating higher prevalence of tigation into the differences between early-onset Alzhei - C1-microglia in the GM could be observed (P = 0.0667; mer’s disease (EOAD, onset < 65y) patients and late-onset Fig. 5f ), which was not the case for C2 and C3-microglia Alzheimer’s disease (LOAD, onset > 65y) patients, (Additional file  1: Fig. S4r, s) However, no difference was Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 11 of 15 observed when looking at the proportion of Aβ-plaques accompanied the pronounced FTL expression. They also infiltrated by C1-microglia (Aβ-mic) (P = 0.5096; Fig. 5g). reflected the morphological appearance of the iron-posi - This suggests that even though a higher percentage of tive microglia identified on the Perl’s staining (Fig. 1a). C1-microglia infiltrate Aβ-plaques (P = 0.0381; Fig.  5h), this is likely due to the increased number of Aβ-plaques Discussion present in the APOE4 carriers. In this manuscript, we confirmed that increased FTL expression reflects an increase in iron accumulation in + + FTL Iba1 ‑microglia have a dystrophic morphological microglia in the cortex of Alzheimer patients. Microglia appearance with increased FTL expression also showed higher Iba1 Finally, we visually evaluated the morphological appear- expression, but loss of homeostatic markers TMEM119 ance of all phenotyped microglia in the same dataset, as and P2RY12, indicative of an activated phenotype. On + + this provides additional information about the activation further investigation this F TL Iba1 phenotype appeared stage of the microglia. Two authors (BK and LdH), evalu- to be increasingly present in Alzheimer patients and the ated the cells according to five distinctive morphological predominant Aβ-plaque infiltrating microglia phenotype. clusters: homeostatic, activated, dystrophic, phagocytic Morphologically they appeared to be in a dystrophic acti- and perivascular macrophages (Fig.  6a), based on pre- vation stage. viously described morphological phenotypes [39]. The Firstly, in this study we confirmed that previously parenchyma of controls was predominantly populated identified iron-positive cells in Alzheimer patients [40, by C6–C11-microglia, which consistently expressed 41] are of microglial rather than astrocytic origin, and TMEM119 and/or P2RY12. These cells presented with show high FTL expression. Subsequently, using mul- homeostatic morphology, showing small circular or tispectral fluorescence and unsupervised clustering, oval cell bodies, with thin highly ramified processes and we identified several FTL -clusters, which were vari- extensive branches (Fig.  6b). Morphological appearance ably present in controls and Alzheimer disease stages. therefore appeared to be in line with the homeostatic C2-microglia, which displayed positivity for all included protein phenotype. Occasionally activated microglia microglia markers, were almost exclusively present in + + were identified, which have larger cell bodies and notice - control patients. Conversely, C1-microglia (F TL Iba1 ) ably fewer branches and ramifications (especially sec - were significantly more present in AD patients, and + + + ond degree) (Fig.  6a). Activated cells generally showed C3-microglia (TMEM119 FTL Iba1 ) were margin- higher Iba1 and FTL expression and were often pheno- ally present in either group. Interestingly, both C1 and typed as C2-microglia (Fig.  6a). Microglia in Alzheimer C3-microglia showed a strong tendency to infiltrate patients, on the other hand, had a much more hetero- Aβ-plaques. C1-microglia were almost exclusively pre- geneous appearance; homeostatic, activated, dystrophic sent in advanced stage Alzheimer patients, whereas and phagocytic microglia could all be observed within C2-microglia were primarily detected in controls (with the coronal sulcus of a single patient (Fig.  6b). Though low Thal/Braak stages), and C3-microglia were variably almost all phenotype clusters and morphological clus- present across controls and Alzheimer patients of all ters could be observed, we focussed on the C1-microglia, stages. Regarding the temporal dynamics of these clus- as they reflected iron-positive  microglia. We found the ters, one could therefore hypothesize that in Alzheimer’s striking majority of