Crystal Structure and Transient Dimerization for the FKBP12 protein from the pathogenic fungus Candida auris

Abstract

International concern over the recent emergence of Candida auris infections reflects not only its comparative ease of transmission and substantial mortality but the increasing level of resistance observed to all three major classes of antifungal drugs. Diminution in virulence has been reported for a wide range of fungal pathogens when the FK506-binding protein FKBP12 binds to that immunosuppressant drug and the binary complex then inhibits the fungal calcineurin signaling pathway. Structure-based drug design efforts have described modifications of FK506 which modestly reduce virulence for a number of fungal pathogens while also lessening the side effect of suppressing the tissue immunity response in the patient. To aid in such studies, we report the crystal structure of Candida auris FKBP12. As physiological relevance has been proposed for transient homodimerization interactions of distantly related fungal FKBP12 proteins, we report the solution NMR characterization of the homodimerization interactions of the FKBP12 proteins from both Candida auris and Candida glabrata.

Graphical Abstract

1. Introduction

Since the first invasive infection of Candida auris reported in 2011 [1], such infections have been found in 39 different countries spanning all continents excepting Antarctica [2]. Significantly, variants of C. auris have emerged that are resistant to all three major classes of antifungal drugs. In recognition of this expanding health challenge, Candida auris became the first fungal pathogen to be included among the five most urgent threats in the Antibiotic Resistance Threats Report published by the Centers for Disease Control and Prevention (CDC) [3]. The potential risk has recently been clearly illustrated in New York, the most severely affected state within the USA, where among the first 331 clinical cases tested for resistance to all major classes of antifungal drugs, all but 1 (99.7%) were resistant to the conventional first-line antifungal drug fluconazole, while 63.4.% were resistant to amphotericin B. Although only 3.9% of the C. auris infections in this study were found to be resistant to echinocandins, each of the three patients who were found to be resistant to all three classes of drugs had begun their therapeutic treatment as echinocandin-sensitive and only developed resistance after being treated with that drug [4]. This level of resistance for both fluconazole and amphotericin B are reflective of C. auris infections derived from the South Asian clade which exhibits higher levels of resistance than the other three clades of this fungal species [5].

The potential risk of an emerging pan-resistance for C. auris is accentuated not only by estimated morbidity rates greater than 30% for untreated patients [6] but also by the comparative ease with which these infections are transmitted when contrasted with other fungal pathogens. A significant factor in this enhanced transmissibility is that C. auris can remain viable on dry plastic surfaces for up to 14 days and on most wet surfaces for up to a week [7]. As a result, C. auris infections have inspired containment and decontamination approaches more closely reminiscent of bacterial pathogens within health care facilities which remain a critical reservoir for clinical fungal infections [8].

In the effort to establish additional classes of broad range antifungal drugs, one seemingly promising alternative therapeutic target is the calcium-calmodulin-calcineurin signaling pathway. Elimination of the fungal calcineurin pathway leads to either complete (e.g., Cryptococcus neoformans [9], Candida gatti [10]) or severe (e.g., Candida albicans [11], Aspergillus fumigatus [12]) suppression of virulence in mouse models. While the serine-threonine phosphatase calcineurin is highly conserved between humans and fungi, there is a greater sequence divergence for the homologous FK506-binding protein FKBP12 which inhibits calcineurin when in complex with the immunosuppressant drug FK506. To facilitate structure-based drug design, crystal structures have been reported for FKBP12 proteins of C. albicans, C. glabrata, and A. fumigatus [13] while extensive NMR resonance assignments have been published for FKBP12 proteins of A. fumigatus and Mucor circinelloides [14] as well as for C. auris and C. glabrata [15].

Of particular relevance are proof-of-concept studies which have recently demonstrated that structure-based design of an acetohydrazine-substituted variant of FK506 showed reduced immunosuppressive effect with some antifungal activity against a range of pathogens including C. albicans, A. fumigatus, and C. neoformans [16]. Crystallographic studies by these authors noted an apparent homodimerization interaction in multiple crystal forms for the FKBP12 proteins from C. albicans and A. fumigatus, and they proposed a potential physiological role for this dimerization based upon the well-known prolyl isomerization activity that is catalyzed by members of this protein family [13]. To help enable similar studies on C. auris FKBP12, we report its crystal structure as well as NMR solution studies on the FKBP12 proteins from both C. auris and C. glabrata to assess the occurrence of selective homodimerization interactions.

