An FEVR-associated mutation in ZNF408 alters the expression of genes involved in the development of vasculature

Abstract

Familial exudative vitreoretinopathy (FEVR) is an inherited retinal disorder hallmarked by an abnormal development of retinal vasculature. A missense mutation in ZNF408 (p.H455Y) was reported to underlie autosomal dominant FEVR in a large Dutch family, and ZNF408 was shown to play a role in the development of vasculature. Nonetheless, little is known about the molecular mechanism of ZNF408-associated FEVR. To investigate this, an in vitro model of ZNF408-associated FEVR was generated by overexpressing wild-type and p.H455Y ZNF408 in human umbilical vein endothelial cells. Cells overexpressing mutant ZNF408 were unable to form a capillary-like network in an in vitro tube formation assay, thereby mimicking the clinical feature observed in patients with FEVR. Intriguingly, transcriptome analysis revealed that genes involved in the development of vasculature were deregulated by the p.H455Y mutation. Chromatin immunoprecipitation showed that p.H455Y ZNF408 has reduced DNA-binding ability, as compared to the wild-type protein. The fact that the p.H455Y mutation disrupts the expression of genes important for the development of vasculature sheds further light on the molecular mechanisms underlying ZNF408-associated FEVR.

Introduction

Familial exudative vitreoretinopathy (FEVR, MIM 133780) is an inherited retinal disorder that was first described by Criswick and Schepens (1). It is primarily characterized by abnormal and incomplete vascularization of the peripheral retina (2–4). Thus far, mutations in NDP (5), FZD4 (6,7), TSPAN12 (8), LRP5 (6,9), ZNF408 (10), ATOH7 (11), RCBTB1 (12) and CTNNB1 (13,14) have been reported in FEVR cases. Besides those of ZNF408, ATOH7 and RCBTB1, the proteins encoded by FEVR genes are involved in the Norrin/β-catenin signaling pathway (15). Next to the genetic heterogeneity, various clinical manifestations can be observed in FEVR, ranging from asymptomatic to complete blindness (16).

We previously discovered a mutation in ZNF408 (c.1363C>T, p.H455Y, NM_024741.2) that segregates in a large autosomal dominant FEVR family (10). The protein encoded by ZNF408 belongs to the PRDM (positive regulatory domain I-binding factor 1/PRD1-BF1 and retinoblastoma-interacting zinc finger protein 1/RIZ1 homology domain containing) family (17). Members of this family typically have an N-terminal PR domain and multiple zinc finger domains. The PR domain belongs to a subfamily of SET methyltransferase domains, whereas zinc finger domains are generally involved in DNA binding and protein–protein interactions (17,18). ZNF408 is predicted to harbor 10 zinc finger domains, each of which consists of two histidine and two cysteine residues. The p.H455Y mutation changes a histidine in the fourth zinc finger domain into a tyrosine (10). We demonstrated that the mutant protein is mislocalized in the cell and knockdown of znf408 in zebrafish resulted in abnormal development of eye and trunk vasculature (10). These data support the causality of the mutation and implicate an important role of ZNF408 in the development of vasculature. Nonetheless, very little is known about the exact molecular function of ZNF408.

To unravel the role of ZNF408 in the development of vasculature and to assess their angiogenic ability, in vitro tube formation assays were performed on endothelial cells overexpressing wild-type or mutant p.H455Y ZNF408. These cells were also subjected to transcriptome analysis to identify genes regulated by ZNF408, as well as the effect of the mutation on gene expression. Furthermore, chromatin immunoprecipitation (ChIP) followed by high-throughput sequencing (ChIP-SEQ) was employed to reveal the binding sites of ZNF408 in the human genome. Our data suggest that ZNF408 indirectly regulates the expression of genes important for the development of vasculature, and that this regulation is disrupted by the p.H455Y mutation.

