Disciplinary Silos in Alzheimer's Disease Research, 1950–2020

Abstract

Why did the assembly of a tripartite mechanistic thesis for Alzheimer's disease — combining locus-coeruleus / NAD+ bioenergetic vulnerability (Phase I), microglial and oligodendrocyte homeostatic collapse (Phase II), and perineuronal-net / extracellular-matrix erosion (Phase III) — require seventy years, when each of its components had been independently described in the published literature by the mid-1980s? This paper offers a structural, evidence-based answer. Drawing on the Oskar Fischer Prize entrant corpus (151 independent submissions, 135 audited against a 235-mechanism trilogy registry), the entrant-corpus convergence on lysosomal/v-ATPase failure (a reproducible bracket of 17–27 of 30 entrants depending on whether the criterion is the named pump or the pathway, from seven distinct upstream triggers), the published trilogy-relevance audit, an original query of the NIH RePORTER award record (FY2000–2020), and a wider survey of the AD literature, we trace the disciplinary geography of three communities — noradrenergic systems neuroscience, glial / immunological neurobiology, and matrix / critical-period plasticity research — and show that they barely intersected with mainstream AD research from 1950 to 2018. Concurrently, the amyloid-cascade community absorbed disproportionate funding, journal pages, and prestige, generating a citation gravity well that distorted the field's epistemic landscape. Our argument is diagnostic, not accusatory. Citation patterns, NIH funding allocations, journal segmentation, and training-pathway separation are presented as the structural conditions under which a real, multi-stream signal could be detected only by 151 independent investigators working outside the consensus and only later assembled into a coherent thesis. The paper closes with counterfactual analysis (the trilogy was assembleable by 2005 had cross-silo dialogue been institutionalised), a scenario-based estimate of the cumulative cost of delay (5–98 million person-years of dementia, depending on assumed treatment coverage), and a structural-reform agenda that the Oskar Fischer Prize itself prefigures. The Collapse Trilogy thesis is rare because the conditions for its assembly are rare. Building those conditions deliberately, this paper argues, is the strongest legacy the Fischer Prize program can leave.

Keywords: disciplinary silos, bibliometrics, sociology of science, Alzheimer's disease, amyloid cascade, locus coeruleus, microglia, perineuronal nets, Oskar Fischer Prize, citation analysis, NIH funding, science policy.

Competing interests. This paper evaluates the Oskar Fischer Prize favourably, concluding in Chapter VII that its institutional structure "worked." The prize was endowed by Dr. James Truchard, who is a co-author of the wider research programme within which this paper was written and to which its corpus belongs; the author is that programme's other principal. Readers should weigh the assessment of the prize accordingly. The bibliometric findings of Chapters V–VI do not depend on it: they stand or fall on the entrant corpus, which is available for independent audit.

---

Contents

Front Matter Abstract · Contents · Note on Method

1. Introduction: Why a Tripartite Thesis Took Seventy Years

Chapter I. The Amyloid Cascade as Hegemon 1.1 The 1992 Inflection Point 1.2 RFAs and the Funding Capture, 1998–2018 1.3 The "Amyloid Mafia" Critique 1.4 The Negative Trial Sequence 1.5 The Lecanemab–Donanemab Apotheosis

Chapter II. The Noradrenergic Silo 2.1 The Locus Coeruleus Research Program, 1950–2010 2.2 Braak & Del Tredici 2011: The Paper That Should Have Ended Hippocampus-Centrism 2.3 Mather, Sara, Aston-Jones: Citations Out of the Mainstream 2.4 NAD+ and the Bioenergetic Stream

Chapter III. The Glial Silo 3.1 Microglia in Immunology, Not Neurology 3.2 The TREM2 / DAM / PAM Eruption, 2013–2020 3.3 Oligodendrocytes, Myelin, and the Forgotten Stream

Chapter IV. The Matrix Silo 4.1 Perineuronal Nets and Critical-Period Plasticity 4.2 The 2020–2025 PNN/AD Crossover 4.3 ECM and MMP Biology Across Stroke and Multiple Sclerosis

Chapter V. The Entrant Corpus as Evidence 5.1 The Fischer Prize Entrant Corpus 5.2 Phase-Overlap Across the Entrant Corpus 5.3 The Endolysosomal Convergence and Its Uncoordinated Character 5.4 NIH Funding Patterns, 2000–2020

Chapter VI. Why Convergence Happened Anyway 6.1 The Seven Upstream Triggers 6.2 Independence as Evidence of Signal

Chapter VII. The Cost of Silos 7.1 Counterfactual Reconstruction 7.2 Cumulative Person-Years of Avoidable Decline

Chapter VIII. Structural Reforms 8.1 Cross-Silo Reviewer Pools 8.2 Mechanism-Agnostic RFAs 8.3 Open Synthesis Platforms 8.4 The Oskar Fischer Prize as Prototype

Chapter IX. Beyond Alzheimer's Disease 9.1 Parkinson's Disease 9.2 Amyotrophic Lateral Sclerosis 9.3 Schizophrenia and Autism

Conclusion. Silos as Designed Outcomes

References

---

Note on Method

This paper is structurally and bibliometrically diagnostic, not historiographic in the narrative sense. The methodology proceeds through four registers. First, archival reconstruction of NIH RFA priorities and journal review patterns is taken from publicly accessible sources (NIH RePORTER, journal editorial statements, congressional testimony) and from secondary historical accounts (Lock, 2013; Whitehouse & George, 2008; Whitehouse, 2014). Second, the Oskar Fischer Prize entrant corpus is treated as a natural experiment in distributed inquiry: 151 independent investigators, working without coordination, generated mechanistic hypotheses for AD between 2022 and 2026. The trilogy-relevance audit (Gustafsson, 2026a) classifies 135 of these against a 235-element registry of canonical mechanisms across the three Collapse Trilogy domains (Phase I bioenergetic, Phase II homeostatic, Phase III synaptic). Third, the synthesis report (Gustafsson, 2026b) summarises the qualitative finding that twenty-five of thirty graded entrants converge on a shared downstream node — lysosomal acidification failure / v-ATPase disassembly — from seven distinct upstream triggers. Fourth, where appropriate we draw on standard bibliometric methods (Bornmann & Daniel, 2008; Mingers & Leydesdorff, 2015) to characterise community structure and citation flow. We make no claim of new empirical bibliometrics; rather, we treat the Fischer corpus as the load-bearing dataset and the secondary literature as scaffolding.

The argument is offered in a diagnostic register. The aim is not to indict individual scientists, programme officers, or journal editors. The aim is to identify the structural features — funding incentives, journal segmentation, training pathways, citation conventions, peer-review norms — that produced and sustained the silos, and to specify reforms that would not require any individual to behave heroically.

---

1. Introduction: Why a Tripartite Thesis Took Seventy Years

The dominant question in the history of Alzheimer's research — historiographically, scientifically, and socially — is not which hypothesis turned out to be correct. It is why the field's center of gravity remained immovable for so long in the face of accumulating contrary evidence. This paper proposes a structural answer: the field's epistemic geography was not a single landscape but an archipelago. The Collapse Trilogy thesis (Gustafsson, 2026c, 2026d, 2026e) — synthesising locus-coeruleus (LC) and NAD+ bioenergetic vulnerability (Phase I), microglial / oligodendrocyte homeostatic collapse (Phase II), and perineuronal-net (PNN) and matrix-metalloproteinase (MMP)-mediated synaptic erosion (Phase III) into a single mechanistic picture — was constructible from existing literature by the mid-1980s. It was not constructed until 2025.

The temporal gap is striking. By 1985, the laboratory had in hand: (i) Mann, Yates, and Marcyniuk's morphometric demonstration that LC neurons are catastrophically depleted in AD (Mann, 1983; Bondareff, Mountjoy, & Roth, 1981); (ii) the foundational characterisation of microglia as a CNS-resident immune lineage (del Río-Hortega, 1932; Streit, 1995, drawing on a tradition that long predated his synthesis); (iii) Brückner, Celio, and others' description of perineuronal nets surrounding parvalbumin interneurons (Celio, 1986); and (iv) the autophagy-lysosomal vacuole accumulation phenotype that Suzuki, Terry, and others had documented in AD ultrastructure since the 1960s (Suzuki & Terry, 1967; later Nixon, 2005). Each of these threads, individually, was a candidate structural hypothesis for AD. None was integrated.

This paper asks a single question: why?

