Regulation of MIF Enzymatic Activity by an Allosteric Site at the Central Solvent Channel

SUMMARY

In proteins with multiple functions, such as macrophage migration inhibitory factor (MIF), the study of its intramolecular dynamic network can offer a unique opportunity to understand how a single protein is able to carry out several nonoverlapping functions. A dynamic mechanism that controls the MIF-induced activation of CD74 was recently discovered. In this study, the regulation of tautomerase activity was explored. The catalytic base Pro1 is found to form dynamic communications with the same allosteric node that regulates CD74 activation. Signal transmission between the allosteric and catalytic sites take place through intramolecular aromatic interactions and a hydrogen bond network that involves residues and water molecules of the MIF solvent channel. Once thought to be a consequence of trimerization, a regulatory function for the solvent channel is now defined. These results provide mechanistic insights into the regulation of catalytic activity and the role of solvent channel water molecules in MIF catalysis.

Graphical Abstract

In Brief

MIF is an inflammatory homotrimeric protein containing a receptor binding site, an enzymatic site, a solvent channel, and an allosteric site. Pantouris et al. unravel how these sites interact and provide insight into how small-molecule therapeutics may be used in MIF-mediated diseases.

INTRODUCTION

Migration inhibitory factor (MIF) has been identified to have multiple, unrelated functions similar to other “moonlighting proteins” (Swope and Lolis, 1999; Kapurniotu et al., 2019; Min et al., 2016). From an evolutionary point of view, these proteins served an initial function and were exapted to serve additional roles. A well-known example are the crystallin proteins in the eye that evolved from various dehydrogenases to serve as structural proteins that increase the transparency and refractive index of light (Asada et al., 2010). These gene products assumed their alternate functions based on a cell that evolved a mechanism to accommodate this purpose. In the case of MIF, various functions are related to its cell biology with respect to its presence in the cytosol, nucleus, and the extracellular space (Calandra and Roger, 2003), yet we lack a full understanding of the intramolecular and intermolecular mechanisms that regulate its multiple and nonoverlapping functions.

MIF is a well-known master regulatory cytokine originally described in the 1960s (Bloom and Bennett, 1966; David, 1966) and named after its ability to inhibit macrophage migration (Nathan et al., 1973). In the cytoplasm, it modulates the activities of JAB-1 (Kleemann et al., 2000), ERK-1/2 (Shi et al., 2006), and p53 (Hudson et al., 1999). In the nucleus, MIF is reported to associate with the apoptosis-inducing factor to function as an endonuclease for genomic DNA degradation resulting in chromatinolysis (Wang et al., 2016). Cytoplasmic MIF is exported to the extracellular space after cell stimulation, resulting in the activation of the type II receptor CD74 and the G protein-coupled receptors CXC chemokine receptor 2 (CXCR2) and CXCR4 (Leng et al., 2003; Bernhagen et al., 2007). Functional regions on MIF responsible for some of these activities were identified using modeling and experimental approaches (Lacy et al., 2018; Meza-Romero et al., 2016; Pantouris et al., 2015; Rajasekaran et al., 2016; Wang et al., 2016; Schober et al., 2008; Lu et al., 2017).

While MIF plays a pivotal role in orchestrating aspects of the innate and adaptive immune response, high levels of this cytokine contribute to a wide range of pathologies, including inflammatory, autoimmune, cardiovascular, and malignant diseases (Tilstam et al., 2017; O’Reilly et al., 2016; Greven et al., 2010). MIF levels are governed by five to eight tandem DNA repeats of CATT at position −794 of the MIF gene promoter region and a G/C SNP at position −173 (Nishihira and Sakaue, 2012; Awandare et al., 2009). Although the majority of published studies suggest that MIF upregulation has a negative impact in disease, in some cases, such as cardiac surgery (Stoppe et al., 2015), MIF plays a protective role.

The crystal structure of human MIF (Sun et al., 1996) reveals a homotrimeric assembly with a solvent channel along the 3-fold axis. The human D-dopachrome tautomerase (D-DT/MIF-2) and the bacterial enzymes 4-oxalocrotanate tautomerase and 5-carboxymethyl-2-hydroxymuconate isomerase (Subramanya et al., 1996), share a similar structural topology, and together comprise the MIF superfamily. The members of this family share a tautomerase or isomerase activity with a catalytic base, Pro1 (Lubetsky et al., 1999), that has an unusually low pK_a of 5.6 in human MIF (Stamps et al., 1998). Despite the distinct physicochemical properties of the active site for the human MIF and D-DT/MIF2, they both catalyze the tautomerization of D-dopachrome (a synthetic molecule), 4-hydoxylphenyl pyruvate (4-HPP), and phenyl pyruvate (Lubetsky et al., 1999; Pantouris et al., 2018; Rajasekaran, 2014). These substrates and catalytic activities were used to identify various inhibitors for MIF (Pantouris et al., 2015; Cho et al., 2010; Kok et al., 2018; Trivedi-Parmar et al., 2018) and D-DT (Pantouris et al., 2018; Rajasekaran, 2014). However, somewhat paradoxically, the physiological substrate of the MIF tautomerase site is unknown, despite universal conservation of the Pro1 active site in MIF species (Figure S1).

Molecular dynamics (MD) simulations were recently utilized to obtain insights into the MIF-induced activation of CD74 (Pantouris et al., 2018). An allosteric site located at one end of the solvent channel was identified to regulate CD74 activation. Here, we use a functional study of MIF mutants, including MD simulations, water molecule analysis, crystallography, and nuclear magnetic resonance (NMR), to understand how MIF catalytic activity is regulated. We identify an intramolecular dynamic mechanism formed among MIF residues and water molecules in the solvent channel that regulates the catalytic activity of MIF.

RESULTS

Mutation of the Allosteric Node and Adjacent Residues Alters MIF Catalytic Activity

MD simulations of wild-type MIF identified that the allosteric site composed of Tyr99 is strongly correlated with the catalytic residue Pro1 (Pantouris et al., 2018). The two sites communicate via the intermediate residues His62 and Met2, which are located on adjacent β strands within the MIF monomer. This dynamic pathway that controls the MIF catalytic activity differs from the corresponding pathway that regulates activation of CD74, both in terms of direction of the signal and the residues involved (Pantouris et al., 2018). To experimentally probe the potential dynamic correlation between Tyr99 and the catalytic activity of MIF, we mutated the key nodes of the dynamic pathway, Tyr99 (Y99F, Y99A, and Y99G) and His62 (H62Y, H62F, H62A, and H62G), and examined the impact of these mutations on the HPP keto/enol tautomerase activity of the protein (Figure 1 and Table 1).

Figure 1. Impact of Tyr99 and His62 Mutations on the Catalytic Activity of MIF.

