COMMENTARY: Molecular polymorphism of Aβ in Alzheimer’s disease

1. Introduction

Alzheimer’s disease is defined pathologically by the presence of senile plaques, which consist primarily of extracellular aggregates of fibrillar Aβ peptide, and neurofibrillary tangles, which are abnormal, intracellular bundles of fibrillar tau protein. The advent of amyloid binding agents as diagnostic imaging probes for Alzheimer’s disease (AD) has made it imperative to understand at a molecular and disease level what these ligands are reporting. In addition to improving the accuracy of diagnosis, we argue that these selective ligands can serve as probes for molecular polymorphisms that may govern the pathogenicity of abnormal protein aggregates.

2. There is a disconnect between amyloid pathology and neurodegeneration/clinical symptoms

A definite histopathologic diagnosis of AD requires the presence of senile plaques. Whether they are directly linked to neurodegeneration or are an epiphenomenon of the pathogenic process, the presence of plaques must be explained by any proposed disease mechanism. Their utility as biomarkers, though, is hampered by the weak correlation of the amount and regional distribution of Aβ deposition with the extent of neurodegeneration that results in functional deficits [5,8,50]. Although some investigators interpret the poor correlation to mean that the senile plaques are secondary to the essential disease process, a great deal of genetic and biochemical data indicate that Aβ is a prime-mover in the pathogenic cascade [12]. The amended Amyloid Cascade Hypothesis (perhaps better designated the Aβ Cascade Hypothesis) holds that soluble, multimeric forms of misfolded Aβ (i.e., Aβ-oligomers) are the most toxic species, whereas insoluble fibrils, which comprise the bulk of senile plaque cores, serve as either a less toxic depot or as a sink in equilibrium with monomers and/or oligomers.

3. Animal models have substantial Aβ deposition but relatively little neurodegeneration

Any postulated Alzheimer’s disease mechanism must be reconciled with the stereotypical Aβ deposition disease hallmarks. It is critical to determine the association of Aβ deposits with the disease to validate the utility of imaging senile plaques in diagnosing AD and/or monitoring disease progression. Nonhuman species have proven to be only partial models of AD. Although APP-transgenic mice, nonhuman primates, and dogs all manifest age-related CNS Aβ deposition that is similar in distribution and quantity to that in humans with AD, these models lack the massive neurodegeneration and neurofibrillary tangles of AD [9,29]. Is it possible that the Aβ deposits in nonhuman animals are structurally different on a molecular/conformational level from AD Aβ deposits, and is this difference directly or indirectly associated with disease mechanism? Understanding the molecular diversity of Aβ multimers could provide important clues to the pathogenesis of AD, and could thereby suggest new therapeutic approaches to the disorder.

4. When is a misfolded protein deposit pathologic?

Insoluble protein deposits have classically been characterized by the presence or absence of fibrillar morphology and a condensed core, as well as by their staining properties with amyloid dyes such as Congo Red (Figure 1 [6]) and the Thioflavines (T and S, Figure 1 [1, 5]). Similar characteristics have been observed in deposits generated by animal models of Aβ deposition, yet neuronal loss is minimal in these instances, suggesting that some as yet unidentified property of the multimeric protein renders the molecule uniquely pathogenic in humans.

Figure 1.

Amyloid-Binding Ligands

• [1]
  Thioflavine T
• [2]
  PIB1
• [3]
  BTA-1
• [4]
  IMPY
• [5]
  Thioflavine S (mixture)
• [6]
  Congo Red
• [7]
  Chrysamine G
• [8]
  X-34
• [9]
  BSB

The prion diseases are the best characterized example of how conformational variants of the same primary amino acid sequence can lead to different disease phenotypes. During the pathogenesis of prionosis, the normal cellular conformation of the prion protein (PrP^C) is converted into β-sheet-rich, pathogenic conformations (PrP^Sc) that are resistant to denaturation and proteolysis. Distinct polymorphic forms of this multimeric PrP^Sc, also called strains, are differently folded molecules that present pathologically different phenotypes and morphologies [1,7]. Diverse structural variants of prions also can be generated in vitro from the same amino acid sequence and starting conformation of prion protein by adjusting the assembly conditions (temperature, pH, agitation) [30,46]. Each strain of prion protein tends to elicit a particular lesion phenotype and clinical course of the individual prion diseases that is dependent on characteristics of both the host and the protein [44,47,48].

