Amyloid‐beta 1‐40 is associated with alterations in NG2+ pericyte population ex vivo and in vitro

Summary The population of brain pericytes, a cell type important for vessel stability and blood brain barrier function, has recently been shown altered in patients with Alzheimer's disease (AD). The underlying reason for this alteration is not fully understood, but progressive accumulation of the AD characteristic peptide amyloid‐beta (Aβ) has been suggested as a potential culprit. In the current study, we show reduced number of hippocampal NG2+ pericytes and an association between NG2+ pericyte numbers and Aβ1‐40 levels in AD patients. We further demonstrate, using in vitro studies, an aggregation‐dependent impact of Aβ1‐40 on human NG2+ pericytes. Fibril‐EP Aβ1‐40 exposure reduced pericyte viability and proliferation and increased caspase 3/7 activity. Monomer Aβ1‐40 had quite the opposite effect: increased pericyte viability and proliferation and reduced caspase 3/7 activity. Oligomer‐EP Aβ1‐40 had no impact on either of the cellular events. Our findings add to the growing number of studies suggesting a significant impact on pericytes in the brains of AD patients and suggest different aggregation forms of Aβ1‐40 as potential key regulators of the brain pericyte population size.

between pericyte number and Ab load has been found in transgenic APP mouse, an in vivo model of AD (Winkler, Sagare & Zlokovic, 2014). Secondly, in vitro studies show that Ab influences survival of cultured brain pericytes. However, it is important to point out that Ab can appear in several different species, whereof  are two of the most abundant forms. Alzheimer 0 s disease is foremost associated with the former specie, as insoluble Ab1-42 fibrils are the major component of Ab plaques and soluble oligomer Ab1-42 is strongly neurotoxic (Dahlgren et al., 2002). The impact of Ab1-42 on brain pericytes in the human brain is scarcely investigated, but a few in vitro studies show reduced cell survival after prolonged Ab1-42 stimulation (Verbeek, de Waal, Schipper & Van Nostrand, 1997;Wilhelmus et al., 2007) and altered shedding of the pericyte adhesion molecule chondroitin sulphate proteoglycan NG2 in the presence of oligomer and fibril Ab1-42 (Schultz, Nielsen, Minthon & Wennstrom, 2014). In comparison with Ab1-42, Ab1-40 has in several human brain tissue studies been shown to have an impact on pericytes. This Ab specie is not as prone to aggregate as Ab1-42, but can nonetheless form depositions, so-called cerebral amyloid angiopathy (CAA), in the vessel walls in the brain (Maia, Mackenzie & Feldman, 2007). The CAA is often associated with pericytes displaying degenerative features (Verbeek, Van Nostrand, Otte-Holler, Wesseling & De Waal, 2000), and thus, it has been hypothesized that aggregated Ab1-40 is particularly toxic for pericytes. Studies showing a reduced cell survival of cultured human brain pericytes in presence of Dutch variant of Ab1-40 (a mutation promoting aggregation) support this hypothesis (Verbeek et al., 1997).
To conclude, several clinical and preclinical studies show that Ab influences the pericyte population, but whether either of the Ab species (i.e. Ab1-40 or Ab1-42) and whether aggregation forms of the Ab species play a role in the pericyte loss seen in AD patients remain to be investigated. We therefore analysed the pericyte population in AD patients, in comparison with nondemented controls, by staining postmortem brain tissue against NG2 and laminin alpha (a)5. These two markers are found on different subset of pericytes: NG2 is expressed by activated pericytes on arterioles/capillaries (Stapor, Sweat, Dashti, Betancourt & Murfee, 2014) and laminin is expressed by mature pericytes on arterioles and capillaries as well as postcapillary venules (Yousif, Di Russo & Sorokin, 2013). The association between pericyte numbers and levels of Ab1-40 or Ab1-42 in brain homogenates was thereafter analysed. We also investigated the direct impact of different aggregations forms of Ab1-40 and Ab1-42 on proliferation, cell survival and caspase 3/7 activity of cultured human brain pericytes. To examine whether hippocampal pericytes are affected in patients with Alzheimer's disease (AD), we counted the number of NG2+ and laminin+ pericytes and measured the length of vessels in each staining. The molecular layer (ML) hippocampal subarea was selected since capillaries (<10 lm in diameter) in the area are easy to distinguish and amyloid-beta plaques are present in this area in later stages of AD. We found significantly fewer NG2+ pericytes as well as fewer NG2+ pericytes/vessel length in the ML of AD patients compared to nondemented controls (F(1, 20) = 1.506, p = .039, g 2 q = 0.196 and F(1, 20) = 5.563, p = .035, g 2 q = 0.204, respectively) ( Figure 1a,b), but no alterations in vessel length. However, neither number of laminin+ pericytes nor laminin+ pericytes/vessel length were significantly altered when the two groups were compared (Figure 1c,d). Representative images of NG2+ pericytes and laminin+ pericytes are shown in Figure 1e-g and an image of hippocampus where ML is outlined is shown in Figure 1h. We also investigated whether APOE-e4 genotype affects the measured pericyte variables, but we found no differences in either NG2+ or laminin+ pericyte numbers, vessel length or number of pericytes/vessel length when comparing APOE-e4 carriers (n = 13) with APOE-e4 noncarriers (n = 12). Finally, we found no association between age and any of the measured pericyte variables and no difference in pericyte variables when males were compared with females or when patients with severe vascular disease where compared with patients without severe vascular disease.

