A G1127S change in calcium-binding epidermal growth factor-like domain 13 of human fibrillin-1 causes short range conformational effects.

Human fibrillin-1, an extracellular matrix glycoprotein, has a modular organization that includes 43 calcium-binding epidermal growth factor-like (cbEGF) domains arranged as multiple tandem repeats. A missense mutation that changes a highly conserved glycine to serine (G1127S) has been identified in cbEGF13, which results in a variant of Marfan syndrome, a connective tissue disease. Previous experiments on isolated cbEGF13 and a cbEGF13-14 pair indicated that the G1127S mutation caused defective folding of cbEGF13 but not cbEGF14. We have used limited proteolysis methods and two-dimensional NMR spectroscopy to identify the structural consequences of this mutation in a covalently linked cbEGF12-13 pair and a cbEGF12-14 triple domain construct. Protease digestion studies of the cbEGF12-13 G1127S mutant pair indicated that both cbEGF12 and 13 retained similar calcium binding properties and thus tertiary structure to the normal domain pair, because all identified cleavage sites showed calcium-dependent protection from proteolysis. However, small changes in the conformation of cbEGF13 G1127S, revealed by the presence of a new protease-sensitive site and comparative two-dimensional NOESY data, suggested that the fold of the mutant domain was not identical to the wild-type, but was native-like. Additional cleavage sites identified in cbEGF12-14 G1127S indicated further subtle changes within the mutant domain but not the flanking domains. We have concluded the following in this study. (i) Covalent linkage of cbEGF12 preserves the native-like fold of cbEGF13 G1127S and (ii) conformational effects introduced by G1127S are localized to cbEGF13. This study demonstrates that missense mutations in fibrillin-1 cbEGF domains can cause short range structural effects in addition to long range effects previously observed with a E1073K mutation in cbEGF12.

The epidermal growth factor-like (EGF) 1 domain is a widely distributed module found in transmembrane and extracellular proteins (1) where it may occur as multiple tandem repeats. The module is characterized by six highly conserved cysteine residues, which normally disulfide bond in a 1-3, 2-4, 5-6 arrangement and stabilize the global fold of the domain. A subset of these domains are distinguished by the presence of a calcium binding consensus sequence (D/N)X(D/N)(E/Q)X m (D/ N*)X n (Y/F), where m and n are variable, and an asterisk indicates a possible ␤-hydroxylation site (2)(3)(4). In tandem repeats of fibrillin calcium-binding EGF (cbEGF) domains, the bound Ca 2ϩ performs a key structural role in restricting interdomain flexibility, which may facilitate protein-protein interactions (5,6) and also protect the modules against proteolytic cleavage in vitro (7,8). The biological importance of the cbEGF domain is highlighted by the number of diseases that are caused by missense mutations within this domain type. These include Marfan syndrome (MFS) and related disorders, congenital contractural arachnodactyly, hemophilia B, familial hypercholesterolemia, "CADASIL" (cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy) (mutations in fibrillin-1 (9), fibrillin-2 (10), factor IX (11), low density lipoprotein receptor (12), and Notch 3 (13), respectively), and protein S (14) and protein C deficiencies (15).
Fibrillin-1, a 350-kDa extracellular matrix glycoprotein, is a major structural component of 10 -12-nm connective tissue microfibrils (16). It has a modular organization and is mainly composed of multiple tandem arrays of EGF domains, the majority of which contain calcium binding sites (Fig. 1). In addition there are seven transforming growth factor ␤-binding protein-like (TB) domains, two hybrid domains, a proline-rich region, and N and C termini with homology to other matrix proteins (17). Mutations within the fibrillin-1 gene (FBN-1) result in a spectrum of connective tissue disorders (fibrillinopathies), which range in severity from MFS, Shprintzen-Goldberg syndrome, and ectopia lentis to familial ascending aortic aneurysm (18). Of particular interest are the structural consequences of a missense mutation, which changes a highly conserved glycine to a serine in cbEGF13 of human fibrillin-1. The G1127S mutation produces a variant of the MFS phenotype and has been identified as a risk factor for ascending aortic aneurysm and dissection (19). Interestingly, the same mutation has been identified in human factor IX (G60S) where it is associated with mild hemophilia B (20). The location of G1127S within the "neonatal" region of fibrillin-1 (Fig. 1), where MFS * This work was supported in part by the Wellcome Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ causing mutations are associated with extreme phenotypic diversity, suggests that structural investigations of this region may yield important insights into the mechanism of disease.
