Abstract
Background:
Neuritic plaques contain neural and microglial elements, and amyloid-β protein (Aβ), but their pathogenesis remains unknown.
Objective:
Elucidate neuritic plaque pathogenesis.
Methods:
Histochemical visualization of hyperphosphorylated-tau positive (p-tau+) structures, microglia, Aβ, and iron.
Results:
Disintegration of large projection neurons in human hippocampus and neocortex presents as droplet degeneration: pretangle neurons break up into spheres of numerous p-tau+ droplets of various sizes, which marks the beginning of neuritic plaques. These droplet spheres develop in the absence of colocalized Aβ deposits but once formed become encased in diffuse Aβ with great specificity. In contrast, neurofibrillary tangles often do not colocalize with Aβ. Double-labelling for p-tau and microglia showed a lack of microglial activation or phagocytosis of p-tau+ degeneration droplets but revealed massive upregulation of ferritin in microglia suggesting presence of high levels of free iron. Perl’s Prussian blue produced positive staining of microglia, droplet spheres, and Aβ plaque cores supporting the suggestion that droplet degeneration of pretangle neurons in the hippocampus and cortex represents ferroptosis, which is accompanied by the release of neuronal iron extracellularly.
Conclusion:
Age-related iron accumulation and ferroptosis in the CNS likely trigger at least two endogenous mechanisms of neuroprotective iron sequestration and chelation, microglial ferritin expression and Aβ deposition, respectively, both contributing to the formation of neuritic plaques. Since neurofibrillary tangles and Aβ deposits colocalize infrequently, tangle formation likely does not involve release of neuronal iron extracellularly. In human brain, targeted deposition of Aβ occurs specifically in response to ongoing ferroptotic droplet degeneration thereby producing neuritic plaques.
INTRODUCTION
Neurofibrillary tangles (NFTs) and neuritic pla-ques (NPs) are distinct neurodegenerative lesions associated with neuronal injury and cognitive impairment during aging and late-onset Alzheimer’s disease (LOAD) [1]. We have been studying the pathogenesis of LOAD by examining the sequential development of pathological lesions in the temporal lobe of non-selected, non-demented humans during preclinical stages. We have found that the sequence of lesion development begins with neurofibrillary degeneration (NFD) and microglial dystrophy, followed by amyloid deposition, neuroinflammation, and increased presence of iron [2, 3]. Our work, which is guided by the microglial dysfunction hypothesis, has suggested that at a critical point in lesion development when preclinical (non-symptomatic) LOAD transitions to clinical (symptomatic) LOAD, main features, such as NFD, amyloid-β (Aβ) deposition, microglial activation, and iron accumulation are converging. This transitional phase during lesion development likely coincides with the clinical entity known as mild cognitive impairment [3–5]. In the current study, we have sought to investigate further the cellular/molecular pathobiology that occurs during this transitional period by performing our postmortem studies in a cohort of individuals most of whom were clinically non-demented, but also included some individuals with clinical dementia. Thus, the current study was focused on cases of late preclinical and early clinical LOAD in order to decipher how NFD, Aβ deposition, microglial activation, and iron accumulation may interact to produce a classical end stage lesion, the NP.
NPs, formerly known as senile plaques, are of critical importance for AD diagnosis and staging [1, 6–9]. They are defined by an amyloid core that is surrounded by and intermingled with p-tau+ dystrophic neurites, as well as reactive microglial cells [6, 10]. It is thought that non-amyloid Aβ deposits (primitive plaques) may be able to convert into neuritic plaques [10], and that neuritic plaques are closely associated with neuronal injury [1], but details of the pathogenic mechanisms that lead to NP formation have remained largely unknown. The amyloid cascade hypothesis proposing that amyloid toxicity and incitement of neuroinflammation at Aβ plaques trigger neurodegenerative changes relies on the neuritic plaque as a key lesion in support but is otherwise problematic when it comes to the neuropathology of LOAD [11–15]. For example, while hypertrophied (activated) microglial macrophages can be present in NPs, they are not associated with other neurodegenerative structures, such as p-tau+ pretangles, NFTs or neuropil threads, and therefore most of the tau pathology observed in humans lacks simultaneous presence of activated microglia rendering moot the claim that neurodegeneration is the result of inflammation. Instead, non-activated and/or dystrophic microglia are commonly seen in association with all forms of tau pathology, including with NPs [3, 17]. Our work and that of many others suggests that aggregated, Congo red-positive amyloid, a foreign body, is the trigger for the characteristic microglial activation (neuroinflammation) observed, which consists of microglia clustered around individual amyloid cores [18–22]. This neuroinflammatory response directed at amyloid is unrelated to the initiation of tau pathology [23], and eventually resolves during more advanced disease leaving behind damaged and exhausted microglia [3, 25].
In the current study, we show that formation of NPs begins with the disintegration of large p-tau+ pretangle neurons in the hippocampus and cerebral cortex into droplet spheres and is followed by Aβ deposition, thus offering novel insights into NP development and AD pathogenesis. The formation of droplet spheres we describe here as originating from neuronal pretangles of mossy and pyramidal neurons is consistent with early immunohistochemical observations using AT8 antibody for studies of cytoskeletal changes associated with AD neurodegeneration [26]. Although the term droplet was not used in that particular study, Braak and del Tredici later on used it describing development of tau pathology in transentorhinal pyramidal neurons during preclinical LOAD [10]. According to these authors, p-tau+ droplets initially form intraneuronally during transformation of pretangle neurons into NFTs. Here we use the term droplet degeneration to distinguish a similar, yet different phenomenon whereby neuronal pretangles with droplets are not just transforming into intracellular or extracellular NFTs, as described by Braak et al. [26], but also into extracellular droplet spheres as a result of neuronal dissolution. The droplet spheres we describe are similar to earlier descriptions of tangle-associated neuritic clusters [27] or clusters of dystrophic neurites [28], yet in our study we did not directly colocalize droplet spheres (neuritic clusters) with NFTs, but show instead that both lesions are forming side-by-side. Our current results suggest that p-tau+ pretangle neurons can follow one of two paths involving either droplet or filamentous transformation. The former leads to formation of neuritic plaques while the latter results in the formation of NFTs.
