Crystal structure of the Alpha subunit PAS domain from soluble guanylyl cyclase

Abstract

Soluble guanylate cyclase (sGC) is a heterodimeric heme protein of ∼150 kDa and the primary nitric oxide receptor. Binding of NO stimulates cyclase activity, leading to regulation of cardiovascular physiology and providing attractive opportunities for drug discovery. How sGC is stimulated and where candidate drugs bind remains unknown. The α and β sGC chains are each composed of Heme-Nitric Oxide Oxygen (H-NOX), Per-ARNT-Sim (PAS), coiled-coil and cyclase domains. Here, we present the crystal structure of the α₁ PAS domain to 1.8 Å resolution. The structure reveals the binding surfaces of importance to heterodimer function, particularly with respect to regulating NO binding to heme in the β₁ H-NOX domain. It also reveals a small internal cavity that may serve to bind ligands or participate in signal transduction.

Introduction

Nitric oxide (NO) is produced in most mammalian cells and serves to regulate blood pressure, wound healing, memory formation, and numerous other physiological processes.¹ The NO receptor is soluble guanylyl/guanylate cyclase (sGC), a large heterodimeric heme protein that is increasingly targeted for drug discovery in the treatment of cardiovascular disease.² Two classes of compounds targeting sGC are now in clinical trial, one that stimulates the heme-containing protein (BAY 63-2521/riociguat),^3,4 and another that functions to replace heme after loss due to oxidation (BAY 58-2667/cinaciguat and HMR1766/ataciguat).^5,6 How NO or drug binding leads to cyclase stimulation and signal transduction in sGC is poorly understood.

sGC is composed of two homologous subunits, α and β. Multiple isoforms of each subunit have been identified; however, the most common isoform is the α₁/β₁ heterodimer.⁷ Each sGC subunit consists of four domains, an N-terminal Heme-Nitric Oxide Oxygen (H-NOX) domain⁸ (also called a SONO domain),⁹ a central Per-ARNT-Sim (PAS) domain,¹⁰ a coiled-coil domain and a C-terminal catalytic cyclase domain.¹¹ NO binding to the heme in the β₁-subunit leads to the formation of a pentacoordinated Fe–NO complex, stimulation of cyclase activity and production cGMP from GTP. Structural insight into the allostery underlying stimulation is lacking. Structures of individual sGC domains such as the β₁ coiled-coil homodimer¹² and the α₁/β₁ heterodimeric cyclase domain¹³ have recently been determined, as have bacterial homologues of the H-NOX and PAS domains.^9,14–16 Yet an understanding of how these domains are arranged in the functional NO sensor remains unknown.

To fill this gap, we have developed sGC from Manduca sexta (tobacco hornworm) for biophysical and biochemical characterization.^17–20 Manduca sexta sGC (Ms sGC) is highly homologous to its mammalian counterparts and responds well to YC-1, the parent compound for riociguat. Using homology modeling, small angle X-ray scattering (SAXS) and chemical cross-linking, we previously determined that Ms sGC lacking the cyclase domains (Ms sGC-NT) is an elongated molecule with a central parallel coiled-coil.²⁰ In this model, the α₁ subunit PAS domain directly contacts the heme-containing β₁ subunit H-NOX domain²⁰ and inhibits NO and CO binding.¹ Here, we present the 1.8 Å crystal structure of the Ms sGC α₁ PAS domain, which reveals the H-NOX binding surface and a small internal cavity.

Results

Crystal structure of the α₁ PAS domain

Ms sGC α₁ PAS protein was obtained from an Escherichia coli expression vector as a SUMO-tagged fusion protein. SUMO cleavage and purification yielded 2–3 mg of highly pure α₁ PAS protein per liter of cell culture. Crystals of the wild-type α₁ PAS domain (residues 279–404) were initially small and could not be improved, possibly due to a requirement for cysteine modification by the arsenic in the cacodylate-containing crystallization buffer.^21–23 To overcome this, we made the triple cysteine mutant C285A, C352A, C374A. This protein crystallized under new conditions, yielding larger crystals with a rhombic dodecahedron morphology and diffraction to 1.8 Å resolution (Table I). Structure solution was by molecular replacement, using the Nostoc punctiforme signal transduction histidine kinase (Np STHK) PAS domain structure (PDB entry 2P04).¹⁵ Four nearly-identical copies of the α₁ PAS domain were present in the asymmetric unit and were generally well ordered except for the loop between beta strands 4 and 5 (residues 357–361; also called Gβ and Hβ, Fig. 1) and the C-termini. All four C-termini were disordered and not included in the final refined models. In the final model, chains A and C included residues 279–391, chain B included residues 279–390 and chain D included residues 279–395.

