The Dbh gene was incorporated into the vector pKKT7-H (a derivative of pKK233, Promega) containing an N-terminal His6 tag (MHHHHHHLVPRGM). Quick-change mutagenesis (Stratagene) was used to change Cys31 to Ser to eliminate potential formation of disulfide bonds. Transfected E. coli BL21 cells were grown in 1L Neidhart’s minimal media (Neidhardt et al., 1974) at 37°C containing 1 g ¹⁵N ammonium chloride; expression was induced by the addition of 1 mM IPTG. Protein was expressed for 5 h; subsequently, the cells were harvested by centrifugation and frozen at −80°C. Dbh was purified from cell lysate by Ni-NTA affinity chromatography under native conditions, and then dialyzed into buffer (20 mM HEPES, 100 mM NaCl, 50 μM EDTA, 50 μM NaN₃, pH 7.5) at 4°C, followed by one change of buffer without EDTA (NMR buffer = 20 mM HEPES, 100 mM NaCl, 50 μM NaN₃, pH 7.5). To prepare the NMR samples, Dbh protein was concentrated to between 0.3–0.5 mM and transferred into a 5-mm Shigemi tube. D₂O was added to the sample for a final concentration of 10% v/v.

Hydrogen exchange: Samples of ¹⁵N-labeled Dbh (0.5 mM or higher concentration) were transferred into deuterated NMR buffer (20 mM HEPES, 100 mM NaCl, 50 μM NaN₃, pD 7.5), using a P10 desalting column equilibrated with the deuterated buffer. The sample was transferred to a Shigemi NMR tube and immediately placed in an 800 MHz Varian Inova NMR spectrometer, containing an xyz triple resonance probe, equilibrated at 35°C or 50°C. After shimming and tuning the magnet, the acquisition of the first 15N-1H TROSY-HSQC (Pervushin et al., 1997) spectrum was started approximately 15 min after insertion of the protein sample into the magnet. Additional ¹⁵N-HSQC spectra were measured sequentially every 147 min and 13 s (2.454 h), except for the last five spectra at 50°C, for which additional scans were taken (2x for the 17th through 20th spectra, 4x for the 21st spectrum) to improve the signal-to-noise ratio. A total of 20 spectra were collected at 35 °C and 50 °C. Due to the additional length of acquisition in the last five spectra at 50°C, the acquisition of the 18th, 19th, 20th, and 21st spectra was started 4.907 h after the start of the previous spectra. Assigned spectra for initial time points at both temperatures are shown in (Supplementary Figure S1). The data were processed using NMRPipe (Delaglio et al., 1995) and analyzed using CCPNMR Analysis (Vranken et al., 2005) (RRID:SCR_016984). The peak intensity was plotted as a function of time and fit to a single-order exponential-decay function (I(t) = I₀ × e^−kt) to extract the exchange rate. Representative time points and rate fits to the experimental data are displayed in (Supplementary Figure S2). To obtain the protection factors the hypothetical exchange rate for the amide proton in a random coil conformation was calculated, corrected for the effect of side chain identity to the left and right of the amide proton according to (Molday et al., 1972; Bai et al., 1994).

Temperature Coefficients (TC): Conventional ¹⁵N-HSQC spectra were taken at 35°C, 45°C, 50°C, 55°C, and 65°C, on samples containing at least 0.5 mM ¹⁵N-labeled Dbh in 20 mM HEPES, 50 mM NaCl, 50 μM EDTA, pH 7.5, with 10% (v/v) D₂O. The data were processed with NMRpipe (Delaglio et al., 1995), and visualized using CCPNMR Analysis 2.5 (RRID:SCR_016984, (Vranken et al., 2005). For each residue, H chemical shift (ppm) was plotted as a function of temperature (K) and fit to a straight line. The slope*1000 for each fit was used as TC for each residue. The standard error of the slope was used as TC uncertainty. Only fits with R ² > 0.75 were used. Non-linear fits were not used, a total of 10 non-linear TC values were discarded. The TC average for each domain only includes amides located in secondary structure regions, and excluded amides in loop regions and amides that are exposed to solvent based on the 1K1S crystal structure.

Ion chemical shift perturbations: All samples were made with 20 mM HEPES, 100 mM NaCl, 50 μM NaN₃, at pH 7.5. The Mg²⁺ sample contained 5 mM MgCl₂. The Mn²⁺ sample contained 5 mM of MnCl₂. Data were analyzed with CCPNMR Analysis (Vranken et al., 2005) (RRID:SCR_016984).

