To determine the impact of sAPP on inflammatory reactions in microglia, we treated the N9 microglial cell line7 with sAPP and measured the activity of the transcription factor NF-κB by electrophoretic mobility-shift assay (EMSA) (Fig. 1). These cells responded to sAPP with activation of NF-κB within 90 min. Induction of a κB-binding transcription factor by sAPP in primary neurons is dependent upon an increase in the concentration of cyclic GMP8 and involves a transcription factor distinct from NF-κB (unpublished results). However, the addition of an inhibitor of cGMP-dependent protein kinase did not block the activation of NF-κB by sAPP in N9 cells, and a cell-permeant cGMP analogue did not mimic the effects of sAPP in these cells. These results indicate that the activation of NF-κB by sAPP in microglial cells occurs through a cGMP-independent mechanism and could thus involve signalling events evoked by regions of sAPP distinct from the carboxy-terminal sequences that stimulate cGMP-dependent neuromodulation (see below).

Figure 1: Activation of NF-κB in N9 microglial cells by sAPPN9 cultures were treated for 90 min with control buffer or 5 nM sAPP (sAPP), and nuclear extracts were assayed by EMSA with a κB DNA probe.
figure 1

a, The sAPP-treated sample was also assayed after incubation with antibodies (Santa Cruz) generated against NF-κB subunits p50 (sAPP/α-p50) or p65 (sAPP/α-p65). b, sAPP also was tested after a 10-min pretreatment of the cultures with 5 μM KT5823, an inhibitor of the cGMP-dependent kinase (sAPP/K); 200 μM 8-Br-cGMP was tested alone (cGMP). The arrow marks a specific, induced band that was either diminished or supershifted by NF-κB antibodies.

To determine whether activation of NF-κB caused an increase in gene expression, we tested the levels of interleukin-1β (IL-1β) and inducible nitric oxide synthase (iNOS) in the N9 cell line and in primary cultures of microglia. A 24-h treatment of primary microglia with 5 nM sAPP increased immunocytochemical staining for IL-1β and iNOS (Fig. 2a); expression of IL-1β and iNOS was also raised, as shown by western blot analysis of sAPP-treated microglia (Fig. 2b). The amounts of IL-1β and iNOS in the N9 cell line responded to sAPP in a dose-dependent manner, with detectable induction at 100 pM (Fig. 2c). Increased expression of iNOS was apparent within 6 h of sAPP addition (data not shown).

Figure 2: Increase in IL-1β and iNOS in microglial cells by sAPP.
figure 2

a, b, Primary microglia were treated for 24 h with 5 nM sAPP (sAPP) or vehicle (Con) and fixed for immunocytochemistry (a) or collected for immunoblot analysis (b) with antibodies against iNOS or IL-1β. Relative molecular mass (Mr) markers are shown on the left (in thousands). c, N9 microglial cells were treated with 5 nM sAPP for 16 h with the indicated concentrations of sAPP; cultures were harvested for immunoblot analysis with antibodies against iNOS (top; Mr 130K) or IL-1β (bottom). The lane of the far right shows the adjacent sample (10 nM-treated) probed with anti-IL-1β that had been preincubated with purified IL-1β. IL-1β precursor (30K) predominated over mature (mat) IL-1β (17K) in N9 lysates, whereas primary microglia appeared to have a much higher proportion of mature IL-1β. Mr (K) markers are shown on the left.

