Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and the most common form of dementia. It is characterized by the extracellular deposition of amyloid plaques and the formation of intracellular neurofibrillary tangles comprised of tau protein [1]. Amyloid plaques are mainly comprised of amyloid-beta (Aβ) peptides. Aβ is formed by enzymatic digestion of Aβ precursor protein (APP), giving rise to Aβ peptides of different lengths [2]. The 40 amino acid long peptide (Aβ40) is the most abundant Aβ form in the human brain followed by Aβ42, an isoform that is more prone to aggregate. Under physiological condition, Aβ exists as an unfolded monomer with studies demonstrating possible physiological roles of this peptide in neurogenesis and lipid homeostasis [3,4,5,6]. However, during AD, partially folded Aβ monomers start to form aggregates of different sizes and solubility, ranging from soluble oligomers with hairpins and protofibrils to insoluble fibrils and finally amyloid plaques [7, 8]. Soluble aggregates (oligomers and protofibrils) have been demonstrated to correlate with AD progression and are suggested to be the most neurotoxic species of Aβ [9,10,11,12,17, 38, 39, 61] was recombinantly linked to soluble NEP (sNEP) and an Fc-region fragment (scFc) of mouse IgG2c antibody, forming the final protein drug (sNEP-scFc-scFv8D3). In addition to the mentioned protein, we designed another recombinant protein (muNEP-scFc-scFv8D3) that was based on a mutated variant of NEP displaying lower efficiency in degrading other NEP substrates compared to Aβ [44].

Following purification, the two NEP proteins were present mainly in the monomeric form with a strong band in SDS-PAGE at around 160 kDa (Fig. 2A). However, SDS-PAGE analysis demonstrated some weaker bands that were bigger than 160 kDa, suggesting the formation of protein multimers. A previous study has demonstrated the dimerization of naturally occurring NEP [62], which could suggest that the formation of sNEP-scFc-scFv8D3 and muNEP-scFc-scFv8D3 multimers is related to the presence of NEP in these designs. Nevertheless, the possibility that the other domains such as scFc contributed to the formation of these multimers cannot be entirely excluded. A rather rapid degradation of the NEP constructs could be detected in vivo, a possible reason for this could be the propensity to form multimers which often leads to more rapid degradation.

The ability of sNEP-scFc-scFv8D3 to degrade Aβ was first assessed in vitro using sNEP as the positive control. In our designs, the N-terminal of sNEP is fused to other protein domains (scFc-scFv8D3). Therefore, we wanted to evaluate whether such protein engineering might interfere with the ability of NEP to degrade Aβ. sNEP-scFc-scFv8D3 preserved NEP properties in binding to and degrading Aβ monomers in vitro (Figs. 3 and 4). These results indicate that the recombinant addition of scFc and scFv8D3 on the N-terminal end of NEP is a suitable strategy to develop NEP-based protein drugs with long half-life and capability to cross BBB, respectively.

sNEP-scFc-scFv8D3 could, similar to sNEP, significantly degrade Aβ1-40 and Aβ1-42 monomers in vitro (Fig. 5) both at the 1:1 and 5:1 ratios. Surprisingly, the recombinant protein based on the mutated NEP protein (muNEP-scFc-scFv8D3) was not able to degrade wt-Aβ at a 5:1 ratio, where the concentration of Aβ was higher, 2.5 μM. Degradation was evident only when a 1:1 ratio of Aβ and NEP proteins was used and with a 0.5-μM concentration of Aβ (Fig. 4). The higher the concentration of Aβ, the faster the aggregation is a well-known fact. It is possible that the concentration of 2.5 μM was high enough to start aggregation already before the first time point analysed and that the muNEP is worse at degrading aggregates than sNEP. muNEP has been previously reported to exert a 20 times higher Aβ degradation capacity compared to wild-type NEP but this was at much higher concentrations of Aβ [44] and a much lower ratio than we had at other concentrations it seemed from the data that it was less effective. Wild-type NEP (sNEP in our study) has the capacity to degrade Aβ at several cleavage sites between amino acids 1 to 35. However, muNEP preferentially degrades Aβ at position 20–21 located at the mid-region of Aβ [44]. This region becomes gradually hidden with the formation of aggregates, further suggesting the inability of muNEP to degrade aggregated forms Aβ.

