Structural Consequences of Cysteine Substitutions C1977Y and C1977R in Calcium-binding Epidermal Growth Factor-like Domain 30 of Human Fibrillin-1*

The largest group of disease-causing mutations affecting calcium-binding epidermal growth factor-like (cbEGF) domain function in a wide variety of extracellular and transmembrane proteins is that which results in cysteine substitutions. Although known to introduce proteolytic susceptibility, the detailed structural consequences of cysteine substitutions in cbEGF domains are unknown. Here, we studied pathogenic mutations C1977Y and C1977R, which affect cbEGF30 of human fibrillin-1, in a recombinant three cbEGF domain fragment (cbEGF29–31). Limited proteolysis, 1H NMR, and calcium chelation studies have been used to probe the effect of each substitution on cbEGF30 and its flanking domains. Analysis of the wild-type fragment identified two high affinity and one low affinity calcium-binding sites. Each substitution caused the loss of high affinity calcium binding to cbEGF30, consistent with intradomain misfolding, but the calcium binding properties of cbEGF29 and cbEGF31 were surprisingly unaffected. Further analysis of mutant fragments showed that domain packing of cbEGF29–30, but not cbEGF30–31, was disrupted. These data demonstrate that C1977Y and C1977R have localized structural effects, confined to the N-terminal end of the mutant domain, which disrupt domain packing. Cysteine substitutions affecting other cbEGF disulfide bonds are likely to have different effects. This proposed structural heterogeneity may underlie the observed differences in stability and cellular trafficking of proteins containing such changes.

domain structure and cbEGF29 -30 domain-domain packing were significantly disrupted by the loss of the 1-3 disulfide bond, no effects on the calcium binding properties of N-terminal cbEGF domain 29 and the C-terminal cbEGF domain 31 were observed, suggesting that both of these cysteine substitutions have localized effects.

Cloning of Wild-type and Mutant cbEGF Domain Constructs from
Human Fibrillin-1-DNA fragments (nucleotides 5918 -6295 of human fibrillin-1 cDNA) encoding the wild-type cbEGF29 -31 domains (residues 1929 -2054) and fragments containing the C1977R and C1977Y mutant sequences were amplified by standard polymerase chain reaction techniques using Pfu polymerase (Stratagene). The DNA was amplified by the use of a forward primer that encodes a factor Xa cleavage site: 5Ј-TAGTAGGGATCCATAGAAGGACGATCAGCAATAG-ATGTTGATGAATGTGC-3Ј and a reverse primer: 5Ј-TAGTAGA-AGCTTCTATTATTGGCACCTTCTTCCACTGGAGGAC-3Ј.
The amplified DNA was cloned into pQE30 (Qiagen), which contains a sequence encoding a six-histidine affinity tag, and used to transform Escherichia coli NM554[pREP4]. The clones were sequenced to confirm that the correct sequence had been amplified and that each mutation had been introduced into the inserted fragments with no other changes. Protein expression, refolding, and purification were carried out as described previously (21)(22)(23)(24), and the identity of the purified proteins was confirmed by electrospray mass spectrometry.
Limited Proteolysis of Wild-type and Mutant cbEGF Domain Constructs-Proteolysis with trypsin (1:50, w/w) or endoproteinase Glu-C (1:125, w/w) was performed as described previously (24). N-terminal sequencing was used to characterize the proteolytic digestion products. Proteolysis performed in 50 mM EGTA or 50 mM CaCl 2 was terminated after 60 -120 min by acidification to pH 2. 50 mM CaCl 2 was used to ensure saturation of the N-terminal cbEGF29 calcium-binding site, which was thought likely to have a weak affinity in the absence of cbEGF28. Samples were purified under nonreducing conditions by reverse phase HPLC. After lyophilization, aliquots of HPLC fractions were analyzed by SDS-PAGE and N-terminal sequencing on an Applied Biosystems 494A Procise sequencer (PE Biosystems).