Table I.

Crystallographic Data

Data measurement
PDB entry	4GJ4
Wavelength (Å)	0.97950
Space group	H32
Unit cell parameters	a = b = 95.42 Å, c = 317.69 Å, α = β = 90°, γ = 120°
Resolution (Å)a	23.7–1.8 (1.86–1.80)
Total reflections	455433 (43841)
Unique reflections	52047 (5122)
Completeness (%)	100.0 (100.0)
Mean I/σI	11.3 (1.9)
Redundancy	8.75 (8.56)
Rmerge (%)	5.5 (69.1)
Refinement
Rwork (%)	19.6 (39.8)
Rfree (%)b	24.2 (42.1)
RMS deviation
Bond lengths (Å)	0.012
Bond angles (°)	1.55
No. of solvent molecules	176
Ramachandran plot
Most favored (%)	89.9
Allowed (%)	10.1

^aOverall (outermost shell).

^bFive percent of data not used in refinement.

Figure 1.

Ribbon diagram of Ms sGC α₁ PAS highlighting secondary structure elements. Helix Fα, which splits into a β strand and α helix as compared with canonical PAS domains, is shown as Fβ and Fα.

The Ms sGC α₁ PAS domain contains a typical PAS fold but one that is modified near the site where ligands often bind in PAS-containing proteins (Fig. 1).¹⁰ The core PAS fold consists of a five-stranded antiparallel beta-sheet with strands arranged in the sequence with order 2-1-5-4-3.¹⁰ The segment connecting strand 1 (also called Bβ, Fig. 1) to strand 5 (Gβ) is quite variable both in length and structure among PAS proteins and often provides a ligand-binding surface. Ligands commonly bind in a pocket formed between the beta 1–5 connecting strand and the interior face of the curved beta sheet. An N-terminal flanking helix is generally also present in PAS-containing proteins.

In α₁ PAS, the beta 1–5 connecting strand displays a unique structure as compared with other PAS domains. In most PAS proteins, this segment includes four helices, generally referred to as Cα, Dα, Eα, and Fα. All four helices are present in Ms sGC α₁ PAS; however, the residues that form the first half of Fα in a typical PAS domain are seen to form a new beta strand in α₁ PAS (referred to as Fβ in Fig. 1), yielding an overall 6-stranded beta sheet with topology 2-1-6-5-4-3. The remaining portion of Fα is in a different orientation than typically observed in PAS domains and makes a critical contact with the β₁ H-NOX domain, based on cross-linking studies.²⁰ The helix includes residues Glu 340 and Lys 343, each of which can be cross-linked to β₁ H-NOX residue Lys 170,²⁰ indicating that the α₁ PAS and β₁ H-NOX domains are in direct contact in sGC.

Of additional interest is an internal cavity found directly behind the Fα helix in the α₁ PAS structure (Fig. 2). This cavity is in a similar position to the ligand-binding site in other PAS domains, overlapping, for example, with the positions for heme in FixL^24,25 and flavin in the FMN containing LOV domains.^26,27 The cavity size is ∼36 Å,³ about two-thirds the size of a benzene ring. While it is tempting to suggest this cavity represents a ligand or protein binding site, a substantial rearrangement in the Fα helix would be required to accommodate anything larger than 3–4 atoms. Nonetheless, large rearrangements are known to occur in other PAS domains leading to ligand binding (see, for example, references28,29) and the cavity found in Ms sGC α₁ PAS may yet have a ligand-binding function. YC-1 appears, however, not to bind to α₁ PAS, but rather to bind to the β₁ H-NOX domain (see footnote *).

Figure 2.

