Deficiency of IKKβ in neurons ameliorates Alzheimer's disease pathology in APP‐ and tau‐transgenic mice

In Alzheimer's disease (AD) brain, inflammatory activation regulates protein levels of amyloid‐β‐peptide (Aβ) and phosphorylated tau (p‐tau), as well as neurodegeneration; however, the regulatory mechanisms remain unclear. We constructed APP‐ and tau‐transgenic AD mice with deletion of IKKβ specifically in neurons, and observed that IKKβ deficiency reduced cerebral Aβ and p‐tau, and modified inflammatory activation in both AD mice. However, neuronal deficiency of IKKβ decreased apoptosis and maintained synaptic proteins (e.g., PSD‐95 and Munc18‐1) in the brain and improved cognitive function only in APP‐transgenic mice, but not in tau‐transgenic mice. Additionally, IKKβ deficiency decreased BACE1 protein and activity in APP‐transgenic mouse brain and cultured SH‐SY5Y cells. IKKβ deficiency increased expression of PP2A catalytic subunit isoform A, an enzyme dephosphorylating cerebral p‐tau, in the brain of tau‐transgenic mice. Interestingly, deficiency of IKKβ in neurons enhanced autophagy as indicated by the increased ratio of LC3B‐II/I in brains of both APP‐ and tau‐transgenic mice. Thus, IKKβ deficiency in neurons ameliorates AD‐associated pathology in APP‐ and tau‐transgenic mice, perhaps by decreasing Aβ production, increasing p‐tau dephosphorylation, and promoting autophagy‐mediated degradation of BACE1 and p‐tau aggregates in the brain. However, IKKβ deficiency differently protects neurons in APP‐ and tau‐transgenic mice. Further studies are needed, particularly in the context of interaction between Aβ and p‐tau, before IKKβ/NF‐κB can be targeted for AD therapies.

Nuclear factor κB (NF-κB), mediating inflammatory responses in various cells, is strongly activated in neurons in the vicinity of Aβ plaques in AD brain. 9,10 NF-κB is sequestered in the cytoplasm in the resting state by binding to its inhibitor, especially, inhibitor of NF-κB (IκB)-α. Once activated, for example, by tumor necrosis factor (TNF)-α and interleukin (IL)-1β, IκBα is phosphorylated by IκB kinase (IKK)-β, ubiquitinated and degraded in the proteasome. Thereafter, NF-κB can enter the nucleus to "turn on" the expression of specific genes. 11 Multiple NF-κB binding sites have been identified in promoters of amyloid precursor protein (APP), β-secretase (BACE1), and a-secretase (ADAM10) genes, [12][13][14][15] suggesting that IKKβ/ NF-κB activation may increase Aβ generation in neurons. However, there are no in vivo studies directly addressing NF-κB in amyloid pathology in AD brains. The role of neuronal IKKβ/NF-κB in the development of tauopathy in AD is unknown. It is also unclear whether neuronal IKKβ/NK-κB affects neurodegeneration in AD, although overexpression of a constitutively active IKKβ in mouse forebrain neurons activates NF-κB and causes neuron loss. 16 In an ischemia stroke mouse model, inhibition of IKKβ/NF-κB signaling prevents neuronal apoptosis and reduces infarct size. 17 It should be noted that NF-κB is constitutively active in glutamatergic neurons of cortex and hippocampus. 18 The activation of IKKβ/NF-κB has the potential to promote neuronal survival, neurite outgrowth, synaptogenesis, and neuronal plasticity. 11 Inhibition of NF-κB in forebrain excitatory neurons by overexpressing dominant-negative IKKβ promotes neuronal apoptosis in closed-head injury mice. 19 Knocking out IKKβ and inhibiting NF-κB in neurons increases the striatal neurodegeneration in the R6/1 mouse model of Huntington's disease. 20 In cultured neurons, pretreatment with Aβ or TNF-α at low concentrations activates NF-κB and protects neurons from the damage caused by Aβ treatments at high concentrations. 9,21,22 Thus, IKKβ/NF-κB activation may serve both toxic and protective effects on neurons, depending on distinct pathophysiological conditions.
Our previous study has shown that IKKβ deficiency in myeloid cells attenuates inflammatory activation and Aβ load in the brain, and improves cognitive function of APP-transgenic mice. 23 In this study, we continued to address the question of whether deletion of IKKβ in neurons also prevents AD pathogenesis in both APP-and tautransgenic mice.

| Animal models and cross-breeding
TgCRND8 APP-transgenic mice (APP tg ) expressing a transgene incorporating both the Indiana mutation (V717F) and the Swedish mutations (K670N/M671L) in the human APP gene under the control of hamster prion protein (PrP) promoter were kindly provided by D. Westaway (University of Toronto). In this mouse strain, the Aβ load does not differ between male and female mice. 24 IKKβ fl/fl mice carrying loxP site-flanked Ikbkb alleles were kindly provided by M. Pasparakis (University of Cologne 25 ). Nex-Cre mice expressing Cre recombinase from the endogenous locus of the Nex gene that encodes a neuronal basic helix-loophelix (bHLH) protein were kindly provided by K. Nave, Max-Planck Institute for Medicine, Göttingen. APP tg , IKKβ fl/fl , and Nex-Cre mice, all on a C57BL6 genetic background, had been cross-bred to build AD animal models with (APP tg IKKβ fl/fl Cre +/− ) and without (APP tg IKKβ fl/fl Cre −/− ) deletion of IKKβ in neurons. In order to investigate physiological function of IKKβ in neurons, we also examined non-APP-transgenic (APP wt ) mice with (APP wt IKKβ fl/fl Cre +/− ) and without (APP wt IKKβ fl/fl Cre −/− ) deletion of neuronal IKKβ. To evaluate the effect of neuronal IKKβ on p-tau-induced phenotype, we cross-bred IKKβ fl/fl and Nex-Cre mice, with P301S tau-transgenic (tau tg ) mice (imported from the Jackson Laboratory, Bar Harbor, MA, USA; Stock number: 008169), which over-express the human tau mutant (P301S) under the direction of mouse prion protein promoter. 3 For this study, APP-transgenic mice and tau-transgenic mice were analyzed by 6 and 9 months of age, respectively, as both AD mice clearly displayed cognitive dysfunction and AD-associated pathologies in the brain at these ages. All animal experiments were performed in accordance with relevant national rules and authorized by the local research ethical committee (permission numbers: 13/2018).

| Morris water maze
The Morris water maze test was used to assess the cognitive function of APP tg or tau tg mice and their APP wt littermates, as previously described. 26 Mice were trained to find the hidden escape platform. There were four trials per training day; with a trial interval of 15 min. Latency time, path length, and velocity were recorded with Ethovision video tracking equipment and software (Noldus Ethovision, Wageningen, The Netherlands). After 6 training days, there were 1 day of rest, and a probe trial on the 8th day. During the probe trial, the platform was removed, and the swimming path was recorded during 5 min. The frequency of entries in the location of original platform were measured.

