Conformational dynamics in specialized C₂H₂ zinc finger domains enable zinc‐responsive gene repression in S. pombe

Abstract

Loz1 is a zinc‐responsive transcription factor in fission yeast that maintains cellular zinc homeostasis by repressing the expression of genes required for zinc uptake in high zinc conditions. Previous deletion analysis of Loz1 found a region containing two tandem C₂H₂ zinc‐fingers and an upstream “accessory domain” rich in histidine, lysine, and arginine residues to be sufficient for zinc‐dependent DNA binding and gene repression. Here we report unexpected biophysical properties of this pair of seemingly classical C₂H₂ zinc fingers. Isothermal titration calorimetry and NMR spectroscopy reveal two distinct zinc binding events localized to the zinc fingers. NMR spectra reveal complex dynamic behavior in this zinc‐responsive region spanning time scales from fast 10⁻¹²–10⁻¹⁰ to slow >10⁰ s. Slow exchange due to cis‐trans isomerization of the TGERP linker results in the doubling of many signals in the protein. Conformational exchange on the 10⁻³ s timescale throughout the first zinc finger distinguishes it from the second and is linked to a weaker affinity for zinc. These findings reveal a mechanism of zinc sensing by Loz1 and illuminate how the protein's rough free‐energy landscape enables zinc sensing, DNA binding and regulated gene expression.

1. INTRODUCTION

Zinc is an essential nutrient for all living organisms and is the second most abundant trace metal in the human body (Chasapis et al., 2012). Zinc plays a catalytic role as a Lewis acid in many enzymes and confers structural stability in a variety of proteins. One of the most prevalent structural domains in eukaryotic transcriptional factors is the “zinc finger” (ZF) (Bonchuk & Georgiev, 2024; Klug, ²⁰¹⁰). Classical C₂H₂‐type ZFs are ~30‐residue domains that form a ββα fold stabilized by a single zinc ion that is coordinated through side chain atoms of conserved cysteine and histidine residues. C₂H₂ ZFs have a consensus sequence of φ‐X‐Cys‐X(2–5)‐Cys‐X(3)‐φ‐X(5)‐φ‐X(2)‐His‐X(2–5)‐H, where φ is a hydrophobic amino acid and X is any amino acid. The conserved hydrophobic residues stabilize the core of each individual ZF domain. In most C₂H₂ ZF proteins, two or more ZFs are arranged in tandem through a canonical TGE(R/K)P linker (Nagaoka et al., 2001; Wolfe et al., ²⁰⁰⁰).

Because canonical C₂H₂ ZF domains typically form very stable scaffolds and have high zinc binding affinity (equilibrium dissociation constants K _d ranging from 10⁻⁹ to 10⁻¹² M) (Kluska et al., 2018; Krizek et al., ¹⁹⁹³; Padjasek et al., ²⁰²⁰), zinc binding by these domains is generally associated with a structural role (Bonchuk & Georgiev, 2024; Laity et al., ²⁰⁰¹; Padjasek et al., ²⁰²⁰). However, in a few transcription factors that help cells to maintain zinc homeostasis ZF domains have been found to function as sensors of intracellular zinc levels (Bird et al., 2003; Choi & Bird, ²⁰¹⁴; Potter et al., ²⁰⁰⁵). In mammals, fish, and reptiles the metal‐responsive transcription factor (MTF‐1) activates target gene expression when zinc is in excess (Heuchel et al., 1994). The six C₂H₂‐type ZFs, part of the DNA‐binding domain of MTF‐1, show increased DNA binding activity upon zinc supplementation (Chen et al., 1998; Heuchel et al., ¹⁹⁹⁴). Experimental evidence suggests that differential zinc‐binding affinity by the six ZF domains underlies zinc‐sensing and gene regulation by MTF‐1. However, structure–function studies to identify the metalloregulatory finger/subset of fingers of MTF‐1 have yielded conflicting observations (Giedroc et al., 2001; Guerrerio & Berg, ²⁰⁰⁴; Potter et al., ²⁰⁰⁵).

Another well‐characterized zinc‐responsive transcription factor is the yeast Saccharomyces cerevisiae protein Zap1, which upregulates the expression of genes involved in zinc uptake under zinc‐limiting conditions. Zap1 is regulated at both the transcriptional and post‐translational levels by zinc and contains multiple zinc‐regulatory domains that can function independently of each other (Wilson & Bird, 2016; Zhao et al., ¹⁹⁹⁸). The most widely studied transcriptional activation domain (AD2) contains two non‐canonical, CWCH₂‐type zinc fingers. Upon zinc binding and folding, the zinc finger pair interact through a network of inter‐finger hydrophobic contacts to form a single structural unit, which is hypothesized to be responsible for reduced AD2 activity and transcription (Bird, 2003; Qiao et al., ²⁰⁰⁶; Wang et al., ²⁰⁰⁶).

In Schizosaccharomyces pombe, the transcription factor Loz1 (Loss of Zinc sensing 1) plays a central role in zinc homeostasis (Yao et al., 2023). Under conditions of excess zinc, Loz1 is recruited to the Loz1 response promoter element (LRE, 5′‐CGNMGATCNTY‐3′; N, any base; M, A or C; Y, pyrimidine), which is both required and sufficient for Loz1‐mediated gene repression (Wilson et al., 2019). Loz1 represses the expression of ~30 genes, including zrt1, which encodes a high‐affinity zinc uptake transporter, and adh1AS, an antisense transcript that controls the expression of the zinc‐dependent alcohol dehydrogenase gene adh1 (Wilson et al., ²⁰¹⁹). Loz1 also negatively regulates its own expression (Corkins et al., 2013; Wilson et al., ²⁰¹⁹). Deletion of the N‐terminal 425 residues of this protein, predicted to be unstructured, leaves a C‐terminal fragment (residues 426–522, here termed Loz1AZZ), containing a pair of classical C₂H₂ ZFs at the extreme C‐terminus, and an upstream 40‐amino acid “accessory domain” that is sufficient for DNA binding in vitro and can confer partial zinc‐dependent repression in vivo (Corkins et al., 2013; Ehrensberger et al., ²⁰¹⁴). Mutations in the ZF domains that disrupt DNA binding abolish the zinc‐dependent activity of Loz1 (Ehrensberger et al., 2014).

Given that the Loz1 zinc‐responsive domain maps to a region containing two C₂H₂‐type ZFs and an uncharacterized “accessory domain,” the goal of this study was to investigate whether each of these domains has a role in sensing zinc ions. The sequence of the canonical ZFs of Loz1 is conserved across Schizosaccharomyces species, but there is little sequence conservation in the accessory domain (Wilson & Bird, 2016). On the other hand, Ser, His and Asp are known to coordinate zinc ions with modest affinities (Auld, 2001) and are abundant in the semi‐conserved accessory domain. We present calorimetry and heteronuclear NMR studies of the Spo Loz1 zinc‐responsive region, which show the ZF domains as the likely zinc sensors. NMR data show that the ZFs exhibit dynamics over a range of time scales and bind zinc with differing affinities. In addition, NMR chemical shift analysis and homology modeling provide insights into the mechanism of recognition of sequence‐specific recognition of LRE DNA. These findings suggest a mechanism by which protein dynamics keep the protein in an inactive state, while zinc binding‐coupled folding of Loz1 zinc finger 1 (ZF1) is a determinant for specific DNA binding and gene repression.

