Introduction
Alzheimer’s disease (AD) is the most frequent cause of dementia in elderly populations and currently affecting an estimated 47 million people worldwide [
1], but as the ages of populations in most countries are increasing, the incidence of AD will significantly increase. The brains of AD-affected subjects show accumulations of amyloid beta (Aβ) plaques and neurofibrillary tangles (NFT), the hallmark pathological features of this disease [
2]. Preventing the formation of these pathological structures is considered the key to preventing cognitive decline, the main clinical feature of AD, but the mechanisms and sequence of events leading to the accumulation of plaques and tangles, neuronal death and cognitive decline are not fully understood. To date, in spite of promising experimental data in AD animal models, therapies to prevent or remove Aβ have generally had limited effects in clinical trials to slow down cognitive decline [
3‐
5]; other approaches are also needed.
Progranulin (PGRN) is a glycosylated protein of 75–80 kDa that can be secreted or transported to lysosomes [
6]. It is expressed in many different tissues and cell-types [
7]. PGRN protein is composed of seven and a half repeats of a highly conserved cysteine-containing motif that can be cleaved into granulin peptides (A-G), some with proinflammatory properties [
8]. In brain, PGRN has been demonstrated to regulate neuroinflammation [
9,
10], neurite branching and outgrowth [
11,
12], and lysosomal function [
13,
14]. The role of PGRN in AD has attracted attention in recent years since the discovery that mutations in GRN, the gene for progranulin, is one cause of frontotemporal dementia (FTD) resulting from frontotemporal lobar degeneration (FTLD) [
15,
16]. In FTD, loss of function mutations in the GRN gene resulting in significantly reduced levels of PGRN protein lead to neurodegeneration [
15]. It has been hypothesized that reduced PGRN results in neurodegeneration due to enhanced neuroinflammation [
10,
17]. The mechanism of reduced PGRN causing enhanced neurodegeneration in FTD and AD has been investigated using gene deletion rodent models but with conflicting results depending on the model [
18‐
22]. Complete loss of PGRN results in enhanced neuroinflammation and disturbance of lysosomal function, but the clinical phenotypes of mice with heterozygous GRN deletion were variable [
13,
23,
24]. Increasing PGRN levels in animal models of FTD, AD and Parkinson’s disease (PD) have been reported to reduce both pathological and clinical features [
19,
23,
25‐
27]. However, there are increased levels of PGRN protein in human AD-affected brains and AD mouse models [
19]. It has been suggested that the onset of AD might be caused by a drop in PGRN levels prior to the end-stage increase, but this has only been demonstrated in AD mouse models not human subjects [
19]. The single nucleotide polymorphism (SNP) rs5848 (T) allele has been associated with an increased risk of AD due to its effect on PGRN protein levels, but these effects were not large [
28,
29]. Biomarker studies of PGRN levels in human cerebrospinal fluid (CSF) and plasma in AD subjects have shown changes with disease progression but limited diagnostic utility [
28,
30]. While most experimental studies on PGRN in brain have employed animal models of FTD with single mutation or GRN gene knockout, the number of studies relating to PGRN and AD are limited, but one feature observed in studies of AD transgenic mice and human brain samples was that PGRN accumulated around Aβ plaques [
19,
31‐
34]. An additional study that employed granulin domain-specific antibodies showed immunoreactivity of neurons, microglia and structures associated with plaques [
35].
Prosaposin (PSAP) is also a lysosomal regulatory protein with significant neuroprotective properties [
36‐
38]. Recent studies have shown biochemical interactions between PGRN and PSAP, with these interactions affecting the trafficking of these proteins to lysosomes [
39‐
41]. There were reduced levels of PSAP in neurons of GRN-deficient mice and in samples from FTD patients with GRN mutations [
42]. Transgenic mice with reduced PSAP expression demonstrated similar pathological and behavioral changes as GRN gene-deficient mice [
42]. PSAP deficiencies in mice led to significant impairment of PGRN trafficking to lysosomes but increased circulating levels of PGRN [
41]. Experimental models of neuronal injury resulted in increased levels of PSAP in neurons and microglia [
38,
42,
43]. The interactions of PGRN and PSAP are complex as both PSAP reduction and overexpression resulted in elevated levels of extracellular/secreted PGRN in different cellular models [
4]. Overexpression of PSAP increased the concentration of PGRN oligomers, while PSAP knockdown increased concentrations of PGRN monomers [
39]. These interactions affecting the levels, localization and aggregation of PGRN might have significant effects on its different biological activities. A recent proteomics study of CSF identified PSAP as a biomarker to discriminate between preclinical AD and control cases [
44].