A previous study has shown that the G1127S mutation resulted in defective folding in vitro when introduced into the single cbEGF domain 13. When present in the covalently linked cbEGF13-14 domain pair, the C-terminal cbEGF14 domain adopted the native fold and retained calcium binding properties despite the fact that the adjacent domain 13 was misfolded. This suggested that the effects of the mutation were localized (21). Here the structural consequences of the G1127S mutation in the cbEGF12-13 domain pair and also in the triple domain fragment, cbEGF12-14, have been investigated, using a combination of protease digestion studies and two-dimensional NMR methods. The data indicate that covalent linkage of cbEGF12 moderates the effect of G1127S in cbEGF13 resulting in localized changes to the structure and calcium binding properties of domain 13, but not cbEGF12 or 14. The implications of such short range effects for the pathogenic mechanism of MFS are discussed.

Cloning of Wild-type and Mutant cbEGF Domain Constructs from
Human Fibrillin-1-DNA fragments (nucleotides 3338 -3595 and 3338 -3721 of human fibrillin-1 cDNA) encoding the wild-type sequences of the cbEGF12-13 domain pair and the cbEGF12-14 triple construct (residues 1069 -1154 and 1069 -1196 respectively, numbering according to Ref. 22) were amplified by standard polymerase chain reaction techniques using Pfu polymerase (Stratagene). The forward primer in cbEGF12 and the reverse primers for cbEGF13 and cbEGF14, together with the cloning procedures used, were as described previously (21,23).
The G1127S mutation was introduced into the pQE30 recombinant plasmids containing the cDNA sequences of the cbEGF12-13 double or cbEGF12-14 triple constructs by PCR-based site-directed mutagenesis. The plasmids were amplified using a forward primer: 5Ј-GAGGTAGT-GTTTGCCATAAC-3Ј and a reverse primer: 5Ј-GGCATAGGAGAG-GATC-3Ј. The amplified DNA was purified from a 0.7% agarose gel with a Qiaex II gel extraction kit (Qiagen) and ligated and transformed into Escherichia coli NM554[pREP4]. Clones were sequenced to confirm the mutation had been introduced with no other changes into the inserted fragments.
Protein purification, refolding, and His-tag cleavage were carried out as described previously (21), except that the cleaved cbEGF12-14 triple construct was further purified on a fast protein liquid chromatography (FPLC) MonoQ column in 50 mM Tris, pH 7.5, using a 0 -0.5 M NaCl gradient prior to a final reverse phase HPLC step. The identity of the purified products was confirmed by mass spectrometry (Table I).
Proteolysis of Wild-type and Mutant cbEGF Domain Constructs-Peptides were dissolved to give a final protein concentration of 3.2 mg/ml in a total volume of 100 l containing 100 mM NaCl, 50 mM Tris pH 7.5, and either 10 mM EGTA or 10 mM CaCl 2 . Incubation with trypsin (EC 3.4.21.4; bovine pancreatic; TPCK-treated; Sigma; 1:1000, w/w) or endoproteinase GluC from Staphylococcus aureus (EC 3.4.21.19; Sigma; 1:100, w/w) was carried out at 37°C. At the times indicated, samples (10 l) were withdrawn, and the reaction was stopped by the addition of 2ϫ SDS sample buffer containing 100 mM dithiothreitol and heating at 95°C for 4 min. The reaction mixtures were analyzed by 16% SDS-PAGE and visualized by Coomassie Blue staining.