The answer, we argue, lies in the disciplinary topography of late-twentieth-century neurodegeneration research. The amyloid-cascade hypothesis (Hardy & Higgins, 1992) acted not merely as a scientific proposition but as an organising attractor — a citation gravity well that pulled funding, prestige, journal pages, and trainees into a single explanatory frame, while the rival communities (LC, glial, matrix) remained loosely coupled enclaves whose members rarely sat on each other's grant panels, rarely co-authored, and rarely cited each other's work. This was not conspiracy; it was the predictable equilibrium of a field whose institutions reward concentration. The Fischer Prize corpus is a natural experiment that demonstrates the consequence: when 151 independent investigators are given permission to ignore the consensus, they converge on a coherent, multi-stream picture that the consensus framework cannot accommodate.

The paper proceeds as follows. Chapter I establishes the amyloid cascade's hegemony as a structural fact. Chapters II–IV map the three silos that the trilogy thesis spans. Chapter V presents the corpus evidence: the Fischer entrant corpus, the mechanistic overlap between its phase communities, and its convergence on a shared downstream node from seven distinct upstream triggers. Chapter VI argues that this uncoordinated convergence is the strongest possible evidence that the silos were preventing the recognition of a real signal rather than constructing a false one. Chapter VII estimates the cumulative cost. Chapter VIII proposes structural reforms. Chapter IX generalises beyond AD. The Conclusion situates the Oskar Fischer Prize as one prototype of the institutional forms a post-silo field would require.

---

Chapter I. The Amyloid Cascade as Hegemon

1.1 The 1992 Inflection Point

Hardy and Higgins published "Alzheimer's Disease: The Amyloid Cascade Hypothesis" in Science in April 1992 (Hardy & Higgins, 1992). The paper was eight hundred words. It proposed, on the strength of familial AD genetics (APP mutations linked to early-onset disease), that amyloid-beta deposition is the proximate cause of AD and that all subsequent pathology — tangles, synapse loss, neurodegeneration, dementia — is downstream. The claim was modest in its empirical content (the data cited would not have supported the strong conclusion); the claim was immodest in its programmatic implications. It offered the field something it badly needed: a single causal mechanism, a single therapeutic target, a single biomarker.

The field accepted the offer. Over the following decade the Hardy–Higgins paper became one of the most heavily cited hypothesis papers in the field, and its framing propagated well beyond the citations themselves. It is not the single most-cited paper in the literature — the most-cited works here are generally instruments and criteria rather than hypotheses, the NINCDS-ADRDA diagnostic criteria and the Braak staging scheme among them. What matters to the argument is not the rank but the propagation, with downstream amplification through review articles, textbook chapters, and grant introductions whose obligatory opening sentence — "Alzheimer's disease is characterised by the deposition of amyloid-beta…" — established the cascade as the field's grammatical default. This was not a neutral choice. It systematically forclosed alternatives by constraining what counted as a legitimate research question.

1.2 RFAs and the Funding Capture, 1998–2018

The funding record is the most concrete manifestation of the cascade's hegemony. NIH issued a series of Requests for Applications (RFAs) over twenty years that explicitly named amyloid-related mechanisms — APP processing, gamma-secretase, beta-secretase, immunotherapy targeting Abeta, anti-aggregation small molecules — as priority targets (NIH, 1998; NIH, 2003; NIH, 2008; NIH, 2014; reviewed in Whitehouse, 2014; Lock, 2013). Concurrently, no RFA in this twenty-year window named the locus coeruleus, NAD+ metabolism, microglial homeostasis, perineuronal nets, oligodendrocyte myelination, or matrix metalloproteinases as a priority target for AD research. (Several such RFAs existed for other disorders — MMPs for stroke, PNNs for amblyopia, microglia for HIV-associated neurocognitive disorder — which is a key part of the silo story we develop in Chapters II–IV.)

The dollar magnitude was substantial, and it is set out with its method in §5.4, where the NIH RePORTER record is queried directly rather than characterised. The gap is real, but for the nearest categories it is not an order of magnitude: tau drew 1.7× less than amyloid, and microglia 3.4× less.

The material claim that survives is narrower, and it is the one §5.4 supports: funding concentrated on the two lesions visible in a stained section of end-stage cortex, and thinned to a fortieth on the mechanisms operating before there is anything to stain. The cascade's hegemony was not merely intellectual but material — but its shape is a boundary between the visible and the invisible, not a simple ratio against all comers.

1.3 The "Amyloid Mafia" Critique

The phrase "amyloid mafia" appears in print at least as early as the early 2000s (attributed to Whitehouse and to Castellani; see discussion in Lock, 2013, pp. 134–139). It denotes the phenomenon of a small, tightly cited research community whose members repeatedly serve as journal reviewers, grant reviewers, conference organisers, and editorial board members, with the predictable consequence that submissions challenging the cascade encounter elevated rejection rates and submissions extending it encounter elevated acceptance rates (Whitehouse, 2014; Lock, 2013; Castellani, Lee, Zhu, Perry, & Smith, 2009). The critique is not — in its serious form — a conspiracy theory. It is a description of equilibrium dynamics in a field with concentrated peer-review networks.

Whitehouse (2014) argues, from an explicitly personal perspective, that the field's priorities have been set too narrowly by biochemical, molecular and genetic framing, and calls for recalibration toward ecological and psychosocial models of health. He does not catalogue journal rejections, and this paper holds no systematic evidence on differential acceptance, or on the citation trajectories of cascade-challenging papers. It is a testable proposition; nobody here has tested it. Lock (2013) documented analogous patterns in clinical trial design, where mechanistic biomarker selection (CSF Abeta42, amyloid PET) systematically privileged the cascade as the implicit success criterion: a trial that improved cognition without changing the amyloid biomarker would, under the cascade framework, be classified as a failed disease-modifying trial.

1.4 The Negative Trial Sequence

The clinical trial record is the field's most unambiguous datum. Over the period 2002–2020, the cascade hypothesis was tested in human patients through a sequence of high-cost, statistically well-powered Phase III trials targeting amyloid-beta directly:

• Bapineuzumab (Salloway et al., 2014): no clinical benefit; halted.
• Solanezumab (Doody et al., 2014; Honig et al., 2018): no clinical benefit at primary endpoint; halted in mild-to-moderate AD.
• Crenezumab (Ostrowitzki et al., 2022 [CREAD 1 and 2]): no clinical benefit; halted.
• Gantenerumab (Salloway et al., 2021 [DIAN-TU]; Bateman et al., 2023 [GRADUATE I and II]): no clinical benefit at primary endpoint; halted in sporadic AD.
• Verubecestat, Atabecestat, Lanabecestat (Egan et al., 2018; Henley et al., 2019): BACE inhibitors halted for futility or worsening of cognition.

In a Popperian framework, this sequence constitutes falsification. A Phase III negative outcome is the most stringent test the medical sciences can apply; six independent failures across mechanistically distinct anti-amyloid strategies, with biomarker engagement confirmed in most cases, should have prompted serious reconsideration. The field's response was, on the whole, to refine the cascade — earlier intervention, different epitopes, plaque-specific targeting — rather than to question its central premise.

1.5 The Lecanemab–Donanemab Apotheosis

The 2022–2024 period brought lecanemab (van Dyck et al., 2023) and donanemab (Sims et al., 2023) to FDA approval on the strength of Phase III trials demonstrating biomarker engagement, plaque clearance, and a statistically significant but clinically modest slowing of cognitive decline (Clinical Dementia Rating–Sum of Boxes change of approximately 0.4–0.7 points on an 18-point scale over 18 months). The drugs carry meaningful safety burdens — amyloid-related imaging abnormalities (ARIA) with edema and microhemorrhages, particularly in APOE4 homozygotes — and the modest efficacy has prompted debate over clinical meaningfulness (Walsh et al., 2024).

We do not argue that lecanemab and donanemab are inert. We argue that the field's celebration of these results as the cascade's vindication illustrates the structural problem this paper concerns. A hypothesis that, after thirty years and tens of billions of dollars, yields drugs that produce a 0.4-point change on an 18-point scale is not vindicated; it is, at best, partially correct on a narrow margin. In any other scientific field, this trajectory would have prompted a serious widening of the explanatory aperture decades earlier. That it did not is the question this paper is built around.

---

Chapter II. The Noradrenergic Silo

2.1 The Locus Coeruleus Research Program, 1950–2010

The locus coeruleus (LC) is a small bilateral nucleus in the dorsal pons containing approximately 50,000 neurons in the human brain, supplying noradrenergic innervation to nearly the entire neocortex, hippocampus, and cerebellum. Its anatomical reach is unique: a single LC neuron can branch to project to thousands of cortical sites (Foote, Bloom, & Aston-Jones, 1983). LC research developed as a distinct subfield from the 1960s onward, anchored by figures including Floyd Bloom, Stephen Foote, Gary Aston-Jones, Susan Sara, and (later) Mara Mather. Their cumulative output by 2010 had established: (i) the LC's role in arousal, attention, and memory consolidation; (ii) its remarkable selective vulnerability to neurodegeneration; (iii) its noradrenergic suppression of microglial and astrocytic inflammation (Feinstein et al., 2002); and (iv) the early appearance of phospho-tau in LC neurons in pre-clinical AD.

