Adenylosuccinate lyase (ADSL) functions in de novo purine synthesis (DNPS) and the purine nucleotide cycle. ADSL deficiency (ADSLD) causes numerous neurodevelopmental pathologies, including microcephaly and autism spectrum disorder. ADSLD patients have normal serum purine nucleotide levels but exhibit accumulation of dephosphorylated ADSL substrates, S-Ado, and SAICAr, the latter being implicated in neurotoxic effects through unknown mechanisms. We examined the phenotypic effects of ADSL depletion in human cells and their relation to phenotypic outcomes. Using specific interventions to compensate for reduced purine levels or modulate SAICAr accumulation, we found that diminished AMP levels resulted in increased DNA damage signaling and cell cycle delays, while primary ciliogenesis was impaired specifically by loss of ADSL or administration of SAICAr. ADSL-deficient chicken and zebrafish embryos displayed impaired neurogenesis and microcephaly. Neuroprogenitor attrition in zebrafish embryos was rescued by pharmacological inhibition of DNPS, but not increased nucleotide concentration. Zebrafish also displayed phenotypes commonly linked to ciliopathies. Our results suggest that both reduced purine levels and impaired DNPS contribute to neurodevelopmental pathology in ADSLD and that defective ciliogenesis may influence the ADSLD phenotypic spectrum.
This article attempts to elucidate adenylosuccinate lyase (ADSL) deficiency in neurodevelopment using a variety of approaches. It provides a better understanding of the underlying mechanisms that lead to microcephaly in some patients affected by ADSL deficiency.https://doi.org/10.7554/eLife.70518.sa0
Adenylosuccinate lyase (ADSL) is a conserved homotetrameric enzyme that catalyzes the eighth reaction in the de novo purine synthesis (DNPS) pathway and also participates in the purine nucleotide cycle (Figure 1—figure supplement 1; Daignan-Fornier and Pinson, 2019). Mutations in ADSL cause adenylosuccinate lyase deficiency (ADSLD), an autosomal recessive disorder characterized by defects in purine metabolism and heterogeneous neurological phenotypes that include lack of eye-to-eye contact, auto-aggressive behavior, speech impairment, mild psychomotor delay, transient contact defects, autism spectrum disorder, epilepsy, and in some cases, microcephaly, encephalopathy, ataxia, or coma vigil (Jaeken and Van den Berghe, 1984; Jurecka et al., 2015). While the incidence of ADSLD has not been fully established, over 100 patients have been diagnosed to date and subcategorized based on their symptoms that range from premature death to milder developmental and behavioral disorders (Jurecka et al., 2015; http://www.adenylosuccinatelyasedeficiency.com/).
ADSLD can be diagnosed by detecting elevated levels of the substrates of ADSL, SAICAR (also referred to as SZMP), and S-AMP, as well as their dephosphorylated riboside forms, SAICAr and S-Ado, in bodily fluids (Jaeken and Van den Berghe, 1984). As normal levels of purine nucleotides were detected in serum from ADSLD patients, the accumulation of S-Ado, and particularly SAICAr, was proposed to play a role in the disease pathology (Jaeken and Van den Berghe, 1984; Jurecka et al., 2015; Stone et al., 1998).
In yeast, ADSL (Ade13) loss provokes genomic instability and is lethal (Daignan-Fornier and Pinson, 2019; Giaever et al., 2002; Pinson et al., 2009; Chen et al., 2016). Lethality in yeast can be rescued by deletion of a number of DNPS enzymes upstream of ADSL, or the transcription factors that regulate the pathway, indicating that the accumulation of metabolic intermediates, rather than impaired DNPS, underlies the toxicity (Daignan-Fornier and Pinson, 2019; Chen et al., 2016). In Caenorhabditis elegans, ADSL loss caused delayed growth, infertility, reduced life span, and locomotion defects. In some studies, growth, life span, and locomotion could be linked to the accumulation of SAICAR (Chen et al., 2016; Marsac et al., 2019; Fenton et al., 2017).
Perfusion of rat brains with SAICAr caused cellular attrition in the hippocampus, leading to the proposition that SAICAr accumulation is neurotoxic, although the potential mechanism remains unknown (Stone et al., 1998). In glucose-deprived cancer cells, SAICAR accumulation was shown to activate PKM2, and a number of other kinases, to promote cancer survival in glucose-limiting conditions, suggesting that purine metabolite accumulation could have distinct signaling outcomes that impact on cell behavior and fate during development (Keller et al., 2012; Keller et al., 2014; Yan et al., 2016). However, despite extensive enzymology and structural information, the underlying mechanisms by which neuropathology arises in ADSLD remain unknown.
To address the potential roles of ADSL deficiency in neurodevelopment, we systematically examined the consequences of ADSL depletion in diploid human cells and in vivo. We found that ADSL depletion in human epithelial cells impaired cell cycle progression, induced DNA damage signaling, and impaired primary ciliogenesis. Deletion of p53 or supplementation with nucleosides rescued cell cycle progression and DNA damage signaling, respectively. In contrast, ciliogenesis defects were unaffected by p53 status or nucleoside supplementation and could be reproduced by SAICAr supplementation or attenuated by the inhibition of phosphoribosylaminoimidazole carboxylase (PAICS), which is required to generate SAICAR/r upstream of ADSL. Depletion of ADSL in chicken or zebrafish embryos impaired neurogenesis and caused developmental defects. Both chicken and zebrafish embryos exhibited microcephaly, which is observed in some ADSLD patients, and increased DNA damage was evident in zebrafish. In addition, fish embryos displayed ciliopathy-related phenotypes and interventions to limit SAICAR accumulation-rescued impaired neurogenesis. Together, our results indicate that ADSL depletion causes context-dependent phenotypes associated with its role in both the purine nucleotide cycle and DNPS, which together impact neurodevelopment.
To investigate the impact of ADSL on cellular homeostasis, we depleted ADSL with a pool of four siRNAs in hTERT-immortalized human retinal epithelial cells (hTERT-RPE-1, referred to henceforth as RPE-1). Depletion of ADSL was effective as we observed 80% depletion of the mRNA and a clear reduction in protein levels (Figure 1A and B). This was accompanied by reduced levels of AMP and GMP, as well as accumulation of S-Ado (Figure 1C, Figure 1—figure supplement 2A and B). SAICAr levels were at the limit of detection in RPE-1 cells and differences could not be assessed (Figure 1—figure supplement 2B), but elevated SAICAr could be readily observed following ADSL depletion in HeLa, potentially due to elevated levels of DNPS enzymes that have been observed in cancer cells (Figure 1—figure supplement 2C; Zurlo et al., 2019; Agarwal et al., 2020; Yamauchi et al., 2022; Taha-Mehlitz et al., 2021). We deconvolved the siRNA pool and identified an effective single siRNA (#2) to further validate results from the siRNA pool (Figure 1—figure supplement 2D).
As ADSL is critical for DNPS, we examined cell growth following ADSL depletion and found reduced levels of proliferation in ADSL-depleted cells compared to controls (Figure 1D). ADSL-depleted cells frequently lacked Ki67 expression, indicating that some cells were exiting the cell cycle, had increased levels of p53, and showed accumulation in G1 phase of the cell cycle (Figure 1E and F, Figure 1—figure supplement 2E). Trypan Blue and β-galactosidase assays indicated that cell death was not increased (Figure 1—figure supplement 2F) and that the Ki67-negative cells were likely quiescent and not senescent (Figure 1—figure supplement 2G). Deletion of TP53 rescued proliferation, prevented G1 arrest, and restored the number of Ki67-positive cells (Figure 1G, Figure 1—figure supplement 2H and I). The reduction in Ki67-positive cells could be rescued by stable expression of an siRNA-resistant allele of ADSL (ADSL*) (Figure 1H, Figure 1—figure supplement 2J). We also checked whether ADSL-depleted RPE-1 cells underwent differentiation by staining with vimentin, a marker of undifferentiated cells (Tamiya et al., 2010), and cytokeratin 20 (CK20), a marker of differentiation, upon ADSL depletion. We did not observe any CK20 signal or a reduction in vimentin-positive cells in the population upon ADSL silencing compared to the controls, arguing against premature differentiation (Figure 1—figure supplement 2K).
To identify the cause of cell cycle exit in ADSL-depleted cells, we performed specific treatments. To restore purine levels, we supplemented cells with a mixture of nucleosides or with adenosine alone. To reduce potential accumulation of SAICAr that is generated by DNPS, we treated with MRT00252040, a small-molecule inhibitor of PAICS, the enzyme required to generate SAICAr upstream of ADSL (Figure 1—figure supplement 1; Hoxhaj et al., 2017). Supplementation of ADSL-depleted RPE-1 cells with nucleosides or adenosine did not prevent p53 induction or cell cycle exit (Figure 1I and J). Similarly, treatment with MRT00252040 did not influence p53 induction or loss of Ki67 (Figure 1K and L). This demonstrated that ADSL depletion in non-transformed human epithelial cells leads to a partial p53-dependent cell cycle exit/arrest that is not rescued by complementing the defects of ADSL depletion with nucleoside supplementation or inhibiting PAICS to prevent SAICAR/r production upstream of ADSL.
Reduced levels of purine nucleotides in ADSL-depleted cells may cause replication stress and DNA damage (Kim et al., 2015; Benedict et al., 2018; Gottifredi et al., 2001). To address this, we examined the levels of chromatin-bound RPA, a surrogate marker of ssDNA accumulation and indicator of replication stress. In ADSL-depleted cells, chromatin-bound RPA levels were increased, indicative of replication stress, although the effect was mild (Figure 2A). Additionally, we observed an increased number of cells with more than five 53BP1 foci per cell, indicative of DNA double-strand break (DSBs) accumulation (Figure 2B). 53BP1 foci were reduced by treatment with a small-molecule inhibitor for ATM (Figure 2C), indicating an active DNA damage response. Supplementation of cells with nucleosides suppressed the appearance of DNA DSBs detected by 53BP1 following ADSL depletion or treatment with the ribonucleotide reductase inhibitor hydroxyurea (HU) that depletes dNTP pools (Figure 2D and E). A similar reduction in 53BP1 foci was also observed with adenosine supplementation of ADSL-depleted cells (Figure 2F). Nucleoside supplementation also reduced the increase in γH2AX, the phosphorylated form of the histone variant H2AX, which is also a marker of DSBs (Figure 2G). In contrast to nucleoside supplementation, the PAICS inhibitor MRT00252040 did not rescue the increased levels of DSBs (Figure 2H). These data indicate that ADSL depletion in cultured cells induces mild levels of DNA damage signaling that can be suppressed by nucleoside supplementation, but not by the inhibition of PAICS. This implicates defects in the purine nucleotide cycle, rather than impaired DNPS or specific metabolite accumulation resulting from ADSL depletion, in replication stress and DSB formation. In addition, it indicates that the partial p53-dependent cell cycle exit is not solely a consequence of DNA damage signaling or purine metabolite accumulation, as it was not rescued by either intervention.
Given the effects of ADSL depletion on cell growth and proliferation, we sought to examine the consequences of its loss in vivo. To this end, we used the chicken embryo system to examine the influence of ADSL depletion on nervous system development. We electroporated one side of the neural tube with plasmid expressing GFP as a transfection marker in combination with either control or ADSL shRNA vectors. After confirming efficient ADSL depletion (Figure 3A), we evaluated neurogenesis by staining with markers for proliferating neural progenitors (SOX2 positive) and post-mitotic neurons (ELAVL3/4 positive). We found that in the ADSL-depleted side both cell populations were reduced when compared to the non-transfected side (Figure 3B) and that the size of the tissue was smaller, suggesting reduced growth and/or increased cell death. Staining for the apoptotic marker cleaved caspase-3 (CC3) revealed no notable differences, suggesting that this was not due to increased cell death (Figure 3—figure supplement 1A).
We then analyzed SOX2 and ELAVL3/4 staining only within the GFP-positive transfected cells and found that ADSL depletion increased the percentage of SOX2-positive progenitors relative to ELAVL3/4-positive neurons (Figure 3C, Figure 3—figure supplement 1B). This suggested that reduced tissue growth was not due to premature differentiation but possibly due to a proliferation defect in the progenitor population. To study cell cycle progression in neural stem cells, we performed fluorescence-associated cell sorting (FACS) analysis of GFP-positive, ELAVL3/4-negative cells following electroporation of control or ADSL shRNA. We found that there was a slight increase in the G2/M population after ADSL depletion (Figure 3D). Further analysis of stained tissue sections showed that ADSL depletion caused a reduction in the fraction of cells that incorporated EdU and an increase in the fraction of cells positive for the G2/M marker phosphorylated histone H3-Ser10 (pH3S10) (Figure 3E). We separated the pH3S10-positive cells into two populations: G2 cells, identified by punctate pH3S10 staining, and mitotic cells, displaying broadly distributed pH3S10 staining. This revealed that only the G2 fraction of cells was increased by ADSL depletion, indicating that ADSL depletion caused a specific delay in G2 phase in the SOX2+ population, rather than during mitosis (Figure 3F). Together, our in vitro data indicate that ADSL depletion leads to a mild induction of DNA damage signaling and impaired cell cycle progression. In vivo, this manifests as reduced cellularity in the developing brain, without a clear induction of cell death or senescence.
As there are non-cycling cells in the brain and ADSL depletion caused cell cycle exit in RPE-1 cells, a condition frequently accompanied by ciliogenesis, we tested the ability of control and ADSL-depleted RPE-1 cells to assemble cilia. Following treatment with siRNA, cells were serum-starved for 48 hr and analyzed by immunofluorescence microscopy. Ki67 staining confirmed that most of the cells in both conditions exited the cell cycle (Figure 4A). We next examined ciliogenesis by staining for the ciliary marker ARL13B and the centrosome marker pericentrin (PCNT). Fewer cells treated with the ADSL siRNA pool had cilia, and the cilia that were present were shorter when compared to controls (Figure 4B). We also observed shorter cilia upon depletion with single siRNAs for ADSL (Figure 4—figure supplement 1A). To exclude the possibility that ciliogenesis was simply delayed, we quantified the number of ciliated cells 72 hr after serum starvation and observed a similar defect (Figure 4—figure supplement 1B). Defective ciliogenesis was rescued by expression of an siRNA-resistant cDNA (ADSL*), but not by nucleoside supplementation (Figure 4C and D). Inhibition of the DNPS pathway with methotrexate (MTX), which impairs steps in DNPS up- and downstream of ADSL (Figure 1—figure supplement 1), had no effect on ciliogenesis in control cells (Figure 4—figure supplement 1C), but rescued both the number of ciliated cells and cilia length in ADSL-depleted RPE-1 cells (Figure 4—figure supplement 1D).
