Essential roles of plexin-B3+ oligodendrocyte precursor cells in the pathogenesis of Alzheimer’s disease

The role of oligodendrocyte lineage cells, the largest glial population in the adult central nervous system (CNS), in the pathogenesis of Alzheimer’s disease (AD) remains elusive. Here, we developed a culture method for adult oligodendrocyte progenitor cells (aOPCs). Fibroblast growth factor 2 (FGF2) promotes survival and proliferation of NG2+ aOPCs in a serum-free defined medium; a subpopulation (~5%) of plexin-B3+ aOPCs was also found. FGF2 withdrawal decreased NG2+, but increased plexin-B3+ aOPCs and Aβ1-42 secretion. Plexin-B3+ aOPCs were distributed throughout the adult rat brain, although less densely than NG2+ aOPCs. Spreading depolarization induced delayed cortical plexin-B3+ aOPC gliosis in the ipsilateral remote cortex. Furthermore, extracellular Aβ1-42 accumulation was occasionally found around plexin-B3+ aOPCs near the lesions. In AD brains, virtually all cortical SPs were immunostained for plexin-B3, and plexin-B3 levels increased significantly in the Sarkosyl-soluble fractions. These findings suggest that plexin-B3+ aOPCs may play essential roles in AD pathogenesis, as natural Aβ-secreting cells.


Introduction
The amyloid hypothesis for Alzheimer's disease (AD) posits a neuron-centric, linear cascade initiated by the abnormal production of longer forms of amyloid β peptides (Aβ), especially Aβ1-42, from amyloid precursor protein (APP), leading progressively to tau pathology, synaptic dysfunction, inflammation, neuronal loss, and, ultimately, dementia 1 . Activated astrocytes and microglia are commonly found as glial nests around senile plaques (SPs) composed of Aβs 2 . Such reactive gliosis occurs as a multicellular neuroinflammatory response to Aβ accumulation and is believed to contribute to the clearance and removal of extracellular Aβs. While much attention has, therefore, been paid to these cells 1 , the roles played by other major glia, such as oligodendrocyte (OL) lineage cells, in the pathogenesis of AD remain largely unknown.
OL lineage cells constitute ~75% of the neuroglial cells in the neocortex and are thus the largest group of non-neuronal cells in the adult human brain 3 . While mature OLs produce myelin and facilitate neuronal transmission, the roles played by adult OL progenitor cells (aOPCs), which are characterized by the expression of the platelet-derived growth factor receptor α subunit (PDGFRα) and the NG2 proteoglycan 4, 6

5
, are still unknown. These aOPCs, which descend from OPCs in the perinatal CNS 6 , are distributed throughout the adult brain, providing up to 5 -10% of the adult CNS cells 7 .
Although they generally continue to proliferate, some generate myelinating OLs in the gray and white matter even during adulthood 7,8 . Furthermore, in the human prefrontal and entorhinal cortex, myelination in the gray matter continues throughout life, peaking around the 5 th -6 th decade 9, 10 .
Braak & Braak found a link between the vulnerability of neurons and the cortical (gray matter) myelination due to the fact that the spreading of neurofibrillary tangles recapitulates the developing pattern of cortical myelination during adulthood in reverse order 11 . This neuropathological finding, though clearly demonstrating the close relationship between adult gray matter OL lineage cells and the AD pathogenesis, has been underestimated, partly due to the lack of appropriate markers and/or in vitro systems to investigate such relationships in humans and rodents. Most recently, Mathys and colleagues illuminate the brain transcript of AD at single-cell resolution and found unexpectedly that, at the early stage of AD, myelination and axonal integrity and repair may be essentially involved in the pathogenesis 12 . However, little is known about how 7 OL lineage cells change with healthy aging and what their role is in the initiation and progression of disease 13 .
Although CNS cell purification and culture from embryonic or postnatal (up to approximately P20) rodent brains are possible using several different protocols, those from adult (> 2 months old) brains remain more challenging and often result in no or low yield(s), except for a few cell types (i.e., adult microglia and hippocampal neural stem/progenitor cells (NSPCs) that reside within a highly specific stem cell niche in the dentate gyrus 14,15,16,17 . Culturing OL lineage cells from adult brains is rather complex.
While OPCs from postnatal ~ adult rodent optic nerves can be purified by using a combination of sequential immunopannings 18 , advancing age, especially after P50, renders any attempt to culture them in vitro increasingly difficult, and in fact nearly impossible 19 . Utilizing surgically resected adult human brains, Antel's group has established a method for purifying and culturing OLs (or OPCs) that can survive for up to 4 weeks in vitro 20, 21 . However, the use of serum limits their precise functional analysis and purity.

