Structure and monomer/dimer equilibrium for the guanylyl cyclase domain of the optogenetics protein RhoGC

RhoGC is a fusion protein from the aquatic fungus Blastocladiella emersonii, combining a type I rhodopsin domain with a guanylyl cyclase domain. It has generated excitement as an optogenetics tool for the manipulation of cyclic nucleotide signaling pathways. To investigate the regulation of the cyclase activity, we isolated the guanylyl cyclase domain from Escherichia coli with (GCwCCRho) and without (GCRho) the coiled-coil linker. Both constructs were constitutively active but were monomeric as determined by size-exclusion chromatography and analytical ultracentrifugation, whereas other class III nucleotidyl cyclases are functional dimers. We also observed that crystals of GCRho have only a monomer in an asymmetric unit. Dimers formed when crystals were grown in the presence of the non-cyclizable substrate analog 2′,3′-dideoxyguanosine-5′-triphosphate, MnCl2, and tartrate, but their quaternary structure did not conform to the canonical pairing expected for class III enzymes. Moreover, the structure contained a disulfide bond formed with an active-site Cys residue required for activity. We consider it unlikely that the disulfide would form under intracellular reducing conditions, raising the possibility that this unusual dimer might have a biologically relevant role in the regulation of full-length RhoGC. Although we did not observe it with direct methods, a functional dimer was identified as the active state by following the dependence of activity on total enzyme concentration. The low affinity observed for GCRho monomers is unusual for this enzyme class and suggests that dimer formation may contribute to light activation of the full-length protein.

The rhodopsin-guanylyl cyclase fusion protein RhoGC 3 was first identified in the phototactic fungus Blastocladiella emer-sonii by Avelar et al. (1). It is a unique protein composed of a type I (microbial) rhodopsin (2,3) domain fused to a guanylyl cyclase catalytic domain showing homology to the rod outer segment guanylyl cyclase from vertebrate retina. It was the first retinylidene protein described that is directly coupled to an enzyme. The protein is localized to the B. emersonii eyespot and is essential for zoospore phototaxis, a pathway known to signal through the second messenger cGMP. RhoGC contains four domains ( Fig. 1): an intracellular N-terminal domain of unknown function (N-term); a type I rhodopsin domain (Rho); a CC domain; and a guanylyl cyclase domain (GC).
RhoGC is of considerable interest as an optogenetic tool for control of cyclic nucleotide-signaling pathways (4,5), and it has now been joined by another type I rhodopsin fusion protein, RhoPDE (6,7), with potential as an optogenetic tool. The gene for RhoPDE was found in the genome of the choanoflagellate Salpingoeca rosetta and encodes a fusion of the rhodopsin domain with a cGMP-specific phosphodiesterase catalytic domain.
We have developed an expression/purification system and performed preliminary characterization for RhoGC in preparation of studies to elucidate the mechanism by which light controls guanylyl cyclase activity in the enzyme (8). We now turn our attention to the individual domains to more fully identify their functional contribution to the full-length RhoGC. We begin here with the guanylyl cyclase domain, GC Rho .
Crystal structures of GC catalytic domains have been determined for bacterial (9), algal (10), and human (11) proteins (see Fig. S1 for sequence alignment). GC domains from the cyanobacterium Synechocystis PCC6803 (GC Cya2 ) (9) and the unicellular algae Chlamydomonas reinhardtii (GC CYG12 ) (10) were crystallized as homodimers, whereas the GC domain (GC Hum ) from the human soluble guanylyl cyclase, sGC (11), was crystallized as both homodimer and heterodimer, although the enzyme is active only in the heterodimeric form. The active sites form at the dimer interface where catalytic residues are contributed by both monomers. Hence, dimerization is necessary for catalysis in this class of enzymes (12,13). All three GC structures solved so far are the substrate-free form and are reported to be an inactive "open" conformation (11). Comparing the open GC structures with the active "closed" conformation of the adenylyl cyclase (AC) catalytic domain (14) revealed that the protomers in the inactive GC dimers are rotated by 7-8°for GC Cya2 and GC CYG12 and by 26°for GC Hum relative to the monomers in the active AC structure.
We have expressed the catalytic domain of RhoGC (GC Rho ) in Escherichia coli, purified the protein by nickel-affinity chromatography, and show here that the isolated protein is constitutively active in catalyzing the formation of cGMP. It is a monomer in solution and crystallizes with a monomer in the asymmetric unit, but it has a non-linear dependence of activity on enzyme concentration that is consistent with transient formation of an active-state dimer. Crystals of a dimeric form of GC Rho were obtained from crystallization trials in the presence of the non-cyclizable substrate analog 2Ј,3Ј-dideoxyguanosine-5Ј-triphosphate (ddGTP), Mn 2ϩ , and added potassium sodium tartrate, but the two protomers are arranged in a head-to-head conformation distinctly different from the canonical head-totail wreath-like structures first described for adenylyl cyclases (13,15) and then later for the guanylyl cyclases (9 -11). We anticipate that both monomer and dimer structures reported here for GC Rho will help guide mutagenesis experiments focused on the mechanism by which light controls activity of the cyclase domain in full-length RhoGC.