200 μL of 0.28 mg/mL Dbh protein in buffered solution (50 mM sodium phosphate, 100 mM NaCl, pH 7.5) were transferred into a CD cuvette with a 1-mm path length. The cuvette was placed in the sample cell Jasco J-810 spectropolarimeter equipped with a Peltier temperature controller, and the cell was calibrated at 35°C. A spectrum was collected at 35°C from 260 nm to 185 nm, with a scan rate of 50 nm/min, data points were collected every 0.5 nm and 10 total scans. The cell was then cooled at a rate of 0.5°C/min to 7°C, with the ellipticity value measured at 208 nm every 2°C, then after a full spectrum was taken at 7°C with the same parameters as above. The sample was then removed from the CD cuvette, the cuvette was cleaned, and a second 200 μL 0.28 mg/mL Dbh protein sample was transferred into the cuvette. The cuvette was placed into the cell equilibrated at 35°C, another full CD spectrum was taken at this temperature, and the sample was heated to 65°C at a rate of 0.5°C/min, with the ellipticity value at 208 nm taken every 2°C. One full spectrum was then taken at 65°C. After adjustments to the chiller unit, a third 0.28 mg/mL Dbh sample was used to collect full CD spectra again at 7°C and then at 2°C.

Protein samples used for NMR measurements were diluted with the NMR buffer solution (50 mM sodium phosphate, 100 mM NaCl, pH 7.5). The protein concentration was then measured by light absorbance at 280 nm and determined to be 1.1—1.3 mg/mL. The protein solution was pipetted into a disposable plastic cuvette with a path length of 1 cm and placed in a Malvern Zetasizer Nano DLS instrument. The temperature was equilibrated at 35°C and measurements were taken for 30 s and repeated ten times. The temperature was then lowered to 5°C, allowed to equilibrate for 10 min at 5°C, then measurements were taken as above.

We used UCSF Chimera to visualize Dbh (Pettersen et al., 2004). Structure 1K1S was used for all analyses (Silvian et al., 2001). The addition of hydrogens and solvent accessibility were obtained using Chimera default settings.

We obtained ¹⁵N-HSQC NMR spectra to determine if Dbh maintains tertiary structure between 35°C and 65°C. The spectra were collected at 35°C, 45°C, 55°C, and 65°C. All spectra remained well dispersed and relatively similar across this range of temperatures (Figure 2). We did observe the normal phenomena of temperature dependent chemical shift changes as temperature increases (Ohnishi and Urry, 1969). We used this temperature dependent chemical shift to calculate TC (discussed later) to determine sensitivity to temperature changes. The retention of well dispersed spectra at temperatures ranging between 35°C–65°C informs us that Dbh retains a well folded and consistent tertiary structure throughout this range of temperatures.

We find that Dbh undergoes cold denaturation of tertiary structure below 25 °C. Cold denaturation is a phenomenon that all proteins are predicted to experience, however, it is extremely rare for it to occur above the freezing point of water (0 °C) (Privalov, 1990). We collected ¹⁵N-HSQC spectra at cold temperatures ranging from 5°C to 25°C. These spectra show a reversible decrease in tertiary structure starting at 25 °C which progresses as temperatures get colder (15°C, 5°C) (Figure 3). To determine if this is a complete loss of structure, we assessed the secondary structure content using CD spectroscopy. Figure 4 shows CD spectra for Dbh at 2°C, 7°C, 35°C, and 65°C. In contrast to the NMR results, CD revealed that the secondary structure of Dbh is consistent over this temperature range, similar to what has been reported for Dpo4 (Sherrer et al., 2012). We used Beta Structure Selection (BeStSeL) (Micsonai et al., 2018) to deconvolute the CD spectra; our assessment of secondary structure content shows only a slight decrease in percent helicity and a small increase in percent beta strand as temperature decreases (Figure 4). The protein is monomeric with a diameter between 6–7 nm at both 5°C and 35°C as determined by dynamic light scattering (DLS), shown in Supplementary Figure S3. Collectively, these results are consistent with a molten globule unfolded state at cold temperatures.

Dbh dynamics were probed by measuring amide HX rates (k _ex ) and calculating protection factors at each temperature. We previously published NMR peak assignments for Dbh at these two temperatures (Moro and Cocco, 2015). Amide hydrogens that are readily exposed to solvent will exchange at a higher rate compared to hydrogens that are involved in secondary structure or sequestered away from solvent (e.g., buried in a hydrophobic core).