Kinetic analyses of wild-type MIF (black bar) and MIF mutants (Y99F, Y99A, Y99G, dark gray bars; H62Y, H62F, H62A, H62G, light gray bars) demonstrate the connection between the allosteric (Tyr99) and catalytic (Pro1) sites. The standard deviation is shown. Related to Table 1.

Table 1.

Kinetic Parameters for wild-type MIF and MIF Mutants Determined from Tautomerase Assays of 4-HPP

	Vmax (μM s−1)	kcat (s−1)	Km (mM)	kcat/Km (mMs−1ss−1)	kcat/Km Changes versus WT MIF (%)
WT MIF	0.32 ± 0.01	6.47 ±0.12	1.06 ± 0.05	6.1 ± 0.36	100 ±6
Y99F	0.60 ± 0.03	11.9 ± 0.64	1.38 ± 0.13	8.7 ± 0.82	143 ± 13
Y99A	0.23 ± 0.02	4.67 ± 0.42	0.99 ± 0.10	4.7 ± 0.21	77 ± 3
Y99G	0.15 ± 0.01	0.59 ± 0.05	1.02 ± 0.20	0.6 ± 0.06	10 ± 1
H62Y	0.43 ± 0.02	8.67 ± 0.46	0.92 ± 0.02	9.4 ± 0.68	154 ± 11
H62F	0.49 ± 0.01	9.86 ±0.12	0.64 ± 0.06	15.5 ± 1.35	254 ± 22
H62A	0.37 ±0.10	7.40 ±1.91	1.64 ± 0.42	4.5 ± 0.53	74 ±9
0H62G	0.52 ± 0.02	2.09 ± 0.08	0.88 ± 0.46	2.4 ± 0.12	39 ± 2

See also Figure 1. These experiments were carried out in triplicate (n = 3) and the data were analyzed using ORIGIN 9.0. The error values are shown as standard deviations.

Based on the catalytic efficiencies (k_cat/K_m) of the MIF variants, we noticed three distinct outcomes. In the first group, we observed increased catalytic activities due to substitution of Tyr99 or His62 by another aromatic amino acid (Tyr or Phe). Compared with wild-type MIF, the catalytic efficiencies of Y99F, H62F, and H62Y were found to be increased to 143%, 254%, and 154%, respectively (Figure 1). The kinetic parameters showed that the enhanced catalytic rates of Y99F and H62Y are related to k_cat values, while for H62F both k_cat and K_m are improved. In the second group, Tyr99 and His62 substitutions to alanine (Y99A and H62A) reduced, to some extent, the catalytic efficiency of MIF to 77% (Y99A) and 74% (H62A) compared to the corresponding efficiency of wild-type MIF (Figure 1). Despite the similar activities, Y99A and H62A revealed distinguishable patterns in relation to their kinetic parameters. For Y99A, the K_m value remained similar to that of wild-type MIF and only k_cat is reduced, while for H62A the opposite effect is observed. The Gly mutations, which is the third distinct group, demonstrated decreased catalytic efficiencies resulting in 10% (Y99G) and 39% (H62G) of the corresponding efficiency of wild-type MIF (Figure 1). The K_m values of Y99G remain unaffected and only k_cat is significantly reduced, while the H62G mutation alters both K_m and k_cat values. Temperature-dependent circular dichroism (CD) spectra indicate that all classes of mutants are folded (Figure S2), with small per-monomer energetic changes that likely only partially account for these functional observations.

Crystal Structures of MIF Mutants Highlight Subtle Differences in Molecular Interactions within the Allosteric Pathway

Crystallographic analysis of the MIF mutants (Y99F, Y99A, Y99G, H62Y, H62F, H62A, and H62G) revealed excellent superposition to wild-type MIF with root-mean-square deviation (RMSD) values that varied between 0.08 and 0.16 Å (Table S1). Overall, we did not observe any major conformational changes that could potentially explain the three distinct catalytic behaviors of the MIF mutants. Y99G, which is the mutant with the lowest catalytic activity (Figure 1), has a superposition agreement to wild-type MIF with an RMSD of 0.10 Å and does not show any significant conformational changes along the dynamic pathway under investigation (Tyr99-His62-Met2-Pro1). We therefore probed whether the solvent channel and residues located at the opening of the channel (Tyr99) or in close proximity to it (His62) were affected. We observed that the size of the solvent channel opening correlates with the catalytic efficiency of the MIF variants. As the side chain volume decreases from Y99F to Y99G (Y99F-Y99A-Y99G) and from H62F/H62Y to H62G (H62F/H62Y-H62A-H62G), the opening of the solvent channel increases and the catalytic activity of the MIF mutants decreases (Figure 2). Interestingly, the two mutants H62Y and H62F, which demonstrate higher catalytic efficiency than wild-type MIF (254% and 154%, respectively), have similar solvent channel openings but their catalytic activities are different (Figure 1). These observations suggest that the channel opening plays a major role in regulating the catalytic activity from the allosteric node. We used crystal structures of these two variants to examine how the interatomic interactions promote these catalytic enhancements between the allosteric and catalytic sites. We observed movements of the Tyr99 side chain toward the Tyr62 (H62Y) and Phe62 (H62F) side chains by 0.9 and 1.2 Å, respectively (Figure 3A). The movement brings the center of Tyr99 and the two aromatic rings for H62F and H62Y to a distance of 4.1 and 4.3 Å, respectively, creating π-π stacking interactions more optimal than those of wild-type (Figure 3B) as confirmed by Arpeggio, a web server that calculates interatomic interactions in protein structures (Jubb et al., 2017). Using the crystal structure of wild-type MIF (PDB: 3DJH), we examined how the intra-subunit π-π stacking interactions can promote the transmission of a dynamic signal between the allosteric and catalytic sites. We found that His62 forms two backbone hydrogen bond interactions with Met2, typical of two adjacent β strands in a β sheet (Figure 3C). Through this network of interactions, a dynamic signal from Tyr99 is able to reach the catalytic site via the β strand system of MIF. Of note, Tyr99 and the catalytic residue Pro1 are located 7.8 Å apart (PDB: 3DJH) (Figure 3D). Therefore, the only way that the catalytic activity of MIF can be influenced by Tyr99 is through intramolecular dynamic signals that involve the β strand system of the protein, something that has been described previously for other proteins (Fenwick et al., 2011, 2014).

Figure 2. Effect of Tyr99 and His62 Mutations at the Opening of MIF Solvent Channel.

Crystallographic analysis of the MIF variants showed that the opening of the solvent channel is clearly affected by the mutation of Tyr99 and His62. The increased opening of the solvent channel is inversely associated with the catalytic activity of MIF. Related to Table S1.

Figure 3. Crystallographic Analysis of MIF Variants.

