Structural Characterization of Thiadiazolesulfonamide Inhibitors Bound to Neisseria gonorrhoeae α-Carbonic Anhydrase

Abstract

Drug-resistant Neisseria gonorrhoeae is a critical threat to public health, and bacterial carbonic anhydrases expressed by N. gonorrhoeae are potential new therapeutic targets to combat this pathogen. To further expand upon our recent reports of bacterial carbonic anhydrase inhibitors for the treatment of N. gonorrhoeae, our team has solved ligand-bound crystal structures of the FDA-approved carbonic anhydrase inhibitor acetazolamide, along with three analogs, in complex with the essential α-carbonic anhydrase isoform from N. gonorrhoeae. The structural data for the analogs presented bound to N. gonorrhoeae α-carbonic anhydrase supports the observed structure–activity relationship for in vitro inhibition with this scaffold against the enzyme. Moreover, the ligand-bound structures indicate differences in binding poses compared to those traditionally observed with the close human ortholog carbonic anhydrase II. These results present key differences in inhibitor binding between N. gonorrhoeae α-carbonic anhydrase and the human carbonic anhydrase II isoform.

Neisseria gonorrhoeae is a Gram-negative human pathogen and the causative agent of the sexually transmitted disease gonorrhea. The World Health Organization estimated 82 million new cases of gonorrhea in 2020,¹ and the U.S. Centers for Disease Control and Prevention (CDC) reported a total of 677,769 new cases in the United States in 2020, with rates of gonorrhea increasing 5.7% from 2019.² This coupled with the increase in multi-drug resistance has led to the CDC to classify drug-resistant N. gonorrhoeae as an urgent threat to public health.³ For the past decade the CDC has recommended a combination treatment of oral azithromycin and intramuscular injection of ceftriaxone.⁴ However, increasing resistance to azithromycin led to the recommendation being altered, removing oral azithromycin; consequently, only a single 500 mg injection of ceftriaxone and no oral therapeutic options remain to treat gonorrhea.⁵

Carbonic anhydrases (CAs) are zinc-metalloenzymes that catalyze the interconversion of carbon dioxide (CO₂) to bicarbonate (HCO₃^–) and a proton.⁶ Human CAs, which consist of only the α-CA class, are important drug targets for treatment of glaucoma⁷ and epilepsy⁸ and as a diuretic.⁹ Bacteria express CAs; however, the diversity of isoforms and essentiality of CAs among different bacterial species is variable. For example, bacteria can encode and express many sub-families, including structurally distinct α-, β-, and γ-CA’s.¹⁰ The α-CA from Helicobacter pylori is essential for survival in the acidic environment in the stomach, and CA inhibitors (CAIs) show activity against H. pylori.^11,12Enterococcus faecium and Enterococcus faecalis express both α- and γ-CA isoforms that appear to be essential and are susceptible to CAIs.¹³Mycobacterium tuberculosis,¹⁴Vibrio cholera,¹⁵Burkholderia species,¹⁶Brucella suis,¹⁷ and many other bacteria also express CAs, with further research necessary to determine the essentiality in each.¹⁸

N. gonorrhoeae is susceptible to human CAIs, including acetazolamide (AZM),¹⁹ and expresses three sub-classes of carbonic anhydrases, with the α-CA (α-NgCA) proven to be essential for bacterial viability.²⁰ Our group developed novel CAIs with efficacy against α-NgCA and antimicrobial activity against N. gonorrhoeae.²¹ From these studies it has been demonstrated that α-NgCA is a primary target of these molecules, and, therefore, it is critical to understand the structural basis by which they interact with this CA for future design.

X-ray crystal structures have been solved for AZM, and other CAIs, bound to several human CA isoforms.^22,23 The structure of AZM bound to α-NgCA has also been reported;²⁴ however, neither structure of the α-NgCA deposited in the Protein Data Bank (PDB ID: 1KOP or 1KOQ) has the ligand AZM bound to the enzyme, leaving a gap in the knowledge for inhibitor binding. To remedy this, we solved ligand-bound structures of AZM, and three anti-gonococcal bacterial CAIs previously reported from our laboratory,²¹ bound to α-NgCA. The results reported herein: (1) provide a structural basis for the observed in vitro SAR of the anti-gonococcal CAIs from our lab (Table 1),²¹ (2) demonstrate differences observed in the binding modes for the 1,3,4-thiadiazolesulfonamides between human CAs and α-NgCA, and (3) provide improved understanding of inhibitor interactions with α-NgCA compared to the similar human carbonic anhydrase II (hCAII).

