Main

The APOE4 variant of apolipoprotein E is the strongest genetic risk factor for AD1. One and two APOE4 alleles increase risk for AD by approximately 4- and 15-fold, respectively, compared to the more-common APOE3 gene that carries lower risk for AD1. Besides accelerating onset and progression of dementia, APOE4 is associated with different brain pathologies. For example, APOE4 accelerates BBB breakdown and degeneration of brain capillary pericytes2,3 that maintain BBB integrity4,5,6 and leads to cerebral blood flow (CBF) reduction7,8 and dysregulation7,9,10. APOE4 is toxic to neurons11 and accelerates tau-mediated neurodegeneration12. Additionally, APOE4 slows down amyloid-β (Aβ) clearance13,14 and accelerates amyloid deposition14,15,16, which promotes development of amyloid pathology.

Recent studies focused on very early stages in the Alzheimer’s continuum in individuals who are cognitively unimpaired or with mild cognitive impairment (MCI) have shown that individuals bearing an APOE4 variant (APOE3/APOE4 or APOE4/APOE4) are distinguished from APOE3 homozygotes by breakdown in the BBB in the hippocampus and medial temporal lobe, regions responsible for memory encoding and cognitive functions17. This finding is apparent in cognitively unimpaired APOE4 carriers and more severe in those with MCI and is independent of Aβ or tau pathology measured in the cerebrospinal fluid or in brain by positron emission tomography17. These findings support the growing evidence suggesting that vascular dysfunction, BBB breakdown and vascular disorder contribute to early cognitive impairment and AD17,18,19,20,21,22,23,24,25,26. On the other hand, accumulation of Aβ in the brain has also been suggested to occur years before cognitive impairment and continues to increase with disease progression27. Although it has been shown that vascular dysfunction contributes to early cognitive impairment in ways that may not be exclusively related to classical AD pathology17,19,20,26, the respective contributions of the BBB pathway and vascular disorder versus amyloid-β pathway to advanced disease stage during progression of neurodegenerative disorder and cognitive decline in AD are still poorly understood.

To address this question, here we studied vascular dysfunction, Aβ pathology, neuronal dysfunction and behavior in older APOE3 and APOE4 knock-in mice28 alone and crossed with the 5xFAD line29. All mice were derived from the same litters, as previously described30. Mice lacking Apoe3 and/or expressing human APOE4 develop early BBB breakdown3,31,32,33 and CBF dysregulation10. On the other hand, the 5xFAD line also develops BBB breakdown34,35,36,37, CBF reductions38 and neuron and synaptic loss at a later stage29,39, whereas APOE3;5xFAD and APOE4;5xFAD mice (also known as E3FAD and E4FAD lines, respectively30) have comparable Aβ pathology at an older age40. These features of the studied models allowed us to interrogate how different pathologies in APOE4 compared to APOE3 mice relate to each other and how they influence neuronal function and behavior.

Results

Blood–brain barrier breakdown in APOE4 and APOE4;5xFAD mice

First, we studied BBB integrity in the cortex and hippocampus in 18–24-month-old APOE3 and APOE4 knock-in controls and APOE3;5xFAD and APOE4;5xFAD mice derived from the same litters, as previously described30. Using a dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) protocol as we previously developed6,41, we found that APOE4 mice compared to the respective APOE3 lines develop accelerated BBB breakdown in the cortex and hippocampus both without and with Aβ pathology (Fig. 1a–c), as shown by 43% and 52% and 29% and 34% increases in the BBB permeability Ktrans values in APOE4 versus APOE3 mice alone and in these mice crossed with 5xFAD mice, respectively (Fig. 1b,c). We also found substantial 35% and 36% and 22% and 20% increases in regional BBB Ktrans values in the cortex and hippocampus of APOE3;5xFAD and APOE4;5xFAD mice compared to their respective APOE3 and APOE4 littermate controls (Fig. 1c), consistent with previously shown BBB breakdown in 5xFAD mice34,35,36,37.

Fig. 1: Blood–brain barrier breakdown in old APOE4 and APOE4;5xFAD mice.
figure 1

a, T2-weighted scans displaying regions of interest: primary somatosensory cortex (Ctx) and hippocampus (Hipp). b,c, Representative Ktrans maps (b) and BBB permeability Ktrans values (c) in the Ctx and Hipp in APOE3 (E3, n = 12; blue empty circles), APOE4 (E4, n = 14; blue-filled circles), APOE3;5xFAD (E3+FAD, n = 13; red empty circles) and APOE4;5xFAD (E4+FAD, n = 16; red-filled circles) mice generated from DCE-MRI scans. d,e, Fibrinogen+ perivascular capillary deposits (red) in the Ctx (d) and quantification of fibrinogen+ deposits in the Ctx and Hipp (e) in E3, E4, E3+FAD and E4+FAD mice. Blue, lectin+ endothelial profiles; n = 5 mice per group; scale bar, 20 μm. f,g, CD13+ pericyte coverage (magenta) of lectin+ endothelial capillary profiles (blue) in the Ctx (f) and quantification of pericyte coverage on capillaries in the Ctx and Hipp (g) in E3 (n = 6), E4 (n = 6), E3+FAD (n = 6) and E4+FAD (n = 5) mice; scale bar, 20 μm. h,i, Negative correlation between fibrinogen deposits and pericyte coverage in the Ctx (h) and Hipp (i) using individual mouse data points from the studied groups. Mice (both sexes) were 18–24 months old. j–m, Representative immunostaining for ZO-1 (red) and lectin+ endothelium (blue) in the Ctx (l, red) and ZO-1 length in the Ctx and Hipp (k) and immunostaining for occludin (green) and lectin+ endothelium (blue) in the Ctx (l, green) and occludin length in the Ctx and Hipp (m) in E3 (blue empty circles), E4 (blue-filled circles), E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice; n = 5 mice/group; scale bar, 20 μm. Boxes in j and l denote sites taken for higher magnification insets shown below merged images. Mice (both sexes) were 18–24 months old. In c,e,g,k and m data are presented as truncated violin plots; continuous line, median; and dotted line, interquartile range. Significance by one-way analysis of variance (ANOVA) followed by the Tukey post hoc test. Significance by two-tailed simple linear regression (h,i); r, Pearson correlation coefficient.

Source data

Full size image

These results were confirmed by increased levels of perivascular brain capillary cortical and hippocampal deposits of blood-derived fibrinogen in APOE4 compared to APOE3 mice both in the presence and absence of Aβ pathology (Fig. 1d,e). Specifically, we found 111% and 122% and 135% and 137% increase in fibrinogen deposits in the cortex and hippocampus in APOE4 versus APOE3 mice alone and in these mice crossed with 5xFAD mice, respectively (Fig. 1e). The effects of Aβ pathology was evident in both genotypes when compared to knock-in littermate controls (Fig. 1d,e), consistent with previously reported BBB breakdown in 5xFAD mice.

In contrast to a previous study showing that younger 8-month-old APOE4;5xFAD female mice had higher BBB leakiness compared to male mice42, the present study did not find the effect of gender on the regional BBB Ktrans values in either APOE4 versus APOE3 mice alone and/or in these mice crossed with 5xFAD mice (Extended Data Fig. 1). Whether the difference between previous studies42 and the current study reflects a difference in age (18–24 versus 8 months old) and/or whether female mice develop less severe BBB impairment with age or alternatively, males develop more rapid BBB changes with age, remains an interesting topic for future studies.

