Search tips
Search criteria 


Logo of acssdACS PublicationsThis JournalSearchSubmit a manuscript
Biochemistry. 2012 January 10; 51(1): 90–99.
Published online 2011 December 5. doi:  10.1021/bi201356v
PMCID: PMC3290515

Structural Characterization of the Caveolin Scaffolding Domain in Association with Cholesterol-Rich Membranes


An external file that holds a picture, illustration, etc.
Object name is bi-2011-01356v_0010.jpg

Members of the caveolin protein family are implicated in the formation of caveolae and play important roles in a number of signaling pathways and in the regulation of various proteins. We employ complementary spectroscopic methods to study the structure of the caveolin scaffolding domain (CSD) in caveolin-1 fragments, while bound to cholesterol-rich membranes. This key domain is thought to be involved in multiple critical functions that include protein recognition, oligomerization, and cholesterol binding. In our membrane-bound peptides, residues within the flanking intramembrane domain (IMD) are found to adopt an α-helical structure, consistent with its commonly believed helical hairpin conformation. Intriguingly, in these same peptides, we observe a β-stranded conformation for residues in the CSD, contrasting with earlier reports, which commonly do not reflect β-structure. Our experimental data based on solid-state NMR, CD, and FTIR are found to be consistent with computational analyses of the secondary structure preference of the primary sequence. We discuss how our structural data of membrane binding Cav fragments may match certain general features of cholesterol-binding domains and could be consistent with the role for CSD in protein recognition and homo-oligomerization.

Caveolae are flask-shaped and cholesterol-rich invaginations in plasma membranes found to participate in many cellular functions across various cell types.13 Their roles include lipid uptake and regulation, molecular transport, cell adhesion, and signal transduction.2,3 The caveolin family of proteins is thought to play a central role in the formation and functioning of these lipid-raft-like domains. There are three isoforms, caveolin-1 (Cav1), caveolin-2, and caveolin-3, of which Cav1 is the most common. Found in a wide array of mammalian cells, Cav1 is implicated in multiple diseases, including cancers and muscular disease.4,5 It contains cytoplasmic N- and C-terminal domains that sandwich the membrane-association domains: the caveolin scaffolding domain (CSD; residues 82–101) and the intramembrane domain (IMD) spanning roughly residues 102–134 (Figure (Figure11).1 The latter is thought to form a unique α-helical hairpin that does not completely traverse the membrane.4,6,7 These domains are assisted in their membrane-binding roles by three palmitoylated cysteines, which are not essential for localization to caveolae,810 but do appear critical for other functionality, such as the transportation of cholesterol.10

Figure 1
Primary sequence and schematic illustration of caveolin-1. Key domains are indicated: the CSD in blue, containing the cholesterol-binding CRAC motif and the residues required for binding of protein partners (underlined). Bold residues indicate sites U– ...

The most prominent domain is the CSD, as residues within it not only are responsible for membrane binding but also are critical for oligomerization, protein interactions, and cholesterol recognition.11,12 Residues within the CSD are required for oligomerization of Cav monomers into homo-oligomers of 14–16 proteins, which themselves assemble into higher-order oligomers during the formation of caveolae.11 On top of this, the F92TVT95 segment within the CSD is important for signaling, as it is required for interaction with binding protein partners (e.g., G-protein α-subunits).11,13,14 This binding event involves a consensus motif in the partner protein with high aromatic content, sometimes referred to as a caveolin-binding motif (CBM). A similar motif is also found in caveolin itself, raising the possibility of structural similarities between protein partner binding and Cav homo-oligomerization.14 Finally, the CSD is ascribed a role in the recognition and binding of cholesterol, which is highly concentrated in caveolae and needed for their formation. More precisely, formation of caveolae strictly requires tight binding of Cav to cholesterol.15 This functionality is thought to localize to a cholesterol recognition/interaction amino acid consensus (CRAC) motif2 in residues V94TKYWFYR101.16 Thus, despite its short length, this 20-residue segment appears to incorporate an array of critical functionalities in overlapping sequence elements.

Unfortunately, its molecular mechanism of action and structural features remain unclear. Many reports indicate significant but varying degrees of α-helical structure in the CSD. A number of experimental studies have reported partially or even extensively α-helical CSD.16,17 This is in part supported by computational analysis of the primary sequence.4,1720 A helical conformation has been employed to rationalize Cav1 binding partner recognition via the CSD21 as well as the cholesterol-binding of CRAC motifs in general.22 Comparisons of the various experimental studies are complicated by differing sample conditions and Cav1 constructs, some of which are rather short fragments.16,20 It has been suggested that use of overly short Cav1 fragments may have caused a loss of helicity due to peptides being truncated within secondary structure elements.17 Experimental conditions also often involve vesicular or micellar samples that lack cholesterol and may not emulate the native, highly cholesterol-rich membrane environment of the caveolae.23 In contrast to most other studies, an integrative bioinformatics study by Spisni et al. suggested an antiparallel β-hairpin for residues 84–94, based on their computational analysis and previous functional and mutational results.24

To explore some of these ambiguities regarding the structure of Cav1, particularly the CSD, we here present results from our studies of Cav1 fragments in cholesterol-rich lipid bilayer samples. Through the use of solid-state NMR (ssNMR), we examine the structure of the peptide fragments while bound to the membrane and also observe their modulation of the membrane structure and dynamics. Solid-state NMR is a powerful tool for membrane protein studies, as it is compatible with a wide range of sample compositions and allows characterization of both lipids and proteins or peptides.2527 This applies to both magic-angle-spinning (MAS) NMR measurements and “static” experiments on pelleted or aligned membranes. Indeed, previous work on shorter CSD fragments employed ssNMR for the characterization of their effect on the membrane.16 We perform static and MAS ssNMR experiments, supported with circular dichroism (CD) and Fourier transform infrared (FTIR) spectroscopy to probe selected Cav1 fragments. These include a short peptide that reflects the isolated CSD (Cav83–102) and a longer construct (Cav82–109), which incorporates part of the IMD. Previous reports suggested that the former has a mixed secondary structure but that the latter has a predominantly α-helical conformation, at least in dodecylphosphocholine (DPC) micelles.16,17 Here, we examine their structures in cholesterol-rich lipid bilayers designed to more closely mimic caveolae, which should provide a better context to probe the cholesterol-binding and oligomerization characteristics of these peptides. Our CD and FTIR data indicate an increased α-helicity in the longer peptide (in agreement with some earlier reports), but ssNMR reveals that this α-helicity is specifically located in the putative first helix of the IMD and that residues within the CSD segment adopt a β-sheet conformation even in the presence of the flanking IMD. This turns out to be consistent with structural propensities of these fragments, as well as full-length Cav1, as indicated by various primary-sequence-based secondary structure prediction algorithms. We conclude by discussing our observations in the light of previous reports, along with possible implications for the structure and function of caveolin and its multifaceted scaffolding domain.