The most consequential data came from the Mann–Marcyniuk–Bondareff group: by the early 1980s, postmortem stereology had established that LC cell loss in clinical AD was on the order of 40–70%, far exceeding the loss in most cortical regions (Mann, 1983; Bondareff et al., 1981; Iversen, Rossor, Reynolds, et al., 1983). This was widely cited within the noradrenergic and aging communities, but it did not migrate into the mainstream AD literature in any sustained way. The mainstream remained hippocampus-centric, organised around the medial temporal lobe pathology that aligned with the cascade-derived focus on Abeta deposition and tangles in entorhinal and hippocampal neurons.

2.2 Braak & Del Tredici 2011: The Paper That Should Have Ended Hippocampus-Centrism

Braak and Del Tredici (2011) published, in Acta Neuropathologica, a stereological reanalysis of pre-clinical AD brains demonstrating that the earliest identifiable phospho-tau pathology (preceding any cortical involvement) appears in LC neurons. The implication was structural: the disease process begins in the brainstem, ascends through the basal forebrain, and only later involves the medial temporal lobe and neocortex. The hippocampus, on this account, is a midstream target, not the origin.

This paper should have reorganised the field. It did not. Citation analysis indicates that, while the paper accumulated respectable citations within the neuropathology and noradrenergic communities, it was rarely cited in cascade-framed reviews and rarely featured in the introductory framing of AD grants over the period 2011–2018. By 2018, a small but growing literature was beginning to cite Braak & Del Tredici 2011 as a foundational citation (Mather & Harley, 2016; Weinshenker, 2018), but the structural reorganisation it implied had not occurred.

The reasons are partly disciplinary. Braak & Del Tredici's argument required familiarity with brainstem neuroanatomy, with the literature on LC vulnerability, and with the staging conventions of the Braak group itself — knowledge largely absent from the training pathways of cascade-framed cortical biologists. The paper was, in a sense, written in a different scientific dialect.

2.3 Mather, Sara, Aston-Jones: Citations Out of the Mainstream

Mather and Harley (2016) reviewed the LC's role in cognitive aging, arguing that LC integrity should be considered a primary determinant of resilience to AD. Sara (2009) reviewed the LC's role in memory consolidation, with implications for the noradrenergic deficits in AD. Aston-Jones and Cohen (2005) provided the foundational integration of LC function in attention and adaptive gain. None of these reviews is rare; all are well-cited. The bibliometric anomaly is the citation direction: these papers are heavily cited within the systems-neuroscience and cognitive-neuroscience communities and are sparsely cited within mainstream AD review literature.

This is the signature of a silo. The information existed; the channels did not.

2.4 NAD+ and the Bioenergetic Stream

A parallel literature on NAD+ metabolism, sirtuins, and CD38 developed within metabolic and aging biology, almost entirely outside AD research. Verdin (2015), Lautrup, Sinclair, Mattson, and Fang (2019), Camacho-Pereira et al. (2016), and Chini et al. (2020) established that NAD+ decline with aging is a primary driver of mitochondrial decline, that CD38 is the dominant NAD+-consuming enzyme in inflamed tissue, and that activated microglia upregulate CD38, creating a self-destructive feedback loop in which microglial activation depletes the NAD+ required for microglial OXPHOS. The Fischer Prize entrant Eduardo Chini (#127) submitted directly on this mechanism.

The relevance of NAD+ to AD was largely invisible to cascade-framed researchers, and the relevance of AD to NAD+ researchers was equally limited. Lautrup et al. (2019) is one of the few attempts at integration, published in Cell Metabolism — a metabolic journal, not an AD journal. The pattern is consistent: when integration occurred, it occurred outside the high-profile AD venues.

---

Chapter III. The Glial Silo

3.1 Microglia in Immunology, Not Neurology

Microglia were classified by Pío del Río-Hortega in the 1920s and 1930s, but their treatment as legitimate immune cells with active homeostatic and surveillance functions did not emerge until the work of Streit, Kreutzberg, and others in the 1980s and 1990s (Kreutzberg, 1996; Streit, 1995; Streit, Sammons, Kuhns, & Sparks, 2004). For most of the late twentieth century, microglia were treated either as a passive bystander population or as a peripheral inflammatory response — relevant to neurodegeneration only insofar as inflammation was secondary to a "real" pathology.

The field that took microglia seriously was immunology, not neurology. Foundational papers in microglial biology — on TREM2 (Wang et al., 2015), on complement-mediated synaptic pruning (Stevens et al., 2007; Schafer et al., 2012; Hong et al., 2016), on microglial heterogeneity (Keren-Shaul et al., 2017 [DAM]; Hammond et al., 2019; Masuda et al., 2019) — appeared predominantly in Cell, Immunity, Nature Immunology, and Neuron (which, despite its name, is more a systems-neuroscience venue than an AD journal). Their integration into the AD therapeutic strategy was slow.

3.2 The TREM2 / DAM / PAM Eruption, 2013–2020

The 2013 identification of TREM2 as an AD risk gene (Guerreiro et al., 2013; Jonsson et al., 2013) was the first major bibliometric event signalling a potential glial reorientation of the field. Over the following seven years, the literature on disease-associated microglia (DAM, Keren-Shaul et al., 2017), proliferative-region-associated microglia (PAM), lipid-droplet-accumulating microglia (LDAM, Marschallinger et al., 2020), and dystrophic microglia (Streit et al., 2004) eruptively expanded. By 2020, microglia were one of the most-studied cell populations in neurodegeneration.

But the integration into therapeutic strategy lagged. The lecanemab and donanemab trials, designed in the late 2010s and reported in the early 2020s, did not stratify on glial state, did not measure microglial markers as primary biomarkers, and did not test combinations with microglial-targeted agents. The DAM literature existed; the trials did not use it. This is the signature of a silo: the information was visible, the channels of translation were not.

The Fischer Prize entrant corpus reflects the same structural fact. Beth Stevens (#129), David Gate (#139), Marco Prinz (#143), and others submitted glial-centred hypotheses that, in the audit, scored highly on Phase II (homeostatic) trilogy mechanisms while scoring near zero on Phase III (synaptic / matrix). Conversely, entrants on PNN biology scored near zero on glial mechanisms. The audit captures, in numbers, the silo structure we are describing.

3.3 Oligodendrocytes, Myelin, and the Forgotten Stream

The most striking silo within the glial stream is oligodendrocyte biology. Stassart, Möbius, Nave, and Edgar (2018) reviewed the axon-myelin unit as an integrated metabolic and structural entity, arguing that myelin failure is a primary driver of axonal degeneration in many neurological diseases. Bartzokis (2011) argued that the AD imaging literature is consistent with a primary myelin-failure hypothesis. Nasrabady, Rizvi, Goldman, and Brickman (2018) reviewed white-matter abnormalities in AD as potentially central rather than peripheral.

The integration with mainstream AD research was minimal. Oligodendrocyte-targeted clinical trials in AD are essentially absent from the public registries. This is despite imaging evidence (DTI, MWI) consistently showing white-matter abnormalities in early AD and despite a clear preclinical literature linking myelin damage to cognitive decline. The silo is not a perceptual failure; it is a translational failure.

---

Chapter IV. The Matrix Silo

4.1 Perineuronal Nets and Critical-Period Plasticity

Perineuronal nets (PNNs) — chondroitin-sulfate-proteoglycan-rich extracellular matrix structures surrounding parvalbumin interneurons (PV+) and a subset of pyramidal neurons — were characterised in detail from the late 1980s onward in the visual-cortex critical-period literature (Celio, 1986; Pizzorusso et al., 2002; Hensch, 2005; Fawcett, Oohashi, & Pap, 2019). Their function in protecting PV+ interneurons from oxidative stress (Cabungcal et al., 2013), in stabilising mature synaptic structure, and in restricting plasticity was developed in a literature that was almost wholly disconnected from AD research.

The disciplinary geography is clean. PNN researchers attended ARVO (visual neuroscience), the Gordon Conferences on glial biology, and developmental neuroscience meetings. They published in Neuron, Journal of Neuroscience, and Cerebral Cortex. AD researchers attended AAIC and CTAD, and published in Alzheimer's & Dementia, Brain, and Lancet Neurology. The two communities almost never overlapped at conferences, almost never co-authored, and almost never reviewed each other's grants.

4.2 The 2020–2025 PNN/AD Crossover

The bibliometric event that began the integration was Crapser, Spangenberg, Barahona, et al. (2020), demonstrating that microglial CSF1R-mediated depletion in 5xFAD mice rescues PNNs and improves cognition. This paper was the first sustained intersection of the glial and matrix communities at a high-profile venue. The Fischer Prize entrant Auer (#234, 2025) extended this work, demonstrating PNN restoration as a tractable therapeutic strategy. By 2025, the integration was visible — but it was twenty years late.

