Controlling gene expression through five zinc finger domains of ZNF18

Abstract

Zinc finger (ZF) proteins are the most abundant transcription factors in vertebrates, and they regulate gene expression through interactions with cis‐acting elements. ZF domains selectively recognize specific sequences to accelerate or repress target genes. Zinc finger protein 18 (ZNF18) contains five CX₂CX₁₂HX₃H‐type ZFs at the C‐terminus, which are expressed in the brain and other organs of the biological system. Bioinformatic study proposed that cyclin‐dependent kinase 1 (CDK1) is in the signaling cascade of ZNF18; although experimental evidence has not yet been reported. In this study, we expressed and purified ZNF18(ZF1‐5), five ZF domains from ZNF18, and investigated metal binding specificity and promoter interactions. ZNF18(ZF1‐5) has specific coordination to Zn²⁺ (K _d  ≤ 18 nM) compared with other xenobiotic metal ions, including Co²⁺, Fe²⁺, and Fe³⁺, with 98.5% of reduced ZF domains after purification. This significantly active ZF can be one of the major reasons for tight coordination affinity. CDK1 rescued the arrested cell cycle induced by DNA damage, resulting in tumorigenesis. Zn²⁺–ZNF18(ZF1‐5) specifically binds to cis‐acting elements of cdk1 (K _d  = 4.63 ± 0.07 nM), mediated by a cell cycle‐dependent element (cde, 5′‐CGCGG) and a cell cycle gene homology region (chr, 5′‐TTGAA). The ZNF18 superfamily was expressed in the brain for the regulation of neuronal development and cell differentiation. Zn²⁺–ZNF18(ZF1‐5) interacted with promoters in the insulin response sequence (IRS) for inhibition of dopamine secretion and cis‐acting element of brain‐2 (BRN2), which controlled astrocyte and cancer development. These results provide the first evidence that ZNF18(ZF1‐5) regulates the cell cycle and neuronal development through transcriptional regulation.

1. INTRODUCTION

Zinc finger (ZF) proteins are essential biological triggers that orchestrate the central dogma, cell signaling, and communication within physiological systems (Lee & Michel, 2014; Yoon et al., ²⁰²⁰). As metalloproteins, ZF proteins typically contain multiple ZF domains that enable them to either initiate or repress the expression of specific genes in eukaryotic cells (Klug, 2010; Tan et al., ²⁰⁰³). These domains interact with diverse binding partners, including nucleic acids, proteins, lipids, and small molecules, for the up‐ or down‐regulation of cellular events in response to intra‐ and extra‐stimuli (Berg & Shi, 1996; Liu et al., ²⁰¹⁷). The ββα secondary structure of ZF domains induces specific hydrogen bonding with phosphate backbones and nitrogenous bases in nucleic acids (Laity et al., 2001; Wolfe et al., ²⁰⁰⁰). The coordination of zinc ions with cysteine (Cys) and histidine (His) residues generates a tetrahedral geometry, although this structural feature disappears in the absence of zinc ions (Berg & Godwin, 1997; Lee et al., ²⁰¹¹). The binding affinities between the ZF domains and Zn²⁺, described by dissociation constants (K _d s), are typically in the sub‐nanomolar range (K _d s ≤ 10⁻⁹), although some ZF proteins exhibit even higher affinities in the picomolar range (K _d s ≤ 10⁻¹² ~ 10⁻¹⁵) (Lee et al., 2011; Michalek et al., ²⁰¹²). These biophysical properties of the ZF domains have been extensively studied to understand their regulatory roles in transcription and translation. Furthermore, elucidating the upstream mechanisms of ZF proteins could provide insights into the design of therapeutic candidates.

Recent reports have proposed a superfamily of ZF proteins in the brain, characterized by four or five ZF domains, each of which has a CX₂CX₁₃HX₃H (C: cysteine, X: amino acids, and H: histidine) sequence (Hwang, Park, et al., 2024). One such example is the Parkin‐interacting substrate (PARIS), which contains three consecutive CX₂CX₁₃HX₃H ZF domains, PARIS(ZF2‐4), in its C‐terminus (Kim et al., 2021). These domains interact with cis‐acting elements in the insulin response sequence (IRS), thereby downregulating peroxisome proliferator‐activated receptor gamma coactivator‐1 alpha (PGC‐1α), which influences the expression of genes that regulate dopamine secretion (Kang et al., 2017; Lee et al., ²⁰¹⁷). Zinc finger protein 18 (ZNF18), another CX₂CX₁₃HX₃H‐type, includes five consecutive ZF domains, ZNF18(ZF1‐5), which share 57% homology with PARIS(ZF2‐4) (Figure S1). Putative irs elements were identified through an interaction study with PARIS(ZF2‐4), and this promoter region needs to be further investigated in ZNF18(ZF1‐5) (Santos et al., 2020; Shin et al., ²⁰¹¹). The serine/threonine kinase, cyclin‐dependent kinase 1 (CDK1), plays a pivotal role in regulating the cell cycle and is a key regulator of the CDK superfamily in mammals (Chang et al., 2021; Xie et al., ²⁰¹⁹). The concentration of CDK1 during each stage of the cell cycle is significantly altered through cell division and transcriptional modification (de Tudela et al., ²⁰¹⁵; Liao et al., ²⁰¹⁷). CDK1 is upregulated in patients with colorectal cancer, whereas ZNF880 expression is downregulated (Dong et al., 2023; Liao et al., ²⁰¹⁷; Wang et al., ²⁰²³). This indicates that ZNF880 represses CDK1, and the sequence similarity between ZNF18(ZF1‐5) and ZNF880(ZF7‐11) suggests that ZNF18 may be associated with CDK1 (Figure S2). Despite these findings, biochemical studies on the relationship between ZNF18 and CDK1 have not yet been described. Meanwhile, studies that have monitored the expression of ZNF18 in the brain have proposed that it is related to autism spectrum disorder (Chahrour et al., 2012; Khaliulin et al., ²⁰²⁴; Sener et al., ²⁰¹⁶). Therefore, biophysical studies on ZNF18 are necessary to clarify its regulatory roles in physiological systems.

