PEDS Advance Access originally published online on September 29, 2006
Protein Engineering Design and Selection 2006 19(12):525-535; doi:10.1093/protein/gzl040
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Published by Oxford University Press.
Dissecting carbohydrate–Cyanovirin-N binding by structure-guided mutagenesis: functional implications for viral entry inhibition
1 Laboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney Diseases National Institutes of Health, Bethesda, MD 20892, USA 2 Special Pathogens Branch, Division of Viral and Rickettsial Diseases National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA 30333, USA 3 Laboratory of Molecular Microbiology, Hospital Universitario 12 de Octubre Madrid, Spain 4 Department of Structural Biology, University of Pittsburgh Medical School Pittsburgh, PA 15260, USA
5To whom correspondence should be addressed. E-mail: amg100{at}pitt.edu; lbarrientos1{at}cdc.gov
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
The HIV-inactivating protein Cyanovirin-N (CV-N) is a cyanobacterial lectin that exhibits potent antiviral activity at nanomolar concentrations by interacting with high-mannose carbohydrates on viral glycoproteins. To date there is no molecular explanation for this potent virucidal activity, given the experimentally measured micromolar affinities for small sugars and the problems encountered with aggregation and precipitation of high-mannose/CV-N complexes. Here, we present results for two CV-N variants, CV-NmutDA and CV-NmutDB, compare their binding properties with monomeric [P51G]CV-N (a stabilized version of wtCV-N) and test their in vitro activities. The mutations in CV-NmutDA and CV-NmutDB comprise changes in amino acids that alter the trimannose specificity of domain AM and abolish the sugar binding site on domain BM, respectively. We demonstrate that carbohydrate binding via domain BM is essential for antiviral activity, whereas alterations in sugar binding specificity on domain AM have little effect on envelope glycoprotein recognition and antiviral activity. Changes in AM, however, affect the cross-linking activity of CV-N. Our findings augment and clarify the existing models of CV-N binding to N-linked glycans on viral glycoproteins, and demonstrate that the nanomolar antiviral potency of CV-N is related to the constricted and spatially crowded arrangement of the mannoses in the glycan clusters on viral glycoproteins and not due to CV-N induced virus particle agglutination, making CV-N a true viral entry inhibitor.
Keywords: Cyanovirin-N/high-mannose oligosaccharides/mutant design/viral env glycoprotein/virucidal agent
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Specific recognition of an oligosaccharide by a protein is a very complex problem, more so than other biologically relevant interactions, such as protein–protein or protein–nucleic acid recognition, since the individual sugar units in a glycan exhibit only a limited repertoire of functional groups and therefore are difficult to distinguish from each other. In addition, the glycosidic bonds between two monosaccharides exhibit considerable flexibility. As a consequence, a high entropic cost limits the binding affinities that can be achieved in carbohydrate interactions. The combination of flexibility and the difficulty in distinguishing monomeric building blocks allows oligosaccharides to mimic each other structurally, rendering specific recognition truly difficult.
The cyanobacterial lectin Cyanovirin-N (CV-N) (Boyd et al., 1997
) exerts its virucidal activity against a variety of enveloped viruses (Boyd et al., 1997; Dey et al., 2000; Barrientos et al., 2003a; O'Keefe et al., 2003; Barrientos et al., 2004a) by recognizing and binding to the N-linked carbohydrates on these highly glycosylated proteins. Results from a variety of laboratories unequivocally established that the Man
(1
2)Man moieties at the termini of the D1 and D3 arms of N-linked high-mannose oligosaccharides (Man-8 and Man-9) on viral surface glycoproteins are the primary molecular targets for CV-N on these viruses (reviewed in Barrientos and Gronenborn, 2005). Interestingly, not only the terminal disaccharide but also the reducing mannose residue or the linkage to it influences the affinity to CV-N (Sandström et al., 2004) and the conformation of the glycosidic linkage is responsible for the observed selectivity between the two binding sites on CV-N, explaining the difference in relative affinity for dimannoses and trimannoses with respect to binding to the two sites on CV-N (Bewley et al., 2002; Shenoy et al., 2002; Sandström et al., 2004).
In solution, CV-N can fold into a compact monomeric bilobal structure with pseudosymmetry (Bewley et al., 1998) or, under specific conditions, into a domain-swapped dimeric structure (Yang et al., 1999a; Barrientos et al., 2002a). The domain-swapped dimer was shown to represent a kinetically trapped folding intermediate at high concentration that converts into the thermodynamically stable monomer at physiological temperatures (Barrientos et al., 2002a). Using the notation introduced in Barrientos and Gronenborn (2002), the two pseudosymmetric domains AM and BM in the monomer consists of residues (1–38/90–101) and (39–89), respectively (Bewley et al., 1998), and the counterpart domains AD, AD, BD and BD in the domain-swapped dimer consist of residues (1–38 and 90'–101'), (1'–38' and 90–101), (39–50 and 51'–89') and (39'–50' and 51–89), respectively (Yang et al., 1999a). NMR titration experiments and high-resolution structures of CV-N in complex with sugars identified one carbohydrate binding pocket in each domain (two in the monomer and four in the dimer) (Bewley, 2001; Bewley and Otero-Quintero, 2001; Barrientos and Gronenborn, 2002; Bewley et al., 2002; Botos et al., 2002; Shenoy et al., 2002): A semicircular cleft in domain AM, comprising residues 1–7, 22–26 and 92–95, and a deeper pocket in domain BM, comprising residues 41–44, 50–56 and 74–78. Both sites, separated by 
40 Å, are chemically and topologically distinct, despite considerable sequence similarity. Each recognizes (12) linked dimannoses, either as an individual oligosaccharide or as part of the terminal arms of the branched Man-8 and Man-9 structures. Domain A exhibits a slight preference for the trimannose and domain B for the dimannose units, respectively (Bewley et al., 2002; Botos et al., 2002; Sandström et al., 2004). In all cases, the reported binding constants are in the micromolar range (Bewley and Otero-Quintero, 2001; Bewley et al., 2002; Shenoy et al., 2002). Specific carbohydrate–protein contacts between the Man(12)Man moiety and polar and charged amino acid residues on CV-N were delineated from the solution structure of monomeric CV-N, complexed with the Man(12)Man dimannoside on domain BM (Bewley, 2001) and from the crystal structure of Man-9 and hexamannoside (HM) bound to domains AD and AD of the domain-swapped dimer (Botos et al., 2002). In both sites, the sugar rings exhibit a stacked conformation, and residues involved in the interaction are as follows: G2/K3/Q6/T7/E23/T25/N93/D95/G96/E101 on domain AM and E41/D44/S52/E56/T57/K74/T75/R76/Q78 on domain BM (underlined in Figure 1B).
