B. Luy, Institut für Organische Chemie und Biochemie der Technischen Universität München, Lichtenbergstr. 4, D-85747 Garching, Germany. Fax: + 49 89 28913210, Tel.: + 49 89 28913275, E-mail: Burkhard.Luy@ch.tum.de and M. K. Gilson, Center for Advanced Research in Biotechnology, 9600 Gudelsky Drive, Rockville, MD 20850, USA. Fax: + 1 301 7386255, Tel.: + 1 301 7386217, E-mail: Gilson@umbi.umd.edu
The pulmonary surfactant prevents alveolar collapse and is required for normal pulmonary function. One of the important components of the surfactant besides phospholipids is surfactant-associated protein C (SP-C). SP-C shows complex oligomerization behavior and a transition to β-amyloid-like fibril structures, which are not yet fully understood. Besides this nonspecific oligomerization, MS and chemical cross-linking data combined with CD spectra provide evidence of a specific, mainly α-helical, dimer at low to neutral pH. Furthermore, resistance to CNBr cleavage and dual NMR resonances of porcine and human recombinant SP-C with Met32 replaced by isoleucine point to a dimerization site located at the C-terminus of the hydrophobic α-helix of SP-C, where a strictly conserved heptapeptide sequence is found. Computational docking of two SP-C helices, described here, reveals a dimer with a helix–helix interface that strikingly resembles that of glycophorin A and is mediated by an AxxxG motif similar to the experimentally determined GxxxG pattern of glycophorin A. It is highly likely that mature SP-C adopts such a dimeric structure in the lamellar bilayer systems found in the surfactant. Dimerization has been shown in previous studies to have a role in sorting and trafficking of SP-C and may also be important to the surfactant function of this protein.
The liquid–air interface in the alveoli of mammalian lungs is coated with a surfactant monolayer that reduces surface tension and thus opposes collapse of the alveoli . The surfactant is composed of lipids (≈ 90%) and the surfactant-associated proteins SP-A, SP-B, SP-C and SP-D. An abnormally low level of surfactant, notably in preterm infants, is associated with respiratory distress syndrome [2,3], a condition that can be ameliorated by instillation of surfactants into the lungs.
The small hydrophobic proteins SP-B and SP-C are important components of the pulmonary surfactant. Thus, SP-B deficiency causes fatal respiratory distress syndrome , and SP-C deficiency in humans is associated with childhood lung disease [5,6] and fatal respiratory distress syndrome . It is also of interest that the therapeutic surfactants now on the market that include SP-B and SP-C are regarded as more effective than those that are purely lipidic . On the other hand, the animal-derived surfactant proteins in current therapeutic preparations pose concerns about immunogenicity and transmission of disease, and efforts are under way to develop novel surfactant replacements with an improved protein component . As a consequence, there is considerable interest in the details of how SP-B and SP-C stabilize the alveolar surface.
The mechanisms by which pulmonary surfactants act are still not fully understood. Current thinking is that their actions are related to multilayer lipid–protein structures that underly the surface monolayer. These subsurface structures grow and contract as material leaves the monolayer during expiration and re-enters it during inspiration . The mechanism of removal presumably involves a process in which the lipidic surface of the monolayer, which is exposed to air when the lungs are expanded, folds against itself and dips below the surface on expiration. SP-B and SP-C appear to play a role in the formation and stabilization of these dynamic, subsurface reservoirs of lipids [10–12]. Experiments on peptides derived from SP-B and SP-C [13–15] and full-length proteins  also provide evidence that the two molecules promote formation of a fluid phase in the monolayer with a net-like topology that isolates patches of a more rigid phase and inhibits alveolar collapse. The molecular basis for these effects is still unclear, however, and development of structural information for SP-B and SP-C is of central importance in establishing their mechanism.
