The sterol regulatory element‐binding protein 2 is dysregulated by tau alterations in Alzheimer disease

Abstract

Disturbed neuronal cholesterol homeostasis has been observed in Alzheimer disease (AD) and contributes to the pathogenesis of AD. As the master switch of cholesterol biosynthesis, the sterol regulatory element‐binding protein 2 (SREBP‐2) translocates to the nucleus after cleavage/activation, but its expression and activation have not been studied in AD which is the focus of the current study. We found both a significant decrease in the nuclear translocation of N‐terminal SREBP‐2 accompanied by a significant accumulation of C‐terminal SREBP‐2 in NFT‐containing pyramidal neurons in AD. N‐terminal‐ SREBP‐2 is also found in dystrophic neurites around plaques in AD brain. Western blot confirmed a significantly reduced nuclear translocation of mature SREBP‐2 (mSREBP‐2) in AD brain. Interestingly, reduced nuclear mSREBP‐2 was only found in animal models of tauopathies such as 3XTg AD mice and P301L Tau Tg mice but not in CRND8 APP transgenic mice, suggesting that tau alterations likely are involved in the changes of mSREBP‐2 distribution and activation in AD. Altogether, our study demonstrated disturbed SREBP‐2 signaling in AD and related models, and proved for the first time that tau alterations contribute to disturbed cholesterol homeostasis in AD likely through modulation of nuclear mSREBP‐2 translocation.

Abbreviations

AD

Alzheimer disease

Aβ

Amyloid beta

ER

endoplasmic reticulum

GFAP

Glial fibrillary acidic protein

NFTs

neurofibrillary tangles

PHF

paired helical filament

SDS‐PAGE

sodium dodecyl sulfate polyacrylamide gel electrophoresis

SREBP

sterol regulatory element‐binding protein

SRE

sterol response element

Introduction

Alzheimer disease (AD) is the most common neurodegenerative disease in the elderly. It is characterized by intraneuronal neurofibrillary tangles (NFTs) containing highly phosphorylated tau and senile plaques composed of Aβ filaments, along with selective neuronal death in affected brain regions. Although the etiology of AD is unknown, it is believed that both environmental and genetic factors are involved among which disturbances in cholesterol metabolism have been implicated 12, 31, 33. Longitudinal studies revealed the positive correlation between higher blood concentration of total and low‐density lipoprotein cholesterol and faster global cognitive decline in late‐life 35. Increased amounts of neuronal cholesterol have been suggested to contribute to inducing and/or aggravating AD 17 and accumulation of sterols has been described in affected brain areas from AD patients and AD mouse models 1, 13, 20, 40, 57, and region‐specific loss of synapses is also related to changes in cholesterol levels 42. Recent large multi‐omics studies suggest that cholesterol homeostasis maybe an integral part in AD pathogenesis 37. Apolipoprotein E (ApoE) is the principal cholesterol‐carrier in the brain and its E4 allele is a major genetic risk factor for AD 26, suggesting the potential involvement of cholesterol transport defect in AD. Hypercholesterolemia is correlated to an increased risk of AD although some studies failed to find such a correlation 50. Retrospective studies suggested beneficial effects of cholesterol‐lowering drugs (ie, statins) in the prevention and the treatment of AD, but clinical trials produced inconsistent results 50. Nevertheless, animal studies have consistently shown that hypercholesterolemia leads to dysfunction of the cholinergic system, cognitive deficits, amyloid‐β (Aβ) and tau pathology, all characteristics of AD 24, 46, 52, which lend strong support for a role of cholesterol disturbance in AD. Indeed, cholesterol is concentrated in lipid rafts where amyloidogenic pathway and fibrillogenesis of Aβ takes place 39, 53. Elevated cholesterol causes increased Aβ production, secretion and fibrillization and facilitates Aβ toxicity 4, 21, 39. Cholesterol deregulation is also a key pathogenic event affecting tau homeostasis 11, 32, 47. NFTs purified from the brains of patients with AD were found to contain cholesterol 22, and tangle‐bearing neurons contain more free cholesterol than adjacent tangle‐free neurons 15, 16.

The CNS contains as much as 23% of total body cholesterol 14. Almost all brain cholesterol is locally synthesized since the blood–brain barrier is nearly impermeable to plasma lipoprotein‐associated cholesterol 6. The membrane‐bound sterol regulatory element‐binding protein 2 (SREBP‐2) is the primary transcription factor for the regulation of cholesterol 8. Low levels of SREBP‐2 are maintained in the endoplasmic reticulum (ER) membrane 5. When cholesterol de novo synthesis is required, SREBP‐2 moves to the Golgi apparatus where it undergoes a regulated processing to release the mature N‐terminal soluble SREBP‐2 (mSREBP‐2) containing the transcriptional activation domain 5. The mSREBP‐2 interacts with importin and is translocated to the nucleus where it binds to sterol response elements (SREs) in the promoter and upregulates the transcription of the target genes mainly involved in cholesterol synthesis including SREBP‐2, LDL receptors and HMG CoA reductase 5, 23.