Expression and purification of GC Rho and GCwCC Rho
Conventional PCR was used to construct genes for the isolated guanylyl cyclase domain, with (GCwCC Rho , residues 397-627) and without (GC Rho , residues 443-627) the connecting linker (CC, residues 397-442), from full-length RhoGC (Fig. 1). Both constructs contained a His 6 tag on the C terminus for purification of the protein by nickel-affinity chromatography. Initial tests in E. coli showed that the proteins were expressed to a high level in T7-express cells under standard induction conditions (1 mM IPTG at A 600 0.4 -0.7 followed by incubation for 3 h at 37°C), but neither protein was found in the soluble fraction after cell lysis and centrifugation. Subsequent attempts varying induction conditions, growth and expression temperature, and a change of cell lines met with limited success. In the end, coexpression with the molecular chaperone proteins GroEL and GroES (16) dramatically increased the amount of both GCwCC Rho and GC Rho in the soluble fraction such that we could purify ( Fig. 2A) on average about 50 mg of protein (determined spectrophotometrically using an extinction coefficient ⑀ 280 nm ϭ 31,065 M Ϫ1 cm Ϫ1 ) per liter of cell culture, a quantity sufficient to undertake biochemical and structural characterization of the proteins.
GCwCC Rho and GC Rho both displayed constitutive guanylyl cyclase activity that varied depending on whether Mn 2ϩ or Mg 2ϩ was used for the metal ion. As is shown in Fig. 2B, GC Rho displayed much higher activity in the presence of Mn 2ϩ than in Mg 2ϩ , consistent with observations from other isolated guanylyl (9 -11, 16) and adenylyl cyclase (12) domains. However, this is not the case when the CC linker was included. GCwCC Rho displays about the same activity in Mn 2ϩ as in Mg 2ϩ , and both are significantly less active than the GC domain alone in Mn 2ϩ . These data seem to suggest that the role of CC might be to attenuate the constitutive activity of the GC Rho domain, as has been suggested for other systems, but this is clearly not the case if GCwCC Rho is compared with GC in the presence of Mg 2ϩ , where GCwCC Rho displays the higher activity. At 10 M enzyme concentration, the activities were as follows: GC Rho with Mn 2ϩ ϭ 21 Ϯ 1 M/s; GCwCC Rho with Mn 2ϩ ϭ 1.0 Ϯ 0.1 M/s; GCwCC Rho with Mg 2ϩ ϭ 0.71 Ϯ 0.06 M/s; and GC Rho with Mg 2ϩ ϭ 0.064 Ϯ 0.002 M/s.
The dependence of activity on GTP concentration for GC Rho in the presence of MnCl 2 is presented in the plot of Fig. S2, where the data are fit with a Hill-type equation using the parameters S 0.5 ϭ 0.9 Ϯ 0.1 mM, V max ϭ 16 Ϯ 1 M/s, and n ϭ 1.2 Ϯ 0.3.
All known class III cyclases function as dimers (12,17,18). They can be either hetero-or homodimers, with active sites formed at the dimeric interface. Given the robust constitutive activity of both GCwCC Rho and GC Rho , we anticipated that the proteins were assembled as homodimers in solution. Contrary to expectations, both proteins ran as monomers by size-exclusion chromatography (SEC) on a Superdex-200 column (Fig.  2C), with molecular mass estimates for the peak elution fractions from standards within 5% error of the calculated molecular mass of monomeric GC Rho (21.5 kDa) and GCwCC Rho (26.5 kDa).

Structure of GC
Initial crystallization trials were conducted with GCwCC Rho . After 3 months, a small rectangular-shaped crystal appeared in a PEG3350 solution. The structure of the enzyme was determined to 1.6 Å resolution by molecular replacement using human soluble guanylyl cyclase (PDB entry 2WZ1) as a search model. The protein had crystallized in space group P22 1 2 1 with one molecule in the asymmetric unit. Although the full-length GCwCC Rho construct was purified and used in the crystalliza- Transmembrane helices were predicted and drawn by Protter (wlab.ethz.ch/ protter/start (Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site)) (34). Cyan, N-term; green, Rho; purple, CC; yellow, GC; blue square, Lys-384, presumed site of covalent attachment to the chromophore; and red circles, Glu-497 and Cys-566, residues controlling specificity for the GTP substrate.

Structure and monomer/dimer equilibrium of guanylyl cyclase
tion trials, electron density was not observed for the CC linker, likely as a result of proteolysis or disorder in the crystal. Repeated trials in the presence of sodium azide to prevent bacterial growth did not yield crystals, and further attempts to optimize crystallization conditions were performed with GC Rho alone. Crystals of GC Rho formed in the same crystallization conditions after about 3 weeks, still in space group P22 1 2 1 with one molecule in the asymmetric unit, and the structure was determined to 1.13 Å resolution and refined to an R work and R free of 0.188 and 0.195, respectively. Importantly, crystal packing did not reveal dimer formation with symmetry-related molecules.
The overall structure of the GC Rho catalytic domain has five ␣-helices and eight ␤-strands (Fig. 3), as has been reported pre-viously for other GC (9 -11) and AC structures (13,15). Clear electron density was observed for residues Thr-443 through Lys-626, with the exception of two missing loop regions not modeled in the final structure: Gly-559 -Asn-562 in the ␤4 -␤5 loop, which has a major role in dimer formation, and Val-608 -Lys-616 in the ␤7-␤8 loop, which carries one of the eight conserved catalytic residues, Lys-612.