SP-C is a 34–35-residue polypeptide the sequence of which (Table 1) can be viewed as consisting of four parts [17,18]: a palmitoylation motif with two surrounding prolines at residues 3–6; two highly conserved cationic residues at positions 10 and 11; a rigid, totally hydrophobic α-helical segment with regularly stacked side chains spanning residues 12–27; and a less hydrophobic, α-helical, C-terminal heptapeptide which is strictly conserved across species. Although all four parts seem to be necessary for the function of SP-C , the only component that has been assigned a clear role is the hydrophobic α-helix, which is believed to anchor SP-C in the lipidic surfactant layer. Here we propose that the role of the C-terminal segment of SP-C is to permit dimerization of SP-C via an AxxxG helix interaction motif similar to the GxxxG motif found in the membrane-spanning helix dimer glycophorin A (GpA) [19,20].
Table 1. Sequences of SP-C variants, and partial sequences of human GpA, Neu and Ros-C. Asterisks indicate residues defining the AxxxG or GxxxG motif in SP-C (A29, G33) and other transmembrane helices.
This study was motivated by the similar observation of two distinct sets of NMR resonances for the C-terminal residues 28–34 of porcine SP-C  and recombinant rSP-C (FFI)  (Table 1) in an organic solvent that mimics the lipidic surfactant environment. Although the second set of resonances in porcine SP-C was originally thought to result from a second oxidation state of the sulfur of Met32 , this cannot explain the existence of dual resonances for rSP-C (FFI) because it lacks Met32 (Table 1). As, in the case of rSP-C (FFI), the relative intensities of the two sets of resonances are clearly concentration dependent, it is hypothesized that the dual resonances result instead from the coexistence of a monomeric and dimeric form of SP-C with its monomer–monomer junction near the C-termini of two peptides. Although such dimerization may be mediated by hydrogen bonds between the C-terminal carboxy groups, which are expected to be neutral under the experimental conditions, the typical hydrogen exchange rates for carboxy groups appear to be inconsistent with this model . It therefore seems more likely that dimerization is mediated by some other type of helix–helix interface. However, this hypothesis proved difficult to explore based on experimentally obtained structural data because it was impossible to assign intermolecular connectivities between the two subunits of the putative dimer because of extensive overlap in the homonuclear NMR spectra and to the weakness of the NOE signal for molecules of this size. We have therefore used a computational method to explore the possible dimerization of SP-C. The suitability of a computational approach is supported by the success of previous computational modeling of the GCN4 leucine zipper  and GpA .
Materials and methods
A fast code for conformational optimization [24,25] was used to carry out extensive searches for the lowest energy conformation of a homodimer of rSP-C (FFI) (Table 1). In accordance with experimental data [17, 21, 26], the calculations treated the peptide as α-helical. Because the data suggested a dimer interface at the C-terminus, residues 1–15 were omitted, decreasing the number of residues to 19 in each α-helical monomer. The artificial N-terminus of the shortened peptide was modeled as un-ionized, and the C-terminus was also treated as neutral, based on the acidic conditions of the NMR sample. An extensive conformational search was carried out with the Chemistry at Harvard Macromolecular Mechanics  force field and a simplified but time-efficient distance-dependent dielectric model (ε = 4rij). The calculations treat one monomer as fixed in space, and the energy of the system is minimized with respect to the position and orientation of the other monomer, along with selected bond rotations in both monomers (see below). The center of mass of the moving monomer was allowed to sample a large range of positions defined by a box of dimensions 36 × 36 × 32 Å, centered on the fixed monomer and with the fixed helix directed along the z-axis. The backbone was kept rigid in this initial search, but all torsional angles in side chains Ile22 to the C-terminal Leu34 were treated as rotatable for both monomers because the NMR restraints of these side chains allow some conformational flexibility. During the search a total of 6 × 106 conformations were generated and used to refine 500 low-energy candidates for the structure of the SP-C dimer. The distribution of energies of the 500 low-energy structures is shown in Fig. 1. The most stable structure is significantly separated from the rest by a marked energy step of 4.5 kcal·mol−1.