For N-terminal sequence analysis of the cleavage sites, samples were digested for 30 min in the presence of EGTA or calcium, the reaction was stopped by boiling for 4 min, and the digest products were purified under non-reducing conditions by reverse phase HPLC. Samples were analyzed by SDS-PAGE, and N-terminal sequencing was carried out using automated Edman chemistry. The samples were absorbed onto PVDF (polyvinlidene difluoride (0.2-m porosity) using a ProSorb cartridge (PE Biosystems, Warrington, UK.) following the manufacturer's protocol. The membrane-bound samples were then excised from the ProSorb cartridge, and N-terminal sequencing was performed on an Applied Biosystems 494A 'Procise' sequencer (PE Biosystems). Cysteine residues reported in the N-terminal sequences were inferred from the published sequence (22).
Structural and Calcium Binding NMR Studies of the cbEGF12-13 Domain Pair-All NMR spectra were recorded on a home-built/GE Omega spectrometer equipped with self-shielded pulsed field gradients at 600 MHz. The wild-type and G1127S mutant cbEGF12-13 domain pair protein samples were dissolved in 550 l of 90% H 2 O, 10% 2 H 2 O, 5 mM Tris-HCl, pH 6.5, to yield final protein concentrations of 2.00 and  1.93 mM, respectively. Added NaCl was not utilized in these investigations because high ionic strength may compromise spectral quality. Two-dimensional NOESY spectra (24,25) were acquired for both samples at 0 mM CaCl 2 and 12.5 mM CaCl 2 to allow qualitative assessment of calcium binding. At 12.5 mM CaCl 2 , the two calcium binding sites of the wild-type domain pair are saturated based on previous calcium binding studies of the cbEGF12-13 domain pair (26), and these calciumloaded spectra were also used to assess the structural consequences of the G1127S mutation. All spectra were recorded with a mixing time of 150 ms at T ϭ 33°C. Water suppression was achieved using field gradients (27). Data were processed using Felix 2.3 (Biosym, Inc.). 1024 complex points were acquired in F 2 and F 1 for each experiment with a spectral width of 8000 Hz in each dimension. Spectra were referenced with respect to the H 2 HO resonance and were zero-filled to 8 K in the F 2 dimension to yield a digital resolution of 0.98 Hz/pt.

Expression and Purification of Wild-type and Mutant
cbEGF Domain Constructs-The predicted structure of the cbEGF12-14 triple construct analyzed in this study, modeled on the coordinates of the cbEGF32-33 domain pair from human fibrillin-1, is shown in Fig. 2 (5,23). This model has been validated by NMR structural studies of the cbEGF12-13 pair. 2 The G1127S mutation in domain 13 was introduced into the cbEGF12-13 double and cbEGF12-14 triple constructs at the position indicated. The wild-type and mutant domain pairs and triple constructs were expressed and purified as described previously (21). After reduction and refolding in vitro followed by reverse phase HPLC, the cbEGF12-13 domain pairs and cbEGF12-14 triple constructs had slightly different elution times. However, there was no difference in the elution profile of each mutant from the corresponding wild-type construct. On in vitro refolding in the presence of Ca 2ϩ each gave one major species, the molecular mass of which was confirmed by mass spectrometry (Table I). Together, the characteristic change in the elution profile on refolding of the pair and triple constructs, the observed Ca 2ϩ -dependent protection against proteolysis, and the NMR analysis of the wild-type cbEGF12-13 domain pair (see below) were all indicative of correctly folded cbEGF domains (28).
Protease Digestion of the cbEGF12-13 Domain Pairs-Digestion and subsequent SDS-PAGE analysis of both the wild-type and G1127S mutant cbEGF12-13 constructs by trypsin in the presence of EGTA (10 mM) or Ca 2ϩ (10 mM) showed that there was significant protection by calcium against proteolysis (Fig.  3A). Although minor differences were apparent between the wild-type and mutant constructs in the presence of EGTA, the degree of protection by Ca 2ϩ appeared equivalent. Similar SDS-PAGE analysis of endoproteinase GluC digests also demonstrated protection by calcium (data not shown). Digestion products obtained in the presence of Ca 2ϩ and EGTA were purified under non-reducing conditions by HPLC, and the Nterminal sequences of each were determined. The amount of each N terminus expressed relative to the authentic N-terminal sequence is shown in Table II. In both the mutant and wild-type cbEGF12-13, three trypsin and one endoproteinase GluC susceptible cleavage sites were located in cbEGF12. The cbEGF12-13 G1127S mutant contained an additional endoproteinase GluC site ( 1134 GSYRC) in cbEGF13. Calcium-dependent protection from proteolysis of all sites in the mutant pair, including this additional site in the mutant domain, was found, and the degree of protection was the same as in the wild-type pair (Table II).
