Introduction

Alzheimer’s disease (AD) entails versatile pathological changes such as synaptic degeneration, neuronal death, chronic inflammation, impaired vasculature function, and reduced plasticity of neural stem cells (NSCs) [1–3]. The cognitive decline that is observed in AD patients and experimental animal models is mainly caused by the reduced neural network integrity [4]. The efforts to rescue the cognitive decline and neuronal death traditionally largely focused on the neuronal compartment and aimed at preventing the death of the neurons, but this has not yielded in desired success in clinics [5,6]. An alternative approach was suggested to complement the treatment in neuronal compartments by increasing the production of new neurons to provide resilience and strength to the diseased circuitry [3,7–9]. Yet, neurogenesis in human brains is rather controversial [10–14]. Although many reports documented the presence of adult neurogenesis in humans [3,15–19], and several studies demonstrated that boosting the neurogenesis might be a viable option for alleviating the cognitive decline [7,20–23], the potential benefits of neurogenic outcome in AD conditions requires further investigation and critical testing. Additionally, in AD conditions, mammalian NSCs reduce the proliferative ability dramatically, and for neurogenesis to become a viable option for treatment of neurological disorders, mammalian NSCs must become plastic first. Therefore, examining how NSCs could be made proliferative and neurogenic during the course of AD could provide important clinical ramifications toward the treatment of this disease [24–26]; however, little is known about the mechanisms by which neural stem cells would enhance their proliferative response [16,27,28]. Recently, we established a zebrafish model, which can recapitulate the symptoms of AD in humans, such as neuronal death, synaptic degeneration, chronic inflammation, and cognitive decline [29–32]. Interestingly, in contrast to humans, the zebrafish brain could enhance NSC proliferation and neurogenesis through a previously unidentified neuro-immune regulation involving Interleukin-4 (IL4). IL4 secreted by dying neurons activated microglia, which in turn activated NSC proliferation [30]. We also found that IL4 could revert the pathological effects on NSC in an in vitro 3D reductionist model of AD [20]. Yet, the mechanisms by which IL4 regulates NSC proliferation and neurogenesis after amyloid toxicity remained unknown. In this manuscript, using single-cell sequencing, we identified a previously undocumented IL4-dependent mechanism that regulates NSC plasticity in the adult zebrafish brain, by which IL4 regulates production of serotonin, which suppresses production of the brain-derived neurotrophic factor (BDNF) in periventricular neurons juxtaposing the NSCs. We found that BDNF is required to activate NSC plasticity, proliferation of NSCs, and neurogenesis through its receptor nerve growth factor receptor A (NGFRA), the blockage of which reduces NSC proliferation. Overall, our results identify a mechanism by which IL4 regulates NSC plasticity through serotonin-dependent expression of BDNF in neurons in the zebrafish brain and demonstrate functional heterogeneity of NSCs based on receptor expression. Our results will provide a conceptual basis for neuro-immune regulation of NSCs, neuronal control of NSC proliferation, and differential response of NSC subtypes to various signals in zebrafish. Such understanding could be instrumental in the efforts to develop novel therapies for AD through increased neurogenesis.

Results

To determine the effects of IL4 on NSC plasticity in the adult zebrafish brain, we administered IL4 through cerebroventricular microinjection, dissected the telencephalon at 1 day post injection (dpi), separated the her4.1: green fluorescent protein (GFP)-positive NSCs from GFP-negative cells, and performed whole-transcriptome profiling in control and IL4-injected brains for both of these cell populations (Fig 1A, S1 Data). When we compared the gene expression profiles in GFP-negative cell populations that contain non-NSC cell types, including the neurons and immune cells, we found that IL4 administration increased the expression of 285 genes and down-regulated 1,435 genes (Fig 1B, S1 Data). To determine the pathways affected by IL4, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses and observed that one of the significantly regulated pathways was tryptophan metabolism (Fig 1C, S1 Fig). Specifically, we observed that the enzymes generating serotonin from tryptophan were down-regulated (Fig 1D, S2 Data), suggesting that IL4 might reduce the production of serotonin (5-HT). To test whether amyloid-beta42 (Aβ42, which induces IL4 expression [30]) and ectopic IL4 would reduce the availability of 5-HT in adult zebrafish brain, we performed immunohistochemistry for 5-HT (Fig 1E) and observed that Aβ42 and IL4 significantly reduces the 5-HT immunoreactivity at 1 dpi in adult zebrafish telencephalon (Fig 1F), and this is due to reduced expression of enzyme tph2, an enzyme responsible for production of serotonin (Fig 1G). We found that Aβ42 or IL4 does not induce the death of 5-HT cells because the cell bodies of serotonergic cells present in superior raphe region are intact and are not, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-positive after Aβ42 or IL4 treatment (S2 Fig). These results demonstrate that Aβ42 and IL4 suppress serotonin production rather than the innervation. Additionally, the effects of IL4 is specific to serotonergic system, because in our deep sequencing analyses we did not observe changes in KEGG analyses in other neurotransmitter pathways, such as dopamine, histamine, or noradrenalin (S2 Data).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. IL4 regulates tryptophan metabolism.