| Correlations between Ab1-40 levels and number of pericytes and number of pericytes/vessel in AD patients
Next, we investigated whether levels of Ab1-40 and Ab1-42 in the hippocampal samples were associated with the pericyte loss in AD patients. We therefore dissolved PFA-fixed brain sections in formic acid and measured the total amount of Ab levels (aggregated + soluble Ab). We found a trend towards higher levels of Ab1-42 and higher Ab1-42/40 ratio in the AD patients compared to nondemented controls (17.91 AE 19.59 vs. 4.64 AE 11.91, p = .061 and 7.39 AE 10.78 vs. 0.78 AE 1.46, p = .058, respectively). In contrast, levels of Ab1-40 were not significantly different when the two groups were compared (10.10 AE 12.31 vs. 5.10 AE 4.98, p = .209).
Correlation analysis further showed that levels of Ab1-40 and number of NG2+ pericytes significantly correlated when the whole cohort was analysed (r = .470, p = .024). This correlation remained significant when the AD patient group was separately analysed (r = .713, p = .014; Figure 2a), but only a trend to significance was seen when the nondemented control group was analysed (r = .504, p = .095). We also found a positive correlation between Ab1-40 levels and the number of NG2+ pericytes/vessel length in AD patients (r = .705, p = .015) (Figure 2b), but a similar association was not found in nondemented controls. Finally, when the whole cohort was analysed, we could see a negative and significant correlation between the Ab1-42/40 ratio and the NG2+ number of pericytes (r = À0.451, p = .035) as well as between Ab1-42/40 ratio and the number of NG2+ pericytes/vessel length (r = À0.544, p = .009). This correlation was lost when the analysis was performed on the AD and NC groups separately. No significant correlations were found between the brain Ab1-42 levels and any of the NG2+ pericyte variables (i.e. the number of pericytes, the length of the vessels and the number of pericytes/vessel), regardless of whether the whole cohort was analysed or the two groups were analysed separately ( Figure S1).
No correlations between Ab1-40, Ab1-42 or Ab1-42/40 ratio and any of the laminin+ pericyte variables were found, regardless of whether the whole cohort was analysed or the two groups were analysed separately. Furthermore, we found no correlation between the measured NG2+ or laminin+ pericyte variables and ABC amyloid stages or Braak stages for NFT.

