First, we developed a thermodynamic model to describe the conformational states of DBD as a function of [Zn²⁺]. We propose a minimal four-state mechanism in which Zn²⁺ binds to a single, high-affinity site in native apoDBD (N) with an equilibrium constant of 1/K_Zn (Equation 1) and to one or more low-affinity sites in unfolded apoDBD (U) with an average equilibrium constant of 1/K_Zn,U (Equation 2). HoloDBD (N·Zn) is the only species with native DNA-binding activity.(1)N+Zn ⇄ N⋅Zn; 1KZn(2)U+ Zn ⇄ U⋅Zn; 1KZn,U

To measure K_Zn, we monitored the increase in Tyr fluorescence (Butler and Loh, 2003) of apoDBD as it binds Zn²⁺ at 10°C (Figure 2A). [Zn²⁺]_free was buffered at the indicated concentrations using chelators of varying Zn²⁺ affinity. Fitting the data to the one-site binding equation yields K_Zn = (1.6 ± 0.3) x 10⁻¹⁵ M. To our knowledge, this is one of the lowest K_Zn values ever reported, with only one other eukaryotic protein possessing comparable affinity (PDZ and LIM domain protein 1; K_Zn = 3.2×10⁻¹⁵ M) (Kochańczyk et al., 2015). Because p53 is a tetramer, we measured K_Zn of the full-length protein to determine if there is cooperativity between the monomers (Figure 2A). Fitting the data to the Hill-binding equation reveals K_Zn = (0.4 ± 0.1) x 10⁻¹⁵ M and n = 0.99 ± 0.02, indicating that each monomer binds zinc independently and that the isolated DBD is an accurate representation of the DBD in the full-length p53 tetramer. We then measured K_Zn,U by competition assay between urea-denatured apoDBD and the fluorescent Zn²⁺ chelator FluoZin-3 (Figure 2B). The apparent K_Zn,U value of (42 ± 7) x 10⁻⁹ M is 10⁷-fold weaker than K_Zn. We generated a Ser mutant of the zinc-binding residue C176 and found its K_Zn,U to be essentially identical ((50 ± 23) x 10⁻⁹ M), indicating that the zinc-binding sites in the native and unfolded proteins are distinct (Figure 2B).

To delineate the relationship between zinc binding and WT DBD folding, we performed urea denaturation experiments at various concentrations of [Zn²⁺]_free that were fixed by a mixture of Zn²⁺ chelators (Figure 2—figure supplement 1A). The overall free energy change for folding to either apo or holo native states (ΔG_fold) is given by (Equation 3; Pace and McGrath, 1980), where ΔG_Zn = -RT·ln(1+K_Zn⁻¹[Zn²⁺]_free), ΔG_Zn,U = -RT·ln(1+K_Zn,U⁻¹[Zn²⁺]_free), and ΔG_apo = -RT·ln(K_apo) (Equation 4).(3)ΔGfold⇄ΔGapo− ΔGZn+ ΔGZn,U(4)U ⇄N; Kapo

Zinc-induced stabilization can be visualized intuitively by plotting ΔG_fold against [Zn²⁺]_free (Figure 2C). Focusing on the WT DBD data, the curve begins as a horizontal line that intersects the y-axis at ΔG_apo, then starts to deflect upward when zinc begins to bind apoDBD, i.e. when [Zn²⁺]_free ≈ K_Zn. ΔG_fold continues to increase linearly with log[Zn²⁺]_free and only levels off when zinc starts to bind unfolded DBD ([Zn²⁺]_free ≈ K_Zn,U). A pure zinc-binding mutant such as R175H weakens ΔG_Zn (Equation 1) without affecting ΔG_apo, shifting the deflection point to higher [Zn²⁺]_free but maintaining the same y-intercept as WT (Figure 2C). Conversely, the signature of pure stability mutants such as A138V, a well-established temperature-sensitive variant (Cuddihy et al., 2008) is a similar deflection point compared to WT but a y-intercept value closer to zero. We observed no trend in the cooperativity parameter (m-value) of the unfolding transitions over 10⁻¹⁶ M < [Zn²⁺]_free <10⁻¹² M (Figure 2—figure supplement 1A), signifying that urea denaturation is adequately described by a two-state transition between unfolded and native apo or native holo states (the m-values of the two are similar [Butler and Loh, 2003]) throughout the tested zinc concentration. We also performed denaturation experiments starting with either apoDBD or zinc-bound DBD and obtained similar results, indicating that metal binding had reached equilibrium during the incubation time (Figure 2—figure supplement 1B). The data are described well by Equation 3 and yield fit parameters of ΔG_apo = 6.4 ± 0.1 kcal mol⁻¹ and K_Zn = (7.0 ± 2.5) x 10⁻¹⁵ M, both of which are in good agreement with direct measurements (Figure 3—source data 1).

Because our DBD experiments must be performed at low temperature to avoid protein aggregation, we sought to gain insight into how ΔG_apo changes with temperature. To quantify this, we performed urea melts of apoDBD at 10 temperatures over the range 2–22°C. The resulting curve is a parabola the narrowness of which is proportional to ΔC_p, the change in heat capacity on protein unfolding at constant pressure (Figure 2D). We fit these data to the Gibbs-Helmholtz equation (Figure 2—figure supplement 1C) to obtain ΔC_p, melting temperature (T_m), and the enthalpy change of unfolding at T_m (ΔH_m). Introduced by Pace and Laurents, 1989, this method has been shown to reproduce ΔC_p, T_m, and ΔH_m obtained by scanning calorimetry for a number of proteins (Pace et al., 1999; Talla-Singh and Stites, 2008). We then combined the full four-state model (Equations 1, 2, 4) with the Gibbs-Helmholtz equation to project the energy landscape of WT DBD as a three-dimensional plot with [Zn²⁺]_free, temperature, and fraction holoDBD (F_holo) on the x-, y-, and z-axes, respectively (Figure 3A).