8
In the present study, we developed a novel method to purify aOPCs from the adult rat brain (> 2 months old) and to culture them for up to several months in a defined medium. This allowed us to identify plexin-B3 as a novel uncharacterized aOPC marker. We also found that plexin-B3 is a unique delayed marker for cortical gliosis following brain injuries. Furthermore, we show that plexin-B3 + aOPCs probably play roles in the pathogenesis of AD, most likely as natural Aβ (especially Aβ1-42)-secreting cells.

Results
Purification and culture of aOPCs from adult rat brain Density gradients have often been used to purify adult microglia and hippocampal neural stem/progenitor cells from adult rodent brains 15,16,17 . This allows less buoyant cells to be separated on density gradients spanning 1.030 -1.065 g/ml 14,15,16 in the case of microglia, and 1.065-1.074 g/ml in the case of NSPCs 17 . In these purification procedures, the fractions with much greater buoyancy (<1.030 g/ml) were always discarded because they were regarded as merely accumulated debris 14,15,17 . However, 9 we found that many unidentified cells were present in this more buoyant fraction (1.029 g/ml) (Fig. 1a).
Preliminary immunolabeling studies revealed that most cells in this fraction were positive for olig2, a transcriptional factor and a marker specific for OL lineage cells, while a smaller population was positive for Iba1, a marker for microglia. To further purify olig2 + cells, we overlaid these fractions on poly-d-lysine-coated dishes for 30 -60 min and then gently washed out the suspension. Because olig2 + cells more quickly and tightly adhered to the dishes, most likely due to their membrane charges, most microglias and debris were removed to the wash suspensions (Sw) (Fig. 1a).
Only in the primary cultures was a mitogenic effect also found in platelet-derived growth factor aa (PDGFaa), although not in epidermal growth factor, nerve growth factor, neurotrophin-3, or in ciliary neurotrophic factor. The effects of 1 0 (Supplementary Fig. 1). After the first passage, however, only FGF2 showed continuously increasing numbers of aOPCs (Fig. 1d). Within a few passages in FGF2, the cultures became more homogeneous; in fact, nearly 100 % of the cells became olig2 + and more than 95 % of those NG2 + .
Western blot (WB) analysis revealed strong expression of OPC markers including NG2, PDGFRα, olig2, and Sox10 (Fig. 1e). Markers for mature OLs, astrocytes, or neurons remained undetected or only detectable at negligible levels ( Taken together, these data suggest not only that OL lineage cells, especially aOPCs, can 1 1 be successfully isolated and cultured from the adult rat hippocampus, but also that FGF2 can effectively maintain aOPC properties for at least several months in vitro.

Plexin-B3 + aOPCs in vitro
In our analyses of microarray and RNA-seq, we found that a transmembrane protein, plexin-B3 24 , was highly expressed in cultured aOPCs. This was further confirmed in the WB analysis (Fig. 2a). Although plexin-family members are generally expressed in neuronal cells, Zhang et al. previously reported in an transcriptome database that plexin-B3 gene was enriched in OL lineage cells, especially in newly formed OLs (NFO) 22 (Zhang et al., 2014). Immunocytochemical studies further revealed that plexin-B3 + cells comprised less than 5% of total cells and all intensely olig2 + (Fig. 2b & c). Plexin-B3 + /olig2 + cells are generally NG2or very weakly NG2 + (Fig. 2b). Moreover, BrdU incorporation studies revealed that they even proliferate in culture (Fig. 2e), suggesting that plexin-B3 is most likely a novel aOPC marker.
Since mitogen withdrawal induces OL differentiation of cultured perinatal OPCs 28 , the effect of FGF2 withdrawal on the expression of the aOPC marker was whereas FGF2 withdrawal decreased NG2 and PDGFRα levels (Fig. 2d).
We found that the 5-day FGF2 withdrawal did not clearly change the transcription factor gene expression profiles of cultured aOPCs ( Supplementary Fig.   4A). In more detailed analyses using the lists of OL lineage-specific genes (the Top 40 genes of OPC, NFO, and MO by Zhang et al. 22   Taken together, these results indicate that FGF2 withdrawal does initiate OL differentiation of cultured aOPCs, albeit incompletely in vitro, and that plexin-B3 is most likely an uncharacterized late OPC marker.

