Prion-like cell lysate seeds plaques in vivo
The seeding of amyloid plaques by intracerebral injection of AD brain material into an AD-model mouse is well established [40]. However, the use of a homogenized brain, even if fractionated, makes it difficult to know what species of Aβ are responsible for the seeding and to determine if the Aβ came from an extra- or intracellular source. We previously reported on a prion-like clonal line of AD Swedish-mutant APP N2a neuroblastoma (SWE) cells that consistently produce and maintain intracellular aggregates of Aβ. We initially induced these aggregates by treating SWE cells with brain lysate from an AD-model transgenic mouse and then performed single-cell cloning to isolate lines with intracellular aggregates of Aβ [25]. We then showed that the lysates of our prion-like cells could induce Aβ aggregation in other SWE cells, demonstrating the presence of prion-like Aβ that could propagate in vitro in the cell line. In the present study, we show that these cell lysates are also capable of seeding in vivo. To this end, we performed unilateral intrahippocampal injections in 7-week-old 5xFAD transgenic mice with SWE cell lysate, prion-like SWE cell lysate, or APP/PSEN1 transgenic mouse brain lysate (positive control), using the contralateral hippocampus as a negative control. The mice were sacrificed 16 weeks post-injection and immunohistochemically processed for analysis (Fig. 1a; all antibodies used are in Table 1 in the Methods). Plaque quantification was done in the dorsal parts of the dentate molecular layer and CA1 stratum lacunosum-moleculare and stratum radiatum as shown in Fig. S2, online resource. No seeding was seen with the injection of control SWE cell lysate (Fig. 1b), while modest seeding was induced around the hippocampal fissure, in the molecular layer of the dentate gyrus, and in the stratum lacunosum-moleculare of CA1 by the injection of prion-like SWE cell lysate (Fig. 1c). In contrast, injection of APP/PSEN1 brain lysate caused robust seeding, including the area around the hippocampal fissure as in the brains injected with prion-like cell lysate. However, the strongest labeling was in the lower blade of the dentate gyrus, in particular in the outer molecular layer corresponding to the perforant path terminals of the lateral entorhinal cortex (LEC) (Fig. 1d). We thus demonstrate that an intracellular source of Aβ can seed amyloid deposition in vivo in a prion-like manner, albeit less robustly than brain extract.
Seeded Aβ plaques follow anatomical pathways
Since the seeded Aβ seemed to associate with select anatomical pathways, the LEC perforant path, and the dorsal fornix of hippocampus, we more closely followed the progression of induced plaque pathology to obtain further insights into the cellular and anatomic mechanisms of seeded Aβ aggregation in vivo. APP/PSEN1 mouse brain homogenate was injected unilaterally into dorsal hippocampus of 7-week-old 5xFAD mice and sacrificed at 4, 6, 10, and 16 weeks post-injection (Fig. 2a). Already at 4 weeks post-injection, Aβ-induced seeding was evident in the injected side in the fibers of the alveus, corpus callosum, fornix, and the external capsule by the presence of immuno-fluorescently labeled Aβ (Fig. 2b). Notably, this initially seeded Aβ was consistently located somewhat anterior to the injection site (the injection was around − 2.5 mm Bregma). While at 4 weeks post-injection, the induced Aβ-signal was mostly confined to white-matter tracts, some induced Aβ could be seen in the brain parenchyma of CA1 stratum oriens adjacent to the alveus (Fig. 2c).
To ascertain what was the original injection material and what were induced plaques, we also injected WT mice with brain homogenate mixed with India ink. We sacrificed these mice 1, 4, and 6 weeks post-injection and observed India ink and human Aβ staining in the anterior parts of the external capsule and some speck-like labeling in the corpus callosum (Fig. S3, online resource). Notably, the Aβ staining always co-localized with the India ink and was faint in WT compared to 5xFAD mice, got fainter with time in the WT mice, and was not evident outside the aforementioned white-matter tracts. Thus, a small part of the injected Aβ in the white-matter tracts is likely from the original injection material in the 5xFAD mice. In contrast, the Aβ in the underlying grey matter outside the white-matter tracts is induced.