2. RESULTS

2.1. Loz1 zinc finger domains exhibit differential affinity toward zinc

To characterize the stoichiometry and thermodynamics of zinc binding to Loz1 we used isothermal titration calorimetry (ITC) to measure heat evolved upon titrating zinc into zinc‐depleted protein. Homology modeling with the Loz1AZZ sequence (Figure 1a) failed to identify structured templates for the accessory domain but returned high‐confidence models for the individual ZFs (Figure 1b). Loz1 constructs with and without the accessory domain (Loz1AZZ and Loz1ZZ, respectively; Figure 1a) were recombinantly expressed in Escherichia coli BL21(DE3) cells and purified in buffers supplemented with zinc (100 μM). Zinc‐free “apo‐Loz1” was obtained by treatment with EDTA followed by dialysis against Chelex‐treated, degassed buffers containing TCEP as a reducing agent to prevent cysteine oxidation. ITC titrations of zinc chloride solutions into apo Loz1AZZ resulted in biphasic thermograms that could be fit by a model with two independent sites (Figure 1c). The zinc affinity for the first binding event was found to be in the nanomolar range (K _d1 = 7 ± 10 nM), whereas the affinity for the second zinc‐binding event was 1.6 orders of magnitude weaker (K _d2 = 300 ± 100 nM). ITC experiments with the construct lacking the accessory domain (Loz1ZZ, res. 461–522) yielded similar thermograms and affinities as those for Loz1AZZ (K _d1 = 4 ± 6 nM and K _d2 = 350 ± 90 nM; Figure 1d), indicating that if zinc binds to the accessory domain, it does so weakly without saturating over the sampled concentration range. The large differential in zinc binding affinity between ZF1 and ZF2, and lack of evidence for zinc binding to the accessory domain, suggest that the two ZFs would respond differently to changing zinc concentrations in the cell, hinting at a role for in zinc sensing by Loz1.

FIGURE 1.

The Loz1 zinc‐responsive element binds two zinc ions with different affinities. (a) The amino acid sequence of the C‐terminal zinc‐responsive element of Loz1 with Zn²⁺‐coordinating residues in blue and conserved hydrophobic core residues in green. The secondary structure schematic represents the predicted fold of the ZF domains. Residues at the −1, 2, 3 and 6 positions of each helix contact DNA bases in canonical complexes (Wolfe et al., 2000). The N‐terminal 425 residues of Loz1 have no recognizable domains. Constructs used in this study are termed Loz1AZZ (residues 426–522) and Loz1ZZ (residues 461–522). (b) Homology model built using as templates (Figure S1) the crystal structures of DNA‐bound ZBTB38 zinc finger 6 (for ZF1) and Y3 zinc finger of Human YY1 protein (for ZF2); colors as in a. (c, d) Representative calorimetric titrations of ZnCl₂ into zinc‐depleted Loz1AZZ (c) and Loz1ZZ (d). Top panel, the baseline‐corrected thermogram; bottom panel, integrated heat (black) and best fit to a binding model with two independent sites (red). Best fit parameters are Loz1AZZ: N ₁ = 1.09 ± 0.13, K _d1 = 7 ± 10 nM, ΔH ₁ = −21 ± 1 kcal mol⁻¹, n ₂ = 1.00 ± 0.18, K _d2 = 300 ± 100 nM, ΔH ₂ = −29 ± 3 kcal mol⁻¹. Loz1ZZ: N ₁ = 0.73 ± 0.07, K _d1 = 3 ± 4 nM⁻¹, ΔH ₁ = −8.02 ± 0.42 kcal mol⁻¹, n ₂ = 1.19 ± 0.08, K _d2 = 345 ± 95 nM, ΔH ₂ = −10.98 ± 0.45 kcal mol⁻¹.

2.2. NMR spectra reveal a disordered accessory domain and slow cis‐trans X‐pro peptide bond isomerization in the ZF1–ZF2 linker

Heteronuclear NMR spectroscopy provided evidence for conformational heterogeneity in the Loz1 zinc‐responsive element. Loz1AZZ was uniformly labeled with ¹⁵N, or ¹³C and ¹⁵N and purified to homogeneity as evidenced by electrospray ionization mass spectrometry under native conditions, which revealed a single species with a mass corresponding to Loz1AZZ bearing >99% ¹⁵N enrichment and two bound zinc atoms (Figure S2). Despite chemical and compositional homogeneity, 2D ¹H‐¹⁵N correlation spectra of zinc‐bound Loz1AZZ (426–522) featured a doubling of several of the resonances, with minor peak intensities of ~20% relative to the major peaks (Figure 2a). In addition, far fewer amide resonances were observed than would be expected for a 97‐residue protein, and many resonances were much broader than would be expected for the small 11.2 kDa protein construct. At lower temperatures (5°C), additional broad signals could be observed in the random coil region of the spectrum (Figure S4), while resonance doubling persisted. Finally, amide resonances exhibited highly variable intensities and line widths, indicative of exchange broadening.

FIGURE 2.

The Loz1 zinc responsive element exhibits an unstructured accessory domain and slow exchange due to cis‐trans isomerization in the TGERP linker. (a) Two‐dimensional ¹H‐¹⁵N HSQC spectrum of Loz1AZZ with assigned backbone amide resonances from the zinc finger region. Doubling of signals is observed for residues in ZF2 and the helical region of ZF1 (red arrows). Absence of signals from the accessory domain indicates it is poorly structured. (b) Strips from 3D CC(CO)NH‐TOCSY (i and iii) and HNCACB (ii and iv) spectra highlighting C^α and C^β chemical shifts of Pro495 and Phe496 residues in the trans and cis conformers. (c) Chemical shift differences (δ_trans‐δ_cis) between the trans and cis Arg494‐Pro495 conformations for affected amide proton and nitrogen resonances in Loz1AZZ. Residues for which only one signal was observed are marked with a gray circle. These data indicate that the structural effects of cis‐trans isomerization extend beyond the ZF1–ZF2 linker. (d) Amide proton temperature coefficients reveal differences in solvent exposure of backbone amides trans (black circles) and cis (red diamonds) conformations of the Arg494‐Pro495 peptide bond.

