Deoxyhypusine synthase deficiency syndrome zebrafish model: aberrant morphology, epileptiform activity, and reduced arborization of inhibitory interneurons

Abstract

DHPS deficiency syndrome is an ultra-rare neurodevelopmental disorder (NDD) which results from biallelic mutations in the gene encoding the enzyme deoxyhypusine synthase (DHPS). DHPS is essential to synthesize hypusine, a rare amino acid formed by post-translational modification of a conserved lysine in eukaryotic initiation factor 5 A (eIF5A). DHPS deficiency syndrome causes epilepsy, cognitive and motor impairments, and mild facial dysmorphology. In mice, a brain-specific genetic deletion of Dhps at birth impairs eIF5A^HYP-dependent mRNA translation. This alters expression of proteins required for neuronal development and function, and phenotypically models features of human DHPS deficiency. We studied the role of DHPS in early brain development using a zebrafish loss-of-function model generated by knockdown of dhps expression with an antisense morpholino oligomer (MO) targeting the exon 2/intron 2 (E2I2) splice site of the dhps pre-mRNA. dhps knockdown embryos exhibited dose-dependent developmental delay and dysmorphology, including microcephaly, axis truncation, and body curvature. In dhps knockdown larvae, electrophysiological analysis showed increased epileptiform activity, and confocal microscopy analysis revealed reduced arborisation of GABAergic neurons. Our findings confirm that hypusination of eIF5A by DHPS is needed for early brain development, and zebrafish with an antisense knockdown of dhps model features of DHPS deficiency syndrome.

Introduction

Post-translational modification is a core cellular strategy to rapidly alter protein activity in response to environmental stimuli. Over 200 post-translational modifications are mediated by enzymes which facilitate diverse reactions including phosphorylation, ubiquitination, glycosylation, palmitoylation, sulfation, methylation, small ubiquitin-like modifier (SUMO)-ylation, and nitrosylation, among others. A unique post-translational modification is hypusination, limited to eukaryotic translation factor 5 A (eIF5A) proteins [1]. Hypusination enables eIF5A activation via enzymatic conversion of a conserved lysine to the novel amino acid hypusine (N^ϵ-4-amino-2-hydroxybutyl(lysine)). This reaction occurs in two steps: (1) lysine residue modification by deoxyhypusine synthase (DHPS), using the polyamine spermidine as a cofactor, to form the intermediate deoxyhypusine; (2) hydroxylation of this residue by deoxyhypusine hydroxylase (DOHH) in an oxygen-dependent reaction [2] to form hypusine.

Hypusinated eIF5A (eIF5A^HYP) has key cellular functions: (1) facilitating initiation, elongation or termination during the translation of cell type-specific transcripts [1, 3,4,5,6]; (2) suppressing ribosomal stalling by stabilizing tRNA-ribosomal P-site interaction, facilitating peptide bond formation for consecutive polyprolines and other tripeptide motifs [7, 8]; and (3) nonsense-mediated decay (NMD) of mRNA transcripts with premature stop codons [9]. eIF5A^HYP is critical to translate long polypeptides [10].

eIF5A, as well as the DHPS and DOHH enzymes required for activation by hypusination, are essential for eukaryotic cell viability and growth. In yeast, eIF5A^HYP controls cell proliferation and is required for polarized cell growth during mating by regulating the translation of polyproline-rich formins [11]. In plants, eIF5A activation by hypusination is essential for growth during development and for environmental stress responses [12].

The DHPS-DOHH-eIF5A pathway relies on an oxygen-sensing mechanism. In yeast, DOHH deficiency impairs hydroxylation of the deoxyhypusine residue in eIF5A, decreasing N-terminal translation of proteins in mitochondrial respiration, oxidative stress response, and protein folding. eIF5A hypusination adapts cellular metabolism to oxygen levels [2].

In mammals, the DHPS-DOHH-eIF5A pathway is implicated in disease pathogenesis and in aging. In the Drosophila brain, hypusinated eIF5A levels decline with age, but can be increased by dietary spermidine, and genetic attenuation of eIF5A^HYP levels induces premature aging (e.g., reduced mitochondrial respiration) [13]. Spermidine supplementation in mice boosts eIF5A hypusination and improves cognitive function [14].

