Postmortem human brains were banked through the willed body donation programs, which existed with government (municipal police department and office of the Red Cross Society of China) and university approval to provide cadavers for teaching anatomy to medical students. Efforts are being taken to develop this platform for human brain banking as a part of the initiatives supporting China Brain Project (Yan et al., 2015; Poo et al., 2016). A subgroup of the elderly cases was recorded being dementia at the time of hospitalization for the care of terminal illnesses (Supplementary Table S1). Brain samples were assessed for AD type-neuropathology using sections from temporal, parietal, frontal and occipital lobes, with the extent of pathology (if present) scored according to Braak’s staging and the NIH guideline (Jellinger and Bancher, 1988; Braak et al., 2006; Montine et al., 2012). Correlated anatomical and biochemical studies were carried out using samples from three groups designated according to age and AD-type neuropathology: (1) Mid-age cases (n = 9) died of non-neurological diseases and free of Aβ/tau pathology in the cerebrum; (2) Aged cases (n = 9) with a history of dementia but Braak’s score of neurofibrillary tangle ≥ IV and the ABC amyloid score ≥ B, defined as AD group, and (3) Aged control group (n = 9) with no Aβ/tau pathology observed in the brain. It should be noted that, for comparative pathological analyses, additional aged cases with cerebral amyloidosis were included (Supplementary Table S1). Through this latter approach, brain samples/cases exhibiting a spectrum of AD-type neuropathology were used to allow a correlated morphometric analysis on Aβ and sortilin pathologies. Brains from sortilin knockout (−/−, n = 4), wildtype (+/+, n = 4) and C57BL/6 mice were used for the purpose of antibody validation (Ruan et al., 2016), with hemi-brains obtained after vascular rinse with cold saline then prepared for immunohistochemistry and immunoblot, respectively. The use of postmodern human brains and laboratory animals was approved by the Ethics Committees of Central South University Xiangya School of Medicine and Chinese Academy of Medical Sciences, Institute of Basic Medical Sciences (human brain study), and the Animal Ethic Committee of SA Pathology at Adelaide, Australia (mouse brain study), in compliance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Postmortem human brains were bisected and cut into ~2 cm-thick frontal slices. Slices from one hemisphere (opposite to handedness) were fixed in formalin for at least 1 week, with the slices from the other hemi-brain stored at −70°C. Temporal, frontal, parietal and occipital lobe blocks were then sampled from the fixed hemisphere, and either embedded with wax or placed in 30% sucrose in 0.01 M phosphate buffer, which were further prepared into paraffin (5 μm thick) or cryostat (40 μm thick) sections. The cryostat sections were collected in phosphate-buffered saline (PBS, 0.01 M, pH 7.2), with consecutive sections placed orderly into 24 wells each containing 4–6 sections/block with equal distance (24 × 40 ≈ 1000 μm). The sections were rinsed with PBS twice to remove the embedding medium, and stored in a cryoprotectant at −20°C before use. Mouse brains were dissected out after vascular perfusion with cold saline, with hemi-brains either snap-frozen for biochemical study or fixed by immersion in 4% paraformaldehyde for anatomical study.

Four sets of consecutive mid-hippocampal temporal lobe sections from each brain, 4–6 cases together in each experiment, were stained immunohistochemically with four antibodies: goat anti-sortilin extracellular domain (diluted at 1:2000, AF3154, immunogenic peptide corresponding to amino acid (a.a.) 76–753 of human sortilin, AF3154-SP, R&D Systems China Co. Ltd., Shanghai, China), rabbit anti-sortilin intracellular C-terminal domain (1:2000, ab16640, immunogenic peptide corresponding to a.a. 800–831 of human sortilin, ab16686, Abcam Trading Shanghai Company Ltd., Shanghai, China), monoclonal mouse anti-Aβ 6E10 (1:5000, #39320, Signet Laboratories Inc., Dedham, MA, USA) and rabbit anti-phosphorylated tau (1:5000, T6819, Sigma-Aldrich, St. Louis, MO, USA). In this set of experiments, the sections subjected to 6E10 labeling were invariably pretreated with formic acid for 1 h at room temperature. To determine the effect of several section pretreatments, including sortilin immunogenic peptide blocking (at 5 and 10 times of primary antibody concentration), heating (65°C), formic acid (100%) and guanidine hydrochloride (HCl; 5 M) treatments, consecutive sections from four brains with extensive amyloid pathology were selected for each treatment paradigm. Also, in each experiment, several adjacent sections were processed together with other sections, excluding the primary antibody in the incubation buffer, which were used to obtain the cutoff level of nonspecific labeling in densitometry.