|
Although considerable information is available to date on the interaction of CV-N with substructures of Man-8 and Man-9, it is not clear whether a single sugar–protein contact is sufficient for the antiviral activity of CV-N. The multisite nature of CV-N (both in the monomeric or dimeric form) and the multivalency of the carbohydrate have been implicated as important factors for activity (Shenoy et al., 2002
); a clean dissection of individual components, however, has not been carried out. The observed high affinity towards carbohydrate epitopes on viral glycoproteins, measured by ELISA is in stark contrast to the solution binding constants for small sugars, found to bind in the micromolar range. In addition, CV-N exhibits antiviral activity in cellular assays at nanomolar concentrations. Studies on a domain AM ‘knockout’ mutant by Chang and Bewley (2002) suggested that the potent antifusion activity of CV-N only requires the carbohydrate binding site on domain BM, termed ‘high affinity’ site by these authors, based on titration data with dimannose. Subsequently, the same laboratory reported that domain AM exhibits ‘high affinity’ binding for the linear trimannoside (Bewley et al., 2002), confirming the observation by Shenoy et al. (2002), thereby rendering the term ‘high affinity’ ambiguous. It also was reported that the domain-swapped dimer appeared more active than the monomer [tetravalency versus divalency; (Kelley et al., 2002)] in a fusion assay, although, in our hands, no significant differences in anti-HIV and anti-Ebola activities were observed between the dimeric and monomeric CV-N in cellular assays (Barrientos et al., 2004b). The present study was conceived to investigate this issue further. Two mutants, CV-NmutDA and CV-NmutDB, were created in which those residues that were implicated in sugar binding were altered with the purpose of changing the carbohydrate binding site on domain AM and eliminating the sugar binging site on domain BM (Figure 1). The resulting proteins were carefully assessed with respect to their biophysical and sugar binding properties and tested in HIV and Ebola assays in parallel with the parent protein, [P51G]CV-N. Our results demonstrate that (i) eliminating the preference of trimannose for domain AM does not affect glycoprotein recognition and virucidal activity. Instead, cross-linking/precipitation at high concentrations was no longer observed; (ii) abolishing the sugar binding site on domain BM results in complete loss of antiviral activity. These new findings were incorporated into the existing models of CV-N sugar/glycoprotein interaction and allowed us to provide a rational explanation for the antiviral potency of CV-N at nanomolar concentrations.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Cloning and expression
The synthetic genes for [P51G]CV-N, CV-NmutDA and CV-NmutDB coding for the amino acid sequences displayed in Figure 1, were inserted into pET26b(+) (Novagen; Madison, WI) as described previously (Mori et al., 1998). The identity of each construct was confirmed by nucleotide sequencing. All proteins were expressed in Escherichia coli BL-21(DE3) and uniformly 15N-labeled using common procedures.
Purification and folding
Proteins were isolated and purified following the procedure described in Barrientos et al. (2004b). In brief, overexpressed proteins were isolated from the periplasmic fraction of E. coli (the expression vector used carries a pelB signal) by twice heating/cooling the cells suspended in PBS buffer (pH 7.4). After removal of insoluble material by centrifugation, the supernatant containing soluble CV-N was dialyzed against 6 M guanidinium hydrochloride (GdnHCl) in 20 mM sodium phosphate buffer, pH 6.0, followed by gel filtration on Superdex-75 (HiLoad 2.6 cm x 60 cm, Amersham Pharmacia Biotech, NJ) in the same buffer. Peak fractions were pooled, and the protein was refolded by dialysis against buffer without GdnHCl. The final purification step for the refolded protein involved chromatography on Superdex-75 in 20 mM sodium phosphate buffer, pH 6.0. In this fashion, all CV-N variants were isolated as pure folded monomers. The purity and identity of the proteins were verified by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and mass spectrometry, and conformational properties were assessed by NMR and multi-angle light scattering. Protein concentration was determined by refractive index detection and UV absorbance (280 nm) as described previously (Barrientos et al., 2004b). All proteins were stored at 4°C.
Thermal and chemical denaturation
Thermal melting was monitored by CD spectrometry, measuring the far-UV molar ellipticity at 200 nm on a J-720 spectropolarimeter (JASCO, Easton, MD), equipped with a thermostatic cell holder (Thermo NESLAB) using 0.1 cm cells. Data were collected as a function of temperature with a scan rate of 0.5°C/min over the temperature range of 5–85°C, and unfolding curves were measured on protein samples (20 µM) containing 10 mM sodium phosphate buffer, pH 6.0.
Equilibrium (un)folding induced by guanidine hydrochloride (GdnHCl) was monitored by steady-state tryptophan fluorescence. Individual protein samples were prepared in 20 mM sodium phosphate buffer, pH 6.0, containing the desired amount of GdnHCl (Ultra pure, US Biochemicals, Cleveland, OH). The GdnHCl concentration of the stock solution was determined by refractive index measurements using a refractometer (Bausch & Lomb, Rochester, NY). Experiments were carried out for protein concentrations of 10 µM to ensure two-state equilibrium between folded monomer and unfolded monomer. Reversibility of unfolding was also assessed by NMR and multi-angle light scattering.