C1-microglia to have a dystrophic disease microglia surround Aβ-plaques and lose P2RY12 morphological appearance. The dystrophic cells show expression, as has been observed previously by others a very distinct phenotype, often with a cloudy or cytor- (transition from C2 to C3) [36, 42]. As of yet we do not rhexic (fragmentation of the cytoplasm) appearance know what the relevance is of the preserved TMEM119- which results in ill-defined processes (Fig.  6a). There is expression. Over time, these microglia take up iron, caus- often deramification and the remaining branches show ing a pronounced increase of FTL expression and loss spheroids and fragmentation. Especially microglia (both of TMEM119. This corresponds to the fact that only C1 and C3) infiltrating Aβ plaques showed highly dys - C1-microglia appeared to correlate with iron-accumulat- trophic morphological characteristics, indicative of an ing microglia. However, our study population is not ideal advanced activated/neurodegenerative state (Fig. 6c). The to dissect the temporal dynamics of these clusters, since dystrophic morphology was also verified using 3D confo - the majority of Alzheimer patients showed advanced dis- cal microscopy, which also showed the same cytorrhexic ease (Braak V/VI) and only two patients showed mild to appearance of microglia surrounding the Aβ-plaques moderate (Braak III/IV). Future work studying these phe- (Fig. 6d). All in all, the finding of a dystrophic phenotype notypes in a larger cohort with a larger range of disease in C1-microglia was in line with the increased Iba1 and stages would be highly relevant to accurately determine decreased TMEM119 and P2RY12 expression, which Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 12 of 15 at what stage of the disease C2-microglia prevalence exhaustion, but also increased iron levels. This is in line decreases and C1 and C3-microglia prevalence increases. with a previous study, which found ferritin levels in the Several qualitative studies had previously identified CSF to not be associated with an inflammatory response increased presence of dystrophic ferritin microglia in in Alzheimer patients and hypothesized ferritin levels to brain tissue of Alzheimer patients [6, 40, 43, 44]. The rather reflect changes in iron associated with tangle and dystrophic morphological appearance was also con- plaque pathology [47]. firmed in this study, though the functional insights of Why iron increases with age and even more profoundly these morphologically defined states remains debat - in neurodegenerative diseases is still largely unknown able. Our spatial analysis revealed a strong tendency of [8, 48]. It is hypothesized to be caused by several fac- + + FTL Iba1 to infiltrate Aβ-plaques; significantly more tors including increased blood–brain barrier perme- than can be expected based on prevalence of the clus- ability and disorganization of the iron-dense myelin ter itself, and more than any other identified microglia sheaths [49–51]. Alongside a general increase of iron cluster. Although some other studies had already looked in the parenchyma, iron was also shown to accumulate into the association of dystrophic ferritin microglia inside Aβ-plaques [51, 52]. Therefore, a possible hypoth - with Aβ-plaques [6, 7, 31, 40, 45], results were inconsist- esis for why iron is sequestered in microglia surrounding ent, as none of these studies so far looked into the rela- Aβ-plaques, could be that the iron is taken up as byprod- tive proportion of these microglia in the total population. uct while attempting to phagocytose the Aβ aggregates. The importance of this is also stressed in a recent study Conversely, considering we only found approximately by Nguyen et  al. [46], in which they found an amyloid- 25% of iron-accumulating C1-microglia to infiltrate responsive microglia (ARM) subset, characterized by Aβ-plaques, iron is more likely sequestered using either CD163, but did not pick up on the Aβ-plaque-infiltrating DMT1 or Transferrin-receptors and stored inside FTL, properties of their identified ferritin microglia. Finally, in an attempt to mitigate the potentially toxic effects we were able to further characterize iron-positive/FTL - of free iron, which in its free form is