When the protein is kept cool (10°C) it is stable even without zinc (ΔG_apo = −6.3 kcal mol⁻¹; Figure 2D), making F_holo effectively unity at all relevant concentrations of free zinc (>10⁻¹⁵ M; Figure 3A). At 37°C, however, the protein is completely unfolded without zinc (ΔG_apo = 6.9 kcal mol⁻¹). Strikingly, the high zinc-binding affinity of WT apoDBD together with the concentration of free zinc in the typical cell (10⁻¹⁰ M) (Krezel and Maret, 2006; Vinkenborg et al., 2009) result in an overall ΔG_fold near zero and F_holo = 0.43 (Figure 3A). This suggests that under normal physiological conditions, WT p53 is balanced on the edge of folding and unfolding and may be pushed in either direction by surprisingly small changes in [Zn²⁺]_free, ΔG_apo, K_Zn, or temperature. For example, raising [Zn²⁺]_free to 10⁻⁹ M or lowering it to 10⁻¹¹ M yields F_holo values of 0.9 and 0.08, respectively. F_holo changes similarly when K_Zn is altered by a factor of 10, ΔG_apo by 1–2 kcal mol⁻¹, and temperature by 2°C.

The conclusion that WT p53 folding is sensitive to [Zn²⁺]_free fluctuations in the physiological regime carries significant implications, and uncertainty of the ΔG_apo extrapolation as well as technical artifacts must be considered. The accuracy of the extrapolation and of our zinc-binding model as a whole can be evaluated by comparing our results to those of Fersht and coworkers, who applied the same Gibbs-Helmholtz analysis to holoDBD stability over a similar temperature range (5–25°C) (Bullock et al., 2000). To connect the two data sets, we calculated F_holo using our energy landscape model with [DBD]_total and [Zn²⁺]_total set to those used in the Fersht group’s experiments (2.5 μM each). We then obtained the theoretical stability of holoDBD (ΔG_holo,theory) using Equation 5 and compared this to the value determined by Fersht et al. (ΔG_holo).(5)ΔGholo,theory= −RT⋅lnFholo1−Fholo

At 10°C, where ΔG_apo and ΔG_holo are both determined experimentally, ΔG_holo,theory (−12.3 kcal mol⁻¹) and ΔG_holo (−10.6 kcal mol⁻¹) are in good agreement. At 37°C, where ΔG_apo and ΔG_holo are extrapolated values, ΔG_holo,theory and ΔG_holo are nearly identical (−2.9 kcal mol⁻¹ and −3.0 kcal mol⁻¹, respectively). Given DBD’s high-affinity for zinc (K_Zn = 1.6 fM), the only way to obtain ΔG_holo = −3.0 kcal mol⁻¹ at 37°C and 2.5 μM total zinc is for apoDBD to be as unstable as the landscape model predicts.

We note, however, that the ΔC_p value we obtain is substantially higher than the figure calculated based on the change in accessible surface area of holoDBD upon unfolding (3.8 kcal mol⁻¹ K⁻¹) (Bullock et al., 2000; Myers et al., 1995). To eliminate the possibility of buffer or denaturant-specific effects, we changed the buffer from Tris to phosphate and the denaturant from urea to guanidine hydrochloride and obtained nearly identical results (Figure 2—figure supplement 1E). There was no trend in m-value that would signify deviation from two-state behavior (Figure 2—figure supplement 1D). We therefore conclude that apoDBD, owing to its large ΔC_p value, possesses a stability that is unusually dependent on temperature.

The energy landscape of mutant DBD can be generated in the same manner as that of WT DBD provided that the mutation does not alter the enthalpy change of unfolding. This scenario is reasonable for the zinc-binding class, because these mutations do not perturb ΔG_apo and thus ΔH and ΔS can be assumed to remain unchanged. Stability-class mutations, however, can affect ΔH, ΔS, or both, making the extrapolation of ΔG_fold to 37°C unreliable.

We then sought to apply this methodology to gain insight into the mechanisms by which tumorigenic mutations cause p53 to lose function, we purified 22 DBD missense mutants commonly found in human cancer. We chose the most frequent somatic mutations in the IARC database with the following additional criteria: (i) when multiple mutations were reported at a single position, only the most common variant was used (to maximize protein coverage); (ii) if the residue was mutated to Trp, we selected the next most frequent alteration at that position (an extra Trp interferes with our fluorescence assays); (iii) when multiple mutations of a zinc-coordinating residue were common, we chose the most isosteric in order to help isolate the effects of Zn²⁺ binding. We performed urea denaturation experiments as a function of buffered [Zn²⁺]_free to determine ΔG_apo and K_Zn at 10°C (Figure 3—figure supplement 1 and Figure 3—figure supplement 2), and we also cross-checked K_Zn using direct Zn²⁺ binding experiments (c.f. Figure 2A) for select mutants. We were able to determine ΔG_apo for all mutants and K_Zn for all but three (Y205C, Y220C, Y163C; vide infra) (Figure 3—source data 1).