Compared to APOE3 mice, APOE4 littermate controls showed loss of brain capillary pericyte coverage in the cortex and hippocampus (Fig. 1f,g). For example, APOE4 compared to APOE3 mice alone and crossed with 5xFAD mice had 24% and 25% and 19% and 27% loss of pericyte coverage in the cortex and hippocampus, respectively (Fig. 1g). These findings were consistent with previously shown loss of pericyte coverage in APOE4 mice3. The effect of Aβ pathology on loss of pericyte coverage was seen in both genotypes, again consistent with findings showing loss of pericytes in AD mouse lines43,44,45. However, the effect was much more pronounced in APOE4 than in APOE3 mice. The regional accumulation of perivascular brain capillary fibrinogen deposits negatively correlated with loss of pericyte coverage (Fig. 1h,i), similar to that reported in pericyte-deficient mouse lines5,41.

The analysis of the BBB tight junction proteins revealed loss in the zonula occludens-1 (ZO-1) and occludin length coverage in APOE4 compared to APOE3 mice both in the presence and absence of Aβ pathology (Fig. 1l,m). We found substantial 34% and 57% and 31% and 42% loss in ZO-1 length and 45% and 42% and 31% and 25% loss in occludin length in the cortex and hippocampus of APOE3 and APOE4 littermate controls alone and in these mice crossed with 5xFAD mice, respectively (Fig. 1k,m). These findings were consistent with previously shown losses of the BBB tight junction proteins in APOE4 (ref. 3) and 5xFAD34 mice.

Cerebral blood flow reductions in APOE4 and APOE4;5xFAD mice

Next, we studied CBF changes in the cortex and hippocampus in 18–24-month-old APOE3 and APOE4 knock-in littermate controls and APOE3;5xFAD and APOE4;5xFAD mice, using a dynamic susceptibility-contrast (DSC)-MRI protocol as we previously developed6,41. We found that APOE4 compared to APOE3 mice develop a moderate, but substantial CBF loss in the cortex and hippocampus both in the presence and absence of Aβ pathology (Fig. 2a,b). This has been shown by 16% and 13% and 24% and 24% reductions in the regional cortical and hippocampal CBF in APOE4 vs. APOE3 mice alone and in these mice crossed with 5xFAD mice, respectively (Fig. 2a,b). Additionally, we found substantial 23% and 23% and 31% and 33% reductions in the cortical and hippocampal CBF in 5xFAD mice crossed to APOE3 and APOE4 mice compared to their respective knock-in littermate controls (Fig. 2b). As with BBB breakdown, sex did not have an effect on regional CBF changes in old mice (Extended Data Fig. 2).

Fig. 2: Cerebral blood flow reductions in old APOE4 and APOE4;5xFAD mice.
figure 2

a,b, Representative CBF maps (a) and regional CBF values (b) in the cortex (Ctx) and hippocampus (Hipp) in E3 (n = 12; blue empty circles), E4 (n = 14; blue-filled circles), E3+FAD (n = 13; red empty circles) and E4+FAD (n = 16; red-filled circles) mice generated from dynamic susceptibility-contrast magnetic resonance scans. c,d, Representative lectin+ endothelial profiles (blue) in the Ctx (c) and microvascular length in the Ctx and Hipp (d) in E3, E4, E3+FAD and E4+FAD mice; n = 5 mice/group; scale bar, 20 μm. e,f, Positive correlation between the CBF and microvascular length in the Ctx (e) and Hipp (f) using individual mouse data points from the studied groups. Mice (both sexes) were 18–24 months old. In b and d, data are presented as truncated violin plots; continuous line, median; and dotted line, interquartile range. Significance by one-way ANOVA followed by the Tukey post hoc test. In e and f, significance was determined by two-tailed simple linear regression.

Source data

Full size image

Compared to APOE3 mice, APOE4 mice had also a moderate but substantial loss of capillary density in the cortex and hippocampus both in the presence and absence of Aβ pathology (Fig. 2c,d), as shown by 17% and 11% and 10% and 9% reductions in the cortical and hippocampal capillary length in APOE4 versus APOE3 littermate control mice and APOE4;5xFAD versus APOE3;5xFAD mice, respectively (Fig. 2d). CBF reductions positively correlated with reductions in the capillary length (Fig. 2e,f). The observed correlations were consistent with previously reported correlations between reductions in the CBF and microvascular density in pericyte-deficient5,6,41 and APOE4 (refs. 3,10) mice.

Cyclophilin A-matrix metalloproteinase-9 pathway

In a search of a mechanism underlying vascular changes in APOE4 compared to the respective APOE3 lines, we focused on the BBB-degrading pro-inflammatory cyclophilin A (CypA)-matrix metalloproteinase (MMP)-9 pathway. When activated by brain capillary pericytes in APOE4 but not APOE3 transgenic mice, this pathway leads to MMP9-mediated breakdown of the BBB via degradation of BBB tight junction proteins, which in turn leads to leakages of blood-derived toxic proteins causing neuronal stress and dysfunction3. Previous work using in vivo multiphoton microscopy of DQ-gelatin has shown increased gelatinase MMP cerebrovascular activity in APOE4 compared to APOE3 mice, which has been confirmed by gelatin zymography of brain tissue, demonstrating an increase in pro-MMP9 and activated MMP9, but not MMP2 (ref. 3).This pathway is also activated in human APOE4 carriers diagnosed with AD, as suggested by neuropathological analysis2 and in living human APOE4 carriers, as shown by increased cerebrospinal fluid levels of CypA, MMP9 and sPDGFR-β, a marker of pericyte injury17.

Immunocytochemical analysis indicated that APOE4 mice compared to respective APOE3 lines have higher levels of CypA+ immunofluorescent signal in CD13+ pericytes in the cortex and hippocampus both in the presence and absence of Aβ pathology (Fig. 3a,b). Specifically, we found 614% and 627% and 45% and 37% increases in CypA+ signal in the cortical and hippocampal pericytes in APOE4 compared to APOE3 control mice alone and in these mice crossed with 5xFAD mice, respectively (Fig. 3b). Increases in the CypA+ immunofluorescent signal in CD13+ pericytes in the cortex and hippocampus in APOE4 compared to APOE3 mice without 5xFAD transgenes (littermate controls) and with 5xFAD transgenes both negatively correlated with the reduced length of the tight junction proteins ZO-1 (Fig. 3c,d) and occludin (Fig. 3e,f), but the effect was more pronounced in mice expressing 5xFAD transgenes, as expected based on the signal intensity data.

