Neural stem and progenitor cells support and protect adult hippocampal function via vascular endothelial growth factor secretion

Adult neural stem and progenitor cells (NSPCs) reside in the dentate gyrus (DG) of the hippocampus throughout the lifespan of most mammalian species. In addition to generating new neurons, NSPCs may alter their niche via secretion of growth factors and cytokines. We recently showed that adult DG NSPCs secrete vascular endothelial growth factor (VEGF), which is critical for maintaining adult neurogenesis. Here, we asked whether NSPC-derived VEGF alters hippocampal function independent of adult neurogenesis. We found that loss of NSPC-derived VEGF acutely impaired hippocampal memory, caused neuronal hyperexcitability and exacerbated excitotoxic injury. We also found that NSPCs generate substantial proportions of total DG VEGF and VEGF disperses broadly throughout the DG, both of which help explain how this anatomically-restricted cell population could modulate function broadly. These findings suggest that NSPCs actively support and protect DG function via secreted VEGF, thereby providing a non-neurogenic functional dimension to endogenous NSPCs.


Introduction
The dentate gyrus (DG) of the hippocampus is one of only a few brain regions that supports resident neural stem and progenitor cells (NSPCs) in adult mammals 1,2 . Adult hippocampal NSPCs have generated clinical interest because hippocampus-mediated memory function declines with many injuries and diseases, such as seizures, trauma, and Alzheimer's disease 3 . Previous research on the functional properties and therapeutic applications of adult NSPCs has focused on their ability to create new neurons that exhibit unique electrophysiological properties 4,5 . NSPCs, in contrast, are not electrophysiologically active. However, preclinical studies have increasingly identified the growth factors and cytokines secreted by stem cells (their secretome) as a major component of the therapeutic effect of stem cell transplants 6,7 . The role that the endogenous NSPC secretome plays in brain health remains more sparsely investigated 3, 8,9 . Nonetheless, several recent studies suggest that NSPCs can shape adult niche properties through their secretome 10-12 . We recently showed that a major component of the adult DG NSPC secretome is the pleiotropic, soluble protein vascular endothelial growth factor (VEGF, or VEGFA) 13 . We showed that self-derived VEGF signaling in radial glia-like neural stem cells (NSCs) is necessary to maintain their quiescence and prevent premature exhaustion 13 . While this work revealed an autocrine role for NSPC-derived VEGF in maintaining neurogenesis, the potential VEGF paracrine effects were unclear. VEGF is a multi-faceted molecule that can modulate multiple cellular processes in the adult CNS and may protect or impair neuronal function depending on the context 14,15 . In contrast to the weeks-long process of neurogenic exhaustion, VEGF paracrine signaling has the potential for rapid, acute regulation of the DG. In this work, we therefore investigated the acute, paracrine role of VEGF derived specifically from NSPCs in the function of the adult mouse DG.

Results
Knockdown of VEGF in NSPCs impairs hippocampal spatial memory Our previous research showed that astrocytes and NSPCs are the primary producers of VEGF in the adult DG with NSPCs contributing ~1/3 of total DG VEGF 13 . To investigate the functional effect of VEGF derived specifically from NSPCs, we crossed NestinCreER T2 mice 16 with VEGF lox mice 17 (Fig. 1A). NestinCreER T2 mice drive LoxP recombination with high selectivity in NSPCs when exposed to the synthetic estrogen tamoxifen (TAM) 16,18,19 . We confirmed this selectivity of recombination in NestinCreER T2 mice crossed with stop-floxed EYFP reporter mice, finding that 97% of EYFP+ cells in the DG were NSPCs and less than 1% were astrocytes 3d after TAM (Supplementary Fig. 1A-D). We further confirmed knockdown specificity in adult NestinCreER T2 ;VEGF lox/lox mice (iKD) and their VEGF lox/lox littermates (Wt). After TAM injection, in situ hybridization with a probe against vegfa coupled with immunolabeling for GFAP confirmed vegfa loss selectively in GFAP+ radial glia-like putative NSCs but not in GFAP+ stellate putative astrocytes (Fig. 1B,C; Supplementary Fig. 1E).
To test the role of VEGF in hippocampal behavioral functions, we treated VEGF iKD mice and Wt littermates with TAM then assessed performance on hippocampus-dependent spatialcontextual memory Y-maze and object location tasks two weeks later (Fig. 1D). We have previously shown that loss of NSPC-VEGF impairs NSC maintenance and thereby neurogenesis, but the loss of new neurons requires > 3 weeks to emerge 13 . In addition, newborn neurons require 3-4 weeks to integrate into local circuitry 20 . This 2-week timepoint for behavioral testing was therefore chosen to be sufficiently close to VEGF knockdown that any changes in production of new neurons would not likely have functional effects on hippocampal circuitry and behavioral changes could be attributed to acute effects of VEGF protein on existing cells.
In a hippocampus-dependent Y maze task, all Wt mice (100%, 10/10) chose to enter a novel, unexplored arm first over a familiar arm, while significantly fewer (54.5%, 6/11) VEGF iKD mice chose the novel arm first (Fig. 1E). Latency to enter the novel arm was longer in VEGF iKD than Wt mice but that measure did not reach statistical significance, nor did time in the novel arm (Supplementary Fig. 1F,G). In the novel object location task, both Wt and iKD mice showed no preference for either of 2 objects during initial exploration (train). During testing, when one object was moved, Wt mice showed a significant increase in investigation preference for the moved object, but VEGF iKD did not (Fig. 1F). There were no differences in total object investigation time, suggesting similar motor ability and motivation to investigate in both groups (Supplementary Fig. 1H). Last, we tested mice in an elevated plus maze to assess anxiety-like behaviors 21 . Wt and iKD mice showed no difference in time in the open arms or entries into the closed arms (Fig. 1G, Supplementary Fig. 1I,J).
