The N1317H Substitution Associated with Leber Congenital Amaurosis Results in Impaired Interdomain Packing in Human CRB1 Epidermal Growth Factor-like (EGF) Domains*

The calcium-binding epidermal growth factor-like (cbEGF) domain is a widely occurring module in proteins of diverse function. Amino acid substitutions that disrupt its structure or calcium affinity have been associated with various disorders. The extracellular portion of CRB1, the human homologue of Drosophila Crumbs, exhibits a modular domain organization that includes EGF and cbEGF domains. The N1317H substitution in the 19th cbEGF domain of CRB1 is associated with the serious visual disorder Leber congenital amaurosis. We have investigated the structure and Ca2+ binding of recombinant wild-type and N1317H CRB1 fragments (EGF18-cbEGF19) using NMR and find that Ca2+ binding is altered, resulting in disruption of long range interactions between adjacent EGF domains in CRB1. From these observations, we propose that this substitution affects the structural integrity of CRB1 in the inter-photoreceptor matrix of the retina, where it is expressed. Furthermore, we identify disease-causing substitutions in other cbEGF-containing proteins that are likely to result in similar disruption of interdomain packing, supporting the hypothesis that the tandem cbEGF domain linkages are critical for the structure and function of proteins containing cbEGF domains.

Unsurprisingly from its localization in humans, mutations in CRB1 have been associated with a number of visual disorders including retinitis pigmentosa with (10,11) or without (11,12) preserved para-arteriolar retinal pigment epithelium, paravenous pigmented chorioretinal atrophy (13), and Leber congenital amaurosis (LCA) 3 (14,15). Of these disorders, LCA is the most severe form of inherited retinal dystrophy and is characterized by severe visual impairment from birth or very early life and a decreased or absent electroretinogram response (16). LCA is generally inherited in an autosomal recessive manner (16). Mutations in CRB1 represent a significant cause (10 -13%) of LCA among the identified LCA loci; missense, nonsense, and splice site mutations together with insertions and deletions have been identified (17,18). The majority of disease-associated missense mutations in CRB1 have been localized to the extracellular region of the protein, and substitutions associated with the various disorders are approximately evenly distributed along its length, with no unambiguous genotype-phenotype correlation ( Fig. 1) (17,18).
CRB1 is a transmembrane protein whose large extracellular portion exhibits a modular domain organization consisting of epidermal growth factor-like, calcium-binding epidermal growth factor-like, and laminin A globular domain-like modules ( Fig. 1) (6). Both Drosophila and human variants possess these domain types, although the extracellular portion of the Drosophila homologue is larger than the corresponding human protein (3,6). Deletion studies in Drosophila have shown the extracellular region to be essential to limit light-dependent photoreceptor degeneration and have shown the highly conserved cytoplasmic tail to participate in a scaffolding complex necessary for the morphogenesis and maintenance of epithelial polarity (19). Proteins homologous to those identified in the cytoplasmic Drosophila scaffold complex have been identified in mammals and shown to interact in a similar way with the cytoplasmic tail of CRB1 (9,20,21). No interacting partner has been identified for the extracellular region of mammalian CRB1 and, therefore, its precise function remains unknown. However, a missense substitution that occurs in this region of CRB1 has recently been associated, in a mouse model, with a down-regulation of a mitotic checkpoint protein (PTTG1) (22).
The epidermal growth factor-like (EGF) domain is a common structural module found in extracellular proteins (23). A subset of this domain type is the calcium-binding EGF (cbEGF) domain, which has additional residues associated with calcium binding (see Fig. 2) (24 -26). The importance of the cbEGF domain in the normal function of proteins has been demonstrated through the Ca 2ϩ dependence of the interaction between, for example, the cbEGF domains of Notch with its ligand Delta (27,28), between fibulin-1 and nidogen (29), and between fibulin-2 and fibrillin-1 (30). Disruption of this domain type through genetic mutations has been implicated in many disease states including Marfan syndrome (31), familial hypercholesterolemia (32), hemophilia B (33), and the ocular disorders Malattia Leventinese and age-related macular degeneration (34,35).