(A) In the H62F and H62Y crystal structures, the larger aromatic residues at position 62 induce a distinct conformational change toward Tyr99, increasing the coupling between the two residues.

(B) Aromatic interactions between Phe62-Tyr99 and Tyr62-Tyr99 as observed from the crystal structures.

(C) His62 forms two backbone hydrogen bond interactions with Met2 that promote communication along the dynamic pathway.

(D) The nearest distance between Tyr99 and Pro1 is 7.8 Å. Tyr99 and Pro1 are shown as sticks. Related to Table S1.

Solvent Channel Water Molecules Mediate MIF Catalysis

The increased catalytic efficiency of the H62Y, H62F, and Y99F mutants is unlikely due to the weak π-π interactions between residues 62 and 99 alone. To determine whether the solvent contributes to the observed differences in catalytic activity across MIF variants, we performed a contact network analysis involving the solvent channel water molecules. We computed the hydrogen bond network of water molecules near the allosteric site for the crystal structures of wild-type MIF and variants (Y99F, Y99A, Y99G, H62Y, H62F, H62A, and H62G). For the MIF mutants with enhanced catalytic activities (Y99F, H62Y, and H62F), we observed an increased hydrogen bond network proximal to residue 99 relative to wild-type MIF (Figure 4A). In contrast, the mutants with reduced catalytic efficiency (Y99A, Y99G, H62A, and H62G) exhibited few-to-no water-mediated hydrogen bonds in the network proximal to the allosteric site. In wild-type MIF, water molecules are positioned to form inter-subunit hydrogen bond interactions with both Tyr99 and His62 and facilitate a bridge between the two residues (Figure 4B). The Y99A, Y99G, H62A, and H62G mutations abolished any ability for bridging residues 99 and 62 by water molecules, thus decoupling the two residues and leading to decreased catalytic activities.

Figure 4. The Role of Water-Mediated Networks in MIF Catalysis.

(A) The degree of water network connectivity between residues 99 and 62 is strongly correlated with the MIF catalytic activity. Water molecules are shown as red spheres and the hydrogen bond as black dotted lines. The central water molecule found in the crystal structures of Y99F, H62F, and H62Y is shown in brown.

(B) In the crystal structure of wild-type MIF, water molecules serve as a bridge to facilitate communication between residues 99 and 62 via hydrogen bond interactions.

(C) A water molecule with very high degree centrality (shown with a red asterisk) serves as a stable anchor that bridges hydrogen bond interactions between residues 99 and 62 in the crystal structures of Y99F, H62F, and H62Y, and increasing their coupling. Related to Figure S3.

To examine the influence of individual water molecules on MIF catalytic activity, we performed a protein-water network centrality analysis computed for wild-type MIF and all variants. From this analysis, we found that the MIF mutants Y99F, H62Y, and H62F have one common water molecule (shown in brown) that is positioned to facilitate both intra- and inter-subunit water bridges between residues 99 and 62 (Figure 4A), in comparison with the inter-subunit water bridges observed in wild-type MIF. For example, in the H62Y mutant, the central water molecule is stably positioned to form water bridges among any of the six neighboring tyrosine residues. The additional water molecules found around the central water molecule help construct a network that connects residues 99 and 62. The denoted central water molecule found in the crystal structures of Y99F, H62Y, and H62F had the second highest degree centrality of any water molecule in the entire protein, thus confirming both its high stability and significance in transmitting dynamic signals between positions 99 and 62 (Figure 4C). With this analysis, we demonstrate that water network arrangements in the solvent channel can play a significant role in MIF catalysis.

Hydrogen bond analysis from 200-ns MD simulations revealed that mutations at His62 or Tyr99 reduced the hydrogen bond interactions connecting the active and Tyr99 sites. Of note, Tyr99 is connected to the active site through hydrogen bonds with Leu61 and Ser63 during the trajectories, which flank His62, and is in turn hydrogen bonded to Met2 of the neighboring antiparallel β strand. In the Y99A mutant, the hydrogen bonding between the backbones of Ala99 and Ser63 was reduced by 16% based on hydrogen bond analysis of the latter 100 ns of the trajectory. In addition, the interaction between Ala99 and Leu61 observed during wild-type trajectories was reduced by 8%. Predictably, mutating Tyr99 to glycine exacerbated this reduction. For Y99G, the hydrogen bonding between Gly99 and Ser63 was reduced by 25% and that between Gly99 and Leu61 by 6%. The H62G mutant was the most perturbed, as residue 62 sits at the center of the interaction path between Tyr99 and the active site. The hydrogen bonding between Tyr99 and Ser63 was reduced by 46%, between Gly62 and Met2 by 18%, and between Leu61 and Tyr99 by 9%, consistent with its drastic reduction in catalytic activity.

Examining the number of water molecules within the channel during MD simulations also shows the clear effect of Tyr99 and His62 mutants (Figure S3; Table S3). The wild-type model had, on average, 41.4 ± 3.5 water molecules in the central channel. The Y99A and Y99G mutants had 46.6 ± 3.8 (+13%) and 52.0 ± 3.7 (+26%), respectively. The H62G mutant showed no significant change in the amount of water molecules in the channel (41.5 ± 3.8), likely due to its placement away from the channel’s opening. However, the H62G mutant increases the root-mean-square fluctuation (RMSF) of Tyr99 from 0.6 Å in wild-type to 1.6 Å, likely contributing the mutant’s loss of hydrogen bonding noted above. There is no similar increase in the RMSF of His62 with the Y99A/G mutants and all simulations had an overall C_α RMSF of 0.6–0.7 Å. Disruptions to intra-subunit correlations are apparent in all MIF mutants studied, and residue networks are altered from those of wild-type (Figure S4). However, these changes are subtle, likely because the allosteric mutations are relatively close to the active site. The C_α of Pro1 and Tyr99 are ~10 Å apart, and are connected by strong hydrogen bonds to the mutually neighboring b4 strand. Thus, the mutational effect does not need to propagate across a large portion of the protein to reach the active site, precluding large rearrangements of dynamic communities. Based on these analyses, the reduced activity in the studied mutants can be attributed to either a disruption of the hydrogen bonds between the residues themselves (as in H62G) or positional alterations within the central water channel (as in Y99A and Y99G).

