Crystal structure of the macrocycle-forming thioesterase domain of the erythromycin polyketide synthase: versatility from a unique substrate channel.

As the first structural elucidation of a modular polyketide synthase (PKS) domain, the crystal structure of the macrocycle-forming thioesterase (TE) domain from the 6-deoxyerythronolide B synthase (DEBS) was solved by a combination of multiple isomorphous replacement and multiwavelength anomalous dispersion and refined to an R factor of 24.1% to 2.8-A resolution. Its overall tertiary architecture belongs to the alpha/beta-hydrolase family, with two unusual features unprecedented in this family: a hydrophobic leucine-rich dimer interface and a substrate channel that passes through the entire protein. The active site triad, comprised of Asp-169, His-259, and Ser-142, is located in the middle of the substrate channel, suggesting the passage of the substrate through the protein. Modeling indicates that the active site can accommodate and orient the 6-deoxyerythronolide B precursor uniquely, while at the same time shielding the active site from external water and catalyzing cyclization by macrolactone formation. The geometry and organization of functional groups explain the observed substrate specificity of this TE and offer strategies for engineering macrocycle biosynthesis. Docking of a homology model of the upstream acyl carrier protein (ACP6) against the TE suggests that the 2-fold axis of the TE dimer may also be the axis of symmetry that determines the arrangement of domains in the entire DEBS. Sequence conservation suggests that all TEs from modular polyketide synthases have a similar fold, dimer 2-fold axis, and substrate channel geometry.

Atomic structure of a glycerol channel and implications for substrate permeation in aqua(glycero)porins.

The structure of a glycerol channel from Escherichia coli at 2.2 A resolution serves as a basis for the understanding of selective transmembrane substrate permeation. In the course of permeation, glycerol molecules diffuse through a tripathic channel with their alkyl backbone wedged against a hydrophobic corner, such that OH groups become acceptors and donors of hydrogen bonds at the same time. The structure of the channel explains the preferential permeability for linear carbohydrates and absolute exclusion of ions and charged solutes. Its gene-duplicated sequence has a structural counterpart in a pseudo two-fold symmetry within the monomeric channel protein.

Structure of a glycerol-conducting channel and the basis for its selectivity.

Membrane channel proteins of the aquaporin family are highly selective for permeation of specific small molecules, with absolute exclusion of ions and charged solutes and without dissipation of the electrochemical potential across the cell membrane. We report the crystal structure of the Escherichia coli glycerol facilitator (GlpF) with its primary permeant substrate glycerol at 2.2 angstrom resolution. Glycerol molecules line up in an amphipathic channel in single file. In the narrow selectivity filter of the channel the glycerol alkyl backbone is wedged against a hydrophobic corner, and successive hydroxyl groups form hydrogen bonds with a pair of acceptor, and donor atoms. Two conserved aspartic acid-proline-alanine motifs form a key interface between two gene-duplicated segments that each encode three-and-one-half membrane-spanning helices around the channel. This structure elucidates the mechanism of selective permeability for linear carbohydrates and suggests how ions and water are excluded.

Crystal structure of the HIV-1 integrase catalytic core and C-terminal domains: a model for viral DNA binding.

Insolubility of full-length HIV-1 integrase (IN) limited previous structure analyses to individual domains. By introducing five point mutations, we engineered a more soluble IN that allowed us to generate multidomain HIV-1 IN crystals. The first multidomain HIV-1 IN structure is reported. It incorporates the catalytic core and C-terminal domains (residues 52-288). The structure resolved to 2.8 A is a Y-shaped dimer. Within the dimer, the catalytic core domains form the only dimer interface, and the C-terminal domains are located 55 A apart. A 26-aa alpha-helix, alpha6, links the C-terminal domain to the catalytic core. A kink in one of the two alpha6 helices occurs near a known proteolytic site, suggesting that it may act as a flexible elbow to reorient the domains during the integration process. Two proteins that bind DNA in a sequence-independent manner are structurally homologous to the HIV-1 IN C-terminal domain, suggesting a similar protein-DNA interaction in which the IN C-terminal domain may serve to bind, bend, and orient viral DNA during integration. A strip of positively charged amino acids contributed by both monomers emerges from each active site of the dimer, suggesting a minimally dimeric platform for binding each viral DNA end. The crystal structure of the isolated catalytic core domain (residues 52-210), independently determined at 1.6-A resolution, is identical to the core domain within the two-domain 52-288 structure.