Mice with T-cell-specific deletions of Dohh and Dhps develop severe intestinal inflammatory disease, supporting the role of hypusination in T cell activation and differentiation, long associated with polyamine synthesis [15]. Conversely, mice with a myeloid-specific deletion of Dhps revealed that eIF5A^HYP promotes a pro-inflammatory macrophage M1-like phenotype [16].

Cancer-related signaling pathways regulated by eIF5A^HYP include MYC, p53, and hypoxia-inducible factor 1-alpha (HIF1A). Overexpression of eIF5A is linked to colorectal, gastric, esophageal, lung, breast, ovarian, cervical, bladder, prostate, and hepatocellular cancers. Therefore, small-molecule inhibitors of hypusination that target DHPS or DOHH are potential anti-neoplastics [17].

In humans, mutations in the DHPS-DOHH-eIF5A pathway cause neurodevelopmental disorders. A rare, autosomal dominant disorder caused by heterozygous pathogenic EIF5A variants results in developmental delay, intellectual disability, microcephaly, and facial dysmorphism [18]. Rare, autosomal recessive biallelic pathogenic missense and truncating DOHH variants cause developmental delay, intellectual disability, microcephaly, facial dysmorphism, and epilepsy [19]. DHPS deficiency causes an ultra-rare, autosomal recessive disorder caused by biallelic pathogenic variants that reduce DHPS enzyme activity (~ 18–25% of normal), with features similar to the eIF5A and DOHH deficiency syndromes [20].

Homozygous knockout mouse models of DHPS, DOHH and eIF5A are early embryonic lethal, underscoring the essential role of this pathway in early development [21,22,23]. In mice, conditional knockout mouse models of these genes support their developmental roles. Mice with conditional genetic deletions of Dhps or Eif5a induced by the Emx1-Cre driver (primarily expressed in the cortex and hippocampus from E9.5 onwards) show gross defects in forebrain development, reduced growth, and premature death [24]. Mice with a brain-specific deletion of Dhps initiated at birth (via intraventricular injection of an adeno-associated virus with CMV-driven Cre expression) exhibited spontaneous seizures, impaired growth, and death before 6 weeks of age. Moreover, proteomic analysis of brain tissue using quantitative mass spectrometry revealed that these brain-specific Dhps knockout mice had changes in numerous proteins involved in neuronal growth, function, and secretion [3].

In patients and animal models, the DHPS-DOHH-eIF5A pathway disorders primarily impact the brain. While conditional mouse knockouts provide valuable data on the role of eIF5A^HYP in postnatal brain development and function, zebrafish models can probe the role of hypusination in early brain development. A zebrafish model for eIF5A deficiency revealed mild microcephaly and micrognathia [18, 25]. A zebrafish model for DHPS deficiency focused on pancreas development [26]. Here, we characterize a zebrafish DHPS deficiency model, assessing the impact of reduced dhps expression on early brain development and activity.

Results

Generation of a zebrafish model for DHPS deficiency

Antisense morpholino oligomers (MOs) were designed targeting the AUG start codon of the zebrafish dhps mRNA (dhps AUG MO), and the E2I2 splice site of the zebrafish dhps pre-mRNA (dhps E2I2 MO) (Fig. 1A). Knockdown efficacy of the dhps E2I2 MO was analyzed by reverse transcriptase PCR (RT-PCR), which confirmed aberrant splicing (loss of exon 2) after microinjection of the dhps E2I2 MO, resulting in the appearance of a 172-bp amplicon and reduction of the 375-bp amplicon versus the control morpholino (Ctrl MO) and uninjected wild-type (Wt) larvae (Fig. 1B). Western blot analysis revealed a reduction in dhps protein levels in knockdown embryos generated using the dhps AUG MO and the dhps E212 MO. dhps knockdown embryos also had strongly reduced levels of both eif5a^HYP and eif5a^TOTAL, with a striking reduction of larger polypeptides (Fig. 1C and D).