5. Polymorphic forms of Aβ

Although the story is not as fully developed for Aβ, polymorphic forms of Aβ peptide fibrils also have been produced in vitro [40,49,52]. Distinct folded structures of two forms differing in hydrogen bonding across the β-turn between residues 25 and 28 have been determined by solid state NMR [40,41]. There is also evidence that pH similar to that in endosomal compartments influences epitope exposure in residues 8-14 of soluble monomeric Aβ peptide, reflecting a distinct, partially folded conformation [27,34]. In vivo, heterogeneous Aβ [3,51] deposits suggestive of Aβ strains have been induced by different exogenous Aβ seeds in APP-transgenic mice [35]. The molecular relationship between any of these molecular forms and the Aβ fibrils and oligomers found in AD brain remains to be determined. Another important question, given the strong correlation between neurofibrillary tangles and degree of dementia in AD [5,8,50], is whether certain Aβ strains might preferentially promote the development of neurofibrillary tangles.

Polymorphism in the molecular structure of Aβ multimers is a potential explanation for the observation that only humans have the severe neuronal loss, neurofibrillary tangles and cognitive decline of AD, and nonhuman species do not. Strain differences might also account for the fact that some cognitively normal humans have large amounts of cerebral Aβ [4], and for other features of the disorder such as inflammation and cerebral β-amyloid angiopathy (CAA) [3,56]. Furthermore, morphological and distributional variants of Aβ deposits coexist within an individual AD brain [51]. These lesions include dense-core senile plaques, CAA, and a multiplicity of diffuse, granular and ribbon-like deposits [13,45,51,53]. Whether these morphologically diverse lesions are the supramolecular manifestations of different Aβ strains remains to be established.

6. Recognizing Aβ Polymorphisms

Classical methods for detecting structural differences in cerebral Aβ deposits do not make fine distinctions at the molecular level. Widely used Aβ sequence-specific antibodies such as 6E10 and 4G8, and techniques such as amyloidotrophic Thioflavine- and Congo Red-staining are of limited utility in distinguishing different types of amyloid deposits. Conformation-specific antibodies that distinguish amyloid fibrils [18,26,36,39] and oligomers are not primary amino acid sequence-dependent (except see [26,36]). However, antibodies that distinguish among fibrillar Aβ strains, similar to those available for PrP, are not yet available. Some of the anti-oligomer antibodies may indeed recognize conformational variants, notably the A11, I11 and OC (officer) antibodies [17,26,55].

Putative structural varieties of Aβ can be distinguished biochemically by differential extraction of AD brain tissue. As AD progresses, it is possible to demonstrate characteristic shifts of Aβ into increasingly insoluble multimers, beginning with a buffer-soluble state, followed by alkali-soluble, SDS-soluble, and finally requiring formic acid to solubilize the peptide. Similar distributions of peptide solubility also are found in the animal models [16,25,37].

A series of fluorescent polythiophene anions and cations have recently been described [6] that report on the spatial (molecular?) heterogeneity of multimeric Aβ. Under appropriate conditions these agents bind the stacks of assembled β-sheets in Aβ and other amyloid fibrils in a sequence-independent fashion [2]. Their fluorescence emission spectra reflect intra- and inter-polythiophene chain contacts which shift to green for tightly packed, regularly oriented fibrils and to red for less organized, less compact fibrils [38]. Synthetic Aβ-amyloid fibrils formed under quiescent conditions fluoresce green with these reagents, while fibrils formed with agitation emit red. These compounds also recognize molecular heterogeneity in Aβ deposits formed in vivo [34]. The fibrillar Thioflavine-positive amyloid deposits in Tg APP/PS1 mice and AD brain are mosaics of the two fibril conformations, indicating that individual fibril deposits consist of polymorphic domains [38]. These compact lesions also are distinct from diffuse or non-fibrillar deposits, which are Thioflavine- and Congo Red-negative. Structural evidence supporting the potential for differential fibril organization comes from solid state NMR, which shows that the synthetic Aβ peptide folds differently in fibrils prepared under quiescent vs. agitated conditions [40].

