Spatiotemporal Regulation of De Novo and Salvage Purine Synthesis during Brain Development

Visual Abstract


Introduction
The spatiotemporally regulated proliferation of neural stem cells and the migration of newborn neurons are crucial for forming the mammalian cerebral cortex.In the embryonic mouse brain, neural stem cells actively and symmetrically proliferate in the ventricular zone (VZ) to expand their original pool, eventually dividing asymmetrically to generate either intermediate progenitor cells (IPs) or neurons.Here, the neural stem cells in the VZ and the IPs in the subventricular zone (SVZ) are collectively referred to as neural stem/progenitor cells (NSPCs).Numerous glutamatergic neurons generated from NSPCs migrate radially to form the stratified layers of the cerebral cortex (Agirman et al., 2017).In addition, many GABAergic inhibitory neurons, generated in the embryonic ganglionic eminence (GE), tangentially migrate, finally integrating in the cerebral cortex (Lavdas et al., 1999;Corbin et al., 2003;Cho et al., 2014).Defects in radial migration frequently cause brain malformation and psychiatric disorders, including mental retardation and epilepsy (Moffat et al., 2015).
Purines are the essential building blocks of DNA/RNA, energy sources of enzymatic reactions (ATP/GTP), and second messengers in various intracellular signaling cascades (cyclic AMP and cyclic GMP).Moreover, adenosine and ATP are involved in extracellular signaling mediated by purinergic receptors, regulating cell functions, such as cell migration, apoptosis, NSPC proliferation, and neuron differentiation (Lin et al., 2007;Huang et al., 2021).
In fact, we recently demonstrated that a defect in purinosome formation leads to malformation of the cerebral cortex (Yamada et al., 2020).Other studies also provided evidence regarding a close relationship between purine production and the mammalian/mechanistic target of the rapamycin complex 1 (mTORC1) signaling pathway (Ben-Sahra et al., 2016;Hoxhaj et al., 2017); however, the molecular interplay between mTORC1 signaling and purine metabolism in cortical development remains unclear.
Furthermore, abnormalities in the purine metabolism are associated with the etiology of many diseases, including gout.For instance, inherited deficiencies in de novo enzymes result in fetal lethality or neurologic diseases in humans (Jurecka, 2009).For instance, a missense mutation in PAICS causes multiple malformations and early neonatal death (Pelet et al., 2019), and ADSL or ATIC deficiency causes various neurodevelopmental phenotypes, including epilepsy, speech impairment, and auto-aggressive behavior (Marie et al., 2004;Jurecka et al., 2015;Dutto et al., 2022).With respect to the salvage pathway, HGPRT deficiency causes the Lesch-Nyhan syndrome, characterized by juvenile gout, dystonia, mental retardation, and compulsive self-injurious behavior (Lesch and Nyhan, 1964;Torres and Puig, 2007).Although HGPRT deficiency is thought to cause functional impairment of the dopamine system in the basal ganglia (Visser et al., 2000), no fundamental therapeutic approach has been established.
pathways is critical for healthy brain development.However, the spatiotemporal regulation of purine production pathways in CNS development remains unknown.Here, we report the expression profile and the functional significance of purine synthesis enzymes in the developing brain.Each purine pathway is driven by its own temporal and regional traits during brain development.

Animals
ICR and C57BL/6J mice (Japan SLC Inc.) were housed under temperature-controlled and humidity-controlled conditions under a 12/12 h light/dark cycle, with ad libitum access to food and water.The date of conception was established using a vaginal plug and recorded as embryonic day zero (E0.5).The day of birth was designated as P0.

Tissue preparation
Embryonic and adult mice of either sex were perfused through the cardiac ventricle with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (pH 7.4), followed by postfixation overnight at 4°C (Yamada and Sakakibara, 2018).Fixed brains were cryoprotected in 30% sucrose in PBS overnight at 4°C and embedded in the optimal cutting temperature compound (Sakura Finetek).Frozen sections of embryonic and postnatal brains were cut into 14-mm thickness sections using a cryostat and collected on MAScoated glass slides (Matsunami Glass).Free-floating sections of adult mice brains were cut at 30-mm thickness.

Immunohistochemistry
Immunohistochemistry (IHC) was performed as described previously (Yamada and Sakakibara, 2018).For antigen retrieval, frozen sections were heated at 90-95°C for 10 min in 10 mM sodium citrate buffer (pH 6.0) using a microwave and treated with 1% H 2 O 2 in PBST (0.1% Triton X-100 in PBS) for 30 min at 20-25°C to suppress the endogenous peroxidase activity.Sections were blocked for 2 h in 10% normal goat serum in PBST and incubated overnight at 4°C with the primary antibodies in the blocking buffer.Then, the sections were incubated with biotinylated anti-rabbit IgG (Vectastain ABC Elite; Vector Lab) in PBST at room temperature for 2 h, followed by incubation with the ABC reagent (Vectastain ABC Elite; Vector Lab) in PBST for 2 h.The horseradish peroxidase signal was visualized using 2.25% diaminobenzidine and 2.25% H 2 O 2 (Peroxidase Stain DAB kit; Nacalai tesque).Each step was followed by four washes in PBST.Free-floating sections were mounted on MAS-coated glass slides, dehydrated, and coverslipped with Entellan New (Merck Millipore).
Double immunostaining was performed as described previously (Yamada and Sakakibara, 2018;Yamada et al., 2020).Frozen sections were blocked for 2 h with 5% normal goat or donkey serum in PBST, followed by incubation with the primary antibodies in blocking buffer at 4°C overnight.After washing with PBST four times and twice with PBS, sections were incubated for 2 h with Alexa Fluor 488-conjugated or Alexa Fluor 555-conjugated secondary antibodies (Thermo Fisher Scientific).After counterstaining with 0.7 nM Hoechst 33342 (Thermo Fisher Scientific), sections were mounted and imaged using a confocal (FV3000, Olympus) or fluorescence inverted microscope (Axio Observer, Zeiss).For BrdU staining, sections were treated with 2 N HCl for 30 min at 42°C for DNA denaturation, followed by treatment with 0.1 M sodium borate buffer (pH 8.5) for 10 min.After washing with PBST four times and twice with PBS, anti-BrdU immunostaining was performed.