Important components of NPs are p-tau+ neural processes, activated microglia, Aβ, and iron [6, 29–31], hence the current study was focused on these aspects to elucidate their participation in NP formation. Our findings show that during early stages of NP formation when droplet degeneration of p-tau+ pretangle neurons first occurs, both amyloid and microglial activation are absent, but iron is present, thus establishing a temporal sequence of events for the involvement of these components. We report that histochemically detectable iron is found in association with pretangles, droplet degeneration spheres, ferritin+ microglia, and amyloid cores, supporting a critical role for iron in AD neurodegeneration, i.e., ferroptosis [32, 33]. Our findings also stress the existence of physiological iron containment mechanisms, such as microglial ferritin expression and Aβ-mediated iron chelation, underscoring the importance of limiting iron neurotoxicity in the CNS.
MATERIALS AND METHODS
Subjects and case selection
Case selection was performed from a pool of approximately 1,000 individuals who came to aut-opsy in 2014–2018 as a matter of routine procedure following their deaths at the Leipzig University Hospital. In their individual contracts that govern medical treatment, upon admission all patients provided written consent to the scientific use of tissue removed and stored after any biopsy or during autopsy. A pro-cedural criterion used was a postmortem interval of less than 48 h. Following removal, whole brains were fixed by immersion in 4% buffered formaldehyde for approximately 2 weeks. Brains were released by the Department of Neuropathology after complete post-mortem evaluation by clinical neuropathologists. Selected brain regions, i.e., the medial temporal lobe, were dissected from 2 cm thick coronal slices and then embedded in polyethylene glycol, followed by sectioning at 100μm on a macrotome, as described previously. The cases listed in Table 1 were selected regardless of clinical dementia status and based on presence of tau pathology ranging from NFT (Braak) stage I to IV. In this way, we assembled a cohort of individuals who were likely to have been in a stage of LOAD that was within the transitional period from non-symptomatic to symptomatic disease. There were no rules for age or for absence/presence of Aβ. The cases in Table 1 are sorted according to increasing age. The majority of cases were individuals whose medical history, including clinical neuropathology evaluation, did not include mention of clinical dementia (CD); they were thus considered non-demented (ND) subjects.
Patient data (sorted according to age)
CD, clinical dementia; ND, non-demented. aNFT (neurofibrillary tangle) stages according to Braak et al., 2006 [35]. bAβ phases according to Thal et al., 2002 [37]. cFerritin staining score: ± occasional (rare) cells; + light staining of moderate numbers of cells; ++ widespread immunoreactivity; +++ widespread, intense immunoreactivity.
Immunohistochemistry
Microglial cells were labeled immunohistochemically using the rabbit polyclonal primary antibody, Iba1, directed against the ionized calcium binding adaptor molecule 1 (Synaptic Systems, Cat. No. 234013, diluted at 1 : 500). This antibody is known to bind to almost all microglial cells irrespective of their activation or degeneration state [34], and also works across several species. Immunolabeling for the iron storage protein, ferritin, was performed using anti-ferritin polyclonal antibody (rabbit anti-horse spleen ferritin, Sigma, F6136, diluted at 1 : 800). A monoclonal antibody against human PHF-tau, clone AT8 (mouse monoclonal, 1 : 2000, Thermo Fisher Scientific, MN1020, diluted at 1 : 500), was used for detecting structures containing hyperphosphorylated tau protein (p-tau), including neuropil threads, pretangles, neurofibrillary tangles, and neural components of neuritic plaques. Aβ was detected using a monoclonal mouse antibody (clone NAB 228, Sigma, A8354, diluted at 1 : 5000).
For immunostaining using peroxidase-based procedures, sections were incubated free-floating in blocking buffer containing 10% goat serum, 0.1% Triton-X 100 in PBS for 1 h before applying the primary antibodies. Primary antibodies, diluted in PBS containing 5% goat serum and 0.1% Triton in PBS, were applied to sections and incubated over-night at 4°C. After several washes, sections were incubated with either a biotinylated goat anti-mouse IgG (Sigma, B7264) or a biotinylated goat anti-rabbit IgG secondary antibody (Sigma, B8895), diluted 1 : 100, for 1 h at room temperature. Sections were then incubated with Avidin D conjugated to horseradish peroxidase (Vector, A-2004) and visualized using DAB H2O2 substrate generating a brown reaction product (Sigmafast 3,3’-diaminobenzidine tablet sets, Sigma, D-4418). Selected sections were counterstained with hematoxylin after immunostaining. Negative controls consisted of omitting either the primary antibodies or incubating primary antibodies with mismatched secondary antibodies (e.g., primary mouse with biotinylated goat-anti-rabbit). Double labelling for Iba1/AT8, ferritin/AT8, and Aβ/AT8 was performed sequentially by completing the substrate reaction (DAB for brown color) with the first primary antibody, followed by two brief (5 min) washes before application of the second primary antibody ending with Vector SG substrate for a black reaction product (Cat. No. SK-4700). Selected sections were counterstained with hematoxylin. All slides were dehydrated through ascending alcohols, cleared in xylene and coverslipped with Canada Balsam. Preparations were examined blindly (knowing only age, sex, and case number) and photographed using a Zeiss photomicroscope. All cases studied were staged for NFD [35, 36], and for phases of Aβ deposition [37], and received an ABC score. Scoring for ferritin was as described in prior studies [3], and as noted in the footnotes of Table 1.