Results

Overexpression of ZNF408 p.H455Y prevents HUVEC from forming capillary-like structures

To study the role of ZNF408, particularly in the development of vasculature, a cellular model for ZNF408-associated FEVR was established. A widely used endothelial cell system, human umbilical vein endothelial cells (HUVEC), was employed, and cells were transduced with lentiviral particles that contain constructs encoding HA-tagged wild-type or p.H455Y ZNF408. These transgenes were stably integrated into the HUVEC genome and allowed inducible overexpression of wild-type and p.H455Y ZNF408 (Fig. 1A and B). The integration of ZNF408 wild-type and p.H455Y was confirmed at the genomic DNA level (S1A Fig.), and their overexpression was validated both at the transcript and protein levels (S1B and S1C Fig.). The phenotype of these cellular models was assessed by in vitro tube formation assay, which is commonly used to examine the angiogenic ability of endothelial cells (19). Unlike overexpression of wild-type ZNF408, which did not interfere with the formation of capillary-like structures in this assay, overexpression of ZNF408 p.H455Y clearly inhibited the formation of these structures (Fig. 1C). This result showed that the in vitro models mimic the abnormal vasculature development observed in patients with FEVR, demonstrating their suitability to study the role of ZNF408 in this process.

Figure 1

Overexpression of ZNF408 p.H455Y resulted in the absence of capillary-like structures in tube-formation assay and disrupts the normal gene expression regulated by wild-type ZNF408.

(A) Lentiviral constructs for wild-type and p.H455Y mutant ZNF408 overexpression. LTR, long terminal repeats; Tet, tetracycline-inducible promoter; WPRE, woodchuck hepatitis virus post-transcriptional regulatory element; SV40, Simian virus 40; BSD, blasticidin resistance gene. (B) Schematic illustration of the experimental setup. NT, non-transduced HUVEC; WT, HUVEC transduced with wild-type ZNF408; MUT, HUVEC transduced with mutant ZNF408. ZNF408 overexpression was induced by doxycycline. (C) Fluorescent images of the capillary-like structures formed by NT and WT cells, but not MUT cells in an in vitro tube formation assay. Scale bars = 1000 μm. (D) PCA of the transcriptome data. Two biological replicates were analyzed in each condition. (E) and (F) MA-plot of the differential expression analysis, comparing WT to MUT and WT to NT, respectively. Red dots represent genes differentially expressed between conditions, and black dots represent no difference in expression. The number of genes with significant difference in expression is depicted on the corner of the plot. The top 10 biological process GO-terms associated to genes belonging to the upper and lower panel of the MA-plot are depicted above and below the MA-plot, respectively. The P-values were adjusted using the Benjamini-Hochberg method.

Overexpression of ZNF408 p.H455Y alters the expression of genes involved in development of vasculature

Next, we performed transcriptome analysis of the generated cells to study the underlying molecular mechanisms of the observed phenotype. Two biological replicates of each experimental condition, non-transduced HUVEC (NT), HUVEC overexpressing wild-type ZNF408 (WT) and HUVEC overexpressing p.H455Y ZNF408 (MUT), were included. The biological replicates clustered together in the principal component analysis (PCA), demonstrating their similarities. Notably, the three different conditions were separated from each other in PCA (Fig. 1D), indicating their distinct gene expression profiles. Furthermore, NT and MUT were located on the same end of PC2 axis, in contrast to WT that was located on the other end of PC2. The distinct gene expression profiles observed are consistent with the heatmap of gene expression in the three experimental conditions (S1D Fig.). Subsequently, differential expressed genes between the conditions were determined. In this analysis, three pairwise comparisons were made, namely a comparison of WT versus MUT (S1 Dataset, Fig. 1E), NT versus WT (S2 Dataset, Fig. 1F) as well as NT versus MUT (S3 Dataset, S1E Fig.). Gene ontology (GO) analysis was performed to investigate the biological processes associated with the differentially regulated genes. The terms nucleosome assembly and protein heterotetramerization were enriched for genes upregulated by WT (Fig. 1F), as compared to NT. The genes associated with these two terms mainly encode for histone H1 and H2. For the genes downregulated by WT, the terms cell adhesion and response to hypoxia were enriched (Fig. 1F), and included ITGA4, CTGF, COL8A1 and VCAM1, or MMP2, ANGPT2, TGFB2 and VCAM1, respectively. Interestingly, the GO terms enriched for the genes differentially upregulated by WT as compared to NT were rather similar to those downregulated by MUT as compared to WT, and vice versa (Fig. 1E and F). These findings indicate that many genes that were upregulated/downregulated by wild-type ZNF408 can be deregulated due to the p.H455Y mutation, suggesting that MUT can disrupt the regulation of gene expression which is controlled by WT.