Other than the above specifications, all sections were treated free-floating first with 5% H₂O₂ in PBS for 30 min and 5% normal horse serum in PBS with 0.3% Triton X-100 for 1 h, followed by incubation with the primary antibodies at 4°C overnight. The sections were then reacted with biotinylated horse anti-mouse, rabbit and goat IgGs at 1:400 for 1 h and ABC reagents (1:400; Vector Laboratories, Burlingame, CA, USA) for 1 h, with the immunoreactivity visualized in 0.003% H₂O₂ and 0.05% 3,3′-diaminobenzidine. Immunolabeling on paraffin sections was processed on-slide following dewaxing, rehydration and the antibody incubation steps described above, which involved Aβ, p-Tau, and sortilin immunolabeling with the rabbit antibody. The immunolabeled sections were used for the scoring of the AD type-pathology (i.e., ABC staging for amyloid pathology and Braak’s I–VI staging for tau pathology (Braak et al., 2006; Montine et al., 2012), and for an overall assessment for the presence of extracellular sortilin labeling relative to the occurrence of Aβ deposition.

Double immunofluorescence was initiated with section pretreatment in PBS containing 5% donkey serum for 30 min. The sections were then incubated overnight at 4°C with the rabbit sortilin antibody (ab16640, 1:1000), together with one of the following: (1) goat anti-sortilin (AF3154, 1:1000); (2) 6E10 (1:4000); (3) mouse anti-SYN (1:1000; MAB329, Millipore, MA, USA); (4) mouse anti-phosphorylated tau (PHF1, courtesy of Dr. P Davis, 1:4000); (5) mouse anti-glial fibrillary acidic protein (GFAP, 1:2000; MAB360, Millipore); and (6) mouse anti-ApoE (1:1000, NE1004, Millipore). Immunofluorescence was visualized with Alexa Fluor^® 488 and Alexa Fluor^® 594 conjugated donkey anti-mouse, anti-rabbit or anti-goat IgGs (1:200, Invitrogen, Carlsbad, CA, USA). The sections were treated with 0.1% Sudan black to block autofluorescence before microscopic examination.

Preparation of tissue lysates and electrophoresis were carried out with caution to minimize the effect of temperature on protein degradation. Neocortex from the middle temporal gyrus of the frozen human cerebral hemispheres and frontal cortex from frozen mouse brains were sampled in the chamber of cryostat, then homogenized on ice by sonication in T-PER extraction buffer (Pierce, Rockford, IL, USA) containing protease inhibitors (Roche, Indianapolis, IN, USA). Resulting lysates were centrifuged at 15,000 g at 4°C, with supernatants collected and protein concentrations measured by DC protein assay (Bio-Rad Laboratories, Hercules, CA, USA). Extracts containing equal amount of total protein were run in SDS-polyacrylamide gel electrophoresis (PAGE) gels (protein loading and SDS concentration to be detailed along with result description). Separated protein products were electrotransferred to Trans-Blot pure nitrocellulose membranes, which were immunoblotted with the aforementioned antibodies (rabbit and goat anti-sortilin diluted at 1:2000, 6E10 at 1:4000, BACE1 at 1:2000 and phosphorylated tau at 1:2000; Cai et al., 2010), and that for loading controls including β-tubulin-III (1:5000, Millipore), β-actin (1:5000, Millipore) or glyceraldehyde-3-phosphate dehydrogenase (GAPDH; 1:5000, Millipore). The membranes were further reacted with HRP-conjugated secondary antibodies (1:20,000; Bio-Rad Laboratories). Immunoblotting signal was visualized with the ECL-Plus detection kit, followed by X-ray film exposure and image capture in a laser scanner.