| Tissue collection for histological and biochemical analysis
Animals were euthanized by over-dose inhalation of isoflurane. The whole brain was collected and divided along the interhemispheric fissure. The left hemisphere was immediately fixed in 4% paraformaldehyde (Sigma-Aldrich GmbH, Taufkirchen, Germany) and embedded in paraffin. The right hemisphere was dissected to remove the cerebellum, brainstem, thalamus, hypothalamus, and olfactory bulb. A 0.5-mm-thick piece of sagittal tissue was cut from the medial side and homogenized in TRIzol (Thermo Fisher Scientific, Darmstadt, Germany) for RNA isolation. The rest of the right hemisphere was snap frozen in liquid nitrogen and stored at −80°C until biochemical analysis was performed.

| Immunohistological analysis
Serial 50-μmthick sagittal sections were cut from the paraffin-embedded hemisphere. Human Aβ in APP tg mouse brains was stained with rabbit anti-human Aβ antibody (1:1000; clone D12B2; Cell Signaling Technology, Frankfurt am Main, Germany), microglia and astrocytes were stained with rabbit anti-ionized calcium-binding adapter molecule (Iba)-1 antibody (1:500; Cat.-No: 019-19 741; Wako Chemicals, Neuss, Germany), and rabbit anti-glial fibrillary acidic protein (GFAP) antibody (1:500; Code-No: Z0334; Agilent Technologies Deutschland GmbH, Waldbronn, Germany) respectively. The VectaStain Elite ABC kit (Vector Laboratories, Burlingame, CA, USA) and diaminobenzidine tetrahydrochloride hydrate (Sigma-Aldrich GmbH) were used for visualization of immunoreactive cells. For each animal, we labeled four serial sections with an interval of 10 layers between each two adjacent sections. In the whole hippocampus and cortex, volumes of Aβ were estimated with the Cavalieri method, and Iba-1 or GFAP-positive cells were counted with Optical Fractionator as described previously 23 on a Zeiss AxioImager.Z2 microscope equipped with a Stereo Investigator system (MBF Bioscience, Williston, VT, USA). Immunohistochemistry with rabbit anti-CD8a antibody (1:100; Cat.-No: HS-361003; Synaptic Systems GmbH, Göttingen, Germany) was performed on four serial brain sections from each tau tg mouse as described above. Because the immunoreactive cells were enriched in the dentate gyrus, we counted all cells only in this brain region and calculated the cell density by dividing with the area of interest.
To evaluate tau pathology in tau tg mice, four serial 50-μmthick sections were chosen as for Aβ analysis. Brain tissues were stained according to our established protocols 26 with a mouse monoclonal antibody against human phospho-tau (Ser202, Thr205) (5 μg/ml; clone: AT8; Thermo Fisher Scientific). Because of low numbers of p-tau-positive cells in the cortex and hippocampus, we did not use stereological analysis, but counted labeled cells in the whole brain region. Data were recorded as the number of labeled cells divided by the full area (in square millimeters) of interest.
To demonstrate the colocalization of Cre and neurons/ astrocytes, we used 30-μmthick sagittal sections that were cut from the dehydrated and cryoembedded left brain hemisphere of 6-month-old APP tg IKKβ fl/fl Cre +/− and APP tg IKKβ fl/fl Cre −/− mice. The guinea pig anti-Cre antibody (1:500; Cat.-No: 257004; Synaptic Systems GmbH) and Alexa Fluor 488-conjugated second antibodies were first incubated with brain sections. After thorough washing, additional antibodies against various cellular markers (mouse anti-NeuN, clone A60, Sigma-Aldrich; or rabbit anti-GFAP, Agilent Technologies Deutschland) were added and were visualized with relevant Alexa Fluor 546-conjugated second antibodies (all second antibodies were from Thermo Fisher Scientific).
To quantify p-tau and total tau (t-tau) proteins, the brain tissue was sequentially homogenized in ice-cold high-salt reassembly buffer (RAB; 0.1 M MES, 1 mM EGTA, 0.5 mM MgSO4, 0.75 M NaCl, 20 mM NaF, and 1 mM PMSF), RIPA buffer and 70% formic acid (FA) as we did in previous studies. 26,27 For the Western blot detection, mouse monoclonal antibodies against p-tau (clone AT8) and t-tau (clone HT7) from Thermo Fisher Scientific were used.

| Quantitative PCR for analysis of gene transcripts
Total RNA was isolated from mouse brains and reverse transcribed. Gene transcripts of pro-and anti-inflammatory markers were quantified with our established protocol 23 using Taqman gene expression assays of mouse Tnfα, Il-1β, Chemokine (C-C motif) ligand 2 (Ccl-2), Il-10, Mannose receptor C type 1 (Mrc1), Chitinase-like 3 (Chi3l3), Glyceraldehyde 3-phosphate dehydrogenase (Gapdh), and 18 s RNA (Thermo Fisher Scientific). Gapdh and 18 s RNA were detected as internal controls. Their threshold cycle (Ct) values in real-time PCRs were significantly correlated (see Figure S1). For the detection of Ikbkb transcript, Taqman gene expression assay (Mm01222249_m1) was used with the amplified PCR product overlapping 6-7 exon boundary of Ikbkb as we did in a previous study. 23 The transcription of following target genes: APP, TAU, Cd200, and Chemokine (C-X3-C motif) ligand 1(Cx3cl1), as well as Ppp2ca and Ppp2cb encoding PP2A catalytic subunit isoform α and β, and Ppp3ca and Ppp3cb encoding PP2B catalytic subunit isoform α and β, was determined using SYBR green binding technique with primers shown in Table 1.

| Construction of knockdown vectors
Two pcDNA6.2-GW/EmGFP-miR vectors (Thermo Fisher Scientific) (kd456 and kd1425) were engineered to contain different select hairpins targeting human IKKβ-encoding gene, IKBKB (Sequence ID: NM_001190720.3) using the protocol we established in the previous study. 28 Sequences of the DNA oligomers are listed in Table 2

| Cell culture and establishment of cell lines
SH-SY5Y neuroblastoma cells were obtained from LGC Standards GmbH (Wesel, Germany) and maintained in DMEM supplemented with 10% fetal calf serum (FCS; PAN Biotech, Aidenbach, Germany). IKKβ-knockdown cell lines were established by transfecting cells with kd-ct, kd456, and kd1425 vectors, and selected with hygromycin until the stable transfection. All genetic modifications were confirmed by Western blot detection of IKKβ proteins using rabbit polyclonal antibody against IKKβ as described above.