The optimum structure was refined via a more focused search with a comparatively realistic treatment of electrostatics. This second search was intentionally guided toward the optimum from the first search by using the fact that the search algorithm stores a list of the low-energy conformations found to date and can use subsets of their stored coordinates in generating new conformations to be tested during the optimization procedure. The desired bias was thus established by inserting the optimum conformation from the first search into the list of found conformations at the outset of the second search. The monomers were afforded greater flexibility in this search: in addition to the flexible dihedrals of the previous optimization, the backbone φ,ψ angles of residues Leu31, Ile32, Gly33 and Leu34 were allowed to vary by ± 15 ° from the original structure because these residues are less well-defined by the NMR data: they form the last turn of the α-helix where the amide protons are not hydrogen-bonded and dα,N(i,i+3) and dα,N(i,i+4) connectivities are absent. However, to somewhat restrict the number of variables in the calculation, valines 23, 24 and 27, which point away from the dimer interface and which shifted little in the results of the initial search, were locked into their optimized conformations. The total number of rotatable dihedrals in both monomers came to 44. The Generalized Born electrostatics model [28–30] was used with a solvent dielectric constant 4 and a molecular dielectric constant of 1.
In further studies, Ala29 was replaced by a Gly residue to assess the effect of replacing AxxxG with the recognized GxxxG dimerization motif. Almost the same procedure as described above was used for the docking study, now allowing side chain flexibility only for residues pointing towards the dimer interface side. The resulting dimer, with the artificial GxxxG motif, is basically identical to the original calculations.
Finally, to determine if wild-type SP-C, which has Met instead of Ile at position 32, can adopt the dimer structure found here for the FFI variant, the same technique was used to optimize the structure of a wild-type SP-C dimer with Generalized Born electrostatics. The resulting geometry is virtually the same as in rSP-C (FFI) (data not shown). In addition, the Met32 side chains of each monomer do not contact the other monomer and hence do not contribute to the binding interface, so no special role could be identified for this residue in the formation of the SP-C dimer.
Results and Discussion
Structural model of the SP-C dimer
Figure 2A provides an overview of the modeled structure of the SP-C dimer that results from the extensive conformational search described in Materials and methods; for comparison, the experimentally determined structure of GpA is also shown. Interestingly, although no symmetry was imposed during the calculations, the modeled dimer is highly symmetrical. The two helices are topologically parallel, with a right-handed crossing and a helix–helix angle of 44 °, as computed with the program interhlx (K. L. Yap, University of Toronto, Toronto, Canada). This angle is consistent with the experimental observation of a 24° angle between the SP-C helix and the normal axis of a membrane bilayer  if the membrane normal is assumed to bisect the angle of the helix–helix dimer. The dimer interface appears to be stabilized by a combination of factors. First, as shown in Fig. 3A, the two monomers are linked by six Cα–H··O hydrogen bonds [32–36] or Cα–H··O contacts, as previously defined : Ala29(O)A–Leu30(Hα)B, Leu30(O)A–Gly33(Hα)B and Gly33(O)A–Leu34(Hα)B and their symmetric pairs. The monomers also pack intimately, forming a serpentine interface of close van der Waals contacts, as highlighted in Fig. 4. Finally, two pairs of peptide groups in monomer A are positioned with their carbonyl carbons directly across from the amide nitrogens of the corresponding peptide groups in monomer B, creating the possibility of attractive electrostatic interactions , as follows: Gly33A–Leu34B, 3.53 Å; Gly33B–Leu34A, 3.59 Å; Ala29A–Leu30B, 4.22 Å; Ala29B–Leu30A, 4.30 Å. Interestingly, the side chain of Ile32 in the dimer is significantly repositioned relative to the monomer structure obtained by NMR studies [17, 21]; this result is consistent with the experimental observation in NMR studies that Ile32 experiences significant changes in side-chain chemical shifts in the two sets of resonances mentioned above .