Immunocytochemistry
Cultured cells were fixed with 4% PFA and permeabilized with 0.1% Triton X-100.For BrdU staining, cells were treated with 2 N HCl for 1 h at 37°C, followed by neutralization with 0.1 M sodium borate buffer (pH 8.5) for 30 min.After blocking with 5% normal donkey serum, cells were incubated with primary antibody for 1 h at room temperature, followed by Alexa Fluor 488-conjugated or 594-conjugated secondary antibodies (Thermo Fisher Scientific) for 2 h.Immunofluorescent signals were acquired using an inverted microscope (Axio Observer, Zeiss) equipped with a CCD camera (AxioCam MRm, Zeiss).

Statistical analyses
Statistical analyses were performed using the R package version 4.2.0.All numerical data are expressed as mean 6 SEM.A one-way ANOVA followed by Weltch's t test with Holm-Bonferroni correction was used in multiple-group comparisons.Welch's t test was used to assess the number of BrdU 1 , pH3 1 , or cleaved-caspase3 1 cells in vivo and pS6/S6 ratio.In in vivo experiments involving inhibitors, three individuals were tested per condition, and three images from each individual were randomly selected for statistical processing.The x 2 test was used to compare the number of BrdU 1 Nestin 1 , BrdU 1 GFP 1 , or cleaved-caspase3 1 Hoechst 1 cells in vitro.*p , 0.05, **p , 0.01, ***p , 0.001 were considered statistically significant.The number of BrdU 1 NSPCs in vitro was counted using MATLAB [version 9.12.0.1884302 (R2022a)].The detailed statistical analyses are shown in Table 1.

Two purine synthesis pathways are activated during brain development
To investigate the expression of purine synthesis enzymes in the developing CNS, immunoblotting analysis was performed on embryonic, postnatal, and adult whole brains.PAICS and FGAMS, both catalyzing the de novo purine synthesis (Fig. 1), were abundant in the embryonic stages from E10.5 to E16.5 during the active NSPC proliferation.Subsequently, from the postnatal to the adult stage, the expression of these de novo enzymes was downregulated (Fig. 2A).Conversely, the expression of HGPRT, a key enzyme catalyzing the salvage pathway (Fig. 1), was relatively low in embryonic brains, gradually increasing in postnatal and adult stages (Fig. 2A).Although HGPRT was expressed as a 25 kDa product with the expected size throughout life, an additional product of 35 kDa was detected during the embryonic and neonatal (E13-P0) period.Whether this slower-migrating band reflects another isoform or a post-translationally modified HGPRT product is unclear.To examine the expression properties of each brain region, we further divided the brain into the cerebral cortex (E13.5-P12) and cerebellum (P2-P12) and then performed the same experiments (Fig. 2B,C).
The expression patterns of PAICS and HGPRT switched around the time of birth (P0) in the developing cerebral cortex, i.e., high PAICS expression was observed in early embryonic stages, whereas HGPRT expression increased after the postnatal period (Fig. 2B,C).Conversely, both PAICS and HGPRT were expressed at high levels in the P2 and P5 cerebellum, in which NSPCs actively divide to produce granule cells (Fig. 2B,C).Subsequently, HGPRT expression continued to increase in the cerebellum until P12 (Fig. 2B,C).These findings indicated that the choice of purine pathway depends on the stage of development and region of the brain.
We next analyzed the expression of these enzymes in primary cultured NSPCs isolated from E12.5 cerebral cortices and the external granular layer (EGL) of the P2 cerebellum.Figure 2D shows that PAICS and FGAMS were abundantly expressed in Pax6 1 NSPCs derived from these two brain regions.In contrast, PAICS and FGAMS expression was decreased in primary cultured neurons (Fig. 2D) isolated from E16.5 cerebral cortices or P4 cerebella and cultured for 10 div until fully differentiated.These observations appear consistent with the in vivo expression profile of FGAMS and PAICS along brain development (Fig. 2A-C).On the other hand, HGPRT was expressed in cultured NSPCs as well as the differentiated neurons (Fig. 2D).We further analyzed the expression of these enzymes in primary cultured GFAP 1 astrocytes, which were differentiated from the embryonic cerebral cortex and postnatal cerebellum.However, the expression of both HGPRT and PAICS was barely detectable in astrocytes (Fig. 2E).The purity of the NSPCs and astrocytes used in these assays was confirmed by the expression of Pax6, a marker for embryonic apical progenitor cells/radial glial cells (RGCs), and GFAP, a marker for astrocyte, respectively (Fig. 2D,E).In the cerebellum, the Pax6 protein is expressed in differentiated granular cells as well as cerebellar NSPCs (Yamasaki et al., 2001).Consistently, we detected Pax6 in the cultured cerebellar neurons (Fig. 2D).Together, these results suggested that the de novo pathway is preferentially driven during the embryonic stage when NSPC proliferation occurs, and that purine metabolism switches from the de novo pathway to the salvage pathway over brain development.