By placing the 22 variants on a plot according to their ΔG_apo and K_Zn values, three classes of mutants emerge: pure zinc-binding, pure stability, and DNA-binding, with a sub-class of mixed zinc-binding/stability phenotype (Figure 3C). Pure zinc-binding mutants exhibit decreased Zn²⁺ affinity but normal stability and thus cluster near a vertical line drawn directly below WT DBD (green box). Members of this class include R175H as well as the direct Zn²⁺ ligating mutants C176S, C242S, and H179R. Significantly, the analysis reveals a number zinc-binding mutations (P152L, L194F, R282Q) that were not previously suspected as being so due to their long distances from the metal-binding pocket (Figure 1).

Pure stability mutants are less stable than WT but suffer no zinc-binding deficiency; these are defined by falling near the horizontal line drawn to the left of WT DBD (red box in Figure 3C). To validate that protein stability can indeed be separated from zinc affinity in this manner, we destabilized WT apoDBD using sub-denaturing amounts of urea and measured K_Zn using the Tyr fluorescence assay. These data (open circles) fall on the horizontal line as predicted by the model. The mixed-phenotype mutants (A138V, V267F, R158H, M237I; orange box in Figure 3C) decrease both stability and zinc affinity. Only three mutations (R248Q, R273H, R280K) do not impair stability or metal affinity (blue box in Figure 3C). These residues are in direct contact with DNA (Figure 1). DNA-contact mutants are more stable than WT, likely because they remove or reposition a positive charge and reduce repulsions in the highly cationic DNA-binding groove (Butler and Loh, 2003; Bullock et al., 2000). Our data confirm that they bind zinc normally and have lost DNA-binding affinity (vide infra).

Figure 3C reveals a remarkable range in magnitude of the stability and zinc-binding deficiencies. The most severe member of each class reduces ΔG_apo by 4.6 kcal mol⁻¹ (E285K) and K_Zn by 5 orders of magnitude (C176S). Additionally, several zinc-binding or mixed mutations (L194F, R282Q, V157F, R158H, P152L) lie far from the canonical zinc-binding pocket, suggesting long-range communication between distant regions of DBD. The subtle nature of DBD structure and energetics is also evident from the observation that adjacent mutations can produce markedly different effects. For example, R248Q is a classic DNA-contact mutant whereas R249S is a pure zinc-binding mutant. This demonstrates the difficulty in inferring physical consequences of p53 mutation based on location and stresses the need to measure properties of each mutant to gain a clear understanding of its impairment.

To gain a better understanding of the effects of p53 mutation on DNA-binding activity, we measured dissociation constants (K_DNA) for binding of the 22 DBD mutants to a panel of fluorescently labeled oligonucleotides bearing 10 p53 recognition elements (p53RE; Supplementary file 1A). The fluorescence anisotropy data fit adequately to the Hill-binding equation. When global fits for individual sequences were performed by linking the Hill parameter and curve amplitudes, the resultant curves fit equally well regardless of DBD mutant, indicating similar binding mechanisms regardless of mutation (Figure 4—figure supplement 1A). Figure 4 represents the K_DNA values relative to that of WT DBD in the form of a heat map. We excluded the two p53REs to which none of the DBD variants bound (WAF1-3’ and EGFR), and assigned K_DNA = 25 μM (the weakest value we could consistently measure) to data sets in which we could detect no interaction (Figure 4—source data 1).

By visual inspection, most of the stability and mixed mutants maintain similar binding affinity for the different p53REs, whereas all of the DNA-contact mutants and most of the zinc-binding mutants lose measurable affinity for all p53REs (Figure 4). The majority of zinc-binding mutants purify with sub-stoichiometric but detectable zinc content (e.g. R175H contains 0.6 equivalents of Zn^{2+ 14}), and consequently should show partial DNA-binding activity. We speculated that the loss of DNA-binding may be caused by Zn²⁺ misligating to non-native sites (K_Zn,U = 42 x 10⁻⁹ M; Figure 2B) that out-compete the native site at the temperature of protein expression (18°C). To test this hypothesis, we attempted to re-establish native metallation status to the zinc-binding mutants P152L, R175H, H179R, and L194F by first removing all bound metal and then remetallating using a EGTA/ZnCl₂ buffering system. This procedure restored measurable DNA-binding affinity (WAF1 oligonucleotide) to all four mutants (Figure 4—figure supplement 1B). The positive control (WT) and negative control (R280K) yielded the expected results of normal and undetected DNA-binding, respectively. These data suggest that zinc misligation contributes to loss of DNA-binding activity for this class of mutants.

Interestingly, while most of the data are explained by global increases and decreases in affinity based on thermodynamic category, there do seem to be mutation-specific effects as well. P152L maintains WT-like affinity for WAF1 5', GADD45, and RGC, but loses all measurable affinity for PUMA, Type IV collagenase, p53RFP, MDM2, and BAX. As another example from the stability mutant category, Y220C maintains affinity for WAF1 5', GADD45, PUMA, and Type IV Collagenase, but seems to lose affinity for p53RFP, MDM2, BAX, and RGC. Additionally, E285K gains affinity for all p53REs, but the increase in affinity ranges from threefold for WAF1 5' to 15-fold for MDM2. E285K may bind more tightly to all DNA sequences due to enhanced electrostatic interactions with the phosphate backbone afforded by the negative-to-positive charge reversal near the active site. These results indicate that there are conserved functional consequences of the thermodynamic impairments we measure for p53 mutants that account for the majority of the functional differences we see in DNA-binding phenotype, but there are also mutation-specific effects that are better explained by idiosyncratic structural changes caused by each mutant that our model does not capture.