Protease Digestion of the cbEGF12-14 Triple Domain Constructs-Trypsin digestion and subsequent SDS-PAGE analysis of both the wild-type and mutant cbEGF12-14 triple constructs in the presence of EGTA (10 mM) and Ca 2ϩ (10 mM) again showed the protection against digestion afforded by calcium (Fig. 3B). The G1127S mutation did not appear to significantly affect the calcium binding properties of the construct, because the calcium-dependent protection from proteolysis was indistinguishable from that of the wild-type on SDS-PAGE. Calcium protection was also evident on digestion by endoproteinase GluC (data not shown).
N-terminal sequence analysis of HPLC-purified digestion products obtained in the presence of EGTA or Ca 2ϩ identified the cleavage sites shown in Table II. On endoproteinase GluC digestion, an extra site 1134 GSYRC, not present in the wildtype construct, was identified in the mutant. This site, located at the turn of the central two-stranded anti-parallel ␤-sheet, was the same site revealed in the cbEGF12-13 G1127S pair. In the case of trypsin, two additional sites, 1126 GSVCH and 1138 CECPP, were seen upon digestion of the mutant construct. The additional site at 1126 GSVCH is adjacent to the mutated residue whereas the other protease sensitive site, 1138 CECPP, occurs distal to the mutation but still within cbEGF13 (Fig. 4). The comparison of the amount of each of these N termini identified on digestion in EGTA or calcium, and expressed relative to the authentic N-terminal sequence is shown in Table II. Calcium-dependent protection from proteolysis of all cleavage sites, including those in the mutant domain, was observed. All cleavage sites were represented in comparable amounts indicative of efficient cleavage of a single population of molecules.
Structural and Calcium Binding NMR Studies of the cbEGF12-13 Domain Pair-Comparative analysis of the calcium-saturated two-dimensional NOESY spectra for the wildtype and G1127S mutant cbEGF12-13 constructs indicated that, unlike the case for the cbEGF13-14 wild-type and mutant spectra (21), a significant number of peaks corresponding to domain 13 were unaffected or only mildly affected by the presence of the G1127S mutation. The fingerprint regions of these spectra are shown in Fig. 5. Most resonances appeared unaltered between the two spectra, and a subset of peaks were only slightly shifted. For example, of the two well resolved sets of triplet peaks in the lower left-hand corner of the spectra, the chemical shifts of the upper set of peaks were identical, whereas the positions of resonances in the lower triplet were slightly shifted. Because these peaks involve connectivities to downfield-shifted C␣H, which are typically involved in ␤-structure, the data suggest that domain 12 in the mutant cbEGF12-13 pair is unaffected by the presence of the G1127S mutation, and that the ␤-sheet region of domain 13 is only mildly affected.