A schematic diagram of ZNF18 shows that its functional domain is located in the N‐terminal region, whereas the five ZF domains are positioned at the C‐terminus (Figures 1a and S3). The SRE‐ZBP, Ctfin51, AW‐1, and Number 18 (SCAN) domains have a conserved motif with a distinct role in protein–protein interactions; specifically, they mediate dimerization by interacting with other SCAN domains to form homodimers or heterodimers (Collins et al., 2001; Nam et al., ²⁰⁰⁴; Schumacher et al., ²⁰⁰⁰). This event stabilizes protein complexes and increases transcriptional activity, and heterodimerization enhances functional diversity by engaging specific cofactors or targeting distinct genes (Edelstein & Collins, 2005; Huang et al., ²⁰¹⁹). The Krüppel‐associated box (KRAB) domain comprises two subdomains, with KRAB‐A mediating transcriptional repression, and KRAB‐B stabilizing KRAB‐A activity (Collins et al., 2001; Kim et al., ¹⁹⁹⁶; Urrutia, ²⁰⁰³). The KRAB domain interacts with the co‐repressor KRAB‐associated protein 1 (KAP1), which facilitates the assembly of a multi‐protein complex including histone methyltransferase, histone deacetylase, and heterochromatin protein 1, leading to a transcriptionally repressive heterochromatin (Groner et al., 2010; Sripathy et al., ²⁰⁰⁶; Stoll et al., ²⁰²²). This mechanism is essential for silencing transposable elements and establishing heritable heterochromatin structures that ensure genomic stability and precise gene expression (Wells et al., 2023; Yang et al., ²⁰¹⁷).

FIGURE 1.

ZNF18 from Homo sapiens. (a) Domain organization of ZNF18. (b) Sequence alignment of the five zinc finger domains of ZNF18. Identical and similar amino acid sequences are shown in blue and green, respectively. Bold letters highlighted in yellow indicate sequences related to the Zn²⁺ coordination site.

Although the biophysical properties of ZNF18(ZF1‐5) have not yet been reported, these five ZF domains exhibit repetitive amino acid patterns. To investigate the functions of these classical ZF domains, we expressed and purified ZNF18(ZF1‐5), which can provide structural details of these proteins (Figures 1 and S3). The coordination spheres and preferences for xenobiotic metals in ZF proteins exhibit different patterns (Besold et al., 2010; Hanas & Gunn, ¹⁹⁹⁶). Previous studies have shown that the zinc ions in PARIS(ZF2‐4) can be replaced by other metal ions, including Co²⁺, Ni²⁺, and Fe²⁺, although this substitution abolishes their affinities to their binding partners (Guerra & Giedroc, 2012; Hwang, Mohammad Mydul Islam, et al., ²⁰²⁴). The distorted secondary structure of ZF domains usually cannot recognize their binding partners owing to their high selectivity (Hartwig, 2001; Predki & Sarkar, ¹⁹⁹⁴). The fundamental characterization of ZNF18 is required to understand the coordination of Zn²⁺ and other xenobiotic metals and their interactions with nucleic acids.

The functional roles of ZNF18(ZF1‐5) require further investigation because of its higher transcriptional control hierarchy, which could influence the expression of specific genes. Although several reports have described their role in controlling downstream transcription factors, no reports have focused on its biochemical characteristics. Therefore, this study investigated its upstream regulatory role in CDK1, a potential therapeutic target for cancer. Additionally, we examined two crucial transcription factors, the IRS and brain 2 (BRN2), which are associated with neuronal development and regulatory functions (Sakaeda et al., 2013; Santos et al., ²⁰²⁰). Classical ZF domains from PARIS repress dopamine release by interacting with a promoter element positioned at PGC‐1α, while BRN2 is associated with glioblastoma, neuroblastoma, and retinoblastoma (Herbert et al., 2019; Shin et al., ²⁰¹¹). The brn2 gene also affects the development of small‐cell lung cancer (SCLC), although further studies are needed (Sakaeda et al., 2013). In this study, we performed a biophysical characterization of ZNF18(ZF1‐5) to elucidate its role as a master regulator of transcription factors. Elucidation of this master protein can be applied to understand the balance of ubiquitous transcription through diverse gene expression and can shed light on how these ZF proteins balance interactions.

2. RESULTS

2.1. Biophysical characterization of ZNF18(ZF1‐5)

2.1.1. Expression and purification of ZNF18(ZF1‐5) and coordination to Co²⁺

ZNF18(ZF1‐5), with 147 amino acids (17,247.66 Da), was over‐expressed in Escherichia coli using isopropyl‐β‐D‐1‐thiogalactopyranoside (IPTG) after codon optimization to address solubility issues (Figure S4a). ZNF18(ZF1‐5) was monitored in the soluble fraction and purified by 2‐step processes (pI: 9.64) to obtain >95% purity for biophysical characterization, as described in Figure S4. UV–visible spectra represent typical ZF domain absorptions (Figure S4b), and the purified apo‐protein was subjected to Co²⁺ titration because of the spectroscopic properties of Zn²⁺, which is a spectroscopic silence with a fully occupied d¹⁰‐orbital (Alies et al., 2016). The Co²⁺ titration demonstrated typical charge‐transfer absorptions at 309 and 355 nm and d–d electronic absorption bands at 574 and 643 nm (Figure 2a,b). These absorptions are consistent with those of other classical ZF proteins, including PARIS and ZIF268 (Hwang, Park, et al., 2024; Lee et al., ²⁰¹¹). The K _d for Co²⁺ binding to apo‐ZNF18(ZF1‐5) was 0.42 ± 0.08 μM, thereby representing the general affinity of Co²⁺ to ZF domains (Figure 2c), although PARIS shows different coordination preferences. Apo‐PARIS(ZF2‐4) shows an extraordinarily low dissociation constant (K _d  = 49.1 ± 7.7 nM) to Co²⁺, a distinctive feature of the ZF domains in PARIS (Hwang, Mohammad Mydul Islam, et al., 2024). Although ZNF18(ZF1‐5) and PARIS(ZF2‐4) have high sequence homology, the metal‐binding preferences of these two ZF domains from the CX₂CX₁₂HX₃H‐type show different patterns (Figure S1).