| Cell culture studies
Since our postmortem studies indicate that Ab1-40 levels are associated with changes in NG2+ pericytes numbers in the AD brain, we found it interesting to investigate the direct impact of different aggregation forms of Ab1-40 on brain pericyte population. For this purpose, we used NG2-and PDGFR-b-expressing HBVP as a cell culture model of brain pericytes and analysed alterations in viability, caspase activity and proliferation of these cells after exposure to different aggregation forms of Ab1-40 and Ab1-42. Analysis of both Ab preparations used in the experiment showed that the fibril-  The area in which pericytes were counted, that is the molecular layer (ML) of hippocampus, is outlined in image (h). Scale bar (e-g) 20 lm. Scale bar (h) 1,000 lm F I G U R E 2 Scatter plots with linear regression lines demonstrating correlations between Ab1-40 levels in formic acid-treated hippocampus sections and number of pericytes/lm 2 in the molecular layer of hippocampus. Scatter plot in (a) demonstrates the correlation between Ab1-40 levels and pericyte numbers/lm 2 within the AD patient group. Scatter plot in (b) shows the correlation between Ab1-40 levels and number of pericytes/vessel length within the AD patient group. Data are analysed using Spearman correlation test approximately 100nm and above, whereas the oligomer preparations contained primarily oligomers ( Figure S4).

| Amyloid-beta 1-40 alters viability in an aggregation-dependent manner
In the viability experiment, we exposed HBVP to monomer and oligomer-and fibril-EP Ab1-40 for either 24 or 96 hr. Already at 24 hr, we found significantly increased necrotic cell death (indicated by increased LDH activity) after fibril-EP Ab1-40 exposure compared to 2.2.2 | Amyloid-beta 1-40 alters caspase 3/7 activity in an aggregation-dependent manner Since previous studies in rat neurons have shown that Ab25-35induced cell death is mediated via caspases (Harada & Sugimoto, 1999;Marin et al., 2000), we found it interesting to investigate whether exposure to Ab1-40 and Ab1-42 also affects caspase activation. We therefore analysed activity of caspases 3 and 7 (the two most prominent executor caspases in the apoptosis pathway (Elmore, 2007)) in HBVP after exposure to the three different aggregation forms of Ab1-40 and Ab1-42. Our analysis showed significantly increased caspase 3/7 activity after 24-hr exposure to fibril-EP Ab1- 99770.50 AE 11318.10, p = .839; Figure 4b).