To further focus on the mechanisms that are disrupted by MUT, we focused on genes whose expression is deregulated by MUT. In total, there were 122 genes that were exclusively upregulated by WT as compared to NT, but downregulated by MUT as compared to WT (WT > MUT, S1 Table, Fig. 2A and C), while there were 102 genes that were uniquely downregulated by WT as compared to NT, but upregulated by MUT as compared to WT (WT < MUT, S2 Table, Fig. 2B and D). Some of these genes were regulated in opposite direction by WT and MUT. For instance, HIST1H1D and HIST1H3C were upregulated by WT as compared to NT, but downregulated by MUT as compared to NT. Another example is VCAM1, which is downregulated by WT as compared to NT, but upregulated by MUT as compared to NT. Consistently, these deregulated genes were similar to the genes that mark the difference on the PC2 axis in the PCA analysis (S3 Table), further highlighting the fact that MUT can disrupt the regulation of these genes. The GO terms enriched for these genes were similar to those enriched in the pairwise comparisons. For instance, nucleosome assembly and protein heterotetramerization, which are relevant to chromatin organization and remodeling, were enriched for genes upregulated by WT (Fig. 2E). Likewise, terms relevant to the development of vasculature were again enriched for genes downregulated by WT. Besides response to hypoxia and cell adhesion, the term angiogenesis was also enriched in this set of genes (Fig. 2F), and included CTGF, COL8A1, ANGPT2, MMP2, FN1 and TGFB2. These data provide novel insights into the function of ZNF408, probably in chromatin organization and remodeling, and also confirm our previous findings of its role in vasculature development.

Figure 2

p.H455Y ZNF408 deregulates the expression of some genes.

(A) Venn diagrams showing the overlap of genes that are exclusively upregulated in WT compared to NT (WT > NT) and downregulated in MUT compared to WT (WT > MUT) and (B) Venn diagrams showing the overlap of genes that are exclusively downregulated in WT compared to NT (WT < NT) and upregulated in MUT compared to WT (WT < MUT). (C) and (D) Heatmap of the expression of these genes described in (A) and (B), respectively. The z-score was calculated based on the FPKM values of these genes. (E) and (F) The enriched biological process GO-terms for the genes described in (A) and (B), respectively. The P-values were adjusted using the Benjamini–Hochberg method.

Given the fact that ZNF408 is mutated in FEVR, we further focused on genes that are relevant to vasculature development. Quantitative PCR (qPCR) was performed to confirm the expression pattern of 15 genes known to be associated with vasculature development including HMOX1, CTGF, ITGA4, ANGPT2 and MMP2 (S2 Fig.). The difference in expression level between WT and MUT was validated for all of them, demonstrating the reliability of the RNA-SEQ data.