Expression, purification, and structural characterization of the bacteriorhodopsin-aspartyl transcarbamylase fusion protein.

We are testing a strategy for creating three-dimensional crystals of integral membrane proteins which involves the addition of a large soluble domain to the membrane protein to provide crystallization contacts. As a test of this strategy we designed a fusion between the membrane protein bacteriorhodopsin (BR) and the catalytic subunit of aspartyl transcarbamylase from Escherichia coli. The fusion protein (designated BRAT) was initially expressed in E. coli at 51 mg/liter of culture, to yield active aspartyl transcarbamylase and an unfolded bacterio-opsin (BO) component. In Halobacterium salinarum, BRAT was expressed at a yield of 7 mg/liter of culture and formed a high-density purple membrane. The visible absorption properties of BRAT were indistinguishable from those of BR, demonstrating that the fusion with aspartyl transcarbamylase had no effect on BR structure. Electron microscopy of BRAT membrane sheets showed that the fusion protein was trimeric and organized in a two-dimensional crystalline lattice similar to that in the BR purple membrane. Following solubilization and size-exclusion purification in sodium dodecyl sulfate, the BO portion of the fusion was quantitatively refolded in tetradecyl maltoside (TDM). Ultracentrifugation demonstrated that BR and BRAT-TDM mixed micelles had molecular masses of 138 and 162 kDa, respectively, with a stoichiometry of one protein per micelle. High TDM concentrations (>20 mM) were required to maintain BRAT solubility, hindering three-dimensional crystallization trials. We have demonstrated that BR can functionally accommodate massive C-terminal fusions and that these fusions may be expressed in quantities required for structural investigation in H. salinarum.

Structure of bovine pancreatic cholesterol esterase at 1.6 A: novel structural features involved in lipase activation.

The structure of pancreatic cholesterol esterase, an enzyme that hydrolyzes a wide variety of dietary lipids, mediates the absorption of cholesterol esters, and is dependent on bile salts for optimal activity, is determined to 1.6 A resolution. A full-length construct, mutated to eliminate two N-linked glycosylation sites (N187Q/N361Q), was expressed in HEK 293 cells. Enzymatic activity assays show that the purified, recombinant, mutant enzyme has activity identical to that of the native, glycosylated enzyme purified from bovine pancreas. The mutant enzyme is monomeric and exhibits improved homogeneity which aided in the growth of well-diffracting crystals. Crystals of the mutant enzyme grew in space group C2, with the following cell dimensions: a = 100.42 A, b = 54.25 A, c = 106.34 A, and beta = 104.12 degrees, with a monomer in the asymmetric unit. The high-resolution crystal structure of bovine pancreatic cholesterol esterase (Rcryst = 21.1%; Rfree = 25.0% to 1.6 A resolution) shows an alpha-beta hydrolase fold with an unusual active site environment around the catalytic triad. The hydrophobic C terminus of the protein is lodged in the active site, diverting the oxyanion hole away from the productive binding site and the catalytic Ser194. The amphipathic, helical lid found in other triglyceride lipases is truncated in the structure of cholesterol esterase and therefore is not a salient feature of activation of this lipase. These two structural features, along with the bile salt-dependent activity of the enzyme, implicate a new mode of lipase activation.

A designed four helix bundle protein with native-like structure.