| Ab1-40 alters proliferation in an aggregation-dependent manner
Next, we investigate the impact of the three different aggregation forms of Ab1-40 and Ab1-42 on proliferation of HBVPs by the use F I G U R E 3 Bar graphs demonstrating alterations in cytotoxicity, measured by LDH assay, in cell culture supernatants from HBVPs after exposure to 10 lM monomer, oligomer-or fibril-EP Ab1-40 and Ab1-42 for 96 hr. Graph in (a) shows significantly increased LDH activity (i.e. increased cell death) in HBVP exposed to fibril-EP Ab1-40, oligomer-EP Ab1-42 and fibril-EP Ab1-42 compared to Ctrl O/F. Graph in (b) shows the significantly decreased LDH activity (i.e. increased viability) in HBVP exposed to monomer Ab1-40 and monomer Ab1-42 compared to Ctrl M. Data were analysed using one-way analysis of variance (ANOVA), followed by Dunnett's post hoc correction (n = 4 comparisons) (oligomer-and fibril-EP Ab1-40 and Ab1-42) or Student's t-test (monomers Ab1-40 and Ab1-42). Results are presented as means AE standard deviations. Significant difference at **p < .01 and ***p < .001 F I G U R E 4 Bar graphs demonstrating alterations in cytotoxicity, measured by caspase 3/7 activity, in HBVPs after exposure to 10 lM monomer, oligomer-or fibril-EP Ab1-40 and Ab1-42 for 24 hr. Graph in (a) shows the significantly increased relative caspase 3/7 activity in HBVP after exposure of fibril-EP of Ab1-40 compared to Ctrl O/F. Graph in (b) demonstrates the significant increased relative caspase 3/7 activity in HBVP after exposure to monomer Ab1-40 compared to Ctrl M. Data were analysed using one-way analysis of variance (ANOVA), followed by Dunnett's post hoc correction (n = 4 comparisons) (oligomer-and fibril-EP of Ab1-40 and Ab1-42) or Student's t-test (monomer preparations of Ab1-40 and Ab1-42). Results are presented as means AE standard deviations. Significant difference at *p < .05 and ***p < .001 a b c d e f F I G U R E 5 Bar graphs demonstrating alterations in Ki67+ cells, indicative of cell proliferation, in HBVPs after exposure to 10 lM monomer, oligomer-or fibril-EP Ab1-40 for 24 hr. Image in (a) shows untreated HBVP stained against Ki67 and DAPI. Image in (b) demonstrates Ab1-40 monomer-exposed HBVP stained against Ki67 and DAPI. Image in (c) demonstrates HBVP stained against Ki67 and DAPI after exposure to oligomer-EP Ab1-40. Image in (d) shows fibril-EP Ab1-40-exposed HBVP stained against Ki67 and DAPI. Graph in e) shows the significantly decreased number of Ki67+ HBVPs after exposure of fibril-EP of Ab1-40 compared to Ctrl O/F. Graph in (f) demonstrates the significantly increased number of Ki67+ HBVPs after exposure of monomer EP of Ab1-40 compared to Ctrl M. Data were analysed using one-way analysis of variance (ANOVA), followed by Dunnett's post hoc correction (n = 2 comparisons) (oligomer-and fibril-EP of Ab1-40) or Student t-test (monomer preparations of Ab1-40). Results are presented as means AE standard deviations. Significant difference at *p < .05, **p < .01 exposure to oligomer-EP Ab1-40 (0.32 AE 0.07 vs. 0.32 AE 0.10, p = .992; Figure 5e).  in line with the previous study by Sengillo and colleagues, demonstrating a pericyte loss in AD patients (Sengillo et al., 2013). In that study, the number of mural cells (an umbrella term for smooth muscle cells and pericytes) was analysed by measuring the amount of PDGFR-b+ (a marker for migrating immature mural cells (Song, Ewald, Stallcup, Werb & Bergers, 2005)) and CD13+ (marker for activated pericytes (Svensson, Ozen, Genove, Paul & Bengzon, 2015)) cells associated with vessels/capillaries in hippocampus. They found a 60% reduction in hippocampal mural cells and a 33% reduction in hippocampal pericytes, which is in the order of magnitude with our own study (approximately 26% reduction in NG2+ pericytes). In view of these findings, we conclude that different subsets of pericytes involved in migration, activation and remodelling (NG2+, PDGFR-b+ and CD13+ pericytes) are affected by AD pathology, whereas mature and quiescent pericytes (laminin+) are less affected.