It was reported that SREBP‐2 has a direct role on BACE1 expression 36. SREBP‐2 overexpression in APP/PS1 mice exhibited exacerbated amyloid pathology and neuronal death with cognitive decline along with hyperphosphorylated tau and NFT formation 3 while SREBP‐2 inhibition reduces amyloid burden in APP/PS1 mice 48. APP expression/processing and Aβ could also impact cholesterol homeostasis 2, 38, 44, 54: APP interacts with SREBP‐1 and inhibits its cleavage/activation 44 and the α‐ and β‐cleaved ectodomains of APP may have opposing effects on SREBP‐2 54. Oligomeric Aβ impairs the cleavage SREBP‐2 and inhibits cholesterol biosynthesis in vitro 32. However, this was not supported in vivo where increased cholesterol synthesis and mitochondrial cholesterol influx along with increased mSREBP‐2 was found in APP/PS1 mice 2. Most recently, one SREBP‐2 polymorphism rs2269657 showed significant associations with LOAD pathological biomarkers 43. To better understand the abnormal cholesterol metabolism in AD, it is important to understand whether and how the expression and activation of SREBP‐2 is changed in AD, which is the main focus of this study. Two specific SREBP‐2 antibodies, either recognizing the N‐terminus or the C‐terminus, were used. We found a significant decrease in the nuclear translocation of N‐terminal SREBP‐2 accompanied by a significant accumulation of C‐terminal SREBP‐2 in NFT‐containing pyramidal neurons in the AD. Reduced nuclear N‐terminal SREBP‐2 was also found in 3XTg AD mice and P301L Tau mice with tau pathology but not in CRND8 APP mice. This study finds that disturbed SREBP‐2 signaling likely contributed to the disrupted cholesterol homeostasis found in AD, and suggests for the first time that tau alterations could disrupt the SREBP‐2 signaling.

Methods and Materials

Immunohistochemistry

Samples of hippocampus were obtained at autopsy under approved IRB protocols from the Brain Bank at Case Western Reserve University and fixed in buffered formalin or methacarn, embedded in paraffin and serial sections were cut. Histopathologically diagnosed cases of AD according to the NINCDS‐ADRDA criteria (n = 27, ages 64–90 years) and age‐matched (n = 15, age 60–86 years) and younger control cases (n = 6, age 17–37 years) were used. Paraffin‐embedded tissue sections were deparaffinized in xylene and rehydrated to Tris‐buffered saline (TBS: 50 mM Tris, 150 mM NaCl, pH = 7.6). After blocking in 10% normal goat serum in TBS for 30 minutes, primary antibodies were applied and incubated for 16 h at 4°C as we previously described 29.

Antibodies used included rabbit anti‐SREBP‐2 N‐terminal antibody (Cayman Chemical), rabbit anti‐SREBP‐2 C‐terminal antibody (Santa Cruz), Rabbit antibody R458 against reticulon, mouse anti‐tau antibodies Alz‐50, PHF‐1 recognizing tau phosphorylated at Ser396/404 (both gift of Peter Davies, Albert Einstein College of Medicine, Bronx NY, USA), AT8 recognizing tau phosphorylated at Ser 202/Thr205 (Pierce) and 4G8 antibody against amyloid‐β (Biolegend). The peroxidase‐anti‐peroxidase method was used and developed using DAB (Dako). Double staining was used to compare SREBP‐2 localization (DAB stain) with Aβ containing plaques using 4G8 (alkaline phosphatase method with Fast Blue as chromagen). An adsorption experiment was performed by incubating diluted SREBP‐2 antibody with 5 micrograms of peptide for 16 h before immunostaining.

To compare the SREBP‐2 C‐terminal antibody with different tau antibodies, serial adjacent sections from three cases of AD were stained and three adjacent fields of the CA1 region were imaged. Following tissue landmark vessels, densitometric quantification was performed with the N‐terminal antibody, comparing the staining intensity in neurons with or without tau‐positive NFT (Zeiss Axiovision).