Notch and APP processing in plexin-B3 + aOPCs in vitro
Since notch signaling is known to inhibit OL differentiation and myelination 28 of cultured perinatal OPCs, we next studied the effects of FGF2 withdrawal on notch signaling in cultured aOPCs. WB analysis revealed that FGF2 dose-dependently increased both of the notch1 and notch1 intracellular domains (NICDs) generated by 1 4 γ -secretase (Fig. 3a), suggesting that cultured aOPCs do possess regulatory systems for notch and γ -secretase. FGF2 dose-dependently increased notch signaling, supporting the notion that notch signaling might inhibit OL differentiation of cultured aOPCs.
We then investigated the processing of APP, another major substrate for We next studied the secretion of Aβs. Aβ1-40 or Aβ1-42 peptides were immunoprecipitated from the conditioned medium with antibodies specific for the carboxyl (C-) terminal of Aβ1-40 or Aβ1-42, respectively, and were analyzed by WB 1 5 with an M3.2 antibody that recognizes rodent Aβ10-15. As a control, secreted APPs (sAPPs) were also immunoprecipitated with the monoclonal antibody 22c11 To more quantitatively measure the levels of Aβ in the medium, we performed ELISAs for Aβx-40 or Aβx-42. An LDH assay was also performed in parallel to adjust cell numbers. To compare Aβ levels between cultured aOPCs and neurons, primary fetal rat hippocampal neurons were also cultured (3 weeks in vitro). We again confirmed that 1 6 Notably, aOPCs with 0 or 1 ng/ml FGF2 secreted more Aβx-42 than cultured fetal rat neurons, resulting in approximately 4-fold higher ratios of Aβx-42 to total Aβ in aOPC than in fetal rat neuron cultures (4.39 and 4.45-fold increases, respectively) ( Fig. 3d).

Brain injuries and plexin-B3 + aOPCs
It is well known that NG2 + aOPCs respond very quickly to brain injuries 35, 36, 37, 38 , a major risk factor for AD 42 . To investigate the effects of brain injury on plexin-B3 + 1 8 aOPCs, we first employed the stab wound model. At 2 to 3 days post stab wound, a dramatic response in NG2 + aOPCs was first observed in and around the stab lesions with increased NG2 immunoreactivities, as well as hypertrophic morphological changes ( Fig. 5a, NG2, 2 Days). Interestingly, no such clear plexin-B3 + aOPC response was noted in the same lesions (Fig. 5a, Plexin-B3, 2 Days).
At 6 to 7 days post stab wound, however, a dramatic increase in the numbers of plexin-B3 + aOPCs was observed. On the walls of the stab wound (Fig. 5a), plexin-B3 was intensely expressed in olig2 + aOPCs, which became hypertrophic and formed glial scars (Fig. 5a). In the gray matter near these glial scars, both laterally and medially, densities of plexin-B3 + aOPCs had significantly increased ( Fig. 5b & Supplementary   Fig. 8A).
Next, we employed the KCl injury model. Topical application of 3 M KCl for 10 minutes induced cortical spreading depression (CSD) within the ipsilateral cortex ( Fig. 5c & d). CSD is a self-propagating wave of cellular depolarization that has been implicated as a fundamental mechanism of progressive cortical injury observed in stroke and head trauma 40,42 . At 2 to 3 days post-KCl application, increased plexin-B3 1 9 immunoreactivity was found only within the CC, just beneath the necrotic lesions of the KCl (Fig. 5c, pink-colored area in 2D); however, these white matter reactions decreased to normal levels within 6 to 7 days.
At 6 to 7 days post-KCl application, however, plexin-B3 + aOPCs increased not only around the necrotic cortical lesions but also in the remote ipsilateral cortex (Fig. 5c & e). Slight microgliosis, though not NG2 + aOPC gliosis or astrocytosis, was observed in the same remote ipsilateral cortical areas in this mild CSD model ( Supplementary Fig.   8B). This delayed cortical gliosis was never induced in the contralateral cortex ( Fig. 5c & e). NaCl application similarly did not induce such gliosis (Fig. 5e, 7DN).

