Psychiatric risk gene NT5C2 regulates protein translation in human neural progenitor cells

Genome-wide significant variants associated with combined risk for major psychiatric disorders on chromosome 10q24 affect the expression of the cytosolic 5’-nucleotidase II (NT5C2, cN-II) in population controls, implicating it as a psychiatric susceptibility gene. Risk alleles are associated with reduced expression of this gene in the developing and adult brain, but the resulting neurobiological risk mechanisms remain elusive. In this study, we provide further evidence for the association of NT5C2 with psychiatric disorders, and use a functional genetics approach to gain a deeper understanding of the function of this risk gene in the nervous system. NT5C2 expression was significantly reduced in the post-mortem brain of schizophrenia and bipolar disorder patients, and its protein predominately expressed in neurons within the adult brain. Using human neural progenitor cells (hNPCs), we found that NT5C2 expression peaked at the neural progenitor state, where the encoded protein was ubiquitously distributed through the cell. NT5C2 knockdown in hNPCs elicited transcriptomic changes associated with protein translation, that were accompanied by regulation of adenosine monophosphate-activated protein kinase (AMPK) signalling and ribosomal protein S6 (rpS6) activity. To identify the effect of reduced neuronal NT5C2 expression at a systems level, we knockdown its homologue, CG32549, in Drosophila melanogaster CNS. This elicited impaired climbing behaviour in the model organism. Collectively, our data implicate NT5C2 expression in risk for psychiatric disorders and in Drosophila melanogaster motility, and further suggest that risk is mediated via regulation of AMPK signalling and protein translation during early neurodevelopment.


62
Genetic variants on chromosome 10q24 are associated with combined risk for 63 schizophrenia, bipolar disorder, major depression, autism, and attention deficit hyperactivity 64 disorder 1 , and constitute the third top association signal in the latest schizophrenia genome-65 wide association study (GWAS) 2 . Previous work from our group has demonstrated that the 66 psychiatric risk variants at this locus exert cis-regulatory effects on the cytosolic 5'-67 nucleotidase II gene (NT5C2, in population controls, reducing expression of this gene 68 in the adult and developing brain 3 . However, the neurobiological mechanisms through which 69 genetic variation at the NT5C2 locus may increase risk for psychiatric conditions remain 70 elusive. Interestingly, NT5C2 has also been implicated in a myriad of other medical 71 conditions, including intellectual disability 4 , Parkinson's disease 5 , spastic paraplegia 6 , 72 cardiovascular disease 7, 8 , and acute lymphoblastic leukaemia 9 , which suggests that NT5C2 73 may regulate fundamental aspects of cell biology that are ultimately dysregulated in these 74 conditions.

76
The NT5C2 gene produces an enzyme that cleaves inorganic phosphate from purine and 77 purine-derived nucleotides such as adenosine, inosine and guanosine monophosphate 78 (AMP, IMP, and GMP, respectively) 10 , or that catalyses the transfer of phosphate groups 79 between purine nucleotides and nucleosides 11 . Purinergic compounds have been shown to 80 regulate cell cycle progression and to act as neurotransmitters, and neurotrophic or 81 neuroprotective agents 12, 13 , which is unsurprising given their role in fundamental metabolic 82 processes such as DNA replication, gene transcription and protein synthesis 14,15 . 83 Interestingly, a purinergic hypothesis of schizophrenia has been proposed, which explains 84 neurodevelopmental and neurochemical aspects of this disorder 16 . One plausible 85 mechanism via which NT5C2 may regulate such pivotal processes is via adenosine 86 monophosphate-activated protein kinase (AMPK) signalling, which is a major regulator of 87 5 cellular energy homeostasis [17][18][19] . It remains unclear, however, whether this mechanism 88 occurs in the context of the nervous system, or the cell types involved.

