Dissecting molecular details and functional effects of the high‐affinity copper binding site in plasminogen activator Inhibitor‐1

Abstract

Plasminogen activator inhibitor‐1 (PAI‐1) is the primary inhibitor for plasminogen activators, tissue‐type plasminogen activator (tPA) and urokinase‐type plasminogen activator (uPA). As a unique member in the serine protease inhibitor (serpin) family, PAI‐1 is metastable and converts to an inactive, latent structure with a half‐life of 1–2 hr under physiological conditions. Unusual effects of metals on the rate of the latency conversion are incompletely understood. Previous work has identified two residues near the N‐terminus, H2 and H3, which reside in a high‐affinity copper‐binding site in PAI‐1 [Bucci JC, McClintock CS, Chu Y, Ware GL, McConnell KD, Emerson JP, Peterson CB (2017) J Biol Inorg Chem 22:1123–1,135]. In this study, neighboring residues, H10, E81, and H364, were tested as possible sites that participate in Cu(II) coordination at the high‐affinity site. Kinetic methods, gel sensitivity assays, and isothermal titration calorimetry (ITC) revealed that E81 and H364 have different roles in coordinating metal and mediating the stability of PAI‐1. H364 provides a third histidine in the metal‐coordination sphere with H2 and H3. In contrast, E81 does not appear to be required for metal ligation along with histidines; contacts made by the side‐chain carboxylate upon metal binding are perturbed and, in turn, influence dynamic fluctuations within the region encompassing helices D, E, and F and the W86 loop that are important in the pathway for the PAI‐1 latency conversion. This investigation underscores a prominent role of protein dynamics, noncovalent bonding networks and ligand binding in controlling the stability of the active form of PAI‐1.

Abbreviations

PAI‐1

plasminogen activator inhibitor‐1

serpin

serine protease inhibitor

tPA

tissue‐type plasminogen activator

uPA

urokinase‐type plasminogen activator

RCL

reactive center loop

VN

vitronectin

IMAC

immobilized metal chromatography

HDX‐MS

hydrogen‐deuterium exchange coupled to mass spectrometry

wt

wild‐type

ITC

isothermal titration calorimetry

IPTG

Isopropyl β‐D‐1‐thiogalactopyranoside

HIC

hydrophobic interaction chromatography

SMB

somatomedin B domain

VMD

visual molecular dynamics

1. INTRODUCTION

Plasminogen activator inhibitor‐1 (PAI‐1) is a member of the serine protease inhibitor (serpin) family, and it serves as the primary inhibitor for both tissue‐type plasminogen activator (tPA) and urokinase‐type plasminogen activator (uPA). ¹ , ² The inhibition of plasminogen activators (PAs) controls the crucial initiation event that triggers the fibrinolysis cascade and leads to the conversion of plasminogen to plasmin. ³ Thus, PAI‐1 plays an essential role in the regulation of the fibrinolytic pathway, with broad‐ranging effects that influence hemostasis, inflammation, tissue remodeling and wound healing. ⁴ , ⁵ Inhibition of PAI‐1 expression results in excessive bleeding and accelerated fibrinolysis. ⁶ , ⁷ On the other hand, elevated PAI‐1 can lead to decreased clot breakdown and promotes fibrosis in various tissues. ⁸ Imbalances in PAI‐1 levels are associated with a variety of pathologies, including cardiovascular disease, inflammatory conditions, thrombosis, and cancer progression. ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷

Whereas PAI‐1 is produced in a variety of cell types, ¹⁸ , ¹⁹ it is mainly stored in platelets. PAI‐1 is also found in the bloodstream and the subendothelial matrix. Active PAI‐1 adopts the canonical serpin fold, characterized by three β‐sheets (a–c) and eight or more α‐helices (a–i) packed mostly on one side of the major β‐sheet A ²⁰ (Figure 1). The anti‐proteolytic inhibitory function of PAI‐1 is achieved through the prominent solvent exposed loop, termed the reactive center loop (RCL), which bridges s5A on the N‐terminal end and s1C on its C‐terminal end. ²¹ The RCL contains the same scissile bond as the natural protease substrate and functions as a “bait” to trap PAs. ²² This bond lies between the P1 and P1' residues, Arg 346 and Met 347, respectively. Upon attack by the protease, an acyl‐intermediate initially forms, as dictated by the well‐known serine protease reaction mechanism. The feature that distinguishes inhibition of the protease by the serpin, in contrast to the subsequent hydrolysis of the acyl‐intermediate by water that occurs with protease substrates, is a dramatic conformational rearrangement that results in a stable, long‐lived acyl bond that covalently tethers the serpin and enzyme. This occurs when the acylated RCL rapidly inserts into the central β‐sheet, repositioning the covalently‐bound protease as it moves to the distal end of the serpin, with a concomitant distortion of the catalytic triad within the active site to an inactive geometry. ²³

FIGURE 1.

Structures of active (left, PDB code: 3Q02) and latent (right, PDB code: 1DVN) PAI‐1. The solvent exposed reactive center loop (RCL) is added as a red dashed line in the active PAI‐1 structure; it is inserted between central β‐sheet s3A and s5A (colored in pink) in the latent structure. The N‐terminal histidines H2 and H3 are represented as gold ball and stick structures. The W86 loop is colored in purple, and helix D is colored in green. All structures were generated in visual molecular dynamics (VMD) ⁵⁵

PAI‐1 is unique among serpins due to its metastable fold, even in the absence of proteases, with a half‐life in its active conformation of only ~1–2 hr. ²⁴ , ²⁵ Under physiological conditions, the RCL spontaneously inserts into the central β‐sheet; this conformational change results in a latent form of PAI‐1 that is inactive because the RCL is no longer surface‐exposed or accessible to proteases ²⁶ , ²⁷ (Figure 1). The rate at which PAI‐1 relaxes to the latent conformation is influenced by environmental factors, including pH, temperature, and ionic strength. ²⁵ , ²⁸ Also, the glycoprotein vitronectin (VN) is an important cofactor that stabilizes PAI‐1, increasing the half‐life by ~50%. VN forms a tight complex with circulating PAI‐1 and localizes the serpin to sites of injury. ²⁹ , ³⁰ The binding between VN and PAI‐1 is a high‐affinity interaction, with a K_d between 0.1 and 1 nM. ²⁹ , ³¹ , ³²

Interestingly, the stabilizing effect of VN on the half‐life of PAI‐1 is modulated by metals. ³³ , ³⁴ We became intrigued with metal interactions with PAI‐1 because immobilized metal chromatography (IMAC) columns charged with transition metals were used to bind to recombinant wild‐type PAI‐1 and afforded a convenient purification strategy. ³⁵ Subsequently, the effects of different transition metals (magnesium, calcium, manganese, iron, cobalt, nickel, and copper) on the stability of PAI‐1 were investigated. ³³ The results indicated that these metals are separated into two groups: Type I metals (magnesium, calcium, and manganese) exhibit a modest stabilization of PAI‐1; Type II metals (cobalt, nickel, and copper) significantly destabilize PAI‐1 alone, but enhance its stability when full‐length VN or the N‐terminal somatomedin B (SMB) domain from VN is present. ³³ Complementary stopped‐flow fluorescence measurements also indicated a high binding affinity between Cu(II) and PAI‐1 (K_d ~ 0.09 μM), ³⁴ within the physiological range of copper in human circulation and tissues. ³⁶ , ³⁷