Fig. 1

Generation of the DHPS deficiency zebrafish model. (A) The first three exons and introns of the dhps gene and selected targeting sites of two antisense morpholino oligomers and PCR primers are shown, as well as the predicted amplicon lengths. (B) RT-PCR analysis confirmed the predicted lengths of the dhps mRNA amplicons from Wt, Ctrl MO and dhps E2I2 MO larvae with two biological repeats. While the length of the PCR amplicon covering the first three exons of dhps mRNA is 375 bp (b) in Wt, Ctrl MO, and dhps E2I2 MO larvae, in the latter two, additional bands appear (a and c). Appearance of the 172-bp amplicon (c), indicating deletion of exon 2 from the dhps mRNA, and reduction in amount of the 375-bp amplicon (b), together confirm partial knockdown of dhps. Appearance of an approximately 450-bp amplicon (a) in dhps E2I2 MO larvae may be due to partial activation of a cryptic splice site induced by antisense blockage of the E2I2 splice site. (C) Analysis of protein expression in dhps knockdown zebrafish. Zebrafish embryos (2 dpf) with knockdown by dhps E2I2 MO or dhps AUG MO were analyzed by Western blot for expression of DHPS, eIF5A^TOTAL, eIF5A^HYP, and total protein (as visualized by REVERT™). (D) Densitometric data for the relative expression of DHPS, eIF5A^TOTAL, and eIF5A^HYP, normalized on the basis of total protein as quantified by REVERT™ staining

Phenotypic analysis of dhps knockdown zebrafish

Phenotypic analysis of zebrafish larvae at 5 days post-fertilization (dpf), microinjected at one-cell stage with exon 2/intron 2 (E2I2) dhps E2I2 MO (7.92 ng) or with dhps AUG MO (11.55 ng), revealed that both antisense MOs caused similar dysmorphologies. dhps knockdown larvae were microcephalic with axis truncation, hyperpigmentation, cardiac edema, uninflated swim bladders, and body curvature, developmental delay and higher death rate at early stages compared to uninjected Wt and Ctrl MO (Fig. 2A). In control groups (Wt, n = 168 and Ctrl MO, n = 148), 96% of larvae exhibited normal development. In the dhps E2I2 MO group (n = 180), 13% of larvae exhibited significant dysmorphology, 38% moderate dysmorphology, and 16% mild dysmorphology. In the dhps AUG MO group (n = 126), 4% of larvae exhibited severe dysmorphology, 7% moderate dysmorphology, and 13% mild dysmorphology (Fig. 2B).

Fig. 2

Phenotypic analysis of dhps knockdown zebrafish at five days post-fertilization (5 dpf). (A) Lateral view photographs of representative Wt, Ctrl MO, dhps E2I2 MO, and dhps AUG MO larvae, scale bar: 0.5 mm. (B) Chi-square analysis revealed significant differences between knockdown and control groups with regard to the percentages of larvae exhibiting normal development, mild and moderate dysmorphology (p<0.0001)

Partial rescue of dysmorphology in dhps knockdown larvae by expression of dhps mRNA

To control for antisense MO specificity, in vitro transcribed wild-type zebrafish dhps mRNA was co-injected with the dhps E2I2 MO to rescue the knockdown phenotype. Co-injection of in vitro transcribed dhps mRNA with dhps E2I2 MO reduced the number of larvae with dysmorphology and developmental delay, increasing in the percentage of normally developed larvae. In the dhps E2I2 MO group (n = 371), only 34% of embryos exhibited normal development, 41% mild dysmorphology, and 17% moderate dysmorphology, while in the group co-injected with dhps E2I2 MO and dhps mRNA (n = 179), 49% of embryos exhibited normal development, 31% mild dysmorphology, and 16% moderate dysmorphology. In control groups (Wt, n = 163 and dhps mRNA, n = 111) more than 94% of embryos exhibited normal development. These results demonstrate a partial rescue of the effects of dhps antisense MO (Fig. 3B, and C).

