Aβ43 in human Alzheimer’s disease: effects of active Aβ42 immunization

Alzheimer’s disease (AD) is characterized by the extracellular accumulation of the Amyloid-β (Aβ) protein in the brain parenchyma as plaques and in cerebrovascular blood vessel walls as cerebral amyloid angiopathy (CAA), and the intraneuronal accumulation of the tau protein. It is hypothesized that CAA is caused by dysfunctional Aβ elimination: a process that may include enzymatic degradation, receptor-mediated clearance across the blood-brain barrier (BBB), and drainage with interstitial fluid along perivascular pathways [33]. Blockage of these pathways may lead to aggregation of Aβ in the vessel walls as CAA [10, 22]. In a subset of CAA patients, Aβ deposits in both arteries and arterioles, and capillaries (CAA type 1), whereas more commonly, cerebrovascular Aβ accumulation is restricted to larger vessels, without affecting capillaries (CAA type 2) [34]. The occipital lobe is most frequently and severely affected by CAA [1, 2], and leptomeningeal vessels are more frequently affected compared to parenchymal vessels [1, 27].

Aβ is a peptide consisting of 38 to 43 amino acids, with Aβ40 and Aβ42 being the most-studied Aβ isoforms. Aβ42 is a major constituent of plaques, whereas some studies show that Aβ40 is the predominant species in CAA [9, 28]. Aβ is produced through cleavage of the amyloid precursor protein (APP) by γ-secretase, which is predominantly initiated at ε cleavage sites situated after Leu49 or Thr48. The resulting Aβ49 or Aβ48 are sequentially cleaved in increments of three amino acids, which leads to the production pathways Aβ49 → Aβ46 → Aβ43 → Aβ40 and Aβ48 → Aβ45 → Aβ42 [7, 26, 32]. Only relatively recently it was found that in addition to Aβ40 and Aβ42, also Aβ43 may play a major role in AD, as plaques contain more Aβ43 than Aβ40 [13, 29, 36]. Interestingly, although Aβ43 seems to be a major component of plaques, no significant levels of Aβ43 have been reported in CAA [13, 36].

The amyloid cascade hypothesis states that aggregation of Aβ into plaques is the initial event in AD, followed by the formation of neurofibrillary tangles, and synaptic and neuronal loss leading to cognitive decline [17]. Therefore, in the past two decades, much attention has been directed towards the possibilities of Aβ immunotherapy to remove Aβ from the brain. Both active and passive Aβ immunotherapy approaches have been shown to be effective in the removal of plaques from the human brain, although so far the effects on slowing or preventing cognitive decline have been mainly disappointing [12, 31, 35].

In the first clinical trial of Aβ immunization, which started in 2000, patients with mild to moderate AD were actively immunized with synthetic full-length Aβ42 (AN1792, Elan Pharmaceuticals [4]). Post-mortem neuropathological examination of 12 immunized AD patients showed variable, sometimes extensive, clearance of Aβ plaques from the cerebral cortex [5, 20, 23, 30]. Interestingly, decreases in parenchymal amyloid were associated with an increased vascular amyloid burden [21, 23, 24]. This lead to the hypothesis that immunization results in solubilization of Aβ plaques, followed by perivascular drainage of the solubilized Aβ that leads to the development of CAA [5].

In this study, we have explored the relatively little studied Aβ43 peptide and compared its cerebral expression levels in AD cases to those of Aβ42 and Aβ40, and studied the effects of active Aβ42 immunotherapy. Aggregation propensity [15] and neurotoxic properties [8] of Aβ43 have been defined as indicators for an important pathogenic role of Aβ43 in AD. Knock-in mice containing a presenilin-1 R278I mutation overproduced Aβ43, and crossing these mice with APP transgenic mice resulted in offspring with elevated Aβ43 levels, impaired short-term memory, and accelerated amyloid pathology [29]. Our results in AD cases also indicated that large amounts of Aβ43 accumulated into plaques, to a degree that is similar to that of Aβ42 accumulation. The cerebrovascular expression of Aβ43 has received sparse attention in literature. It has been demonstrated in brain tissue of five AD patients that Aβ43 and Aβ42 are absent from CAA [13]. Using mass spectrometry, it was demonstrated that Aβ43 levels in SDS preparations of brain tissue, containing blood vessels or parenchymal deposits other than plaque cores, are substantially lower compared to Aβ40 and Aβ42 [36]. Together, these data suggested that Aβ43 does not readily accumulate in the vasculature. In our study, however, we observed cerebrovascular accumulation of Aβ43 in AD, although less vessels were positive for Aβ43 compared to Aβ42 and Aβ40. Of note, despite low levels of cerebrovascular Aβ43, we observed a high abundance of Aβ43 in plaques. Interestingly, we found a correlation between Aβ peptide length and plaque load (Aβ43 > Aβ42 > Aβ40), indicating that longer Aβ peptides have an increased tendency towards accumulation in the brain parenchyma. In contrast, an inverse correlation was observed between Aβ peptide length and CAA load (Aβ40 > Aβ42 > Aβ43), suggesting that shorter peptides have an increased tendency to accumulate in the cerebrovasculature.

We found a higher plaque load of Aβ43 and Aβ42 versus Aβ40 (Fig. 2g), which likely drives the positive association between plaque load and peptide length (Fig. 5a). It is known that Aβ42 and Aβ40 differ in that Aβ40 is mainly located in plaque cores whereas Aβ42 is present also in the diffuse component of plaques [14], and the immunohistochemical pattern of Aβ43 in our study largely resembles that of Aβ42 (Fig. 2). It is possible, though, that a non-significant number of plaques contain more Aβ43 than the other isoforms, or that the distribution of Aβ43 extends beyond the area stained for Aβ42 in some plaques. This point has not been addressed in this study and requires meticulous analysis of expression patterns in individual plaques, which was outside the scope of this study. Similarly, it is likely that Aβ40, Aβ42, and Aβ43 are – a least to some extent – colocalised in blood vessel walls. Future studies including double or triple Aβ staining in brain tissue of AD and iAD patients are needed to provide insight into the exact co-localisation of various Aβ isoforms in plaques and CAA.