To further confirm these findings, we repeated these tests in a separate cohort of mice 3 days after TAM, reasoning that deficits due to acute loss of VEGF protein could be evident soon after recombination, given the generally short half-life of secreted proteins in vivo ( Supplementary  Fig. 1K). Y-maze first arm choice was not different between Wt and iKD mice, though this analysis was complicated by the failure of the Wt mice to show a preference for the novel arm ( Supplementary Fig. 1L-N). In the novel object location test, in contrast, Wt mice showed significant preference for the moved object while the iKD mice did not, replicating our finding at 2 weeks (Supplementary Fig. 1O,P). In the EPM, iKD mice showed a small but significant increase in time in the open arms, but there was no difference in the time in open arms, compared to Wt mice (Supplementary Fig. 1Q,R). No difference in closed arms entries was found (Supplementary Fig. 1S). Overall, findings in both the Y maze and object location task suggest that NSPC-derived VEGF acutely supports hippocampal spatial memory, while general object exploration and elevated plus maze findings further suggest that these effects are likely not confounded by any changes in anxiety-like behavior.
Knockdown of VEGF in NSPCs causes DG hyperexcitation We next sought to clarify the circuit mechanism by which NSPC-VEGF supports memory function. VEGF is known to signal via VEGFR2 expressed by neurons, generally in support of memory function 22,23 . In many brain regions, enhanced memory is linked to enhanced neuronal activity. The DG, however, is heavily inhibited by local interneurons, and this inhibition is hypothesized to underpin its role in spatial memory pattern separation 24-27 . We therefore investigated the hypothesis that NSPC-derived VEGF supports hippocampal memory via suppression of DG excitability using slice electrophysiology. Ex vivo hippocampal slices were prepared from Wt and iKD mice 3d after the last TAM injection ( Fig. 2A). Perforant path stimulation revealed a significant increase in I-V slope in the DG of VEGF iKD mice compared to Wt littermates, suggesting increased basal synaptic transmission in iKD mice DGs (Fig. 2B). LTP induced by high frequency stimulation (HFS) of the perforant path was also significantly higher in VEGF iKD mice DGs than Wt littermate controls (Fig. 2C). As picrotoxin is present within the bathing solution, these results suggest greater plasticity in excitatory circuits within the DG. No difference was found in paired-pulse ratio (Fig. 2D) indicating that the effects observed are likely due to postsynaptic, rather than presynaptic, changes. Together, these data suggest greater excitatory transmission and plasticity following loss of NSPC-derived VEGF supporting the idea that there is hyperexcitability within the DG.
VEGF disperses widely throughout the DG from a point source In our previous work 13 , we showed that VEGF in the DG is synthesized primarily by astrocytes and NSPCs. In the adult DG, these two cell types are concentrated in separate DG cell layers, with NSPCs residing exclusively in the subgranular zone and astrocytes residing primarily in the more acellular spaces of the DG like the hilus and molecular layer. VEGF is a secreted, soluble protein, and its potential to regulate other cells will depend critically on its spatiotemporal spread through tissue. Computational estimates suggest that VEGF decreases by up to 12% every 10 µm away from its source in mature muscle tissue 28 . Given that DG subregions can span 50-200+ µm, such a rapid decay in VEGF through tissue would severely limit what cell types NSPC-derived VEGF could reach. We therefore sought to quantify dispersal of VEGF through live adult mouse DG to determine if it is plausible that NSPC-derived VEGF could regulate cells throughout the DG subregions.
To visualize VEGF dispersal through live DG tissue, we used genetic code expansion coupled with bioorthongonal non-canonical amino acid tagging to create tagged forms of both major VEGF isoforms synthesized by NSPCs and astrocytes: VEGF120 TAG and VEGF164 TAG . We chose this approach over a more traditional approach, such as generating a fluorophore-fusion protein, because most fluorophores (e.g. GFP, 27 kDa) would more than double the molecular weight of VEGF (~17-22 kDa). This tagging approach, in contrast, involves integrating a single noncanonical amino acid in a recombinant VEGF protein, which can be detected by spontaneous and highly specific inverse-demand Diels-Alder cycloaddition click reaction with tetrazines [29][30][31][32] . In brief, we designed expression constructs for each isoform of VEGF cDNA with an in-frame amber stop codon in exon 2. We generated tagged VEGF isoforms using these constructs coupled with expression of a mutant pyrrolysyl-tRNA synthetase/tRNA Pyl pair (PylRS/tRNApyl) that mediates incorporation of a noncanonical amino acid (TCO*A) at amber stop codons 29 (Fig.  3A). Isolated proteins were found at the appropriate weight for VEGF120 and VEGF164 dimers respectively (~34 and ~44 kDa) using 3 different methods: 1) at the total protein level, 2) reacted with streptavidin IR800 (to reveal TCO*A-tetrazine-biotin presence) and 3) using VEGF immunoreactivity (Supplementary Fig. 2A). Tagged VEGF monomers were also found at the predicted weight when separated under reducing conditions (~17 and ~22 kDa) (Supplementary Fig. 2B).