(A) Schematic view of the experimental pipeline for whole-transcriptome sequencing IL4 treatment. (B) MA-plot for DEGs. (C) GO-term analyses on DEGs. (D) Modified KEGG-pathway view of tryptophan metabolism. Green indicates the enzymes down-regulated by IL4. (E) IHC for 5-HT in control (left), Aβ42-injected (middle), and IL4-injected (right) brains. Single-channel images show 5-HT. Red insets are high-magnification images of arrowed regions. (Dm: dorsal-medial) (F) Quantification of 5-HT-span area density under the conditions of E. (G) qRT-PCR results for tph1a, tph1b, and tph2 in control, amyloid-injected, and IL4-injected zebrafish brains. Beta-actin used for normalization. n = 3 animals for experiments. Scale bars equal 100 μM. Data are represented as mean ± SEM. See also S1 Fig and S2 Fig. See S2 Data and S7 Data for supporting information. Aβ42, amyloid-beta42; DEG, differentially expressed gene; ECM, extracellular matrix; FC, fold change; FDR, false discovery rate; GFP, green fluorescent protein; GO, gene ontology; IHC, immunohistochemistry; KEGG, Kyoto Encyclopedia of Genes and Genomes; IL4, interleukin-4; MA, Bland-Altman mean-average plot; qRT-PCR, quantitative real-time polymerase chain reaction; 5-HT, serotonin.

https://doi.org/10.1371/journal.pbio.3000585.g001

Because Aβ42/IL4 enhances NSC proliferation in the adult zebrafish brain [29,30,32,33] and they down-regulate 5-HT (Fig 1E and 1F), we hypothesized that 5-HT could be negatively affecting NSC plasticity. To test this, we injected 5-HT into the adult zebrafish brain and analyzed the proliferation of NSCs (S100 calcium-binding protein B (S100β) and and proliferating cell nuclear antigen (PCNA) double positive cells) at 1 dpi (Fig 2A–2C). We observed a significant reduction (reduced by 32%) of NSC proliferation after 5-HT injection (Fig 2C). This reduction has implications in neurogenesis from NSCs, because the analysis of 5-bromo-2'-deoxyuridine (BrdU)-labeled newborn neurons at 14 dpi of PBS (Fig 2D) or 5-HT (Fig 2E) showed that 5-HT reduces the neurogenic outcome by 12% (Fig 2F). To verify our results, we used a transgenic reporter line that marks NSCs with GFP driven under the her4.1 promoter and injected Aβ42, IL4, or 5-HT followed by immunohistochemistry (IHC) for GFP and PCNA at 1 dpi (Fig 2G). Consistent with our previous results, Aβ42 and IL4 injection increased NSC proliferation by 35% and 42%, respectively, whereas 5-HT reduced the number of proliferating progenitors by 18% (Fig 2H), indicating that 5-HT has a negative impact on NSC plasticity in zebrafish telencephalon. To further test whether the negative effects of 5-HT would be reversed by Aβ42 or IL4, we co-injected 5-HT with Aβ42 and IL4 and observed that the reduction in NSC proliferation by 5-HT can be abrogated by Aβ42 or IL4 (S3 Fig), indicating that 5-HT signaling acts antagonistically to IL4 in NSCs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Serotonin regulates NSC plasticity indirectly through HTR1 signaling in periventiruclar neurons.

(A, B) IHC for S100β and PCNA in control (A) and 5-HT-injected (B) brains. (C) Quantification of proliferating glia in conditions of panels A and B. (D, E) IHC for BrdU and HuC/D for newborn neurons at 14 dpi after BrdU treatment at 2 and 3 dpi (D) and BDNF injection (E). (F) Quantification of newborn neurons. (G) IHC for GFP (driven by glial promoter her4.1) and PCNA in control, IL4-injected, Aβ42-injected, and 5-HT-injected brains. (H) Quantification of proliferating glia in conditions of panel G. (I-K) IHC for her4.1-driven GFP and 5-HT. The composite image (I) and single fluorescent channels for her4.1:GFP (J) and 5-HT (K). (L) ISH for htr1a (panel Lʹ: close-up image of framed region in panel L). (M) ISH for htr1d (panel Mʹ: close-up image of framed region in panel M). (N) IHC for pERK and her4.1-driven GFP in control and 5-HT-injected brains. (O) Quantification of pERK-positive periventricular neurons. (P) Working hypothesis on the indirect regulation of 5-HT on NSC plasticity. n = 4 animals for experiments. Scale bars equal 100 μM. Data are represented as mean ± SEM. See also S3 Fig. See S7 Data for supporting information. Aβ42, amyloid-beta42; BDNF, brain-derived neurotrophic factor; BrdU, 5-bromo-2'-deoxyuridine; dpi, days post injection; GFP, green fluorescent protein; HTR1, 5-hydroxytryptamine receptor 1; HuC/D, ELAV (embryonic lethal, abnormal vision, Drosophila)-like 3/4; IHC, immunohistochemistry; IL4, interleukin-4; ISH, in situ hybridization; NSC, neural stem cell; pERK,; phosphorylated extracellular signal-regulated kinase; PCNA, proliferating cell nuclear antigen; PVZ, periventricular zone; S100β, S100 calcium-binding protein B; 5-HT, serotonin.