Solution NMR Highlights the Dynamic Character of the Allosteric Pathway

The ¹H¹⁵N TROSY-HSQC NMR spectra of MIF mutants revealed an increase in the flexibility of the entire protein along the series H62Y-H62F-H62A-H62G and Y99F-Y99A-Y99G (Figures 5A, 5B, S5, and S6). The residue-residue and residue-water interactions we observed for Y99F, H62Y, and H62F are absent from the remaining mutants and consistent with the NMR findings. While the chemical shift perturbation profiles of MIF variants are rather modest, a high degree of line broadening due to motion of the protein on the millisecond timescale is observed in variants with low catalytic activity. We monitored dynamic signatures of critical relay points along the Tyr99-His62-Met2-Pro1 pathway and found that mutation of Tyr99 or His62 to Gly broadens the correlated node residues, a classic signature of heightened dynamics at those sites (Lisi and Loria, 2016), suggesting coupling of the N terminus of MIF toTyr99 in a manner consistent with MD simulations (Pantouris et al., 2018). A reciprocal effect is observed at Tyr99 in the presence of Pro1/Met2 mutations (Figure S7). Interestingly, Tyr99 correlations from MD simulations, as well as mutations at Tyr99, propagate to the CD74 and tautomerase sites; however, mutations at Pro1 affect Tyr99 but not the CD74 site in NMR experiments (Figure S7). We also observe a significant reordering of the solvent channel network along a continuum of mutations that may facilitate the enhanced (diminished) tautomerase function. In cases where MIF catalytic efficiency is only marginally attenuated (Y99A, H62A), fewer flexible residues are observed, occurring only at sites outside of the allosteric relay. Mutations that enhance MIF activity show very small, if any, changes to the structure or dynamics of the entire protein. The NMR results are also consistent with His62 playing a role in propagating a dynamic signal between the active and regulatory sites, as shown by substantial exchange broadening in resonances corresponding to His62 (Figure 5A). Additional signatures of enhanced flexibility in Y99G and H62G MIF are localized at the monomer-monomer interfaces near the catalytic N terminus, indicating that changes at the active site are largely dynamic, rather than structural, in nature. It should be noted that, despite relatively small changes in raw CD spectra indicating folded structures of MIF mutants in solution, Y99G is ~1.3 kcal/mol, on a per monomer basis, less thermostable than wild-type MIF while Y99F is ~0.32 kcal/mol more thermostable (Figure S2). These energetic differences track with the dynamic character of the MIF mutants, particularly the NMR exchange broadening that is typically ascribed to heightened intra-residue dynamic signaling in allosteric proteins. Furthermore, the wild-type MIF structure itself displays significant fast timescale flexibility by NMR (Figure S8), suggesting that, even though X-ray crystallography, CD, and NMR indicate stable folded structures, MIF is intrinsically flexible. Based on these data, it is clear that mutations introducing aromatic interactions to these sites strengthen hydrogen bonding networks essential for catalysis, while those that promote structural flexibility disrupt the residue-residue and residue-solvent interactions.

Figure 5. NMR Probe of Perturbations to the MIF Allosteric Pathway.

(A) Combined ¹H¹⁵N chemical shift perturbations (black lines) caused by mutations at His62 (top row) and Tyr99 (bottom row). Blue dashed lines represent 1.5σ above the 10% trimmed mean of all shifts and pink bars denote sites of exchange broadening in each NMR spectrum.

(B) Close-up views of ¹H¹⁵N HSQC spectral overlays of wild-type MIF (red) and MIF mutants (blue). Resonances corresponding to allosteric nodes Tyr99 and Met2 are monitored for His62 variants, while resonances corresponding to His62 and Met2 are monitored for Tyr99 variants. Related to Figures S2‖S6.

Ligand-Induced Dynamic Signals Are Propagated from the Catalytic Site to Tyr99

Since our data show a dynamic association between the active site Pro1 and the allosteric Tyr99 residue, we explored whether substrates or inhibitors at the active site have any effect on Tyr99. Nineteen high-resolution human MIF structures from the Protein DataBank (PDB) with all three active sites of the homotrimer occupied by a ligand were examined (Table S2). The ligand library consisted of a substrate (hydroxyphenylpyruvate solved at 2.5 Å), four competitive inhibitors (resolutions between 1.8 and 2.0 Å), and 14 covalent inhibitors bound to Pro1 (resolutions between 1.53 and 2.07 Å). The PDB entries, RMSD values to wild-type MIF, and resolutions of the 19 structures used in this study are provided (Table S2). Depending on the ligand, the side-chain hydroxyl group of Tyr99 can vary relative to other ligand-dependent positions of Tyr99 by up to 1.4 Å (Figure 6A). Larger fluctuations are limited due to the presence of the side chain of His62. To understand the correlation between Tyr99 fluctuation and ligand binding, we explored the amino acid interaction network around this area. In addition to the two backbone hydrogen bond interactions between His62 and Met2 and the p-p interactions between Tyr99 and His62, we found an inter-subunit hydrogen bond interaction formed between the side chains of His62 and Tyr99 (Figure 6B). Thus, binding of a ligand to the catalytic site causes movements of Pro1, which in turn affects the Met2 backbone atoms. These movements are transmitted to His62, and from there to Tyr99 (Figure 6C). The ligand-dependent conformational changes of Tyr99 are also consistent with the dynamic connection between the two sites (Figure 5).

Figure 6. Active Site Ligands and the Tyr99 Allosteric Site.

(A) Ligands that bind to the active site pocket of MIF induce dynamic motions of Tyr99 side chain. We used 19 high-resolution crystal structures in which the ligands occupy all three active sites. Pro1 is shown as spheres and Tyr99 as sticks for each monomer.

(B) His62 forms two backbone hydrogen bond interactions with Met2 (β1 strand) and a side-chain hydrogen bond interaction with Tyr99 (β5 strand) from the adjacent monomer. The hydrogen bond interactions are shown as black dotted lines. His62 also forms aromatic interactions (shown in red dotted lines) with Tyr99 from the same monomer.

(C) The ligand-induced dynamic effect on Pro1, Met2, His62, and Tyr99 is shown by superposition the 19 crystal structures. Related to Table S2.

DISCUSSION

We previously performed MD simulations to provide insight into the MIF-induced activation of CD74 (Pantouris et al., 2018). This method identified an allosteric site, which we experimentally tested, verified, and determined its intramolecular pathway to regulate CD74 activity. In this new study, we experimentally demonstrate that the MIF allosteric site, which is located at one opening of the solvent channel, also regulates the protein’s catalytic activity. Crystallographic analysis show that Tyr99 and Pro1 are dynamically connected via aromatic interactions and a hydrogen bond network that involves His62 and Met2. The residues Tyr99, His62, and Met2, located on β5, β4, and β1 strands, respectively, implicate the β strand system of MIF in the regulatory mechanism. The dynamical connection between the allosteric and catalytic sites is supported by mutagenesis, protein-water network analyses, and crystallographic and NMR studies. Nineteen MIF-ligand crystal structures obtained from PDB independently support this conclusion. Interestingly, the established MIF homolog D-DT (MIF-2) shares Pro1, but with a Phe2-Ser62-Phe99 sequence. D-DT activates CD74, but with no sequence homology to MIF at its binding site (Figure S1). Thus, it is unclear if the MIF homolog supports an allosteric pathway encompassing these functionalities. This possibility is currently being investigated in a related study.