Experimental Procedures

Sample Preparation

Appropriately Fmoc- and side-chain-protected 13C, 15N-labeled amino acids were purchased from Cambridge Isotope Laboratories (Andover, MA) and Isotec (Sigma-Aldrich, St. Louis, MO). Site-specifically 13C, 15N-labeled and unlabeled peptides (Table S1) were synthesized by Fmoc solid-phase chemistry and purified via reversed phase HPLC to greater than 95% purity by the W.M. Keck Facility at Yale University and New England Peptides (Gardner, MA). 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) was obtained from Avanti Polar Lipids (Alabaster, AL), and cholesterol was obtained from Sigma-Aldrich (St. Louis, MO); both were used without further purification. Membrane samples containing a 1:1 molar ratio of phospholipid/cholesterol were prepared following modifications of previously published protocols.16,28 Peptide dissolved in trifluoroethanol (TFE) was combined with codissolved phospholipid and cholesterol in chloroform/methanol (2/1, v/v). A dry peptide/lipid/cholesterol film, obtained by drying under a N2-stream and overnight exposure to high vacuum, was resuspended in excess HEPES buffer (20 mM HEPES with 1 mM EDTA and 150 mM NaCl at pH 7.0), sonicated for 10–15 min, and subjected to several cycles of freeze–thawing. Negatively stained transmission electron microscopy and dynamic light scattering measurements (not shown) indicated the resulting vesicles were typically ~100 nm in diameter.

Solid-State NMR

Membrane samples, with or without peptide (typically 10 mol %), were pelleted into 4 mm or 3.2 mm zirconia MAS rotors (Bruker Biospin, Billerica, MA) by centrifugation and were kept fully hydrated and unfrozen at all times. All experiments were carried out on a wide-bore Bruker Avance I spectrometer operating at 600 MHz 1H Larmor frequency (14.3T) using either a Bruker wide-bore 4 mm CPMAS triple channel or standard-bore 3.2 mm CPMAS EFree HCN probe. The MAS experiments employed a spinning rate ωr/2π = 8 kHz, while cooling the sample using precooled gas, at all times maintaining the sample temperature above the phase transition temperature of the lipids. Static and MAS 31P spectra were acquired with a wide-bore 4 mm double-channel (1H,31P) CP MAS probe (Bruker) in the absence and presence of sample spinning, applying 50 kHz two-pulse phase modulation (TPPM) 1H decoupling.29 Line shape analysis and integration were performed using Topspin software (Bruker Biospin, Billerica, MA). 31P chemical shifts were referenced to phosphoric acid via external referencing to hydroxyapatite.30 Assignments of peptide resonances were primarily carried out through analysis of 2D 13C–13C experiments employing 1H–13C cross-polarization (CP) followed by 25–100 ms DARR mixing31 with 83 kHz TPPM decoupling during acquisition and evolution. Long distance contacts were observed via 2D 13C–13C experiments utilizing 400 ms PDSD mixing with 71–83 kHz TPPM decoupling during acquisition and evolution. Assignments of the POPC and cholesterol resonances were done with reference to previously published literature and are supported by 2D 1H–13C HETCOR experiments (not shown).32,33 Spectra were processed and analyzed using the Bruker Topspin, NMRPipe,34 Sparky,35 and CCPNMR/Analysis programs.3613C chemical shifts were referenced to dilute aqueous DSS via external referencing to adamantane, and 15N chemical shifts were indirectly referenced to liquid ammonia.37 Additional experimental details are available in Table S2.

FTIR and CD Spectroscopy

For FTIR, samples containing 5 or 10 mol % Cav82–109 or Cav83–102 in 1:1 POPC/cholesterol or DPC micelles were obtained as pellets via centrifugation. The pellet was placed between two CaF2 windows on an MB series spectrophotometer (ABB Bomem, Quebec City, QC, Canada) and analyzed with PROTA software (Biotools Inc., Jupiter, FL). Spectra were recorded at 4 cm–1 resolution and at room temperature (averaging over 400 scans). Buffer correction was applied by subtracting the buffer spectrum interactively until a flat baseline was obtained between 1700 and 1800 cm–1. Peak volumes were estimated by fitting with Gaussian line shapes using the PeakFit routine in SigmaPlot (Systat Software, San Jose, CA).

Cav82–109 and Cav83–102 containing samples for CD were prepared in either POPC vesicles (1:100 peptide–lipid ratio) or DPC micelles (4 mM) in a 13.3 mM phosphate buffer at pH ~6.25, resulting in final peptide concentrations of 10–25 μM. CD spectra were recorded at 25 °C using a Jasco J-810 spectropolarimeter with a 1 mm path length quartz cuvette (Jasco Inc., Easton, MD), averaging over 10 scans with a scanning speed of 50 nm/min in steps of 1 nm.

Sequence Analysis and Secondary Structure Prediction

The secondary structure of full-length Cav1 and fragments Cav82–109 and Cav83–102 was predicted via selected algorithms. The PSIPRED (, Proteus2 (, and PredictProtein ( algorithms were accessed via their respective Web sites.3840


Peptide Incorporation into Model Membranes Using Static and MAS ssNMR

In order to confirm the incorporation of the peptides into the lipid membranes and to probe their localization within the membrane, we applied a number of ssNMR experiments. In analogy with previous work,16 we examined the peptides’ effect on the lipids through basic 13C and 1H MAS NMR experiments. In our hands, the POPC and cholesterol chemical shift changes upon peptide introduction were relatively small and did not unequivocally identify more significant effects on particular parts of the lipids. In various independently prepared samples, introduction of CSD peptides does consistently cause a broadening of the carbon line widths, likely due to a decrease in membrane fluidity. Figure Figure2a–c2a–c shows the natural abundance 13C signals from 1:1 POPC/cholesterol in presence of 10 mol % Cav82–109 (a), with 10 mol % Cav83–102 (b), or without peptide (c). It is notable that a single signal is observed for each lipid and cholesterol site, indicating a single homogeneously structured population of the lipids.