Various disease models were also examined: cortical sections from Down’s syndrome (n = 2, ages 54 and 61 years, CWRU Brain Bank) and a case with the Swedish double APP mutation (gift of Raj Kalaria) were examined. CRND8 mice (transgenic for human APP695 Swedish KM670/671NL and Indiana mutation V717F) 58 were obtained from University of Toronto and a breeding colony has been maintained at CWRU. 3XTg mice (APP Swedish, P301L, PSEN1 M146V) 27 were obtained from Jaxon Laboratory and a breeding colony has been maintained at CWRU. All animals were group housed and provided ad libitum access to food and water and maintained on a 12 h light/dark cycle. Brain tissues were harvested after sacrifice at the different ages following protocols approved by the Institutional Animal Care and Use Committee (IACUC) of CWRU. Brain sections from the 3XTg mice and non‐transgenic WT of various ages with equal males and females per age group, male 9 and 12 month WT and CRND8 mice, 8 month non‐transgenic and P301L transgenic mice overexpressing human tau protein (gift of Dr. Shu‐Hui Yen of Mayo Clinic) 59, and a mouse model overexpressing reticulon protein (RTN3) (gift of Dr. Riqiang Yan of Cleveland Clinic), which develops dystrophic neurites 28 were also examined.

Nuclear fraction and western blotting

Samples of frozen cortical gray matter of AD (n = 6; ages 65–90, mean 78.3; 4 male and 2 female) and age/gender‐matched control cases (n = 6; ages 63–85, mean 76.8; 4 male and 2 female) sourced from the Brain Bank at Case Western Reserve University and from the NIH NeuroBioBank were homogenized and lysed with RIPA Buffer (Cell Signaling) plus 1 mM phenylmethylsulfonyl fluoride (Sigma) and Protease Inhibitor Cocktail (Sigma) and centrifuged for 10 minutes at 16 000 × g at 4°C. Highly purified nuclear fractions were prepared following a previously published method 7. Protein concentrations were determined by the bicinchoninic acid assay method (Pierce, Rockford, IL, USA). Equal amounts of proteins (10 µg) were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS‐PAGE) and transferred to immobilon membranes. After blocking with 10% non‐fat dry milk, primary and secondary antibodies were applied and the blots developed with enhanced chemiluminescence (Millipore) as we described previously 55. As loading control, GAPDH (Chemicon) was used for total lysates and Lamin A/C (Cell signaling) was used for nuclear fractions. Blots were quantified with Image J software.

Quantitative PCR analysis

Quantitative real‐time PCR analysis was performed to determine mRNA levels in cortical tissues from AD and controls. Total RNA was isolated with TRIzol (Invitrogen) and treated with DNase I (Takara). RNA samples were reverse‐transcribed using 6‐nucleotide random primers and M‐MLV reverse transcriptase (M0253, NEB). Quantitative PCR reactions were performed with SYBR™ Green Master Mix (43‐856‐12, Applied Biosystems) and specific primers on the StepOne 96‐well Real‐Time PCR System (Applied Biosystems). The mRNAs levels were normalized to the GAPDH expression level. The PCR reaction was performed at 95°C for 5 minutes followed by 45 cycles at 95°C for 10 s, 60°C for 30 s. Each set of PCR reaction was performed in triplicate and CT values of each PCR reaction were obtained. The ratios of gene expression were calculated relative to the control set in the experiments. The following primer pairs were used: SREBP‐2, 5′‐GGCTACGGTGCAGACAGTTG‐3′ (sense) and 5′‐GGTCAGAATGGTCCCACTGC‐3′ (antisense); HMGCoAR, 5′‐TTGACCTTTCCAGAGCAAGC‐3′ (sense) and 5′‐GCAAGAACTGACATGCAGCC‐3′ (antisense); GAPDH, 5′‐AAGGCTGTGGGCAAGG‐3′ (sense) and 5′‐TGGAGGAGTGGGTGTCG‐3′ (antisense).

Statistical analysis

Statistical analysis was performed using the SPSS 17.0 software. Data were analyzed by Student’s t‐test. All the data were expressed as means ± SEM. p < 0.05 was considered as statistically significant.

Results

Decreased nuclear translocation of mSREBP‐2 in AD brain

The activation of SREBP‐2 involves its cleavage into N‐terminal cleavage product of mSREBP‐2 and nuclear translocation. To study the activation of SREBP‐2 in AD, we used an N‐terminal specific antibody for SREBP‐2 which detects both full‐length SREBP‐2 and the N‐terminal cleavage product of mSREBP‐2. In the hippocampus sections of young (aged 37 years and under) and age‐matched (aged 60–86 years) normal human control cases, immunocytochemical studies revealed the N‐terminal antibody‐positive immunostaining localized to pyramidal neurons demonstrating both nuclear and cytoplasmic staining (Figure 1A,B). Some neurons showed strong nuclear staining (Figure 1A,B, arrows). In AD cases, however, fewer pyramidal neurons showed specific, but weaker, cytoplasmic localization and some dystrophic neurites (DNs) around plaques were also stained (Figure 1C, DNs‐arrowhead). Strikingly, in none of the AD cases were neuronal nuclei stained, and many neurons instead had clear, unstained nuclei (Figure 1C). Adsorption of the antibody with its peptide antigen completely abolished the neuronal nuclear staining (Figure 1D,E). In the white matter, smaller glial nuclei and cells with typical astrocyte morphology were also stained in the majority of control cases (Figure 1F), while the AD cases showed much less glial cell staining (Figure 1G). Comparison of the N‐terminal SREBP‐2 immunostaining with AT8, an antibody specific for phosphorylated tau, on adjacent serial sections in AD cases, revealed that AT8‐positive NFTs had no or much reduced SREBP‐2 immunoreactivity (Figure 1H,I). Densitometric analysis found that the NFT‐containing neurons had significantly less SREBP‐2 compared to AT8‐free neurons (Figure 1J).