Since defective ciliogenesis caused by ADSL depletion was rescued by MTX, but not by nucleoside supplementation, we next examined whether specific inhibition of PAICS was sufficient to induce the phenotype. Following ADSL depletion, we treated cells with the PAICS inhibitor MRT00252040 (Hoxhaj et al., 2017). Similar to MTX, this rescued ciliogenesis as the number and length of cilia were similar in control and ADSL-depleted cells (Figure 4E). As inhibition of PAICS impairs DNPS upstream of SAICAR production, we treated cells with SAICAr and observed that this recapitulated the ciliogenesis defect observed in ADSL-depleted cells (Figure 4F and G). SAICAr treatment did not activate AKT activity, which has recently been shown to inhibit ciliogenesis due to serum starvation (Figure 4—figure supplement 1E; Walia et al., 2019). To exclude additional indirect effects as a cause for the reduction in the number of ciliated cells, such as cell cycle progression defects caused by DNA damage, p53 activation or defective cilia removal from interphase cells, we repeated the experiment in p53 KO cells and examined cilia resorption following the addition of serum. While the overall percentage of ciliated cells was slightly lower in p53 KO cells, depletion of ADSL recapitulated the result obtained in RPE-1 wt cells, a reduction in ciliated cells and cilia length compared to controls (Figure 4—figure supplement 1F). In addition, the kinetics of cilia removal following serum add back was similar in both controls and ADSL-depleted cells (Figure 4—figure supplement 1G). We concluded that impairing DNPS specifically at the ADSL-dependent step and/or SAICAr accumulation caused by ADSL depletion impairs the generation of primary cilia.
To understand the origin of the ciliogenesis defect, we examined centriole configurations since mother centrioles, after conversion to basal bodies, template formation of the primary cilium. Centrosomes in ADSL-depleted cells had normal levels of PCNT and a normal number of centrioles (Figure 5A, Figure 5—figure supplement 1A). However, we found that the removal of CP110 from the mother centriole, a key step in early ciliogenesis, was impaired in serum-starved, ADSL-depleted cells. Compared to controls, a larger number of ADSL-depleted cells contained centrosomes with two CP110 foci (Figure 5B). This could be phenocopied by administration of SAICAr and was rescued by PAICS inhibition (Figure 5C and D). To determine if the retention of CP110 could underlie the phenotype, we co-depleted CP110 with ADSL using three different siRNAs. All three siRNAs silenced CP110, as verified by Western blot (Figure 5—figure supplement 1B) and partially depleted CP110 at centrioles (Figure 5—figure supplement 1C). In non-serum-starved conditions, CP110 siRNA-treated cells had fewer than the two centriolar CP110 foci typically observed in control cells (Figure 5—figure supplement 1D). The remaining centriolar signal was associated with daughter centrioles (distal to the base of the cilium in ciliated cells; Figure 5—figure supplement 1C). Co-depletion of CP110, using three independent siRNAs, with ADSL rescued the ciliogenesis defect (Figure 5E). These data demonstrated that ADSL deficiency or SAICAr administration impairs primary ciliogenesis, and this can be rescued by CP110 depletion or inhibition of PAICS, but not by supplementation of purine levels.
To test whether ADSL deficiency caused ciliary defects in vivo, we employed a zebrafish model. As CRISPR/Cas9-mediated gene knockout did not yield viable mutants, we used two different antisense morpholino oligonucleotides (MO) to deplete Adsl in zebrafish embryos. Adsl is ubiquitously expressed at early embryonic stages and, by the 18-somite stage, highly expressed in several areas of the developing brain, including the midbrain and mesencephalon (Figure 6—figure supplement 1A–L). Antibody staining demonstrated expression of Adsl in neurons, which was abolished upon injection of either MO (Figure 6—figure supplement 2). Examination of embryo morphology 48 hr post fertilization (hpf) revealed pericardial edema, kinked tail, hydrocephalus, and pinhead (microcephaly) phenotypes (Figure 6A–E). Defects in head size, which are consistent with the clinical presentation of ADSLD patients, were further corroborated by staining for skull formation that is coordinated with brain development. Alcian Blue staining showed that nearly 50% of the Adsl-depleted embryos exhibited weak or absent staining (Figure 6F). Defects in skull formation could be largely rescued by zebrafish Adsl or human ADSL expression but not expression of a human ADSL R426H mutant, the most frequently observed ADSLD mutation (Figure 6F; https://www.adenylosuccinatelyasedeficiency.com/). Examination of DNA damage signaling in the developing neural tube revealed an increase in γH2AX-positive cells. Similar to what was observed in RPE-1 cells, treatment with nucleosides suppressed DNA damage signaling (Figure 6G). These data demonstrated that Adsl depletion strongly impaired normal zebrafish development, leading to DNA damage that could be suppressed with nucleoside supplementation and several phenotypes consistent with ciliary defects.
Figure 6—source data 1
Figure 6—source data 2
Figure 6—source data 3
Figure 6—source data 4
As the observed phenotypes were potentially indicative of defects in cilium function, we examined heart looping by staining for cardiac myosin light chain 2 (cmcl2) mRNA. Adsl-depleted embryos showed higher frequencies of defects, including inverse looping and to a lesser extent no loops (Figure 7A). Inverse heart looping may be indicative of laterality impairment (situs inversus) that can arise due to ciliary defects. To corroborate this possibility, we examined liver placement by staining for angiopoietin-like 3 (angptl3). A significant increase in inverse liver placement was observed in Adsl-depleted embryos compared to controls, supporting a general defect in laterality (Figure 7B). To further investigate the laterality defects, we examined left-right asymmetry at the 20-somite stage, staining for the mRNA of the left lateral plate mesoderm marker southpaw (spaw). Consistent with the altered distribution of cmcl2 and angptl3, asymmetric spaw mRNA localization was changed in about 40% of Adsl-depleted embryos. Most of these embryos showed symmetric patterning and a smaller fraction of no or only weak staining. The correct asymmetric distribution of spaw mRNA could be largely restored by expression of mRNA encoding zebrafish Adsl (Figure 7C).
Figure 7—source data 1
Figure 7—source data 2
Figure 7—source data 3
Figure 7—source data 4
Figure 7—source data 5
To test if impaired laterality may involve ciliary defects, we examined the Kupffer’s vesicle (KV, organ of laterality). While KV area and cilia number were not significantly affected by Adsl depletion, cilia length was reduced in Adsl ATG MO-treated embryos, a phenotype that was partially rescued by co-injection of RNA encoding zebrafish Adsl (Figure 7D–G). These data, in combination with additional phenotypes, including laterality defects and hydrocephalus, support the role of ADSL in promoting proper cilia formation or function in vivo.
As ciliogenesis defects were linked to impaired DNPS and excess SAICAr in human cells (Figure 4), we examined the effects of inhibiting the DNPS pathway during zebrafish development. We quantified the effects of Adsl depletion on differentiating neuronal cells by staining for the marker Elavl3/4. Similar to what we observed in the chicken neural tube, depletion of Adsl caused a significant reduction in Elavl3/4-positive cells that could be rescued by the co-injection of RNA encoding zebrafish Adsl (Figure 8A). We next treated control and Adsl-depleted embryos with MTX to attenuate the DNPS pathway upstream of ADSL and reduce SAICAR production. Treatment with MTX completely rescued the reduction in Elavl3/4-positive cells in the neural tube, indicating that this was not a result of impaired DNPS per se, but likely a consequence of intermediate metabolite accumulation (Figure 8B). Similar results were observed with a second morpholino targeting Adsl (Figure 8—figure supplement 1). We next examined the effect of MTX treatment on Sox2-positive neural progenitors in the developing forebrain (Figure 8C, upper right panel). Depletion of Adsl reduced the number of Sox2-positive cells, and this was rescued by co-treatment with MTX (Figure 8C). In contrast, supplementation with nucleosides did not rescue the reduced number of Sox2-positive progenitors in Adsl-depleted embryos (Figure 8D). These data indicate a specific role of impaired DNPS, and likely SAICAr accumulation, in the neural progenitor defects associated with Adsl depletion.
Figure 8—source data 1
Figure 8—source data 2
Figure 8—source data 3
Figure 8—source data 4
Despite a detailed understanding of the enzymology of DNPS and the purine nucleotide cycle, the specific cell and organismal effects underlying the complex etiology of ADSLD remain unclear. Our results uncovered multiple phenotypes associated with ADSL depletion in human RPE-1 cells, as well as developing chicken and zebrafish embryos (Appendix 1). These can largely be rescued by interventions that suppress the generation of metabolites in the DNPS pathway or restore purine levels, complementing defective DNPS or purine nucleotide pathway activity. DNA damage signaling was suppressed by nucleoside supplementation in human cells and zebrafish, suggesting that this was caused by purine deficiency (Figure 1A). In contrast, defects in primary ciliogenesis resulting from ADSL depletion were rescued by PAICS inhibitor or MTX treatment, but not nucleosides, and phenocopied by SAICAr administration. The inability to rescue these phenotypes with nucleoside supplementation indicated that they most likely resulted specifically from loss of ADSL, and potentially SAICAr accumulation, rather than a general deficiency in purine supply. As SAICAr levels were barely detectable in RPE-1, it could indicate that a very small amount of SAICAr is sufficient to provoke the phenotypes or that signaling elicited specifically from loss of the ADSL protein is responsible. In ADSL-depleted RPE-1 cells, we could also detect p53 activation and defects in cell cycle progression, in the absence of cell death or senescence, and these phenotypes were insensitive to nucleoside supplementation or DNPS modulation, indicating the involvement of additional pathways.
Microcephaly, which is present in a subset of more severe ADSLD cases (Jurecka et al., 2008; Jurecka et al., 2012; Mouchegh et al., 2007), was observed in zebrafish embryos following Adsl depletion. Similarly, ADSL depletion in chicken embryos led to a reduction in neural tube size. Together, the results suggest that they are potentially valuable models for understanding the etiology of microcephaly in ADSLD (Jurecka et al., 2015). DNA damage, p53 activation, and defects in cilia function, which we observed following ADSL depletion, have all been implicated in neurodevelopmental disorders, such as Seckel syndrome, which is associated with microcephaly (Stracker et al., 2020). In Seckel syndrome mice expressing hypomorphic, humanized alleles of Atr, replication stress and DNA damage preceded neuroprogenitor cell death (O’Driscoll et al., 2003; Murga et al., 2009). This was accompanied by extensive p53 activation, but co-deletion of p53 exacerbated the cellular and organismal phenotypes, indicating a protective effect of p53 induction. Mutations in centrosomal proteins, such as CENPJ/SAS4/CPAP or CEP63, were also implicated in Seckel syndrome (Al-Dosari et al., 2010; Sir et al., 2011). In contrast to mice expressing hypomorphic Atr, progenitor loss in mice with CEP63 or CENPJ/SAS4 deficiency resulted from mitotic delays and the activation of the USP28-53BP1-dependent mitotic surveillance pathway (Bazzi and Anderson, 2014; Insolera et al., 2014; Marjanović et al., 2015; McIntyre et al., 2012; Lin et al., 2020; Fong et al., 2016; Lambrus et al., 2016). In this case, the phenotype was completely rescued by p53 co-deletion, revealing p53-dependent cell death as a main driver of the phenotype. Despite some phenotypic similarities with Seckel syndrome at the cellular level, including replication stress and p53 activation, we did not detect increased cell death as a result of ADSL depletion. This indicated that the reduced cellularity caused by ADSL depletion occurs in a manner mechanistically distinct from Seckel syndrome driven by ATR deficiency or centrosome duplication defects.
GMPS (GMP synthase), an essential enzyme in the purine synthesis pathway (Figure 1—figure supplement 1), was identified as a target for p53 repression following p53 activation using Nutlin-3a (Holzer et al., 2017). In addition, GMPS was demonstrated to promote p53 stabilization in response to genotoxic stress in a manner that requires the USP7 ubiquitin protease (Reddy et al., 2014). While we cannot rule out a role for this pathway in p53 stabilization following ADSL depletion, we could not rescue p53 activation with nucleoside supplementation and induction of p53 via the GMPS pathway was demonstrated to induce cell death, which we did not observe clearly in any of the systems we analyzed.
In addition to p53-dependent and -independent cell death, defects in cilium removal and premature progenitor differentiation have also been clearly implicated in microcephaly in Seckel syndrome and primary microcephaly (Gabriel et al., 2016; Farooq et al., 2020). We examined the removal of cilia by adding back serum to serum-starved RPE-1 cells depleted for ADSL and observed normal rates of cilia removal (Figure 4—figure supplement 1G), suggesting that impaired cilia resorption was likely not impairing cell cycle exit and indirectly impacting ciliogenesis. Analysis of differentiation in RPE-1 cells depleted for ADSL also did not support a premature differentiation phenotype (Figure 1—figure supplement 2K), but we recognize that RPE-1 cells may not respond to purine depletion in the same way as neural progenitors or other cell types in vivo.
In RPE-1 and chicken embryos, we observed cell cycle delays following ADSL depletion, as well as an overall reduction in SOX2-positive progenitor cells in both in vivo systems. As a result, differentiated ELAVL3/4-positive cell numbers were also reduced in both chicken and fish systems. Cell cycle delay was further supported by the observation that, contrary to overall Sox2-positive cell numbers, the number of transfected, Sox2-positive cells in chicken embryos was increased relative to controls. In agreement with our findings, nutrient restriction was shown to arrest the proliferation of neural progenitors of Xenopus larvae and zebrafish reversibly in G2, suggesting that most of the cells were quiescent (McKeown and Cline, 2019), consistent with the lack of evidence for cell death or senescence in ADSL-depleted systems. While arrest in Xenopus progenitors did not require mTOR signaling, nutrient-dependent cell cycle reentry was mTOR dependent. Purine deficiency inhibits the mTORC1 pathway, which regulates protein synthesis in response to nutrient availability (Hoxhaj et al., 2017), and inactivation of mTORC1 results in microcephaly in mouse models, but this was attributed mainly to increased levels of cell death (Cloëtta et al., 2013). Thus, while we cannot rule out a role for dysregulated mTOR signaling in the phenotypes associated with ADSLD, the origin of microcephaly is likely to be mechanistically distinct to that in mTORC1-deficient mouse models. We therefore propose that the reduction in brain size resulting from ADSL depletion is largely due to the impaired cell cycle progression observed in SOX2-positive progenitors and delayed differentiation.
