Proteomic-based evidence for adult neurogenesis in birds and mammals as indicated from cerebrospinal fluid.

Adult neurogenesis is the life-long process of neural stem cell proliferation, differentiation into neurons, migration, and incorporation into the existing neuronal circuits. After decades of research, it is now widely accepted that mammals and birds retain the capacity to regenerate neurons even after their subadult ontogeny. Cerebrospinal fluid participates in the regulation of the neurogenic niches of the vertebrate brain through signaling pathways not fully elucidated. Proteomic studies of cerebrospinal fluid have the potential to allow the in-depth characterization of its molecular composition. Comparative studies help to delineate those pathways that are universally critical for the regulation of neurogenesis in adulthood. In this review, we performed literature-based data mining in studies using liquid chromatography-tandem mass spectroscopy that analyzed cerebrospinal fluid samples from healthy adult humans (Homo sapiens); mice (Mus musculus); sheep (Ovis aries); chickens (Gallus gallus); and two parrot species, the budgerigar (Melopsittacus undulatus) and cockatiel (Nymphicus hollandicus). We identified up to 911 proteins represented in cerebrospinal fluid, involved in various pathways regulating adult neurogenesis. However, only 196 proteins were common across humans, mice, and birds. Pathway components involved in nervous system development, cell migration, and axonal guidance were commonly evident in all species investigated so far. Extensive bioinformatic analysis revealed that the universally over-represented pathways involved L1 cell adhesion molecule protein interactions, cell-adhesion molecules, signals regulating extracellular matrix remodeling, regulation of insulin growth factor signaling, axonal guidance, programmed cell death, immune signaling, and post-translational modifications. Most of the reported proteins are part of extracellular vesicles enriched in cerebrospinal fluid. However, the information presently available is still highly fragmentary, and far more questions persist than are answered. Technological advances will allow cerebrospinal fluid comparative proteomic research to delve into the fundamental processes of adult neurogenesis and eventually translate this research into any regenerative interventions.

Introduction

The central nervous system communicates dynamically with all tissues through the cerebrospinal fluid (CSF) and choroid plexus system (Illes, 2017). Considered the third circulation, CSF is a colorless fluid ensuing mechanical protection, waste products elimination from the central nervous system, homeostatic and nutrient support and it is the secretory product of the four choroid plexuses located within each of the brain ventricles (Spector et al., 2015). The choroid plexuses are highly vascularized structures and act as sensors, integrating physiological signals (Lun et al., 2015). The CSF composition derives from the ultrafiltration of plasma in choroid plexuses at approximately 80%. The rest originates from the interstitial fluid of the brain (Reiber, 2001). Once produced, CSF steadily circulates throughout the central nervous system (Lun et al., 2015). During development, CSF provides active signaling cues orchestrating brain morphogenesis (Zappaterra and Lehtinen, 2012; Fame and Lehtinen, 2020; Gato et al., 2020) and, although its composition changes over the lifespan, CSF maintains its role as a dynamic regulator of the activity of cell proliferating areas through its derived factors (Gato et al., 2020). Neurogenic loci at the subventricular zone (SVZ) are in direct contact with the CSF-choroid plexus system (Lindsey and Tropepe, 2006). Neurogenesis and gliogenesis are the transitions from proliferative and multipotent neural stem cells (NSCs) to fully differentiated neurons and glia, respectively. These processes occur throughout a lifetime, following a specific cellular dynamic regulated by a wide range of intrinsic and extrinsic factors (Obernier and Alvarez-Buylla, 2019). In the SVZ, NSCs interact with the brain ventricular system and CSF through an apical process containing a primary cilium that mediates signal transduction (Tong et al., 2014; Obernier and Alvarez-Buylla, 2019). Ablation or dysfunction of these processes leads to disruption of Hedgehog signaling and decreased neurogenesis in the ventral SVZ (Tong et al., 2014). Thus, the anatomical relationship between choroid plexus-CSF and neurogenic niches in the brain indicates their interaction as well as its promoting effect on neural development, adult neurogenesis, and nervous system plasticity.

The advent and continuous improvement of proteomic technologies have facilitated a recent outburst of studies extending our understanding of CSF composition. Presently, the CSF proteome has been characterized only in a few species. In humans, different age groups were investigated mostly in the context of the discovery of biomarkers for the emergence of neurodegenerative diseases. Comparing multiple species would allow a deeper understanding of the evolution of adult neurogenesis and provide an insight into its basic biological mechanisms. For example, comparative studies between birds and mammals have advanced our knowledge on adult neurogenesis. Such studies elucidated the site of neuronal generation and stem cell characteristics, the factors controlling the proliferation of the precursors, and the migration of the differentiated neurons (Doetsch and Scharff, 2001). Birds are good models to study adult neurogenesis due to naturally occurring memory- and learning-related behaviors such as song and vocalization learning, food caching, spatial memory, and seasonality in neuronal turnover (Barnea and Pravosudov, 2011). However, so far, no thoroughly comparative research in vertebrate CSF proteomics has been conducted and a comprehensive description of avian and mammalian CSF factors implicating adult neurogenesis is lacking. In this study, we survey the current evidence on CSF-represented proteins involved in neurogenesis, detected in different adult avian and mammalian CSF proteomes while sparing information on their expression. Applying comparative bioinformatic approaches on literature-reported datasets, we address the degree of interspecific consistency in CSF proteomic composition, highlighting the key role of components involved in adult neurogenesis. Furthermore, we discuss the current state of knowledge on the major signaling pathways related to adult neurogenesis.