Conventional double‐ and triple‐resonance NMR experiments enabled us to assign ~95% of the resonances observed in two‐dimensional ¹H‐¹⁵N HSQC spectra (Figure 2a). Most of the resonances could be assigned to the two ZFs, while no resonances were assigned to residues in the accessory domain. Backbone resonance assignments included 50 out of 55 (91%) amides in the zinc finger and linker region, 80% of the C′ and 95% of the C^α resonances. Some residues from predicted loop regions of the zinc fingers (T471, S478, S511 and N508) and S481 from ZF1 helix could not be assigned, likely due to exchange broadening. Three‐dimensional heteronuclear TOCSY‐type experiments (Table S1) were used to obtain additional sidechain proton and carbon assignments and 80% of aliphatic protons from the zinc finger residues could be assigned. Secondary chemical shifts were generally consistent with the predicted ββα fold for each zinc finger (Figure S5). NMR spectra of the construct lacking the accessory domain (Loz1ZZ) were nearly superimposable on those of Loz1AZZ (Figure S3), with differences limited to small perturbations of resonances from adjacent residues Val466 and Arg467, and the absence of two unassigned signals. These observations indicate that the accessory domain exhibits intermediate time scale disorder and is dispensable for the structural integrity of the zinc fingers and their zinc‐sensing function.

Doubling of the amide resonances results from slow cis‐trans isomerization of Arg‐Pro peptide bond in the TGERP linker between the two zinc finger domains. Doubled resonances mapped to the end of the ZF1 helix, the TGERP⁴⁹⁵ linker, and many residues in ZF2. The chemical shift difference between C^β and C^γ of Pro⁴⁹⁵ is consistent with a trans‐conformation of the major state; the minor state Pro C^γ resonance was not observed, but the C^β chemical shift is comparable to that for cis‐conformation (Schubert et al., 2002; Shen & Bax, ²⁰¹⁰) (Figure 2b). This assignment is substantiated by the significant upfield shift in ¹H and ¹⁵N resonances of Arg494 (i − 1 residue) in the cis‐state (Figure 2c). Further, the upfield shift (δ_trans‐δ_cis = 1.6 ppm) of C^α resonance for the i − 2 residue is similar in sign and magnitude to the C^α chemical shift change (δ_trans‐δ_cis = 1.2 ppm) of the isomerizing Pro, diagnostic of the cis state (Alderson et al., 2018). To support the conclusion that X‐Pro isomerization at Pro495 is responsible for the observed signal doubling we prepared samples in which each of the two proline residues in Loz1AZZ was replaced with Ala: P495A and P480A. Two‐dimensional NMR spectra of P480A‐Loz1 exhibited the same signal doubling, while only one set of signals is observed for the P495A variant (Figure S6). Although canonical DNA‐binding C₂H₂ ZF proteins feature TGE(R/K)P linkers (Nagaoka et al., 2001; Persikov et al., ²⁰⁰⁹), slow cis‐trans isomerization has not previously been reported for this class of proteins. Substitution of Arg494 in the linker with Lys also did not eliminate the slow exchange behavior (Figure S7). No exchange cross‐peaks were observed in ZZ exchange (Farrow et al., 1994) spectra probed up to 900 ms, consistent with exchange rates much slower than 1 s⁻¹ (Grathwohl & Wüthrich, 1981; Reimer et al., ¹⁹⁹⁸).

NMR spectra of major and minor states reveal that structural perturbations from cis‐trans isomerization of the Arg494‐Pro495 peptide bond are not limited to the linker. Amide chemical shift perturbations from cis‐trans isomerization (i.e., signal doubling) are observed throughout ZF2 to the C‐terminal residue Leu522 (Figures 2c and S8). We measured amide proton temperature coefficients (Δδ/ΔT in ppb K⁻¹) by recording ¹H‐¹⁵N HSQC spectra at five‐degree intervals from 278 to 308 K (Figure S9); these report on secondary structure stability via the hydrogen bond strength and its thermal expansion, with smaller Δδ/ΔT being associated with stronger intramolecular hydrogen bonds, whereas the largest negative Δδ/ΔT are associated with hydrogen bonding to solvent (Baxter & Williamson, 1997; Cordier & Grzesiek, ²⁰⁰²; Trainor et al., ²⁰²⁰). Temperature coefficients of amide protons in the cis and trans species (Figure 2d) show the largest differences for residues in the β‐hairpin region of ZF2, consistent with their proximity; however, just as many amides experience a positive and negative change in Δδ/ΔT. These data indicate that cis‐trans isomerization of the X‐Pro peptide bond in the TGERP linker has structural consequences on the entirety of ZF2, not just the linker.

2.3. ZF1 has lower zinc affinity than does ZF2

To identify the zinc finger domain with weaker zinc affinity, making it a candidate for direct zinc sensing, we performed NMR‐based zinc titrations (Figure 3). [U‐¹⁵N]‐Loz1AZZ was stripped of bound zinc using EDTA for ITC experiments. The ¹H‐¹⁵N HSQC spectrum of zinc‐free Loz1AZZ exhibited a high degree of line broadening, with few well‐resolved signals (Figure 3b, left). Most signals cluster in a narrow range of ¹H frequencies of 8–8.5 ppm, with little signal dispersion in both dimensions, characteristic of aggregation in the sample, or conformational heterogeneity on a ms‐μs time scale (Rehm et al., 2002). Titrating zinc into the protein solution resulted in the incremental appearance of signals corresponding to the zinc‐bound, folded form of the protein, while broad resonances from the apo form diminished (Figure 3b, center and right). At one equivalent of added zinc, signals from ZF2 residues were fully populated (Figure 3c). Signals from ZF1 did not completely saturate even after the addition of 2 molar equivalents of zinc. These data indicate that ZF1 has the lower affinity zinc site identified by ITC (Figure 1).

FIGURE 3.

Loz1 ZF1 coordinates zinc with weaker affinity. (a) Loz1AZZ amino acid sequence with a schematic representing the predicted secondary structure. (b) ¹H‐¹⁵N HSQCs of Loz1AZZ in zinc‐free apo form (left), 0.5 equivalents of zinc (middle), and equimolar zinc (right). Apo‐Loz1AZZ backbone amide signals are clustered around 8.3 ppm, diagnostic of a disordered protein. Upon addition of limiting zinc, resonances from folded ZF2 appear in the spectrum whereas ZF1 resonances appear upon titration with additional zinc. (c) Representative plots of normalized intensities of residues from ZF1 and ZF2 against [Zn]:[Loz1AZZ] molar ratio. The dashed line corresponds to the equimolar zinc‐to‐protein ratio (third panel in b) at which point ZF2 intensities saturate while ZF1 intensities continue to increase, indicating weaker zinc binding to ZF1.

