Fulfilling the neuropathological criteria of AD
We performed neuropathological examinations of the brains from four patients who carried the Arctic AβPP mutation and whose clinical picture complies with AD. We could demonstrate that the neuropathological hallmarks of AD, Aβ plaques and deposition of hp-tau, were prominent features in all four patients’ brains. However, the Arctic AD pathology has certain features that deviate from the common AD pathology , as it is referred to in both of the latest consensus criteria [23–25].
When the ABC system of the new 2012 NIA-AA criteria [23, 25] is applied, the distribution of Aβ plaques, i.e. NIA-AA component A = 3, since the accumulation of Aβ is florid even in cerebellum (= Thal phase 5, ). As for the neurofibrillary tau pathology (component B), the presence of NFTs throughout the neocortical regions fulfils the criteria of Braak stage V-VI , i.e. NIA-AA component B = 3. Braak stage V-VI is also met, if the Brain Net Europe (BNE) staging is applied , as the neocortical NTs are abundant in the striate area (occipital cortex; Additional file 2: Figure S7a). The grading of component C in the Arctic brains according to CERAD as recommended in NIA-AA criteria is somewhat problematic, because of the exceptional plaque structure. Although Aβ plaques are compact, they are devoid of fibrillar amyloid cores and only rarely harbour robust DNs. Nevertheless, we propose that they do fulfil the CERAD criteria for neuritic plaques, since they display delicate hp-tau positive NTs and occasional DNs. The frequency of plaques should be considered frequent, i.e. the component C = 3 .
Taken together, our neuropathological findings in the Arctic AD patients’ brains meet both the 2012 NIA-AA criteria for high level of AD- related changes (A3, B3, C3) [23, 25] and the previous NIA-RI criteria for a high likelihood of AD . The very rare α-synuclein positive neurons encountered (not shown) most likely did not contribute to the patients’ clinical picture. Neither did we find any significant ischemic pathology, although CAA was relatively prominent. Thus, the dementia caused by the Arctic AβPP mutation is due to AD, albeit with unique features of AD neuropathology.
Interestingly, the large, round cortical Arctic plaques bear resemblance to the cotton wool plaques, such as the Aβ deposits found in AD patients with PS1 mutations [2–4]. Similarly, the cotton wool plaques and the Arctic plaques are devoid of amyloid cores and harbour very few hp-tau-positive DNs . However, cotton wool plaques have a homogeneous composition (Additional file 1: Figure S9), whereas the Arctic plaques are composed of variably truncated and spatially differentially distributed Aβ and therefore when immunostained with different Aβ antibodies they appear targetoid.
Amount of Aβ deposition
The extent of Aβ deposits in the Arctic AD patients’ brains was massive, which may explain why the weights of Sw1, Sw2 and Am2 brains were considerably higher than what is commonly recorded in sporadic AD. Alzheimer brains are usually atrophic, but no studies have systematically compared the brain weight with the Aβ burden. Interestingly, the brain weights in three of the five PS1Δ9 AD patients with large cotton wool plaques of similar abundance as Aβ plaques in our Arctic AD patients’ brains were within normal limits .
Hierarchical order of Aβ deposition
In sporadic AD, Aβ deposition has been suggested to occur hierarchically in five phases . The distribution of Aβ deposits in the Arctic AD patients’ brains corresponds to Thal’s most advanced phase 5, for which the requirement is the presence of Aβ also in cerebellum. Aβ deposits were highly abundant in all Arctic AD patients’ cerebellum, but they were few in brain stem nuclei–the criterion for phase 4. These features suggest that, in these brains, cerebellum may have become involved at an earlier phase or at a faster pace than the brain stem. Thus, the Arctic AβPP mutation may alter the hierarchy according to which the Aβ deposits are generally seen to emerge in AD patients’ brains. Perhaps abundant extracellular perivascular Aβ aggregates in Arctic AD brain overwhelm the perivascular drainage pathways  in cerebellum leading to an early pathology. Moreover, the pattern of Aβ deposition and distribution in our Arctic patients’ cerebellum was exceptional (see next section).
Variation in the distribution of Aβ deposition
The pattern and composition of Aβ deposits in the Arctic AD brain revealed additional topographic variability. The ring plaques, originally discovered by Bielschowsky silver and immunostaining with abAβx-42
, were − with the use of additional Aβ antibodies − shown to be targetoid rather than ring-shaped (as preliminarily demonstrated in our previous article  and now in greater detail in this article). These plaques were almost only observed in cerebral cortex, whereas in other anatomical locations the plaques were usually of more irregular and diffuse types.