ZNF408 p.H455Y has reduced DNA-binding ability

The data obtained from the transcriptome analysis indicate a regulatory role for ZNF408 in the development of vasculature. Due to the presence of 10 zinc finger domains in ZNF408, we hypothesized that this protein regulates gene expression by binding to regulatory genomic regions. Therefore, a ChIP followed by high-throughput sequencing (ChIP-SEQ) experiment was performed. HEK293T cells were transfected with HA-tagged wild-type and p.H455Y mutant ZNF408, and the immunoprecipitation was performed using antibodies targeting either the HA-tag or the ZNF408 protein (Fig. 3A). There were 1548 and 3074 binding sites detected for wild-type ZNF408, while there were only 56 and 58 binding sites detected for p.H455Y ZNF408 in the HA-ChIP and ZNF408-ChIP, respectively. The overlapping binding sites between HA-ChIP and ZNF408-ChIP were considered the most reliable, which were 730 for wild-type ZNF408 and 10 for p.H455Y ZNF408. Interestingly, these results showed that p.H455Y ZNF408 has remarkably less binding sites compared to wild-type ZNF408 (Fig. 3B), examples of which are shown for ANXA5, GPR180 and LZTS2 (S3a Fig.). To validate the ChIP-SEQ results, the identified regulatory regions were cloned upstream of the luciferase sequence in the pGL3-Enhancer vector. The overexpression of these constructs, together with constructs encoding wild-type or p.H455Y ZNF408, showed activation of luciferase gene transcription by wild-type ZNF408, but not by p.H455Y ZNF408 (S3b Fig.), thereby confirming the reduced DNA-binding ability of p.H455Y that was observed in the ChIP-SEQ experiment.

Figure 3

ZNF408 p.H455Y has reduced DNA-binding ability.

(A) Schematic illustration of the experimental setup. NT, non-transduced HEK293T; WT, HEK293T transfected with wild-type ZNF408; MUT, HEK293T transfected with mutant ZNF408. (B) Number of peaks detected in ChIP using HA antibody, ZNF408 antibody and the overlap of both. (C) The distance of the overlapping peaks to TSS. (D) The top 10 biological process GO-terms of genes associated with ZNF408 binding sites. The P-values are adjusted with false discovery rate method.

The binding sites of wild-type ZNF408 were predominantly located around the transcription start site (TSS) (Fig. 3C) and associated with 778 genes (S4 Dataset). GO analysis showed that these genes are involved in diverse biological processes, such as nucleosome assembly, regulation of gene expression, megakaryocyte differentiation and protein processing (Fig. 3D). Despite the clear loss of binding demonstrated by p.H455Y ZNF408, only 15 of these genes are differentially regulated by wild-type and p.H455Y ZNF408 in HUVEC (S4 Table). Two of these 15 genes, HMOX1 and ITGA4, are involved in biological processes relevant to the development of vasculature. Intriguingly, gene expression analysis showed that wild-type ZNF408 upregulated the expression of HMOX1 while downregulating that of ITGA4, processes that were reversed by the p.H455Y mutation (S2 Fig.). ZNF408 binds to a region around 8 kb upstream the TSS of HMOX1, whereas it binds approximately 350 bp downstream of the TSS in ITGA4 (S4A and S4B Fig.). Cloning of these regions upstream of luciferase sequence followed by transactivation assay in HEK293T cells validated direct binding of wild-type ZNF408 to the ITGA4 TSS (S4C and S4D Fig.).

Discussion

In this study, the role of ZNF408 in the development of vasculature was investigated. An in vitro model of ZNF408-associated FEVR was generated by overexpressing wild-type or p.H455Y mutant ZNF408 in HUVEC. The overexpression of p.H455Y ZNF408 disrupted the ability of HUVEC to form a capillary-like network in vitro, which mimics the aberrant retinal vasculature development observed in patients with FEVR. Transcriptome analyses of the cellular models showed that the mutation altered the normal gene expression regulated by wild-type ZNF408, which may underlie the observed cellular phenotype. ChIP in HEK293T cells showed that the p.H455Y mutation reduced the DNA-binding ability of mutant ZNF408.

