Soluble APP functions as a vascular niche signal that controls adult neural stem cell number

The vascular niche is now firmly established as a microenvironment that sustains tissue progenitors in close proximity to blood vessels (Rafii et al., 2016). In the neurogenic regions of the adult brain – the subventricular zone (SVZ) of the lateral ventricle and the subgranular zone (SGZ) of the dentate gyrus of the hippocampus (Kriegstein and Alvarez-Buylla, 2009; Ming and Song, 2005; Zhao et al., 2008) – neural stem cells (NSCs) are located in close proximity to specialized capillary endothelial cells (ECs) (Palmer et al., 2000; Shen et al., 2008; Tavazoie et al., 2008). Emerging evidence suggests that these ECs provide signals that direct NSC quiescence and growth (Rafii et al., 2016). In the SVZ, endothelial neurotrophin 3 (NT-3) and ephrin B2 control NSC quiescence as a soluble paracrine and cell-cell contact-dependent signal, respectively (Delgado et al., 2014; Ottone et al., 2014), and endothelial brain-derived neurotrophic factor (BDNF) (Leventhal et al., 1999), pigment epithelium-derived factor (PEDF; also known as Serpinf1) (Ramirez-Castillejo et al., 2006), betacellulin (Gomez-Gaviro et al., 2012) and placental growth factor 2 (PlGF-2) (Crouch et al., 2015) stimulate NSC growth. These studies raise an intriguing question about how multiple vascular signals coordinate NSC quiescence and growth.

Amyloid precursor protein (APP) is widely expressed throughout the adult mouse brain, including in neurogenic regions such as the SVZ (Caille et al., 2004) and SGZ (Guo et al., 2012; Wang et al., 2014), and affects embryonic and adult neurogenesis in mice (Baratchi et al., 2012; Caille et al., 2004; Ma et al., 2008; Wang et al., 2014, 2016). The secreted form of APP (soluble APP, sAPP) has been reported to enhance proliferation of SVZ-NSCs (Baratchi et al., 2012; Caille et al., 2004; Ohsawa et al., 1999) and SGZ-NSCs (Baratchi et al., 2012). Interestingly, infusion of sAPP increased the proliferation of epidermal growth factor (EGF)-responsive neurosphere-forming SVZ-NSCs in vivo (Caille et al., 2004), whereas App deficiency enhanced the proliferation of neural progenitors in the SGZ (Wang et al., 2014, 2016). Despite the significant roles of APP in adult neurogenesis, the potential involvement of APP as a vascular niche signal in maintaining SVZ-NSCs has not been studied.

In this study, we attempted to identify EC-derived soluble signals that control NSC number in the SVZ. We found that brain microvascular EC line (bEND3)-derived conditioned medium (CM) increased the number of SVZ-derived neurospheres and decreased the size of individual neurospheres in culture. One of the 29 proteins we identified from the bEND3-CM, sAPP, was shown to enhance neurosphere-forming potential but suppress NSC growth in culture. Furthermore, our extensive studies in conventional and cell type-specific mutant mice clearly demonstrate that endothelial APP negatively regulates NSC number in the SVZ.