| DISCUSSION
As mentioned in the introduction, previous studies have suggested Ab, in particular Ab1-40, to be toxic for pericytes (Verbeek et al., 1997;Wisniewski, Wegiel, Wang & Lach, 1992). We therefore Also, fibril morphology and distribution have been used to distinguish the different forms, where individually dispersed fibrils with a curvilinear feature have been defined as protofibrils (Hartley et al., 1999;O'Nuallain et al., 2010;Paranjape, Gouwens, Osborn & Nichols, 2012;Relini et al., 2010). Our fibril preparations contained straight, aggregated fibrils, which were approximately 100 nm and above. In addition, our OMAB and ThT analysis demonstrated lower amount of oligomer and higher amount of b-sheets in fibril-EP compared to oligomer-EP. We therefore draw the conclusion that our fibril-EP contains high amounts of fibrils, but due to the length of our fibrils, we cannot exclude the possibility that a portion of those fibrils are protofibrils.
Our results further showed that the HBVP viability, 3/7 activity and proliferation were not affected by the oligomer form of Ab1-40.
This is particularly interesting since the oligomer form of Ab1-42, and not the fibril form of this specie, is regarded as the most toxic form (Benilova, Karran & De Strooper, 2012 (Benilova et al., 2012), highlighting the different impact of the two Ab species in regard to oligomer toxicity.
In contrast to fibril-EP Ab1-40, our in vitro results showed that monomer Ab1-40 both had a rescuing and mitogenic impact on HBVPs. Similar beneficial effect of Ab1-40 has been described before. For example, Ab1-40 prevents degradation of cell membrane lipids by inhibiting auto-oxidation of CSF and plasma lipoproteins (Kontush et al., 2001), an event known to be increased in AD patients (Bassett et al., 1999). The Ab1-40 peptide is further the most effective Ab specie to rescue cultured neurons which are dying in response to endogenous Ab inhibition (Plant, Boyle, Smith, Peers & Pearson, 2003). Our own cell viability results, showing decreased cell death as well as decreased caspase 3/7 activity after monomer Ab1-40 exposure, support the idea that Ab1-40 is beneficial for cell survival. But our results also show that it is foremost the monomers that have this ability. Apparently, the Ab1-40 monomers also have a strong mitogenic impact on HBVPs as our proliferation assay showed marked increased cell division in response to the stimulus. This finding is backed up by a previous study demonstrating a mitogenic property of monomer Ab1-40 on cultured human skin fibroblasts (Theda, Drews, Zitnik, Oshima & Martin, 2016), a cell type sharing both origin and certain markers with the pericyte (Birbrair et al., 2013). Interestingly, decreased caspase activity has been associated with hyperproliferation of pancreatic b-cells in mice (Woo et al., 2003). It may thus be that the decreased activity of caspase 3/7 in SCHULTZ ET AL.
| 7 of 13 response to monomer Ab1-40 seen in our study contributes to both the increased cell viability and the increased cell division. Finally, it is important to point out that cell cultures can never replicate a biological system and although pericyte proliferation does occur in the human brain (Fernandez-Klett et al., 2013;Goritz et al., 2011;Matsushita et al., 2015;Paul et al., 2012), the conditions and rate in the brain are not comparable to the same in a Petri dish. Hence, we would like to stress that our cell model is useful foremost as a tool to demonstrate that differences in pericyte response to Ab depend on both aggregation form and specie.

| CONCLUSIONS
To conclude, our results confirm the previous study demonstrating pericyte alterations in AD patients and highlight Ab1-40 as a potential regulator of brain pericyte population. Moreover, our in vitro studies show that the regulatory role of Ab1-40 is dependent on aggregation form, where the monomeric appears to have a rescuing and mitogenic impact, oligomers are quiescent and fibrils induce toxicity. This is particularly interesting given the common notion that the oligomer form of the kin peptide, Ab1-42, is considered to be the culprit behind neuronal loss in AD. Hence, our results point out the differences between Ab species and their aggregation forms in terms of toxicity on pericytes, a finding important to take into consideration when targeting the aggregation process of Ab as a treatment for AD.

| Individuals included in the study
The study includes frozen samples of midlevel hippocampus from clinically and postmortem-verified patients with AD (n = 13) and nondemented controls (n = 9) (Netherlands Brain Bank (NBB)).
Neuropathological assessments were performed by NBB according to Braak stages of neurofibrillary tangles (NFT) spreading (Braak & Braak, 1995) and Lewy bodies (LB) (Braak et al., 2003) and ABC staging for amyloid (Ab) (Schultz, Hubbard, Rub, Braak & Braak, 2000). Individuals which after clinical evaluation were considered nondemented and after neuropathological NFT evaluation were scored as 3 or less were considered as nondemented controls. None of the patients included in the study displayed CAA or LB pathology in hippocampus. Individuals with hypertension in combination with either of the following criteria (i) transient ischaemic attack/stroke, (ii) valvular heart disease, (iii) ischaemic heart disease or (iv) arrhythmias were considered as severe cardiovascular disease (CVD) patients. Single nucleotide polymorphisms at positions rs429358 and rs7412 of the APOE gene were determined by polymerase chain reactions using allele-specific primers. Demographic data and the neuropathological assessment of the individuals are presented in Table 1. Written informed consent for the use of brain tissue and clinical data for research purposes was obtained from all patients or their next of kin in accordance with the International Declaration of Helsinki.