The results obtained in this study suggest a role of ZNF408 in various biological processes. Nucleosome assembly and protein heterotetramerization are two terms that were highly enriched for genes upregulated by wild-type ZNF408. The genes associated with these terms are mainly encoding for histones H1 and H2, implying a role for ZNF408 in chromatin remodeling. Interestingly, the term nucleosome assembly was also highly enriched for the genes identified in the ChIP-SEQ experiment and is mainly associated with genes encoding histone H1, H2 and H3. Although little overlap was observed between genes identified by ChIP-SEQ and the transcriptome analysis, our results suggest the involvement of ZNF408 in chromatin remodeling, which may directly or indirectly contribute to the observed vascular phenotype.

Response to hypoxia, cell adhesion and most importantly angiogenesis are processes that are directly relevant to the development of vasculature. Genes related to these terms were downregulated by wild-type ZNF408, which was deregulated by p.H455Y ZNF408. Given that the overexpression of wild-type ZNF408 did not inhibit in vitro tube formation, we hypothesize that wild-type ZNF408 promotes the vasculature development by downregulating genes inhibiting angiogenesis, explaining why reversal of this process by p.H455Y ZNF408 led to a disruption of in vitro tube formation. Indeed, based on literature, some of the genes associated with these terms (e.g. CTGF and ANGPT2) are proposed to act as angiogenesis inhibitors in some circumstances (20,21).

Hypoxia has been widely described as one of the inducers of angiogenesis, including in the retina (22). Hypoxic retinal astrocytes in avascular areas express the vascular endothelial growth factor (VEGF) and can direct endothelial cells to a specific site to form a vascular network (23,24). The downregulation of genes associated to a response to hypoxia by ZNF408 implies a role for this protein in regulating this response, albeit via an unknown mechanism. The deregulation of these genes by p.H455Y ZNF408 suggests a prevention of compensatory hypoxia-induced angiogenesis in FEVR patients, despite the presence of an avascular peripheral retina that would normally induce hypoxia.

Since ZNF408 is predicted to have 10 zinc finger domains, we hypothesized that it regulates gene expression by binding to genomic regulatory regions and therefore performed ChIP-SEQ experiment to determine ZNF408 binding sites. A relatively small number of peaks were detected in the ChIP-SEQ experiment. This, together with the difference in cellular systems used—HEK293T in ChIP-SEQ experiment and HUVEC in transcriptome analysis—may contribute to the little overlap with the differentially expressed genes. The small number of peaks also suggests that ZNF408 does not bind directly to DNA to regulate gene expression. Despite DNA binding being the primary role of many zinc finger domains, binding activity to other ligands such as RNA and protein has been described as well (25,26). Moreover, ZNF408 belongs to the PRDM family. Although some members of this family have been shown to bind DNA directly, other members of the family are known to regulate gene expression either by acting as direct histone methyltransferase or by recruiting other histone modifying enzymes. PRDM family proteins have also been shown to bind to transcription factors to target gene promoters or to act as non-DNA binding cofactors (17,18). These not only further suggest the role of ZNF408 in chromatin remodeling, but also support the hypothesis that ZNF408 acts in concert with other molecules to regulate gene expression instead of directly binding to its target genes. Further investigation on ZNF408 interacting partners will enable a more detailed understanding of the mechanism by which this protein regulates gene expression.