FIGURE 2.

Dynamics of metal‐coordination in ZNF18(ZF1‐5). (a) UV–Vis spectra showing Co²⁺ titration to apo‐ZNF18(ZF1‐5). (b) Increase in the d–d transition band at 643 nm as Co²⁺ coordinates to apo‐ZNF18(ZF1‐5). (c) Determination of the K _d value by curve fitting with the changes in absorption at 643 nm observed during Co²⁺ titration. (d) Non‐specific d–d transition bands at 695 and 743 nm during Co²⁺ titration. (e) Proposed model of Co²⁺ coordination in ZNF18(ZF1‐5) with Cys₄ mis‐coordination at low Co²⁺ equivalents and proper Cys₂His₂ coordination with excess Co²⁺. Experiments were performed with 10 μM ZNF18(ZF1‐5) in 100 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM TCEP, N₂ (g)‐filled glovebox, and 25°C.

The oxidation rate of ZF domains is significantly affected by the amino acid sequences, and this oxidation is critical for biological functions due to the Zn²⁺ coordination being necessary for the proper folding of each ZF domain (Hübner & Haase, 2021; Lee & Michel, ²⁰¹⁰). ZIF268, which contains three classical ZF domains with cysteine thiols (–SH), underwent rapid oxidation, reaching approximately 90% oxidation within 1 min; whereas >90% of PARIS(ZF2‐4) remained reduced (Lee et al., 2011). The ββα secondary structure potentially represents the secondary structure of classical ZF domains, and the tetrahedral Zn²⁺ coordination is essential for this folding with reduced –SH from Cys (Lee et al., 2011). The oxidation of ZNF18(ZF1‐5) was evaluated using cobalt‐binding assays, and more than 98% of the thiols remained reduced (Figure 2b). These results indicated that the ZF domain of ZNF18 was not easily oxidized, suggesting that the CX₂CX₁₃HX₃H motif is a suitable candidate for in vitro systems. Resistance to the oxidation of ZF domains is crucial for applications because rapid oxidation induces metal elimination, which causes loss of folding of the ZF domain.

2.1.2. Misfolding between Co²⁺–ZNF18(ZF1‐5)

Non‐specific d–d absorption was monitored at 695 and 743 nm by Co²⁺ titration of apo‐ZNF18(ZF1‐5) as shown in Figure 2b. These absorptions quantitatively increased to two equivalents per ZF domain and were diminished by additional Co²⁺ titration (Figure 2d). These electronic configurations, based on the CCCC‐type ZF domains from CP‐1, GATA‐1, and XPA, were identified as four d–d absorptions with additional transitions at 742 and 695 nm (Kopera et al., 2004; Krizek et al., ¹⁹⁹³; Michalek et al., ²⁰¹¹). Co²⁺ coordination in the ZF domains of ZNF18(ZF1‐5) caused unexpected coordination involving cysteines from other domains (Figure 2e). Apo‐ZNF18(ZF1‐5), which contains five classical ZF domains (CCHH), exhibited CCCC coordination at low Co²⁺ equivalents (0 < Co²⁺/ZNF18(ZF1‐5) ≤ 2); however, this mis‐coordination was reorganized into proper folding by the titration of five equivalents of Co²⁺. Absorptions at 695 and 743 nm can be generated because of the relatively large number of amino acids from the five ZF domains with structural flexibility; however, equimolar concentrations of metal ions induced a structurally stable CCHH coordination to metal ions. Unexpected CCCC coordination can be generated through intra‐ and inter‐CCCC coordination, where one Co²⁺ ion can coordinate with two cysteines from different ZF domains or even from other ZNF18(ZF1‐5) molecules. Notably, over two equivalents of Co²⁺ demonstrated that the CCCC mis‐coordination is changed to proper CCHH coordination due to the electronic configurations from the stable secondary structure of the ZF domains.

2.1.3. Zn²⁺ binding affinity

The Zn²⁺ titration to Co²⁺−ZNF18(ZF1‐5) resulted in a sharp decrease in d–d transitions, with an upper‐limit K _d of 18 nM (Figure 3). The K _d of PARIS(ZF2‐4) was 7.4 ± 9.2 nM, another CX₂CX₁₂HX₃H‐type ZF domain, which has a similar binding affinity to Zn²⁺ (Hwang, Mohammad Mydul Islam, et al., 2024). The metal‐binding specificity of PARIS(ZF2‐4) demonstrates tight binding to Co²⁺ and a general binding preference for Zn²⁺. Compared with the K _d s of metal‐coordination in PARIS(ZF2‐4), the five ZF domains of ZNF18(ZF1‐5) have a general metal preference, although the protein belongs to the same superfamily. Although Co²⁺ was used as a spectroscopic probe, its binding specificity was not as tight as that of Zn²⁺; the Co²⁺–ZF domains cannot interact with their cognate nucleic acids through hydrogen bonds (Krizek et al., 1991; Lee et al., ²⁰¹¹). ZNF18(ZF1‐5) and PARIS(ZF2‐4) have similar affinities for Zn²⁺, whereas their Co²⁺ binding affinities are quite different. This difference may be attributed to the sequence homology in the primary structures of the two ZF domains (Figure S1). The similar metal preferences and oxidation patterns of PARIS(ZF2‐4) and ZNF18(ZF1‐5) are induced by similar amino acid patterns, and these ZF proteins are expressed in neuronal systems (Chahrour et al., 2012; Shin et al., ²⁰¹¹). Suppression of these ZF proteins should be further investigated through interaction studies with cis‐acting elements (Jen & Wang, 2016).

FIGURE 3.