https://doi.org/10.1371/journal.pbio.3000585.g002

Our previous results indicated that a subset of NSCs express IL4 receptor and can therefore be directly regulated by IL4 [24,30,34]. Therefore, we aimed to investigate whether 5-HT would affect the same subset of NSCs or whether it affects a distinct population. To determine whether or not 5-HT has a direct effect on NSCs, we determined the spatial organization of NSCs (her4.1:GFP, Fig 2I and 2J) and 5-HT innervation (Fig 2I and 2K). We observed that NSCs are located apically and are separated by periventricular zone (PVZ) neurons before the proximal front of the serotonergic innervation in the parenchyma (Fig 2I), suggesting that the effect of 5-HT on NSCs could be indirect. If this hypothesis was true, NSCs would not express 5-HT receptors. To determine which cells expressed 5-HT receptors, we performed in situ hybridization (ISH) for 5-HT receptors (htr genes) and observed that among 7 serotonin receptor genes and in total 21 isoforms of those genes (S2 Data), only htr1a and htr1d gave ISH signals in adult zebrafish telencephalon (Fig 2L–2Mʹ). We observed that serotonin receptor genes are not expressed in the progenitor cells (S3 Fig), supporting previous findings [35] and suggesting that serotonin does not act directly on NSCs. To further investigate this hypothesis, we performed electrophysiological recordings from 9 her4.1:GFP-positive NSCs and 7 adjacent her4.1:GFP-negative cells (neurons; S3 Fig). We found that out of 9 recorded NSCs (3 of them in the presence of tetrodotoxin [TTX]), none responded to serotonin (S3 Fig). For the neurons, we recorded 8 (2 in the presence of TTX), and 7 of them responded to serotonin by increasing the frequency of the recorded excitatory postsynaptic currents (EPSCs) (S3H Fig). This further confirms that serotonin is indeed affecting periventricular neurons but not NSCs in the adult zebrafish brain.

Among the serotonin receptor genes, htr1a was expressed only in ventral regions (Fig 2L and 2Lʹ), whereas htr1d was present in periventricular region immediately adjacent to the ventricular zone containing the NSCs that span the medial and dorsal regions of the telencephalon (Fig 2M and 2Mʹ). These findings suggested that the effect of 5-HT on NSCs is not direct, and it could be mediated through periventricular neurons expressing htr1d (Fig 2N). If 5-HT would act on periventricular neurons, 5-HT would activate its downstream effector phosphorylated extracellular signal-regulated kinase (pERK) [36] in periventricular regions. Indeed, we found that compared with controls in which nuclear pERK is in few cells at the periventricular zone (PVZ) (Fig 2O), 5-HT injection increases the number of pERK+ nuclei (by 37%) only in the PVZ region (Fig 2N). Quantification of pERK+ nuclei also confirmed that 5-HT increases pERK+ nuclei on the contrary to Aβ42/IL4 (Fig 2O), indicating that 5-HT directly affects periventricular neurons but not the NSCs (Fig 2P).

To determine the mechanistic link between 5-HT and NSC plasticity through hrt1+ cells (Fig 2P), we performed single-cell transcriptomics in control and 5-HT-treated adult zebrafish telencephalon by unbiased clustering, determination of cell types and differential gene expression analyses using the methodology we have recently developed [34] (S4A Fig; S3 Data). After quality control analyses (S4B Fig), we obtained t-distributed stochastic neighbor embedding (tSNE) clusters, which dissolve into 4 major cell types: neurons (eno2, gap43, map1aa, sypb-positive), glia (gfap, her4.1, msi1, s100b-positive), oligodendrocytes (aplnrb, olig2-positive), and immune cells (pfn, lcp1-positive; S4C–S4E Fig). Heat maps reveal marker genes expressed predominantly in those cells (S4D Fig). To determine the cells that are responsive to 5-HT, we plotted the cells expressing htr1 and observed that htr1 expression was confined to neurons (S4F Fig). This expression pattern was consistent with ISH results of htr1 genes (Fig 2M, S3G Fig) and was not overlapping with il4r.1-expressing cells that were exclusively glia and immune cells ([30] and S4F Fig).

To determine the mechanism by which 5-HT-responsive cells affect glia proliferation, we devised an analysis pipeline in which we dissected the htr1d+ cells from the rest and performed differential expression analysis between control and 5-HT-treated brains (Fig 3A; S4G Fig). We found that 3,166 genes changed their expression levels in htr1d+ cells after 5-HT treatment (Fig 3B, S4 Data). We hypothesized that a possible regulation between htr1d+ neurons and NSCs could be through a paracrine ligand–receptor crosstalk. We found that in htr1d+ cells, 40 ligands change their expression levels (Fig 3C), 3 of which also change their expression reciprocally after 5-HT and IL4/ Aβ42 treatment (Fig 3D, and data from [34]). The highest up-regulation was observed for bdnf (Fig 3D). From single-cell analyses, we found that bdnf was predominantly expressed in neuronal clusters (90% in neurons and 8% in glial cells; Fig 3E, S4H Fig), whereas its receptor ntrk2 [37,38] was expressed mainly in neurons (68% neuronal and 28% glia; Fig 3E, S4H Fig), and another BDNF receptor ngfra [37–39] was mostly glial (93% in glia and 5% in neurons; Fig 3E; S4H Fig). Therefore, we hypothesized that 5-HT dependent regulation of bdnf expression might signal to the glial cells through ntrk2 and ngfra in the adult zebrafish brain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Single-cell sequencing after serotonin treatment in the adult zebrafish brain and data analyses.