The data obtained in our ChIP-SEQ experiment showed ZNF408 binding sites near two differentially regulated genes which are relevant to the development of vasculature, namely HMOX1 and ITGA4. HMOX1 encodes for heme oxygenase 1 (HO-1), which is an enzyme that catalyzes the conversion of heme into carbon monoxide, free iron and biliverdin. HO-1 has a dual role in the development of blood vessels, in which its activation is required in VEGF-induced angiogenesis, while its inhibition promotes inflammatory angiogenesis (reviewed in (27,28)). Overexpression of wild-type ZNF408 in HUVEC resulted in an upregulation of HMOX1. Interestingly, overexpression of HMOX1 inhibits the expression of VCAM-1, a cell adhesion molecule that, together with its receptor α4 integrin (the protein encoded by ITGA4), mediates TNFα-induced angiogenesis (27,29). Induction of VCAM-1 expression by TNFα during inflammation triggers leukocyte migration through the cell junctions, which subsequently may lead to endothelial cell barrier breakdown (30,31). Blockade of α4 integrin has been shown to reduce leukocyte adhesion and thereby prevent vascular leakage in a diabetic retinopathy model (32). The overexpression of wild-type ZNF408 also resulted in a downregulation of ITGA4, VCAM1 and of genes associated to leukocyte migration, while their expression was deregulated by the p.H455Y mutation. This suggests an involvement of ZNF408 in protecting the endothelial cell barrier, thereby preventing vascular leakage. Additionally, binding of ZNF408 to the proximity of ITGA4, but not HMOX1, was further validated by a transactivation assay in HEK293T cells. This indicates that ZNF408 represses the expression of ITGA4 by binding directly to its TSS, whereas HMOX1 activation is more likely due to an indirect effect of ZNF408 overexpression.

Mutations in ZNF408 have also been reported in retinitis pigmentosa (RP), another type of inherited visual impairment in which the patients suffer from photoreceptor degeneration (33). Immunohistochemistry revealed that ZNF408 is expressed in both retinal vasculature as well as the photoreceptor layer of the human retina (33). The hypothesis that ZNF408 works together with other molecules to regulate gene expression may contribute to distinct phenotypes that can result from different ZNF408 mutations. However, it remains unclear what determines the exact outcome of the different mutations, as those reported in both FEVR and RP are spread throughout the gene (10,33–36), implying that there is not yet a clear correlation between the exact position of the mutation and the corresponding phenotype.

In summary, the results obtained in this study indicate that ZNF408 is involved in the regulation of genes involved in the development of vasculature as well as other biological processes. Our data also imply that ZNF408 mainly acts in concert with other molecules to regulate gene expression. This process is clearly disrupted by the FEVR-associated p.H455Y mutation, leading to the observed cellular phenotype, and thereby increases our understanding on the molecular mechanisms underlying ZNF408-associated vitreoretinopathy.

Materials and Methods

Cell culture

HUVEC (a kind gift from Dr William Leenders) were cultured in Endothelial Growth Medium 2 (EGM2, Promocell) supplemented with 1% penicillin/streptomycin or other required antibiotics. Culture dishes were coated with 0.01 mg/ml bovine fibronectin (Promocell) for at least 1 h at 37°C before use for HUVEC culture. HEK293T cells (ATCC) were cultured in DMEM (Sigma) supplemented with 10% FCS, 1% sodium pyruvate and 1% penicillin/streptomycin. HUVEC and HEK293T cells were maintained at 37°C and 5% CO₂.

Lentivirus production

To generate lentiviral constructs that enable inducible expression of N-terminally HA-tagged ZNF408, the 2K7_bsd lentivirus vector (37) was used. A tetracycline inducible promoter was cloned into the pDONR-P4P1R entry clone, whereas full-length human ZNF408 cDNA (wild-type and p.H455Y mutant) fused with a sequence encoding HA-tag were cloned into the pDONR201 entry clone. LR Clonase II Plus enzyme mix (Thermo Fisher Scientific) was used to insert the promoter and fusion gene into the 2K7_bsd lentiviral vector. To produce lentivirus particles, 5 μg of 2K7_bsd_Tet_HA-ZNF408 construct was co-transfected with 3.2 μg psPAX2 and 1.8 μg pMD2.G into HEK293T cells using the CaPO₄ (final concentration of 250 mM CaCl₂, 0.5 mM Tris pH 7.5, EDTA, 25 mM HEPES-NaOH pH 7.3, 280 mM NaCl, 1.5 mM NaPO₄) transfection method. Medium was refreshed 8 h post-transfection and the supernatant containing lentivirus particles was collected 48 h post-transfection. Lentiviral particles of tetracycline-controlled transcriptional activator (rtTA, pLVX-EtO) were produced following similar procedures.