Fig. 3

Partial rescue of dhps knockdown larvae by expression of dhps mRNA at 1 day post-fertilization (1 dpf). (A) Map of pIVT construct with zebrafish dhps cDNA. (B) mRNA overexpression partially shifted the more severe phenotypic categories of dhps knockdown to less severe categories. Chi-square analysis showed significant differences between control, knockdown and rescue groups with regard to the percentages of embryos exhibiting normal development, mild and moderate dysmorphology (P<0.0001). (C) Definition of observed phenotypic categories and sample size of each group

Electrophysiological analysis of dhps knockdown zebrafish larvae

Local field potential (LFP) recordings from the optic tectum revealed epileptiform events in 4-dpf dhps knockdown larvae. LFP recording of brain activity of 4-dpf dhps knockdown larvae revealed spontaneous electrographic discharges with high amplitude (≥ three-fold baseline) versus controls (Fig. 4A). Quantification of ictal-like events revealed that uninjected Wt larvae (n = 20) had a mean of 0.25 ± 0.14(SEM), Ctrl MO larvae (n = 22) had mean of 0.68 ± 0.24(SEM) and dhps E2I2 MO larvae (n = 23) had a mean of 3.74 ± 1.27(SEM). A significant increase was observed in the dhps knockdown group using the one-way ANOVA test (p ≤ 0.02) (Fig. 4B).

Fig. 4

LFP recording of dhps knockdown larvae. (A) Snapshot of LFP signal recorded from 4-dpf Wt, Ctrl MO and dhps E2I2 MO larvae. (B) One-way ANOVA revealed that the frequency of ictal-like events in dhps knockdown larvae was significantly increased compared to control larvae (p ≤ 0.02). Each recording was performed over a period of 20 min. (C) Each pie chart shows percentage of larvae with more than 3 ictal-like events during 20 min of recording. While in Ctrl MO larvae less than 10% of larvae had more than 3 ictal-like events, in dhps E2I2 MO-injected larvae, more than 43% of larvae had more than 3 ictal-like events

Reduced GABAergic neuronal arborization in dhps knockdown larvae

Confocal microscopy analysis of dhps knockdown larvae derived from a transgenic reporter line with GABAergic-specific expression of mCherry assessed the effects of reduced dhps function on GABAergic neuron development. In zebrafish larvae, GABAergic neurons project arbors from the optic tectum to the tectal neuropil (Fig. 5A). Confocal microscopy revealed significantly reduced arborization of GABAergic neurons in dhps knockdown larvae (Fig. 5B). Quantification of the GABAergic neuronal arborization by Sholl analysis revealed a decreased dendritic arborization (p<0.0001, Table 1) in larvae with reduced dhps expression (Fig. 5C).

Fig. 5

GABAergic neuronal dendritic arborization. (A) Larval head with location of optic tectum (blue) and neuropil (light yellow) where that GABAergic neurons project their arbors. (B) Confocal microscopy revealed that dhps knockdown larvae have less complex dendritic arborization in the GABAergic neurons. (C) GABAergic neuronal arbors were quantified in zebrafish larvae at three different ages (3, 5, and 7 dpf). At all three developmental ages, via one way ANOVA there was a significant reduction in the number of neuronal arbors in the dhps knockdown larvae (dhps E2I2 MO) versus both wild-type control larvae (Wt) and control MO-injected larvae (Crtl MO) (see also Table 1 below)

Table 1 Quantification of the number of neuronal arbors in wild-type control larvae (wt), control MO-injected larvae (Cntl MO), and dhps knockdown larvae (dhps E2I2 MO) at 3, 5 and 7 dpf

Discussion

Our zebrafish loss-of-function model of DHPS deficiency by antisense knockdown of dhps expression mimicked features of the human DHPS deficiency syndrome. The zebrafish showed developmental delayed with epileptiform discharges, as well as microcephaly, axis truncation, and body curvature. Electrophysiological analysis of dhps knockdown larvae showed increased epileptiform activity, while confocal microscopy analysis revealed significantly reduced arborisation and complexity in GABAergic neurons.

The epileptiform discharges in dhps knockdown larvae paralleled the epileptiform activity on electroencephalography (EEG) and seizures in DHPS deficiency syndrome patients [20], and in mice with a brain-specific knockout of Dhps induced by postnatal intraventricular injection of a Cre-expressing AAV vector [3]. Seizures were not reported in patients with loss-of-function EIF5A mutations [18], nor in an eIF5A-deficient zebrafish model [18, 25]. Patients with DOHH mutations have seizures [19]. Our findings suggest that for the seizure phenotype, the brain is more sensitive to reduced eIF5A hypusination caused by mutations in DHPS or DOHH, while patients with neurodevelopmental disorders caused by autosomal dominant EIF5A mutations do not exhibit seizures as remaining eIF5A^HYP levels may be sufficient to prevent epileptiform activity and seizures.