Ribbon drawing of Ms sGC α₁ emphasizing the small internal pocket and inter-domain contact residues (cross-eyed stereo view). The small internal pocket found in the structure is highlighted in purple and the C285A, C352A, C374A mutations are shown in yellow. Also shown are residues Glu 340 and Lys 343, which can be cross-linked to the β₁ H-NOX domain; residue Lys 286, which can be cross-linked to the β₁ PAS domain; and residue Glu 366, which can be cross-linked to the β₁ coiled-coil.

The four copies of α₁ PAS in the asymmetric unit are quite similar, displaying similar internal cavity volumes and pairwise RMS deviations in Cα positions of 0.4–0.6 Å. Superpositioning of Ms sGC α₁ PAS with Np STHK, which was used for molecular replacement, leads to an RMSD of 1.4 Å for 93 core residues (29% identity) when aligned using secondary structure matching (SSM).³⁰ Superimposing Ms sGC α₁ PAS with heme-containing FixL (PDB entry 1EW0)²⁵ and FAD-containing PAS1 of NIFL (PDB entry 2GJ3)²⁶ reveal RMSD values of 2.7 Å for 87 core residues (10% identity) and 2.7 Å for 86 core residues (11% identity), respectively. The key difference between the FixL and NIFL PAS1 structures is the position of the Fα helix.

Discussion

The α₁ PAS crystal structure provides constraints for understanding domain arrangement in sGC. The overall fold is typical for PAS domains but displays a unique arrangement for the most variable region in the family, the segment connecting beta sheet strands 1 and 5. In this segment, the Fα helix is split into a sixth beta strand and a shorter helix with a new orientation with respect to most other PAS domains, but similar to that of Np STHK.¹⁵

In Ms sGC-NT, a parallel coiled-coil provides a platform on which the other domains assemble. Direct cross-links between the α₁ PAS Fα helix and the β₁ H-NOX domain near the heme pocket suggest the two domains are in direct contact, providing a means for allosteric regulation of the protein. Both domains also cross-link with the coiled-coil domain. Our working hypothesis is that YC-1 binding disrupts the α₁ PAS/β₁ H-NOX interaction, leading to a closed H-NOX domain and tighter CO and NO binding.

PAS domains often form homo- or hetero-oligomers as part of their function¹⁰ and the possibility that sGC forms an α₁/β₁ PAS dimer has been previously proposed. The most compelling data are based on the oligomer formed by Np STHK, which shares sequence homology with the sGC PAS domains.¹⁵ Np STHK forms a homodimer involving a hydrophobic patch near the N-terminus and a strand swap that allows Leu 8 from one chain to cover the hydrophobic patch of the other chain in the dimer. Several hydrogen bonds help stabilize the dimer while removal of the first seven residues in the protein abolishes the dimer. Intriguingly, the rat β₁ PAS domain also dimerizes. In contrast, our construct for Ms α₁ PAS runs as a monomer over a sizing column and appears to be monomeric in solution. The protein crystallizes as a rhombic dodecahedron, but this arrangement is likely an artifact of crystallization and of no physiological significance. The N-terminal hydrophobic patch at the heart of the Np STHK dimer interface is also found in Ms α₁ PAS; however, it does not lead to a dimer interface. Our structure is six residues shorter at the N-terminus than that for the rat β₁ PAS construct and, conceivably, this could alter dimer formation much as it did in Np STHK. Nonetheless, our cross-linking data for Ms sGC-NT include a link between α₁ Lys 286 and β₁ Glu 196, located at the N-termini of the two PAS domains, indicating the two PAS domains are in contact in the intact heterodimer.²⁰

The Ms sGC α₁ PAS structure also displays a small internal cavity behind the Fα helix (Fig. 2). Our binding data indicate YC-1 does not bind to this domain but do not rule out another role for this pocket in ligand binding. For ligand binding to occur, the beta 1–5 connecting segment, which includes the Fα helix, would need to rearrange, allowing the pocket to open up. There is precedent for such rearrangements in PAS domain proteins. For example, human PAS kinase has a dynamic Fα helix in the PAS A domain that allows for small molecule entry into the hydrophobic protein core, near to where the internal cavity lies in sGC α₁ PAS.²⁸ Importantly, the ligand-binding pocket is collapsed in the apo protein and only forms upon ligand binding. A second example is that of histidine kinase CitA, which uses a PAS domain for sensing citrate.²⁹ In CitA, the beta 1–5 connecting segment is poorly ordered in the absence of citrate binding, but becomes well ordered in the complex with citrate bound in the protein interior. The loop connecting beta strands 4 and 5 also rearranges upon citrate binding, shifting inward. In Ms sGC α₁ PAS, the beta 4–5 loop is poorly ordered and could serve a similar role in ligand binding or in signal transduction.