Binding of Zn²⁺ to ZNF18(ZF1‐5) analyzed through competitive binding. (a) UV–Vis spectra showing Zn²⁺ back‐titration to Co²⁺–ZNF18(ZF1‐5). (b) Decrease in the d–d transition band at 643 nm as Co²⁺ is replaced by Zn²⁺ in ZNF18(ZF1‐5). (c) Determination of the K _d value by curve fitting with the changes in absorption at 643 nm observed during the Zn²⁺ back‐titration. Experiments were performed with 10 μM ZNF18(ZF1‐5) in 100 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM TCEP, N₂ (g)‐filled glovebox, and 25°C.

2.1.4. Coordination of xenobiotic metal ions to apo‐ZNF18(ZF1‐5)

Some ZF domains coordinate with Co²⁺, Fe²⁺, and Fe³⁺, although metal substituted–ZF domains do not show selectivity to their biological partners (Guerra & Giedroc, 2012; Lee & Michel, ²⁰¹⁴). Charge transfers at 309 and 335 nm from Fe²⁺ to apo‐ZNF18(ZF1‐5) were induced by titration (Figure 4a), and these peaks were also generated from Fe²⁺–PARIS(ZF2‐4) (Hwang, Mohammad Mydul Islam, et al., 2024). The intensities of charge transfers from Fe²⁺–ZNF18(ZF1‐5) increased to five equivalents to a ZF domain, as most cysteine residues are active to maintain the reduced state. The Co²⁺ titration assay measuring ZF oxidation demonstrated that PARIS(ZF2‐4) had an approximately 90% reduced state, and the redox‐sensitive metal ion, Fe²⁺, was equivalently coordinated to apo‐PARIS(ZF2‐4) (Hwang, Mohammad Mydul Islam, et al., 2024). These results confirmed that the ZF domains from CX₂CX₁₂HX₃H‐type were less sensitive to oxidation, which is crucial for retaining the secondary structure essential for DNA interactions.

FIGURE 4.

Oxidation state‐dependent interaction of iron with apo‐ZNF18(ZF1‐5). (a) UV–Vis spectra showing the coordination of Fe²⁺ to apo‐ZNF18(ZF1‐5). (b) Determination of the K _d value by the curve fitting of the absorbance at 309 nm during Fe²⁺ titration to apo‐ZNF18(ZF1‐5). (c) UV–Vis spectra showing Fe³⁺ titration to apo‐ZNF18(ZF1‐5). Experiments were performed with 10 μM ZNF18(ZF1‐5) in 100 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM TCEP, N₂ (g)‐filled glovebox, and 25°C.

The coordination of Fe²⁺ to apo‐ZNF18(ZF1‐5) was investigated to understand binding affinity (K _d  = 0.38 ± 0.04 μM, Figure 4b) that shows lower dissociation constants compared to apo‐PARIS(ZF2‐4) with K _d = 2.2 ± 0.1 μM (Hwang, Mohammad Mydul Islam, et al., 2024). This result indirectly suggests that ZNF18(ZF1‐5) has two additional ZF domains and remains stable under oxidized conditions. TTP‐2D, a non‐classical ZF domain, has tandem CCCH sequences, and Fe²⁺ coordinates to TTP‐2D, and Fe²⁺–TTP‐2D can interact with its cognate RNA with specificity (vide infra), although Fe²⁺‐PARIS(ZF2‐4) did not recognize its target DNA (Lee & Michel, 2010). The oxidized Fe³⁺ did not coordinate with apo‐ZNF18(ZF1‐5) as observed in Figure 4c, and apo‐PARIS(ZF2‐4) represents the same coordination to Fe³⁺. The metal‐coordination trends of the CX₂CX₁₂HX₃H‐type ZF domains can be explained by ZNF18(ZF1‐5) and PARIS(ZF2‐4), and the results show significant differences between these two ZF domains. PARIS(ZF2‐4) has a significant coordination preference for Co²⁺, whereas ZNF18(ZF1‐5) has a traditional binding affinity (Yoon & Lee, 2021). The lower K _d of ZNF18(ZF1‐5) compared to PARIS(ZF2‐4), which typically coordinates with Zn²⁺, can be attributed to the stable oxidation states of these ZF domains.

2.2. Exploring the interaction between ZNF18(ZF1‐5) and the cdk1 promoter

2.2.1. Specific recognition of cdk1 by ZNF18(ZF1‐5)

A recent analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) proposed that ZNF18 is a hub gene in the tumorigenesis network; this association results from the signaling consequences of CDK1 (Zhu, 2023). CDK1, a pivotal regulator of the cell cycle, is upregulated in patients with cancer, and CDK1 inhibitors are promising targets (Han et al., 2023; Wang et al., ²⁰²³). ZNF880, a CX₂CX₁₂HX₃H‐type ZF protein with high sequence homology (62%) to ZNF18, has been studied as a comparative model (Figure S2). Colorectal cells have high CDK1 levels along with reduced ZNF880 expression, whereas normal cells display lower CDK1 levels with relatively higher ZNF880 expression (Dong et al., 2023). These results indirectly suggest that ZNF18 modulates the expression of CDK1, a transcription factor, during the cell cycle, which is a crucial step in apoptosis.