As a result of the previous reports of increased PGRN expression in AD brains in contrast to the deficits occurring in FTD due to GRN mutations, detailed investigations using immunohistochemistry and biochemical techniques were carried out to address the question how increased expression of PGRN, a documented protective molecule, could be associated with pathology in AD. We employed a series of human brain samples from non-demented cases with low plaque and high plaque pathology, along with samples from demented AD cases with high plaque and tangle pathology to study the progression of changes of PGRN and PSAP expression and their interactions. We identified that PGRN and PSAP expression were increased in AD cases, and their interaction could be demonstrated in human brain samples. The interaction with PGRN and PSAP occurred early in plaque development being detectable in plaques present in the low plaque control cases, and PGRN associated with Aβ plaques in all cases were positive to differing extents for PSAP. Overall, these results suggest that the protective and inflammatory modulating properties of PGRN might not be functional in AD, and PSAP bound-aggregated PGRN associated with plaques might lack the biological activities associated with this protein. This can be the basis for further experimental studies, but could be an important issue when considering PGRN supplementation if the protein becomes sequestered into non-active or pathological forms associated with plaques.
Materials and methods
Human brain samples
All human brain tissue samples used in this study were obtained from the Banner Sun Health Research Institute Brain and Body Donation Program (Sun City, Arizona, U.S.A.) as part of the Arizona Study of Aging and Neurodegenerative Diseases (AZSAND) [
45]. The operations of the Brain and Body Donation Program have received continuous approval of different Institutional Review Boards (IRB). Current operations have been reviewed by Western IRB (Puyallup, WA, U.S.A.). Tissue studies carried out in the U.S.A. were considered non-human subject research under exemption 4 (C.F.R 46.101). Tissue studies carried out in Japan were approved by Shiga University of Medical Science Ethical Committee (Certificate no. 29–114). A summary of demographic details of cases used in this study is presented in Table
1. The details of cases used for immunohistochemistry are described in Table
1a, while those cases used for protein analysis are described in Table
1b. The cases used for protein analysis were all included in the larger group used for immunohistochemistry.
Table 1Demographic details of cases used in study
A. Demographic information of cases used for immunohistochemistry |
| Gender (M:F) | Mean age ± SD | PMI | APOE4 | Plaque score±SEM | Tangle score±SEM | BRAAK score |
LP (n = 16) | 8/8 | 84.75 ± 6.96 | 2.93 ± 0.88 | 0% (0/32) | 2.54 ± 2.12 | 4.92 ± 2.33 | I-IV |
HP (n = 15) | 7/8 | 85.62 ± 5.98 | 2.9 ± 0.64 | 10.7% (3/28) | 10.49 ± 2.55 | 4.5 ± 1.93 | II-IV |
AD (n = 14) | 6/8 | 80.64 ± 5.58 | 3.22 ± 0.98 | 35.7% (10/28) | 13.33 ± 3.15 | 12.46 ± 3.95 | V-VI |
B. Demographic information of cases used for protein analysis |
| Gender (F/M) | Mean age ± SD | PMI | APOE4 | Plaque score | Tangle score | BRAAK score |
LP (n = 12) | 6/6 | 85.91 ± 8.93 | 3.09 ± 1.02 | 4.5% (1/22) | 1.33 ± 1.93 | 5.43 ± 2.44 | I-IV |
HP (n = 9) | 6/3 | 87.22 ± 8.22 | 2.72 ± 0.28 | 12.5% (2/16) | 12.05 ± 1.58 | 5.38 ± 2.02 | II-IV |
AD (n = 11) | 4/7 | 80.27 ± 3.82 | 3.79 ± 0.47 | 36.7% (8/22) | 14.36 ± 0.67 | 13.5 ± 1.96 | V-VI |
Human brain autopsy and neuropathological diagnosis
At autopsy, brains were sectioned into 1 cm thick coronal slices. Tissue taken from the right hemisphere of each brain donor was frozen on dry ice, while coronal slices from the left hemisphere were fixed for 2 days in 4% paraformaldehyde or 10% formalin, followed by cryoprotection in phosphate buffered glycol/glycerol solution. Frozen brain slices were stored at − 70 to − 80 °C and retrieved for dissection when samples were required for biochemical studies.