suggested to par- microglia by analyzing co-expression of several other take in Fenton’s reaction to form hydroxyl radicals and microglia markers on a single cell level. This revealed that cause toxic oxidate stress [50]. When iron is taken up by C1-microglia, with the highest FTL protein expression microglia, it first becomes part of the labile iron pool, and increased Iba1 expression, showed complete loss where it can produce reactive oxygen species damaging of expression of homeostatic markers TMEM119 and the mitochondria and other cell organelles [53]. Studies + + P2RY12. Although we acknowledge that our F TL Iba1 ( performed using peripheral tissue cells showed the non- − − P2RY12 TMEM119 )-microglia were only characterized CNS equivalent of microglia, macrophages, to respond using four protein-markers, which is only a fraction com- to intracellular iron accumulation by also activating the pared to the total amount of genes used to define specific NLRP3 inflammasome [54]. Accordingly, in  vitro and transcriptomic states such as the DAM/HAM-states, we in  vivo studies have shown that exposure to a combina- do want to highlight the similarities. The DAM/HAM- tion of iron and Aβ induces the production of cytokine subsets showed FTL among the highest upregulated IL-1β and a switch to glycolytic metabolism in microglia, genes, with coinciding downregulation of TMEM119 both of which can be interpreted as NLRP3-inflammas - and P2RY12 [3, 4]. Additionally clustering around ome activation [55, 56]. NLRP3-inflammasome activa - Aβ-plaques was also reported as a characteristic feature tion in microglia was shown to be able to modify disease of DAM  microglia [3], as is observed for the identified progression in two different Alzheimer mouse models + + FTL Iba1 -microglia. [57, 58]. Our data support the in vitro and mouse model To date, the reason for the observed increase of FTL- evidence that iron and Aβ can act together to accelerate expression remains disputed. With FTL being the long- disease progression via microglial inflammasome activa - term storage component of ferritin, its expression is tion, by showing that in human brain tissue of Alzhei- likely to be increased in response to increased intracel- mer patients, microglia are exposed to a combination lular labile iron concentrations. Yet, ferritin is also widely iron and Aβ. Finally, these findings are also in line with recognized as an acute phase reactant and it has also recent clinical studies, in which iron was found to act as been suggested that microglia upregulate ferritin as a a potential disease modifier by accelerating deterioration response to exhaustion, caused by the attempt to phago- in Alzheimer patients with high Aβ load [12, 13]. cytose aggregated Aβ [45]. However, our findings show Thanks to the possibility to visualize up to six protein + + that the identified FTL Iba1 -microglia closely reflected markers on the same section using mIF, we could bet- microglia with high levels of the metal iron, and therefore ter study the great heterogeneity in microglia phenotype suggest that the observed increased FTL-expression at and its spatial relationship with pathology. A limitation least does not merely reflect inflammatory activation or of mIF compared to other high-dimensional techniques Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 13 of 15 such as single-cell or imaging mass cytometry is the microglia functioning and consequentially accelerate dis- limited number of markers available to characterize the ease progression. complex microglial activation states. However, single- cell mass cytometry lacks the spatial component, which is essential when studying the relation with Aβ. Imaging Conclusion mass cytometry, on the other hand, does capture the spa- In summary, we showed that our multispectral immu- tial distribution, however to date does not enable high- nofluorescence pipeline allowed for accurate identifica - throughput analysis and offers limited resolution. Since tion of specific microglia clusters, and more importantly microglia have very complicated and variable morphol- for the spatial analysis with respect to pathological hall- ogy, solely evaluating protein expression directly sur- marks. In this specific