Vitamin B6 (VB6)-dependent epilepsy was first reported in 1954; however, the underlying genetic cause(s) was unknown until 2005 when the first mutation was identified in an autosomal recessive gene pyridox(am)ine 5’-phosphate oxidase (PNPO). Gene PNPO encodes a rate-limiting enzyme in the synthesis of pyridoxal 5’-phosphate (PLP), the biologically active form of VB6 and a co-factor required for the synthesis of several neurotransmitters such as dopamine, serotonin, and GABA. Since 2005, mutations in PNPO have been increasingly reported, mostly in neonatal epileptic encephalopathy (NEE) patients but also in early-onset epilepsy. Moreover, PNPO has been recently included as one of the sixteen epilepsy genes involved in the common epilepsies, suggesting that PNPO deficiency may contribute to epilepsy in general. In contrast to the increasingly recognized significance of PNPO deficiency in epilepsy, our understanding of the neurobiological mechanisms of PNPO deficiency is limited. Moreover, PNPO gene, PNPO mutations, and PNPO deficiency have never been systematically studied due to the lack of animal models. Based on a nutritional conditional lethal phenotype, I have previously identified a Drosophila PNPO gene (sugarlethal, sgll) and a hypomorphic sgll allele (sgll95). This identification opened up opportunities of using Drosophila as a model system to study PNPO deficiency and to functionally characterize PNPO mutations found in patients. In the first part of my thesis, I report the establishment of Drosophila models of PNPO-deficiency-induced epilepsy. Using both behavioral and electrophysiological approaches, we for the first time show that PNPO deficiency leads to seizures in an animal model as it does in humans. Moreover, seizures and the conditional lethality are correlated with low internal PLP levels and can be rescued by ubiquitous expression of wild-type (WT) human PNPO (hPNPO), demonstrating that human and Drosophila PNPOs are functionally conserved. Furthermore, examination of the spike patterns from electrophysiological recordings reveals a potential involvement of GABA dysfunction in seizures caused by PNPO deficiency. Lastly, cell type-specific sgll knock-down indicates an important role of brain-expressed sgll in survival and in maintaining normal brain functions. In the second part, I took advantage of the functional conservation of human and Drosophila PNPOs to generate hPNPO knock-in (KI) lines using CRISPR/Cas9. In each KI line, the endogenous sgll gene was replaced by one of three disease-causing hPNPO cDNAs. Based on in vitro studies, these three mutant alleles show different severity of impaired PNPO activity. A WT hPNPO KI line was generated as a control. Establishing these KI lines has allowed me to discover that severe hPNPO deficiency leads to lethality in early development, intermediate hPNPO deficiency results in conditional lethality and seizures, whereas mild hPNPO deficiency shortens lifespan. At the molecular and cellular level, in addition to the reported impairment in catalytic activity, different hPNPO mutations reduce the stability of hPNPO protein, change its cellular localization, or affect hPNPO transcription. Lastly, I report a dominant-negative effect of one mutant hPNPO on WT hPNPO protein. Based on these fly studies and previously published structural studies by other groups, I predict that a number of seizure-causing hPNPO mutations will have dominant-negative effects and therefore, individuals who are heterozygous for these mutations may be susceptible to diseases. The third part of my thesis is devoted to human studies. I tested whether mild PNPO deficiency confers susceptibility to epilepsy in adult individuals as suggested by our fly studies. Specifically, I examined whether mild PNPO deficiency is over-represented in adult patients with generalized epilepsy. We collected saliva samples from these patients, extracted genomic DNA, and amplified the coding exons using PCR. By sequencing the PCR products, I collected the information of all single nucleotide polymorphisms (SNPs) in them but mainly focused on R116Q, a mutation/variant that can cause NEE and early-onset epilepsy, and is common in the general population. So far, a total of 36 samples have been collected. Sequencing results show that there are three variants in these samples: two of them are synonymous and the third one is R116Q. Among these 36 samples, six are heterozygous for R116Q, so the allele frequency is 8.33 %, which is higher than that in any races in the general population. Power analysis based on the current effect size show that a total of ∼ 300 samples will be needed to reach a power level of 0.8. Genes and mutations that cause human genetic diseases have been increasingly identified in patients since the introduction of whole genome sequencing and whole exome sequencing. Valid animal models are indispensable to functionally examining identified genes and mutations. My thesis work represents a concrete example of using Drosophila models to study the fundamental biology of disease-causing genes and to uncover the molecular and cellular mechanisms of disease-causing mutations identified in patients. These studies expand our understanding of disease-causing genes and mutations, and may also lead to better treatments in the future.