The missense mutations in cbEGF domains of CRB1 can be divided into two broad categories, cysteine and non-cysteine sequence changes ( Fig. 1) (15,17,18). Substitutions that involve the replacement of a cysteine are likely to lead to misfolding, changes in calcium binding, and an increase in proteolytic susceptibility (36,37). The second category includes conservative and non-conservative substitutions (with respect to charge and size) distributed throughout the cbEGF module. Interestingly, in contrast to cbEGF domains from other proteins such as fibrillin-1, no mutations have been identified in CRB1 that directly affect a calcium-binding residue.
Structural studies on cbEGF-containing fragments from fibrillin-1, low density lipoprotein receptor, and Notch have identified a calcium-dependent interdomain interface that is important in maintaining a rigid, rod-like arrangement of tandem domains (27,38,39). This interface may also facilitate protein-protein interactions and protect against proteolytic cleavage (39 -41). In the absence of bound Ca 2ϩ , this interface does not exist, and the rigidity of the molecule is lost (42). The interdomain interface involves the conserved aromatic residue located between cysteines 5 and 6 of the N-terminal domain (Phe-1289 in EGF18 of CRB1) and two residues located in the turn in the major ␤-hairpin between cysteines 3 and 4 of the C-terminal domain (Leu-1316 and Asn-1317 in cbEGF19 of CRB1) (see Fig. 2a). The proper formation of this interdomain interface has been implicated in high affinity calcium binding (42,43), and it has been suggested that alteration or disruption of this interface may have some involvement in disease (38,40). In CRB1, missense mutations that might affect this putative interdomain interface are observed (Fig. 1).
Here, we have undertaken an investigation to assess the effect of a disease-causing missense mutation in CRB1 on calcium binding and the interface between tandemly linked domains. The calcium binding properties of the 19th EGF module (cbEGF19) were studied in the minimum structure approximating native conditions (the tandemly linked domain pair EGF18-cbEGF19) both for the wild-type fragment and for a fragment into which the LCA-associated N1317H missense mutation (15) has been introduced. The data reveal a site of moderate affinity for Ca 2ϩ in the wild-type fragment and reveal that this affinity is significantly reduced in the mutant construct. The effect of this particular substitution can be further exacerbated by changes in the protonation state of His-1317, providing further evidence of the nature and importance of the native interdomain interface. We consider these results in the context of the calcium environment in the region of the retina to which CRB1 has been localized and propose a model for the molecular basis of the effects of this substitution.

Production of Wild-type and N1317H Variant Recombinant
Fragments EGF18-cbEGF19-A cDNA fragment (nucleotides 3898 -4143, numbering according to GenBank TM accession AY043325 (6)), encoding CRB1 domains EGF18 and cbEGF19 (residues 1255-1336), was amplified from a plasmid derived from a human fetal brain cDNA library using Pfu polymerase (Stratagene) and was cloned into bacterial expression vector pQE30 (Qiagen). Recombinant protein was expressed as a His 6 affinity-tagged fusion in Escherichia coli NM554 and purified using Ni 2ϩ affinity chromatography. The protein fragment was then reduced with dithiothreitol, purified by reverse-phase HPLC, and refolded in vitro, and the His 6 affinity tag was removed by cleavage with Factor Xa as has been described for production of similar cbEGF-containing protein fragments (44). The purity and identity of the final oxidized product were confirmed by reducing and non-reducing SDS-PAGE analysis and electrospray ionization mass spectrometry. The missense mutant N1317H was created from the corresponding wild-type plasmid described above. Protein expression, refolding, and purification were carried out as described above for the wildtype protein. Limited Proteolysis of the Wild-type EGF18-cbEGF19 Domain Pair-Proteolysis with elastase (1:5, w/w) was performed as described previously (45) in 50 mM Tris-HCl buffer at pH 8.0 with 150 mM NaCl. Reaction mixtures contained 10 mM CaCl 2 or 10 mM EGTA, and the mixtures were incubated at room temperature. Aliquots were removed at time intervals up to 4 h; the reaction was stopped by the addition of reducing protein-loading dye with boiling for 5 min, and samples were kept on ice until SDS-PAGE analysis and staining with Coomassie Blue. The identification of cleavage sites was performed by N-terminal sequencing of aliquots taken after 2.5 h of incubation with elastase and purified under non-reducing conditions by reverse phase HPLC.