Table 1. Inhibition Constants and ΔG Values for CAIs against α-NgCA and hCAII.

^aCO₂ hydration assay inhibition. Values determined from the mean of one experiment in triplicate. Values reported are ± standard error of the mean, from a prior manuscript.²¹

^bCalculated using ΔG = −RT ln(K_i), where R = 8.314 J/K·mol), T = 298 K, and K_i is the mean value.

CAIs have been solved in complex with CAs using both co-crystallization and crystal soaking. We attempted both strategies but were unable to obtain structures using co-crystallization. Soaking α-NgCA crystals with inhibitors provided data sets with clear inhibitor density present in the active site. Crystallization conditions and procedures are reported in the experimental section of the Supporting Information. The α-NgCA crystals belong to the P2₁ space group, with four identical protein subunits in the asymmetric unit, with a root-mean-square deviation (rmsd) of 0.37 Å for the Cα atoms. One molecule of inhibitor was observed to occupy the binding site on each protein subunit in the asymmetric unit. (X-ray data in Tables S1 and S2.)

The α-NgCA structure solved by our team (PDB: 8DPC) was refined to 2.41 Å resolution and is similar to the α-NgCA structure deposited and described by Huang et al. (PDB: 1KOQ) with a rmsd of 0.31 Å for the Cα atoms.²⁴ The active-site residues for hCAII and α-NgCA are quite similar, particularly with two conserved Thr residues and a Gln residue, compared to other hCAs, such as hCAI, with a His in place of one of the Thr residues that impacts the binding of inhibitors. For purposes of comparison, we will focus on hCAII and α-NgCA due to the conservation of these key residues and wealth of structural data. The overall architectures of α-NgCA and hCAII (PDB: 1CA2) are similar and have been previously compared by Huang et al. in detail.²⁴ One notable difference in the overall structures relevant to the ligand binding discussion presented in this publication is the presence of two 3₁₀-helices in human CAs, depicted on hCAII from Thr125 to Ala134 (Figure S1B,D, lavender ribbon and surface), that are absent in α-NgCA (Figure S1A,C). This sequence in hCAII creates a hydrophobic cleft (Figure S1) and contains Phe131, known to interact with AZM and other CAIs. Conversely, the absence of these helices in α-NgCA exposes a hydrophobic channel lined by side chains Leu176, Pro179, Pro180, Thr181, and Val185 (Figure S1C, gold surface; electrostatic surface potential shown in Figure S2). This deletion of the helix-forming residues is conserved across bacterial α-CAs studied.^11,13,25,26

A structure of AZM bound to α-NgCA (Figure 1, PDB: 8DYQ) was solved and refined to 2.15 Å resolution with density for AZM in each protein subunit (Figure 1). The thiadiazole core is oriented in the same direction typically observed for AZM with α-CAs, with the diazo portion directed toward two threonine residues on the right-hand side of the active site. Comparing with AZM bound to hCAII,²⁷ the conserved Thr199 and Thr200 form the same interactions and orient AZM in the same direction and plane (Figure 1C). This binding pose is the placement traditionally observed for 1,3,4-thiadiazole CAIs against α-CAs.²⁸

Figure 1.

AZM bound to α-NgCA and hCAII (PDB: 8DYQ). (A) Full view of α-NgCA structure (gray ribbons) with AZM (cyan sticks). (Inset) Omit |2F_o – F_c| electron density map (yellow mesh) for AZM (cyan sticks) bound to α-NgCA (gray ribbon) at 2.15 Å resolution and contoured to the 1.0σ level. (B) Binding interactions for AZM with α-NgCA: hydrogen bonds (yellow dashes) and distance measurement to Gln90 (blue dashes). (C) Overlay of AZM binding poses with α-NgCA (gray ribbon/sticks) and hCAII (salmon ribbon/sticks). AZM: cyan sticks = α-NgCA; green sticks = hCAII. Residue positions are labeled; hCAII residues in parentheses. Phe131 is on the 3₁₀-helix from hCAII (lavender sticks).