Confocal microscopy analysis of dhps knockdown zebrafish using a transgenic reporter line revealed significantly reduced GABAergic neuron aborization. Loss of inhibitory GABAergic inputs may contribute to neuronal hyperexcitability, but mouse model studies are needed to confirm these findings. Moreover, cell type analyses in knockdown zebrafish brains could determine if this phenotype results from overall neuronal loss or a selective loss of GABAergic neurons. Brain magnetic resonance imaging scans of DHPS deficient patients are normal [OD et al., unreported data], suggesting that the brain morphology phenotypes in our zebrafish and other mouse models [3, 24] are more pronounced than the human phenotype.

Since hypusination of eIF5A is a key regulator of autophagy [14, 27], and autophagy is critical during axonal and presynaptic development [28], impaired autophagy during zebrafish and mouse brain development in the DHPS deficiency syndrome models should be explored as a pathogenic mechanism. Our analysis of protein expression in dhps knockdown zebrafish reveal a striking reduction in expression of larger proteins. Ongoing studies using proteomic and transcriptomic analysis of embryonic brains with reduced hypusination beginning during early development may identify key molecular and cellular changes at different developmental stages and may elucidate the role of eIF5A^HYP in the translation of certain proteins (e.g., long polypeptides [10] and neurosecretory factors [5]). We need to better understand the reductions in total eIF5A and eIF5A^HYP levels seen in dhps knockdown zebrafish, which are similar to those in conditional cell-specific mouse models of DHPS loss [4, 5].

Overexpression of dhps mRNA in wild-type embryos did not cause a deleterious phenotype, suggesting that gene therapy strategies that involve overexpression of DHPS may show benefit for DHPS deficiency syndrome patients. Conversely, only partial rescue of the dhps knockdown phenotype was achieved by overexpression of dhps mRNA, which may be due to the limited half-life of dhps mRNA after microinjection at the single-cell stage.

Our overall findings support that hypusination of eIF5A is important for early brain development, and zebrafish with reduced dhps expression are a useful model for DHPS deficiency syndrome. Future experiments with this model will evaluate anti-seizure medications using the seizure phenotype, as well as other therapeutic modalities and endpoints. Generation of zebrafish models of neurodevelopmental disorders caused by mutations in the DHPS-DOHH-eIF5A pathway that incorporate specific patient mutations into the genetically engineered lines could dissect the phenotypic differences observed between these related diseases.

Methods

Zebrafish husbandry

Wild-type (Wt) adult zebrafish (Danio rerio; AB strain; CVE-KIT) were maintained at 28.5 °C on a 14-h/10-h light/dark cycle under standard aquaculture conditions, and fertilized eggs were collected via natural spawning. Embryos were raised in embryo medium (E3; 1.5 mmol/L HEPES, pH 7.6, 17.4 mmol/L NaCl, 0.21 mmol/L KCl, 0.12 mmol/L MgSO4, and 0.18 mmol/L Ca[NO3]2), under the same conditions as adults. For all zebrafish experiments conducted at NCMM, larvae up to 7 days post-fertilization (dpf) were used.

Antisense morpholino oligomers (MOs) and microinjections

11.55 ng of a translation-blocking MO (dhps AUG MO: 5’ GGTTATGGATGTAAATCCGGCTTTT) targeting the AUG start codon of the zebrafish dhps mRNA and 7.92 ng of splice site-blocking MO targeting the exon 2/intron 2 splice site of the zebrafish dhps pre-RNA (dhps E2I2 MO: 5′ CACGATCAGTCTGTCACTCACCATC) were used to achieve partial knockdown of zebrafish dhps. Fluoresceinated standard control MO was used as a negative control (Ctrl MO) (11.55 and 7.92 ng respectively). MOs were designed and synthesized by Gene Tools LLC (Philomath, Oregon, USA) and injected into 1-2-cell stage embryos [26].