Adult wildtype mice were unilaterally stereotaxically infused with VEGF120 TAG , VEGF164 TAG or the vehicle control into DG then perfused 10 min, 1hr, 3h or 6h later. The dispersion area of VEGF TAG -bearing proteins was quantified using automated region detection software. At the epicenter of the infusion, we found that both VEGF isoforms reached maximal dispersion ipsilateral to the injection site 1-3h after infusion, with minimal spread to the contralateral DG ( Fig. 3B,C; Supplementary Fig. 2C). There was no significant difference between VEGF isoforms, though VEGF164 TAG appeared to have slightly less dispersal area than VEGF120 TAG at later timepoints, likely reflecting the known lower solubility of VEGF164 compared to VEGF120. To quantify the spatial dispersal of VEGF TAG though the DG, we focused on the decay of signal over space from the epicenter. We found that the decay half-life of spatial dispersal was similar between isoforms: 227.7 µm (95% CI 95.94 to 906.3 μm) for VEGF120 TAG and 312.2 µm (95% CI 131.8 to 1612 μm) for VEGF164 TAG (Fig. 3D). Representative areas show that both isoforms largely fill the area of the DG within that half-life (Fig. 3E, Supplementary Fig. 2D). These findings suggest that VEGF in the adult DG has substantial spatial dispersal from a point source. The location of NSPCs within the SGZ is therefore unlikely to be a barrier to widespread signaling of their secreted VEGF. DG VEGF derives from astrocytes and NSPCs following excitotoxicity Several studies suggest that VEGF may help prevent DG hyperexcitation and neuronal damage in pathological conditions [33][34][35] . Our findings that loss of NSPC-VEGF increases DG excitability and that VEGF likely can disperse widely throughout the DG from isolated sources suggests a possible role for NSPCs in reducing widespread pathological excitotoxicity, such as that which occurs with seizure-related activity. We therefore next investigated the hypothesis that VEGF protects the adult mouse DG from seizure-related excitotoxic damage.
To determine the relative contribution of NSPCs to total DG VEGF after seizure-induced excitotoxic injury, we used the glutamatergic agonist kainic acid (KA). First, we treated VEGF-GFP transcriptional reporter mice with KA (or vehicle) and quantified GFP expressing astrocytes, NSCs and intermediate progenitor cells (IPCs) 1, 3 and 7d later. The density of VEGF-GFP+ astrocytes and NSCs increased with KA, peaking around 3-7d after injury, in parallel with an increase in total astrocytes and NSCs (Fig. 4A,B; Supplementary Fig. 3A,B).
We next asked whether the proportion of total DG VEGF produced by NSPCs changes with KA. We measured VEGF mRNA and protein in whole DG 1 or 7d after KA in VEGF iKD and Wt mice . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; https://doi.org/10.1101/2023.04.24.537801 doi: bioRxiv preprint ( Fig. 4C). At 1d after KA, KA increased vegf120 and vegf164 mRNA by ~2-fold in Wt mice. iKD mice showed significant blunting of the KA-response specifically in the vegf120 isoform, but not in the vegf164 isoform ( Fig. 4D). At 7d after KA, DG VEGF protein (including both 120 and 164 isoforms) was significantly lower in iKD mice compared to Wt mice (Fig. 4E). Combined, these findings suggest that NSPCs contributed measurable amounts of VEGF to total DG VEGF, particularly of the vegf120 isoform, both during an acute VEGF surge and a later recovery phase.
Knockdown of NSPC-derived VEGF exacerbates excitotoxic injury We next examined the injury response to KA-induced excitotoxicity. Using wildtype mice, we found that KA caused large increases in expression of proinflammatory genes c1qa, il1a and tnf that peaked between 1 and 3d after KA (Supplementary Fig. 4A-D). Expression of these genes, especially tnf, was most consistently high in the DG and CA regions of the hippocampus, with the cortex and subventricular zone (SVZ) showing no significant response to KA. This relative isolation of injury response to hippocampal regions is consistent with numerous observations of the hippocampus showing heightened sensitivity to injury relative to other brain . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; regions 36 . We also confirmed that KA induced increases in DG microgliosis as revealed by immunolabeling for CD68 (Supplementary Fig. 4E,F) and astrogliosis as revealed by immunolabeling for GFAP (Supplementary Fig. 4G,H), both of which peaked 3-7d after KA.
Based on the above results, we first examined VEGF iKD and Wt mice 1d after KA to quantify the early phase of excitotoxic response (Fig. 5A). VEGF iKD and Wt mice both showed motor seizure activity after KA, with a mode Racine score of 3 for both groups (Supplemental Fig.  5A). cFos labeling in the granule cell layer, a surrogate measure of neuronal activity, was 66.6% greater in VEGF iKD mice than Wt littermates (Fig. 5B,C). Mice were also injected with BrdU during TAM injections, to mark dividing cells. Almost no BrdU+ cells co-labeled with cFos in any group, suggesting that cells born during TAM-induced recombination were not yet functionally integrated in circuitry, as would be expected based on their cellular age (Supplemental Fig.  5B). These findings suggested that loss of NSPC-derived VEGF causes an acute increase in excitability of existing neurons in the DG.
We next used Fluorojade C labeling to quantify neuronal degeneration. Fluorojade C was only detected in DG of KA treated mice, with VEGF iKD mice being significantly more likely to show Fluorojade C+ degenerating cells than Wt littermates: 16.6% (2/12) of Wt mice treated with KA had detectable Fluorojade C in the DG while 60% (9/15) of VEGF iKD mice had detectable Fluorojade C (Fig. 5D,E). VEGF iKD mice also showed significantly greater signs of DG neuroinflammation in response to KA than Wt littermates as revealed by CD68 immunolabeling (Fig. 5F,G) and expression of c1qa and tnf (though not of il1a) (Fig. 5H). GFAP immunolabeling showed no change in response to KA at this 1 day timepoint (Supplementary Fig. 5C,D) .