Immunolabeled sections were examined on an Olympus BX51 microscope and a Nikon confocal microscope equipped with digital imaging systems (CellSens Standard and Nikon-EZ-C1, respectively). Light microscopic images were taken using 2× to 40× objectives, with low magnification images montaged to re-construct the labeling over the temporal lobe regions. Confocal images were taken at 20× and 40× through three scans covering ~6 μm tissue depth. Images prepared for densitometry were obtained from comparable locations in adjacent sections using the same light intensity and digital exposure pre-settings, for all cases and all types of immunolabeling. Quantification of immunolabeled profiles was carried out following incorporation of region-matched images into a single file, which was converted into gray-scale TIFF format. Subsequently, either the areas or optic densities of the profiles/anatomic regions of interest were measured with the OptiQuant software (Packard Instruments, Meriden, CT, USA). Optic densities over immunoblotted protein bands were obtained with the same software.

All image data were exported into Excel spreadsheets and re-arranged according to groups, with specific densities calculated by background subtraction, and areal fractions or relative density levels calculated against internal references, whenever applicable. Finally, all data were entered into Prism spreadsheets, analyzed statistically, and graphed if applicable (GraphPad Prism 4.1, San Diego, CA, USA). Statistical analyses were conducted using either paired two-tailed t-test or a nonparametric test (Kruskal–Wallis with Dunn’s multiple comparison) with the GraphPad Prism software. Details regarding group pairing, sample size and statistics for individual tests will be incorporated in the following sections along with result description as well as in figure legend. The minimal significant level of difference was set at p < 0.05. Figures were assembled with Photoshop 7.1.

We validated the specificity of the sortilin antibodies to the extracellular (goat) and intracellular C-terminal (rabbit) parts using sortilin knockout (−/−) and wildtype (+/+) mouse brains (Figures 1A–E; Ruan et al., 2016). Both antibodies marked a ~100 kDa band in immunoblot of sortilin+/+, but not −/−, lysates, representing a presence of the full-length protein in the wild-type but not knockout mouse tissues (Figures 1B,D). By extending the film exposure time, two bands at ~15 kDa and ~40 kDa became visible in the immunoblot with the C-terminal antibody of the sortilin+/+, but not sortilin−/−, lysates (Figure 1D), indicating the existence of specific sortilin-derived peptides in the former preparation at very low concentration. In immunohistochemistry, the two antibodies exhibited identical, essentially neuronal, labeling in the sortilin+/+ mouse forebrain (Figures 1C,E), whereas the labeling appeared background like in sortilin−/− counterpart (Figures 1B,D). We also confirmed a fully co-localized immunofluorescent labeling by the two antibodies, occurring primarily in the somata and dendrites of cortical and hippocampal pyramidal neurons as well as the granule cells of dentate gyrus (DG) in wild-type rodents, including at granular profiles inside the somata and large dendrites (Figures 1F–M).

The two sortilin antibodies also exhibited identical labeling pattern in mid-age human cerebrum (Figure 2). Thus, immunolabeling in the neocortex occurred in the gray matter, mainly over layers II–VI and heaviest in layers VI, with little reactivity in the WM (Figures 2A,H). At higher magnifications, pyramidal and polymorphic neurons in layers V/VI and II/III were among the most distinctly labeled (Figures 2A,B,D). The somata and apical dendrites of the subicular and hippocampal pyramidal neurons were clearly visualized, with granular elements visible in the cytoplasm (Figures 2C,E). In the DG, both the granule cell layer (GCL) and molecular layer (ML) showed heavy reactivity (Figures 2A,F). Hilar mossy cells and CA3 pyramidal neurons, including their dendritic trees, were well-labeled (Figures 2A,F,G,H,J), with the thorny excrescences (TE) on the dendrites clearly identifiable at high magnifications (Figures 2F,G,I,J). Overall, axonal profiles were not identifiable around the somata of cortical and hippocampal neurons (Figures 2D,E), nor in the cerebral WM (Figures 2A,H) and in the mossy fiber terminal field (Figures 2A,F,H).

The sortilin antibodies against the extracellular and intracellular domains exhibited sharply different labeling patterns in the brains with AD-type pathology. Thus, the C-terminal antibody visualized plaque-like lesions, whereas the extracellular domain antibody did not (Figures 3A–H; Supplementary Figures S1, S2). Notably, in the samples with extensive plaque lesions, the neuronal labeling by the C-terminal antibody became less impressive, compared to that by the extracellular domain antibody (Figures 3A–D; Supplementary Figures S1, S2). This phenomenon appeared to likely reflect a competitive situation among the antigen epitopes in the section to bind the available antibody, because we found that the neuronal labeling could be enhanced by raising the concentration (from 1:2000 to 1:1000 or 1:500) of the rabbit antibody in the incubation buffer (data not shown).