NMR Spectroscopy-Samples for NMR spectroscopy contained 300 M protein in a 600-l volume of matrix solution (99.9% D 2 O containing 5 mM Tris-DCl and 150 mM NaCl at pH 7.0 or 7.5). All NMR experiments were performed at 35°C on a home-built 500-MHz spectrometer in the Department of Biochemistry NMR facility. Calcium titrations were performed by adding small aliquots (1-5 l) of CaCl 2 solutions in D 2 O. The pK a of His-1317 in the EGF18-cbEGF19 mutant fragment in the absence of Ca 2ϩ was determined by measuring the chemical shifts of the H ␦2 and H ⑀1 peaks arising from His-1317 in the aromatic region of one-dimensional spectra as a function of pH. One-dimensional NMR spectra were collected with a spectral width of 5494.51 Hz, 4096 complex points, and 512 scans. Data were zero-filled to 8192 points, resulting in a digital resolution of 0.67 Hz/point. Two-dimensional DQFCOSY were collected with a spectral width of 5494.51 Hz in F 2 and F 1 , with 1024 complex points in F 2 , with 512 complex t 1 increments, and with 48 scans per increment. Two-dimensional NOESY were collected with 1024 complex points in F 2 , with 256 complex t 1 increments, 96 scans per increment, and a mixing time of 150 ms. All spectra were processed using Felix 2.3 (Accelrys, Inc.). DQFCOSY and NOESY spectra were zero-filled to 2048 points in each dimension, yielding a digital resolution of 2.68 Hz/point. Assignment of peaks in one-dimensional and two-dimensional spectra was performed by analysis of the DQFCOSY and NOESY spectra in conjunction with sitedirected mutagenesis (F1289Y) and comparison of wild-type and N1317H-mutant spectra.
Determination of Calcium Dissociation Constants-Changes in chemical shifts of intra-and interdomain markers for Ca 2ϩ binding were measured in one-dimensional and two-dimensional spectra and fitted to the equation where ⌬ is the observed chemical shift change at each value of [Ca 2ϩ ] free and ⌬ o is the maximum observed chemical shift change (44). For the wild-type EGF18-cbEGF19 fragment, changes in the chemical shift of the H ⑀ resonance of the aromatic packing residue Phe-1289 from EGF18 were monitored in two-dimensional DQFCOSY spectra. The uncertainty in the calculated K d value for the single Ca 2ϩ -binding site in wild-type cbEGF19, arising from errors in chemical shifts, was determined using standard error propagation formulae and Monte Carlo simulations. For the mutant N1317H fragment, changes in the chemical shifts of the H ⑀1 and H ␦2 peaks arising from the substituted His-1317 residue were monitored by onedimensional spectra in addition to changes in the Phe-1289 H ⑀ peak from two-dimensional DQFCOSY spectra. The K d value was determined by averaging the values obtained for each of these three markers; the uncertainty in the K d is defined by the standard deviation from the mean.