Figure 1.

Immunocytochemical analysis of SREBP‐2 N‐terminal antibody revealed reduced nuclear translocation of SREBP‐2 in the pyramidal neurons in AD hippocampus. In both young (A) and age‐matched (B) control individuals, pyramidal neurons in the hippocampus contain SREBP‐2 detected by the N‐terminal antibody, having both cytoplasm and nucleus clearly stained and some neurons have strongly stained nuclei (arrows). In the same hippocampal area, however, a smaller number of the pyramidal neurons are readily detected by the N‐terminal antibody with only cytoplasm stained and nuclei clearly unstained in cases of AD (C) and a subset of dystrophic neurites associated with amyloid plaques are also stained (C, arrowhead). Adsorption with the peptide antigen finds that the staining (D) is completely abolished (E, * label landmark vessels in adjacent sections). In the white matter of control cases, many smaller, presumably glia, nuclei and occasional typical astrocytes (arrow) are stained (F), while this pattern appears diminished in AD cases (G). Immunostaining adjacent serial sections with SREBP‐2 (H) and tau antibody AT8 (I) finds that few NFT show SREBP‐2 colocalization (asterisks). AT8‐positive NFT are encircled in I and labels transposed onto the image in H. Landmark vessels are also outlined and asterisks mark those neurons that co‐stained with AT8. Quantification of SREBP‐2 immunoreactivity in pyramidal neurons from multiple adjacent sections from 3 AD cases finds that levels of SREBP‐2 is greatly reduced in NFT‐containing neurons co‐stained with AT8 (*P < 0.001). (Scale bars = 50 µm).

Western blot analysis of the human cortical homogenates with the N‐terminal SREBP‐2 antibody revealed several bands between 55 and 68 kDa and one weak band around 125 kDa (Figure 2A). The bands between 55 and 68 kDa were also found in the nuclear fraction (Figure 2D) which likely represent the mSREBP‐2. The weaker 125 kDa band (Figure 2A) likely represents the full‐length protein. Western blot and quantification analysis revealed significant reduction in the levels of full‐length SREBP‐2 (Figure 2B) and mSREBP‐2 (Figure 2C) in total lysate between AD and control. Similarly, in the nuclear fractions, the levels of mSREBP‐2 were also decreased significantly in AD (P < 0.05; Figure 2E). Consistent with a reduced level of nuclear translocated mSREBP‐2 in AD, the real‐time PCR analysis revealed decreased expression of SREBP‐2 (P < 0.05) and its downstream effector of HMGcoAR (P = 0.07) in the cortical tissues from AD compared with age‐matched controls (Figure 2F).

Figure 2.

Biochemical analysis revealed reduced nuclear translocation of SREBP‐2 and SREBP‐2 signaling in AD brain. A‐C. Western blot of total lysate from human cortical homogenates showed strong bands around 55–68 kDa (ie, mSREBP‐2) and a weak band around 120 kDa (ie, full length or FL SREBP‐2) (A). The longer exposure image of this weak band was shown on the top. Quantification analysis revealed decreased protein levels of FL SREBP‐2 (B) and mSREBP‐2 (C) between AD and controls. GAPDH was used as a loading control. D,E. Western blot of purified nuclear fraction (D) and quantification (E) revealed significantly decreased levels of mSREBP‐2 in AD compared with age‐matched controls. LaminA/C was used as a loading control. F. Quantitative PCR analysis revealed decreased mRNA levels of SREBP‐2 and HMGCoAR in AD compared with age‐matched controls. (*P < 0.05, ***P < 0.001. N = 6 in each group).

Decreased nuclear translocation of mSREBP‐2 was correlated with tau alterations

We then explored what pathologic factor(s) may underlie decreased nuclear translocation of mSREBP‐2 in AD brain. It was shown that APP expression/processing and Aβ could affect SREBP‐2 signaling 38, 54, so we examined mSREBP localization in the CRND8 APP transgenic mice (Figure 3A).However, in both old CRND8 mice with many amyloid plaques, and age‐matched nonTg littermate control mice hippocampal neurons displayed a nuclear staining pattern.

Figure 3.