We carried out an overall assessment of the C-terminal antibody labeling in sections from more than 70 brains with and without amyloid and tau pathology (including 10 Caucasian AD cases; Supplementary Table S1). Overall, the plaque-like sortilin labeling was found consistently among the cases that also had cerebral amyloid pathology. Thus, as assessed between consecutive temporal lobe sections, sortilin and Aβ labeled plaques both occurred over the neocortex, subiculum and hippocampal areas with their distributions matchable in reference to given subregions and lamina, noticeable at low magnification (Figures 3A,B,E,F; Supplementary Figures S1, S2). As with Aβ plaques, sortilin plaques lacked a spatial parallelism with tau pathology over the temporal lobe regions (Figures 3A–H; Supplementary Figures S1, S2) (therefore we did not quantitatively analyze sortilin relative to tau pathology). While assessing the overall pattern of extracellular sortilin/Aβ labeling, it should be notified that, along with the compact-like lesions diffuse Aβ deposition was present in local areas of the cortex, e.g., around layer I, the cell islands in layer II of the entorhinal cortex and the WM. In comparison, immunoreactivity of the C-terminal sortilin antibody in the above anatomic locations appeared to be not changed (relative to adjacent regions; Figures 3I,J; Supplementary Figures S2F,I). Also as a general note, 6E10 was found to label β-amyloidosis locally around the pia and blood vessels in the gray and WMs, whereat no increased extracellular sortilin deposition was found at these locations (Figures 3K,L).

At high magnifications more clearly in immunolabeled sections with hematoxylin counterstain, the plaque-like lesions detected with the sortilin C-terminal antibody consisted of extracellularly deposited fibrillary products (Figures 3M–P). The plaques were largely round, varying in size, often with the fibrillary material densely packed across the entire plaque. However, some plaques contained less densely stained fibrils (Figure 3N). Coexisting with the above, labeled extracellular fibrils were seen to deposit over small areas about the sizes of the neuronal nuclei, or as mini-plaques of the sizes of the neuronal perikarya (Figure 3O). Large sortilin plaque profiles could be ring-like with the deposits packed on the periphery, leaving an empty or pale center that sometimes occupied by cell nuclei (Figure 3P).

We used immunogenic peptide absorption to verify the specific labeling of the sortilin antibodies in the human brain sections. The labeling of the rabbit antibody was diminished to the background levels (defined using sections processed in the absence of primary antibody), when the immunohistochemistry was proceeded by co-incubation with the C-terminal immunogenic peptide at either 5 (data not shown) and 10 times of the concentration of the primary antibody (Figures 4A,B,D,I). Such an effect did not occur by including the extracellular domain peptide in the incubation buffer (Figures 4A,C,I). Vice versa, the extracellular immunogenic peptide could completely block the labeling of the goat antibody, but did not affect the immunolabeling of the rabbit C-terminal antibody (Figures 4E–H,J).

In a subset of the aged and AD cases we noticed sortilin/Aβ plaques preferentially distributed along the ML of the DG, with distinctly labeled plaques also present in the subiculum and adjoining CA1 area (Figures 3K, 5). Therefore, we carried out correlative areal ratio analyses of sortilin and Aβ plaques across the ML and in a defined area of the CA1/subiculum region. The fractional areas of the sortilin and Aβ plaques, expressed as % area occupied, were plotted relative to individual cases (Figures 5A,B,E). A strong positive correlation (P < 0.0001, R² = 0.932) between the fractional areas of sortilin and Aβ plaques in the ML was found based on the analysis of sections from 20 individuals (Figure 5F). In the CA1/subiculum transitional region selected for quantification, the fractional areas of sortilin and Aβ plaques were also positively correlated (P < 0.0001, R² = 0.983, n = 24) in a case-dependent manner (Figures 5E,F). We further assessed if the overall amounts of sortilin and Aβ deposition were correlated in the neocortex, using a densitometric approach (Figure 5G). The resulting densitometric data indicated a positive correlation (P < 0.0001, R² = 0.880, n = 15) between sortilin and Aβ deposition in reference to individual brains (Figure 5H). It should be noted that the total optic density measured in a region reflected that contributed by labeled plaques as well as neurons. In the case of Aβ immunolabeling, diffuse plaques and vascular β-amyloidosis present in the measured areas would also contribute to the yielded total optic density. Nonetheless, the densitometric correlation reflected a trend of overall parallelism between sortilin and Aβ extracellular deposition in the neocortex.