Expression of purine synthesis enzymes in the developing and adult CNS
The distribution and localization of PAICS, FGAMS, and HGPRT were examined immunohistochemically in the developing and matured mouse brain.At E13.5, three enzymes were strongly expressed in the VZ/SVZ of the cerebral cortex, where NSPCs localize (Fig. 3A-C).These immunoreactivities in the VZ/SVZ considerably decreased until E18.5 and P0 (Fig. 3A-F,J-L), decreasing until disappearing in the SVZ of the adult cerebral cortex (Fig. 3P-R).The development of the cerebellum proceeds rapidly from late embryonic stages.Granule neurons are produced from NSPCs in the superficial EGL covering the cerebellar cortex and migrate to form the internal granule layer (IGL; Yacubova and Komuro, 2003).We observed robust HGPRT immunoreactivity in the EGL at E18.5 and P0; meanwhile, the PAICS and FGAMS expression level in the EGL was relatively weak (Fig. 3G-I,M-O, arrows).These observations might indicate that embryonic NSPCs actively drive purine synthesis pathways to meet the high demand at the beginning; as neurogenesis progresses though, the activity of the purine synthesis pathway would decline.We further detected HGPRT expression in CD31 1 endothelial cells of blood vessels in the brain parenchyma from E13.5 to P0.5 (Fig. 3S; Extended Data Fig. 3-1), but HGPRT was not observed in blood vessels in the adult brain.These results suggested a specific role of the salvage pathway in vasculogenesis in the embryonic brain.
In the adult brain, PAICS, FGAMS, and HGPRT were widely detected in neuronal somata and neuropil of discrete areas (Extended Data Figs.3-2 and 3-3).In the brain stem, the locus coeruleus, Edinger-Westphal nucleus, and vestibular nucleus expressed these three enzymes.Moreover, strong immunoreactivities for PAICS and FGAMS were differentially observed in several nuclei of the brain stem, including the lateral cerebellar, red, facial, medial vestibular, and ambiguous nucleus (Extended Data Figs.3-2 and 3-3).On the other hand, HGPRT was preferentially expressed in numerous nuclei in the forebrain and diencephalon, including the posterior interlaminar thalamic nucleus, polymorph layer of the dentate gyrus, arcuate nucleus, dorsomedial hypothalamic nucleus, zona incerta, lateral septal nuclei, and neurons in the nigrostriatal bundle (Extended Data Figs.3-2 and 3-3).These findings implied that the purine pathway to be activated depends on the developmental stage and brain region involved.

Inhibiting de novo purine synthesis affects NSPC proliferation
We sought to investigate the effects on brain development of purine synthesis inhibition during a specific period in embryonic and postnatal life.To this end, using genetically engineered animals, such as knockout mice and conditional knockout mice of purine synthesis enzymes, was limited in their ability to transiently suppress the enzyme activity in the mid or late embryonic period, times of active neurogenesis in the cerebral cortex.In this study, we used several inhibitors that target specific enzymes of purine synthesis pathways.These inhibitors allowed us to target a specific embryonic period and to easily adjust the degree of inhibition during corticogenesis depending on inhibitor concentration.
First, the role of purine pathways in NSPCs was explored using various specific inhibitors for the de novo and salvage enzymes.Accordingly, primary cultured NSPCs prepared from E12.5 cerebral cortex were incubated for 48 h with or without the following inhibitors (Fig. 1): MMF, an inhibitor of the inosine monophosphate dehydrogenase, which is the rate-limiting enzyme in de novo synthesis of guanosine nucleotides (Allison, 2005) or forodesine hydrochloride, an inhibitor of the purine nucleoside phosphorylase mediating the salvage pathway (Evans et al., 2018).Immunostaining with anti-Nestin, an NSPC marker and anti-BrdU revealed that MMF treatment significantly reduced the number of Nestin 1 BrdU 1 NSPCs, while forodesine treatment did not affect NSPC proliferation (Fig. 4A-D).MMF treatment led to no morphologic changes or enhanced differentiation into neurons in Nestin 1 NSPCs, suggesting that MMF exclusively affects the mitotic potential of NSPCs.Similar results were obtained in the NSPCs isolated from the cerebellar EGL (Extended Data Fig. 4-1).Next, we confirmed that the reduced NSPC proliferation was caused by repression of the de novo pathway using PAICS shRNAs.The silencing effect and specificity of PAICS shRNAs (#1 and #3) were validated in our previous study (Yamada et al., 2020).Consistent with the inhibitory effect of MMF on NSPC division, knocking down PAICS significantly decreased the number of BrdU-incorporated NSPCs (Paics shRNA #1, 37%, n ¼ 100; Paics shRNA #3, 6%, n ¼ 100, arrows) relative to the control (control shRNA, 67%, n ¼ 100, arrowhead; Fig. 4E-G).Based on these results, we concluded that activation of the de novo pathway is critical for NSPC proliferation.

The development of the cerebellum requires both purine synthesis pathways
Next, we examined the significance of the purine pathways on in vivo neurogenesis in the cerebellum.To this end, embryos at E16.5 or pups at P2 mice were treated with forodesine, MMF, or MTX.MTX is known to impede the de novo purine pathway at multiple steps by inhibiting the dihydrofolate reductase, which catalyzes the folate biosynthesis pathway (Fig. 1; Cronstein, 1997).Gross anatomic analysis 2 or 4 d after inhibitor administration showed no significant brain dysplasia.However, immunostaining with BrdU and Ki67, which marks cells with proliferative capacity, showed that the fraction of BrdU 1 and Ki67 À cells increased in the E18.5 cerebellar IGL treated with forodesine, MMF, and MTX compared with control (Fig. 5A-E; control, 20.6 6 2.7%, n ¼ 7; forodesine, 40.3 6 2.7%, n ¼ 8; MMF, 34.4 6 2.9%, n ¼ 7; MTX, 49.6 6 3.2%, n ¼ 9).Comparable results were observed in the P6 cerebellar IGL treated with forodesine or MMF (Fig. 5F-I; control, 27.1 6 1.5%, n ¼ 9; forodesine, 38.4 6 3.2%, n ¼ 7; MMF, 42.8 6 2.7%, n ¼ 9).These findings suggest that inhibition of purine synthesis in the cerebellum may accelerate the exit from the cell cycle in NSPCs, affecting cell fate determination and leading to impaired cerebellar development.Considering the abundant HGPRT expression in EGL (Fig. 3I,O), proper development of the cerebellum requires the salvage as well as the de novo pathways.