Materials and Methods

Ms sGC α₁ PAS expression and purification

All chemicals were obtained from Sigma-Aldrich, restriction enzymes from New England Biolabs, and purification columns from GE Healthcare unless otherwise indicated. Ms sGC α₁ PAS with a N-terminal His-tagged SUMO fusion was cloned into the pETHSUL vector,³¹ expressed in E. coli strain BL21 (DE3) pLysS, and purified after cleavage of the SUMO tag with SUMO hydrolase, as described elsewhere (see footnote *). The final material was concentrated to 10–15 mg/mL using a Vivaspin concentrator (Sartorius Stedim Biotech) and stored at –80°C. A final yield 2–3 mg of highly pure protein was obtained per liter of cell culture.

Crystallization

Initial crystallization conditions for Ms sGC α₁ PAS were found using a PHOENIX protein crystallization robot (Art Robbins Instruments) and commercially available screens (Hampton Research and Qiagen). Crystals formed in a 96-well Intelli-Plate using sitting drop vapor diffusion at 4°C and precipitants of 1.4–1.6 M ammonium sulfate, 50 mM sodium cacodylate (pH 5.5–6.5) and 15 mM magnesium acetate tetrahydrate. Protein at 10–15 mg/mL was mixed with precipitant at ratios of 1:1 and 1:2. Cubic crystals appeared within 24–48 h after plate setup but failed to grow beyond 100 μm in size. Diffraction quality hexagonal crystals for Ms sGC P35α (cysteine triple mutant) were obtained by hanging drop vapor diffusion at 4°C using a precipitant solution of 1.5 M lithium sulfate, 0.1 M Hepes (pH 7.5). Small crystals were also observed from 4.3 M NaCl, 0.1 M Hepes (pH 7.5) and from 25% PEG 3350, 0.2 M NaCl, 0.1 M Hepes (pH 7.5). Total of 90% saturated lithium sulfate was used as the cryoprotectant and crystals were flash frozen in liquid nitrogen.

Data collection, structure solution, and refinement

X-ray diffraction data for Ms sGC α₁ PAS (wild type) cubic crystals were measured remotely on SSRL beamline 9–2 (Stanford) using a MAR325 detector at T = 100 K and λ = 0.97950 Å. The data were processed in space group P2₁3 to 3.7 Å resolution using CrystalClear.³² The unit cell parameters were a = b = c = 143.26 Å, and α = β = γ = 90°.

Diffraction data for hexagonal crystals of Ms sGC α₁ PAS (triple mutant) were also measured remotely on SSRL beamline 7-1 (Stanford) using a MAR325 detector at T = 100 K, λ = 0.97950 Å and were processed to 1.8 Å with CrystalClear in hexagonal space group H32 (Table I). There were four molecules in the asymmetric unit. The structure was determined using molecular replacement as implemented in MrBUMP^33,34 and search models generated from the structure of the Nostoc punctiforme signal transduction histidine kinase HNOXA domain (PDB entries 2P04 and 2P08),¹⁵ which yielded an ensemble model. Model building and refinement were performed using programs COOT and REFMAC5.^35,36 Figures were prepared using PyMOL (W. L. DeLano, http://www.pymol.org) and UCSF Chimera.³⁷ Model quality was evaluated with PROCHECK.³⁸ Cavity volume was computed using CASTp.³⁹

Atomic Coordinates

The atomic coordinates and structure factors have been deposited with the Protein Data Bank (PDB entry 4GJ4).

Glossary

H-NOX domain

heme-nitric oxide/oxygen binding domain

PAS domain

Per-ARNT-Sim domain

SAXS

small angle X-ray scattering

sGC

soluble guanylyl cyclase

Ms sGC

Manduca sexta sGC

Ms sGC-NT

Manduca sexta sGC lacking the catalytic domains.