Gene map analysis indicated that cis‐acting elements in the promoter of cdk1 consisted of two important sequences, including cell cycle‐dependent elements (CDE, positioned −22 to −18) and a cell cycle gene homology region (CHR, positioned −12 to −8), as described in Figures 5a and S5. Mutational studies have suggested that the depletion of CDK1 transcription depends on these two elements (Badie et al., 2000; Fajas et al., ²⁰⁰⁰). To understand the activity of ZNF18(ZF1‐5) as a transcription factor, fluorescein‐labeled CDE and CHR sequences containing 29‐ and 21‐base pair (bp) elements were used, as described in Figures 5b and S6. The wild‐type cis‐acting element with 29‐bp shows the most selective interaction (K _d  = 4.63 ± 0.07 nM) to Zn²⁺–ZNF18(ZF1‐5); this result represents the first evidence that five ZF domains interact with the cdk1 promoter, suggesting a potential role in transcriptional regulation of CDK1 (Figure 5b). CDE and CHR mutational studies were performed to evaluate binding specificity, and these mutated 29‐bp sequences maintained tight interactions. The CDE (cdk1‐M1) and CHR (cdk1‐M2) mutated sequences show 5.04 ± 0.08 nM and 4.9 ± 0.2 nM K _d values, respectively (Figure 5b). Both the CDE‐ and CHR‐mutated sequences, cdk1‐M3 and cdk1‐M4, lost their specific interactions compared with the wild‐type sequences. Only the CDE‐ or CHR‐mutated sequence had almost the same binding affinity because the properties of fluorescence anisotropy are influenced by one‐element interactions. The expression levels of CDK1 are cell cycle‐dependent and modified according to the stage of the cell cycle. This effect could be modified by suppressing CDK1 expression (Badie et al., 2000; Liao et al., ²⁰¹⁷; Wang et al., ²⁰²³). Furthermore, short‐sequence 21‐bp cis‐acting elements were measured using ZNF18(ZF1‐5) as shown in Figure S6. The wild‐type 21‐bp (cdk1‐21‐WT) exhibited selectivity with a K _d of 1.67 ± 0.06 nM, which is slightly tighter than that of the 29‐bp element. This indicates that additional sequences can retard the recognition by ZNF18(ZF1‐5).

FIGURE 5.

Interaction between ZNF18(ZF1‐5) and the cdk1 promoter. (a) Schematic diagram of CDK1 inhibition by ZNF18 through binding to the CDE and CHR elements. (b) Binding of ZNF18(ZF1‐5) to the cdk1 promoter and its mutated variants. Native and mutated sequences are highlighted in bold black and red, respectively. The K _d values, measured through curve fitting, are presented in the inlet table as averages with standard deviations. Experiments were performed with 5 nM DNA in 25 mM MOPS (pH 7.4), 50 mM NaCl, 1 mM DTT, and 25°C.

Further studies were conducted to investigate the influence of replacing zinc ions in ZNF18(ZF1‐5) with other metal ions, including Co²⁺, Fe²⁺, and Fe³⁺, on its binding affinity for the cdk1 promoter (Figure S7). This substitution completely abolished the binding affinity of ZNF18(ZF1‐5) for the cdk1‐WT promoter, consistent with previous findings from PARIS(ZF2‐4) (Hwang, Mohammad Mydul Islam, et al., 2024). These results revealed the essential role of zinc ions in maintaining the structural integrity and functionality of ZNF18(ZF1‐5). The loss of binding affinity likely results from the distorted secondary structure of ZF domains, which hinders proper recognition of the DNA binding site. Modifying this metal coordination could be explored as a potential therapeutic strategy to modulate ZNF18 activity in pathological conditions.

2.2.2. Interaction studies of cde and chr promoters

The CDE and CHR elements were further investigated independently to evaluate their binding specificity to the ZF domains (Figure 6). The wild‐type promoter of CDE (cde‐WT, 5′‐TAGCGCGGTGAG‐F) and CHR (chr‐WT, 5′‐GAGTTTGAAACT‐F) selectively interacted with ZNF18(ZF1‐5) with K _d  = 1.71 ± 0.07 nM and K _d  = 1.56 ± 0.05 nM (Figure 6), respectively. The cde sequence has a specific promoter element, 5′‐CGCGG, and mutational studies demonstrated that the cytosine residue was crucial for ZNF18 interactions (Figure 6a). The binding specificity of chr‐WT indicates that thymine and adenine 5'‐TTGAA are crucial for the interaction with ZNF18(ZF1‐5) (Figure 6b). These mutational studies showed that chr‐M2 (5′‐TTTAA) had a tight interaction with Zn²⁺‐ZNF18(ZF1‐5), with K _d  = 1.68 ± 0.04 nM, whereas the mutated chr‐M1 (5′‐GGGGG) lost its interaction. Both elements had a high affinity for ZF domains and specific sequences for recognition in the cdk1 promoter sequences. CDK1 is considered the controlling protein because the inhibition of CDK1 arrests the G2/M cycle and stimulates apoptosis (Badie et al., 2000). These results revealed that the ZF domain of ZNF18 is a potential therapeutic target for the treatment of tumors.

FIGURE 6.

Binding of ZNF18(ZF1‐5) to cis‐acting elements and its mutated variants. (a, b) Interaction of ZNF18(ZF1‐5) with (a) CDE and (b) CHR. Native and mutated sequences are highlighted in bold black and red, respectively. The K _d values, measured through curve fitting, are presented in the inlet table as averages with standard deviations. Experiments were performed with 5 nM DNA in 25 mM MOPS (pH 7.4), 50 mM NaCl, 1 mM DTT, and 25°C.

2.3. Neurological factor, ZNF18(ZF1‐5), for gene regulation in the brain

2.3.1. Regulation of dopamine release through insulin response sequences

ZNF18(ZF1‐5) shares the CX₂CX₁₂HX₃H classical domain pattern with PARIS(ZF2‐4), although the first domain of PARIS has a non‐classical ZF domain with CCHC (Yoon & Lee, 2021). Previous biophysical studies have identified that PARIS(ZF2‐4) interacts with the irs sequence (5′‐TATTTT) with a K _d of 38.9 ± 2.4 nM (Hwang, Mohammad Mydul Islam, et al., 2024). Consequently, dopamine depletion causes pathological progression, including Parkinson's disease and neural metabolic disorders (Lee et al., 2017). In this study, the functional roles of ZNF18(ZF1‐5) were investigated to identify its binding partners and interaction details with the brn2 promoter and irs, as shown in Figure 7. A fluorescent probe attached to the irs element interacts with ZNF18(ZF1‐5) with K _d  = 0.82 ± 0.03 nM, indicating that the five ZF domains in ZNF18 have tighter interactions than those of the three ZF domains from PARIS. The expression of ZNF18 has been monitored in the human body except in the cardiac system (Guo et al., 2005), and this result suggests that ZNF18 can interact with the promoter region of PGC‐1α, potentially influencing dopamine secretion (Figure S8). Mutational studies using irs‐M1 and irs‐M2 showed that they lost their specific and selective interactions with ZNF18(ZF1‐5) as shown in Figure 7a. This finding provides the first evidence that the ZF domains of ZNF18 exhibit a higher binding affinity to the irs element compared to PARIS, which suggests a potential regulatory role for ZNF18 in dopamine secretion. The lower K _d of ZNF18 can be explained by its five ZF domains, and the structural information from ZIF268 demonstrates that DNA interaction requires positively charged amino acids, including arginine and histidine residues, in the negatively charged phosphate backbone of DNA (Negi et al., 2011; Yang et al., ²⁰¹¹). The five classical ZF domains have a higher binding affinity than that of a small number of ZF domains. The non‐classical ZF domain is located in the N‐terminal region of PARIS, and the functions of this CCHC domain require further investigation. One possible mechanism is degradation by E3 ligase because over‐expressed PARIS interacts with parkin in the cytosol (Shin et al., 2011). Further investigations are required to elucidate the mechanisms underlying the regulation of ZNF18.