At 6 weeks post-injection, induced Aβ in 5xFAD mice was apparent in the injected side of the dorsal fornix and became more pronounced in the corpus callosum, alveus, and external capsule (Fig. 2b). At this time-point, we also observed robust plaque-like structures outside of the white-matter tracts, primarily in the CA1 stratum oriens underlying the corpus callosum. Interestingly, in one animal in this 6-week post-injection group, we also observed Aβ plaques in the border zone of the outer molecular layer of the dentate gyrus and the stratum lacunosum-moleculare of CA1, i.e., surrounding the hippocampal fissure (Fig. 2d). At 10 weeks post-injection, induced Aβ aggregates were found mainly in the dentate gyrus (Fig. 2b), although one mouse exhibited aggregates in the corpus callosum in addition to spread in the hippocampus (Fig. S4, online resource). At 16 weeks post-injection, all mice displayed prominent induced plaques concentrated in the molecular layer of the dentate gyrus that expanded into more posterior areas, along with aggregates/spread in the fimbria of the hippocampus (Fig. 2b, e) and the stratum oriens of CA3. Interestingly, none of the 16 weeks post-injection mice displayed prominent Aβ-aggregates in the fornix, alveus, or the external capsule, supporting the conclusion that such aggregates that were observed in these fiber tracts after shorter incubation times represent a transient phenomenon, peaking at 6–10 weeks post-injection (Fig. S5, online resource).
Loss of NeuN in stratum oriens
Due to the early appearance of seeded Aβ aggregates in parts of the stratum oriens of CA1, we conducted a closer examination of the neurons in that area. Remarkably, there was a decrease in NeuN labeling in this layer near to where the injection had seeded Aβ aggregates compared to the uninjected side (Fig. 3a, b). There was also greater heterogeneity in the intensity of NeuN labeling near induced Aβ aggregates with numerous nuclei that were weakly NeuN-positive (Fig. 3b). The loss of NeuN with intact DAPI suggests damage but not death of the neurons [11]. In the parts of stratum oriens of the injected side where plaques had not been induced, there was no decline of NeuN-positive cell labeling (Fig. 3c). The neurons in CA1 stratum oriens are interneurons and early loss of interneurons has been described in AD and the 5XFAD mouse [8]. We noted that the induced plaque-like labeling in the stratum oriens represented smaller deposits of aggregated Aβ than typical larger amyloid plaques (Fig. 3d). This wisp-like Aβ labeling, which among other markers was negative for GFAP, may be consistent with dystrophic neurites, which have been shown to accumulate the earliest small Aβ aggregates by immuno-EM [33, 35].
Dynamic intraneuronal Aβ changes in entorhinal cortex layer II and decreased intraneuronal Aβ in CA1 pyramidal neurons near plaques
In the dentate gyrus, much of the Aβ seeding corresponded to the terminal fields of afferents originating in LEC layer II neurons (a major component of the perforant path). We therefore further examined the cell bodies of LEC layer II neurons in brains that had Aβ seeding associated with the perforant path axonal terminals (Fig. 4a). Interestingly, the levels of intraneuronal Aβ in soma of LEC layer II were increased in the injected versus uninjected side at 6 weeks post-injection but then decreased in the injected versus uninjected side at 16 weeks post-injection (Fig. 4b). There was no difference in total Aβ immunofluorescence of the LEC between 6 and 16 weeks in the uninjected sides, which however is a statistically less powered comparison, since one then compares between different mice and sections rather than internally with the contralateral hemisphere of the same section. This is consistent with the Aβ of the injected LEC changing between 6 and 16 weeks, but not the uninjected side.
We also noted that the levels of Aβ in the cell bodies of the CA1 pyramidal neurons were decreased in the injected side, where plaques had been induced in the adjacent stratum oriens (Fig. 4c, d, e). One possibility was that the pyramidal cells were losing their intracellular Aβ to the nearby induced plaques. The induced plaques are located near the cell bodies and basal dendrites of the CA1 neurons. Another possibility is that Aβ is redistributed from CA1 cell bodies to processes in the injected side. It has been shown that primary AD-transgenic neurons accumulate aggregated Aβ in their processes with time in culture [33].
Extra- and intracellular pools of Aβ exist in an equilibrium that regulates Aβ production
Thus far, our in vivo work showed that intracerebral injection of prion-like seeds of Aβ not only induced plaques, but also affected intracellular (IC) Aβ. Furthermore, the decline of IC Aβ in CA1 pyramidal neurons suggested a possible equilibrium between IC and interstitial fluid (ISF) Aβ. In our previous work with the prion-like Aβ N2a cell line, we noted changes in APP processing with the accumulation of IC Aβ, namely increased β-cleavage but no change in α-cleavage [25]. For these reasons, we next sought to investigate how IC aggregation of prion-like Aβ might affect APP processing and the equilibrium between extracellular (EC) and IC Aβ. To this end, we manipulated levels of IC and EC Aβ in the prion-like cells and parent SWE N2a cells.