The molecular targets of SAICAR remain largely unclear, but its accumulation has been linked to activation of PKM2, and other kinases, in the context of glucose deficiency in cancer (Keller et al., 2012; Keller et al., 2014; Yan et al., 2016). Recent work also implicated ADSL in the activation of MYC, which plays a major role in controlling metabolism and proliferation in cancer cells (Zurlo et al., 2019). Considering that ciliogenesis is a highly regulated process and tightly coordinated with the cell cycle, modulation of central signaling pathways, including AKT and mTOR that control energetics and proliferation, may be one way by which SAICAr could affect cilia (Hoxhaj et al., 2017; Walia et al., 2019). In particular, AKT was shown to regulate a key step in ciliary vesicle formation and AKT activation exerts an inhibitory effect on ciliogenesis induced by serum starvation (Walia et al., 2019). We examined AKT activation following SAICAr treatment and did not observe any detectable increase, indicating that the ability of SAICAr to inhibit ciliogenesis occurs through a distinct mechanism, consistent with the fact that CP110 depletion, which occurs prior to vesicle formation, rescued ciliation in ADSL-depleted cells (Figure 5E). ADSL has also been identified as a proximal interactor of numerous centrosomal proteins, including CEP104, CEP128, CEP135, CEP152, CEP89, and centrobin, suggesting that it may play specific regulatory roles involving direct interactions or its enzymatic activity locally at centrosomes (Gupta et al., 2015). To address this, future work will be needed to determine the cellular interactomes of ADSL and SAICAr, and the signaling they elicit, to identify the disease-relevant targets.
To our knowledge, this is the first demonstration of a specific purine metabolite impairing ciliogenesis. While ciliopathy-like features have not been described for the pathology of ADSLD, we observed a robust rescue of the neural progenitor population in zebrafish following MTX treatment, suggesting that SAICAr and its effects on cilia may be involved. Consistent with effects on cilia in vivo, we observed shorter cilia in the KV of zebrafish, as well as several ciliopathy-related phenotypes (Figures 6A–F ,–7A–G), consistent with impairment of ciliogenesis or cilia function in vivo. Adsl-depleted zebrafish also presented with defects in skull cartilage formation that is coordinated with brain size, in part through cilia-based sonic hedgehog signaling (Xavier et al., 2016). Primary cilia and Hedgehog signaling have well-established roles in regulating multiple progenitor populations in the developing brain (Bangs and Anderson, 2017). However, we also note that severe defects in primary cilia function or Hedgehog signaling cause more drastic reductions in progenitor proliferation and numbers than we observed in either system, potentially consistent with the milder ciliogenesis effects observed in vivo (Breunig et al., 2008; Han et al., 2008). Moreover, we used ADSL knockdown in our experimental systems, and it is currently unclear to what extent the observed ciliary defects would be recapitulated by ADSL mutations in patients and in what tissues and cell types. If ADSL deficiency would be less severe or only a subset of tissues would be affected, patients may not present classic and widespread ciliopathy features. Together, our work provides a cell-level analysis of ADSL deficiency, identifies specific cellular defects, and ascribes these to defects in specific steps of DNPS or the purine nucleotide cycle. Highlighting the complex etiology of ADSLD, our results add further support to the notion that SAICAr may play a key role in pathological outcomes and establish a framework for deciphering the underlying molecular mechanisms.
Human immortalized hTERT-RPE-1 WT (ATCC), TP53 knockout (kind gift from Brian Tsou), RPE-1-expressing pLenti-EGFP and pLenti-ADSL*EGFP (siRNA-resistant mutant) cells were cultured in Dulbecco’s modified Eagle medium-F12 (DMEM-F12; Thermo Fisher Scientific) supplemented with 10% (v/v) fetal bovine serum (MilliporeSigma) and 100 U/ml penicillin–streptomycin at 37°C and 5% CO2 in humidified atmosphere. For cilia experiments, silenced RPE-1 cells were serum starved for 48 hr in OptiMEM (Thermo Fisher Scientific). Hela (ATCC) and AD293 (Agilent) cells were cultured in DMEM with high glucose (Thermo Fisher Scientific) supplemented with 10% (v/v) fetal bovine serum (MilliporeSigma) and 100 U/ml penicillin–streptomycin at 37°C and 5% CO2 in humidified atmosphere. All cell lines were authenticated by STR testing (ATCC) and routinely tested for mycoplasma and found negative (Universal Mycoplasma Detection Kit; ATCC).
1 mg/ml SAICAr (Carbo Synth) was added to the cells for 96 hr to mimic ADSL depletion. 60 μM nucleosides (100X Embryomax, Merck Millipore) were added from the first silencing to the end at 1× in the culture medium. MRT00252040 (kindly provided by Simon Osborne, LifeArc, London, UK) dissolved in DMSO was used at 2 μM and MTX (MilliporeSigma) at 4 μM as described in Hoxhaj et al., 2017. ATM inhibitor (KU-55933; Selleckchem) was used at 5 mM for 24 hr before fixation. Doxorubicin (MilliporeSigma) was used as positive control for senescence at 1 μg/ml for 6 days. Adenosine (MilliporeSigma) was dissolved in water at a concentration of 0.8 mg/ml and diluted 100× to reach a final concentration of 80 μg/ml for 96 hr. HU (from MilliporeSigma) has been freshly dissolved in water and used at a final concentration of 2 mM in cell culture for 6 hr.
Staining of human cells was performed with the following primary antibodies: α-ADSL (MilliporeSigma, rabbit, 1:100 IF, 1:1000 western), α-ARL13B (Santa Cruz Biotechnology, mouse monoclonal C5, 1:100), PCNT (Novus Biologicals, rabbit, 1:400), α-p53 (Cell Signaling, mouse monoclonal 1C12, 1:100), α−RPA2 (Calbiochem, mouse monoclonal Ab-3, 1:100), a-53BP1 (Novus Biologicals, rabbit, 1:400), α-pSer139-H2A.X (Santa Cruz Biotechnology, rabbit, 1:100), α-AKT (Cell Signaling, rabbit polyclonal, 1:2000), α-pAKT-S473 (Cell Signaling, rabbit polyclonal, 1:2000), α-vinculin (Cell Signaling, rabbit polyclonal, 1:2000), α-actin (MilliporeSigma, mouse monoclonal AC-40, 1:1500), α-vimentin (Abcam, rabbit, 1:100), α-CK20 (DaKo, mouse, 1:200), α-Ki67 (Novocastra, mouse, 1:500), α-centrobin (a kind gift from Ciaran Morrison, mouse, 1:500 [Ogungbenro et al., 2018]), α-centrin (EMD Millipore, mouse, 1:1000), α-CP110 (a kind gift from Andrew Holland, rabbit, 1:1000), α-mouse-Alexa Fluor 594 (Molecular Probes, goat, 1:400), and α-rabbit-Alexa Fluor 488 (Molecular Probes, goat, 1:400). Staining of chicken tissues was performed with the following primary antibodies: α-ELAVL3/4 (Molecular Probes, mouse, 1:500), α-β-TubulinIII-Tuj1 (Covance, mouse, 1:1000), Pax6 (DSHB, mouse, 1:250), SOX2 (Invitrogen, rabbit, 1:500), pH3S10 (Millipore, rabbit, 1:500), and cleaved caspase-3 (Millipore, rabbit, 1:500). Staining of zebrafish tissues was performed with the following primary and secondary antibodies: α-ELAVL3/4 (GeneTex, rabbit, 1:1000), α-acetylated-alpha-tubulin (Santa Cruz Biotechnology, mouse monoclonal 6-11B-1, 1:1000), α-SOX2 (Abcam, rabbit, 1:1000), α-γH2AX (GeneTex, rabbit, 1:400), α-ADSL (MilliporeSigma, rabbit, 1:200), α-PKCζ (Santa Cruz Biotechnology, rabbit, 1:500), α-digoxigenin-AP Fab fragments (Roche, sheep, 1:5000 ISH), α-mouse-Alexa Fluor 568 (Molecular Probes, donkey, 1:1000), and α-rabbit-Alexa Fluor 488 (Molecular Probes, donkey, 1:1000).
RPE-1 (hTERT-RPE-1; ATCC) were transfected with 100 nM siRNAs (MilliporeSigma or Dharmacon) with Lipofectamine RNAiMAX (Thermo Fisher Scientific) in Opti-MEM (Gibco) without antibiotics for one or two rounds of 48 hr, depending on the gene to be silenced. We used siGFP (GGCUACGUCCAGGAGCGCCGCACC) and siGL2 (siC) (CGUACGCGGAAUACUUCGA) as negative controls. In this study, we used a smart pool (four siRNAs) against ADSL (Dharmacon) or single oligos siADSL#2 5′-CAAGAUUUGCACCGACAUA-3′ (MilliporeSigma). The siRNA-resistant mutant was designed to be resistant to siADSL#2. For rescue experiments with siCP110, we used three oligos (#1 5′-GCAAAACCAGAAUACGAGAUU-3′, #2 5′-CAAGCGGACUCACUCCAUATT-3′, and #3 5′- TAGACTTATGCAGACAGATAA-3′ [MilliporeSigma] for 24 hr).
RPE-1 cells (ATCC) were seeded in a six-well plate, silenced for 96 hr, washed twice in PBS, and resuspended in 300 μl of Tri-Reagent (MilliporeSigma). RNA was isolated by centrifugation followed by chloroform extraction, isopropanol precipitation, washing twice in 75% ethanol, and resuspended in 20 μl DEPC-treated water (Thermo Fisher Scientific). Total RNA was quantified with a Nanodrop 8000 Instrument (Thermo Fisher Scientific). 1 μg of total RNA was used for the reverse transcription reaction performed by High-Capacity RNA-to-cDNA Kit (Applied Biosystems), according to the manufacturer’s recommendations, in a 2× RT buffer mix, supplemented with dNTPs, random primers, and RT enzyme in a final volume of 20 μl. Quantitative real-time PCR (qRT-PCR) was performed using the comparative CT method and a Step-One-Plus Real-Time PCR Instrument (Thermo Fisher). Amplification of the 16 ng of cDNA was done in triplicate with TaqMan Universal PCR Master Mix (Thermo Fisher) for ADSL and GAPDH.
The siRNA-resistant mutant, ADSL*, was produced by introducing five different silent mutations using the QuikChange mutagenesis kit (Thermo Fisher) with the following primers: forward, 5′-GGTTTGCCAGGAGGCGTAGGTCTTTGCAAATTGTGTGCACTGATGCCCCCA-3′; reverse, 5′-CCAAACGGTCCTCCGCATCCAGAAACGTTTAACACACGTGACT ACGGGGGT-3′. Constructs were checked by sequencing (Macrogen), and expression was checked by Western blot and immunofluorescence into the pLenti-CMV-eGFP-BLAST (659-1) plasmid, a gift from Eric Campeau and Paul Kaufman (Addgene plasmid #17445; http://n2t.net/addgene:17445; RRID:Addgene_17445) (Campeau et al., 2009) using the primers containing XhoI and EcoRI overhangs (ADSL-XhoI forward 5′-AAAACTCGAGCGATGGCGGCTGGAGGCGATCAT-3′ and ADSL-EcoRI reverse 5′-TTTTGAATTCCAGACATAATTCTGCTTCA-3′). The siRNA-resistant mutant was produced by introducing five different silent mutations using the QuikChange mutagenesis kit (Thermo Fisher) with the following primers: forward, 5′-GGTTTGCCAGGAGGCGTAGGTCTTTGCAAATTGTGTGCACTGATGCCCCCA-3′; reverse, 5′-CCAAACGGTCCTCCGCATCCAGAAACGTTTAACACACGTGACT ACGGGGGT-3′. Constructs were checked by sequencing (Macrogen), and expression was checked by Western blot and immunofluorescence. For lentivirus preparation: 6 × 106 AD293 cells were plated in 15 cm culture dishes and transfected with 20 μg pLenti-CMV-EGFP empty and pLenti-CMV-ADSL*-EGFP, 2 μg pRSV-REV, a gift from Didier Trono (Addgene plasmid # 12253; http://n2t.net/addgene:12253; RRID:Addgene_12253), 6 μg pMDLg/pRRE, a gift from Didier Trono (Addgene plasmid # 12251; http://n2t.net/addgene:12251; RRID:Addgene_12251), and 2 μg pCMV-VSV-G, a gift from Bob Weinberg (Addgene plasmid # 8454; http://n2t.net/addgene:8454; RRID:Addgene_8454), plasmids with 160 μl PEI pH 7.0 (Polyscience Euro), and 150 mM NaCl (Dull et al., 1998; Stewart et al., 2003). After 48 hr, the medium containing the viruses was cleared with a 0.45 mm filter (Millipore) and added to the target cells. Three days after the infection, cells were selected with blasticidin (Invitrogen) for 7 days.
Silenced RPE-1 cells were seeded on 18 mm round coverslips after 96 hr of silencing and fixed accordingly with the antibody requirements, with 4% paraformaldehyde (PFA) for 10 or 30 min, followed by permeabilization in 0.1% Triton-PBS for 5 min and stored in 100% EtOH. For RPA2 and γH2AX detection, pre-extraction was performed using cold 0.2% Triton X-100 in 1× PBS on ice for 5 min, before fixation and permeabilization as described before. Cells were incubated with the blocking solution of 3% bovine serum albumin (MilliporeSigma) in PBT for 30 min. Primary antibodies (listed below) were diluted in the same blocking solution and incubated for 1 hr at room temperature (RT). After three washes, cells were incubated with Alexa Fluor-conjugated 594 and 488 secondary antibodies (Thermo Fisher Scientific) at 1:400 dilution for 1 hr at RT. DAPI was used to visualize the DNA. Slides were imaged using Orca AG camera (Hamamatsu) on a Leica DMI6000B microscope equipped with 1.4 100× oil immersion objective. AF6000 software (Leica) was used for image acquisition. Image processing and quantification was performed with ImageJ software. Intensities were measured in images acquired with the same exposure settings and subtracting the background for each image.