Stable cells generation

HUVEC were seeded on a 12-well plate with 2.5 × 10⁵ cells/well density. After 6–8 h, they were transduced with supernatant containing lentiviral particles of 2K7_bsd_Tet_HA-ZNF408 and rtTA with 1:1 ratio overnight. The following day, the medium was changed to EGM2 without selection antibiotics. Ninety-six h post-transduction, 4 μg/ml blasticidin (Sigma) and 1000 μg/ml G418 (Sigma) were added to EGM2 to select for positively transduced cells. The selection medium was applied every 48 h for 6 days. The obtained stable cells were further cultured in medium containing 25% of the antibiotics dose used for selection.

Stable cells validation

The stable cells were validated for the integration of 2K7_bsd_Tet_HA-ZNF408 (wild-type and p.H455Y mutant) and rtTA constructs at the genomic DNA level. DNA was isolated from cell pellet using DNeasy Blood and Tissue kit (Qiagen) following manufacturer’s instructions. Twenty nanograms DNA was used as template for amplification. The expression of HA-ZNF408 was examined at RNA and protein level. Total RNA was isolated using Macherey-Nagel RNA isolation kit according to the kit manual. iScript kit (Bio-Rad) was used to synthesize cDNA from 250 ng RNA. Subsequently, GoTaq Green Master Mix (Promega) was used in the qPCR. Primers used in the genomic DNA amplification and qPCR are listed in S5 Table. Western blot analysis was performed to check the expression at protein level. Mouse anti-HA (Sigma, 1:1000) and rabbit anti-ZNF408 (Biorbyt, 1:500) were used to detect the expression of HA-ZNF408 at the protein level. Tubulin was stained as loading control using mouse anti-tubulin (Abcam, 1:2000). Donkey anti-mouse IRDye 680 RD and donkey anti-rabbit 800 CW (LI-COR, 1:20000) were used as secondary antibodies. The blot was scanned on LI-COR Odyssey CLx system.

In vitro tube formation assay

Non-transduced HUVEC as well as stable HUVEC were seeded at the density of 1.5 × 10⁵ cells in a 12-well plate format. Forty-eight h post-seeding, the overexpression of wild-type and p.H455Y ZNF408 was induced by doxycycline treatment (100 ng/ml and 500 ng/ml, respectively) for 48 h. The cells were then trypsinized and seeded on 96-well plate coated with Matrigel (BD Biosciences) at the density of 1.5 × 10⁴ cell per well. After 20 h, the cells were stained with Calcein Red (Thermo Scientific) at a 6 μM final concentration, followed by imaging on EVOS cell imaging station.

RNA-SEQ

Non-transduced HUVEC or the stably transduced HUVECs were seeded in duplicate at a density of 2.5 × 10⁵ cells in 6-well plate format. Overexpression of wild-type and p.H455Y ZNF408 was induced as mentioned previously. The cells were harvested for total RNA isolation using Macherey-Nagel RNA isolation kit, following manufacturer’s instructions. Ribo-Zero rRNA removal kit (Illumina) was used to eliminate rRNA from the sample. RNA fragmentation as well as first and second strand cDNA synthesis was performed as described in (38), with the deviation that RNA fragmentation was performed for 3 min at 95°C instead of 1.5 min. The obtained cDNA was used for next generation sequencing on Illumina NextSeq 500 platform.