Figure 2
SSNMR shows effects of peptides on the membranes. 1H–13C CPMAS spectra of 1:1 POPC/cholesterol in presence of (a) 10 mol % Cav82–109 (unlabeled), (b) 10 mol % Cav83–102 (unlabeled), or (c) without peptide. Selected assignments ...

To probe the behavior of the phospholipid head groups, we applied static 31P NMR experiments that are informative of the lipids’ macroscopic assembly into distinct lipid phases and the dynamics of individual lipids. Static 1D 31P NMR spectra of 1:1 POPC/cholesterol mixtures are dominated by broad signals with a line shape typical of uniaxial motional averaging of the 31P chemical shift anisotropy (CSA) (Figure (Figure2f)2f) as part of the complex motions that characterize phospholipid bilayers.41 The apparent uniaxial rotation reflects rapid Brownian motion of the phospholipids within the fluid bilayer. In addition to the bilayer phase, certain samples also feature a small isotropic peak (close to 0 ppm), constituting at most 5% of the overall intensity (as determined by line shape simulations). Such isotropic 31P NMR signals indicate that a small portion of the phospholipid head groups undergoes rapid isotropic motional averaging, as is found in small vesicles, micelles, or highly curved nonbilayer phases.42,43 Note that these isotropic signals are not specifically correlated to the presence of peptide but are also seen in lipid-only samples. Compared to POPC bilayers alone (panel d), introduction of high levels of cholesterol (panel f) leads to a notable decrease of the 31P CSA.44 This is most likely due to an increased wobbling motion of the PC headgroup,41 which can be rationalized based on the fact that cholesterol mostly occupies the hydrophobic core of the membrane, lacks a large headgroup, and thus allows for more space and motion of the PC headgroup.45

In the presence of the Cav1 fragments, the PC head groups still show a dominant bilayer line shape (Figure (Figure2g,h).2g,h). The Cav1 peptides do significantly modulate the observed inhomogeneous and homogeneous line broadening of the lipid 31P signals. They have a concentration-dependent (Figure S1) narrowing effect on the 31P CSA. In analogy to cholesterol’s effect, this may indicate incorporation into the membrane’s hydrophobic core with less interaction with the lipid head groups. This decrease in the inhomogeneous broadening is more pronounced for the Cav82–109 peptide (Figure (Figure2h)2h) than the isolated CSD (Figure (Figure2g).2g). Addition of the peptide also causes an increase in the homogeneous line width (Figure (Figure2e,g,h).2e,g,h). Such effects have previously been seen with other membrane-binding compounds and are most likely due to an interference with the overall motion of the phospholipids.46 In POPC (Figure (Figure2e)2e) and, to a lesser extent, POPC/cholesterol bilayers (Figure (Figure2g,h),2g,h), the rapid Brownian motion of each lipid in the fluid bilayer helps to reduce the 31P–1H dipolar interactions, which can otherwise cause an “intrinsic broadening” of the 31P signal of individual lipids.46,47 The incorporation of the Cav1 fragments interferes with the rapid lipid motion and thus results in an increased homogeneous broadening of the 31P NMR line shape.

The above observations indicate an intimate interaction of these peptides with the lipid membranes, without signs of peptide-induced heterogeneity. The peptides seem to partition preferentially into the hydrophobic part of the bilayer, providing more space for POPC headgroup motion, while reducing the overall fluidity of the membrane.

Peptide Secondary Structure Content

Having probed the peptides’ incorporation into the mixed membranes, we also characterized the secondary structure of the membrane-associated peptides. FTIR studies were performed on Cav82–109 and Cav83–102 peptides in a 1:1 POPC/cholesterol mixture, prepared identically to the NMR samples (Figure (Figure33). Cav82–109 features C=O vibrations characteristic for both α-helix (1654 cm–1) and β-strand (1623 and 1693 cm–1).48 Based on peak deconvolution (see Figure Figure3),3), relative amounts of approximately 60% α-helical and 40% β-sheet are estimated for Cav82–109 (Table S3). In contrast, the shorter Cav83–102 displays little α-helical content (Figure (Figure3b).3b). The phospholipid’s ester C=O vibration is also observed at 1730 cm–1.49

Figure 3
FTIR spectra of (a) 10 mol % Cav82–109 and (b) 10 mol % Cav83–102 in 1:1 POPC/cholesterol. Deconvolutions for estimating secondary structure content are shown as dashed for their respective data sets. Results are consistent with a mixed ...

The FTIR results thus indicate a notable increase in α-helical content in the longer peptide, an observation that is consistent with the work by Le Lan et al., employing CD spectroscopy.17 These earlier studies were done in DPC rather than cholesterol/POPC membranes. We obtained similar FTIR results in DPC micelles (not shown) and CD spectra in POPC also indicated a low level of helicity in the shorter peptides and an increased α-helical content for the longer peptide (Figure S2).

Thus, both CD and FTIR indicate an increase in α-helical content in Cav82–109, but it appears that a substantial fraction remains non-α-helical. These data do not indicate the localization of the helix and could be explained by sample heterogeneity.

Site-Specific Structure Analysis by MAS ssNMR

To site specifically probe the mixed secondary structure of Cav82–109, MAS ssNMR was applied to partly 13C,15N-labeled peptides. Two Cav82–109 peptides with 13C,15N-labeling of residues in both the CSD and IMD were incorporated into 1:1 POPC/cholesterol membranes. The first peptide (p1) featured U–13C,15N-labeled residues in sites Ala87, Phe92, Val94, and Lys96, and the second (p2) was labeled in residues Gly83, Val94, Phe99, Leu102, and Ala105 (Table S1). Probing residues Phe92 and Val94 yields structural information on the FTVT protein binding segment,13 while analogously measuring Val94, Lys96, and Phe99 provides insight into the CRAC motif.16 Labeling of Val94 in both peptides permitted verification of unchanged Val94 shifts between different samples, which was indeed the case (Figure S3). Chemical shift assignments were obtained using 2D 13C–13C DARR31 experiments (e.g., Figure Figure4),4), supplemented with 2D 15N–13C correlations (not shown), and are available in Table S4. In most cases single resonances were observed, indicating the presence of a single well-defined conformation, although motional averaging that is fast on the NMR time scale cannot be excluded. We do observe weaker additional resonances for Ala87 and Val94 (in both peptides), indicating a less-populated secondary conformation of those residues. Reduced intensities of certain peaks (e.g., the labeled Phe side chains) are consistent with conformational exchange of the amino acid side chains.