Fluorescence spectra were recorded on a Model LS-50B spectrofluorometer (Perkin-Elmer, Wellesley, MA), equipped with a temperature-controlled water bath (Thermo NESLAB, Portsmouth, NH). Measurements were performed with a scan speed of 120 nm/min and data intervals of 0.5 nm. An excitation wavelength of 280 nm was used, and the intrinsic fluorescence (I330/I360) was recorded. The equilibrium constant, K, and the free energy of folding, 
G, were calculated using the following equation: G = –RTlnK = –RTln(fu/(1 – fu)); with R the gas constant (1.987 cal·deg–1·mol–1), T the absolute temperature (K) and fu the apparent fraction of unfolded molecules.
NMR sugar titration experiments
Titration experiments were performed with protein samples containing 100 µM (with dimannose and trimannoside) or 20 µM (with Man-9) 15N-labeled CV-N in 20 mM sodium phosphate buffer (pH 6.0), 0.05% sodium azide and 90% H2O/10% D2O. A series of 1H-15N HSQC spectra was recorded after addition of sugar aliquots from stock solutions of 4.3 mM dimannose (Sigma), 4.5 mM synthetic trimannoside (kindly provided by Daniel M.Ratner and Peter H.Seeberger) and 425 µM Man-9 (Glyko, Inc.). The molecular sizes of the complexes at the end of each titration were assessed based on heteronuclear T2 values for the amide protons and nitrogens atoms. These values were previously correlated with analytical centrifugation data (Shenoy et al., 2002; Barrientos et al., 2003b). Dimannose and trimannoside was quantitated as previously described (Shenoy et al., 2002) and the amount of Man-9 was used according to the amount indicated in the manufacture's product specification sheet (Glyko, Inc.). Titration experiments were also monitored by native gel electrophoresis. The titration was performed as described above, except that the protein concentration of 15N-labeled CV-N was 8.7 µM and no D2O was added. Aliquots (10 µl) at a given sugar:protein ratio were saved for gel electrophoresis under native conditions.
NMR experiments
NMR experiments were recorded at 20°C using an UltraShield Bruker DRX600 spectrometer equipped with a cryo-probe. Spectra were processed and analyzed with NMRPipe (Delaglio et al., 1995).
Native gel electrophoresis experiments
Aliquots (10 µl) of the sugar:protein complex mixtures were added to 2x native sample buffer (Invitrogen), and samples were run on native 10–20% Tris–glycine gels (Invitrogen) for 1-1/2 h. As reference for the position of dimeric CV-N, Q50CV-N was used, since this variant exists exclusively as domain-swapped dimer under native conditions (Kelley et al., 2002; Barrientos et al., 2004b).
Size exclusion chromatography and multi-angle light scattering
Light scattering data were obtained using separation on an analytical superdex-75 column (1.0 x 30 cm; Amersham Biosciences, Piscataway, NJ) with in-line multi-angle light scattering (DAWN EOS, Wyatt Technology, Inc., Santa Barbara, CA), refractive index (OPTILAB DSP, Wyatt Technology, Inc., Santa Barbara, CA) and UV (Waters, Inc.,) detection. 20–100 µg of proteins in 75–150 µl of 20 mM sodium phosphate buffer (pH 6.0) containing 0.02% sodium azide were applied to the pre-equilibrated S75 column at a flow rate of 0.5 ml/min at room temperature and eluted in the same buffer.
Isothermal titration experiments
Calorimetric titrations were performed using a VP-ITC isothermal titration calorimeter (MicroCal, LLC; Northampton, MA). The protein solution (30 µM) was placed in the calorimeter cell (1.44 ml active volume), stirred at 310 r.p.m. and 15 µl aliquots of the sugar solution (270 µM) were added at 2 min intervals from a 295 µl stirring syringe. A total of 20 injections were performed. Titrations were carried out at 30°C and all solutions contained 50 mM sodium phosphate buffer, 0.2 M NaCl, 0.02% NaN3 (pH 7.5). The isotherm was fit using the Origin 7.0 software with the standard One Site model (MicroCal). Values for enthalpy of binding, the apparent number of binding sites (n) on the protein and the binding affinity were derived from the fit. Other thermodynamic quantities were calculated using the standard expressions: G = –RT ln Ka; G = H – TS.
Preparation of reduced and carboxymethylated [RCM]HIV-1 gp120
HIV-1 gp120 (100 µg) (Intracel Corp.) was dialyzed against 100 mM ammonium bicarbonate, pH 8.0, containing 6 M GdnHCl. The sample was heated at 60°C for 20 min. DTT was added to a concentration of 10 mM, and the sample was incubated for 4 h at room temperature. The sample was then treated with 25 mM iodoacetic acid in the dark for 30 min at room temperature. The reaction was quenched with excess DTT, and the sample was diluted with PBS containing 1% NP40 to give a final glycoprotein concentration of 10 pmol/100 µl.
Preparation of EboZV GP1-Fc
GP1 from Ebola Zaire virus (EboZV) GP was fused with human IgG-Fc by subcloning the fragment into pEF-Fc (Kindly provided by J.M.Casasnovas, Centro Nacional de Biotecnología, CSIC, Madrid). The construction was transfected into 293T cells and supernatants were collected after 72 h.
Biological assays
Binding of CV-N mutants to glycosylated soluble HIV-1 gp120 and [RCM]HIV-1 gp120 (Intracel Corp.) was assessed by ELISA as described in Barrientos et al. (2003a).
The lectin-EboZV GP1-Fc binding assay was performed as follows: Jurkat cells were incubated with an 1:10 dilution of the supernatants containing EboVZ GP1-Fc in the presence or absence of CV-N and its variants for 30 min at 4°C, washed twice with PBS–2% FBS and incubated under the same conditions with a PE-conjugated goat anti-human IgG-Fc antibody (Immunotech, Marseille, France). After fixation with formaldehyde, cells were analyzed in an EPICS-XL cytometer using the Expo32 software.