Y163C, Y205C, Y220C, and Y234C are all destabilized (ΔG_apo = 3.5–5.0 kcal/mol), but, unlike all other members of the stability class except for M237I, their ΔG_fold values fail to increase with [Zn²⁺]_free (Figure 3—figure supplement 1). The flat profiles of these plots can potentially be explained by very weak K_Zn values, but two observations argue against this interpretation. First, direct measurement of K_Zn by Tyr fluorescence yields a value of (1.9 ± 0.2) x 10⁻¹⁵ M for Y234C (Figure 3—source data 1), the only Tyr→Cys mutant that was able to survive the zinc removal procedure without aggregating. Second, Y205C, Y220C, and Y234C retain WT-like DNA-binding activity, whereas most of the zinc-binding mutants are compromised in this regard. We hypothesized that the extra Cys residue forms a new Zn²⁺ interaction site in the unfolded state, thereby bringing K_Zn,U closer to K_Zn and siphoning the native state of the stabilization energy it would normally receive by binding Zn²⁺. To test this hypothesis, we denatured Y234C apoDBD in 6 M urea and measured K_Zn,U by Tyr fluorescence. Unfolded Y234C binds zinc twice as tightly as WT [K_Zn,U = (20 ± 4) x 10⁻⁹ M; n = 6], consistent with the hypothesis with the caveat that the unfolded state in urea is likely different from that in buffer. As an additional test, we made the Y234A mutant and found that this mutation restores the relationship between ΔG_fold and [Zn²⁺]_free as predicted by the model (Figure 3—figure supplement 2), with a K_Zn value identical to that of WT within experimental error (Figure 3—source data 1). These results demonstrate that Y234A is a stability-class mutant and suggest that other Tyr→Cys mutants may also be members of this category. The data also indicate that introducing Cys can encourage zinc misligation in the unfolded state and destabilize native p53 through manipulation of K_Zn,U in our model.

We previously demonstrated that the small-molecule zinc metallochaperone ZMC1 can reactivate p53 mutants of the zinc-binding class in cultured cells. ZMC1 forms a 2:1 complex with Zn²⁺ and acts as an ionophore to transport Zn²⁺ into the cell, whereupon it buffers [Zn²⁺]_free to 10–20 nM (Yu et al., 2014). When ZMC1 was added to human cancer cell lines homozygous for p53^R175H or one of the direct zinc ligation mutations (p53^C176F, p53^C238S, p53^C242S), cell toxicity was observed with EC₅₀ values well below that of p53^WT and p53-null controls. This enhanced ZMC1 sensitivity was shown to be due to a p53-mediated apoptotic program (Yu et al., 2012). Given the ZMC mechanism, we previously surmised that the spectrum of mutants to which ZMCs were amenable was limited to those with impaired zinc binding. However, the relationship of zinc binding to the energy of protein folding we observed in our model allowed us to hypothesize that raising intracellular zinc concentrations could increase protein stability enough to rescue wild-type conformation of some stability mutants. To test this, we used ZMC1 as a tool to raise intracellular concentrations of zinc in an array of zinc-binding, stability, and mixed classes of p53 mutants generated by site specific mutation in plasmids and expressed in H1299 (p53-null) cells. We also used human tumor cell lines that endogenously express mutant p53, when available. Cell survival curves are shown in Figure 5 and ZMC1 EC₅₀ values are summarized in Table 1.

The mutants tested in the pure zinc-binding class are R175H, C176S, L194F, P152L, and R282Q. Cells transfected with R175H, C176S, and L194F show >100 fold enhanced sensitivity to ZMC1 (EC₅₀ = 0.002 μM – 0.005 μM) relative to untransfected and vector-only controls (Table 1). A p53^L194F tumor cell line is also sensitive to ZMC1 (Figure 7—figure supplement 1A). EC₅₀ values of R282Q and P152L are not detectable in the tested range, suggesting that they are not functionally reactivated by ZMC1. R249S belongs to the zinc-binding class and we previously found that Hs700T cells (p53^R249M) were not sensitive to ZMC1 (Yu et al., 2014). It is possible that mutations of R249, being adjacent to the DNA-contact residue R248, distort the structure of the DNA-binding groove regardless of zinc-binding status. The negative control group consists of the DNA-contact mutants R273H and R280K. As expected, these variants show no increased sensitivity to ZMC1, indicating that their DNA-binding defects cannot be ameliorated in a zinc-dependent fashion.

The most discriminating test of the thermodynamic model is whether the pure stability-class (Y234C, Y234A, V272M, E285K) and mixed stability/zinc-binding class (M237I) of p53 mutants can be reactivated by elevating intracellular zinc. V272M is marked by pronounced ZMC1 sensitivity, with an EC₅₀ value (0.002 μM) lower than that observed for any of the zinc-binding class mutants. This result demonstrates that a mutation whose sole consequence is to promote unfolding of p53 can be rescued by zinc. Y234C, E285K, and M237I, however, fail to show ZMC1 sensitivity. The negative result for Y234C agrees with the biophysical data (Figure 3—figure supplement 2), which suggested that the extra Cys residue forms a competing, non-native zinc-binding site in the unfolded state. We tested that hypothesis using the Y234A mutant. As predicted by the zinc misligation model, the EC₅₀ value of Y234A drops to a value comparable to that of the pure zinc-binding mutants R175H and L194F (Table 1). To confirm the result for M237I, we tested two human cancer cell lines that express p53^M237I. One cell line (T98G) was insensitive to ZMC1 and the other (SUM149PT) showed a partial response in which cell viability remained at ~40% at 10 μM ZMC1 (Figure 7—figure supplement 1B). The BXPC3 human cancer cell line bearing p53^Y220C was also insensitive to ZMC1 (Figure 7—figure supplement 1B). 