150,000 RPE-1 cells were plated in six-well plates and silenced with control or siADSL oligos (MilliporeSigma or Dharmacon) for 72 hr, when they were counted and plated again in the same amount for the second round of silencing. After 3 days, cells were counted as second timepoint (144 hr, 6 days) and seeded for a third timepoint (9 days). Cells were cultured in the presence of serum for all the experiments. The ΔPDL (difference in population doubling levels) was obtained by using the formula: log(N1/N0)/log2, where N1 is the number of cells at the timepoint we collected them and N0 is the initial number of cells plated (Pantoja and Serrano, 1999). For detecting cell death, cells in suspension were collected in the growth medium and the attached ones were trypsinized and resuspended in complete medium to block trypsin activity. Cells were then mixed in 0.4% Trypan Blue solution (Gibco). The number of blue-positive cells and total cell number was quantified at the microscope.
10 * 106 cells were collected after 96 hr of siRNA depletion, as previously described, and fixed in cold 100% EtOH dropwise and stored at –20°C for 24 hr. Cells were stained with propidium iodide (MilliporeSigma) in PBS and RNase, and 10,000 cells were analyzed at the Parc Cientific de Barcelona flow cytometry facility using a Gallios (Beckman Coulter) instrument. The data were collected and analyzed using FlowJo v10.7.2.
RPE-1 cells were seeded in a six-well plate, and after 96 hr of silencing they were trypsinized, washed once in PBS, and resuspended in a 2× SDS lysis buffer (2× SDS lysis buffer contained 4% SDS, 20% glycerol, 120 mM Tris/HCl pH 6.8, 1X protease [Roche], and phosphatase inhibitors [MilliporeSigma]). Protein concentration was quantified using the DC Protein Assay (Bio-Rad), and proteins separated by SDS-PAGE and transferred to 0.2 μm nitrocellulose membrane (Amersham Protran) or 0.45 μm PVDF membrane (MilliporeSigma) depending on the molecular weight. Membranes were blocked in 5% milk in PBT (PBS containing 0.2% Tween-20) for 30 min and then incubated with primary antibodies for 1 hr at RT. After three washes in PBS containing Tween-20 0.02%, membranes were incubated with secondary antibodies conjugated to HRP and protein bands were visualized by ECL-Plus (MilliporeSigma).
RPE-1 were silenced for 96 hr with siControl and siADSL#2, then fixed in ice-cold X-gal fixative solution (containing 4% formaldehyde, 0.5% glutaraldehyde, 0.1 M sodium phosphate buffer pH 7.2) for 4 min. After two washes in PBS, X-gal (Roche) was diluted 1:100 at a final concentration of 1 mg/ml in X-gal solution (containing 5 mM K3Fe(CN)6, 5 mM K4Fe(CN)6, 2 mM MgCl2 in PBS). Incubation was performed at 37°C for 8 hr in the dark. Two washes in PBS were performed before taking the images. Doxorubicin was used as a positive control.
For Figure 1C and Figure 1—figure supplement 2A, metabolites were extracted from snap-frozen cell pellets by adding 300 μl ACN:MeOH:H2O (5:4:1, v:v:v) solution and vortexing samples for 30 s. Samples were immersed in liquid N2 to disrupt cell membranes followed by 30 s of ultrasonication. These two steps were repeated three times. Then, samples were incubated at −20°C for 1 hr, centrifuged at 15,200 rpm for 10 min at 4°C, and the supernatant was collected into an LC-MS vial.
For cell media, metabolites were extracted from 200 μl of lyophilized media by adding 300 μl acetyonitrile:MeOH:H2O (5:4:1, v:v:v) and vortexing samples for 30 s. Then, samples were incubated at −20°C for 20 min, centrifuged at 15,200 rpm for 10 min at 4°C, and the supernatant was collected into an LC-MS vial. Samples were injected (1 μl of media and 10 μl of cell extracts) in a UHPLC system (1290 Agilent) coupled to a triple quadrupole (QqQ) MS (6490 Agilent Technologies) operated in multiple reaction monitoring (MRM) and positive (POS) or negative (NEG) electrospray ionization (ESI) mode. Source parameters were gas temperature (°C) = 270; gas flow (l/min) = 15; nebulizer (psi) = 25; capillary (V) = 3500 (POS) and 2500 (NEG). Metabolites were separated using an ACQUITY UPLC BEH HILIC 1.7 μm, 2.1 × 150 mm chromatography (Waters) at a flow rate of 0.4 ml/min. The solvent system was A = 50 mM ammonium acetate in water and B = acetonitrile. The linear gradient elution started at 5% A (time 0–1 min), 50% A (time 1–4 min), 5% A (time 4–4.5 min), and finished at 5% A (time 7 min). MRM transitions are shown in Appendix 2.
The ribosides SAICAr and SAdo and the internal standards 13C4-SAICAr and 13C4-SAdo were synthesized according to previously published procedures (Baresova et al., 2016; Zikánová et al., 2005). SAICAr and S-Ado analysis was performed essentially as previously described (Mádrová et al., 2018). For Figure 1—figure supplement 2B and C, RPE-1 or HeLa cells were sonicated in 10 mM Tris pH 8,2, 2 mM EDTA, 10 mM KCl, 1 mM DTT, and 4% glycerol, centrifugated, metabolites extracted from 50 μl of lysate diluted to protein concentration 1 mg/ml with 200 μl of extracting solution (ACN:MeOH 1:1, v:v, 0.125 M formic acid, 15 nM 13C4-SAICAr and 13C4-SAdo), and incubated on dry ice for 30 min. After centrifugation, the supernatant was deep-freeze-dried and pellets were resuspended in 50 μl of water, briefly centrifuged, and analyzed by LC-MS/MS.
The reverse-phase column Prontosil 120-3C18 AQ 150 × 3 mm (Bischoff Chromatography) and Agilent 1290 Infinity LC System coupled with an API 4000 triple quadrupole mass spectrometer with an electron ion source (Agilent Technologies) were used. The mobile phase consisted of 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B). The gradient elution was performed as follows: t = 0.0 min, 100% A; t = 2.5–6.5 min, linear gradient to 20% B; t = 6.5–7.5, linear gradient from 20% to 60% B; t = 7.5–9.0 min, 90% B and then a regeneration of the column for 12 min. The flow rate was set to 0.3 ml/min for the first step and increased to 0.4 ml/min. The injection volume was 5 μl. MRM transitions for SAICAr, SAdo, SAICAR, SAMP, and their corresponding internal standards are shown in Appendix 2.
The chromatographic conditions were the same as in the HPLC-HRMSn analysis mentioned above. Detection was performed on an Orbitrap Elite operating in positive ionization mode with the same setting as above. The detection method was divided into four time segments. Full scan analysis within the mass range m/z 70–1000 was performed in the first (0.0–3.0 min) and fourth (12.0–17.0 min) segments. The selected ion monitoring (SIM) method was applied in the second segment (3.0–7.0 min) for the analysis of ribosides (m/z 177–417) and in the third segment (7.0–12.0 min) for the analysis of ribotides (m/z 257–497) to enhance the sensitivity towards these metabolites (except for the measurement of SAdo, which had m/z ranges 379–389). The resolution was set to 60,000 FWHM. The mass error was below 3 ppm. All cell lines were measured in hexaplicate, and the intensity values are presented as averages. The identities of the accumulated compounds in both cell lysates and media were confirmed by MS2 fragmentation analysis. Fragmentation spectra were produced via CID with the fragmentation energy set to 30 units of normalized collision energy.
In vitro data were analyzed with an unpaired two-sided t-test when two samples were compared, while one-way ANOVA was used to compare more than two samples in the same graph (GraphPad Prism 6.0, GraphPad Software Inc). Values of p<0.05 were considered statistically significant (*p<0.05; **p<0.01; ***p<0.001; ****p<0.0001). Two or more independent experiments were performed for each condition, and this is indicated in individual figure legends. Statistical analysis for percentages of positive cells: counts of cells belonging to the positive and negative condition were used to fit a generalized linear model with binomial distribution. The R  packages ‘lme4’ (Jaeken and Van den Berghe, 1984) and ‘multcomp’ (Jurecka et al., 2015) were used to fit the model and compute raw and adjusted p-values with the ‘Shaffer’ method.
To generate a template for the generation of an antisense in situ probe, a 921 bp fragment of the Danio rerio adsl open-reading frame was cloned into pCRII via TOPO TA cloning (Invitrogen). To have a template for the generation of capped mRNA using the AmpliCap SP6 High Yield Message Maker Kit (Cellscript), the whole open-reading frame of zebrafish adsl was cloned with an N-terminal Flag-tag into pCS2+ using EcoRI and XhoI.
Zebrafish embryos were fixed with 4% buffered PFA at the indicated stages. Antibody staining was performed as described (Jaffe et al., 2010) using the primary antibodies previously described (see also section ‘Antibodies’) and detected with Alexa Fluor-labeled secondary antibodies (1:1000, Molecular Probes).
The number of fertilized eggs per clutch determined the size of experimental groups with clutches having been randomly and equally divided into treatment groups. No additional statistical methods have been applied to predetermine sample size. All zebrafish experiments were done at least three times with eggs from different mating tanks or different mating days. Embryo numbers are given in the legends. All statistical analyses were performed with GraphPad Prism 7 and 8, respectively. Data were tested for normality and analyzed accordingly by parametric or nonparametric tests. Graphs display, if not indicated otherwise, individual datapoints and medians in case of nonparametric datasets. An α level of <0.5 was considered significant.
Zebrafish were maintained in a 14 hr light and 10 hr dark cycle in a standardized, water recycling housing system (Tecniplast) with automatic monitoring and adjustments of pH, conductivity, and temperature. Fertilized eggs were generated by natural matings of the wild-type strains EK or AB. Eggs were incubated at 28.5°C and allowed to develop until the desired stages. In order to achieve Adsl knockdown, a translation blocking antisense MO (Adsl ATG MO) (5′-TCCCTCCATGCCTGCAGCGGTTAAA) was used or a MO that targets the exon-intron boundary at exon 4 of Adsl (Adsl SplMO) (5′-CCAACTGTGGGAGAGAGCGACTGTA). A standard control MO was also used in all experiments. MOs (GeneTools Inc) were injected at the 1–2-cell stage directly into the yolk. In addition, noninjected wild-type embryos served as internal control for clutch quality. For pharmacological manipulation, zebrafish embryos were immersed in embryo water containing 1% DMSO or 1% DMSO and 100 µM MTX (Cayman Chemical) from 10 until 24 hr post fertilization (hpf) or 50 µM nucleosides. All zebrafish maintenance and procedures have been approved by the Veterinary Care Unit at Ulm University and University of Tübingen, respectively, and the animal welfare commissioner of the regional board for scientific animal experiments in Tübingen, Germany. Zebrafish experiments were performed according to the European Union Directive 86/609/EEC for the protection of animals used for experimental and other scientific purposes.
Zebrafish were fixed overnight at 4°C at the indicated stages using 4% buffered PFA, dehydrated with a gradual methanol series, and stored at –20°C until further use. For in situ hybridization (ISH), embryos were rehydrated in a methanol series containing PBST (PBS containing 0.1% Tween-20) and processed according to standard protocols (Thisse and Thisse, 2014). Genes of interest were detected using DIG-labeled in situ probes, which were in vitro transcribed from linearized plasmids carrying fragments of the gene of interest: adsl (GenBank no.199899.2), angiopoietin-like 3 (angptl3, GenBank no. AF379604). The probes against cardiac myosin light chain 2 (cmcl2) and spaw have been described before (Burkhalter et al., 2013).
4 days post fertilization (dpf), old zebrafish embryos were fixed for 2 hr at RT using 4% buffered PFA. After rinsing with PBS, embryos were washed for 10 min with 50% EtOH in PBS before the staining solution (0.02% Alcian Blue [MilliporeSigma], 70% EtOH, 50 mM MgCl2) was added, and the embryos were incubated overnight at RT. On the next day, embryos were rinsed with H2O and subsequently bleached for 20 min at RT with opened lid of the reaction tube (bleaching solution: 1.5% H2O2 in 1% KOH). A clearing series was performed (30 min 20% glycerol/0.25% KOH, 2 hr 50% glycerol/0.1% KOH). Stained embryos were stored at 4°C in 50% glycerol/0.1% KOH.
To count neural progenitors, anterior views of 24 hpf embryos were taken using a fluorescent whole-mount microscope. The number of Sox2-positive cells within the forebrain was determined. To count differentiated neural cells, dorsal views of embryos were captured by fluorescent whole-mount microscopy and the number of ELAVL3/4-positive cells per 100 µm was counted. γH2AX-positive cells were counted over a distance of 300 µm in the neural tube. Cilia were counted and measured after acquiring confocal z-stacks of flat-mounted tails of six somite stage (ss) embryos. The Simple Neurite Tracer in Fiji was used to trace and measure cilia through the whole z-stack. ImageJ was also used to trace and measure the outline of the KVs.
Live zebrafish embryos and those processed by ISH or for cartilage staining were imaged using a M125 whole-mount microscope equipped with a Leica IC80 HD camera. Zebrafish embryos undergoing immunofluorescence stainings were assessed with an M205 FCA and a DFC 9000 GT sCMOS camera. Confocal z-stacks were acquired on a TCS SP5II with LAS AF software (all microscopes and software: Leica).