Data and Methods

Proteomic dataset search strategy

A literature survey was performed through PubMed and Web of Science databases, searching for peer-reviewed articles in the English language and published in the last 5 years, to cover studies involving the latest proteomic methods and instrumentation. We included only studies performed in healthy adult homoiothermic amniotes, i.e., birds and mammals including humans), reporting the identified proteins from the whole CSF proteome. Using the keywords [(cerebrospinal fluid OR CSF) AND (proteom*) AND (Bird* OR Avian OR Mammal* OR Human*) AND (mass spectrometry OR MS/MS)] 234 non-duplicate studies were found (date of the last search: 14/5/2021). The articles were selected for further evaluation based on abstract analysis. Restrictions based on methodological approaches were not applied, but only publications reporting identified CSF proteins based on at least two matching peptides were included.

Data extraction

In total, we selected 18 studies for data mining (), including the previous work from the authors on three bird species (Voukali et al., 2021). Details on the characteristics of the included studies are enclosed in . We focused on studies published during the last five years to comply with the latest instrumentation technologies. Two unpublished new datasets of budgerigar CSF proteome completed the retrospective data. We extracted the overall output list of the liquid chromatography-tandem mass spectroscopy identified CSF proteins from each proteomic study. Most studies (n = 12) investigated the human CSF proteome. Other species represented mouse (n = 5) and sheep (n = 1). Some human studies investigated biomarkers comparing the CSF proteome of healthy individuals to patients with various conditions. These included neuronal ceroid lipofuscinosis, amyotrophic lateral sclerosis, Alzheimer’s disease, Parkinson’s disease, or unipolar major depressive disorder (Sleat et al., 2017; Bereman et al., 2018; Sathe et al., 2018; Thompson et al., 2018; Dayon et al., 2019; Johnson et al., 2020; Rotunno et al., 2020; Wang et al., 2020). From these studies, we extracted only the data for the healthy control subjects. Three studies reported comprehensive descriptive data on CSF proteome and included only healthy participants (Schilde et al., 2018; Zhao et al., 2018; Macron et al., 2020). In total, data from more than 340 healthy subjects were included (ages 19–90+). Three mouse studies included expression data from various animal models of degenerative diseases compared to wild-type animals (Pigoni et al., 2016; Hosp et al., 2017; Sleat et al., 2019; Wang et al., 2020) and one study reported data from wild-type healthy mice (Tüshaus et al., 2020). Tüshaus et al. (2020) used both data-dependent acquisition and data-independent acquisition methods of analysis. Both datasets were used herein because the protein lists were not completely overlapping. Equally to what we applied to human studies, only data from wild-type animals were extracted from a total of 34 mice (4–52 weeks of age). Most studies used the C57BL/6NCrl mouse strain. Sheep adult CSF proteome data were extracted from 21 animals of various ages (Chen et al., 2018). Bird studies involved 7 chickens, 35 budgerigars, and 5 cockatiels (adult animals, unspecified age) (Voukali et al., 2021, unpublished data). In sum, 24 datasets from more than 442 individuals from six species, three mammalian and three avians, were subsequently analyzed ().

Flow diagram for article search.

Electronic search recovered 234 articles after exclusion of duplicates. After selection criteria were implemented, 18 studies, yielding 22 datasets and additional 2 unpublished datasets from the authors summed to 24 datasets included in the final analysis. CSF: Cerebrospinal fluid.

The total reference gene products annotated with the gene ontology (GO) term Neurogenesis (GO:0022008) were retrieved from the Gene Ontology database (http://amigo.geneontology.org, Ashburner et al., 2000). The list of these proteins is included in .

Enrichment analysis of cerebrospinal fluid proteins with respect to neurogenesis-related genes

The statistical analysis was carried out in the R software version 4.0.3 (R Core Team, 2021) unless otherwise specified. Downstream bioinformatic analysis initially included over-representation analysis of GO Biological Process in DAVID (Ashburner et al., 2000; Huang et al., 2009) to address which pathways are enriched in the CSF. We used the reported identified proteins in the CSF from each dataset as the target and the whole Homo sapiens genome as the background. From the resulting enriched pathways, all gene IDs annotated with the GO term Neurogenesis (GO:0022008), thereby called neurogenesis-related proteins (NG) were identified. Then, we estimated the NG ratio to the total CSF identified proteins for each dataset.