We determined k _ex by measuring peak volumes as a function of time after transfer into D₂O and fit the results to calculate the exchange rate (Supplementary Figure S2). PFs were calculated by k _rc /k _ex , where k _rc is the exchange rate of a backbone amide in a random coil and k _ex is the experimental exchange rate fit from the NMR peak volumes. k _rc was calculated based on in Bai, et al. (Bai et al., 1994). Obtaining PFs from multiple amides in each domain gives us a thorough understanding of the dynamics throughout the Dbh structure. Our Dbh construct is composed of 353 residues, 15 of which are prolines that do not have a corresponding amide hydrogen. In addition, Dbh has four distinct small domains with significant solvent accessible area. Consequently, many amides exchanged immediately upon transfer into D₂O. At 35°C we were able to observe signals of 82 residues that remain in D₂O. Eighteen residues exchanged too fast to obtain sufficient points to calculate an exchange rate, and 26 remained unchanged (stable) throughout the experiment (Supplementary Table S1). In contrast, at 50°C we obtained HX information on 43 residues. Seven of these exchanged too fast to obtain sufficient data to reliably fit exchange rates and 21 remained unchanged (stable) throughout the experiment (Supplementary Table S1).

The palm domain is composed of residues 1–19, 78–171. This is a critical domain because it holds the active site for nucleotide extension. At 35°C we observe protection for residues in the beta-sheet and two alpha helices that comprise the palm domain (Figure 5). We observe a minor decrease in protection at 50 °C. Much of the decrease in protection is in proximity to the major helices in the palm domain, which include helices E and F and strand 6. In this region of the palm domain we observe seven residues that lose complete protection (84, 85, 88, 91, 136–138) and five residues lose over 70% of their protection (89, 92, 106, 108, 124, 135) with the increase in temperature.

The fingers domain is composed of residues 20–77. The protection factor data shows this domain is more dynamic at 50 °C than at 35°C. At 35°C signals from two residues (20, 68) are present immediately after transfer to D₂O but exchange too rapidly to determine exchange rates/PFs. The remaining seven observable residues (19, 28, 30, 33 50, 55, 74) are stable enough to obtain an exchange rates and protection factors (Figure 5; Supplementary Table S1). Two of the residues (50, 74) have relatively high PFs which indicate resistance to local unfolding at 35°C. Both residues are involved in secondary structure hydrogen bonding. Solvent accessibility surface area (SASA) calculations showed these amides are not exposed to solution. These residues are found 12.9 Å away from each other which indicates far reaching stability of the domain at 35 °C. In contrast, at 50°C all residues in the finger domain exchange too fast to measure. These two HX experiments reveal that the fingers domain of Dbh is significantly more dynamic at 50°C than at 35°C.

The thumb domain is composed of residues 172–236. At 35°C the thumb domain contains six observable residues (175, 178, 184, 194, 201, 206) at the start of the HX experiments. Two residues (194, 201) have high protection factors at 35°C. Residue 194 is involved in a helical section of the protein and residue 201 is in an unstructured part of Dbh between two helices. Both residues 194 and 201 amides have no exposure to solvent (SASA = 0 Å).

The LF domain is composed of residues 237–344 and becomes slightly more dynamic at 50 °C than at 35 °C. Two residues (287, 290) were extremely stable (no exchange) throughout the experiment at 35 °C. The only regions that showed a uniform decrease in protection in the LF were helix N (258–276) and helix O (307–324). All other decreases in protection factors were evenly distributed throughout the LF domain (Figure 5).

Temperature coefficients are calculated from the chemical shift change with temperature. It has been shown that amides not involved in intramolecular hydrogen bonding (exposed to solvent) and amides that unfold in response to temperature changes have TC values below −4.6 ppb/K (Baxter and Williamson, 1997; Cierpicki and Otlewski, 2001; Cierpicki et al., 2002). We implemented TC analysis and obtained values that agreed with the crystal structure. We see that most residues with TC values below −4.6 ppb/K belong to amides that are not involved in hydrogen bonding and TC values more positive than −4.6 ppb/K belonged to hydrogen bonding amides in secondary structure (Supplementary Figure S4; Table 1). However, we did find several exceptions; for example, helix E (residue 77–94) of the palm domain gives TC values more negative than −4.6 ppb/K. This includes residues 82, 86, and 89 all of which are involved in hydrogen bonding and have no solvent exposure based on the crystal structure 1K1S. These data indicate that this particular region of the palm domain is sensitive to increases in temperature with unfolding and increasing exposure to solvent. Helix F of the palm domain also shows this unfolding (amide exposure to solvent) with temperature. In the thumb domain, although the crystal structure shows that residue 176 is involved in an intramolecular hydrogen bond within a helix, its TC value corresponds to an amide that also becomes exposed to solvent (<-4.6 ppb/K) as the temperature is raised.

