Unexpected binding mode of the sulfonamide fluorophore 5-dimethylamino-1-naphthalene sulfonamide to human carbonic anhydrase II. Implications for the development of a zinc biosensor.

The three-dimensional structure of human carbonic anhydrase II (CAII) complexed with the sulfonamide fluorophore 5-dimethylamino-1-naphthalene sulfonamide (dansylamide) has been determined to 2.1-Å resolution by x-ray crystallographic methods. Unlike other arylsulfonamide inhibitors of CAII, the naphthyl ring of dansylamide binds in a hydrophobic pocket in the active site, making van der Waals contacts with Val-121, Phe-131, Val-143, Leu-198, and Trp-209. Interestingly, a conformational change of Leu-198 is required to accommodate dansylamide binding, which rationalizes the enhanced dansylamide affinity measured for certain Leu-198 variants (Nair, S. K., Krebs, J. F., Christianson, D. W., and Fierke, C. A. (1995) Biochemistry 34, 3981–3989). Modeling studies indicate that a second binding mode, in which the fused aromatic ring is rotated out of the hydrophobic pocket, is sterically feasible. Both experimentally observed and modeled binding modes have implications for new leads in the design of avid CAII inhibitors. Finally, the structure of the CAII-dansylamide complex has implications for its exploitation in zinc biosensor applications, and possible routes toward the optimization of fluorophore design are considered on the basis on this structure.

Sensors based upon biological macromolecules are increasingly used for the selective detection of numerous chemical entities, including the detection of trace quantities of metal ions. The advantages of metalloprotein-based biosensors over classical fluorometric chemical indicators for such applications include greater recognition and affinity for the target metal ion, exquisite discrimination among transition metals, and kinetically rapid biosensor-analyte association and dissociation. Recently, Thompson and Jones (1993) have exploited the selective recognition of zinc by the metalloenzyme carbonic anhydrase II (CAII) 1 in the design of an enzyme-based zinc biosensor. The physical basis of this sensor is the binding of zinc to the CAII apoenzyme, followed by binding of the fluorophore inhibitor 5-dimethylamino-1-naphthalene sulfonamide (dansylamide, K d ϭ 0.93 M (Nair et al., 1995); see Fig. 1) to the zinc-containing CAII holoenzyme. This leads to a titratable change in the fluorescence emission wavelength and intensity (Chen and Kernohan, 1967); nanomolar concentrations of zinc can be detected and quantified by ratiometric methods by measuring the emission of free dansylamide at 580 nm and the emission of bound dansylamide at 470 nm when excited at 290 nm.
The zinc ion of CAII resides at the base of a 15-Å deep cleft, where it is liganded by the imidazole side chains of His-94, His-96, and His-119 (Håkansson et al., 1992). Hydroxide ion, which is the catalytic nucleophile in the CO 2 hydration mechanism, completes a tetrahedral metal coordination polyhedron. Zinc-bound hydroxide donates a hydrogen bond to the hydroxyl group of Thr-199, which in turn donates a hydrogen bond to Glu-106 (Eriksson et al., 1986(Eriksson et al., , 1988aHåkansson et al., 1992). The binding of sulfonamide inhibitors displaces zinc-bound hydroxide and maintains the tetrahedral metal coordination polyhedron by the coordination of an ionized sulfonamide NH group, which also maintains the hydrogen bond interaction with Thr-199 (e.g. see Eriksson et al. (1988b) and Vidgren et al. (1990)). Although the crystal structures of a number of CAIIarylsulfonamide complexes have been determined (Eriksson et al., 1988b;Vidgren et al., 1990;Bunn et al., 1994;Boriack et al., 1995), none of the arylsulfonamides studied to date contain fused aromatic rings as found in the naphthyl moiety of dansylamide. Hence, the molecular details of CAII-dansylamide complexation cannot be predicted based on these structures. An understanding of these details is important not only for optimizing the design of a CAII-based biosensor, but also for exploring new regions of inhibitor affinity in the active site of this pharmaceutically important metalloenzyme (CAII inhibitors are highly effective in glaucoma therapy (Baldwin et al., 1989)). Indeed, CAII-inhibitor affinity is routinely measured by competition against the CAII-dansylamide complex (Chen and Kernohan, 1967).