Perl’s Prussian blue iron stain
Selected sections from cases showing intense ferritin immunoreactivity, were subjected to Prussian blue staining following completion of AT8 immunostaining with DAB substrate. Sections of PEG embedded tissue were cut at 30–50μm for this purpose since 100μm sections were found to be problematic. The sections were incubated free-floating in a 1 : 1 mixture of 1% HCl and 1% aqueous potassium ferrocyanide for approximately 30 min, rinsed briefly and differentiated, and then mounted and coverslipped.
Quantitative image analysis
Sections double stained with AT8 and anti-Aβ antibodies from four individuals were used to enumerate amyloid deposits, pretangles, NFTs, NPs, and overlays of Aβ and NFTs. Whole sections were fully digitized at 20x magnification using a digital slide scanner (Pannoramic Scan II, 3D HISTECH Ltd., Budapest, Hungary) in MRXS format. The scanner software (Pannoramic Scanner, version 1.23, 3D HISTECH Ltd., Budapest, Hungary) was operated in extended focus mode (30 levels with 1.2μm axial distance) to combine images from several adjacent focal planes into one image with maximum depth of sharpness (0.24μm pixel size). Annotations of polygonal ROIs (regions of interest: dentate gyrus, dentate hilus, CA3, CA2, CA1, subiculum, entorhinal cortex, perirhinal cortex, temporal cortex) and POIs (points of interest: Aβ deposits, neurofibrillary tangles, pretangles, neuritic plaques (AT8/Aβ double stained structures) and overlapping POIs (NFT/Aβ coexisting structures) were created using CaseViewer (Version 2.4, 3D HISTECH Ldt., Budapest, Hungary).
Annotation data were imported into Mathematica (Version 12.2, Wolfram Research Inc., Champaign, IL, USA) and all geometric coordinates were extracted. Tissue regions within ROIs were exported from MRXS data sets as PNG images using OpenSlide (https://openslide.org/). Images were imported into Mathematica, actual tissue was detected by global thresholding (removal of holes, incisions, etc.) and tissue area was calculated. The number of POIs per ROI were counted and respective POI densities (events per mm2) were calculated.
Statistical analysis was performed using GraphPad Prism (Version 9.1.1 for Windows, GraphPad Software, San Diego, California USA, http://www.graphpad.com). Descriptive statistics were calculated, bar charts and scatter plots were generated. All data were examined for normal distribution using Shapiro-Wilk test. Group comparisons were performed using the non-parametric Friedman test for related samples. Level of significance was set at p < 0.05. Post-hoc Dunn-Bonferroni correction was applied for multiple comparisons.
RESULTS
Immunostained sections of 100μm thickness were studied microscopically and photographed. Thick sections of this kind offer both advantages and disadvantages in that they produce good three-dimensional depth of the microscopic field offering greater detail, yet they also make it difficult to focus on multiple structures of interest at the same time. Therefore, some of the figures shown provide two focal planes to allow simultaneous visualization of more than one structure. In addition, 100μm sections show highly variable background staining depending on the staining protocol used and from one case to another, which is evident throughout the histopathology presented here.
All cases in our study showed considerable levels of tau pathology ranging from a minimum of NFT stage I to NFT stage IV. All cases, except case 1, contained Aβ deposits up to Thal phase A2. Figures 1 2 show p-tau+ pretangle pyramidal and mossy neurons in the hippocampus of non-demented individuals. All pretangles displayed normal neuronal morphology and uniform p-tau immunoreactivity throughout the somatodendritic compartment allowing good visualization of dendritic trees and dendritic spines. The latter were prominent in mossy cells of the dentate hilus revealing thorny excrescences that are characteristic of these neurons [38, 39]. In preparations from these same cases that were double-stained for p-tau and Aβ, neuronal pretangles in Ammon’s horn and in the dentate hilus were not colocalized with Aβ deposits, i.e., there was no direct overlap between p-tau+ and Aβ+ structures.
Neuronal pretangles in CA2/CA3 sectors of the hippocampus. a, b) AT8-positive hippocampal pyramidal neurons at low and high magnifications, respectively. Neuronal morphology appears normal revealing somata with dendrites. Note presence of numerous neuropil threads in the background of moderate density. Case 2, Table 1. c) Two AT8-positive pyramidal pretangles in CA2 show diffuse cytoplasmic p-tau and distinct labeling of the nuclear envelope (arrows). 84-year-old ND female (case 6). d) A single AT8-positive pretangle pyramidal neurons in CA3 shows cluster of dendritic spines on proximal dendrite (arrow); 81-year-old female (case 5). Scale bar: a 400μm; b,c 100μm; d 200μm
Neuronal pretangles in the dentate hilus of the hippocampus. a, b) AT8-positive mossy neurons at very low and high magnifications, respectively, in 89-year-old ND female (case 10). Dendritic spines are apparent in b. c) Elaborate clusters of dendritic spines on proximal dendrites form thorny excrescences characteristic of mossy neurons (arrow); 84-year-old ND female (case 6). d) Two focal planes (d1, d2) at high magnification of the same mossy neuron reveal detail of dendritic spines; 62-year-old ND female (case 2). Scale bar: a 800μm; b, c 100μm; d 50μm
Droplet degeneration was evident in all cases at NFT stages II and higher, with the notable exception of case 1, which had very few neuritic plaques but abundant NFTs (Table 1). Case 9 staged at NFT II, revealed large numbers of AT8+ hilar mossy neurons that were seen to be at various stages of dissolution (Fig. 3). These neurons were in the process of breaking apart and forming spheres of p-tau+ degeneration droplets. Some of the degenerating mossy neurons were only partially degraded in that dendritic extensions and spines could still be discerned, leaving little doubt about the origin of the droplets that were forming; other neurons had already completely disintegrated into droplet spheres. Slides from case 9 double stained with AT8 and Aβ antibodies revealed little direct colocalization between Aβ deposits and AT8+ degenerating neurons in the dentate gyrus (Fig. 4). The majority of Aβ deposits present were in the molecular layer (Fig. 4a), which showed faint and non-specific AT8 background staining and no discernable neuronal structures. A few diffuse Aβ deposits were present also in the hilus where they overlapped with some of the developing droplet spheres but not with AT8+ pretangles. (Fig. 4). However, not all droplet spheres colocalized with Aβ deposits and some remained without (Fig. 4b-d). The fact that many mossy neurons were only partially degraded in this individual (Figs. 3 4) indicated that droplet degeneration was at an early stage. Case 3 revealed early droplet degeneration occurring in the hippocampal CA1/2 sector in the absence of Aβ deposits (Fig. 5). The nearby entorhinal and temporal cortices within the same section revealed droplet spheres that were encased in diffuse or cored Aβ deposits and displayed features typical of NPs (e.g., Fig. 9g, h).