Characteristics of plexin-B3 expression in AD brains
Our in vitro and in vivo findings led us to consider a new idea, namely that plexin-B3 + aOPCs constitute one of the Aβ-secreting cells in AD. To more directly test this idea, we stained paraffin-embedded human brain sections from patients with AD and normal controls (Supplementary Table 3A). In normal control brains, we could not get any clear plexin-B3 immunoreactive structures in the paraffin sections. In AD brains, however, 2 0 we found that plexin-B3 antibodies stained almost all SPs, including cored ones ( Fig. 6a & b, Supplementary Fig. 9, 10 & 11). These plexin-B3 + structures were specific ( Fig. 6c & d, Supplementary Fig. 11A), and co-immunolabeled with antibodies for total Aβ (4G8) and Aβ1-42 ( Fig. 6e~h & Supplementary Fig. 10). Interestingly, Aβ + areas of cored SPs were always slightly larger than the corresponding plexin-B3 + cell body areas ( Fig. 6e~h, merge, see also Supplementary Fig. 10), indicating the extracellular precipitation of Aβs. We also confirmed that these plexin-B3 + SPs were closely associated with, but clearly distinct from, microglia and astrocytes (Fig. 6i~l).
WB analyses of Sarkosyl-soluble and -insoluble fractions of the frozen normal control and AD brains (Supplementary Table 3B) further revealed that the expression of plexin-B3 was detected only in the Sarkosyl-soluble fractions (Fig. 6m).
The specificity of the antibody was further confirmed by WB ( Supplementary Fig. 11B & C). We found that the levels of plexin-B3 increased in AD brains in association with the accumulation of abnormally hyperphosphorylated tau at Ser396 in the Sarkosyl-insoluble fractions (Fig. 6m~p). This idea is fairly consistent with Braak's neuropathological findings 11 {Braak, 1996 #1377} (Braak & Braak, 1996) and the recent findings in the unbaised single-cell transcriptome analyses of AD brains 12 . If our scenario is true, fine control of cortical myelination in association with proper APP processing in the OL lineage cells in aged brains would be one of the essential requirement of effective AD therapy ( Supplementary Fig. 12C), although the exact roles of APP in aOPCs, or its processing in OL differentiation, remain completely unknown.
Cortical plexin-B3 + aOPC gliosis induced by CSD (Fig. 5c,  2 times during 10 min) (Fig. 5d) was sufficient to induce delayed plexin-B3 + aOPC gliosis in the remote ipsilateral cortex, suggesting that plexin-B3 is one of the most sensitive glial markers for brain injuries. Interestingly, an uncompetitive pan-NMDA-R blocker, memantine, which has been approved for the treatment of dementia, has proven to be partially protective against CSD 43 . The preventive use of memantine-type derivatives in MCI or during the early phase of AD seems promising.
This study also sheds light on several signal transduction pathways, including FGF2 and plexin-B3-semaphorine, as potential therapeutic targets for AD. For example, our in vitro data clearly indicate that the loss of FGF2 signaling is amyloidgenic in aOPCs (Fig. 3). This, however, does not simply mean that FGF2 replenishment confers therapeutic benefits for AD, since it also promotes aOPC proliferation, thereby inhibiting OL differentiation ( Fig. 1d & f) and likely disturbing normal myelination.
Temporal up-and down-regulation of FGF2 may also be involved in the induction of plexin-B3 + aOPC gliosis in the CSD models 44,45 . In AD brains, elevated FGF2 levels have been reported 46,47 , further highlighting the need for clarifying the exact roles played by FGF2 in the pathogenesis of AD.