| Preparation of membrane components from brain tissues and cultured cells
To measure βand γsecretase activity, membrane components were purified according to the published protocol. 29 Briefly, brain tissues or SH-SY5Y cell pellets were transferred into sucrose buffer (10 mM Tris/HCl, pH 7.4, including 1 mm EDTA and 200 mM sucrose) and homogenized on ice. The homogenate was centrifuged at 1000g for 10 min at 4°C to delete nuclei. The resulting postnuclear supernatant was transferred to a new tube and centrifuged again at 10 000g at 4°C for 10 min. Finally, the resulting supernatant was centrifuged at 187 000g in an Optima MAX Ultracentrifuge (Beckman Coulter GmbH, Krefeld, Germany) for 75 min at 4°C. The resulting supernatant was discarded, and pellets were re-suspended using cannulas of decreasing diameter in sucrose buffer.

| βand γsecretase activity assays
βand γsecretase activities were measured by incubating the crude membrane fraction with secretasespecific FRET substrates according to our established methods. 29 For measurement of βsecretase activity, the crude membrane fraction was resuspended in 500 μl βsecretase assay buffer (0.1 M sodium acetate, pH 4.5). The final concentrations for the βsecretase assay were: 0.1 mg/ml membrane protein (12.5 μg protein per well in 96-well plates), 10% dimethyl sulfoxide, and 8 μM βsecretase substrate IV (Calbiochem, Darmstadt, Germany). For measurement of γsecretase activity, the crude membrane fraction was resuspended in 500 μl γsecretase assay buffer (50 mM Tris/HCl, pH 6.8, 2 mM EDTA). Final concentrations for the γsecretase assay were: 1 mg/ml membrane protein (125 μg protein per well in 96-well plates) and 8 μM γsecretase substrate (Calbiochem). For both secretase assays, kinetics was performed at 37°C and fluorescence intensity in each well was measured for 8 h with intervals of 5 min with Synergy Mx Monochromator-Based Multi-Mode Microplate Reader (BioTek, Winooski, USA). Fluorescence intensity of the first cycle was considered as background and subtracted for each well.
To further analyze the βsecretase, SH-SY5Y neuroblastoma cells were lysed in RIPA buffer and detected with Western blot using rabbit monoclonal antibody against BACE1 (clone D10E5; Cell Signaling Technology) as described above.

| Statistics
Data were presented as mean ± SEM. For multiple comparisons, one-way or two-way ANOVA followed by

Kd456
Top TGCTG TCA TGA AGG TAT CTA AGC GCA GTT TTG GCC ACT GAC  TGA CTG CGCTTATACCTTCATGA   Bottom CCTGT CAT GAA GGT ATA AGC GCA GTC AGT CAG TGG CCA  AAA CTG CGC TTAGATACCTTCATGAC   Kd1425  Top  TGCTG TTT CGG AGG AGA TTC ATC ATG GTT TTG GCC ACT GAC  TGA

| Establishment of APP-transgenic mice with deletion of IKKβ in neurons
To delete IKKβ specifically in neurons, we cross-bred TgCRND8 APP-transgenic mice (APP tg ) to Ikbkb-floxed mice (IKKβ fl/fl ) and Nex-Cre knock-in mice (Nex-Cre +/− ) as we did in previous studies. 23,26 The cellular specificity of Nex-Cre-mediated gene recombination had been described in detail by our and other groups. 26,30 In this study, we co-stained Cre recombinase with various cell markers in brains of APP tg IKKβ fl/fl Cre +/− mice. As shown in Figure 1A,B, Cre recombinase was expressed only in NeuN-positive cells (neurons) but not in GFAP-immunoreactive cells (astrocytes), confirming previous observations. As a control, there were no Creimmunoreactive cells in the brains of APP tg IKKβ fl/fl Cre −/− mice ( Figure 1C). Using real-time RT-PCR, we observed that transcription of IKKβ-encoding gene, Ikbkb, was down-regulated in the brains of APP tg IKKβ fl/fl Cre +/− mice compared to APP tg IKKβ fl/fl Cre −/− littermates ( Figure 1D; t test, p = .002). Quantitative Western blot further showed that the protein level of IKKβ in homogenates of cortex and hippocampus from APP tg IKKβ fl/fl Cre +/− mice was significantly lower than that in APP tg IKKβ fl/fl Cre −/− mice ( Figure 1E,F; IKKβ/Vinculin: 0.218 ± 0.058 vs 1.056 ± 0.259; t test, p = .004). To investigate whether IKKβ deficiency regulated NF-κB activation, we detected IκB in brain homogenates and observed that the protein level of IκB was significantly increased in the brains of 6-month-old APP tg IKKβ fl/fl Cre +/− mice compared with APP tg IKKβ fl/fl Cre −/− controls ( Figure 1G,H; IκB/β-actin: 0.717 ± 0.098 vs 1.271 ± 0.136; t test, p = .004), suggesting that IKKβ deficiency in neurons inhibits NF-κB activation in the brain.

IKKβ reduces cerebral Aβ load in APP-transgenic mice
Aβ is a key molecule leading to neurodegeneration in AD. 31 We used the stereological Cavalieri method to measure Aβ volume, adjusted relative to the volume of analyzed tissues, in 6-month-old APP-transgenic mice. The volume of immunoreactive Aβ deposits in APP tg IKKβ fl/fl Cre +/− mice (1.094% ± 0.121% in the hippocampus and 1.136% ± 0.124% in the cortex) was significantly lower than that in APP tg IKKβ fl/fl Cre −/− mice (1.623% ± 0.109% in the hippocampus and 1.582% ± 0.148% in the cortex; Figure 2A-C; t test, p = .006 and .039 respectively). Western blot analysis using human Aβ-specific antibody was performed to determine levels of Aβ monomers and dimers in the homogenate of cortex and hippocampus derived from 6-month-old APP tg IKKβ fl/fl Cre +/− and APP tg IKKβ fl/fl Cre −/− littermate mice. As shown in Figure 2D-F, deletion of IKKβ in neurons reduced dimeric Aβ by 35% (t test, p = .019). Thus, our study suggested that deletion of IKKβ in neurons reduces Aβ load in the brain of APP-transgenic mice.
In the following experiments, we examined how neuronal IKKβ regulates Aβ level in the brain. Six-month-old APP tg IKKβ fl/fl Cre +/− and APP tg IKKβ fl/fl Cre −/− mice did not differ in either gene transcription or protein expression of APP ( Figure 3A-C; t test, p > .05). Interestingly, we observed that the activity of βsecretase but not γsecretase, was significantly decreased in 6-month-old APP tg IKKβ fl/fl Cre +/− mice compared with APP tg IKKβ fl/fl Cre −/− mice ( Figure 3D,E; two-way ANOVA, p = .046).
To further ask whether IKKβ regulates βand γsecretase activity, we constructed two SH-SY5Y cell lines (kd456 and kd1425) with knock-down of IKBKB gene ( Figure 3F,G; one-way ANOVA followed by posthoc test, p < .05). Compared with the control cells (kd-ct), IKKβ deficiency significantly reduced both the protein expression of BACE1 ( Figure 3H,I; one-way ANOVA followed by post-hoc test, p < .05), and βsecretase activity in kd456 and kd1425 cells ( Figure 3J; two-way ANOVA followed by post-hoc test, p < .05). Deficiency of IKKβ did not change γsecretase activity in kd456 and kd1425 cells compared with kd-ct cells ( Figure 3K; two-way ANOVA, p > .05).