(A) Schematic workflow for single-cell sequencing and data analyses. (B) Distribution plot for DEGs in htr1d-expressing cells after 5-HT treatment. (C) Ligands selected from panel B. (D) Plots for ligands that change oppositely in 5-HT and IL4/Aβ42 treatment. (E) Feature plots for bdnf and its receptors ntrk2 and ngfra. (F) Spatial map of NSCs/PCs in the adult zebrafish brain as previously described in Cosacak et al., 2019. (G) In silico interaction map for BDNF in amyloid versus control, IL4 versus control, and 5-HT versus control comparisons. (H) In silico interaction map for NGFRA in Aβ42 versus control, IL4 versus control, and 5-HT versus control comparisons. In panels G and H, black arrows: interactions unchanged with treatment, cyan arrows: interaction lost with treatment, magenta arrows: interaction gained/emerged with the treatment. See also S4 Fig and S5 Fig. See S3 Data and S4 Data for supporting information. Aβ42, amyloid-beta42; BDNF, brain-derived neurotrophic factor; CVMI, cerebroventricular microinjection; DEG, differentially expressed gene; GFP, green fluorescent protein; IL4, interleukin-4; NGFRA, nerve growth factor receptor A; NSC, neural stem cell; PC, progenitor cell; 5-HT, serotonin.

https://doi.org/10.1371/journal.pbio.3000585.g003

To further investigate whether BDNF signaling through Ntrk2 or Ngfra would affect NSCs, we used an in silico interaction map analysis that we recently developed [34]. According to this analysis, if BDNF had a potential interaction between neuronal and glial (progenitor) clusters, we would see an in silico interaction between these cells. Alternatively, if ntrk2 or ngfra could constitute a crosstalk between neuronal and glial clusters, we would be able to see an interaction (see Cosacak and colleagues [34] for details of interaction mapping). For interaction analyses, we used the spatial organization map of adult zebrafish NSCs in the telencephalon (Fig 3F), distinguished the cell types based on our previous findings (Cosacak and colleagues [34]) by using machine learning algorithms, and compared the 3 treatments with controls (IL4 versus control, amyloid versus control, and 5-HT versus control; Fig 3G and 3H; S5 Fig). For mapping, we used the following 3 criteria: (1) we took either the bdnf as a starting point (i.e., the cells expressing bdnf are matched with the cells expressing any bdnf receptor, or the bdnf receptors ntrk2 and ngfra were taken as starting points [i.e., only ntrk2 or ngfra-expressing cells were matched with cells expressing any ligand for those receptors (Fig 3H for ngfra and S6A Fig for ntrk2)]); (2) charted the potential interactions as arrows starting from bdnf-expressing clusters to bdnf receptor-expressing clusters (Fig 3G) or to bdnf receptor-expressing clusters from clusters expressing their ligands (Fig 3H), and (3) depending on the change of the interaction after any treatment, we color-coded the interactions (black arrows indicate interactions that are unchanged by the treatment, cyan arrows indicate the interactions that are lost after treatment, and magenta arrows indicate interactions that emerged after a particular treatment). After these analyses, we found that especially 2 clusters of progenitor cells (PC0 and PC2, which are located to dorsal and medial region of the telencephalon; Fig 3F and Cosacak and colleagues [34]) were among the clusters that had the highest number of arrows pointing toward them (Fig 3G). Interestingly, the majority of the potential interactions were newly generated after amyloid or IL4 treatment, whereas those interactions are mainly lost after 5-HT treatment (Fig 3G, green circles), suggesting that the regulation of progenitor cells (NSCs) by bdnf could be promoted by amyloid or IL4 treatment and suppressed by 5-HT, which is consistent with our hypothesis. Additionally, to support our findings, we performed independent mapping analyses of bdnf receptors ntrk2 and ngfra. We constructed interaction maps for these receptors and found that ngfra might be the main receptor for bdnf signaling in adult zebrafish telencephalon because 2 NSC clusters (PC0 and PC2) were at the center of the interaction maps with many potential interactions (Fig 3H, S4 Fig), whereas the ntrk2 interaction map provided only a small number of interactions to another glial cluster (S4 Fig, S6A Fig). We observed that IL4 activates many interaction routes to NSC clusters whereas 5-HT almost diminishes all the interactions between NSCs and other cells (Fig 3H). These in silico analyses suggest that IL4 promotes the interactions between neurons and NSCs whereas 5-HT suppresses such interactions.