Eggs from white leghorn chickens were incubated at 37.5°C in an atmosphere of 45% humidity, and the embryos were staged according to HAMBURGER and HAMILTON, 1951. Chick embryos were electroporated with column-purified plasmid DNA (3 μg/μl for shRNAs) in H2O containing Fast Green (0.5 μg/μl). Briefly, plasmid DNA was injected into the lumen of HH12 or HH16 neural tubes, electrodes were placed on either side of the neural tube, and electroporation was carried out by applying five 50 ms square pulses using an Intracel Dual Pulse (TSS10) electroporator set at 25 V. Transfected embryos were allowed to develop to the specific stages and then dissected under a fluorescence dissection microscope.
shRNAs were generated using pSHIN plasmid (a GFP expressing evolution of pSUPER): shCONTROL sequence (CCGGTCTCGACGGTCGAGT) and shADSL sequence (GAGCTGGACAGATTAGTGA). The knockdown efficiency of shRNAs was assessed by RT-qPCR in electroporated chicken embryonic fibroblast cultures (Herrera et al., 2014).
Embryos were fixed overnight at 4°C in 4% PFA, and immunostaining was performed on vibratome sections (60 μm) following standard procedures. After washing in PBS-0.1% Triton X-100, the sections were incubated overnight with the appropriate primary antibodies diluted in a solution of PBS-0.1% Triton supplemented with 10% bovine serum albumin. After washing in PBS-0.1% Triton, sections were incubated for 2 hr at RT with the appropriate Alexa-conjugated secondary antibodies diluted in a solution of PBS-0.1% Triton supplemented with 10% bovine serum albumin. After staining, the sections were mounted and examined on a Leica SP5 or a Zeiss Lsm 780 multiphoton microscope. For EdU incorporation, 200 μl of EdU solution (1 mM) was added on the vitelline membrane of each embryo 2 hr before fixation in 4% PFA. EdU was detected in sections using the Click-iT EdU imaging kit (Invitrogen).
HH-12 chicken embryos were electroporated with shCONTROL or shADSL plasmids, and 48 hr post electroporation (hpe), a single-cell suspension was obtained by digestion for 10–15 min with Trypsin-EDTA (MilliporeSigma) and labeled with Hoechst and α-ELAVL3/4 antibody used with Alexa 647-conjugaded anti-mouse secondary antibody. Alexa 647, Hoechst, and GFP fluorescence were determined by FACSAria Fusion cytometer (BD Biosciences), and the data were analyzed with FlowJo software (Tree Star) and Multicycle software (Phoenix Flow Systems; cell cycle profile analysis).
Quantification of cleaved caspase-3 immunofluorescence intensity was done using ImageJ software. Tuj1+ and Tuj1- areas on the electroporated side and the respective areas on the non-electroporated side were delimitated by polygonal selection, and the mean intensity of cleaved caspase-3 immunofluorescence was quantified as mean gray values. At least three different images were used to calculate the mean value per embryo. Each mean value was normalized to the mean value obtained for the respective non-electroporated area of the same embryo.
Summary of phenotypes identified in the systems used. Corresponding figures are indicated, as well as interventions that rescue or recapitulate the phenotype.
|In vitro (RPE-1)|
|Phenotype||Defined by||Complemented with||Recapitulated by||Figures|
|Cell cycle exit||Ki67 positivity||ADSL expression, p53-KO||-||Figure 1E, H, I and L, Figure 1—figure supplement 2I|
|Cell cycle arrest (G1)||Increased P53 levels, flow cytometry, cumulative growth||P53 deletion||-||Figure 1D, F, G, J and K, Figure 1—figure supplement 2E and H|
|Elevated DNA damage signaling||γH2AX, 53BP1 foci||ATM inhibitor (53BP1), nucleosides, adenosine||Hydroxyurea, ionizing radiation||Figure 2B–H|
|Elevated replication stress||Chromatin bound RPA||Nucleosides, adenosine||-||Figure 2A|
|Ciliogenesis defect||IF analysis of ARL13B, centriole markers, CP110||PAICS inhibition, MTX treatment, CP110 depletion||SAICAr addition||Figure 4B–G, Figure 4—figure supplement 1C–E, Figure 5A–E|
|In vivo (chicken)|
|Phenotype||Defined by||Complemented with||Recapitulated by||Figure|
|Microcephaly||Reduced neural tube size||-||-||Figure 3B|
|Cell cycle arrest (G2)||Edu/H3S10 IF, Flow cytometry||-||-||Figure 3C–F|
|Reduced Sox2+ progenitors||SOX2 IF, flow cytometry||-||-||Figure 3B and C|
|Reduced differentiated cells||ELAVL3/4 IF||-||-||Figure 3B and C, Figure 3—figure supplement 1B|
|In vivo (fish)|
|Phenotype||Defined by||Complemented with||Recapitulated by||Figure|
|Microcephaly||Pinhead, hydrocephalus, pericardial edema, kinked tail, defective skull formation||Fish or human ADSL expression||-||Figure 6A–F|
|Reduced Sox2+ progenitors||SOX2 IF||MTX treatment||-||Figure 8B and C|
|Reduced differentiated cells||ELAVL3/4 IF||MTX treatment, ADSL expression||-||Figure 8A and B, Figure 8—figure supplement 1|
|Ciliogenesis defect||Left-right patterning defect, reduced cilia length||ADSL expression||-||Figure 6A–F, Figure 7A–G|
|Elevated DNA damage signaling||gH2AX IF||Nucleosides||-||Figure 6G|
Multiple reaction monitoring transitions. See details in Materials and methods.
|Multiple reaction monitoring transitions|
|Metabolite||First transition||Second transition||Polarity|
Most data generated or analysed during this study are included in the manuscript and supporting source data files. Additional source data is available via Figshare, https://doi.org/10.25452/figshare.plus.c.5793614.
Inactivation of mTORC1 in the developing brain causes microcephaly and affects gliogenesisThe Journal of Neuroscience 33:7799–7810.https://doi.org/10.1523/JNEUROSCI.3294-12.2013
A third-generation lentivirus vector with a conditional packaging systemJournal of Virology 72:8463–8471.https://doi.org/10.1128/JVI.72.11.8463-8471.1998
A series of normal stages in the development of the chick embryoJournal of Morphology 88:49–92.https://doi.org/10.1002/aja.1001950404
Proteomic Analysis Reveals GMP Synthetase as p53 Repression Target in Liver CancerThe American Journal of Pathology 187:228–235.https://doi.org/10.1016/j.ajpath.2016.09.022
Cortical neurogenesis in the absence of centriolesNature Neuroscience 17:1528–1535.https://doi.org/10.1038/nn.3831
An infantile autistic syndrome characterised by the presence of succinylpurines in body fluidsLancet (London, England) 2:1058–1061.https://doi.org/10.1016/S0140-6736(84)91505-8
Clinical, biochemical and molecular findings in seven Polish patients with adenylosuccinate lyase deficiencyMolecular Genetics and Metabolism 94:435–442.https://doi.org/10.1016/j.ymgme.2008.04.013
SAICAR stimulates pyruvate kinase isoform M2 and promotes cancer cell survival in glucose-limited conditionsScience (New York, N.Y.) 338:1069–1072.https://doi.org/10.1126/science.1224409
Mutant p53: Multiple Mechanisms Define Biologic Activity in CancerFrontiers in Oncology 5:249.https://doi.org/10.3389/fonc.2015.00249
A USP28-53BP1-p53-p21 signaling axis arrests growth after centrosome loss or prolonged mitosisThe Journal of Cell Biology 214:143–153.https://doi.org/10.1083/jcb.201604054
Loss of CPAP in developing mouse brain and its functional implication for human primary microcephalyJournal of Cell Science 133:jcs243592.https://doi.org/10.1242/jcs.243592
Nutrient restriction causes reversible G2 arrest in Xenopus neural progenitorsDevelopment (Cambridge, England) 146:dev178871.https://doi.org/10.1242/dev.178871
Centrobin controls primary ciliogenesis in vertebratesThe Journal of Cell Biology 217:1205–1215.https://doi.org/10.1083/jcb.201706095
A primary microcephaly protein complex forms a ring around parental centriolesNature Genetics 43:1147–1153.https://doi.org/10.1038/ng.971
Lentivirus-delivered stable gene silencing by RNAi in primary cellsRNA (New York, N.Y.) 9:493–501.https://doi.org/10.1261/rna.2192803
Succinylpurines induce neuronal damage in the rat brainAdvances in Experimental Medicine and Biology 431:185–189.https://doi.org/10.1007/978-1-4615-5381-6_36
Epithelial-mesenchymal transition and proliferation of retinal pigment epithelial cells initiated upon loss of cell-cell contactInvestigative Ophthalmology & Visual Science 51:2755–2763.https://doi.org/10.1167/iovs.09-4725
In situ hybridization on whole-mount zebrafish embryos and young larvaeMethods in Molecular Biology (Clifton, N.J.) 1211:53–67.https://doi.org/10.1007/978-1-4939-1459-3_5
Zebrafish Models of Neurodevelopmental Disorders: Limitations and Benefits of Current Tools and TechniquesInternational Journal of Molecular Sciences 20:E1296.https://doi.org/10.3390/ijms20061296
Akt Regulates a Rab11-Effector Switch Required for CiliogenesisDevelopmental Cell 50:229–246.https://doi.org/10.1016/j.devcel.2019.05.022
Hedgehog receptor function during craniofacial developmentDevelopmental Biology 415:198–215.https://doi.org/10.1016/j.ydbio.2016.02.009
Preparation of 5-amino-4-imidazole-N-succinocarboxamide ribotide, 5-amino-4-imidazole-N-succinocarboxamide riboside and succinyladenosine, compounds usable in diagnosis and research of adenylosuccinate lyase deficiencyJournal of Inherited Metabolic Disease 28:493–499.https://doi.org/10.1007/s10545-005-0493-z
Fadel TissirReviewing Editor; Université Catholique de Louvain, Belgium
Anna AkhmanovaSenior Editor; Utrecht University, Netherlands
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
[Editors' note: this paper was reviewed by Review Commons.]https://doi.org/10.7554/eLife.70518.sa1
Reviewer #1 (Evidence, reproducibility and clarity (Required)):
The presented work by Stracker and Co attempts to elucidate ADSL deficiency in neurodevelopment using a variety of exciting approaches. The presented experiments are thorough, and an orthogonal experiment supports each finding. Overall, the manuscript is technically solid and brings new ideas and finding to the field. in vivo data with Chicken and Fish experiments of high quality. In my opinion, the manuscript requires are clarifications with minor sets of experiments as below.
Authors claim that upon ADSL depletion, cells exit the cell cycle as well as p53 dependent arrest. Do they mean a proportion of cells have exited (G0), and the rest got stuck G2 to M transition via p53 dependent manner? I believe these two may not go well hand to hand. If so, it has to be dissected with proper quantifications.
Our analysis of cell cycle exit in RPE1 cells is based on Ki67 positivity (marker of cycling cells) in the population. To address this point, we performed flow cytometric analysis of cells using propidium iodide to allow the assessment of cell cycle profiles using DNA content. In wild type RPE1 cells, we observe a higher proportion of cells in G1 with a corresponding reduction of cells in S/G2 phase of the cell cycle, consistent with p53 induction driving a G1 arrest in fibroblasts. In contrast, this difference in cell cycle was not observed in p53 deficient RPE1 cells, indicating the arrest is p53 dependent following ADSL depletion. This data is included in new Figure 1—figure supplement E and H in the revised manuscript.
As we previously mentioned, we expected that RPE1 cells in culture may behave differently than NPCs in vivo. However, we believe the phenotype remains informative and demonstrates a clear impact of ADSL depletion on cell cycle progression, consistent with other data.
However, for me, it appears that these two phenomena are unrelated, as I could see from their Chicken experiment (which is done cleanly) that SOX2 positive cells get stuck at G2. Cell cycle exit would have caused depletion of NPCs, leading to NPCs’ overall reduction with the simultaneous increase in neurons. To prove their cell cycle exit theory, the authors should also conduct serum starvation and stimulation experiments.
As we previously indicated, we are not reporting that chicken NPCs are necessarily exiting the cell cycle in the ADSL depleted chicken embryos. NPCs arrest in G2, and we see slightly more Sox2+ NPCs and a consequent reduction in differentiated cells. We tried extensively to optimize p53 and Ki67 staining for use in the chicken samples to address this further but were unsuccessful.
We believe the NPCs in vivo are a better system for addressing the mechanisms underlying microcephaly but feel that reporting the RPE1 data is also important, as it may reflect defects in other cell types depleted for ADSL. We can also obtain more detail in RPE1s due to the nature of the in vitro system that is more amenable to controlled manipulation. We will clarify this point in the text related to several of the additional comments below.
We cannot perform the requested serum starvation/stimulation experiments in the NPCs in vivo, as it is not technically possible in this system, but we included this data for RPE1 cells in new Figure 4—figure supplement 1F.
Overall, It appears that the cells have G2 arrest predominantly, and cell cycle exit defect is probably not present (at least their data does not sufficiently prove it, see below).
We agree and again want to reiterate that while the 2 different systems both show a clear cell cycle progression phenotype, they do not behave in exactly the same way. They are different cell types, different organisms, and comparing the cultured cells to a subpopulation in vivo has many caveats that we indeed recognize (ex. constant serum stimulation in culture vs growth factor gradients and limitations in vivo). As mentioned, we tried extensively to optimize p53 and Ki67 for use in the chicken embryos, purchasing several different antibodies and working with our histochemistry core facility, but were unable to achieve staining that was reliable for quantification. We therefore are technically unable to address cell cycle exit in the chicken system but agree that the in vivo data suggests it is unlikely in NPCs (at least in the time frame of our analysis), a point we will address in the revised discussion.
Figure 4: Measuring cell cycle parameters will help assess whether there is an undisrupted cell cycle progression in the p53 KO lines.
As previously described, we carried out cell cycle analysis in RPE1 lines to examine the distribution following ADSL depletion and found that indeed, wild type cells arrested in G1 in a p53 dependent manner (new Figure 1—figure supplement 2E and H). However, we recognize that this may not directly inform about the potential role of p53 in vivo. This is unfortunately a limitation of the chicken embryo system.
Figure 5B should use cilia imaging to ensure that the authors look at the proper cell stage. Interestingly, SAICAr impairs the removal of CP110. The details are missing such that the concentrations used (at the figure panel itself).