We constructed a generalized linear mixed model with binomial data distribution to test whether the number of NG differed between taxa. The taxon (human, mouse, sheep, or bird) was included as a fixed effect, while the study identity was included as a variable with the random intercept. This model was compared with the null model based on the change of deviance with an accompanied change in degrees of freedom (analysis of variance) using Chi2 statistics. Additionally, to examine whether the ratio between NG and other proteins was associated with the total number of CSF identified proteins identifications, we performed Spearman’s correlation. Finally, to check if the NG were more consistently represented across the studies than other protein classes, we compared the ratios of the proteins consistently reported across all studies to all proteins identified in any study between the NG and other proteins. For this test, we adopted the Wilcoxon Rank paired test. Venn diagrams were contracted using the online tool BioVenn (Hulsen et al., 2008).

To identify which sub pathways are implicated in NG protein groups, over-representation analysis was performed on the subset pool of NG proteins consistent to birds, humans, and mice as target and all the retrieved NG (total NG) as background, using the R package clusterProfiler and GO, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome databases (Ashburner et al., 2000; Yu et al., 2012; Kanehisa et al., 2017; Fabregat et al., 2018). We also repeated these analyses for each species separately. For all pathway enrichment analyses, the terms with both p-value and false discovery rate lower than 0.05 were considered as significantly enriched.

Results

Overall patterns of the amniote CSF proteome composition

Our final dataset comprised of proteomic studies in humans, mice, sheep, and birds. We found variable numbers of identified proteins between the studies. The median number of proteins identified across the studies was 1030.5 (range 341–5941), 1058.5 (range 778–4688), 483, 1337 (range 622–1734), 452 for humans, mice, chickens, budgerigars, and cockatiels respectively. The least reported identifications were for sheep (264 proteins).

Different sample preparation approaches and detection techniques are probably accountable for the differences in the number of protein identifications between the studies. For example, a study using Tandem-mass-tag labeling and fragmentation at triple-stage mass spectrometry (MS3) level reported more identifications than other approaches (Sleat et al., 2019). The introduction of the MS3 method and extensive LC fractionation facilitates deep proteomic coverage (Ting et al., 2011). In addition, some studies used immunoaffinity-based depletion of the most abundant proteins, but some did not. Although antibody-based depletion may induce experimental variations, this method is often utilized to enhance the detection of low-abundance proteins (Tu et al., 2010). This is particularly relevant in biological fluids, where for example, the high concentrations of albumin may mask other proteins such as cytokines present in minimal quantities in healthy individuals (Geyer et al., 2017).

Enrichment analysis performed on the complete CSF proteome from each dataset revealed pathways implicated in various functions including metabolism, immune response, and signaling (). In specific, the enrichment analysis consistently revealed that GO terms related to neurogenesis, nervous system development, axon development, neuron projection development, and cell migration are among the most highly statistically significant and frequently represented terms across all the species and datasets ().

Gene Ontology (GO) Biological Process terms significantly enriched in all cerebrospinal fluid (CSF) proteomes of human, mouse, sheep, chicken, budgerigar, and cockatiel following over-representation analysis.

The Vonoroi tree map shows the common GO terms across all the 24 datasets included in this study, clustered according to their biological relevance in different colors. Each leaf corresponds to a GO term. Most GO terms concern functions related to nervous system development (0022604, regulation of cell morphogenesis; 0010769, regulation of cell morphogenesis involved in differentiation; 1990748, cellular detoxification; 0051960, regulation of nervous system development; 0007399, nervous system development; 0007417, central nervous system development; 0048666, neuron development; 0031175, neuron projection development; 0031102, neuron projection regeneration; 0030182, neuron differentiation; 0048812, neuron projection morphogenesis; 0010647, positive regulation of cell communication; 0097485, neuron projection guidance; 0010720, positive regulation of cell development; 0051050, positive regulation of transport; 0007596, blood coagulation; 0000902, cell morphogenesis; 0016477, cell migration; 0000904, cell morphogenesis involved in differentiation; 0051604, protein maturation; 0048667, cell morphogenesis involved in neuron differentiation; 0048468, cell development; 0061564, axon development; 0007411, axon guidance; 0002253, activation of immune response; 0010001, glial cell differentiation; 0022008, neurogenesis; 0050767, regulation of neurogenesis; 0045664, regulation of neuron differentiation). All shown pathways are significantly over-represented (P < 0.05, false discovery rate < 0.05).

The NG proteins in individual datasets and their ratio to total CSF identifications are shown in .