MMF suppresses NSPC proliferation and delays neuron migration in the cerebral cortex
To assess the impact of the purine pathway on neurogenesis in the embryonic cerebral cortex, MMF or forodesine was administered at E12.5, followed by BrdU administration at E13.5, and embryo analysis at E14.5.In the control, many BrdU-incorporated NSPCs exited the cell cycle and left the SVZ/VZ; about half of BrdU 1 cells were dispersed within the intermediate zone (IZ) and cortical plate (CP) at E14.5 (Fig. 6A,D; CP, 8.2 6 1.1%; IZ, 42.9 6 2.0%; SVZ/VZ, 48.9 6 1.3%, n ¼ 9).In contrast, MMF-treated embryos exhibited a substantial accumulation of BrdU 1 cells in the SVZ/VZ, with few BrdU 1 cells being located in the IZ/CP (Fig. 6C,D; CP, 0.1 6 0.1%; IZ, 27.7 6 3.5%; SVZ/VZ, 72.2 6 3.5%, n ¼ 9), indicating suppression of radial migration of newborn neurons.On the other hand, forodesine treatment had only a small effect on neuron migration (Fig. 6B,D; CP, 2.5 6 0.8%; IZ, 40.4 6 3.1%; SVZ/VZ, 57.1 6 3.1%, n ¼ 9).To determine whether the BrdU 1 cells remaining in the VZ/SVZ had lost their mitotic potential, we performed immunostaining with an antibody to phospho-histone H3 (pH3), which marks cells in late G 2 -M phase.Compared with controls, the number of pH3 1 NSPCs in the VZ facing the lateral ventricles was significantly reduced in MMF-treated embryos (Fig. 6E,G,H, arrows; control, 31.1 6 3.5, n ¼ 11; MMF, 19.0 6 1.4%, n ¼ 9).Meanwhile, the number of pH3 1 NSPCs in the VZ remained unchanged in forodesine-treated embryos (Fig. 6F,H; 34.2 6 1.9%, n ¼ 9).Immunostaining with anti-PAICS, FGAMS, and HGPRT showed that the expression and subcellular localization of each enzyme were not affected by treatment with these inhibitors, as all these enzymes were expressed in the cytoplasm in the VZ/SVZ (Fig. 6I-Q).Furthermore, treatment with these inhibitors did not induce early premature differentiation into astrocytes, as demonstrated by immunostaining with anti-GFAP (Extended Data Fig. 6-1).These results suggested that inhibition of the de novo pathway, but not of the salvage pathway, suppressed NSPC mitosis and the subsequent neuron migration from the VZ.

De novo purine pathway controls mTORC1/S6K/S6 signaling
We attempted to elucidate the molecular mechanism by which the de novo purine synthesis pathway regulates cortical development.Among various intracellular signaling pathways, we focused on mTOR, a serine/threonine kinase involved in the proliferation and growth of cells through the regulation of numerous cellular processes (Szwed et al., 2021).Previous studies have shown that mTORC1 stimulates purine nucleotide production (Ben-Sahra et al., 2016) as well as the inhibitory effect of purine nucleotide depletion on mTORC1 signaling (Hoxhaj et al., 2017).In addition, mTORC1 mutation affects the development of the cerebrum, suggesting a role of mTORC1 signaling in NSPCs (Tarkowski et al., 2019;Andrews et al., 2020).mTORC1 signaling includes two downstream cascades, namely eukaryotic initiation factor 4E (eIF4E)/binding protein 1 (4E-BP1) and S6K/S6 (Fig. 7C; Morita et al., 2013;Ben-Sahra et al., 2016).These two downstream cascades play an essential role in protein translation.Specifically, mTORC1 pS6K, which further phosphorylates various substrate proteins, including ribosomal subunit S6, thereby promoting translation.However, the molecular interplay between the two mTORC1 signaling cascades and purine metabolism in cortical development remains unclear.
Brain lysates were prepared from E12.5 embryos continuously treated with MTX, allopurinol, or MTX plus allopurinol, and the expression of the downstream proteins of mTORC1 signaling was examined.Allopurinol is an inhibitor of xanthine oxidase, which converts xanthine to uric acid, thus promoting purine synthesis via the salvage pathway (Fig. 1; Day et al., 2017).PAICS expression was decreased by MTX, but HGPRT expression showed no change; thus, MTX specifically inhibits de novo purine synthesis.For the S6K/S6 signaling cascade, we analyzed the expression of S6K, pS6K, S6, and pS6.As shown in Figure 7A, pS6K, S6K, and pS6 expression was completely inhibited by MTX treatment.Compared with the control, S6 protein expression remained unchanged in MTX-treated brains, indicating that MTX completely inhibits S6 phosphorylation and activation in the embryonic brain (Fig. 7B).These severe MTX-induced impairments of the S6K/S6 signaling cascade were restored by the concurrent administration of allopurinol and MTX (Fig. 7A), which eventually restored the phosphorylation level of S6 to control levels (Fig. 7B).Regarding another downstream pathway of mTORC1, 4E-BP1 inhibits cap-dependent translation by binding to the translation initiation complex eIF4E (Sun et al., 2019).Three 4E-BP1 isoforms (a, b , and g ) exist in mammalian cells that represent the phosphorylation status of 4E-BP1, with a being hypophosphorylated and b /g being hyperphosphorylated (Gingras et al., 1998).We found a mobility shift in MTX-treated brains from 4E-BP1-b /g to 4E-BP1-a (Fig. 7A, arrows; Extended Data Fig. 7-1B), although quantitative analysis revealed little difference in total 4E-BP1 protein expression (Extended Data Fig. 7-1A).A previous report consistently indicated that the accumulation of 4E-BP1-a was induced by MTX treatment in cultured HeLa cells (Hoxhaj et al., 2017).Considering that 4E-BP1-a can efficiently interact with and inhibit eIF4E (Gingras et al., 1998), the MTX-induced dephosphorylation of 4E-BP1 might repress eIF4E-dependent translation in the embryonic brain.
These results indicate that the de novo purine synthesis pathway is strongly associated with mTORC1/S6K/S6 signaling and is related to mTORC1/4E-BP1/eIF4E signaling to some extent.As the activation of the salvage pathway by allopurinol reversed the inhibition of mTORC1/S6K/S6 signaling, we considered that a steady-state level of purines supplied by the de novo pathway is essential for the maintenance of mTORC1 signaling within NSPCs in vivo (Fig. 7C).