To examine calcium binding properties of each construct, the aromatic regions of the two-dimensional NOESY spectra acquired at 0 and 12.5 mM CaCl 2 were overlaid for both the wild-type and G1127S mutant cbEGF12-13 domain pair (Fig.  6). Chemical shift changes of the H␦* resonances of the consensus, aromatic, calcium binding residues for the N-and Cterminal domains of each pair, i.e. Phe 1093 and Tyr 1136 , respectively, were used to assess calcium binding. The overlay of the wild-type NOESY spectra at zero and saturating calcium showed the expected, native binding properties previously observed when K d values were determined for the two sites in this domain pair (26). Comparison of the wild-type spectral overlay with that of the G1127S mutant revealed that both domain 12 and 13 retained calcium binding properties in the mutant cbEGF12-13 pair. Domain 12 appeared to retain native calcium binding properties because the Phe 1093 peak movement was identical to that seen in the wild-type spectrum. The peak movement seen for Tyr 1136 in cbEGF13 was slightly altered, however, with a larger chemical shift change associated with binding by the mutant domain. Collectively, these comparative data indicate that when preceded by cbEGF12, the G1127S mutant cbEGF13 domain preserves a native-like but not identical fold. DISCUSSION In a previous study, it was shown that G1127S caused misfolding of isolated domain 13 with loss of calcium binding properties, whereas, in a cbEGF13-14 domain pair, cbEGF14 was unaffected. In this study, the effects of the G1127S mutation on the cbEGF12-13 domain pair have been assessed, and the results indicate that domain 12 is unaltered, and the con-

TABLE II
Protease digestion products N-terminal sequences identified in purified digestion products were quantitated and expressed relative to the native N-terminal sequence of cbEGF12-14 (or cbEGF12-13). The cbEGF12-13 pair contained the same sites in cbEGF12 as the triple construct (the relative amount of each site is given in parentheses). Endoproteinase GluC revealed the site at 1134 GSYRC in cbEGF13. All sites were protected by calcium. The protease susceptible sites identified in cbEGF13 of the cbEGF12-14 GS mutant construct were protected by calcium to a similar extent as those present in cbEGF12 and cbEGF14 of the wild-type construct. Local Structural Effects of G1127S in Human Fibrillin-1 sequences of this mutation are again localized to domain 13. The demonstration that domain 13 in the 12-13 pair retains the ability to bind calcium is unlike the situation previously seen for both the single mutant cbEGF13 domain and the mutant cbEGF13-14 domain pair. As calcium binding by cbEGF domains is used as a probe for correct refolding, the fact that cbEGF13 continues to bind calcium in the mutant cbEGF12-13 pair suggests that this domain preserves a native-like fold. This hypothesis is also supported by a comparative analysis of the calcium-saturated wild-type and mutant two-dimensional NOESY spectra for the cbEGF12-13 pair, which show only minor differences (see Fig. 5).
These NMR studies indicate that the G1127S mutation has a less severe effect on folding when preceded by cbEGF12. Measurement of the calcium binding affinities of the two sites in the cbEGF12-13 wild-type pair by NMR and fluorescence spectros-copy has previously demonstrated that covalent linkage of an N-terminal cbEGF domain had a stabilizing effect on the adjacent C-terminal domain (26,6). The observation in this study that cbEGF12 moderates the effect of the G1127S folding mutation also suggests an effect of N-terminal linkage, because structural features associated with cbEGF domains (␤-sheet, calcium binding) were absent in cbEGF13 or cbEGF13-14 mutant constructs.
In parallel with NMR analyses, protease digestion studies have also been used to probe the structural and calcium binding properties of the wild-type and G1127S-containing fibrillin fragments. It has previously been shown that tandem repeats of wild-type cbEGF domains are susceptible to proteolytic cleavage at specific sites on removal of calcium by EGTA, whereas in the presence of calcium, protection from proteolysis is observed (7,29). An increased susceptibility to proteolysis in vitro also results from the introduction of specific calcium binding mutations in both recombinantly expressed polypeptides (30) and domain pairs (29). A comparison of protease digestion profiles together with the identification of how far reaching the loss of calcium-dependent protection from proteolysis is can therefore be used to assess the short versus long range consequences of various mutations. As this approach can be applied to relatively large fragments and only requires a small amount of material, it can be successfully applied to the investigation of large numbers of mutations in a more native setting.