F I G U R E 1
IKKβ is efficiently deleted in neurons of APP tg IKKβ fl/fl Cre +/− mice. Brain sections from 6-month-old APP tg IKKβ fl/fl Cre +/− (IKKβ ko) and APP tg IKKβ fl/fl Cre −/− (IKKβ wt) mice were co-stained for Cre-recombinase in green and NeuN or GFAP in red (A-C). Cre is present in the nuclei of NeuN but not GFAP-positive cells in both the hippocampus (A) and cortex (B) of IKKβ ko mice. As a control, Cre is absent in the brain of IKKβ wt mice (C). Transcripts of Ikbkb gene in the brain of 6-month-old IKKβ ko and wt mice were determined with quantitative RT-PCR (D; t test, n ≥ 4 per group). The protein levels of IKKβ and IκB in brain homogenates from these two groups of AD mice were further detected with quantitative Western blot. IKKβ significantly decreases IKKβ and increases IκB (E-H; t test, n ≥ 8 per group).

| Neuronal deficiency of IKKβ decreases neuroinflammation in APP-transgenic mice
Neuroinflammation is another key mechanism in AD pathogenesis. 4 We counted Iba1-and GFAP-positive cells in the hippocampus of 6-month-old APP tg IKKβ fl/fl Cre +/− and APP tg IKKβ fl/fl Cre −/− littermates using our established protocol. 23 We observed that deletion of IKKβ in neurons significantly reduced the number of both Iba-1 and GFAPpositive cells compared to neuronal IKKβ-wild-type APP tg mice ( Figure 4A-D; t test, p < .05).
In further experiments, we measured inflammatory gene transcripts in the brains of 6-month-old APPtransgenic and non-APP-transgenic mice. IKKβ deficiency significantly decreased the transcription of Ccl2 and Il-10 genes ( Figure 4E,H; t test, p < .05), but did not change the transcription of other inflammatory genes tested (e.g., Tnfα, Il-1β, Mrc1, and Chi3l3; Figure 4F,G,I,J; t test, p > .05). IKKβ deficiency did not alter the transcription of Cd200 and Cx3cl1 genes in the brains compared with neuronal IKKβ-wild-type APP tg mice ( Figure 4K,L; t test, p > .05). CD200 and CX3CL1 are released by neurons and regulate microglial activation. 32

| Neuronal deficiency of IKKβ attenuates cognitive deficits and synaptic impairments in APP-transgenic mice
After we observed that neuronal deficiency of IKKβ reduced Aβ and microglia/astrocytes in the brain of APPtransgenic mice, we asked whether IKKβ deficiency protected neurons. We used the Morris water maze test to F I G U R E 2 Deficiency of IKKβ in neurons reduces cerebral Aβ in APP-transgenic mice. Six-month-old APP tg IKKβ fl/fl Cre +/− (IKKβ ko) and APP tg IKKβ fl/fl Cre −/− (IKKβ wt) mice were analyzed for cerebral Aβ load after immunohistochemical staining of human Aβ (A). The Aβ volume was estimated with Cavalieri method and adjusted by the relevant brain volume. IKKβ deficiency in neurons significantly reduced the cerebral Aβ volume (B and C; t test, n ≥ 7 per group). The cerebral Aβ in APP tg mice was also evaluated by detecting Aβ in the brain homogenate with quantitative Western blot (D). Normalization of Aβ against GAPDH shows reduced amount of dimeric Aβ but not monomeric Aβ after deletion of IKKβ in neurons (E and F; t test, n ≥ 5 per group).

F I G U R E 3
Deficiency of IKKβ reduces βsecretase activity in brains and cultured neuronal cells. Expression levels of APP in brains of 6-month-old APP tg IKKβ fl/fl Cre +/− (IKKβ ko) and APP tg IKKβ fl/fl Cre −/− (IKKβ wt) mice were evaluated by RT-PCR (A; t test, n ≥ 9 per group) and quantitative Western blot (B and C; t test, n = 6 per group). Membrane components were further prepared from IKKβ wt and ko AD mice and incubated with fluorogenic βand γsecretase substrates. IKKβ deficiency in neurons reduces βbut not γsecretase activity in the brains of APP-transgenic mice (D and E; two-way ANOVA, n = 9 and 5 for wt and ko mice respectively). IKKβ-deficient cell lines were established by stably transfecting SH-SY5Y cells with kd456 and kd1425 knock-down vectors. Compared with control cells transfected with kd-ct vector, protein levels of IKKβ and BACE1 are significantly decreased (F-I; one-way ANOVA followed by Bonferroni post-hoc test, n ≥ 5 per group). The following βand γsecretase assays showed that kd456 and kd1425 cells significantly decreases βbut not γsecretase activity compared with kd-ct cells (J and K; two-way ANOVA followed by Bonferroni post-hoc test, n ≥ 4 per group).
examine the cognitive function of 6-month-old APP wt and APP tg littermate mice with and without deletion of IKKβ in neurons. As shown in Figure 5A-C, the swimming time and distance to reach the platform for all tested mice significantly decreased when the training time increased (two-way ANOVA, p < .05); the swimming velocity did not differ between IKKβ-deficient and wild-type APPtransgenic mice or for the same mice on different training dates (two-way ANOVA, p > .05).
Six-month-old APP tg mice with normal IKKβ expression (APP tg IKKβ fl/fl Cre −/− ) traveled longer distances than APP wt mice to reach the escape platform during the acquisition phase ( Figure 5A; two-way ANOVA, p = .029). Neuronal deficiency of IKKβ (APP tg IKKβ fl/fl Cre +/− ) significantly improved APP tg mice in searching for and finding the platform compared to APP tg IKKβ fl/fl Cre −/− littermates ( Figure 5A,B; two-way ANOVA followed by post-hoc test showing the difference in traveling distance and time between APP tg IKKβ fl/fl Cre −/− and APP tg IKKβ fl/fl Cre +/− mice: p < .001). However, in the 5-min probe trial, designed to test the memory of the mice, we did not detect a difference between any two groups of mice in the frequency that the mice visited the region where the escape platform was located ( Figure 5D; one-way ANOVA, p > .05).
We further used Western blot analysis to quantify cleaved caspase-3 and evaluate the apoptosis of brain cells in 6-month-old APP-transgenic mice. As shown in Figure 5E,F, the protein level of cleaved caspase-3 was significantly higher in APP-transgenic mice than in APP-wild-type littermates (one-way ANOVA followed by post-hoc test; p = .037). Neuronal deficiency of IKKβ significantly decreased the protein level of cleaved caspase-3 in APP tg IKKβ fl/fl Cre +/− mice compared with APP tg IKKβ fl/fl Cre −/− controls ( Figure 5F; one-way ANOVA followed by post-hoc test; p = .034).
In our previous study, we observed that protein levels of synaptic proteins, PSD-95 and Munc18-1, in the brain homogenate of 6-month-old TgCRND8 APP-transgenic mice are significantly lower than that in APP-wild-type controls. 23 In this study, we did observe that the protein level of Munc18-1 tended to decrease in APP-transgenic mice compared with APP-wild-type littermates ( Figure 5I