2.4. Loz1 exhibits heterogeneity on fast and intermediate time scales

Doubling of NMR signals due to cis‐trans isomerization revealed structural heterogeneity on slow time scales (>1 s), while non‐uniform intensities and broad lines were indicative of internal motions on intermediate time scales (μs‐ms). We first probed fast (ps‐ns) time scale dynamics of the major state of Loz1AZZ by recording ¹⁵N spin relaxation rates R ₁, R _1ρ and {¹H}‐¹⁵N heteronuclear NOE spectra; R ₂ values were computed from R ₁‐corrected R _1ρ values (Figures 4 and S10). The low signal‐to‐noise for the data from the cis state precluded its analysis. Heteronuclear NOE values for trans Loz1AZZ were generally uniform for residues in the predicted helices, with values near the 10% trimmed‐mean for the entire dataset of 0.63 at 600 MHz, excluding linker residues. Values decreased as far as 0.44 for residues in the linker and the zinc‐binding loops, indicating increased ps‐ns flexibility for those regions. The R ₁/R ₂ ratio was found to differ between ZF1 and ZF2, suggestive of anisotropic tumbling (Bruschweiler et al., 1995; Fushman et al., ¹⁹⁹⁹; Tjandra et al., ¹⁹⁹⁵; Tjandra et al., ¹⁹⁹⁷). However, similar data collected on Loz1ZZ exhibited similarly elevated R ₂ values for ZF1, despite the absence of the accessory domain, suggesting that the elevated R ₂ arises from exchange broadening by an underlying dynamic process at the μs‐ms timescale.

FIGURE 4.

¹⁵N relaxation data reveal elevated R ₂ values for ZF1. Loz1AZZ {¹H}‐¹⁵N heteronuclear NOE, R ₁, R ₂ and R ₂/R ₁ ratios from data recorded at 600 MHz. Error bars for the heteronuclear NOE data were obtained from peak height uncertainties based on average noise levels in the NMR spectra. Uncertainties in R ₁ and R ₂ values were obtained from replicates. Dashed lines represent the 10% trimmed‐mean values excluding the linker region. The hetNOE and R ₁ values could be described by a uniform trimmed mean value, whereas ZF1 and ZF2 exhibited different mean R ₂ values.

Amide ¹⁵N Carr–Purcell Meiboom–Gill relaxation dispersion (CPMG RD) (Hansen et al., 2008) experiments at two magnetic fields revealed intermediate exchange dynamics in Loz1AZZ. Indeed, the effective transverse relaxation rate R _2,eff of many amide resonances were found to depend strongly on the rate at which the CPMG refocusing pulses were applied (i.e., ν_CPMG) (Figure 5). These dispersion curves, with exchange contributions to transverse relaxation R _ex as high as 13 s⁻¹, mapped to all assigned amides in ZF1 and the linker region, as well as Asp499 and Gly504 in ZF2 (Figure 5a). Weak (<2 s⁻¹), or no dispersions were observed in most of ZF2. Similar relaxation dispersion profiles were observed for the same residues for Loz1ZZ, indicating that these dynamics are inherent to ZF1, not a consequence of transient interactions with the accessory domain (Figure S10).

FIGURE 5.

Loz1 ZF1 undergoes complex dynamics at μs‐ms time scales. (a) Exchange contributions to transverse relaxation R _ex for Loz1AZZ, recorded at 800 MHz, obtained by two‐field fitting of CPMG dispersion curves for each residue with the Carver‐Richards two‐site exchange model. An asterisk identifies assigned residues for which good fits of the dispersion curves could not be obtained, or had R _ex values close to zero; the dashed line is the cut‐off below which dispersions were considered unreliable. (b) Left, global fit of a group of residues undergoing “slow” exchange with a k _ex of 532 s⁻¹. Right, global fit of residues undergoing “fast” intermediate exchange, with a fitted k _ex of 1328 s⁻¹. (c) Homology model of ZF1 with “slow” exchanging residues in blue, and “fast” exchanging residues in orange. Residues that are either unassigned or did not yield reliable fits to a two‐state exchange model are gray.

Relaxation dispersions in Loz1AZZ could not be explained by concerted exchange between two well‐defined states. Global fitting of dispersion curves for ZF1 with the Carver–Richards two‐state exchange model (Kleckner & Foster, 2012) resulted in poor fits and large residuals. However, residues could be separated into two groups, each of which could be fit to a global two‐site model—the region close to the zinc coordinating region of the ZF with a faster exchange rate (k _ex = 530 ± 40 s⁻¹) and slower exchange rate (k _ex = 1330 ± 40 s⁻¹) for residues away from the zinc coordinating site. (Figure 5b,c). This behavior is consistent with two independent motional modes, or perhaps exchange between more than two states, the fitting of which is beyond the precision of the available data and Carver–Richards approach (Kleckner & Foster, 2011). These observations highlight a strong parallel between weak zinc affinity and intermediate time‐scale conformational exchange.

2.5. DNA‐bound Loz1AZZ exhibits a single conformation

To test whether the cis‐trans isomerization observed in free Loz1AZZ persists when bound to DNA, 2D ¹H‐¹⁵N NMR spectra were recorded in the presence of a 14 base pair DNA duplex containing the LRE element (5′‐GC GACGATC ATTGG‐3′). The Loz1AZZ–LRE complex was assembled in NMR buffer by titrating [U‐¹⁵N]‐Loz1AZZ into the dsDNA. Upon the formation of the DNA complex, chemical shift perturbations were observed throughout the spectrum (Figure 6a), consistent with an extensive protein‐DNA surface and remodeling of the inter‐finger interface (Foster et al., 1998). Increased uniformity of amide signal intensity is consistent with quenching of μs‐ms exchange upon binding to DNA. Doubling of amide resonances was also not observed, consistent with a canonical trans Arg‐Pro peptide bond conformation in the protein‐DNA complex. As in the absence of DNA, fewer than 95 protein amide resonances were observed, indicating that the accessory domain does not adopt a well‐defined structure when bound to an LRE. Homology modeling of the two zinc finger domains and docking to a model DNA duplex (Figure 6b) places residues at the recognition positions −1, 2, 3 and 6 in position to recognize the consensus nucleotides in the LRE with Ile485 and Val509 packing in the inter‐finger and DNA interface.

FIGURE 6.

DNA binding quenches both fast and slow dynamics in Loz1AZZ. (a) Overlay of ¹H‐¹⁵N HSQC spectra Loz1AZZ in the absence (black) and presence (green) of near equimolar LRE DNA. The DNA substrate sequence is shown with the consensus nucleotides in bold. Indicated amide assignments in the bound state are inferred from proximity to isolated signals in the spectra of the free protein. The absence of signal doubling indicates the cis Arg‐Pro conformer is no longer populated, while more uniform peak shapes are indicative of reduced motion on the μs‐ms time scale. (b) Model of zinc‐loaded Loz1AZZ bound to the consensus DNA. Side chains of residues at the −1, 2, 3 and 6 positions of each helix are shown as sticks, and the consensus base positions in the bottom strand are indicated in bold.