Here, we report the crystal structure of the CAII-dansylamide complex determined at 2.1-Å resolution, and we demonstrate a unique binding mode for the naphthyl moiety of the inhibitor previously unobserved in any other CAII-arylsulfonamide complex. This has clear implications for new leads in the design of pharmaceutically important CAII inhibitors. The structure of the CAII-dansylamide complex also rationalizes the fluorescence quenching behavior observed in the complex (Chen and Kernohan, 1967) as well as inhibitor affinity variations observed among CAII variants Nair et al., 1995). Finally, this structure provides a foundation for the design of second-generation sulfonamide-based fluorophores with optimal properties for zinc biosensor applications.

EXPERIMENTAL PROCEDURES
Lyophilized CAII was purchased from Sigma and affinity purified by the method of Khalifah (1971). Attempts to obtain crystals of the CAII-* This work and the x-ray data acquisition facility were supported by a grant from the Office of Naval Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Supported in part by National Institutes of Health Postdoctoral Fellowship F32 CA62613.
§ To whom correspondence should be addressed. Tel.: 215-898-5714; Fax: 215-573-2201; E-mail: chris@xtal.chem.upenn.edu. 1 The abbreviations used are: CAII, carbonic anhydrase II; dansylamide, 5-dimethylamino-1-naphthalene sulfonamide. dansylamide complex by typical cocrystallization methods (Bunn et al., 1994) were unsuccessful, and crystal soaking experiments in various precipitant buffers failed to yield appreciable occupancy for a bound inhibitor. However, a modified sample preparation scheme facilitated the crystallization of the complex. Samples of dansylamide at 5 mM concentration in 5% acetone were serially added to a dilute sample of CAII (0.03 mM) until Ͼ95% inhibitor binding was achieved as determined by loss of esterase activity against p-nitrophenylacetate. The complex was then eluted through a Bio-Rad PD-10 desalting column to remove excess inhibitor and concentrated in an Amicon stir-cell to 0.3 mM. Crystallizations were then performed by the sitting drop method. A 10-l drop containing the CAII-dansylamide complex in 50 mM Tris-HCl (pH 8.0 at 24°C) was added to a 10-l drop of precipitant buffer containing 1.75-2.5 M ammonium sulfate and 50 mM Tris-HCl (pH 8.0 at 24°C) and equilibrated against 1 ml of the precipitant buffer in the reservoir. After 3 days, each sitting drop well was streak-seeded with micro-crystals of uncomplexed CAII using a thin glass wire. These streak seeds produced small well-formed crystals (Stura and Wilson, 1990), which were then used as macroseeds in sitting drop wells containing precipitant buffer plus saturated methyl mercuric acetate (necessary for the formation of diffraction-quality parallelepipedons). Crystals of the CAII-dansylamide complex appeared with typical dimensions of 0.2 ϫ 0.2 ϫ 0.8 mm within a week.
Crystals of the CAII-dansylamide complex were mounted and sealed in 0.5-mm glass capillaries with a small portion of mother liquor. Complete x-ray diffraction data to 2.1-Å resolution were collected at room temperature on an R-Axis IIc image plate detector using Cu-K␣ radiation generated by a Rigaku RU-200 HB rotating anode operating at 50 mV and 100 mA. Data frames of 2.0°oscillation about the spindle axis were collected, with exposure times of 15 min/frame, for a total angular rotation range of 90°about . Raw data frames were analyzed with the MOSFLM suite of programs (Nyborg and Wonacott, 1977). These crystals were isomorphous with crystals of the native enzyme (Eriksson et al., 1988a;Håkansson et al., 1992) and belonged to space group P2 1 with unit cell dimensions of a ϭ 42.7 Å, b ϭ 41.7 Å, c ϭ 73.0 Å and ␤ ϭ 104.6°as determined by the Kabsch autoindexing algorithm (Kabsch, 1978). Data were scaled and corrected for Lorentz and polarization effects using ROTOVATA, and replicate and symmetry-related reflections were merged using AGROVATA from the CCP4 suite of programs (Collaborative Computing Project, 1994); relevant data collection statistics are recorded in Table I.