Droplet degeneration of pretangle mossy neurons in the dentate hilus of the hippocampus in 87-year-old ND female (case 9). a-c) Very low to intermediate magnifications show that all mossy cells in the hilus are at various stages of dissolution into droplets and fibrous elements. Incompletely degraded pretangle neurons with partially remaining dendrites are still identifiable. Disintegrating neurons produce numerous p-tau+ droplets. c1, c2) Slightly different focal planes of the same field reveal details of early droplet degeneration in the hilus. d-f) Higher magnification shows dissolution of pretangle mossy neurons into droplets (dashed boxes). An incompletely degraded neuron with partially remaining dendrite is shown in d (arrow). A long dendritic segment emanating from a droplet sphere and forming a knob-like ending is shown in e (arrow). Multiple remaining dendritic segments are shown in f (arrow). g) Very high magnification of a droplet sphere reveals countless droplets of varying sizes. AT8 immunostaining. Scale bar: a 800μm; b 400μm; c 200μm; d-f 100μm; g 50μm
Double staining with AT8 (black) and anti-Aβ (brown) in dentate gyrus reveals partial colocalization of degeneration droplets with Aβ. a) Low power view shows that most Aβ deposits are located in the molecular layer (arrows). b1, b2) Two focal planes of the same field show two droplet spheres, one is encased in Aβ (arrow), the other one devoid of Aβ (circle). c) Two droplet spheres are partially colocalized with diffuse Aβ (black arrows). Two partially degraded AT8+ mossy neurons do not show Aβ staining (white arrows). d) Aβ deposit is seen associated with two droplets; and an adjacent droplet sphere lacks Aβ (circle). Scale bar: a 400μm; b-d 100μm
Early droplet degeneration in CA1/2 of hippocampus occurs in the absence of Aβ deposits. a-c) Single staining with AT8. a) Multiple intracytoplasmic droplet-like accumulations of p-tau+ densities are seen in a pyramidal neuron. b) Large droplet shows remains of pyramidal cell soma (arrow) next to droplet sphere and degenerating dendrite. c) Degenerating p-tau+ axon emerges from droplet sphere (arrow). d-f) Double staining with AT8 and anti-Aβ shows lack of Aβ deposits over droplet spheres. d) Large portion of pyramidal soma containing unstained nucleus (arrow) is part of droplet sphere. e) Large fragment of neuronal soma (arrow) is part of droplet sphere. f) Multiple circular droplets comprise a droplet sphere. 70-year-old ND male. Scale bar: a-f 50μm
Double-label immunohistochemistry with AT8 (black) and anti-Aβ (brown) reveals specificity of Aβ deposition for droplet spheres. a, b) Entorhinal cortex reveals extensive tau pathology with NFTs and droplet spheres. Aβ staining is seen in association with droplet spheres only. c, d) Areas containing multiple NFTs show lack of overlap between NFTs and Aβ despite presence of numerous Aβ deposits. White arrow in c indicates overlap between Aβ and p-tau+ droplet sphere. d) Two droplet spheres colocalize with Aβ, two pretangles do not. e) Side-by-side occurrence of droplet sphere and NFT shows the specificity of Aβ encasement for droplet sphere. f) Early Aβ encasement (partial Aβ deposition) of droplet sphere. g, h) Temporal cortex. Detail shows p-tau+ dystrophic neurites and droplets in Aβ plaques with core. Case 10 (a-e); case 12 (f,g); Case 3 (h). Scale bar: a, c, d 200μm; b, e, f 100μm; g, h 50μm
In another individual (case 10) where preclinical pathology (ABC score) was more advanced than in case 9 (Table 1), droplet degeneration was also a prominent feature in the CA sectors of Ammon’s horn, subiculum, and entorhinal cortex (Fig. 6). Here widespread tau pathology was present, including countless neuropil threads, pretangles, NFTs, as well as droplet spheres. Figure 6 shows that numerous hippocampal pyramidal neurons were undergoing degenerative changes, which included acquisition of a primal droplet intracellularly within pretangles (Fig. 6d), formation of numerous p-tau+ droplets in and around pretangle pyramidal neurons (Fig. 6c), and transformation of pretangles into NFTs and droplet spheres (Fig. 6b). Some of these droplet spheres were quite large suggesting that more than one degenerating neuron contributed to their formation. Double staining with AT8 and Aβ antibodies showed that in case 10 most but not all droplet spheres had become encased with diffuse Aβ (Fig. 6e) and many had advanced to resemble classical neuritic plaques displaying droplets, as well as fibrous p-tau+ elements together with Aβ (Fig. 6f). These results suggested that neuronal pretangles progress in the degenerative process via droplet and/or filamentous transformation turning into either NPs or NFTs. The idea was supported by further examination of AT8/Aβ double labeling in other individuals (cases 7 and 1). The predominance of NFTs in case 1 is reflected in its ABC score showing a high NFT stage of III-IV in the absence of Aβ and NPs (Table 1). Close examination of neuronal pretangles in case 7 showed a heterogenous mixture of neurons that had progressed to various stages of degeneration. Many of them showed both droplet and filamentous elements (Fig. 7). Others developed predominantly droplets, and yet others showed predominantly fila-mentous changes. Regarding filamentous transformation, it is important to note that the intensely p-tau+ fibrils remained intracellular and covered by the neuronal membrane (Fig. 7e-g). Intracellular accumulations of dense p-tau+ accumulations coalesced into larger fibrous tangles. Pretangle neurons undergoing transformation into droplet spheres developed both intracellular and extracellular degenerative droplets arising from somata and dendritic projections (Fig. 7a-c). By the time complete droplet spheres had formed, neither the nucleus nor the membrane of pretangle neuron were discernable except in a few instances (e.g., Fig. 5d), leaving behind a sphere of degeneration droplets in the extracellular space. Figure 8 illustrates schematically the two modes of degenerative transformations of neuronal pretangles.