To further test the thermodynamic model, and to gain insight as to why some mutants fail to regain their cell-killing functions in the presence of ZMC1, we assayed the extent to which elevated zinc can refold mutant p53 in cells. Refolding was monitored by the reduction in staining by the antibody PAB240, which binds to a cryptic epitope only exposed on p53 unfolding or misfolding (residues 212–217 [Stephen and Lane, 1992]). For the zinc-binding class of mutants, the variants that exhibit low EC₅₀s (R175H, C176S, L194F) show reduced binding to PAB240 after ZMC1 treatment (Figure 6A and Figure 7), implying that they undergo zinc-mediated refolding. The human cancer cell line T47D bearing the p53^L194F mutation also showed refolding of mutant p53 by reducing PAB240 staining and increasing PAB1620 staining (Figure 7—figure supplement 1C). Antibody staining of P152L and R282Q did not change after ZMC1 treatment, consistent with their lack of sensitivity to the drug (Figure 6A). We also found that ZMC1 did not refold R249M in Hs700T cells (Yu et al., 2014). It is possible that these mutants misfold and/or aggregate in the cell to a conformation that is not amenable to zinc binding or refolding. Another potential explanation for why ZMC1 does not refold certain mutants in cells when there is evidence of refolding in vitro is related to zinc homeostatic mechanisms. Zinc homeostatic genes can be dysregulated in cancer, for example in breast cancer subtypes (luminal, basal, triple-negative) in which cellular zinc levels are elevated compared to normal mammary epithelial cells, and expression of zinc homeostatic genes is perturbed (Chandler et al., 2016). Overexpression of metallothioneins or zinc exporting proteins (ZnTs), which muffle ZMC1’s metallochaperone activity, may also provide a mechanism of resistance. As expected, staining intensities of the DNA-contact mutants R273H and R280K do not change after ZMC1 treatment. R280K cells appear dark in the absence of ZMC1, consistent with the DNA-contact phenotype of this mutant, but R273H stains brightly. This result suggests that the R273H mutation induces misfolding or aggregation in addition to loss of DNA-binding affinity by direct contact.

For the pure stability-class mutants, the PAB240 refolding results (Figure 6A) agree well with the EC₅₀ data (Table 1): the mutants that regain cell-killing activity in the presence of ZMC1 (Y234A and V272M) show reduced PAB240 antibody staining, and the mutants that are insensitive to ZMC1 (Y234C and E285K in H1299 cells and Y220C in a cancer cell line (Figure 7—figure supplement 1B)) react equally well with PAB240 before and after drug treatment. The mixed stability/zinc-binding mutant M237I is an interesting exception. M237I shows the greatest decrease in PAB240 staining after ZMC1 treatment of all mutants tested, yet M237I cells are insensitive to the drug. Increasing intracellular zinc therefore appears to refold M237I, as predicted by our model, but fails to restore its apoptotic activity. In agreement, the human cancer cell lines with p53^M237I mutation also showed refolding after treatment (by decreasing PAB240 staining after treatment of ZMC1) (Figure 7—figure supplement 1E) but was not sensitive to ZMC1 (Figure 7—figure supplement 1B). The M237I mutation may compromise p53 function by an additional mechanism such as introducing a structural defect in the folded protein or perturbing a binding interaction between p53 and another protein. We further tested the DNA-binding of p53^M237I to the p53RE in CDKN1A promoter in cells using a luciferase reporter assay and found that the M237I mutant was unable to activate transcription (Figure 6—figure supplement 1).

As immunofluorescence is sometimes criticized for its possible interference with protein structure by fixation of cells, we sought to confirm refolding of the p53 mutants by performing immunoprecipitation using PAB240 followed by western blot for p53 proteins. Consistent with the immunofluorescent staining of the individual cells, R175H, Y234A, V272M showed decreased p53 protein by PAB240 pull-down after ZMC1 treatment, while the p53 band intensities remained similar for Y234C, P152L and E285K (Figure 6B). As a positive control, we treated the temperature-sensitive A138V with ZMC1 and observed a 2-fold decrease in PAB240 pull-down (Figure 6—figure supplement 2).

To determine if the conformational change observed with the mutants results in restoration of WT p53 transcriptional function, we compared mRNA levels of the p53-responsive genes PUMA and NOXA before and after 24 hr of ZMC1 treatment, using the transfected cells described above. Mutants for which no increase in transcription is observed for either of the two probe genes (blue or white bars in Figure 6C) are P152L and E285K (stability class), M237I (mixed class), and R273H (DNA-contact class). All these mutants are insensitive to ZMC1-mediated cell killing and fail to refold as judged by the PAB240 test (except M237I). Mutants for which ZMC1 treatment enhances transcription of at least one gene (one or more beige or red bars in Figure 6C) are R175H, C176S, L194F, R282Q (zinc-binding class), Y234A, Y234C, and V272M (stability class), and R280K (DNA-contact class). Thus, all the mutants that exhibit low EC₅₀ values and undergo ZMC1-dependent refolding also have their transcriptional activities at least partially restored by the drug. Similarly, the human cancer cell line with p53^L194F also showed decreased mutant p53 protein (Figure 7—figure supplement 1C) and the induction of the gene expression of CDKN1A and PUMA (Figure 7—figure supplement 1E), indicating reactivation of p53 transcriptional function. The transcriptional assay, however, also finds that ZMC1 elevates mRNA levels of at least one gene for two stability mutants (Y234C and R282Q; Figure 6C) that failed to show enhanced cell killing or refolding in the presence of ZMC1. The increase in mRNA levels appears to be insufficient to bring about apoptosis.