To confirm that KA did not alter the specificity of TAM-induced recombination, we quantified expression of stop-floxed EYFP in NestinCreER T2 mice 1d after KA. Almost all EYFP+ DG cells showed either NSC phenotype (56%) or IPC phenotype (41%), with <1% showing astrocytic phenotype (Supplementary Fig. 5E-G). In addition, to confirm that the apparent exacerbation of KA-induced injury due to loss of NSPC-VEGF was not an artefact of Cre expression, we injected VEGF wt/wt ;NestinCreER T2 mice (Cre+) and VEGF wt/wt littermates (Cre-) with TAM and KA. Cre+ and Cre-mice did not differ in Fluorojade C labeling or CD68 immunolabeling, suggesting that Cre expression alone does not drive exacerbation of excitotoxic injury in the DG (Supplementary Fig. 5H,I).
At 7d after KA, we found evidence of continued exacerbation of injury in VEGF iKD mice compared to Wt littermate controls (Fig. 6A). VEGF iKD mice showed greater frequency of Fluorojade C+ degenerating neurons (43% or 6/14) than in Wt mice (7% or 1/15) ( Fig. 6B,C). CD68 immunoreactivity in the DG was no longer different between iKD and Wt mice (Fig. 6D,E), but GFAP immunoreactivity showed a KA-induced increase that was exacerbated in VEGF iKD mice compared to Wt mice (Fig. 6F,G). In agreement with previous work showing that excitotoxic injury causes vascular remodeling in the DG that can be reflected as reduced vessel coverage 37,38 , we found that KA induced a loss of vessel coverage using immunolabeling for the endothelial cell marker CD31. However, this loss was similar in Wt and iKD mice (Fig. 6H,I), suggesting that gross vascular remodeling after KA injury is not dependent on VEGF derived from NSPCs.
Combined, we found substantial acute and long-lasting exacerbation of neuronal degeneration and neuroinflammation in the DG of VEGF iKD mice compared to Wt littermates in response to excitotoxic injury.

Discussion
The stem cell secretome has recently emerged as a potent mechanism by which endogenous and transplanted stem cell populations can regulate adult tissue physiology. As efforts to harness stem cells as therapeutics advance, this functional dimension requires consideration 3, 8,39 . Here, we demonstrate a beneficial role of endogenous hippocampal NSPCderived VEGF in promoting hippocampal memory function and protecting against excitotoxic damage secondary to seizures, likely by suppressing neuronal excitability in DG circuitry.
Several other components of the DG NSPC secretome have been reported, primarily with functional roles in regulating NSPCs or neurogenesis 40 . For example, NSPCs in the adult hippocampus have been shown to secrete proteins that support either their own proliferative maintenance (e.g. IGF 41 , VEGF 13 , MFGE8 11 ) or the maturation of immature neuronal populations (e.g. PTN 10 ). Less is known about functional paracrine signaling from endogenous NSPCs. One recent study showed that NSPCs in the other major adult neurogenic niche, the SVZ, support striatal neuronal function via secretion of IGFBPL1 12 . We have previously shown that NSPCs in the SVZ do not synthesize VEGF 13 , making it unlikely that the cells in this other niche use VEGF similarly to what we found in the DG. However, these complimentary findings with IGFBPL1 further highlight the potential for direct modulation of local . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; https://doi.org/10.1101/2023.04.24.537801 doi: bioRxiv preprint neuronal function by an endogenous NSPC secretome. More research is needed to better identify other components of the endogenous NSPC secretome and functional relevance. Previous research presents often opposing findings about how VEGF affects hippocampal function. These opposing findings have garnered VEGF a reputation as a "double-edged sword" 42 , particularly in the context of excitotoxic injury. Generally, in response to hippocampal injury, VEGF signaling to neurons is considered neuroprotective, while its effects on vascular cells is considered detrimental 14 . Our findings suggest that VEGF derived from endogenous NSPCs is beneficial to hippocampal memory function and suppresses hyperexcitation. This beneficial effect extends to the response to excitotoxic injury, providing protection against neuronal injury and gliosis. We did not detect any effect of NSPC-derived VEGF on vascular remodeling after excitotoxic injury, suggesting that VEGF from another source (or other factors) mediate this vascular response. Future research will be necessary to determine the mechanisms underlying DG vascular remodeling after excitotoxicity.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; Historically, VEGF production in the brain has been attributed to astrocytes 43 . Here, we show that, while astrocytes are the majority VEGF-expressing population by cell number in the DG, NSPCs still contribute substantial quantities of VEGF, with the most notable contribution being to the vegf120 isoform in response to kainic acid excitotoxicity. This finding is consistent with our previous finding that NSPCs synthesize both vegf120 and vegf164 13 , while astrocytes are known to synthesize vegf164 primarily 44 . We also found that both major VEGF isoforms (VEGF120 and VEGF164) had wide and generally similar dispersal properties in mouse DG in vivo. The wide dispersal of VEGF protein that we observed suggests that the restriction of NSPCs to the subgranular zone is not a likely impediment to NSPCs affecting DG cells broadly via secreted VEGF. It also further suggests that VEGF derived from stem cell transplants may have the potential to affect large portions of hippocampal tissue, even if cells do not migrate long distances themselves.
VEGF120 and VEGF164 differ primarily in their relative solubility. VEGF164 carries C-terminal heparin-binding domains that partially bind in the extracellular matrix, while VEGF120 does not 45 . As a result, VEGF120 is frequently described as soluble while VEGF164 is described as semisoluble. The similarity in dispersal between VEGF isoforms that we observed here therefore seems in opposition to these known solubility differences. However, computational estimates reveal that the dominant limiter of VEGF dispersal in vivo is binding and sequestering by VEGF receptors, not isoform solubility 28 . These two VEGF isoforms have similar receptor binding capabilities, possibly explaining the general similarities in their dispersal and clearance characteristics in the present study 44 .