The early studies were followed by the use of hydrogen‐deuterium exchange coupled to mass spectrometry (HDX‐MS), which indicated that Cu(II) binding not only destabilizes PAI‐1, but also increases dynamics within PAI‐1 in key structures/sequences involved in the active‐to‐latent transition. ³⁸ Conversely, concurrent binding of Cu(II) and VN to PAI‐1 leads to decreases in dynamics within those same regions, consistent with the observation that Cu(II) stabilizes PAI‐1 bound to VN. ³⁸

We have sought to identify the sites for Cu(II) binding to PAI‐1. Our initial predictions derived from the examination of the crystal structure (pdb code: 3Q02), which features a zinc ion coordinating with N‐terminal residues H2 and H3. We suspected that the N‐terminus of PAI‐1 could function as an intrinsic “His‐tag” to bind transition metals and provide the mechanistic basis for the IMAC step in PAI‐1 purification. Thus, a PAI‐1 variant with replacements at the two N‐terminal histidines, denoted H2AH3A PAI‐1, was constructed to explore this potential binding site. Cu(II) binding to wild‐type (wt) PAI‐1 and H2AH3A PAI‐1 was evaluated by isothermal titration calorimetry (ITC) and kinetic measurements. ³⁹ The results indicated that H2 and H3 are indeed involved in a high‐affinity Cu(II)‐binding site. Interestingly, binding of Cu(II) to this site was shown to prolong the active conformation of PAI‐1 rather than promote the PAI‐1 latency transition when VN is absent. ³⁹ This binding site with H2 and H3 is responsible for the synergistic effects of Cu(II) and VN that stabilize PAI‐1 beyond the effect of VN alone. ³⁹

With the roles of H2 and H3 at the N‐terminus defined, we now face additional questions regarding metal binding to PAI‐1: 1. What other residues complete the metal coordination sphere with these two N‐terminal histidines? 2. Are other histidines in PAI‐1 likewise involved in binding to metal? 3. Is there a concentration dependence of metal effects on PAI‐1, both in the presence and absence of VN? 4. What is the mechanism for effects of metals upon binding to the high‐affinity site containing H2 and H3? To answer these questions, protein engineering was used to identify other residues in PAI‐1 hypothesized to bind Cu(II) and mediate functional effects on PAI‐1, both in the presence and absence of VN. ITC was utilized to characterize the thermodynamic features of Cu(II) binding to these variants. Kinetics to measure PAI‐1 stability and gel‐based copper titration assays were also performed to evaluate the roles of targeted residues in mediating Cu(II) effects on PAI‐1 stability. The comprehensive strategy that utilized this variety of experimental approaches was important, as it revealed unanticipated insights into the mechanism of metal‐induced changes in PAI‐1 stability that would not have been apparent from the results of only one method in isolation.

2. RESULTS

2.1. Rationale for studies to localize additional metal‐binding residues in PAI‐1

Coordination chemistry of metal–ligand binding suggests that Cu(II) prefers strong binding with the side chains of histidines over other residues. The d9 electron system of Cu(II) prefers a binding geometry coordinated by 4, 5 or 6 ligands, in square planar, square pyramidal, or axially distorted octahedral geometries, respectively. ⁴⁰ A high‐affinity Cu(II) binding site also requires a minimal coordination number of 4, for example, including three nitrogen donors (histidines or other) and 1 oxygen donor. ⁴¹

Residues in the vicinity of H2 and H3 were assessed as potential partners in coordinating metal. Potential metal‐binding clusters were identified, proximal to the N‐terminal high‐affinity metal binding site with H2 and H3, that could include residues E81, H364 or H10 (Figure 2). Molecular dynamics queries were used to further evaluate the potential for these amino acids to contribute to a coordination sphere for Cu(II). A comparison of the dynamic PAI‐1 structure (obtained from 10 ns molecular dynamics [MD] simulation) versus the static structure from x‐ray crystallography (pdb code: 3Q02) revealed significant movement of the N‐terminal H2, H3 residues towards E81A and H364, generating a much smaller pocket with appropriate geometry to coordinate zinc (Figure 2). The distances between zinc and all residues in this coordination sphere are shorter in the MD structure compared to the crystal structure. Most notably, the distance between zinc and the OE1 atom of E81 (labeled as d1 in Figure 2) decreased from 8.93 Å to 1.95 Å. A comparison of the distances between the metal and the various amino acids in the crystal structure and MD simulated structure is given in Table 1.

FIGURE 2.

Molecular dynamics simulation on PAI‐1 predicts a tight metal‐binding pocket. The crystal structure of active PAI‐1 (silver) (pdb code: 3Q02) and the PAI‐1 structure obtained from 10 ns MD simulation (pink) are superimposed. The potential high‐affinity metal binding site is shown, including residues H2, H3, H10, E81, and H364. Residues in the crystal structure are represented in thin line structures, whereas residues from the MD structure are shown in thicker bond structures. Histidines are colored in gold and glutamate is colored in green. Zinc is shown as a cyan ball. The distance between zinc and the OE1 atom of E81 is labeled as d1. The distance between zinc and the NE2 atom of H2 is labeled as d2. The distance between zinc and the NE2 atom of H3 is labeled as d3. The distance between zinc and the NE2 atom of H364 is labeled as d4. The distance between zinc and the NE2 atom of H10 is labeled as d5. All distances measured in the crystal structure are shown in black; the corresponding distances measured in the MD structure are shown in red

TABLE 1.

Geometry of metal ligation within a putative high‐affinity binding site in PAI‐1

Label	Point‐to‐point distance definition	Distance in crystal structure	Distance in structure obtained from MD simulation
d1	ZN:E81(OE1)	8.93 Å	1.95 Å
d2	ZN:H2(NE2)	3.41 Å	2.23 Å
d3	ZN:H3(NE2)	2.95 Å	2.21 Å
d4	ZN:H364(NE2)	9.41 Å	8.83 Å
d5	ZN:H10(NE2)	13.32 Å	15.64 Å

Note: Distances (in Å) from metal to individual PAI‐1 residues were measured to compare the geometry of the meta‐binding pocket in the crystal structure (pdb code: 3Q02) and the structure obtained from 10 ns MD simulations, as shown in Figure 2.

As noted above, histidine (H10) also lies in close proximity to H2 and H3 and could alternatively contribute to this N‐terminal metal‐binding site. The MD simulation also shows a change in the distance between zinc and H10 in the binding pocket. However, the distance from the zinc to the NE2 atom of H10 (labeled as d5 in Figure 2) actually increases slightly, from 13.32 to 15.64 Å, an opposite change to that observed with H364 and E81. Based on these MD simulations, we designed experiments to test the role of residues E81, H364 and H10 in high‐affinity binding to Cu(II). Although MD analysis suggested that H364 was a more likely candidate to contribute to the N‐terminal metal‐binding cluster, H10 was in close enough proximity to H2 and H3 to warrant investigation in our mutagenesis studies.