Fig. 3: Increased levels of cyclophilin A and matrix metalloproteinase-9 in pericytes in old APOE4 and APOE4;5xFAD mice correlate with the loss of tight junction proteins.
figure 3

a, Representative high-magnification confocal images of CD13+ pericytes (green), cyclophilin A (CypA; red) and lectin+ endothelium (blue) in the cortical capillaries in situ in E3, E4, E3+FAD and E4+FAD mice. Orthogonal views show colocalization of CypA with CD13+ pericytes; scale bar, 10 μm. b, Quantification of CypA+ pericytes in the cortex (Ctx) and hippocampus (Hipp) in E3 (blue empty circles), E4 (blue-filled circles), E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice; n = 5 mice/group. c–f, Negative correlation between CypA+ pericytes and ZO-1 length (c,d), and CypA+ pericytes and occludin length (e,f) in the Ctx and Hipp. g, Representative high-magnification confocal images of CD13+ pericytes (green), MMP9 (red) and lectin+ endothelium (blue) in the cortical capillaries in situ in E3, E4, E3+FAD and E4+FAD mice. Orthogonal views show colocalization of MMP9 with CD13+ pericytes; scale bar, 10 μm. h, Quantification of MMP9+ pericytes in the Ctx and Hipp in E3 (blue empty circles), E4 (blue-filled circles), E3+FAD (red empty circles) and E4+FAD (red empty circles) mice; n = 5 mice/group. i–l, Negative correlation between MMP9+ pericytes and ZO-1 length (i,j) and MMP9+ pericytes and occludin length (k,l) in the Ctx and Hipp. m,n, Positive correlation between CypA+ pericytes and MMP9 pericytes in the Ctx (m) and Hipp (n). Mice (both sexes) were 18–24 months old. Immunostaining for CypA or MMP9 and CD13, ZO-1 and lectin-endothelial staining and/or occludin and lectin-endothelial staining was performed on adjacent sections. In b and h data are presented as truncated violin plots; continuous line, median; and dotted line, interquartile range. Significance by one-way ANOVA followed by the Tukey post hoc test. In c–f and i–n, significance was obtained by by two-tailed simple linear regression. In c–f and i–n, the number was based on individual mouse values (n = 10) for E4 versus E3 mice in the absence of FAD transgenes (blue-filled and empty circles, respectively) and presence of FAD transgenes (red-filled and empty circles, respectively).

Source data

Full size image

We also found increased MMP9+ immunofluorescence signal in CD13+-pericytes (Fig. 3g,h) by 406% and 412% and 47% and 61% in the cortex and hippocampus in APOE4 compared to APOE3 control mice alone and in these mice crossed with 5xFAD mice, respectively (Fig. 3h). Increases in the MMP9+ immunofluorescent signal in CD13+ pericytes in the cortex and hippocampus also negatively correlated with the reduced length of ZO-1 (Fig. 3i,j) and occludin (Fig. 3k,l), as determined by immunostaining on adjacent sections. As expected, there was a positive correlation between the CypA+ and MMP9+ immunofluorescent signals in CD13+ pericytes, both in the cortex and hippocampus, as shown by the correlation analysis using individual data points from both APOE4 compared to APOE3 mice without (littermate controls) and with 5xFAD transgenes (Fig. 5m,n).

Altogether, these data demonstrate activation of the CypA-MMP9 pathway in pericytes in old APOE4 and APOE4;5xFAD mice compared to the respective APOE3 lines suggesting a possible mechanism for the APOE4-mediated BBB breakdown and loss of capillary density.

Aβ pathology and vascular dysfunction

As previously reported40, we also found that levels of Aβ42, the principal Aβ peptide in 5xFAD mice29,37, in the cortex and hippocampus were not substantially different between 18–24-month-old APOE4;5xFAD and APOE3;5xFAD mice (Fig. 4a). The tissue analysis confirmed comparable Aβ cortical and hippocampal load in these mice (Fig. 4b,c). As reported40, Aβ42 levels in female APOE3;5xFAD and APOE4;5xFAD mice were substantially higher compared to the respective male mice (Fig. 4d). Consistent with previous reports in 5xFAD lines29,37, Aβ40 levels in both APOE4;5xFAD and APOE3;5xFAD mice were substantially lower than Aβ42 levels (Extended Data Fig. 3a). As with Aβ42, we also found substantially higher Aβ40 levels in female compared to male mice in both APOE4 and APOE3 mice crossed with 5xFAD mice (Extended Data Fig. 3b).

Fig. 4: Aβ pathology in old APOE3;5xFAD and APOE4;5xFAD mice and Aβ-independent vascular changes.
figure 4

a, Aβ42 levels in the cortex (Ctx) and hippocampus (Hipp) in E3+FAD (n = 14; red empty circles) and E4+FAD (n = 17; red-filled circles) mice. b,c, Aβ immunostaining (b) and Aβ load in the Ctx or Hipp (c) in E3+FAD (n = 5) and E4+FAD (n = 6) mice. Scale bar, 500 μm. d, Aβ42 levels in the Ctx and Hipp in male and female E3+FAD (n = 14, 9 ♂ and 5 ♀; red empty circles) and E4+FAD (n = 17, 12 ♂ and 5 ♀; red-filled circles) mice. e–h, Lack of correlation between the BBB permeability Ktrans values and Aβ42 levels in the Ctx and Hipp (e,f) and between regional CBF values and Aβ42 levels in the Ctx and Hipp (g,h). Mice (both sexes) were 18–24 months old. In d–g, there were n = 29 (13 E3+FAD and 16 E4+FAD) individual mouse data points from both groups. In a and c, data are presented as violin plots; continuous line, median; and dotted line, interquartile range. Significance was obtained by unpaired two-tailed Student t-tests. In d, significance was obtained by one-way ANOVA followed by Tukey post hoc test. In e–h, significance was obtained by two-tailed simple linear regression. NS, not significant.

Source data

Full size image

Interestingly, the increased BBB permeability Ktrans values and a moderate CBF loss in APOE4 compared to APOE3 mice crossed with 5xFAD mice did not correlate with elevated Aβ42 cortical or hippocampal levels (Fig. 4e–h) or Aβ40 levels (Extended Data Fig. 3e,f), suggesting that Aβ pathology is likely not causing vascular dysfunction in old APOE4;5xFAD when compared to APOE3;5xFAD mice.

APOE4-mediated neurodegeneration

Next, we found that old APOE4 mice compared to APOE3 mice develop accelerated neuron and axon loss without and with Aβ pathology (Fig. 5a–f). This has been shown by 18% and 34% and 11% and 22% loss of neuronal counts and 20% and 27% and 27% and 43% decrease in the neuritic density in the cortex (Fig. 5b,c) and hippocampus (Fig. 5e,f) in APOE4 versus APOE3 control mice alone and in these mice crossed with 5xFAD mice, respectively. Consistent with previous reports showing early neuronal dysfunction in APOE4 transgenic lines3 and in 5xFAD mice showing neuron and synaptic loss29,39, we found substantial 17% and 23% and 10% and 10% reduction in neuronal counts and 30% and 37% and 36% and 51% reduction in neuritic density in the cortex and hippocampus of 5xFAD;APOE3 and 5xFAD;APOE4 mice compared to their respective APOE3 and APOE4 knock-in littermate controls (Fig. 5b,c,e,f). Notably, loss of neurons correlated with the increased BBB Ktrans values reflecting BBB breakdown (Fig. 5g,h), but did not correlate with elevated Aβ42 levels (Fig. 5i,j).

Fig. 5: Neurodegenerative changes in old APOE4 and APOE4;5xFAD mice correlate with BBB breakdown but not with Aβ pathology.
figure 5

a–f, NeuN+ neurons (red) and SMI312+ neurites (green) and quantification of NeuN+ neurons and SMI312+ neuritic density in the cortex (Ctx) (a–c) and hippocampus (Hipp) (d–f) in E3 (n = 5; blue empty circles), E4 (n = 5; blue-filled circles), E3+FAD (n = 6 (b); n = 5 (c); n = 5 (e); n = 5 (f); red empty circles) and E4+FAD (n = 5; red-filled circles) mice. Scale bars, 20 μm. g,h, Negative correlation between the number of NeuN+ neurons and BBB permeability Ktrans values in the Ctx (g) and Hipp (h) using individual mouse data points from the studied groups. i,j, Lack of correlation between the number of NeuN+ neurons and Aβ42 levels in the Ctx (i) and Hipp (j) in E3+FAD (n = 6 (i); n = 5 (j)) and E4+FAD (n = 5 (i); n = 5 (j)) individual mouse data points from both groups. Mice (both sexes) were 18–24 months old. In b,c,e and f, data are presented as violin plots; continuous line, median; and dotted line, interquartile range. Significance by one-way ANOVA followed by the Tukey post hoc test. In g–j, significance was obtained by two-tailed simple linear regression. NS, not significant.

