In the present study, our main findings showed that the Aβ42/Aβ40 ratio in both CSF and serum were reduced during early cerebral amyloidosis in AppNL−F/NL−F knock-in mice. Significant changes in the CSF Aβ42/Aβ40 ratio occurred when cerebral Aβ plaque burden started to become more widespread in cortical and hippocampal regions, while the corresponding ratio in serum was altered at a somewhat later time point. Furthermore, the initial decline of the CSF Aβ42/Aβ40 ratio coincided with increased concentrations of soluble and insoluble Aβ42 in cortical brain tissue. In both fluid compartments, the reduction in the Aβ42/Aβ40 ratio quickly started to stabilize towards a plateau although both insoluble and soluble forms of Aβ steadily continued to increase as the mice aged. Accordingly, we found inverse hyperbolic associations between cerebral Aβ and the Aβ42/Aβ40 ratio in both CSF and serum. These associations tended to be greater for the measures in CSF compared with serum. In general, similar results were obtained for CSF and serum Aβ42, but not Aβ40, when compared to those obtained for the Aβ42/Aβ40 ratio.
The Aβ42/Aβ40 ratio in CSF was significantly reduced earlier than in blood in AppNL-F/NL-F knock-in mice
The Aβ42/Aβ40 ratio in human CSF declines at least a decade before cognitive symptoms due to both sporadic and familial AD develop [
7,
8], and recent studies using highly sensitive biochemical assays suggest that the corresponding measure in plasma also is reduced during preclinical sporadic AD [
15,
18,
20]. Our findings in the
AppNL−F/NL−F knock-in mice are in good agreement with these studies, demonstrating an age-dependent decline in CSF and serum Aβ42/Aβ40 ratios, changes that occur prior to the time at which Aβ-dependent spatial memory deficits develop in these mice [
28]. The Aβ42/Aβ40 ratio in CSF was steadily reduced from 12 months of age, while the corresponding ratio in serum significantly declined somewhat later, at 16 months of age. The decline in serum Aβ42/Aβ40 ratio was partially due to elevated concentrations of Aβ40 in the oldest age groups. This is in contrast to previous studies in sporadic AD, in which the concentrations of this Aβ peptide have been reported to remain unaltered during the preclinical stage of the disease [
15,
18]. Nevertheless, considering that the increase in serum Aβ40 coincided with the decrease in serum Aβ42 and that we found no age-dependent effect on serum Aβ40 in control
AppNL/NL knock-in mice, it is possible that the concentrations of Aβ40 are elevated in response to AD-related pathological processes in
AppNL−F/NL−F knock-in mice. Future studies should address the potential impact of the Beyreuther/Iberian mutation and the relevance of this finding for sporadic and familial AD.
In
AppNL−F/NL−F knock-in mice, the Aβ42/Aβ40 ratio in serum was positively correlated with that in CSF when assessed over all age groups, although the association was relatively modest. These results are well in line with previous findings in which the Simoa platform was used to measure plasma Aβ42 and Aβ40 in a clinical cohort consisting of cognitively healthy individuals as well as patients with mild cognitive impairment (MCI) and AD dementia [
15]. The modest association may to some extent be explained by physiological confounding factors that influence the measurement of Aβ in blood compared with CSF, such as degradation in the liver or by circulating enzymes, matrix effects, and renal clearance [
33]. Moreover, it is likely that the production of these peptides in peripheral organs significantly contributes to the circulating pool, as it has been estimated that a maximum of 30–50% of Aβ present in blood derives from the central nervous system [
34]. The choice of analytical platform may also play a role, as the use of methods based on IP-MS has recently been reported to generate higher associations between the Aβ42/Aβ40 ratio in the two fluid compartments compared to different Simoa immunoassays [
35].
CSF and serum Aβ42/Aβ40 ratios in relation to cerebral amyloidosis
In agreement with previous findings [
28], a few isolated extracellular Aβ plaques first appeared in cortical areas of the brain at 6 months of age in
AppNL−F/NL−F knock-in mice. At 9 months, the cerebral Aβ plaque burden remained sparse and did not associate with changes in the Aβ42/Aβ40 ratio in either CSF or serum. Instead, the decline of this biomarker in the two fluid compartments at 12 and 16 months, respectively, occurred in relation to a more pronounced Aβ plaque load in both cortex and hippocampus. In a recent study in autopsy-confirmed AD cases, the authors reported that a decline in the CSF Aβ42/Aβ40 ratio was initiated in Thal phase 2 [
36], which is characterized by the presence of Aβ deposits in neocortex and allocortical regions [
37]. No changes in CSF Aβ42/Aβ40 ratio were found in cases in which the pathology was restricted to the neocortex,
i.e., in Thal phase 1 [
36]. The study also showed that a reduced CSF Aβ42/Aβ40 ratio was associated with a moderate Aβ plaque burden, as estimated in accordance with CERAD (Consortium to Establish a Registry for Alzheimer’s Disease) recommendations, a finding that is similar to what has been observed for the Aβ42/Aβ40 ratio in plasma [
38]. These results are in line with those from the present study, suggesting a temporal sequence of events in which initial deposition of Aβ aggregates in restricted brain regions is followed by a decline in CSF and serum Aβ42/Aβ40 ratios once the Aβ pathology is somewhat more widespread but still relatively low to moderate. In addition, studies have suggested that CSF Aβ42—alone or in ratio with Aβ40 [
12,
39‐
42]—as well as the Aβ42/Aβ40 ratio in plasma [
17] are significantly changed before the threshold for abnormal fibrillar dense-core plaque burden in the brain, as measured by amyloid PET, is reached. It is possible that the sensitivity of amyloid PET to detect fibrillar Aβ species in the brain in early preclinical AD is limited, as the results from the present study suggest a sparse burden of thioflavin S-positive fibrillar dense-core plaques in the brain prior to changes in the investigated fluid biomarkers.