2.2. Effects of Cu(II) on the rate of conversion of PAI‐1 to the latent form are concentration‐dependent

The effects of Cu(II) binding on PAI‐1 stability have been studied previously by our lab using a variety of buffer conditions. ³³ , ³⁹ The results indicated that free PAI‐1 is destabilized by Cu(II) binding, whereas PAI‐1 is actually stabilized by Cu(II) binding when SMB is also bound. These effects were observed using fixed concentrations of Cu(II) with PAI‐1. To more fully understand these unusual metal effects, we investigated the Cu(II) effects over a broader Cu(II) concentration range, aiming to establish whether different effects were observed at sub‐saturating vs. saturating metal concentrations. This was prudent because the mutagenesis strategy may perturb binding constants, which could confound interpretation of kinetic measurements made only at a single metal concentration.

As shown in Figure 3, the effect of Cu(II) to destabilize or stabilize PAI‐1, both in the absence and presence of SMB, does in fact vary according to metal concentration. At low Cu(II) concentrations, the results indicate that wtPAI‐1 is actually stabilized, both in the absence and presence of SMB. This result clearly expands on our previously published work, which used significantly higher concentrations of Cu(II) and indicated the destabilization of PAI‐1 without SMB bound. At intermediate Cu(II) concentrations, wtPAI‐1 is destabilized by Cu(II) binding in the absence of the SMB domain, but stabilized when SMB is present, in keeping with our previous observations. ³⁹ Finally, at the highest Cu(II) concentration tested, wtPAI‐1 is destabilized by Cu(II) binding both in the absence and presence of SMB. This also is a new finding, contrasted with the prior characterization of destabilization of PAI‐1 by Cu(II) alone, and stabilization by Cu(II) with SMB bound to PAI‐1.

FIGURE 3.

Stability of wtPAI‐1 over a broad concentration range of Cu(II). The half‐life for PAI‐1 conversion from the active to latent form is plotted on the y‐axis versus Cu(II) concentration on the x‐axis. Horizontal dashed lines depict the value of the half‐life of wtPAI‐1 with no Cu(II) present. Half‐lives from stability measurements in the absence of SMB are shown as closed circles, and those from measurements in the presence of SMB are shown as open circles

Clearly, the previously unrecognized concentration dependence of the Cu(II) effect on the stability of PAI‐1 is an important consideration as we analyze properties of PAI‐1 variants. As a baseline for comparison with the new variants pursued here, we evaluated how the half‐life of the H2AH3A PAI‐1 was affected by Cu(II) +/− SMB over the same broad range in Cu(II) concentrations. The half‐lives for wtPAI‐1 (Figure 4a) and H2AH3A PAI‐1 (Figure 4b) in the absence of metals are indistinguishable, both in the presence and absence of the SMB domain. Furthermore, as noted previously, ³⁹ the half‐life at high concentrations of metals is decreased for H2AH3A PAI‐1 compared to wtPAI‐1, both in the presence and absence of the SMB domain. However, at no concentration does H2AH3A PAI‐1 exhibit an increase in its half‐life, as observed with wtPAI‐1. In contrast to wtPAI‐1, which is stabilized at low concentrations of metals (with or without the SMB domain) and also is stabilized by all except the highest concentration of Cu(II) when bound to SMB, there are no conditions under which H2AH3A PAI‐1 is stabilized by Cu(II). Results are summarized in Table 2.

FIGURE 4.

Half‐lives of wtPAI‐1 and H2AH3A PAI‐1 on Cu(II) concentrations and/or SMB binding. Data for wtPAI‐1 are shown in panel (a) and for H2AH3A PAI‐1 in panel (b). PAI‐1 was incubated at 0.1 μM in 50 mM MOPS,100 mM (NH₄)₂SO₄, 0.1 mM EDTA, pH 7.4 at 37°C. A total of eight conditions were tested, corresponding to four Cu(II) concentrations (0, 70, 170, and 270 μM), each in the absence or the presence of SMB. At various time points, PAI‐1 is mixed with 0.1 μM tPA to react with all PAI‐1 remaining in the active form. tPA activity is measured by addition of 1 mM (final) Spectrozyme tPA substrate and the absorbance of substrate cleavage is monitored at A₄₀₅ for 5 min. PAI‐1 inhibitory activity is plotted versus incubation time and fit to an exponential decay function to determine the rate of latency transition. Experiments were performed in triplicate

TABLE 2.

Effects of Cu(II) and SMB on the stability of PAI‐1 variants

[Cu(II)] PAI‐1 variant	0 μM	70 μM	170 μM	270 μM
wt PAI‐1	78.9 ± 3.2	123.9 ± 4.7	50.9 ± 1.6	25.0 ± 1.7
wt PAI‐1 + SMB	122.5 ± 1.5	194.7 ± 3.4	171.4 ± 4.4	90.5 ± 1.2
E81A	82.5 ± 1.8	82.3 ± 2.9	8.0 ± 0.3	5.7 ± 0.3
E81A + SMB	127.3 ± 0.9	127.7 ± 3.7	84.4 ± 4.8	55.5 ± 3.1
H364A	126.4 ± 5.8	126.8 ± 8.4	17.6 ± 1.0	13.3 ± 1.1
H364A + SMB	1,514.7 ± 7.1	1,523.1 ± 12.5	412.4 ± 9.7	259.5 ± 12.3
H2AH3A	72.7 ± 5.6	75.9 ± 6.8	13.9 ± 0.7	5.5 ± 0.2
H2AH3A + SMB	126.3 ± 2.7	119.5 ± 2.1	77.6 ± 2.5	37.3 ± 3.3
H10A	85.3 ± 4.0	83.0 ± 2.8	45.3 ± 4.8	25.3 ± 0.3
H10A + SMB	131.6 ± 1.6	211.3 ± 2.0	143.8 ± 5.0	103.5 ± 4.5

Note: The half‐lives (in minutes) for wtPAI‐1 and E81A, H364A, H2AH3A, and H10A PAI‐1 are reported at various four Cu(II) concentrations, with and without SMB.

2.3. Two variants, E81A and H364A PAI‐1, exhibit Cu(II) effects similar to H2AH3A PAI‐1

Based on the molecular dynamics simulation (Figure 2), we hypothesized that E81, H364 or H10 are additional ligands interacting with copper as a part of the Cu(II) coordination sphere including H2 and H3. We would predict the same pattern of copper effects as with H2AHA PAI‐1 to be observed for other residues that contribute to the metal‐binding site housing the two N‐terminal histidines in PAI‐1.

Stabilities of the H364A, E81A and H10A variants over a broad concentration range of Cu(II) were measured (Figure 5, Table 2). These measurements show that the E81A (Panel a) and H364A (Panel b) variants exhibit effects of Cu(II), both with and without SMB, that are similar to H2AH3A PAI‐1. Specifically, the E81A and H364A PAI‐1 variants are unaffected by low concentrations of copper, in contrast to wtPAI‐1, which is stabilized at low concentrations of copper, both in the presence and absence of SMB. At intermediate Cu(II) concentrations, E81A PAI‐1 and H364A PAI‐1 are both destabilized, regardless of the presence of SMB, differing significantly from the stability profile of wtPAI‐1 at intermediate Cu(II) concentrations that confer destabilization in the absence of SMB and stabilization in the presence of SMB. Also, at the highest Cu(II) concentration, E81A PAI‐1 and H364A PAI‐1 are both more destabilized than wtPAI‐1.

FIGURE 5.