Comparison with other guanylyl cyclases
Synechocystis guanylyl cyclase Cya2. The catalytic domain (GC Cya2 ) of the Cya2 guanylyl cyclase from the cyanobacterium Synechocystis PCC6803 crystallized as a homodimer, and its structure was determined at 2.3 Å resolution (9). A superpose of

Structure and monomer/dimer equilibrium of guanylyl cyclase
the GC Rho monomer and molecule A of GC Cya2 shows that the two proteins are very similar in overall fold with r.m.s.d. for C␣ atoms of 0.9 Å. The ␣3-␤4 loop in GC Rho is four residues shorter than the corresponding loop in GC Cya2 , and the ␣3 helix is two turns longer in GC Cya2 . This structural feature appears to be unique to GC Cya2 because the respective regions in Chlamydomonas (10) and human sGCs (11) were similar to that of the GC Rho monomer. The GC Rho ␤4 -␤5 and ␤7-␤8 loops are each one residue shorter than in GC Cya2 , whereas GC Cya2 has an extended C terminus.
C. reinhardtii guanylyl cyclase CYG12-The catalytic domain (GC CYG12 ) of the soluble guanylate cyclase from the green algae C. reinhardtii crystallized as a homodimer, and its structure was solved at 2.6 Å resolution (10). A superpose of the GC Rho monomer and molecule A of GC CYG12 shows that the two proteins are very similar in overall fold with r.m.s.d. for C␣ atoms of 1.2 Å. The ␣2-␤2 and ␣5-␤7 loops in GC Rho are one and two residues shorter, respectively, than the corresponding regions in GC CYG12 . The C. reinhardtii protein also has an additional short ␣6 helix, which is not observed in other GC structures.
Human soluble guanylyl cyclase-The catalytic domain (GC Hum ) of the human soluble guanylate cyclase crystallized as a heterodimer, and its structure was solved at 1.6 Å resolution (11). A superpose of the GC Rho monomer with the ␣-subunit of the human enzyme shows that the two proteins are very similar in overall fold with r.m.s.d. for C␣ atoms of 1.2 Å but that they differ in conformation of the ␣3-␤4 loop. In addition, the ␣1-␣2, ␣2-␤2, and ␣6 -␤7 loops in GC Rho are five, three, and eight residues shorter than respective regions in GC Hum .

Dimeric GC Rho
Given that the active state of GC Rho was expected to be a dimeric enzyme, we explored other crystallization conditions that might trap a dimer in the asymmetric unit. Because the enzyme displayed significant guanylyl cyclase activity when assayed in solution (Fig. 2B), we reasoned that substrate and metal ion might facilitate formation of the dimer. We used the substrate analog 2Ј,3Ј-dideoxyguanosine 5Ј-triphosphate (ddGTP), which lacks the ability to be cyclized due to the absence of a hydroxyl group at the 3Ј-position of the sugar ring. GC Rho was mixed with ddGTP and Mn 2ϩ and then monitored for formation of dimer by analytical ultracentrifugation (AUC). Sedimentation velocity profiles showed the existence of two separate sedimenting species consistent with the molecular mass expected for monomeric and dimeric GC Rho (Fig. 4A). Notably, GC Rho sediments exclusively as monomer in the absence of ddGTP and Mn 2ϩ . When a similar sample was applied to the Superdex-200 column, three major species were observed, corresponding to dimer, monomer, and free nucleotide (Fig. 4B). The percentage of dimer in the SEC experiment was less than that observed in AUC likely due to slight differences in sample preparation.

Crystal structure of the dimer
GC Rho crystallized in space group C222 1 from PEG3350 in the presence of 1 mM ddGTP, 10 mM MnCl 2 , and 200 mM potassium sodium tartrate additive. The structure was solved to 1.7 Å resolution by molecular replacement using the GC Rho monomer ( Fig. 3) as search model and yielded two molecules in the asymmetric unit (Fig. 5A). The final model was refined to R work and R free of 0.164 and 0.196, respectively. Clear electron density was observed for residues Met-442 to Met-627 in molecule A and Ala-445 to Lys-626 in molecule B. The ␤7-␤8 loop region residues Glu-610 -Gly-615 in molecule A and Lys-612-Lys-614 in molecule B were highly disordered and not modeled in the final structure.
Superposition of the GC Rho monomer with both monomers of the dimer showed that there are no significant structural