Activation of mTOR signaling abrogates inhibition of the de novo pathway
Inhibition of de novo synthesis in NSPCs often induces cell death in addition to reducing mitotic potential (Fig. 4).We determined whether this apoptosis could be rescued by mTOR activation.For this purpose, NSPCs were treated with purine inhibitors along with MHY1485, an activator of the mTOR signaling pathway (Choi et al., 2012).The primary cultured NSPCs prepared from the E12.5 cerebral cortex were co-incubated with or without MHY1485, and the apoptotic rate was evaluated via immunostaining for cleaved-caspase3 at 6 or 48 h.Treatment with MMF resulted in a significant induction of apoptosis in Nestin 1 NSPCs, whereas treatment with forodesine showed no effect (Fig. 8A-C,E).This deleterious effect of MMF was entirely abolished by co-treatment with MHY1485 (Fig. 8D,E).Notably, neither MMF nor forodesine affected cell death in GFAP 1 astrocytes (Fig. 8F-J).These results suggested that apoptosis caused by inhibition of the de novo pathway does not simply result from impaired proliferative ability; instead, it represents a phenomenon attributable to cellular characteristics unique to NSPCs, such as multipotency, wherein the activation of mTORC1 signaling is critical for maintenance or survival.A-E, Primary cultured NSPCs from the E12.5 telencephalon were treated for 48 h with DMSO as a control (A), forodesine (50 mM; B), MMF (10 mM; C), or MMF and MHY1485 (10 mM; D).NSPCs were immunostained with anti-Nestin (green) and anti-cleaved-caspase3 (red) antibodies.E, Quantification of apoptotic NSPCs.The number of cleaved-caspase3 1 Nestin 1 cells was counted and represented as a ratio to the number of Hoechst 1 Nestin 1 cells at 6 (white bar) or 48 h (black bar).F-J, Astrocytes differentiated from NSPCs were treated for 48 h with DMSO (F), forodesine (50 mM; G), MMF (10 mM; H), or MMF and MHY1485 (10 mM; I).Astrocytes were immunostained with anti-GFAP (green) and anti-cleaved-caspase3 (red) antibodies.J, Quantification of apoptotic astrocytes.The number of cleaved-caspase3 1 GFAP 1 cells was counted and represented as a ratio to the number of Hoechst 1 GFAP 1 cells.ns, not significant, *p , 0.05, ***p , 0.001, x 2 test with the Holm-Bonferroni correction.control, n ¼ 313 (E, 48 h), 236 (E, 6 h), 322 (J); forodesine, n ¼ 159 (E, 6 h), 237 (E, 48 h), 746 (J); MMF, n ¼ 284 (E, 6 h), 224 (E, 48 h), 354 (J); MMF and MHY1485, n ¼ 100 (E, 6 h), 148 (E, 48 h), 292 (J).Scale bars, 50 mm.