Reverse transcriptase PCR (RT-PCR)

Efficiency of knockdown was determined by RT-PCR, using primers that amplify across the predicted deletion: 5′ GCGCTGTGAAATGTGAGTGAAAC and 5′ GTTTGACGTGTAGCCCAGGAAT. The PCR amplicon was 385 bp in the control embryos and 172 bp in the dhps MO-injected embryos, and was visualized by standard agarose gel electrophoresis [26].

Western blot analysis

Zebrafish embryos were evaluated by Western blot analysis, adapting methods previously described [3]. Specifically, 30–40 zebrafish embryos (2 dpf) were lysed in 400 µL of buffer containing 50 mM Tris, pH 8.0, 150 mM NaCl, 0.05% deoxycholate, 0.1% IGEPAL CA-630, 0.1% SDS, 0.2% sarcosyl, 10% glycerol, 1 mM DTT, 1 mM EDTA, 10mM NaF, protease inhibitors (#11836170001, Roche), phosphatase inhibitors (#4906845001, Roche), 2 mM MgCl2, and 0.05% v/v Benzonase (Millipore) and were intermittently vortexed to facilitate protein extraction. Protein was quantified using the DC Protein Assay Kit II (#5000112, Bio-Rad) followed by SDS-PAGE (4–20% gel). Separated protein (20 µg) was transferred to PVDF membranes and blocked in Odyssey Blocking Buffer (#927-40100, LI-COR Biosciences) at room temperature for 1 h. Membranes were incubated with REVERT (#926-11016; LI-COR Biosciences) to permit visualization of total protein. Subsequent incubation with primary antibodies diluted in Intercept Blocking Buffer (#927-70001; LI-COR Biosciences) was performed overnight at 4 °C. Membranes were washed twice with TBST buffer prior to incubation with near infrared, fluorescent dye-conjugated secondary antibodies at room temperature for 1 h. Following additional washes with TBST buffer, the membranes were imaged using an Odyssey CLx Imaging System and images were analyzed using the CLx Image Studio Version 5.2 Software (LI-COR Biosciences).

The following primary antibodies were used at the dilutions indicated: mouse anti-deoxyhypusine synthase (1:2000; Santa Cruz, #sc-365077), mouse anti-eIF5A^TOTAL (1:2000; BD Biosciences, #611977), and rabbit anti-eIF5A^HYP (1:5000; Millipore, #ABS1064-I). Densitometric data are graphed as relative expression.

mRNA rescue

To generate dhps RNA, zebrafish dhps cDNA was cloned into the pIVT expression construct [Addgene plasmid 122139; 29], which was linearized through restriction at the 3’ end of the ORF and used as a template to generate dhps mRNA using T7 RNA polymerase in an in vitro transcription reaction.

Local field potential (LFP) recording

Recordings were obtained from tecta at 4 dpf [30]. Seizure detection was performed through visual inspection and automated using a custom-written R script to minimize bias and artifacts due to muscle contractions. Recorded frequencies were divided 1-100, 100–250, and 250–500 Hz bands. If amplitude exceeded 3× background, the event was considered a seizure based on high-frequency oscillations (> 100 Hz) [31]. Power spectrum was analyzed using Clampfit 10.7 software (Molecular Devices). 20-minute-long recordings were used to compute the power spectrum from larvae at 4 dpf, and each condition was averaged per group [32].

Confocal microscopy and quantification of arbors

To visualize GABAergic neurons, wild-type and dhps E2I2 MO-microinjected embryos were treated from 1 dpf with 0.003% phenylthiourea to prevent pigmentation. Larvae were anesthetized in 0.001% tricaine (Sigma), fixed for 3 h at room temperature with 4% paraformaldehyde, mounted on glass slides, and imaged using confocal microscopy. A dorsal z-stack of the optic tectum was collected using a 40x lens and a z-resolution of 0.44 μm. For Sholl analysis [33, 34], images were filtered using the 3D-Median filter in ImageJ. A z-projection of the tectum was generated, and the resulting image was converted to a thresholded binary image. Arborization extent was quantified using Sholl analysis (plug-in; http://imagej.net/Sholl_Analysis). The number of intersections was normalized against the number of neurons in the imaged area; this value was statistically analyzed [32].