For the infection experiments, HIV virus and EboZV-GP pseudotyped lentiviruses were handled under biosafety level 3 conditions, according to standard recommendations. EboZV GP pseudotyped lentivirus particles were used to study the in vitro potency of CV-N to inhibit EboZV GP-mediated gene transduction into HeLa cells (Barrientos et al., 2004a). The lentiviral vector pNL4-3.Luc.R–E– was obtained from Dr Nathaniel Landau (He et al., 1995) through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, and was used for the production of EboZV GP pseudotypes using a transient transfection protocol with 293T cell as previously described (Yang et al., 1999b). Supernatant fluids were obtained 48 h after transfection, filtered (0.45 µm) and stored frozen (–80°C). Supernatant fluids containing EboZV GP-pseudotyped lentiviruses were treated with different concentrations of CV-N and its variants for 20 min at RT and added to HeLa cells monolayers in 24-well plates at an M.O.I of 0.1 using DMEM–10% FBS. Infectivity was measured 48 h after infection by a luciferase assay using reagents from Promega (Madison, WI) in a Berthold Sirius luminometer (Berthold, Munich, Germany).
The anti-HIV activity was characterized by determining the percentage of cells p24 positive using FACS [antiviral activity kindly provided by Laurel Lagenaur (NIAID) and Qiang Xu (OSEL, Inc.)]. Samples were diluted in DMEM in a V bottom dish, HIV Ba-L was added, equ. M.O.I. = 0.1, followed by incubation with virus for 1 h at 37°C and addition to 10e6 IL-2 stimulated PBMC. Indinavir was added to prevent cell to cell spread. The medium was changed at 24 h and samples were fixed and stained for p24 via FACS.
|
Results and discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Glycoprotein binding by CV-N can proceed via two originally proposed plausible modes, involving one of the two sugar binding sites or both sites simultaneously. In order to elucidate the mode of action of CV-N and guide our attempts to further develop the protein as a microbicidal agent, it is important to clarify the interaction with the high-mannose sugars, Man-8 and Man-9, the pertinent glycan structures on the viral envelopes. In vitro studies of CV-N binding to these high-mannose sugars at concentrations necessary for biophysical studies are hampered by cross-linking between carbohydrate and protein, resulting in precipitation of high molecular weight aggregates. Multivalent and multisite binding is a hallmark of lectin–carbohydrate interactions (Leikina et al., 2005
; and selected reviews Lee and Lee, 2000; Lundquist and Toone, 2002) and the question, therefore, arose whether such interactions are essential for CV-N activity. We employed structure-guided mutagenesis to investigate this point, by modifying the sugar binding sites on domain AM and BM. Based on all structural data on CV-N to date, we designed and created two specific binding site mutants, referred to as CV-NmutDA and CV-NmutDB. Using these variants, the role of avidity versus affinity in antiviral activity was studied and compared with monomeric [P51G]CV-N. We selected [P51G]CV-N as the CV-N molecule of choice, since it is thermodynamically more stable than wild-type CV-N and has performed consistently and statistically better than wild-type CV-N in a number of biological assays in our hands for a wide range of cell lines and viruses [including HIV (Mori et al., 2002; Barrientos et al., 2004b), influenza (O'Keefe et al., 2003) and Ebola (Barrientos et al., 2004b)]. Mutant design
The amino acid sequences of [P51G]CV-N, CV-NmutDA and CV-NmutDB are summarized in Figure 1, indicating those positions that were altered. All three CV-N variants contain the P51G mutation for enhanced stability, previously shown to be benign for antiviral activity (Barrientos et al., 2002a; Mori et al., 2002; Barrientos et al., 2004b). Rather than completely obliterating the binding site on domain AM, we changed the primary sequence in such a manner as to make it more similar to that in domain BM [the sequence identity between the two sequence repeats (residues 1–50 and 51–101) was increased by 18%, i.e. from 32 to 50%]. This modification eliminates the preference for trimannoside, while retaining the ability to recognize the Man(12)Man epitope (see below). The reasons for our choice were as follows: (i) Chang and Bewley (2002) had already reported a site A ‘knock out’ and (ii) by eliminating the unique selectivity for the trisaccharide Man(12)Man(12)ManOMe, i.e. the D1 arm of Man-9 or Man-8, we were hoping to prevent cross-linking of proteins that simultaneously interact with the D1 (via domain AM) and D3 (via domain BM) arms of the sugar. We therefore grafted most of those residues involved in the domain BM–carbohydrate interaction onto equivalent position in domain AM (Q6E, E23K, R24T, T25R and A92E). In addition, several more changes were introduced to avoid unfavorable contacts (K3N, N26A, G27L, G28Q, N30V, G65S and D88N). The design of the second mutant, CV-NmutDB, was less complex. Abolishing the carbohydrate binding pocket on domain BM was achieved by targeting the residues that are crucial for protein–carbohydrate interaction, changing them into alanines or glycine: E41A, N42A, T57A, R76A and Q78G.
Biophysical characterization
The importance of biochemical and biophysical properties of a mutant, such as protein stability and conformational integrity, cannot be underestimated when interpreting binding studies and biological assays (Barrientos et al., 2002a,b, 2003b, 2004b). We therefore routinely assess the structure of a protein variant by NMR spectroscopy. For both mutants, CV-NmutDA and CVNmutDB, the 1H–15N HSQC NMR spectra exhibited all the characteristics of the wild-type CV-N fold, confirming that the overall structures of these mutants are very similar to native CV-N. In addition, all proteins used in the current study were monomeric, as assessed by multi-angle light scattering and native gel electrophoresis. Unfolding/refolding experiments for different protein concentrations indicated that both variants preferentially refolded into monomeric species at concentrations of 50 µM and below. Above 50 µM, folded monomer, domain-swapped dimer and misfolded/aggregated material were observed for CV-NmutDA, similar to previous results with wt-CVN (Barrientos et al., 2002a, 2004b); for CV-NmutDB, on the other hand, only folded monomeric and domain-swapped dimeric protein was obtained in these refolding experiments, without any noticeable misfolding and/or aggregation, reminiscent of our findings for the stabilized [P51G]CV-N mutant (Barrientos et al., 2002a, 2004b).