Droplet degeneration in hippocampal CA1/2 sectors occurs amidst pretangles, transforming into NPs and NFTs in 89-year-old ND female (case 10). a) Low magnification reveals great density of tau pathology consisting of neuropil threads, pretangles, NFTs, and droplet spheres. b) Droplet spheres (stars) occur side-by-side with pretangles and NFTs. c) A pyramidal neuron is undergoing transformation into a droplet sphere (dashed box). d) A pretangle neuron reveals singular droplet in somatodendritic compartment (arrow). AT8 immunostaining. e, f) Double staining for p-tau (black) and Aβ (brown) shows that most droplet spheres show classic NP morphology revealing colocalization of Aβ and p-tau within the plaque (black arrow in e). Only few droplet spheres remain without Aβ (white arrow in e). Scale bar: a 400μm; b, e, f 200μm; c, d 100μm
Hippocampal and cortical pyramidal neurons undergo droplet and filamentous transformations. a) Mossy neuron pretangle in dentate hilus showing extensive droplet formation on its distal dendrites (arrow). b) Pretangle neuron showing formation of a dendritic droplet (arrow). P-tau staining along soma’s edge is dense and filamentous. c1, c2) Two focal planes showing a pretangle with two perinuclear droplets (arrow) next to pyramidal neuron forming an NFT (star). 84-year-old CD female (case 7). d) Large numbers of NFTs are present in the temporal cortex of a 59-year-old ND male without Aβ deposits. e) Neuronal pretangle shows a large p-tau+ primal droplet in perinuclear soma (arrow). f) Dense accumulations of filamentous inclusions (arrows) remain inside a pyramidal neuron with intact plasma membrane. Transformation into NFT seems imminent. g) Well-developed intracellular NFT is seen together with primal droplet (arrow). Scale bar: a, d 100μm; b, c, e-g 50μm
Schematic summarizing acquisition of droplet and filamentous inclusions in pyramidal pretangle neurons and subsequent transformations into either droplet spheres or intracellular neurofilament tangles as precursors to NPs or NFTs, respectively. The transformations of pretangles occur in the absence of colocalized Aβ deposits.
The deposition of diffuse Aβ was often specific for the extracellular droplet spheres resulting from neuron dissolution, but NFTs were infrequently seen to be colocalized with Aβ deposits, i.e., there was rarely encasement by or direct overlap of NFTs with diffuse Aβ in double-stained preparations (Fig. 9). In many sections, Aβ deposits were widespread and seemingly distributed at random showing very little overlap with any p-tau+ structures (Fig. 9c). However, the Aβ deposits that colocalized with droplet spheres overlapped precisely with the spheres. Some droplet spheres displayed only partial encasement with Aβ suggesting that Aβ production was just beginning (Fig. 9f). Figure 9 shows examples from entorhinal and temporal cortices of several cases illustrating the specificity of Aβ encasement of droplet spheres. NFTs rarely colocalized with Aβ even when numerous deposits were in the vicinity (Fig. 9c). Many of the Aβ deposits present in areas without droplet spheres did not colocalize with any tau pathology at all. To validate the qualitative histological observations through objective quantification, we performed image analysis comparing densities of total NFTs and Aβ deposits (cumulative) with densities of Aβ deposits that overlapped with NFTs (“Overlap” in Fig. 10). The combined results from four cases (Fig. 10c) revealed a significant difference (p = 0.0049), validating histological observations that overlays of NFTs and Aβ deposits were rare and supporting the interpretation that they were likely accidental in nature.
Quantitative analysis of AD hallmark lesions shows significant difference between overlap and cumulative appearances of NFTs and Aβ deposits. a) Micrograph shows double-staining with AT8 (black) and anti-Aβ (brown) in a representative example of a section scanned for quantification. b) Overview of individual hallmark lesions counted: 1 Aβ deposit; 2 neurofibrillary tangle; 3 pretangle; 4 neuritic plaque; 5 Overlay NFT/Aβ deposit. c) Quantitative analysis of AD lesions represented as event types. The five histograms on the left display case-specific events relative to the total area of scanned brain slices evaluated. Clear overrepresentation of Aβ deposits+ NFTs (cumulative) versus overlaps is shown for all analyzed brain sections (case 7 was analyzed in two samples from both rostral and caudal hippocampal levels). The scatter plot on the right shows the distribution of individual events for all cases. Since a rostral and a caudal brain slice were evaluated from case 7, the mean value of both was plotted. Again, relative values were calculated (absolute frequencies related to the evaluated area). Friedmann test between Cumulative and Overlap shows a significant difference with p = 0.0049. d) Numerical values of lesion densities (events/mm2) for each case based on the respective events.