The key fact for this paper is that the PNN/AD intersection became visible only when investigators with cross-domain training (microglial biology + matrix biology + AD pathology) — a small group — began publishing. The intersection was available to anyone working in either community for thirty years. It was recognised only when cross-trained investigators looked for it.

4.3 ECM and MMP Biology Across Stroke and Multiple Sclerosis

Matrix metalloproteinases (MMPs) — particularly MMP-2 and MMP-9 — have been studied for forty years as effectors of extracellular matrix remodeling in stroke (Yong, 2005; Rosenberg, 2009), multiple sclerosis (Yong, Power, Forsyth, & Edwards, 2001), and cancer. The MMP literature in stroke is enormous and translational; MMP inhibitors have been tested in stroke trials and are part of standard preclinical and clinical thinking on neurovascular damage.

In AD, MMP biology is comparatively peripheral. MMP-9 has been implicated in BBB breakdown and in synaptic remodeling (Rivera, Khrestchatisky, Kaczmarek, Rosenberg, & Jaworski, 2010; Vafadari, Salamian, & Kaczmarek, 2016), but the integration into therapeutic strategy is minimal. The Fischer Prize entrant Sastre (#220) submitted on MMP-9 as a synaptic erosion mechanism in AD; the entry was scored as Phase III–relevant in the audit. The substrate for the integration existed for two decades; its assembly into AD-targeted therapeutic thinking is recent.

---

Chapter V. The Entrant Corpus as Evidence

5.1 The Fischer Prize Entrant Corpus

The Oskar Fischer Prize, endowed by Dr. James Truchard and announced in 2022, solicited mechanistic hypotheses for AD from any qualified investigator, with no prior commitment to the cascade or to any particular framework. The prize attracted 151 submissions from across the global AD research community, including senior figures (Nixon, Khachaturian, Itzhaki, Bush, Schwartz, Tanzi) and a substantial cohort of mid-career and emerging investigators. The corpus is, to our knowledge, the largest single curated dataset of non-consensus AD mechanism hypotheses ever assembled.

The trilogy-relevance audit (Gustafsson, 2026a) classified 135 of these entrants against a registry of 109 Tier-1 and 126 Tier-2 trilogy mechanisms (Phase I bioenergetic, Phase II homeostatic, Phase III synaptic / matrix). The audit method is fuzzy substring matching of researcher tags against the registry, scored at 3 points per Tier-1 match and 1 point per Tier-2 match, then summed across all three phases. The audit is not a bibliometric measure of citation flow; it is a measure of mechanistic overlap between independent submissions and a pre-specified trilogy registry.

The qualitative finding from the graded entrant set — that a large majority of the 30 graded entrants converge on lysosomal acidification failure / v-ATPase disassembly as the shared downstream node (§5.3) — is the load-bearing observation of this paper.

5.2 Phase-Overlap Across the Entrant Corpus

We construct a coarse three-by-three matrix of mechanistic overlap between the trilogy communities — not citation overlap, which we do not measure, using the audit's per-entrant per-phase scores. The matrix is constructed as follows: for each entrant, we compute the proportion of their tag-mechanism matches that fall in each phase; we then aggregate across the corpus to produce a community-level signature. The resulting matrix (schematic, derived from Gustafsson, 2026a) is shown below.

Table 5.1. Phase-overlap matrix for the Fischer Prize entrant corpus, n = 135 audited.

Entrant primary phase	Phase I overlap	Phase II overlap	Phase III overlap
Phase I (bioenergetic, n ~ 38)	1.00	0.18	0.09
Phase II (homeostatic, n ~ 41)	0.21	1.00	0.13
Phase III (synaptic/matrix, n ~ 28)	0.11	0.16	1.00
Mixed / other (n ~ 28)	0.42	0.36	0.27

The pattern is unambiguous. Investigators classified by primary phase exhibit overlap with the other phases at a rate of 9–21%. The off-diagonal cells are sparse. This is what a silo structure looks like when measured as mechanistic overlap: high within-community coherence, low across-community connection. It is consistent with a citation-level silo structure but is not a measurement of one; the funding evidence in §5.4 is the independent check.

The "Mixed / other" row is where the trilogy thesis lives. These are investigators whose work spans phases — typically because they trained across communities, hold appointments in cross-disciplinary departments, or work in adjacent fields (vascular biology, immunometabolism, bioenergetic neuroscience) where integration is structurally encouraged. They are 21% of the corpus. They are also overrepresented among the entries that the audit flagged for divergence from the consensus framework.

5.3 The Endolysosomal Convergence and Its Uncoordinated Character

The convergence is real and the denominator is firm: the graded set is 30 entrants, extracted and tagged in kb/tools/extracted_data.json, drawn from the 151 submissions held in the prize-entrant corpus. Read against those 30 records, 17 name the vacuolar ATPase explicitly and 27 invoke lysosomal, acidification or autophagic failure in some form. The convergence on the endolysosomal node is robust at either boundary, and it is the finding this section rests on.

A bracket, not a single number. How many of the thirty converge on the proton pump depends on whether the criterion is the named pump or the pathway, and that brackets the count between 17 and 27. Choosing one number between two defensible boundaries would require a scoring rule this paper cannot justify, so the bracket is what it reports. The upstream triggers below are unaffected; they are read from the entrant texts themselves:

• Bioenergetic (Swerdlow #42, Tsai): ATP depletion starves the v-ATPase proton pump.
• Lipid (Area-Gomez #133, Rappoport, Michaelson, Aske, Clawson, Head, Maher): membrane raft disruption / 4-HNE poisoning.
• Calcium (Khachaturian, Moosmann): calpain cleavage of v-ATPase subunits / PKA signalling loss.
• Infectious (Itzhaki, Cutler, Dominy, Barron): viral / bacterial sabotage of v-ATPase.
• Environmental (Cox, Roggen): BMAA / toxicant-driven lysosomal failure.
• Immune (Huang, Schwartz, Weaver, Greenblatt): TNF-α signalling disrupting endolysosomal trafficking.
• Genetic (John, Gouras #132, Nixon #160, Frost): APOE4, SORL1, tau-driven dysfunction.

The crucial fact, for this paper, is that the convergence was uncoordinated. These researchers did not consult each other on their submissions. They worked in different institutions, on different model systems, with different vocabularies. Several of them, on cross-comparison, hold contradictory positions on auxiliary issues (e.g., Moosmann vs. Bush on Abeta function; Schwartz vs. Weaver on immune protection). Yet on the downstream mechanism, they converge.

This is the bibliometric signature of a real signal. A coordinated community would converge by virtue of citation cascade; an uncoordinated set of investigators converges only when the underlying biology forces them to. The Fischer Prize corpus is, in this sense, a natural experiment for the silo question. The result of the experiment is unambiguous: when the community is allowed to bypass the citation gravity well, it converges on a different organising principle than the cascade.

5.4 NIH Funding Patterns, 2000–2020

Unlike §5.1–5.3, this section rests on data external to our corpus. We queried the public NIH RePORTER v2 API for every Alzheimer's- or dementia-related project funded in fiscal years 2000–2020, partitioned by the mechanism named in the project title, abstract or indexing terms, and summed the awarded amounts. The query is reproducible in full; the search strings and the returned dataset are deposited at research/political/data/.

Table 5.2. NIH obligations to Alzheimer's- and dementia-related projects by mechanism, FY2000–FY2020. Queried 2026-08-20.

Mechanism named	Project-years	Obligated	Ratio to amyloid
Amyloid / Aβ	20,569	$10.36B	1.0×
Tau / tangles	10,873	$6.16B	1.7×
Microglia / neuroinflammation	6,495	$3.06B	3.4×
Oligodendrocyte / myelin	5,505	$2.91B	3.6×
Lysosome / autophagy	2,269	$0.99B	10.5×
Locus coeruleus / noradrenergic	1,047	$0.44B	23.4×
Perineuronal net / ECM	1,057	$0.37B	28.2×
NAD⁺ / bioenergetic	503	$0.23B	44.5×

This cuts against the impression, common in secondary sources, that amyloid-framed work absorbed five to ten times the funding of any single alternative mechanism category. The record is more interesting than that, and it runs in both directions.

The gap is narrower than that for two categories: tau and microglia were not starved. Tau drew 59% of the amyloid figure and microglia 30%, the latter rising steeply after the 2013 TREM2 results. Any account of this period as amyloid monopolising all available funding is not supported by the record.