In agreement with the literature, Figure 4A confirms that serum starvation causes essentially all cells to exit the cell cycle and produce a cilium. Thus, we can conclude that essentially all cells in Figure 5B are in G0. We have reviewed the legends and methods and added all pertinent details to both for clarity.
It is essential to perform this experiment in cycling cells to avoid the effect of serum starvation. Again, cell cycle parameters should be measured whether there is a defective proliferation (or G2 arrest).
We established in Figure 1—figure supplement 2E that there is not a G2 arrest in RPE-1, but rather a p53-dependent G1 arrest. As the reviewer is likely aware, serum starvation is used in this assay to drive cells into G0, which is permissive for cilium assembly. This allows us to compare ciliogenesis in control and ADSL knockdown cells, independently of any cell cycle effects (in cycling cells cilia are only transiently assembled in G1 and only in roughly 20% of cells in RPE1). We feel that performing this experiment in cycling cells will not be very informative due to the low percentage of ciliated cells and the fact that cell cycle defects and DNA damage signaling in ADSL-depleted cells will further complicate the analysis.
However, we examined cilia disassembly following serum add back in RPE-1 (roughly as performed in (Gabriel et al., 2016)) and we did not observe a defect in timely disassembly, as has been implicated in NPC attrition in Seckel Syndrome. We have included this data in Figure 4—figure supplement 1F and indicated this point in the text and discussion.
The experiments in Figure 5 are very intriguing and bring valuable insights. Numerous graphs without representative figures make it hard to interpret. I recommend putting the findings in a tabular form and provide respective imaging data.
We added additional representative data for experimental conditions to the new Figure 5. As many of the graphs are repeats with different siRNAs to control for off target effects, we did not add representative panels for each condition to avoid crowding the figure, as the effects of the CP110 siRNAs are fairly uniform regardless of the siRNA used.
In the Discussion section, maybe the authors can help me get it straight. ADSL depletion accompanied by cilia loss is suggested to elevate the cellular level of SAICAr. In their orthogonal experiment, adding the very same compound impairs CP110 removal, a negative regulator of ciliogenesis. The authors should at least discuss any potential mechanisms for CP110 removal by this SACAr, an important aspect.
The reduction of ADSL in human patients, as well as numerous experimental systems, was shown to elevate SAICAR/r and S-Amp/S-Ado levels (substrates of ADSL in their monophosphate and riboside forms, respectively). We have
focused on the riboside forms that are what is predominantly detected in patients and cell lines. We observed that upon ADSL depletion, ciliogenesis is impaired (after serum starvation). Nucleoside supplementation does not rescue the phenotype, indicating that a lack of purines does not underlie the phenotype. The administration of SAICAr phenocopies this, indicating that the accumulation of SAICAr, due to a disruption in DNPS specifically at the ADSL step, is therefore likely responsible for the effects on ciliogenesis observed following ADSL depletion. This is further corroborated by the fact that both MRT00252040 and MTX, that inhibit DNPS upstream of ADSL, can rescue the defect, while DNPS remains impaired. However, we cannot rule out that the physical depletion of ADSL per se leads to some unexpected signaling or effects on the stability of other proteins.
To address this further, we carried out an analysis of SAICAr levels following
ADSL depletion in RPE-1 cells. Unfortunately, these cells have very low levels of SAICAr and we are technically unable to detect it at levels where we can reliably assess differences (new Figure 1—figure supplement 2B-C). In experiments in HeLa cells performed in parallel, we can readily detect SAICAr and its increase following ADSL depletion – note that levels in HeLa are roughly 250X higher. We would be very surprised if a similar relative increase was not occurring in RPE-1 but due to differences in core metabolism, we are simply below the limits of reliable detection of differences in that cell line. While we still believe that SAICAr accumulation is the most likely explanation, we have recognized this caveat in the interpretation of the data and altered the text accordingly. In addition, we revised the title to Pathway, rather than Metabolite, specific effects, to take into account other possibilities.
How SAICAr would impair CP110 removal and ciliogenesis remains an important and interesting question that we have tried extensively to address. We have revised the discussion to highlight several possibilities but will mention a few here.
SAICAR has been shown to activate PKM2, that can then phosphorylate numerous other kinases in the context of cancer cells (Keller et al., 2014, 2012). There is not a clear connection between PKM2 and ciliogenesis that we are aware of. However, its effect on a particular kinase or other enzyme could influence the process indirectly through signal transduction.
One putative PKM2 target identified was AKT3. AKT signalling was recently implicated in controlling ciliogenesis upstream of CP110 removal. We considered that ADSL depletion and SAICAr administration could then potentially lead to increased AKT3 activation that would impair ciliogenesis in the absence of serum. However, we cannot detect any indication that AKT is activated by SAICAr treatment, suggesting that this is not the underlying mechanism. However, it remains possible that aberrant kinase activation by a yet unidentified kinase (or kinases) underlies the phenotype. In future work, it would be of interest to address this with phosphoproteomics but this is out of the scope of the current manuscript.
Alternatively, ADSL may interact with and regulate a protein involved in ciliogenesis. Proximity labeling using multiple centrosomal baits identified ADSL as an interactor, suggesting that ADSL (or a subset of the ADSL pool) may localize to the cilia(Gupta et al., 2015). We attempted to validate this by examining ADSL localization in IF, but could not detect a clear centrosomal signal. However, it remains possible that a subset of the ADSL interacts directly with centrosomal proteins to regulate their functions in some way.
We have extensively revised the discussion to include these possibilities and accompanying references.
The authors discuss p53 dependent cell death in Seckel syndrome (or when CPAP is eliminated). This is an important point that stem cell death causing microcephaly that is so far observed only in mouse models (few works that used CPAP elimination). It is somehow misinterpreted in the field that stem cell death is a significant cause of microcephaly. In the present work, this is not the case which is new evidence. It is also true that when Seckel syndromic microcephaly was modeled in human brain organoids (By Gabriel et al. modeling microcephaly due to CPAP mutation and WDR67 KO). Importantly, these experiments used disease-relevant cell types that did not show any cell death, but the microcephaly was due to cilia abnormality. It is disappointing why the authors ignore these essential works, which support their current work. Addressing these clarifications should result in a solid manuscript for publication.
We appreciate the reviewer pointing out these details, it is clearly relevant to discuss, as we indeed did not see apoptosis in vivo. Due to involvement in other work addressing apoptosis mechanisms, our discussion was admittedly limited to this area. As previously discussed, we did examine cilia removal after serum starvation and did not see defects in ADSL depleted cells. While we cannot discount that defects in cilia are related to the cell cycle issues we observed in chicken embryos, we think it is unlikely to be the main issue here based on this experiment. In the revision, we extended the discussion and add additional related references to explore this issue further.
It would be helpful if the authors explain what is SAICAr and S.Ado in the abstract itself. The abstract is very convoluted for a non-specific reader. I got the meaning of the abstract after reading the manuscript.
Based on several other comments, we have tried to make the abstract less technical in this regard and we added a schematic (new Figure 1—figure supplement 1) that shows the relationship between the enzymes and metabolites as suggested by another reviewer to address this point (see below). We also added a phenotype table (Appendix 1) to allow faster comparison of cellular and organismal phenotypes and the ID of the relevant figure panels.
Reviewer #1 (Significance (Required)):
Reviewer #2 (Evidence, reproducibility and clarity (Required)):
The authors chose to investigate the link between ADSL deficiency and microcephaly by studying the impact of ADSL experimental depletion on cell proliferation, fitness and fate, using a combination of in vitro and in vivo approaches and different cell and model organisms, i.e. RPE1 cells, chicken and zebrafish. The authors propose that both nucleotide depletion and intermediate metabolite production are responsible for neurodevelopmental defects. Furthermore, they show impaired cell cycle progression, defective primary ciliogenesis, as well as DNA damage after ADSL depletion. The manuscript is of good quality, well written and referenced. The data and methods are clearly presented, illustrated and discussed. I have however two main criticisms regarding the manuscript.
My first criticism concerns the conclusions taken on the consequences of reduced purine levels.
1. The contribution of reduced purine levels after ADSL depletion is only relying on the capacity of experimental nucleoside supplementation to rescue or not the phenotype of interest. Nonetheless, there are no clear information throughout the manuscript on the choice of dose used to supplement the purine deficiency and supposedly restore nucleoside concentration, which is to my sense a critical point. More importantly, we cannot appreciate whether such addition of nucleosides is readily compensating the deficiency without generating another imbalance. The authors should evaluate quantitatively whether their experimental conditions by nucleoside supplementation are sufficient to restore or at least to which extent they restore purine-based nucleoside concentration.
The nucleoside supplementation experiments were performed using a concentration based on numerous other experiments that used this intervention to address replication stress (Halliwell et al., 2020; Ruiz et al., 2015; Kanu et al., 2016; Parsels et al., 2018). This is a media supplement that has been widely used in the replication stress field and sufficient to rescue genetic and pharmacologically induced replication stress. Our results indicate that this is sufficient to functionally restore nucleotide levels, including after treatment with Hydroxyurea (new Figure 2E), however we recognize that they are most likely in excess of what they were prior to treatment.
We repeated key experiments with the addition of only Adenosine, as suggested by the reviewers (see new Figure 1—figure supplement 1) and we observed a similar rescue of specific phenotypes associated with ADSL depletion, indicating this is likely due to the restoration of limiting purine nucleotides, as we proposed. We have also added new data in zebrafish that shows that nucleoside supplementation of embryos is not sufficient to rescue the reduction in Sox2+ progenitor levels. This is consistent with data from human patients that have normal purine levels and the fact that we can partially rescue progenitors using MTX, that targets the pathway upstream of ADSL, thus implicating metabolic intermediates such as SAICAr rather than purine deficiency.
We do not feel it is feasible or necessary to repeat all experiments to measure precise levels of individual nucleosides/tides, as this is unlikely to make the data more informative and we do believe it will not be trivial to achieve consistent physiological levels in this type of cell culture addback experiment.
2. The effects of nucleoside supplementation can also be questioned after observing that nucleoside addition in control conditions appears sufficient to decrease the percent of Ki67 positive cells (Figure 1I), p53 positive cells (Figures 1J) and the percent of gH2Ax nuclear intensity (Figure 1D). The authors should explain this observation in the manuscript.
It is true that we observed minor alterations in the percentages of these parameters in control cells following nucleoside supplementation. In the case of Ki67, these effects are marginal (Figure 1I) and not statistically different. The fact that p53 and gH2AX are also slightly reduced (Figure 1J and 2D) is not entirely surprising and it is consistent with the fact that nucleosides may suppress culture stress arising from atmospheric oxygen. Similar observations have been made previously and this is why nucleoside supplementation is used in culture of ESCs and iPS cells.
Nucleosides of course have many other cellular roles and we do not see any possible way that it is feasible to ensure they are only used in DNA replication, regardless of how carefully the concentration is controlled. However, I think that the important point here is that the clear difference between control and ADSL depletion remains with regards to reduced Ki67 and increased p53 despite the minor effects on control RPE-1.
In contrast to p53 and Ki67, nucleosides suppress DNA damage signaling that we
believe is arising from mild replication stress. Treatment of cells with hydroxyurea (HU), a ribonucleotide reductase inhibitor, is a common intervention used to induce replication stress by limiting the nucleotide pool. We tested this nucleoside intervention on HU treated cells and found it was also sufficient to strongly reduce DNA damage (as has been previously published)(new Figure 2E). Therefore, we know that this intervention is capable of rescuing a strong inhibition of the ribonucleotide reductase.
As mentioned, we examined key endpoints with the restoration of only Adenosine to determine if this is phenotypically similar and obtained similar results, indicating that restoration of the purine pool is the most likely explanation.
My second main criticism regarding the manuscript is that the authors describe the different phenotypes that they observe, without exploring at all whether these defects are independent or related one to each other. This is an issue, as we can clearly expect from the literature that some phenotypes are related one to each other. Furthermore, some conclusions are not convincing, some phenotypes are not enough explored and some experiments are clearly lacking.
We agree that the phenotypes may be inter-related but we feel that we made a strong effort to determine which phenotypes result from reduced purine pools (ex. DNA damage/p53/cell cycle arrest) vs those that result from impaired DNPS and intermediate metabolite accumulation (ex. Impaired ciliogenesis). These have been substantiated with rescue experiments using nucleosides, SAICAr administration and small molecule inhibitors of the pathway upstream of ADSL.
Moreover, we attempted to determine how these impact particular aspects of the pathology observed in vivo by complementation experiments with nucleosides or inhibition of DNPS/SAICAR production where feasible. We clearly did not convey this as effectively as intended and to address this we have included a schematic model of the pathways (Figure 1-supplement 1), as well as a summary of phenotypes and interventions (Appendix 1) in the revised manuscript to try to clarify this point further. We hope this addresses the reviewer’s comments but without specific examples, we are not fully confident we understand the issue completely.
3. The authors mention that DNA damage signalling, evaluated after 53BP1 and gH2Ax immunostaining, is mildly affected, but can be suppressed by nucleoside
supplementation (Figure 2). It is difficult to reach this conclusion with the data presented in Figure 2. If this is a sound conclusion for 53BP1, only a very small population of cells (maybe 10% of cells) show differences in gH2Ax staining intensity between control and ADSL siRNA conditions (Figure 2D). One possibility is that DNA damage is repaired and only transiently detected during the cell cycle.
We have 2 issues at play in our opinion. The first is the use of siRNA- this leads to different levels of depletion in individual cells and ADSL was difficult to deplete, requiring 2 subsequent transfections of siRNA. Unfortunately, ADSL deletion is lethal and using a stable shRNA leads to growth defects prior to the experiment, making it difficult to compare to controls. It is likely that cells with the strongest depletion will not exhibit DNA damage due to the fact they exit cell cycle (replication is necessary to elicit replication stress). Likewise, cells with poor depletion will not suffer a significant reduction in purine levels. This would be consistent with the fact that we see a subpopulation of cells that exhibit DNA damage markers. A more continuous detection method would be ideal but is not feasible due to the siRNA depletion protocol needed to efficiently reduce ADSL.