We first examined SWE cells without any prion-like induced IC Aβ. To remove the pool of EC Aβ, we replaced the media with conditioned media (CM) from untransfected N2a cells. We then performed Western blots and densitometric quantification, adjusted to actin, to assess levels of Aβ, the APP C-terminal fragment (CTF) C99, and full-length APP (flAPP) in cells and media at 3, 6, or 24 h after the media change (Fig. 5a). After 3 h, there was a greater than 80% decrease of IC Aβ (Fig. 5b), which could be accounted for by secretion of the IC Aβ. At the 6-h time-point, the IC Aβ levels had recovered to the same levels as control, while the EC levels were still low (Fig. 5c). Finally, 24 h after changing the media, the IC and EC Aβ levels reached levels similar to the controls (Fig. 5d). These changes are consistent with a prior report, showing that depletion of the EC Aβ pool is followed by depletion of the IC pool, after which the latter recovers first [24]. Interestingly, we also observed significantly increased C99 levels 6 h after the media change with untransfected N2a cell CM (Fig. 5e). C99 is a fragment of flAPP that is generated by β-secretase, and further cleavage of C99 by γ-secretase produces Aβ. In contrast to the C99 levels after 6 h, C99 levels were not altered at 3 h (data not shown) nor 24 h after media change (Fig. 5e). Finally, under no condition did we observe any significant change in the levels of flAPP. We also performed the 6-h low Aβ CM experiment in primary neurons and saw that C99 was similarly increased (Fig. 5f).
To examine whether decreased lysosomal degradation of C99 rather than increased β-cleavage of flAPP could cause the C99 increase with low CM Aβ, we repeated the media change experiment above but pre-treated the cells with the lysosomal inhibitor chloroquine. If decreased lysosomal degradation was the reason for increased C99 after 6 h of low Aβ media, then blocking degradation should eliminate the C99 difference between cells treated with low Aβ compared to baseline media (control). As expected, chloroquine increased Aβ in all conditions. However, there were more C99 in the low EC Aβ condition compared to control (Fig. 5g), supporting the conclusion that decreased lysosomal degradation was not the primary cause of the elevated C99 with low Aβ media. Further supporting this point, cells treated with the γ-secretase inhibitor DAPT, which inhibits the production of Aβ from C99, thereby increasing C99, still had elevated levels of C99 after 6 h of low Aβ media compared to control media (Fig. 5g), indicating that lower γ-cleavage is also not the primary reason for the increased levels of C99. Thus, it is likely that β-cleavage of APP is upregulated in response to low levels of EC Aβ. We summarize our findings from our Aβ equilibrium studies in N2a SWE cells in a schematic (Fig. 5h).
The addition of synthetic Aβ has been reported to increase Aβ production via increased β-cleavage in a cell line overexpressing APP [45]. The addition of exogenous Aβ will increase both EC and IC Aβ. To expand on this finding, we treated SWE cells with 1 µM of synthetic Aβ1-42 for 3, 6, or 24 h. Three hours after adding 1 µM of EC Aβ, there was no change in the levels of C99. However, 6 h post-treatment, there was a significant increase in C99. These levels were reduced by the 24-h time-point, but were still much higher than under control conditions (Fig. S7b, online resource). Notably, when measured 24 h after addition of Aβ1-42, the IC Aβ remained around 30 times higher in the treated cells compared to control (Fig. S7c, online resource). Taken together, these findings underscore the equilibrium of the extra- and intracellular pools of Aβ and show that both reduced and elevated EC Aβ can stimulate β-cleavage to elevate Aβ production.