Misfolding and aggregation appears intimately linked with thermodynamic stability, with the latter clearly impacting the integrity of the molecule during the course of the biological assays. This connection was noted in our studies of the circular permuted variant of CV-N, cpCV-N (Barrientos et al., 2002b, 2003b). cpCV-N exhibits the same global fold as wtCV-N and is capable of sugar and glycoprotein binding, albeit with significantly lower antiviral potency (1000-fold), possibly contributed to by its considerably lowered stability (G 2 kcal/mol compared to wtCV-N). The thermal stabilities of CV-NmutDA and CV-NmutDB were assessed by CD spectroscopy and the melting temperatures are summarized in Table I. CV-NmutDA and CV-NmutDB exhibit Tm values of 58.7°C and 71.9°C, respectively, virtually identical to monomeric wild-type CV-N101aa (61.3°C) and [P51G]CV-N (71.5°C), respectively (Table I). Stability towards chaotropic agents was evaluated by GdnHCl titrations and unfolding curves were analyzed assuming a two-state model. Free energies of folding were GCVNmutDA = 4.0 ± 0.5 kcal/mol and GCVNmutDB = 9.6 ± 0.5 kcal/mol (Table I), similar to values of wt-CVN and [P51G]CV-N, respectively. Therefore, both mutants were suitable for comparative studies of antiviral activity.
|
Incidentally, the gain in thermodynamic stability for [P51G]CV-N over wtCV-N had little effect on the intrinsic antiviral potency. However, [P51G]CV-N performed better than wild-type CV-N in a number of biological assays in our hands for a wide range of viruses and cell lines (Mori et al., 2002
; O'Keefe et al., 2003; Barrientos et al., 2004b), most likely due to the more robust biochemical properties of [P51G]CV-N over the long periods of incubation at elevated temperature (37°C) and complex assay conditions. Interaction of CVNmutDA, CVNmutDB and [P51G]CV-N with trimannoside, dimannose and Man-9
The interaction of CVNmutDA with trimannoside, dimannose and Man-9 was investigated by chemical-shift mapping using 1H–15N HSQC spectroscopy, native gel electrophoresis and ITC. All experiments were carried out in parallel with [P51G]CV-N under identical conditions in order to compare binding parameters. Chemical shift titrations of CV-N with different sugars have been extensively reported and for the present purpose we selected one representative resonance from each domain (G96 for AM and N53 for BM), rather than presenting overlays of the entire spectra. Peak volumes or weighted averaged chemical shift changes were followed for increasing sugar concentrations for slow exchange or fast exchange regimes, respectively.
Titration curves for trimannoside binding to CVNmutDA, CVNmutDB and [P51G]CV-N are displayed in Figure 2. In all cases, the peak intensities from residues in both domains decreased with increasing sugar concentrations, i.e. only slow to intermediate exchange was observed. For trimannoside binding to CVNmutDA (Figure 2A), no difference between resonances associated with domains AM and BM was noted. This is in contrast to the observation with [P51G]CV-N, for which G96 clearly was affected early, indicating a preferred binding to domain AM. Thus, in the CVNmutDA variant the preference for the trisaccharide Man(12)Man(12)ManOMe has been abolished and similar affinities for both domains exist. For CVNmutDB (Figure 2B), only resonances on domain AM were affected throughout the entire titration and a true, sigmoidal titration curve of G96 (domain AM) was observed, starting as early in the titration as noted with [P51G]CV-N. This behavior is a reflection of the specificity for the trimannoside in this domain. Even for an excess up to six equivalents sugar over protein, no change in any resonance of domain BM was seen. We therefore are confident that the sugar binding site on domain BM has been completely abolished (at least for micromolar concentrations) in this variant and that the only interactions left involve residues in domain AM. Interestingly, conformational heterogeneity was noted for resonances from domain BM (peak doubling) for both [P51G]CV-N and CV-NmutDA, suggesting that prior to saturation of the two sites on the protein, the two sets of (1-2) linked dimannose units in trimannoside can interact with similar affinity on domain BM, resulting in slightly different chemical shifts for the 1:1 and 1:2 protein:trimannoside complex.
|
The titration of CVNmutDA and [P51G]CV-N with dimannose revealed that initially (up to 0.5 equivalent), only resonances in domain BM of CVNmutDA were strongly affected and slow exchange between free and sugar-bound protein was observed. The binding curves for dimannose interaction are provided in Figure 3. At higher sugar concentrations (>0.5 equivalent) also resonances in domain AM experienced changes, albeit in the fast to intermediate exchange regime, reaching saturation at
2 equivalents, clearly demonstrating tighter binding of this sugar to domain BM as compared to domain AM. Essentially identical behavior was observed for [P51G]CV-N, indicating preferential binding of the Man(12)Man dimannose to domain BM for both proteins.
|
Titration of CVNmutDA with Man-9 by NMR, even at very low protein concentration (20 µM), resulted in extreme line broadening and the ultimately ‘disappareance’ of resonances, similar to our previous observations with [P51G]CV-N. However, in contrast to the [P51G]CV-N interaction, which was accompanied by precipitation of sugar–protein complexes due to multisite/multivalent cross-linking (Shenoy et al., 2002
), here, at concentrations of 20 µM and below, no precipitation was observed. We further investigated complex formation between CV-N variants and Man-9 using native gel electrophoresis (Figure 4). Whereas for [P51G]CV-N, at a concentration of 8.7 µM, protein is no longer observed in the soluble fraction after equimolar addition of Man-9 (Figure 4A), for CV-NmutDA, even after addition of a 15-fold excess of sugar, no loss of protein is apparent, and a band consistent with an average monomeric CVNmutDA/sugar complex is observed (Figure 4B). This suggests that the severe line broadening for resonances in the NMR spectra of CV-NmutDA upon Man-9 addition is caused by exchange between the sugar bound in either of the two protein carbohydrate recognition sites. Figure 5 schematically depicts these various possibilities. This finding of multiple modes of mannose binding is similar to results obtained for oligomannose complexes with the neutralizing antibody 2G12 (Calarese et al., 2005).