Our studies show that Mg²⁺ interacts with Dbh in the absence of DNA. We analyzed ¹⁵N-HSQC spectra of Dbh in the presence and absence of Mg²⁺(Figure 6). When ligands, such as Mg²⁺, bind proteins they change the local chemical environment, resulting in chemical shifts changes for peaks of residues involved in ligand binding and for residues near the binding site or affected structurally by the binding event. Residues with shifted peaks in the Dbh/Mg²⁺ spectra inform us on local Mg²⁺ binding. We found shifted peaks clustered in three distinct locations: 1) at the active site in the palm/finger domains, 2) the thumb domain, and 3) the little finger domain (Figure 7). We found that ten of the residues that shift in the presence of Mg²⁺ are located in the active site. Seven of these are found in the palm domain and three in the finger domain (Table 2). Although they are in different domains, all residues are in proximity to the active site residues of this Y-family polymerase (Figure 7). Mg²⁺ also binds the thumb domain in the absence of DNA since the ¹⁵N-HSQC of Dbh in the presence of Mg²⁺ shows three peaks shift in the thumb domain (Table 2; Figure 7). In addition, we found that Mg²⁺ binds Dbh’s LF domain in the absence of DNA. When comparing the spectra with and without Mg²⁺ we observe that four peaks from the LF domain shift in the presence of Mg²⁺. All four residues are within 14 Å from each other in the three-dimensional structure (Table 2; Figure 7).

We confirmed ion binding sites by comparing ¹⁵N-HSQC spectra with and without Mn²⁺. Mn²⁺ binds proteins in a similar manner as Mg²⁺, however, Mn²⁺ is paramagnetic. This increases relaxation times, leading to the broadening and disappearance of peaks for residues near the binding site (16–25 Å) (Pintacuda et al., 2004). By overlaying the Mn²⁺ and apo spectra we should see the absence of peaks of residues near the metal binding sites (Supplementary Figure S5). The paramagnetic effect of Mn²⁺ is more far-reaching than chemical shift perturbations from Mg²⁺; consequently, the number of affected residues is increased in the Mn²⁺ samples compared to Mg²⁺ (Table 2). We observed the disappearance of 37 residues that correlate well with our predicted Mg⁺² binding sites except for one (Gly117), which is not within any of our predicted binding sites (Supplementary Figure S6).

Although the high protein concentrations required for NMR studies complicate accurate calculations of dissociation constants (Jarmoskaite et al., 2020), we can compare relative affinities of Mg²⁺ binding sites using NMR. We measured chemical shift changes for the signals with the largest peak position difference upon Mg²⁺ binding in each site (Supplementary Figure S7). We find that all three binding sites have an affinity for Mg²⁺ in the low mM concentration range of the ion. The amount of endogenous Mg²⁺ in S. acidocaldarius has not been reported, but we do know that the concentration of Mg²⁺ in E. coli is between 30–100 mM (Milo and Phillips, 2015). We do not have experimental data on Mg²⁺ binding of the protein at physiological concentrations. However, we can say binding of Mg²⁺ to the protein at NMR concentrations is within the Mg²⁺concentration range of available ions in the cell. Thus, it is possible that Dbh binds Mg²⁺ under physiological conditions.

NMR ¹⁵N-HSQC chemical shifts reveal that Dbh is thermally stable; however, tertiary structure denatures at and below 25°C. The protein maintains secondary structure down to 2°C as determined by CD (Figure 4) and remains compact and monomeric (DLS). Together these data show that Dbh retains tertiary structure between 35°C–65°C, and transitions to a molten globule at and below 25°C. Cold denaturation is unusual for a large protein and could be related to the requirement for this protein to function at very high temperatures.

Although many proteins crystallize well in the cold, it is notable that Dbh crystals formed best at room temperature (Silvian et al., 2001; Wilson and Pata, 2008; Wilson et al., 2013). In addition, the loss of tertiary structure at 25°C correlates well with the loss of catalytic efficiency observed in a primer extension study at 22°C compared to higher temperatures. Potapova, et al., reported a 40-fold change in reaction rate going from 65°C to 22°C (Potapova et al., 2002). However, kinetic theory predicts a two-fold change in rate for every 10°–if the transition state remains the same. If the mechanism were maintained across the temperature range, an approximate change in temperature of ∼40°C should have resulted in an approximate 16-fold change in rate of synthesis, much smaller than the 40-fold rate change that Potapova, et al. reported. It is likely that structural changes resulting from cold denaturation contribute to an additional 2.5-fold diminishment of catalytic efficiency measured at 22°C.