A key difference between AZM bound to hCAII versus α-NgCA is the lack of an interaction with a Gln90 side chain in the active site of the bacterial isoform (Figure 1B). This commonly observed interaction between AZM and Gln92 in hCAII has a distance of 3.1 Å in the AZM/hCAII (PDB: 3HS4),²⁷ within the range of a moderate strength hydrogen bond.²⁹ In α-NgCA, the carbonyl group drops in the active site, increasing the measured distance from the nitrogen in the Gln side chain to 3.9 Å in the AZM/α-NgCA complex (Figure 1B, blue dashes)—a distance at the upper limit for a relatively weak interaction.²⁹ This is likely due to the absence of the 3₁₀-helix in α-NgCA compared to hCAII, as noted by the position of Phe131 to prop the carbonyl up on hCAII (Figure 1C). To assess if this is a lost interaction or crystallization artifact catching this pose, we compared inhibition data for AZM and two additional matched molecular pair (MMP) analogs lacking the carbonyl (Table 1). Analogs 1 and 2, each lacking the carbonyl, displayed similar K_i values against α-NgCA compared to AZM, all around 74 nM, suggesting the carbonyl is not involved in forming contacts with the Gln90. Alternatively, the same analogs lacking the carbonyl displayed 3.5–5-fold reduced inhibition against hCAII compared to AZM, confirming the carbonyl hydrogen bond with the Gln92 side chain contributes to binding in hCAII.

The Gibbs free energy of binding (ΔG) was determined using the K_i values, ΔG = −RT ln(K_i), to approximate the contribution of the hydrogen bond (Table 1). Comparing ΔG between AZM and 2 with hCAII, the difference ΔΔG = ΔG_AZM-hCAII – ΔG_2-hCAII is −4.39 kJ/mol (Table 1), approximating the value of the carbonyl hydrogen bond with Gln92 in hCAII. A similar comparison of AZM binding between the proteins α-NgCA versus hCAII results in ΔΔG = ΔG_AZM-hCAII – ΔG_AZM-NgCA = −5.89 kJ/mol, suggesting the hydrogen-bond interaction between the two proteins and potentially additional variables contribute to this difference. Taken altogether, the biochemical data reinforces the structural binding observations for AZM in binding to α-NgCA and hCAII.

The complex of 3 with α-NgCA was solved and refined to 2.80 Å resolution (Figure 2, PDB: 8DQF). Immediately it was noted that the density for the thiadiazole core was flipped, with density around the sulfur atom directed toward the threonine residues (Figure 2A), contrary to the traditional binding pose observed for AZM. A ligand-bound structure for this molecule in complex with hCAII has been reported (PDB: 7JNZ),³⁰ providing a comparator for hCAII in which the thiadiazole maintains the traditional binding orientation with the nitrogen atoms directed toward Thr199 and Thr200 of hCAII (overlay in Figure S3). The alternate thiadiazole orientation for the same analog may be an attribute of binding α-NgCA compared to hCAII or may possibly arise from differences in crystallization strategy (i.e., co-crystallization vs crystal soaking). The structure reported here was obtained through crystal soaking. We sought to confirm the method used for the 3-hCAII structure by Andring et al.³⁰ They utilized both co-crystallization and soaking techniques with different analogs. However, the specifics for 3 are not provided in their publication; therefore, we cannot speculate if a difference in crystallization procedure accounted for the alternative binding pose. Notably, in the 3-hCAII complex the cyclohexane ring of 3 interacts with Phe131, pushing the cycloalkane to occupy a P2 hydrophobic site with an Ile91 and Val121 (Val113 in α-NgCA) (Figure S3A). The hydrophobic cleft in hCAII props the cyclohexane ring upward, similar to that observed with AZM, while the lack of this cleft in α-NgCA exposes Val113, Leu115, and Pro121 for hydrophobic interactions lower in the hydrophobic channel (overlay in Figure S2C,D).

Figure 2.

Binding mode for 3 to α-NgCA (PDB: 8DQF). (A) Omit |2F_o – F_c| electron density map (yellow mesh) for 3 (salmon sticks) bound to α-NgCA (gray ribbon) at 2.80 Å resolution and contoured to the 1.0σ level. Amino acid residues shown as sticks/labeled; Zn²⁺ shown as the gray sphere. (B) Interactions for 3/α-NgCA: hydrogen bonds (yellow dashes) and hydrophobic interactions (magenta dashes). (C) Surface representation of α-NgCA (gray surface) in complex with 3 (salmon sticks) with a hydrophobic channel (gold surface). (D) Overlay of 3 (salmon sticks) and AZM (cyan sticks) binding to α-NgCA (gray ribbon and sticks). Black arrow indicates movement of 3 toward Gln90 compared to AZM.