|
|
Titration of CVNmutDB with Man-9 (see Supplementary Figure S1 available at PEDS online) resulted in almost identical effects in the HSQC spectra as seen with the trimannoside (Figure 2B), indicating that only the binding site on domain AM was involved in sugar binding. Interestingly, even a 12-fold molar excess of Man-9 did not result in further changes in the spectrum and a superposition of the two spectra, with and without Man-9, is provided in Figure 6 (20 µM protein concentration). This suggests that no significant binding site with micromolar affinity is present on domain BM in CVNmutDB, leaving the site on domain AM as the sole site for interaction. Relaxation data supports the notion that the overall size of the complex is in the 10–13k Da range, confirming the 1:1 stoichiometry. Supporting evidence is provided by native gel electrophoresis, in which no higher molecular weight complexes were observed (Figure 4C). Overall, the CV-NmutDB titrations represent the simplest case—only binding to domain AM is possible, whereas for CV-NmutDA the dimannose and trimannoside can bind to domain BM and domain AM. For Man-9, again only domain AM is available for the interaction in CV-NmutDB, although different binding modes may be possible at varying concentrations of both protein and glycan (EM and AMG, data not shown).
|
Given the fact that for CVNmutDB only one sugar binding site was observed by NMR, we carried out isothermal titration experiments with Man-9 to determine its binding constant. In contrast to previous results with wild-type and [P51G]CV-N, for which Man-9 interaction resulted in aggregation and precipitation of sugar–protein complexes (Shenoy et al., 2002
), no such effects were observed for this variant, rendering this molecule an ideal analog for structural and thermodynamic investigations. Calorimetric titrations revealed that binding was driven by enthalpic contributions (negative H value), as determined from the observed isotherm (Figure 7) with a strong unfavorable entropic contribution (negative TS values; G = H – TS), consistent with losses in rotational, translational and conformational freedom of the oligosaccharide and/or CV-N upon complex formation. Analysis of the binding isotherm yielded a G value of –7.4 kcal/mol (Table II) with a satisfactory fit to a one-site model. The recovered parameters were H = –11.1 kcal/mol and an equilibrium dissociation constant Kd = 4.3 + 0.3 µM for the Man-9:CVNmutDB interaction. Indeed, overall the enthalpy of binding was favorable enough to compensate for the unfavorable binding entropy, resulting in moderately tight binding, similar to findings with subfragments of Man-9 and wild-type CV-N. We also performed the equivalent ITC experiment with CVNmutDA and Man-9. Due to the presence of exchange in a multisite/multivalent fashion between the protein and the glycan (see schematics in Figure 5), we did not extract binding parameters. However, no indication for nanomolar binding was observed. (EM and AMG, data no shown).
|
|
Interaction of CV-NmutDA, CV-NmutDB and [P51G]CV-N, with viral glycoproteins
The binding of CVNmutDA and CVNmutDB to the HIV env glycoproteins gp120 was probed by ELISA and compared with binding by monomeric [P51G]CV-N (Table III). These experiments were carried out with commercially available HIV-1 gp120, for which the presence of both Man-8 and Man-9 had been confirmed. Essentially identical curves were observed for CVNmutDA and [P51G]CV-N with midpoints at 4 nM protein concentration. For the CVNmutDB variant, however, no gp120 binding was observed below 1 µM. These results prove that (i) the binding to HIV-1 gp120 requires an intact binding site on domain BM, (ii) the loss of the trimannoside preference of domain AM in CVNmutDA has no detrimental effect on the gp120 interaction and (iii) the sugar microenvironment (clustering) is crucial for the high-affinity interaction between HIV-1 gp120 and CV-N, since complex formation is observed in the ELISA assays for CVNmutDA, despite its low micromolar affinity binding to Man-9.
|
We also investigated whether structural protein determinants on HIV-1 gp120 contribute to the apparent high-affinity interaction between the glycoprotein and CV-N. The glycoprotein was unfolded, reduced and carboxymethylated, thereby ensuring that no native, folded three-dimensional architecture remained. No differences in the ELISA data for native, intact gp120 and the denatured glycoprotein binding by [P51G] CV-N was observed, clearly confirming that the carbohydrates on gp120 are the predominant target of CV-N (Table III).
Binding of Ebola glycoprotein (EboZV GP1-Fc) to all three CV-N variants was assessed by FACS assays (Figure 8). Increasing amounts of [P51G]CV-N dramatically enhanced GP1-Fc binding to Jurkat cells (a cell line nonpermissible for infection), whereas no such effects were observed with CV-NmutDA or CV-NmutDB. Therefore, this effect seems to be governed by a distinct modality, different from that crucial for antiviral activity, and may involve mannosylated cell surface proteins interacting with the binding site on domain AM at higher protein concentrations. Similar observations were noted using HIV-1 gp120-Fc [data not shown; and by Hong et al. (2002) on studies with wtCV-N and gp120-Fc].