One hypothesis to explain the relatively low abundance of Aβ43 in the cerebrovasculature compared to other isoforms in AD, is a more efficient Aβ43 clearance at the BBB. Analysing the fate of Aβ43 following immunotherapy may provide insight in this potential efficient clearance of Aβ43. Similar to Aβ40 and Aβ42, the Aβ43 load in the parenchyma is decreased in response to Aβ42 immunotherapy. This is not surprising, as the majority of antibodies produced following AN1792 immunization are directed against the N-terminus of Aβ, which is shared by all three Aβ isoforms [19]. Interestingly, despite the absence of a gross significant difference in cerebrovascular expression of the three peptides in the two cohorts, Aβ43 was the only isoform that did not display the gross - yet not significant - increase of leptomeningeal CAA that was seen for Aβ42 (275%) and Aβ40 (165%) after immunization. This is noteworthy, considering the fact that Aβ43 is the isoform that was most rigorously removed from plaques (5.5 times lower Aβ43 load compared to a 2.3- and 2.5- fold reduction for Aβ42 and Aβ40 respectively), indicating that following immunotherapy Aβ43 is readily released from plaques. However, despite its increased availability for transport towards the vasculature, Aβ43 is less prone to accumulate at the vasculature than Aβ42. This suggests that Aβ43 might be more efficiently cleared at the BBB. A potential alternative explanation for the relatively low abundance of Aβ43 in the AD vasculature could be its high propensity to aggregate [15, 29], which may prevent Aβ43 from reaching the vasculature. There are, however, several arguments against that scenario. First, if its high aggregation propensity would explain the low abundance of Aβ43 in CAA, it would be expected that after solubilization of Aβ43 from plaques, its levels in CAA would increase. This is, however, not the case. Second, a high aggregation propensity does not necessarily prevent vascular accumulation of Aβ, as can be deduced from observations on the Dutch mutant Aβ40(E22Q) which is both very prone to aggregate and widely observed in the cerebrovasculature in mutation carriers [11, 18]. Therefore, the low abundance of Aβ43 in CAA seems to be more consistent with our hypothesis of a more efficient clearance of this Aβ isoform across the BBB.

We used immunohistochemistry to assess the antigen load of the different Aβ isoforms. Although this is a widely accepted method for relative protein quantification [6, 29, 30], a possible limitation of our study is that comparison of antigen load stained by different antibodies is not fully possible. However, by carefully selecting thresholds for every antibody, we ensured that only specific immunoreactivity for each antibody was captured, allowing for relative assessment of the expression of the different Aβ isoforms. However, immunohistochemistry does not yield a measure of absolute protein levels, and the observed differences between Aβ isoforms are relative.

Using immunoassays, we showed that the 21F12 antibody, often used for its assumed Aβ42-specificity, also recognized Aβ43, albeit more weakly. This observation has been shared and acknowledged by many researchers through the use of the term “Aβ42(43)-specific” when describing the 21F12 antibody. Many antibodies thought to be specific for Aβ42 also recognize longer variants, a fact that should be kept in mind when interpreting results obtained using such antibodies. We do not expect that this cross-reactivity has substantially affected our results, as the vascular and parenchymal changes of Aβ42 accumulation due to immunotherapy have been described before and our findings are in line with these previous studies [5, 23].

Earlier studies in our cohort have shown a decrease in the number of plaques and an increase of vascular Aβ42 and Aβ40 in the early years after immunotherapy [5, 30]. We confirmed significantly lower levels of Aβ42 and Aβ40 in plaques of iAD patients compared to unimmunized patients. However, we did not detect significantly higher numbers of Aβ42- and Aβ40-affected vessels in the immunized group. A possible explanation might be that in the current study only one brain region, the middle temporal gyrus, was included in the analysis, whereas in earlier publications multiple brain regions were studied, including the frontal, middle and temporal gyrus, and thus we might be less powered to detect these changes. Another, more important, explanation may be that, in the current study, we have expanded the iAD population with patients who survived longer after immunization than in previous studies. It is possible that the effects of active immunization on cerebrovascular Aβ levels may be transient or less pronounced in the long term. Indeed, we found indications of a negative correlation between Aβ42 levels in CAA and survival time (parenchymal Aβ42: r = − 0.605, p = 0.022; leptomeningeal Aβ42: r = − 0.629, p = 0.021), although the small number of cases precludes any firm conclusion.

Another potential confounder may be that, in line with previous observations, the extent of Aβ release from plaques is positively associated with the level of the immune response [12]. In addition, we observed a trend towards lower levels of Aβ in CAA in cases with a strong immune response (data not shown), indicating that, in addition to plaque removal, immunotherapy may lead to the removal of vascular Aβ, as has been demonstrated before in mouse models [3, 25]. However, again, the small number of cases and the limitation of only one studied brain region precludes us from drawing firm conclusions on these potential relations. Further studies on the long-term effects of active immunization on CAA levels are needed.

We show that Aβ43 is not only an important constituent of plaques but also associated with CAA, albeit at lower levels than Aβ42 and Aβ40, suggesting that the expression pattern of Aβ pathology is isoform-specific. Furthermore, neuropathological analysis from the first human immunotherapy study also yielded preliminary evidence that Aβ43 may be more efficiently cleared at the BBB than Aβ42 and Aβ40, although mechanistic studies are needed to confirm this suggestion.