All donated brains received full neuropathological diagnosis including reference to pre-mortem clinical history of each case. Consensus clinical and neuropathological criteria were used to diagnose AD, Dementia with Lewy bodies (DLB) or Parkinson’s disease (PD) in donated cases [
46,
47]. To assess severity of AD pathology in each case, tissue sections from 5 brain regions (entorhinal cortex, hippocampus, frontal cortex, temporal cortex and parietal cortex) were stained with thioflavin-S, Gallyas and Campbell-Switzer histological stains, and assessed semi-quantitatively for the density of neurofibrillary tangles and amyloid plaques. These methods of assessing pathological load are carried out by the neuropathology department of the Banner Sun Health Research Institute Brain and Body donation program on each donated brain as part of diagnostic procedures. In brief, for each case, each brain region was ranked on a scale of 0–3 based on 0 being no plaques or tangles, 1 being few plaques or tangles, 2 being moderate numbers of plaques and tangles and 3 being numerous plaques and tangles. By combining the measures across these 5 brain regions, assessment of total AD pathology can be ranked on an ordinal scale of 0–15 for plaques and tangles [
48]. The cases were classified into low-plaque non-demented (LP) (plaque score < 6), high-plaque non-demented (HP) (plaque score 6–14) and AD with dementia (plaque score > 12). The severity of Lewy body pathology as a score of 0–40 was assessed in 10 different brain regions using immunohistochemistry for phosphorylated alpha-synuclein according to the Unified Staging Scheme for Lewy body disorders [
49].
Apolipoprotein E genotyping
Apolipoprotein E genotypes were determined for most cases using a polymerase chain reaction (PCR)/restriction endonuclease fragment polymorphism method employing DNA extracted from cerebellum to discriminate between APOE2, APOE3 and APOE4 alleles [
50]. Results in Table
1 are presented as number of APOE4 alleles out of total numbers of APOE alleles identified in each group.
Immunohistochemistry
Paraformaldehyde or formalin-fixed tissue sections from temporal cortex (middle temporal gyrus) were used for localization of progranulin (PGRN)-positive cells identified with antibody AF2420 (R&D Systems, Minneapolis, MN, U.S.A.), and for colocalization with Aβ peptide and phosphorylated tau, and markers of microglia (IBA-1, CD45), astrocytes (GFAP), endothelial cells (CD31), lysosomal proteins (LAMP-1, CD68, prosaposin, cathepsin D) and others (sortilin, beta-secretase-1 (BACE1), TMEM106B, neurofilaments, synaptophysin) according to our previously published procedures [
51,
52]. Antibodies used in this study are listed in Table
2. For this procedure, 25 μm brain sections were processed using a free-floating method. Sections were rinsed three times in phosphate-buffered saline containing 0.3% Triton-X100 (PBSTx) (0.1 M Phosphate buffer, pH 7.4, 0.137 M NaCl, 0.3% Triton-X100 (Nacalai-Tesque, Kyoto, Japan)), and reacted in PBSTx containing 1% hydrogen peroxide (30 min) to remove endogenous peroxidase activity, rinsed three times in PBSTx and then incubated in optimal dilutions of antibody overnight with shaking at room temperature (RT). Sections were then rinsed three times, incubated in biotinylated anti-species immunoglobulin (Vector Laboratories, Burlingame, CA, U.S.A.) at 1:1000 for 2 h at room temperature, rinsed three times and then incubated with avidin-biotin-peroxidase complex (ABC, 1:1000, Vector Laboratories). Localization of bound antibody was visualized using avidin-biotin horseradish peroxidase (HRP) enzyme complex (ABC-Vector Laboratories) histochemistry and nickel ammonium sulfate-enhanced diaminobenzidine-HCl (100 μg/ml) (Dojindo, Kumamoto, Japan) as substrate to produce a dark purple reaction product. To detect a second antigen, reacted sections were quenched in 1% hydrogen peroxide in PBSTx for 30 min, rinsed and then reacted with the second antibody in the same manner. The second antibody was detected using the same procedure, but with diaminobenzidine-HCl (200 μg/ml) without nickel ammonium sulfate as substrate to produce a brown reaction product. Sections were then mounted on microscope slides, counterstained with neutral red, dehydrated and coverslipped with permanent mounting agent.