RESULTS
Expression and Characterization of EGF18-cbEGF19 Wildtype and N1317H Missense Mutant Fragments-The EGF18-cbEGF19 wild-type and mutant constructs of CRB1 were expressed and purified using established methods (44). The addition of Ca 2ϩ was found to be essential for the in vitro refolding of the fragments to form the disulfide bond-stabilized fold. After removal of the His 6 affinity tag by cleavage with Factor Xa, a single peak was observed after HPLC purification. Masses determined experimentally using electrospray ionization mass spectrometry (9132.00 Ϯ 0.07 Da for the EGF18-cbEGF19 wild-type and 9154.46 Ϯ 0.08 Da for the N1317H mutant) were in good agreement with the predicted masses for the oxidized form of each protein (9132.12 and 9155.20 Da, respectively). NMR spectroscopy confirmed the presence of a stably folded single conformation for both wild-type and mutant proteins; peaks in the one-dimensional spectrum were well dispersed and at significantly different positions from those expected for an unstructured protein. In two-dimensional NOESY spectra, downfield shifted H ␣ -H ␣ cross-peaks confirmed the presence of antiparallel ␤-sheet, a structural feature common to all EGF domains (37).
cbEGF domains in fibrillin-1 demonstrate a protection against proteolytic cleavage in the presence of saturating Ca 2ϩ concentrations (37,41,45). The effect of Ca 2ϩ binding in the wild-type EGF18-cbEGF19 fragment was assessed by limited proteolysis with elastase; a time course of proteolysis shows cleavage both in the absence and in the presence of Ca 2ϩ , but significant protection is observed in the presence of Ca 2ϩ (Fig.  2c). Edman degradation of digested protein reveals four sites of elastase cleavage (Fig. 2a). The majority of these occur in the non-calcium-binding domain EGF18, with a single site observed in cbEGF19. It is interesting to note that differences in the degree of proteolysis between the Ca 2ϩ -free (10 mM EGTA) and Ca 2ϩ -saturated (10 mM Ca 2ϩ ) samples were observed for the sites in both EGF18 and cbEGF19 (Fig. 2a). This indicates that Ca 2ϩ binding in cbEGF19 leads to a more stable structure throughout the molecule.
NMR Reveals Formation of an Interdomain Interface Associated with Moderate Affinity Calcium Binding in Wild-type EGF18-cbEGF19-NMR spectroscopy has been used very successfully in previous studies of cbEGF domains of fibrillin-1 to probe changes in structure associated with calcium binding at the level of individual residues. It has also been used to determine calcium affinity by following changes in chemical shift of consensus aromatic residues (37,40,42,44). The aromatic region of the one-dimensional NMR spectrum of the EGF18-cbEGF19 domain pair in the absence and presence of Ca 2ϩ are compared in Fig. 3a. The observed changes can be assigned to specific residues using two-dimensional DQFCOSY experiments (Fig. 3, b and c). Cross-peaks arising from the single Tyr and Trp residues (Tyr-1269 and Trp-1293, Fig. 2a), both located in EGF18, can be assigned unambiguously; these peaks do not show significant changes in chemical shift upon the addition of Ca 2ϩ (Fig. 3, b and c). Only two of the four Phe residues give rise to resolved cross-peaks in the absence of Ca 2ϩ (Fig. 3b). Phe-1289, located in EGF18, is predicted by homology with other EGF/cbEGF domains to be involved in a Ca 2ϩ -dependent interdomain-packing interaction with Leu-1316 and Asn-1317 of cbEGF19 (Fig. 2b) (27,38,40). The observed upfield chemical shift change of 0.043 ppm for H ⑀ of Phe-1289 upon the addition of Ca 2ϩ confirms the formation of the interface in this domain pair from CRB1. The other Phe spin system observed in the absence of Ca 2ϩ is assigned to Phe-1277; this residue does not show a Ca 2ϩ -dependent chemical shift change consistent with its location in EGF18, away from the interdomain-packing region (Fig. 2a). Upon the addition of Ca 2ϩ , two additional pairs of cross-peaks arising from Phe-1319 and Phe-1327 of cbEGF19 appear in the spectrum (Fig. 3c). These peaks sharpen significantly and shift upfield as the Ca 2ϩ concentration is increased. Analysis of the behavior of peaks in other regions of the spectrum, arising from Ala, Thr, Leu, and Val residues, indicates a general sharpening of peaks arising from cbEGF19 as Ca 2ϩ is added. This is likely to result from a stabilization of the structure of the cbEGF19 domain when Ca 2ϩ is bound, consistent with the observed protection from proteolytic cleavage upon the addition of Ca 2ϩ .