We aimed to address the possible differences in the representation of NG-identified proteins among taxa. Our comparative analysis has shown that human, mouse, sheep, and avian (chicken, budgerigar, and cockatiel) proteomes were similar regarding their NG content (generalized linear mixed model: Chi3,5 = 3.96, P > 0.05; ). Interestingly, the chicken was similar to the sheep based on the single observations available. In addition, we tested for the effect of the total identified proteins on the number of identified NG and found a significant negative linear association (Spearman’s rank correlation: r = –0.463, S22 = 3365.20, P = 0.023, ). Therefore, studies with a higher number of identified proteins contained a lower ratio of NG. This is evident especially in human samples from studies utilizing techniques allowing a higher protein coverage (for example Johnson et al., 2020; Wang et al., 2020), in contrast to those reporting a restricted number of CSF total proteins (for example Bereman et al., 2018; Franzen et al., 2020; Rotunno et al., 2020). This negative relationship suggests that regardless of the species the NG could be more consistently represented than other protein classes even in studies with low total numbers of identified proteins. To test this possibility, we compared the ratios of the proteins consistently reported across all studies to all proteins identified in each study between the NG and the proteins annotated otherwise. Given the effect of the total number of identified proteins on the frequency of identified NG, we excluded from further analysis all data sets with less than 400 protein identifications, where the total protein richness is limited (Bereman et al., 2018; Chen et al., 2018; Franzen et al., 2020; Rotunno et al., 2020). We retained 20 datasets overlapping in 38 consistently represented gene products. Out of these, 12 were NG, and 26 were proteins annotated otherwise. Our test confirmed that the NG- are, indeed, significantly more consistently represented in homoiothermic amniote CSF than other proteins (Wilcoxon Rank paired test: V = 210, P < 0.001; ). This expressional consistency indicates that the NG are a constitutively important part of CSF. Other functionally relevant protein subsets apparently vary more profoundly between individuals, species, and studies. Finally, we compared the overlaps in all reported NG and other proteins for human, mouse, and avian proteomes. The pools of the NG from humans, mice, and the bird species overlapped at 21.5% (196 proteins) for the NG and 15.7% (1251 proteins) for the other proteins (). Overall, up to 911 NG were reported to be expressed in the CSF from the species analyzed herein. The detailed lists of these pools are in . These findings suggest that despite the interspecific variation, all homoiothermic amniote taxa share in their adult CSF proteome an important core fraction of NG.

Proportions and overlaps of neurogenesis-related proteins (NG) in the cerebrospinal fluid (CSF) of healthy adult birds, human, mouse, and sheep compared to other CSF proteins (Others).

(A) Comparison of the proportions of the number of protein identifications annotated with the Gene Ontology term Neurogenesis (GO:0022008, NG) to the total number of the CSF protein identifications (CSF total) per taxon (shown in different colors). (B) Significant negative linear relationship between the total number of CSF protein identifications and the proportion of NG to CSF tot (species indicated by different colors; Spearman’s rank correlation, P = 0.023). (C) Comparison of the ratios of proteins consistently reported across all studies with more than 400 protein identifications to all proteins identified in each particular study (Consistency ratio), between the NG proteins and Others, showing higher proportion of the consistently represented proteins among the NG proteins. The two asterisks indicate high statistical significance of the difference (Wilcoxon ranked paired test, ***P < 0.001). (D) Venn diagrams depicting the overlaps of protein identifications in the proteome of birds (blue), mice (green) and humans (red) for NG proteins and others. Unpublished data.

Next, we addressed which sub pathways are enriched among the 196 proteins overlapping across the taxon-specific pools of the NG. Over-representation analysis of GO Biological Process revealed 71 enriched pathways including cell component assembly, regulation of cell morphogenesis, developmental cell growth, synaptic signaling (). Most proteins were part of extracellular vesicle, exosome, and post-synapse signaling as found from GO Cell Component analysis (). Analysis of GO Molecular Function showed classification based on binding to cytoskeletal components, signal transduction, and cell-adhesion proteins (). Enrichment analysis on signaling pathways of KEGG and Reactome supported our results of the GO over-representation analysis. The KEGG pathway analysis revealed over-representation of cell-adhesion molecule pathways (P < 0.001, q < 0.001). The proteins implicated are shown in red in . Furthermore, Reactome pathway analysis revealed 18 enriched pathways related to cell-adhesion, insulin growth factor (IGF) signaling, axon guidance, immune system, cell death, and post-translational protein phosphorylation (, ). Although not fully identical, similar results were found for all species when examined separately. All pathway analysis results are detailed in .

Results of the Gene Ontology (GO) enrichment analysis of the commonly represented cerebrospinal fluid (CSF) gene products involved in neurogenesis for Biological Process (BP), Cellular Component (CC) and Molecular Function (MF).

(A) A top 30 functional enrichment analysis results for BP. Dots represent term enrichment with color coding scaled on the degree of statistical significance. The sizes of dots represent the number of proteins identified and the gene ratio refers to the proportion of the proteins implicated in the pathway out of the overall number of the analyzed proteins. (B) Network plot showing the linkages of the top 5 categories of CC enriched terms and the gene products. The size of the dot corresponding to each term represents their number of protein content. (C) Plot depicting the relationships of MF significant terms with the identified proteins. The proteins are labeled by their gene symbols. Enriched pathways shown were statistically significant at P < 0.05, false discovery rate < 0.05.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of the gene products involved in neurogenesis commonly identified in the cerebrospinal fluid (CSF) proteomes of human, mouse, and birds.

The cell adhesion molecule pathways (hsa04514) were enriched by KEGG analysis. Proteins are labeled by their gene symbols. Gene symbols in red were commonly associated with the enrichment of this pathway in the CSF of human, mouse, and birds. The enriched pathways are statistically significant at P < 0.05, false discovery rate < 0.05.