Intracellular aggregation of prion-like Aβ disrupts the equilibrium between extra- and intracellular Aβ
In N2a SWE cells, reducing EC Aβ quickly lowered IC Aβ levels (Fig. 5b). Remarkably, when we repeated these experiments with our prion-like cell line, which have intracellular aggregates of Aβ (Fig. 6b), the changes in IC Aβ levels due to low Aβ media seen in the parent SWE line were no longer present. At the 3 h post-media change time-point, CM from untransfected cells (low Aβ) did not reduce IC Aβ in the prion-like cells (Fig. 6b, c). Notably though, C99 was still elevated after 6 h of low EC Aβ (Fig. 6b, d), indicating that low EC Aβ drives the increase in C99 and not low IC Aβ. It should also be noted that the prion-like cells constitutively have 3–4 times the amount of IC Aβ and up to 10 times higher levels of C99 compared to the parent SWE cells [25]. Thus, in our cell model, under conditions of both high IC Aβ and low EC Aβ, the production of Aβ could be permanently upregulated. We summarize our findings for this altered Aβ equilibrium in prion-like cells in a schematic drawing (Fig. 6e).
Induced Aβ aggregation in primary neurons and Aβ redistribution from soma to terminals
We next examined whether we could seed Aβ aggregation in primary neurons and how it affects the intraneuronal distribution of Aβ. To this end, we treated AD-transgenic primary neurons in culture with APP/PSEN1 transgenic mouse brain lysate. As controls, we treated transgenic primary neurons with brain lysate from either WT or APP KO mice and WT neurons with brain lysate from APP/PSEN1 or APP KO mice. While our previous study had shown that N2a cells could tolerate 3000×g supernatant from mouse brain lysate [25], primary neuronal cultures were more sensitive and died after the addition of this lysate regardless of whether it was from WT, APP KO, or APP/PSEN1 brain. To overcome this issue, we used brain lysate supernatant obtained from ultracentrifugation at 100,000×g. While this process removes the vast majority of Aβ, it has been shown that this fraction can still induce significant seeding in vivo [7]. We, therefore, treated APP/PSEN1 mouse primary neurons with 0.5% of media volume of 100,000×g supernatant from 21-month-old WT, APP KO, or APP/PSEN1 mouse brain at 12 and 19 days in vitro (DIV) with the final added Aβ concentration in the picomolar range. The primary neuron cultures were then either fixed or collected and pelleted at 28 DIV. To visualize potential seeding, we performed immunofluorescence with antibodies against Aβ42 and MAP2 (Fig. 7a, b). In the cultures treated with APP/PSEN1 brain lysate, we observed increased Aβ42-labeling in processes but reduced labeling in the soma. We also noted beading of the MAP2 labeling, indicating dendritic beading, which has been associated with excitotoxicity, dystrophic neurites, and AD [10, 30, 38]. In contrast, WT neurons treated with APP/PSEN1 brain lysate did not show dendritic beading or redistribution of Aβ42 (Fig. S7, online resource). We next performed dot blots on the collected pellets of primary neurons with the antibody OC (Fig. 7c) that reacts against aggregated amyloid proteins [16]. Remarkably, there was almost a twofold increase in OC labeling in cell lysates from the AD-transgenic cultures treated with APP/PSEN1 mouse brain lysate compared to controls (Fig. 7d). Taken together, we provide evidence that a minute amount of brain-derived Aβ can induce aggregation of Aβ in primary neurons, appears to redistribute Aβ from soma into processes, and causes dendritic beading.
Finally, to determine whether a similar redistribution to terminals occurs in vivo, we injected adeno-associated virus (AAV) with a vector encoding mCherry into the LEC of 5xFAD mice crossed with transgenic mice that express GFP in dentate granule cells. We observed distended mCherry-positive LEC layer 2 axonal terminals within developing plaques in the outer molecular layer of the dentate (Fig. 7e); this is in line with evidence showing the importance of axonal transport from LEC to plaque formation in the outer molecular layer of the dentate by perforant path lesioning [20, 31]. Furthermore, different sizes of plaques were evident, with the size proportional to the number of distended mCherry-positive LEC axonal terminals present, consistent with a progression in plaque size depending on the associated number of dystrophic neurites (Fig. 7e). The insets highlight these potential stages of plaque formation: in the largest plaque, Aβ labeling is seen in the center (core) but also co-localizing within mCherry-positive LEC axon dystrophies just internal to mCherry-positive but Aβ negative outer dystrophies, while the medium-sized plaque shows no core and mostly mCherry-positive/Aβ-negative dystrophies except in the middle (Fig. 7e, inset). Even a single mCherry dystrophy is evident, which, however, only has minimal Aβ labeling. In contrast, dentate granule cell dendritic GFP is not pronounced in these axonal dystrophy-derived plaques, consistent with the absence of intracellular Aβ accumulation in the dentate granule cells (Fig. 7f).