| Neuronal deficiency of IKKβ reduces phosphorylated tau protein in the brain of tau-transgenic mice
APP-transgenic mice cannot model all pathological changes of AD, such as those associated with p-tau. It was reported that p-tau mediates toxic effects of Aβ in AD pathogenesis. 33,34 To investigate the effects of neuronal IKKβ on tau-associated pathologies, we cross-bred IKKβ fl/fl , Nex-Cre +/− , and P301S tau-transgenic (Tau tg ) mice 3 to create neuronal IKKβ-deficient and wild-type Tau tg AD mice. We counted AT8-positive cells in cortex and hippocampus of 9-month-old Tau tg IKKβ fl/fl Cre −/− and Tau tg IKKβ fl/fl Cre +/− mice. The total number of AT8immunoreactive cells adjusted to the investigated area in neuronal IKKβ-deficient Tau tg IKKβ fl/fl Cre +/− mice (2.25 ± .39/mm 2 in cortex, and 3.94 ± 0.65/mm 2 in hippocampus) was significantly fewer than that in IKKβwild-type Tau tg IKKβ fl/fl Cre −/− mice (3.73 ± 0.55/mm 2 in cortex and 8.13 ± 0.88/mm 2 in hippocampus; Figure 6A-C; t tests, p = .044 and .001 respectively).
We also extracted tau proteins from 9-month-old Tau tg IKKβ fl/fl Cre +/− and Tau tg IKKβ fl/fl Cre −/− mice with RAB, RIPA, and FA buffers as we did in previous studies. 26,27 Western blots revealed that ratios of p-tau/t-tau were significantly lower in RAB and RIPA fractions derived from Tau tg IKKβ fl/fl Cre +/− mice, than from Tau tg IKKβ fl/fl Cre −/− mice ( Figure 6D-F; t test, p < .05). In the FA fraction, deletion of neuronal IKKβ tended to decrease the ratios of p-tau/t-tau in Tau tg mice, although the difference was not significant ( Figure 6G; t test, p = .051). Moreover, we observed that the protein levels of p-tau adjusted by βactin in RIPA and FA fractions of Tau tg IKKβ fl/fl Cre +/− mice were also decreased F I G U R E 4 Deficiency of IKKβ in neurons inhibits inflammatory activation in the brains of APP-transgenic mice. Six-month-old APP tg IKKβ fl/fl Cre +/− (IKKβ ko) and APP tg IKKβ fl/fl Cre −/− (IKKβ wt) mice, and 6-month-old non-APP-transgenic mice (APP wt IKKβ fl/fl Cre −/− ; APP wt ) as controls, were analyzed for the neuroinflammatory activation. Microglia and astrocytes were stained with immunohistochemistry using antibodies against Iba1 and GFAP (A and C; in brown color) and counted with the stereological probe, Optical Fractionator. Deficiency of IKKβ significantly reduces Iba1-and GFAP-positive cells in the hippocampus (B and D; t test, n ≥ 4 per group). The transcripts of both pro-and anti-inflammatory genes, as well as Cd200 and Cx3cl1 genes, in the brain of 6-month-old APP tg and APP wt with different expression of IKKβ in neurons, were further detected with real-time PCR. Transcription of Ccl-2 and Il-10 genes, but not Tnfα, Il-1β, Mrc1, Chi3l3, Cd200, and Cx3cl1 genes was reduced by deficiency of IKKβ in APP-transgenic mice (E-L; t test, n ≥ 4 per group).
by IKKβ deficiency compared with Tau tg IKKβ fl/fl Cre −/− mice ( Figure 6D,H,I; t test, p < .05). As βactin was not detectable in RAB fraction, we could not determine the protein levels of p-tau in RAB fractions. Additionally, we found that the protein level of t-tau in RIPA fraction of brain homogenate was significantly higher in Tau tg IKKβ fl/fl Cre −/− mice than in Tau tg IKKβ fl/fl Cre +/− littermates ( Figure 6K,L; t test, p < .05); however, IKKβ deficiency did not change the transcription of TAU gene ( Figure 6J; t test, p > .05), which suggested that deficiency of IKKβ might also increase the degradation of tau protein.

| Neuronal deficiency of IKKβ potentially increases the dephosphorylation of p-tau in tau-transgenic mice
As IKKβ deficiency in neurons attenuated p-tau in the brain of tau-transgenic mice, we hypothesized that IKKβ regulated the phosphorylation of tau protein in AD mice. We quantified phosphorylation levels of GSK3β and p38-MAPK, two important kinases phosphorylating tau proteins in neurons. 26,35 As shown in Figure 7A-D, neuronal IKKβ-deficient and wild-type 9-month-old tau-transgenic mice did not differ in phosphorylation of any of the enzymes (t test, p > .05), suggesting that the decrease in p-tau in IKKβ-deficient tau mice was probably not due to the reduction in p-tau generation. Interestingly, we observed that IKKβ deficiency significantly up-regulated the transcription of Ppp2ca, but not Ppp2cb, Ppp3ca, and Ppp3cb in the brain of tau-transgenic mice ( Figure 7E-H; t test, p < .05), which was in line with a previous observation that NF-κB activation inhibits expression of catalytic subunit of PP2A (PP2Ac). 36 Western blot experiments verified the finding by showing that the cerebral protein level of PP2Ac was significantly higher in 9-month-old Tau tg IKKβ fl/fl Cre +/− mice than in Tau tg IKKβ fl/fl Cre −/− littermates ( Figure 7I,J; t test, p < .05). Thus, neuronal deficiency of IKKβ potentially increases the dephosphorylation of p-tau in tau-transgenic mice.

| Neuronal deficiency of IKKβ promotes autophagy in the brains of both APP-and tau-transgenic mice
We have recently observed that activated autophagy is a mechanism mediating the degradation of BACE1 in APPtransgenic mice 37 and p-tau in tau-transgenic mice. 27 Interestingly, we observed that neuronal deficiency of IKKβ significantly increased the ratios of LC3B-II/I in RIPA-soluble brain homogenates of both 6-month-old APP-transgenic mice and 9-month-old tau-transgenic mice, indicating that IKKβ deficiency enhanced autophagy in the brains of AD mice ( Figure 8A-D; one-way ANOVA followed by post-hoc test for APP mice and t test for tau mice; p < .05 for both mice), which was in accordance with a previous finding that inhibition of NF-κB in neurons promotes autophagy in the brains of TDP-43transgenic mice. 38 Compared with non-APP-transgenic mice, APP-transgenic mice inhibited autophagy in the brain ( Figure 8A,B; one-way ANOVA followed by post-hoc test; p < .05). However, the protein levels of Beclin1 and SQSTM1/p62 did not differ between Tau tg IKKβ fl/fl Cre +/− and Tau tg IKKβ fl/fl Cre −/− mice ( Figure 8E,F; t test, p > .05).