Reduced nuclear translocation of mSREBP‐2 was associated with tau alterations, instead of amyloid changes, in the hippocampus of animal models. A. In 12‐month‐old CRND8 mice with extensive amyloid pathology (as detected by 4G8, right panel), hippocampal neuronal nuclei are stained by N‐terminal SREBP‐2 antibody (middle panel), similar to that in age‐matched Non‐Tg littermate control mice (left panel). B. In the study of 3XTg AD mice and WT littermate controls, immunocytochemistry revealed clear nuclear staining in the pyramidal neurons in the hippocampus of 23‐month‐old WT control mice and in younger 3XTg mice (ie, 5 and 12 months old). After NFT pathology begins to appear, fewer neurons show a nuclear localization (ie, 16 months old), and by the age of 18 months, no neuronal nuclei are seen. Tau pathology in adjacent sections of the same 16‐, 18‐ and 19‐month‐old 3XTg AD mice is demonstrated by staining with AT8 (B). C. In the tau‐only overexpressing mouse model, P301L, the nuclear translocation is evident in the hypothalamus and preoptic nuclei in the WT (left panel) but not in Tg mice (middle panel). NFT in the same brain region is demonstrated by staining with AT8 (right panel). Scale bars = 50 µm.

To examine if tau alterations may be a candidate for the aberrant cytoplasmic accumulation of N‐terminal SREBP‐2 in neurons, we examined an aging series of the 3XTg mice (Figure 3B), which express mutations in APP, PS1 and tau and develop tau pathology in older mice. Indeed, the N‐terminal SREBP‐2 stained hippocampal neuronal nuclei in WT up to 23 months of age and young transgenic mice without tau pathology. However, starting at 16 months of age, when NFT‐like pathology is present (detected using AT8 in Figure 3B), fewer CA1 neurons showed the nuclear localization and by age of 19 months when many AT8‐positive NFTs are present, no neuronal nuclei exhibit SREBP‐2 (Figure 3B). To completely exclude the potential contribution from APP or Aβ, we also examined P301L mice (Figure 3C), which only overexpress mutant tau and develop tau pathology. In the non‐Tg mice, nuclei were well stained in the hypothalamus and preoptic nuclei. However, in the Tg mice that developed NFTs, as clearly labeled with AT8, neurons in the same brain areas demonstrated cytoplasmic staining with clearly unstained nuclei.

Localization of N‐terminal SREBP‐2 in neuritic pathology in AD and AD‐related diseases and mouse models

In addition to the neuronal localization of N‐terminal SREBP‐2 in human brain (Figure 1), some DNs around amyloid plaques are also stained in the hippocampal tissues from AD patients (Figures 1C and 4A). In fact, DNs in human cases of Down’s syndrome (Figure 4B) and familial AD cases with the Swedish double mutations of APP (Figure 4C), were also stained (SREBP‐2‐brown, and amyloid‐blue). In both the CRND8 APP transgenic mice (Figure 4D) and the 3XTg mice (Figure 4E), DNs surrounding amyloid plaques were also well stained. Reticulon‐3 (RTN3) plays a critical role in the development of neuritic pathology in AD 28. Transgenic mice overexpressing RTN3 develop dystrophic neurites independent of APP overexpression 28. Indeed, in these mice, the N‐terminal SREBP‐2 antibody strongly labeled the DNs in the CA1 region (Figure 4F), very similar to immunostaining pattern of RTN3 (Figure 4G) 28.

Figure 4.

N‐terminal SREBP‐2 accumulated in dystrophic neurites (DN) around amyloid plaques in the brain tissues from patients with AD or related human diseases as well as in animal models. Human brain was compared by using double staining for SREBP‐2 (brown) and amyloid monoclonal antibody 4G8 (blue) (A–C). In human disease tissue of AD (A), Down’s syndrome (B) and an APP mutation case bearing the Swedish double mutation (C), DNs were seen accumulated around plaques with the N‐terminal antibody. In transgenic mouse models overexpressing amyloid precursor protein (ie, CRND8 and 3XTg), the N‐terminal antibody detects the neuritic component surrounding plaques in both aged CRND8 mouse (D) and in aged 3XTg AD mice (E). In a 7‐month‐old RTN3 Tg mouse, a model of dystrophic neurites independent of APP, DNs also accumulate the N‐terminal SREBP‐2 (F) similar to the DNs stained with anit‐reticulon antibody (G). Scale bars = 50 µm.

NFT localization of SREBP‐2 C‐terminal fragments in AD brain

The C‐terminal specific antibody detects full‐length and cleaved products that should not be translocated to the nucleus. By immunostaining, C‐terminal SREBP‐2 was found in pyramidal neurons in the hippocampus in all control individuals (Figure 5A), while in the control cases over the age of 60 years, the occasional NFTs are stained (Figure 5A, arrows). In no case were neuronal nuclei apparently stained. In all of the AD cases examined, many neurons with the morphology of NFTs (Figure 5B,C) were intensively stained as well as many dystrophic neurites associated with amyloid plaques (Figure 5B,D). Indeed, immunostaining adjacent serial sections with various tau antibodies (ie, PHF1, AT8 and Alz‐50) revealed that many neurons contain both SREBP‐2 and tau (Figure 5E–H). Quantification showed approximately 40%–50% of tau‐positive neurons demonstrate SREBP‐2 immunoreactivity (Figure 5I).