Inhibiting de novo pathway causes forebrain-specific malformations
To determine how the chronic inhibition of the purine pathway impacts brain formation in vivo, different inhibitors (MMF, forodesine, MTX, and allopurinol) were continuously administered to embryos from E9.5 to E11.5, with or without MHY1485, and their brain architecture was analyzed at E12.5.As shown in Figure 9, embryos treated with forodesine, MHY1485, or allopurinol alone appeared to follow normal brain development (Fig. 9A,C,E,F; Extended Data Fig. 9-1).On the other hand, a drastic anomaly was reproducibly observed in MMF-treated brains, in which the lateral ventricles widely opened into the third ventricle, probably because of forebrain hypoplasia (Fig. 9B).The combined administration of MMF and MHY1485, an activator of the mTOR signal, could restore this gross malformation (Fig. 9D).MTX-treated embryos exhibited a more severe defect with complete loss of the forebrain region (Fig. 9G).This systematic abnormality was extremely severe and could not be reversed by the combined administration of MTX and allopurinol, thereby enhancing the salvage pathway (Fig. 9H).
These observations further supported that the de novo purine synthesis pathway is predominantly involved in the development of the cerebral cortex, as the de novo synthesis inhibition impacted forebrain formation in early neural development.
We performed a detailed histologic analysis to gain insights into the cellular mechanism of the brain abnormality caused by the suppression of de novo purine synthesis.As shown in Figure 10, MMF-treated embryos showed a significant increase in the number of cleaved-caspase3 1 apoptotic cells in the rostral region of the cerebral cortex, whereas a considerably lower number of apoptotic cells were detected in the caudal forebrain region (Fig. 10A,B,AF; control, 7.2 6 1.2, n ¼ 4; MMF rostral area, 47.3 6 7.7, n ¼ 8; MMF caudal area, 18.8 6 2.7, n ¼ 8).Co-treatment with MHY1485 successfully rescued the morphologic phenotype induced by MMF and reduced the number of apoptotic cells, thereby restoring the normal histologic structure of the brain (10.8 6 2.9, n ¼ 9; Fig. 10C,AF).As illustrated in Figure 9G, the consecutive administration of MTX led to severe cerebral   E, J, O, T, Y) was administered to pregnant mice between E9.5-11.5 and analyzed at E12.5.Horizontal frozen sections were immunostained with antibodies to cleaved-caspase3 (red; A-E), Pax6 (green) and DCX (red; F-J), Tbr2 (green; K-O), Ki67 (red; P-T), pH3 (green; U-Y), Nestin (green; Z-AB), or GSH2 (green; AC-AE).The forebrain region treated with MMF was divided by dotted lines into the rostral (Pax6 -GSH2 1 ) and caudal area (Pax6 1 GSH2 -).The asterisk in G denotes the expanded cortical layer filled with DCX 1 immature neurons.AF, AG, Quantification of cleaved-caspase3 1 (AF) or pH3 1 (AG) cells in the cerebral cortex (number of cells per 500-mm width of the cortex).In embryos treated with MMF, cells were counted individually in the rostral and caudal areas.ns, not significant, *p , 0.05, **p , 0.01, Welch's t test followed by Holm-Bonferroni correction.Scale bars, 100 mm (A-AE).AH, AI, Expression of cleaved-caspase3.Embryos were treated with MTX or MTX and MHY1485 during E9.5-E11.5, and the brains were subjected to malformations accompanied with an increase in cell death and positivity for cleaved-caspase3.This cell death was entirely ameliorated by treatment with MHY1485, as determined by Western blotting using anti-cleaved-cas-pase3 (Fig. 10AH,AI), implicating that the MTX-induced brain phenotype depends on the mTORC1 pathway.In contrast, neither the continuous inhibition of the salvage pathway by forodesine nor the activation of the salvage pathway by allopurinol had any effect on brain formation (Fig. 10D,E).In addition, forodesine treatment did not affect cell death (Fig. 10D,AF; 16.4 6 2.2, n ¼ 6).Interestingly, allopurinol treatment resulted in a marked increase in the number of apoptotic caspase3 1 cells throughout the cerebral cortex (Fig. 10E,AF; 42.9 6 5.6, n ¼ 9).Because uric acid is a major antioxidant and its neuroprotective effects reduce the risk of neurologic diseases (Kachroo and Schwarzschild, 2014), suppressing uric acid production with allopurinol can potentially affect the developing brain.
Next, we performed double immunostaining with anti-Pax6 and anti-DCX, specific markers for RGCs in the cerebral cortex and postmitotic newborn neurons, respectively.The embryonic forebrain typically contains a VZ occupied by Pax6 1 NSPCs and an IZ/CP occupied by DCX 1 neurons, as shown in the DMSO-treated control.The structure of the embryonic forebrain treated with forodesine, allopurinol, or MHY1485 was equivalent to that of the control brain, and the spatial distribution of the two cell populations (Pax6 1 NSPCs and DCX 1 neurons) appeared unchanged (Fig. 10F,I,J; Extended Data Fig. 10-1).In contrast, in the forebrain of MMF-treated embryos, the distribution of Pax6-expressing and DCX-expressing cells was disrupted and drastically affected: Pax6 expression remained in the caudal area of the forebrain but was completely absent in the rostral and dorsal parts (Fig. 10G).With respect to DCX 1 neurons, although their distribution was relatively preserved throughout the forebrain, the DCX 1 cortical layer was significantly expanded and thickened at the boundary between the caudal and rostral area (Fig. 10G, asterisk).Furthermore, the expression of T-box brain protein 2 (Tbr2; Englund et al., 2005), an IP marker in the SVZ, was lost in the rostral area of MMF-treated embryos, similar to Pax6 (Fig. 10K-O).Nonetheless, the number and distribution of Ki67 1 or pH3 1 proliferating cells in rostral area were unaffected (Fig. 10P-Y,AG; control, 32.7 6 3.0, n ¼ 4; MMF rostral area, 27.9 6 1.6, n ¼ 8; MMF caudal area, 27.3 6 1.8, n ¼ 8; forodesine, 25.9 6 1.8, n ¼ 6; allopurinol, 24.6 6 3.1, n ¼ 9).Interestingly, co-injection of MMF and MHY1485 increased the number of pH3 1 proliferating cells in VZ (Fig. 10W,AG; 50.9 6 4.9, n ¼ 9).Immunostaining with Nestin, a universal RGC marker in all brain regions, unexpectedly revealed that Nestin 1 RGCs were uniformly distributed in rostral and caudal areas in MMF-treated brains (Fig. 10Z-AB; Extended Data Fig. 10-1F).Since the rostral region of the MTX-treated cortex was occupied by Nestin 1 but Pax6 À RGCs, we considered that this region corresponds to a nonneocortical area region that contains a different subtype of NSPCs.
Previous studies have shown that Pax6 and Tbr2 are expressed in NSPCs of the embryonic cerebral neocortex, while GSH2, another homeobox gene, is expressed explicitly in the NSPC population residing in lateral GE, representing the striatal primordium (Yun et al., 2001;Corbin et al., 2003;Englund et al., 2005).In fact, immunostaining MMF-treated brains with an anti-GSH2 antibody revealed that GSH2 was exclusively expressed in the rostral area, suggesting that this area is a structure derived from the lateral GE (Fig. 10AC,AD).Importantly, this phenotype of mislocalization of GSH-positive cells was completely reverted to an intact cortical architecture on co-treatment with MHY1485 (Fig. 10AE).
Consolidating these histologic observations, we concluded that loss or hypoplasia of the rostral neocortex in MMF-treated brains may have caused dorsal displacement and bulging of ventral forebrain regions, including the striatal GE, resulting in abnormal forebrain-like structures (Fig. 10AJ).Consistent with the results in cultured NSPCs, these findings strongly suggest that the NSPC population is more susceptible to deficits in the purine de novo pathway regulating mTORC1 signaling pathway in dorsal forebrain regions, such as the neocortex, than in other brain regions.The large number of caspase3 1 apoptotic cells in the GE-derived region of the rostral area (Fig. 10B,C) also might reflect a shift in dependence on the de novo and mTORC1 pathways along the anteroposterior and dorsoventral brain axis.