Source data

Full size image

APOE4-related behavioral deficits

Finally, we found that APOE4 compared to APOE3 mice develop behavioral deficits on daily activity tests as shown by a substantial drop in the nest construction score both in the presence and absence of Aβ pathology (Fig. 6a). A drop in the nesting scores correlated with the increased BBB Ktrans values and diminished CBF (Fig. 6b–d), but did not correlate with elevated Aβ42 levels in the cortex or hippocampus (Fig. 6e,f). APOE4 compared to APOE3 mice also developed hippocampal-associated memory deficits as shown by their worse performance on the novel object location (NOL) test both in the presence and absence of Aβ pathology (Fig. 6g). Deficits in NOL exploratory preference correlated with the increased BBB permeability Ktrans values in the hippocampus (Fig. 6h) and diminished CBF in the cortex (Fig. 6i), but not with elevated Aβ42 levels in either of these regions (Fig. 6j,k).

Fig. 6: Behavioral deficits in old APOE4 and APOE4;5xFAD mice correlate with vascular dysfunction but not with Aβ pathology.
figure 6

a, Nesting scores in E3 (n = 17; blue empty circles), E4 (n = 25; blue-filled circles), E3+FAD (n = 18; red empty circles) and E4+FAD (n = 18; red-filled circles) mice. b,c, Negative correlation between nesting scores and BBB permeability Ktrans values in the cortex (Ctx) (b) and hippocampus (Hipp) (c). d, Positive correlation between nesting scores and CBF values in the Ctx. In b–d, n = 48 individual mouse data points from the studied groups. e,f, Lack of correlation between nesting scores and Aβ42 levels in the Ctx (e) and Hipp (f); n = 24 individual mouse data points from the studied groups. g, NOL in E3 (n = 15), E4 (n = 21), E3+FAD (n = 15) and E4+FAD (n = 17) mice. h, Negative correlation between NOL and Ktrans values in Hipp. i, Positive correlation between NOL and CBF values in Ctx. In h and i, n = 42–44 individual mouse data points from the studied groups. j,k, Lack of correlation between NOL values and Aβ42 levels in Ctx (j) and Hipp (k); n = 20 individual mouse data points from two groups. Mice (both sexes) were 18–24 months old. In a and g, data are presented as violin plots; continuous line, median; and dotted line, interquartile range. Significance by one-way ANOVA followed by the Tukey post hoc test. In b–f and h–k, significance was obtained by two-tailed simple linear regression. NS, not significant.

Source data

Full size image

Inhibition of the CypA-MMP9 pathway in APOE4;5FAD mice

Next, we treated 10–12-month-old APOE4;5xFAD and APOE3;5xFAD mice with a non-immunosuppressive CypA inhibitor Debio-025, which has been used in humans for non-neurological indications such as hepatitis C46. Additionally, Debio-025 is currently undergoing testing for SARS-CoV-2 infection47 and is in phase 2 clinical trial against COVID-19 (www.clinicaltrialsarena.com/news/debiopharm-begins-alisporivir-france).

Debio-025 was administered for 30 d (10 mg kg−1 d−1 intraperitoneally (i.p.)) using a dosing regimen as previously reported in mouse models of muscular dystrophy and myopathy48,49 and in the low-density lipoprotein receptor-related protein 1 (LRP1) endothelial-specific knockout mice50. We started Debio-025 at 10–12 months of age, when APOE4;5xFAD compared to APOE3;5xFAD mice develop accelerated BBB breakdown, behavioral deficits and substantial loss of neuronal networks, as reported42,51 (Fig. 7a) The goal of this experiment was to determine whether protecting BBB integrity by suppressing the CypA-MMP9 pathway in pericytes can prevent further neuron loss and behavioral changes in APOE4;5xFAD relative to APOE3;5xFAD mice.

Fig. 7: Debio-025 inhibits the CypA-MMP9 pathway in pericytes and improves BBB integrity in APOE4;5xFAD mice.
figure 7

a, A schematic of the time course of Debio-025 experiment. b, Confocal images of CD13+ pericytes (green), CypA (red) and lectin+ endothelium (blue) in the cortical capillaries in E4+FAD mice treated with vehicle or Debio-025. Orthogonal views, colocalization of CypA with CD13+ pericytes; scale bar, 10 μm. c, Quantification of CypA+ pericytes in the cortex (Ctx) and hippocampus (Hipp) in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group. d, Confocal images of CD13+ pericytes (green), MMP9 (red) and lectin+ endothelium (blue) in cortical capillaries in E4+FAD mice treated with vehicle or Debio-025. Orthogonal views, colocalization of MMP9 with CD13+ pericytes; scale bar, 10 μm. e, Quantification of MMP9+ pericytes in the Ctx and Hipp in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group. f,g Immunostaining for ZO-1 (red) and lectin+ endothelium (blue) in the Ctx in E4+FAD mice treated with vehicle or Debio-025 (f, red) and ZO-1 length in the Ctx and Hipp (g) in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group; scale bar, 20 μm. h–i, Immunostaining for occludin (green) and lectin+ endothelium (blue) in the Ctx (h, green) and occludin length in the Ctx and Hipp (i) in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group; scale bar, 20 μm. Boxes in f and h denote sites taken for higher magnification insets shown below merged images. j, BBB permeability Ktrans values in the Ctx and Hipp in E3+FAD vehicle-treated (n = 6; red empty circles), E3+FAD Debio-025-treated (n = 8; red empty circles), E4+FAD vehicle-treated (n = 8; red-filled circles) and E4+FAD Debio-025-treated (n = 8; red-filled circles) mice. k, Quantification of fibrinogen+ pericapillary deposits in the Ctx and Hipp in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group. l,m, CD13+ pericyte coverage (magenta) of lectin+ endothelial capillary profiles (blue) in the Ctx (l; scale bar, 20 μm) and quantification of pericyte capillary coverage in the Ctx and Hipp (m) in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group (scale bar, 20 μm). Mice (both sexes) were 10–12 months old. In c,e,g,i–k and m data are presented as truncated violin plots; continuous line, median; and dotted line, interquartile range. Significance was determined by one-way ANOVA followed by the Tukey post hoc test.

Source data

Full size image

Immunocytochemical analysis indicated that a 30-d treatment with Debio-025 compared to vehicle substantially reduced CypA+ and MMP9+ immunofluorescent signal in CD13+ pericytes in the cortex and hippocampus of APOE4;5xFAD mice down to the levels seen in APOE3;5xFAD mice, respectively (Fig. 7b–e). Consistent with inhibition of the CypA-MMP9 BBB-degrading pathway3, treatment of APOE4;5xFAD mice with Debio-025 increased the length coverage of tight junction proteins ZO-1 and occludin that are both well-known substrates for enzymatic degradation by MMP9 (ref. 3) (Fig. 7f–i). However, Debio-025 did not have an effect on CypA+ and/or MMP9+ signal in CD13+ pericytes, nor on ZO-1 and occludin length in APOE3;5xFAD mice. Inhibition of the CypA-MMP9 pathway in APOE4;5xFAD mice improved BBB integrity as shown in vivo by DCE-MRI (Fig. 7) and ex vivo by reduced perivascular capillary fibrinogen deposits (Fig. 7g) and was associated with improved pericyte coverage (Fig. 7h,i).