The in vitro Aβ degradation results (Fig. 4) demonstrated the ability of sNEP-scFc-scFv8D3 to efficiently degrade both Aβ1-40 and Aβ1-42 monomers. Interestingly, sNEP-scFc-scFv8D3 could degrade Aβ1-40 faster than Aβ1-42 at high Aβ concentrations (Fig. 4A–C). NEP cleavage sites are located within the first 35 amino acids of Aβ [63], meaning that the enzyme should have an equal degradation efficiency on both Aβ isoforms. NEP is more efficient in degrading monomers, and since Aβ1-42 is more prone to aggregate, it is more likely that this isoform started to aggregate during the experiment, hence slower degradation of this isoform compared to Aβ1-40. Our results are in agreement with [64], demonstrating NEP ability to degrade synthetic Aβ1-40 and Aβ1-42 peptides in vitro. Similar to our results, the mentioned study demonstrated NEP ability to degrade Aβ1-40 faster compared to Aβ1-42. On the contrary, we have recently demonstrated a selective degradation of Aβ42 (but not Aβ40) in the hippocampus of AD mice mediated by enhanced NEP expression following treatment with a brain-penetrating somatostatin peptide [17]. However, our previous study demonstrated the degradation efficiency of membrane-bound NEP, while the current study is based on the use of the soluble part of NEP (sNEP). These studies suggest NEP-mediated degradation of Aβ to be dependent on several factors such as the availability and concentration of Aβ, its aggregation state and the form of NEP (membrane-bound or soluble).

A bit to the contrary of the discussion above, both NEP constructs degraded the fast-aggregating arctic Aβ the best in the conditions used. A possible reason for this can be that arctic Aβ is known to form protofibrils but less of insoluble aggregates [46]. The structure of these protofibrils has not been determined, but they possibly retain the beta hairpin that is also present in oligomers and the first thing that forms when monomers start to aggregate. This hair pin converts to larger beta sheets in the larger aggregates. In the modelling of NEP binding site to Aβ, it seems like this hair pin still could be accessible while the beta sheets of the aggregates would not be. That is, it is possible that NEP can degrade the aggregates formed by arctic Aβ, but not other types of aggregates. That this effect is the strongest for muNEP could be due to that the altered binding of muNEP compared to NEP is exactly on the amino acid 22 which is the aa that is mutated in the arctic Aβ.

By recombinantly linking the TfR binding moiety (scFv8D3) to our constructs, we could detect high brain uptakes of the NEP protein drug in tg-ArcSwe mice. Brain uptake was 20 times higher compared to the sNEP protein lacking scFv8D3 (Fig. 5A), which is a similar increase as seen with other scFv8D3 fusions to large proteins [17, 39]. In addition, high concentrations were detected in the hippocampus (Fig. 5B), a brain area where Aβ pathology starts [45], and where strong Aβ immunostaining is detected in young tg-ArcSwe mice (Fig. 5C). We have previously also analysed how much of free iodine that enters the brain to see if degradation of constructs is likely to have an effect on the measured uptake but we see that very little free iodine can enter the brain and are therefore not likely to significantly contribute to the signal [65].

The addition of scFc domains prolonged the blood half-life of the NEP-based recombinant proteins compared to the short half-life of soluble NEP. Nevertheless, significantly lower plasma concentrations, twofold shorter blood half-life and fivefold lower drug exposure (AUC blood) of sNEP-scFc-scFv8D3 and muNEP-scFc-scFv8D3 were displayed compared to scFc-scFv8D3 (Fig. 6C–F). Binding of NEP present in these designs to Aβ, but also other NEP substrates, in the blood followed by subsequent degradation could be one of the possible reasons of the low concentrations of sNEP-scFc-scFv8D3 and muNEP-scFc-scFv8D3 detected in the blood. The high liver concentrations of sNEP-scFc-scFv8D3 at the early time points (Fig. 5E) further suggest protein degradation. In addition to the low concentrations in the blood, the brain concentrations of NEP proteins drastically decreased 72 h post-injection (Fig. 6B), opposite to what has been previously reported with other Aβ binders containing scFv8D3 [41], suggesting fast clearance of the NEP proteins from the brain. The same pattern of low NEP protein concentration was also evident in the lung and heart (Fig. 6G), two organs containing several NEP substrates [43, 66, 67].