High-resolution crystallography and MD simulations have revealed the importance of water molecules in mediating protein folding, function, and signaling (Levy and Onuchic, 2004; Venkatakrishnan et al., 2019). Crystallographic and kinetic analyses of the MIF mutants led us to identify afunctional role for the solvent channel. In the crystal structures of Y99F, H62Y, and H62F, the dense hydrogen bonding network created proximal to the allosteric site and around a stable central water molecule greatly strengthens the allosteric signal from Tyr99 to Pro1 and increases the catalytic efficiency of these MIF mutants. These mutants display enhanced aromatic interactions between positions 99 and 62, and well-ordered water molecules in the solvent channel compared with the wild-type protein. In contrast, the absence of the aromatic interactions in the Y99G and H62G mutants result is a poorly connected hydrogen bonding network in the solvent channel and a dramatic increase in the dynamic properties of Y99G and H62G.

Collectively, our findings demonstrate that at least two of the MIF activities, CD74 activation and enzymatic activity, are regulated by the same allosteric node via separate intramolecular dynamic signals that are transmittable to distal functional regions (Figure 7). For the MIF enzymatic activity, we found a Tyr99-His62-Met2-Pro1 pathway connecting the allosteric and catalytic sites. Experimental evidence from NMR suggests that the relationship between these two sites/pathways is quite complex. The tautomerase site is sensitive to Tyr99 mutations, and reciprocity of the allosteric signal (at Tyr99) is observed in tautomerase mutants (Pro1, Met2) in the form of chemical shifts and line broadening at His62 and Tyr99 (Figures S6 and S7). MD simulations show connectivity from Tyr99 to both the CD74 and tautomerase sites (Pantouriset al., 2018), and mutations atTyr99 (this work) also perturb chemical shifts at both functional sites. However, mutations at Pro1 only affect the Tyr99 site, as no chemical shift perturbations are observed along the CD74 pathway (Figure S7). One explanation for this experimental result could be that the MIF C-terminal binding site is not directly coupled to Pro1, and that, while they share the Tyr99 node, MIF utilizes two distinct pathways to signal these regions (Figure 7). In addition, many other distal residues in MIF alter CD74 activation, suggesting that this activity may encompass a broader network than was previously known. Although it is clear that Tyr99 is a critical node in both processes, the fact that neutrophil recruitment (a reporter of CD74 activation) is completely abolished in Y99A MIF (Pantouris et al., 2018) but tautomerase is attenuated only 33%, suggests that the influence of Tyr99 differs in these processes.

Figure 7. Allosteric Pathway(s) Influencing MIF Activity.

(A) Communication from the solvent channel (red arrow) to the N-terminal tautomerase site (orange sticks) was established in this work. Reciprocity of this effect is shown by NMR studies of Pro1 mutants (green arrow). Tyr99 (green sticks) is also a central node in the CD74 activation network (blue arrow, sticks) shown by MD simulations in Pantouris et al. (2018).

(B) Allosteric pathways mapped onto a MIF monomer highlight the majority of the CD74 network residing on the C terminus of an adjacent subunit (pink cartoon). Related to Figures S7 and S8.

We also found mutants with increased catalytic activity to have decreased dynamics as observed by NMR, slightly more favorable energetics, and a new solvent network around the allosteric site. Other mutants of Tyr99 and His62 with decreased enzymatic activity have substantially increased dynamics, poorer structural energetics, and lacked the solvent network. The intrinsic flexibility of the MIF structure is modulated along a continuum with mutations at or near the allosteric node. Quantitating rates of conformational exchange and dynamic equilibrium populations of MIF residues in future studies may enable us to definitively identify sites that share similar motional parameters indicative of a higher degree of coupling in allosteric pathways (Lisi et al., 2016). This would, potentially, allow us to address the question of whether or not Tyr99 is a node in two distinct or one contiguous pathway(s). Our study demonstrates that the MIF solvent channel is not merely a result of protein folding but contains a critical allosteric node that also serves an important role for at least the MIF enzymatic activity.

STAR★METHODS

RESOURCES AVAILABILITY

Lead Contact

Further information and requests for reagents should be directed to and will be fulfilled by the corresponding authors.

Materials Availability

Plasmids generated in this study are availabe from Elias Lolis, George Lisi, or Georgios Pantouris and have been deposited to Addgene.

Data and Code Availability

Coordinates and structure factors have been deposited and are available at the ODB website (https://www.rcsb.org/) under the accession PDB codes 6OYE (MIF Y99F), 6OY8 (MIF Y99G), 5UMK (MIF H62Y), 6OYG (MIF H62F), 5UMJ (MIF H62A), 6OYB (MIF H62G). Other PDB coordinates used for Y99 analysis are 3DJH (Crichlow et al., 2009), 1CA7 (Lubetsky et al., 1999), 1GCZ(Orita et al., 2001), 1LJT (Lubetsky et al., 2002), 1MFI (Taylor et al., 1999), 3L5U (McLean et al., 2010), 4F2K (Tyndall et al., 2012), 3B9S (Winner et al., 2008), 3WNT (Spencer et al., 2015), 3CE4 (Crichlow et al., 2009), 3SMB (Crichlow et al., 2012), 3SMC (Crichlow et al., 2012), 4OYQ (Spencer et al., 2015), 3JSF (McLean et al., 2009), 3JSG (McLean et al., 2009), 3JTU (McLean et al., 2009), 4P01 (Pantouris et al., 2015), 4TRF (Pantouris et al., 2015), 4POH (Pantouris et al., 2015) and 4PLU (Pantouris et al., 2015).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

BL21-Gold(DE3) competent cells were cultured at 37°C to OD₆₀₀ ~0.6–0.8, followed by an additional 4–6 hours at 37°C post-induction.