We then found that Debio-025 compared to vehicle prevented further loss of neurons and axon density in the cortex and hippocampus of APOE4;5xFAD mice (Fig. 8a–c) and improved performance on NOL and novel object recognition (NOR) hippocampal tests (Fig. 8d,e) bringing them nearly to the levels seen in APOE3;5xFAD mice, but did not have an effect on neuronal phenotype or behavior in APOE3;5xFAD mice. Notably, age-matched APOE3 control mice not expressing 5xFAD transgenes had still notably higher neuronal counts, neuritic density and better performance on hippocampal NOR test than Debio-025-treated APOE4;5xFAD mice or APOE3;5xFAD mice (Extended Data Fig. 4). These data suggest that Debio-025 ameliorates APOE4-mediated loss of neurons and behavioral deficits relative to APOE3, but does not protect against neurodegenerative changes caused by 5xFAD transgenes and subsequent direct Aβ neurotoxicity, as previously shown in 5xFAD mice29,37,52. Consistent with these findings, Debio-025 had no effect on Aβ42 or Aβ40 levels in the cortex and hippocampus of APOE4;5xFAD or APOE3;5xFAD mice (Fig. 8f,g), confirming that protection of neuronal phenotype in APOE4;5xFAD mice is independent of Aβ pathology.

Fig. 8: Debio-025 prevents further loss of neurons and improves cognitive function in APOE4;5xFAD mice without affecting Aβ42 or Aβ40 levels in the brain.
figure 8

a–c, Cortical NeuN+ neurons (red) and SMI312+ neurites (green) (a) and quantification of NeuN+ neurons (b) and SMI312+-neuritic density (c) in the cortex (Ctx) and hippocampus (Hipp) in E3+FAD (red empty circles) and E4+FAD (red-filled circles) mice treated with vehicle or Debio-025; n = 5 mice per group. Scale bar, 20 μm. d,e, NOL (d) and NOR (e) in E3+FAD vehicle-treated (n = 8; red empty circles), E3+FAD Debio-025-treated (n = 9; red empty circles), E4+FAD vehicle-treated (n = 7; red-filled circles) and E4+FAD Debio-025-treated (n = 8; red-filled circles) mice. f,g, Aβ42 (f) and Aβ40 (g) levels in the Ctx and Hipp in E3+FAD vehicle-treated (n = 8; red empty circles), E3+FAD Debio-025-treated (n = 8; red empty circles), E4+FAD vehicle-treated (n = 8; red-filled circles) and E4+FAD Debio-025-treated (n = 8; red-filled circles) mice. Mice (both sexes) at 10–12 months old were the same as in Fig. 7. In b–g, data are presented as violin plots; continuous line, median; and dotted line, interquartile range. Significance was obtained by one-way ANOVA followed by Tukey post hoc test.

Source data

Full size image

Discussion

Our findings show that old APOE4 knock-in mice and APOE4;5xFAD mice, compared to their respective APOE3 lines develop an accelerated BBB breakdown associated with loss of pericyte capillary coverage and reductions in CBF associated with diminished capillary density both in the presence and absence of Aβ. Tissue analysis revealed reductions in the tight junction proteins ZO-1 and occludin coverage length, which correlated with activation of the CypA-MMP9 BBB-degrading pathway in pericytes. As previously reported, this pathway is activated in pericytes in transgenic APOE4 lines3, as well as at the BBB endothelial cells and pericytes in human APOE4 AD carriers as shown by postmortem tissue analysis2 and in living APOE4 carriers, as shown by cerebrospinal fluid analysis17. When activated, the CypA-MMP9 pathway leads to MMP9-mediated enzymatic degradation of the BBB tight junction proteins causing BBB breakdown3,50.

As recently reported40, we found comparable Aβ pathology in old APOE3;5xFAD and APOE4;5xFAD mice, but with higher severity in females than in males in both genotypes. Notably, elevated Aβ42 levels, the major Aβ species in 5xFAD mice29,30,37,40 and of Aβ40, did not correlate with BBB breakdown or CBF reductions in these mice, suggesting that APOE4-mediated vascular dysfunction in old mice is Aβ-independent. Moreover, neuron loss and behavioral deficits in APOE4 compared to APOE3 lines correlated well with losses in BBB integrity and CBF, but not with Aβ pathology, as indicated by the lack of correlation between neuron loss and Aβ42 levels, or behavioral deficits and Aβ42 levels.

Loss of neurons and neuritic density in 5xFAD mice crossed with APOE3 and APOE4 mice is consistent with findings of neuron and synaptic loss in older 5xFAD mice, as reported29,39. Strong correlation between BBB dysfunction and neurodegenerative and behavioral changes in APOE4 compared to the respective APOE3 lines, is consistent with enhanced brain capillary leakages of blood-derived neurotoxic proteins such as albumin53, thrombin, plasminogen and/or iron-containing proteins3,5 or fibrinogen41, all capable of causing neuron and synaptic dysfunction and/or eventually neuron death by different but complementary mechanisms. Additionally, greater pericyte loss in APOE4 compared to APOE3 lines in the absence and presence of Aβ pathology results in a greater loss of pericyte-derived neurotrophic growth factors, such as pleiotrophin6, which might further accelerate neuron loss particularly in the presence of diminished CBF, as we recently reported6.

How different AD pathologies interact to contribute to an advanced AD disease stage with age has been understudied and poorly understood. The present findings indicate that vascular dysfunction, particularly the BBB breakdown, plays a important role accelerating neurodegenerative process and cognitive impairment in old Alzheimer’s mice independently of Aβ and therefore it could be an important new therapeutic target for treating an advanced disease stage for which anti-Aβ strategies have proven to be ineffective. Indeed, treatment of APOE4;5xFAD mice with CypA inhibitor Debio-025 suppressed the CypA-MMP9 pathway in pericytes, which in turn improved BBB integrity and prevented further neuron loss and behavioral deficits despite having no effect on Aβ pathology. These results suggest that interventions directed at the BBB repair and improvement of vascular dysfunction are potentially viable new approaches to slow down and/or arrest neurodegenerative process and cognitive decline in an advanced AD stage. As Debio-025 has been used in humans for non-neurological applications46 this approach may hold potential to treat cognitive impairment in APOE4 carriers that show activation of the CypA-MMP9 pathway in the cerebrospinal at an early disease stage17 and at the BBB at a later disease stage with fully developed AD as shown by neuropathological analysis2.

One limitation of the present Debio-025 experiment, however, is that it has been performed in somewhat younger 10–12-month-old APOE4;5xFAD and APOE3;5xFAD mice, compared to the rest of the study that has been carried out in older 18–24-month-old mice. Although, at 10–12 months of age APOE4;5xFAD compared to APOE3;5xFAD mice develop accelerated BBB breakdown, behavioral deficits and notable loss of neuronal networks42,51 and both genotypes develop substantial Aβ pathology30,40, as we have also demonstrated, it remains to be seen whether Debio-025 treatment will have similar striking effects in 18–24-month-old mice. This, however, is a challenging experiment because of increased mortality of APOE4;5xFAD mice at that old age as reported40 and as we also observed.