Effects of Cu (II) and/or SMB binding on half‐lives of PAI‐1 variants. Data for E81A PAI‐1 are shown in panel (a); H364A PAI‐1 in panel (b); and H10A PAI‐1 in panel (c). PAI‐1 variants were incubated at 0.1 μM in 50 mM MOPS,100 mM (NH₄)₂SO₄, 0.1 mM EDTA, pH 7.4 at 37°C. A total of eight conditions were tested, corresponding to four Cu(II) concentrations (0, 70, 170, and 270 μM), each in the absence or the presence of SMB. At various time points, PAI‐1 is mixed with 0.1 μM tPA to react with all PAI‐1 remaining in the active form. tPA activity is measured by addition of 1 mM (final) Spectrozyme tPA substrate and the absorbance of substrate cleavage is monitored at A₄₀₅ for 5 min. PAI‐1 inhibitory activity is plotted versus incubation time and fit to an exponential decay function to determine the rate of latency transition. Experiments were performed in triplicate

An obvious difference between the E81A and H364A variants is an increase in the half‐life of H364A PAI‐1 that results from the single amino acid substitution; the half‐life is even more highly elevated in the presence of SMB. In the absence of Cu(II), E81A PAI‐1 exhibits a half‐life similar to wtPAI‐1, both in the presence and absence of SMB, whereas H364A PAI‐1 has a half‐life that is modestly increased in the absence of SMB, but which is over 10‐times longer than the half‐life for wtPAI‐1 when SMB is present. With this extended half‐life, the effect of Cu(II) on the H364A PAI‐1:SMB complex is marked. At high Cu(II) concentrations with SMB, the rate of conversion to the latent form of H364A is 5.8‐times faster than with SMB alone, while the rate of conversation of wtPAI‐1 is only 1.4‐times faster with Cu(II) and SMB than it is without added metal.

In contrast to the results with E81A and H364A PAI‐1, results with the H10A PAI‐1 variant (Figure 5c) show stability patterns more akin to wtPAI‐1 (Figure 4a) than the H2AH3A variant (Figure 4, panel B) (and see Table 2). The only difference in stability measured over this broad set of conditions comparing H10A PAI‐1 and wtPAI‐1 is that H10A PAI‐1 is not stabilized at low concentrations of Cu(II).

In total, our results indicate that two variants, E81A and H364A PAI‐1, are similar to H2AH3A PAI‐1 in terms of Cu(II) effects on their stability profiles across all concentrations measured. Each of these substitutions, E81A, H364A, and H2AH3A, eliminates the stabilizing effect of low Cu(II) concentration, both in the absence and presence of the SMB domain. Binding of Cu(II) promotes conversion of these forms of PAI‐1 to the latent conformation, but, in contrast to wtPAI‐1, there is no stabilization observed of either variant with both SMB and Cu(II) bound. These results are consistent with the predictions of the MD simulations. These results suggest that E81 and H364 may coordinate copper along with H2 and H3. H10 does not appear to be part of this coordination sphere.

2.4. Changes in Cu(II) effects on the PAI‐1 variants are also observed in gel‐based sensitivity assays

We previously developed a gel‐based copper sensitivity assay that is complementary to the kinetic measurements of latency to assess how PAI‐1 stability is affected by Cu(II) over a wide range of concentrations. ³⁹ This assay identifies copper concentrations that invoke an accelerated rate of latency conversion on PAI‐1. As described previously, ³⁹ PAI‐1 samples were incubated for 30 min with Cu(II) concentrations ranging from 0–1,000 μM. After the 30‐min incubation with metal, the PAI‐1 sample was then combined with tPA to test for remaining active PAI‐1 in the mixture. The samples were analyzed by non‐reducing SDS‐PAGE and stained with Coomassie Blue, and the band intensities of the PAI‐1/tPA complex at each copper concentration were quantified using gel densitometry (Figure 6a). At low concentrations of Cu(II), the data on wtPAI‐1 indicate that it exhibits robust complex formation with tPA because most of the serpin is still active 30 min after addition of metal. It should be noted that the lowest concentrations of Cu(II) actually increase wtPAI‐1 activity, consistent with the finding from kinetic assays that small amounts of copper stabilize PAI‐1 in the presence or absence of the SMB domain. At intermediate copper concentrations, there is clearly a mixture of inactive wtPAI‐1 and active wtPAI‐1 that forms a complex with tPA, and at high metal concentrations, all PAI‐1 is latent so that no PAI‐1/tPA complex is observed.

FIGURE 6.

Copper titration gel assay for PAI‐1. Panel (a) shows a sample gel for analysis of samples at the various concentrations levels. wtPAI‐1 was incubated with varying concentrations of CuSO₄ (ranging from 5 to 1,000 μM) at 37°C for 30 min and then mixed with an equimolar concentration of single‐chain tPA. Lanes 1–3 show control samples for PAI‐1, tPA, and the PAI‐1/tPA mixture, respectively. Lanes 4–16 correspond to samples of PAI‐1/tPA mixtures with various concentrations of CuSO₄ as indicated. The last lane contains protein standards, with sizes indicated to the right of the gel. Panel (b) PAI‐1/tPA complex formation is plotted as a function of the total Cu(II) concentration. Results are shown for wtPAI‐1 (closed circles), H2AH3A PAI‐1 (open circles), E81A PAI‐1 (closed squares), H364A PAI‐1 (closed triangles), and H10A PAI‐1 (open triangles)

We measured the gel‐based Cu(II) titration profiles to probe the roles of H10, E81, and H364 and compare them to results observed with H2AH3A PAI‐1, with a damaged Cu(II)‐binding site (Figure 6b). The gel sensitivity assay results for H10A PAI‐1 are similar to those observed for wtPAI‐1, with little difference in the amount of active PAI‐1 measured in this assay across the Cu(II) concentration range comparing the H10A variant and wtPAI‐1. In contrast to the results with wtPAI‐1, H10A PAI‐1 activity is not increased at low Cu(II) concentrations. This observation is consistent with the stability kinetic measurements (Figure 5) that showed no activation of H10A PAI‐1 by metal in the absence of the SMB domain.

The latency conversion of E81A PAI‐1 is more sensitive to Cu(II) binding compared to wtPAI‐1, similar in most respects to the PAI‐1 variants with N‐terminal histidine replacements at H2 and/or H3 that had been characterized previously ³⁹ and repeated here. Also, the stabilization of PAI‐1 at low Cu(II) concentrations is not seen in the E81A variant or with H2AH3A PAI‐1; this result is consistent with the kinetic assays that show no Cu(II) stabilization of these PAI‐1 variants over the entire Cu(II) concentration range, regardless of the presence of the SMB domain.

H364A PAI‐1 is somewhat more sensitive to Cu(II) than wtPAI‐1 in the gel assay, and the stabilization of H364A PAI‐1 is also not seen at low Cu(II) concentrations, an observation once again consistent with the stability assays presented above. The results of this gel assay for H364A PAI‐1 are more similar to those with H10A PAI‐1 or wtPAI‐1 and could, on first inspection, appear to suggest that H364A PAI‐1 is less sensitive to Cu(II) than either E81A PAI‐1 or H2AH3A PAI‐1. However, because the stability assays (Figure 5) have established that H364A PAI‐1 has a much longer half‐life than any of the other forms of PAI‐1, an alternative interpretation of this result is more likely. Because of the longer half‐life of H364A PAI‐1, more active PAI‐1 persists at the 30‐min time point across the Cu(II) concentration range. It must be recognized that the Cu(II) sensitivities measured with this method are influenced by the half‐lives of the latency conversion of the PAI‐1 variants in the assay and, thus, are not simple measures of affinity for the metal. Because the distribution of species on the gel is a function of both affinity for metal and rates of latency conversion of metal‐bound‐PAI‐1 and free PAI‐1, a direct comparison is not straightforward. The other PAI‐1 variants (H2AH3A, H10A, E81A and wtPAI‐1) have similar half‐lives and thus can be directly compared in terms of metal sensitivity, with H2AH3A and E81A variants much more sensitive than either H10A or wtPAI‐1.

