Anti-Aβ immunotherapy represents a major approach in current AD drug development but has had only mitigated clinical success so far. We present here the biochemical and pharmacological properties of a novel antibody, SAR228810, and its murine precursor, SAR255952. They have been selected and engineered to address some of the proposed limitations of current clinical mAbs: 1) specificity for soluble protofibrillar and fibrillar Aβ assemblies/conformations associated with AD pathology to increase efficacy and brain bioavailability; and 2) drastically reduced effector functions to further limit ARIA risks, thereby enabling higher tolerated doses.
Due to its biophysical properties, Aβ peptide (the Aβ40 or Aβ42 isoforms) can form a very broad spectrum of assemblies ranging from monomeric to low and high-n oligomers, soluble protofibrils, and insoluble diffuse deposits to aggregated plaques [
11]. However, these different conformations lack a robust biophysical characterization, are highly dependent on experimental conditions, and the presence of some of them in pathological tissues has not been demonstrated [
11]. It remains highly debated which Aβ conformation(s) should be neutralized/cleared for therapeutic benefit. Starting with brain pathology studies, soluble oligomeric forms of Aβ have attracted renewed attention. Initially, in case–control studies, brain soluble Aβ levels were correlated with AD dementia [
12,
13]. More recently in a series of autopsy cases selected with a similar brain amyloid load, high brain soluble oligomeric Aβ was shown to differentiate demented (CDR > 1) versus non-demented (CDR = 0) cases [
14]. Brain soluble oligomeric Aβ was shown to have a strong functional impact, inhibiting acute development of long-term potentiation in vitro and in vivo and leading to neuronal neurite disruption [
15‐
18]. These Aβ high molecular weight soluble species are still under characterization [
19,
20] but would represent key candidates to be neutralized for therapeutic benefit.
SAR228810 shows a high selectivity versus monomeric Aβ with no detectable binding to monomeric Aβ by surface plasma resonance and with approximately 100-fold selectivity in ELISA assays, a likely underestimation since monomeric Aβ preparation is subject to early aggregation on adhering to plastic wells in the ELISA format. In comparison, gantenerumab and BAN2401 retain some limited affinity for monomeric Aβ. SAR228810 selectivity versus monomeric Aβ is further confirmed in vivo where it does not increase plasma Aβ levels unlike monomeric Aβ-binding mAs such as bapineuzumab (used here as a positive control), solanezumab, or crenezumab [
4] that bind to plasma Aβ and prevents its degradation/clearance in liver [
41]. In immunohistology, SAR228810 (up to 5 μg/ml, 35 nM) did not display binding to diffuse Aβ deposits common in elderly cases and which are not related to AD pathology, further documenting that SAR228810 does not bind to Aβ assemblies devoid of β-sheet conformation. In comparison, gantenerumab, in line with its low affinity for monomeric Aβ, has been reported to bind to both diffuse Aβ deposits and compact plaques at 1 μg/ml (7 nM) [
8], in addition to cross-reactivity on neurons [
44], documenting significant differences in binding profile with SAR228810.
Of interest, SAR228810 neutralizes the neurotoxicity of synthetic oAβ42 in cultures and prevents development of synaptic dysfunction in vivo in amyloid transgenic mice. While the three protofibrillar Aβ mAbs neutralized oAβ42 neurite toxicity, only SAR228810 prevented neuronal apoptosis, further suggesting differences in the binding profiles of the three mAbs across the wide spectrum of Aβ oligomer/protofibril conformers. However, it will be important in the future to compare the binding profiles of those mAbs, including aducanumab, with the different multimeric Aβ assemblies and in particular their capacity to neutralize AD brain-derived synaptotoxic oAβ.