2. METHODS

2.1. DNA constructs, expression and purification

Genes for the FK506-binding protein FKBP12 from Candida auris and Candida glabrata were chemically synthesized (Genscript) as described previously [15]. Ser 12 of the C. auris sequence (NCBI assembly ASM301371v2) has been reported as leucine in other sequencing studies. Protein expression and purification procedure for both proteins followed that previously described [15,17]. Bacterial growth was carried out in deuterated minimal medium containing 0.25% [U-²H]glycerol (Cambridge Isotopes) and 0.1% ¹⁵N ammonium chloride (Cambridge Isotopes). Following protein purification, the backbone amide positions were back-exchanged into protonated buffer at pH 9 for three hours at 25°C, and the sample was then neutralized.

2.2. X-ray crystallographic analysis

Concentrated C. auris FKBP12 (86 mg/ml) in 25 mM Tris-acetate, pH 8.0, 2 mM DTT was diluted to 21.5 mg/ml with a buffer of 20 mM Tris-HCl, pH 8.0, 200 mM NaCl. Initial crystallization conditions were established using the Hampton Research Crystal Screen I & II at room temperature in hanging drops. Upon optimization, large crystals were grown by mixing 2 μ l of protein solution with an equal volume of reservoir solution containing 54% saturated ammonium sulfate, 0.1 M HEPES pH 7.5, and 2% isopropanol. The crystals belong to space group P 4₃ 2 2 with cell parameters: a=b =74.80 Å, c=72.29 Å, α=β=γ=90˚. With one molecule per asymmetric unit, the crystal solvent content is 70%. For data collection, crystals were cryo-protected by a solution containing crystallization buffer supplemented with 20% glycerol, and then snap-cooled in liquid nitrogen. Diffraction data were collected at 100K at the beamline 19-ID NYX of the National Synchrotron Light Source II at the Brookhaven National Laboratory. Data processing and scaling with autoPROC utilized the XSCALE scaling pathway in XDS [18]. The C. auris FKBP12 crystal structure was obtained with the PHASER molecular replacement program within the PHENIX suite [19] using C. glabrata FKBP12 (PDB code: 5HT1) as the search model. Structural refinement was carried out using PHENIX, with model rebuilding using Coot [20]. The surface area was calculated using the Areamol program within the CCP4 suite [21]. Figures were created using UCSF CHIMERA.

2.3. Solution NMR spectroscopy

The protein samples were concentrated and then equilibrated into a pH 6.50 buffer containing 25 mM sodium phosphate, 2 mM dithiothreitol and 2 mM tris(2-carboxyethyl)phosphine and 6% ²H₂O by centrifugal ultrafiltration. The protein samples were adjusted to 1.00 mM based on 280 nm absorbance (e of 8480 for C. auris and 9970 for C. glabrata). A series of two-fold dilutions into the NMR buffer were 5 generated. NMR data were collected at 37°C on a Bruker Avance III 14.1 T (600 MHz) spectrometer equipped with a triple-resonance cryoprobe and pulsed-field Z-gradient. NMR data were processed and analyzed using FELIX software (Felix NMR). A weighting value of 0.2 was used for the ¹⁵N chemical shifts in the dissociation equilibrium analysis [22]. Analysis of the concentration-dependent chemical shift perturbation data was carried out using Origin 8.6 software (OriginLab).