Interestingly, in claustrum the pattern of Aβ deposits was similar to that in cerebral cortex. This is consistent with a previous observation of cotton wool plaques in PS1Δ9 AD patients’ brains . In addition, the tau pathology in claustrum was similar as in cerebral cortex. Claustrum is generally considered to be a part of the basal ganglia and the presence of plaques in this region would thus correspond to Thal’s phase 3. However, this pattern was strikingly different from that in the other subcortical grey matter nuclei. For example, globus pallidus was virtually negative for Aβ plaques, while only small diffuse plaques could be found in putamen, thalamus and caudate nucleus. This discrepancy may be related to the fact that, of these regions, only claustrum has bidirectional connections with almost all cortical regions .
The pattern of cerebellar Aβ deposits differed remarkably from the commonly observed restriction of deposits to the molecular layer in sporadic AD  and from the abundance of cored plaques in PS1Δ9 AD patients’ cerebellum. In the Arctic AD patients’ brains, the abundant Aβ deposition next to the Purkinje cells may indicate active Aβ production in these cells. Moreover, the apparent pattern of Aβ deposition along the perivascular drainage pathways (also depicted in NIA-AA article by Hyman et al. 2012 ) may be explained by a relatively profuse transport of extracellular Aβ through the molecular layer to the subarachnoid space .
Differential truncation of Aβ
The composition of Aβ deposits in both sporadic and familial AD brains has been shown to be highly variable, with both N- and C-terminal truncations and modifications [32–37].
Our mass spectrometric analyses, in which we immunoprecipitated both wt and mutated Aβ from the temporal cortex of the patient Sw2 with antibodies against two distinct Aβ epitopes (ArcAβ17–24 and wtAβ17–24), demonstrated that the deposits contained variably truncated and modified Aβ species, both wt and Arctic Aβ. Furthermore, our immunohistochemical stainings revealed that although all four patients carried the same AβPP mutation, both the distribution and the composition (truncations or modifications) of the Aβ deposits showed considerable inter- and intra-individual variability. This should not actually be surprising, since it is most likely that the “machinery and milieu” in the AβPP production, processing and/or Aβ aggregation are not identical in the four individual Arctic AD patients, nor in different anatomic regions within each patient’s brain. Whether a similar variability in the composition of Aβ deposits exists in other forms of FAD has been the subject of several studies [5, 34–37].
The variability in the composition of deposited Aβ also sets new requirements for the antibodies used in diagnostic work. Since truncations can occur at many different sites of the Aβ peptide, some antibodies give virtually negative parenchymal staining. Thus, either an antibody against the mid-portion of Aβ preferably against an epitope located C-terminally to aa 11 (to avoid problems with 11pE truncated Aβ species) and N-terminally to aa 40 (to avoid problems with the different C-terminal truncations), or a mixture of different antibodies should be recommended for routine analyses. In our hands, the antibody abAβ17–24 (clone 4G8) appeared to best fulfil such requirements (even though it also recognizes AβPP
), whereas the other widely used mid-portion antibody abAβ8–17 (clone 6F/3D) yielded inconsistent results. For instance, it gave only vague staining of the cerebellar parenchymal deposits, although it stained blood vessels in this region clearly positively.
It is known that the Arctic mutation interferes with the processing of AβPP by making it less prone for α-secretase cleavage, while elevating β-secretase cleaved fragments [12, 38]. Natural processing of AβPP at the β′-site (at Aβ aa 10/11) is also favoured over the β-site (at Aβ aa −1/1) in situations when accession to it by BACE1 enzyme is affected, e.g. through structural twists in AβPP . Moreover, N-terminally truncated and modified Aβ peptides (e.g. AβpE3-x and AβpE11-x) have been shown to be significantly increased in the brains of AD patients with various PS1 mutations [36, 40]. Our data show that the Arctic mutation like the PS1 mutations referred to above may increase cleavages at aa 3 and 11 since we observed 3pE40arc, 11pE40arc and 11pE42arc in our MS measurements (Figure 3e, Additional file 5: Table S1 and ), and considerable immunopositivities with ab3pE and ab11pE antibodies. However, since we also observed a heterogeneous population of N-terminally truncated Aβ peptides, additional (primary/secondary) cleavages are also likely to occur in the Arctic AD brain. Aβ-peptides might also aggregate in a way that exposes cleavage sites and facilitates peptide truncation and their modifications as a secondary process.