It has been shown9 that the carboxy-terminal 15 residues that distinguish the products of α- and β-secretase activity (sAPP-α and sAPP-β, respectively) are critical for the cGMP-dependent neuroprotective effects of sAPP-α. We have confirmed these findings in a paradigm of neuronal death induced by 18 h of glucose deprivation of primary hippocampal neurons, in which treatment with 10 nM sAPP-α gave a survival rate of 222.3 ± 19.6% of control (neurons deprived only of glucose) and sAPP-β enhanced survival to only 126.5 ± 12.0% of control. However, such experiments have been done only in cultures lacking significant numbers of microglia. To test for any indirect effects of sAPP on neuronal survival mediated through microglia, we investigated whether there was a relation between specific sAPP sequences and microglial activation. As an index of activation, the culture medium of primary cultures of microglia was assayed for the presence of nitrite, a stable product of nitric oxide formation. Although there was a moderate difference, both sAPP-α and sAPP-β markedly increased nitrite concentrations in these cultures (Fig. 3). The neuroprotective effect of sAPP-α is retained by a construct containing residues 444–612 (βAPP695 numbering)9; however, this construct was unable to activate microglia efficiently (Fig. 3). These differences indicate that different domains of sAPP are involved in the protection of neurons and the activation of microglia, consistent with the lack of involvement of cGMP in activation. We next assayed this indirect neurotoxic effect of sAPP by applying soluble factors to microglia, then removing them before exposing the microglia to primary neurons. Primary microglia were plated onto a permeable membrane suspended in the bottom of a culture-well basket. Cultures were pretreated for 24 h with 5 nM sAPP-α, sAPP-β, or sAPPα444–612, then washed and transferred to 35-mm wells containing primary hippocampal neurons. Viability of all neurons decreased by 18% over the next 48 h, and those exposed to untreated microglia were additionally compromised by a further 34% (Table 1). Toxicity was even greater in the presence of microglia that had been pretreated with sAPP-α or sAPP-β; this correlation of microglial activation with neurotoxicity was extended to the reduced activity of sAPPα444–612 in both assays.

Figure 3: Structural requirements for sAPP stimulation of microglia.
figure 3

Constructs containing the coding sequences of human sAPPα, sAPPβ, and sAPPα444–612 were expressed in E. coli and purified to apparent homogeneity. Primary microglia were treated with 0.1–10 nM of each protein. After 24 h, the culture medium was tested for the presence of nitrite with Griess reagents. Values represent the mean ± s.e.m. for triplicate determinations in a single experiment representative of three performed. Results for sAPPα444–612 were significantly different from those for sAPPα and sAPPβ (P and P , respectively); the difference between sAPPα and sAPPβ was not significant.

Table 1 Indirect neurotoxicity of sAPPs

We have previously shown that the bioactivities of sAPP can be regulated differentially by two forms of human apolipoprotein E (ApoE) that are encoded by gene alleles in disequilibrium with Alzheimer's disease (AD). Specifically, interaction of ApoE3 with sAPP inhibits an activity associated with the amino-terminal 443 residues; ApoE4 was a less potent inhibitor10. Incubation with ApoE3 for 45 min also inhibited the ability of sAPP-α to increase nitrite production in N9 cells (Fig. 4a) and to evoke microglia-mediated neurotoxicity (Fig. 4b); ApoE4 was less effective than ApoE3 in both respects. Co-precipitation experiments revealed that ApoE3 could bind sAPP-α (Fig. 4c) but not sAPPα444–612, suggesting that ApoE3 may bind and mask the sAPP domains responsible for activating microglia. The weaker binding of ApoE4 to sAPP-α (Fig. 4c) could explain why it does not affect sAPP-α activity.

Figure 4: Interaction between ApoE and sAPP.
figure 4

a, Nitrite production in N9 cultures stimulated for 24 h with vehicle, 3 nM ApoE3, ApoE4 or sAPPα that had been incubated alone or with equimolar ApoE3 or ApoE4. b, Indirect neurotoxic effect on microglia treated as in a and assayed as for Table 1. c, ApoE3 (lanes 1, 3, 5) or ApoE4 (lanes 2, 4) was incubated alone (lanes 3, 4) or with equimolar polyhistidine-tagged sAPPα (lanes 1, 2) or sAPPα444–612 (lane 5). Precipitation with Ni2+-affinity resin was followed by immunoblotting with anti-ApoE (Boehringer–Mannheim); a sample of ApoE3 was included on the gel as a positive control (std). d, Mammalian-expressed sAPPα was incubated alone (lane 1) or with equimolar ApoE3 (lanes 2, 3). The mixture was precipitated with protein A (lane 3) or anti-ApoE protein A (lane 2), followed by immunoblotting with anti-APP. Positions of Mr markers kD are indicated on the right; *P versus sAPP.