Structure and monomer/dimer equilibrium of guanylyl cyclase
changes in the monomer units, with r.m.s.d. for C␣ atoms of 0.4 and 0.3 Å for monomer A and monomer B, respectively, although minor changes were observed in part of the dimer interface around the ␣1 helix (Asn-460 -Ser-469) and ␤4 -␤5 loop (Leu-558 -His-564). Nonetheless, the monomer units do not form the signature head-to-tail wreath-like conformation observed in other GC (9 -11) and AC (13-15) dimeric structures. Instead, the two monomers are in a head-to-head arrangement. Superposition of the GC Rho dimer with the active conformation of the mammalian AC heterodimer (PDB entry 1CJU (14)), using monomer A as a reference, shows that monomer B of GC Rho is rotated by 101°and displaced by 6.3 Å with respect to AC monomer B (Fig. 5B). The large rotation of monomer B compared with the AC structure suggests that the GC Rho dimer is non-functional.
The GC Rho dimer has a buried surface area of 1273 Å 2 and geometric surface complementarity (S c ) (19) of 0.63. The buried surface area is comparable in magnitude to the buried surface area for the C. reinhardtii cyclase GC CYG12 (1370 Å 2 ; 3ET6), for the cyclase GC Cya2 from Synechocystis (1548.9 Å 2 ; 2W01), and for the human sGC heterodimer GC Hum (1230 Å 2 ; 3UVJ). The shape complementarity statistic, S c , is also similar in magnitude to those of other adenylyl and guanylyl cyclases (AC, 0.59; GC CYG12 , 0.64; GC Cya2 , 0.67; GC Hum , and 0.74).
Part of the dimer interface is formed by an unusual interaction of ␤-strands ␤4b and ␤5 of monomer A with those of monomer B. In other GC dimeric structures (9 -11), this ␤-sheet interacts with strands ␤2 and ␤3 to form the dimer interface. Significantly, monomers A and B of GC Rho are connected by a disulfide bridge involving the catalytically important residue Cys-566 (8,20,21) in ␤5 from both monomers (Fig. 5C), reinforcing the conclusion that this is an inactive form of the protein.
After initial refinement, difference Fourier density appeared at the cavity formed by the catalytically important residues Asp-457, Asp-501, and Arg-545 (12)(13)(14) from both monomers A and B (Fig. 5D). The density was further resolved into two 22 F o Ϫ F c peaks and an overlapping, elongated ϳ8 F o Ϫ F c peak. Both 22 peaks were modeled with a Mn 2ϩ ion. The 8 peak electron density was not fit well by ddGTP, cGMP, or PP i but was modeled well by the tartrate additive. Mn 2ϩ A is hexacoordinate with Asp-457 O␦2 of molecule B, Asp-501 O␦1 of molecule A, O2 and O6 atoms of the tartrate molecule, and two water molecules as ligands. Mn 2ϩ B is also hexa-coordinate with Asp-457 O␦1 and Asp-501 O␦1 and O␦2 of molecule B, O1 and O3 atom of the tartrate molecule, and a single water molecule as ligands. Tartrate is perfectly sandwiched between the side chains of the catalytically important residue R545 of both monomer A and monomer B. Apart from metal ion coordination, tartrate also forms hydrogen bonds with amino acids D457, T462, and D501 of molecule A, D457, I458, T462, and D501 of molecule of B, and with several water molecules.

Dependence on Mn 2؉
Discovery of the disulfide cross-link in the GC Rho dimer caused us to revisit the AUC experiments where we could show that no dimer species was formed in the presence of a reducing agent (5 mM DTT). Likewise, subsequent crystal trials with ddGTP and Mn 2ϩ in the presence of DTT produced only crystals with monomer in the asymmetric unit and cell dimensions and space group identical to those observed in the absence of nucleotide. Therefore, we conducted a series of experiments to determine what factors were important for formation of the disulfide cross-linked GC Rho dimer as assayed by SDS-PAGE. As is shown in Fig. 6, the single factor most important was the

Structure and monomer/dimer equilibrium of guanylyl cyclase
presence of Mn 2ϩ . The presence of GTP, cGMP, and inorganic pyrophosphate in the absence of metal ion were without effect. Mg 2ϩ could substitute for Mn 2ϩ in this reaction but with reduced efficiency. Notably, addition of the CC domain dramatically decreased the amount of dimer formed under all conditions.
It should be noted that disulfide bond formation required a lengthy incubation period. The reactions shown in Fig. 6 included a 20-h incubation to mimic the original AUC and SEC conditions (Fig. 4). No cross-linked dimer band was observed when the incubation period was omitted.

Structure of the GC AC domain
Previously, we constructed a mutant of RhoGC in which the activity of the enzymatic domain was changed from a guanylyl cyclase to an adenylyl cyclase (8). This required the introduction of two mutations in the GC domain of the protein, E497K and C566D. Given that the reactive Cys-566 is replaced by Asp, it was of interest to determine a crystal structure for the isolated "adenylyl cyclase" domain, GC AC , of the mutant. GC AC displays the same substrate specificity (Fig. 7A) as was observed for the full-length mutant (8) and was used for crystal trials under a host of different conditions. Although GC AC did not form crystals under the same conditions as for the GC Rho dimer, it did crystallize from PEG20000 in the presence of 10 mM potassium hydrogen tartrate with the same space group and unit cell dimensions as found with the GC Rho monomer. The structure was solved by molecular replacement using the GC Rho monomer as a search model. The final structure was determined to a resolution of 1.4 Å with R work /R free of 0.190/0.208. A single molecule was in the asymmetric unit with overall structure identical to that of the GC Rho monomer. The r.m.s.d. for C␣ atoms from superposition of the two structures is only 0.1 Å suggesting that the mutations had minimal effect on secondary and tertiary structure of the protein. Amino acids Gly-559 and Asp-560 in the ␤4 -␤5 loop and Glu-610 -Gly-615 in the ␤7-␤8 loop were highly disordered and not modeled in the final structure. The calculated omit map clearly revealed the two mutations in the protein (Fig. 7B), and Lys-497 and Asp-566 were modeled in the structure where their side chains were found to form a salt bridge.