|
Antiviral activity of CV-NmutDA, CV-NmutDBand [P51G]CV-N
Purified proteins were tested side by side in HIV-1 infection and EboVZ GP-mediated transduction assays. For anti-HIV activity, HIV-1 Ba-L virus and PBMC cells were used and the percentage of cells positive for p24 was determined by FACS analysis. As shown in Table IV, [P51G]CV-N (EC50 of 6.25 nM) exhibited the highest potency, with CV-NmutDA less effective by a factor of 4 (EC50 of 25 nM). In striking contrast, CV-NmutDB lacked any detectable activity. Likewise, the in vitro potency with respect to anti-Ebola activity using EboZV GP-pseudotyped virus in HeLa cell transduction revealed EC50 values of 40.5 ± 21.5 nM and 200 ± 50 nM, respectively, for [P51G]CV-N and CV-NmutDA. Again, no detectable activity was observed for CV-NmutDB. Therefore, within the experimental error of these cellular assays, [P51G]CV-N and CV-NmutDA exhibit comparable antiviral activities. The ca 4- to 5-fold difference may simply reflect the differences in stability of the CV-NmutDA variant compared with [P51G]CV-N under assay conditions.
|
Concerns about problems associated with cross-linking of virus to the cell surface that could potentially enhance infectivity via a similar mechanism used by DC-SIGN, prompted us to further investigate this issue. Infection assays (with HeLa cells) were performed in the presence of large amounts of CV-N (protein concentrations of 1 and 10 µM) and inhibition of infection was still observed. Therefore, within the low micromolar range and within the therapeutic window (<1 µM) no such detrimental side effects are apparent.
Model of CV-N binding to glycans and viral glycoproteins
In order to derive a comprehensive and consistent picture of CV-N/glycan interactions it seems prudent to summarize all previously published results and integrate these data into our current model. Early experiments aimed at measuring equilibrium binding constants for CV-N binding to Man-8 and Man-9 were curtailed by aggregation and precipitation of the protein–carbohydrate complexes, preventing the extraction of useful data. Employing nonamanoside and hexamannoside sugars, that lack the chitobiose moiety at the reducing end of the oligosaccharide structure, prevented precipitation, although we clearly observed a multisided/multivalent interaction between CV-N and the nonamannoside (Shenoy et al., 2002). The resulting complexes were polydisperse with an average size of 22 KDa as determined by NMR relaxation experiments and analytical ultracentrifugation, and exchange between components was evident by NMR. Native gel electrophoresis revealed two bands of equal intensity at sizes of 11 KDa and 22 KDa, confirming the monomer–dimer equilibrium for the proteinaceous components in the protein:sugar complex. One possible explanation for the observed difference in behavior between Man-9 and nonamannoside may be realated to a difference in flexibility of the D1 and D3 arms of the sugar, such that in the nonamannoside, the trimannose and dimannose moieties behave more like free trimannose and dimannose units (Shenoy et al., 2002). Therefore, domains AM and BM can bind the trimannose (D1 arm) and dimannose (D3 arm) units similar to the free sugars, with binding constants in the micromolar range, weakly enough to prevent tight aggregate formation. In contrast, the case of Man-9, the D1 and D3 arms may be in a more rigid conformation and binding can lead more readily to cross-linking and precipitation. Alternatively, a very week interaction with the two GlcNac units in Man-9 could exist, too weak to be observed with the free chitobiose units, but adding to the overall affinity when linked to the nonamannoside unit in Man-9. We also note that domain AM's elongated cleft provides a better fit for three stacked rings (Man12Man12Man); therefore, explaining the slight preference for (12) linked trimannoses. In the case of domain BM, only two stacked sugar rings can be accommodated into the deep binding pocket, irrespective of the linkage between the second mannose and a third unit (Bewley and Otero-Quintero, 2001; Bewley et al., 2002; Shenoy et al., 2002), and therefore two distinct complexes in the trimannoside titration can be formed prior to saturation (e.g. doubling of resonances from domain BM as discussed above).
How do the binding data fit the activity profile? All data up to now indicate that recognition of high-mannose sugars such as Man-8, Man-9 or derivatives thereof by CV-N, involves (12) linked dimannose and trimannose epitopes on both, the D1 and D3 arms, with the dimannose unit being the smallest unit within the sugar moiety necessary for interaction with CV-N. Indeed, the hexamannoside contains only a single dimannose unit and binding to CV-N occurred exclusively to domain BM in a 1:1 complex (Shenoy et al., 2002). Therefore, in the context of a brached oligomannose, it appears that domain BM interacts preferentially with the dimannose end of the glycan. The mutational work by Chang and Bewley (2002) and our present work now further dissect the protein side of the interaction. Abolishing the site on domain BM with domain AM intact, clearly abrogates activity (as presented here), whereas the equivalent disruption in domain AM seemed to have no effect (Chang and Bewley, 2002). In addition, our results show that elimination of preferential binding of the trimannose unit on domain AM does not abolish overall binding or antiviral activity, although cross-linking at high protein concentration is prevented. Therefore, the observed cross-linking and precipitation of complexes that occurs in vitro at micromolar concentrations, is most likely not related to antiviral activity since it is only observed outside the therapeutic window.
How can the observation of micromolar affinities for free glycans be reconciled with nanomolar viral glycoprotein interactions and antiviral activity? Most in vitro binding studies were carried out with oligosaccharides devoid of the GlcNAc moieties, and although no binding to CV-N was detected for individual GlcNAc units, one cannot exclude the possibility that in the context of Man-8 or Man-9 these units exert modulating effects. Indeed, the nonamannoside appeared to interact less strongly with CV-N than Man-9, resulting in soluble aggregates, rather than precipitate (Shenoy et al., 2002). Therefore, conformational variability and flexibility of the terminal sugar units of the D1 and D3 arms of Man-8 or Man-9 may possibly be reduced by the addition of a GlcNAc unit in the free sugars and certainly in the bound form by attachment to a viral glycoprotein. Such conformational restriction would decrease the unfavorable binding entropy, resulting in tighter overall binding.