Retrospective analysis11 and preliminary clinical trials12 have indicated that anti-inflammatory drugs may delay the onset or progression of AD. The crucial inflammatory events could involve microglia, which express markers of activation in AD. These activated microglia are clustered near amyloid plaques, particularly near subtypes of plaques that contain dystrophic neurites expressing high levels of β-APP13. Our data indicate that sAPP can activate microglia, enhancing their neurotoxicity. Considered together with the defective neuroprotection by sAPP-β, our results indicate that events favouring processing of β-APP by β-secretase might reduce the amounts of neuroprotective sAPP-α while retaining a detrimental activity mediated through microglia. The isoform-specific modulation by ApoE of detrimental microglial activation suggests that ApoE4 may allow more frequent neurodegenerative microglial reaction to insults, especially those that increase β-APP expression. For instance, β-APP increases after head injury (where AD or AD-like changes are synergistic with ApoE genotype14,15), in epilepsy (where microglia also are activated16), and in normal ageing17,18,19 (when the interaction between ApoE4 and AD is most pronounced20). Although our results suggest that the influence of ApoE results from its interaction with sAPP, direct immunomodulation by ApoE21 could explain effects on microglia. Amyloid-β (Aβ) can activate microglia22, an effect mediated through receptors for advanced-glycation end-products23. The effects of Aβ on microglia show synergy with other cytokines24, which might include sAPP. This suggests a role for sAPP even if it does not accumulate in AD brain, as indicated by measurements in cerebrospinal fluid25,26. Our results do not exclude a contact-mediated effect on microglia by holo-βAPP. It will be important pharmaceutically to identify the neurotoxin produced here: previous studies in which microglia were activated by amyloid plaques implicated a novel microglia-produced amine acting through glutamate receptors27.


Cells and reagents. Primary microglia were subcultured to 95% purity from neonatal rat mixed glial cultures by differential adherence and panning techniques. Unless otherwise indicated, they were subcultured in minimal essential medium (MEM) supplemented to 10% with fetal bovine serum (FBS). The N9 cell line is a myc-immortalized murine microglial cell line generated by P. Ricciardi-Castagnoli; they were maintained in MEM/10% FBS and switched to serum-free MEM 18–24 h before stimulation. Primary cultures of hippocampal neurons were established from E18 rats as described8. Unless otherwise indicated, sAPP was purified (>98% homogeneity) from the conditioned medium of HEK293 transfectants as described8. The protein produced in this system is generated from βAPP695 and has a C terminus consistent with α-secretase processing28. Bacterially expressed protein for structural comparisons was generated from sequence coding for Val20–Lys612 of human βAPP695 (‘sAPPα’) placed in a pTrcHis (InVitrogen) expression vector. This vector tags expressed proteins N-terminally with a polyhistidine sequence to allow one-step purification on a nickel-affinity column. A second construct (‘sAPPβ’) was generated with a stop codon after Met 596. The third construct (‘sAPPα444–612’) was made from sAPPα by deletion of coding sequences aminoterminal to Asp 444. To exclude the possibility of contamination by bacterial endotoxin, some assays of these bacterially expressed proteins were done in the presence of 10 μg ml−1 polymyxin-B sulphate. Human recombinant ApoEs were obtained from Pan Vera and were not de-lipidated or exposed to reducing agents during purification. Antibodies included anti-iNOS monoclonal (Transduction Laboratories), hamster anti-murine IL-1β monoclonal (Genzyme), and anti-ApoE monoclonal (Chemicon). Co-incubations of sAPP and ApoE were at room temperature for 45 min (polyhistidine-tagged sAPP) or 60 min (HEK293-expressed sAPP). For physiological assay, proteins were co-incubated at 30 nM each; for precipitation, the co-incubation concentrations were 300 nM (polyhistidine-tagged sAPP) or 450 nM (HEK293-expressed sAPP). Precipitation reactions have been described10.

EMSA. Nuclear extracts were prepared according to ref. 29. 5 μg extracted protein from each treatment condition was incubated with a 32P-labelled, κB DNA probe in EMSA buffer (50 mM Tris-HCl, pH 7.4, 20% glycerol, 50 mM NaCl, 5 mM MgCl2, 2.5 mM EDTA, 0.5% Nonidet P-40, 5 mM β-mercaptoethanol, and 250 μg ml−1 poly(dI-dC)). Electrophoresis was done as described8.

Nitrite assay. For determination of nitrite, 100 μl culture medium was removed and mixed with an equal volume of 0.5% sulphanilamide and 0.05% naphthylethyleneamine dihydrochloride in 0.25% phosphoric acid. After 10 min, the colour reaction was measured in a spectrophotometer at 540 nm and the readings calibrated from standards containing known amounts of nitrite. Data are presented as the mean ± s.e.m. for triplicate determinations within one of at least three similar experiments.

Statistics. Data were analysed by ANOVA with Scheffe post hoc, and P-values ≤ 0.05 were assumed to indicate significance.