The finding that pretangle neurons undergo frank neuronal disintegration and dissolution into droplet spheres prompted questions regarding involvement of microglial cells, as they have long been known to phagocytize dead neurons [40]. We therefore per-formed double labeling with AT8 and Iba1 to address this issue and the results are shown in Fig. 11. In every instance of successful double-labeling there was no morphologically apparent phagocytic res-ponse of Iba1+ microglia to any p-tau+ structures, including droplet spheres, pretangles, or NFTs (Fig. 11). In contrast, sections double labeled with AT8 and anti-ferritin revealed a striking microglial reaction. Figure 12a-d show that p-tau+ droplet spheres and surrounding territories were inundated with numerous ferritin-positive microglial cells displaying enlarged, ferritin+ cell bodies. This display of microglial hypertrophy suggests that the cells were producing large amounts of ferritin to sequester free iron possibly internalized via phagocytosis of Aβ-iron aggregates. Further study using Aβ/ferritin double-labeling confirmed colocalization of Aβ with ferritin by showing that amyloid cores were ferritin+, and that ferritin+ microglia contained Aβ (Fig. 12e-g). In order to correlate p-tau immunoreactivity with presence of iron we performed staining with Prussian blue (Fig. 13a-d). We found the blue reaction product indicative of redox-active iron to be associated with pretangles (where it was largely obscured due to intensely dark AT8 staining), droplet spheres, amyloid plaque cores, and microglia (Fig. 13 e,f) demonstrating close congruence with ferritin staining in these cells.
Double-label immunohistochemistry with AT8 (black) and Iba1 (brown). a–c) Dentate hilus shows multiple AT8+ mossy neurons at various stages of dissolution into droplet spheres, as well as numerous, regularly spaced and ramified Iba1+ microglia. Microglia are distributed evenly throughout the area and do not form clusters to indicate ongoing phagocytosis. 87-year-old ND female. d) A droplet sphere is intermingled with and surrounded by ramified, non-activated microglia in CA1. 89-year-old ND female. e, f) AT8+ mossy neuron pretangles are surrounded by Iba1+, non-activated microglia. 84-year-old ND female. Scale bar: a 400μm; b 200μm; c-f 100μm
Double-label immunohistochemistry with AT8 (black) and anti-ferritin (brown) illustrates association between droplet spheres and ferritin+ microglia in CA1 hippocampus. a–d) All panels show a droplet sphere in the center and numerous NFTs in surrounding areas. Microscopic fields are populated by numerous ferritin+ microglia, many of which show hypertrophy (arrows in a-d). 89-year-old ND female. e-g) Double-label immunohistochemistry with ferritin (black) and Aβ antibodies (brown) in entorhinal cortex. Ferritin+ microglia make up the core of an amyloid plaque (stars in e, f). Ferritin+ microglial cell with enlarged cell body reveals intracellular Aβ immunoreactivity (arrows in f, g), suggesting phagocytosis of Aβ. 89-year-old CD male. Scale bar: a–f 100μm; g 50μm
Localization of iron using Prussian blue staining. a) P-tau+ neuronal pretangle shows faint intracytoplasmic Prussian blue. b) Large droplet sphere shows p-tau+ debris intermingled with blue iron staining. AT8 immunostaining. c, d) Aβ deposits reveal blue iron staining in their cores. Aβ immunostaining. e, f) Dystrophic microglia are positive with Prussian blue. Entorhinal cortex, 89-year-old ND female. Scale bar: a 100μm; b-f 50μm
DISCUSSION
The current study provides novel insights into the formation of neuritic plaques by detailing a sequence of events that begins with p-tau+ pretangle neurons transforming into either NFTs or spheres of p-tau+ degeneration droplets, showing that both NFTs and NPs originate from neuronal pretangles. The transformation of pretangles into droplet spheres represents obvious neuronal disintegration and therefore neuron loss; it provides perhaps the most direct evidence for neuronal death in AD. Neuronal droplet degeneration differs fundamentally from NFT formation in that it is linked to deposition of Aβ, which selectively encases droplet spheres but not NFTs. This suggests that targeted Aβ deposition occurs in response to neuronal pretangles degenerating into droplet spheres. The concomitant presence of hypertrophic ferritin+ microglia and iron-containing amyloid plaques suggest that microglial ferritin expression and Aβ deposition represent endogenous iron-containment mechanisms designed to limit the extracellular spread of free iron resulting from ferroptotic droplet degeneration. The associated release of iron together with p-tau likely creates potentially toxic neuronal debris that triggers selective phagocytosis in ferritin+ but not Iba1+ microglia, implying failure or inability of non-ferritin+ microglia to remove this material. Deposition of diffuse Aβ may serve to limit iron toxicity via chelation. At the same time, extracellular iron contributes to aggregation of diffuse (soluble) Aβ into insoluble amyloid [14, 41–43], which may induce selective microglial phagocytosis of iron-containing amyloid aggregates. The final outcome of these interactions is reflected in the composition of neuritic plaques, well-known end stage lesions in AD. The formation of NFTs may represent yet a third mechanism of iron containment via iron entombment. A summary of our findings and conclusions about an iron-driven sequence of events in AD neurodegeneration is shown in Fig. 14.
Summary diagram showing proposed cellular mechanisms underlying AD neurodegeneration. Iron is a critical component at all stages: pretangles, NFTs, droplet degeneration (ferroptosis), and NPs. Gradually developing iron overload in the aging CNS is counteracted continuously by ferritin-mediated sequestration within microglial cells.