study we identified dystrophic + + − − rounding the nucleus is insufficient, and high-resolution FTL Iba1 TMEM119 P2RY12 -microglia to be signifi - images are required for proper segmentation and pheno- cantly more present in Alzheimer’s disease patient, and typing. Secondly, as we are studying relatively rare acti- to be the predominant Aβ-plaque infiltrating microglia vated microglia subtypes that will not be present in every cluster. Finally, in correspondence with the increase of + + ROI or even subject, we required high-throughput quan-FTL-expression, FTL Iba1 -microglia showed massive titative analysis methods. The mIF-mic panel, together iron-loading. with our optimized microglia segmentation pipeline for 2D-images, enabled accurate segmentation and analysis Supplementary Information of > 60,000 cells to carefully identify the FTL -microglia The online version contains supplementary material available at https ://doi. org/10.1186/s4047 8‑021‑01126 ‑5. in an unbiased fashion. In this study, we adopted an unsupervised learning Additional file 1. approach to generate distinct clusters in our dataset, and avoid bias in the identification in clusters, as can be pre - sent in more classical IHC studies. However, as already Acknowledgements We would like to thank all patients who donated their brain to the Leiden indicated in the results section, even though distinct University Medical Center (LUMC), Netherlands Brain Bank (NBB) or the Normal clusters were identified, the low degree of separation on Aging Brain collection Amsterdam (NABCA), and prof. A.J.M. Rozemuller for the t-SNE mapping and similarity on the associated heat- neuropathological evaluation of the brains. We would also like to thank I.M. Hegeman‑Klein for technical assistance with histological and immunohisto ‑ map, suggest these clusters may be more of a continuum chemical techniques. rather that distinct subsets. This is in line with other transcriptomic and proteomic studies, in which they also Author contributions B.K. and L.v.d.W. conceived and designed the project. B.K., M.I. and N.F.C.C.d.M showed the microglia clusters to be more of a continuum, designed the antibody panel for microglia multispectral immunofluorescence even when studying substantially more genes or proteins (mic‑mIF). A.S., O.D. and B.K. created the microglia segmentation pipeline. A.S. [5, 59, 60]. However, employment of distinct clusters created the spatial analysis tools for mic‑mIF data under supervision of B.P.F.L, J.D. and T.H.. B.K and L.d.H. performed morphological evaluation of microglia. allows for studying the extreme ends of the continuum B.K. and A.S., analysed and interpreted the mic‑mIF data. B.K., A.S., W.M.C.v.R‑M, of the clusters to find meaningful changes in activation T.H., and L.v.d.W. wrote the manuscript. All authors read and approved the final state. Finally, to verify that we were not looking at arbi- manuscript. trary differences in expression levels, we visually checked Funding distinguishability of all independent clusters on the asso- B.K. is supported by an MD/PhD‑ grant from the Leiden University Medical ciated immunohistochemical images and merged clusters Center. In addition, he has received funding from an early career fellowship from Alzheimer Nederland ( WE.15‑2018‑13) and a Eurolife Scholarship for Early where this was not possible, as illustrated in Additional Career researcher. A.S. has received funding through Leiden University Data file 1: Fig. S3. Science Research Programme. LvdW received funding from The Netherlands Future studies looking into the effect of iron and Aβ in Organization for Scientific Research (NWO) Innovational Research Incentives Scheme ( VIDI 864.13.014). humanized models such as iPSC-derived microglia would be extremely valuable to decipher the functional effect of Availability of data and materials this combination, and the influence of Alzheimer-asso - The data that support the findings of this study are available from the cor ‑ responding author upon reasonable request. ciated genetic risk variants such as APOE. In addition, since microglia, as well as iron accumulation, are shown Competing interests to be involved in many different neurodegenerative and The authors have no conflicts of interest