Brain bioavailability
As previously reported for IgG [
45], the murine version of SAR228810 had detectable but low brain penetration, representing approximately 0.03–0.1% of plasma levels. In transgenic mice, 7 days following dosing at 10 mg/kg, SAR255952 concentrations in CSF were in the range of 150–300 ng/ml. In live imaging in mice bearing amyloid plaques, we could document a slow accumulation over days of labeled SAR255952 to the periphery of cerebral amyloid plaques while parenchymal levels decreased and antibody was cleared from plasma. Finally, after chronic treatment in transgenic mice, a strong amyloid plaque-associated IgG1 immunoreactivity could be detected, indicating SAR255952 localization to cerebral plaques.
Due to the measurable but low brain penetration, the antibody selectivity among the different Aβ pools/conformations (which have extremely different ranges of concentrations in the brain as mentioned in the Background section) has a direct impact on its bioavailability in vivo to neutralize/clear each of the cerebral Aβ pathological assembly types. For instance, at the clinical dose of 2 mg/kg, CSF bapineuzumab concentration was 45 ng/ml (0.25 nM) [
46], well below the concentration of monomeric Aβ in human CSF (approximately 10 ng/ml, 2.5 nM), and, therefore, most bapineuzumab molecules would be Aβ-bound when reaching the brain parenchyma with limited brain bioavailability to neutralize soluble oligomeric and protofibrillar forms. We would like to emphasize the issue of low therapeutic IgG bioavailability for central nervous system (CNS) indications. It is useful to recall that, for the rare disease amyloid IgG light chain (LC) amyloidosis affecting peripheral organs, the aggregated LC antibody NEOD0001 has recently shown some early signs of efficacy at a monthly dose of 24 mg/kg [
47], while IgG penetration in target organs (the kidney and heart) is known to be in the order of 0.1 to 0.5 compared with plasma levels, at least ten-fold higher than for the brain [
45]. The recent positive clinical results for aducanumab decreasing brain amyloid in patients were dose-dependent and strongly significant at the highest doses of 6 and 10 mg/kg [
7], and similarly for BAN2401 active only at the highest dose (10 mg/kg) tested ([
10]), although both had ARIA occurrence. Indeed, the Aβ immunotherapy field is now recognizing this major limitation with a recent trend to largely increase doses when tolerated [
48].
The issue of therapeutic antibody bioavailability is further exacerbated by dose limitations due to ARIA adverse effects observed in clinical studies for antibodies binding to vascular amyloid and with effector functions (bapineuzumab from the dose of 2 mg/kg, gantenerumab and aducanumab in clinical studies [
26,
49]). To limit the risk of ARIA, SAR228810 was engineered with a double mutant human IgG4 Fc domain to endow drastically reduced effector functions which did not compromise its activity at preventing amyloid plaque development in vivo. Using murine versions, SAR255952 (aglycosylated mIgG1 with very low effector functions) was even moderately more potent than a 3D6 mIgG2a control (murine bapineuzumab) and only minimally less potent than the complete mIgG1 version of SAR255952 (data not shown), the latter two possessing significant high effector functions. Along with amyloid plaque lowering, microglial and astrocyte inflammation were decreased as well as plaque-associated dystrophic neurites, leading to improved synaptic function in the hippocampus. In the animal model used, the impact on cognitive deficit could not be assessed, however. Of note, the human antibody SAR228810 was also shown in vivo to have similar activity as murine SAR255952 using immunotolerized mice. As expected, even at high doses, SAR255952 did not induce brain microhemorrhages and vasculitis while murine bapineuzumab did, similar to its humanized form inducing ARIA in patients.