A 108 amino acid protein was designed and constructed from a reduced alphabet of seven amino acids. The 2.9 A resolution X-ray crystal structure confirms that the protein is a four helix bundle, as it was designed to be. Hydrogen/deuterium exchange experiments reveal buried amide protons with protection factors in excess of 1 x 10(6) in the range characteristic of well protected protons in functional folded proteins (10(3)-10(8)) rather than protons in rapid exchange (0-10(2)). The protein is monomeric at 1 mM, the concentration at which the exchange experiments were undertaken, indicating that the exchange factors are due to a unique stable tertiary structure fold, and not due to any higher order quaternary structure. Thermodynamic analysis provides an estimate of the free energy of folding of -9.3 kcal mole-1 at 25 degrees C, consistent with the free energy of folding derived from the protection factors of the most protected protons, indicating that global unfolding is required for exchange of the most protected protons.

Structure at 2.5 A of a designed peptide that maintains solubility of membrane proteins.

A 24-amino acid peptide designed to solubilize integral membrane proteins has been synthesized. The design was for an amphipathic alpha helix with a "flat" hydrophobic surface that would interact with a transmembrane protein as a detergent. When mixed with peptide, 85 percent of bacteriorhodopsin and 60 percent of rhodopsin remained in solution over a period of 2 days in their native forms. The crystal structure of peptide alone showed it to form an antiparallel four-helix bundle in which monomers interact, flat surface to flat surface, as predicted.

Two-dimensional crystallization of Escherichia coli-expressed bacteriorhodopsin and its D96N variant: high resolution structural studies in projection.

Highly ordered two-dimensional (2-D) crystals of Escherichia coli-expressed bacteriorhodopsin analog (e-bR) and its D96N variant (e-D96N) reconstituted in Halobacterium halobium lipids have been obtained by starting with the opsin protein purified in the denaturing detergent sodium dodecyl sulfate. These crystals embedded in glucose show electron diffraction in projection to better than 3.0 A at room temperature. This is the first instance that expressed bR or a variant has been crystallized in 2-D arrays showing such high order. The crystal lattice is homologous to that in wild-type bR (w-bR) in purple membranes (PM) and permit high resolution analyses of the structure of the functionally impaired D96N variant. The e-bR crystal is isomorphous to that in PM with an overall averaged fractional change of 12.7% (26-3.6-A resolution) in the projection structure factors. The projection difference Fourier map e-bR-PM at 3.6-A resolution indicates small conformational changes equivalent to movement of approximately < 7 C-atoms distributed within and in the neighborhood of the protein envelope. This result shows that relative to w-bR there are no global structural rearrangements in e-bR at this 3.6 A resolution level. The e-D96N crystal is isomorphous to the e-bR crystal with a smaller (9.2%) overall averaged fractional change in the structure factors. The significant structural differences between e-D96N and e-bR are concentrated at high resolution (5-3.6 A); however, these changes are small as quantified from the 3.6 A resolution e-D96N-e-bR Fourier difference map. The difference map showed no statistically significant peaks or valleys within 5 A in projection from the site of D96 substitution on helix C. Elsewhere within the protein envelope the integrated measure of peaks or valleys was < approximately 3 C-atom equivalents. Thus, our results show that for the isosteric substitution of Asp96 by Asn, the molecular conformation of bR in its ground state is essentially unaltered. Therefore, the known effect of D96N on the slowed M412 decay is not due to ground-state structural perturbations.

Bacteriorhodopsin D85N: three spectroscopic species in equilibrium.