90
The present work provides the most in-depth characterisation of the distribution, expression 91 and function of NT5C2 in the brain and in cultures of human neural progenitor cells (hNPCs). 92 First, to confirm that NT5C2 is a psychiatric risk gene, we explore its expression in RNA 93 sequencing (RNA-seq) data from the brain of schizophrenia, major depression and bipolar 94 disorder patients. This analysis revealed that NT5C2 expression is significantly reduced in 95 schizophrenia and bipolar disorder patients relative to unaffected controls. We also used 96 post-mortem human brain tissue to identify the major cell types expressing NT5C2 in the 97 adult brain, revealing this protein is more expressed in neurons relative to glial cells. To gain 98 insight into the molecular mechanisms that this gene influences, we investigated the 99 expression, sub-cellular distribution and function of NT5C2 in hNPCs. This revealed that the 100 gene is highly expressed during neurodevelopment and that the NT5C2 protein is 101 ubiquitously distributed in the soma and cellular processes of neural progenitors. Using an 102 RNA interference-mediated approach (RNAi), reduced expression of NT5C2 in hNPCs 103 elicited transcriptomic changes associated with protein translation regulation, which were 104 accompanied by differential regulation of AMPK signalling and ribosomal protein S6 (rpS6) 105 activity. Finally, to elucidate the impact of reduced NT5C2 expression at a systems level, we 106 utilised a Drosophila melanogaster (D. melanogaster) model in which the fly homologue of 107 NT5C2, CG32549, was knocked-down either ubiquitously or specifically within the nervous 108 system. Collectively, our work describes the hitherto unknown pattern of NT5C2 distribution 109 and expression in the adult brain and hNPCs, implicating it in the regulation of AMPK    125 To study expression of NT5C2 in the brain of psychiatric patients, we analysed RNA-seq 126 data from the Stanley Neuropathology Consortium 20 . This cohort consisted of a collection of 127 matched hippocampus samples from subjects diagnosed with bipolar disorder, 128 schizophrenia, or major depression, and unaffected controls (n = 15 each). Reads were 129 trimmed using Trimmomatic 0.36 21 , and mapped to Ensembl genes (build 38, v93) using 130 kallisto 22 . Differential expression was calculated using Wald tests in DESeq2 23 whilst 131 controlling for the effect of demographics, and corrected using a false discovery rate (FDR) 132 cut-off of 10% (q < .10). Raw and normalised counts, and factors and covariates considered 133 in the analysis are available in Supplementary Tables 1-3. Raw RNA-seq data is available 134 upon request via http://sncid.stanleyresearch.org.

136
Immunohisto-and cytochemistry 137 We used immunolabelling to quantify the NT5C2 knockdown in hNPCs and identify its sub-138 cellular distribution in hNPCs and brain tissue. Brain sections were deparaffinised and 139 7 submitted to antigen retrieval and autofluorescence removal protocols (please see   156 To identify gene and protein expression and phosphorylation differences associated with 157 NT5C2 function, we isolated total RNA or protein from in vitro cultures using TRI Reagent and total rpS6 (54d2) and phospho-rpS6 (Ser235/Ser236) (Cell Signalling,Danvers,165 Massachusetts, United States).

167
Confocal microscopy 168 We imaged fluorescently labelled cultures or brain sections using confocal microscopy to 169 identify the distribution of NT5C2 in the adult brain, and the sub-cellular distribution of this 170 protein in hNPCs. Imaging was performed at the Wohl Cellular Imaging Centre, King's 171 College London, using a Nikon A1R (Nikon, Amsterdam, Netherlands) or a Leica SP5 172 Confocal Microscope (Leica, Wetzlar, Germany). Images were taken as z-stacks of 8-10 173 plans, and exported to Fiji, where background subtracted images and maximum intensity 174 projections were generated. To identify cell-type expression of NT5C2 in the brain, co-175 localisation was defined as percentage of co-localised clusters relative to total number of 176 clusters detected per image. High throughput analysis was performed using an ImageJ  To identify differences between more than two independent groups, we used one-way 208 ANOVAs followed by Tukey post hoc tests if values were normally distributed (e.g. co-209 localisation between NT5C2 and different markers, or the effect of two independent siRNAs 210 in hNPCs relative to control cultures); or Kruskal-Wallis tests followed by Dunn's post hoc 211 tests, if values were not normally distributed (e.g.: the effect of the knockdowns on total and 212 phosphorylated levels of AMPK and rpS6). To compare differences between two groups, we 213 performed t-tests if values were normally distributed (e.g. expression differences between 214 hNPCs and cultures, the effect of the knockdown or overexpression in hNPCs or HEK293T); 215 or Mann-Whitney tests if they were not normally distributed (e.g. survival and climbing 216 success ratios associated with CG32549 knockdowns in Drosophila); correction for multiple 217 testing was performed using the Bonferroni method. The Fisher's exact test to calculate 218 significance of the gene overlaps was performed in R using the package 'GeneOverlap'.