Using PFs we determined the regions/domains of the protein that are resistant to exchange and compared these regions of rigidity at a temperature close to the cold denaturation transition (35°C) and one closer to the optimal functional temperature (50°C). There are two mechanisms that could affect rates of amide exchange: changes in local unfolding kinetics or a structural change whereby an amide becomes more or less occluded from solvent. For proteins where the structure does not change significantly, differences in the rates of amide hydrogen exchange can reflect the influence of the temperature on the local unfolding of that amide hydrogen bond.

As previously mentioned Dbh functions in extremely hot environments (∼75°C), therefore the dynamics we observed at temperatures close to Dbh’s natural environment should reflect the natural state of the Dbh polymerase. In general, our protection factor data show that Dbh protein is considerably more dynamic at 50°C than at 35°C. We mapped PFs at 35°C and 50°C on the crystal structure of Dbh (Figure 5) and a clear pattern of dynamics and stability emerged. PFs show that the palm domain and the little finger domain are the most stable regions of Dbh at both temperatures. In contrast, the dynamics of the thumb and finger domain are significantly different between 35°C and 50°C. At 35°C we observe nine residues protected in the finger domain and six residues protected in the thumb domain. This protection is lost at 50°C indicating that these domains become more flexible as the temperature is increased toward the physiological temperature of S. acidocaldarius.

The palm domain is the most stable domain in Dbh (Figure 5). The LF domain is the second most stable part of Dbh, followed by the fingers domain. Compared to the other domains the thumb domain seems to be most dynamic regardless of temperature. This is revealed by a lack of protected amides. Nevertheless, like the fingers domain, the thumb domain shows the same trend with lower PFs at 50°C than 35°C. At 50°C there are no protected residues in the Dbh thumb domain. TC data show that residue 176 is sensitive to temperature increase (TC value = −6.61), indicating unfolding with temperature. Together the protection factor data and TC calculations show that the thumb domain is the most flexible domain in the protein and that this flexibility increases with temperature.

TC values for proteins have been shown to be sensitive to temperature dependent global and local unfolding (Ohnishi and Urry, 1969; Andersen et al., 1997; Baxter and Williamson, 1997). Historically TC data for protein is analyzed at the individual amide level. This correlates with our PF data that show that a small region of the palm domain becomes slightly more flexible when going from 35°C to 50°C (Figure 5). Together these two methods allow us to conclude that although the palm domain is the most protected domain at both temperatures, helix E (77-94) and helix F (121–138) of the palm domain do adopt flexibility as temperature increases.

Most recently, analysis of TC data has been performed in group averages of helix secondary structures (Doyle et al., 2016). It has been shown that this type of analysis is a rich source of information on temperature dependent global stability (Doyle et al., 2016). We averaged our TC data by secondary structures, and by domain (Table 1). When averaging by secondary structure (i.e., strand and helix) only limited structural information is obtained. In contrast, averaging the entire domains shows a clear trend that correlates with our hydrogen exchange PF data. We see the most dynamic domain the thumb domain has the most negative TC average of −3.28 ppb/K, followed by the second most dynamic finger domain with a TC domain average of −3.19 ppb/K. For the more stable domains the TC averages are close, but do not correlate with our PF data. TC average for the palm domain was −2.77 ppb/K and −2.69 for the LF domain.

Other groups have correlated TC and HX to obtain structural information on proteins (Andersen et al., 1997; Baxter and Williamson, 1997). It has been shown that protected amides (slow exchangers) average a less negative TC average compared to lower protected amides (fast exchangers) who have a more negative TC average (Andersen et al., 1997). Our TC average analysis only included amides in defined secondary structure (strands and helix). The analysis excluded amides in unstructured regions and amides that did not hydrogen bond and were exposed to water. This was done to avoid artificially decreasing our average TC by including naturally fast exchanging amides.