METHOD DETAILS

Mutagenesis and Expression

All the MIF mutations, encoded in the pET-11b, were produced by the QuikChange II site-directed mutagenesis kit (Agilent Technologies) using the cDNAof wild type (wt) MIF as a template (Bernhagen et al., 1994; Lacy et al., 2018). Mutagenesis of Y99Aalso used the QuikChange II kit with mutagenic Y99A primers and PfuTurbo DNA Polymerase in a PCR to extend DNA synthesis from the primers all around pET11b-wtMIF plasmid. DpnI treatment digested the parental plasmid with the wt MIF cDNA that contained methylated and hemimethylated DNA, prior to transformation of Escherichia coli BL21-Gold (DE3) competent cells (Pantouris et al., 2018). The same procedure was used mutants in this study. Briefly, 100 μl of Escherichia coli BL21-Gold (DE3) competent cells were mixed with DpnI-treated pET-11b/mutant DNA (25-50 ng) and incubated on ice for 30 minutes, transferred into a 42°C water bath for 45 seconds, placed on ice for two minutes, supplemented with SOC medium to 1 mL, and transferred to a 37°C shaker incubator for growth with vigorous shaking for 1 hour. The cells were spun using a microcentrifuge, the supernatant was removed and the cells were resuspend in LB, and plated on LB agar with 100 mg/ml ampicillin (Amp). Five clones from each mutation were used to purify plasmid to confirm by PCR that MIF was ligated into the plasmid. DNA sequencing was used to verify each plasmid contained the expected mutant MIF. To test for expression of MIF mutations, each E. coli BL21-Gold (DE3) cells verified to have a MIF mutant in pET-11b was cultured in 5 ml LB with 100 mg/ml Amp (LB/Amp) and induced by 1 mM of isopropyl b-D-1-thiogalactopyranoside (IPTG) at a cell density of O.D.₆₀₀ of 0.6. Levels of expression were assessed using a NuPAGE 4-12 % Bis-Tris gels after 4 hours after addition of IPTG. Althoug theTyr-99 and His-62 mutations were performed at different times during this project, the expression of the MIF mutations appeared similar. One clone per mutant was chosen for larger scale purification for each MIF mutations with 1 L LB/Amp medium using a shaking incubator at 37°C. The large scale growth procedure was similarto the small scale growth protocol with the exception that after a 4-hour IPTG induction, the cells were centrifuged and stored in −80°C until further use.

Protein Purification

Frozen cells (for WT MIF and the seven mutants) were thawed and resuspended in 20 mM Tris, 20 mM NaCl, pH 7.4 containing a mini EDTA-free protease inhibitor cocktail tablet (Sigma-Aldrich) and lysed by sonication. The cell debris was removed by centrifugation at 4°C and the supernatant was filtered using 0.22 mm filter units (Millipore). The filtrate was loaded onto Q-Sepharose and SP ion exchange chromatography columns connected in series. Wild-type MIF and MIF mutants were eluted in the flow-through with ~95 % purity. After concentrating the samples using 10 kDa molecular weight cut-off centrifugal filters (Millipore), the proteins were loaded onto size-exclusion column (16/60 Superdex 75) to remove the remaining ~5% of contaminants. Purification was carried out using 20 mM Tris, pH 7.4 with 20 mM NaCl as the running buffer. Protein concentration was determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific).

Enzyme Kinetics

A 100 mM stock solution of 4-hydroxyphenylpyruvate (4-HPP) in 0.5 M ammonium acetate, pH 6.2 was freshly prepared and equilibrated for 24 hours at room temperature to produce the predominating keto-form under this condition. Kinetic assays were carried out using keto form of 4-HPP at 0 - 2 mM. For each concentration, 10 μL of 4-HPP was mixed with 130 μL borate at a final concentration of 0.420 M in a UV transparent flat bottom 96-well plate. The reaction was initiated by addition of 10 μL of protein at a final concentration of 50 nM (for wild-type MIF and Y99F, Y99A, H62Y, H62F, H62A) or 250 nM (for H62G and Y99G). Absorbance measurements for the conversion to the enol form of HPP (the enol-borate complex has an extinction coefficient of ε₃₀₆ = 11,400 M⁻¹ cmr⁻¹) were recorded at 10 second intervals for a total of 360 seconds using a Tecan Infinite M200. All experiments were performed at 25°C. Initial velocities were plotted as a function of 4-HPP concentration and the Michaelis-Menten parameters (V_max, K_m, k_cat) were determined by nonlinear regression using ORIGIN 9.0. Each assay was carried out in triplicate in three separate experiments.

Crystallization and Structure Determination

Crystallization of Y99A was described elsewhere (Pantouris et al., 2018). The MIF mutant proteins (Y99F, Y99G, H62Y, H62F, H62A and H62G) were crystallized in 24-well hanging drop trays. Crystallization was accomplished by mixing the stock protein solution concentrated to ~18 mg/ml with 20 mM Tris, pH 7.5,2 M ammonium sulfate and 3% 2-propanol at various volume ratios. In all cases the drop size was 4ml. Trays were stored at 20°C and crystals were formed overnight. Individual crystals were transferred to a new drop containing the mother liquor enriched with 28% glycerol for cryoprotection, mounted on a loop, and flash frozen in liquid nitrogen. Crystal screening and complete data set collection occurred at the Macromolecular X-ray Facility at the Yale School of Medicine using a Rigaku Pilatus 200K Detector with a Rigaku 007 rotating copper anode X-ray generator (wavelength=1.5418 Å). All the data sets were collected at a temperature of 100K. The HKL2000 program suit (Otwinowski and Minor, 1997) was used for integration and scaling. The CCP4 supported program PHASER (McCoy et al., 2007) was used to obtain solutions for the MIF mutants. For molecular replacement, the crystal structure of wild-type MIF (PDB entry: 3DJH) was used as the starting model. Refinement of the structures were carried out using Refmac (Winn et al., 2003) and COOT (Emsley et al., 2010). Ramachandran analyses of the MIF mutants showed 0% outliers and 98.8%, 99.1%, 99.1%, 98.5%, 98.5%, or 98.8% residues in the preferred regions, for Y99F, Y99G, H62F, H62Y, H62A and H62G, respectively. The crystal structures of wild-type MIF, the MIF mutants, and the nineteen MIF-ligand complexes obtained from PDB were superimposed using the CCP4-supported program SUPERPOSE (Winn et al., 2011), and visualized and analyzed in PyMOL (DeLano, 2002). The interatomic interactions around the allosteric site were examined using Arpeggio (Jubb et al., 2017). The detailed statistics of the Tyr99 and His62 mutants are presented in Tables S1 and S2, respectively. The rmsd values obtained for the MIF mutants and nineteen MIF-ligand complexes vary between 0.08-0.16 Å and 0.20-0.36 Å, correspondingly.