Our findings of neuroprotection via NSPC-VEGF add a new dimension to consider when investigating the role of adult neurogenesis in seizure response. Parent et al. 46 originally described aberrant neurogenesis in the adult rodent DG after induced seizures. They reported a surge in production of new neurons that migrated to the hilus, rather than the granule cell layer, and sent axonal projection to inappropriate DG cell layers. These findings led to the hypothesis that aberrant newborn neurons may contribute to development of epileptic circuits in the hippocampus, a hypothesis that has recently been supported by findings in humans 47 . Nonetheless, contrasting rodent studies suggest a potentially complex role for new neurons in seizures, with some neuronal subsets contributing to epileptogenesis and others preventing it 48,49 . Our findings add this growing literature, suggesting that NSPCs themselves play a beneficial role in protecting the DG. Attempts to harness endogenous neurogenic processes in any kind of therapeutic intervention for epilepsy will benefit from considering these multiple functional domains of the NSPCs and their progeny.
There are several limitations of the present study. First, though we identify neuronal hyperexcitability after NSPC-VEGF loss, we do not identify whether this happens via direct VEGFR2 signaling on excitatory granule cells, via direct VEGFR signaling on another neuronal subtype, or via signaling to another non-neuronal intermediary cell. Knockdown of VEGFR2 in hippocampal neurons impairs memory function 22,23 , making a direct loss of VEGFR2 signal a likely mechanism, but more research is needed to confirm this hypothesis. A challenge moving forward will be how to distinguish effects from NSPC-derived VEGF versus VEGF derived from other cells (e.g. astrocytes) on target cells/receptors. Second, our estimates of VEGF dispersal in vivo derive from an artificial source of VEGF and therefore may differ from how VEGF disperses when secreted from endogenous cells. A more accurate assessment would require tagging of proteins synthesized in vivo, which remains a challenge for the future.
In summary, we show that NSPCs support and protect the hippocampus from injury via production of VEGF. These findings provide a functional role for NSPCs independent of their capacity to generate new neurons. This new functional role is worthy of consideration in attempts to develop therapies leveraging endogenous NSPCs. It also warrants consideration within the ongoing debate about the existence of adult neurogenesis in humans. Much of this debate has focused on generation of new neurons as the crucial functional output of the neurogenic lineage 7,50-54 . The present work adds to an emerging literature suggesting that NSPCs themselves may have functional roles which could persist even in the absence of generation of new neuronal progeny.

Immunolabeling
Immunolabeling was performed similar to our previous work 18,56 . After perfusion, brains were harvested and fixed 24h in 4% paraformaldehyde in 0.1 M phosphate buffer followed by equilibration in 30% sucrose in PBS, both at 4°C. Brains were sliced on a freezing microtome (Leica) in a 1 in 12 series of 40 µm slices and stored in cryoprotectant at −20°C. Slices were rinsed, incubated in blocking solution (1% normal donkey serum, 0.3% triton X 100 in PBS) for 30 min at room temperature then transferred to primary antibody in blocking solution overnight at +4˚C with rotation. The next day, slices were washed and incubated in secondary antibody (1:500) in blocking solution for 2h at room temperature with rotation, followed by 10 min in Hoechst (1:2000 in PBS, Fisher #H3571). After rinsing, slices were mounted on superfrost plus microscope slides precleaned (Fisher, #12-550-15) and protected with Prolong Gold antifading medium (Fisher, #P36934) with coverglass. For BrdU labeling, all other antigens were labeled first, followed by 10 min in 4% paraformaldehyde, rinsing and 30 min in 2N HCl at 37˚C. Sections were then rinsed blocked and labeled for BrdU. For EdU labeling, click labeling was . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Microscopy and image quantification All images using co-labeling were acquired on a Zeiss Axio Observer Z1 microscope with Apotome for optical sectioning or a confocal laser-scanning microscope (Zeiss) using 20x air, 40x air or 63x oil objective as needed. Full z-stacks were acquired for analysis. Quantification of astrocytes, NSCs and IPCs was done by 1-2 blinded observers in z-stacks to identify GFAP+SOX2+ stellate cells (astrocytes), GFAP+SOX2+ radial glia like cells (NSCs) and SOX2+ cells lacking GFAP co-label (IPCs). Cell counts were corrected for area sampled as appropriate (SGZ, GCL, Hilus as described in figures) to yield a density per area. Images of Fluorojade C, CD68, Iba1, GFAP and CD31 as single labels were acquired using an automated slide scanner (Hamamatsu Photonics). CD68, Iba1, GFAP and CD31 single labels were all quantified using thesholded area analysis in ImageJ. For RNAscope analysis, Hoechst 33342/GFAP were used to identify stellate astrocytes and radial glia-like NSCs then Vegfa puncta were counted inside the nucleus of each cell.