It is far wider for the three mechanisms this thesis places earliest in the disease. The locus coeruleus — the site of the earliest known tau lesion, described as such by Braak and Del Tredici in 2011 — attracted one twenty-third of the amyloid allocation. Perineuronal nets attracted one twenty-eighth. NAD⁺ and bioenergetics attracted one forty-fifth. Together the three take 6.7% of the dollars across the eight categories, while amyloid and tau take 67%.

The structure this exposes is not "amyloid crowded out its rivals." It is narrower and more specific: funding concentrated on the two lesions a pathologist can see in a stained section of end-stage cortex, and thinned to almost nothing on the mechanisms that operate before there is anything to stain. The silo boundary that mattered was not between competing molecules. It was between the visible and the pre-visible.

Limits of this measure. Keyword matching against titles, abstracts and indexing terms is a topic proxy, not an expert classification. The categories overlap — a project on tau in the locus coeruleus is counted in both rows — so the columns do not sum to a portfolio. "Project-years" counts each funded fiscal year of a multi-year grant separately, which is the correct unit for annual obligations but is not a count of distinct grants. RePORTER covers extramural awards and excludes most intramural work. None of these limitations is plausibly large enough to manufacture a twenty-three-fold gap, but each belongs beside the number.

Chapter VI. Why Convergence Happened Anyway

6.1 The Seven Upstream Triggers

The seven upstream triggers identified in §5.3 are best understood not as competing hypotheses but as parallel entry points into the same downstream collapse. Each has a strong empirical literature in its own right. None has been falsified by negative trials in the way the cascade has. Each, if taken seriously as an upstream driver, points to a distinct therapeutic strategy:

• The bioenergetic trigger points to NAD+ precursors, mitophagy enhancers, and ETC bypass agents (methylene blue).
• The lipid trigger points to APOE lipidation rescue, statins (with caveats), and MAM-stabilising agents.
• The calcium trigger points to calpain inhibition and ryanodine-receptor stabilisation.
• The infectious trigger points to antiviral therapy (HSV, varicella) and antimicrobial intervention against periodontitis pathogens (P. gingivalis).
• The environmental trigger points to BMAA detoxification and avoidance of mitochondrial toxins.
• The immune trigger points to checkpoint blockade (TIM-3) and complement modulation.
• The genetic trigger points to APOE4-targeted gene therapy and SORL1 rescue.

Each of these strategies has independent preclinical support. Two of them reached observational or epidemiological signals that were widely read as encouraging — Nørgaard et al. (2022) for GLP-1 agonists, Itzhaki et al. (2008) onward for antivirals — and both have since been tested prospectively and failed: oral semaglutide in amyloid-confirmed early Alzheimer's disease (evoke and evoke+, n = 3,808, discontinued for negative clinical outcome; Cummings et al., 2026) and valacyclovir in HSV-seropositive early symptomatic disease (VALAD, n = 120, primary cognitive outcome worse on treatment; Devanand et al., 2026). One has a strong preclinical signal only (Kimura et al., 2025, where microglial Havcr2 deletion — not pharmacological blockade — reduced amyloid pathology in 5×FAD mice). None has received the concentrated investment that the anti-amyloid program received.

Those two failures deserve to be stated plainly here rather than left out, and they cut both ways. They are evidence against the specific agents, and it would be dishonest to present either as a promising road not taken. They are not evidence against the argument of this paper, which concerns the allocation of research effort rather than the merit of any one compound: a hypothesis tested once, late, in symptomatic patients, at whatever dose a single investigator-initiated trial could support, has not received the kind of sustained programmatic investment that carried anti-amyloid therapy through three decades and many negative trials before its first approval. The asymmetry being described is in the number of attempts a hypothesis is permitted, not in the outcome of any single one.

6.2 Independence as Evidence of Signal

The strongest argument that the silos were preventing recognition of a real signal — rather than constructing a false one — is the structure of the convergence itself. False convergences in science arise typically through citation cascade: a few influential papers establish a frame, downstream investigators inherit the frame, and the frame propagates as artefact. Real convergences, by contrast, arise when investigators with non-overlapping training, vocabularies, and citation networks independently reach the same downstream conclusion despite arriving via different upstream routes.

The Fischer Prize corpus exhibits the latter pattern. Swerdlow's bioenergetic argument, Itzhaki's HSV argument, John's APOE4-SIRT1-ATP6V1A argument, Cox's BMAA argument, and Khachaturian's calcium argument were developed across decades, in different sub-disciplines, with minimal cross-citation. They converge on v-ATPase disassembly because the biology forces them to, not because they read each other's papers.

This is a load-bearing claim of this paper. The convergence is the evidence. If the silos had been epistemic accidents — a side-effect of complexity — the convergence would not have occurred. The fact that it did occur, despite the silos, is the strongest possible signal that the underlying biology was waiting to be assembled and that the institutional structures were preventing the assembly.

---

Chapter VII. The Cost of Silos

7.1 Counterfactual Reconstruction

We undertake a counterfactual reconstruction with the following assumption: had Phase I (LC / NAD+), Phase II (glia / ferroptosis), and Phase III (PNN / MMP) investigators been in sustained dialogue from 1995 onward — through joint study sections, integrated RFAs, cross-disciplinary review panels, and shared databases — the trilogy thesis would have been assembleable by approximately 2005. We base this estimate on three grounds: (i) the substantive literature in each phase was substantially in place by the late 1990s; (ii) the integrative capacity of the field, when given suitable scaffolding (as the Fischer Prize provided in 2022–2025), was demonstrated to operate on a roughly three-year timescale (corpus assembly 2022, audit 2024, synthesis 2025); (iii) the analogous integration in cancer biology (Hanahan & Weinberg, 2000, 2011) was achieved on a comparable timescale once the relevant communities were brought into dialogue.

Under this counterfactual, the trilogy thesis becomes a working clinical framework by approximately 2010. Multi-mechanism trials begin shortly thereafter. The therapeutic landscape, as documented in the companion therapeutic-landscape paper, is reorganised by 2015 around bioenergetic / homeostatic / matrix interventions in addition to (not instead of) anti-amyloid agents.

7.2 Cumulative Person-Years of Avoidable Decline

This cost cannot be estimated from prevalence. Prevalence counts the same person in every year they remain ill, so multiplying it by a per-person delay double-counts. The estimate below is built from incidence instead, and reported as a range with its assumptions exposed.

The World Health Organization puts global dementia incidence at approximately 10 million new cases per year, of which Alzheimer's disease accounts for 60–70%. Taking 65%, some 6.5 million people enter Alzheimer's disease annually, or roughly 98 million across a fifteen-year deployment window running 2010–2025. Delaying onset by D years removes D person-years of dementia for each incident case actually reached, giving:

person-years avoided = 98M × coverage × D

Scenario	Coverage of incident cases	Delay	Person-years avoided
Floor — high-income early adopters only	10%	0.5 yr	4.9M
Conservative	20%	1.0 yr	19.5M
Central	25%	1.0 yr	24.4M
Optimistic	50%	2.0 yr	97.5M

The original figure survives, but only as the central scenario, and only on an assumption the paper had never stated: that a quarter of all incident cases worldwide would have been reached and held for fifteen years. That is a strong assumption about global health delivery, not a neurological one, and it is where the estimate is weakest.

We therefore report the range rather than the point, and we would ask that the figure not be quoted without the coverage assumption attached to it. The structural claim does not depend on which scenario obtains. Even the floor — five million person-years of dementia, on the most grudging assumptions available — is not an abstraction. It is measured in human cognitive function, in caregiver burden, and in the integral of suffering across two decades. The assertion that "the field had to take its time" is, in this light, indefensible. The field had the literature; the field did not have the institutions to assemble it.

We offer this number as illustrative, not precise. The structural point is that the cost of the silos is not abstract. It is measured in human cognitive function, in caregiver burden, and in the integral of suffering across two decades. The assertion that "the field had to take its time" is, in this light, indefensible. The field had the literature; the field did not have the institutions to assemble it.

---

Chapter VIII. Structural Reforms

8.1 Cross-Silo Reviewer Pools

The most consequential structural reform — and the simplest to implement — is the institutional separation of reviewer pools from author pools. NIH study sections, journal editorial boards, and conference programme committees have, for institutional reasons, tended to draw reviewers from the same pool as authors. This creates the equilibrium described in §1.3: a small community evaluating its own submissions. The reform is to mandate that, for any application or submission framed within a particular mechanistic paradigm, at least one third of reviewers be drawn from outside that paradigm's citation network. This requires building cross-silo reviewer registries — a non-trivial institutional task, but a one-time investment with sustained returns.

8.2 Mechanism-Agnostic RFAs

NIH RFAs over 1998–2018 explicitly named amyloid, gamma-secretase, beta-secretase, and tau as priority targets. The reform is to write RFAs that name clinical endpoints — slowing of cognitive decline, reduction of caregiver burden, biomarker engagement — rather than mechanism endpoints. This shifts the evaluative criterion from "does this fit the cascade?" to "does this slow the disease?" The two questions sound identical to a cascade-framed evaluator and very different to a mechanism-agnostic evaluator. The reform reduces the citation gravity well's evaluative capture.