3. DISCUSSION

We used NMR and binding assays to characterize the domains of Loz1 that are sufficient for its zinc‐responsive function. Calorimetric and NMR titrations localized the likely zinc‐sensing function to the first C₂H₂ zinc finger in Loz1. Although primary sequence considerations and ITC data suggest that the accessory domain may also respond to zinc concentrations, the absence of the titratable signature for the accessory domain suggests it is not the major determinant. On the other hand, ZF1 binds zinc weakly (K _D ~ 300 nM) (Figures 1 and 3) compared with ZF2, and to other ZF domains whose zinc binding affinity has been characterized (typical K _D values range from 10⁻⁹ to 10⁻¹² M) (Kluska et al., 2018; Padjasek et al., ²⁰²⁰; Rich et al., ²⁰¹²). Moreover, ZF1 exhibits complex dynamics on the timescale of 10⁻⁶–10⁻³ that leads to line broadening and relaxation dispersion in NMR experiments (Figure 5) (Kleckner & Foster, 2011). Motions on these timescales are relevant to processes that may enable non‐concerted binding and release of zinc, such as histidine switching (Zhu et al., 2017) or ligand flipping (Chandra et al., 2007). These atypical properties reveal a likely mechanism for zinc‐responsive gene repression wherein elevated cellular zinc levels promote zinc‐dependent folding of ZF1. The structured sensory finger (ZF1), together with the more weakly responsive accessory domain, could then cooperate with the structural finger (ZF2) to make high‐affinity base‐specific and backbone interactions with LRE DNA sequence.

In addition to the dynamic properties that make ZF1 a sensor, we found evidence for slow cis‐trans isomerization of the Arg‐Pro peptide bond in the TGERP linker between the two zinc fingers. Low‐level population of a cis conformer is not unusual in flexible polypeptides containing Pro residues (Alderson et al., 2018; Reimer et al., ¹⁹⁹⁸). However, this behavior has not previously been noted for other C₂H₂ zinc finger proteins with canonical TGE(K/R)P linkers (Bonchuk & Georgiev, 2024).

The TGERP linker leading into ZF2, in place of the more common TGEKP, and the replacement of a Leu at the third hydrophobic position with a Met are notable differences compared with the canonical C₂H₂ ZF. However, replacing Arg494 with Lys did not eliminate the isomerization of Pro495. Conceivably, the more flexible Met sidechain could alter the dynamics and stability of ZF2, allowing for more facile Pro isomerization; however, the third zinc finger of Xenopus laevis TFIIIA also features a Leu/Met substitution at that position and its NMR spectra provide no evidence that the substitution alone is destabilizing (Foster et al., 1998). NMR‐based analyses of other non‐canonical C₂H₂ ZFs similarly have not reported cis‐trans isomerization in the linkers (Bernard et al., 2013; Boisvert et al., ²⁰²²; Lee et al., ²⁰⁰⁶; Nunez et al., ²⁰¹¹; Schmiedeskamp et al., ¹⁹⁹⁷; Tochio et al., ²⁰¹⁵; Wienk et al., ²⁰⁰⁹).

NMR studies on classical C₂H₂ zinc finger proteins have generally found the ZF domains to not interact significantly in the absence of DNA, but because of restraints imposed by the short linkers tumble anisotropically, with correlated diffusion (Bédard et al., 2017; Bruschweiler et al., ¹⁹⁹⁵; Tsui et al., ²⁰⁰⁰). In the major trans conformation of Loz1AZZ the zinc‐fingers exhibit different R ₂/R ₁ ratios, consistent with lack of inter‐finger interactions. We estimate the population of the cis‐proline conformation to be ~15%–20% based on peak intensities, which is higher than what it typically found in model peptides and unfolded proteins (Alderson et al., 2018; Reimer et al., ¹⁹⁹⁸). When bound to DNA, only a single conformation of Loz1 is observed, consistent with the trans conformation observed with other such protein‐DNA complexes (Elrod‐Erickson & Pabo, 1999; Houbaviy et al., ¹⁹⁹⁶; Hudson et al., ²⁰¹⁸; Nagaoka et al., ²⁰⁰¹; Peisach & Pabo, ²⁰⁰³; Wolfe et al., ²⁰⁰⁰).

While the biological significance of the cis‐trans X‐Pro isomerization in the Loz1 TGERP linker has yet to be examined, evidence from other systems suggests a potential role in regulating zinc‐activated repression (Boisvert et al., 2022; Gurung et al., ²⁰²³; Wang et al., ²⁰⁰⁶). Post‐translational modifications in canonical TGE(R/K)P linker sequences, such as Thr phosphorylation during mitosis in Ikaros, Sp1 (Dovat et al., 2002), and Yin Yang 1 (YY1) protein, or acetylation of Lys in YY1 (Rizkallah et al., 2011; Rizkallah & Hurt, ²⁰⁰⁹) or erythroid Krüppel‐like factor (EKLF), have been associated with modulation of DNA‐binding activity (Kluska et al., 2018). In cells, peptidyl propyl cis/trans isomerases catalyze the interconversion of the cis‐ and trans‐forms of peptide bonds, often in response to specific environmental or cellular signals (Belova et al., 2022; Sebák et al., ²⁰²²). As these regulated conformational changes can affect protein folding, stability and activity, cis‐trans isomerization of X‐Pro bonds can function as a molecular switch to control DNA binding, subcellular localization, enzymatic activity, homo‐dimerization, and protein–protein interactions (Belova et al., 2022; Gurung et al., ²⁰²³; Hanes, ²⁰¹⁵). As alterations in the ratio of trans to cis forms would potentially influence the levels of Loz1 available to bind to DNA under a given growth condition, X‐Pro cis‐trans isomerization in Loz1 presents a potential additional layer for regulated zinc finger function. Future studies will be required to test this hypothesis and to determine whether this might be a general mechanism to modulate the activity of zinc finger proteins with TGE(R/K)P linkers.

Analysis of the Loz1‐LRE docking model (Figure 6) contributes to an expanded understanding of the diversity of DNA sequence recognition mechanisms employed by C₂H₂ ZF proteins (Zhang et al., 2024). The model is overall consistent with a dominant role for residues at the −1, +3 and +6 positions of each α‐helix (Figure 1) in a direct readout of the DNA sequence. Prior ChIP‐Seq and in vivo reporter assay experiments identified a consensus of Loz1 DNA recognition element (LRE) of 5′‐GnnGATC‐3′ (Wilson et al., 2019). In the docking model, the sidechain guanidinium of Arg479 at the −1 position of ZF1 is positioned near the 5′ G of this sequence where it could participate in Hoogsteen‐face recognition of the N7 and O6. In ZF2, Asn512 at the +3 and Arg515 at the +6 position are positioned for canonical recognition of the GA step on the complementary strand, 5′‐GATCnnG‐3′, while a hydrophobic contact between Ile485 (ZF1 +6 position) and Val509 (ZF2–1 position) coincide with the C5 methyl group of T on that complementary strand. ZF1 has Ser482 at position +3 where it could make favorable but perhaps not highly selective interactions major groove. While these structural features await experimental validation, the correspondence between a naïve homology model and the known sequence preference suggests that major features of the model may be correct and provide insights into the mechanism for site selection by Loz1.