The ability of the recombinant NEP proteins to degrade Aβ in vivo was investigated using tg-ArcSwe mice treated with a single therapeutic dose of 30 nmol/kg. sNEP-scFc-scFv8D3 and muNEP-scFc-scFv8D3 significantly decreased aggregated Aβ concentration in plasma but not of monomers (Fig. 8A, B). The significant reduction in aggregated plasma Aβ levels did not significantly affect the levels of Aβ aggregates, except oligomers in the brain (Fig. 7). In the brain, we saw a small but significant reduction of Aβ monomers in the TBS-T fraction and of Aβ oligomers as detected by the A11 antibody that detects oligomers with a beta hairpin in the TBS fraction. This beta hairpin is lost when the aggregates grow and form fibrils. The reduction we see in hairpin containing Aβ goes in line with the possibility that neprilysin is efficient in degrading hairpin containing oligomers as we discussed above. Since no significant reduction in the amount of Aβ42 monomers could be detected in the brain, we assume that it is Aβ40 that is reduced. It is also the main type of Aβ in the ArcSwe mice.

There is also a possibility that the effects seen in the brain are due to the reduction of Aβ in the blood, but since it is mainly a reduction in aggregates in the blood and not monomers, we think that is less likely.

Our findings are in accordance with previous studies using similar proteins consisting of recombinant NEP linked to either an albumin protein [68] or the Fc region of IgG antibodies [69]. NEP proteins used in these studies demonstrated no alteration in Aβ concentration in the brain despite the significant Aβ reduction detected in the periphery. However, none of these studies used a BBB transporter and was based on the sink hypothesis of Aβ equilibrium between the brain and periphery. In our study, we used scFv8D3 as the BBB transporter which facilitated the delivery of recombinant NEP proteins into the brain. In addition, treatment with sNEP-scFc-scFv8D3 was associated with a selective reduction (p value 0.03) in membrane-bound Aβ monomers, but not aggregates. These findings are similar to previous studies demonstrating more efficient degradation of Aβ monomers by NEP compared to oligomers [33].

The in vitro experiments demonstrated that NEP-based proteins were capable of degrading arctic-Aβ, opposite to what has been previously reported [70]. Based on the in vitro results, we chose the tg-ArcSwe AD mouse model for the in vivo experiments. Nevertheless, treatment with scFv8D3-fused NEP proteins failed to display any effects on the concentration of arctic-Aβ in tg-ArcSwe mice, but this can be attributed to the ELISA that is set up. With the ELISA used, one will only detect aggregates that have both the N-terminal and the middle available for binding, and likely mostly aggregates will be detected due to the coating with the 3D6 antibody that binds both aggregates and monomers, but that likely will favour aggregates due to the avidity effect. The short retention time of the injected NEP in the brain will likely reduce the possible effect the NEP have had in the brain.

Limitations

One limitation can be the small sample size, especially in a treatment study where n = 3–4 per treatment group. However, since we see this study as a proof of concept demonstrating the treatment effects of BBB-penetrating formats of NEP, we followed the 3R principles and used the minimal number of animals needed to display a statistically significant effect.

Conclusion

Overall, the present study shows the potential of using TfR-mediated transcytosis to successfully deliver NEP proteins into the brain. By adding a scFc and a scFv8D3, our new NEP-based protein designs displayed longer blood half-life and 20 times higher brain uptake compared to sNEP. The retained enzymatic activity in vitro and the significant reduction of Aβ monomers and Aβ oligomers in vivo following a single intravenous injection highlight the potential of recombinant NEP-based proteins in AD although a modified construct with better brain retention would be desired.