Indirect evidence for functional dimerization
Indirect evidence for formation of an active dimer was obtained by following the dependence of enzyme activity on the concentration of protein in the assays. As shown in Fig. 8, guanylyl cyclase activity showed a distinctly nonlinear dependence on total enzyme concentration that was well fit by Equation 1 under "Experimental procedures" for a monomer/dimer equilibrium in which only the dimer is active. The parameters used to fit Equation 1 to the data are presented in Table 1. The inclusion of DTT in the reaction had no effect on activity demonstrating that the disulfide cross-linked dimer observed in the crystal structure does not form during the short incubation times used here in the activity assays. It is evident from the data in Fig. 8 and Table 1 that the effect of the CC domain on quaternary structure of the protein is a rather modest 3-4-fold decrease in K D for formation of the dimer.
To eliminate concern that the C-terminal His 6 tag might interfere with GC Rho dimerization, a new construct was generated with an N-terminal His 6 tag and TEV-proteolysis site inserted between the His 6 tag and the remainder of the GC Rho catalytic domain, and the dependence of activity on enzyme concentration repeated with the TEV-proteolyzed enzyme. The data were again well fit by Equation 1 with an obviously non-linear dependence of activity on enzyme concentration giving a K D value of 64 Ϯ 9 M, well within experimental error of the K D values for GC Rho reported in Table 1.

Active-state dimer
Because we could not crystallize an active-state dimer of GC Rho , we modeled what the dimer might look like based upon superposition with the published structure of the mammalian adenylyl cyclase VC 1 IIC 2 catalytic domain in complex with Mn 2ϩ and ddATP (Fig. S4) (14). The active-site residues at the dimer interface align nicely between the two structures, although side chain conformations show some differences, as is expected given that GC Rho crystallized as a monomer.

Discussion
The purpose of this study was to provide an initial purification and characterization of the isolated guanylyl cyclase domain of RhoGC, with (GCwCC Rho ) and without (GC Rho ) the CC domain, in preparation for a more in-depth mechanistic exploration of the full-length protein. Genes for His 6 -tagged versions of both proteins were co-expressed in E. coli with GroEL and GroES (16) to improve the yield of properly folded protein in the soluble fraction of cellular extracts. The proteins were purified by nickel-affinity chromatography in a yield of about 50 mg/liter cell culture.
Both proteins displayed robust guanylyl cyclase activity in a reaction that was dependent on added metal ion, Mg 2ϩ or Mn 2ϩ . As has been observed with other cyclases, GC Rho was significantly more active with Mn 2ϩ than with Mg 2ϩ (9 -12, 16). Interestingly, including the CC domain in the GCwCC Rho construct eliminated this difference in activity with the two metal ions. GCwCC Rho displays a k cat (2.5 s Ϫ1 ) that is about 4-fold less than the GC Rho domain with Mn 2ϩ (10 s Ϫ1 ) but significantly more than GC Rho with Mg 2ϩ (0.16 s Ϫ1 ) and, perhaps most importantly, about the same as the apparent k cat for light-stimulated guanylyl cyclase activity for the full-length RhoGC (2.2 s Ϫ1 ) (8). The latter comparison suggests that a function of the remaining domains in RhoGC (i.e. N-term and Rho) is to suppress constitutive activity of GCwCC Rho , such that the full-length protein expresses guanylyl cyclase activity only after activation with light. This appears to be the case also for ligand-activated guanylyl cyclases, where regulatory domains suppress constitutive activity of the catalytic domains (9 -11, 16).
Although there were clear similarities of GC Rho and GCwCC Rho in comparison with other guanylyl cyclases, a striking difference was encountered in crystal structures of the GC Rho domain where the protein crystallized under standard conditions with only a single subunit in the asymmetric unit. This was unexpected because both GCwCC Rho and GC Rho display significant guanylyl cyclase activity. All other guanylyl cyclases crystallize as dimers with two monomers in the asymmetric unit (9 -11), and the active sites of adenylyl and guanylyl cyclases are known to be formed at the interface of the two protomers in a dimeric complex (17,18). The unusual crystallization results for GC Rho were supported by SEC and AUC experiments showing that the protein was 100% monomer in solution. To be sure, monomers of other guanylyl cyclases have been observed in solution studies (9,16), but the monomer/ dimer equilibria are more heavily shifted toward the dimer, and, as noted above, the other guanylyl cyclases all crystallized with a dimer in the asymmetric unit. There are two reports in which adenylyl cyclases crystallized as a monomer: one for a Phe to Arg mutant involving a residue in the dimer interface of the Mycobacterium tuberculosis guanylyl cyclase Rv1625c catalytic domain (22), and the other for the catalytic domains of the adenylyl cyclases, GRESAG4.1 and GRESAG4.3, from Trypanosoma brucei (23). Rv1625c forms an inactive dimer in solution (22) whereas the T. brucei enzymes display weak subunit affinity and transient formation of catalytically active dimers in solution (23).
We were able to crystallize a GC Rho dimer in the presence of ddGTP, Mn 2ϩ , and added potassium sodium tartrate. The two subunits of the dimer are nearly identical in overall fold to the isolated GC Rho monomer and to the individual subunits found in dimer structures of other guanylyl cyclases, but the arrangement of the two subunits is very different from the canonical wreathlike structure first described for adenylyl cyclase (13)(14)(15) and later found in the guanylyl cyclases (9 -11). The subunits of  Table 1. The enzyme concentration is total enzyme added to the reaction, expressed in terms of the micromolar concentration of monomer. An expanded view of the data for GCwCC Rho -Mn is shown in Fig. S3A. A control experiment showing that the non-linear behavior is not a result of non-specific protein interactions at low concentrations is shown in Fig. S3B.