Contrary to AZM, molecule 3 is observed to form a hydrogen bond between the carbonyl and Gln90 in α-NgCA. The new Gln90 interaction may be a result of the aforementioned hydrophobic interactions below the active site. The density for the cyclohexane is not fully complete; however, the modeled ligand based on the partial density indicates directionality toward the hydrophobic residues Val113, Leu115, and Pro121 (Figure 2B,C). This brings the cyclohexane within 4.0 Å of both Val113 and Leu115 side chains, facilitating hydrophobic interactions, and subsequently pulls the ligand from the traditional interactions with Thr178 toward Gln90, perhaps resulting in the core flip of 180°. The carbonyl is also pulled closer toward Gln90 compared to AZM, thus providing access to the hydrogen bond (Figure 3D).

Figure 3.

Binding mode for 6 to α-NgCA (PDB: 8DR2). (A) Omit |2F_o – F_c| electron density map (yellow mesh) for 2 (orange sticks) bound to α-NgCA (gray ribbon) at 2.81 Å resolution and contoured to the 1.0σ level. Amino acid residues are shown as sticks/labeled; active-site Zn²⁺ shown as the gray sphere. (B) Density map side view of 6 bound to α-NgCA. (C) Surface representation of α-NgCA (gray surface) in complex with 6 (orange sticks) with the hydrophobic channel (gold surface) shown. (D) Proposed interactions for 6 (orange sticks) with α-NgCA (gray ribbon and sticks). Hydrogen bonds (yellow dashes), hydrophobic contacts (magenta dashes), and distance from amide carbonyl to Gln90 (blue dashes) are depicted.

Inhibition data corroborates these structural observations for molecule 3. Comparison of K_i data for AZM and 3 indicates a roughly 7.5-fold improvement against α-NgCA: K_i from 74.1 ± 3.2 nM for AZM to 9.8 ± 0.4 nM for 3 (Table 1). This is likely attributed to the hydrogen bond with Gln90 and additional hydrophobic contacts now observed for 3 compared to AZM. Moreover, the MMP analog lacking the carbonyl, analog 4, exhibited a 4.3-fold reduced potency with a K_i value of 42.1 ± 2.9 nM, implying the loss of the carbonyl–Gln90 interaction observed in the 3/α-NgCA complex. Finally, we observed that phenyl derivatives with analog 5 displayed a K_i value of 78.6 ± 5.2 nM, 8-fold less potent than 3, suggesting that alkyl derivatives are favored over aromatics for the interactions with the hydrophobic side chains of Val113 and Leu115.

The calculated ΔΔG value for the analogs bound to α-NgCA using the respective K_i values for 3 and the non-carbonyl-containing 4 is ΔΔG = ΔG_3-NgCA – ΔG_4-NgCA = −4.82 kJ/mol (Table 1). The same comparison for hCAII yields ΔΔG = ΔG_3-hCAII – ΔG_4-hCAII = −4.48 kJ/mol. These values are similar to the ΔΔG value −4.39 kJ/mol calculated above, attributed to the loss of the carbonyl hydrogen bond of AZM when binding hCAII, suggesting the hydrogen-bond energy resides in the 4.5 kJ/mol range.

Lacking the hydrogen bond, 4 still maintains an edge over AZM for inhibition of α-NgCA, with a ΔΔG value −1.88 kJ/mol. This residual activity difference is nearly identical to the ΔΔG value −1.87 kJ/mol comparing 4 to 2 and likely represents the contribution in binding energy from the additional hydrophobic interactions with Val113 and Leu115. Additionally, quantification of the difference in binding energy between the MMP cyclohexane derivative 3 and phenyl derivative 5 indicates a ΔΔG = ΔG_3-NgCA – ΔG_5-NgCA = −6.69 kJ/mol. Structural data is needed to assess to what degree the ΔΔG value is attributed to the alkyl versus aromatic groups or if there is a subsequent loss of other interactions due to the phenyl moiety. Attempts to solve a ligand-bound structure for 5 have been unsuccessful up to this point. Nonetheless, the overall biochemical analysis supports the ligand binding pose for 3 in complex with α-NgCA.