Generating tagged VEGF constructs VEGF120 TAG and VEGF164 TAG were synthesized by Genscript using the known sequences of mouse VEGF120 and VEGF164 cDNA. A TAG stop codon inserted at bp100-102, corresponding to the . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; https://doi.org/10.1101/2023.04.24.537801 doi: bioRxiv preprint location after the 33 rd amino acid, within the second exon. VEGF120 and VEGF164 have identical sequences through this point in their coding sequence. This location was chosen because this is the location of an additional amino acid in human VEGF compared to mouse VEGF isoforms. Human and mouse isoforms of VEGF show great functional overlap, leading us to hypothesize that an insertion here would not disrupt the conformation of mouse VEGF. Human VEGF protein sequence from signal peptide (in parentheses) to amino acid 40 is (MNFLLSWVHW) SLALLLYLHH AKWSQAAPMA EGGGQNHHEV. The parallel mouse VEGF protein sequence is (MNFLLSWVHW) TLALLLYLHH AKWSQAAPTT EGEQKSHEVI. The TAG-inserted constructs would code for (MNFLLSWVHW) TLALLLYLHH AKWSQAAPTT EGE*QKSHEVI, where * indicates the TAG amber stop codon. These constructs were inserted downstream of an Ef1α promoter in a viral expression backbone, with a 6x His tag added at the end of the coding sequence. These constructs are in the process of being deposited at Addgene for public use.
Synthesis of VEGF TAG proteins Human embryonic kidney (HEK293T) cells (ATCC, #CRL-3216) were cultured in standard conditions on uncoated flasks: DMEM with high glucose, pyruvate, GlutaMAX (Fisher, #10-569-010) with 10% FBS (Fisher, #10-437-028), 1x MEM non-essential amino acids (Fisher, #11-140-050) and Pen/Strep (Fisher, #15140-122). In commercial production, recombinant proteins are often synthesized using bacterial cells. However, post-translational modifications can differ greatly between prokaryotes and eukaryotes. We therefore chose to use mammalian cells (HEK cells) to generate proteins that were as similar to endogenous mammalian VEGF isoforms as possible. Cells were transfected using lipofectamine 3000 (Fisher, #L3000015) according to manufacturer instructions with: VEGF120 TAG , VEGF164 TAG and/or PylRS/tRNApyl (Addgene #105830 from 29 ). TCO*A (SiChem, #SC-8008) was diluted 1:4 in 10X PBS then spiked into culture dishes to yield a final concentration of 1 mM. One day later, media was changed to serum-free media with 1 mM TCO*A. 1 day later, CM was collected and concentrated 10x with Amicon Ultra 3 kDa filter (Millipore, #UFC900308) according to manufacturer instructions. We chose to isolate protein from CM to capture the secreted form of VEGF after all posttranslational modifications prior to secretion were complete. CM concentrate was incubated with High-Capacity Ni-IMAC Magnetic Beads (Invitrogen, # A50588) to select His-tagged proteins according to manufacturer instructions. Bead supernatant was click reacted with tetrazine-biotin (50 µM) then dialyzed 3h and then overnight in 0.01 M PBS using G2 Dialysis Cassettes 3.5K MWCO (PI87722, Fisher) to eliminate free tetrazine-biotin. Dialysate was then concentrated with Amicon Ultra 3 kDa filter (Millipore, #UFC900308) according to manufacturer instructions and stored at -80˚C until use. Vehicle control was CM from HEK cells transfected with PylRS/tRNApyl only and processed in parallel with tagged CM samples.
To confirm the efficacy and specificity of tagging, samples were separated by PAGE in 4-15% Mini-PROTEAN® TGX™ Precast Protein Gels (BioRad, #4561084) for 1hr at 120V in Tris/Glycine/SDS running buffer (BioRad, #161-0772). Proteins were transferred to nitrocellulose overnight in 20% methanol in Tris/Glycine transfer buffer (VWR, #97061-382). Total protein was visualized using the LI-COR REVERT™ Total Protein Stain (Licor, #926-11021) with IR imaging on a Licor Odyssey Clx (Licor). Total protein was reversed, and membranes were reacted with streptavidin-IR800 (VWR, #103011-496) for 1h at room temperature, followed by rinsing in TBS-t and blocking in 5% milk in TBS-t for 1h at room temperature. Membranes were then incubated in primary antibody in blocking solution overnight at 4˚C. The next day, membranes were rinsed 3x in TBS-t, incubated in secondary antibody diluted 1:10,000 in blocking solution and rinsed 3 more times before visualization on a Licor Odyssey Clx imager.
Stereotaxic infusion of VEGF TAG proteins . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; Adult male and female C57Bl6 mice were anesthetized by isoflurane inhalation (Akorn, 4% induction, 2% maintenance) in oxygen. After mounting in a stereotaxic frame (Stoelting), alcohol and betadine were used to sterilize the scalp. After exposing the skull using a scalpel blade, a unilateral hole was drilled at approximate 2.0 mm posterior to bregma. Hamilton syringes were lowered to the following position: A/P -2.0 mm, M/L +2.0 mm, D/V -1.7 mm relative to bregma. 2.5 ng of VEGF TAG or matched volume of vehicle control sample was infused unilaterally (right and left side counterbalanced across animals) at a rate of 0.1 µl/min using an automated injector (Stoelting). Mice were administered carprofen prior to surgery. If mice were in the 1h, 3h or 6h groups, they also received buprenorphine immediately post-surgery. At their designated timepoint, mice were perfused as described above. A total of 59 animals received unilateral infusion (Veh: 12 mice; VEGF120 TAG : 24 mice, and VEGF164 TAG: 23 mice).