Lack of fibrillar Aβ and amyloid cores is a characteristic feature of both Arctic plaques and PS1Δ9 cotton wool plaques. The reason for this feature in these two genetic variants of AD is unknown. Certain properties of the mutated forms of Aβ such as posttranslational modifications and altered propensity to oligomerize/aggregate, could offer an explanation for the limited formation of structurally ordered Aβ fibrils in Arctic AD patients , whereas the underlying reason in the PS1Δ9 AD patients is even less clear, because the cotton wool plaques in such brains contain wild-type Aβ only. It might be that in these two types of FAD Aβ aggregation is too rapid and does not lead to Congo-red positive, fibrillar Aβ deposits.
It has been shown that the Arctic mutation leads to an accelerated oligomerization and disordered fibrillogenesis of Aβ, measured both in vitro
[11, 41–44] and in vivo
[11, 45, 46]. The diameter of Arctic Aβ fibrils correlated with decreased neuronal viability . Recent in vitro experiments on the aggregation process of ArcAβ1–40
 demonstrated that at least four types of fibrils can be identified. The intermediate phase of spherical aggregates appeared at earlier time points and ArcAβ1–40 fibrils polymerized more rapidly and at lower concentrations than wt Aβ1–40 fibrils. At late stages fragmentation and clustering of ArcAβ1–40, but not of wtAβ1–40, fibrils were observed . The results of these experiments are in agreement with the suggestion that spherical aggregates (containing abundant β-hairpin and/or β-sheet structures), and/or Aβ oligomers have a pathogenic role in the AD brain. Especially the larger soluble oligomers, i.e. protofibrils, are known to have neurotoxic properties. However, an alternative Aβ aggregation pathway, different from simple assembly of spherical aggregates and protofilaments into fibrils, has also been proposed , which may contribute to the distinct morphology of Aβ plaques in Arctic AD patients (i.e. Congo-red negativity of the Arctic AD deposits). Moreover, co-incubation of ArcAβ with wtAβ1-40 led to kinetic stabilization of Arctic protofibrils . An increase in the ratio of ArcAβ to wtAβ in Arctic AD may result in the rapid accumulation of neurotoxic protofibrils and acceleration of the disease process .
The large plaques in both the Arctic and PS1Δ9 AD patients’ brains were found to embrace neurons. More specifically, in all four Arctic brains seemingly viable neocortical pyramidal neurons and cerebellar Purkinje cells could be identified within several plaques of variable Aβ composition. Interestingly, the perikarya of these neurons seemed intact and did not appear to be under way to develop neurofibrillary pathology. This sparing of neuronal perikarya is in accordance with the notion that intraneuronal Aβ triggers neuron loss in AD . In several animal models it has been demonstrated that extracellular Aβ plaques do not seem to instigate neuronal death, whereas accumulation of intraneuronal Aβ correlated well with the loss of neurons , including a transgenic model expressing pyroglutamate Aβ3pE-42
We showed that even apparently intact axons traversed the plaques. These observations are phenomena, similar to those in PS1Δ9 AD patients, i.e. axons are not pushed aside to wind around the accumulated Aβ. However, the relatively low number of neurofilament positive intraplaque axons may indicate that axons traversing Arc plaques suffer from some degree of degeneration, as it was interpreted to occur within PS1Δ9 AD patients’ cotton wool plaques . The intraplaque accentuation of NTs (i.e. axons containing hp-tau) in both Arc and PS1Δ9 plaques supports the suspicion of axonal degeneration. Accumulation of pathogenic species of microtubule associated tau protein can impair axoplasmic transport and consequently contribute to synaptic loss, which may be pivotal in the pathogenesis of AD . If the synaptic contacts are lost, sparing of the perikarya or axons en route cannot prevent functional loss (and further degeneration). The fate of axons within Aβ plaques obviously merits further analysis.
Furthermore, the non-fibrillar type of Aβ deposits in the Arctic brain induced only a limited reactive response. Although the density of astrocytic processes was increased within the Arctic Aβ deposits (clearly visible around Purkinje cells), the astrocytic cell bodies appeared not to cluster around the non-fibrillar Aβ deposits. Likewise, in PS1Δ9 AD patients’ brains astrocytes did not cluster around the non-fibrillar cotton wool plaques, although in these brains the scarce cored plaques were surrounded by an increased number of astrocytes . Similarly, the Arctic non-fibrillar plaques did not appear to attract microglial cells. The presence of preserved neuronal perikarya and axons as well as the lack of activated glial cells strengthen the perception of extracellular non-fibrillar Aβ deposits as being relatively non-toxic. Thus, it is conceivable that Aβ oligomers, which are considered pathogenic, exert their effects already within the neurons or by being diffusely distributed (not in plaques) in the parenchyma.