Table 2Information of primary antibodies used for the study
Progranulin | PGRN | R&D | AF2420 | Goat/Polyclonal | IHC | 1:4000 |
| | Systems | | | FIHC | 1:1500 |
| | | | | WB | 1:2000 |
| | | | | IP | 2 μg |
Amyloid beta | 6E10 | Biolegend | 803001 | Mouse/Monoclonal | IHC | 1:2000 |
(1–16) | | | | | FIHC | 1:1000 |
| | | | | WB | 1:1000 |
CD45 | CD45/HI30 | Biolegend | 304001 | Mouse/Monoclonal | IHC | 1:2000 |
| | | | | FIHC | 1:1000 |
PHF-Tau (Ser202/Thr205) | AT8 | ThermoFisher | MN1020 | Mouse/Monoclonal | IHC | 1:2000 |
PHF-Tau | AT180 | Thermo | MN1040 | Mouse/Monoclonal | FIHC | 1:2000 |
(Thr231) | | Fisher | | | WB | 1:2000 |
IBA1 | IBA1 | Fujifilm | 019–19,741 | Rabbit/Polyclonal | FIHC | 1:1000 |
CD31 | CD31 JC/70A | Abcam | Ab9498 | Mouse/Monoclonal | FIHC | 1:500 |
GFAP | GFAP | BD bioscience | 556330 | Mouse/Monoclonal | FIHC | 1:1000 |
LAMP1 | LAMP1 | Sigma | L1418 | Rabbit/Polyclonal | FIHC | 1:1000 |
CD68 | CD68 | Biolegend | 916104 | Mouse/Monoclonal | FIHC | 1:1000 |
Prosaposin | PSAP | R&D | AF8520 | Rabbit/Polyclonal | FIHC | 1:2000 |
| | Systems | | | WB | 1:10000 |
| | | | | IP | 2 μg |
| Sortilin NT3 | Abcam | Ab16640 | Rabbit/Polyclonal | FIHC | 1:1000 |
| | | | | WB | 1:2000 |
| TMEM106B | Bethyl Lab | A303- | Rabbit/Polyclonal | FIHC | 1:1000 |
| | | 439A-1 | | WB | 1:1000 |
Cathepsin D | cathepsin D | Cell signaling | #2284 | Rabbit/Polyclonal | WB | 1:1000 |
BACE1 | BACE1 | R&D | MAB931 | Mouse/Monoclonal | FIHC | 1:1000 |
| | Systems | | | WB | 1:2000 |
Pan NF | SMI312 | Biolegend | 837904 | Mouse/Monoclonal | FIHC | 1:1000 |
Synaptophysin | SVP-38 | Sigma | S5768 | Mouse/Monoclonal | FIHC | 1:500 |
goat IgG | goat IgG | R&D | AB-108-C | Goat/Polyclonal | WB | 1:2000 |
| | Systems | | | IP | 2 μg |
Multi-color fluorescent confocal immunohistochemistry was carried out to verify cellular co-localization of PGRN-expressing cells with certain other antigenic markers, as described previously [
51,
54]. Tissue sections were incubated with optimal dilutions of antibodies at room temperature overnight with shaking. After three washes (10 min each) in PBSTx, sections were incubated with optimal concentrations of fluorescent-labeled secondary antibodies. Bound primary antibodies were detected with Alexa Fluor 488-donkey anti-goat IgG, Alexa Fluor 568-donkey anti-rabbit or anti-mouse IgG or Alexa Fluor 647-donkey anti-mouse IgG or anti-rabbit IgG (all from ThermoFisher, U.S.A.). Sections were counterstained with Sudan Black (1% solution in 70% ethanol for 3 min) to quench tissue auto-fluorescence, and with DAPI (ThermoFisher, U.S.A.) to reveal nuclei. Sections were coverslipped with fluorescent mounting media (Vector Laboratories) and imaged using an Olympus FV1000 confocal microscope and system software. Some images were acquired using a Leica SP8 confocal microscope and this is indicated on the appropriate figure legend. All images presented are z-stacks of multiple scans (5 scans). These were examined for saturation using software. For imaging of plaques for fluorescent intensity measurements and three-dimensional imaging, z-stacks were acquired to encompass the entire structure (15–20 scans with step-size of approximately 0.46 μm) using the same laser settings.
Progranulin and Prosaposin antibody validation
To validate the specificity of the PGRN goat antibody (R&D Systems #AF2420), antibody was incubated overnight at 4 °C with recombinant human PGRN protein (R&D Systems #2420-PG, amino acids 18–593) in a mass ratio of 1:200. Similarly, the PSAP rabbit antibody (R&D Systems #AF8520) was incubated with recombinant human PSAP protein (Sino-Biologicals, Beijing, China, #16224-H08H) in the same ratio. Control and protein-absorbed antibodies were diluted to the optimal concentrations for immunohistochemistry and reacted with sections using the above-described enzyme immunohistochemistry procedure. PGRN-absorbed antibody prepared in the same manner was also used for western blots.
Quantification of Progranulin-positive plaques
To quantify numbers and areas of PGRN-positive plaques, brain sections double-stained for PGRN and Aβ by two-color DAB enzyme histochemistry were used. Photomicrographs were taken with a 10x objective in 3-fields per case. Images were enhanced to maximize color separation between PGRN immunoreactivity (purple) and Aβ plaques (brown). Field selection was performed by choosing 3 evenly-spaced fields encompassing all of the cortical grey matter layers of each case. Images were imported to Adobe Photoshop software (Adobe Inc., San Jose, CA, U.S.A.) and a grid layer consisting of 90,000 pixels per area (field) was created, and 10 fields were measured for a total area of 900,000 pixels/case. The following measures for each section were made; total number of plaques, number of PGRN-associated-plaques, percentage area covered by PGRN-associated plaques and mean area of PGRN-associated plaques as pixels/field.