Next, a structure of analog 6 in complex with α-NgCA was solved and refined to 2.81 Å resolution (Figure 3, PDB: 8DR2). The thiadiazole core was again observed to be flipped in the active site as compared to AZM and similar to 3. This analog is a congener of the previous analog 3 with a methylene inserted between the carbonyl of the amide and the cyclohexane ring. The density around this cyclohexane ring in 6 is better defined and projects deeper into the hydrophobic channel compared to that in 3 (Figure 3C and Figure S4). This extension effectively pulls the amide carbonyl away from Gln90, increasing the distance to approximately 3.9 Å and presumably losing the productive contact with the Gln side chain (Figure 3D). However, the dynamics of the ligand within the active site could still provide weak contact to the Gln90 while bound. The cyclohexane moiety remained approximately 4.2 Å away from Leu115 and 4.5 Å from Val113, which could feasibly maintain hydrophobic, albeit weaker, interactions with these side chains. Additionally, the hydrophobic tail appears to pull the central thiadiazole core slightly lower for 6 compared to 3 (Figure S4).

Contrary to the binding pose for 6 in which the carbonyl–Gln90 contact appears to be absent, the biochemical analysis suggests the carbonyl does still factor in binding for the analog (Table 1). MMP analysis between analogs 6 (contains carbonyl) and 7 (lacking carbonyl) indicates an approximately 1.8-fold decrease in K_i potency when the carbonyl is removed, equating to a ΔΔG of binding to α-NgCA of +1.94 kJ/mol for 7 compared to 6. This value is less than half the proposed value of the carbonyl–Gln90 interaction described above. One explanation may be that in solution the carbonyl in analog 6 still forms a weak transient hydrogen bond with Gln90. A second possible explanation is the carbonyl in 6 still forms a hydrogen bond with Gln90 but upon its removal the additional methylene is able to form a compensatory hydrophobic interaction with Val113. Regardless, the structural and biochemical data agree that the contribution of the carbonyl is not as strong as observed for analog 3, perhaps due to the hydrophobic tail projecting lower into the channel and pulling the carbonyl with it.

The final complex to report is analog 8 bound to α-NgCA and refined to 2.59 Å resolution (Figure 4, PDB: 8DRB). This analog has two differences compared to the previous analogs shown in complex with α-NgCA: (1) an additional methylene within the linker, and (2) a pendant phenyl group as opposed to the cyclohexane moieties. Like 3 and 6, the central thiadiazole core is rotated 180° in the active site. The amide carbonyl, now situated 3.4 Å from the Gln90 side chain, accepts a hydrogen bond from Gln90. The extended linker and phenyl pendant group adopt a confirmation in which the phenyl ring is folded back toward the bottom of the hydrophobic channel, thus placing the linker in a high-energy pseudo-cis conformation. This seats the phenyl ring in proximity to Val113 and Leu115, shown previously to partake in hydrophobic interactions. The phenyl group being held in proximity to these residue side chains, coupled with the adopted conformation of the two-methylene linker, appears to push the carbonyl upward into proximity of Gln90 to form the hydrogen-bond interaction.

Figure 4.

Binding mode for 8 to α-NgCA (PDB: 8DRB). (A) Omit |2F_o – F_c| electron density map (yellow mesh) for 8 (dark gray sticks) bound to α-NgCA (gray ribbon) at 2.59 Å resolution and contoured to the 1.0σ level. Amino acid residues shown as sticks/labeled; active-site Zn²⁺ shown as the gray sphere. (B) Density map side view of 8 bound to α-NgCA. (C) Surface representation of α-NgCA (gray surface) in complex with 8 (dark gray sticks) with hydrophobic channel (gold surface) shown. (D) Proposed interactions for 8 (dark gray sticks) with α-NgCA (gray ribbon and sticks). Hydrogen bonds (yellow dashes) and hydrophobic contacts (magenta dashes) are depicted.