The stability mutant V272M is a particularly interesting case because it is efficiently refolded by ZMC1 (Figure 6A) and is threefold more sensitive to the drug than the pure zinc-binding mutant R175H (Figure 8). This behavior is consistent with the thermodynamic model. The energy landscape of R175H indicates that raising intracellular [Zn²⁺]_free from 100 pM to 100 nM increases the percentage of folded R175H from 0.0028% to 2.8% (Figure 3B). V272M is destabilized by 2.0 kcal mol⁻¹ and if we assume that this relatively small perturbation does not change the enthalpy of unfolding, then V272M will achieve 2.8% refolding at only 10–20 nM [Zn²⁺]_free. Both IF and IP confirm ZMC1 refolds V272M, whereas ZMC1 fails to induce refolding of the negative control (R273H) (Figure 6A; Figure 6B). Finally, ZMC1 elicits high level of PUMA and NOXA expression in V272M (Figure 6C). These findings demonstrate that V272M is robustly activated by ZMC1.

E285K, the most severe of the stability class of mutants, failed to refold in the presence of ZMC1 (Figure 6A; Figure 6B) and cells transfected with p53^E285K were not sensitive to the drug (Table 1). We speculated that the extreme instability of E285K prevented it from refolding despite the elevated intracellular zinc concentrations afforded by ZMC1 treatment, as well as its increased affinity for most p53REs relative to WT. We asked whether reducing temperature could act synergistically with ZMC1 to reactivate E285K. The temperature of p53^E285K H1299 cell cultures was lowered to 22°C for 4 hr in the presence of ZMC1 to allow for temperature-assisted refolding, then returned to 37°C for cell-killing assays. This procedure resulted in 3.8-fold decrease in EC₅₀ compared to the control in which 37°C was maintained throughout (Figure 7A, Figure 7B). Of note, we observed a similar effect for R175H in H1299 (2.4-fold decrease) and TOV112D cells (1.6-fold decrease), suggesting that low temperature acts synergistically with zinc-binding and stability mutants alike, as predicted by the thermodynamic model. We further determined that reducing temperature to 22°C successfully induced ZMC1-mediated refolding of E285K (Figure 7C).

ZMCs are currently in pre-clinical development and represent a viable strategy to reactivate mutant p53 in the clinic. One of the attractive features of the ZMC program is that the spectrum of patients that will potentially respond to the drugs is known (those that harbor zinc-binding class p53 mutations). To determine if this spectrum should include individuals that have stability-class mutations, we sought to obtain pre-clinical evidence for this using the xenograft tumor assay. We treated the animals with the tumors derived from the stable cell lines with V272M and E285K mutations with ZMC1 and found that the growth of the V272M tumors was inhibited but growth of E285K tumors was not (Figure 9; Figure 9—figure supplement 1), implying that ZMC1 can be used to reactivate certain p53 stability mutations, in addition to the zinc-binding deficient mutations.

We hypothesized from the energy landscape map that at physiological conditions of temperature and available zinc concentration, WT p53 may exist in approximately equal populations of folded and unfolded molecules. To address this question in context of the cell, we employed three cell lines expressing WT p53: MCF7, U2OS, and H460. We immunoprecipitated p53 protein from cell lysates with PAB240 or PAB1620 and blotted with the pan-p53 antibody DO-7 (Figure 10). Because the two antibodies may have different affinities for p53, the percentages of unfolded and folded p53 cannot be determined quantitatively from this experiment. The intensities of the PAB240 bands, however, are comparable to those of the PAB1620 bands, suggesting that a significant fraction of WT p53 is unfolded in all three cell lines.

Samples sizes were adjusted to provide smooth interpolation across the experimental range; experimental range spanned at least one order of magnitude greater the highest and lower than the smallest central value (e.g. K_d, C_m, EC₅₀) unless constrained by physical limitations (solubility, buffer range, etc.). Number and type of experimental replicates are indicated in figure legends. Cell biological replicates consist of separate plates of cells from a common source. Data from replicate experiments were pooled prior to analysis. All data are shown and excluded outliers are marked. In nonlinear data, potential outliers were identified on inspection, and excluded if they met both of the following conditions: (1) removal of the outlier improved goodness of fit as judged by the average relative standard error of the fit parameters, and (2) upon exclusion and re-fitting, the excluded point lies outside the 95% prediction interval of the fit. For multiple outliers, the Holm-Bonferroni correction was applied when calculating the 95% prediction interval. R code for outlier testing is available at https://codeberg.org/AlanBlayney/mutant-p53. Curve fitting and data plotting were performed with SigmaPlot 13.0, GraphPad Prism, KaleidaGraph 4.5, and R 3.6.1, with additional annotation in Adobe Illustrator. Three-dimensional plots were generated in R using the ‘lattice’ package (R Development Core Team, 2020; Sarkar, 2008).