219
Statistical analyses were performed in R or in IBM SPSS.

222
NT5C2 expression is reduced in the brain of schizophrenia and bipolar disorder 223 patients 224 We previously demonstrated that NT5C2 expression is reduced in the brain of unaffected 225 controls due to cis-regulatory effects associated with psychiatric risk alleles located on  Tables 7-12). This analysis revealed that NT5C2 was less 232 expressed in SCZ (P < .001, false discovery rate (FDR) corrected P = .01, fold-change = 233 0.56) and BD patients (P < .001, corrected P = .02, fold-change = 0.69), but not in MDD 234 patients (P > .05) ( Figure 1A). These findings corroborate a role for NT5C2 in mediating 235 susceptibility to psychiatric disorders, particularly those associated with psychotic features.

238
To understand how genetic variation affecting NT5C2 expression may confer risk for 239 psychiatric disorders, we investigated which neuronal cell types expressed this gene. First, 240 we examined single-cell RNA-seq data from the mouse cortex 30 to predict which cell type(s) 241 highly expressed NT5C2. This analysis revealed a significant difference in the cell-type 242 specific expression of NT5C2 within neuronal and non-neuronal cells (One-way ANOVA, F 243 (3, 1008) = 11.11, P < .001). Post-hoc analysis confirmed that NT5C2 is more abundant in To investigate whether this distribution pattern also occurs in humans, we performed a 248 series of immunocolocalisation experiments using autopsy brain tissue and confocal 249 microscopy. We confirmed specificity of an antibody raised against NT5C2 by using it to .001, n = 4 control subjects). Post-hoc analysis confirmed that the mean percentage co-267 localisation values significantly differed between neuronal and non-neuronal markers 268 (MAP2: 7.48% ± 2.02 (standard deviation); PARVALB: 6.89% ± 2.09; GFAP: 3.13% ± 1.09; 269 IBA1: 1.44% ± 0.93; Tukey post hoc tests: P < .001 for all comparisons; Figure 1F), but not 270 within these categories (i.e., GFAP vs. IBA1, P > .05). These data corroborate the single-271 cell RNA-seq data from the mouse brain (Supplementary Figure 1), and confirm our 272 qualitative observations using immunoperoxidase staining (Supplementary Figure 3). 273 Overall, these data suggest that whilst NT5C2 expression at the message and protein level 274 is found in both neurons and glial cells, it is clearly enriched in neurons relative to glia. This Next, we assessed the sub-cellular distribution of NT5C2 in hNPCs derived from human 296 induced pluripotent stem cell (hiPSC) and CTX0E16 cells. First, we ectopically expressed a 297 myc-tagged NT5C2 construct in hiPSC-NPCs. This revealed that ectopic myc-NT5C2 was 298 abundantly expressed in the cell soma and was present in punctate structures along neurites 299 ( Figure 2C). We further confirmed the ability of our antibody raised against NT5C2 to detect 300 myc-NT5C2 (Supplementary Figure 4). Next, we examined the distribution of endogenous 301 NT5C2 in hNPCs derived from hiPSCs ( Figure 2D) and from the CTX0E16 cell line ( Figure   302 2E). Similar to the distribution of ectopic protein, endogenous NT5C2 was found dispersed 303 throughout the cell, and was present in punctate structures within the cell soma and along 304 neurites; virtually all imaged cells expressed NT5C2. Taken together, these data suggest 305 that NT5C2 is highly expressed and ubiquitously distributed in hNPCs.    