Chemical shift changes of ¹H-¹⁵N HSQC reveal three ion binding sites of Mg²⁺ and Mn²⁺, even in the absence of bound DNA. The binding regions include two that have been reported previously in crystal structures of other Y-family polymerases in the presence of DNA (Irimia et al., 2010; Pata, 2010) and a novel binding region on the LF domain of Dbh. There is currently an active debate on whether the DNA synthesis mechanism requires two or three ions for synthesis. The third ion has been suggested to be involved in the active site interacting with the incoming nucleotide (Yang et al., 2016). Our experiments did not involve DNA or free-floating nucleotide. However, a third binding site is apparent even without DNA bound.

Twenty-three residues that disappeared in the presence of Mn²⁺ are near the palm/finger Mg²⁺ binding site, three are found near the LF Mg²⁺ binding site, and ten were in the thumb Mg²⁺ binding site (Table 2; Figure 7). Notably, all residues that shift in presence of Mg²⁺ spectra also disappear in the Mn²⁺ spectra (Table 2) except for residue Met282. It is possible that Met282 chemical shift changes with Mg²⁺ are a result of long-range conformational change and this residue may not be involved directly in ion binding. Regardless, the Mn²⁺ data serves as a secondary verification of our proposed ion-binding sites.

Notably, apo crystal structures of Dbh do not show binding of metal ions (PDB: 1IM4 (Zhou et al., 2001), 1K1Q (Silvian et al., 2001), 1K1S (Silvian et al., 2001)). In contrast, Dbh in complex with DNA has been shown to bind one Ca²⁺ (Wilson and Pata, 2008). The chemical shift changes we observed indicate that Mg²⁺ binds Dbh in the active site in the absence of DNA in solution; however, these ions bind weakly since they do not co-crystallize with the protein. The active site of Y-family polymerases consist of the palm domain surrounded by the finger domain and to a lesser extent the thumb domain (Wilson and Pata, 2008; Irimia et al., 2010; Pata, 2010). One Ca²⁺ binds in the active site in Dbh/DNA structures (Wilson and Pata, 2008). This Ca²⁺ ion is chelated by Asp7, Phe8, and Asp105. Our solution-state NMR data show that Mg²⁺ interacts with these same residues in Dbh even without DNA bound. Furthermore, these chelating residues are surrounded by residues Asp9, Ile139, and Thr140; we find that these also experience chemical shift changes in the presence of Mg²⁺. In addition, our data shows that Mg²⁺ interacts with three residues that are near the active site but shifted towards the finger domain, these include Ala13, Glu16, and Thr45. It may be that Mg²⁺ is positioned more towards the finger domain when Dbh is in solution and not in the presence of DNA or that these residues are structurally adjusted when the ion binds.

As mentioned earlier other Y-family such as Dpo4 polymerase have been shown to crystalize with two Mg²⁺ ions near this active site. This interaction can be attributed to a second Mg²⁺ which is almost always shown interacting with DNA and the protein. Published studies (Irimia et al., 2010) show Dpo4 chelating Mg²⁺ with residues 181–183 (182–184 Dbh). Our work suggests a slightly different binding site near residues 178, 201, and 204 in the absence of DNA. Two residues in this vicinity are positively charged Arg204 and Lys202. The peak corresponding to Arg204 is shifted in the presence of Mg²⁺, but the peak corresponding to Lys202 is found in a crowded region of the spectra which could not be accurately measured. Both charged residues have potential salt bridge partners, however, one appears to be too far for the average salt bridge distance of 4 Å (204Arg-205Asp (5.2Å) and 202Lys-180Asp (2.7 Å)). If they are indeed forming salt bridges the introduction of Mg²⁺ in this region could interrupt the salt-bridge partners which could lead to peak shifts caused by the repulsion forces of two like charges.

Our results show that the Dbh LF has a Mg²⁺ binding site. Dpo4 has been shown to bind a Mg²⁺ ion in the LF domain near Asp294 (PDB 2XCA) (Irimia et al., 2010). However, we found the Dbh binding site to be 23.5 Å away from the reported binding site of Dpo4. Dbh and Dpo4 sequences have the lowest sequence similarity between their LF domains with only 42.1% similarity. These amino acid changes could result in a movement of the ion binding site. Furthermore, a sequence alignment across 34 Y-family homologs showed that two of the shifted residues (Gly299, Lys301), are commonly mutated in other Y-family polymerases (Wu et al., 2017). We suspect that the distinct sequence identity of the LF domain may create a Mg²⁺ binding region unique to Dbh. However, another explanation for the lack of Mg2+ binding in the LF in other crystal structures might be caused by low occupancy Mg2+. It is difficult to visualize using crystallography if the ion has lower than 30% occupancy due to the low electron number of Mg2+ (Yang et al., 2016).