To verify our in silico analyses for interaction mapping on single-cell transcriptomics data and to investigate the changes in the expression of bdnf, we performed ISH on control (Fig 4A), 5-HT-injected (Fig 4B), and IL4-injected brains (Fig 4C). The expression of bdnf in control brains was in the PVZ proximal to the NSCs, which confirmed our single-cell transcriptomics data. The expression of bdnf was almost abolished by 5-HT injection (Fig 4B), whereas it is enhanced by IL4 (Fig 4C), suggesting that 5-HT suppresses NSC plasticity through reducing the bdnf signaling.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Serotonin regulates periventricular bdnf expression and NFkB signaling in NSCs.

(A–C) ISH for bdnf in control (A), 5-HT-injected (B), and IL4-injected (C) brains. Red rectangles are enlarged to the right of the main panels. Note significant reduction of bdnf expression after 5-HT. (D, Dʹ) IHC for PCNA and S100β in control (D) and BDNF-injected (Dʹ) brains. (E) Quantification of proliferating glial cells after BDNF injection. (F) IHC for BrdU and HuC/D in control and BDNF-injected brains. (G) Quantification of newborn neurons at 14 dpi after BrdU treatment at Day 2 and Day 3 after PBS and BDNF injection. (H) ISH for ntrk2, which is expressed in periventricular neurons but not in NSCs in the vz. (I) ISH for ngfra, which is expressed in vz. (J) IHC for S100β, NFkB-driven GFP, and PCNA in control brains. To the right of the larger panel are single-fluorescence channels. (K) High-magnification of the medial region of panel J without DAPI. (K1–K3) Single fluorescent channels of panel K. (L) IHC for S100β, NFkB-driven GFP, and PCNA in BDNF-injected brains. To the right of the larger panel are single-fluorescence channels. (M) High-magnification of the medial region of panel L without DAPI. (M1–M3) Single-fluorescence channels of panel M. Scale bars equal 100 μM. Data are represented as mean ± SEM. See also S4–S10 Figs. See S7 Data for supporting information. BDNF, brain-derived neurotrophic factor; BrdU, 5-bromo-2'-deoxyuridine; dpi, days post injection; GFP, green fluorescent protein; HuC/D, ELAV (embryonic lethal, abnormal vision, Drosophila)-like 3/4; IHC, immunohistochemistry; IL4, interleukin-4; ISH, in situ hybridization; NFkB, nuclear factor 'kappa-light-chain-enhancer' of activated B-cells; NSC, neural stem cell; PCNA, proliferating nuclear cell antigen; pvz, periventricular zone; S100β, S100 calcium-binding protein B; vz, ventricular zone; 5-HT, serotonin.

https://doi.org/10.1371/journal.pbio.3000585.g004

Given that 5-HT suppresses bdnf expression and NSC proliferation whereas IL4 enhances bdnf expression and NSC proliferation, we hypothesized that bdnf would also enhance NSC plasticity by increasing cell proliferation. To test this, we injected BDNF into adult zebrafish brains (Fig 4D and 4E) and observed that BDNF indeed increases the number of proliferating NSCs by 27% at 1 dpi (Fig 4E). The enhanced proliferation of NSCs by BDNF is also translated into increased neurogenesis because BDNF-injected brains produce more neurons (increase by 19%) compared with control-injected zebrafish brains (Fig 4F and 4G), indicating that BDNF enhances NSC plasticity in the adult zebrafish brain.

If BDNF would act directly on NSCs, its receptor must have been present in the target cells. Therefore, we performed ISH for ntrk2 and ngfra in adult zebrafish brains. We observed that ntrk2 mRNA is expressed mainly in the periventricular region of the telencephalon (Fig 4H and S6B Fig), which is supported by IHC staining for NTRK2 protein (S6C Fig). On the contrary, ngfra is expressed mainly in the ventricular zone where NSCs reside (Fig 4I, S6 Fig). These findings are perfectly matching with our single-cell transcriptomics data (Fig 3E, S4H Fig) and suggest that ngfra is the primary receptor for bdnf in NSCs and ntrk2 is the primary receptor in neurons. To test this hypothesis, we injected BDNF into the adult zebrafish brain and determined the downstream effectors of NTRK2 and NGFRA signaling pathways. To determine the cells responding to BDNF/NTRK2 signaling, we detected the downstream effector phosphorylated protein kinase B (pAkt) [36] and observed that after BDNF injection, pAkt is almost exclusively present in periventricular cells and parenchymal neurons (few speckles in ventricular region constitute less than 0.1% of the pAkt signal when compared with the intensity and number of cells outside the ventricular region; S6D and S6E Fig). On the contrary, compared with control-injected brains, BDNF injection increased the activity of the NFkB reporter [40], which is the downstream effector of NGFR signaling [41] in the ventricular region where NSCs reside (Fig 4J–4M3). Additionally, consistent with our previous result, we observed that at 1 dpi, NFkB signaling and the number of NFkB-positive proliferating NSCs (S100β /PCNA/NFkB:GFP-triple positive cells) increased by 47% after BDNF injection, 28% after Aβ42 injection, and 25% after IL4 injection but reduced by 10% after 5-HT injection (S6 Fig). These results indicate that NGFR-mediated intracellular signaling is the primary route for the 5-HT-dependent BDNF activity on NSCs in the adult zebrafish brain.