NMR Spectroscopy

Samples for NMR spectroscopy were expressed under identical conditions to those utilized for X-ray or kinetic studies. Isotopically enriched ¹⁵N-MIF was grown in deuterated M9 minimal medium containing CaCl₂, MgSO₄, and MEM vitamins and utilizing glucose (¹²C₆H₁₂O₆, natural abundance) and ammonium chloride (¹⁵NH₄Cl, Cambridge Isotope Labs) as the sole carbon and nitrogen sources, respectively. Cells were adapted to be grown in D₂O as follows. A small culture of MIF was grown for 8 −10 hours in LB medium. The resulting cloudy suspensions were then used to inoculate 25 mL of LB containing 50% D₂O, which was grown for 8 – 10 hours and then used to inoculate a final LB culture (~50 mL) made with 95% D₂O. After 7 hours of incubation at 37°C, these cells were collected by centrifugation and resuspended in 1 L of M9 minimal medium. MIF was grown at 37°C to an OD₆₀₀ ~ 0.8 – 1.0 and induced with 1 mM IPTG. After an additional 4 – 6 hours of incubation, cells were harvested by centrifugation and stored at −80°C. MIF was purified identically to the manner described earlier. ¹H¹⁵N-TROSY HSQC NMR spectra were collected at 30°C on a Bruker Avance NEO 600 MHz spectrometer. The ¹H and ¹⁵N carriers were set to the water resonances and 120 ppm, respectively. All NMR spectra were processed with NMRPipe (Delaglio et al., 1995) and analyzed in Sparky (Lee et al., 2015).

Circular Dichroism (CD) Spectroscopy

CD spectra and thermal unfolding experiments were collected on a JASCO J-815 spectropolarimeter equipped with a variable temperature Peltier device. Denaturation curves of 15 – 20 mM MIF were recorded at 218 nm in a 1 mm quartz cuvette over a temperature range of 20 – 90°C (293 K-363K). Thermodynamic parameters were extracted via nonlinear curve fitting in GraphPad Prism, and free energy analysis was performed with calculated values of the unfolding heat capacity, DCp (Figure S2).

Computing Water-Mediated Interactions

Hydrogen bonds were computed in crystal structures with GetContacts (Fonseca, 2017). We applied a simple distance cutoff of < 3.5 Å between donor and acceptor atoms. No angle criteria were imposed since crystal structures generally lack hydrogen atoms. A water-mediated interaction is defined to occur between a pair of residues if they form hydrogen bonds to a common water molecule or a pair of water molecules held together by a hydrogen bond. All water-mediated interactions were computed throughout the wild-type MIF, H62A, and H62Y crystal structures to compare protein-water interaction networks.

Protein-Water Graph Network Analysis

Previous studies have shown that residues critical for protein function tend to have high centrality measures within the non-covalent residue interaction network (del Sol et al., 2006; Thibert et al., 2005; Vishveshwara et al., 2009). Here, we perform a similar network centrality analysis that involves both the protein and solvent to assess the significance of individual water molecules in assisting protein function. The protein-water interaction graph network is represented by G(V, E) where V is the set of nodes including all residues and water molecules in the system, and E is the set of hydrogen bonds forming edges between nodes. Degree centrality was computed using the NetworkX Python library (Hagberg et al., 2008). The degree centrality of each node v in G is defined as the fraction of nodes in G that shares an edge with v. We rank each node by degree centrality to identify the water molecules with greatest functional significance.

Molecular Dynamics Simulations

Four models of MIF (wild-type, Y99A, Y99G, and H62G) were made using the PDB structure 3DJH (Crichlow et al., 2009). In the case of Y99A, Y99G, and H62G, the mutations were made uniformly on all three MIF monomers. The models were each placed in a periodic water box and neutralized with NaCl. Simulations were run for 200 ns at 300 K using NAMD (Phillips et al., 2005) and the CHARMM36m force field (Huang et al., 2017), and the last 100 ns were analyzed. Hydrogen bond and rmsf analyses were conducted using the respective VMD plugins (Humphrey et al., 1996). Hydrogen bonds were defined as an acceptor-donor distance of no more than 3.0 A and an acceptor-donor-hydrogen angle no more than 30°. The average number of hydrogen bonds in the solvent channel per MD frame was averaged across the three MIF monomers, and the percentage change in hydrogen bonding was determined as (Mutant-WT)/WT. Overall rmsf values were calculated from the alpha carbons of aligned trajectories and the rmsf of His62 and Tyr99 were calculated using the entire residue of interest, excluding hydrogens. In order to calculate the number of waters in the central channel of each mutant, the mouth at either end of the channel was defined using the alpha carbons of Val42 and Tyr99. The number of waters within these limits was averaged for the final 100 ns of the trajectory using 100 frames/ns for a total of 10,000 frames. The total (and percent) change in water molecules was determined similarly to the method described for hydrogen bonds (vide supra).