Structural and Calcium Binding NMR Studies of the cbEGF29 -31 Constructs-1 H NMR experiments were performed at 500 MHz using a home-built spectrometer in the Oxford Centre for Molecular Sciences NMR facility. Wild-type and C1977R and C1977Y mutant cbEGF29 -31 samples were dissolved in 550 l of matrix solution (99.9% D 2 O containing 5 mM Tris-HCl and 150 mM NaCl, pH 6.5) to produce final protein concentrations of 0.37 mM. Before data collection, 6 l of a 10% dioxane solution was added to the samples for chemical shift referencing. Calcium titrations were performed by adding small aliquots (5-20 l) of CaCl 2 solutions in D 2 O in 100 M to 1 mM increments. Samples containing 10 mM EDTA were also prepared to ensure that the protein had no bound calcium. Differences observed in the NMR spectra, collected in the absence of calcium and the presence of 10 mM EDTA, indicated small amounts of contaminating calcium in the sample. In these cases, a correction was made in the K d determination. The sample was maintained at pH 6.5 Ϯ 0.2 and 150 mM NaCl was present in the sample buffer to maintain approximate physiological ionic strength (I ϭ ϳ0.15) throughout the experiments. All of the experiments were conducted at 35°C. The K d value of the low affinity site in cbEGF29 was calculated as described previously (21).
One-dimensional NMR data were collected with a spectral width of 6666.67 Hz, 4,096 complex points, and 256 acquisitions. Two-dimensional spectra were recorded with a spectral width of 6666.67 Hz and 1,024 complex points in F 2 . In NOESY and total correlation spectroscopy experiments, 96 acquisitions and 256 complex points in F 1 were collected. In correlated spectroscopy experiments, 64 or 80 acquisitions and 400 complex points in F 1 were collected. Mixing times of 150 and 40 ms were used in the NOESY and total correlation spectra, respectively. Water suppression in all experiments was achieved using presaturation during the 1.2-s recycle delay. All NMR data were processed using Felix 2.3 (Accelrys, Inc., San Diego, CA). Two-dimensional spectra were zero-filled to 4,096 complex points in the F 2 dimension and to 2,048 complex points in F 1 to yield a digital resolution of 0.0033 ppm/point in F 2 and 0.0066 ppm/point in F 1 . A table containing the assignments of aromatic peaks for wild-type cbEGF29 -31 is provided in the supplementary data.
Measurement of Calcium Dissociation Constants (K d ) using Chromophoric Chelators-K d values of the high affinity sites were determined using the method of Linse et al. (25). The dissociation constant of the chromophoric chelator 5,5Ј-Br 2 BAPTA was determined using data from four blank titrations. The value of 1.6 M is in agreement with previous measurements (28). Calcium-free solutions of proteins in Ca 2ϩ -free buffer (5 mM Tris, pH 7.5, 150 mM NaCl) were titrated with Ca 2ϩ -stock buffer (5 mM Tris, pH 7.5, 1 mM CaCl 2 , 135 mM NaCl) in the presence of the chromophoric chelator 5,5Ј-Br 2 BAPTA. All of the titrations were performed with 25-30 M chelator and 25-30 M protein at room temperature (approximately 23°C) using a Shimadzu UV mini 1240 spectrophotometer. The K d values were calculated by least squares fitting to the data as described previously using in-house software (25)(26)(27)(28). The microscopic K d values for the wild-type protein were calculated from the macroscopic K d values using the relationships in Linse et al. (25) assuming two noninteracting calcium-binding sites. The K d measurements were repeated four times to provide an estimate of the experimental error. The statistical F test was used to determine whether a model with two calcium-binding sites gave a statistically significant improvement in 2 compared with a one-site model.

RESULTS
Purification of cbEGF29 -31 Wild-type, C1977R, and C1977Y Triple-domain Constructs-The cbEGF29 -31 wild-type and mutant triple-domain constructs of human fibrillin-1 ( Fig. 1) were expressed and purified using a previously described method (21)(22)(23)(24). After reduction and refolding in vitro, each construct was purified by reverse phase HPLC, and the major species was collected. The presence of Ca 2ϩ in the refolding buffer was found to be essential for the production of a single peak in the HPLC chromatogram for all three constructs; refolding in the presence of EGTA resulted in multiple peaks, suggesting multiple conformations. Analysis of purified material on nonreducing and reducing SDS-PAGE gels (data not shown) confirmed the presence of a single, major protein species. The purified proteins were analyzed by electrospray mass spectrometry, and the experimental molecular masses were found to agree well with the predicted values ( Table I). The analysis of the C1977Y and C1977R constructs by mass spectrometry showed an increased mass equivalent to a covalently bound cysteine residue. This is consistent with oxidation of the unpaired thiol in cbEGF29 -31 by free cysteine present in the refolding reaction mixture.