Atomic coordinates of the 1.54-Å structure of native CAII (Håkansson et al., 1992) were retrieved from the Protein Data Bank (Bernstein et al., 1977), and these coordinates, less active site solvent molecules, were refined against structure factors obtained from the corrected intensity data. Refinement utilized the simulated annealing protocol of Brü nger et al. (1985) installed on a Silicon Graphics IRIS workstation. During the course of refinement, residue conformations and solvent molecules were routinely examined using simulated annealed omit maps calculated with Fourier coefficients 2͉F o ͉ Ϫ ͉F c ͉ or ͉F o ͉ Ϫ ͉F c ͉ and phases derived from the in-progress model. Model building was performed with the graphics programs CHAIN (Sack, 1988) and O (Jones et al., 1991). The inhibitor molecule was modeled into density after the R factor dropped below 0.19; force field parameters for the inhibitor molecule were generated using the XPLO2D utility provided with the O package (Jones et al., 1991).
Initial attempts to constrain the 5-dimethylamine of dansylamide in the naphthyl plane (the rationale being that an sp 2 -hybridized nitrogen would be conjugated with the aromatic ring) were unsuccessful. Surprisingly, the electron density map indicated that the dimethylamine moiety was approximately perpendicular to the plane of the aromatic ring. A search of small molecule crystal structures contained in the Cambridge Structural Data Base (Allen et al., 1979) revealed four additional examples of naphthyl or quinoline dimethylamines exhibiting similar conformations (Antolini et al., 1986;Foitzik et al., 1991;Staab et al., 1991). This confirmed the observed conformation of the dansylamide dimethylamino group, which persisted throughout refinement.
Only minimal adjustments to the protein model were necessary during the course of refinement. Refinement converged smoothly to a final crystallographic R-factor of 0.185 with R free ϭ 0.233 (Brü nger, 1992). The final model had excellent stereochemistry with root mean square deviations from ideal bond lengths and angles of 0.007 Å and 1.6°, respectively; relevant refinement statistics are recorded in Table I. The root mean square error in atomic coordinates is calculated to be 0.25 Å based on relationships derived by Luzzati (1952) and 0.39 Å based on the SIGMA-A estimation (Srinivasan, 1966;Read, 1986). Refined atomic coordinates have been deposited in the Protein Data Bank (Bernstein et al., 1977).

RESULTS AND DISCUSSION
An electron density map of the CAII-dansylamide complex is found in Fig. 2. As expected, the ionized sulfonamide NH group of dansylamide displaces zinc-bound hydroxide, maintains tetrahedral metal coordination geometry, and donates a hydrogen bond to Thr-199. Additionally, a sulfonamide SϭO group accepts a hydrogen bond from the backbone NH group of Thr-199. This binding mode is consistent with structures of other CAIIarylsulfonamide complexes (Eriksson et al., 1988b;Vidgren et al., 1990;Bunn et al., 1994;Boriack et al., 1995), and it appears that the coordination interaction with zinc substantially contributes to high affinity binding (K d ϭ 0.93 M (Nair et al., 1995)). However, the aromatic naphthyl group of dansylamide adopts an orientation previously unobserved in any other CAIIarylsulfonamide complex. Rather than extending out into the active site cleft, the naphthyl ring is situated within the hydrophobic substrate association pocket in much the same location occupied by the competitive inhibitor phenol (Fig. 3; Nair et al., 1994). In comparison with typical arylsulfonamides such as the oligoglycol derivative 4-((phenylalanyltriethyleneglycol)carboxy)benzene sulfonamide (Boriack et al., 1995), the fused aromatic ring of dansylamide is rotated ϳ54°about the sulfonamide N-S-C-C dihedral angle (Fig. 4). The aromatic ring  of dansylamide makes van der Waals contacts with pocket residues Val-121, Phe-131, Val-143, . The dimethylamino group of dansylamide makes no hydrogen bond interactions with the protein.
Although the overall structure of CAII remains virtually unchanged by inhibitor binding (root mean square deviation of C ␣ coordinates ϭ 0.17 Å between native CAII (Håkansson et al., 1992) and its dansylamide complex), there are local differences in active site side chain conformations that accompany dansylamide binding. A conformational change is observed for His-64, the catalytic proton shuttle (Steiner et al., 1975;Tu et al., 1989); the imidazole side chain undergoes a rotation of 58°a bout side chain torsion angle 1 to the "out" conformation (data not shown), as observed in certain CAII structures (e.g. see Nair and Christianson (1991)). Although such a rotation is typically observed in structures of CAII-inhibitor complexes where the inhibitor sterically requires it (Baldwin et al., 1989;Smith et al., 1994), there is no such requirement for dansylamide binding. Interestingly, a comparable His-64 rotation is observed in complexes of certain CAII variants with acetazolamide, even though this inhibitor similarly does not require this rotation (Nair et al., 1995).