In double immunofluorescence, only the rabbit antibody, but not the goat antibody, labeled plaques, while there was a colocalization of the labeling by the two antibodies in neuronal somata (Figures 6A1–4). Colocalization of extracellular sortilin and Aβ labeling at individual compact-like plaques was clearly seen in the hippocampal and cortical regions, shown for examples in the ML of the DG (Figures 6B1–4) and the subiculum (Figures 6C1–4). Notably, while the immunofluorescence appeared to be parallel in intensity between the two markers across the entire area of a plaque (Figures 6B1–4); it could be stronger for one marker than the other over subareas of the same plaque (Figures 6C1–4). Sortilin and SYN immunofluorescence co-existed at apparently neuritic plaques, with SYN-labeled DNs surrounded by or intermingled with extracellular sortilin labeling (Figures 6D1–4). Extracellular sortilin labeling was also found to coexist with neuritic elements with weak p-Tau immunofluorescence in plaques (Figures 6E1–4). Hypertrophic astrocytes with bright GFAP immunofluorescence were present in the vicinity of some sortilin labeled plaques (Figures 6F1–4). Moreover, sortilin-labeled plaques colocalized with extracellular ApoE labeling (Figures 6G1–4).

Tissue processing and antigen retrieval, e.g., temperature, formic acid and guanidine HCl, could affect Aβ aggregation or Aβ immunodetection. We therefore explored if sortilin labeling could be affected by some Aβ-enhancing antigen retrieval methods. These experiments were carried out by batch-processing consecutive frontal (blocked from the precentral gyrus) and parietal (postcentral gyrus) neocortical sections subjected to untreated and pretreated conditions. The sections from four brains with stage C Aβ pathology were used, with densitometric data obtained from the same gyral region.

Compared to control, sections preheated at 65°C for 1 h and 3 h had more intense Aβ immunolabeling, especially along the pia, over layer I, in the WM and at individual sites of cerebral vasculature (i.e., cerebral amyloid angiopathy, CAA; Figures 7A–C). The mean density of Aβ labeling measured over layers I–VI increased to 433.4 ± 134.3% and 470.9 ± 146.2% in the sections preheated for 1 h and 3 h, respectively, relative to control (100 ± 37.3%), exhibiting an overall difference among the three groups (P = 0.0003, KWI = 16.1), with significant difference between the control and both preheated groups, but not between the groups heated for 1 h and 3 h, by post hoc test (Figure 7D). On the contrary, section preheating did not alter the pattern and intensity of sortilin labeling (Figures 7E–H).

Section pretreatment with formic acid for 1 h and 3 h also greatly enhanced Aβ labeling relative to untreated controls, with the effect readily maximized by 1 h treatment therefore presented below. Quantitatively, the mean density of Aβ labeling in formic treated sections increased significantly (P < 0.0005, paired t-test), to over 4-fold (433.4 ± 134.3%) relative to untreated sections (100 ± 32.2%; Figures 7I,J,M). Again, formic acid treatment did not affect the distribution or intensity of sortilin labeling (Figures 7K–M). Pretreatment of sections with guanidine HCl for either 1 h or 3 h (not shown) appeared to enhance Aβ labeling slightly (however, not reached significance), relative to untreated control (Figures 7N,O,R). The distribution or the overall amount of sortilin labeling was not different between the treated and untreated sections (Figures 7P–R).