Signaling pathways of the gene products involved in neurogenesis commonly identified in the cerebrospinal fluid proteomes of human, mouse and birds.

(A) Dot plot of all significantly enriched Reactome pathways. Dots represent term enrichment with color coding scaled on the degree of statistical significance. The sizes of dots represent the number of proteins identified and the gene ratio refers to the proportion of the proteins implicated in the pathway out of the overall 196 analyzed proteins. (B) Network plot showing the relationship of all significantly enriched pathways in terms of their involved proteins. The size of the dot corresponding to each term represents their number of protein content. Proteins are labeled by their gene symbols. Regulation of IGF*: Regulation of Insulin-like Growth Factor transport and uptake by Insulin-like Growth Factor Binding Proteins. The enriched pathways are statistically significant at P < 0.05, false discovery rate < 0.05.

Discussion

In this review, we investigated the existing literature on healthy homoiothermic amniote adult CSF. We extracted related species-specific proteomic data in mammals and birds. Our analysis comprised protein identifications in CSF from human, mouse, sheep, chicken, and two parrot species, the budgerigar, and the cockatiel. Studies reported varying numbers of proteins associated with the CSF proteome, plausibly related to different approaches in sample preparation and detection techniques. Our over-representation pathway enrichment analysis showed that neurogenesis, nervous system development, axon development, neuron projection development, and cell migration were among the most highly statistically significant terms compared to other biological processes. In addition, these pathways had the highest protein coverage across all the species and datasets. Following systematic evaluation of the reviewed proteomic datasets, we found that different mammalian and avian species are comparable in their proportions of CSF proteins implicated in neurogenesis signaling, and based on single observations, the chicken is more similar to the sheep. However, the paucity of proteomic studies in birds and different mammalian species hinder further insights into any similarities and differences. We also observed that the NG are a consistently important part of the CSF when compared to other functionally important protein subsets that vary more among individuals, species, and studies. As far as we are aware, this is the first proteomic study comparing CSF candidate factors involved in adult neurogenesis in different species.

Adult neurogenesis is a complex, dynamic, and multistep process orchestrated by tight regulation of intrinsic and extrinsic factors, like what occurs during development. It remains elusive how active and widespread this process is across taxa and brain regions (Accogli et al., 2020). During the sequence of molecular and cytoarchitecture events, CSF supports, maintains, and promotes the proliferation of the neurogenic niches (Fame and Lehtinen, 2020; Gato et al., 2020). Interindividual variation in rare protein representation, post-translational modifications, age, blood contamination (You et al., 2005), or circadian rhythm (Nilsson et al., 1992) affect the CSF molecular composition. Also, adult neurogenesis depends on various factors such as aging, stress, exercise, and diet (Aimone et al., 2014). Therefore, in our analysis, we focused only on the CSF proteins related to neurogenesis that are consistently found across human, mouse, and avian studies.

Pathway enrichment analysis of the commonly reported avian and mammalian NG proteins revealed known pathways such as those comprising IGF signaling, cell-adhesion and extracellular matrix (ECM) molecules, immune signaling, axonal guidance, migration, and cell death. Various soluble factors have been reported to be part of the CSF molecular profile, like cytokines and chemokines (IL1beta, CSF1, CX3CL1) (Kokovay et al., 2012; Silva-Vargas et al., 2016), growth factors, and associated proteins such as IGF2, transforming growth factor (TGF) beta2, fibroblast growth factor (FGF) 2, vascular endothelial growth factor (VEGF) A, leukemia inhibiting factor, endothelial growth factor receptor, IGF binding protein 2 (Jin et al., 2002; Bauer and Patterson, 2006; Mathieu et al., 2010; Fame and Lehtinen, 2020), carriers (retinol-binding protein 4), fetuin-A, enzymes, hormones (adiponectin), ECM re-modelers (SPP1, SERPINA1, MMP3, TIMP1) (Silva-Vargas et al., 2016; Obernier and Alvarez-Buylla, 2019). Other constituents for intercellular communication such as extracellular vesicles trafficking cargo of microRNAs, mRNAs, proteins, or lipids are also present in CSF (Losurdo and Grilli, 2020).

In our analyzed datasets, the presence of mitogens in the CSF was evidenced by the presence of growth factors such as IGFs and their binding proteins, a finding that is well documented in CSF during development and adulthood (Lehtinen et al., 2011). In addition, other proteins implicated in growth factor pathways were commonly enriched in CSF, such as FGF (FGFR1, FGFR2), VEGF, and TGF-beta (TGFB1, TGFB2) pathways. Some members of the L1 family of cell adhesion molecules (L1CAM) were among the most significantly over-represented consistent proteins. L1CAMs are neural recognition molecules of the immunoglobulin superfamily, implicated in nervous system development, neurite outgrowth, and neuron migration (Schmid and Maness, 2008). Furthermore, we detected over-represented pathways involved in the formation and maintenance of synapses including N-cadherin (CDH2), neural cell-adhesion molecule, neuroligins and neurexins, and axonal guidance by enrichment of netrins (e.g. NTNG1, NTN1, NTN4), reelin (RELN), Slits (SLIT1, SLIT2, SLIT3), semaphorins (e.g. SEMA6D, SEMA4G, SEMA7A, SEMA4B, SEMA3D) and ephrins (e.g. EPHA4, EPHA5, EPHB2, EPHB3) and their receptors (e.g. DCC, ROBO1, ROBO2). A comprehensive description of axonal guidance molecules and pathways has been reviewed (Russell and Bashaw, 2018).