to declare. All co ‑authors have seen and agree with the contents of the manuscript and there is no financial inter ‑ neuro-immunological disease such as Parkinson’s disease est to report. and multiple sclerosis, it would be worthwhile looking into this interaction as a common pathway in neurode- Consent for publication Not applicable. generation. Like for Alzheimer disease, iron could inter- act with the accumulating protein of interest to affect Kenkhuis et al. acta neuropathol commun (2021) 9:27 Page 14 of 15 Ethics approval and consent to participate multispectral imaging without tyramide signal amplification. J Pathol Clin All material has been collected with written consent from the donors and Res 5:3–11 the procedures have been approved by the Medical Ethical committee of the 18. Abdolhoseini M, Kluge MG, Walker FR, Johnson SJ (2019) Segmentation, LUMC and the Amsterdam UMC. tracing, and quantification of microglial cells from 3D image stacks. Sci Rep 9:8557 Author details 19. Dzyubachyk O, Van Cappellen WA, Essers J, Niessen WJ, Meijering E (2010) Department of Human Genetics, Leiden University Medical Center, Albinus‑ Advanced level‑set ‑based cell tracking in time ‑lapse fluorescence micros‑ dreef 2, 2333 ZA Leiden, The Netherlands. Department of Radiology, Leiden copy. IEEE Trans Med Imaging 29:852–867 University Medical Center, Leiden, The Netherlands. Department of Pathol‑ 20. Otsu N (1979) Threshold selection method from gray‑level histograms. ogy, Leiden University Medical Center, Leiden, The Netherlands. Department IEEE Trans Syst Man Cybern SMC‑9:62–66 of Intelligent Systems, Delft University of Technology, Delft, The Netherlands. 21. Li CH, Lee CK (1993) Minimum cross entropy thresholding. Pattern Recog‑ Department of Cell and Chemical Biology, Leiden University Medical Center, nit Pergamon 26:617–625 Leiden, The Netherlands. 22. Beucher S, Lantuejoul C (1979) Use of watersheds in contour detec‑ tion. Int Work Image Process Real‑ Time Edge Motion Detect Rennes Received: 21 December 2020 Accepted: 30 January 2021 132:2.1–2.12 23. Sommer C, Straehle C, Kothe U, Hamprecht FA (2011) Ilastik: interactive learning and segmentation toolkit. In: Proceedings of the international symposium on biomedical imaging, pp 230–233 24. Dice LR (1945) Measures of the amount of ecologic association between species. Ecology 26:297–302 References 25. Levine JH, Simonds EF, Bendall SC, Davis KL, Amir EAD, Tadmor MD et al 1. Jansen IE, Savage JE, Watanabe K, Bryois J, Williams DM, Steinberg S et al (2015) Data‑ driven phenotypic dissection of AML reveals progenitor‑like (2019) Genome‑ wide meta‑analysis identifies new loci and functional cells that correlate with prognosis. Cell 162:184–197 pathways influencing Alzheimer’s disease risk. Nat Genet 51:404–413 26. Hintze JL, Nelson RD (1998) Violin plots: a box plot‑ density trace syner‑ 2. Lambert JC, Ibrahim‑ Verbaas CA, Harold D, Naj AC, Sims R, Bellenguez C gism. Am Stat 52:181–184 et al (2013) Meta‑analysis of 74,046 individuals identifies 11 new suscepti‑ 27. Van Der Maaten L, Hinton G (2008) Visualizing data using t‑SNE. J Mach bility loci for Alzheimer’s disease. Nat Genet 45:1452–1458 Learn Res 9:2579–2605 3. Keren‑Shaul H, Spinrad A, Weiner A, Matcovitch‑Natan O, Dvir ‑Szternfeld 28. Höllt T, Pezzotti N, van Unen V, Koning F, Eisemann E, Lelieveldt B et al R, Ulland TK et al (2017) A unique microglia type associated with restrict‑ (2016) Cytosplore: interactive immune cell phenotyping for large single‑ ing development of Alzheimer’s disease. Cell 169(1276–1290):e17 cell datasets. Comput Graph Forum 35:171–180 4. Mathys H, Davila‑ Velderrain J, Peng Z, Gao F, Mohammadi S, Young JZ 29. Somarakis A, IJsselsteijn ME, Luk SJ, Kenkhuis B, de Miranda NFCC, et al (2019) Single‑ cell transcriptomic analysis of Alzheimer’s disease. Lelieveldt BPF et al (2021) Visual cohort comparison for spatial single‑ cell Nature 570:332–337 omics‑ data. IEEE Trans Vis Comput Graph 27:733–743 5. Zhou Y, Song WM, Andhey PS, Swain A, Levy T, Miller KR et al (2020) 30. Somarakis A, Van Unen V, Koning F, Lelieveldt BPF, Hollt T (2019) ImaCytE: Human and mouse single‑nucleus transcriptomics reveal TREM2‑ visual exploration of cellular microenvironments for imaging mass