Structure and monomer/dimer equilibrium of guanylyl cyclase
GC Rho dimer are displaced by 6 Å and rotated by 101°with respect to the arrangement of subunits in the adenylyl cyclase dimer (14) such that the conserved active-site residues are dispersed in the protein and are unlikely to support binding of nucleotide. In addition, the GC Rho dimer contains a disulfide bond involving Cys-566, an active-site residue known to be critical for activity of the protein (8). Thus, it is clear that this dimer cannot be part of the catalytic cycle of the enzyme. Unusual dimer structures have been described for both the Rv1264 (24) and Rv1625 (22) adenylyl cyclases from M. tuberculosis, and in this context the domain-swapped head-to-head dimer observed for WT Rv1625 (22) shares some similarity with the head-to-head dimer we see here for GC Rho . It is unlikely that the disulfide in GC Rho would form under intracellular reducing conditions, and it seems most probable that the disulfide forms subsequent to dimerization as a consequence of a high local concentration of the two Cys residues in an oxidizing environment. Thus, the disulfide could function as a kinetic trap for the dimer. The dimer structure does not appear to result from two monomers that were brought together in a random and haphazard association as would be expected if initial formation of a disulfide cross-link were followed by formation of the dimer. To begin, the GC Rho dimer has a buried surface area and geometric surface complementarity (S c ) comparable in magnitude to those of other guanylyl and adenylyl cyclases. The C2 axis of the GC Rho dimer is rotated by 90°w ith respect to the canonical AC dimer such that six of the conserved active-site residues, three identical residues from each subunit (Asp-457, Asp-501, and Arg-545), come together to form a ligand-binding pocket for tartrate and two Mn 2ϩ ions. The Asp residues bind the Mn 2ϩ ions much as do the corresponding residues in adenylyl cyclase, and the conserved Arg residues bind to the tartrate carboxylates much as the corresponding residue in adenylyl cyclase binds the terminal phosphate in ddATP (14). Interestingly, the disulfide bond and Mn 2ϩ /tartrate ligand-binding site are located on the C2 axis, at opposite ends of the GC Rho dimer interface. Nonetheless, the disulfide-linked dimer is likely a crystallization artifact, and it remains to be determined experimentally whether this particular oligomeric state has any relevance for regulation of guanylyl cyclase activity.
Irrespective of whether or not the observed dimer is involved in GC Rho function, it seems clear that the catalytically active state must be a dimer, similar in arrangement of subunits to that of adenylyl cyclase (13,14). We have not been able to provide evidence for such a dimer by direct methods such as SEC, AUC, and crystallography, but we could provide evidence through the non-linear dependence of guanylyl cyclase activity on enzyme concentration. It is likely that Mn 2ϩ and substrate assist in formation of the dimer under these conditions, but even so, the K D value for formation (20 -80 M) is much higher than for other guanylyl cyclases (e.g. the K D value for the catalytic domain of the mammalian soluble guanylyl cyclase is 0.45 M) (16). The extremely high K D value suggests that formation of the active-state dimer might be part of the light-activation mechanism. Future work in the laboratory will explore the role of domain dimerization in the lightdependent activation of full-length RhoGC.

Materials and methods
Unless specified otherwise, all materials were as described previously (8).