| Neuronal deficiency of IKKβ regulates neuroinflammation in tau-transgenic mice
As we did for APP-transgenic mice, we counted Iba1 and GFAP-immunoreactive cells in the brains of 9-month-old Tau tg mice. Similarly, we observed that deficiency of neuronal IKKβ significantly decreased the numbers of microglia and astrocytes in the hippocampus and cortex of Tau tg IKKβ fl/fl Cre +/− mice compared with Tau tg IKKβ fl/fl Cre −/− littermates ( Figure 9A-E; t test, p < .05).
In following experiments, we quantified transcripts of inflammatory genes in 9-month-old Tau tg and Tau wt littermate mice. As shown in Figure 9F,J, the transcription of pro-inflammatory Tnfα gene was significantly downregulated, while the transcription of anti-inflammatory Mrc1 gene was up-regulated in Tau tg IKKβ fl/fl Cre +/− mice F I G U R E 5 Deficiency of IKKβ in neurons improves cognitive function and attenuates synaptic impairments in APP-transgenic mice. Six-month-old APP-transgenic (APP tg ) and non-transgenic (APP wt ) mice with (ko) and without (wt) deletion of neuronal IKKβ were examined for cognitive function with Morris water maze test. During the training phase, APP tg IKKβ-wt mice reached the escape platform with significantly longer traveling distance than APP wt IKKβ-wt littermate mice (A; two-way ANOVA followed by Bonferroni post-hoc test, n is shown in the figure). Deletion of IKKβ in neurons (APP tg IKKβ-ko) significantly reduced the traveling time and distance to the escape platform compared with APP tg IKKβ-wt mice (A, B; two-way ANOVA followed by Bonferroni post-hoc test). IKKβ deficiency affected the swimming speed neither of APP tg mice, nor for each mouse at different training time points (C; two-way ANOVA, p > .05). However, APP tg IKKβ-wt mice swam much faster than non-APP transgenic (APP wt IKKβ-wt) mice with unknown reasons (C; two-way ANOVA followed by Bonferroni post-hoc test). In the probe trial, APP tg and APP wt mice with and without IKKβ deficiency did not differ in the frequency, with which the mice visited the region where the platform was previously located (D; one-way ANOVA, p > .05, n ≥ 6 per group). Western blotting was used to detect cleaved caspase-3, and the amount of synaptic structure proteins, Munc18-1, SNAP25, synaptophysin, and PSD-95 in the brain homogenate of 6-month-old APP tg and APP wt mice (E-K). Transgenic expression of APP increases cleaved caspase-3, and neuronal deficiency of IKKβ recovers it (F; one-way ANOVA, n ≥ 7 per group). Deficiency of IKKβ in neurons was associated with a higher level of PSD-95 and Munc18-1 in APP-transgenic mice (H and I; one-way ANOVA followed by Bonferroni post-hoc test, n ≥ 5 per group). compared with Tau tg IKKβ fl/fl Cre −/− littermates (t test, p < .05). However, neuronal deficiency of IKKβ did not change the transcription of Tnfα and Mrc1 gens in Tau wt mice ( Figure 9F,J; t test, p > .05). The transcription of other tested genes (e.g., Il-1β, Ccl-2, Il-10, and Chi3l3) was not changed by IKKβ deficiency in neurons in both Tau tg and Tau wt mice ( Figure 9G-I,K; t test, p > .05). It is known that active and healthy neurons inhibit microglial activation through releasing CD200 and CX3CL1. 32 Interestingly, we observed that the transcriptional level of Cd200 but not Cx3cl1 was significantly higher in Tau tg IKKβ fl/fl Cre +/− mice than in Tau tg IKKβ fl/fl Cre −/− littermate mice (Figure 9L,M; t test, p < .05).
CD8-positive lymphocytes have been shown to exacerbate AD pathology in tau-transgenic mice. 39 We continued to stain CD8-positive cells in brains of 9-month-old Tau tg IKKβ fl/fl Cre +/− and Tau tg IKKβ fl/fl Cre −/− littermate mice. CD8-positive cells were mainly located in the dentate F I G U R E 6 Deficiency of IKKβ in neurons reduces cerebral p-tau in tau-transgenic mice. Nine-month-old Tau tg IKKβ fl/fl Cre +/− (IKKβ ko) and Tau tg IKKβ fl/fl Cre −/− (IKKβ wt) mice were analyzed for cerebral p-tau load after immunofluorescent labeling with AT8 antibody (A). The p-tau-positive cells were counted and adjusted by the relevant brain area. IKKβ deficiency in neurons significantly reduces the cerebral p-tau-positive cells (B and C; t test, n ≥ 10 per group). Tau proteins were extracted from 9-month-old tau-transgenic mice with RAB, RIPA, and FA buffers and detected with Western blots for both phosphorylated and total tau (p-tau and t-tau, respectively) (D-I). The ratios of p−/t-tau are significantly lower in RAB and RIPA fractions of IKKβ-ko mice than in IKKβ-wt littermates (E and F; t test, n ≥ 5 per group). When adjusted to βactin in the same sample, the protein levels of p-tau in RIPA and FA fractions were also significantly reduced by deficiency of IKKβ, compared with IKKβ-wt tau-transgenic mice (H and I; t test, n ≥ 5 per group). In additional experiments, expression levels of tau were determined by quantitative RT-PCR assay of TAU gene transcripts and quantitative Western blot of total tau protein in RIPA-soluble brain homogenate fractions. Neuronal deficiency of IKKβ does not change TAU transcription (J; t test, n ≥ 8 per group), but significantly decreases protein level of t-tau (K and L; t test, n = 7 per group).
gyrus, especially in the perivascular space ( Figure S2B). After counting cells, we found that neuronal deficiency of IKKβ tended to decrease the density of CD8-positive lymphocytes in the dentate gyrus compare to IKKβ-wildtype tau-transgenic mice; however, the reduction was not statistically significant ( Figure S2A-C).