FIGURE 7.

Regulation of dopamine release and neuronal development by ZNF18(ZF1‐5). (a) Interaction of ZNF18(ZF1‐5) with irs and its mutated variants. (b) Interaction of ZNF18(ZF1‐5) with putative nucleotides to determine the promoter element through fluorescence anisotropy. Putative promoter regions of brn2 are classified (F1–F3) to investigate the interaction between ZNF18(ZF1‐5) and promoter sequences. Experiments were performed with 5 nM DNA in 25 mM MOPS (pH 7.4), 50 mM NaCl, 1 mM DTT, and 25°C.

2.3.2. Neuronal developments and brain tumorigenesis through BRN2 regulation

ZBTB20 has five classical ZF domains in the same family as ZNF18 (Figure S9), and ZBTB20 is expressed in the brain and other organs for transcriptional regulation (Sutherland et al., 2009; Tonchev et al., ²⁰¹⁶). ZBTB20 represses BRN2 expression as a master neural transcription factor (Nagao et al., 2016). BRN2 regulates the expression of astrocytes and glioblastoma cells, and immunoprecipitation revealed its regulation in glioblastoma and neuroblastoma (Herbert et al., 2019). In addition, BRN2 is a higher regulator of SCLC than lineage‐specific transcription factors such as achaete‐scute homolog‐like 1 (ASCL1) and NeuroD1 (ND1) (Ishii et al., 2013; Sakaeda et al., ²⁰¹³). Therefore, the network between ZNF18 and BRN2 needs to be further investigated to complete the transcriptional processes.

The promoter sequences of brn2 were positioned between −1204 and −1254 as described in Figure S10. The binding region of the brn2 element has not yet been investigated, and the sequences were classified into three regions based on the number of base pairs. Fluorescence anisotropy (FA) results indicated that ZNF18(ZF1‐5) selectively interacted with brn2‐F3 (5′‐GGGGCTAACGGAAAAG) with K _d  = 1.46 ± 0.03 nM (Figure 7b). The brn2‐F3 sequence is located at −1204 to −1219 bp, and this study provides the first evidence that ZNF18(ZF1‐5) interacts with this region. The FA results suggest that ZNF18(ZF1‐5) may influence the expression of BRN2 through tight interactions; although further studies are needed to clarify whether this interaction suppresses the expression of BRN2. The robust interactions of ZNF18(ZF1‐5) with the brn2 promoter region highlight the potential role of its five ZF domains in specific dsDNA interactions.

3. DISCUSSION

The pivotal roles of ZF proteins have been extensively investigated, although many aspects require further investigation in the areas of molecular biology and biochemistry (Cassandri et al., 2017). Bioinformatic studies have proposed that ZNF18 is located on the hub of the network in critical gene expression (Zhu, 2023). Our results provide insights into the biochemical properties of ZNF18, including its coordination environments and potential implications for gene regulation. Another CX₂CX₁₂HX₃H‐type ZF protein, PARIS, has a similar binding affinity for Zn²⁺ with nanomolar (nM) K _d values. The tight coordination of ZNF18 can be induced by the five ZF domains compared to that of the three ZF domains from PARIS; however, the properties of each ZF domain cannot be ruled out. Some classical and non‐classical ZF domains with two or three ZFs have a tight binding affinity for a nanomolar (nM) range of K _d s to target DNA or RNA (Lee & Michel, 2014; Pritts et al., ²⁰²¹). These ZF proteins have KRAB and SCAN domains for dimerization or multimer complexes, which increase their binding affinity to biological molecules (Al‐Naama et al., 2020). Complex formation driven by KRAB and SCAN can effectively regulate the suppression of gene expression, but further studies are required to uncover the binding affinity between transcription factors and nucleic acids (Huang et al., 2019; Nam et al., ²⁰⁰⁴; Stoll et al., ²⁰²²; Urrutia, ²⁰⁰³).

The structural information obtained through AlphaFold2 provided interesting results for ZNF18(ZF1‐5) as shown in Figure 8. The proposed structure from N‐ to C‐terminus demonstrates the traditional folding of rod‐shaped ZF domains to contact specific DNA. The distribution based on the charge of the side chains in these fingers demonstrates that the palm inside the domains presents a blue color and positive charge due to the Arg, His, and Lys residues, thereby inducing ionic interactions with the phosphate backbones (Figure 8 and Videos S1 and S2). These patterns were well‐organized from the first ZF domain in the N‐terminal region to the fifth ZF domain to enhance binding specificity. Orientational rigidity of the side chains was induced by the Zn²⁺ coordination to Cys and His residues with a tetrahedral geometry. Additionally, hydrophobic residues, including Phe, Leu, and Ile, were positioned between the second Cys and third His, stabilizing the fold of the ZF domains (Eom et al., 2016). These residues are essential for the regulation of positively charged residues among diverse combinations of amino acid sequences, which can modify the number and positions of amino acids (Eom et al., 2016). Discovering the structures of ZF proteins has significant limitations owing to the difficulties of crystallization for X‐ray crystallography, and their application to cryoelectronic microscopy is challenging owing to their low molecular weight (Benjin & Ling, 2020). To overcome these limitations, complex generation through SCAN‐ or KRAB‐induced multimerization is a potential candidate.