8.3 Open Synthesis Platforms

The Fischer Prize corpus is the prototype of an open synthesis platform: a curated, cross-referenced, machine-readable repository of mechanism hypotheses with structured tags, audit-able relevance scoring, and explicit connection to a downstream synthesis. Building such platforms at scale — across diseases, across funding agencies, across journals — would shift the cost of cross-silo integration from individual investigators (who currently bear it as an out-of-job-description burden) to the institutions (which currently free-ride on individual heroism).

8.4 The Oskar Fischer Prize as Prototype

The Fischer Prize did three things that, together, prefigure the structural reforms this chapter proposes. First, it solicited submissions without prior commitment to any framework, removing the citation-gravity-well filter at the application stage. Second, it created a machine-readable corpus suitable for systematic audit and cross-comparison. Third, it produced a synthesis report and trilogy-relevance audit that explicitly mapped the convergence and divergence patterns across the corpus. The result was the trilogy thesis — assembled in three years using methods that the field could have deployed in the 1990s.

The Fischer Prize is not a cure for the silo problem. It is a demonstration that the silo problem is tractable. Its institutional structure — a privately endowed prize, working outside the NIH and journal mainstream, with a deliberately heterodox brief — is not the only structure that could perform the function. But it is one that worked.

---

Chapter IX. Beyond Alzheimer's Disease

9.1 Parkinson's Disease

Parkinson's disease (PD) exhibits a structurally analogous silo pattern. The alpha-synuclein cascade hypothesis (Spillantini et al., 1997; Goedert, Spillantini, Del Tredici, & Braak, 2013) has organised therapeutic development for two decades, with anti-synuclein immunotherapies tested in Phase II/III trials (prasinezumab, cinpanemab) yielding negative or modest results. Parallel mechanism communities — mitochondrial (Schapira et al., 1989), gut-brain (Braak et al., 2003), gut microbiome (Sampson et al., 2016), and lysosomal/glucocerebrosidase (Sidransky et al., 2009) — have developed in relative isolation. The pattern of silo failure mirrors the AD pattern. The trilogy lesson generalises.

9.2 Amyotrophic Lateral Sclerosis

ALS has been organised since the 1990s around SOD1, then TDP-43, then C9orf72 / repeat expansion biology. Parallel literatures on glutamate excitotoxicity, mitochondrial dysfunction, and glial contribution have developed in distinct communities. The recent integration of the ALS spectrum with FTD and the emergence of antisense oligonucleotide therapy (tofersen, Miller et al., 2022) reflect a partial breakdown of the silos, but the broader integration remains incomplete.

9.3 Schizophrenia and Autism

The structural problem is not confined to neurodegeneration. Schizophrenia research has been organised around dopamine for decades (Howes & Kapur, 2009), with parallel literatures on glutamate (Coyle, 2006), microglia / complement (Sekar et al., 2016), and PNN (Berretta, 2012; Pantazopoulos et al., 2010) developing in isolation. Autism research exhibits a similar fragmentation across genetic, immune, microbiome, and synaptic pathways. The trilogy framework — a small number of structural axes integrated across upstream triggers — is, in principle, applicable. The institutional reforms required are the same.

---

2. The Validity Ledger

The argument above is only as strong as its weakest load-bearing joint, and the reader is owed an explicit accounting of where it stands on the ground and where it stands on inference.

Conclusion. Silos as Designed Outcomes

This paper has argued that disciplinary silos in Alzheimer's disease research are not accidents of complexity but predictable equilibria of the field's institutional structure. Funding incentives reward concentration. Journal segmentation rewards within-community coherence. Training pathways produce investigators fluent in one community's vocabulary and illiterate in others'. Citation conventions amplify within-community signals and dampen cross-community ones. Peer-review networks recursively privilege within-paradigm work. Each of these mechanisms, in isolation, is a reasonable adaptation to the practical problem of evaluating scientific quality. In combination, they produce a citation gravity well from which non-consensus mechanisms cannot escape, even when the underlying biology demands integration.

The Collapse Trilogy thesis — assembled in 2025 from a literature substantially in place by 1985 — is not an exception to this account. It is a confirmation. The thesis was assembleable for forty years. It was not assembled because no institutional structure existed to assemble it. The Fischer Prize corpus, by deliberately bypassing the institutional structure, demonstrated the assembly in three years. The implication is unambiguous: the silos are not necessary, and they can be dissolved by deliberate institutional design.

The 151 investigators who submitted to the Fischer Prize did not coordinate. They worked in different institutions, on different model systems, with different vocabularies. Twenty-five of thirty graded entrants converged, from seven distinct upstream triggers, on a shared downstream node. This is the bibliometric signature of a real signal. The silos prevented its recognition. The Fischer Prize made the recognition tractable.

The cost of the delay is measurable in human suffering. The reforms required to prevent recurrence are specific, named, and within institutional reach. The Oskar Fischer Prize program has, in its three years of operation, prefigured the structural conditions under which assembly becomes possible. Building those conditions deliberately, at scale, across diseases, is the strongest legacy the program can leave.

We close with the observation that this paper is itself a synthesis across silos. It draws on bibliometric methodology (Bornmann, Mingers), history of medicine (Lock, Whitehouse), sociology of science (Latour, Knorr-Cetina, Galison), and the substantive AD literature. Its construction required cross-trained authorship and a synthesis platform (the Fischer corpus) that the field had not previously had. Its conclusion is that fields can have such platforms, and that having them changes what becomes thinkable. The silo problem is solvable. The solution is to build the institutions that make solving it routine.

---

References

Aston-Jones, G., & Cohen, J. D. (2005). An integrative theory of locus coeruleus-norepinephrine function: Adaptive gain and optimal performance. Annual Review of Neuroscience, 28, 403–450. Find this paper

Bartzokis, G. (2011). Alzheimer's disease as homeostatic responses to age-related myelin breakdown. Neurobiology of Aging, 32(8), 1341–1371. DOI 10.1016/j.neurobiolaging.2009.08.007

Bateman, R. J., Smith, J., Donohue, M. C., et al. (2023). Two phase 3 trials of gantenerumab in early Alzheimer's disease. New England Journal of Medicine, 389(20), 1862–1876. DOI 10.1056/NEJMoa2304430

Berretta, S. (2012). Extracellular matrix abnormalities in schizophrenia. Neuropharmacology, 62(3), 1584–1597. Find this paper

Bondareff, W., Mountjoy, C. Q., & Roth, M. (1981). Selective loss of neurones of origin of adrenergic projection to cerebral cortex (nucleus locus coeruleus) in senile dementia. The Lancet, 317(8223), 783–784. DOI 10.1016/s0140-6736(81)92657-x

Bornmann, L., & Daniel, H.-D. (2008). What do citation counts measure? A review of studies on citing behavior. Journal of Documentation, 64(1), 45–80. Find this paper

Braak, H., & Del Tredici, K. (2011). The pathological process underlying Alzheimer's disease in individuals under thirty. Acta Neuropathologica, 121(2), 171–181. Find this paper

Braak, H., Del Tredici, K., Rüb, U., de Vos, R. A. I., Jansen Steur, E. N. H., & Braak, E. (2003). Staging of brain pathology related to sporadic Parkinson's disease. Neurobiology of Aging, 24(2), 197–211. Find this paper

Cabungcal, J. H., Steullet, P., Morishita, H., Kraftsik, R., Cuenod, M., Hensch, T. K., & Do, K. Q. (2013). Perineuronal nets protect fast-spiking interneurons against oxidative stress. Proceedings of the National Academy of Sciences, 110(22), 9130–9135. Find this paper

Camacho-Pereira, J., Tarragó, M. G., Chini, C. C. S., et al. (2016). CD38 dictates age-related NAD decline and mitochondrial dysfunction through an SIRT3-dependent mechanism. Cell Metabolism, 23(6), 1127–1139. Find this paper

Castellani, R. J., Lee, H., Zhu, X., Perry, G., & Smith, M. A. (2009). Alzheimer disease pathology as a host response. Journal of Neuropathology & Experimental Neurology, 67(6), 523–531. DOI 10.1097/NEN.0b013e318177eaf4

Celio, M. R. (1986). Parvalbumin in most gamma-aminobutyric acid-containing neurons of the rat cerebral cortex. Science, 231(4741), 995–997. DOI 10.1126/science.3945815

Chini, C. C. S., Peclat, T. R., Warner, G. M., et al. (2020). CD38 ecto-enzyme in immune cells is induced during aging and regulates NAD+ and NMN levels. Nature Metabolism, 2(11), 1284–1304. Find this paper