Widespread neuronal loss has not been known as a prominent feature of sporadic AD [44], yet the current findings reveal frank neuronal dissolution during the transition from late preclinical AD to clinical dementia demonstrating obvious neuronal death and therefore loss. We show that the initial neuronal disintegration occurs in the absence of colocalized Aβ deposits on site, and our findings clearly point towards iron overload within neurons as the likely cause of neuronal death, i.e., ferroptosis [32, 33]. Ferroptosis is a form of cell death that involves membrane rupture [33], and it is different from the formation of NFTs where the neuronal membrane remains intact during filamentous transformation. We show that formation of NFTs and ferroptotic droplet degeneration can occur simultaneously in neighboring hippocampal neurons, suggesting that levels of redox-active iron are greatly increased in regions with this dual pathology. Our observations revealing an abundance of ferritin+ microglia during, and likely prior to ferroptosis further implicate aging-related iron overload as a key event.
NFTs and NPs were previously shown to contain redox-active iron in humans with AD using a histochemical modification of Prussian blue staining similar to the one employed here [45]. Our current protocol of Prussian blue did not result in discernable blue staining of NFTs, but Smith et al. [45] showed iron to be present in NFTs, as well as in droplet spheres which they termed neuritic plaques. They showed neuritic plaques and NFTs occurring side-by-side in the hippocampus, a scenario very similar to what we show here using p-tau immunostaining (Fig. 6). These histopathological findings suggest that all p-tau+ structures contain ferric iron (Fe+3), which may account for tau phosphorylation by controlling activity of kinases [46], as well as for tau aggregation and toxicity [47–49]. Iron and associated oxidative stress have long been implicated as critical factors in AD pathogenesis [14, 50–53], and have provided an alternative to a widely perceived dominant role of Aβ in AD neurodegeneration [13, 54]. This alternate point of view is supported strongly by the experimental work of Bishop and Robinson in rats who showed that intracerebral injections of iron, as well as co-injections of iron and Aβ, caused significantly greater damage and neuronal loss than injections of Aβ alone [55, 56]. They also found that co-injections of iron and human Aβ were less toxic than co-injections of iron and rat Aβ (which was the same as iron alone), prompting them to conclude that Aβ deposition may be a neuroprotective response to elevated iron, and NPs a reflection of such neuroprotection. The current findings in human brain showing selective encasement by Aβ of iron-containing droplet spheres (Figs. 4, 9) strongly support the conclusions drawn from these experiments and offer in situ evidence for Aβ-mediated neuroprotection [57]. The work by Bishop and colleagues also showed that neurons normally accumulate histochemically detectable iron intracellularly and at concentrations much higher than present extracellularly, and that among all CNS cell types microglia are the most efficient at iron accumulation [58, 59]. These findings fit well with our current and previous demonstrations of ferritin expression in microglia, presence of iron in neuronal pretangles, droplet spheres, and in nascent neuritic plaques. They support our assertion that neuronal dissolution into droplet spheres results in the release of neuronal iron extracellularly accompanied by its rapid sequestration into ferritin-expressing microglia, and subsequent chelation by Aβ. The mechanisms determining whether neurons overloaded with iron become NFTs or undergo ferroptosis remain unknown at this time. We hypothesize that intraneuronal iron concentrations and/or ratios of ferric over ferrous ion may be critical in this regard.
A consistent finding from several individuals in our cohort (Table 1) was that the formation of NPs was at different stages in neighboring, as well as within the same brain regions. Early ferroptotic droplet degeneration of pyramidal or mossy neurons in the hippocampus revealed large irregular droplets and the remains of partially degraded neuronal somata, showing that neuronal breakup was occurring at the time the brain was preserved in fixative (Figs. 3, 5). At that same moment in time, some of the droplet spheres in the CA1,2 hippocampus were already beginning to become encased with Aβ, but others were not (Figs. 4–6, 9), suggesting that encasement happens gradually but soon after neuron dissolution. The adjacent entorhinal and temporal cortices in the same section revealed that here nearly all spheres containing smaller droplets and dystrophic neurites were fully encased with diffuse or dense core Aβ deposits (Figs. 4, 9), the latter being coincident with the presence of Congo-red+ amyloid. These events were especially evident in one individual at NFT stage III-IV characterized by presence of abundant NPs, high ferritin expression, and Thal Aβ phase A1, i.e., absence of Aβ in the hippocampus (case 3, Table 1). The observations suggest that ferroptotic neuronal death is a fleeting event that rapidly triggers subsequent deposition of Aβ. A different situation was present in the youngest individual in our cohort (case 1, Table 1), found to be also at NFT stage III-IV but showing a predominance of NFTs with very few NPs, minimal ferritin expression, and no Aβ deposits (A0). Tellingly, the striking difference in ferritin and Aβ expression between cases 1 and 3 appears to correlate with presence/absence of NPs, further supporting ferroptosis as the mechanism of droplet degeneration and subsequent NP formation. Since case 1 represents the youngest individual in our cohort, one is tempted to speculate that low levels of iron were related to the relative youth of this individual, since brain iron is known to increase with aging [42]. In case 1, the proposed mechanism of permanent entombment of iron in NFTs may thus be adequate for controlling increased levels of intraneuronal iron not high enough to cause neuronal breakup.