Further evidence for the formation of the interdomain interface in EGF18-cbEGF19 is found in NOESY spectra collected in the presence of Ca 2ϩ (Fig. 3d). NOE cross-peaks are evident between the consensus packing aromatic from EGF18 (Phe-1289) and the side chain of Leu-1316 located in the ␤-turn of cbEGF19; these NOEs are not observed in the absence of Ca 2ϩ . Similar NOEs involving homologous residues have been observed in the cbEGF12-13 and cbEGF32-33 pair of fibrillin-1 and in the cbEGF11-13 domains of Notch (27,39,40).
Ca 2ϩ dissociation constants for cbEGF domains are often determined by monitoring the change in chemical shift of peaks arising from the consensus aromatic residue as a function of Ca 2ϩ (44). Due to line-broadening effects in the NMR . Phe-1289, from EGF18, and Asn-1317, from cbEGF19, which are predicted to interact in the interdomain interface, are shown in blue and magenta, respectively. c, comparative time courses of digestion of EGF18-cbEGF19 were performed using elastase in 10 mM EGTA (EG) and 10 mM CaCl 2 (CA) at pH 8. A reduced level of digestion in the presence of Ca 2ϩ is indicative of protection afforded by Ca 2ϩ binding to cbEGF19. The molecular masses (kDa) of prestained markers are indicated. The band at ϳ30 kDa arises from the added elastase. spectra in the absence of Ca 2ϩ and at low Ca 2ϩ concentrations, the typical intradomain consensus marker for Ca 2ϩ binding to a cbEGF domain (in this case Phe-1319 in cbEGF19) could not be monitored in the EGF18-cbEGF19 fragment. Instead, chemical shift changes of the interdo-main-packing aromatic residue from domain 18 (Phe-1289) were monitored (Fig. 4a); least squares fitting and error analysis (see "Experimental Procedures") yielded a dissociation constant for Ca 2ϩ of 220 Ϯ 40 M.
The N1317H Substitution Results in a Significant Reduction in Calcium Affinity in EGF18-cbEGF19-A similar analysis was performed for the EGF18-cbEGF19 N1317H mutant fragment. Ca 2ϩ -dependent changes are observed in one-dimensional and two-dimensional spectra, indicating that Ca 2ϩ binding is preserved in this mutant protein (Fig.  3, e-g). However, in the two-dimensional DQFCOSY spectra recorded at various concentrations of Ca 2ϩ , the cross-peak arising from Phe-1289 undergoes a significantly smaller overall chemical shift change in saturating Ca 2ϩ than in the corresponding wild-type spectra (0.043 and 0.014 ppm in the wild-type and mutant proteins, respectively). This may indicate a change in the specific nature of the interdomain-packing interaction in the mutant protein.
Since the native sequence of EGF18-cbEGF19 does not contain histidine residues, peaks arising from the substituted His-1317 could be unambiguously assigned in onedimensional NMR spectra (Fig. 3e). Both the H ␦2 and the H ⑀1 peaks of His-1317 were observed to shift upfield in a Ca 2ϩ -dependent manner. The shifts experienced by these protons were significantly larger than those observed for Phe-1289 (0.27 and 0.79 ppm for H ␦2 and H ⑀1 , respectively), and they thus provided additional markers of Ca 2ϩ binding. Least squares fitting of these markers yielded an average dissociation constant for Ca 2ϩ of 1.1 Ϯ 0.4 mM (Fig. 4b), indicating that Ca 2ϩ affinity is reduced ϳ5-fold in the N1317H variant.