Western blot was used to explore the putative sortilin products found deposited at neuritic plaques, with an attempt to identify potential dependent effect of the presence of amyloid pathology as well as brain aging. Thus, temporal neocortical samples (n = 9/group, from the frozen hemisphere) were obtained from the cases in which the fixed contralateral hemispheres were pathological examined (see Supplementary Table S1). Cortical lysates were immunoblotted with the two sortilin antibodies, and antibodies to APP, BACE1 and p-Tau (as pathological controls). The rabbit sortilin antibody blotted the principal band migrated at ~100 kDa, which appeared thicker in the lysates from the AD and Aged relative to Mid-age groups. Another prominent band blotted by this antibody migrated at ~15 kDa, which was present in lysates from the AD and Age groups, denser in the former, while it was absent or minimal in the lysates from Mid-age subjects (Figure 8A). The medians of normalized (to β-actin signal) densities of the 100 kDa band were significantly different in the AD (mean ± SD = 209.6 ± 9.4%, same format below), Aged (192.2 ± 29.3%) and Mid-age (147.2 ± 38.8%) groups (P = 0.0220, Kruskal-Wallis statistic index, i.e., KWI, 7.6), with Dunn’s Multiple Comparison Test reported a statistical difference between the AD and Mid-age groups (Figure 8E1). The median densities of the 15 kDa product were different in the three groups (P < 0.0001, KWI = 21.3), with post hoc test indicating a significant increase in the AD (190.2 ± 34.7%) relative to Aged (44.9 ± 27.1%) and Mid-age (6.8 ± 7.7%) groups (Figure 8E2). A light ~40 kDa band was also visualized by the rabbit antibody in some AD and Aged cases, but not in the Mid-age group (Figure 8A) (not quantified as it was not consistently present). Different from the rabbit antibody, the goat sortilin antibody only blotted the ~100 kDa band (Figure 8B, quantitative data not shown), which showed a parallel trend in density between the cases as that blotted by the rabbit antibody. Immunoblotted p-Tau products occurred in large amounts in the lysates from the AD cases, appearing as a smear of bands ≤70 kDa (Figure 8C). The median density of the p-Tau products was increased in the AD group (1455.1 ± 745.0%) relative to Aged group (109.4 ± 205.5%) and Mid-age group (2.5 ± 3.8%) (P < 0.0001, KWI = 20.2) (Figure 8E3). The bands of full-length APP protein (~100 kDa) blotted with the 6E10 antibody appeared to be denser among the AD cases (Figure 7E), with an overall difference of its levels between AD (183.9 ± 13.2%), Aged (151.6 ± 27.4%) and Mid-age (88.1 ± 43.9%) groups (P = 0.0003, KWI = 16.0), elevated in the AD relative to the Mid-age groups by post hoc test (Figure 8E4). The amount of BACE1 protein (~70 kDa) appeared to be increased among the AD cases relative the other two groups (Figure 8D), with densitometry indicating an overall difference (P = 0.0003, KWI = 16.0) between the AD (138.7 ± 15.1%), and Aged (118.2 ± 22.1%) and Mid-age (105.9 ± 16.2%) groups, and significant difference between the AD and Mid-age groups by post hoc test (Figure 8E5). Amounts of β-actin (used as a total protein loading control) and β-tubulin (used as a neuronal protein control) appeared to be comparable between the samples (Figure 8D).

Sortilin is a type I transmembrane protein consisted of 831 a.a. residues in Homo sapiens, with a long extracellular N-terminal sequence (755 a.a. in length), a single-pass transmembrane domain (a.a. 756–779) and a short intracellular C-terminal tail (a.a. 780–831) (Westergaard et al., 2004; Hermey, 2009). First cloned from human brain (Petersen et al., 1997), sortilin is also called as neurotensin receptor-3 for an initially identified role in neurotensin signaling (Mazella et al., 1998; Sarret et al., 2003). Follow-up cell biology studies indicate that it plays a general role in sorting/trafficking of membrane proteins between intracellular organelles (Strong et al., 2012; Mortensen et al., 2014; Kjolby et al., 2015). Sortilin is also known as an ApoE receptor, and may play a key role in lipoprotein metabolism and involve in some cardiovascular and metabolic disorders (Carlo, 2013; Coutinho et al., 2013; Kjolby et al., 2015). In the nervous system, sortilin is shown to regulate the signaling of nerve growth factors and progranulin, which may relate to disease mechanisms underlying some neurological and mental disorders (Nykjaer et al., 2004; Chen et al., 2005; Hu et al., 2010; Yang et al., 2011, 2013).

In this study, we have characterized the normal expression pattern of sortilin in adult human cerebrum with two antibodies targeting separately to the extracellular and intracellular domains. The specificity of the antibodies has been vigorously checked with three approaches, i.e., gene knock-out system, immunogenic peptide absorption and primary antibody omission. Since both antibodies essentially only label the ~100 kDa band in immunoblot of adult/normal rodent and human brain lysates, the immunolabeling seen microscopically would reflect the expression of full-length sortilin in vivo. Based on morphological characteristics, cortical and hippocampal principal neurons express high levels of the full-length protein, particularly in the somatodendritic compartment. The labeled granular elements seen in the somata and dendrites would represent intracellular organelles wherein protein sorting/trafficking takes place (Petersen et al., 1997; Nielsen et al., 2001; Gelling et al., 2012).