Several ECM proteins (e.g., laminins, tenascins, agrin) known to bind to growth factors and other signaling molecules were also commonly present in avian and mammalian CSF. ECM molecules regulate the activity of growth factors (Kim et al., 2011). Components of the CSF proteome might also regulate the ECM components of NSCs, matrix elasticity, or stiffness, which in turn influences cell proliferation, migration, and survival in neurogenic niches. Recently, characterization of ECM proteome in SVZ and medial sub-ependymal zone showed a significant correlation between higher tissue stiffness and NSC responsiveness to neurogenesis in mouse brain (Kjell et al., 2020). Extrinsic cues like hyaluronan and proteoglycan link protein 1, tenascins, S100 proteins, and annexins regulate the adult mammalian NSCs. In addition, collagens and laminins confer increased stiffness to ECM (Swift et al., 2013; Kjell et al., 2020). To note, bone morphogen proteins that also confer mechanical stiffness were not present in the avian CSF proteome in the studied datasets, in contrast to mammals, which merits further study.

These pathways are possibly regulated by programmed cell death which is an essential balancing step during development and adulthood triggering birth (Doetsch and Scharff, 2001). Immune input in the absence of inflammation supports homeostasis in the absence of infection and exerts neuromodulatory actions (Salvador et al., 2021). NSCs are sensitive to leukemia inhibiting factors, cytokines, and complement cascade proteins that act out in triggering NSCs proliferation (Chintamen et al., 2021). For example, CSF-borne IL1beta upregulates the expression of vascular cell adhesion protein 1, which maintains the architecture of NSCs (Kokovay et al., 2012). Cytokines (CSF1, IL6ST), chemokines (e.g., CXCL12), apolipoproteins (APOA1, APOD), post-translational protein modification signaling pathways (e.g., CDK5, CDC42), and proteases were also present in the analyzed avian and mammalian healthy adult CSF proteomes. However, few proliferating cells survive to migrate, and microglia may elicit cell death or clear the remains of apoptotic cells by phagocytosis, participating in a negative feedback loop to regulate the neurogenic niche (Sierra et al., 2010).

Moreover, most of the reported common NG proteins belonged to extracellular vesicles or exosomes implying that they constitute cargo of exosomes residing in CSF. Intercellular communication through extracellular vesicles is a field of current active research (Losurdo and Grilli, 2020). CSF extracellular vesicles have been reported to contain brain-derived proteins, amyloid precursor protein, the prion protein, and DJ-1, alix and syntenin-1, heat shock proteins, and tetraspanins (Chiasserini et al., 2014).

Several studies have provided evidence that CSF can promote neurogenesis in NSCs (Lehtinen et al., 2011; Zappaterra and Lehtinen, 2012; Fame and Lehtinen, 2020). Mature primary rat hippocampal neuronal cultures and human brain slices exposed to human CSF maintained their synaptic function and overall viability better than a conventional medium (Perez-Alcazar et al., 2016; Schwarz et al., 2017; Wickham et al., 2020).

Moreover, the secretome from the lateral ventricle choroid plexus, the primary producer of the CSF, promoted the formation and proliferation of multipotent neurospheres derived from ventral SVZ cells and quiescent and activated NSCs and transit-amplifying cells (Silva-Vargas et al., 2016). More recently, it was shown that ventricular CSF from aged individuals continues to influence NSC proliferation, motility, and differentiation in vitro (de Sonnaville et al., 2020). Human adult healthy CSF improved the maturation of neuronal circuits in human-induced pluripotent stem cells (a human 3D neural in vitro model), causing an immediate and lasting increase of neurite net and synapse formation, astroglial and neuronal differentiation (Izsak et al., 2020). Noteworthy, it has been proposed that NSCs of the adult human SVZ are more prone to tumor formation than those of the adult hippocampus partly because they contact the CSF (Fontán-Lozano et al., 2020). In another recent study, human NSCs expressed diversified proteome in various stages of their differentiation, as shown by alterations in Hypoxia-inducible factor 1, wingless related (Wnt), and VEGF signaling pathways, increased expression of neuropilin 1 (NRP1) and catenin b-1 (CTNNB1, present in human and mouse datasets in this review) as well as IL6 and VEGF regulation of proliferation and survival of the differentiating cells (Červenka et al., 2021). Whether the upregulation of these proteins is influenced by the CSF and changes in differentiating NSCs parallel CSF proteome alterations warrants further study.