In the NOESY spectrum recorded for the N1317H mutant, cross-peaks between Phe-1289 and Leu-1316 are still present, indicating the formation of an interdomain interface in the mutant protein under Ca 2ϩ -saturating conditions (Fig. 3h). However, the observed NOEs are weaker than those observed for the wild-type protein, indicating a FIGURE 3. Aromatic region of NMR spectra of wild-type and N1317H EGF18-cbEGF19. a, one-dimensional spectra collected in the absence (top) and presence (bottom) of 13 mM Ca 2ϩ (both at pH 7.5 in 5 mM Tris-DCl and 150 mM NaCl) are shown for wild-type EGF18-cbEGF19. b and c, the observed changes in the spectra, which are indicative of Ca 2ϩ binding to cbEGF19, are highlighted more clearly in DQFCOSY spectra collected in the absence (b) and the presence (c) of 13 mM Ca 2ϩ . d, the interdomain interaction in wild-type EGF18-cbEGF19 in 13 mM Ca 2ϩ is identified by the observed NOEs between the side chains of Phe-1289 (EGF18) and Leu-1316 (cbEGF19). e, one-dimensional spectra collected in the absence (top) and presence (bottom) of 25 mM Ca 2ϩ (both at pH 7.5 in 5 mM Tris-DCl and 150 mM NaCl) are shown for N1317H EGF18-cbEGF19. The resonances belonging to His-1317 are labeled (*). f and g, the observed changes in the spectra, which are indicative of Ca 2ϩ binding to cbEGF19 in the mutant protein, are highlighted more clearly in DQFCOSY spectra collected in the absence (f) and the presence (g) of 25 mM Ca 2ϩ . h, the interdomain interaction in N1317H EGF18-cbEGF19 in 25mM Ca 2ϩ , is identified by the observed NOEs between the side chains of Phe-1289 (EGF18) and Leu-1316 (cbEGF19). The comparative weakness of the cross-peaks in h suggests that the interface is not formed as rigidly as in the wild-type construct.
weakening of the interdomain interaction, consistent with the lower Ca 2ϩ affinity.
Calcium Affinity of the N1317H Variant Can Be Decreased Further by a Decrease in pH in the Physiological Range-The imidazole side chain of histidine is ionizable with a nominal pK a of ϳ6.5. This value can vary in folded proteins due to the effects of the local chemical environment. Because of the proximity of the histidine pK a to the pH range under which the Ca 2ϩ binding properties of the mutant fragment were determined, the pH effects on the dissociation constant for Ca 2ϩ in this fragment were assessed. Under Ca 2ϩ -free conditions, both the H ␦2 and the H ⑀1 of His-1317 experience large pH-dependent changes in chemical shift; least squares fitting of the chemical shifts of these peaks yielded an average pK a of 7.1 for His-1317. The Ca 2ϩ titrations described above were conducted at a constant pH of 7.5. Because the pK a of His-1317 is close to this value and within the physiological pH range, the Ca 2ϩ titration was repeated at pH 7.0. The dissociation constant in this instance was determined to be 1.6 Ϯ 0.4 mM, a value higher (indicating weaker affinity) than that determined at pH 7.5 and ϳ8-fold higher than the dissociation constant measured for the corresponding wild-type EGF18-cbEGF19 fragment measured at pH 7.0 (190 Ϯ 40 M). This result is consistent with a weakening of the interdomain interface when His-1317 is protonated, leading to a lowering of the Ca 2ϩ affinity between pH 7.5 and 7.0.