Figure 5.

SREBP‐2 C‐terminal fragments accumulated in the neurofibrillary tangles in AD brain. Immunocytochemical analysis of SREBP‐2 C‐terminal antibody in the hippocampal tissues from AD and controls. In control individuals, pyramidal neurons in the hippocampus contain SREBP‐2 detected with the C‐terminal antibody (A) and in the aged cases, some neurons have higher levels of SREBP‐2 associated with NFTs (A, arrows). In the hippocampus of AD cases, many neurons have high levels of C‐terminal SREBP‐2 and indeed have the appearance of NFT (B, higher magnification C). Also, dystrophic neurites accumulated around amyloid plaques are apparently stained (B, higher magnification D). Correlation with tau staining on adjacent sections finds that many of the NFT detected with tau antibodies are well stained with the C‐terminal SREBP‐2 antibody. Many of the NFT labeled with the phosphorylated tau specific antibody PHF1 (F, marked with outlines), have C‐terminal SREBP‐2 (E, showing the PHF‐positive NFT outline superimposed). Also, many of the NFT detected with the conformational specific tau antibody Alz‐50 (H, marked with outlines) also have C‐terminal SREBP‐2 (G, showing the Alz‐50‐positive NFT outlines superimposed). Quantification of 3 fields from 3 cases using this staining and imaging method finds that between 40% and 50% of tau‐positive NFTs accumulate the SREBP‐2 detected using the C‐terminal specific antibody (I). Scale bars = 100 µm (A,B), 50 µm (C–H).

Only one specific band about 70 kDa was detected in human cortical homogenates by the C‐terminal specific antibody (Figure 6A), which likely represents the C‐terminal cleavage product. Quantification showed this C terminal fragment increased significantly in AD (P < 0.05; Figure 6B).

Figure 6.

Western blot analysis revealed increased levels of SREBP‐2 C‐terminal fragments in AD brain. A. Western blot of total lysate from human cortical homogenates showed strong bands around the expected molecular weight about 70KD in AD and control samples. GAPDH was used as an internal loading control. Quantification analysis, normalized to GAPDH levels, shows the C‐ terminal fragments of SREBP‐2 increased significantly in AD (*P < 0.05; B). Data are means ± SEM of three independent experiments.

In aged cases of Down’s syndrome (Figure 7B) and familial AD cases with the Swedish double mutations of APP (Figure 7C), the C‐terminal antibody detected NFT and dystrophic neurites associated with amyloid plaques similar to AD (Figure 7A). The NFT pathology in both the P301L mice (Figure 7D) and the 3XTg mice (Figure 7E) were readily stained by the C‐terminal SREBP‐2 antibody. Interestingly, unlike in human brain, DN in the 3XTg mice was barely stained (Figure 7E). A lack of detection of DN by C‐terminal antibody was also noted in both the CRND8 APP transgenic mice (Figure 7F) and the RTN3 transgenic mice (Figure 7G).

Figure 7.

Neurofibrillary changes, but not neuritic changes, were consistently labeled by SREBP‐2 C‐terminal antibody in the brain tissues from patients with AD or related human diseases as well as in animal models. NFTs and DNs were stained using the C‐terminal antibody in AD (A), DS (B) and a familial AD case with Swedish APP mutations (C). NFTs were also stained by C‐terminal SREBP‐2 antibody in P301L Tau Tg mice (D) and in 3XTg AD mice (E). However, the DNs present in 3XTg AD mice (E), CRND8 APP Tg mice (F) and RTN3 Tg mice (G) were not stained by C‐terminal SREBP‐2 antibody. Scale bars = 50 µm.

Increased cytoplasmic C‐terminal SREBP‐2 correlated with reduced nuclear mSREBP‐2

Since the N‐terminal and C‐terminal SREBP‐2 antibodies demonstrated distinct staining profiles among the AD and control cases, we compared their staining patterns in individual neurons in control and AD tissue sections. In a representative older control case, many pyramidal neurons exhibiting strong nuclear labeling with the N‐terminal antibody had only weak cytoplasmic staining with the C‐terminal antibody on the adjacent section (Figure 8A,B, same neurons marked with black arrows). The few neurons with much stronger N‐terminal SREBP‐2 cytoplasmic staining also demonstrated strong cytoplasmic staining with C‐terminal antibody (Figure 8A,B marked with arrowhead). In AD cases, strong nuclear labeling with the N‐terminal antibody was never apparent, but rather many of the same neurons with increased cytoplasmic staining (Figure 8C) also showed strong cytoplasmic staining with the C‐terminal antibody (Figure 8D).