DBD genes (p53 residues 94–312, no expression or purification tags) were cloned into pET23a plasmids and transformed into E. coli BL21(DE3) cells. Cultures were grown at 37°C to OD₆₀₀ = 0.6, induced with isopropyl β-D-1-thiogalactopyranoside, and expressed for 12–15 hr at 18°C. Following cell lysis and centrifugation, soluble DBDs were purified in 20 mM Tris (pH 7.2), 5 mM β-mercaptoethanol using SP-sepharose then heparin HiTrap cation exchange columns (GE Life Sciences) (11, 21). Mutant proteins were judged to be > 98% pure by SDS-PAGE with Coomassie brilliant blue staining. Apo proteins were generated by lowering pH to 4.8 using acetic acid in the presence of a large excess of EDTA, restoring pH to 7.5 using Tris, and removing EDTA and zinc with a DG10 desalting column (Bio-Rad) (9). Full-length p53 was expressed as a fusion construct with a cleavable N-terminal expression tag consisting of HisTag fused to ribose-binding protein derived from Thermoanaerobacter tengcongensis. FL-p53 was purified by nickel-NTA chromatography (Qiagen) following manufacturer’s protocols. The HisTag and ribose-binding protein purification tag was then removed using human rhinovirus 3C protease, at which point FL-p53 was purified identically to DBD with the addition of a final clean-up step on a Superdex S200 size exclusion column (GE Life Sciences). FL-p53 was >90% pure by SDS-PAGE with Coomassie brilliant blue staining. All in vitro experiments were performed in 50 mM Tris (pH 7.2), 0.1 M NaCl, 10 mM β-mercaptoethanol unless otherwise noted.

ZnCl₂ stocks were dissolved in 30 mM HCl and their concentrations determined by titration with 4-(2-pyridylazo)resorcinol using ε₅₀₀ = 66,000 M⁻¹ cm⁻¹ for the PAR-Zn²⁺ complex. [Zn²⁺]_free concentrations were fixed at the indicated values by mixing 2 mM chelator (EDTA, N-(2-Hydroxyethyl)EDTA, EGTA, or diethylenetriaminepentaacetic acid) with 0.005–1.75 mM ZnCl₂. All chelators were purchased from Sigma Aldrich (St. Louis, MO). These total zinc concentrations were always in excess of protein concentration (1 μM). [Zn²⁺]_free values were calculated using the MAXCHELATOR software suite (http://somapp.ucdmc.ucdavis.edu/ pharmacology/bers/maxchelator/). Concentrations of Zn²⁺-chelators were determined by titrating the chelator solutions against pre-formed PAR-Zn²⁺ complex and measuring the decrease in absorbance in buffer. Concentrations of FZ3 solutions (Life Technologies, Norwalk, CT) were determined by equivalence point titration with Zn²⁺ in buffer.

Samples for urea denaturation experiments were prepared using a Hamilton Microlab 500 diluter, with final urea concentrations measured by index of refraction (Butler and Loh, 2003). For urea denaturation studies in the presence of zinc, samples contained chelator and ZnCl₂ at the concentrations indicated above. Trp fluorescence was measured using either a Fluoromax-4 spectrofluorometer (Horiba Scientific, Edison, NJ) in a 5 mm x 5 mm quartz cuvette or in a SpectraMax i3x plate reader (Molecular Devices, Sunnyvale, CA) in 96-well UV-Microplates (Thermo Scientific, Waltham, MA) with comparable results (λ_ex = 280 nm, λ_em = 355 nm). The resultant curves were fit to the 2-state linear extrapolation model with linear baselines to obtain ΔG and m-values (Pace, 1975; Schellman, 1975). To increase the accuracy of our measurements, m-values of all curves were pooled (m_pool = 3.10 ± 0.12 kcal mol⁻¹ M⁻¹, n = 213, mean ± SE) and ΔG was calculated from the midpoint of denaturation (C_m) of each curve, which is more accurately determined than m, according to the equation ΔG = C_m·m_pool. ΔG values were then fit to Equation 3 to obtain ΔG_apo and K_Zn. For variable temperature urea denaturation studies, samples were equilibrated at the indicated temperature for 15–24 hr, either in urea or GdnHCl denaturant. ΔG values were obtained by the pooled m-value method described above (in GdnHCl, m_pool = 7.01 ± 0.92 kcal mol⁻¹ M⁻¹, n = 6, mean ± SE). ΔG values were plotted as a function of T and fit to the Gibbs-Helmholtz equation (Figure 2—figure supplement 1C). K_Zn competition assays were carried out by incubating unfolded apoDBD (6 M urea) at the indicated concentrations with 15 nM ZnCl₂ and 30 nM FZ3 for 1 hr at room temperature. FZ3 fluorescence was then scanned and IC50 was obtained by fitting the data to Equation 6:(6)Fobs=A1+exp(IC50 − xb)where F_obs is observed fluorescence, A is the curve amplitude, x is log[DBD], and b is an empirical steepness parameter. K_Zn was then calculated from the Munson-Robdard solution to the Cheng-Prusoff equation (Equation 7; Munson and Rodbard, 1988):(7)KZn= IC501+ FZ3(y0+2)2K(y0+1)+y0−Ky0y0+2where FZ3 is the total concentration of FZ3, K is the dissociation constant of the FZ3·Zn²⁺ complex (15 nM per the manufacturer), and y0 is the ratio of bound FZ3 to free FZ3 in the absence of DBD (equal to unity in our conditions).

K_Zn measurements by intrinsic fluorescence were performed by incubating apoDBD or apo-FL-p53 with chelator/ZnCl₂ mixtures (as described above) for 16 hr at 10°C. Zinc binding was monitored by the slight increase in fluorescence at 306 nm (DBD) or 350 nm (FL-p53), with excitation at 280 nm. Data for DBD were fit to the single-site binding equation (Equation 8):(8)Fobs= F0+ A[ Zn ]freeKZn+ [ Zn ]freewhere F₀ is the fluorescence baseline and A is the amplitude of the curve. To determine if there was cooperativity of zinc binding to the FL-p53 tetramer, those data were fit to the Hill-binding equation (Equation 9):(9)Fobs= F0+ A([ Zn ]free)n(KZn)n+ ([ Zn ]free)nwhere n is the Hill parameter. Reported K_Zn values and Hill parameters were averaged from independent replicates.