If our hypothesis was true and BDNF would regulate NSC proliferation through NGFRA and NFkB signaling, knocking-down the ngfra receptor after BDNF injection would suppress the increase in NSC proliferation by BDNF as well as the increase in NFkB signaling. To test this hypothesis, we knocked down ngfra by using morpholino oligonucleotides and determined the extent of proliferating NSCs (Fig 5A–5A5). Compared with control brains, BDNF increased the NSC proliferation by 38%; control morpholino did not alter this increase (remained 37% increase) whereas ngfra morpholino diminished (reduced down to 13%) the increased NSC proliferation upon BDNF injection (Fig 5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Ngfra signaling regulates NSC plasticity independent of Il4ra signaling.

(A–A5) IHC for S100β and PCNA in control (A), control morpholino-injected (A1), ngfra morpholino-injected (A2), BDNF-injected (A3), BDNF + control morpholino-injected (A4), and BDNF + ngfra morpholino-injected (A5) brains. (B) Quantification for the relative number of proliferating glial cells. (C) Co-representation of ngfra and ilr4 expressions on glial tSNE plot. (D) DEG plots for ngfra-positive and il4r-positive neural stem cells after IL4 treatment. (E) Pie chart distribution of unique DEGs. Note that the overlapping genes constitute only 13.6% of all DEGs. (F, G) IHC for S100β, NFkB-driven GFP, and PCNA in BDNF + control morpholino-injected (F–F3) and BDNF + ngfra morpholino-injected brains (G–G3).(H) Quantification graph for relative numbers of proliferating stem cells with active NFkB signaling. Scale bars equal 100 μM. Data are represented as mean ± SEM. See also S10 Fig. See S5 Data, S6 Data, and S7 Data for supporting information. BDNF, brain-derived neurotrophic factor; DEG, differentially expressed gene; GFP, green fluorescent protein; IHC, immunohistochemistry; IL4, interleukin-4; NFkB, nuclear factor 'kappa-light-chain-enhancer' of activated B-cells; Ngfr, nerve growth factor receptor; NSC, neural stem cell; PCNA, proliferating nuclear cell antigen; S100β, S100 calcium-binding protein B; tSNE, t-Distributed Stochastic Neighbor Embedding.

https://doi.org/10.1371/journal.pbio.3000585.g005

Because BDNF/NGFRA signaling is affected by IL4 that also acts directly on NSCs through IL4R [30], we hypothesized that the effect of IL4 on NSCs could be mediated through IL4R (directly by IL4/IL4R interaction) and NGFR (through 5-HT/BDNF/NGFRA/NFkB axis) distinctly, and IL4R-positive and NGFRA-positive glia would constitute 2 functional subtypes of NSCs. To determine whether IL4R-positive and NGFRA-positive NSCs are distinct subtypes of NSCs in adult zebrafish brain, we plotted both cell populations on the same tSNE plot (Fig 5C). We observed that only 3.2% of ngfra-positive NSCs were also il4r1-positive (24.5% are only il4r-positive and 72.3% are only ngfra-positive), indicating that these 2 populations are likely to represent 2 functional subtypes of NSCs. We further hypothesized that if ngfra-positive and il4r-positive NSCs would constitute different subtypes, their response to particular treatments would also lead to distinct differential gene expression profiles. To address this question, we determined the overlap between differentially expressed genes in il4r-positive and ngfra-positive NSCs (Fig 5D). We determined that, after IL4 treatment, 423 and 895 genes are differentially expressed in il4r+ (S5 Data) and ngfra+ (S6 Data) NSCs, respectively (Fig 5D). Only 14% of the DEGs are common in these cell populations whereas the rest are differentially expressed in only one of the cell types (Fig 5E). Similarly, after BDNF treatment, ngfra knockdown reduced the number of proliferating NSCs where NFkB signaling is active (Fig 5F–5H). These results indicate that neuron-glia interaction through BDNF regulates NSC proliferation and neurogenesis through NGFRA/NFkB signaling (Fig 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Schematic view of the findings on how neuron-glia crosstalk regulates Alzheimer-induced neurogenesis in adult zebrafish brain.

Aβ42, amyloid-beta42; bdnf, brain-derived neurotrophic factor; htr1, 5-hydroxytryptamine receptor 1; IL4, interleukin-4; NFkB, nuclear factor 'kappa-light-chain-enhancer' of activated B-cells; NGFRA, nerve growth factor receptor A; NSC, neural stem cell; PCNA, proliferating cell nuclear antigen; pERK, hosphorylated extracellular signal-regulated kinase; pSTAT6, phosphorylated signal transducer and activator of transcription 6, interleukin-4 induced; PVZ, periventricular zone; ngfra, nerve growth factor receptor A; Tph2, tryptophan hydroxylase; 5-HT, serotonin.

https://doi.org/10.1371/journal.pbio.3000585.g006

To explore the evolutionary conservation of our findings in healthy and AD conditions, we determined the expression of BDNF, NTRK2, and p75/NTR (NGFRA ortholog in mouse) in wild-type mouse and APP/PS1dE9 AD model (S7–S9 Figs). Compared with 12-month-old control mouse brains, age-matched APP/PS1dE9 mouse displayed reduced SOX2 (neurogenic competency and NSC maintenance marker [42,43]) and increased GFAP (glial marker [44,45]; S7 Fig, S8 Fig, S9 Fig) that is indicative of reduced neurogenic ability and increased gliosis. We found that in mouse cortex and dentate gyrus (DG), BDNF is mainly expressed by nonglial cells (S7 Fig, S8 Fig), which is supported by previous studies [46,47], and this pattern is not altered in AD brains (S7 Fig, S8 Fig).