Figure 8.

Increased cytoplasmic C‐terminal SREBP‐2 correlated with reduced nuclear mSREBP‐2. In aged control individuals, neurons with strong nuclear localization using the N‐terminal antibody (A, black arrows) have only weak cytoplasmic staining with the C‐terminal antibody (B, same neurons marked with black arrows). However, the one neuron with much stronger C‐terminal staining (B, arrowhead at top) appears to show both nuclear and cytoplasmic staining using the N‐terminal antibody (A, arrowhead). In cases of AD, many of the pyramidal neurons with increased levels of SREBP‐2 detected with the N‐terminal antibody (C) also have increased accumulation of SREBP‐2 detected with the C‐terminal antibody (D). Many of the same cells are labeled with the black arrows. *mark landmark vessels in the adjacent sections. Scale bars = 50 µm.

Discussion

The major finding of the current study is the decreased nuclear localization of SREBP‐2 detected by N‐terminal SREBP‐2 antibody in the AD brain by both immunocytochemical and western blot analysis: immunostaining revealed clear nuclear localization in pyramidal neurons in the hippocampal tissues from young and age‐matched control brain (Figure 1A,B,D) but predominant cytoplasmic localization in these neurons in AD with clearly unstained nuclei (Figure 1C). Western blot analysis demonstrated that although this antibody recognizes both full‐length and the cleaved SREBP‐2 (Figure 2A), only the 50–68 kDa cleaved products, representing the active mSREBP‐2, translocate to nucleus (Figure 2D). Furthermore, we confirmed that the mSREBP‐2 levels are significantly reduced in the nuclear fraction of the cortical homogenates from AD brain compared to that from control brain (Figure 2E). Our finding is unlikely an artifact since the antibody works as expected and absorption with its immunizing peptide completely abolished the immunoreactivity (Figure 1D,E). More importantly, the western blot result not only confirmed the specificity of the antibody but also was consistent with the immunocytochemical findings. Collectively, these data clearly demonstrated that nuclear mSREBP‐2 is significantly reduced in AD pyramidal neurons. As the master transcription factor for cholesterol biosynthesis, when cholesterol level is low, mSREBP‐2 is produced in the Golgi apparatus and is translocated to the nucleus, where it activates transcription of target genes involved in cholesterol synthesis 5, 23. Consistent with reduced nuclear mSREBP‐2, we found that the expression of its target genes, SREBP‐2 and HMGcoAR, are also reduced in AD brain tissues (Figure 2F). The decreased nuclear mSREBP‐2 in the pyramidal neurons thus indicates reduced cholesterol biosynthesis in these neurons in AD. Given that there is accumulated cholesterol in pyramidal neurons in AD 15, 16, our data suggest that this is likely due to either increased import from other cells such as astrocytes or decreased cholesterol elimination in pyramidal neurons through oxidization or both. However, since only in the white matter area of controls cases were many smaller nuclei representing glial cells, and some cells resembling astrocytes, commonly seen labeled (Figure 1F), yet reduced in AD cases (Figure 1G), suggesting that cholesterol biosynthesis may also be reduced in astrocytes in AD, makes the increased import of cholesterol from astrocytes less likely. On the other hand, cholesterol elimination through oxidation is catalyzed by the enzymes cholesterol 24‐hydroxylase (CYP46A1) and cholesterol 27‐hydroxylase, both of which were reduced in AD 9. Consistent with this notion, oxidized cholesterol is reduced in the plasma and CSF of AD 10. More recent study also demonstrated that tau caused CYP46A1 defects both in vitro and in vivo 10. Therefore, given the decreased cholesterol biosynthesis in both pyramidal neurons and astrocytes as suggested by our novel finding, it appears that a decreased cholesterol elimination likely plays a major role in the neuronal accumulation of cholesterol in AD.

The next important question is what caused reduced nuclear mSREBP‐2 in AD neurons. Several recent studies demonstrated that APP and/or Aβ negatively impacted cholesterol biosynthesis 38. However, we found normal nuclear mSREBP‐2 translocation in pyramidal neurons in old CRND8 APP transgenic mice with extensive amyloid plaque pathology (Figure 3A), suggesting that APP or Aβ may act on molecules downstream of mSREBP‐2. The negative association between the presence of AT8‐positive NFTs and the nuclear localization and expression of SREBP‐2 in AD (Figure 1H–J) suggests that tau pathology may be involved.