Often it is difficult with immunolabeling to determine whether a protein is localized to synaptic structures under the light microscope, especially in human brain samples obtained with postmortem delay. The finding of sortilin immunolabeling to the TE of hilar mossy cells and CA3 pyramidal neurons in normal human brain is remarkable. The visualization of this complex dendritic spine formation by both the extracellular and intracellular domain antibodies strongly suggests that the full-length sortilin protein is actually enriched at postsynaptic sites. Therefore, sortilin could play a physiological role in sorting/recycling of postsynaptic proteins, which could be important for synapse function and plasticity.

Recent studies have explored sortilin in relevance to AD and age-related dementia at genomic, protein and cell biology levels. So far, no data have shown that sortilin gene mutation or polymorphic variation increases the risk of developing AD (Zeng et al., 2013). Notably, a late study reports that the SNP rs17646665 located in a non-coding region of SORT1 gens is associated with a reduced risk for AD (Andersson et al., 2016). Three studies (to the best of our knowledge) have analyzed the 100 kDa sortilin protein in postmortem human brains, with its levels found either preserved in AD patients and subjects with mild cognitive impairment relative to people with normal cognition at death (Mufson et al., 2010), or elevated among AD relative to control elderly (Finan et al., 2011; Saadipour et al., 2013). Cell biology and proteomics studies suggest a regulatory interplay between sortilin and the amyloidogenic proteins. Thus, sortilin can mediate APP and BACE1 trafficking (Finan et al., 2011; Saadipour et al., 2013; Yang et al., 2013), while BACE1 and γ-secretase may catalyze sortilin proteolysis (Nyborg et al., 2006; Hemming et al., 2009).

In the present study, we show that a sortilin C-terminal antibody visualizes extracellularly deposited products in the human brains with cerebral amyloid pathology. The smallest amount of microscopically detectable deposits appears fibrillary, present in the neuropil covering areas about the sizes of neuronal nuclei and somata. The deposits appear to accumulate locally to form mini-plaques, dense-packing and ring-like mature-looking plaques. Neither the isolated deposits nor plaques occur preferentially surrounding the neuronal somata. Also, the deposits are not detectable inside neuronal somata and large dendrites. Quantitative analyses based on fractional areal measurement of individual plaques or collective densitometry of labeled profiles indicate a correlation between sortilin and Aβ deposition in the human cerebrum in a brain region/lamina- and case-dependent manner. Double immunofluorescent characterizations with a panel of markers co-existing at neuritic plaques, i.e., SYN, p-Tau, GFAP and ApoE (Kida et al., 1995; Zhang et al., 2009; Cai et al., 2010, 2012; Sadleir et al., 2016), show a colocalization of sortilin deposits in this type of amyloid plaques.

As elaborated in the “Introduction” Section, senile or neuritic plaques are historically referred to the silver stained lesions consisted of DNs and amorphous material, i.e., the amyloid product recognized later. In Aβ antibody stain, compact plaques in primitive, dense-packing or cored forms appear to closely represent neuritic plaques (while the so-called “burn-out” plaques may lack or contain only a small amount of DNs; Delaère et al., 1991). Aβ antibodies also label diffuse plaques as well as vascular and meningeal amyloidosis that often co-exist with compact plaques in the brain. Antigen retrieval methods, especially formic acid treatment, are routinely applied to Aβ immunohistochemistry. Diffuse plaques can show fairly dense labeling and mix with compact plaques in the same microscopic field. Thus, microscopic differentiation between the two types may not be always conclusive. In fact, inter-laboratory and inter-experimental inconsistency exists in the assessment of plaque lesions (Alafuzoff et al., 2008). Our comparative analyses on the effect of antigen retrieval indicate that section pretreatment with formic acid or heating essentially does not alter the amount and pattern of sortilin deposition, although dramatically enhances Aβ labeling and affects plaque appearance. Thus, the extracellular sortilin immunolabeling appears to be highly reproducible when used to visualize neuritic-like plaques in postmortem human brain.