On endoproteinase GluC digestion of the cbEGF12-13 pair, an additional cleavage site was revealed in cbEGF13 of the mutant fragment, which showed calcium-dependent protection from proteolysis. These results are consistent with the NMR analysis of this domain pair, suggesting that the mutant domain has a degree of disruption but is not severely misfolded. A comparison of the digestion patterns of the cbEGF12-14 triple construct with those of the corresponding wild-type provided additional insights. In addition to the endoproteinase GluC site in the mutant domain, also identified in the cbEGF12-13 pair, two tryptic sites were identified in the mutant triple construct indicative of further subtle changes to the mutant domain, which did not detectably affect its calcium binding properties.
The presence of the cleavage site 1134 GSYRC identified in both the 12-13 and 12-14 mutant constructs (Fig. 4) and located distal to the mutation suggests a conformational change around the ␤-sheet region of domain 13. Additional tryptic cleavage sites 1138 CECPP and 1126 GSVCH, observed only in 12-14 are supportive of this, although because of the proximity of these sites to the mutation (in the case of C 1138 via its disulfide bond to C 1124 ) we cannot formally exclude a sequence-specific effect of the mutation that enhances substrate recognition by the protease in the triple domain construct. However, an alternative explanation that the addition of a C-terminal cbEGF domain may influence the structure of the preceding mutant domain is also possible. The position and calcium-dependent properties of the cleavage sites in the flanking domains 12 and 14 were the same in the mutant and wild-type constructs, indicative of a structural change localized to cbEGF13. The protease data are therefore consistent with the NMR data for the cbEGF12-13 (this study) and cbEGF13-14 (21) domain pairs and validate this as a useful method for probing structural effects of mutations. A summary of the structural and calcium binding consequences of the G1127S mutation for cbEGF13, the cbEGF13-14 pair (21), the cbEGF12-13 pair and the cbEGF12-14 triple construct (present study) are schematically represented in Fig. 7. Further comparative NMR analyses of the structure and dynamics of both the wild-type and mutant constructs will allow the degree of disruption caused by the mutation to be defined more precisely.
The tryptic sites identified in cbEGF12 of both the pair and triple constructs were identical to those found in digests of a recombinantly expressed cbEGF10-22 construct containing a calcium binding mutation E1073K in cbEGF12 (30). E1073K is associated with neonatal MFS, the most severe form of the disease. In this case, the sites within mutant domain 12 showed enhanced susceptibility to proteolysis compared with the wild-type, in contrast to the present study in which the sites in mutant domain 13 containing the G1127S mutation were protected by calcium. In addition a cleavage site was also revealed N-terminal to cbEGF11, indicating a longer range structural effect of the calcium binding mutation. Collectively these data indicate that MFS-causing mutations in this region cause variable intramolecular effects on fibrillin-1 structure.
Because the effect of the G1127S mutation is confined to domain 13 and this domain retains a native-like fold, fibrillin-1 monomers containing this mutation are likely to be secreted by the cell and the effect of the mutation exerted either on or after incorporation into the microfibril. The small localized changes in structure and calcium binding in domain 13 could disrupt protein binding sites involved in the assembly process or the properties of the assembled microfibril. The protection afforded by calcium against proteolytic degradation of tandem repeats of cbEGF domains in fibrillin-1 (7,29) suggests that missense mutations may result in increased proteolysis in vivo. Although cbEGF13 containing the G1127S mutation appears to be more susceptible to proteolysis than the wild-type in vitro, calcium-  (21), the cbEGF12-13 domain pair, and the cbEGF12-14 triple construct (this study) are summarized. Calcium atoms are depicted as gray circles. The rectangle used to represent the state of cbEGF13 in the cbEGF12-13 and cbEGF12-14 G1127S mutant constructs highlights the difference between the native-like and native fold. A question mark is used to represent the state of cbEGF13 in the mutant triple construct, because protease sites not detectable in the cbEGF12-13 pair were observed in this construct (see text for details). dependent protection was observed for all the additional protease sites revealed in the domain (Table II), therefore increased proteolytic susceptibility seems less likely to be involved in the pathogenic mechanism than for the E1073K mutation.
In summary, interdisciplinary studies utilizing both high and low resolution methods have proved effective in identifying the structural consequences of the G1127S mutation. Further studies will now focus on identifying the functional effects of this and related mutations on fibrillin assembly.