3. Results and Discussion

During the NMR resonance assignment of the FKBP12 proteins of C. auris and C. glabrata [15], it was noted that the ¹H resonances for the aromatic ring of Phe 91 (Phe 94 in C. glabrata) were markedly broadened (Fig. 1). Unlike the well-studied human FKBP12 protein for which three distinct regions of the protein exhibit linebroadening effects that arise from conformational exchange dynamics of transitions occurring in the ~100 μ s timeframe [17,23,24], such exchange linebroadening effects are much more limited in both the C. auris and C. glabrata proteins. In crystal structures of related FKBP12 proteins, this phenylalanine sidechain and the homologous histidine sidechain of the human protein appear to be highly solvent-exposed with no obvious intramolecular interactions that might retard the kinetics of the normally rapid conformational transitions of such an unimpeded sidechain so as to shift those dynamics into the roughly millisecond timescale as is required to give rise to this type of linebroadening effect. In light of the proposed weak dimerization of the FKBP12 proteins from C. albicans and A. fumigatus [13], an alternative explanation for the observed linebroadening could be a moderately slow intermolecular dissociation reaction with kinetics occurring in the ~ms timeframe. The distinct behavior of this aromatic sidechain gains particular interest as the design study which generated the acetohydrazine-substituted variant of FK506 identified this phenylalanine residue as a critical site for fungal-selective drug interactions, proposing a direct steric interaction with the site of modification [16].

Fig. 1.

Selective ¹H resonance linebroadening for the aromatic protons of Phe 91 in C. auris FKBP12. The 2D ¹H-¹H TOCSY spectrum for a section of the aromatic spectral region illustrates broadened and weakened resonances for the sidechain of Phe 91, potentially indicative of a conformational transition at this site exhibiting dynamics in the sub-millisecond timeframe. Off-diagonal crosspeaks are identified by the ¹H resonance having its frequency in the direct observe dimension (x-axis).

To enable analogous studies for C. auris, the protein was crystallized and diffraction analysis was extended to 1.87 Å (Table 1). The most fruitful structural comparison was drawn against C. glabrata FKBP12 for which both the C. auris and human proteins generate an ungapped alignment throughout the sequence. Reflecting the broad divergence of genomic sequences that have been identified among fungi assigned to the Candida genus, the sequence identity between the C. auris and C. glabrata FKBP12 proteins is only 77% as compared to a 60% sequence identity with the human protein [15].

Table 1.

Data collection and crystallographic refinement statistics

Data collection
Space group	P 43 2 2
Cell dimension (Å)	a = b = 74.80, c = 72.29
Resolution (Å)	33.45–1.87 (1.94–1.87)a
Total reflections	209247 (22490)
Unique reflections	17303 (571)
Multiplicity	12.1 (13.3)
Completeness (%)	91.89 (33.75)
Mean (I)/sigma (I)	20.76 (0.74)
Wilson B-factor	45.46
R-merge	0.07096 (3.45)
R-meas	0.07417 (3.58)
R-pim	0.02119 (0.96)
CC1/2	0.999 (0.29)
CC*	1 (0.67)
Reflections used in refinement	16027 (571)
R-work	0.202 (0.340)
R-free	0.206 (0.359)
CC(work)	0.965 (0.618)
CC(free)	0.966 (0.566)
No. of non-hydrogen atoms	924
macromolecule	841
ligand	5
solvent	78
Protein residues	111
RMS (bonds)	0.009
RMS (angles)	0.95
Ramachandran
favoured (%)	97.25
allowed (%)	2.75
outliers (%)	0.00
Rotamer outliers (%)	0.00
Clashscore	1.78

^aStatistics for the highest-resolution shell are shown in parentheses.

The crystal structure of the unliganded C. auris FKBP12 protein (PDB code 6VSI) superimposes upon the crystal structure of the FK506-bound C. glabrata protein (PDB code 5HUA[13]) to a backbone heavy atom rmsd of 0.57 Å over residues 8 to 111 (Fig. 2). Despite the absence of a drug molecule in the crystal form, the aromatic ring of Phe 91 in the C. auris protein occupies nearly the identical position to that in the FK506-bound C. glabrata protein (Fig. 2). A similarly preserved positioning is observed for the sidechain of Leu 94. For the C. glabrata protein, the homologous sidechains are packed against the tetrahydropyran ring of FK506. In contrast, for the unliganded state of the C. glabrata protein (PDB code 5HT1[13]), the backbone of this segment is splayed away by 2 Å from its FK506-bound form. Differences in crystal lattice packing appear unlikely to explain this variation as the space group and cell parameters for the unliganded C. auris and C. glabrata structures are quite similar, while those of the FK506-bound C. glabrata FKBP12 are highly disparate. Nine of the 21 residues that lie at the interface between FKBP12 and the calcineurin CnA/CnB heterodimer in the crystal structure of the bovine ternary complex (PDB code 1TCO) are not conserved in the C. auris FKBP12 sequence [16,25]. Of these nine differing residues, only Phe 91 and Leu 94 are positioned to make direct contact with FK506 as well.