Our data show that activation of the CypA-MMP9 pathway leads to reductions in tight junction proteins causing BBB breakdown in APOE4 and APOE4;5FAD mice, but whether nonspecific caveolar transcytosis is also affected by APOE4, as in some other models of BBB breakdown54, remains presently unknown and should be addressed by future studies. It has been also reported that epidermal growth factor (EGF) can increase capillary length and improve BBB permeability in APOE4;5FAD mice42, but whether the EGF–EGFR signaling pathway can play a role in spontaneous BBB breakdown as observed in APOE4;5FAD mice has not been studied. Additionally, future studies should address at the cellular and molecular level how APOE4 leads to CBF reductions and whether loss of CBF is related to loss in pericyte coverage and/or reductions in microvascular density, as suggested by our correlation data and/or is caused by activation of the CypA-MMP9 pathway in pericytes or alternatively by a different underlying mechanism.

The present study confirms that Aβ pathology increases BBB breakdown and CBF deficits in APOE3;5xFAD mice compared to APOE3 control mice, which is to be expected given that 5xFAD mice alone develop BBB breakdown34,35,36,37 and CBF deficits38, as a result of direct Aβ vasculotoxicity. Interestingly, we also found activation of the CypA–MMP9 pathway in pericytes in APOE3;5xFAD mice relative to APOE3 control mice. Whether this could be attributed to direct toxic effects of Aβ on pericytes43 and/or to Aβ-mediated proteasomal degradation of LRP155 resulting in loss of this receptor which is required for suppression of the CypA–MMP9 pathway by apoE3 (ref. 3), remains to be determined by future studies. Namely, studies in pericyte cultures and transgenic mice show that apoE3 requires LRP1 to maintain CypA synthesis within low physiological range, whereas poor interaction of apoE4 with LRP1 in pericytes leads to activation of this pathway via transcriptional activation of CypA triggering NF-κB transcriptional activation of MMP9, which increases active MMP9 levels degrading BBB tight junction proteins3. Furthermore, silencing LRP1 in vivo in APOE3 knock-in mice blocks apoE3-mediated suppression of CypA leading to activation of the CypA–MMP9 pathway and BBB breakdown similarly as seen in APOE4 mice3. As LRP1 is reduced in blood vessels in AD in humans2,55,56,57 and in AD Aβ models55,58,59,60, it is possible that LRP1 loss from blood vessels and pericytes may contribute to activation of the CypA–MMP9 pathway in APOE3;5xFAD mice.

In conclusion, our findings show that APOE4 accelerates advanced-stage vascular dysfunction, BBB breakdown and neurodegeneration in AD mice via the CypA pathway in pericytes independently of Aβ. These findings have implication not only for better understanding the pathogenesis of an advanced-stage vascular and neurodegenerative disorder in AD, but also for potential treatment of APOE4 carriers with advanced AD for whom we do not have yet an effective apoE-based therapy to offer.

Methods

Animals

As previously described30, APOE3;5xFAD and APOE4;5xFAD mice, also commonly called E3FAD and E4FAD mice, respectively, were generated by crossing heterozygous 5xFAD± line (Tg6799 on a C57BL/6 and SJL background) that coexpress five FAD mutations (APP K670N/M671L, I716V, V717I and PS1 M146L, L286V) under the control of the neuron-specific mouse Thy-1 promoter29 with homozygous human APOE3+/+ and APOE4+/+ targeted replacement mice (on C57BL/6 background)28. APOE3;5xFAD and APOE4;5xFAD littermate controls or noncarriers derived from these litters were 5xFAD−/−;APOE3+/+ and 5xFAD−/−;APOE4+/+, respectively, which we refer to as APOE3 and APOE4 knock-in control mice. Breeding pairs were generously provided by M. J. LaDu (University of Illinois at Chicago).

This study used both male and female mice at 18–24 months of age. The survival rate of mice at age 18 months was 90% and 80% for APOE3 and APOE4 knock-in littermate controls and 75% and 55% for APOE3;5xFAD and APOE4;5xFAD lines and was driven by early death of female APOE4 mice, as previously shown40. All survivors between 18 and 24 months of age were used in the study. Group sizes varied from n = 12–16 for MRI studies, n = 17–25 for behavioral studies, n = 14–17 for Aβ assays and n = 5–6 for tissue immunohistology analysis for all four groups APOE3, APOE4, APOE3;5xFAD and APOE4;5xFAD. All procedures were approved by the Institutional Animal Care and Use Committee at the University of Southern California using US National Institutes of Health guidelines. All animals were randomized for their genotype information and were included in the study. The operators responsible for experimental procedure and data analysis were blinded and unaware of group allocation throughout the experiments.

Treatment with Debio-025

Debio-025 (MedChem Express, cat. no. HY-12559), a non-immunosuppressive CypA inhibitor, was administered in 10–12-month-old APOE3;5xFAD and APOE4;5xFAD mice generated as above (n = 7–11 per group as indicated in Fig. 8 legend) at 10 mg kg−1 d−1 i.p. for 30 d following a dosing regimen as previously reported in mouse models48,49,50. Control mice received vehicle for 30 d. NOL and NOR behavioral tests were performed 30 d after Debio-025 or vehicle treatment before the DCE-MRI study. Mice were killed after 30 d and brains collected for tissue analysis.

Magnetic resonance imaging

As we previously reported6,41, mice were scanned with either a Biospec 7T system (300 MHz, Bruker) at the California Institute of Technology or a MR Solutions 7T positron emission tomography–MR system (MR Solutions Ltd.) at the Zilkha Neurogenetic Institute (University of Southern California). The Bruker system was equipped with the standard B-GA12 gradient set (~114-mm inner diameter; 400 mT.m−1 maximum gradient) and a 35-mm internal diameter quadrature volume coil (M2M Imaging). The MR Solutions system was equipped with the MRS cryogen-free MRI system (bore size ~24 mm, up to 600 mT.m−1 maximum gradient) and a 20-mm internal diameter quadrature bird cage mouse head coil. Comparable sequences and parameters were used with both MR scanners, as described below.

As we previously reported6,41, mice were anesthetized by 1–1.5% isoflurane/air. Respiration rate (80 ± 10 breaths min−1) and body temperature (36.5 ± 0.5 °C) were monitored during the experiments using an abdominal pressure-sensitive probe and a rectal temperature probe. The isoflurane dose and heated air flow was adjusted continuously to ensure stable and reproducible depth of anesthesia. The sequences were collected in the following order: T2-weighted (two-dimensional-fast spin echo, TR/TE 4,900/45 ms, 34 slices, slice thickness 500 μm, in-plane resolution 70 × 70 μm2) to obtain structural images; DCE protocol for the BBB permeability assessment; and finally, DSC imaging for CBF. Total imaging time was approximately 40 min per mouse.