A conformational change of Leu-198 is required for dansylamide binding but not for typical arylsulfonamide binding to CAII (Fig. 4). This provides the first structural rationalization for enhanced dansylamide affinity measured for certain Leu-198 CAII variants (Nair et al., 1995); in variants lacking a large side chain at position 198, the conformational change of this residue does not present a barrier to dansylamide binding and affinity is therefore improved. The role of adjacent residue Cys-206 in modulating dansylamide affinity is also rationalized ; in the native enzyme (Eriksson et al., 1988a;Håkansson et al., 1992), Cys-206 might sterically hinder the movement of Leu-198 required for dansylamide binding. Thus, either the conformational change of Cys-206 (e.g. as induced by Hg 2ϩ binding (Tilander et al., 1965;Eriksson et al., 1986Eriksson et al., , 1988a) or the substitution of larger side chains at position 206 (which would sterically eliminate the barrier for the Leu-198 conformational change) should enhance dansylamide binding. Determination of the three-dimensional structures of  variants and/or their dansylamide complexes may provide further insight on this proposal.
Given the required rearrangement of Leu-198 and the unexpected orientation of the inhibitor in the CAII-dansylamide complex, other possible orientations of the naphthyl ring have been modeled and compared with the experimentally observed conformation to explore further structure affinity relationships in the CAII active site. Since this is the first structure of a FIG. 2. Simulated annealing omit map of the CAII-dansylamide complex contoured at 2.5 . The map is calculated with Fourier coefficients ͉F o ͉-͉F c ͉ and phases derived from the final atomic model minus the atoms corresponding to the inhibitor subject to a cycle of simulated annealing refinement. The final refined atomic coordinates are superimposed, and residues Ile-91,  are indicated. The zinc ion appears as a gray sphere. Although somewhat noisy, this simulated annealing omit map minimizes model bias on the observed electron density.

FIG. 3. Least-squares superposition of the CAII-dansylamide (thick bonds) and phenol (Nair et al., 1994) (thin bonds) complexes.
Enzyme residues Ile-91,  are indicated; the zinc ion appears as a gray sphere, and zinc-bound solvent appears as a dark sphere. Note that the aromatic rings of both inhibitors are nearly coplanar as they bind in the active site hydrophobic pocket.
bound arylsulfonamide bearing an ortho substituent to the sulfonamide moiety (i.e. the fused aromatic ring), both experimental and modeling data derived from this structure may illuminate new structure-activity relationships that may be relevant to the design of pharmaceutically useful CAII inhibitors (Maren, 1987). Notably, previous structure-activity relationships indicate that an ortho substituent on an arylsulfonamide moiety compromises enzyme-inhibitor affinity (Burbaum et al., 1995;Vedani and Meyer, 1984;Hansch et al., 1985;Menziani et al., 1989); the three-dimensional structure of the CAII-dansylamide complex suggests that this is not an absolute relationship if alternate conformations of Leu-198 and zinc-bound sulfonamide are attainable. Holding zinc-bound sulfonamide fixed in its experimentally observed position, the naphthyl group of dansylamide can be rotated about the C-S bond into an alternate conformation, which does not suffer from severe steric clashes and engages in van der Waals interactions with Asn-62, ); a comparable binding mode has been suggested by Vedani and Meyer (1984). However, this alternate conformation pulls the naphthyl ring out of and adjacent to the hydrophobic pocket and increases its solvent-exposed surface area. As for the ex-perimentally observed orientation, the modeled conformation does not correspond to the binding mode of any arylsulfonamide complexed with CAII. The 180°relationship between the observed and modeled dansylamide conformations has significant implications for new leads in the design of CAII-targeted therapeutic agents as well as redesigned fluorophores for zinc biosensor applications. For example, it is possible that a 1,8di-substituted-10-anthracene sulfonamide would bind to CAII with optimal sulfonamide-zinc coordination stereochemistry, while its three fused aromatic rings would exploit favorable van der Waals interactions simultaneously within and adjacent to the hydrophobic pocket.
