The proteinopenia hypothesis: Loss of A β 42 and the onset of Alzheimer ’ s Disease

The dominant protein-lowering strategy in Alzheimer ’ s Disease (AD) has failed to provide a clinically-meaningful treatment for patients. We hypothesize that the loss of functional, soluble A β 42 during the process of aggregation into amyloid is more detrimental to the brain than the corresponding accrual of insoluble amyloid.


The prevailing proteinopathy hypothesis
The proteinopathy paradigm posits that when proteins become aggregated into an amyloid state, referred to as pathology, they not only define neurodegenerative diseases but cause them-that pathology is pathogenesis.In AD, such proteinopathy applies to the transformation of monomeric Aβ, particularly the more fibrillogenic variant with 42 amino acids Aβ 42 , into amyloid.This paradigm suffers from two major anomalies: (1) aggregated proteins are poor predictors of functional decline, and (2) amyloid clearance has yielded futility, harm, or, with lecanemab and donanemab, no improvement but statistically significant slowing in cognitive decline of a magnitude below the established threshold for clinical meaningfulness (Ackley et al., 2021).
Most amyloid-positive individuals do not develop dementia within a normal life span (Brookmeyer and Abdalla, 2018).The population-based amyloid-and-dementia prevalence curves diverge for most of the seventh and eighth decades of life, becoming parallel only by the age of 85 years, when the amyloid-to-dementia ratio reaches 5:1 (Fig. 1) (Jack et al., 2019).That is, four-fifths of amyloid-positive individuals remain cognitively normal within a normal lifetime.Conversely, the ratio between low soluble Aβ 42 and dementia is close to 1:1, as most individuals with cerebrospinal fluid (CSF) levels of Aβ 42 below 800 pg/ml by ELISA (INNOTEST Aβ 42 ; Innogenetics) are clinically demented (Andreasen et al., 2001).In short, high brain amyloid is compatible with normal neurological function whereas low soluble Aβ 42 is not.Among all AD biomarkers, including high phosphorylated tau, low soluble Aβ 42 has the greatest sensitivity for AD detection, 96 %, with a receiver operating characteristic area under the curve of 0.91 (Shaw et al., 2009).
In support of their hypothesis, Hardy and Higgins invoked patients with Down syndrome whose trisomy in the region of chromosome 21 overexpresses the amyloid precursor protein (APP).Mutations in the APP gene, which codes for APP, lead to excessive amyloid formation and AD-like dementia.Since then, carriers of APP and the other early-onset AD-causing mutations presenilin-1 (PSEN1) and presenilin-2 (PSEN2) have been thought to have high levels of Aβ 42 ( Tcw and Goate, 2017;Potter et al., 2013).However, mutation carriers have lower Aβ 42 levels compared to non-mutation carriers (Fagan et al., 2021).Cognitive impairment appears in AD and Down's syndrome when soluble Aβ 42 levels fall below a compensation threshold (Fagan et al., 2021;Portelius et al., 2014).The reduction in Aβ 42 levels begins as early as 25 years before symptom onset in mutation carriers (McDade et al., 2018).Furthermore, most PSEN1 mutations lower Aβ production, (Sun et al., 2017) as do the Osaka and Arctic APP mutations, remarkably in the  absence of detectable brain amyloid (Tomiyama and Shimada, 2020;Schöll et al., 2012).Other genetic risk factors for AD point in the same direction.For example, carriers of the protective ApoE ε2 variant have higher levels of soluble Aβ 42 compared to ApoE ε4 carriers who are at high risk of developing AD (Conejero-Goldberg et al., 2014).Finally, the direct pharmacological lowering of soluble Aβ 42 , accomplished through β-secretase (BACE1) inhibitors, has been shown to worsen cognition (Egan et al., 2019;Sperling et al., 2021).
If low soluble Aβ 42 is detrimental, high Aβ 42 should be protective.This has been supported cross-sectionally in a cohort of amyloid-positive individuals (Sturchio et al., 2021) and longitudinally in amyloid-positive carriers of an AD-causing mutation (APP, PSEN1, or PSEN2) (Sturchio et al., 2022).Levels of Aβ 42 above 800 pg/ml are associated with normal cognition regardless of any level of brain amyloid (Sturchio et al., 2021).The higher the brain amyloid levels, the higher the soluble Aβ 42 levels predicting normal cognition, with an effect size greater than for changes in total or phosphorylated tau (Sturchio et al., 2022).Of all 17 previous anti-amyloid treatments, lecanemab and donanemab are the first to show a statistically significant, if clinically modest, slowing of cognitive decline (van Dyck et al., 2023).We hypothesize that what distinguishes them from prior monoclonal antibodies is not their success in clearing insoluble Aβ but the marked increase they induced in soluble Aβ 42 levels (lecanemab increases CSF levels of Aβ 42 by 288 pg/ml (van Dyck et al., 2023) whereas other monoclonal antibodies do so to a lower extent; solanezumab, for instance, increases such levels by only 37 pg/ml (Doody et al., 2014)).Of note, the destabilization of aggregated species has been proposed as a direct action of anti-Aβ monoclonal antibodies (Imbimbo et al., 2023).An in vitro study has shown that monoclonal antibodies bind to Aβ aggregates leading to disaggregation of the fibrils and partial restoration of Aβ solubility (Solomon et al., 1997).