Quantification of co-localization of Progranulin and Prosaposin with Aβ plaques
The amounts of colocalization of PGRN- and PSAP-immunoreactivity associated with Aβ plaques were quantified using double-stained confocal sections (PGRN and PSAP) for 3 LP, 3 HP and 3 AD cases. Sections were imaged using an Olympus FV1000 confocal microscope. The settings for laser intensities and number and thickness of scans were determined based on optimal results for LP cases, and these settings were maintained for all sections of HP and AD cases. The images were analyzed using EzColocalization plugin for ImageJ image analysis software [
55,
56]. This was used to determine the Pearson colocalization coefficient for 6 separate plaques analyzed for each case (total no. plaques analyzed =18/disease group). Using the same images, the fluorescent intensities of PGRN and PSAP and areas of PGRN/PSAP plaques were also measured for each case. Three-dimensional imaging and colocalization of PGRN and Aβ interactions, and PGRN and PSAP interactions, were carried out using Imaris 8 (Bitplane AG, Switzerland) and Meshlab v2016_12 software (
www.meshlab.net).
Western blotting
Extracts from brain samples were prepared by sonicating each sample in 5-volumes of RIPA buffer (ThermoFisher Scientific; 20 mM Tris-HCl, pH 7.5. 150 mM NaCl, 1% NP40, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate) supplemented with protease and phosphatase inhibitors (Nacalai-Tesque). These samples were used without centrifugation for preparation of total protein extracts for western blotting. A similar procedure was used to extract proteins from cell pellets of THP-1-derived macrophages and PGRN-overexpressing HEK-293 cells. Total protein concentration of each sample was determined using a MicroBCA assay kit with bovine serum albumin as standard. For SDS gel electrophoresis of brain protein samples, brain protein extracts (1 μg/ul) were dissolved in 4xSDS gel sample buffer (Wako Chemicals-FujiFilm, Japan) with or without reducing agent (0.1 M dithiothreitol), heated to 95 °C for 10 min and loaded onto 4–20% gradient pre-cast gels (Nacalai-Tesque, Kyoto, Japan). Gel electrophoresis was carried out at 100 V in Tris-glycine buffer except for Aβ proteins, which employed Tris-tricine buffer (pH 8.5, 100 mM Tris, 100 mM tricine and 0.1% SDS) as gel running buffer. Separated proteins were transferred to nitrocellulose and processed for detection. Membranes were blocked in 5% skimmed milk dissolved in Tris-buffered saline with 0.1% Tween 20 (TBST – 20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.1% Tween 20) and incubated in optimal dilutions of antibody (see Table
2) in 2% milk in TBST overnight at room temperature. Membranes were washed 3 times with TBST and incubated 2 h in HRP-labeled anti-goat, rabbit or mouse IgG (ThermoFisher) at 1:10,000. After a further 3 washes, membranes were exposed to Chemi-Lumi One Super Chemiluminescent substrate (Nacalai-Tesque) and imaged using an ImageQuant LAS 4000 system (GE LifeSciences, U.S.A.). Images were adjusted and band intensities measured using Image Studio Lite software (LI-COR, Lincoln, NE, USA). After initial detection, all membranes (except the immunoprecipitated samples) were reprobed with an HRP-conjugated antibody to β-actin (Abcam, Cambridge, MA. USA) for normalization purposes.
Increased sensitivity and resolution of PGRN-immunoreactive bands were obtained when western blot membranes were fixed in paraformaldehyde (PFA) vapor. A modification of the procedures described to increase detection of α-synuclein was used [
57‐
59]. Dried membranes were exposed to vaporized PFA, rather than by immersion fixation, in a sealed chamber at 60 °C for 30 min. After the PFA vapors had been vented, the membranes were processed using the above-described western blot detection method.
Enzymatic Deglycosylation
Enzymatic deglycosylation of PGRN was performed according to the manufacturer’s protocol using PNGase F (New England Biolabs, Beverley, MA, U.S.A.) [
54]. Twenty μg of protein lysates from cells and brain samples were diluted in denaturation buffer and heated at 100 °C for 10 min. PNGase F enzyme (1 unit) was added to those samples along with enzyme buffers and NP-40. Following incubation of 1 h at 37 °C, samples were diluted in 4xSDS sample buffer and analyzed by the described SDS-PAGE/western blot method.