For HIV-1 gp120, sugar mapping experiments indicated that Man-8 and Man-9 are present on the glycoprotein. Although no quantitative data as to the relative proportions of these two sugars are available, it is likely that the degree of Man-9 and Man-8 modification as well as the clustering of carbohydrates in three-dimensional space may play the most important role in CV-N envelope glycoprotein interaction. Indeed, HIV-1 Env gp120 is among the most heavily glycosylated proteins in nature, far more heavily glycosylated than envelopes of other retroviruses of similar size (e.g. HTLV-1, MuLV) or cell surface glycoproteins. This explains the high efficacy of CV-N in HIV inactivation.
|
Conclusions |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Based on all available data it appears that the presence of an intact sugar binding site on domain BM is necessary for CV-N/viral glycoprotein interaction and antiviral potency of CV-N. Complete loss of activity ensues upon obliteration of the domain BM binding site (leaving domain AM intact) and no detrimental effect in activity is seem upon altering the binding site in domain AM (leaving domain BM intact). Nanomolar antiviral potency appears to be related to a specific, possible more rigid orientation of the Man
12Man units at the termini of high oligomannoses and/or a particular positioning within the glycan clusters of HIV-1 gp120 and EboZV GP. The precise spatial arrangement of the glycans within the glycan cluster of the heavily glycosylated viral glycoproteins appears to define the functional grouping that allows to recognize these glycoproteins with high specificity and avidity. By binding to the sugars on the viral surface, CV-N might (i) mask the carbohydrates attached to the region of the glycoprotein that is important in the early stages of viral–cell fusion or (ii) physically interfere with receptor and co-receptor binding sites in the postattachment steps of viral entry or (iii) lock the glycoprotein into a fusion-incompetent state, thereby rendering it unable to undergo the crucial conformational changes necessary for interaction with the co-receptors. It is unlikely that CV-N's mode of action includes cross-linking/precipitation of virus in the pre-incubation periods, and indeed no precipitation of virus was observed upon incubation of virus stock solutions with CV-N. Further research will clarify and help guide our ongoing attempts to develop CV-N into a more effective microbicidal agent.
|
Footnotes |
|---|
Edited by Flemming Poulsen
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
We thank Pierre Rollin (NCID) for his continued support and helpful discussions. We are grateful to Daniel M.Ratner and Peter H.Seeberger for supplying the original trimannoside sample. We thank Laurel Lagenaur (NIAID) and Qiang Xu (OSEL, Inc.) for kindly providing independent anti-HIV activity for our CV-N variants using an IC p24 assay (Table IV) and John Louis (NIDDK) for generous advise, expertise and discussions. This research was supported by the Intramural AIDS Targeted Antiviral Program of the Office of the Director of the National Institutes of Health (to A.M.G.) and G03173 [GenBank] and PI030300 from Fondo de Investigación Sanitaria (FIS-ISCIII) (to R.D.). L.G.B. was supported in part by the Research Participation Program at the Centers for Disease Control and Prevention, National Center for Infectious Diseases, Division of Viral and Rickettsial Diseases administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the US Department of Energy and CDC.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Barrientos L.G. and Gronenborn A.M. (2002) Biochem. Biophys. Res. Comm. 298:598–602.[CrossRef][ISI][Medline]
Barrientos L.G. and Gronenborn A.M. (2005) Mini Rev. Med. Chem. 5:21–31.[ISI][Medline]
Barrientos L.G., Louis J.M., Botos I., Mori T., Han Z., O'Keefe B.R., Boyd M.R., Wlodawer A., Gronenborn A.M. (2002a) Structure 10:673–686.[Medline]
Barrientos L.G., Louis J.M., Hung J., Smith T.H., O'Keefe B.R., Gardella R.S., Mori T., Boyd M.R., Gronenborn A. M. (2002b) Proteins 46:153–160.[CrossRef][ISI][Medline]
Barrientos L.G., O'Keefe B.R., Bray M., Sanchez A., Gronenborn A.M., Boyd M.R. (2003a) Antiviral Res. 58:47–56.[CrossRef][ISI][Medline]
Barrientos L.G., Louis J.M., Ratner D., Seeberger P., Gronenborn A.M. (2003b) J. Mol. Biol. 325:211–223.[CrossRef][ISI][Medline]
Barrientos L.G., Lasala F., Otero J.R., Sanchez A., Delgado R. (2004a) J. Inf. Dis. 189:1440–1443.[CrossRef][ISI][Medline]
Barrientos L.G., Lasala F., Delgado R., Sanchez A., Gronenborn A.M. (2004b) Structure 12:1799–1807.[Medline]
Bewley C.A. (2001) Structure 9:931–940.[Medline]
Bewley C.A. and Otero-Quintero S. (2001) J. Am. Chem. Soc. 123:3892–3902.[CrossRef][ISI][Medline]
Bewley C.A., Gustafson K.R., Boyd M.R., Covell D.G., Bax A., Clore G.M., Gronenborn A.M. (1998) Nat. Struct. Biol. 5:571–578.[CrossRef][ISI][Medline]
Bewley C.A., Kiyonaka S., Hamachi I. (2002) J. Mol. Biol. 322:881–889.[CrossRef][ISI][Medline]
Botos I., O'Keefe B.R., Shenoy S.R., Cartner L.K., Ratner D.M., Seeberger P.H., Boyd M.R., Wlodawer A. (2002) J. Biol. Chem. 277:34336–34342.
Boyd M.R., et al. (1997) Antimicrob. Agents Chemother. 41:1521–1530.[Abstract]
Calarese D.A., Lee H.-K., Huang C.-Y., Best M.D., Astronomo R.D., Stanfield R.L., Katinger H., Burton D.R., Wong C.-H., Wilson I.A. (2005) Proc. Natl Acad. Sci. USA 102:13372–13377.
Chang L.C. and Bewley C.A. (2002) J. Mol. Biol. 318:1–8.[CrossRef][ISI][Medline]
Delaglio F., Grzesiek S., Vuister G.W., Zhu G., Pfeifer J., Bax A. (1995) J. Biomol. NMR 6:277–293.[ISI][Medline]
Dey B., Lerner D.L., Lusso P., Boyd M.R., Elder J.H., Berger E.A. (2000) J. Virol. 74:4562–4569