As we previously reported6,41, DCE-MRI imaging protocol was performed within the dorsal hippocampus territory and included measurement of pre-contrast T1-values using a variable flip angle fast low-angle shot (FLASH) sequence (FA = 5, 10, 15, 30 and 45°, TR/TE = 18 ms/4 ms, slice thickness 1 mm, in-plane resolution 85 × 85 μm2), followed by a dynamic series of 325 T1-weighted images with identical geometry and a temporal resolution of 4.6 s (FLASH, TR/TE 18 ms/4 ms, flip angle 15°, slice thickness 1 mm, in-plane resolution 85 × 85 μm2). Using a power injector, a bolus dose (140 μl) of 0.5 mmol kg−1 gadolinium-diethylenetriamine pentaacetic acid (Gd-DTPA), BioPAL, Inc., diluted in saline 1:6) was injected via the tail vein at a rate of 600 μl min−1. DCE images were collected for 5 min before and 20 min after the injection. DSC-MRI imaging was performed on the exact same geometry. A dynamic series of 100 T2*-weighted images was used, with a temporal resolution of 1.5 s (FLASH, TR/TE 18 ms/4 ms, slice thickness 1 mm, flip angle 15°, in-plane resolution 170 × 270 μm2). A second bolus dose (140 μl) of Gd-DTPA (1:1) was injected via the tail vein at a rate of 1,000 μl min−1. DSC images were collected over 120 s after the injection.

MRI analysis

T1 mapping

As we described previously6,41, T1 relaxation times were estimated using the variable flip angle method, before Gd-DTPA injection, with a series of FLASH images with varying FA and constant TR and TE using the standard saturation recovery equation.

Blood–brain barrier permeability

As we previously reported6,41, post-processing of the collected DCE-MRI data was performed using in-house DCE processing software (Rocketship) implemented in Matlab61. The unidirectional blood-to-brain transfer constant, Ktrans, to intravenously injected gadolinium-based contrast agent was determined in both primary somatosensory cortex and hippocampus in mice using a modified method, as previously reported in both humans17,23,26 and mice6,41 with the post-processing Patlak analysis62. As reported previously6,17,23,26,41, we determined the arterial input function (AIF) in each mouse from the common carotid artery.

As we previously reported6,41, the total tracer concentration in the tissue, Ctissue (t), can be described as a function of the arterial vascular concentration CAIF (t), the intravascular blood volume vp and the BBB permeability transfer constant Ktrans that represents the flow of the tracer from the intravascular to the extravascular space using equation (1) below.

$$C_{\rm{tissue}}\left( t \right) = K_{\rm{trans}}\mathop {\int}\limits_0^t {C_{\rm{AIF}}} \left( \tau \right)d\tau + v_p \cdot C_{\rm{AIF}}\left( t \right)$$
(1)

Cerebral blood flow

As reported previously, DSC-MRI data were processed using our in-house Matlab code6,41,63. In brief, assuming a linear relationship between the signal drop induced by Gd-DTPA and concentration, these quantities can be related via:

$$S\left( t \right) = S_0 \times e^{ - r_{2^ \ast }.C\left( t \right).{\rm{TE}}}$$
(2)

Where S(t) is the signal intensity at time t after bolus injection for any given voxel, S0 is the mean pre-contrast signal intensity, r2* relaxivity constant of the contrast agent used, C(t) is the concentration of contrast as a function of time and TE is the time echo of the acquisition sequence. From the previous formula, the conversion from signal to contrast agent concentration is straightforward and occurs via:

$$C\left( t \right) = \frac{{ - 1}}{{r_{2^ \ast }.{\rm{TE}}}}\ln \left( {\frac{{S\left( t \right)}}{{S_0}}} \right)$$
(3)

As we previously reported6,41, the profile of this concentration curve is heavily influenced by the manner in which the tracer bolus is injected into the mouse. To define the shape of the bolus curve, a representative AIF was obtained for each mouse individually. The AIF was obtained from the image data via manual delineation, typically from the common carotid arteries (same as for Ktrans mapping). By defining the residual function, R(t), which represents the fraction of tracer presently circulating at time t, the relationship between tracer concentration and blood flow can be modeled as a convolution between R(t) and the AIF:

$$C_t\left( t \right) = \frac{\rho }{{Kh}}F\mathop {\int}\limits_0^t {C_a} \left( \tau \right)R\left( {t - \tau } \right)d\tau$$
(4)

Where Ct(t) is the concentration of contrast agent in the tissue, F is regional CBF, κH is the ratio of capillary to artery hematocrit (a value of 0.45 was used), ρ is tissue density (1.04 g ml−1), Ca(t) is AIF time course. To solve equation (4), we evoked an oscillation-limited circulant singular value decomposition approach64. Using this deconvolution, R(t) and F values were obtained and regional CBF (ml 100 g−1 min−1) was computed using the equation:

$$\mathrm {CBF} = \frac{{\kappa _H}}{\rho }.\max \left( {R\left( t \right)} \right)$$
(5)

Behavior

Nesting

Nest construction test was performed as previously reported41,43,59. Two hours after the beginning of the dark cycle, the animals were individually placed in clean home cages with a single nestlet. Nests were assessed the next morning and evaluated following the five-point scale as we described in detail59.

Novel object location

This was performed as we have previously reported5,6,43. Briefly, animals were placed in a 30-cm3 box and allowed to habituate to the testing area for 10 min. Animals were then placed back in their cages and two identical approximately 5 × 5-cm objects were placed in the top left and right corner of the testing area. Animals were allowed to explore the two objects in the testing area for 5 min before being returned to their cages. After a 1-h interval one of the objects was relocated and the animals were allowed to explore the testing area once again for 3 min. After each trial, the testing area and the objects were thoroughly cleaned with 70% ethanol solution. All the trials, including habituation, were recorded with a high-resolution camera and the amount of time each animal spent exploring the objects was analyzed. Any animals that presented a preference for either of the two identical objects, before replacement with the novel location, were eliminated from the analysis.