dependent and TREM2‑independent cellular responses in Alzheimer’s cytometry data. IEEE Trans Vis Comput Graph 1:1 disease. Nat Med 26:131–142 31. Meadowcroft MD, Connor JR, Yang QX (2015) Cortical iron regulation and 6. Lopes KO, Sparks DL, Streit WJ (2008) Microglial dystrophy in the aged inflammatory response in Alzheimer’s disease and APPSWE/PS1ΔE9 mice: and Alzheimer’s disease brain is associated with ferritin immunoreactivity. a histological perspective. Front Neurosci 9:255 Glia 56:1048–1060 32. Krasemann S, Madore C, Cialic R, Baufeld C, Calcagno N, El Fatimy R et al 7. Grundke‑Iqbal I, Fleming J, Tung YC, Lassmann H, Iqbal K, Joshi JG (1990) (2017) The TREM2‑APOE pathway drives the transcriptional phenotype Ferritin is a component of the neuritic (senile) plaque in Alzheimer of dysfunctional microglia in neurodegenerative diseases. Immunity dementia. Acta Neuropathol 81:105–110 47:566–581 8. Damulina A, Pirpamer L, Soellradl M, Sackl M, Tinauer C, Hofer E et al 33. Banerjee P, Paza E, Perkins EM, James OG, Kenkhuis B, Lloyd AF et al (2020) (2020) Cross‑sectional and longitudinal assessment of brain iron level in Generation of pure monocultures of human microglia‑like cells from Alzheimer disease using 3‑ T MRI. Radiology 296:192541 induced pluripotent stem cells. Stem Cell Res 49:102046 9. Spotorno N, Acosta‑ Cabronero J, Stomrud E, Lampinen B, Strandberg 34. Van Wageningen TA, Vlaar E, Kooij G, Jongenelen CAM, Geurts JJG, Van OT, van Westen D et al (2020) Relationship between cortical iron and tau Dam AM (2019) Regulation of microglial TMEM119 and P2RY12 immuno‑ aggregation in Alzheimer’s disease. Brain 143:1341–1349 reactivity in multiple sclerosis white and grey matter lesions is dependent 10. Bulk M, Kenkhuis B, van der Graaf LM, Goeman JJ, Natté R, van der Weerd on their inflammatory environment. Acta Neuropathol Commun 7:66 L (2018) Postmortem T2*‑ weighted MRI imaging of cortical iron reflects 35. Bennett ML, Bennett FC, Liddelow SA, Ajami B, Zamanian JL, Fernhoff NB severity of Alzheimer’s disease. J Alzheimer’s Dis 65:1125–1137 et al (2016) New tools for studying microglia in the mouse and human 11. van Duijn S, Bulk M, van Duinen SG, Nabuurs RJA, van Buchem MA, van CNS. Proc Natl Acad Sci USA 113:E1738–E1746 der Weerd L et al (2017) Cortical iron reflects severity of Alzheimer’s 36. Mildner A, Huang H, Radke J, Stenzel W, Priller J (2017) P2Y 12 receptor is disease. J Alzheimer’s Dis 60:1533–1545 expressed on human microglia under physiological conditions through‑ 12. Ayton S, Fazlollahi A, Bourgeat P, Raniga P, Ng A, Lim YY et al (2017) out development and is sensitive to neuroinflammatory diseases. Glia Cerebral quantitative susceptibility mapping predicts amyloid‑β‑related 65:375–387 cognitive decline. Brain 140:2112–2119 37. Zrzavy T, Hametner S, Wimmer I, Butovsky O, Weiner HL, Lassmann H 13. Ayton S, Wang Y, Diouf I, Schneider JA, Brockman J, Morris MC et al (2019) (2017) Loss of “homeostatic” microglia and patterns of their activation in Brain iron is associated with accelerated cognitive decline in people with active multiple sclerosis. Brain 140:1900–1913 Alzheimer pathology. Mol Psychiatry 66:1–10 38. Ayton S, Faux NG, Bush AI, Initiative ADN (2015) Ferritin levels in the cer‑ 14. Braak H, Braak E (1991) Neuropathological stageing of Alzheimer‑related ebrospinal fluid predict Alzheimer’s disease outcomes and are regulated changes. Acta Neuropathol 66:239–259 by APOE. Nat Commun 6:6760 15. Braak H, Braak E (1995) Staging of alzheimer’s disease‑related neurofibril‑ 39. Streit WJ, Xue QS, Tischer J, Bechmann I (2014) Microglial pathology. Acta lary changes. Neurobiol Aging Neurobiol Aging 16:271–278 Neuropathol Commun 2:1–17 16. Hyman BT, Phelps CH, Beach TG, Bigio EH, Cairns NJ, Carrillo MC et al 40. Connor JR, Menzies SL, St. Martin SM, Mufson EJ, (1992) A histochemi‑ (2012) National Institute on Aging‑Alzheimer’s Association guidelines cal study of iron, transferrin, and ferritin in Alzheimer’s diseased brains. J for the neuropathologic assessment of Alzheimer’s disease. Alzheimer’s Neurosci Res. 31:75–83 Dement NIH Public Access 8:1–13 41. Zeineh MM, Chen Y, Kitzler HH, Hammond R, Vogel H, Rutt BK (2015) Acti‑ 17. IJsselsteijn ME, Brouwer TP, Abdulrahman Z, Reidy