RNA-SEQ data analysis

Sequence reads were mapped to the human genome (build hg19) using the STAR algorithm (39). Number of reads per gene were counted using HTseq (40) followed by differential gene expression analysis using DESeq2 package (41). The cutoff used to determine a significant difference in gene expression was set at a Benjamini-Hochberg adjusted P-value of <0.05 and a fold-change of ≥1.5. The fragments per kilobase of transcript per million mapped reads (FPKM) of differentially expressed genes were calculated using Cuffdiff (42). GO-terms enrichment analysis was performed on the lists of differentially expressed genes by DAVID with default settings (43,44). The list of enriched GO-terms was summarized using REVIGO, with the settings: small list (allowed similarity = 0.5), Homo sapiens database and SimRel for semantic similarity measure (45).

Real-time qPCR

Some of the differentially regulated genes detected in the RNA-SEQ data were validated by means of real-time qPCR using GoTaq Green Master Mix (Promega). qPCR was performed on three biological replicates and Student’s t-test was used to determine significance difference. The primers used for this are listed in Table S1.

ChIP-SEQ

Full-length human wild-type and p.H455Y mutant ZNF408 cDNA were cloned into pcDNA3-HA vector to generate a construct encoding ZNF408 with the HA-tag at the N-terminus. These constructs were transiently transfected into HEK293T cells cultured in 15-cm dishes using Fugene HD reagent (Promega). Forty-eight h post-transfection, the cells were crosslinked using 1% formaldehyde at room temperature for 10 min. The crosslinking was stopped with 0.125 M glycine and the nuclear fraction was harvested as described in Denissov et al. (46). DNA was fragmented by sonication using Bioruptor (Diagenode). The settings of the sonication were high power, 16 cycles of 30 s ON and 30 s OFF. In each immunoprecipitation reaction, 100 μl sonicated nuclear lysate, 30 μl Prot/AG beads (Santa Cruz) and 1–2 μg antibody were used. The antibodies used for ChIP were mouse monoclonal anti-HA (1 μg per reaction, Sigma) and rabbit polyclonal anti-ZNF408 (2 μg per reaction, Biorbyt). Immunoprecipitation reactions were incubated at 4°C overnight. Immunoprecipitated DNA was purified by phenol/chloroform extraction. Next generation sequencing was performed on Illumina HiSeq 2000 using 5 ng of the immunoprecipitated DNA.

ChIP-SEQ data analysis

The sequence reads were uniquely mapped to the human genome build hg19 using bwa algorithm (47). Peak calling was performed using MACS2 (48) with default settings and a P-value threshold of 1E-05. Data from non-transfected HEK293T cells were used as control file. The functional significance of the detected peaks was assessed using GREAT with basal plus extension settings (49). The list of enriched GO-terms was summarized using REVIGO, with the settings: small list (allowed similarity = 0.5), Homo sapiens database and SimRel for semantic similarity measure (45).

Transactivation assay

Putative ZNF408 binding sites on the human genome were amplified and subsequently cloned into Gateway adapted pGL3-Enhancer vector. The constructs were co-transfected with pcDNA3-HA, pcDNA3-HA-ZNF408 wild-type or pcDNA3-HA-ZNF408 p.H455Y vector into HEK293T cells using Fugene HD transfection reagent (Promega). The transfection was performed in triplicates on 24-well format. Forty-eight h after transfection, the cells were lysed and luciferase expression was measured using Dual Luciferase Assay kit (Promega) following the manufacturer’s instructions. Luciferase measurements were performed in duplicate using 5 μl of cell lysate. Student’s t-test was performed to assess significant differences between wild-type and mutant.

Statistical analysis

Statistical tests used in the data analysis of the different experiments are described in detail in the sections that belong to the corresponding experiments.

Acknowledgements

We would like to thank William Leenders for providing the HUVEC used in this study, Martin Oti for helpful discussions on data analysis, Simon van Heeringen for helpful discussions on motif finding, Bert van den Heuvel for providing endothelial cell RNA as well as Saskia van der Velde-Visser and Marlie Jacobs-Camps for cell culture assistance.

Conflict of Interest statement. None declared.

Funding

Radboudumc PhD Grant (to D.W.K).

Author notes

These authors contributed equally to this work.