The transfer of fluorescence energy between CAII tryptophans and bound dansylamide is commonly used to avoid excitation of free dansylamide during binding assays (Chen and Kernohan, 1967). The ϳ75% quenching of tryptophan fluorescence in the presence of dansylamide (Chen and Kernohan, 1967) may be due to either partial quenching of all six tryptophans of CAII (Henderson et al., 1976) or, instead, complete quenching of a few of these residues. Of the six tryptophans in the CAII sequence, Trp-209 is the closest to bound dansylamide; the centroid-centroid separation of the two aromatic groups   the zinc ion appears as a gray sphere. Unlike typical arylsulfonamide inhibitors, the naphthyl ring system of dansylamide binds in the hydrophobic pocket due to a rotation about the sulfonamide N-S-C-C dihedral angle. Note that a conformational change of Leu-198 is required to accommodate dansylamide binding but the binding of typical arylsulfonamide inhibitors.
FIG. 5. Van der Waals surface representation of the CAII-dansylamide complex active site with superimposed atomic coordinates of the protein (white), the experimentally observed conformation of dansylamide (red), and the modeled conformation of dansylamide (yellow) with the naphthyl ring rotated about the carbon-sulfur bond by ϳ180°. The zinc is represented by a green sphere. This suggests that 1,8-di-substituted-10-anthracene sulfonamide derivatives may bind with high affinity to the CAII active site, and such compounds may be useful in the design of novel fluorescent probes for a CAII-based zinc biosensor. is 8.4 Å. However, complete quenching of solely Trp-209 would not account for the experimentally observed loss of fluorescence. In fact, the remaining five tryptophans are sufficiently close (9.0 -21.1 Å) that partial quenching of all tryptophans is most likely consistent with the observed quenching behavior.
It is clear that sulfonamide-bearing compounds represent the optimal fluorophores in sensitive and specific CAII-based zinc biosensors for three reasons. First, the sulfonamide NH 2 group requires deprotonation to bind tightly to zinc, and this is likely achieved by zinc-bound hydroxide of the native holoenzyme. The pK a of zinc-bound solvent and hence its reactivity are optimally tuned for this chemistry only in the zinc holoenzyme. Second, sulfonamide binding is most avid to the zinc holoenzyme (Pesando and Grollman, 1975), although some binding is also observed to the Co 2ϩ and Cd 2ϩ holoenzymes (Coleman, 1967;Pesando, 1975). This is easy to rationalize in light of the three-dimensional structure of the CAII-dansylamide complex; only in the zinc holoenzyme will the proper energetic balance be struck between optimal sulfonamide nitrogenzinc coordination separation and hydrogen bond interaction with Thr-199. If the metal ion were larger or smaller than zinc, interactions with Thr-199, as well as van der Waals interactions with other residues in the active site, would be compromised. Finally, as noted by Thompson and Jones (1993), the optimal feature of CAII-sulfonamide binding for biosensor applications is that sulfonamide binding generally requires the zinc holoenzyme. Dansylamide does not bind avidly to the apoenzyme due to loss of the favorable zinc interactions described above, and this is prerequisite for measuring a fluorescence signal proportional to bound zinc concentrations.
This work sets the foundation for the design of second-generation fluorescent probes in the continued development of CAII-based biosensors. For example, our modeling work suggests that substituents at the 4-position of dansylamide would interact with a hydrophobic patch in the active site cleft defined by the side chains of Val-121, Phe-131, and Ile-91 (data not shown). Binding interactions with these residues may enhance fluorophore affinity relative to that achieved with unmodified dansylamide (Thompson and Jones, 1993), and such properties may contribute to improved sensitivity. Additionally, the exploration of 1,8-di-substituted-10-anthracene sulfonamide-based fluorophores, as suggested by our modeling studies, may yield a probe molecule with fluorescence wavelength properties optimized for use in fiber optic arrays. Future studies will focus on the structure-assisted redesign of both the fluorophore and CAII Ippolito and Christianson, 1994;Kiefer and Fierke, 1994;Ippolito et al., 1995;Kiefer et al., 1995;Lesburg and Christianson, 1995;Nair et al., 1995) in the optimization of a CAII-based zinc biosensor.