NTRK2 is expressed in the cortex and DG, again mainly in nonglial cells but few GFAP-positive astrocytes were NTRK2-positive in both the wild-type and AD mouse brains with no clear change in the expression pattern between healthy and diseased brains (S7 Fig). Overall, NTRK2 is expressed mainly by neurons but also few astrocytes and microglia (S8 Fig). We found that p75/NTR is mainly expressed in neurons in the cortex and the DG, but NSC niche in the DG (subgranular zone[(SGZ)]) does not express p75/NTR (S7 Fig, S9 Fig). In both wild-type and APP/PS1dE9 mouse, we could detect BDNF, NTRK2, and p75/NTR expression; however, increased number of GFAP-positive astrocytes AD brains did not correlate with the expression of these proteins, suggesting that BDNF signaling might not regulate NSC proliferation in mouse brains.

To test this hypothesis, we investigated the effect of BDNF in wild-type and AD mouse brains by injecting BDNF into the mouse brains (S7 Fig, S10 Fig). BDNF injection was performed in one hemisphere of the mouse brain, whereas the other hemisphere was used as a control with PBS injection. We found that BDNF injection increased the overall proliferation levels in the brain compared with PBS-injected hemispheres in both wild-type and APP/PS1dE9 brains (S7 Fig). In addition, we also checked the cell types that increase their proliferation levels.

To identify the BDNF-responsive proliferating cells, we performed co-staining with astrocyte marker GFAP and found that the cells positive for Ki67 (antigen identified by monoclonal antibody Ki-67, a marker for proliferating cells) are GFAP negative, suggesting that BDNF-responsive proliferating cells are not astrocytes (S7 Fig, S10 Fig). Next, we performed co-staining of Ki67 with microglial marker Iba1 (ionized calcium binding adaptor molecule 1) and found that almost all Ki67-positive cells are Iba1-positive, suggesting that BDNF induces microglial proliferation resulting in microgliosis but not NSC proliferation (S7 Fig, S10 Fig). This finding is consistent with previous reports [7,48,49]. Additionally, because of the lack of p75/NTR expression in NSCs present in the SGZ of the mouse hippocampus, BDNF/p75NTR signaling, which enhances proliferative output of NSCs and neurogenesis in zebrafish brain, is not an active signaling mechanism in mouse brains. With these results, we propose that zebrafish utilizes special circuit mechanism that uses serotonin-BDNF signaling to enhance the NSC plasticity and to induce neurogenesis through neuronal intermediates; however, BDNF signaling is not regulating NSC plasticity and neurogenesis in mammalian brains whereas its effect is mainly on the neuronal survival and regulation of immune reaction [36,41,49–52].

Discussion

Our results identify a previously uncharacterized regulatory circuit mechanism that involves neuronal intermediates for NSC plasticity of zebrafish brain in AD conditions. This mechanism involves the suppressive effect of serotonin on periventricular neurons that express bdnf, which directly regulates a subset of NSCs expressing BDNF receptor ngfra. amyloid-induced IL4 suppresses the initial serotonergic output and therefore potentiates the bdnf expression and stem-cell proliferation. We propose that induced expression of IL4 generates a disease-associated NSC niche environment that is supportive of NSC proliferation.

Serotonin affects neurogenesis, stem-cell proliferation, and normal functioning of neural circuitry; however, its effects on NSCs and neurogenesis are controversial [53]. In zebrafish, serotonin positively affects NSC proliferation in midbrain but not the hypothalamus [54] and promotes spinal motor neuron regeneration [55]. Although depletion of serotonin reduces adult neurogenic outcome in rats [56], elevated levels of serotonin in serotonin transporter (SERT)-deficient mice did not affect NSC proliferation and neurogenesis [57,58]. In mesencephalic and hippocampal progenitors of mouse brain, inhibition of serotonin receptors or knocking out serotonin-producing enzymes to generate hyposerotonergic phenotypes had positive effects on neurogenesis [59–62] ascertaining a negative role for serotonin on NSC plasticity. These findings indicate that the effects of serotonin on NSCs are context dependent and may run through intermediary secondary signaling mechanisms, which supports our findings. Additionally, in the majority of the studies, the specific receptors for serotonin were not investigated in NSCs; therefore, our study provides a detailed delineation of the signaling cascades that link serotonergic input to the regulation of stem cell proliferation via intermediates in our study by BDNF.