Unlike in previously reported in-vitro phagocytosis studies, multiple lines of evidence support that effector functions of anti-Aβ mAb are not necessary for activity on amyloid plaques in vivo (discussed in [
28]). Using in-vivo multiphoton microscopy, Bacskai and colleagues [
50] have demonstrated that 3D6 F(ab’)
2 fragments (that lack the Fc region of the antibody and therefore effector functions) led to clearance of nearly half of amyloid deposits in APP mice within 3 days, similar to the results obtained with full-length 3D6. These data could be extended to the chronic setting where intraperitoneal chronic treatment with the F(ab’)2 fragment of an Aβ mAb significantly reduced cerebral amyloid plaques, similar to full IgG mAb [
51]. Similarly, brain expression of anti-Aβ single chain antibody variable domain (lacking a Fc domain) was also efficacious in amyloid transgenic models [
52]. These findings are also consistent with early active immunization data where Aβ immunization was as effective at reducing amyloid deposition in APP mice with FcR-γ chain knockout (FcR-γ
−/
−) as in FcγR-sufficient APP mice [
53]. Additionally, the deglycosylated anti-Aβ 2H6 (therefore with low effector functions) had an activity comparable to fully glycosylated 2H6 with regard to clearance of brain amyloid plaque and reversing behavioral deficits in the transgenic model [
54]. The deglycosylated IgG, however, had the clear advantage of not inducing brain microhemorrhages, similar to that demonstrated in the present report with SAR228810/SAR255952. A similar strategy was adopted for crenezumab which was engineered with a human IgG4 isotype while retaining efficacy in animal studies [
55]. A low effector function variant of bapineuzumab (AAB003) developed ARIA-E at higher doses than its parent antibody, confirming that reduced effector function would lower ARIA risk but was discontinued for a lack of effect on biomarkers, suggesting that both epitope and level of effector function could be important [
56].
Conceptually, if the therapeutic intent is to neutralize the brain soluble form of synaptotoxic Aβ, there would be a limited requirement for IgG effector function that could even be detrimental in the AD brain which presents a significant inflammatory-primed cerebral environment. Indeed, in aging and in chronic neurodegenerative diseases, modest challenges can lead to more profound CNS inflammation and cognitive deficits than in healthy young individuals, a phenomenon named microglia priming [
57]. In acute studies, a full IgG anti-Aβ mAb induced a significant increase in brain immune cells while the corresponding F(ab’)2 fragment did not, while maintaining efficacy on amyloid plaques [
51]. Similarly for tau mAbs, it was demonstrated that effector-less IgG maintained efficacy without leading to the neurotoxic activation of microglia triggered by effector-competent IgG [
58]. Overall, these results suggest that targeting amyloid in vivo might not require Fc-dependent effector functions.
Regarding the mechanism of action of SAR228810, its binding to soluble protofibrillar Aβ assemblies could prevent further aggregation and synaptotoxicity. In addition, by binding/coating existing plaques, SAR228810 would have the potential to block/prevent secondary Aβ nucleation generated at the plaque surface [
59] that could be related to the periplaque protofibrillar Aβ42 “hot spots” associated with axonal dystrophies [
60]. It could also prevent leakage from plaques of the oAβ assemblies associated with dementia that have recently been identified in the human AD brain [
14]. In this regard, blocking Aβ monomer production, such as with a BACE inhibitor, might be ineffective at preventing the release of toxic species from plaques.
Translating to human AD clinical studies where patients have a high cerebral amyloid deposition many years prior to the onset of symptoms, we are aware that the efficacy of SAR228810 was demonstrated only as a prevention of plaque accumulation and synaptic dysfunction but not for reduction of preexisting amyloid plaque load in the very aggressive APP transgenic line used. There are actually very limited data in the literature demonstrating that in amyloid transgenic mice peripheral administration of an anti-Aβ mAb can decrease preexisting amyloid burden. Even for aducanumab, that has demonstrated a clear dose- and time-dependent decrease of amyloid positron emission tomography (PET) burden in patients, its murine chimeric version could not decrease preestablished amyloid plaque load after systemic administration [
42], while being able to prevent plaque accumulation when administered preventatively [
7]. The transgenic strains have been designed with very high Aβ production to develop plaques in a few months versus many decades in humans, and only in conditions where Aβ production was stopped or blocked concomitantly could immunotherapy decrease brain amyloid [
61,
62]. Of interest in the therapeutic conditions mentioned above, aducanumab improved neuronal calcium levels [
42] that could be linked to the cognitive effect in patients, possibly separate from amyloid clearance itself. In the future, it will be important to compare the different high molecular weight Aβ assembly preferring antibodies in more detail and in particular for neutralizing the soluble AD brain synaptotoxic species.