Collectively, the data obtained from the human brains through comparative immunohistochemical and immunoblot assessments with the extracellular and intracellular domain antibodies allow us to conclude that the extracellularly deposited sortilin products must be the fragments rather than the parent protein. Specifically, the 15 kDa fragments detected only by the C-terminal antibody appear to be the major culprit behind the extracellular lesions. The levels of these fragments are dramatically elevated in the cortical lysates from AD human brains with amyloid pathology relative to aged/mid-age controls without microscopically overt amyloidosis.

As an initially observed human pathology, many mechanistic questions await to address, such as how and why sortilin deposition occurs. Our current data allow some preliminary interrogation on a few issues, in reference to existing literature. The first one involves whether sortilin deposition occurs as a direct consequence of amyloidogenesis. While sortilin deposits coexist with Aβ at neuritic-like plaques in the gray matter, they are not evidently present along with meningeal and vascular β-amyloidosis or in areas with typical diffuse Aβ deposition in the gray and WMs. Accordingly, sortilin deposition could not be coherently explained as a passive event resulted from absorptive binding by Aβ. Neither sortilin full-length nor deposited products are labeled in the DNs. Therefore, it appears unlikely that the sortilin deposits derive from the DNs via secretase-mediated proteolysis of the parent protein. In comparison, increased BACE1 labeling can be found in DNs of neuritic plaques, swollen axons in the WM as well as in vascular and meningeal cells, implicating a potential role of Aβ overproduction by local cellular profiles for site-specific cerebral β-amyloidosis (Cheng et al., 2014; Deng et al., 2014; Yan et al., 2014; Xue et al., 2015; Devraj et al., 2016).

The second issue is about the anatomic origin of the sortilin deposits. Labeling of the full-length protein is localized primarily to neuronal profiles, enriched in the somatodendritic rather than axonal profiles (as discussed in the first part of “Discussion” Section). The giant spine formations, TE, are distinctly labeled. The labeling is also intense in the ML of the DG and the stratum radiatum of Ammon’s horn/subiculum likely in dendritic elements. Thus, dendritic branches and spines could be a major origin of the sortilin deposits. Such a notion fits well with the overall region-specific distribution of sortilin plaques in the neuropil of the gray matter, including the ML and stratum radiatum wherein neuronal somata are rare but plaques are distinct. Synaptic degeneration is considered the most important anatomical basis underlying cognitive decline in AD. Plaque formation appears to be a continuing process tightly associated with synaptic degeneration, with dendritic spine loss and axonal terminal degeneration/dystrophy at the lesion sites (Gonatas et al., 1967; Masliah et al., 1991; Ferrer et al., 1998; Adalbert and Coleman, 2013; Yan et al., 2014; Herms and Dorostkar, 2016; Sadleir et al., 2016). Sortilin fragments might derive largely from degenerating dendritic spines and branches therefore accumulate preferentially at neuritic-like plaques.

The third issue relates to the biochemical process that might underlie the formation of sortilin fragments. Some in vitro and in vivo data suggest that the extracellular domain of sortilin can shed off as soluble products (Navarro et al., 2002; Molgaard et al., 2016; Ogawa et al., 2016). The ~15 kDa products are exclusively immunoblotted by the C-terminal antibody, with the plaque labeling completely diminished by the corresponding blocking immunogenic peptide. With this information and in reference to the molecular weight (~15 kDa), the deposited sortilin products could be predicted to contain the C-terminal tail, transmembrane domain and a short part of the extracellular sequence (roughly a.a. 670–831 according to online protein analysis tools), without considering the situation of protein modification (the C-terminal region contains two putative phosphorylation sites). The sortilin fragments might be formed after the extracellular domain is cleaved off, and aberrantly accumulate site-specifically along with dendritic/spine degeneration of local neurons during plaque pathogenesis.

In summary, we have identified sortilin fragments as a new component of the abnormally deposited protein products in senile plaques in the human cerebrum. This sortilin extracellular pathology is fairly preferentially associated with neuritic-like amyloid plaques, but rarely occurs around the pia, at vascular wall and in regions with typical diffuse Aβ deposition. The plaque-like sortilin labeling is not affected by antigen retrieval methods that enhance Aβ immunoreactivity. The ~15 kDa fragments appear to be a major form of the deposited components, which is predicted to derive largely from the C-terminal region likely including the transmembrane part of the parent protein. Mechanistically, we speculate that these fragments are formed along with the degeneration of largely the dendritic spines and fine branches.