The overall results with the Cu(II) titration assays in gels provided consistent results with the stability assays and suggest that H2, H3, and E81 may contribute to the high‐affinity metal‐binding site and/or mediate copper effects on the stability of PAI‐1. Consistently with the kinetic assays, the gel sensitivity measurements suggest that H10A is not involved in the high‐affinity binding site. Conclusions about the role of H364A are not definitive from this gel‐based assay because it is a snapshot from a single point in time (not an equilibrium measurement) and thus is sensitive to the kinetics of the latency conversion.

2.5. Thermodynamics of Cu(II) binding to PAI‐1 variants using ITC

To resolve the ambiguities from the kinetic and gel‐based metal sensitivity assays, we investigated the thermodynamic properties of metal binding to the PAI‐1 variants with ITC, used in our previous work to characterize Cu(II) binding to wtPAI‐1 and the H2AH3A variant. ³⁹ This technique provides an important complementary measure to the stability kinetics and gel sensitivity assays because it is a direct measurement of metal binding affinity. By monitoring heat exchange during titration of Cu(II) aliquots into a PAI‐1 solution, ITC provides full binding profiles, and analyses yield binding affinity, stoichiometry, enthalpy and entropy of interaction. Active wtPAI‐1 exhibits a tight Cu(II)‐binding site, whereas latent wtPAI‐1 contains a tight binding site as well as a weaker one. ³⁹ Previous work has demonstrated that substitutions for histidines at positions 2 and 3 significantly alter the ITC profiles for Cu(II) binding to PAI‐1. ³⁹ Notably, H2AH3A PAI‐1 exhibits only a single weak binding site in both the active and latent states. ³⁹

Results of ITC experiments monitoring Cu(II) binding to wtPAI‐1 and the variants tested are shown in Figures 7 and 8 and summarized in Table 3. Titrations of active wtPAI‐1 and latent wtPAI‐1 agree with previously published work ³⁹ and are shown in Figure 7, panels A and B for comparison with the other PAI‐1 variants tested. Binding to active H10A PAI‐1 is shown in Figure 7c and latent H10A PAI‐1 is shown in Figure 7d. Not surprisingly, H10A PAI‐1 exhibits binding behavior similar to wtPAI‐1, consistent with the other data presented above that show insignificant changes in copper effects. In a similar fashion to wtPAI‐1, the active conformation of H10A PAI‐1 exhibits a single high affinity binding site with a K _d ~ 90 nM, averages ΔG _obs and − TΔS _obs of −9.2 and 9.3 kcal/mol, respectively (Table 3). Also in line with latent wtPAI‐1, latent H10A PAI‐1 binds Cu(II) with two distinct binding transitions, best fit with a two‐site non‐symmetric binding model and yielding average K _obs terms of 1.5 × 10⁷ and 2.3 × 10⁴ M⁻¹ and ΔH _obs terms of −11.9 and − 24.2 kcal/mol. These analyses indicate the presence of high and low affinity binding sites, with average K _d values of 74.6 nM and 36.0 μM, respectively. Moreover, the best fits to the data indicate some cooperativity between the two binding sites, suggesting allosteric effects (Table 3).

FIGURE 7.

Isothermal titration calorimetry of wtPAI‐1 and H10A PAI‐1. Cu(II) titration of active wtPAI‐1 using ITC is shown in panel (a); latent wtPAI‐1 in panel (b); active H10A PAI‐1 in panel (c); and latent H10A PAI‐1 in panel (d). ITC data are baseline corrected in NITPIC software and fit to either a single‐site model (active wtPAI‐1 and H10A PAI‐1) or two‐site binding model (latent wtPAI‐1 and H10A PAI‐1) using SEDPHAT. The data and fit are represented in GUSSI by plotting heat of injection (kcal/mol) against the copper/PAI‐1 molar ratio

FIGURE 8.

Isothermal titration calorimetry of H364A and E81A PAI‐1. Cu(II) titration of active H364A using ITC is shown in panel (a); latent H364A PAI‐1 in panel (b); active E81A PAI‐1 in panel (c); and latent E81A PAI‐1 in panel (d). ITC data are baseline corrected in NITPIC software and fit to either a single‐site model (active and latent H364A PAI‐1) or two‐site binding model (active and latent E81A PAI‐1) using SEDPHAT. The data and fit are represented in GUSSI by plotting heat of injection (kcal/mol) against the copper/PAI‐1 molar ratio

TABLE 3.

Thermodynamics properties for Cu(II) binding to wtPAI‐1 and variants at 10°C

PAI‐1 variant	Model (n)	K obs, M−1	K d a	ΔG obs, (kcal/Mol) b	ΔH obs , (kcal/mol)	−TΔS obs, (kcal/Mol) c
Active wt	1	1.9 (±0.9) × 107	59.5 (±27.6) nM	−9.4 (±0.3)	−17.1 (±0.1)	7.7 (±0.3)
Latent wt	2	1.9 (±0.2) × 107 1.7 (±1.0) × 104	52.0 (±7.1) nM 73.6 (±45.6) μM	−9.4 (±0.1) −5.4 (±0.4)	−12.1 (±0.1) −12.9 (±1.0)	2.6 (±0.1) 7.5 (±0.6)
Active H10A	1	1.2 (±0.3) × 107	87.8 (±23.3) nM	−9.2 (±0.2)	−18.4 (±0.1)	9.3 (±0.2)
Latent H10A	2	1.5 (±0.6) × 107 2.3 (±1.3) × 104	74.6 (±32.3) nM 36.0 (±7.5) μM	−9.3 (±0.3) −5.6 (±0.4)	−11.9 (±0.1) −24.2 (±4.6)	2.6 (±0.2) 18.6 (±4.2)
Active E81A	2	5.7 (±0.9) × 107 2.5 (±1.8) × 104	17.0 (±2.8) nM 53.7 (±39.1) μM	−10.1 (±0.1) −5.6 (±0.5)	−17.5 (±0.1) −15.7 (±5.6)	7.4 (±0.1) 10.1 (±5.2)
Latent E81A	2	2.8 (±0.1) × 107 1.2 (±0.1) × 104	36.0 (±0.1) nM 83.0 (±7.8) μM	−9.6 (±0.1) −5.3 (±0.1)	−14.7 (±1.0) −15.3 (±2.0)	5.0 (±1.0) 10.1 (±2.1)
Active H364A	1	3.5 (±1.2) × 105	3.0 (±1.0) μM	−7.3 (±0.2)	−33.1 (±3.8) d	25.4 (±4.4) d
Latent H364A	1	4.8 (±0.7) × 104	21.2 (±3.3) μM	−6.1 (±0.1)	−43.9 (±8.6) d	37.8 (±8.7) d
Active H2AH3A e	1	4.2 (±0.9) × 104	24.0 (±5.0) μM	−6.0 (±0.1)	−20.0 (±12.0) d	14.0 (±12.0) d
Latent H2AH3A e	1	7.7 (±4.6) × 104	13.0 (±6.0) μM	−6.3 (±0.4)	−16.4 (±0.6) d	10.0 (±0.4) d

^a K _d calculated from the following relationship: K_d = 1/K_obs.