Protein expression and purification
Two constructs containing the GC domain of RhoGC were used in these studies: one composed of the GC domain alone ("GC Rho "; amino acids 443-627), and the other that included the CC domain as well ("GCwCC Rho "; amino acids 397-627). Both were cloned into a pET15b vector for expression in E. coli BL21(DE3) cells and included a C-terminal His 6 tag for purification on a nickel-affinity column. Both genes also included a Val codon after the first methionine to improve bacterial expression. The BL21(DE3) cell line harbored an additional pGro7 plasmid (Takara Bio Inc.) containing the genes for the GroEL and GroES molecular chaperone proteins under the control of the AraB promoter. Competent cells were transformed with either pET15b-GC Rho or pET15b-GCwCC Rho according to standard procedures and incubated overnight at 37°C on LB-agar plates containing 100 g/ml ampicillin and 20 g/ml chloramphenicol. Single colonies were used to inoculate small growth cultures (10 ml of LB containing 100 g/ml ampicillin and 20 g/ml chloramphenicol), which were grown for 14 -18 h with shaking at 220 rpm and 37°C. One ml of the overnight culture was used to inoculate a liter of LB containing both antibiotics. 500 mg of L-(ϩ)-arabinose (Acros Organics) was added upon inoculation to begin induction of GroEL/ES. Cells were grown at 37°C until an A 600 0.4 -0.7 was reached, at which point expression of the protein was induced by addition of IPTG to a final concentration of 250 M. Cells were then incubated with shaking at 20°C for 20 h and harvested by centrifugation at 4000 rpm in a Beckman Coulter JLA-8.1000 rotor for 15 min at 4°C. Cell pellets were stored at Ϫ80°C until needed.
The frozen cell pellets were thawed on ice and resuspended to a final volume of 50 ml with lysis buffer (25 mM HEPES, pH 7.0, 100 mM NaCl, and 20 mM imidazole) containing 1 mM PMSF. The cell suspension was sonicated on ice with a Misonix Sonicator 3000 at 70-watt power with a cycle of 20 s on and 20 s off for a total of 4 min. Sonicated cells were centrifuged in a Beckman Coulter JA-20 rotor at 16,000 rpm for 30 min at 4°C to pellet cell debris, and the supernatant fraction was passed through a 0.22-m filter before loading onto a pre-equilibrated 5-ml prepacked HiTrapFF nickel-Sepharose column (GE Healthcare) at 1 ml/min. The column was washed with 10 column volumes of lysis buffer followed by 20 volumes of lysis buffer containing 40 mM imidazole, and the protein was eluted with an 80-ml linear gradient of 40 -500 mM imidazole in lysis buffer. Protein was monitored by absorbance at 280 nm, and fractions containing GC Rho (or GCwCC Rho ) were identified by SDS-PAGE and pooled before concentration with an exchange of buffer to remove imidazole (10-kDa molecular mass cutoff Amicon Ultra Centrifuge Filter from EMD Millipore). Purified GC Rho and GCwCC Rho were frozen in small concentrated aliquots at Ϫ80°C until needed.
Size-exclusion chromatography (SEC) on an Ä kta FPLC system (Amersham Biosciences, Uppsala, Sweden) equipped with Structure and monomer/dimer equilibrium of guanylyl cyclase a Superdex-200 10/300 GL gel filtration column (GE Healthcare) was used to determine the oligomeric state of the proteins. 300 l of 200 M purified GC Rho (or GCwCC Rho ) was loaded onto the Superdex column that had been pre-equilibrated with 25 mM HEPES, pH 7.0, and 100 mM NaCl at 4°C, and the column was run at a constant rate of 0.5 ml/min. Protein was monitored by absorbance at 280 nm, and elution volume was compared with a standard curve using soluble protein standards to estimate the molecular mass of the peak fraction.
A separate GC Rho construct, GC NHisTEV , was designed with a TEV-protease site (ENLYFQG) inserted immediately after the N-terminal His 6 tag and followed by a Met and the remainder of GC Rho sequence to determine whether the affinity tag interfered with dimerization of the protein in solution. The protein was expressed and purified as presented above. The purified protein was proteolyzed with a 30:1 ratio of GC NHisTEV to TEVprotease for 9 h at 4°C in 25 mM HEPES buffer, pH 7.0, containing 100 mM NaCl, 1 mM DTT, and 0.5 mM EDTA in a total volume of 1 ml. The cleaved protein was separated from undigested GC NHisTEV and TEV-protease by nickel-affinity chromatography, as confirmed by SDS-PAGE and anti-His 6 Western blot analysis. Both TEV-cleaved and -uncleaved GC NHisTEV were shown to be monomeric in solution by SEC on a Superdex-200 column equilibrated in 25 mM HEPES, pH 7.0, 100 mM NaCl running at 0.5 ml/min at 4°C.

Analytical ultracentrifugation
Sedimentation velocity experiments were designed to identify factors (e.g. salt, enzyme concentration, divalent cation, etc.) that could impact the oligomeric state of isolated GC Rho protein (25). All centrifugation steps were carried out using a ProteomeLab Optima XL-A ultracentrifuge (Beckman Coulter). In all cases, 395 l of protein was loaded into double sector centerpieces balanced by 400 l of reference solution containing all components of the sample solution except the protein. All cell housings were loaded into a four-hole An-60 Ti rotor and spun at 35,000 rpm at 22°C for 15 h. Absorbance at 280 nm was measured to track protein concentration during sedimentation. Data were processed and analyzed using SEDFIT software (National Institutes of Health, Version 14.4fb). Data were fit using a continuous c(s) distribution model based on theory for a soluble and globular (spherical) protein with self-association. Parameters used to fit the raw absorbance data to a confidence level of 95% included assumptions for average protein partial specific volume (0.73) as well as for buffer density (1.00 g/ml) and buffer viscosity (0.01002 units) being identical to pure water as no viscous agents such as glycerol were used to stabilize the protein in solution.

Enzymatic activity assays
Guanylyl cyclase activity was measured using reversed-phase (RP)-HPLC to follow formation of cGMP and disappearance of GTP. 100-l reactions were prepared in 50 mM Tris buffer, pH 7.6, containing 50 mM NaCl and 0.5 mM EDTA. Reactions contained GTP and metal ion (MgCl 2 or MnCl 2 ) at the concentrations indicated in the figures and were initiated by addition of enzyme. 20-l aliquots were quenched at specified time points (30 s to 5 h) by combining with an equal volume of 1 N HCl.
Precipitated protein was removed with 0.22-m Spin-X centrifugal filters (Corning Costar) by centrifugation at 5000 rpm for 5 min in a tabletop centrifuge. The samples were then neutralized with 20 l of 1 M potassium phosphate, pH 8, and applied to a 250 ϫ 2.1-mm ACE 5 C18-AR reversed-phase 5 m column connected to an Agilent 1260 Infinity HPLC system with a G136D 1260 multiwavelength detector. Nucleotides were separated by isocratic elution with a 100 mM potassium phosphate buffer, pH 6.2, at a flow rate of 0.4 ml/min and monitored by absorbance at 254 nm. Peaks were integrated with OpenLab CDS ChemStation software and compared with peaks from standards for GTP and cGMP of known concentration.
Assays for adenylyl cyclase activity were performed identically except that the reaction contained ATP instead of GTP, and the HPLC column running buffer was 100 mM potassium phosphate, pH 6.2, containing 10% (v/v) methanol.
The concentration of stock solutions was determined spectrophotometrically using extinction coefficients of 13,700 M Ϫ1 cm Ϫ1 at 252 nm for GTP, 12,320 M Ϫ1 cm Ϫ1 at 260 nm for cGMP, 15,400 M Ϫ1 cm Ϫ1 at 259 nm for ATP, and 15,000 M Ϫ1 cm Ϫ1 at 260 nm for cAMP.
The dependence of activity on enzyme concentration was evaluated through a non-linear least-squares fit (MATLAB) of the rate data to Equation 1 for a reaction in which the only active species is a dimer of the enzyme in a monomer/dimer equilibrium, where ϭ reaction rate (M/s); k cat ϭ turnover number (s Ϫ1 ); E ϭ total enzyme concentration (M); and K D is the equilibrium dissociation constant for dimerization of the enzyme, defined as K D ϭ [monomer] 2 / [dimer] with units of M. K D and k cat are parameters determined through the fitting process. These reactions contained 10 mM GTP and 20 mM metal ion (Mn 2ϩ or Mg 2ϩ ), which was saturating for all enzyme concentrations tested.