Ground-state absorbance measurements show that BR from Halobacterium halobium containing asparagine at residue 85 (D85N) exists as three distinct chromophoric states in equilibrium. In the pH range 6-12 the absorbance spectra of the three states are demonstrated to be similar to flash-induced spectral intermediates which comprise the latter portion of the wild-type BR photocycle. One of the states absorbs maximally at 405 nm, has a deprotonated Schiff base, and contains predominantly the 13-cis form of retinal, identifying it as a close homologue of the M intermediate in the BR photocycle. The other species possess absorbance maxima with correspondence to those of the wild-type N (570 nm) and O (615 nm) photointermediates. The retinal composition of the O-like form was found to be dominated by all-trans isomer. The pH dependence of the concentrations of the equilibrium species corresponds closely with the pH dependence of the M, N, and O photointermediates. These data support kinetic models which emphasize the role of back-reactions during the photocycle of bacteriorhodopsin. Energetic and spectral characterization of the D85N ground-state equilibrium supports its use as a model for elucidating molecular transitions comprising the latter portion of the BR photocycle.

Effects of Asp-96----Asn, Asp-85----Asn, and Arg-82----Gln single-site substitutions on the photocycle of bacteriorhodopsin.

Bacteriorhodopsin (BR) with the single-site substitutions Arg-82----Gln (R82Q), Asp-85----Asn (D85N), and Asp-96----Asn (D96N) is studied with time-resolved absorption spectroscopy in the time regime from nanoseconds to seconds. Time-resolved spectra are analyzed globally by using multiexponential fitting of the data at multiple wavelengths and times. The photocycle kinetics for BR purified from each mutant are determined for micellar solutions in two detergents, nonyl glucoside and CHAPSO, and are compared to results from studies on delipidated BR (d-BR) in the same detergents. D85N has a red-shifted ground-state absorption spectrum, and the formation of an M intermediate is not observed. R82Q undergoes a pH-dependent transition between a purple and a blue form with different pKa values in the two detergents. The blue form has a photocycle resembling that for D85N, while the purple form of R82Q forms an M intermediate that decays more rapidly than in d-BR. The purple form of R82Q does not light-adapt to the same extent as d-BR, and the spectral changes in the photocycle suggest that the light-adapted purple form of R82Q contains all-trans- and 13-cis-retinal in approximately equal proportions. These results are consistent with the suggestions of others for the roles of Arg-82 and Asp-85 in the photocycle of BR, but results for D96N suggest a more complex role for Asp-96 than previously suggested. In nonyl glucoside, the apparent decay of the M-intermediate is slower in D96N than in d-BR, and the M decay shows biphasic kinetics. However, the role of Asp-96 is not limited to the later steps of the photocycle. In D96N, the decay of the KL intermediate is accelerated, and the rise of the M intermediate has an additional slow phase not observed in the kinetics of d-BR. The results suggest that Asp-96 may play a role in regulating the structure of BR and how it changes during the photocycle.

Resonance Raman spectra of bacteriorhodopsin mutants with substitutions at Asp-85, Asp-96, and Arg-82.

Detergent solubilized bacteriorhodopsin (BR) proteins which contain alterations made by site-directed mutagenesis (Asp-96----Asn, D96N; Asp-85----Asn, D85N; and Arg-82----Gln, R82Q) have been studied with resonance Raman spectroscopy. Raman spectra of the light-adapted (BRLA) and M species in D96N are identical to those of native BR, indicating that this residue is not located near the chromophore. The BRLA states of D85N and especially R82Q contain more of the 13-cis, C = N syn (BR555) species under ambient illumination compared to solubilized native BR. Replacement of Asp-85 with Asn causes a 25 nm red-shift of the absorption maximum and a frequency decrease in both the ethylenic (-7 cm-1) and the Schiff base C = NH+ (-3 cm-1) stretching modes of BRLA. These changes indicate that Asp-85 is located close to the protonated retinal Schiff base. The BRLA spectrum of R82Q exhibits a slight perturbation of the C = NH+ band, but its M spectrum is unperturbed. The Raman spectra and the absorption properties of D85N and R82Q suggest that the protein counterion environment involves the residues Asp-85-, Arg-82+ and presumably Asp-212-. These data are consistent with a model where the strength of the protein-chromophore interaction and hence the absorption maximum depends on the overall charge of the Schiff base counterion environment.