Unfortunately, a nearest-neighbor analog for 8 lacking the carbonyl was not available for the biochemical analysis to confirm the carbonyl hydrogen-bond contact. However, the 4.4-fold boost in potency versus α-NgCA compared to 6 suggests a strong interaction has been gained (Table 1). It is also worth noting that the biochemical data again suggests a preference for the cycloalkane 9 over the phenyl derivative 8 by almost 12-fold in terms of K_i values, a trend similar to that observed previously comparing cyclohexane derivative 3 to phenyl 5 (8-fold difference). These values result in a ΔΔG = ΔG_9-NgCA – ΔG_8-NgCA = −8.18 kJ/mol compared to −6.69 kJ/mol for 5 compared to 3. While the entire difference in binding between 9 and 8 may not be attributed solely to the difference between the phenyl and the cyclohexane interactions with Val113 and Leu115, there is clearly a preference for the cycloalkane over the aromatic tail group. The large ΔΔG may also be a factor of the high-energy conformation of the linker region in 8. It is possible that the linker of the bound structure for 9 may not be in the same conformation, allowing for a lower energy conformation and thus improving the binding affinity. Efforts are ongoing to solve the structure of 9 in complex with α-NgCA but have been unsuccessful to this date. Future crystallography studies and isothermal titration calorimetry will assist in elucidating the binding thermodynamics between the phenyl and cyclohexane tails. Nonetheless, the clear preference for cycloalkyl groups rather than phenyl in binding to α-NgCA is a robust SAR trend observed over several nearest-neighbor analog pairs and corroborated by the structural studies.

In conclusion, this study presents ligand-bound structures for AZM and analogs 3, 6, and 8 in complex with α-NgCA to further elucidate the observed SAR. It was demonstrated that AZM has a slightly different binding mode when in complex to α-NgCA compared to the closely related human ortholog hCAII, and this observed difference was corroborated by in vitro inhibition constants against the two CAs. Notably, the absence of the 3₁₀-helix in α-NgCA creates a hydrophobic channel below the active site not present in hCAII, in which the alkyl and phenyl tails of analogs 3, 6, and 8 are shown to thread down to access new hydrophobic contacts (overlay of molecules in Figure S4). This appears to influence the bonding mode in two ways: (1) the ability of the carbonyl in the analogs to form a hydrogen bond with Gln90 in α-NgCA and (2) positions the ligands away from the conserved Thr side chains on the right-hand of the active site, thus allowing for the heterocyclic core to rotate 180°. Conversely, hCAII also contains a hydrophobic surface area defined by Phe131, Ile91, and Val121, and analogs 3, 6, and 8 still demonstrated potent inhibition toward hCAII. However, this hydrophobic surface is structurally different compared to that of α-NgCA and was shown in previous reports to prop the amide tail up for AZM and 3 to maintain access to the Gln92 hydrogen bond in hCAII. Notably, this series of ligand-bound structures are corroborated, and further explained, by observed in vitro biochemical inhibition data against α-NgCA and hCAII and provide tools for future design of bacterial carbonic anhydrase inhibitors with improved anti-gonococcal activity.

Glossary

Abbreviations

AZM

acetazolamide

α-NgCA

Neisseria gonorrhoeae alpha-carbonic anhydrase

hCAII

human carbonic anhydrase II

CA

carbonic anhydrase

CAI

carbonic anhydrase inhibitor

MMP

matched molecular pair

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.2c00471.

• X-ray data collection statistics, electrostatic surface potential for hCAII and α-NgCA, and binding modes for analog 3 to α-NgCA and hCAII; overlay comparison of novel inhibitor binding modes; and experimental section (PDF)

• Protein Data Bank validation reports for PDB entries 8DPC, 8DYQ, 8DQF, 8DR2, 8DRB (ZIP)

Author Contributions

Small-molecules synthesis: M.S.Y., W.A., and K.J.H. Biochemical inhibition: A.N. and C.T.S. Crystallography, soaking, data collection, structural determination, and model building: A.K.M., C.D., and R.Y. Project coordination/planning: M.N.S. and D.P.F. Manuscript draft: A.K.M. and D.P.F. All authors have given approval for the final version of the manuscript.

Work funded by Purdue College of Pharmacy (D.P.F.), the Purdue Institute for Drug Discovery (D.P.F./M.N.S), the National Institute for Allergy and Infectious Diseases (Grant 5R01AI148523, D.P.F./M.N.S.; and 1R01AI153264, D.P.F./M.N.S.), and the Italian Ministry for University Research (Grant FISR2019_04819 C.T.S.). Research reported in this publication was partially supported by the National Institute of Allergy And Infectious Diseases of the National Institutes of Health under Award Number T32AI148103. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science user facility operated for the DOE Office of Science by Argonne National Laboratory under the Contract No. DE-AC02-06CH11357.

The authors declare no competing financial interest.