To improve our understanding of the nature of vascular niche signals for NSC maintenance, we used an established neurosphere culture (passaged neurospheres) from adult mouse SVZ cells (Fig. 1A). Passaged neurospheres were cultured with a medium conditioned by bEND3 cells, which have been reported to support NSCs (Ottone et al., 2014; Shen et al., 2004). We found that the bEND3-CM increased the number of SVZ-derived neurospheres and decreased the size of individual neurospheres (Fig. 1B-E). After cells had been treated with bEND3-CM, secondary neurospheres formed in normal growth medium at a significantly higher number but with a smaller size (Fig. 1F,G), suggesting that the bEND3-CM enhances neurosphere-forming potential but suppresses NSC growth in culture. The bEND3-CM treatment did not affect multipotency of SVZ-NSCs, as differentiation of neurospheres into neurons, astrocytes and oligodendrocytes was observed (Fig. 1H-M). No significant pro-differentiative effects of bEND3-CM on NSCs were observed in culture (Fig. S1). We further examined the expression of key transcription factors such as Hes1/5, Sox2, Ascl1 (also known as Mash1) and Id1-4, which are characteristic of NSCs (Jung et al., 2010; Nam and Benezra, 2009; Ramirez-Castillejo et al., 2006). The bEND3-CM did not affect the expression of these genes, apart from a significant reduction of Hes5 expression and a slight increase of Id1 expression (Fig. 1N; data not shown), corroborating the observation that the bEND3-CM-treated neurospheres retain NSC features. The effects of the bEND3-CM on the size of individual neurospheres were not due to an increase in cell death (Fig. 1O-Q). Rather, phospho-histone H3⁺ proliferating cells were significantly decreased in the bEND3-CM-treated Sox2⁺ NSCs (Fig. 1R-T). The effect of bEND3-CM on the number of neurospheres is unlikely to be due to increased cell viability, because the number of apoptotic and necrotic cells remained unchanged after the bEND3-CM treatment (Fig. 1U). Combined, these results suggest that EC-derived signals enhance neurosphere-forming potential but suppress NSC growth in culture.

To identify EC-derived soluble factors, the CM proteins were fractionated by heparin-affinity chromatography. Based on the fractions' effects on the number and size of neurospheres, we focused on the fraction eluted by 300 mM NaCl (Fig. S2A). Liquid chromatography-mass spectrometry (LC-MS) analysis of the fraction revealed 29 proteins, some of which were examined for their effects on a neurosphere culture derived from acutely dissociated SVZ cells (Fig. S2B,C). Of these, we focused on sAPP because it was also identified from our secreted molecule profiling of the adult SVZ ECs (Lee et al., 2012). Of note, blocking antibodies against sAPP did not affect the number or the size of neurospheres in the bEND3-CM-containing culture, suggesting that other factor(s) may cooperate with sAPP (Fig. S2D,E).

Among different fragments of sAPP (Caille et al., 2004; Ohsawa et al., 1999; Olsen et al., 2014; Willem et al., 2015), the soluble N-terminal fragment (aa 18-286) and the 770 aa form sAPP(770) negatively regulate NSC growth but do not affect the number of neurospheres in the culture from acutely dissociated SVZ cells (Fig. S3). We further evaluated the effect of sAPP(770) on NSC behaviors in the established neurosphere culture (passaged neurospheres). As seen in the bEND3-CM treatment, sAPP(770) increased the number and decreased the size of both primary (Fig. 2A-D) and secondary (Fig. 2E,F) neurospheres while retaining their multipotency (Fig. 2G-L). These data suggest that sAPP(770) enhances neurosphere-forming potential but suppresses NSC growth in culture. sAPP(770) did not affect the expression of the key transcription factors, except for a significant reduction of Hes5 expression (Fig. 2M; data not shown). sAPP(770) treatment decreased the number of proliferating cells but did not appear to affect cell death in Sox2⁺ NSCs (Fig. 2N-S). The number of apoptotic and necrotic cells remained unchanged after sAPP(770) treatment (Fig. 2T). These results are similar to the effects of bEND3-CM on NSCs in culture.

Previous studies showed that APP is expressed in the SVZ (Caille et al., 2004). To confirm APP expression by an independent method, we performed RT-PCR experiments on acutely isolated ECs from mouse SVZ and cortex by fluorescence-activated cell sorting (FACS). We detected App variants in both SVZ and cortex ECs (Fig. S4), indicating that the SVZ ECs express APP, but its expression is not specific to SVZ ECs.