Insoluble forms of Aβ found predominantly in plaques can be measured biochemically in FA extract from brain tissue homogenates. As expected, cortical FA-soluble Aβ42 increased in an age-dependent manner with significant changes from 9 months of age, which is the same time from which we also observed a sparse burden of cortical Aβ plaques in
AppNL−F/NL−F knock-in mice. In addition, cortical TBS-soluble Aβ42 increased from 12 months of age and thereby coincided with the initial decline in the CSF Aβ42/Aβ40 ratio. Although both insoluble and soluble forms of Aβ steadily increased over the studied time period, the reduced Aβ42/Aβ40 ratio in both CSF and serum quickly started to stabilize toward a plateau. Indeed, we found an inverse hyperbolic association between cerebral amyloidosis and the Aβ42/Aβ40 ratio in both CSF and serum, which is consistent with multiple cross-sectional studies in humans investigating the association between Aβ42—alone or in ratio with Aβ40—in the two fluid compartments and amyloid PET [
9,
12,
17,
43‐
45]. Our results are also in good agreement with longitudinal studies suggesting that once a decline in CSF Aβ42 has occurred in early preclinical AD, the concentration remains fairly stable as the disease progresses [
8,
46,
47]. Together, these findings may to some extent challenge the proposed hypothesis that the reduction in the investigated fluid biomarkers is due to the deposition of Aβ into extracellular plaques [
48], as cerebral plaque load is only linearly associated with the Aβ42/Aβ40 ratio in CSF and blood during a very limited time-frame in the early disease stage. However, our findings that no changes in CSF or serum Aβ were found in
AppNL/NL mice with age confirm that biological processes related to Aβ pathology are required for these biomarkers to decline. Future studies should further address the underlying cause of these changes in the preclinical stage of AD.
The inverse correlation with cerebral amyloidosis tended to be greater for the Aβ42/Aβ40 ratio in CSF when compared with serum. These results imply that the Aβ42/Aβ40 ratio in CSF more reliably may reflect Aβ pathology in the brain than the corresponding ratio in blood in preclinical AD. In agreement with these findings, a study conducted by Schindler
et al. reported that the Aβ42/Aβ40 ratio in CSF was a better predictor of and showed a greater correlation with amyloid PET than the Aβ42/Aβ40 ratio in plasma in a cohort consisting of mainly cognitively healthy individuals [
17]. Furthermore, although the ratio between plasma Aβ42 and Aβ40 has shown higher correspondence with cerebral amyloidosis than Aβ42 alone when studied in clinical cohorts [
16], the correlations between cerebral amyloidosis and these two measures in serum were similar in
AppNL−F/NL−F knock-in mice. As the concentrations of Aβ42 and Aβ40 in blood may be affected by comorbidities and other confounding factors [
15,
49], the limited biological variation in the
AppNL−F/NL−F knock-in mice compared to a human study population may to some extent explain these results.
A few studies have previously investigated Aβ changes in CSF [
24,
25] and blood [
26] in relation to cerebral amyloidosis over time using transgenic mouse models that overexpress mutant human
APP under the control of certain promotors. In line with our own findings in
AppNL−F/NL−F knock-in mice, these studies have reported a decline in Aβ42—alone or in ratio with CSF Aβ40—in these fluid compartments that is initiated shortly after the onset of Aβ plaque deposition and inversely associates with the burden of Aβ in the brain. In one of the studies, increased concentrations of both CSF Aβ42 and Aβ40 prior to plaque deposition were observed, suggesting that the biphasic profile of this biomarker change potentially could be used for early identification of cognitively healthy individuals who are at risk of developing AD dementia [
25]. Although we have observed similar findings in the well-characterized
APP-overexpressing 3xTg mouse model (Additional file
1: Fig. S6 and Supplementary methods), this initial increase was not found in
AppNL−F/NL−F knock-in mice. The concomitant increase in CSF Aβ42 and Aβ40 in
APP-overexpressing mice suggests an elevated production or cleavage of APP and one may speculate that this early change to some extent is a result of an age-related overexpression of
APP in these models. However, further studies are needed to elucidate these differences and their translational implication.