Indeed, similar to what is seen in AD, there is a loss of nuclear mSREBP‐2 in old 3XTg mice after the appearance of tau pathology (Figure 3B), and also in P301L Tg mice which only have tau alterations and not amyloid pathology (Figure 3C), thus strengthening the case for a causal role of tau alterations in reduced nuclear mSREBP‐2 signaling. However, it should be emphasized that lack of nuclear translocation of SREBP‐2 was also observed in neurons without clear AT8‐positive NFTs (Figure 1C), suggesting that tau alterations, earlier than the formation of AT8‐positive NFTs, such as pre‐tangles could contribute to SREBP‐2 alterations, but the tau pathology further exacerbate the deficits as evidenced by more significant reduction in SREBP‐2 in the NFT‐bearing neurons (Figure 1J). Nevertheless, it remains to be determined how tau expression and/or pathology inhibit SREBP‐2 signaling. Prior studies demonstrated that Tau overexpression and pathology cause Golgi fragmentation 34 and nucleocytoplasmic transport in AD 18, implicating that tau could negatively impact Golgi function and thus SREBP‐2 cleavage or impair nuclear mSREBP‐2 translocation. However, our western blot analysis demonstrated that the reduction of mSREBP‐2 appears proportional to the reduction of total level of SREBP‐2, suggesting that the cleavage of SREBP‐2 may not be affected (Figure 2A,B). Also, the extent of nuclear mSREBP‐2 reduction appears similar to the extent of mSREBP‐2 reduction in the total lysate, suggesting that nuclear translocation may not be affected either (Figure 2E). Therefore, it is most likely that tau expression and/or pathology may affect SREBP‐2 signaling through transcriptional regulation. In this regard, we found that tau expression inhibited the transcription activity of estrogen receptor and its downstream signaling 56. Indeed, the mRNA levels of the SREBP‐2 target genes including SREBP‐2 and HMGcoAR are reduced in AD (Figure 2F).

Another interesting finding of this study is that N‐terminal SREBP‐2 localizes to DNs in AD brain and related human diseases, as well as animal models of amyloidosis (ie, CRND8 APP Tg and 3XTg mice) and neuritic pathology (ie, RTN3 Tg mice) (Figure 4). The development of DNs is believed to contribute to cognitive impairment of AD 28 and our recent studies demonstrated that DNs originate from tubular ER 49. SREBP‐2 is localized in ER and translocated to Golgi apparatus to produce mSREBP‐2 during activation. It would be of interest to explore whether disturbed SREBP‐2 signaling or cholesterol metabolism contributes to neuritic dystrophy pathology.

While our finding of the NFTs staining of C‐terminal‐specific SREBP‐2 antibody is striking (Figure 5B,C), its significance is unclear. Western blot demonstrated that the antibody mainly recognizes the 70 kDa C‐terminal fragments (Figure 6) and therefore, this result likely suggests an accumulation of the C‐terminal fragments in the NFTs. Such an accumulation is accompanied by increased cytoplasmic mSREBP‐2 (Figure 8). mSREBP‐2 dimerizes before its translocation into nucleus 5. It is not clear whether the accumulation of C‐terminal fragments in NFTs contributes to the retention of mSREBP‐2 in the cytoplasm. The cleavage of SREBP‐2 takes place in the Golgi apparatus. Intraneuronal NFTs usually displaced Golgi apparatus 51 and components of Golgi apparatus sometimes are found in NFTs 30. As a component of Golgi apparatus, it is possible that SREBP‐2 or its cleavage products may be sequestered and accumulated in NFTs due to deformation of Golgi apparatus.

In summary, our study demonstrated disturbed SREBP‐2 signaling in AD and related models and suggested for the first time that tau alterations may contribute to disturbed cholesterol homeostasis in AD by impacting SREBP‐2 regulation. Studies in the recent decade established chronic neuroinflammation as an additional hallmark pathology of AD in addition to amyloid plaques and neurofibrillary tangles. Early studies demonstrated intimate connections of cholesterol metabolism and homeostasis with glial function and neuroinflammation 19, 45. For example, immune pathway and cholesterol metabolism and homeostasis involve many of the same genes including ApoE, TREM2 and ABCA7. Recent studies suggested SREBP‐2 as a signaling hub integrating cholesterol metabolism with NLRP3 inflammasome assembly and inflammation activation 25, while SREBP‐2‐deficiency could also enhance pro‐inflammatory signals in response to inflammatory insults 41. Thus, SREBP‐2 dysregulation downstream to tau alterations could potentially cause broader detrimental effects than cholesterol dis‐homeostasis and amplify toxic effects of tau through additional pathways such as neuroinflammation.

Author Contributions

XZ conceived and directed the project, designed experiments, interpreted results and wrote the manuscript. CW and FZ conceived and carried out experiments and wrote the manuscript. WW, KS, ASR and SLS carried out experiments. CFP, GP, LS, BS and RY contributed to interpretation of results and provided feedback on the manuscript. All authors had final approval of the submitted version.

Conflict of Interest

The authors have no conflicts of interest to declare.