^b ΔG _{ob s} calculated from ΔG_obs = −RT ln K_obs.

^c −TΔS _obs term calculated from ΔG_obs = ΔH_obs − TΔS_obs.

^d Parameters measured at the low sensitivity limit of the ITC and likely have significant error associated with them.

^e Values previously published (Bucci JC, McClintock CS, Chu Y, Ware GL, McConnell KD, Emerson JP, Peterson CB (2017) J Biol Inorg Chem 22 (7):1123–1,135).

The ITC results for active and latent H364A PAI‐1 are shown in Figure 8a,b, respectively, characterized by markedly different titrations than observed with either wtPAI‐1 or H10A PAI‐1. H364A PAI‐1 exhibits only a single weak Cu(II)‐binding site in both the active and latent forms (Table 3), similar to that previously observed with H2AH3A PAI‐1. ³⁹ As discussed in our previous study, ³⁹ ΔH _obs terms for the weak binding events are likely inaccurate due to the low sensitivity limits of ITC. The corresponding K _d values are calculated to be 3.0 μM for the active conformation and 21.2 μM for the latent conformation of H364A PAI‐1. The results are essentially the same as observed with H2AH3A PAI‐1 ³⁹ and likely represent the same weak binding site that is manifested in both conformations of the PAI‐1 variant. The similarity in ITC titrations of H364A PAI‐1 and H2AH3A PAI‐1 is consistent with the hypothesis that H364 contributes to the metal‐binding site, along with the previously identified N‐terminal residues.

Surprisingly, the E81A form of PAI‐1 exhibits quite different results with Cu(II) titrations of the active (Figure 8c) and latent (Figure 8d) forms. In contrast to observations with the H364A variant, the ITC results indicated that both active and latent E81A PAI‐1 contain two Cu(II)‐binding sites with markedly different affinities. Indeed, this result closely resembles the Cu(II) binding behavior of latent wtPAI‐1 (Figure 7b), and mathematical fits of the titration data for active and latent E81A PAI‐1 give similar values to those from ITC with latent wtPAI‐1 (Table 3). The data indicate a tight Cu(II)‐binding site with a K _d of 17.0 nM and a weaker binding site with a K _d of 53.7 μM. Again, some cooperativity in binding between the two sites is apparent from the mathematical fits to the data.

3. DISCUSSION

Understanding copper effects on PAI‐1 stability has been a focus for our laboratory recently because copper reserves in physiological settings will influence the delicate balance needed for active PAI‐1 in circulating systems and tissues. Furthermore, imbalances in copper homeostasis impact prominent disease states that have also been tied to PAI‐1 function, including platelet malfunction, heart disease, and inflammation. ⁴² , ⁴³ In this study, we probed further to determine the molecular mechanisms for metal effects on PAI‐1. In doing so, we first gained new insights about mixed responses PAI‐1 exhibits over a range in Cu(II) concentrations, with and without added SMB. Surprisingly, we observed that PAI‐1 is actually stabilized at low Cu(II) concentrations. This finding is novel, as our other studies had only evaluated higher concentrations of Cu(II) that clearly destabilized PAI‐1. ³³ , ³⁹ It also clarified our observation in SDS‐gel‐based Cu(II) sensitivity assays that showed an increase in PAI‐1 stability at low metal concentrations.

3.1. New insights into metal binding and stabilization of the active conformation of PAI‐1

The H2AH3A PAI‐1 variant, generated previously to study the N‐terminal binding site, was shown to alter the Cu(II) effects on PAI‐1 stability and the thermodynamics properties of Cu(II) binding with PAI‐1. ³⁹ Prior work showed that the N‐terminal site corresponds to the high‐affinity metal‐binding site in PAI‐1 and is essential for the synergistic effect of metal and SMB in stabilizing PAI‐1. In this study, we investigated other histidine residues, H10 and H364, near the N‐terminal H2 and H3, for their possible roles as Cu(II) binding ligands. We also investigated the oxygen‐containing residue E81, which is proximal to the N‐terminal histidine site in the three‐dimensional fold of PAI‐1. Kinetic studies evaluating copper effects on the rate of latency conversion indicated that only the E81A and H364A PAI‐1 exhibit similar profiles as H2AH3A PAI‐1, while H10A PAI‐1 exhibits only small insignificant differences compared to wtPAI‐1. For H10A PAI‐1, there is no stabilization in the low Cu(II) concentration range without added SMB, but other effects of metal and SMB remain intact. The subtle difference in Cu(II) effects on H10A PAI‐1 compared to wtPAI‐1 are likely attributable to local structural changes in the vicinity of Helix A. Thus, the stability kinetics profiles of these variants suggested that E81 and H364, together with H2 and H3, could be involved in the tight Cu(II) binding that stabilizes PAI‐1.

ITC was also performed to study the thermodynamics properties of Cu(II) binding with these variants, and it revealed a different picture. H364A PAI‐1 exhibits a similar Cu(II) binding profile as H2AH3A PAI‐1, with a single weak Cu(II)‐binding site, consistent with the hypothesis that this residue is part of the coordination sphere for metal. On the other hand, E81A PAI‐1 exhibits Cu(II) binding at two sites, a high affinity site with a K _d of 17.0 nM and a weaker affinity site with a K _d of 53.7 μM. This was a surprising result, since the stability kinetics and metal sensitivity assay using SDS‐PAGE showed impaired responses to Cu(II) binding of E81A PAI‐1.

How can these two results be reconciled? Previous studies ³⁸ , ⁴⁴ have illustrated the way structural dynamics play essential roles in PAI‐1 conformational changes during the latency conversion. Fluctuations within key PAI‐1 regions, including helices D, E, F and the W86 loop, are thought to be essential for the PAI‐1 latency conversion. HDX‐MS and computational simulations show major differences in dynamics in these regions comparing the active and latent forms of PAI‐1. ⁴⁴ Furthermore, helices D and F exhibit profound deformations during micro‐second simulations, ⁴⁴ consistent with a temporal role of fluctuations within these structural elements in early transitions that lead to latency. E81 is located within helix D in close proximity to the W86 loop, within one of the most dynamic regions in PAI‐1 that is involved in an early unfolding event in the pathway to latency of PAI‐1.