FIGURE 8.

Predicted structure of Homo sapiens ZNF18(ZF1‐5) by AlphaFold2. The molecular surface is colored according to the charge distribution: Red and blue indicate negative and positive charges, respectively.

This study suggests that ZNF18 may be involved in transcriptional regulation of multiple genes, including CDK1, IRS, and BRN2 (Figure 9). CDK1 controls the cell cycle and reverses cell cycle arrest, causing uncontrolled cell growth (Liao et al., 2017; Wang et al., ²⁰²³). IRS and BRN2 are associated with various neurodegenerative diseases, including Parkinson's disease, autism, glioblastoma, neuroblastoma, and SCLC (Khaliulin et al., 2024; Sener et al., ²⁰¹⁶). Although we have identified the binding mechanisms of ZNF18 to these genes, it is essential to conduct functional studies to determine its specific roles in cellular and physiological pathways. These investigations could offer a more comprehensive understanding of ZNF18's biological significance and therapeutic applicability.

FIGURE 9.

Physiological function of ZNF18 in regulating the cell cycle, tumor progression, and neurodegeneration.

4. MATERIALS AND METHODS

4.1. General methods and chemicals

The chemicals required to perform various experiments were purchased and used to express ZNF18(ZF1‐5). Luria‐Bertani (LB) broth (Duchefa Biochemie), kanamycin monosulfate (Goldbio), and isopropyl‐β‐D‐1‐thiogalactopyranoside (IPTG, Goldbio) were used for stimulating cell growth and protein expression. Purification was performed using SP‐sepharose (Cytiva) and Superdex75 (Cytiva) coupled with ÄKTA Pure 25 L (Cytiva). The following chemicals were used for the purification process: 3‐(N‐morpholino) propanesulfonic acid (MOPS, Fisher bioreagents), sodium chloride (NaCl, DAEJUNG), magnesium chloride anhydrous (MgCl₂, DAEJUNG), dithiothreitol (DTT, GoldBio), glycerol (Junsei), DNase (Takara), zinc sulfate heptahydrate (Sigma‐Aldrich), and phenylmethylsulfonyl fluoride (PMSF, Thermo Scientific). The supernatant was filtered using an ultracentrifuge (Hanil Science and BECKMAN), and the concentration of the purified ZF domains was measured using a UV–Vis spectrophotometer (Agilent Technologies, Cary 60) in a cuvette (Hellma Analytics).

4.2. Expression and purification

4.2.1. Gene cloning and expression of ZNF18(ZF1‐5)

The znf18(zf1‐5) gene derived from Homo sapiens has a molecular weight of 17,247.66 Da and an isoelectric point of 9.64 (Figure S4a). Codon‐optimized znf18(zf1‐5) was synthesized and inserted into the pET30a(+) vector to construct znf18(zf1‐5)‐pET30a(+), as shown in Figure S3b (NCBI number: BC036096.2). To express ZNF18(ZF1‐5), znf18(zf1‐5)‐pET30a(+) cells were transformed into Rosetta(DE3) cells at 37°C with shaking at 200 rpm for 12–16 h. The cells were further incubated in 500 mL LB broth (50 μg/mL kanamycin) to reach an OD₆₀₀ of 0.6. The expression of ZNF18(ZF1‐5) was induced using IPTG (0.1 mM) with zinc sulfate heptahydrate (0.1 mM) at 18°C with shaking at 180 rpm for 18 h. The induced cell culture was centrifuged at 11,355×g at 4°C for 20 min.

4.2.2. Purification of ZNF18(ZF1‐5)

The cell pellet obtained from 3 L culture of over‐expressed ZNF18(ZF1‐5) was dissolved in buffer A (25 mM MOPS, 50 mM NaCl, 5 mM Mgcl₂, 1 mM DTT, 0.01 μL/mL DNase I, and 0.002 mg/mL PMSF; pH 7.4) and sonicated for 40 min (15 s on and 45 s off). The whole cell lysate was centrifuged at 28,306×g at 4°C for 1 h and filtered through a 0.22 μm membrane syringe filter. The supernatant was applied to an SP‐Sepharose column, washed with activating buffer B (25 mM MOPS, 50 mM NaCl, 1 mM DTT, and 5% glycerol; pH 7.4), and eluted using a linear gradient of 50–1000 mM NaCl. The presence of ZNF18(ZF1‐5) was confirmed by 15% sodium dodecyl sulfate–polyacrylamide electrophoresis (SDS‐PAGE) followed by Coomassie brilliant blue staining. Fractions containing the target proteins were concentrated at 4°C and 2095×g using a 10 kDa cut‐off membrane filter (Merck Millipore). The concentrated proteins were further purified by loading them onto a Superdex 75 column equilibrated with activating buffer B. Finally, the samples were reconcentrated using a 10 kDa cut‐off membrane filter at 4°C and 2095×g. The concentrated eluate was subsequently quantified via UV–Vis spectrometry, with extinction coefficients (ε₂₈₀) of 13,575 M⁻¹ cm⁻¹ for ZNF18(ZF1‐5), and stored at −88°C.

4.3. Metal binding studies

4.3.1. Preparation of apo‐ZNF18(ZF1‐5)

Apo‐ZNF18(ZF1‐5) was prepared by adding 2 M HCl to ZNF18(ZF1‐5) at 25°C for 20 min to eliminate coordination with zinc ions. The mixture was filtered using a 3 kDa cut‐off membrane filter to eliminate metal ions. Apo‐ZNF18(ZF1‐5) was washed at least five times with exchange buffer C (100 mM HEPES, 150 mM NaCl, and 1 mM DTT, pH 7.4). The concentration of apo‐ZNF18(ZF1‐5) was calculated using UV–vis spectrometry with extinction coefficients (ε₂₈₀) of 12,910 M⁻¹ cm⁻¹.