We attempted to perform ADSL immunofluorescence, that would allow a single cell based analysis of depletion levels, but this signal is too weak and diffuse to use as a marker in these experiments to “normalize” the effects in any way. For this reason we provided population level graphing of gH2AX intensity or the presence of 53BP1 foci, as is standard in the field. It is clear that there is wide variability in the levels (which may also reflect cell cycle phase and timing of the damage related to DNA repair, as the reviewer points out, as well as the level of ADSL depletion in that particular cell).
The second issue is that the phenotype is mild. We provided IR treatment as a benchmark, given that this is what is commonly used to elicit damage markers. It is clear that there is a low amount of DNA damage compared to an exogenous insult like IR. That being said, we can reliably detect an increase in the intensity of both damage markers in the siADSL population that is reduced by nucleoside supplementation (notably this is also observed in vivo in zebrafish, further substantiating the phenotype – Figure 7G) and was recently reported following PAICS inhibition in leukemia cells. As cells are repair proficient, as the reviewer points out, this is potentially an underestimate of the phenotype due to repair of lesions.
In the revised manuscript we now provide additional example images in Figure 2, as well as an assessment of RPA levels in chromatin (Figure 2A: proxy measure of ssDNA) that accompanies replication stress to further address this point.
As cell cycle progression is also impaired after ADSL depletion (Figures 1D and 1G), the authors should evaluate at which cell cycle stage high levels of gH2Ax staining can be detected to explore this possibility.
As indicated in the response to reviewer 1, we carried out an additional analysis of the cell cycle using flow cytometry (new Figure 1—figure supplement 2E). In our experience, IF is a more sensitive method for quantitative gH2AX analysis and we repeated key experiments with the addition of Adenosine to address previous comments but were unfortunately not able to optimize coupling gH2AX with cell cycle analysis due to limitations arising from Covid restrictions and termination of fellowships. While this would be interesting, we remain unclear exactly what this would clarify. Similar to what is well documented with HU that depletes nucleotide pools, we expect that the primary issue is in S-phase (consistent with RPA results now presented) but damage can arise there due to fork stalling and breakage, in G2 due to condensation of under-replicated regions, in mitosis due to breakage of under-replicated regions and in G1, in the form of 53BP1 foci that mark those broken regions.
Furthermore, it is difficult to conclude about a rescue by nucleoside addition, as both control and ADSL siRNA conditions show decrease in gH2Ax intensity level. The authors need to explain this point in the manuscript. To strengthen their conclusions about DNA damage signalling, they have the possibility to evaluate the activation of classical signalling pathways that involve the kinases Chk1 and Chk2.
We respectfully disagree with the reviewer on this point. The fact that nucleoside addition reduces basal gH2AX levels in the WT cells is not entirely unexpected, as most cultured cells in atmospheric oxygen culture conditions have some level of replication stress that would be reduced by nucleoside addition. This nucleoside mix is sold commercially to address precisely this phenotype in normal stem cells.
The point we wish to make is that the clear difference between WT and ADSL depletion is now erased, indicating that the defect in ADSL depleted cells was likely due to insufficient nucleotide supply, rather than SAICAR accumulation, and this is consistent with new results using only Adenosine. Restoring nucleosides (likely in excess) erases the difference in damage markers between control and ADSL depleted lines, with the caveat that nucleosides may cause other off target effects.
Given how mild the phenotypes are, we do not expect that we will see a strong CHK1/2 response and have been unable to address this experimentally. We note that we did demonstrate that treatment with ATM inhibitors suppressed the 53BP1 increase, providing evidence that ATM signaling is active to some extent (Figure 2C).
4. From the literature, nucleotide depletion can trigger replication stress, and especially replication fork stalling in S phase as a source of DNA damage. This can trigger G2 block as observed in the chicken spinal cord (Figure 3D), due to DNA damage signalling activation before mitotic entry. Can the authors provide evidences of the likelihood of this scenario? Do they observe replication stress, or any sign of replication fork stalling? One possibility is to use RPA (Replication Protein A) loading as a readout of replication stress.
Replication stress can indeed induce a G2 arrest but this depends largely on the amount of damage being produced. For example, low dose aphidicolin induces clear levels of replication stress without eliciting a potent checkpoint response and under-replicated DNA enters mitosis where it can lead to strand breaks. This can also induce p53 and cause a G1 arrest, consistent with the data in RPE1 cells. As previously described, we now provide analysis of RPA, as well as cell cycle progression (Figure 2A and Figure 1—figure supplement 2E), that indicate that we have mild replication stress that is rescued by nucleosides/Adenosine and a G1 arrest that is p53 dependent.
5. Decreased % of ciliated cells and decrease in cilia length can be observed before mitosis in normal conditions, as cilium undocking from the plasma membrane and resorption occurs in general before mitotic entry. Can the authors exclude the possibility that the "defects" in primary ciliogenesis are not only related to the block in G2 phase that they observe (Figure 3F)? Further, CP110 is normally degraded in G2 phase by an F box protein. Any block in G2 phase would therefore be at the origin of CP110 accumulation on both centrioles. Can the authors explore this possibility?
This is a good point and one that we previously considered. First, we have not observed a block in G2 in RPE-1 cells. Unlike the NPCs in vivo, RPE-1 arrest in G1 in a p53-dependent manner and this is now confirmed by the cell cycle analysis presented in Figure 1—figure supplement 2E. As mentioned in the response to reviewer 1, we examined cilia removal in detail following serum addition post ciliation (new Figure 4—figure supplement 1G). This did not indicate any defects in the cilia removal process that is cell cycle dependent as the reviewer indicated. Additionally, we know that the cells being analyzed have left the cell cycle efficiently by loss of Ki67 staining (Figure 4A), so there is unlikely to be a stalled G2 population among the cells being analyzed for ciliogenesis. We therefore think that the cell cycle arrest observed in cycling RPE-1s does not account for the ciliogenesis defect.
1. Ki67 immunofluorescent staining panels provided in Figure 1H and 1L are not accurately illustrating differences in the percent of Ki67 positive cells between the different conditions.
We have tried to improve these images and indicate clearly what is being scored as positive vs negative cells (keep in mind that the DAPI shows all the cells and one must go back and forth a bit). We simply cannot get enough cells in the field to provide a single image that is mathematically representative of the total scoring in each case but believe the current images are highly representative of the populations observed. We are open to suggestions for better ways to annotate this.
2. 53BP1 immunofluorescent staining, which appears as very small dots, cannot be appreciated from the panels provided in Figure 2C.
We have again tried to improve the 53BP1 example images in what is now Figure 2B and 2D, including some enlarged cells so that examples of foice can be seen clearly. We hope this addresses the reviewer’s concern.
3. The dose used for nucleoside supplementation should be mentioned in the method section.
We agree and apologize for this omission, it has been corrected, thanks for pointing this out.
4. The title of the manuscript contains too many abbreviations and does not give a clear idea of the main findings of the manuscript. I recommend to change it.
We have considered this point, as well as other related comments from reviewers and changed the title to: Pathway specific effects of ADSL deficiency on neurodevelopment
Reviewer #2 (Significance (Required)):
The manuscript by Dutto et al. aims to provide a better understanding of the underlying mechanisms that lead to the microcephaly phenotype that can be observed in some human patients affected by ADSL (Adenylosuccinate lyase) deficiency. As ADSL is an enzyme involved in the de novo purine synthesis, patients suffer from global defects in purine metabolism, but the link with neurological symptoms, and especially microcephaly when present, remains totally unexplored. First, the authors provide a characterization of the impact of ADSL depletion on cell proliferation, fitness, fate and outcome and propose that the consequences are context-dependent. Second, as the accumulation of intermediate metabolites like SAICAr has already been proposed as neurotoxic, they clarify the respective implication of purine depletion and SAICAr accumulation on the different cell phenotypes observed and on neurodevelopment. Beyond the conceptual advance provided by this work for the etiology of microcephaly, and the ADSL deficiency disorder in particular, the manuscript will be of interest for the developmental and cell biology communities regarding the impact of metabolism impairment at the cell and organism level.
I am not personally an expert in cell metabolism, but in development and cell cycle progression.
Reviewer #3 (Evidence, reproducibility and clarity (Required)):
The evidence presented of the ADSL deficiency effect on cell proliferation, through DNA damage increase and p53 activation is highly compelling, as is the effect on ciliogenesis. The interpretation and conclusions however, in our view, raise two major points, that need to be addressed. The good news is that these don't necessary require new experiments.
The intermediate products in the purine biosynthesis pathway are metabolized in the monophosphate form, particularly the ADSL substrates SAICAR (a.k.a. SZMP) and SAMP. To the best of our knowledge the monophosphate forms are the biologically active and physiologically relevant forms. Throughout the manuscript the observed results are ascribed to the riboside forms SAICAr and sAdo – including in the title of the manuscript – which are their degradation products, and the measurements of the monophosphate forms not mentioned and not presented. Given the metabolic profiles were performed using LC-MS this information will be probably available to the authors by re-analyzing their data, and presumably will not require further experiments. This issue needs to be addressed throughout the manuscript, the data presented does not grant ascribing the phenotypes to the riboside forms without establishing the status of the monophosphate forms.
We agree with this point with some caveats. The degradation (riboside) products are more readily detected in ADSLD patient serum and cell lines and specifically SAICAr has been implicated in pathological outcomes experimentally(Stone et al., 1998), however there is no data we know of indicating whether either form is more relevant for any particular phenotype in vertebrates.
We specifically used SAICAr for the cilia experiments and see that it phenocopies ADSL depletion, although we did not test SAICAR in parallel, and thus do not know if there is a substantial difference in their effect. As SAICAr accumulates in human patients and was specifically implicated in neuropathology(Stone et al., 1998) in mammals, we chose to examine it in isolation. We will revise the text to ensure we are not making conclusions that do not take this relevant point into consideration.
We attempted many times to measure the SAICAR/r levels without success, although reduced AMP, GMP and increased S-Ado were readily detected. We revisited this again with the help of the Zikanova lab and again carried out MS on RPE-1, also examining HeLa cells in parallel, where ADSL loss has been shown experimentally to induce SAICAR/r. While we can readily detect the expected modulation of SAICAr (and other metabolites) in HeLa cells, we simply cannot detect SAICAr in RPE1 to a level that is sufficient for meaningful comparisons – it is ~250X lower in RPE1 than HeLa. We have tried several optimization steps but his is a technical limitation we are unfortunately not able to overcome. This data does however give us some pertinent information regarding the fact that the PAICS inhibitor has the expected impacts on its accumulation using HeLa cells. A similar issue in detection was encountered previously in human fibroblasts during the characterization of a PAICS mutation and there is unfortunately not anything further we can do with currently available methods. Given this technical limitation, we have revised the conclusions to ensure we are not making an absolute cause-effect conclusion and accounting for the possibility that specifically depleting ADSL could have other effects not directly related to the accumulation of substrates.
The ADSL activity in converting SAMP into AMP is common to both de novo synthesis of AMP and AMP synthesis through the recycling pathway. This is relevant particularly when discussing the potential effects of sAdo/SAMP accumulation, which may be due to recycling rather than de novo synthesis. This point needs to be clarified throughout the manuscript.
There are many discrepancies in the literature related to the definition of the different pathways. To define this in our opinion, we included a schematic in new Figure 1—figure supplement 1 that breaks down the roles of ADSL in DNPS and the purine nucleotide cycle. Both DNPS and recycling pathways feed into the purine nucleotide cycle as the reviewer alludes to. We have gone through the text and revised our discussion of these pathways regarding the interpretation of the data but would ask for specific passages that the reviewer feels are incorrect if any remain. We have made very few conclusions regarding the potential effects of SAMP/S-Ado accumulation and have not used that to attempt to induce phenotypes, so the observation that this increases upon ADSL depletion is simply a confirmation to us that we are effectively impairing its activity.
Specific points (that require new experiments):
In a number of instances the observed effects are ascribed to changes in the abundance of specific metabolites, however no evidence is provided that such changes occur (as measured by LC-MS).
We agree that it is important to try and confirm the effects of several interventions used in the paper to make key conclusions. We address the individual points below.
These are specifically: reduction of SAICAr/SAICAR levels with MRT00252040 drug (e.g. lines 122-130);
Our data indicates the drug is working specifically as expected in HeLa but due to technical limitations, we cannot determine this clearly for RPE1 (new Figure 1figure supplement 2B-C). We expect that a similar proportional effect is taking place in RPE1 but due to the much lower levels, we cannot make any quantitative statements and unfortunately do not have any additional methods by which to address this. All of the experiments using MRT00252040 have been further corroborated by the addition of SAICAr that leads to the same outcome of ADSL depletion. Given that we cannot definititely address changes in SAICAr levels in RPE-1, we have adjusted the text to reflect other possibilities.
Increase in SAICAr/SAICAR levels upon SAICAr treatment (Figure 4);
We are unclear what this would clarify. Given that we have a clear phenotypic effect, what does the LC-MS measurement of this added in excess inform us about and why would we measure SAICAR when we are adding SAICAr? Given the cost of this reagent and scale we require for MS, we cannot consider this experiment without a clearer rationale.
Increased levels of AMP and GMP upon treatment with nucleosides throughout the manuscript;
As discussed earlier, we are using a well-established intervention and we have experimentally validated that this is sufficient to suppress damage induced by HU (new Figure 2E). We have not seen anyone ever report nucleotide levels in the context of this type of experiment of adding nucleosides. We remain unclear again how quantifying this is essential given the number of phenotypic endpoints we provide supporting the fact that purine pools are restored, including adenosine add back that is now included in the revised manuscript.
Levels of SAICAr/SAICAR upon methotrexate treatment (e.g. Figure S3). These measurements are particularly important in the instances where negative results occur, compounds added in the external medium don't necessarily enter all cells or reach all tissues.
As mentioned, we are not able to meaningfully measure SAICAr levels in RPE1 due to detection limits. We also do not agree that measuring metabolite levels will address whether all cells or tissues have incorporated the drug (they surely have not). MTX has been used for decades and the effects are well documented, including that it significantly impacts AMP and GMP levels in cell culture at similar concentrations to what we used (Hoxhaj et al., 2017). While we cannot determine that MTX is inducing SAICAr due to the much lower levels in RPE-1, there is little doubt this is being incorporated and acting as expected in our opinion. In zebrafish, we are certain it is not being taken up by all cells/tissues, there is again no meaningful way to address this experimentally. However, the results of its addition are clear and reproducible in fish depleted for ADSL.