Limited Proteolysis of cbEGF29 -31 Mutant Proteins Indicates Disruption of cbEGF30 but Not cbEGF31-In the presence of Ca 2ϩ , fibrillin cbEGF domains are resistant to proteolysis (6,8,9,19,20,24,29). Digestion of cbEGF constructs in the presence and absence of Ca 2ϩ and subsequent identification of cleaved N termini can therefore be used to identify the structural effects of amino acid substitutions on each domain. Comparative digestion by trypsin in the presence of EGTA (50 mM) or CaCl 2 (50 mM) followed by SDS-PAGE analysis revealed significant protection by Ca 2ϩ against proteolysis in the wildtype construct but increased susceptibility to proteolysis in the two mutant constructs ( Fig. 2A). Similar SDS-PAGE analysis of endoproteinase Glu-C, elastase, and chymotrypsin digests also demonstrated loss of protection by Ca 2ϩ in the mutant constructs, whereas the wild-type constructs were protected by Ca 2ϩ against proteolysis (data not shown).
N-terminal sequence analysis was performed on the HPLCpurified trypsin digestion products obtained in the presence of EGTA or Ca 2ϩ as described previously (24). The proteins were purified under nonreducing conditions, and quantitative Nterminal sequencing was used to determine the amount of each cleavage product (Table II). Comparative analysis of these data are shown in Fig. 2B. Upon tryptic digestion, four cleavage sites were mapped in the protein constructs. In the wild-type digests, the cleavage sites 1983 KCAPG and 1998 CICPP in cbEGF30 and 2038 CLCPE in cbEGF31 were only detected in the absence of CaCl 2. This indicated that the cbEGF30 and cbEGF31 domains of the wild-type construct were protected by Ca 2ϩ against proteolysis, consistent with a native calciumbinding fold for each of these domains. The positions of these sites are indicated on a schematic structure and a three-dimensional model of the triple construct in Fig. 3. The fourth cleavage site, 2052 RCQ, was not protected against proteolysis by Ca 2ϩ in any of the proteins. This sequence is located close to the C terminus of cbEGF31 and most likely reflects the flexibility commonly observed at the termini of recombinant domain constructs (30).
In the tryptic digests of the C1977Y and C1977R mutants, the same four cleavage sites were observed in the absence of Ca 2ϩ as were seen in the wild-type protein. One of these sites,  Table II. 2038 CLCPE located in cbEGF31, was significantly protected against proteolysis in Ca 2ϩ and suggested that the C-terminal cbEGF31 domain of the mutant proteins contained a nativelike calcium-binding fold. However, in contrast to the wild-type protein, the cleavage sites 1983 KCAPG and 1998 CICPP in cbEGF30 were detected in significant amounts in Ca 2ϩ (Table  II and Fig. 2B). This demonstrated new proteolytic susceptibility in cbEGF30 and suggested a loss of Ca 2ϩ binding to this domain. No cleavage sites in the cbEGF29 domain were identified with trypsin or with the subsequent use of endoproteinase Glu C, chymotrypsin, or elastase (data not shown), and it was therefore not possible to determine whether or not the C1977Y and C1977R substitutions affected the structural integrity of the N-terminal cbEGF29 domain.

One-dimensional NMR Analysis Identifies Both Low and High Affinity Calcium-binding Sites in Wild-type and Mutant
Proteins-NMR spectroscopy has been used previously in many studies to probe the structure and calcium binding properties of wild-type and mutant cbEGF domains (5, 7, 21-24, 30 -32). One-dimensional and two-dimensional 1 H NMR spectra of the wild-type and mutant cbEGF29 -31 constructs were collected in the absence and presence of varying amounts of Ca 2ϩ . All three proteins gave NMR spectra consistent with a folded protein; upfield shifted methyl resonances, which are characteristic of the close packing of methyl and aromatic groups, and downfield shifted H ␣ resonances, which are characteristic of the antiparallel ␤-sheet secondary structure of a cbEGF domain (21), were observed. The spectra of the mutant constructs were very similar to those of the wild-type protein, suggesting the presence of folded domains within the mutant proteins.