Fig. 2.

Crystal structure of the unliganded FKBP12 protein from C. auris. As displayed in ribbon representation, the backbone of this protein (cyan - PDB code 6VSI) closely aligns with that of the FK- 506 (green)-inhibited FKBP12 protein of C. glabrata (yellow - PDB code 5HUA[13]). The homologous sidechains of Phe 91 and Leu 94 of C. auris and Phe 94 and Leu 97 of C. glabrata are displayed to illustrate their close positioning to the tetrahydropyran ring of the inhibitor.

Preliminary ¹⁵N NMR relaxation measurements indicated an effective global rotational correlation time that was well above what had been previously observed for the comparably sized human protein, consistent with some degree of protein association. To provide maximal spectral resolution and sensitivity, a sample of the C. auris protein was prepared with uniform ²H and ¹⁵N enrichment. ¹H-¹⁵N 2D TROSY spectra were collected at pH 6.50 and 37°C on a set of C. auris FKBP12 samples serially diluted in buffer from 1.00 mM to 0.004 mM protein (Fig. 3). To maintain a comparable dynamic range for each spectrum, the lowest contour level displayed was scaled to the most intense peaks that undergo negligible shift migration during dilution. A subset of amide resonances exhibit substantial changes in frequency. Each of these perturbed residues shows a pattern of comparatively large changes in chemical shift at the higher concentrations and progressively smaller increments at the lowest concentrations. This pattern is consistent with the highest concentrations having approximately equal fractions of the total protein in the monomer and dimer states, while by the lowest concentrations the monomeric state predominates.

Fig. 3.

Concentration-dependent chemical shift perturbations in the ¹H-¹⁵N amide spectral region of the FKBP12 protein from C. auris. Starting at a concentration of 1.00 mM (red), 2D TROSY spectra were collected on a series of consecutive 2-fold dilutions down to 0.004 mM, displayed in yellow, gold, green, aqua, blue, indigo, purple, magenta. Peak intensities of each spectrum were scaled the most intense unperturbed resonances. The ten most strongly shifting amide resonances are identified by residue position.

A strong indication that these concentration-dependent chemical shift perturbation effects reflect the same dynamical process that gives rise to the sidechain resonance linebroadening (Fig. 1) is provided by the most strongly shifted amide resonance for Arg 89 (Fig. 3). While the amide resonance signal of Arg 89 is comparatively weak at the highest concentrations, its intensity becomes more similar to the other resonances at the lowest concentrations. This pattern of intensities reflects the fact that for transitions occurring in the sub-millisecond timeframe, the magnitude of the conformational exchange linebroadening effect is proportional to p_ap_bDw² where p_a and p_b denote the population of the monomer and dimer states and w is the difference in chemical shift between these two states [26]. As a result, the maximum exchange effect will occur when the populations of the monomer and dimer states are equal.

As the chemical shift effects arising from a rapid monomer-dimer equilibrium exchange are directly proportional to the fraction of the protein that exists in the free monomer state, the dissociation constant can be determined from the concentration-dependent changes [27], as illustrated for the ¹H^N resonance of Phe 91 (Fig. 4A). Since the differential ¹H and ¹⁵N chemical shifts for a given protein amide are presumed to reflect the same transition, chemical shift perturbation studies commonly use a weighted average of these two values for quantification purposes [22]. By applying this analysis to the amide resonances of C. auris FKBP12 that exhibit concentration-dependent shifts larger than 0.05 ppm, a value of 1.07 +/− 0.04 mM was derived for the dimer dissociation constant.

Fig. 4.