Discussion
Purine synthesis pathways are activated during brain development We revealed that PAICS and FGAMS, two de novo purine synthesis enzymes, were abundant in the embryonic cerebral cortex and downregulated toward postnatal and adult stages.On the other hand, HGPRT, which promotes the salvage pathway, exhibited the opposite trend during cortical development.These results indicate a switch in continued Western blot analysis at E12.5.AH, Representative Western blotting illustrating the accelerated apoptosis in MTX-treated brains and the rescue effect of MHY1485.The blot was reprobed with a-Tubulin antibody (bottom) to examine quantitative protein loading.AI, Quantified comparison of cleaved-caspase3 expression.Data are presented as the mean 6 SEM.AJ, Schematic model of brain abnormalities caused by de novo pathway inhibition.Control (left upper panel) or MMF-treated brains (left lower panel) were cut in the horizontal (middle panels) or sagittal plane (right panels).Horizontal sections were sliced through the orange line shown in the 3D brain (left).MMF-treated brains show a loss or hypoplasia of the rostral neocortex and a dorsal expansion of striatal GE (blue), resulting in abnormal forebrain structures.In MMF-treated embryos, the rostral neocortex containing Pax6 1 VZ disappeared; instead, the GE containing GSH2 1 VZ ectopically appeared on the dorsal surface.lv, lateral ventricles; 3v, third ventricle; 4v, fourth ventricle; GE, ganglionic eminence.Data demonstrating that administration of MHY1485 alone does not affect embryonic brain development is shown in Extended Data Figure 10-1.
the main pathway of purine synthesis from the de novo pathway to the salvage pathway during brain development.This switch probably reflects that the de novo pathway is driven to meet the massive purine demand caused by active cell proliferation during the embryonic period and that the salvage pathway becomes dominant after birth, when the energy cost stabilizes after the peak of neurogenesis and reduced demand for intracellular purines.
The inhibition of de novo purine synthesis by drugs in vivo caused proliferation defects in NSPCs and delayed the migration of immature neurons in the embryonic neocortex.Previous studies have indicated that purinergic signaling is essential for NSPC maintenance and neuronal migration in the neocortical SVZ during brain development (Lin et al., 2007;Liu et al., 2008;Yamada et al., 2020).Cell cycle progression is closely associated with the properties of NSPCs.For instance, NSPCs residing in the embryonic VZ often exhibit a cellular behavior known as interkinetic nuclear migration (INM), which refers to the periodic movement of the cell nucleus depending on the cell cycle phase.Although the nuclei of NSPCs occupy different positions along the apical/basal axis within the cerebral wall, mitosis (M-phase) exclusively occurs when the nucleus approaches the ventricular surface (apical side; Baye and Link, 2008;Taverna and Huttner, 2010).Dysregulation of INM is reported to induce the aberrant division of NSPCs on the basal side of the VZ (Tamai et al., 2007).Similarly, on inhibition of de novo synthesis, we observed an abnormal increase in the number of pH3 1 Mphase cells that lie beyond the VZ (Fig. 6G).The de novo purine pathway might play an essential role in INM by controlling cell cycle progression.A premature exit from the cell cycle caused by the inhibition of de novo synthesis might delay the radial migration of newborn neurons during brain development.
Additionally, the elaboration of the vascular system within the embryonic cerebral cortex might regulate the demand level for de novo purines in NSPCs.Previous studies have indicated the presence of highly vascularized and avascular regions in the developing neocortex and that the nervous system and the vasculature of the brain are closely associated during the development of the cerebral cortex (Komabayashi-Suzuki et al., 2019).An avascular region forms on the VZ from E12.5 to E17.5, where dividing undifferentiated RGCs remain in contact with sprouting neovascular tip cells, which gradually shrinks over development (Bjornsson et al., 2015;Komabayashi-Suzuki et al., 2019).This avascular hypoxic niche is supposedly required to maintain the stemness of NSPCs as well as stem cells in other developing organs (Bjornsson et al., 2015;Komabayashi-Suzuki et al., 2019).As the embryonic NSPCs in the VZ avascular region cannot receive sufficient purines or purine metabolites through blood vessels, it seems logical that NSPCs are strongly dependent on the de novo pathway during cortical development.Meanwhile, HGPRT is expressed in CD31 1 blood vessels during angiogenesis at embryonic stages; a colocalization is not observed at the adult stage (Fig. 3S; Extended Data Fig. 3-1).Thus, the salvage pathway in vascular endothelial cells may supply purines to the surrounding differentiating and migrating neurons in the IZ/CP.In addition, we expected that glial cells, especially astrocytes, which increase in number from late embryonic stages through cell division, would use the salvage pathway.However, HGPRT was rarely detected in astrocytes (Fig. 2E).As a recent study revealed that hypoxanthine, an intermediate metabolite of the purine salvage pathway, is involved in morphologic changes in microglia (Okajima et al., 2020), microglia might represent the major cell type that actively utilizes the salvage pathway from late embryonic stages to the postnatal stage.
Unlike the embryonic cerebral cortex, the developing cerebellum displayed abundant expression of the salvage enzyme HGPRT in addition to PAICS during the early postnatal period (Fig. 2B,C).Both de novo and salvage inhibitors affected the proliferation of cerebellar granular cells, indicating an essential role of both pathways in cerebellar development.One possible reason is that the de novo pathway alone may not be sufficient to supply purines since cerebellar NSPCs require large amounts of purines to undergo numerous symmetric cell divisions.Another possible reason is that the cerebellar NSPCs, located in the EGL, could drive the salvage pathway by receiving a supply of purine metabolites from the blood vessels of the meninges covering the brain surface.Over the past decade, a link between purine synthesis pathways and various cancers has been demonstrated (Di Virgilio and Adinolfi, 2017;Yin et al., 2018).In general, the intracellular concentration of purine metabolites and the activity of de novo pathway enzymes are enhanced in cancer cells.Changes in the ratio of purine metabolites in tumor cells affect tumor growth, invasion, and metastasis.A switch in purine synthesis pathways that depends on the extracellular environment may not only exist in NSPCs but also in cancer stem cells and other tissue stem cells.This possibility could provide a valuable perspective for anticancer drug development.