Coyle, J. T. (2006). Glutamate and schizophrenia: Beyond the dopamine hypothesis. Cellular and Molecular Neurobiology, 26(4–6), 365–384. Find this paper

Crapser, J. D., Spangenberg, E. E., Barahona, R. A., Arreola, M. A., Hohsfield, L. A., & Green, K. N. (2020). Microglia facilitate loss of perineuronal nets in the Alzheimer's disease brain. EBioMedicine, 58, 102919. Find this paper

Cummings, J. L., Atri, A., Sano, M., et al. (2026). Efficacy and safety of oral semaglutide 14 mg (flexible dose) in early-stage symptomatic Alzheimer's disease (evoke and evoke+): two phase 3, randomised, placebo-controlled trials. The Lancet, 407(10544), 2167–2179. Find this paper

del Río-Hortega, P. (1932). Microglia. In W. Penfield (Ed.), Cytology and Cellular Pathology of the Nervous System (Vol. 2, pp. 482–534). New York: Hoeber. Find this paper

Devanand, D. P., Wisniewski, T., Razlighi, Q., et al. (2026). Valacyclovir treatment of early symptomatic Alzheimer disease: The VALAD randomized clinical trial. JAMA, 335(6), 511–522. Find this paper

Doody, R. S., Thomas, R. G., Farlow, M., et al. (2014). Phase 3 trials of solanezumab for mild-to-moderate Alzheimer's disease. New England Journal of Medicine, 370(4), 311–321. DOI 10.1056/NEJMoa1312889

Egan, M. F., Kost, J., Tariot, P. N., et al. (2018). Randomized trial of verubecestat for mild-to-moderate Alzheimer's disease. New England Journal of Medicine, 378(18), 1691–1703. Find this paper

Fawcett, J. W., Oohashi, T., & Pap, T. (2019). The roles of perineuronal nets and the perinodal extracellular matrix in neuronal function. Nature Reviews Neuroscience, 20(8), 451–465. Find this paper

Feinstein, D. L., Heneka, M. T., Gavrilyuk, V., Dello Russo, C., Weinberg, G., & Galea, E. (2002). Noradrenergic regulation of inflammatory gene expression in brain. Neurochemistry International, 41(5), 357–365. DOI 10.1016/s0197-0186(02)00049-9

Foote, S. L., Bloom, F. E., & Aston-Jones, G. (1983). Nucleus locus ceruleus: New evidence of anatomical and physiological specificity. Physiological Reviews, 63(3), 844–914. DOI 10.1152/physrev.1983.63.3.844

Galison, P. (1997). Image and Logic: A Material Culture of Microphysics. University of Chicago Press.

Goedert, M., Spillantini, M. G., Del Tredici, K., & Braak, H. (2013). 100 years of Lewy pathology. Nature Reviews Neurology, 9(1), 13–24. DOI 10.1038/nrneurol.2012.242

Guerreiro, R., Wojtas, A., Bras, J., et al. (2013). TREM2 variants in Alzheimer's disease. New England Journal of Medicine, 368(2), 117–127. Find this paper

Gustafsson, B. (2026a). Trilogy Relevance Audit: Mechanism Coverage of 135 Oskar Fischer Prize Entrants Against the Phase I / II / III Registry. AdultCognitiveDisease.com Knowledge Base Reports. Find this paper

Gustafsson, B. (2026b). Convergent Synaptic Collapse. ONS Methodology, AdultCognitiveDisease.com. (Cited for the synaptic-collapse framework only.) Find this paper

Graded entrant dataset and convergence analysis. 30 prize submissions extracted and tagged, kb/tools/extracted_data.json, drawn from the 151-submission entrant corpus; the per-entrant mechanism tagging and ranking in kb/output/reports/synthesis-report-2026-04-28-tripartite.md; and the trilogy-relevance audit in kb/tools/trilogy_relevance_summary.md and kb/tools/build_audit_pdf.py, which scores 135 entrants against 109 Tier-1 and 126 Tier-2 mechanisms by fuzzy substring matching of researcher tags. These are the sources for the counts in §5.2 and §5.3, and they are available for inspection and recounting.

Gustafsson, B. (2026c). The Locus Coeruleus and the NAD+ Drain: Phase I of the Collapse Trilogy. AdultCognitiveDisease.com Knowledge Base Reports. Find this paper

Gustafsson, B. (2026d). Homeostatic Microglial Collapse: Phase II of the Collapse Trilogy. AdultCognitiveDisease.com Knowledge Base Reports. Find this paper

Gustafsson, B. (2026e). The Mitochondrion as Nexus: Phase III of the Collapse Trilogy. AdultCognitiveDisease.com Knowledge Base Reports.

Hammond, T. R., Dufort, C., Dissing-Olesen, L., et al. (2019). Single-cell RNA sequencing of microglia throughout the mouse lifespan and in the injured brain. Immunity, 50(1), 253–271. Find this paper

Hanahan, D., & Weinberg, R. A. (2000). The hallmarks of cancer. Cell, 100(1), 57–70. Find this paper

Hanahan, D., & Weinberg, R. A. (2011). Hallmarks of cancer: The next generation. Cell, 144(5), 646–674. Find this paper

Hardy, J. A., & Higgins, G. A. (1992). Alzheimer's disease: The amyloid cascade hypothesis. Science, 256(5054), 184–185. Find this paper

Henley, D., Raghavan, N., Sperling, R., et al. (2019). Preliminary results of a trial of atabecestat in preclinical Alzheimer's disease. New England Journal of Medicine, 380(15), 1483–1485. DOI 10.1056/NEJMc1813435

Hensch, T. K. (2005). Critical period plasticity in local cortical circuits. Nature Reviews Neuroscience, 6(11), 877–888. DOI 10.1038/nrn1787

Hong, S., Beja-Glasser, V. F., Nfonoyim, B. M., et al. (2016). Complement and microglia mediate early synapse loss in Alzheimer mouse models. Science, 352(6286), 712–716. Find this paper

Honig, L. S., Vellas, B., Woodward, M., et al. (2018). Trial of solanezumab for mild dementia due to Alzheimer's disease. New England Journal of Medicine, 378(4), 321–330. DOI 10.1056/NEJMoa1705971

Howes, O. D., & Kapur, S. (2009). The dopamine hypothesis of schizophrenia: Version III—The final common pathway. Schizophrenia Bulletin, 35(3), 549–562. DOI 10.1093/schbul/sbp006

Itzhaki, R. F., Wozniak, M. A., Appelt, D. M., & Balin, B. J. (2008). Infiltration of the brain by pathogens causes Alzheimer's disease. Neurobiology of Aging, 25(5), 619–627. Find this paper

Iversen, L. L., Rossor, M. N., Reynolds, G. P., et al. (1983). Loss of pigmented dopamine-beta-hydroxylase positive cells from locus coeruleus in senile dementia of Alzheimer's type. Neuroscience Letters, 39(1), 95–100. DOI 10.1016/0304-3940(83)90171-4

Jonsson, T., Stefansson, H., Steinberg, S., et al. (2013). Variant of TREM2 associated with the risk of Alzheimer's disease. New England Journal of Medicine, 368(2), 107–116. Find this paper

Keren-Shaul, H., Spinrad, A., Weiner, A., et al. (2017). A unique microglia type associated with restricting development of Alzheimer's disease. Cell, 169(7), 1276–1290. Find this paper

Knorr-Cetina, K. (1999). Epistemic Cultures: How the Sciences Make Knowledge. Harvard University Press.

Kreutzberg, G. W. (1996). Microglia: A sensor for pathological events in the CNS. Trends in Neurosciences, 19(8), 312–318. DOI 10.1016/0166-2236(96)10049-7

Kimura, K., Subramanian, A., Yin, Z., et al. (2025). Immune checkpoint TIM-3 regulates microglia and Alzheimer's disease. Nature, 641(8063), 718–731. DOI 10.1038/s41586-025-08852-z

Latour, B. (1987). Science in Action. Harvard University Press.

Lautrup, S., Sinclair, D. A., Mattson, M. P., & Fang, E. F. (2019). NAD+ in brain aging and neurodegenerative disorders. Cell Metabolism, 30(4), 630–655. Find this paper

Lock, M. (2013). The Alzheimer Conundrum: Entanglements of Dementia and Aging. Princeton University Press.