Co-Immunoprecipitation
RIPA-brain extracts prepared for western blots were centrifuged at 14,000 g/30 min to prepare samples for immunoprecipitation assays. Immunoprecipitations were carried out using protein G- or protein A-coupled magnetic beads (G-Biosciences, St. Louis, MO, U.S.A.) conjugated with test antibodies. For each sample, 10 μl of Protein G or Protein A magnetic beads were collected and washed with PBS 0.01% Tween 20 (PBST) using a magnetic stand, which were then mixed with 2 μg of antibody (PGRN goat polyclonal, PSAP rabbit polyclonal, or normal goat IgG) for 30 min with constant mixing. Unbound antibodies were removed by washing beads with PBST, then 200 μg of brain protein extract or 100 μg of cell protein extracts were added to the antibody-coupled beads. Samples were mixed with antibody-conjugated beads for 18 h at 4 °C with rotation, washed three times with RIPA, and then eluted into SDS sample buffer without reducing agent at 80 °C. The omission of reducing agent and lower denaturation temperature prevented the eluted immunoglobulin molecules from being denatured to molecular sizes that interfere with detection of target proteins. Samples were separated through SDS polyacrylamide gels and transferred to nitrocellulose membranes as described and detected with different antibodies by western blot.
Progranulin-expressing cell culture
Protein extracts from macrophage-like cells produced from the THP-1 monocytic cell line and neuronal cells produced from LAN-5 neuroblastoma cells were used PGRN-containing samples for antibody validation studies,. THP-1 monocytes (TIB-202) obtained from the American Type Culture Collection (Manassas, VA, U.S.A.), were cultivated in suspension culture using RPMI media (Nacalai-Tesque) supplemented with 10% fetal bovine serum (FBS), and differentiated into adherent macrophage-like cells by treatment with 25 nM phorbol myristate acetate (PMA – Sigma Aldrich, St. Louis, MO, U.S.A.) for 3 days in RPMI with 5% FBS. LAN-5 neuroblastoma cells (provided by Dr. R.C. Seeger, Children’s Hospital of Los Angeles, CA, U.S.A.) were used as a human neuronal model [
60]. Cells were cultured in RPMI with 10% FBS and differentiated in RPMI with 5% FBS containing 10 μM retinoic acid (Nacalai-Tesque). A PGRN-overexpressing stable-transfected HEK cell line was also prepared. HEK cells were transfected with plasmid expressing PGRN protein fused with a green fluorescent protein sequence (gift from Dr. Morimura, Shiga University of Medical Science, Japan) and selected for resistance to G418 (500 μg/ml). Cells were collected and analyzed by western blot or immunoprecipitation for expression of PGRN.
Data analysis
Western blot data and plaque measurement data were analyzed by one-way Analysis of Variance (ANOVA) with Newman-Keuls post-hoc test for significance between paired groups. Significant differences were assumed if P values of less than 0.05 were obtained. All statistical analyses were carried out using Graphpad Prism Version 7 software (Graphpad software, La Jolla, CA, U.S.A.).
Discussion
The majority of experimental and neuropathological studies of PGRN have focused on the consequences of GRN gene mutations or deletion in rodent models. Mutation in one GRN allele resulting in lower levels of PGRN protein is a cause of frontotemporal lobar degeneration (FTLD) and the accompanying clinical syndrome frontotemporal dementia (FTD) [
62]. As PGRN has multiple cellular properties, the mechanisms that cause neurodegeneration have not definitively been identified. PGRN deficiencies in animal models are associated with increased neuroinflammation [
17], increased synaptic pruning [
21,
63], and dysregulation of lysosomal function [
14,
64]. The situation for PGRN in AD is different as increased levels of PGRN in AD brains measured by ELISA has been reported [
19], and confirmed in this study by western blot measurements. Another study showed no significant increase in PGRN protein in frontal cortex of AD brains by western blot, but did detect increased levels of PGRN mRNA in these samples [
34]. The overall aim of this study was to further investigate using human brain samples some of the features of PGRN interactions with neuropathology identified in mice models of AD, particularly the nature of PGRN associations with Aβ plaques. The common features of PGRN in human AD brains and AD mouse models are PGRN immunoreactive structures associated with Aβ plaques, and cellular expression of PGRN by neurons and microglia [
19,
31‐
34,
65]. From this study, we showed the interaction of PGRN with PSAP, a molecule with similar properties to PGRN, to be a major feature of PGRN-associated with amyloid plaques.
Increased levels of PGRN in AD brains could be considered a reparative feature to prevent further neuropathology as experimental studies have shown that supplementation of PGRN in Grn haploinsufficient mice reduced microglial activation, neuronal lipofuscinosis and improved lysosomal function [
25]. Viral transduction of PGRN into an AD transgenic mouse model reduced amyloid accumulation, neuroinflammation and synaptic loss [
27], and this treatment protected dopaminergic neurons in a toxin-induced PD mouse model [
26]; however, there have been conflicting results from other studies of PGRN in AD mouse models. In one study, reducing PGRN levels resulted in impaired microglial phagocytosis and increased amyloid plaque deposition, while overexpressing PGRN in microglia had the opposite consequence [
19]. Using a different AD mouse model (APP/PS1), deficiency of PGRN was associated with reduced deposition of diffuse amyloid due to enhancement of microglial phagocytosis caused by PGRN deficiency increasing expression of microglia TYROBP genes [
21]. This study and another observed increased tangle-associated phosphorylated tau with PGRN deficiency in P301L tau mutation mice [
66].