K_DNA values were obtained using 5’-Cy3 labeled oligonucleotides (Eurofins Genomics, Louisville, KY; see Supplementary file 1A for sequences). Annealed oligonucleotides (50 nM) were incubated with 5–10,000 nM DBD on ice for 1 hr in 50 mM Tris pH 7.2, 100 mM NaCl, 1 mM TCEP, and 0.005% Tween-20 in black 96-well plates. Uncalibrated fluorescence anisotropy was measured using SpectraMax i3x equipped with the rhodamine fluorescence polarization module (G-factor = 1, λ_ex\λ_em = 535 nm/595 nm) (Molecular Devices, Sunnyvale, CA). Data for all trials of all mutants for a given sequence were subjected to global curve fitting with Equation 10, linking A and n across all trials.

The folding landscape of WT DBD was calculated from the four-state model, assuming that unfolded DBD binds only a single zinc ion, using Equation 10:(10)Fholo= [Zn]free⋅(KZn)−11+ [Zn]free⋅(KZn)−1+ 1+ [Zn]free⋅(KZn,U)−1Kapo 

Energy landscapes of the zinc-binding class of DBD mutants were generated from the WT landscape by subtracting the difference in the ΔG_apo value of the mutant from that of WT DBD (both measured at 10°C) at each temperature, and using the mutant K_Zn value in Equation 10. K_Zn values were assumed to be independent of temperature, and K_Zn,U values for WT and mutant DBD were assumed to be identical. R code for the generation of landscape plots is available at https://codeberg.org/AlanBlayney/mutant-p53.

TOV112D, H1299, T47D, SUM149PT, T98G, Calu-3, U2OS and V138 cell lines were cultured in DMEM with 10% FBS. BXPC3, MCF7 and H460 were cultured in RPMI with 10% FBS. TOV112D, H1299, T47D, SUM149PT, T98G, Calu-3, MCF7, U2OS and H460 were purchased from American Type Culture Collection (ATCC). V138A was a gift from Dr. Arnold Levine. Cell lines were authenticated by examination of morphology, genotyping by PCR and growth characteristics. All cell lines were determined to be Mycoplasma free using the Universal Mycoplasma Detection Kit (ATCC 30–1012K, identifier 30–1012K). Mycoplasma testing methods and results are included as Supplementary file 2. Vectors for expressing p53 mutants were generated by site-directed mutagenesis using the Q5 Site-directed mutagenesis kit (New England Biolabs), following the manufacturer’s instructions. The initial plasmid with human TP53 R273H mutant was purchased from OriGene. The oligonucleotides for the mutations are listed in Supplementary file 1B. ZMC1 was synthesized by Rutgers Molecular Design and Synthesis group, Office of Research and Economic Development (Zaman et al., 2019).

Cells at 80–90% confluence were transfected with Lipofectamine 3000 (Thermo Fisher), following the manufacturer’s instructions. Expression of p53 protein was confirmed by Western blot. For generation of stable cell lines, H1299 cells were transfected with a vector encoding various p53 mutants generated by site-directed mutagenesis. Cells were then selected for G418 resistance. Single positive clones were isolated and stably maintained in G418-containing medium.

5,000 cells per well were cultured in 96-well plates such that 50% confluence was reached after one day. At this point, serial dilutions of ZMC1 were added and incubation continued for 3 days. Viability was then measured by Calcein AM (Trevigen, Gaithersburg, MD). Variable temperature incubations were performed as described in the text. Growth inhibition was quantified as absolute EC₅₀ by fitting the average data from three replicates to a three-parameter sigmoid.

Immunofluorescent staining was performed as described previously (Yu et al., 2012). The conformation of folded and misfolded/unfolded p53 were recognized by the antibodies PAB1620 (1:50) and PAB240 (1:400) (EMD Millipore) respectively, with overnight binding. The secondary antibody, goat anti-mouse IgG, was incubated for 40 m. Fluorescent staining intensity was quantified using ImageJ software (NIH).

For immunoprecipitation, cells were harvested and lysed using RIPA buffer. The lysates were incubated with protein A/G beads with PAB240/PAB1620 or IgG overnight at 4°C. The IP products were detected by Western blot using the pan-53 antibody DO-7. IgG was used as a negative control. The input total lysates were detected by western blot with p53 antibody DO-1 and actin served as the internal loading control. IgG, DO-7, DO-1, actin antibody, and protein A/G beads were purchased from Santa Cruz Biotechnology.

The p53 recognition element in the CDKN1A promoter region, constructed in pGL3 vector, was a gift of Dr. Carol Prives (Columbia University, New York, NY). It was transfected into the cells in 96-well plate, followed by the treatment of 1 μM ZMC1 for 48 hr. The luciferase reporter assay was then performed using Dual-Glo luciferase assay system (Promega) following the manufacturer’s instructions.

Mice were housed and treated according to guidelines established by the Institutional Animal Care and Use Committee of Rutgers University, who also approved all mouse experiments (animal protocol PROTO99900044, approval date 10/16/2019 – 10/15/2022). Nude mice (NCR nu/nu) were purchased from Taconic Biosciences. Xenograft tumors were generated from the stable tumor cell lines H1299-V272M and H1299-E285K (1 × 10⁷ cells/tumor site/mouse). Tumor dimensions were measured every 1–4 d and their volumes (V) were calculated by using the formula: V = (length × width² × π)/6. Tumors were allowed to grow to 50 mm³ at which point the mice were randomly allocated to treatment and control groups and ZMC1 or vehicle was administered.