Figure 4
2D 13C–13C ssNMR spectra using 25 ms DARR mixing, providing 1–2 bond transfers. Measurement on samples (a) p1 and (b) p2 are shown; both obtained at 600 MHz 1H frequency and 8 kHz MAS at 283 K. Aliphatic-to-carbonyl (left) as well as intra-aliphatic ...

We examined the residue-specific secondary structure by comparing our chemical shifts to the corresponding shifts of known protein structures, in an approach known as chemical shift indexing (CSI).50,51 The difference of 13C chemical shifts from random coil shifts51 correlates in a predictable fashion to the local secondary structure. As shown in Figure Figure5,5, these data indicate that residues Ala87, Phe92, Val94, and Lys96, spanning the protein binding and CRAC motifs, adopt a β-sheet conformation, whereas residues Phe99 (part of the CRAC motif), Leu102, and Ala105 (in the IMD) adopt an α-helical structure. Gly83, near the N-terminus of this peptide fragment, appears to lack a clearly defined secondary structure.

Figure 5
(a) SSNMR indicates α-helical conformation (red) for labeled sites in the IMD, while most CSD residues have a β-conformation (blue). Residues of the CRAC motif (V94–Y100) span the helix–strand boundary, and the protein ...

Given concerns about the effect of truncation of secondary structure elements,17 we also examined a longer peptide Cav76–109 (p3), which in the native sequence is flanked on both sides by prolines (P75 and P110), which often act as secondary structure breakers. Proline effects on membrane-spanning helices seems to be more subtle, yet they do tend to destabilize these α-helices.52 We incorporated isotopic labels at Gly83 and Ala87 as probes of the CSD secondary structure and found that these residues had the same chemical shifts (and thus structures) as in the Cav82–109 fragment (Figure S3).

The β-sheet and α-helical structural elements in Cav82–109 were further probed in 2D 13C–13C experiments with 400 ms PDSD mixing. These conditions permit 13C–13C magnetization transfer over longer distances and revealed not only intra-residue contacts within labeled residues but also specific inter-residue contacts. As summarized in Figure Figure5,5, residue Val94 is observed to interact with both Phe92 and Lys96, consistent with ii + 2 contacts typical of β-sheet structure. Similarly, residue Leu102 interacts with residue Ala105, reflecting an ii + 3 contact characteristic of α-helical structure (and one that would not likely be observed in a β-strand). The observation of these ii + 2 and ii + 3 contacts are all consistent with the secondary structure elements identified from the CSI analysis above and further solidify the secondary structure assignments (Figure (Figure55).

Structure Prediction

Our observations of a partially β-sheet structure localized within the CSD (specifically including the labeled residues between positions 87 and 96) seem to match results from a computational study that predicted an antiparallel β-hairpin motif in residues 84–94.24 However, a number of other studies report a preference for α-helical structure from primary sequence analysis.4,19 To explore this in more detail, we have also applied a number of prediction algorithms on full-length Cav1 and selected fragments. The results of predictions using the PSIPRED algorithm38 are shown in Figure Figure6,6, suggesting a preference for an extended (possibly β-sheet) conformation throughout the CSD segment, followed by an α-helical structure in the IMD, not only in full length Cav1 but also in the Cav82–109 fragment (gray box in Figure Figure6).6). Moreover, consistent with our and earlier experimental results,17 truncation of the α-helical segment (i.e., ending at residue 102 rather than just before the native helix-breaker Pro110) is predicted to destabilize the helical conformation (e.g., in Cav83–102). However, there appears to be little indication of a fully α-helical CSD. Similar results were obtained with the PROTEUS2 and PredictProtein prediction algorithms,39,40 both of which suggest mixtures of β- and α-structure within the CSD (see Figure S4). Overall, the results of all three algorithms are also somewhat similar to the consensus results presented by Spisni et al, but differ from the reports of more extensive helical preferences. Of course, it remains unseen how reliable any of these generally globular-protein-based algorithms will prove to be for this oligomerizing, cholesterol-binding and likely monotopic membrane protein.

Figure 6
Secondary structure prediction of Cav1 and fragments. PSIPRED analysis of full-length Cav1 (Cav1–178) indicates a propensity for β-sheet structure in the CSD (underlined). Predictions for fragments Cav82–109 and Cav83–102 ...


On the basis of a combination of ssNMR, CD, and FTIR measurements, we have characterized the secondary structure of Cav1 fragments, while bound to cholesterol-rich model membranes. These experiments were designed to examine the CSD, as it harbors many Cav1 functionalities. The Cav82–109 peptide, incorporating both the CSD and the first half of the IMD, was found to contain a mixture of β-sheet and α-helical structure. The latter was localized within the putative first α-helix of the helical hairpin of the IMD, consistent with previous work and structural models and predictions.1720 More surprisingly, our observations indicate a β-strand conformation for the majority of labeled residues within the CSD, based on ssNMR shift analysis and inter-residue contacts, as well as the detection of mixed secondary structure by CD and FTIR. Computational analysis of the primary sequence predicts that part of the CSD prefers a β-sheet conformation and matches both the location of the experimental transition to α-helix and the observed lack of helicity in short peptides.17,24 Figure Figure77 contains a graphical schematic that summarizes these observations. The observed helicity of the first half of the IMD is consistent with reports proposing models where the P110 forms the turn in a “re-entrant” helix stabilized by interacting smaller residues in both its helical segments.4,6,7 This domain’s wedge shape that fails to traverse the entire bilayer may well contribute to the remarkable membrane curvature of the caveolae.53

Figure 7
Schematic illustration of secondary structure distribution of membrane-bound Cav1. The wedge-shaped α-helical hairpin is too short to traverse the entire bilayer and occupies a single monolayer. The CSD is shown as largely β-stranded, ...