Our prior HDX‐MS work with Cu(II) and PAI‐1 showed that there is increased mobility in the peptide containing E81 upon binding of copper, consistent with the increased rate for the latency conversion in the presence of copper. ³⁸ This increased H‐D exchange in the vicinity of E81 is also observed upon copper binding to the H2AH3A PAI‐1 variant. ³⁸ In fact, the enhancement in the H‐D exchange behavior is more pronounced with the H2AH3A PAI‐1 variant compared to wtPAI‐1, suggesting that metal binding to the residual weaker Cu(II)‐binding site separate from the N‐terminal cluster has the predominant influence on helix D fluctuations. Indeed, it is this weaker site that appears to drive the faster conversion to latency. ³⁹ Consistently the binding of the SMB domain or SMB and Cu(II) simultaneously to PAI‐1 results in depressed dynamic behavior within the same Helix D peptide region. ³⁸

Thus, we propose that the mechanism for copper and SMB effects requires changes in protein dynamics that are influenced by E81, although direct involvement of this residue in the Cu(II) coordination sphere with the N‐terminal residues H2 and H3 may not be required. A reasonable model is that Cu(II) binding at the high‐affinity metal‐binding site, including H2, H3, and H364, influences the available population of conformational fluctuations of helix D, wherein E81 resides. Restriction of the conformational alternatives of the helix D‐W86 loop region to only a subset is consistent with the smaller overall increase in dynamics in the vicinity of E81 observed upon Cu(II) binding to wtPAI‐1 compared to the H2AH3A variant lacking the high‐affinity metal‐binding site. ³⁹ The conformational bias in this selection appears to disfavor the rapid collapse to the inactive, latent structure.

If the aspartate side chain from residue 81 is part of the tetravalent coordination site with H2, H3 and H364, it is possible that the oxygen donor from the carboxylate could be replaced with other atoms in the E81A PAI‐1 variant lacking the aspartate side chain. For instance, several water molecules appear available near the metal‐binding site in the crystal structure ⁴⁵ and could replace aspartate to coordinate the metal. Alternatively, E81 may solely play a role in transducing the effects of metal binding to distal regions of PAI‐1 through restricted conformational dynamics and may not be directly involved in Cu(II) coordination. This set of experiments cannot distinguish between these possibilities.

3.2. A single amino acid substitution leads to stabilization of PAI‐1

The work pursued in this set of studies revealed another unexpected finding – that the half‐life of PAI‐1 in the active conformation is increased due to a single amino acid replacement of H364 with alanine. Furthermore, SMB binding to H364A PAI‐1 increases its half‐life to a value that is almost 12‐times longer than the half‐life of wtPAI‐1 bound to SMB. How might this single amino acid substitution affect the stability of PAI‐1 so significantly? A comparison of the crystal structures available for active and latent PAI‐1 provides some insight. Interactions with amino acids in close proximity to H364, which differ in the active and latent forms of PAI‐1, provide a likely mechanism underlying the effect of this substitution. One significant difference noted between the active and latent structures is the movement of the W86 loop towards H364 that occurs in the latency conversion. A new interaction is formed between N87 and H364 in the latent form of PAI‐1 (Figure 9). Previous atomistic computer simulations and HDX‐MS studies ⁴⁴ have indicated that the W86 loop is extremely dynamic in active PAI‐1. The fluctuations of the W86 loop are proposed to lead to the progressive detachment of β‐strand 2A from β‐strand 3A and act as the “first breach” in the conversion of PAI‐1 to the latent form.

FIGURE 9.

Structural model of PAI‐1 highlighting the region surrounding H364. The three‐dimensional structure of PAI‐1 in the vicinity of H364 is shown for the active (left, PDB code: 1DVM) and latent (right, PDB code: 1DVN) forms of PAI‐1. The W86 loop is colored in purple, and helix‐D is colored in green. H364, N87, and E81 are shown in ball and stick structures, and W86 is shown in a line structure representation. The distances between H364 and N87 are shown for comparison in the active (10.44 Å) and latent (3.49 Å) structures. The distances between H364 and E81 are also shown and compared in the active (4.43 Å) and latent (5.70 Å) structures

A cursory analysis of available structures for active and latent PAI‐1 shows that the distance between the side‐chain carbonyl oxygen of N87 and the side‐chain imidazole nitrogen of H364 is uniformly shorter within the latent structures, consistent with the formation of a hydrogen bond between the imidazole nitrogen on residue 364 and carbonyl oxygen from the amide group at position 87. ²⁸ , ⁴⁵ , ⁴⁶ , ⁴⁷ Such bonding could guide the W86 loop fluctuations and the β‐strand opening toward the latent PAI‐1 structure; substitution of alanine for H364 would disrupt the bonding so the restricted movements of the flexible W86 loop would not be incurred. The rate of the conformational changes leading to the PAI‐1 latency conversion is thus proposed to be slower in the absence of this interaction between H364 and N87.

Interestingly, H364 was identified a number of years ago as the single residue responsible for the pH‐dependent variation in stability exhibited by PAI‐1. ⁴⁸ The half‐life of PAI‐1 increases as pH decreases, and the protonation of the imidazole side chain to provide a positive charge at H364 was suggested as the mechanism for this effect. ⁴⁸ The original model suggested that the positively charged residue near the N‐terminus and in the vicinity of β‐sheet B could prevent insertion of the RCL into the central β‐sheet C. Our analyses indicate an alternative interaction that could stabilize active PAI‐1 – an ionic bond between the H364 and E81 (Figure 9). Lower pH would favor formation of a salt bridge between a protonated imidazole on H364 with the carboxylate on E81 to stabilize the active conformation of PAI‐1. Clearly, this interaction would form only in the absence of metal. Thus, pH and metal coordination would influence H364 interactions with one of two residues near the W86 loop, E81 or N87, alternatively favoring the active or latent conformation.

In summary, the coordination chemistry for Cu(II) binding to the high‐affinity binding site containing two residues near the N‐terminus of PAI‐1 has been further elucidated and shown to involve H364, distal in sequence but proximal in the three‐dimensional fold of the protein. The mechanism of metal‐binding effects on the stability of PAI‐1 appears to involve the carboxylate side chain of E81 as a key lever in regulating copper effects. Without this single carboxyl group, PAI‐1 is extremely sensitive to inactivation by metals. The differences between H364 and E81 contributions to metal effects on PAI‐1 were only apparent from utilizing a combination of kinetic analysis and thermodynamic binding methods. Also, the single substitution of alanine for histidine at position 364 increases the half‐life of PAI‐1 in the active conformation significantly, an effect that is exacerbated upon the binding of the SMB domain resulting in a half‐life that is increased more than 10‐fold compared to wtPAI‐1 with SMB bound. Effects of metals, vitronectin and/or amino acid substitutions are mediated through W86 loop fluctuations and perturbations of dynamic networks. These changes ultimately are manifest in the changes in overall stability of the serpin, which result from the metastable fold of active conformation of PAI‐1.

4. MATERIALS AND METHODS

4.1. Molecular dynamics simulation of PAI‐1

The PAI‐1 crystal structure (PDB code: 3Q02) was minimized, solvated, minimized again, equilibrated, and simulated in the NAMD program ⁴⁹ using the CHARMM36 force field ⁵⁰ to obtain the dynamics structure. A periodic boundary condition was used. The particle mesh Ewald ⁵¹ algorithm was used to treat long‐range electrostatic interactions, and the SHAKE algorithm ⁵² was applied to constrain bonds involving hydrogen atoms. Nonbonding interactions were truncated at 12.0 Å. The initial structure was first minimized and then solvated by placing a water box with dimensions of 10 Å in each direction from the atom with the largest coordinate in that direction. The solvated system was subjected to another energy minimization and heating to a temperature of 310 K, followed by 0.5 ns of equilibration. A short molecular dynamics simulation of 10.0 ns was performed to observe local side chain motions. Simulation was performed at a constant temperature of 310 K and a constant pressure of 1 atm. The simulation time step was 2 fs.