Changes in the aromatic region of the NMR spectrum (6 -8 ppm), specifically shifts of resonances assigned to the calciumbinding consensus aromatic residue at the N terminus of the cbEGF domain, have been used previously to characterize Ca 2ϩ binding (21,23,31,33,34). In this study, equivalent residues within cbEGF 29 -31 are Phe 1954 , Tyr 1996 , and Phe 2036 (Fig. 3, asterisks). Initially, in the absence of specific assignments for these residues, the aromatic region of one-dimensional spectra of the wild-type and mutant proteins in selected Ca 2ϩ concentrations was examined (Fig. 4). In the wild-type protein, spectral changes were observed in both low (200 M) and high (50 mM) concentrations of Ca 2ϩ , indicative of both high and low affinity calcium-binding sites. The spectra of the mutant proteins also showed characteristics of both high and low affinity calcium-binding sites. However, some of the spectral changes observed for the wild-type protein were not observed for the FIG. 3. Schematic representation of the structure of cbEGF29 -31 and a three-dimensional model of cbEGF29 -31. A, structure of cbEGF29-31 indicating the bound calcium ions (red) and the aromatic residues used as structural markers in the NMR study (tyrosines are green, and phenylalanines are blue). The calciumbinding consensus aromatics are marked by asterisks. Major cleavage sites identified from tryptic digests of wild-type, C1977R, and C1977Y constructs are indicated with arrows on this diagram. Protease analyses were performed in both 50 mM EGTA and 50 mM CaCl 2 . The cleavage sites in all three constructs were conserved, and the sites that lose calcium protection in the mutant constructs are shown with black arrows. The cleavage site that demonstrates calcium protection in both the mutant and wild-type constructs is indicated with a white arrow. The disulfide bonds in cbEGF30 are indicated with dashed lines because these may not form correctly in the mutant constructs. The mutated residue and the disrupted disulfide bond in the C1977R and C1977Y constructs are indicated in red. B, predicted tertiary structure of cbEGF29-31 with the position of aromatic markers and cleavage sites indicated as in A. This model has been constructed using the co-ordinates from the cbEGF32-33 domain pair (5). The calcium ions are shown as red spheres. This figure was rendered (38) from MOLSCRIPT (39). mutants, indicating a change in calcium binding properties.

Domain-specific Assignments of Aromatic Resonances in cbEGF29 -31 Identification of a Low Affinity Calcium-binding
Site in cbEGF29 of Wild-type and Mutant Proteins-More detailed, domain-specific information about the calcium binding affinities of cbEGF29 -31 was obtained using two-dimensional correlated spectroscopy, total correlation spectroscopy, and NOESY experiments and the aromatic residues within cbEGF29 -31. In addition to the calcium-binding consensus aromatic residue, which can be used to measure intradomain calcium binding, another consensus aromatic residue closer to the C terminus of each domain (Tyr 1962 in cbEGF29 and Tyr 2004 in cbEGF30; Fig. 3) is predicted by homology to be involved in an interdomain packing interaction with the following domain in the calcium-bound protein (5, 7). The chemical shift of this residue is therefore sensitive to calcium binding to the following domain. The spin systems of all three tyrosines and two of the three phenylalanine residues in cbEGF29 -31 were identified using correlated and total correlation spectra collected in 10 mM EDTA and in 50 mM CaCl 2 ; integration of one-dimensional spectra indicated that the peaks of the third phenylalanine residue overlap at ϳ6.7 ppm and do not give resolved cross-peaks in two-dimensional spectra. Residue-specific assignments for the tyrosine and phenylalanine spin systems (see supplemental data) were made by analysis of their behavior in calcium titrations, comparison of the wild-type and mutant spectra and comparison with spectra obtained for a related triple-domain construct cbEGF28 -30. 2 One of the peaks assigned to Phe 1954 in cbEGF29 is observed to shift in the one-dimensional NMR spectra shown in Fig. 4 at calcium concentrations up to 50 mM; this is indicative of a low affinity calcium-binding site. The Phe 1954 peak shows similar behavior in the C1977Y and C1977R mutant spectra (Fig. 4), indicating that low affinity calcium binding is maintained in the cbEGF29 domain. The chemical shift changes of the Phe 1954 peak from its starting position in 10 mM EDTA were plotted against the free calcium concentrations in the protein The method used to calculate free calcium concentration and derive the values for dissociation constants has been previously described (18). Calculated K d values were ϳ1.5 mM for the wild-type construct, ϳ1.8 mM for the C1977R construct, and ϳ2 mM for the C1977Y construct. samples to obtain K d values (21). The estimated K d values were within the 1.5-2 mM range in all three constructs (Fig. 5), as expected for an N-terminal cbEGF domain. It is likely in the native protein, however, that covalent linkage of cbEGF28 will enhance the affinity of cbEGF29 for calcium.