In this report, cellular and pathological localization of PGRN was carried out with a well-characterized PGRN antibody raised against a glycosylated recombinant fragment of PGRN corresponding to almost full-length PGRN. This antibody will be able to recognize multiple epitopes of PGRN and appeared to have excellent sensitivity for detection of PGRN in tissue samples. It had been shown to have excellent specificity with no staining or polypeptide bands detected against PGRN knock-out cell samples in contrast to other antibodies [
61]. We confirmed its specificity by peptide absorption with immunohistochemistry and western blot analyses. This antibody detected polypeptides of 75–80 kDa in brain samples, the expected size for full-length PGRN. Our studies showed that sensitivity of detection was enhanced in the absence of reducing agents and by fixation of western blot membranes with paraformaldehyde. With this antibody, we demonstrated expression by microglia, neurons and in the cerebrovasculature but not by astrocytes. Staining of structures closely associated with Aβ plaques but not neurofibrillary tangles was observed. In a previous study, the predominant PGRN polypeptide detected using a peptide-derived monoclonal antibody had a molecular weight of approximately 55 kDa, which we showed corresponded to unglycosylated PGRN [
34]. This study did not detect the abundant species of PGRN of 75–80 kDa in brain detected in this report. These authors also detected neurofibrillary tangles positive for PGRN, while we observed no significant association of PGRN-immunoreactivity of neurofibrillary tangles identified using two different antibodies to separate epitopes of phosphorylated tau.
A key issue to address in this study is whether the PGRN immunoreactive structures being identified in brain sections are full-length PGRN or proteolytically-processed granulin peptides. Western blot results with the goat antibody to PGRN identified full-length PGRN as the most abundantly present in all brain samples, while lower molecular weight granulin peptides were not readily detectable, or were at very low abundance. This antibody has been shown to recognize granulin peptides if present in samples [
61]. A recent study used granulin-domain antibodies to identify immunoreactive structures in AD brains similar to what we have characterized [
35]; however the granulin sub-domain antibodies used would have the ability to also detect PGRN, so the issue of the amount of PGRN compared to granulin in plaque structures will require further investigation.
It had been hypothesized that a deficit in PGRN might be an early feature of AD in the prodromal stage [
19], and the increase occurred later in the disease as pathology developed. We investigated this using pathologically-staged LP, HP and AD cases. We examined the levels of PGRN protein by western blot, and quantified the numbers and sizes of PGRN-associated amyloid plaques in each of these groups. A limitation to this study was that there were fewer protein extract samples available compared to fixed tissue sections, but all of the cases with available protein extract also had tissue sections. There was no difference in total brain PGRN protein levels between LP and HP cases, but there was a difference between these groups in the number and area of plaques with PGRN-immunoreactive structures. The presence and increase in numbers of PGRN-positive plaques suggest that once Aβ plaques form, PGRN becomes associated with them. In the LP cases, 7 of the cases had no Aβ plaques, while the remaining 9 with plaques all had PGRN-associated with Aβ plaques. In all of the groups, the plaques without PGRN-associated structures tended to have a diffuse morphology and were negative for staining with thioflavin-S, while the plaques with PGRN-associated structures had thioflavin-S positive, aggregated morphologies (Fig.
3d).
We investigated the nature of PGRN structures associated with plaques in terms of their interaction with lysosomal proteins as it had been observed using 5xFAD AD model mice that most PGRN associated with plaques was present within aberrant accumulations of lysosomes [
32]. In these mice, most plaques had significant amounts of LAMP-1, CD68 and other lysosomal proteins associated with them, with PGRN and LAMP-1 showing significant colocalization. Our study demonstrated LAMP-1 immunoreactivity associated with plaques but only limited colocalization with PGRN immunoreactivity. CD68 immunoreactivity associated with plaques colocalized with PGRN immunoreactivity, but the majority of plaque-associated PGRN did not colocalize with these lysosomal proteins. The question arising from these observations is whether PGRN associated with plaques was enhancing plaque development, promoting its removal or was not in a biologically active form. On account of our biochemical studies showing that PGRN immunoprecipitated from brain samples pulled down PSAP, the involvement of this protein with PGRN in plaques became an additional feature of this study.