F I G U R E 7 Deficiency of IKKβ in neurons increases
PP2Ac expression in the brain of tau-transgenic mice. Brains from 9-month-old Tau tg IKKβ fl/fl Cre +/− (IKKβ ko) and Tau tg IKKβ fl/fl Cre −/− (IKKβ wt) mice were homogenized in RIPA lysis buffer. Phosphorylated (p-) and total (t-) GSK3β and p38-MAPK were detected with Western blot (A and C). The ratios of p−/t-GSK3β and p−/t-p38-MAPK are not different between IKKβ-wt and ko tau-transgenic mice (B and D; t test, p > .05, n = 7 per group). Transcripts of Ppp2ca, Ppp2cb, Ppp3ca, and Ppp2cb genes were also measured in the brain of tau-transgenic mice with quantitative PCR (E-H), showing that IKKβ deficiency significantly upregulates the transcription of Ppp2ca gene, but not other genes tested (E-H; t test, n = 4 per group). Quantitative Western blot of PP2Ac in RIPA fraction of brain homogenate verified that IKKβ deficiency in neurons increases the cerebral protein level of PP2Ac in tau-transgenic mice (I and J; t test, n ≥ 9 per group).

| Neuronal deficiency of IKKβ does not improve cognitive function and neuroprotection in tau-transgenic mice
We have observed that the cognitive function of tautransgenic mice is impaired in the Morris water maze test, 27 so we used water maze to evaluate whether neuronal deficiency of IKKβ improves the cognitive function of 9-month-old tau mice. As shown in Figure 10A,B, the swimming time and distance to reach the platform for all tested mice significantly decreased when the training time increased (two-way ANOVA testing the effect of training time, p < .05). However, Tau tg IKKβ fl/fl Cre +/− and Tau tg IKKβ fl/fl Cre −/− littermate mice differed in neither the traveling time nor distance to reach the escaping platform in the training phase (two-way ANOVA testing effect of genotypes, p > .05). Similarly, in the probe trial, IKKβ deficiency did not change the frequency with which Tau tg IKKβ fl/fl Cre +/− mice visited the initial region for platform during the training phase compared to Tau tg IKKβ fl/fl Cre −/− littermates ( Figure 10C; t test, p > .05). F I G U R E 8 Deficiency of IKKβ in neurons increases autophagic activity in brains of both APP-and tau-transgenic mice. Brain homogenates were prepared from 6-month-old APP-transgenic mice and 9-month-old tau-transgenic mice with (IKKβ ko) and without (IKKβ wt) deletion of IKKβ in neurons. Six-month-old non-APP-transgenic (APPwt) mice were used as controls. Quantitative Western blot was used to detect LC3B, beclin1, and SQSTM1/p62 (A and C). IKKβ deficiency significantly increases ratios of LC3B-II/I, but not protein levels of p62 and Beclin1 in both AD mouse models (B, D-F; one-way ANOVA followed by Bonferroni post-hoc test for APP mice, n ≥ 6 per group; and t test for tau mice, n = 8 per group).

F I G U R E 9
Deficiency of IKKβ in neurons inhibits inflammatory activation in the brain of tau-transgenic mice. Brain sections from 9-month-old Tau tg IKKβ fl/fl Cre +/− (IKKβ ko) and Tau tg IKKβ fl/fl Cre −/− (IKKβ wt) mice were stained with immunohistochemistry for Iba1 and GFAP (A and D, in brown color). Deficiency of IKKβ significantly reduces the number of Iba1-positive cells in both the hippocampus and cortex (B and C; t test, n ≥ 8 per group) and GFAP-positive cells in the hippocampus (E; t test, n = 4 per group). The transcripts of both pro-and anti-inflammatory genes, as well as Cd200 and Cx3cl1 genes, in brains of tau-transgenic (tg) and non-transgenic (wt) mice with (ko) and without (wt) deficiency of neuronal IKKβ were further detected with real-time PCR (F-M). Transcription of Tnfα gene is significantly down-regulated, while the transcription of Mrc1 gene is up-regulated in Tau tg IKKβ fl/fl Cre +/− mice compared with Tau tg IKKβ fl/fl Cre −/− mice (F and J; one-way ANOVA followed by Bonferroni post-hoc test, n ≥ 6 per group). Moreover, transcription of Cd200 is also increased in tau tg mice by neuronal deficiency of IKKβ (L; t test, n ≥ 6 per group).
Moreover, we observed that neuronal deficiency of IKKβ altered neither the protein levels of Munc18-1, PSD-95, SNAP25, and synaptophysin, nor the phosphorylation level of cAMP response element-binding protein (CREB) as detected with quantitative Western blot ( Figure 10D-K; t test, p > .05). CREB activation promotes neuronal survival and plasticity. 40 Surprisingly, we observed that neuronal deficiency of IKKβ increased the protein level of cleaved caspase-3 in the brain homogenate of 9-month-old tautransgenic mice ( Figure 10L,M; t test, p < .05), suggesting that IKKβ deficiency in neurons promotes apoptosis in the brain of tau mice.