4.2. Protein expression and purification

The Agilent QuikChange Site‐Directed Mutagenesis protocol was used to introduce mutations in PAI‐1, which were confirmed by DNA sequencing. Recombinant human wtPAI‐1 and its variants were cloned in pET‐24d vector and transformed into Rosetta 2(DE3)pLysS E.coli cells by heat shock. Colonies containing the recombinant plasmid were selected on agar plates containing kanamycin and chloramphenicol. A single colony was cultured in Terrific Broth in 1 Liter at 37°C, with shaking at 250 rpm. The growth temperature was reduced to 15°C when the optical density (OD₆₀₀) reached 1.0; protein expression was induced with 1 mM Isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) and cultures were grown for an additional 20 hr. Cell cultures were harvested by centrifugation for 30 min at 10,000 x g and 4°C, and cell lysis and protein purification of PAI‐1 variants was conducted as described previously. ³³ , ³⁹ For PAI‐1 variants with replacements of histidine, buffers for cell lysis and initial ionic exchange column were adjusted to pH 5.5 due to the change in pI that results from amino acid replacements. Whereas the yield of H364A PAI‐1 was low using IMAC, as well as Hydrophobic interaction chromatography (HIC) as an alternative, the yield for all other histidine variants and E81A PAI‐1 is similar as wtPAI‐1. The final step in purification was size exclusion chromatography using Sephacryl S‐100 resin. ³³ Latent PAI‐1 was generated by diluting PAI‐1 to 5 μM in S100 buffer (50 mM sodium phosphate, 300 mM NaCl, 1 mM EDTA, pH 6.25) and incubating at 37°C for 6–7 days.

4.3. PAI‐1 activity measurements

The activity of wtPAI‐1 variants was measured by the ability of the serpin to inhibit tPA using a chromogenic assay, as previously reported. ³⁹ PAI‐1 (prepared in Buffer A: 50 mM MOPS, 100 mM [NH₄]₂SO₄, 0.1 mM EDTA, pH 7.4 at 37°C) and two‐chain tPA (Molecular Innovations, final concentration of 100 nM, prepared in Buffer B: 50 mM MOPS, 100 mM [NH₄]₂SO₄, 2 mM EDTA, 1% bovine serum albumin, pH 7.4 at 37°C) were first mixed at ratios ranging from 0.2 to 4. The mixtures were incubated at room temperature for 30 min. tPA substrate (Spectrozyme tPA, BioMedica Diagnostics, final concentration of 1 mM, prepared in Buffer B) was then added to the PAI‐1/tPA mixture to determine the residual tPA activity, based on substrate lysis and product detection at 405 nm. Purified wtPAI‐1 and the PAI‐1 variants generated in this study fully inhibited tPA function at ~1.25 equivalents.

4.4. Stability assays

The half‐life of conversion of PAI‐1 to the latency form was measured by an adaptation of previously reported stability assays. ³³ PAI‐1 (100 nM, in Buffer A) was incubated at 37°C water bath. An aliquot of PAI‐1 was removed and mixed with two‐chain tPA (Molecular Innovations, 100 nM, in Buffer B) at specific time points. Residual tPA activity in each mixture was measured by adding excess Spectrozyme tPA (BioMedica Diagnostics, 1 mM, in Buffer B). Each reaction was monitored for 5 min to measure the absorbance of p‐nitroaniline at 405 nm in a 96‐well plate using a BioTek SYNERGY 4 microplate reader. The reaction slope was plotted against PAI‐1 incubation time to calculate tPA activity. The negative slope was taken and normalized (defining 100% activity at t = 0 time point and 0% activity at the longest time point, at which PAI‐1 was completely latent) to yield normalized PAI‐1 activity. The normalized PAI‐1 activity was then plotted as a function of incubation time and fit to a single exponential decay using GraphPad Prism software to determine the half‐life of PAI‐1 latency conversion. Each measurement was conducted in triplicate, and an incubation time of over 10 half‐lives was covered. For PAI‐1 stability with Cu(II) assays, PAI‐1 was mixed and incubated with Cu(II) (various concentrations tested, in Buffer A). For assays with VN somatomedin B domain (SMB) incubation, PAI‐1 and SMB were mixed at ratio 1:2 before adding Cu(II) or incubation at 37°C.

4.5. Copper titration gel assay

A gel based metal titration assay was performed to measure Cu(II) effects on PAI‐1 stability as described previously. ³⁹ CuSO₄ (in Buffer A) at various concentrations was added to PAI‐1 (final concentration of 4 μM, in Buffer A) to create a serial dilution of Cu(II) in the mixture (ranging from 5 to 1,000 μM). The mixtures were incubated at 37°C for 30 min and then mixed with single‐chain tPA (Molecular Innovations, 4 μM, in Buffer A). Non‐reducing SDS‐PAGE gel analysis was performed to separate complexed, cleaved, or uncleaved/latent PAI‐1 upon mixing with tPA at each Cu(II) concentration. Gel pictures were recorded using a ChemiDoc XRS molecular imager (Biorad), and gel band densitometry of the PAI‐1/tPA complex bands was performed using Imagelab (Biorad). The band intensity of each sample was normalized to 100% with an untreated PAI‐1/tPA complex sample, and to 0% using the gel background. The normalized band intensities were plotted as a function of the logarithm of Cu(II) concentration. Each assay was run in duplicate to calculate averages and errors.

4.6. Isothermal titration calorimetry

Isothermal titration calorimetry was conducted to characterize the thermodynamic parameters of PAI‐1 binding with Cu(II). PAI‐1 samples were dialyzed into 100 mM MOPS, 250 mM (NH₄)₂SO₄, pH 7.4 at 10°C. A stock of 500 mM CuSO₄ in H₂O was prepared, confirming the concentration by atomic absorbance spectroscopy. Copper solutions were then prepared by diluting the 500 mM CuSO₄ stock into the dialysate buffer. Experiments were performed on a (Malvern) MicroCal VP‐ITC, which was equilibrated to 10°C. 2.2 mL of PAI‐1 (20 μM) and 4 mL of CuSO₄ solution were first degassed under vacuum for 10 min at 5°C. After degassing, the PAI‐1 solution was loaded into the ITC cell, and the copper solution was loaded into the ITC syringe. The experiment was conducted by a preliminary injection of 2 μL CuSO₄ into PAI‐1 cell, following by each additional injection of a 4 μL injection at 120 or 240 s intervals. Experiments were performed at least in duplicate. The ITC data were baseline corrected using NITPIC software ⁵³ (University of Texas, Southwestern), and fit to either a single‐site or two‐site non‐symmetric binding model in SEDPHAT ⁵⁴ (National Institutes of Health). The data and fits were represented using GUSSI software (University of Texas, Southwestern).

AUTHOR CONTRIBUTIONS

Yuzhuo Chu: Conceptualization; data curation; formal analysis; investigation; methodology; software; validation; visualization; writing‐original draft. Joel Bucci: Conceptualization; methodology; writing‐original draft. Cynthia Peterson: Conceptualization; funding acquisition; methodology; project administration; supervision; visualization; writing‐original draft; writing‐review and editing.

Chu Y, Bucci JC, Peterson CB. Dissecting molecular details and functional effects of the high‐affinity copper binding site in plasminogen activator Inhibitor‐1. Protein Science. 2021;30:597–612. 10.1002/pro.4017