Although no information on the structure of GpA was used in the calculations, the interface of the modeled SP-C homodimer strikingly resembles that of the GpA homodimer, which has been determined experimentally [19,20], as illustrated in Fig. 3B. In particular, the GpA interface possesses six Cα–H··O hydrogen bonds precisely analogous to those in SP-C: Gly79(O)A–Val80(Hα)B, Val80(O)A–Gly83(Hα)B and Gly83(O)A–Val84(Hα)B and their symmetric pairs. As previously noted, the association of the GpA helices is mediated by a GxxxG motif in each helix (Gly79, Gly83) [19,20]. In SP-C, the corresponding pattern is AxxxG (Ala29, Gly33). Initially, we thought that the smaller helix–helix angle of GpAs (35 ° vs. 44 °) appeared to be attributable to the smaller size of Gly79 in GpA relative to Ala29 in SP-C. However, docking studies using monomers with Ala29 replaced by a Gly residue resulted in an interhelical angle of 43.2 °, which is virtually identical with the original angle determined for the SP-C dimer. The helix–helix interface must therefore be mainly determined by surrounding residues at the dimerization site, as is the case with Thr87 in GpA, which probably forms a hydrogen bond to the backbone carboxy group of Val84 .
Another difference between SP-C and GpA is that the helix–helix interface of SP-C is very near the C-termini of the helices, whereas the interface of GpA is relatively central (Fig. 2). The stability of the two dimers therefore seems quite different. Owing to its position at the C-terminus, the contact surface of the SP-C dimer is considerably smaller than the GpA dimer and also lacks the equivalent to the Thr87–Val84 hydrogen bond. Furthermore, the dimer of GpA is stabilized in a membranous bilayer by its hydrophilic ends, whereas the C-termini of SP-C are just buried in the hydrophobic core. All this is reflected by SDS/PAGE experiments in which the GpA dimer is clearly visible, and SP-C without additional chemical cross-linking appears to be monomeric (Fig. 8C in ). Besides the reduced stability, the C-terminal position of the interface in SP-C along with the relatively large interhelical angle causes SP-C to have a much more V-shaped appearance than the more compact X-shaped GpA, as seen in Fig. 2A,C.
The GxxxG and AxxxG motifs belong to one of the two basic types of helix–helix contacts recently identified in membrane proteins . Here, small residues, especially Gly, Ser, Ala and Thr, create smooth surfaces which allow the close approach of the helices' backbones. The importance of such contacts is supported by statistical analyses showing enrichment of small residues at helix–helix interfaces [39,40], with an especially high occurrence of G–G contacts . Moreover, the GxxxG motif emerged spontaneously in an experimental system where dimerization was applied as a selection criterion for randomized transmembrane helices , and GxxxG was also found to be overrepresented in protein segments identified as transmembrane helices by sequence analysis . The GxxxG motif, as well as variants in which the first G is replaced by other small amino acids, appear to play a role in the dimerization of epidermal growth factor receptors  and the receptor tyrosine kinases Neu and Ros-C (chicken) [44–46], and also in the function of the αIIbβ3 integrins . Interestingly, GxxxG, AxxxA and GxxxA sequences have recently been implicated in helix–helix contacts in water-soluble proteins [48,49].
Experimental correlations and functional implications
The natural environment of SP-C is the surfactant, a complicated structure of monolayer and lamellar bilayer systems consisting mainly of the phopholipids dipalmitoylphosphatidylcholine (DPPC) and dipalmitoylphosphatidylglycerol. This environment favors the formation of a dimeric structure in several ways. For one thing, the bulk concentration of SP-C in the surfactant appears to be of the same order of magnitude as the ≈ 1 mm concentration used in the NMR studies [21, 26]. [Given the weight percentage of SP-C in surfactant (1%), and the specific volume of DPPC in a membrane (about 1000 Å3), the bulk concentration of SP-C in the membrane is about 2 mm.] The concentration of SP-C is even higher in surfactant prepared by lung lavage and in several therapeutic preparations . Furthermore, it is likely that SP-C molecules in the surfactant are oriented with their positively charged residues (10 and 11) close to the lipids' polar head groups and their hydrophobic α-helices pointing into the lipidic part of the layer. This orientation positions the AxxxG dimerization motifs near each other, increasing their local concentration relative to bulk and thereby increasing their tendency to dimerize. The palmitoylated residues Cys4 and Cys5 of wild-type SP-C are probably situated near the polar head groups as well, in which case the relatively short fatty acid chains cannot interfere with the proposed dimerization site but may even support the dimer formation by filling in the gap between the two monomeric units.