E, Ramalheiro A, vated iron‑ containing microglia in the human hippocampus identified Heeren AM et al (2019) Cancer immunophenotyping by seven‑ colour Kenk huis et al. acta neuropathol commun (2021) 9:27 Page 15 of 15 by magnetic resonance imaging in Alzheimer disease. Neurobiol Aging 52. Nabuurs RJA, Hegeman I, Natté R, van Duinen SG, van Buchem MA, van 36:2483–2500 der Weerd L et al (2011) High‑field MRI of single histological slices using 42. Walker DG, Tang TM, Mendsaikhan A, Tooyama I, Serrano GE, Sue LI et al an inductively coupled, self‑resonant microcoil: application to ex vivo (2020) Patterns of expression of purinergic receptor P2RY12, a putative samples of patients with Alzheimer’s disease. NMR Biomed 24:251–357 marker for non‑activated microglia, in aged and Alzheimer’s disease 53. Dixon SJ, Stockwell BR (2014) The role of iron and reactive oxygen species brains. Int J Mol Sci 21:66 in cell death. Nat Chem Biol 25:9–17 43. Kaneko Y, Kitamoto T, Tateishi J, Yamaguchi K (1989) Ferritin immunohis‑ 54. Nakamura K, Kawakami T, Yamamoto N, Tomizawa M, Fujiwara T, Ishii T tochemistry as a marker for microglia. Acta Neuropathol 79:129–136 et al (2016) Activation of the NLRP3 inflammasome by cellular labile iron. 44. Jellinger K, Paulus W, Grundke‑Iqbal I, Riederer P, Youdim MB (1990) Brain Exp Hematol 44:116–124 iron and ferritin in Parkinson’s and Alzheimer’s diseases. J Neural Transm 55. McIntosh A, Mela V, Harty C, Minogue AM, Costello DA, Kerskens C et al Park Dis Dement Sect 2:327–340 (2019) Iron accumulation in microglia triggers a cascade of events that 45. Streit WJ, Braak H, Del Tredici K, Leyh J, Lier J, Khoshbouei H et al (2018) leads to altered metabolism and compromised function in APP/PS1 mice. Microglial activation occurs late during preclinical Alzheimer’s disease. Brain Pathol 29:606–621 Glia 66:2550–2562 56. Nnah IC, Lee C, Wessling‑Resnick M (2020) Iron potentiates micro ‑ 46. Nguyen AT, Wang K, Hu G, Wang X, Miao Z, Azevedo JA et al (2020) APOE glial interleukin‐1β secretion induced by amyloid‐β. J Neurochem and TREM2 regulate amyloid‑responsive microglia in Alzheimer’s disease. 154:177–189 Acta Neuropathol 1:3 57. Ising C, Venegas C, Zhang S, Scheiblich H, Schmidt SV, Vieira‑Saecker A 47. Ayton S, Janelidze S, Roberts B, Palmqvist S, Kalinowski P, Diouf I et al et al (2019) NLRP3 inflammasome activation drives tau pathology. Nature (2020) Acute phase markers in CSF reveal inflammatory changes in 575:669–673 Alzheimer’s disease that intersect with pathology, APOE ε4, sex and age. 58. Heneka MT, Kummer MP, Stutz A, Delekate A, Schwartz S, Vieira‑Saecker A Prog Neurobiol 47:101904 et al (2013) NLRP3 is activated in Alzheimer’s disease and contributes to 48. Acosta‑ Cabronero J, Betts MJ, Cardenas‑Blanco A, Yang S, Nestor PJ (2016) pathology in APP/PS1 mice. Nature 493:674–678 In vivo MRI mapping of brain iron deposition across the adult lifespan. J 59. Sankowski R, Böttcher C, Masuda T, Geirsdottir L, Sagar, Sindram E et al Neurosci 21:36 (2019) Mapping microglia states in the human brain through the integra‑ 49. Farrall AJ, Wardlaw JM (2009) Blood–brain barrier: ageing and microvas‑ tion of high‑ dimensional techniques. Nat Neurosci 22:2098–2110 cular disease—systematic review and meta‑analysis. Neurobiol Aging 60. Böttcher C, Schlickeiser S, Sneeboer MAM, Kunkel D, Knop A, Paza E et al 25:337–352 (2019) Human microglia regional heterogeneity and phenotypes deter‑ 50. Ward RJ, Zucca FA, Duyn JH, Crichton RR, Zecca L (2014) The role of mined by multiplexed single‑ cell mass cytometry. Nat Neurosci 22:78–90 iron in brain ageing and neurodegenerative disorders. Lancet Neurol 13:1045–1060 Publisher’s Note 51. Bulk M, Abdelmoula WM, Nabuurs RJA, van der Graaf LM, Mulders CWH, Springer Nature remains neutral with regard to jurisdictional claims in pub‑ Mulder AA et al (2018) Postmortem MRI and histology demonstrate dif‑ lished maps and institutional affiliations. ferential iron accumulation and cortical myelin organization in early‑ and late‑ onset Alzheimer’s disease. Neurobiol Aging 62:231–242 Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions

Journal

Acta Neuropathologica Communications – Springer Journals

Published: Feb 17, 2021

There are no references for this article.