Two-dimensional NMR Identifies High Affinity Calcium Binding in cbEGF31 of Wild-type and Mutant Proteins-
NOESY spectra collected at low Ca 2ϩ concentrations (100 -500 M) show behavior characteristic of slow chemical exchange (Fig. 6); cross-peaks (colored red and dark blue in Fig. 6) corresponding to the Ca 2ϩ -free species observed in EDTA decrease in intensity as Ca 2ϩ is added and new cross-peaks (colored in violet in Fig. 6) corresponding to the Ca 2ϩ -bound species increase in intensity upon the addition of Ca 2ϩ . In addition, a number of exchange peaks increase in intensity as Ca 2ϩ is added and then disappear when the high affinity calciumbinding sites are saturated. These exchange peaks (colored green and cyan in Fig. 6) arise as a result of the interconversion of Ca 2ϩ -free and Ca 2ϩ -bound forms of the protein during the mixing time of the NOESY experiment, and they are useful for the purposes of assignment because they correlate the chemical shifts of a particular proton in the free and bound forms (35).
This type of slow exchange behavior, characteristic of high affinity calcium binding in the micromolar range, can be seen clearly for Tyr 1996 of cbEGF30 and for Phe 2036 of cbEGF31 in wild-type spectra. The exchange peak for Phe 2036 (shown in green in Fig. 6A) is visible at a lower Ca 2ϩ concentration than the exchange peak for Tyr 1996 (shown in cyan in Fig. 6B). These residues are markers for Ca 2ϩ binding in their respective domains. Thus, the cbEGF31 domain appears to have a higher Ca 2ϩ affinity than the cbEGF30 domain within the cbEGF29 -31 wild-type construct. In the C1977R (Fig. 6) and C1977Y (data not shown) mutant spectra, the exchange peak and high affinity behavior is observed for Phe 2036 , indicating that high affinity binding is maintained in cbEGF31 in these mutant proteins.
The behavior of the peaks corresponding to Tyr 1996 in the mutants, however, differs from that observed in the wild-type protein. No exchange peak or calcium-dependent chemical shift changes are observed for Tyr 1996 in the spectra of the mutants. This indicates that calcium binding within the cbEGF30 domain is lost as a result of the cysteine substitutions. Multiple cross-peaks, corresponding to Tyr 1996 , are observed in the spectra of the mutants, which may suggest a mixture of conforma-FIG. 6. Aromatic region of two-dimensional 1 H NMR spectra of the cbEGF29 -31 wild-type (WT) and C1977R constructs. Twodimensional NOESY spectra were recorded at varying concentrations of CaCl 2. Wild-type spectra collected with 0 (A), 200 (B), or 400 M CaCl 2 (C) and C1977R spectra collected with 0 (D), 100 (E), and 200 M CaCl 2 (F) are presented as an example. Minor differences in the cross-peaks in 0 CaCl 2 spectra compared with the 10 mM EDTA spectra were observed, which indicated a small amount of residual CaCl 2 present in the samples. The cross-peaks for Phe 2036 in the Ca 2ϩ -free and Ca 2ϩ -bound forms are colored in red and violet, respectively. The cross-peaks for Tyr 2004 in the Ca 2ϩ -free and Ca 2ϩ -bound forms are colored in gray and orange, respectively. The cross-peaks for Tyr 1996 in the Ca 2ϩ -free form are colored dark blue; the Ca 2ϩ -bound peak is located close to the diagonal and is indicated by an arrow in B and C. The exchange peaks, which link the free and bound states of the respective aromatic residue, are observed. The exchange peaks for Tyr 1996 and Phe 2036 are colored cyan and green, respectively. The cross-peaks for Phe 1954 , which shows low affinity behavior, are labeled. The mutant spectra are similar to the wild-type spectra with the exception of the region (indicated with a bracket) where multiple peaks are observed for Tyr 1996 , indicating the possibility of multiple species for the cbEGF30 domain. tions in the cbEGF30 domain (Fig. 6).