The interaction between serotonin and BDNF is partially understood. In stress-related disorders, depletion of serotonin reduces brain BDNF levels [63–65]. On the contrary, hyposerotonergic mouse (Tph2−/− or Pet1−/−) showed increased BDNF levels in the hippocampus and prefrontal cortex [60–62,66], and humans with SERT deficiency (hyperserotonin) reduced the availability of BDNF [67], which enhances NSCs proliferation and neurogenesis [68,69]. Our findings support the negative role of serotonin on BDNF signaling and NSC plasticity.

The effects of BDNF in AD are mainly on neuronal survival rather than neurogenesis or NSC plasticity. Although BDNF expression correlates with increased neuronal survival [70–72] and may have positive impact on increased net hippocampal neurogenesis and better cognitive functioning [73,74], several studies showed that these effects are not directly related to enhanced neurogenesis [48,49]. In fact, in a recent study, in 5X-FAD AD model of mice, increasing adult neurogenesis and simultaneous BDNF treatment increases cognitive output yet adult neurogenesis or BDNF alone are not sufficient to do so [7], suggesting that a BDNF-dependent neuronal survival cascade is required to counteract the symptoms of AD after an independent stimulation of neurogenesis. In zebrafish, in contrast to mammals, BDNF directly regulates NSC plasticity and neurogenesis. Our results in APP/PS1dE9 mice also supported the findings in mice [7] and indicate that zebrafish has an intrinsic and natural ability to kick-start NSC proliferation and neurogenesis, which could teach us how to modulate mammalian brains to initiate “regeneration” of lost neurons in AD.

BDNF expression is abundant in mammalian brains [75], and its main receptor is TrkB/Ntrk2, which is predominantly expressed in neurons [76] in the cortex and DG as well as in few astrocytes and microglia as determined by single-cell sequencing [46,47]. BDNF also has binding affinity towards p75 receptor, which is primarily expressed by neurons in mammals [46,47]. Yet, in zebrafish, ngfra was predominantly expressed in NSCs and were responsible for NSC plasticity. Zebrafish glial cells responded to BDNF via ngfra receptor and activated proliferation and neurogenesis; whereas, mouse glial cells responded to BDNF by reactive astrogliosis and microgliosis. Therefore, our results also propose that the BDNF activity could be specialized to distinct receptor signaling pathways in vertebrate brains. For instance, BDNF-NGFRA signaling could be a factor underlying the regenerative and plastic nature of the NSCs in regenerating organisms such as zebrafish but not in nonregenerating organisms like mammals, in which BDNF mostly acts through TrkB/Ntrk2 signaling. Further studies of this regulatory mechanism in other regenerating organisms would be instrumental in testing our hypothesis. Furthermore, inducing expression of ngfra in stem-cell populations in mammalian brains could be a potential way to impose neurogenic plasticity to astrocytes in AD conditions.

We found that IL4 suppresses the serotonin production in neurons that do not express il4r, suggesting that IL4 regulates serotonin availability indirectly. Indeed, proinflammatory cytokines such as IL-1β, TNFα, IFNα, and IFNγ can increase 5-HT levels and the expression of 5-HT transporter SERT, which positively correlates with 5-HT levels in the brain [77–83]. Theoretically, anti-inflammatory factors would be expected to reduce 5-HT levels in the brain because of their role in suppression of proinflammatory factor expression. Indeed, IL4 reduces the 5-HT or SERT levels [84–87], supporting our findings. Therefore, in the adult zebrafish brain, AD conditions elicit an inflammatory regulation by IL4 on serotonin, which translates into a neurotransmitter response in a specific neuronal subtype that directly regulates a subpopulation of NSCs.

We identified 2 functionally distinct populations of NSCs that are both responsive to IL4 but through different receptor signaling (IL4R/STAT6 and BDNF/NGFRA) and through distinct gene regulation (Fig 5D and 5E). This finding suggests that NSC heterogeneity is an important determinant of how a certain type of neuropathology and a particular signaling pathway would affect certain subtypes of cells differentially (e.g., in AD conditions, the regulation of proliferation by IL4 in 2 distinct modes [Fig 6]). This understanding would be instrumental in designing targeted therapy options for AD in humans and may lead to more detailed analyses of NSC subtypes in experimental mammalian models of AD. Additionally, the role of IL4 and serotonin in regulating diverse signaling mechanisms in NSCs opens up interesting research avenues as to whether modulation of neuroinflammatory milieu in mammalian brains would kick-start a “regeneration” response by mobilizing the endogenous reservoir of NSCs, which is a controversial subject [8,12–14,23,88,89]. We propose that zebrafish can be used to address neurogenesis-related questions in disease conditions and could serve as a useful experimental model for investigating the molecular mechanisms of NSC plasticity.

Materials and methods

Supporting information

Acknowledgments

We would like to thank to John F. Rawls and Steven Renshaw for NFkB:GFP reporter zebrafish line; Tomohisa Toda, Gerd Kempermann, Emre Yaksi, and Nathalie Jurisch-Yaksi for critical comments on the manuscript; Andreas Dahl, Susanne Reinhardt, and Andreas Petzold for single-cell sequencing.