Existing Structural Data and Models for the CSD

Previous reports on Cav and the CSD fail to reach a consensus on the structure of the CSD. Studies by Fernandez et al. via CD and sequence analysis of a Cav1–101 fragment in solution proposed the presence of an α-helix spanning residues 79–96.19 Solution NMR and CD on solubilized Cav82–101 also indicated a predominantly helical structure spanning residues Ile84–Tyr97.21 However, the latter experiments were largely done in the presence of the helicity-enhancing cosolvent TFE, which was reportedly required to avoid aggregation. CD spectroscopy on Cav83–102 peptides associated with POPC vesicles16 and Cav82–101 peptides in DPC micelles17 indicated only a partially α-helical structure for the isolated CSD. CD on a longer construct Cav82–109 (also employed here) revealed a large increase in α-helicity, and solution NMR on the micelle-associated peptides localized the α-helical structure to residues 83–88 and 93–97 of the CSD and 102–108 within the IMD.17 In some of these studies computational analysis of the primary sequence supported a predominantly α-helical CSD conformation, even leading to a predominantly helical full-length model of Cav.4,19 On the other hand, an in silico study by Spisni et al. has proposed an antiparallel β-hairpin structure in residues 84–94 of the CSD, which is remarkably consistent with our findings.24

Despite the lack of consensus, our observations on membrane-bound peptides recapitulate a number of features reported above. We observe that the peptides are characterized by a partially α-helical conformation that is most pronounced in the C-terminal residues (beyond the CSD proper). The key difference in this study is the experimental identification of a β-stranded conformation within the CSD (in cholesterol-rich bilayers) and the fact that we site-specifically delineate the location of the different secondary structures. Our ssNMR results show that the observed mixed secondary structure content, with substantial nonhelical structure, is not simply due to sample heterogeneity (with some peptide largely helical but others unstructured) but rather reflects the presence of distinct domains within the peptides.

Our work is also distinct in that we probe the structure of the peptide in the presence of high levels of cholesterol to mimic the caveolar membrane and to facilitate examination of the CSD–cholesterol interaction. MAS and static ssNMR experiments clearly show that the lipid mixture responds to the integration of the peptides without generating peptide-induced heterogeneity in the lipids. Thus, we see a membrane-bound peptide and not simply the coprecipitation of noninteracting peptides and lipids.

Scaffolding Domain Functionalities

In the following sections, we will examine our observations in view of the different roles ascribed to the CSD, in an attempt to place our results in context and discuss potential implications for each of these CSD functionalities.

1. CSD Binding Partners

One critical caveolin function localized within the CSD relates to interactions with its numerous binding partners, such as protein G α-domains and Src family kinases.13 Within the partner protein, this is mediated by the CBM, with a consensus sequence (ΦXΦXXXXΦXXΦ) rich in aromatic residues (Φ). The CBM of the protein G α-subunit forms a β-hairpin (Figure (Figure88a)13. Within the CSD, residues F92TVT95 are involved, which adopt a β-stranded conformation in our samples. In various studies,11,13,14,54,55 isolated short CSD peptides appear to emulate the effect of Cav binding to its binding partners, suggesting that the Cav tertiary structure is not required. Often a membrane is not present in those studies, raising the question of whether a membrane-bound CSD conformation is at all involved and thus how our data correlate to this CSD role.

Figure 8

Secondary structure motifs in proposed homologous protein domains. (a) Structure of the aromatic-residue-rich caveolin binding sequence as originally identified in Gα proteins,13 featuring a β-hairpin structure. (b, c) β-Strand-rich structures ...

2. Oligomerization

The CSD is also involved in the formation of Cav homo-oligomers, and oligomerization is a critical functionality as it appears to be required in the formation of caveolae. We note that a membrane-embedded β-stranded CSD may be prone to self-assemble into β-sheets in order to eliminate membrane-exposed non-hydrogen-bonded backbone sites. Indeed, a bioinformatics study by Spisni et al. has suggested that such a β-hairpin structure in residues 84–94 could play a role in homo-oligomer interactions.24 It has also been noted that Cav1 appears to have a motif in residues 92–100, coinciding with the CSD, which strongly resembles the CBM.13 This may indicate a mechanism for Cav self-oligomerization that is analogous to its recognition of the binding partner proteins. If so, this seems consistent with a potential CSD β-hairpin structure, given the hairpin conformation of the CBM (see above).

Our current data do not inform us on the oligomeric (or monomeric) nature of the peptides under investigation. Long-mixing 13C–13C experiments could be employed to identify intermolecular interactions and detect and characterize the supramolecular assemblies.56,57 However, the long-mixing experiments on the current samples have failed to produce data consistent with intermolecular interactions, including experiments performed on samples prepared from mixtures of peptides p1 and p2 (not shown). This may be due to the sparse labeling and the relatively low signal intensities for some of the samples. More indirectly, the apparent rigidity of the peptides and pronounced effect on membrane fluidity could be indicative of the formation of peptide oligomers. However, this may also be explained in part by the high level of cholesterol present. Note that our data do show clear effects on the membrane dynamics and that the peptides are homogeneously associated with the membrane, resulting in a single population of lipid NMR as well as peptide signals. This indicates that any oligomerization that has occurred is taking place in a membrane-bound state and thus may be mimicking the homo-oligomer formation that Cav displays in cholesterol-rich caveolae.

3. Cholesterol Binding

The binding of cholesterol is thought to be mediated by residues V94TKYWFYR101 that form a putative CRAC motif. The original CRAC motif was identified in the translocator protein TSPO.58 Subsequent solution NMR studies of various CRAC motifs have proposed a predominantly α-helical CRAC structure,17,20,22 although other work suggested a lower level of α-helicity.16 This difference has been credited to a loss of helical structure due to truncation in the shorter peptides used in the latter studies.17 The above structural work involved short peptides lacking protein context, but a recent study reported on CRAC motifs in the protein PDC-109, for which a known homologous structure exists.59 While this seemingly provides much-needed high-resolution structures of these motifs in a native context, a more detailed inspection reveals an array of different structures (Figure (Figure8b,c8b,c and Figure S5). This may act as a reminder that these are putative motifs and that some may not be functional cholesterol binding domains. Nonetheless, while PDC-109 fails to provide an unambiguous structure for the CRAC motif, it may provide examples of possible β-rich CRAC conformations (Figure (Figure8b,c).8b,c). Both our experimental and computational data indicate a specific transition of β-sheet to α-helical structure to occur within the CRAC motif, near the motif’s conserved central Tyr97, reminiscent of earlier computational predictions.24 Intriguingly, at least one of the PDC-109 CRAC motifs (Figure (Figure8c)8c) shows a similar motif, with the central Tyr at the C-terminal end of a β-hairpin and near the start of a flanking helical segment. The observed sheet–helix transition is also reminiscent of the tendency for CRAC motifs to occur directly adjacent to transmembrane helices.60

Given the reported strong interactions between caveolin and cholesterol, one key point of interest will be to characterize this interaction in more detail in future work. Interestingly, in the current work we have not seen large effects for the peptide NMR signals of the omission of cholesterol (Figure S1). This suggests that the observed secondary structures are not directly correlated to cholesterol binding, raising the question of whether this indicates a more subtle effect of cholesterol binding, the need for other parts of Cav, or a role for other lipids enriched in caveolae (e.g., sphingomyelin).