Immunohistochemistry

As we previously reported6,41, mice were anesthetized i.p. with 100 mg kg−1 ketamine and 10 mg kg−1 xylazine and transcardially perfused with 20 ml phosphate-buffered saline (PBS) containing 5 mM ethylenediaminetetraacetic acid (EDTA). Brains were removed, snap-frozen and one hemisphere was embedded into OCT compound (Tissue-Tek) on dry ice (the other hemisphere was used for biochemical assays; see Human Aβ40 and Aβ42 specific ELISA section). Brains were cryosectioned at a thickness of 20 μm and then fixed in 4% PFA as previously described. Sections were blocked with 5% normal donkey serum (Vector Laboratories)/0.1%Triton-X/0.01 M PBS for 1 h and incubated with primary antibodies diluted in blocking solution overnight at 4 °C. After incubation with primary antibodies, sections were washed in PBS and incubated with fluorophore-conjugated secondary antibodies (see below) and then mounted onto slides with fluorescence mounting medium (Dako). The following primary and secondary antibodies were used, respectively: for pericyte coverage, polyclonal goat anti-mouse aminopeptidase N/ANPEP (CD13) (R&D systems AF2335, 1:100 dilution) and Alexa Fluor 488- or 568-conjugated donkey anti-goat (Invitrogen, A-11055 or A-11057, 1:500 dilution); for fibrin/fibrinogen, polyclonal rabbit anti-human fibrinogen (Dako A0080, 1:500 dilution) and Alexa Fluor 488-conjugated donkey anti-rabbit (Invitrogen, A-21206,1:500 dilution); for tight junction proteins (zonula occludens), rabbit anti-mouse ZO-1 (Invitrogen 40–2200, 1:200 dilution) and Alexa Fluor 568-conjugated donkey anti-rabbit (Invitrogen A-10042, 1:500 dilution); and for occludin, mouse anti-mouse occludin (Invitrogen clone OC-3F10 no. 33-1500, 1:100 dilution) and Alexa Fluor 488-conjugated donkey anti-mouse (Invitrogen A-21202, 1:500 dilution); for CypA, rabbit anti-mouse CypA (Abcam ab42408, 1:100 dilution) and Alexa Fluor 488 or 568-conjugated donkey anti-rabbit (Invitrogen A-21206 or A-10042, 1:500 dilution); for MMP9, rabbit anti-mouse MMP9 (Abcam, ab38898, 1:100 dilution) and Alexa Fluor 488 or 568-conjugated donkey anti-rabbit (Invitrogen A-21206 or A-10042, 1:500 dilution); for amyloid-β, rabbit anti-human β-amyloid (Cell Signaling 8243S, 1:500 dilution) and Alexa Fluor 647-conjugated donkey anti-rabbit (Invitrogen A-31573,1:500 dilution); for neurofilament, mouse monoclonal anti-axonal neurofilament marker (SMI312) (BioLegend SMI312, 1:500 dilution) and Alexa Fluor 488-conjugated donkey anti-mouse (Invitrogen A-21202, 1:500 dilution); and for neuronal, polyclonal rabbit anti-NeuN (Millipore ABN78, 1:500 dilution) and Alexa Fluor 488 or 568-conjugated donkey anti-rabbit (Invitrogen, A-21206 or A-10042; 1:500 dilution). To visualize brain microvessels, sections were incubated with Dylight 488, 594 or 647-conjugated Lycopersicon esculentum lectin (Vector Labs, DL-1174; 1:200 dilution) for 1 h. Sections were imaged with a Zeiss LSM 510 confocal laser-scanning microscope or BZ9000 fluorescence microscope. Z-stack projections and pseudo-coloring were performed using ZEN software (Carl Zeiss Microimaging). Image post-analysis was performed using ImageJ software.

Microscopy analysis

For extravascular brain capillary leakages, pericyte coverage, tight junction protein analysis, CypA and MMP9 quantifications and NeuN+ neuron counts and neurofilament (SMI312)+ axons, 4–6 randomly selected fields in the somatosensory cortex region and/or the CA1 region of the hippocampus were analyzed for each animal in 3–4 nonadjacent sections (~100 μm apart) and averaged per mouse. For Aβ load, whole cortices and hippocampi were used, corresponding to the areas used for Aβ ELISA.

Extravascular leakages

For detection of extravascular brain capillary fibrinogen deposits, an antibody that detects both fibrinogen and fibrinogen-derived fibrin polymers was used. Ten-micron maximum projection z-stacks were reconstructed and the fibrinogen+ perivascular signal on the abluminal side of lectin+ endothelial profiles on microvessels ≤6 µm in diameter was analyzed using ImageJ software Integrated density analysis measurement tool, as we previously described41.

Pericyte coverage

Ten-micron maximum projection z-stacks were reconstructed and the areas occupied by CD13+ pericytes on lectin+ endothelial profiles on microvessels ≤6 µm in diameter was analyzed using ImageJ as we previously described65.

Immunofluorescent ZO-1 and occludin tight junction analysis

The length of ZO-1+ and occludin+ immunofluorescent signal on lectin+ endothelial brain capillary profiles (<6 µm in diameter) in the cortex and the hippocampus was determined using the ImageJ Area measurement tool and expressed in mm of ZO-1 and occludin length per mm2 of the total area of lectin+ endothelial capillary profiles, as we previously described6.

Microvascular length

Ten-micron maximum projection z-stacks were reconstructed and the length of lectin+ endothelial capillary profiles (≤6 μm in diameter) was measured using the ImageJ plugin ‘Neuro J’ length analysis tool, as previously described41,65. The length was expressed in mm of lectin+ endothelial capillary profiles per mm3 of brain tissue.

Immunofluorescent CypA detection

The area occupied by CypA+ immunofluorescent signal on CD13+ pericytes was expressed as the percentage of CD13+ pericyte area using the ImageJ Area measurement tool as we previously described3. Orthogonal views of confocal images of CypA, the pericyte marker CD13 and lectin+ endothelium showing colocalization of CypA with CD13+ pericytes were generated from a single-plane projection from a 20 single-plane z-stack using ZEN software (black edition). For the correlation analyses between CypA+ pericytes and ZO-1 length and CypA+ pericytes and occludin length immunostaining for CypA and CD13, ZO-1 and lectin-endothelial staining and occludin and lectin-endothelial staining was performed on adjacent sections.

Immunofluorescent MMP9 detection

The area occupied by MMP9+ immunofluorescent signal on CD13+ pericytes was expressed as the percentage of CD13+ pericyte area using the ImageJ Area measurement tool as we previously described3. Orthogonal views of confocal images of MMP9, the pericyte marker CD13 and lectin+ endothelium showing colocalization of MMP9 with CD13+ pericytes were generated from a single-plane projection from a 20 single-plane z-stack using ZEN software (black edition). For the correlation analyses between MMP9+ pericytes and ZO-1 length and MMP9+ pericytes and occludin length immunostaining for MMP9 and CD13, ZO-1 and lectin-endothelial staining and occludin and lectin-endothelial staining was performed on adjacent sections.

Aβ load

Primary somatosensory cortex and hippocampus were analyzed separately and the percentage area occupied by amyloid was quantified using Otsu-thresholding in ImageJ, as previously described37.

NeuN+ neuronal nuclei counting

NeuN+ neurons were quantified using the ImageJ Cell Counter analysis tool, as previously described6.

Neurofilament (SMI312)+ axons

As we previously described6,41, 10-μm maximum projection z-stacks were reconstructed and SMI312+ signal was subjected to threshold processing and analysis using ImageJ.

Human Aβ40 and Aβ42 ELISA

Cortices and hippocampi were isolated and snap-frozen from one brain hemisphere (the other hemisphere was used for immunohistochemical assays; see Immunohistochemistry section). Cortices and hippocampi were sequentially extracted with sonication, first extracted with ten volumes of Tris-buffered saline with complete protease inhibitor cocktail (Roche), centrifuged at 20,000g for 30 min at 4 °C, followed by homogenization of pellet in ice-cold guanidine buffer (5 M guanidine hydrochloride/50 mM Tris HCl, pH 8). Guanidine extracts were diluted 1:200 and 1:20,000 for measurement of human Aβ40 and Aβ42, respectively by a Meso Scale Discovery assay (K15200E-1).

Statistical analysis

Sample sizes were calculated using nQUERY assuming a two-sided α-level of 0.05, 80% power and homogeneous variances for the samples to be compared, with the means and common s.d. for different parameters predicted from published data and our previous studies. For comparison between two groups, an F test was conducted to determine the similarity in the variances between the groups that are statistically compared and statistical significance was analyzed by unpaired two-tailed Student’s t-test. Lilliefors test was used to test normality of the data (XLSTAT). For multiple comparisons, the F test was also used to determine the equality of variances between the groups compared and one-way ANOVA followed by Tukey’s post hoc test, which was used to test statistical significance. For correlation analyses, we employed Pearson’s correlation coefficient to measure the strength of the linear relationship between two variables. All analyses were performed using GraphPad Prism v.8.4.2 software and by an investigator blinded to the experimental conditions. Data are presented as violin plots with median and interquartile range as indicated in figure legends. A P value >0.05 was considered statistically not significant.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.