The association of iron and ferritin-expressing microglia within neuritic plaques has long been known, yet the contributions of iron and microglia to AD neurodegeneration have remained unclear [29, 61]. Current results with Prussian blue provide affirmation that ferritin expression coincides with presence of intracellular ferric ion in microglia (Fig. 13) [53], supporting our contention that ferritin expression in microglia serves a neuroprotective role through iron sequestration by microglia, which however has adverse effects on microglia themselves, reflected morphologically by dystrophy [2, 62]. Dystrophic (senescent) microglia can be seen in association with neuritic plaques [3, 16], and p-tau has been implicated in promoting microglial degeneration in human brain [63]. Our current and previous findings support this idea, but they also include iron and amyloid [3] in promoting microglial senescent degeneration. Earlier work from our laboratories had shown that exposure of cultured rodent microglia to isolated human amyloid cores, known to be rich in iron [41] (Figs. 12 13), promotes telomere shortening indicative of microglial senescence [25]. Thus, high levels of iron are toxic to neurons, as well as to microglia, and likely to be the main driving force in promoting both neuronal and microglial degeneration in LOAD. Our current findings suggest that ferritin expression and iron sequestration in microglia occurs in a specific subpopulation of microglia displaying cytoplasmic hypertrophy. This set of cells is therefore likely to be reactive, while the majority of Iba1+ (and ferritin-negative) cells are non-hypertrophic (ramified, resting) microglia. We attribute microglial hypertrophy to massive uptake and storage of the extracellular iron released suddenly by neuronal ferroptosis. Iron uptake by microglia may be partially linked to uptake of amyloid, since we were able to observe Aβ staining in dystrophic, ferritin+ microglia (Fig. 12). However, such intracellular staining of amyloid within microglia is a rare event we have observed only once, which suggests a limited ability of microglia to retain ingested amyloid and degrade this insoluble material [3, 65]. The fact that we did not observe Aβ staining within Iba1+ microglia further supports this thought. Ferritin expression in microglia occurs before droplet spheres and/or neuritic plaques form, as shown in the current study, as well as in our prior work [3] that included numerous non-demented individuals where single (un-clustered) ferritin+ microglia were present in the absence of neuritic plaques and Aβ deposits. This suggests that microglia begin sequestering iron early as brain iron overload develops gradually with increasing age. Microglia are therefore the first line of defense against rising iron levels in the brain, a notion consistent with current understanding of microglial neuroprotection [15, 66].
Our findings show that formation of NPs begins with neuronal death occurring in the absence of colocalized Aβ deposits. This is contrary to the assumption that NPs form by conversion of diffuse Aβ plaques into neuritic ones, a process thought to be mediated by neuroinflammation (activated microglia) [30, 67–70]. This conversion of diffuse Aβ deposits into NPs has not been shown but rather assumed to occur in humans. In transgenic mouse models overexpressing Aβ and forming abundant Aβ plaques, and also showing robust microglial activation [63, 71], formation of p-tau+ neuritic plaques does not occur. Furthermore, it is well documented that during preclinical stages of AD, tau pathology precedes Aβ deposition by many years [12, 72]. Our current findings not only confirm this temporal relationship revealed in large cross-sectional population studies, they also offer a possible explanation as to why Aβ deposition occurs in the first place, i.e., they suggest a neuroprotective role for Aβ, which is opposite to its alleged role as a neurotoxic inducer of tau pathology [73, 74]. Aβ-mediated neuroprotection may be provided in that Aβ helps contain the toxic spill of iron and p-tau caused by ferroptosis. Iron chelation by Aβ works concurrently and synergistically with microglial iron sequestration via ferritin. The current observations are supportive of previously expressed views by others suggesting that Aβ serves as a neuroprotective bioflocculant [57].
We show that Aβ encasement of iron-containing droplet spheres occurs in a highly localized fashion sharply demarcating droplet degeneration spheres, raising questions about cellular sources of Aβ. Multiple hypotheses regarding the locus of Aβ production have been put forth [75]. In light of the current findings, the one hypothesis that seems to be most relevant is neuronal lysis (see Fig. 2e in [75]). However, all six hypotheses proposed by Fiala [75] have in common the idea that amyloid is neurotoxic, either directly or indirectly via microglial activation, and that it causes neurofibrillary degeneration secondarily. The current study clearly does not support the idea of neurotoxic Aβ being a primary event in AD neurodegeneration.
We have previously determined that microglial activation and attempted phagocytosis of Aβ is initiated only when Aβ aggregates and transforms into insoluble amyloid [3]. We show here that droplet spheres without concurrent Aβ presence do not elicit a phagocytic response by microglia. This apparent failure of microglia to phagocytize p-tau+ neuronal debris was unexpected because in situations of experimentally induced neuronal death, linked to increased iron uptake [76], microglia respond vigorously to neuronal degeneration by activation and transformation into hypertrophic brain macrophages that avidly phagocytize neuronal debris [40]. The fact that this macrophage response does not occur with p-tau+ debris suggests a number of possibilities, for example, failure of microglia to recognize p-tau+ structures, toxicity of p-tau+ structures which may deter microglia not expressing ferritin, or perhaps microglial phagocytic anergy. Toxicity and phagocytic anergy both seem likely because tau toxicity has been suggested by other findings [63], and defective phagocytosis is thought to play a role in AD [77, 78]. Impaired microglial phagocytosis of amyloid is also suggested by the current results showing that Aβ internalization within microglia is a rare event in the AD brain.
In conclusion, the current study offers evidence supporting a role for excess iron rather than for Aβ in the initiation of LOAD neurodegeneration. Aβ deposits are an accompaniment or side effect of neurodegeneration, as stated by Alzheimer himself more than a hundred years ago [15, 79], and the role of Aβ in AD might be more akin to neuroprotection than to neurotoxicity. Similarly, the role of activated microglia and an allegedly neurotoxic neuroinflammatory response, frequently cited as a potential cause of neurodegeneration, is instead a natural cellular response to neuronal death by ferroptosis and related amyloid production.
Footnotes
ACKNOWLEDGMENTS
The authors wish to thank Angela Ehrlich and Heidrun Kuhrt for their expert technical assistance. The work was supported by the Deutsche Forschungsgemeinschaft (DFG) Collaborative Research Center 1052-A9. We also thank the Cooper family of Indialantic, FL for their support of dementia research.