The alternative proteinopenia hypothesis
An alternative hypothesis can be conceived when considering the upstream end of the amyloid formation process (Fig. 2).As Aβ 42 monomers transform into amyloid (measurable in the brain), the Aβ 42 pool decreases (measurable in the CSF) (Fagan et al., 2006).Aβ 42 is a neuropeptide evolutionarily conserved throughout the animal kingdom with well-documented neurotrophic effects and roles in modulating synaptic plasticity (Zhou et al., 2022).There have been at least 25 studies supporting these and other beneficial functions of Aβ 42 and other Aβ species (see Supplementary Table ).In experiments performed at physiological (picomolar) concentrations, Aβ exerts neurotrophic (Whitson et al., 1989(Whitson et al., , 1990) ) effects; in experiments performed at supra-physiological (nanomolar to micromolar) concentrations, Aβ 42 is toxic (Walsh and Selkoe, 2007).This toxicity is, thus, physiologically artificial.Furthermore, Aβ knockdown and knockout animals can display a neuronal phenotype in the absence of amyloid (Kent et al., 2020).Therefore, we hypothesize that the loss of soluble Aβ 42 to protein aggregation also represents the loss of its neurotrophic properties and is, therefore, more detrimental to the brain than the corresponding accrual of insoluble amyloid (Espay et al., 2021).
Amyloid formation is a physical process of liquid-to-solid phase transition, from soluble, freely moving protein molecules into solid elongated fibrils of repeating protein units tightly bound and stacked on top of each other in a cross-β conformation.In this conformation, the side chains are involved in extensive intermolecular self-interactions instead of being available to perform specific functions (Sawaya et al., 2021).Because of insolubility and colloidal stability, amyloids are non-reactive and, therefore, generally non-toxic, explaining the lack of correlation between brain amyloid levels and neurodegeneration.Notably, the physical process of phase transition of proteins will not take place unless two physicochemical requisites are fulfilled: a. Supersaturation, as high protein concentration creates the crowding and molecular proximity needed for developing intermolecular amyloid bonds rather than adopting the native conformation of the free protein.b. Nucleation, the rate-limiting step in forming the first stable amyloid "nucleus," after which amyloid formation proceeds spontaneously, transforming the available proteins into amyloids.
Triggers or catalysts of protein aggregation should thus be understood within this physical framework.Protein overexpression, as is the case with APP duplications, will increase the concentration of proteins, facilitating their phase transition.With ongoing aggregation, there is corresponding consumption of the soluble protein substrate, as observed in Down's syndrome patients who reach the threshold of dementia when their levels of soluble Aβ 42 become too low.In the case of normal protein expression, nucleation catalysts can play a more decisive role.These catalysts can be preformed amyloid nuclei (seeds) or aberrant surfaces from microbes (e.g., viruses), (Protto et al., 2022) which are more common.

Testing the hypothesis: restoring Aβ 42
Developing ways to restore soluble Aβ 42 via pharmacological induction, enhancement, or replacement represents a testable rescue strategy for slowing disease progression in AD.Indeed, the administration of exogenous synthetic Aβ 42 monomers has been shown to rescue the degenerative and behavioral phenotype in both double knockout and APP/PS1/Tau triple transgenic mice models (Duan et al., 2022).We propose testing this hypothesis by designing a clinical trial to evaluate whether increasing the cerebrospinal fluid (CSF) levels of Aβ 42 to correct its depletion can improve the overall function of patients with both sporadic and familial AD.Patients with mild-to-moderate sporadic or familial AD and CSF Aβ 42 levels < 800 pg/ml would be ideal candidates to recruit into a placebo-controlled, randomized clinical trial evaluating an Aβ 42 -raising intervention using validated safety, clinical, and brain volume endpoints.The hypothesis tested is that those reaching CSF Aβ 42 levels > 900 pg/ml will show improvements in clinical (e.g., Clinical Dementia Rating Scale Sum of Boxes) and brain volumetric endpoints compared to those on placebo, in whom CSF Aβ 42 levels would be expected to further decline (Fig. 3).
While amyloid clearance still occupies an important place within the AD research pipeline, future strategies could also feasibly test the safety and efficacy of Aβ 42 restoration strategies if adequate investment is allocated to evaluating this alternative hypothesis.

Declaration of Competing Interest
AJE has received grant support from the NIH and the Michael J Fox Foundation; personal compensation as a consultant/scientific advisory board member for Neuroderm, Amneal, Acadia, Acorda, Bexion, Kyowa Kirin, Sunovion, Supernus (formerly, USWorldMeds), Avion Pharmaceuticals, and Herantis Pharma; personal compensation as honoraria for speakership for Avion; and publishing royalties from Lippincott Williams & Wilkins, Cambridge University Press, and Springer.He cofounded REGAIN Therapeutics and is co-inventor of the patent "compositions and methods for treatment and/or prophylaxis of proteinopathies."KH has nothing to disclose.TD has nothing to disclose.KPK has nothing to disclose.

Fig. 1 .
Fig. 1.Predicted versus actual relationship between amyloid and dementia prevalence.The curves diverge throughout the seventh and eight decades of life, becoming parallel (without converging) by around age 85.If amyloid were toxic the predicted brain amyloid-to-dementia ratio toward the end of a lifetime should be ~1:1 (left); the actual ratio is ~5:1 (right).Conversely, the ratio of low cerebrospinal fluid (soluble) Aβ 42 levels to dementia is close to 1:1.Virtually no patient with Alzheimer's disease has high Aβ 42 levels but most amyloid-positive individuals remain normal throughout their lifespan.Data on population prevalence adapted from data from the Mayo Clinic Study of Aging cohort (amyloid PET curve, n = 1524) (Jack et al., 2019).

Fig. 3 .
Fig. 3. Rescue medicine.Outline of a future clinical trial aimed at restoring the levels of soluble Aβ 42 above the threshold of compensation to test the proteinopenia hypothesis.CDR-SB: Clinical Dementia Rating Scale Sum of Boxes; ADL: activities of daily living.