Preservation of cbEGF30/31 but Not cbEGF29/30 Domain Packing Is Observed in C1977R and C1977Y Mutant Proteins-By structural homology to other fibrillin-1 cbEGF domain pairs, residues Tyr 1962 in cbEGF29 and Tyr 2004 in cbEGF30 (Fig. 3) should be involved in interdomain packing interactions in the calcium-bound form of the protein, and their NMR chemical shift behavior in the calcium titration should reflect the affinity for Ca 2ϩ of the following domains, cbEGF30 and cbEGF31, respectively. The behavior of the peaks corresponding to Tyr 1962 in the wild-type protein is consistent with the high affinity binding of calcium in cbEGF30 (Fig. 4). In the C1977R and C1977Y spectra, the peaks corresponding to Tyr 1962 do not change as calcium is added (Fig. 4). This is consistent with the loss of calcium binding in the cbEGF30 domain and indicates that no packing interaction between cbEGF29 and cbEGF30 is formed in the mutant proteins. The peaks corresponding to Tyr 2004 show high affinity calcium binding in the spectra of the wild-type and mutant proteins (Fig. 6). This indicates that the interdomain packing interaction between cbEGF30 and cbEGF31 is maintained in the C1977Y and C1977R mutant proteins and is consistent with the high affinity site identified in cbEGF31 using Phe 2036 as a marker of calcium binding.

Estimation of K d Values for High Affinity Ca 2ϩ -binding Sites
Using Chemical Chelation-Quantitative information on K d values in cbEGF30 and cbEGF31 could not be gained using NMR because of difficulties in measuring peak heights for the free and bound cross-peaks. Therefore, calcium titration assays were performed using the chromophoric chelator, 5,5Ј-Br 2 BAPTA (25-28,36) (Fig. 7). For the wild-type fragment, the data gave a good fit to a model with two calcium-binding sites with K d values of 0.3 Ϯ 0.05 and 3.7 Ϯ 1.0 M (values expressed as the means Ϯ S.D.). The NMR spectra indicate that the higher affinity calcium-binding site (0.3 M) is located in cbEGF31, and the lower affinity site is located in cbEGF30. The C1977R and C1977Y constructs were each found to have only a single high affinity calcium-binding site, with K d values of 0.7 Ϯ 0.15 and 0.8 Ϯ 0.6 M, respectively. This result is consistent both with the NMR studies, which show a loss of calcium binding in the cbEGF30 domain in the mutant proteins, and the limited proteolysis data. Although minor effects on the calcium binding affinities of the adjacent cbEGF domains can-not be excluded, collectively these results show that the C1977R and C1977Y mutations do not significantly affect the calcium binding properties of cbEGF29 and cbEGF31. DISCUSSION Cysteine substitutions in cbEGF domains are a significant cause of disease in a number of different extracellular and transmembrane proteins. Although predicted to cause domain misfolding, there have been few studies of their structural consequences. Here we have used low and high resolution techniques to examine the effects of two pathogenic amino acid changes, C1977R and C1977Y, in human fibrillin-1. Each substitution was expressed in a triple-domain construct, cbEGF29 -31, where the mutant cbEGF30 domain is placed in a native-like context. Initial comparative analysis by limited proteolysis identified structural changes within cbEGF30, consistent with a degree of misfolding in the mutant domain. Further analysis by NMR revealed that despite the disruption of the 1-3 disulfide bond in cbEGF domain 30, relatively localized structural effects of each Cys substitution were observed.
By observing spectral changes of assigned aromatic resonances within cbEGF29 -31 on the addition of Ca 2ϩ , it was possible to identify a normal interdomain packing interaction of cbEGF30 -31. However, the absence of spectral changes for Tyr 1962 , located in the minor ␤-sheet of cbEGF29, indicated that the interdomain packing interaction between cbEGF29 and cbEGF30 was missing in the mutant proteins. Despite this disruption, calcium binding to cbEGF29 in the mutant proteins was indistinguishable from wild-type protein, suggesting that each cysteine substitution did not affect the N-terminal region of this domain. In addition, a high affinity calcium-binding site was observed in cbEGF31 of both mutant and wild-type proteins, which gave rise to slow exchange peaks in NMR spectra. Measurement of the K d value of this site in each protein using chromophoric chelation gave similar values of 0.3-0.8 M. The retention of a high affinity calcium-binding site in cbEGF31 of the mutant proteins, together with the observed hydrophobic packing interaction between cbEGF30 -31 characteristic of tandem cbEGF pairs, confirmed that the cysteine substitutions did not result in disruption of the C-terminal region of the cbEGF29 -31 protein.