The above shows that there have been indications of a role for β-stranded conformations in CSD and that such a structure may make sense from a functional point of view. One seemingly common feature is the presence of β-hairpins, as also suggested by Spisni et al., who proposed that an antiparallel β-hairpin is formed by residues 84–94.24 Our ssNMR data reveal the localized β-conformation, but having thus far failed to observe long-distance inter- or intramolecular interactions, it remains uncertain whether this represents a single β-strand or shorter β-strands within e.g. a β-hairpin. The relatively sparse labeling applied here makes it impossible to exclude the hairpin motif based on these data, as it is quite possible that labeled sites in a β-hairpin would be too far apart. Whether in a hairpin or not, it may well be that the β-strands play a role in the oligomerization process, as membrane-embedded strands could be driven to self-assemble to prevent exposure of the peptide backbones to the hydrophobic membrane environment.24


Our experimental and computational analysis of the structural features of Cav1 fragments seems to indicate a more significant propensity for β-sheet structures with the CSD than apparent from the existing literature. We have discussed that such β-rich structures could be consistent with its various functionalities, while highlighting the need for further characterization. One important consideration is that the caveolins fulfill a variety of functions, which could be independent of each other, occur in different locations within the cell, and may involve different Cav conformations. For instance, it has been pointed out that palmitoylation is nonessential for the localization in caveolae but that it is critical for the cholesterol trafficking by Cav.10 It seems likely that the Cav conformation within caveolae could be significantly different from its structure in a cholesterol-bound cytosolic state. Thus, it is important to investigate Cav structure in different contexts, including cholesterol-rich membranes as examined here. Thus, far, we lack a clear understanding of the structural effects of cholesterol binding, as we did not observe a large structural change in the presence of cholesterol. Clearly, additional experimental and structural data will be critical to elucidate the conformation and oligomeric states of Cav and to provide an understanding of the effect of disease-causing mutations and the structural underpinning of its many functional roles within the cell.


We thank Mike Delk for technical assistance and Ron Wetzel for the use of laboratory equipment.



caveolin-binding motif
cholesterol recognition/interaction amino acid consensus
chemical shift anisotropy
caveolin scaffolding domain
chemical shift indexing
dipolar-assisted rotational resonance
intramembrane domain
magic angle spinning
proton-driven spin diffusion
solid-state NMR
two-pulse phase modulation.

Funding Statement

National Institutes of Health, United States


This work was supported in part by Grant UL1 RR024153 from the National Center for Research Resources (NCRR) and startup funds from the University of Pittsburgh.

Author Contributions

Author Contributions

These authors contributed equally to this work.

Author Present Address

Author Present Address

Department of Chemistry & Biochemistry, 251 Nieuwland Science Hall, University of Notre Dame du Lac, Notre Dame, IN 46556.

Supporting Information Available

Supporting Information Available

Table of labeled peptides; detailed experimental conditions for ssNMR experiments; peptide concentration and cholesterol dependence of 13C and 31P NMR spectra; secondary structure content estimates from FTIR; 13C chemical shift assignments; additional FTIR and CD spectra; additional 2D 13C–13C spectra and secondary structure predictions; structures of related proteins. This material is available free of charge via the Internet at