To investigate whether APP plays a similar role in regulating the NSC number in vivo, we performed a label-retention assay using 5-bromodeoxyuridine (BrdU) in the adult SVZ of App null mutants (Zheng et al., 1995) and their control littermates. BrdU⁺ label-retaining cells are defined as SVZ-NSCs because the cells have a relatively long cell-cycle time, co-express Sox2 (Fig. S5A,B) and GFAP (data not shown), but do not express any differentiation markers including S100 (Fig. S5C,D). Consistent with previous studies (Shen et al., 2008; Tavazoie et al., 2008), BrdU⁺ label-retaining NSCs were localized in close proximity to PECAM1⁺ capillaries in the SVZ (Fig. 3A). App null mutants exhibited a significant increase in the number of BrdU⁺ label-retaining NSCs in the SVZ, compared with control littermates (Fig. 3B-D). In support of this observation, the mutants yielded more primary neurospheres than did control littermates (Fig. 3E-G). However, there was no significant difference in the number of secondary neurospheres on subcloning (Fig. 3H-J), suggesting that APP negatively regulates NSC number in a non-cell-autonomous manner.

To address directly whether EC-derived APP controls NSC number in vivo, we bred App^flox/flox mice (Callahan et al., 2017) with an EC-specific Cre driver, Tie2-Cre (Kisanuki et al., 2001), and an NSC/astrocyte-specific Cre driver, GFAP-Cre (Zhuo et al., 2001), producing App^flox/flox;Tie2-Cre mice and App^flox/flox;GFAP-Cre mice, respectively. As seen in the conventional App null mutants, more BrdU⁺ label-retaining NSCs were detected in the SVZ of App^flox/flox;Tie2-Cre mice (Fig. 4A-C). In contrast, no change in NSC number was seen in the SVZ of App^flox/flox;GFAP-Cre mice (Fig. 4D-F). These data clearly indicate an important role for endothelial APP in determining NSC number in the SVZ.

No significant change in the pinwheel-like arrangement of GFAP⁺ NSCs, β-catenin⁺ ependymal cells (Fig. S6A,B) or the SVZ microvascular network (Fig. S6C-E) was observed in App mutants. Moreover, the mutants showed no significant change in the amount of doublecortin (DCX)⁺ neuroblasts (Fig. S6F-H) and Mash1 (Ascl1)⁺/Ki67 (Mki67)⁺ transit-amplifying cells in the SVZ (Fig. S6I-K). Likewise, no significant change in the amount of NeuN (Rbfox3)⁺ post-mitotic neurons in the olfactory bulb was observed (Fig. S6L-O). One possible explanation is that the ∼20% increase in the number of SVZ-NSCs might be overridden by the subsequent expansion of SVZ neural progenitors and the migration of SVZ neuroblasts to the olfactory bulb. These relatively subtle but significant phenotypes in conventional and App mutants could indicate that other APP family members may compensate for the loss of APP. Previous studies in mice suggest functional redundancy between APP and APP-like protein 2 (APLP2): APLP2 can take over APP functions in vivo (van der Kant and Goldstein, 2015). Indeed, APLP2 was found in the 29 proteins from the bEND3 derived-CM fraction (Fig. S2B) and was also identified from our secreted molecule profiling of the adult SVZ ECs (Lee et al., 2012). It would be interesting to study the functional redundancy between APP and APLP2 by analysis of App^−/−;Aplp2^−/− double knockout mice.

We present here a normal cell biological role of APP as a negative regulator of NSC growth to control NSC number in the adult SVZ. Given that the major source of SVZ-neurosphere-forming cells are actively dividing in vivo and include EGF-responsive activated NSCs but not BrdU⁺ label-retaining quiescent NSCs (Codega et al., 2014), the observations that sAPP treatment increases the number of neurospheres in culture (our studies) and neurosphere-forming cells in vivo (Caille et al., 2004) can be explained by the positive effect of sAPP on the neurosphere-forming potential of activated NSCs. By contrast, sAPP negatively regulates the growth of quiescent NSCs in vivo (our studies). These results raise one intriguing question: if sAPP treatment enhances the neurosphere-forming potential in activated NSCs, why does App deficiency lead to an increase in the number of neurosphere-forming cells in the SVZ? One possible explanation is that sAPP may be sufficient, but not necessary, to enhance the neurosphere-forming potential in vivo: in the App mutants, an increase in the number of quiescent NSCs may lead to an increase in the number of neurosphere-forming activated NSCs. Identification of NSC receptors for sAPP will provide a clearer understanding of the mechanisms by which sAPP regulates the cellular processes of quiescent and activated NSC behaviors.