Different purine synthesis pathway dependence of NSPCs according to brain regions
The immunohistochemical analysis of adult mouse brains indicated different dominant purine synthesis pathways according to brain region .In addition, this study showed that NSPCs have a different predominance of purine synthesis pathways depending on the brain region.Accordingly, disturbance or loss of Pax6, Tbr2, and DCX expression in the neocortex was accompanied by a cortical malformation in which GSH2 1 GE cells emerged in the dorsal forebrain region.The purine demand in NSPCs likely differs between the dorsal (cerebral neocortex) and ventral forebrain (GE).Considering that NSPCs in the neocortex and GE produce predominantly glutamatergic excitatory and GABAergic inhibitory neurons, respectively, the different cell lineage of each NSPC may explain the differences in the requirement of de novo purine synthesis.Indeed, we observed that cell death occurred widely in cells, including neurons in the CP in the rostral GE-like structures formed in MMFtreated embryos (Fig. 10B), suggesting that the cells in GE-like structures belong to distinct cell lineages from their surrounding structures.Alternatively, mitotic delay in NSPCs alters fate specification or viability, as previously described (Pilaz et al., 2016).However, it remains unclear why de novo purine production is higher in the neocortex than in other brain regions.There may exist a relationship between purine synthesis pathways and the evolution of mammals with a huge neocortex.
The de novo purine synthesis is involved in cortical formation through mTOR signaling Our results indicated that inhibiting de novo purine synthesis by MMF or MTX during corticogenesis in vivo results in brain malformation.Clinical case reports indicate that prenatal exposure to MTX leads to fetal death or fetal MTX syndrome, which is characterized by CNS anomalies, including alobar holoprosencephaly (Seidahmed et al., 2006;Corona-Rivera et al., 2010).However, the molecular mechanism underlying the brain abnormalities induced by MTX remains unknown.We revealed that the brain malformation caused by inhibiting de novo purine synthesis was accompanied with a marked inhibition of mTORC1/S6K/S6 signaling and partial impairment of mTORC1/4E-BP1/eIF4E signaling.This malformation was rescued by activation of mTOR signaling.Our results are consistent with previous studies showing that inhibition of purine synthesis in tumor cells mainly suppresses mTORC1/S6K/S6 signaling (Emmanuel et al., 2017;Hoxhaj et al., 2017).Some patients with cortical malformations, such as hemimegalencephaly and focal cortical dysplasia, have been shown to carry mutations in mTOR (Tarkowski et al., 2019).Deficiency in mTOR causes morphologic abnormalities in the brain, including a thinner cerebral cortex because of suppressed NSPC proliferation (Ka et al., 2014), suggesting a close correlation between the mTOR pathway and disruption of cortical development.These findings further supported our notion that de novo purine synthesis and the mTORC1/S6K/S6 signaling pathway tightly regulate each other spatiotemporally to control neocortical development.

Limitations
In this study, we revealed a switch from de novo to salvage purine synthesis as brain development proceeds, and the de novo pathway plays a vital role in early embryonic corticogenesis, cooperating with the mTOR signaling pathway.We indicated that purine sensibility or predominance of purine synthesis pathways differs in NSPCs depending on the brain region.However, we could not clarify the factor that drives the switch from de novo to salvage pathways as well as the molecular mechanisms of brain malformation caused by de novo purine synthesis inhibitors.Our future study will focus on regional and lineage-specific differences in purine metabolism in the brain.For example, genetic manipulation of the de novo synthesis enzymes using Emx1-Cre transgenic mice will evidently provide molecular and cellular clues regarding the role of this pathway in determining the properties of NSPCs in the dorsal telencephalon.These studies might shed light on the significance of adaptive changes in purine metabolic pathways during the evolution of the mammalian cerebral neocortex.

Figure 2 .
Figure2.Developmental expression of the enzymes of the de novo and salvage purine synthesis pathways in the nervous system.PAICS, FGAMS, and HGPRT expression was assessed via western blotting using embryonic and postnatal brains (A-C) or primary cultured cells (D-E).A, Developmental changes in the expression of each enzyme in whole-brain lysates prepared from different developmental stages (E10.5-adult).B, PAICS, HGPRT, GFAP, and Pax6 expression in the cerebral cortex (ctx; E13.5-P12) or cerebellum (cbl; P2-P12).C, The line charts representing the expression of PAICS and HGPRT during ctx (upper) and cbl development (bottom).The bands obtained in three independent mouse brains in each developmental stage and the brain region of blots were quantified using a-Tubulin as an internal standard, and the mean values were plotted for each age.D, Expression of FGAMS, PAICS, HGPRT, and Pax6 in cultured NSPCs or differentiated neurons prepared from the ctx (E12.5 or E16.5) and cbl (P2 or P4).E, Expression of PAICS, HGPRT, GFAP, and Pax6 in primary cultured NSPCs or astrocytes.a-Tubulin was used as a loading control (bottom panels).

Figure 5 .
Figure 5.The development of the cerebellum is cooperatively regulated by both purine synthesis pathways.A-D, Confocal images of the E18.5 cerebellum stained with the Ki67 (green) and BrdU (red).Control DMSO (A), forodesine (B), MMF (C), or methotrexate (MTX; D) was administered at E16.5, followed by labeling with BrdU at E17.5, and the cerebellum was analyzed at E18.5.A'-D', Higher magnification of A-D.E, Quantified comparison of the percentage of Ki67 À BrdU 1 cells to total BrdU 1 cells in GL at E18.5.F-H, Confocal images of the P6 cerebellum stained with the Ki67 (green) and BrdU (red).Control DMSO (F), forodesine (G), or MMF (H) was administered to P2 pups, followed by BrdU labeling at P4, and the cerebellum was immunostained at P6. F'-H', Higher magnification of F-H.I, Quantified comparison of the percentage of Ki67 À BrdU 1 cells to the total BrdU 1 cells in the GL at P6. Data are presented as means 6 SEM; *p , 0.05, **p , 0.01, ***p , 0.001, Welch's t test followed by Holm-Bonferroni correction.EGL, external granule cell layer; GL, internal granule cell layer.Scale bars, 100 mm.

Figure 7 .
Figure 7.The de novo purine pathway affects mTORC1/S6K/S6 signaling.A, Immunoblot analysis of E12.5 brains treated with each inhibitor at E9.5-E11.5.Each panel presents the expression of mTOR signaling proteins (pS6K, S6K, pS6, S6, and 4E-BP1) and purine synthesis enzymes (PAICS, FGAMS, and HGPRT).Arrows denote the a, b , and g isoforms of 4E-BP1.The blot was reprobed with a-Tubulin antibody (bottom) to examine quantitative protein loading.B, Quantified comparison of the pS6/S6 ratio.Data are presented as the mean 6 SEM; ns, not significant, **p , 0.01, Welch's t test followed by the Holm-Bonferroni correction.C, Schematic diagram of the relationship between mTOR signaling and purine nucleotides.Data supporting the alteration of 4E-BP1 expression by inhibiting the de novo purine synthetic pathway is shown in Extended Data Figure 7-1.