Supplementary Material


  • Razani B.; Woodman S. E.; Lisanti M. P. (2002) Caveolae: from cell biology to animal physiology. Pharmacol. Rev. 54, 431–467. [PubMed]
  • Anderson R. G. W. (1998) The caveolae membrane system. Annu. Rev. Biochem. 67, 199–225. [PubMed]
  • Parton R. G.; Simons K. (2007) The multiple faces of caveolae. Nat. Rev. Mol. Cell Biol. 8, 185–194. [PubMed]
  • Parton R. G.; Hanzal-Bayer M.; Hancock J. F. (2006) Biogenesis of caveolae: a structural model for caveolin-induced domain formation. J. Cell Sci. 119, 787–796. [PubMed]
  • Engelman J. A.; Zhang X.; Galbiati F.; Volonte D.; Sotgia F.; Pestell R. G.; Minetti C.; Scherer P. E.; Okamoto T.; Lisanti M. P. (1998) Molecular genetics of the caveolin gene family: implications for human cancers, diabetes, Alzheimer disease, and muscular dystrophy. Am. J. Hum. Genet. 63, 1578–1587. [PubMed]
  • Ostermeyer A. G.; Ramcharan L. T.; Zeng Y.; Lublin D. M.; Brown D. A. (2004) Role of the hydrophobic domain in targeting caveolin-1 to lipid droplets. J. Cell Biol. 164, 69–78. [PubMed]
  • Aoki S.; Thomas A.; Decaffmeyer M.; Brasseur R.; Epand R. M. (2010) The role of proline in the membrane re-entrant helix of caveolin-1. J. Biol. Chem. 285, 33371–33380. [PubMed]
  • Li S.; Song K. S.; Lisanti M. P. (1996) Expression and characterization of recombinant caveolin. Purification by polyhistidine tagging and cholesterol-dependent incorporation into defined lipid membranes. J. Biol. Chem. 271, 568–573. [PubMed]
  • Dietzen D. J.; Hastings W. R.; Lublin D. M. (1995) Caveolin is palmitoylated on multiple cysteine residues. Palmitoylation is not necessary for localization of caveolin to caveolae. J. Biol. Chem. 270, 6838–6842. [PubMed]
  • Uittenbogaard A.; Smart E. J. (2000) Palmitoylation of caveolin-1 is required for cholesterol binding, chaperone complex formation, and rapid transport of cholesterol to caveolae. J. Biol. Chem. 275, 25595–25599. [PubMed]
  • Li S.; Couet J.; Lisanti M. P. (1996) Src tyrosine kinases, Galpha subunits, and H-Ras share a common membrane-anchored scaffolding protein, caveolin. Caveolin binding negatively regulates the auto-activation of Src tyrosine kinases. J. Biol. Chem. 271, 29182–29190. [PubMed]
  • Schlegel A.; Schwab R. B.; Scherer P. E.; Lisanti M. P. (1999) A role for the caveolin scaffolding domain in mediating the membrane attachment of caveolin-1. The caveolin scaffolding domain is both necessary and sufficient for membrane binding in vitro. J. Biol. Chem. 274, 22660–22667. [PubMed]
  • Couet J.; Li S.; Okamoto T.; Ikezu T.; Lisanti M. P. (1997) Identification of peptide and protein ligands for the caveolin-scaffolding domain. Implications for the interaction of caveolin with caveolae-associated proteins. J. Biol. Chem. 272, 6525–6533. [PubMed]
  • Couet J.; Sargiacomo M.; Lisanti M. P. (1997) Interaction of a receptor tyrosine kinase, EGF-R, with caveolins. Caveolin binding negatively regulates tyrosine and serine/threonine kinase activities. J. Biol. Chem. 272, 30429–30438. [PubMed]
  • Murata M.; Peränen J.; Schreiner R.; Wieland F.; Kurzchalia T. V.; Simons K. (1995) VIP21/caveolin is a cholesterol-binding protein. Proc. Natl. Acad. Sci. U. S. A. 92, 10339–10343. [PubMed]
  • Epand R. M.; Sayer B. G.; Epand R. F. (2005) Caveolin scaffolding region and cholesterol-rich domains in membranes. J. Mol. Biol. 345, 339–350. [PubMed]
  • Le Lan C.; Neumann J. M.; Jamin N. (2006) Role of the membrane interface on the conformation of the caveolin scaffolding domain: a CD and NMR study. FEBS Lett. 580, 5301–5305. [PubMed]
  • Levin A. M.; Coroneus J. G.; Cocco M. J.; Weiss G. A. (2006) Exploring the interaction between the protein kinase A catalytic subunit and caveolin-1 scaffolding domain with shotgun scanning, oligomer complementation, NMR, and docking. Protein Sci. 15, 478–486. [PubMed]
  • Fernandez I.; Ying Y.; Albanesi J.; Anderson R. G. W. (2002) Mechanism of caveolin filament assembly. Proc. Natl. Acad. Sci. U. S. A. 99, 11193–11198. [PubMed]
  • Le Lan C.; Gallay J.; Vincent M.; Neumann J. M.; de Foresta B.; Jamin N. (2010) Structural and dynamic properties of juxta-membrane segments of caveolin-1 and caveolin-2 at the membrane interface. Eur. Biophys. J. 39, 307–325. [PubMed]
  • Levin A. M.; Murase K.; Jackson P. J.; Flinspach M. L.; Poulos T. L.; Weiss G. A. (2007) Double barrel shotgun scanning of the caveolin-1 scaffolding domain. ACS Chem. Biol. 2, 493–500. [PubMed]
  • Jamin N.; Neumann J. M.; Ostuni M. A.; Vu T. K. N.; Yao Z.-X.; Murail S.; Robert J.-C.; Giatzakis C.; Papadopoulos V.; Lacapère J.-J. (2005) Characterization of the cholesterol recognition amino acid consensus sequence of the peripheral-type benzodiazepine receptor. Mol. Endocrinol. 19, 588–594. [PubMed]
  • Ortegren U.; Karlsson M.; Blazic N.; Blomqvist M.; Nystrom F. H.; Gustavsson J.; Fredman P.; Strålfors P. (2004) Lipids and glycosphingolipids in caveolae and surrounding plasma membrane of primary rat adipocytes. Eur. J. Biochem. 271, 2028–2036. [PubMed]
  • Spisni E.; Tomasi V.; Cestaro A.; Tosatto S. C. E. (2005) Structural insights into the function of human caveolin 1. Biochem. Biophys. Res. Commun. 338, 1383–1390. [PubMed]
  • McDermott A. (2009) Structure and dynamics of membrane proteins by magic angle spinning solid-state NMR. Annu. Rev. Biophys. 38, 385–403. [PubMed]
  • Hong M. (2006) Oligomeric structure, dynamics, and orientation of membrane proteins from solid-state NMR. Structure 14, 1731–1740. [PubMed]
  • Opella S. J.; Marassi F. M. (2004) Structure determination of membrane proteins by NMR spectroscopy. Chem. Rev. 104, 3587–3606. [PubMed]
  • Smith S. O.; Eilers M.; Song D.; Crocker E.; Ying W.; Groesbeek M.; Metz G.; Ziliox M.; Aimoto S. (2002) Implications of threonine hydrogen bonding in the glycophorin A transmembrane helix dimer. Biophys. J. 82, 2476–2486. [PubMed]
  • Bennett A. E.; Rienstra C. M.; Auger M.; Lakshmi K. V.; Griffin R. G. (1995) Heteronuclear decoupling in rotating solids. J. Chem. Phys. 103, 6951–6958.
  • Luo W.; Hong M. (2010) Conformational changes of an ion channel detected through water-protein interactions using solid-state NMR spectroscopy. J. Am. Chem. Soc. 132, 2378–2384. [PubMed]