Medisch Ethische
Toetsingscommissie (METc) of VU University has approved the procedures of brain tissue collection, and the regional ethical review board in Lund has approved the study.

| Immunostaining procedures of brain tissue
The frozen hippocampal samples were fixed in 4% paraformaldehyde (PFA) for 4 hr and left in phosphate-buffered saline (PBS) containing 30% sucrose until they were sunken. The tissue was then cut into 40lm-thick sections using a Microtome Leica and stored free floating in antifreeze cryoprotectant solution at À20°C until analysis. In order to examine the number of pericytes, the brain sections were immunohistochemically (IHC) stained against either the pericyte marker NG2  Braak staging of neurofibrillary tangles (NFT) and Lewy bodies (LB) and ABC staging of amyloid (Ab).
to cell experiments, fibril-and oligomer-enriched preparations (EP) of Ab1-40 and Ab1-42 were generated according to previously published protocol (Brannstrom et al., 2014). The Ab1-40 and the Ab1-42 peptides were solubilized in 10 mM NaOH pH 11, and then the pH was adjusted to pH 7.0 by further diluting the solution to a concentration of 100 lM in phosphate buffer. Fibril-EP was generated by a 72-hr incubation with agitation in 37°C and oligomer-EP was generated by 20-min incubation with agitation RT. Monomer prepa-

| Cell culture treatment
The medium was removed 2 hr prior to experiments and replaced with serum-free medium in order to synchronize the cells. After serum starvation, the medium was once again removed and replaced with either serum-free medium (for the toxicity experiments) or repeated independently three times. The cells were incubated in 37°C for 24 or 96 hr depending on the assay. Cell culture supernatant was collected after treatment, centrifuged (275 9 g 5 min, 4°C), aliquoted and stored at À80°C until used. Cells were lysed using cell lysis kit (Sigma-Aldrich), aliquoted and stored at À80°C until used.

| Analysis of cytotoxicity
To evaluate cytotoxic effect, we used two different cytotoxicity assays (i) extracellular lactate dehydrogenase activity (LDH) (measures foremost necrosis) and (ii) caspase 3/7 activity (measures foremost apoptosis). The levels of LDH activity were measured in conditioned medium using LDH assay kit (Sigma-Aldrich). In 96-well plates (Nunc), 100 ll HBVPs cell-free supernatants were mixed with 100 ll reagent buffer, agitated 30 s 300-450 rpm and then incubated 30 min at RT. The absorbance was measured at 490 nm using Microplate Spectrophotometer Infinite M200 and the Magellan version 3.5 software. Caspase 3/7 activity was analysed using Caspase-Glo 3/7 luminescence assay according to the manufacturer's instructions (Promega, Fitchburg, WI).
Cells were incubated 1 hr at RT in dark with secondary antibody (1:500, rabbit-anti-goat Dylight 459). Vectashield Set mounting medium with DAPI (Vector Laboratories) was used to mount the cells.
Proliferation was analysed by counting the number of Ki67+ cells of total DAPI+ cells.

| Statistical analysis
Statistical analysis was performed using SPSS software (version 24 for Mac, SPSS Inc., Chicago, IL). The Kolmogorov-Smirnov test was used to assess normal distribution. Since number of pericytes, vessel length and the ratio of pericytes/vessel length were normally distributed, differences between diagnose groups were analysed by the use of independent-samples t-test. Levels of Ab1-40 and Ab1-42 as well as Ab1-42/40 ratio were not normally distributed, and thus, Mann-Whitney U-test was used when comparison analysis of these values was performed. Consequently, correlations between the investigated variables in the brain study were examined using the Spearman correlation test. Differences between Ctrl O/F and oligomer-and fibril-EP in the in vitro study were analysed using one-way analysis of variance (ANOVA), followed by Dunnett's post hoc correction (comparisons for n = 4), whereas the comparisons between monomer preparations of Ab1-40 and Ab1-42 and their control were analysed using Student's t-test. Results are presented as means AE standard deviations, and a value of p < .05 level was considered statistically significant.

CONF LICT OF I NTEREST
The authors declare that they have no competing interests.