Part of the surfactant is believed to exist as a lipid monolayer, and in vitro studies of SP-C in lipid monolayers reveal that the main α-helix of SP-C adopts a tilt angle of ≈ 70 ° relative to the membrane normal. It seems possible that SP-C still exists as a dimer under these conditions, as tilting the entire structure may just bury the hydrophobic helix in the lipid's acyl chains, but it is also likely that the dimer is disrupted.
SP-C is known to oligomerize under various conditions. Higher-order aggregates, however, are usually of β-amyloidogenic structure and should not be confused with the specific α-helical dimer presented in this article. A number of experimental studies provide evidence for a dimeric form of SP-C. Indeed, a decade ago direct evidence for dimerization of human wild-type SP-C was provided by electrospray ionization MS data based on the detection of specific odd-charged molecular ions ([M + 5H]5+) [51,52]. In addition, cross-linking studies on mature SP-C using bismaleimidohexane show a distinct dimer at pH 7.4 (Fig. 8C of ). Finally, in very recent studies a specific dimer could be unambiguously identified by high-resolution Fourier-transform ion cyclotron resonance MS and light-scattering methods (, A. Seidl, G. Maccarone, N. Youhnovski, K. P. Schaefer and M. Przybylski, unpublished data). It was also found that the dimer appears only at low to neutral pH and is mainly α-helical (verified by CD spectra), whereas tetramers and higher-order oligomers of nonhelical conformation were only detected at higher pH. Evidence that the binding site of the α-helical dimer is positioned close to the C-terminus can be found in the resistance of Met32 to CNBr cleavage  and the concentration-dependent dual set of NMR resonances of the C-terminal heptapeptide segment already mentioned . All experimental results fit well with the computational homodimeric model described in this work.
It is interesting to speculate on the possible significance of SP-C dimerization for its role in the pulmonary surfactant, in addition to the apparent role of homomeric association in trafficking . One possibility is that the ‘V’ shape of the dimer allows it to stabilize the membrane curvature required to form the multilamellar structures that underly the surfactant monolayer (e.g.  and references therein). Also, the specific shape of the SP-C dimer may be important in promoting the selective squeeze-out of non-DPPC lipids during surface film compression , or in the formation and 2D patterning of the liquid-expanded and liquid-condensed phases observed in surfactant preparations (e.g.  and references therein). Finally, it is conceivable that compression and expansion of the surface film shifts the monomer–dimer equilibrium of SP-C by mass action and that this shift buffers and stabilizes the physical characteristics of the surface.
The computational analysis of SP-C described here reveals a dimer with a helix–helix interface that strikingly resembles that of GpA, except that it is based on an AxxxG motif rather than GxxxG, and that the dimerization takes place near the C-terminus rather than in the center of a membrane-spanning helix. This result is consistent with the existence of dual chemical shifts at the C-terminus of rSP-C (FFI), which cannot be explained by alternative oxidation states of a methionine residue, and with a growing body of biophysical and biological data. In particular, recent experimental evidence strongly suggests that dimerization is important in the trafficking of SP-C. The potential that it is also important for surfactant function should be borne in mind in developing therapeutic pulmonary surfactants.
We thank S. O. Smith (State University of New York at Stony Brook) for kindly providing coordinates of glycophorin A and M. Przybylski (University of Konstanz, Germany) for providing his results on SP-C dimerization in advance of publication. This work was supported by a grant from the National Institutes of Health (GM61300). B. L. thanks the Fonds der Chemischen Industrie, the Alexander von Humboldt Foundation, and the Deutsche Forschungsgemeinschaft (Emmy Noether LU 835/1-1) for financial support.