DISCUSSION
The study of the calcium binding properties of the EGF18-cbEGF19 domain pair from CRB1 is facilitated by the presence of a single calcium-binding site. The EGF18 domain preceding cbEGF19 represents a native context in which the Ca 2ϩ binding in cbEGF19 can be examined. The Ca 2ϩ dissociation constant determined at pH 7.5 (220 Ϯ 40 M) represents a moderate affinity for Ca 2ϩ when compared with values determined for other cbEGF-containing proteins, which can be of nanomolar affinity (42,46).
Immunohistochemical localization of CRB1 reveals its location in the mammalian retina to be adjacent to the outer limiting membrane, bordering on the subretinal space (9). In response to stimulation by light from a dark-adapted state, a transient, localized surge in extracellular Ca 2ϩ has been observed; however, some uncertainty remains concerning the precise initial and final concentrations since the extracellular calcium levels ([Ca 2ϩ ] o ) can depend on pigmentation, the animal model used, and the depth of penetration of the subretinal space by the ion-selective electrode (47)(48)(49) in addition to modulation of the subretinal space volume by other factors (50). For the present analysis, a resting value of [Ca 2ϩ ] o is taken as ϳ1.5 mM, in the middle of the typical tissue [Ca 2ϩ ] o ranging from 1 to 2 mM. The pH is taken to lie between 7.0 and 7.2 in the dark-adapted state, as estimated from values recorded in retinal layers encompassing the region to which CRB1 is localized (51).
Studies on fragments from fibrillin-1 and Notch-1 have suggested that Ca 2ϩ binding plays a role in the dynamics of the proteins, most notably with respect to the rigidity of the interdomain interface (27,42). In the calcium-saturated state, a rigid interdomain interface is detected for cbEGF pairs, whereas significant interdomain dynamics are observed in the absence of calcium. Introduction of the substitution N1317H to cbEGF19 results in a 5-fold decrease in the affinity for Ca 2ϩ , at pH 7.5. Under the conditions of resting [Ca 2ϩ ] o estimated above, this decrease in Ca 2ϩ affinity would be manifested as a decrease in the percentage of Ca 2ϩ -bound protein containing a rigid interdomain interface from ϳ90% in the wild-type protein to the significantly lower value of only ϳ60% in the mutant.
In the solution structures of the cbEGF12-13 and cbEGF32-33 domains of fibrillin-1 and the cbEGF11-13 fragment from Notch-1, the interdomain interaction is mediated through hydrophobic effects (27,39,40). Replacement of a residue with a polar side chain in the native sequence (Asn-1317) with a bulkier polar residue (His-1317) would be expected to have a deleterious effect on the formation of the interdomain interface, as characterized by a higher dissociation constant for calcium, particularly if the histidine imidazole side chain is positively charged. The local chemical environment of His-1317 in EGF18-cbEGF19 results in a side chain pK a of 7.1, a value within the physiological pH range and within the estimated range of pH for this region of the retina (51). Decreasing the pH of the solution from 7.5 to 7.0 would result in a large increase in the proportion of His-1317 side chains in the charged proto- and H ⑀1 (square) peaks shift with the addition of Ca 2ϩ and were monitored to determine the Ca 2ϩ K d . In addition to the His markers in the one-dimensional spectra, the H ⑀ peak from Phe-1289 in twodimensional DQFCOSY spectra (triangles), used to monitor binding in the wild-type protein, was monitored. Least squares fitting of the data yielded an average K d value of 1.1 Ϯ 0.4 mM. nated state; this would lead to a further weakening of the interdomain interface reflected by a decrease in the Ca 2ϩ affinity. Our observation of an increase in the K d from 1.1 Ϯ 0.4 mM at pH 7.5 to 1.6 Ϯ 0.4 mM at pH 7.0 supports this hypothesis. This decrease in pH would result in a further decrease in the percentage of Ca 2ϩ -bound mutant protein from ϳ60 to ϳ50%. The pH of the subretinal space is expected to increase by up to 0.2 pH units upon light stimulation of the dark-adapted state (51). For the N1317H mutant protein, this would lead to an increase in Ca 2ϩ affinity upon illumination, whereas for the wild-type protein, which does not contain an ionizable group at the interdomain interface, the percentage of Ca 2ϩ -bound protein would not be sensitive to such a light-induced change in pH.