C57BL/6 mice (The Jackson Laboratory), App mutants (Zheng et al., 1995) (The Jackson Laboratory), App^flox/flox mutants (Callahan et al., 2017), Tie2-Cre mice (Kisanuki et al., 2001), GFAP-Cre mice (Zhuo et al., 2001) (The Jackson Laboratory), and Flk1-GFP BAC Tg mice (Ishitobi et al., 2010) have been reported elsewhere. All experiments were carried out according to the guidelines approved by the National Heart, Lung, and Blood Institute Animal Care and Use Committee.

bEND3 cells (ATCC) were cultured in the ATCC formulated Dulbecco's Modified Eagle's Medium (DMEM, Gibco/Thermo Fisher Scientific) containing 10% fetal bovine serum (Hyclone Laboratories) and Penicillin-Streptomycin (Gibco/Thermo Fisher Scientific). Preparation of bEND3-derived CM is provided in supplementary Materials and Methods.

The secreted forms of control alkaline phosphatase (AP) and APP were prepared for the neurosphere assay. For details, see supplementary Materials and Methods.

A total of 500 ml of bEND3 CM was applied to Heparin-Agarose (5 ml bed volume, Sigma). After washing of the column by PBS, the bound proteins were eluted by 30 ml of different concentration of NaCl solution (from 200 mM to 500 mM, every 50 mM). The eluted proteins were concentrated with buffer exchange to PBS through the centrifugal filter unit (Amicon ultra 10K MWCO, Millipore). Each fraction was tested in the neurosphere assay.

Neurospheres were established and cultured as described previously (Ferron et al., 2007) with slight modifications. For details, see supplementary Materials and Methods.

Viability of NSCs was assessed as described previously (Ferron et al., 2007). For details, see supplementary Materials and Methods and Table S1.

BrdU labeling and whole-mount SVZ immunohistochemistry were performed essentially as described previously (Shen et al., 2008; Tavazoie et al., 2008). All confocal microscopy analysis was carried out on a Leica TCS SP5 confocal (Leica). For details and antibody information, see supplementary Materials and Methods and Table S1.

To isolate SVZ and cortex endothelial cells, we used Flk1-GFP endothelial cell reporter mice (Ishitobi et al., 2010). All sorting was performed with a FACS Aria (Becton Dickinson). For details, see supplementary Materials and Methods.

Total RNA was isolated using TRIzol reagent (Ambion) or the RNeasy Mini Kit (QIAGEN) according to the manufacturer's instructions, and reverse-transcribed using SuperScript III (Invitrogen) for PCR. For details and primer information, see supplementary Materials and Methods.

Quantification data are represented as means±s.d., means±s.e.m. or box-and-whisker plots. P-values were calculated by Student's t-test, with P<0.05 considered significant. For comparisons across more than two groups, data were analyzed using one-way analysis of variance (ANOVA) with Tukey-HSD multiple comparison test or two-way ANOVA with Sidak correction for multiple comparisons.

We thank Y. Chen and M. Gucek for LC-MS assistance; J. Hawkins and the staff of NIH Bldg50 and Bldg14F animal facility for assistance with mouse breeding and care; P. J. McCoy and the staff of the Flow Cytometry Core for FACS assistance; K. Gill for laboratory management and technical support; and R. Reed and F. Baldrey for administrative assistance. Thanks also to R. Adelstein for editorial help and discussion; W. Li for technical help; and S. Kim, B.-N. Koo, J. Msefya, L. Henderson and other members of the Laboratory of Stem Cell and Neuro-Vascular Biology for valuable help and discussion.