Detection and quantification of VEGF TAG proteins in vivo Fixed brains were sliced into 40 µm coronal slices and stored in cryoprotectant at -20˚C. Slices were later rinsed 3x with PBS before mounting the entire rostral-caudal extent of the hippocampus on superfrost plus microscope slides precleaned (Fisher, #12-550-15). After drying overnight, slices were incubated in streptavidin-IR800 (VWR, #103011-496) diluted 1:2000 in PBS for 1 hr then rinsed 3 more times in PBS before drying and coverslippping with Invitrogen ProLong Gold Antifade Reagent (Fisher, #P36934). Slides were imaged on a Licor Odyssey Clx imager. Area of IR800 signal was detected and measured using the Licor Small Animal Imaging Analysis key (Licor, #2000-502), with detection thresholds set at standard deviation 4.0 relative to tissue background and search limit 50 pixels. The needle track was used to set the epicenter of the infusion, which included the observed center of the track plus 4 slices rostral and caudal. To correct for area detected from tissue autofluorescence in the DG, vehicle sample average area per slice was subtracted from all areas to express signal beyond background. Of the 59 animals infused, 56 survived to perfusion and needle tracks were confirmed to be in the hippocampus for 42 mice: Veh: 2 10 min mice, 2 1h mice, 2 3h mice, 3 6h mice; VEGF120 TAG : 4 10 min mice, 4 1h mice, 4 3h mice, 4 6h mice, and VEGF164 TAG: 5 10 min mice, 4 1h mice, 4 3h mice, 4 6h mice. Vehicle mice showed no differences across timepoints and were collapsed into a single group for analysis. RNAscope© in situ hybridization 3d after the last TAM injections, Wt and VEGF iKD mice were transcardially perfused with ice cold PBS and 4% paraformaldehyde. Tissue processing is further described in 57 . Briefly, brains were harvested and fixed overnight in 4% paraformaldehyde followed by serial equilibration in 10-30% sucrose. Tissue was snap frozen in OCT and stored at -70˚C until sectioning in to 12 µm slices on a cryostat (Leica). Thaw-mounted sections were stored at -70˚C. RNAscope Multiplex Fluorescent v2 Assay (Advanced Cell Diagnostics, Newark, CA, United States) was performed with probe against mouse Vegfa (Mm-Vegfa-ver2, ACD 41226) according to manufacturer recommendations followed by immunolabeling for GFAP. Immunolabeling was similar to that described above with the following exceptions: blocking = 10% normal donkey serum in TBS-1% BSA, antibody diluent = TBS-1% BSA, washes = TBS-t.
Behavioral testing Mice were all handled by experimenters for at least 3 days before beginning testing. Both male and female experimenters were used and the same team members habituated the mice to handling and ran behavioral tests. The Y-maze test was conducted using an opaque, light-grey, arena made of extruded acrylic sheets with three identical arms of equal dimension (37 cm × 6 cm × 13 cm) angled 120° away from each other from a central point. Outside the arena, spatial cues were set up on each of the four sides of the rectangular table upon which the maze was set up. Mice were released into a designated "release arm," and an opaque curtain was set up . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; between the testing area and the experimenters. Four mice were trained or tested at once, with a camera set up centrally on the ceiling. During the Y-maze training (15 min), mice were able to roam freely in the release arm and one other arm, with access to the third arm blocked by a rectangular insert made of the same material as the arena. During the ITI (2 h), mice were returned to their experimental cages and placed back into the procedure room. Following the ITI, mice were returned to the Y-maze for 5 min with access to all three arms open. Once the test was complete, mice were once again returned to their home cages. Between each group of mice, each Y-maze arena was thoroughly wiped with 70% alcohol and completely dried.
After Y maze testing, mice were habituated to an open field, which was an opaque, light-grey, rectangular arena (40 cm × 39 cm × 30 cm) made of extruded acrylic sheet. During the habituations (3 x 5 min), each mouse was released into the designated release corner of the arena, with access to the entire, empty space. Four mice were habituated at once, with a camera set up centrally on the ceiling. Between each habituation, each spent a 50-minute ITI singly housed in their experimental cage in the procedure room. Between each group of mice, each OLT arena was thoroughly wiped with 70% alcohol and completely dried. At the end of the third habituation, mice were returned to their original cages.
The day after Y maze and open field habituation, mice were returned to the open field for novel object location testing (OLT), which was performed similarly to our previous work 58 . During training, the arena contained two objects, each one 9 cm away from each of the two closest walls of the arena. A 6.5 cm-tall, plastic cone filled with orange sand and a 9 cm-tall, plastic elephant toy were used as the objects. Each mouse was released into a designated release corner into the same arena in which they were habituated and allowed to explore for 10 min. Four mice were trained or tested at once, with a camera set up centrally on the ceiling. During the ITI (1 h), mice were returned to their home cages and placed back into the procedure room. The arenas were cleaned and one object was moved to a new corner-adjacent location. After the ITI, mice were returned to the arenas and allowed to explore for 10 min. Mice were then returned to their home cages. Between each group of mice, each OLT arena and all objects were thoroughly wiped with 70% alcohol and completely dried.
After the OLT task, mice were placed in the closed arm of an elevated plus maze and allowed to explore for 10 min. The elevated plus maze consisted of four arms-two open (6 cm × 34.5 cm) and two enclosed (6 cm × 34.5 cm × 21.5 cm). The arms were angled 90° from a central platform (6 x 6 cm).
Behavioral analysis Behavior was scored manually by 1-2 blinded observers in video recordings of all tasks. In the Y maze, the duration of time spent in each arm, frequency of visits to each arm, and latency to investigate the novel arm were all scored manually. Percent time in the novel arm was calculated relative to total time novel and familiar arm (excluding time in the release arm). In OLT, investigation time with each object was scored manually. Object investigation was defined as a mouse's nose being towards and within 2 cm of the object, as described in our previous work 58 . Climbing the object was not considered as object investigation. Percent investigation time was calculated as (investigation time of moved object) / (investigation time of both stationary and moved objects). In EPM, time in open arms and entries in to closed and open arms were scored manually. An entry into an arm was counted when the mouse placed two or more paws into the new area.