  • Takegoshi K.; Nakamura S.; Terao T. (2001) C-13-H-1 dipolar-assisted rotational resonance in magic-angle spinning NMR. Chem. Phys. Lett. 344, 631–637.
  • Forbes J.; Bowers J.; Shan X.; Moran L.; Oldfield E.; Moscarello M. A. (1988) Some new developments in solid-state nuclear magnetic resonance spectroscopic studies of lipids and biological membranes, including the effects of cholesterol in model and natural systems. J. Chem. Soc., Faraday Trans. 1 84, 3821–3849.
  • Guo W.; Hamilton J. A. (1996) 13C MAS NMR studies of crystalline cholesterol and lipid mixtures modeling atherosclerotic plaques. Biophys. J. 71, 2857–2868. [PubMed]
  • Delaglio F.; Grzesiek S.; Vuister G. W.; Zhu G.; Pfeifer J.; Bax A. (1995) NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6, 277–293. [PubMed]
  • Goddard T. D., and Kneller D. G.Sparky, 3rd ed.; University of California: San Francisco.
  • Vranken W. F.; Boucher W.; Stevens T. J.; Fogh R. H.; Pajon A.; Llinas M.; Ulrich E. L.; Markley J. L.; Ionides J.; Laue E. D. (2005) The CCPN data model for NMR spectroscopy: development of a software pipeline. Proteins 59, 687–696. [PubMed]
  • Harris R. K.; Becker E. D.; De Menezes S. M. C.; Granger P.; Hoffman R. E.; Zilm K. W. (2008) Further conventions for NMR shielding and chemical shifts (IUPAC Recommendations 2008). Magn. Reson. Chem. 46, 582–598. [PubMed]
  • Jones D. T. (1999) Protein secondary structure prediction based on position-specific scoring matrices. J. Mol. Biol. 292, 195–202. [PubMed]
  • Rost B.; Yachdav G.; Liu J. (2004) The PredictProtein server. Nucleic Acids Res. 32, W321–326. [PubMed]
  • Montgomerie S.; Cruz J. A.; Shrivastava S.; Arndt D.; Berjanskii M.; Wishart D. S. (2008) PROTEUS2: a web server for comprehensive protein structure prediction and structure-based annotation. Nucleic Acids Res. 36, W202–209. [PubMed]
  • Dufourc E. J.; Mayer C.; Stohrer J.; Althoff G.; Kothe G. (1992) Dynamics of phosphate head groups in biomembranes. Comprehensive analysis using phosphorus-31 nuclear magnetic resonance lineshape and relaxation time measurements. Biophys. J. 61, 42–57. [PubMed]
  • Van der Wel P. C. A.; Pott T.; Morein S.; Greathouse D. V.; Koeppe R.; Killian J. (2000) Tryptophan-anchored transmembrane peptides promote formation of nonlamellar phases in phosphatidylethanolamine model membranes in a mismatch-dependent manner. Biochemistry 39, 3124–3133. [PubMed]
  • Costello A. L.; Alam T. M. (2010) Investigating the impact of cholesterol on magnetically aligned sphingomyelin/cholesterol multilamellar vesicles using static (31)P NMR. Chem. Phys. Lipids 163, 506–513. [PubMed]
  • Holland G. P.; McIntyre S. K.; Alam T. M. (2006) Distinguishing individual lipid headgroup mobility and phase transitions in raft-forming lipid mixtures with 31P MAS NMR. Biophys. J. 90, 4248–4260. [PubMed]
  • Huang J.; Feigenson G. W. (1999) A microscopic interaction model of maximum solubility of cholesterol in lipid bilayers. Biophys. J. 76, 2142–2157. [PubMed]
  • Benetis N.; Kyrikou I.; Zervou M.; Mavromoustakos T. (2005) Static CP P-31 NMR multilamellar bilayer broadlines in the absence and presence of the bioactive dipeptide beta-Ala-Tyr or Glu. Chem. Phys. 314, 57–72.
  • Seelig J. (1978) P-31 nuclear magnetic-resonance and head group structure of phospholipids in membranes. Biochim. Biophys. Acta 515, 105–140. [PubMed]
  • Kong J.; Yu S. (2007) Fourier transform infrared spectroscopic analysis of protein secondary structures. Acta Biochim. Biophys. Sin. 39, 549–559. [PubMed]
  • Reis O.; Winter R.; Zerda T. W. (1996) The effect of high external pressure on DPPC-cholesterol multilamellar vesicles: a pressure-tuning Fourier transform infrared spectroscopy study. Biochim. Biophys. Acta 1279, 5–16. [PubMed]
  • Wishart D. S.; Sykes B. D. (1994) The 13C Chemical-Shift Index: A simple method for the identification of protein secondary structure using 13C chemical-shift data. J. Biomol. NMR 4, 171–180. [PubMed]
  • Zhang H.; Neal S.; Wishart D. S. (2003) RefDB: a database of uniformly referenced protein chemical shifts. J. Biomol. NMR 25, 173–195. [PubMed]
  • Senes A.; Engel D. E.; DeGrado W. F. (2004) Folding of helical membrane proteins: the role of polar, GxxxG-like and proline motifs. Curr. Opin. Struct. Biol. 14, 465–479. [PubMed]
  • Morris R.; Cox H.; Mombelli E.; Quinn P. J. (2004) Rafts, little caves and large potholes: how lipid structure interacts with membrane proteins to create functionally diverse membrane environments. Subcell. Biochem. 37, 35–118. [PubMed]
  • Volonte D.; Galbiati F. (2009) Inhibition of thioredoxin reductase 1 by caveolin 1 promotes stress-induced premature senescence. EMBO Rep. 10, 1334–1340. [PubMed]
  • Toya Y.; Schwencke C.; Couet J.; Lisanti M. P.; Ishikawa Y. (1998) Inhibition of adenylyl cyclase by caveolin peptides. Endocrinology 139, 2025–2031. [PubMed]
  • Lewandowski J. R.; van der Wel P. C. A.; Rigney M.; Grigorieff N.; Griffin R. G. (2011) Structural Complexity of a Composite Amyloid Fibril. J. Am. Chem. Soc. 133, 14686–14698. [PubMed]
  • Li J.; Hoop C. L.; Kodali R.; Sivanandam V. N.; van der Wel P. C. A. (2011) Amyloid-like fibrils from a domain-swapping protein feature a parallel, in-register conformation without native-like interactions. J. Biol. Chem. 286, 28988–28995. [PubMed]
  • Li H.; Papadopoulos V. (1998) Peripheral-type benzodiazepine receptor function in cholesterol transport. Identification of a putative cholesterol recognition/interaction amino acid sequence and consensus pattern. Endocrinology 139, 4991–4997. [PubMed]
  • Scolari S.; Müller K.; Bittman R.; Herrmann A.; Müller P. (2010) Interaction of mammalian seminal plasma protein PDC-109 with cholesterol - Implications for a putative CRAC domain. Biochemistry 49, 9027–9031. [PubMed]
  • Epand R. M. (2008) Proteins and cholesterol-rich domains. Biochim. Biophys. Acta 1778, 1576–1582. [PubMed]
  • Wah D. A.; Fernández-Tornero C.; Sanz L.; Romero A.; Calvete J. J. (2002) Sperm coating mechanism from the 1.8 A crystal structure of PDC-109-phosphorylcholine complex. Structure 10, 505–514. [PubMed]

Articles from ACS AuthorChoice are provided here courtesy of American Chemical Society