Wild-type and mutant bacteriorhodopsins D85N, D96N, and R82Q: purification to homogeneity, pH dependence of pumping, and electron diffraction.

Bacterioopsin, expressed in Escherichia coli as a fusion protein with 13 heterologous residues at the amino terminus, has been purified in the presence of detergents and retinylated to give bacteriorhodopsin. Further purification yielded pure bacteriorhodopsin, which had an absorbance ratio (A280/A lambda max) of 1.5 in the dark-adapted state in a single-detergent environment. This protein has a folding rate, absorbance spectrum, and light-induced proton pumping activity identical with those of bacteriorhodopsin purified from Halobacterium halobium. Protein expressed from the mutants D85N, D96N, and R82Q and purified similarly yielded pure protein with absorbance ratios of 1.5. Proton pumping rates of bacteriorhodopsins with the wild-type sequence and variants D85N, D96N, and R82Q were determined in phospholipid vesicles as a function of pH. D85N was inactive at all pH values, whereas D96N was inactive from pH 7.0 to pH 8.0, where wild type is most active, but had some activity at low pH. R82Q showed diminished proton pumping with the same pH dependence as for wild type. Bacteriorhodopsin purified from E. coli crystallized in two types of two-dimensional crystal lattices suitable for low-dose electron diffraction, which permit detailed analysis of structural differences in site-directed variants. One lattice was trigonal, as in purple membrane, and showed a high-resolution electron diffraction pattern from glucose-sustained patches. The other lattice was previously uncharacterized with unit cell dimensions a = 127 A, b = 67 A, and symmetry of the orthorhombic plane group pgg.

Wild-type and mutant bacterioopsins D85N, D96N, and R82Q: high-level expression in Escherichia coli.

The integral membrane protein bacterioopsin, found in the extremely halophilic archaebacterium Halobacterium halobium, was expressed in Escherichia coli as a fusion protein containing 13 heterologous amino acids at the amino terminus. The expressed protein was localized primarily to the E. coli cytoplasmic membrane (greater than 80%) and had an in vivo half-life of 26 min. The amount of bacterioopsin in E. coli crude lysates was quantitated immunologically from Western blots and was expressed at 10-20-fold higher levels than seen previously (i.e., 17 mg/L; 5.6% of the total protein). Three distinct forms of the protein were detected immunologically: two of the forms were generated by the removal of either one or four amino acid residues at the amino terminus; the third form remained unaltered.

Effects of detergent environments on the photocycle of purified monomeric bacteriorhodopsin.

Time-resolved difference spectra have been obtained for the photocycle of delipidated bacteriorhodopsin monomers (d-BR) in six different detergent micelle environments that were prepared by two new detergent-exchange techniques. A global kinetic analysis of the photocycle spectra for d-BR in each detergent environment was performed. Comparison of these results with those obtained for the photocycle of bacteriorhodopsin in purple membrane (PM) shows that there is one fewer kinetically distinguishable process for monomeric BR between the decay of the K intermediate and the rise of the M intermediate. Assuming a sequential pathway occurs in the photocycle, it appears that the equilibrium between the L and M intermediates is reached much more rapidly in the detergent micelles. This is attributed to a more direct interaction between Asp-85 and the proton on the nitrogen of the Schiff base of retinal for BR in the detergents. Equilibrium concentrations of late photocycle intermediates are also altered in detergents. The later steps of the photocycle, including the decay of the M intermediate, are slowed in detergents with rings in their hydrocarbon region. This is attributed to effects on conformational changes occurring during the decay of M and/or other later photocycle intermediates. The lifetime of dark adaptation of light-adapted d-BR in different detergent environments increases in environments where the lifetime of the M intermediate increases. These results suggest that the high percentage of either unsaturated or methyl-branched lipids in PM and the membranes of other retinal proteins may be important for their effective functioning.

Preparative purification of functional bacteriorhodopsin by high-performance size-exclusion chromatography.