4.3.2. Co²⁺ titration and Zn²⁺ back‐titration

The binding affinities between ZNF18(ZF1‐5) and the metal ions were measured based on d–d transition Abs using a UV–Vis spectrophotometer. All experiments were performed in an anaerobic chamber (Coy Laboratory). Cobalt(II) chloride hexahydrate was added to apo‐ZNF18(ZF1‐5) (10 μM) in the buffer D (100 mM HEPES, 150 mM NaCl, and 1 mM TCEP; pH 7.4). The absorption band at 643 nm was monitored; the results were fitted using a 1:1 binding model with Origin software. The concentration of the reduced protein was obtained by measuring the Abs of the d–d transition at 643 nm and the five Co²⁺ coordinated extinction coefficients (ε₆₄₃: 3250 M⁻¹ cm⁻¹) using the following formula (Lee et al., 2011):

$$\%\text{Reduced}=\frac{\left[{\text{Protein}}_{\text{reduced}}\right]}{\left[{\text{Protein}}_{\text{total}}\right]}=\frac{\left[\mathrm{Co}\left(\mathrm{II}\right)-\mathrm{ZNF}18\left(\mathrm{ZF}1-5\right)\right]}{\left[\mathrm{apo}-\mathrm{ZNF}18\left(\mathrm{ZF}1-5\right)\right]}$$

The dissociation constant between ZNF18(ZF1‐5) and Zn²⁺ was measured via back‐titration with zinc sulfate heptahydrate (Sigma‐Aldrich) after adding it to 10 μM Co²⁺–ZNF18(ZF1‐5). The decrease in Abs at 643 nm following Zn²⁺ replacement was measured and fitted using a 1:1 binding model with OriginPro 2019b and KaleidaGraph 5.0.

4.3.3. Fe²⁺ and Fe³⁺ titration

Ammonium iron(II) sulfate was added to 10 μM apo‐ZNF18(ZF1‐5) in buffer D. The absorption band was monitored at 309 nm; the results were fitted using the 1:1 binding model with Origin. Iron(III) chloride hexahydrate was added to 10 μM apo‐ZNF18(ZF1‐5) in buffer D in an anaerobic chamber.

4.3.4. Binding equation model (1:1) in metal binding studies

The dissociation constants (K _d s) between the protein and metal ions were obtained using a 1:1 binding model. The overall calculations were performed using the following equation:

$$M+\mathrm{P}\rightleftharpoons \mathrm{MP}$$

$$K_d=\frac{\left[M\right]\left[P\right]}{\left[\mathit{MP}\right]}$$

$$F_{\text{bound}}=\frac{M_{\text{total}}+P_{\text{total}}+K_d-\sqrt{{\left(M_{\text{total}}+P_{\text{total}}+K_d\right)}^2-4M_{\text{total}}{P}_{\text{total}}}}{2P_{\text{total}}}$$

where M indicates the concentration of metal ions and P indicates the protein concentration.

4.4. Binding affinity of the protein–DNA interaction

4.4.1. FA for measuring protein‐nucleic acid interactions

FA values were measured using an FP8300 spectrofluorometer (JASCO) with cuvettes (JASCO, J/3 type). 6‐Carboxyfluorescein (6‐FAM) fluorescein dye was attached at the 3′‐end of the DNA (Integrated DNA Technology; IDT). Excitation and emission absorbances were measured at 499 and 518 nm, respectively. The measurements for all experiments were performed with the following parameters: bandwidth, 5 nm; response, 0.5 s; and PMT voltage, 700 V at 25°C. The fluorescence‐tagged 5 nM DNA in 1,600 μL of buffer E (25 mM MOPS, 50 mM NaCl, 1 mM DTT, and pH 7.4) was titrated by adding proteins, and anisotropy was measured thrice after 2 min of reaction time.

4.4.2. Binding equation model (1:1) in the protein–DNA interaction

The dissociation constants (K _d s) between the protein and DNA were obtained using a 1:1 binding model. The overall calculations were performed using the following equation:

$$F_{\text{bound}}=\frac{r-r_{\text{free}}}{r_{\text{bound}}-r_{\text{free}}}$$

where F _bound (fraction bound) indicates the fraction of the bound protein to DNA, r _free indicates the anisotropy of unbound DNA, and r _bound indicates the anisotropy of protein‐bound DNA when saturated. The dissociation constant (K _d s) was calculated using the following 1:1 binding model (Hwang et al., 2023):

$$P+D\rightleftharpoons \mathit{PD}$$

$$K_d=\frac{\left[P\right]\left[D\right]}{\left[\mathit{PD}\right]}$$

$$F_{\text{bound}}=\frac{P_{\text{total}}+D_{\text{total}}+K_d-\sqrt{{\left(P_{\text{total}}+D_{\text{total}}+K_d\right)}^2-4P_{\text{total}}{D}_{\text{total}}}}{2D_{\text{total}}}$$

where P indicates the protein concentration; and D indicates the DNA concentration.

AUTHOR CONTRIBUTIONS

Soyeon Park: Conceptualization; methodology; investigation; formal analysis; writing – original draft; writing – review and editing. Yunha Hwang: Conceptualization; methodology; investigation; validation; formal analysis; writing – original draft; writing – review and editing. Ki Seong Eom: Conceptualization; methodology; formal analysis; validation; writing – review and editing. Jin Sung Cheong: Conceptualization; methodology; writing – review and editing; formal analysis; validation. Seung Jae Lee: Project administration; funding acquisition; supervision; conceptualization; investigation; resources; validation; methodology; writing – original draft; writing – review and editing; visualization.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest requiring disclosure.

Supporting information

DATA S1. Supporting information.

VIDEO S1. Front view of the AlphaFold2‐predicted Homo sapiens ZNF18(ZF1‐5) structure.

VIDEO S2. Side view of the AlphaFold2‐predicted Homo sapiens ZNF18(ZF1‐5) structure.

Park S, Hwang Y, Eom KS, Cheong JS, Lee SJ. Controlling gene expression through five zinc finger domains of ZNF18 . Protein Science. 2025;34(9):e70278. 10.1002/pro.70278

DATA AVAILABILITY STATEMENT

All experimental data are available within the article and ESI. The data that support the findings of this study are available from the corresponding author upon reasonable request.