The studies reviewed in this contribution were not directly comparable due to methodological differences (e.g., different sample preparation and detection techniques) and their selection and treatment of study subjects (e.g., subjects’ precise age, housing, history). The degree of senescence of the animals in the individual studies analyzed could be an important issue because it varied, although only research in adults was considered. The studies in birds and mice involved younger adults, while most studies in humans involved more senior individuals. For these reasons, the quantitative expression data were omitted from our analysis. It remains open how the identified proteins are regulated at the systemic level, their lifetime in CSF and how individual proteins contribute to the interindividual variability. Nevertheless, our study forms an essential comparative basic reference for the common vertebrate CSF candidate factors participating in adult neurogenesis.

Conclusion

In summary, our literature survey, database mining, and extensive bioinformatic analysis identified up to 911 proteins comprising CSF-derived diffusible factors related to neurogenesis in adult humans, mice, sheep, and birds. Systematic evaluation of the reviewed datasets showed that proteins implicated in neuronal regeneration pathways constituted an important and consistent fraction of CSF independently of the species. Focusing on NG factors common to avian and mammalian CSF proteomes, the universally enriched pathways were associated with regulation of nervous system development comprising L1CAM protein interactions, cell-adhesion molecules, signals regulating ECM remodeling, regulation of IGF signaling, axonal guidance, programmed cell death, immune signaling, and post-translational modifications and most proteins were part of extracellular vesicles enriched in CSF. The list of these proteins is not exhaustive, but the recent progress in genomics and proteomics technologies now promisingly allows comprehensive comparative research aimed at the characterization of CSF composition differences in various species. It is plausible that these factors act at multiple levels and in an intertwined manner. It remains unclear how the common pathways are regulated to contribute to the inter- and intra-specific variability observed in vertebrates. Therefore, further comparative proteomic research in CSF is imperative. Overall, this meta-analysis provides a synopsis of the molecular composition of CSF that is preserved in homoiotherms and defines the determinants of neurogenesis found in the proteome of healthy adult samples. This information is relevant especially to studies using in vitro models in neurological research. Despite its limitations, our study provides valuable information for the proteomic community as a starting point for translational studies focusing on CSF and neurological disorders during the multi-omics era. Presently, comparative proteomics is developing our understanding of animal phenotypic variation and profile as an actual and promising field of investigation. In-depth comparative insights into the molecular CSF composition unravel the basic mechanisms driving and diversifying adult neurogenesis, facilitating further progress in restorative and regenerative interventions to the central nervous system.

Additional files:

Additional Table 1: Details and characteristics of the included studies.

Additional Table 1

Details and characteristics of the included studies

Dataset	Study reference	doi	Species	Healthy subjects	Total subjects	Type of study	Site of cerebrospinal fluid (CSF) extraction	Age	Identified proteins	Labelingmethod	Other information
1	Zhao et al 2018	10.1002/prca.201800008	Human	14	14	Descriptive	Lumbar puncture	24–55 years, median 28	1271	iBAQ	Immunoaffinity depletion
2	Sleat et al 2016	10.1021/acs.jproteome.7b00460	Human	7	14	Comparative	Lumbar puncture	53-80	1470	iBAQ
3	Bereman et al 2018	10.1038/s41598-018-34642-x	Human	30	63	Comparative	Lumbar puncture	Adult, age unspecified	357	LabelFree	Abundant protein depletion
4	Macron et al 2020	10.1016/j.dib.2020.105704	Human	Pool	Unspecified	Descriptive	Lumbar puncture	Adult, age unspecified	3174	TMT	Abundant protein immuno-depletion
5	Sathe et al 2018	10.1002/prca.201800105	Human	5	10	Comparative	Lumbar puncture	Adult, age unspecified	2244	TMT	Abundant protein immuno-depletion
6	Schilde et al 2018	10.1371/journal.pone.0206478	Human	12	12	Descriptive	Lumbar puncture	average 65	610	LabelFree	Intra-and interindividual variability
7	Thomson et al 2018	https://doi.org/10.1002/ana.25143	Human	20	101	Comparative	Lumbar puncture	58.5 +/- 8.6	677	LabelFree
8	Dayon et al 2019	10.1021/acs.jproteome.8b00809	Human	48	120	Comparative	Lumbar puncture	70.2	790	TMT	Immunoaffinity depletion
9	Franzen et al 2020	10.1186/s12888-020-02874-9	Human	8	15	Comparative	Lumbar puncture	19-54	425	LabelFree
10	Johnson et al 2020	10.1038/s41591-020-0815-6	Human	72	420	Comparative	Lumbar puncture	64-90+	3331	TMT	Abundant protein immuno-depletion
11	Rotunno et al 2020	10.1038/s41598-020-59414-4	Human	115	196	Comparative	Lumbar puncture	35-81	341	LabelFree	Data Independent Acquisition
12	Wang et al 2020	10.1186/s13024-020-00384-6	Human	9	20	Comparative	Lumbar puncture	Adult, age unspecified	5941	TMT	No depletion
13	Wang et al 2020	10.1186/s13024-020-00384-6	Mouse	5	11	Comparative	Cisterna magna	9–12 months	1058	TMT	No depletion
14	Pigoni et al 2016	10.1186/s13024-016-0134-z	Mouse	7	14	Comparative	Cisterna magna	Adult, age unspecified	1329	LabelFree
15	Hosp et al 2017	10.1016/j.celrep.2017.10.097	Mouse	9	20	Comparative	Cisterna magna	5, 8, and 12 weeks	778	iBAQ + LabelFree
16	Sleat et al 2019	10.1074/mcp.RA119.001587	Mouse	9	27	Comparative	Cisterna magna	4 to 52 weeks of age	4688	TMT	MS3 level
17	Tushaus et al 2020 a	10.15252/embj.2020105693	Mouse	4	4	Descriptive	Cisterna magna	Adult, age unspecified	908	LabelFree	Data Dependent Acquisition
18	Tushaus et al 2020 b	10.15252/embj.2020105693	Mouse	4	4	Descriptive	Cisterna magna	Adult, age unspecified	984	LabelFree	Data Independent Acquisition
19	Chen et al 2018	10.1016/j.exger.2018.04.012	Sheep	21	21	Comparative	cisterna magna	1 and 10 year old	264	iTRAQ
20	Voukali et al 2021	10.1038/s41598-021-84274-x	Chicken	7	7	Descriptive	cisterna magna	adult, age unspecified	483	LabelFree
21	Voukali et al 2021	10.1038/s41598-021-84274-x	Budgerigar	5	5	Descriptive	cisterna magna	adult, age unspecified	622	LabelFree
22	Voukali et al 2021	10.1038/s41598-021-84274-x	Cockatiel	5	5	Descriptive	cisterna magna	adult, age unspecified	452	LabelFree	Mapping against budgerigar proteome
23	Unpublished data	N/A	Budgerigar	14	30	Comparative	cisterna magna	adult, age unspecified	1337	LabelFree
24	Unpublished data	N/A	Budgerigar	16	30	Comparative	cisterna magna	adult, age unspecified	1662	LabelFree