4
In contrast, almost nothing is known about the roles of plexin-B3 in physiological OL differentiation 48 or in AD pathogenesis. Plexin-B3, which is expressed mainly in postnatal brains 49 , is a high-affinity receptor specific for semaphorines 24, 48 .
Mice lacking plexin-B3 display normal CNS morphology and behaviour 49 . Even the fate of plexin-B3 + aOPCs in CSD-induced cortical gliosis remains unclear at present.
Since plexin-B3 act at least as a temporal aOPC marker expressed during the initiation of OL differentiation, plexin-B3 + aOPCs may differentiate into mature OLs for the recovery, remodeling, and remyelination of neuronal circuits after cortical injuries.
However, we hypothesize that such mechanisms may be largely destroyed in AD ( Supplementary Fig. 12). Further studies are clearly needed to understand the fundamental mechanisms of adult cortical myelination, the roles of plexin-B3 + aOPCs in health and disease, the pathophysiology of AD-type de-or dysmyelination, and its therapeutic interventions.
In conclusion, we not only provide both a novel culture method for aOPCs and an aOPC marker plexin-B3, but also demonstrate several lines of evidence suggesting that plexin-B3 + aOPCs may be one of the major Aβ-secreting cells in AD.

5
The culture method will be useful for discovering novel functions of aOPCs as well as the regulatory mechanisms underlying adult OL differentiation. Furthermore, these findings will shed light on new AD pharmacotherapies targeting plexin-B3 + aOPC differentiation and cortical (oligodendro)gliosis.

All protocols were approved by the Tokyo Metropolitan Institute of Medical Science
Animal Care and Use Committee.

Western blot
The

RNA extraction
Cultured aOPCs were washed with cold PBS and frozen at -80°C until shipping. Total RNA was purified using an RNeasy Mini Kit (QIAGEN, Limburg, Netherlands) in accordance with the manufacturer's instructions. RNA quality was accessed by Bioanalyzer analysis.

Microarray
Samples were shipped to Agilent Array Services (Hokkaido System Science, Sapporo, Japan). RNA was amplified into cRNA and labeled according to the Agilent One-Color

Microarray-Based Gene Expression Analysis Protocol (Agilent Technologies, Santa
Clara, CA). The samples were hybridized to Rat GE 4x44K v3 array slides, and the 3 0 arrays were then scanned using an Agilent Microarray Scanner (Agilent Technologies).
The scanned images were analyzed using the standard procedures described in the Agilent Feature Extraction software 9.5.3.1 (Agilent Technologies).
To compare rat and mouse genes, rat orthologs were checked manually through PubMed (http://www.ncbi.nlm.nih.gov/pubmed) or the Rat Genome Database (http://rgd.mcw.edu/). We excluded genes from the lists if no rat ortholog of the gene was found or if it was not listed in the Agilent Rat GE 4x44K v3 array. When more than one expression data was obtained for a gene, the largest expression data was used.
Complete lists of the cell-type-specific genes appear in Supplementary Table 4 -6, and the top genes appear in Fig. 1f & Supplementary Fig. 2.

RNA-sequencing
RNA extraction and RNA-sequencing (RNA-seq) were performed in TaKaRa RNA-seq Services (TaKaRa, Kusatsu, Japan), including library preparation, fragmentation and PCR enrichment of target RNA. Samples with an RNA integrity number greater than 8 were used for library construction. Sequencing libraries were prepared using a TruSeq 3 1

RNA Sample Prep Kit (Illumina) according to manufacturer's instructions and then
sequenced by the HiSeq 2500 platform (Illumina) to obtain 100 bp paired-end reads.

ELISA
A two-site sandwich ELISA 50 was also used for the measurement of Aβ levels. Cultures of embryonic rat hippocampal cells were prepared as previously described 51 .