SDS-PAGE analysis of disulfide cross-links
Samples of GC Rho and GCwCC Rho were incubated for 20 h under various conditions, as indicated in Fig. 6, to test for the formation of intermolecular disulfide bonds. Each sample contained 1 M GC Rho or GCwCC Rho in 25 mM HEPES, pH 7.0, and 100 mM NaCl. Metal ion (Mn 2ϩ or Mg 2ϩ ), GTP substrate, and cGMP and inorganic pyrophosphate products were included as indicated in the figure. Each sample was diluted 1:2 with gel load buffer, with or without 5 mM DTT, and then loaded onto a 10% Mini-PROTEAN TGX polyacrylamide gel (Bio-Rad) for analysis. Protein bands were visualized with Coomassie Brilliant Blue stain.

Crystallization
Crystallization trials were performed by sitting drop vapor diffusion at room temperature using Hampton (Hampton Research, CA) and Jena Bioscience (Jena Bioscience, Jena, Germany) sparse matrix crystallization screens. Drops were set with a Phoenix robot (Art Robbins Instruments, CA) by mixing Structure and monomer/dimer equilibrium of guanylyl cyclase protein (GC Rho , GCwCC Rho , or GC AC ; 20 mg/ml) in 25 mM HEPES, pH 7.0, containing 100 mM NaCl with crystallization mother liquor in a 1:1 ratio. Crystals of GC Rho appeared after about 3 weeks in 25% w/v PEG3350, 100 mM BisTris, pH 5.5, and 200 mM NaCl, whereas GC AC crystals formed in 15% w/v PEG20000 and 10 mM potassium hydrogen tartrate. These crystals all contained a single subunit in the asymmetric unit. In some trials, the protein solution also contained 10 mM MgCl 2 (or MnCl 2 ) and 1 mM ddGTP. In this case (MnCl 2 ), crystals appeared in 20% (w/v) PEG3350 and 200 mM potassium sodium tartrate and contained a disulfide cross-linked dimer in the asymmetric unit.

Data collection, processing, and refinement
Crystals were soaked in reservoir solution containing 15% glycerol as cryoprotectant before flash-freezing in liquid nitrogen. Diffraction data were collected at 100 K with beamline 8.2.1 at the Advanced Light Source (Lawrence Berkeley National Laboratory, Berkeley, CA) using an ADSC Q315R CCD detector (Area Detector Systems Corp.). The best crystals diffracted up to 1.1 and 1.4 Å resolution for the GC Rho and GC AC monomers, respectively, and 1.7 Å resolution for the GC Rho dimer. Data sets were integrated using iMosflm version 7.2 (26) and scaled using SCALA version 3.3 (27) from the CCP4 software suite version 7.0 (28,29). Diffraction data were processed in space group P22 1 2 1 4 for the GC Rho and GC AC monomers and C222 1 for the GC Rho dimer. Complete data collection statistics are listed in Table 2.
The GC Rho monomer structure was solved by molecular replacement using PHASER version 2.6 (30) with the structure of the catalytic domain of homodimeric human soluble GC (PDB entry 2WZ1) molecule A as a search model. The GC AC monomer and GC Rho dimer structures were solved using the final refined model of the GC Rho monomer as a search model. Rigid body refinement followed by positional and B-factor refinement was carried out using phenix.refine (31) from the PHENIX software suite version 1.11 (32). Simulated annealing was included in earlier refinements to minimize the initial model bias. Manual model building was done using COOT version 0.8 (33). Water molecules were included in the final refinement after satisfying the criteria of 3 F o Ϫ F c and 1 2F o Ϫ F c . Several iterative cycles of refinement were carried out before final submission of data. Data collection and final refinement statistics are given in Table 2. Data sets for the GC Rho monomer (PDB entry 6AO9), GC AC (PDB entry 6AOA), and GC Rho dimer (PDB entry 6AOB) have been submitted to the Protein Data Bank. All crystal structure figures in this paper were prepared using PyMOL version 1.8 (Schrödinger LLC, Portland, OR).
Author contributions-D. D. O. conceived and coordinated the study. R. P. K. and B. R. M. were responsible for all aspects of the experimental program. J. F. was involved in the initial development of the expression and purification system for the guanylyl cyclase domain. M. M. T. was involved in the development of the HPLC method for enzymatic assays and in the construction of the GCAC mutant. D. H. Z. performed assays for dependence of activity on enzyme concentration (Fig. 8). M. O. L. developed and performed disulfide cross-linking assays (Fig. 6). D. D. O., R. P. K., and B. R. M. wrote the paper. All authors discussed and commented on the manuscript.