QUANTIFICATION AND STATISTICAL ANALYSIS

Kinetic experiments were carried out in triplicate (n=3) and the data were analyzed using ORIGIN 9.0. The error values are shown as standard deviation in Table 1. For X-ray structures, we monitored the R_free during model building into electron density and used the Ramachandran plot to fix any outliers before submitting structures to the PDB/RCSB (https://www.rcsb.org). All the crystallographic statistics for the structure and data are shown in Table S1. The PDB/RCSB also analyzes a variety of metrics, compares them to other structures in the database at the same resolution, and describes them in the “Full wwPDB X-ray Structure Validation Report” before accepting these structures. These reports are available with each PDB code in the database. The NMR chemical shift perturbations were analyzed as a 10% trimmed of all data sets (n=7). Shifts were deemed significant if the magnitude was greater than 1.5 standard deviation units above the 10% trimmed mean. Instances of line broadening related to dynamical analysis were quantified as greater than 50% reduction in the peak height, measured within the NMRFAM-Sparky software. These are plotted in Figure S6. Thermodynamic parameters were extracted from nonlinear curve fitting using GraphPad Prism in the temperature-dependent circular dichroism experiments and reported in Figure S2.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
BL21-Gold (DE3) E. coli/PET-11b with human WTMIF	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF Y99A	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF Y99F	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF Y99G	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF H62Y	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF H62A	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF H62G	This paper	N/A
BL21-Gold (DE3) E. coli/PET-11b with human MIF H62F	This paper	N/A
Chemicals, Peptides, and Recombinant Proteins
4-hydroxyphenylatepyruvate (HPP)	Sigma-Aldrich	Cat# 14286
Ammonium Acetate	Sigma-Aldrich	Cat# A1542
Ammonium Chloride (15N, 99%)	Cambridge Isotopes Laboratories	Cat# NLM-467-1
cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail	Sigma-Aldrich	Cat# 11836170001
Deuterium Oxide (D2O, 98%)	Cambridge Isotopes Laboratories	Cat# DLM-4-99.8-1L
WTMIF	This paper	N/A
MIF Y99A	This paper	N/A
MIF Y99F	This paper	N/A
MIF Y99G	This paper	N/A
MIF H62Y	This paper	N/A
MIF H62A	This paper	N/A
MIF H62G	This paper	N/A
MIF H62F	This paper	N/A
PfuTurbo DNA Polymerase	Agilent	Cat# 600250
Critical Commercial Assays
QuikChange II site-directed mutagenesis kit	Agilent	Cat# E4780
Pierce BCA Protein Assay Kit	Thermo-Fisher	Cat# 23225
Deposited Data
MIF WT	(Crichlow et al., 2009)	PDB: 3DJH
MIF Y99A	(Pantouris et al., 2018)	PDB: 5EIZ
MIF Y99F	This paper	PDB: 6OYE
MIF Y99G	This paper	PDB: 6OY8
MIF H62Y	This paper	PDB: 5UMK
MIF H62A	This paper	PDB: 5UMJ
MIF H62G	This paper	PDB: 6OYB
MIF H62F	This paper	PDB: 6OYG
WT MIF (apo)	(Crichlow et al., 2009)	PDB: 3DJH
WT MIF-substrate (p-hydroxyphenyl pyruvate) complex	(Lubetsky et al., 1999)	PDB: 1CA7
WT MIF-competitive inhibitor complex	(Orita et al., 2001)	PDB: 1GCZ
WT MIF-competitive inhibitor complex	(Lubetsky et al., 2002)	PDB: 1LJT
WT MIF-competitive inhibitor complex	(Taylor et al., 1999)	PDB: 1MFI
WT MIF-competitive inhibitor complex	(McLean et al., 2010)	PDB: 3L5U
WT MIF-covalent inhibitor complex	(Tyndall et al., 2012)	PDB: 4F2K
WT MIF-covalent inhibitor complex	(Winner et al., 2008)	PDB: 3B9S
WT MIF-covalent inhibitor complex	(Spencer et al., 2015)	PDB: 3WNT
WT MIF-covalent inhibitor complex	(Crichlow et al., 2009)	PDB: 3CE4
WT MIF-covalent inhibitor complex	(Crichlow et al., 2012)	PDB: 3SMB
WT MIF-covalent inhibitor complex	(Crichlow et al., 2012)	PDB: 3SMC
WT MIF-covalent inhibitor complex	(Spencer et al., 2015)	PDB: 4OYQ
WT MIF-covalent inhibitor complex	(McLean et al., 2009)	PDB: 3JSF
WT MIF-covalent inhibitor complex	(McLean et al., 2009)	PDB: 3JSG
WT MIF-covalent inhibitor complex	(McLean et al., 2009)	PDB: 3JTU
WT MIF-covalent inhibitor complex	(Pantouris et al., 2015)	PDB: 4PO1
WT MIF-covalent inhibitor complex	(Pantouris et al., 2015)	PDB: 4TRF
WT MIF-covalent inhibitor complex	(Pantouris et al., 2015)	PDB: 4P0H
WT MIF-covalent inhibitor complex	(Pantouris et al., 2015)	PDB: 4PLU
Experimental Models: Cell Lines
BL21-Gold(DE3) Competent Cells	Agilent	Cat#230130
Oligonucleotides
See Table S4
Recombinant DNA
pET11b	Novagene	Cat# 69436
pET11b-MIF Y99A	This study	N/A
pET11b-MIF Y99F	This study	N/A
pET11b-MIF Y99G	This study	N/A
pET11b-MIF H62Y	This study	N/A
pET11b-MIF H62A	This study	N/A
pET11b-MIF H62G	This study	N/A
pET11b-MIF H62F	This study	N/A
Software and Algorithms
Arpeggio	Harry Jubb	https://bitbucket.org/harryjubb/arpeggio/src/master/
CHARMM36 Force Field	University of Maryland School of Pharmacy	http://mackerell.umaryland.edu/charmm_ff.shtml
COOT	SGBrid	https://sbgrid.org/software/titles/coot
GetContacts	GitHub	https://github.com/getcontacts/getcontacts
HKL2000	HKL Research, Inc.	https://www.hkl-xray.com/
MATLAB	MathWorks	https://www.mathworks.com/products/matlab.html
Nanoscale Molecular Dynamics	University of Illinois at Urbana-Champaign	http://www.ks.uiuc.edu/Research/namd/
NetworkX Python library	GitHub	Harberg et al. (2008)
NMRPipe	University of Maryland/NIST IBBR	http://www.ibbr.umd.edu/nmrpipe/
PHASER	CCP4	http://www.ccp4.ac.uk/html/phaser.html
Prism	GraphPad	https://www.graphpad.com/scientific-software/prism/
PyMOL	Schrödinger	https://pymol.org/2/
Python3	Python Software Foundation	https://www.python.org/
Refmac	SGBrid	https://sbgrid.org/software/titles/refmac
Sparky	NMRFAM	http://www.nmrfam.wisc.edu
SUPERPOSE	CCP4	http://www.ccp4.ac.uk/html/superpose.html
Visual Molecular Dynamics	University of Illinois at Urbana-Champaign	http://www.ks.uiuc.edu/Research/vmd/
Prism	Graphpad	www.graphpad.com/scientific-software/prism
Origin 9.0	https://www.originlab.com/	https://www.originlab.com/
Other
600 MHz Nuclear magnetic resonance spectrometer	Bruker	Model# Bruker Avance NEO series
Circular dichroism spectropolarimeter	Jasco Inc.	Model# J-815
Microplate reader	TECAN	Model# Infinite M200
Pilatus Detector/007 rotating copper anode X-ray generator	Rigaku, Inc.	https://medicine.yale.edu/xray/instruments/pilatus/

Highlights.

• An allosteric gating residue of a solvent channel affects the MIF enzymatic site

• Solvent channel water molecules mediate MIF catalysis

• A mechanism between the allosteric and enzymatic site is defined

SIGNIFICANCE.

Macrophage migration inhibitory factor (MIF) is a critical immunoregulatory cytokine and enzyme that is strongly implicated in inflammatory diseases, such as asthma, arthritis, acute respiratory distress syndrome, and sepsis, as well as several cancers. Attenuation of these disease phenotypes is well documented in MIF-knockout mice; however, the molecular mechanisms of the pathophysiology of MIF are unclear. Furthermore, efforts to target MIF with inhibitors or therapeutics are complicated by its nearly ubiquitous cellular localization and protective role in innate immunity. Additional mechanistic studies are therefore required to leverage control over the pathological effects of MIF. We recently identified Tyr99 of the MIF solvent channel, formed along the 3-fold axis of its trimeric structure, to be a critical mediator of MIF-induced activation of its CD74 receptor. Here, we confirm Tyr99 as an allosteric site in MIF, and reveal the structural, dynamic, and biochemical determinants of an allosteric pathway that couples Tyr99 to the MIF catalytic residue, Pro1. In addition to residue-residue interactions, we identified a functional role for solvent channel water molecules in MIF allostery, suggesting that this previously unexplored channel is not simply a consequence of MIF trimerization. The atomistic characterization of this novel allosteric network may provide alternate routes of therapeutic inhibition of MIF.