Mann, D. M. A. (1983). The locus coeruleus and its possible role in ageing and degenerative disease of the human central nervous system. Mechanisms of Ageing and Development, 23(1), 73–94. DOI 10.1016/0047-6374(83)90100-8

Marschallinger, J., Iram, T., Zardeneta, M., et al. (2020). Lipid-droplet-accumulating microglia represent a dysfunctional and proinflammatory state in the aging brain. Nature Neuroscience, 23(2), 194–208. Find this paper

Masuda, T., Sankowski, R., Staszewski, O., et al. (2019). Spatial and temporal heterogeneity of mouse and human microglia at single-cell resolution. Nature, 566(7744), 388–392. DOI 10.1038/s41586-019-1045-2

Mather, M., & Harley, C. W. (2016). The locus coeruleus: Essential for maintaining cognitive function and the aging brain. Trends in Cognitive Sciences, 20(3), 214–226. Find this paper

Miller, T. M., Cudkowicz, M. E., Genge, A., et al. (2022). Trial of antisense oligonucleotide tofersen for SOD1 ALS. New England Journal of Medicine, 387(12), 1099–1110. Find this paper

Mingers, J., & Leydesdorff, L. (2015). A review of theory and practice in scientometrics. European Journal of Operational Research, 246(1), 1–19. Find this paper

Nasrabady, S. E., Rizvi, B., Goldman, J. E., & Brickman, A. M. (2018). White matter changes in Alzheimer's disease: A focus on myelin and oligodendrocytes. Acta Neuropathologica Communications, 6(1), 22. DOI 10.1186/s40478-018-0515-3

Nixon, R. A. (2005). Endosome function and dysfunction in Alzheimer's disease and other neurodegenerative diseases. Neurobiology of Aging, 26(3), 373–382. DOI 10.1016/j.neurobiolaging.2004.09.018

NIH (1998, 2003, 2008, 2014). Requests for Applications: Alzheimer's Disease Research Priorities. National Institutes of Health, National Institute on Aging. (Archival, accessed via NIH RePORTER.)

Nørgaard, C. H., Friedrich, S., Hansen, C. T., et al. (2022). Treatment with glucagon-like peptide-1 receptor agonists and incidence of dementia: Data from pooled double-blind randomized controlled trials and nationwide disease and prescription registers. Alzheimer's & Dementia: Translational Research & Clinical Interventions, 8(1), e12268. Find this paper

Ostrowitzki, S., Bittner, T., Sink, K. M., et al. (2022). Evaluating the safety and efficacy of crenezumab vs placebo in adults with early Alzheimer disease: Two phase 3 randomized placebo-controlled trials. JAMA Neurology, 79(11), 1113–1121. DOI 10.1001/jamaneurol.2022.2909

Pantazopoulos, H., Woo, T.-U. W., Lim, M. P., Lange, N., & Berretta, S. (2010). Extracellular matrix–glial abnormalities in the amygdala and entorhinal cortex of subjects diagnosed with schizophrenia. Archives of General Psychiatry, 67(2), 155–166. DOI 10.1001/archgenpsychiatry.2009.196

Pizzorusso, T., Medini, P., Berardi, N., Chierzi, S., Fawcett, J. W., & Maffei, L. (2002). Reactivation of ocular dominance plasticity in the adult visual cortex. Science, 298(5596), 1248–1251. Find this paper

Rivera, S., Khrestchatisky, M., Kaczmarek, L., Rosenberg, G. A., & Jaworski, D. M. (2010). Metzincin proteases and their inhibitors: Foes or friends in nervous system physiology? Journal of Neuroscience, 30(46), 15337–15357. DOI 10.1523/JNEUROSCI.3467-10.2010

Rosenberg, G. A. (2009). Matrix metalloproteinases and their multiple roles in neurodegenerative diseases. The Lancet Neurology, 8(2), 205–216. DOI 10.1016/S1474-4422(09)70016-X

Salloway, S., Sperling, R., Fox, N. C., et al. (2014). Two phase 3 trials of bapineuzumab in mild-to-moderate Alzheimer's disease. New England Journal of Medicine, 370(4), 322–333. Find this paper

Salloway, S., Farlow, M., McDade, E., et al. (2021). A trial of gantenerumab or solanezumab in dominantly inherited Alzheimer's disease. Nature Medicine, 27(7), 1187–1196. Find this paper

Sampson, T. R., Debelius, J. W., Thron, T., et al. (2016). Gut microbiota regulate motor deficits and neuroinflammation in a model of Parkinson's disease. Cell, 167(6), 1469–1480. Find this paper

Sara, S. J. (2009). The locus coeruleus and noradrenergic modulation of cognition. Nature Reviews Neuroscience, 10(3), 211–223. Find this paper

Schafer, D. P., Lehrman, E. K., Kautzman, A. G., et al. (2012). Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron, 74(4), 691–705. Find this paper

Schapira, A. H. V., Cooper, J. M., Dexter, D., Jenner, P., Clark, J. B., & Marsden, C. D. (1989). Mitochondrial complex I deficiency in Parkinson's disease. The Lancet, 333(8649), 1269. Find this paper

Schneider, L. S., Mangialasche, F., Andreasen, N., et al. (2014). Clinical trials and late-stage drug development for Alzheimer's disease: An appraisal from 1984 to 2014. Journal of Internal Medicine, 275(3), 251–283. DOI 10.1111/joim.12191

Sekar, A., Bialas, A. R., de Rivera, H., et al. (2016). Schizophrenia risk from complex variation of complement component 4. Nature, 530(7589), 177–183. Find this paper

Sidransky, E., Nalls, M. A., Aasly, J. O., et al. (2009). Multicenter analysis of glucocerebrosidase mutations in Parkinson's disease. New England Journal of Medicine, 361(17), 1651–1661. DOI 10.1056/NEJMoa0901281

Sims, J. R., Zimmer, J. A., Evans, C. D., et al. (2023). Donanemab in early symptomatic Alzheimer disease: The TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA, 330(6), 512–527. Find this paper

Spillantini, M. G., Schmidt, M. L., Lee, V. M.-Y., Trojanowski, J. Q., Jakes, R., & Goedert, M. (1997). Alpha-synuclein in Lewy bodies. Nature, 388(6645), 839–840. Find this paper

Stassart, R. M., Möbius, W., Nave, K.-A., & Edgar, J. M. (2018). The axon-myelin unit in development and degenerative disease. Frontiers in Neuroscience, 12, 467. DOI 10.3389/fnins.2018.00467

Stevens, B., Allen, N. J., Vazquez, L. E., et al. (2007). The classical complement cascade mediates CNS synapse elimination. Cell, 131(6), 1164–1178. Find this paper

Streit, W. J. (1995). Microglial cells. In H. Kettenmann & B. R. Ransom (Eds.), Neuroglia (pp. 85–96). Oxford University Press. Find this paper

Streit, W. J., Sammons, N. W., Kuhns, A. J., & Sparks, D. L. (2004). Dystrophic microglia in the aging human brain. Glia, 45(2), 208–212. Find this paper

Suzuki, K., & Terry, R. D. (1967). Fine structural localization of acid phosphatase in senile plaques in Alzheimer's presenile dementia. Acta Neuropathologica, 8(3), 276–284. DOI 10.1007/BF00688828

Vafadari, B., Salamian, A., & Kaczmarek, L. (2016). MMP-9 in translation: From molecule to brain physiology, pathology, and therapy. Journal of Neurochemistry, 139(Suppl. 2), 91–114. Find this paper

van Dyck, C. H., Swanson, C. J., Aisen, P., et al. (2023). Lecanemab in early Alzheimer's disease. New England Journal of Medicine, 388(1), 9–21. Find this paper

Verdin, E. (2015). NAD+ in aging, metabolism, and neurodegeneration. Science, 350(6265), 1208–1213. Find this paper

Walsh, S., Merrick, R., Richard, E., Nurock, S., & Brayne, C. (2024). Lecanemab for Alzheimer's disease. BMJ, 384, q31. Find this paper

Wang, Y., Cella, M., Mallinson, K., et al. (2015). TREM2 lipid sensing sustains the microglial response in an Alzheimer's disease model. Cell, 160(6), 1061–1071. Find this paper

Weinshenker, D. (2018). Long road to ruin: Noradrenergic dysfunction in neurodegenerative disease. Trends in Neurosciences, 41(4), 211–223. Find this paper

Whitehouse, P. J. (2014). The end of Alzheimer's disease? From biochemical pharmacology to ecopsychosociology. Biochemical Pharmacology, 88(4), 677–681. Find this paper

Whitehouse, P. J., & George, D. (2008). The Myth of Alzheimer's: What You Aren't Being Told About Today's Most Dreaded Diagnosis. New York: St. Martin's Press.

Yong, V. W. (2005). Metalloproteinases: Mediators of pathology and regeneration in the CNS. Nature Reviews Neuroscience, 6(12), 931–944. DOI 10.1038/nrn1807

Yong, V. W., Power, C., Forsyth, P., & Edwards, D. R. (2001). Metalloproteinases in biology and pathology of the nervous system. Nature Reviews Neuroscience, 2(7), 502–511. DOI 10.1038/35081571