(A) Prediction of the dimerization dissociation constant from the concentration-dependent chemical shift values for the amide ¹H resonance of Phe 91. The rmsd fit yields a dimer dissociation constant of 1.08 mM. (B) The crystal lattice dimerization interaction of C. auris FKBP12. The ten residues exhibiting the largest concentration-dependent chemical shift perturbations are shown in red, orange, and gold to indicate the decreasing magnitude of those perturbations. On either side of the center are the p-stacked aromatic rings of Phe 91 from each monomer which project into the active site cleft of the opposite monomer.

While only a monomer of C. auris FKBP12 is contained with the asymmetric unit of the crystal, a symmetry partner is present in the lattice which forms an interface that is centered upon a pi-stacking interaction between the Phe 91 aromatic rings from the two monomers (Fig. 4B). The Phe 91 aromatic ring of one monomer is inserted into the active site of the other monomer so as to lie behind the Phe 91 ring of that monomer. When compared to the structure of FK506-inhibited C. glabrata FBKP12, it is apparent that the inserted Phe 91 ring is occupying the position of the tetrahydropyran ring of the inhibitor (Fig. 2). In the structure of the crystallographic model for the C. auris dimer, the area of the dimer interface is 806 Ų, consistent with a modestly stable binding interaction [28]. The ten residues containing the most strongly shifted backbone amides are highlighted in that dimerization model (Fig. 4B). In addition to the strongly shifted Phe 91 that is positioned at the center of the dimer interface, the other nine residues undergoing substantial concentration-dependent shifts either lay directly at the interface or are positioned nearby. While the quantitative structural interpretation of amide chemical shift perturbation effects remains problematic, the consistency of the structural distribution of the highlighted residues with the magnitude of their concentration-dependent chemical shift perturbations strongly support the interpretation that the dimerization interaction occurring in solution is well represented by the interface seen in the crystallographic analysis.

An analogous series of concentration-dependent 2D ¹H-¹⁵N TROSY spectra were collected on the [U-²H,¹⁵N]-labeled C. glabrata FKBP12 protein (Supplementary Fig. S1). The ten residues exhibiting the largest concentration-dependent perturbations map along the interface of the crystal lattice-modeled putative homodimer [13] (Supplementary Fig. S2). Eight of these are homologous to the set of most concentration-dependent residues in the C. auris protein. The direction and magnitude of the chemical shift perturbation for these eight residues were also quite similar. The derived K_D value of 0.91 ± 0.06 mM further supports the interpretation that C. glabrata FKBP12 undergoes a qualitatively similar dimerization interaction in solution.

A Pro–Gly dipeptide lies between Phe 91 and Leu 94 in the C. auris FKBP12 as well as for the homologous positions in C. glabrata and human proteins (Fig. 2). In contrast, a Pro-Pro sequence occurs at this site in the C. albicans and A. fumigatus proteins. In the crystal structure for the dimer of the A. fumigatus FKBP12 protein, one monomer has a cis-linkage for the second proline residue while the other monomer has a trans-proline linkage [13]. This isomerizing proline residue is positioned in the catalytic site in such a way as to prompt those authors to propose a possible physiologically relevant self-catalysis of prolyl isomerization. The presence of both cis- and trans-linkages at the homologous Pro-Pro sequence at the tip of the β₄-β₅ loop have previously been noted in crystal structures of the FK1 domain of human FKBP52 [29,30]. On the other hand, the homologous Gly 93 in C. auris FKBP12 and that of C. glabrata [13] are positioned well away from the presumed catalytically active residues of the opposite monomer. More exceptionally, the homologous glycine residue in human FKBP12 populates a cis-peptide linkage at the level of 12% [31], a 400-fold elevated population relative to typical non-prolyl peptide linkages [32]. Clearly, the conformational energetics of the β₄-β₅ loop can provide a substantial differential stabilization of a cis-peptide linkage at this position. However, no evidence has been published to demonstrate self-catalysis of peptide isomerization at this position in these FKBP domain proteins.

X-ray structure determination of FKBP12 from fungal pathogen Candida auris

FKBP12 provides a potential target for selective inhibition of fungal virulence

C. auris FKBP12 forms a weak but selective homodimerization interaction in solution