To support the association between NT5C2, AMPK signalling and rpS6 activity, we carried 397 out complementary experiments in HEK293T cells using a pNT5C2-myc overexpression 398 plasmid. Exogenous expression of NT5C2 in HEK293T cells resulted in a mean ~64% 399 decrease in phosphorylated control: 223.00 ± 76.99,400 overexpression: 81.05 ± 30.14, t(15) = 4.88, P < .001, Bonferroni corrected P < .001; Figure   401 4C), whilst no difference in total AMPK levels were observed (t(12) = 1.16, P > .05; Figure   402 4C). These data are consistent with our data indicating that NT5C2 is a negative regulator 403 of AMPK signalling. Next, we examined levels of total and phosphorylated rpS6 in the 404 presence or absence of ectopic NT5C2. This revealed a mean ~28% decrease in total rpS6 405 abundance (control: 159.10 ± 48.52, overexpression: 108.8 ± 48.52, t(16) = 2.88, P = .011, 406 corrected P = .044 ; Figure 4C), and a mean 300% increase in rpS6 phosphorylation 407 (control: 31.03 ± 10.66, overexpression: 124.10 ± 8.20, t(16) = 20.76, P < .001, corrected P 408 < .001; Figure 4C). This effect of exogenous NT5C2 on rpS6 phosphorylation was opposite 409 to that observed in hNPCs, corroborating the complex nature of the intracellular cascades Our NT5C2 knockdown studies in hNPCs, together with the developmental profile of the 419 gene in human brain (Figure 2A) support an important role for NT5C2 in early brain 420 development. However, these molecular studies do not afford an insight into the potential 421 impact of reduced developmental expression of NT5C2 at a systems level. Interestingly, the 422 NT5C2 protein shares 60.5% sequence identity and 80.2% sequence similarity with the D. 423 melanogaster homologue, CG32549 (Supplementary Figure 7), suggesting that they likely 424 exert the same or similar function. Thus, we reasoned that it would be possible to gain an 425 insight into the functional impact of reduced NT5C2 expression in vivo, by modelling the 426 knockdown of CG32549 on a complex and polygenic behaviour, such as climbing. This is a 427 polygenic psychomotor trait driven by an interaction between cognitive function and physical 428 activity, which is easily observable and measurable in D. melanogaster, and that has been hypothesise that this mechanism may also occur in the adult brain, although it may be 472 susceptible to additional regulatory mechanisms. We further hypothesise that the psychiatric 473 risk mechanism pertaining NT5C2 expression in the adult brain is more likely to occur in 474 neurons, where this protein is more expressed relative to glial cells (Figure 1). This is 475 consistent with the recent suggestion that expression of psychiatric risk genes is cell-type 476 20 specific, with particular enrichment of neuronal cell type 32 . Ultimately, the association 477 between NT5C2 function and protein translation regulation corroborates the idea that this 478 gene serves a fundamental role in cell biology, as it is implicated in a multitude of disease 479 states 1, 2, 4-9 .

481
The transcriptomic analysis of the knockdown in hNPCs also revealed that NT5C2 may 482 govern genes that regulate the cytoskeleton (Figure 4) considering the heterogeneity inherent to all psychiatric conditions. We partly addressed this 503 issue by controlling for gene expression changes associated with the effect of 504 demographics; this analysis could be further improved by significantly increasing sample 505 size. Second, we obtained a modest knockdown in the loss-of-function experiment in 506 hNPCs, which is likely due to the proliferative nature of these cells. We partly addressed this 507 issue by testing hypotheses generated based on the microarray results using western 508 blotting and a different cell type (HEK293T cells; Figure 4), which supported our findings.  Genome-wide association study identifies eight loci associated with blood pressure.