Deletion experiments established that the Loz1 accessory domain (residues 426–460) is part of the minimal construct required for Loz1‐mediated transcriptional repression of zrt1 and adh4 in vivo (Corkins et al., 2013; Ehrensberger et al., ²⁰¹⁴). The results reported here indicate that it does not appear to function significantly in zinc sensing or site recognition in vitro. Similarity in NMR spectra and zinc binding affinities for AZZ and ZZ constructs indicates that the accessory domain does not become structured upon binding zinc, nor does it alter the affinity of the stoichiometry of zinc binding. In the yeast regulatory protein ADR1, an additional 20‐residue proximal accessory region becomes structured upon DNA binding and contributes to complex stability (Bowers et al., 1999), whereas crosslinking experiments with the Drosophila protein Tramtrack reveal that a similar domain feature contributes to both specific and non‐specific DNA binding (Kamashev et al., 2000). In mammalian MTF‐1, ZF5 and ZF6 are dispensable for zinc‐induced DNA binding activity and reporter gene activation but are necessary to trigger chromatin‐remodeling and activation of the native MT‐1 promoter (Jiang et al., 2003). Absent a role in zinc sensing, the Loz1 accessory domain function may instead play a role in stabilizing the DNA‐bound state or facilitating target gene repression within cells.

The atypical properties of the Loz1 zinc finger domains may provide insight into how the activity of other regulatory factors can be altered in response to cellular zinc status. The human protein ZNF658, which contains 21 zinc finger domains, was shown to inhibit the transcription of multiple genes and act on rRNA involved in ribosomal biogenesis, in response to cellular zinc levels (Ogo et al., 2015). Zinc finger linker sequences have been shown to play a role in zinc‐sensing in MTF‐1. Those studies showed that the replacement of a non‐canonical RGEYT linker between zinc fingers 1 and 2 with a canonical TGEKP linker resulted in the loss of zinc sensing and constitutive binding of mouse MTF‐1 to its target DNA (Li et al., 2006). Another zinc finger containing protein that may be regulated by zinc occupancy is the APC/C inhibitor EMI2, which is part of the cytostatic factor complex that helps maintain the mature, unfertilized mammalian egg in metaphase II arrest (Bernhardt et al., 2012). During fertilization, transient increases in calcium trigger the rapid expulsion of zinc from the cell, which in turn leads to egg activation and the resumption of meiosis. As this event is dependent upon EMI2, changes in the occupancy of its zinc‐binding domains could directly coordinate the meiotic progression of the mammalian egg. In another example, pervasive exchange dynamics in one C₂H₂ ZF domain of the promyelocytic leukemia protein, PML, is linked to its role in response to oxidative stress (Bregnard et al., 2023). Thus, the “canonical” C₂H₂ ZF domain indeed has more “tricks up its sleeve” than its humble conserved structural elements would suggest.

In summary, our observations reveal a potential mechanism for Loz1 zinc‐responsive gene repression and provide insight into the properties of zinc finger domains that enable them to have a sensing function. In eukaryotes, the importance of zinc as a regulator of cellular function has been well‐established (Maret, 2013, 2017). In addition to homeostasis mechanisms that maintain optimal zinc levels, in certain cell types dynamic zinc signals, zinc waves, and zinc fluxes can result in rapid and transient changes the cellular zinc concentrations. Since all these processes have the potential to affect the activation of various zinc‐binding proteins, knowledge of properties that lead to differences in affinities of zinc‐binding sites may help identify other proteins and cellular processes whose activity is regulated by zinc.

4. MATERIALS AND METHODS

4.1. Plasmid construction and expression

Loz1AZZ and Loz1ZZ protein constructs were generated by PCR amplifying Loz1 and cloning the resulting PCR products into the NdeI and BamHI sites of a pET21a vector (Novagen). Inserts were amplified in 25 μL PCR reactions using Taq DNA polymerase (New England Biolabs), Loz1 genomic DNA as the template, and the following oligonucleotide primers: Loz1‐REV, 5′‐catggatccCTACAAACCATGAATGCGTTGA‐3′, Loz1AZZ, 5′‐cggcatatgTCCAATTATTCTGATCATCAC‐3′, and Loz1ZZ, 5′‐catcatatgCGCAAAATTGCACAATCCC‐3′; restriction sites are underlined, Loz1‐template sequences are upper case. Ligated plasmids were transformed into E. coli BL21(DE3) competent cells using electroporation and plated onto LB agar plates supplemented with 100 μg/mL of carbenicillin. A single isolated colony was used to inoculate 100 mL of LB broth containing the 100 μg/mL of the antibiotic and grown to an OD of 1 at 600 nm. Exactly 10 mL of this starter culture was used to inoculate 1 L of M9 minimal medium (6.6 g NaH₂PO₄, 3 g of K₂HPO₄, 1 g NaCl, 2 mM MgSO₄, 100 μM CaCl₂, 1× MEM Vitamin Mix (Gibco), 100 μg/mL carbenicillin) supplemented with 1 g of ¹⁵NH₄Cl (Cambridge Isotope Laboratories) and 4 g of D‐glucose for U‐[¹⁵N]‐protein, or 1 g ¹⁵NH₄Cl and 2 g ¹³C‐D‐Glucose (Martek Isotopes) for U‐[¹⁵N,¹³C]‐protein. Cultures for expressing Loz1AZZ were grown at 37°C to an OD 600 of ~0.8 and expression was induced by addition of 1 mM IPTG supplemented with 100 μM ZnCl₂. Cells were harvested by centrifugation (4200g for 10 min at 4°C) 4 h after induction. Loz1ZZ expression hampered bacterial growth, therefore cultures were grown at 37°C to an OD 600 of 1 to accumulate cell density and were induced overnight at 18°C or 25°C to reduce cellular metabolic activity. Loz1 point mutants P480A, P494A and R494K were generated using the pET21a vector encoding Loz1AZZ as the template for QuikChange site directed mutagenesis (Agilent) using the primers listed in Table S2.