The authors use a mix of nucleosides to "rescue" ADSL deficiency, we would suggest to add an experimental group with Adenosine only in the same experimental conditions, as Adenosine alone should restore AMP (and even GMP levels). This experiment is not absolutely required, but would be interesting, we would leave it to the consideration of the authors.
We thank the reviewer for this great suggestion which was carried out. Adenosine rescues the same phenotypes of the nucleoside mix, indicating that, while there may be some off target effects, the main effect of these interventions is to restore the purine pool.
Suggestions for text:
We would suggest a schematics of the purine biosynthesis pathway, placing the two reactions catalyzed by ADSL, on Figure 1. It's very helpful, even for specialists.
We agree and have added this as indicated previously in Figure 1—figure supplement 1 of the revised manuscript, as described previously in this document.
There is an ambiguity in the literature concerning some purine pathway intermediates including ADSL substrates. SAICAr – riboside form – is also referred in the literature as SAICAR, and SAICAR – monophosphate form – is also referred to SZMP. We would suggest, to avoid ambiguity" to mention once in the text "SAICAR (monophosphate from a.k.a. SZMP)".
We added this for clarification in the text.
State once in the text the which nucleosides compose the nucleoside mix, and in the Materials and methods detail the composition and concentrations of each nucleoside.
This information was partially included in the methods section, however we agree that the precise details are important. We updated this with additional details in the revised manuscript. We used a commercial mix (Embryomax) that has been used previously in numerous papers for attenuating replication stress. This contains Cytidine (0.73 g/L), Guanosine (0.85 g/L), Uridine (0.73 g/L), Adenosine (0.8 g/L) and Thymidine (0.24 g/L)(Halliwell et al., 2020; Ruiz et al., 2015; Kanu et al., 2016; Parsels et al., 2018).
Line 547 "morpholino targeting" should read "morpholino targeting Adsl".
We have corrected this, thank you.
Statistics and experimental design:
We would suggest consult with a statistician and revise the manuscript accordingly.
Here are the issues we detected.
– t-test not suitable for experiments with n=4, better use a non-parametric test (e.g. metabolic profiling Figure 1C).
We have consulted our statistician and we respectfully disagree with the reviewer on this point. The adequacy of a test depends on the distribution generating the data and not on the number of observations. When a distribution cannot be assumed given the nature of the experiment or normality cannot be tested, a nonparametric distribution should be used. Nevertheless, for Figure 1C and similar experiments which include continuous positive measurements we decided to apply a logarithmic transformation to ensure normality. Revised statistical analysis is presented in the revised manuscript.
– For normalisation of western blots, total protein is more suitable than actin.
We included another representative blot with a corresponding Ponceau panel and included this in new Figure 1B.
– Percentages and n (e.g. Ki67 or p53 positive cells) is ambiguous, we would suggest to mention both how many replicates were performed (which as we understand correspond to the stated n) and how many cells were scored (could be an average per replicate).
We prefer to graph the percentages, as we are combining replicates of independent experiments. We revised legends to indicate clearly the n, as well as the cell numbers where applicable in the revised manuscript.
– Anova and t-student test are suitable for continuous variables, not for percentages (averaging percentages does not make the variable continuous), contingency tables and Fisher's exact test, or statistical tools based on the same principle, are more suitable.
We applied a generalized linear model with a binomial distribution in order to capture the nature of the positive/negative values of the cells in the experiment, and to take into consideration multiple biological replicates. Resulting coefficients, p-values and methods are included in the updated manuscript.
– Whenever multiple comparisons are performed, a Bonferroni correction should be applied. If that was included in the statistical analysis it should be stated.
We thank the referee for this suggestion but do not feel it is necessary to apply a correction in most experiments. Each of our experiments measure the percentage of positive cells under different and independent conditions. Moreover, the hypotheses of interest were decided before generating the data and all associated p-values are reported regardless of their significance. We believe that conceptually this discards the need for adjusting for multiple hypothesis testing.
Also, for those cases where multiple contrasts are extracted from the same linear model and dataset, we have applied the Shaffer correction method as implemented in the "multcomp" R package.
– Where the experiments performed blindly? If so it should be stated. If not we would suggest for future work that treatment/control groups are blinded for the observer scoring the results.
Experiments were scored with the samples blinded in all cases. We will indicate this in the methods section as requested.
Reviewer #3 (Significance (Required)):
The manuscript RC-2021-00642 by Dutto et al. presents a characterization of ADSL deficiency in three different model systems. It provides compelling evidence of ADSL deficiency causing proliferation defects, through p53 activation, and in cilliogenesis. The defects observed particularly in the nervous system provide indications of the etiology of patients affected by ADSL deficiency. Hence this research provides new and relevant information concerning the understanding not only of ADSL deficiency but other genetic diseases linked to defects in the purine biosynthesis pathway. Beyond the medical aspect this article provides rather novel insights onto the impact of metabolism, and in particular biosynthetic pathways, have in developmental processes. There is very little research on this topic on the literature, particularly in the developmental models reported: zebrafish and chick embryo. Overall the paper is of interest both to the field of genetic metabolic diseases and developmental biology.
Gabriel, E., A. Wason, A. Ramani, L.M. Gooi, P. Keller, A. Pozniakovsky, I. Poser, F. Noack, N.S. Telugu, F. Calegari, T. Šarić, J. Hescheler, A.A. Hyman, M. Gottardo, G. Callaini, F.S. Alkuraya, and J. Gopalakrishnan. 2016. CPAP promotes timely cilium disassembly to maintain neural progenitor pool. EMBO J. 35:803–819. doi:10.15252/embj.201593679.
Gupta, G.D., É. Coyaud, J. Gonçalves, B.A. Mojarad, Y. Liu, Q. Wu, L. Gheiratmand, D.
Comartin, J.M. Tkach, S.W.T. Cheung, M. Bashkurov, M. Hasegan, J.D. Knight, Z.-Y. Lin, M. Schueler, F. Hildebrandt, J. Moffat, A.-C. Gingras, B. Raught, and L. Pelletier. 2015. A Dynamic Protein Interaction Landscape of the Human Centrosome-Cilium Interface. Cell. 163:1484–1499.
Halliwell, J.A., T.J.R. Frith, O. Laing, C.J. Price, O.J. Bower, D. Stavish, P.J. Gokhale, Z. Hewitt, S.F. El-Khamisy, I. Barbaric, and P.W. Andrews. 2020. Nucleosides Rescue Replication-Mediated Genome Instability of Human Pluripotent Stem Cells. Stem Cell Rep. 14:1009–1017. doi:10.1016/j.stemcr.2020.04.004.
Hoxhaj, G., J. Hughes-Hallett, R.C. Timson, E. Ilagan, M. Yuan, J.M. Asara, I. BenSahra, and B.D. Manning. 2017. The mTORC1 Signaling Network Senses Changes in Cellular Purine Nucleotide Levels. Cell Rep. 21:1331–1346.
Kanu, N., M.A. Cerone, G. Goh, L.-P. Zalmas, J. Bartkova, M. Dietzen, N. McGranahan,
R. Rogers, E.K. Law, I. Gromova, M. Kschischo, M.I. Walton, O.W. Rossanese, J. Bartek, R.S. Harris, S. Venkatesan, and C. Swanton. 2016. DNA replication stress mediates APOBEC3 family mutagenesis in breast cancer. Genome Biol. 17:185. doi:10.1186/s13059-016-1042-9.
Keller, K.E., Z.M. Doctor, Z.W. Dwyer, and Y.-S. Lee. 2014. SAICAR induces protein kinase activity of PKM2 that is necessary for sustained proliferative signaling of cancer cells. Mol. Cell. 53:700–709. doi:10.1016/j.molcel.2014.02.015.
Keller, K.E., I.S. Tan, and Y.-S. Lee. 2012. SAICAR stimulates pyruvate kinase isoform M2 and promotes cancer cell survival in glucose-limited conditions. Science.
Parsels, L.A., J.D. Parsels, D.M. Tanska, J. Maybaum, T.S. Lawrence, and M.A. Morgan. 2018. The contribution of DNA replication stress marked by highintensity, pan-nuclear γH2AX staining to chemosensitization by CHK1 and WEE1 inhibitors. Cell Cycle. 17:1076–1086. doi:10.1080/15384101.2018.1475827.
Piazza, I., K. Kochanowski, V. Cappelletti, T. Fuhrer, E. Noor, U. Sauer, and P. Picotti. 2018. A Map of Protein-Metabolite Interactions Reveals Principles of Chemical Communication. Cell. 172:358–372.e23. doi:10.1016/j.cell.2017.12.006.
Ruiz, S., A.J. Lopez-Contreras, M. Gabut, R.M. Marion, P. Gutierrez-Martinez, S. Bua,
O. Ramirez, I. Olalde, S. Rodrigo-Perez, H. Li, T. Marques-Bonet, M. Serrano, M.A. Blasco, N.N. Batada, and O. Fernandez-Capetillo. 2015. Limiting replication stress during somatic cell reprogramming reduces genomic instability in induced pluripotent stem cells. Nat. Commun. 6:8036. doi:10.1038/ncomms9036.
Stone, T.W., L.A. Roberts, B.J. Morris, P.A. Jones, H.A. Ogilvy, W.M. Behan, J.A.
Duley, H.A. Simmonds, M.F. Vincent, and G. van den Berghe. 1998.
Succinylpurines induce neuronal damage in the rat brain. Adv. Exp. Med. Biol. 431:185–189. doi:10.1007/978-1-4615-5381-6_36.https://doi.org/10.7554/eLife.70518.sa2
- Ilaria Dutto
- Jens Lüders
- Travis H Stracker
- Melanie Philipp
- Melanie Philipp
- Jens Lüders
- Travis H Stracker
- Olga Souckova
- Václava Škopová
- Marie Zikánová
- Sebastian Pons
- Travis H Stracker
- Jordann A Smak
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank the members of the Lüders, Roig and Stracker labs for input, A Riera for chemistry advice, C Morrison for Centrobin antibody, A Holland for CP110 antibody, B Tsou for p53-deficient RPE-1 cells, D Zafra for assistance, CSO Attolini and the IRB Barcelona Biostatistics and Bioinformatics Core Facility for statistical analysis, C Donow and S Burczyk for excellent help with zebrafish maintenance, and LifeArc for supplying MRT00252040. ID was funded by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 754510, THS, JL, and SP were funded by the Ministry of Science, Innovation and Universities (MCIU; PGC2018-095616-B-I00 to THS, PGC2018-099562-B-I00 to JL, and BFU2017-83562-P to SP), the 2017 SGR 1089 (AGAUR), FEDER, the Centres of Excellence Severo Ochoa award, and the CERCA Programme. THS was supported by the NIH Intramural Research Program, National Cancer Institute, Center for Cancer Research. MP was funded by grants from the Deutsche Forschungsgemeinschaft (DFG PH144/4-1 and PH144/6-1). MZ, OS, and VS were supported by Charles University, program PROGRES Q26/LF1. We would like to thank Biocev, First Faculty of Medicine, Charles University, for the opportunity to use their department’s equipment.
- Anna Akhmanova, Utrecht University, Netherlands
- Fadel Tissir, Université Catholique de Louvain, Belgium
- Preprint posted: November 23, 2020 (view preprint)
- Received: May 26, 2021
- Accepted: December 22, 2021
- Accepted Manuscript published: February 8, 2022 (version 1)
- Accepted Manuscript updated: February 9, 2022 (version 2)
- Version of Record published: February 24, 2022 (version 3)
- Version of Record updated: March 18, 2022 (version 4)
- Version of Record updated: September 16, 2022 (version 5)
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Having its genome makes the mitochondrion a unique and semiautonomous organelle within cells. Mammalian mitochondrial DNA (mtDNA) is a double-stranded closed circular molecule of about 16 kb coding for 37 genes. Mutations, including deletions in the mitochondrial genome, can culminate in different human diseases. Mapping the deletion junctions suggests that the breakpoints are generally seen at hotspots. ‘9 bp deletion’ (8271–8281), seen in the intergenic region of cytochrome c oxidase II/tRNALys, is the most common mitochondrial deletion. While it is associated with several diseases like myopathy, dystonia, and hepatocellular carcinoma, it has also been used as an evolutionary marker. However, the mechanism responsible for its fragility is unclear. In the current study, we show that Endonuclease G, a mitochondrial nuclease responsible for nonspecific cleavage of nuclear DNA during apoptosis, can induce breaks at sequences associated with ‘9 bp deletion’ when it is present on a plasmid or in the mitochondrial genome. Through a series of in vitro and intracellular studies, we show that Endonuclease G binds to G-quadruplex structures formed at the hotspot and induces DNA breaks. Therefore, we uncover a new role for Endonuclease G in generating mtDNA deletions, which depends on the formation of G4 DNA within the mitochondrial genome. In summary, we identify a novel property of Endonuclease G, besides its role in apoptosis and the recently described ‘elimination of paternal mitochondria during fertilisation.
Obesity is generally associated with insulin resistance in liver and muscle and increased risk of developing type 2 diabetes, however there is a population of obese people that remain insulin sensitive. Similarly, recent work suggests that mice fed high carbohydrate diets can become obese without apparent glucose intolerance. To investigate this phenomenon further, we fed mice either a high fat (Hi-F) or high starch (Hi-ST) diet and measured adiposity, glucose tolerance, insulin sensitivity, and tissue lipids compared to control mice fed a standard laboratory chow. Both Hi-ST and Hi-F mice accumulated a similar amount of fat and tissue triglyceride compared to chow-fed mice. However, while Hi-F diet mice developed glucose intolerance as well as liver and muscle insulin resistance (assessed via euglycaemic/hyperinsulinaemic clamp), obese Hi-ST mice maintained glucose tolerance and insulin action similar to lean, chow-fed controls. This preservation of insulin action despite obesity in Hi-ST mice was associated with differences in de novo lipogenesis and levels of C22:0 ceramide in liver and C18:0 ceramide in muscle. This indicates that dietary manipulation can influence insulin action independently of the level of adiposity and that the presence of specific ceramide species correlates with these differences.