7. Amyloid probes and polymorphic Aβ

A probe of fibrillar Aβ, the N- ¹¹CH₃-labeled benzothiazole ligand 6-OH-BTA-1 (Pittsburgh compound B; ¹¹C-PIB) (Figure 1 [2]), a brain-penetrant analog of BTA-1 (Figure 1, [3]), is retained in Aβ-rich brain regions of living MCI and AD subjects at a high affinity binding site, but it washes rapidly out of unaffected regions or appropriate control brains [20]. Positron-Emission-Tomography (PET) images of these subjects can be processed into a map of putative Aβ-deposition in the living brain. Neurofibrillary tangles (hyperphosphorylated tau) [22] and Lewy Bodies (α-synuclein) [10], which also contain β-sheet structure, have low affinity for PIB, so they are not expected to contribute significantly to the signal at imaging concentrations of PIB used in AD subjects. Frontotemporal dementia (a primary tauopathy) gives detectable, but much weaker, PIB signals in some subjects [42], but the retention of label is distinguishable by its regional distribution. The possible contribution of a small amount of Aβ plaque pathology in these cases has not been excluded. PIB uptake is increased in disease-relevant brain regions in MCI subjects [11,19]. CAA, which is found to varying degrees in most AD subjects, also binds PIB [15]. When ³H-PIB is applied to AD histological sections, dense-core plaques, CAA and some diffuse Aβ deposits bind the agent, but NFTs do not [14,31]. Studies are underway to determine whether PIB retention can be used to follow the progression of AD and whether it will be useful for monitoring the effectiveness of therapeutic interventions.

8. PIB detects differences between animal models and AD

The sulfonated Congo Red [6] and Thioflavine S [5] dyes are used routinely to analyze AD brain pathology. Congo Red and its congeners (Chrysamine G, X-34, BSB; Figure 1 [7-9]), as well as Thioflavine S, bind to sites distinct from the high affinity site that is occupied by PIB [32,54]. In fibrils made with synthetic Aβ1-40 or Aβ1-42 peptide, and in plaque-rich transgenic mouse and AD brain, Congo Red binds fibrils in a 1:1 stoichiometry with Aβ [21,28]. PIB binding may be more selective on a molecular level than are Congo Red and Thioflavine S. The Klunk group made the seminal observation that when ¹¹C-PIB of the same specific radioactivity and concentration (~nM) used for human studies was used to image APP/PS1 transgenic mice (Tg2576 X M146L-PS1) with an Aβ plaque load equal to or exceeding that in AD brain, the plaques did not bind enough radioligand to produce an image [21]. However, the murine Aβ plaque pathology was revealed by staining sections with 100 nM 6-CN-BTA-1 (a close analog of PIB [6-OH-BTA-1]), X-34, or with anti-Aβ antibodies. Subsequent quantitative binding studies comparing brain homogenates from AD, non-demented humans, and plaque-bearing transgenic mice revealed that the transgenic mice had a greatly reduced number of high affinity (nM) PIB binding sites per mole of Aβ peptide (< 1:100 compared to 1:2 in AD). These high affinity sites would be the only sites significantly labeled at the concentrations of ¹¹C-PIB usually achieved in vivo, hence the weak signal in transgenic mice. The high affinity binding stoichiometry was even lower (1:370) on fibrils of Aβ1-40 or Aβ1-42 formed from synthetic peptide [21]. The structural basis at the molecular level for these differences remains to be elucidated.

9. Implications

An intriguing possibility is that the high affinity PIB binding site represents a molecular polymorph of Aβ that constitutes a high proportion of deposits in AD brain, but a low proportion in APP transgenic mouse brain and in fibrils of synthetic peptide. Support for this hypothesis comes from imaging studies with ten-fold higher specific activity ¹¹C-PIB, which is sufficient to detect the lower percentage of high affinity binding sites in the APP23 Tg mouse and in histological sections of Tg2576 and Tg2576 X A260V-PS1 brains [33]. In vivo binding, homogenate binding, and autoradiographic studies of the Tg2576 X M146L-PS1 mice with ¹²⁵I-IMPY ([4], an analog of PIB) of similar high specific radioactivity also detect a low abundance, high affinity binding site that is associated with plaques and cerebrovascular amyloid fibrils, but only minimally with neurofibrillary tangles [23,24].

Whether the material in brain that is selectively labeled by PIB is a cause or effect of the pathogenic process in AD, it appears to be a marker for an aspect of the disease that is only minimally recapitulated by synthetic peptides and current APP transgenic mouse models. Aged nonhuman primates and dogs naturally deposit human-sequence Aβ in the brain and display some behavioral dysfunction, but, like APP-transgenic mice, they do not exhibit extensive neuronal loss or dementia. Our preliminary studies of aged nonhuman primates with heavy Aβ-amyloidosis [43] support the hypothesis that PIB recognizes an enriched human molecular morphotype of Aβ. PIB binding, along with other new molecular probes, could be useful tools for assessing efforts to expand current models beyond recreating Alzheimer’s histopathology to more accurately phenocopying Alzheimer’s disease.