4.2. Protein purification and sample preparation

All buffers were filtered and degassed under vacuum using a bottle top filter before use. Dialysis buffers were additionally degassed by bubbling argon through the buffer. Cell pellet from 1 L culture was lysed using sonication on ice in 35 mL Buffer B (50 mM Tris pH 7.5, 1 M NaCl, 10 mM β‐mercaptoethanol, 100 μM ZnCl₂) with half a tablet of cOmplete, Mini EDTA‐free Protease Inhibitor Cocktail (Roche, cat no. 11836170001) and 0.1% (v/v) Tween‐20. The cell debris was pelleted using centrifugation (27,000g for 45 min at 4°C) and the soluble fraction containing the protein was diluted four‐fold with Buffer A (50 mM Tris pH 7.5, 10 mM β‐mercaptoethanol, 100 μM ZnCl₂) to reduce salt concentration. The diluted soluble fraction was then loaded at 1 mL/min in Buffer A onto a 5 mL HiTrap SPFF column (GE Healthcare Life Sciences) and eluted at 2 mL/min with a 70 mL gradient to 100% Buffer B; proteins eluted ~55% Buffer B. The fractions containing the protein were identified from Coomassie‐stained 8%–16% SDS‐PAGE gradient gels (Genscript), pooled and diluted three‐fold with Buffer A. The diluted peak fractions were loaded at 2 mL/min in Buffer A onto a 5 mL HiTrap Heparin column (GE Healthcare Life Sciences) and eluted at 2 mL/min with a 50 mL gradient from 30% to 100% Buffer B; proteins eluted ~70% Buffer B. The fractions containing the protein were concentrated to 2 mL using centrifugal filters (Millipore‐Sigma, 3k MWCO) and then loaded on a Superdex 75 16/60 size exclusion column (GE Healthcare Life Sciences) pre‐equilibrated with Buffer S (20 mM Tris pH 7.5, 100 mM NaCl, 2 mM TCEP, 100 μM ZnCl₂, 0.02% NaN₃). The protein was eluted at 0.75 mL/min and protein fractions were identified using SDS‐PAGE. Fractions containing the protein were pooled and concentrations determined using Pierce BCA Protein Assay Kit—Reducing Agent Compatible (Thermo Fischer Scientific, cat no. 23250). Integrity, purity and extent of isotopic labeling were assessed by SDS‐PAGE with Coomassie staining (Figure S2) and MALDI‐TOF mass spectrometry (Bruker Microflex) or electrospray ionization (Q Exactive EMR+, Thermo). Mass spectra confirmed expected molecular weights, plus the plasmid‐encoded N‐terminal methionine.

Loz1 point mutants P480A, P494A and R494K were expressed using the same procedure as with the wild‐type protein except for the P480A mutant, which partitioned to the insoluble fraction of the cell lysate. To isolate the P480A Loz1, the insoluble lysis product was washed three times by resuspending the pellet on ice in Wash Buffer (50 mM Tris pH 7.5, 100 mM NaCl, 4 mM TCEP, 2 M urea, and 10% w/v Triton X‐100) and pelleted at 27,000g for 30–40 min at 4°C, then one additional time without urea or Triton X‐100. The resulting pellet was then resuspended in Extraction Buffer (50 mM Tris pH 7.5, 100 μM ZnCl₂, 6 M guanidinium HCl, 100 mM NaCl) and centrifuged at 27,000g, 4°C for 80 min. The supernatant was then collected, diluted 20‐fold with Buffer A (50 mM Tris, 100 mM salt, 4 mM TCEP, 100 μM ZnCl₂) before purifying on 5 mL HiTrap SPFF column as described above.

4.3. Calorimetry

Zinc‐free Loz1 was prepared by treatment with a 10‐fold molar excess of EDTA over protein, followed by two‐step dialysis (500× dilution, each) using 3.5K MWCO G2 cassettes (Thermo Scientific, cat no. 87723) against degassed Buffer H (50 mM HEPES pH 7.5, 100 mM NaCl, 10 mM TCEP) containing Chelex, for at least 12 h each, at 4°C. HEPES was chosen for ITC experiments, instead of Tris, because of its lower ionization enthalpy (4.9 vs. 11.4 kcal mol⁻¹) (Goldberg et al., 2002), and broad suitability for studies involving metals (Ferreira et al., 2015); nevertheless, calorimetric studies of buffer‐divalent metal interactions suggest that MES may be a better choice for future experiments (Xiao et al., 2020). Isothermal titration calorimetry was performed using a VP‐ITC instrument (Malvern). Zinc solutions were prepared by dilution of zinc chloride stock (10 mM in water) in ITC buffer. Protein concentrations in the ITC cell were 4–7 μM, and zinc concentrations in the syringe were 80–200 μM. Zinc concentrations were determined using an Agilent Atomic Absorption Spectrometer 240FS. ITC data were recorded in 5 μL first injection followed by 25 injections of 10 μL, spaced 420 s apart, at 298 K. Origin 7.0 (OriginLab) was used to integrate and fit the data to a binding model with two independent sets of sites to extract the thermodynamic parameters, n, ΔH and K _d for each site.

4.4. NMR

NMR spectra were recorded at 25°C on 600, 800 or 850 MHz Bruker Avance III spectrometers fitted with 5 mm cryoprobes on samples ranging from 0.5 to 1 mM concentration in NMR Buffer (20 mM Tris pH 7.5, 100 mM NaCl, 2 mM TCEP, 100 μM ZnCl₂, 0.02% NaN₃, 10% (v/v) ²H₂O and 200 μM DSS as an internal standard; heteronuclear shifts were referenced indirectly to DSS from the corresponding gyromagnetic ratios). Data were processed using NMRpipe (Delaglio et al., 1995) or NMRFx (Johnson, 2018) and analyzed using NMRViewJ (Johnson, 2018).

4.5. Resonance assignments

Data from 3D HNCO, HNCA, HNCOCA, HNCACB, CBCACONH experiments were used to assign backbone H^N, N, C′, C^α and C^β atoms. The 3D HCCH‐TOCSY, HCCH‐COSY, H(CCO)NH‐TOCSY and (H)CC(CO)NH‐TOCSY spectra were used to obtain additional sidechain proton and carbon assignments, respectively. The chemical shift index for C_α and C_β were obtained from TALOS‐N software (Shen, 2015). Chemical shift perturbations (CSPs) were calculated using the amide chemical shifts according to:

$$\mathrm{\Delta }\delta =\sqrt{{\left({\mathrm{\Delta }\delta }_{\mathrm{{\rm H}{\rm N}}}\right)}^2+{\left(0.2{\mathrm{\Delta }\delta }_{\mathrm{N}}\right)}^2}$$

where Δδ is the compounded chemical shift perturbation, Δδ _HN is the chemical shift perturbation of the amide proton and Δδ _N is the chemical shift perturbation of the amide nitrogen.

4.6. NMR relaxation

¹{H}‐¹⁵N heteronuclear NOE, R ₁ and R ₁ρ data were recorded at 600 and 850 MHz for Loz1AZZ, and at 800 MHz for Loz1ZZ, following published methods (Palmer et al., 2001). R ₁ and R ₁ρ data were acquired in an interleaved fashion with relaxation delays of 40, 200 (×2), 520, 720 (×2) ms and 2, 34, 68 (×2), 100 (×2) ms, respectively using a recycle delay of 2 s. ¹{H}‐¹⁵N hetNOE values were obtained from ratios of signal intensities for spectra collected in presence and absence of ¹H pre‐saturation during a 4 s recycle period. The relaxation rates were obtained by fitting peak intensities to an exponential decay function.