It is interesting to speculate how the native-like properties of the C-terminal region are retained in the mutant proteins. Our data suggest that, although the mutant domain was misfolded as a consequence of the 1-3 disulfide disruption, the 5-6 disulfide bond of cbEGF30 may have been formed correctly. This would then allow the packing interaction of cbEGF30 and cbEGF31, which is required for high affinity calcium binding to cbEGF31, to form (5, 7). Thus, it can be speculated that if cysteine substitutions disrupting the 5-6 disulfide bond of cbEGF30 were to be introduced, different structural effects, including destabilization of the interdomain interface and a decrease in calcium affinity, might be observed for the C-terminal domain. In support of this, a recent study by Vollbrandt et al. (20) demonstrated that a C750G substitution that affects the C-5 residue of cbEGF7 (and therefore disrupts the 5-6 disulfide bond) caused increased proteolytic susceptibility of cbEGF8. Although calcium binding to this fragment was not measured, it is likely that the increased proteolysis observed occurs as a result of the disruption of domain packing between cbEGF7 and 8 and the decrease in calcium binding affinity in cbEGF8. As observed previously for mutations that specifically affect calcium-binding residues within fibrillin-1 cbEGF domains, cysteine substitutions also have the potential to cause considerable structural heterogeneity. This may result in a variety of pathogenic mechanisms, despite the apparent similarity of the mutation. Protease digestion studies and NMR analyses have previously been used to probe the structural and calcium binding properties of wild-type and mutant fibrillin-1 cbEGF-containing fragments. It has been shown that the effect of the G1127S substitution in cbEGF13 on protease susceptibility is confined to the mutant domain, and this domain retains a native-like fold (22). In contrast, a calcium-binding mutation E1073K in cbEGF12, which causes severe neonatal MFS, demonstrated a longer range structural effect where a new protease-sensitive site in the N terminus of the adjacent domain was revealed (6). The effects of the cysteine substitutions studied here appear unique, because the calcium binding properties of the mutant domain and the 29 -30 domain interface are disrupted, but effects on the N-terminal cbEGF29 calcium-binding site are not observed.
In addition to identifying structural consequences of specific cysteine substitutions, this study has provided further insight into the calcium binding properties of fibrillin-1. Previous studies have indicated high affinity calcium-binding sites in cbEGF13 and 14 (Ͻ30 and Ͻ100 M) (24) of the "neonatal region" and a moderate affinity site in cbEGF33 (ϳ350 M) (21). The analyses performed on the wild-type fragment in this study demonstrate that other fibrillin cbEGF domains such as cbEGF30 and 31, located away from the neonatal region, have a high affinity for calcium. Thus, although the high affinity observed in the neonatal region may be important for function, it is not a unique property of this region.
Pulse-chase studies on MFS patient fibroblasts containing cysteine substitutions in fibrillin-1 cbEGF domains have commonly reported normal synthesis and a delay in secretion of fibrillin-1 that leads to severe reduction of matrix deposition (14 -18). It was suggested that the delay in secretion might be due to targeted recognition of misfolded protein. In contrast, a smaller number of cases (24% of those studied) have been found in MFS patients with cysteine substitutions where normal secretion of fibrillin-1 has been observed (14,18). The C1977R substitution is one example where pulse-chase analysis of patient fibroblasts harboring this mutation showed normal synthesis, secretion, and deposition of fibrillin-1 compared with normal fibroblasts (Ref. 37 and data not shown). One may speculate that this is due to the relatively localized effects of the mutation.
Our data, and that of others (20), indicating the structural heterogeneity of this group of mutations suggests that both the disulfide bond affected and the degree of structural destabilization are factors that determine the pathogenic role of the mutant protein. The misfolded domain of a mutant protein may be recognized and targeted to degradation pathways by intracellular mechanisms or may be retained within the cell. In contrast, partially misfolded fibrillin-1 may escape quality control surveillance in the cell. On encountering the extracellular space, mutant proteins may be rapidly degraded by proteases in the surrounding environment or may subsequently disrupt a specific protein-protein interaction required for the assembly of fibrillin-1 or interactions of microfibrils with other cell-matrix components. The structural effects of misfolding mutations in other proteins containing tandem repeats of EGF domains may also be heterogeneous and lead to more complex pathogenic mechanisms than previously thought.