| DISCUSSION
Neuroinflammation is a hallmark of AD pathology. Uncontrolled inflammatory activation exacerbates amyloid pathology and tauopathy, leading to neurodegeneration in the AD brain. However, the molecular mechanisms by which neurons respond to neuroinflammation are still unclear. In this project, we knocked out IKKβ-encoding gene, Ikbkb, specifically in neurons of both APP-and tautransgenic mice and observed that neuronal deficiency of IKKβ attenuates Aβ and p-tau load, and modifies inflammatory activation in the brain. Deficiency of IKKβ improves cognitive function and prevents neurodegeneration in APP-transgenic mice; however, it does not confer an efficient neuroprotection in tau-transgenic mice.
The effect of neuroinflammation on Aβ production has been extensively studied. Activation of BACE1 coincides with focal glial inflammatory activation in APP-transgenic mice. 41 In APP-transgenic mice or the systemically lipopolysaccharide-administered mice, pharmacological treatments with acetyl-11-keto-βboswellic acid, 42 bee venom, 43 and carbon monoxide 44 simultaneously suppress BACE1 expression and NF-κB activation in the brain, indicating the correlation between BACE1 and NF-κB. We deleted IKKβ specifically in neurons, which inhibited NF-κB activation in the brain of APP-transgenic mice. Using this approach, we directly demonstrated the role of IKKβ/NF-κB in BACE1 activation and Aβ production in the brain. Our SH-SY5Y cell culture experiments showed that deficiency of IKKβ decreases protein levels of BACE1, verifying our in vivo finding.
Aβ and inflammatory activation induce tau phosphorylation, 5,45 and trigger tau spreading along axonal projections in AD brains. 46 Our experiments showed that deletion of IKKβ in neurons attenuates both phosphorylated and total tau in brains of tau-transgenic mice. Because neuronal deficiency of IKKβ did not alter transcription of TAU gene, or activation of p38α-MAPK and GSK3β, two key kinases that phosphorylate tau in AD brain, 26,35 we hypothesized that the reduction in tau proteins might be due to increased dephosphorylation of tau and/or degradation of tau. PP2A and -2B are the enzymes dephosphorylating p-tau. 47 NF-κB activation inhibits the expression of the catalytic subunit of PP2A (PP2Ac) in pancreatic cancer cells. 36 We observed that IKKβ deficiency in neurons upregulates both the transcription of Ppp2ca gene, which encodes isoform A of PP2Ac, and the protein level of PP2Ac in the brain of tau-transgenic mice. Thus, the reduction of p-tau in IKKβ-deficient AD mice may result from the increased dephosphorylation of p-tau.
It was interesting to find that deficiency of IKKβ in neurons enhances autophagy in the brains of our APPand tau-transgenic mice. IKKβ/NF-κB signaling mediates the expression of autophagy-associated proteins, for example, Beclin1, LC3B, and Atg5, and is thought to promote autophagy 20,48,49 ; however, IKKβ/NF-κB activation also induces the expression of Bcl-2 and Bcl-xL, 50 which bind to Beclin1 and block autophagy. 49 In TDP-43transgenic mice, neuron-specific expression of a superrepressor form of IκB (IκB S32A, S36A ) enhances autophagy, decreases TDP-43 accumulation, and improves motor performance. 38 Thus, enhancing autophagy in IKKβ-deficient neurons could promote degradation of BACE1 and tau in neurons, as we observed in previous studies. 26,27,37 Deficiency of IKKβ in neurons reduces microglia and astrocytes in APP-and tau-transgenic mice; however, how it regulates inflammatory activation in AD is unclear. Not surprisingly, deficiency of IKKβ decreases Aβ and subsequently inhibits both pro-and anti-inflammatory activation (e.g., down-regulation of Ccl-2 and Il-10 transcription) in APP-transgenic mice. As we previously observed, Aβ is a ligand of CD14 and TLR2, which induces F I G U R E 1 0 Deficiency of IKKβ in neurons does not affect cognitive function and even increases apoptosis in the brains of tau-transgenic mice. In the water maze test, 9-month-old Tau tg IKKβ fl/fl Cre +/− (IKKβ ko) and Tau tg IKKβ fl/fl Cre −/− (IKKβ wt) littermate mice did not differ in traveling latency and distance to reach the escape platform during the training phase (A and B; two-way ANOVA, p > .05, n ≥ 6 per group), nor in the frequency with which the mice visited the region where the platform was previously located during the probe trial (C; t test, p > .05, n ≥ 6 per group). Western blotting was used to detect cleaved caspase-3, phosphorylation level of CREB, and the amount of synaptic structure proteins, Munc18-1, SNAP25, synaptophysin, and PSD-95 in the brain homogenate of tau-transgenic mice (D-M). Deficiency of IKKβ in neurons does not alter the protein levels of various synaptic proteins (E-H; t test, p > .05, n ≥ 4 per group), nor the ratio of phosphorylated (p-) to total (t-) CREB (J and K; t test, p > .05, n ≥ 6 per group). Surprisingly, IKKβ deficiency significantly increases the protein level of cleaved caspase-3 in the brains of 9-month-old tau-transgenic mice (M; t test, n ≥ 9 per group).
both types of inflammatory activation in microglia. 51,52 Tauopathy can primarily damage neurons and then triggers microglial activation, 53 although tau proteins also activate microglia. 54 In our tau-transgenic mice, IKKβ deficiency decreases pro-inflammatory Tnfα transcription while increasing anti-inflammatory Mrc1 transcription, in which IKKβ-regulated neuronal status may be one of the determinants. CD200 and CX3CL1 released by healthy and active neurons control microglial activation. 32 IKKβ deficiency increases transcription of Cd200 gene in the brains of tau mice, but not APP-transgenic mice, perhaps providing further evidence that neuronal activity regulates neuroinflammation in tau mice.
We observed that IKKβ deficiency protects neurons in APP-but not in tau-transgenic mice, which may also be due to the different pathogenic mechanisms of Aβ and p-tau in AD. In APP-transgenic mice, Aβ oligomers activate N-methyl-D-aspartate receptor (NMDARs) directly, 55 and indirectly by blocking astrocytic re-uptake of glutamate and accumulating glutamate in the perisynaptic space, 56 both of which lead to calcium overload of neurons. Elevated calcium activates NF-κB possibly by interaction between CaMKII and IKKβ. 57,58 The rapid, particularly sustained, accumulation of p65/NF-κB and IκBα in the nucleus is associated with neuron death. 59 IKKβ deficiency may prevent calcium-induced NF-κB activation and neurotoxicity in APP-transgenic mice. In tau-transgenic mice, p-tau destabilizes the cytoskeleton and disrupts the axonal transport, leading to axonal degeneration and neuronal death. 60 Tau accumulation damages mitochondria, induces endoplasmic reticulum stress and dysregulates neuronal GABAergic and cholinergic signaling. 53 IKKβ deficiency may not prevent these pathogenic processes; however, compromise NF-κB activation-mediated neuroprotection in tau-transgenic mice.
Our study has shown that neuronal deficiency of IKKβ serves diverse effects on neuroprotection in APP and tautransgenic AD mice. It is obviously a limitation of our study that we studied pathogenic role of neuronal IKKβ in APP-or tau-transgenic mice, instead of in an AD model with both Aβ pathology and tauopathy. It is known that Aβ and p-tau synergistically contribute to AD pathogenesis. Tau appears to mediate the neurotoxic effects of Aβ. Deletion of endogenous tau abolishes Aβ-induced neurotoxicity. 33,34,61 Thus, further studies are needed to clarify the pathogenic role of neuronal IKKβ, particularly in the context of interaction between Aβ and p-tau.
In summary, deficiency of IKKβ in neurons reduces Aβ and p-tau pathologies in APP-and tau-transgenic mice. Possible mechanisms are that IKKβ deficiency: (1) decreases BACE1 expression and Aβ generation, (2) increases PP2Ac expression and p-tau dephosphorylation, and (3) enhances neuronal autophagy, which promotes BACE1 and p-tau degradation. However, IKKβ has also neuroprotective effects. Although our study deciphered the pathophysiological role of neuronal IKKβ/NF-κB in AD, it remains unclear whether IKKβ can serve as a therapeutic target for AD patients. A following study is needed to address the pathogenic role of neuronal IKKβ in an AD model with both amyloid pathology and tauopathy.

AUTHOR CONTRIBUTIONS
Yang Liu conceptualized and designed the study, acquired funding, conducted experiments, acquired and analyzed the data, and wrote the manuscript. Laura Schnöder, Wenqiang Quan, Ye Yu, Inge Tomic, Qinghua Luo, and Wenlin Hao designed and conducted the experiments, acquired data, and analyzed the data. Guoping Peng and Dong Li provided advice and edited the manuscript. Klaus Fassbender provided laboratory equipment and supervised the study. All authors read and approved the final manuscript. through HOMFOR2022 (to LS). Qinghua Luo holds a scholarship from China Scholarship Council (CSC; 201906820011). Open Access funding enabled and organized by Projekt DEAL.

DISCLOSURES
The authors have declared that they have no competing interests.

DATA AVAILABILITY STATEMENT
All data generated or analyzed during this study are included in this published article. Raw data are available upon reasonable request.

ETHICS APPROVAL
All animal experiments were performed in accordance with relevant national rules and authorized by Landesamt für Verbraucherschutz, Saarland, Germany (permission number: 13/2018).