Additional Table 2: Cerebrospinal fluid protein identification lists per dataset.

Cerebrospinal fluid protein identification lists per dataset

Additional Table 3: All gene products annotated with the gene ontology (GO) term neurogenesis (GO:0022008) as retrieved from the GO database.

All gene products annotated with the gene ontology (GO) term neurogenesis (GO:0022008) as retrieved from the GO database

Additional Table 4: Results of Gene Ontology (GO) Biological Process pathway enrichment analysis performed on the complete CSF proteome from each dataset.

Results of Gene Ontology (GO) Biological Process pathway enrichment analysis performed on the complete cerebrospinal fluid proteome from each dataset

Additional Table 5: Reported Gene Ontology Neurogenesis (GO:0022008) proteins per species/taxon.

Reported Gene Ontology Neurogenesis (GO:0022008) proteins per species/taxon

Additional Table 6: Enrichment pathway analysis results of the reported Gene Ontology Neurogenesis (GO:0022008) proteins common to humans, mice, and birds and per species.

Enrichment pathway analysis results of the reported Gene Ontology Neurogenesis (GO:0022008) proteins common to human, mouse and birds and per species

Table 1

Proportions (NG-ratio) of neurogenesis-related-proteins (NG) out of the total identified proteins from cerebrospinal fluid proteome (CSF total) per dataset, classified by species

Dataset	Species	NG	CSF total	NG-ratio (%)	Reference
1	Human	254	1271	19.98	Zhao et al., 2018
2	Human	210	1468	14.31	Sleat et al., 2016
3	Human	77	356	21.63	Bereman et al., 2018
4	Human	479	3173	15.1	Macron et al., 2020
5	Human	406	2244	18.09	Sathe et al., 2018
6	Human	132	610	21.64	Schilde et al., 2018
7	Human	109	676	16.12	Thomson et al., 2018
8	Human	159	790	20.13	Dayon et al., 2019
9	Human	61	376	16.22	Franzen et al., 2020
10	Human	389	3262	11.93	Johnson et al., 2020
11	Human	72	342	21.05	Rotunno et al., 2020
12	Human	675	5940	11.36	Wang et al., 2020
13	Mouse	167	1092	15.29	Wang et al., 2020
14	Mouse	191	1409	13.56	Pigoni et al., 2016
15	Mouse	134	826	16.22	Hosp et al., 2017
16	Mouse	519	4688	11.07	Sleat et al., 2019
17	Mouse	163	971	16.79	Tushaus et al., 2020
18	Mouse	168	1025	16.39	Tushaus et al., 2020
19	Sheep	30	264	11.36	Chen et al., 2018
20	Chicken	57	483	11.80	Voukali et al., 2021
21	Budgerigar	94	622	15.11	Voukali et al., 2021
22	Cockatiel	89	452	19.69	Voukali et al., 2021
23	Budgerigar	214	1337	16.00	Unpublished data
24	Budgerigar	259	1662	15.58	Unpublished data