This study has made a number of observations concerning PSAP in relation to AD pathology. With the human brain samples available, we were able to examine PGRN and PSAP changes early in the pathological stages of plaque and tangle formation. Using these samples for co-immunoprecipitation studies of human brain samples, we showed that precipitating PGRN consistently pulled down PSAP but with little difference between the different disease groups. This is the first demonstration of PGRN/PSAP interactions in human brain samples. These PGRN immunoprecipitated samples were negative for sortilin [
67,
68]; this finding was unexpected as sortilin is enriched in plaques and regulates PGRN levels [
53,
69]. Other proteins that did not interact with PGRN were TMEM106B [
34,
70], cathepsin D [
71], EphA2 [
72] and BACE-1 [
32]. Other proteins have been identified to interact with PGRN that were not assessed. These include phospholipase D3 (PLD3) that colocalizes with PGRN on neuritic plaques [
73], and Toll-like receptor-9 whose signaling in macrophages is regulated by granulin [
74]. However, we could detect PSAP binding to cathepsin D and BACE-1 in brain samples. This will be investigated in further studies. The possible role of PSAP in AD has not been adequately addressed. There were no changes in microglial or neuronal PSAP immunoreactivity in AD sections compared to those from FTLD cases due to GRN mutations where neuronal PSAP was reduced, but microglia and astrocyte expression was increased [
42]. Our findings from this study were increased levels of PSAP in AD brains, with significant positive correlation between PGRN and PSAP levels in all samples. We confirmed PGRN and PSAP interactions in neurons and microglia, but most significantly might be the colocalization of PGRN and PSAP associated with plaque structures. Increased levels of PSAP have been shown to increase the oligomerization of PGRN [
42]. The interaction of PGRN and PSAP into aggregated structures may result in loss of biological activities, a feature that will need to be investigated. Even though there are increased amounts of PGRN and/or PSAP around plaques, being bound into these structures may prevent cellular signaling needed for their protective properties. It had been hypothesized that increased PGRN should stimulate associated microglia to phagocytose and remove the plaques, however if the PGRN is in a form that does not permit cellular endocytosis, namely bound with PSAP, excess amounts of inactive PGRN protein might be hindering plaque removal or promoting Aβ deposition. We observed that PGRN/PSAP-positive Aβ plaques in AD cases appeared to extend beyond the zones of PGRN/PSAP deposits. The biological properties of PGRN bound with PSAP in extracellular locations have not been investigated.
The significance of PSAP in neurodegenerative diseases is just being appreciated. A recent study employing new proteomics methods identified PSAP as a CSF biomarker for distinguishing preclinical AD from AD [
44]. As PGRN has been studied as a biomarker for CSF and plasma but with limited diagnostic utility [
30], improved diagnostic results might occur by combining both of these factors in biomarker discovery studies. We found most PSAP expression in brain samples was in neurons, which strongly colocalized with PGRN. PSAP expression by microglia and astrocytes was very limited in AD and aged brains but we observed that most PGRN in plaques colocalized with PSAP. Induction of microglial and astrocytic PSAP expression was reported in acute injury. Using the acute cortical stab wound model in mice, a 10-fold increase in PGRN was detected and a 50% increase in PSAP [
42]. Experimental studies have shown that PSAP can regulate the levels and aggregation state of PGRN. Reducing levels of PSAP resulted in increased levels of PGRN in vitro, and PSAP gene-deficient mice had higher levels of PGRN. However, PSAP overexpression also induced increased amounts of aggregated PGRN, but not GRN mRNA [
39]. One of the initial questions about PGRN in Aβ plaques was whether the presence of this factor affected the aggregation and/or removal of the plaques. In vitro experiments have shown that PGRN can stimulate microglial phagocytosis but the PGRN associated with plaques, which is detectable in sections from non-AD low plaque cases, does not appear to promote the removal of plaques. PGRN/PSAP plaques are infiltrated by microglia but these cells do not appear capable of removing Aβ. Our results seem to indicate that Aβ deposition increases in AD brains irrespective of the presence of PGRN and PSAP.
In summary, we have described features of PGRN and PSAP in a staged series of human brain tissue samples, and in particular PGRN-positive plaques. The most noticeable feature was the interactions in plaques of PGRN with PSAP. It is possible that PGRN/PSAP interactions with plaques could result in sequestration of toxic forms of Aβ, however, to address this as a mechanism, further studies will determine whether PGRN interactions with PSAP affect its neuroprotective and anti-inflammatory properties. This could be an important issue when determining whether PGRN supplementation will be useful. If the excess PGRN protein becomes absorbed by plaques in AD cases, it might not have the expected neuroprotective properties. The possible role for PSAP supplementation can also be considered. Both PGRN and PSAP have similar growth factor and lysosomal regulatory functions, but the consequences of PSAP supplementation has not been explored.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.