Local field excitatory postsynaptic potentials (fEPSPs) were recorded from the molecular layer of the dentate gyrus with borosilicate glass electrodes (1.5-3 MΩ, filled with aCSF) and evoked by electrical stimulation of the afferent fibers of the medial perforant path (MPP) located in the upper blade of the DG. Stimulation pulses (100 µS duration, every 30 s) were generated by an isolator (Iso-flex, A.M.P.I.) under computer control and delivered through a custom-made twisted nichrome wire stimulating electrode. Input-output (I/O) curves were generated for each slice with stimulation intensity varying from 0-1.0 mA at each 0.1 mA step. The stimulation intensity which evoked ~ 50% of the maximum response, without emergence of population spikes, was chosen to investigate LTP and the Paired-Pulse Ratio (PPR). The PPR were recorded at 6 time intervals (30,50,100,150,200,250 mSec). Synaptic field potentials were low-pass filtered at 1 kHz and digitally sampled at 50 kHz with Axopatch 200B amplifier and Digidata 1440 A interface. The baseline was monitored for 20 min to establish a consistent response. High frequency stimulation consisted of four sets of 100 pulses, each delivered in 1 second (100Hz) and administered in 20 second intervals. The LTP recording lasted 60 minutes after high frequency stimulation. The slope (20-80%) of the evoked fEPSPs was measured and normalized to the average baseline 5 min prior to delivery of high frequency stimulation. Data were monitored on-line and analyzed off-line using Clampex 10.6 software.
VEGF ELISA After PBS perfusion, whole DG was dissected and flash frozen on dry ice then stored at -20˚C. Protein was extracted from whole DG by homogenization in RIPA buffer (Fisher, #PI89900) with Halt Protease and Phosphatase Inhibitor (Fisher, #PI78440). Samples were incubated on ice for 10 min then freeze-thawed using dry ice and wet ice respectively 3 times. Samples were centrifuged at 14,000 rpm for 5 min at 4˚C. Total protein content of supernatant was determined using a BCA kit (Fisher, #PI23225). 40 µg total protein was diluted in assay buffer and quantified using the R&D Systems Duo set VEGF ELISA according to manufacturer instructions.

Real time qPCR
After PBS perfusion, whole DG was dissected and flash frozen on dry ice then stored at -80˚C. RNA was extracted using the RNeasy Mini Kit (Qiagen, #74104) according to manufacturer's instructions. RNA was DNase I (Invitrogen, # 18068015) treated and then converted to cDNA using the SuperScript III Frist Strand Synthesis System (Fisher, # 18080051). cDNA diluted 1:5 in water was quantified using either SYBR Green I (Roche, # 04707516001) and a LightCycler 480 (Roche) or SsoAdvanced-Universal SYBR Green Supermix (Biorad, # 1725274) and a BioRad CFX96 Touch Real-Time PCR Detection System. ΔΔCt values were calculated relative to housekeeping genes (actin and or sdha) and converted to log2fold change over control.
Statistical analysis Statistical analysis was performed using GraphPad Prism. Specific tests used are detailed in figure legends. Sample sizes were always individual mice unless otherwise noted and are detailed in figure legends. In general, if 2 groups were compared, student's t-tests or Mann-Whitney tests were used. If more than 2 groups were being compared within one factor, ANOVA was used with posthoc error corrected tests. If more than 2 groups were being compared with 2 or more factors, 2 way ANOVA was used with posthoc error corrected tests. Y maze arm first arm entry was compared using Fisher's exact test. Spatial dispersal of tagged VEGF was compared using 95% confidence intervals because we had no apriori hypothesis in this experiment. P<0.05 was defined as significant.

Supplemental Figure 5. NSPC-VEGF knockdown exacerbates excitotoxic injury. A)
Average, maximum and mode seizure score of Wt and iKD mice according to the Racine scale in the 4h following KA injection. T-tests within each category all ns. Mean ± SEM and individual mice shown. N = 7-8 mice/genotype. B) Density of BrdU+cFos+ cells in the granule cell layer 1d after KA injection in Wt and iKD mice. 2 way ANOVA treatment p < 0.0001. C) Fold change in GFAP immunoreactivity (ir) thresholded area in the DG of Wt and iKD mice 1d after KA relative to Wt-Veh mice. 2 way ANOVA all ns. Mean ± SEM and individual mice shown. N = 12-15 mice/group. D) Representative images of GFAPir in the DG of Wt and iKD mice. Scale = 20 µm. E) Schematic of treatments. NestinCreER T2+/-;LoxP-STOP-LoxP-EYFP +/mice received 5d TAM injections followed by a single KA injection 3d after the last TAM injection. Mice were perfused 1d after KA. F) Pie graph showing percent of EYFP+ cells in the DG that where phenotypic NSCs, IPCs, astrocytes or other 1d after KA. N = 3 mice. G) Representative image of EYFP col-labeling with Sox2 and GFAP used to identify astrocytes, NSCs and IPCs. Scale = 20 µm. H) Number of NestinCreER T2-/and NestinCreER T2+/mice showing any FJC labeling in the DG 1 d after KA. Fisher's exact test ns. I) Fold change in CD68 immunoreactivity (ir) thresholded area in the DG of NestinCreER T2-/and NestinCreER T2+/mice 1d after KA relative to NestinCreER T2-/-Veh treated mice. 2 way ANOVA treatment p = 0.0011. Mean ± SEM and individual mice shown. N = 9-14 mice/group. . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 24, 2023. ; https://doi.org/10.1101/2023.04.24.537801 doi: bioRxiv preprint