¹⁵N relaxation dispersion CPMG data were recorded at 600 and 800 MHz proton frequencies for each Loz1AZZ and Loz1ZZ construct following published methods (Hansen et al., 2008). The experiments on both instruments were acquired using a recycle delay of 2 s and varying CPMG pulse train frequency: for 800 MHz data, 82.6 μs pulses in pulse train of 50 ms with 19 points (0, 20, 40 (×2), 60, 80, 100, 120, 160, 200, 280, 440, 600, 760, 920, 1080, 1240 (×2), 1400 Hz); and for 600 MHz data, 87.2 μs pulses in a pulse train of 40 ms with 21 points (0, 25, 50 (×2), 75, 100, 125, 150, 175, 200, 250, 300, 350, 400, 500, 600, 700, 800, 900 (×2), 1000 Hz). Dispersion data from both 600 and 800 MHz were fit simultaneously with the Carver–Richards two‐state exchange model for individual residues and globally for groups of residues, using GUARDD (Kleckner & Foster, 2012). Good quality fits had small and random residuals.

4.7. NMR‐monitored zinc titrations

Zinc‐free protein was generated by treatment with 10‐fold molar excess EDTA over protein, followed by overnight dialysis (500× dilution) using 3.5K MWCO G2 cassettes (Thermo Scientific, cat no. 87723) against degassed NMR‐buffer without ZnCl₂, at 4°C. Zinc‐coupled binding/folding of apo Loz1AZZ construct was monitored by recording 2D {¹H}‐¹⁵N HSQC spectra over a range of added zinc concentrations. Exactly 10 mM zinc chloride solution in NMR buffer, made by diluting 1 M zinc chloride stock (in water), was titrated into 200 μM [U‐¹⁵N]‐Loz1AZZ and incubated at room temperature for 15 min before data collection. Fractional peak intensities were calculated from the ratio of each peak intensity to the maximum peak intensity measured from all {¹H}‐¹⁵N‐HSQC experiments for that residue.

4.8. DNA substrate and DNA‐protein complex preparation

ssDNA oligos: 5′‐d(GCGACGATCA TTGG)‐3′ and 5′‐d(CCAA TGATCGTCGC)‐3′ (IDT, Inc.) were resuspended in DNA buffer (20 mM Tris pH 7.5, 50 mM NaCl) and annealed to obtain dsDNA substrate by heating equimolar mixture of oligos to 95°C in a water bath for 15 min and cooling in water bath overnight to room temperature. One dimensional ¹H spectra revealed eight imino signals (4 G‐C and 4 A‐T pair). PD‐10 column (GE Healthcare Life Sciences) was used to exchange the dsDNA into NMR buffer. Concentration of DNA was measured using UV–Vis spectrometry absorbance at 260 nm. Exactly 2 mM of purified Loz1AZZ in NMR buffer was titrated into 130 μM DNA substrate to reach an equimolar concentration.

4.9. Homology modeling and docking

Each Loz1 zinc finger (ZF) domain was individually submitted for homology modeling using the SwissModel webserver (Waterhouse et al., 2018). The template (Figure S1) for modeling ZF1 was ZF 1 of ZBTB38 (residues 1010–1033) bound to DNA (PDB: 6E93), and ZF2 was modeled using ZF 3 of YY1 (residues 349–377) bound to DNA (PDB: 1UBD) (Houbaviy et al., 1996; Hudson et al., ²⁰¹⁸). Comparison of these models to those generated with AlphaFold2 (https://alphafold.ebi.ac.uk/entry/Q9UTA1) (Jumper et al., 2021; Varadi et al., ²⁰²⁴) produced backbone RMSDs of 1.3 and 0.7 Å, respectively. The resulting models were superposed onto zinc finger 2 and 3 of the YY1‐DNA complex using the align function of PyMOL (Schrödinger). The LRE DNA model was generated by first using the X3DNA webserver to obtain the helical parameters from the YY1 DNA substrate, and then rebuilding the LRE consensus sequence by mutating positions C6, C9 and T12 of the YY1 substrate to G6, G9 and C12 of the LRE using the X3DNA webserver in conjunction with previously obtained DNA helical parameters (Li et al., 2019). After superposing the two zinc fingers of Loz1 onto the modeled LRE, the chain break present from homology modeling each finger independently was closed using the kinematic closure protocol in Rosetta (Stein & Kortemme, 2013). Knowledge of canonical zinc finger recognition of DNA substrates was used to generate a set of ambiguous and unambiguous restraints from Loz1 ZZ to the LRE DNA substrate, including restraints to maintain correct geometries for coordinating zinc in each ZF (Touw et al., 2016). These restraints and homology models were submitted to the HADDOCK 2.4 webserver for docking and the lowest energy model from the best scoring cluster (RMSD, HADDOCK Energy, and restraint energy violation) was selected as the Loz1–LRE complex model for analysis (Van Zundert et al., 2016).

4.10. Accession codes

Schizosaccharomyces pombe Loz1, GenBank: CAB61785.2, PomBase (Rutherford et al., 2024): SPAC25B8.19c, NCBI: NP_594479.2.

AUTHOR CONTRIBUTIONS

Vibhuti Wadhwa: Conceptualization; investigation; writing – original draft; methodology; validation; visualization; writing – review and editing; formal analysis; data curation. Cameron Jamshidi: Conceptualization; investigation; writing – original draft; methodology; validation; visualization; writing – review and editing; formal analysis; data curation. Kye Stachowski: Conceptualization; investigation; visualization; data curation. Amanda J. Bird: Project administration; conceptualization; investigation; funding acquisition; writing – original draft; writing – review and editing; supervision; resources. Mark P. Foster: Conceptualization; investigation; funding acquisition; writing – original draft; methodology; validation; visualization; writing – review and editing; formal analysis; project administration; supervision; resources; data curation.

FUNDING INFORMATION

This work was funded by NIH grant R01 GM105695 to AB. Native mass spectrometry experiments were supported by NIH grant P41GM128577 to Vicki Wysocki. Kye Stachowski was supported by NIH grant R01GM122432 to MPF. ITC experiments were made possible by NIH grant R01GM063615S1. This study made use of NMRBox, which is supported by NIH grant P41GM111135.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1. Supporting Information.

Wadhwa V, Jamshidi C, Stachowski K, Bird AJ, Foster MP. Conformational dynamics in specialized C₂H₂ zinc finger domains enable zinc‐responsive gene repression in S. pombe . Protein Science. 2025;34(2):e70044. 10.1002/pro.70044

DATA AVAILABILITY STATEMENT

The NMR data that support the findings of this study are openly available in Biological Magnetic Resonance Data Bank at https://bmrb.io/. Assignments for zinc‐bound Loz1AZZ were deposited with BMRB ID 52876.