High-performance size-exclusion chromatography (HPSEC) was used to produce stable bacteriorhodopsin essentially free of native lipids. The purified bacteriorhodopsin was shown to be highly functional when reconstituted into phospholipid vesicles. Purple membrane was washed in the detergent 3-[( 3-cholamidopropyl)dimethylammonio]-2,2-hydroxy-1-propanesulfonate (CHAPSO) to remove a large fraction (65%) of the membrane lipids, solubilized in Triton X-100 and purified on a Bio-Sil TSK G3000SW column using a CHAPSO mobile phase. Pooled column fractions of bacteriorhodopsin from 25-mg sample loads show a 280/548 nm absorbance ratio of 1.5-1.6 and contain less than 4% endogenous lipids. This HPSEC method requires much less expensive synthetic detergent and is much faster than open column methods [cf. L.J.W. Miercke, P.E. Ross, R.M. Stroud and E.A. Dratz, J. Biol. Chem., 264 (1989) 7531-7535].

Purification of bacteriorhodopsin and characterization of mature and partially processed forms.

Bacteriorhodopsin (BR) essentially free of native lipids has been prepared in a highly stable state. Purple membrane was solubilized in Triton X-100 and BR was purified by size exclusion chromatography using 3-[cholamidopropyl)dimethylammonio]-2-hydroxyl-1-propanesulfonic acid (CHAPSO) detergent at pH 5. Molar ratios of phospholipid/BR ranged from 0.4 to 0.05 corresponding to 94-98% phospholipid removal. Purified BR has an absorbance ratio (A280nm/A548nm) of 1.5-1.6 in the dark-adapted state which is the highest purified BR/protein ratio reported to date. The purified BR in CHAPSO shows maximum stability in the pH range 5.0-5.5. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis profiles of native purple membrane and solubilized BR from most Halobacterium halobium JW-3 cultures show 3 higher molecular weight bands in addition to BR. Immunological staining and amino acid sequencing indicates that these additional proteins are partially processed forms of the BR precursor protein. The BR preprotein contains 13 additional amino acids on the NH2 terminus which are removed by post-translational processing in at least four steps. Isoelectric focusing separated most delipidated and non-delipidated BR samples into 8 bands. Incomplete BR post-translational processing BR is thought to be largely responsible for the multiplicity of isoelectric BR species. The principal components have pI values of 5.20 and 5.24 and both have absorption maxima at 550 nm, characteristic of detergent-solubilized BR. BR in Triton X-100 or nonylglucoside, delipidated BR in CHAPSO, and BR in intact purple membrane all have a dark-adapted ratio of 13-cis to all-trans-retinal of 1.9:1.

Isoelectric focusing studies of bacteriorhodopsin.

Purified bacteriorhodopsin (BR) samples show a minimum of four isoelectric forms in immobilized pH gradient isoelectric focusing gels. The bands occur as doublets with isoelectric points (pI) centered at 5.20 (principal species) and 5.60. In typical preparations additional bands may be observed at 4.90, 5.07 and 5.50. Purple membrane (PM) was proteolyzed with papain to calibrate the pI shift produced by changing the number of charges on the protein. Asp-242 is removed during the first cleavage between residues 239 and 240 resulting in the loss of a single negative charge and a shift of the principal doublet by +0.35 pH units to pI 5.55. The second papain cleavage occurs between residues 231 and 232 which removes Glu-232, -234 and -237 and shifts the pI by +0.60 pH units to pI 6.10. The +0.60 pH shift upon the second papain cleavage is consistent with the loss of two negative charges and is supported by prior evidence that at least one of the three glutamate residues lost during the second proteolysis step is protonated and neutral in the intact protein. The native and proteolyzed products of BR retain the characteristic 550 nm absorption maxima for solubilized BR. A model for the structural origin of the pI heterogeneity of BR species in proteolyzed PM is presented.