Image analysis
Fluorescent images were observed using high-resolution confocal microscopies (Zeiss, Oberkochen, Germany). In some cases, the entire area of the immunostained section was digitalized with a virtual slide system (VS120) (Olympus, Tokyo, Japan). Plotting was performed manually on transparent layers, overlaid onto the original virtual slide images (x 10 objective). For cell counting, defined areas (e.g., Supplementary Fig. 8) were captured by a microscope equipped with AxioCam MRc 5 (x 20 objective) (Zeiss).
For the quantitative immunofluoresent histogram, cultured cells were imaged and analyzed using the Zen 2011 imaging software (Zeiss). For definition of the region 3 3 of interest (ROI), the whole cell territory was first outlined in light blue line under the phase contrast image and then lacked areas outlined by yellow lines were subtracted (Fig. 3f). The signal intensity of the ROI was measured and adjusted by the area (= Mean Intensity).

Animal studies
SD rats (Charles River Laboratories, Yokohama, Japan) were deeply anesthetized with pentobarbital (50 mg/kg) and placed in a stereotactic frame.
For stab wound injuries, 4 male animals (3 months) were underwent a stab wound in the right cortex (Bregma AP -3.6 mm, ML 2.4 mm). Rats were killed at 2 (n = 2) and 7 (n = 4) days after the lesion.
For KCl injuries, animals were assigned to the following groups: sham-operated controls (n = 2), 2 ~ 3-days (2D, n = 3), 6 ~ 7-days (7D, n = 4), or 17-days (n = 3) after 3 M KCl, and 3-days (n = 1) and 7-days (n = 1) after 3 M NaCl treatment. A right parietal trepanation (AP -5.7 mm, L 2.7 mm) was carried out and the dura was incised for the topical application of a cotton ball soaked with 3 M KCl or 3 4 NaCl. After 10 min, the application site was rinsed with saline and the animals were allowed to survive.
To detect proliferating cells in vivo, 10 mg/ml BrdU was administered intraperitoneally (50 mg/Kg) twice/day for 5 consecutive days.
For immunohistochemistry, the animals were anesthetized and transcardially

Human studies
The autopsied human brains (Supplementary Table 3A) were fixed in 10% formalin and embedded in paraffin wax. Five to 20-μm-thick sections from the hippocampus were obtained. For plexin-B3 immunohistochemistry, serial pretreatment with heat (110°C in 0.01 M citrate buffer for 10 min) and trypsin (0.05 % for 10 min) was performed. 3 6 Samples were then incubated in primary antibodies diluted in PBS containing 5% BSA and 0.03% Triton X-100 at 4°C for overnight, and biotinylated secondary antibodies For WB analysis of human brains (Supplementary Table 3B), biochemical fractionation from frozen postmortem brain homogenates was performed as previously described 52 . Briefly, unfixed frozen human brain blocks (Brodmann area 6) were homogenized in 9 volumes of extraction buffer A68 (10 mM Tris-HCl, pH 7.5, 1 mM EGTA, 10% sucrose, 0.8 M NaCl) containing a protease inhibitor cocktail (Calbiochem, 3 7 San Diego, CA). After adding 20% (w/v of water) Sarkosyl to concentrations of 2% w/v, the homogenates were left for 30 min at 37 °C, followed by a 10min spin at 15,000rpm.
The resulting supernatants were further centrifuged at 50,000rpm for 20min. The final supernatants retained as Sarkosyl-soluble fractions and the pellets were solubilized in PBS as Sarkosyl-insoluble fractions. To examine the effects of postmortem interval on the plexin-B3 expression, frozen rat brains dissected from the rats with 0 hr and with 70 hrs (at 4 °C) postmortem intervals were fractionated and analyized as described above.

Data analysis
The data are expressed as the mean ± SD unless otherwise indicated. Statistical comparisons were performed using paired Student's t-tests.  Immunocytochemistry of a typical cored-SP-like plexin-B3 + aOPC cultured in 0 ng/ml FGF2 for 5 days. Note that plexin-B3 + aOPC is also M3.2 positive. Scale bar: 15 μm.
(f) Quantitative analysis of plexin-B3 and M3.2 (full length APP or intracellular Aβ) immunoreactivities. An example of region of interest for a single cell is shown in the left panels. Note that, as plexin-B3 immunoreactivities increased, M3.2 immunoreactivities increased (right). Mean Intensity: the integrated total value of the signal intensities adjusted by the single cell area.