The decrease in the occupancy of the cbEGF19 calciumbinding site in the mutant N1317H fragment is accompanied by a change in the nature of the interface formed between the domains. Interdomain NOEs between Phe-1289 and Leu-1316 indicate that an interface mediated through the same residues as the wild-type fragment is formed in the mutant protein.
However, its character is changed as indicated by the significantly smaller chemical shift change observed for the crosspeak arising from Phe-1289 in domain 18 and by the lower intensity of the interdomain NOEs, even at saturating Ca 2ϩ concentrations. These observations are consistent with a less extensive interface in the mutant protein.
In other proteins containing tandemly linked cbEGF domains, missense mutations associated with disease have also been localized to regions that, by homology, could affect the interdomain interface, either by substitutions that directly replace residues involved in the formation of the interface or by changes in adjacent residues, which could affect the orientation of the residues that are involved in the interface (Fig. 5). Substitutions that affect residues immediately adjacent to the conserved aromatic residue in the N-terminal side of the interface and that affect the two residues in the turn of the major ␤-hairpin in the C-terminal side of the interface occur. Many of these substitutions involve the introduction of a bulkier charged amino acid in place of a glycine. In previous studies, it has been postulated that alteration or disruption of the interdomain-packing interaction could result from the G2627R substitution in fibrillin-1 (40) and from the G352D substitution in low density lipoprotein receptor (38); these substitutions are associated with Marfan syndrome and familial hypercholesterolemia, respectively. In the present study, we have demonstrated experimentally, for the first time, that a disease-associated mutation of a residue in this region does lead to perturbation of the interdomain interface.
Precisely how the alteration in the interdomain interface structure and flexibility impairs function in N1317H EGF18-cbEGF19 is unknown since binding partners of CRB1 that interact with its extracellular portion have yet to be identified. One possibility is that the interface between the domains comprises a site of protein binding, and the inability of the mutant fragment to achieve the rigid, extended conformation of the wild-type protein, even under conditions of saturating Ca 2ϩ , could abrogate binding. An extension of this hypothesis could be suggested if the extracellular region of the protein is involved in a similar heterologous protein scaffold as has been determined for the cytoplasmic tail. In this case, separate binding partners may be misaligned as a result of the non-rigid interface. Finally, longer range intramolecular effects could be involved if, for example, a ligand binding event closer to the N terminus of the protein had to be communicated via this interface. Which of these mechanisms may prove to be critical in the development of the deleterious effect on CRB1 function remains to be determined. However, these data suggest that in proteins containing cbEGF domains tandemly linked to other modules, Ca 2ϩ binding together with the proper formation of the interdomain interface is important for normal function and that disruption of this interface may be associated with disease. FIGURE 5. Disease-causing amino acid substitutions predicted to disrupt the conserved interdomain interface between contiguous cbEGF/cbEGF or EGF/cbEGF domains pairs. Amino acid substitutions adjacent to the conserved aromatic residue between cysteines 5 and 6 would affect the N-terminal face of the interdomain interface; in EGF18, the conserved aromatic residue is Phe-1289. Amino acid substitutions involving the two residues in the turn in the major ␤-hairpin between cysteines 3 and 4 would affect the C-terminal face of the interdomain interface; in cbEGF19, Leu-1316 and Asn-1317 are located in these positions. Among the substitutions illustrated, mutations in CRB1 give rise to LCA, those in fibrillin-1 (FBN1) result in Marfan syndrome, and those in low density lipoprotein receptor (LDLR) give rise to familial hypercholesterolemia. Numbering and missense substitutions are according to the Human Gene Mutation data base. The consensus residues of the cbEGF domain are highlighted in gray.