Identification of Amino Acid Residues That Determine pH Dependence of Ligand Binding to the Asialoglycoprotein Receptor during Endocytosis*

The rat hepatic asialoglycoprotein receptor mediates clearance of galactose- andN-acetylgalactosamine-terminated glycoproteins by endocytosis, binding ligands through a C-type, Ca2+-dependent carbohydrate-recognition domain (CRD) at extracellular pH and releasing them at lower pH in endosomes. At physiological Ca2+ concentrations, the midpoint for ligand release from the CRD of the major subunit of the receptor is pH 7.1. In contrast, the midpoint is pH 5.0 for a galactose-binding derivative of the homologous C-type CRD of serum mannose-binding protein, which would thus not efficiently release ligand at an endosomal pH of 5.4. Site-directed mutagenesis of the CRD from the major subunit of the asialoglycoprotein receptor has been used to identify residues that are essential for efficient release of ligand at endosomal pH. The effects of changes to residues His256, Asp266, and Arg270 singly and in combination indicate that these residues reduce the affinity of the CRD for Ca2+, so that ligands are released at physiological Ca2+ concentrations. The proximity of these three residues to the ligand-binding site at Ca2+ site 2 of the domain suggests that they form a pH-sensitive switch for Ca2+ and ligand binding. Introduction of histidine and aspartic acid residues into the mannose-binding protein CRD at positions equivalent to His256 and Asp266 raises the pH for half-maximal binding of ligand to 6.1. The results, as well as sequence comparisons with other C-type CRDs, confirm the importance of these residues in conferring appropriate pH dependence in this family of domains.

The asialoglycoprotein receptor, which mediates clearance of galactose or N-acetylgalactosamine-terminated glycoproteins from circulation, is a prototype for receptor-mediated endocytosis. Related but distinct forms of the receptor are found in mammalian liver and in peritoneal macrophages (1,2). The rat hepatic receptor is composed of three subunits, of which the major subunit, rat hepatic lectin 1 (RHL-1), 1 represents 70 -80% of the total mass of the receptor. The second subunit, RHL-2/3, consists of two species that are differentially glyco-sylated forms of a second, homologous polypeptide (3). RHL-1 and RHL-2/3 bind carbohydrate ligands through COOH-terminal, Ca 2ϩ -dependent carbohydrate-recognition domains (Ctype CRDs).
Structural analysis of wild-type and mutant forms of the homologous C-type CRD from serum mannose-binding protein suggests that carbohydrate ligands bind to the CRDs of RHL-1 at a conserved Ca 2ϩ designated site 2. The primary interaction with sugar is through hydroxyl groups that form coordination bonds with the Ca 2ϩ and hydrogen bonds with protein side chains that also ligate the Ca 2ϩ (4). The molecular basis for galactose binding to C-type CRDs has been particularly well investigated in a mutant CRD from serum mannose-binding protein which has been engineered to bind galactose. Three single amino acid replacements (E185Q, N187D, and H189W) and insertion of a glycine-rich loop result in a modified CRD designated QPDWG, which has affinity and selectivity similar to RHL-1 (5). The crystal structures of this modified CRD and of a further mutant that binds N-acetylgalactosamine with high affinity have been solved (6,7). These structures provide a basis for modeling the CRD of RHL-1.
Following binding of ligand to the asialoglycoprotein receptor at the cell surface, the receptor-ligand complex is endocytosed via clathrin-coated pits and directed to endosomes, where the complex dissociates. The carbohydrate ligand is targeted for degradation in lysosomes while the receptor recycles to the cell surface with a vacant binding site (1). Because the RHL-1 subunit alone can perform receptor-mediated endocytosis (8), it provides a useful tool for analyzing the molecular basis for the endocytosis process. Endosomal pH is an important determinant of recycling of the asialoglycoprotein receptor as well as other endocytic receptors (9). The dissociation process that occurs in endosomes can be mimicked in vitro by ligand release at pH 5.4. In the case of the chicken hepatic lectin, an homologous endocytic receptor, pH modulates structural transitions between several distinct states of the CRD (10). At endosomal pH, the structural change causes an approximately 10-fold reduction in affinity for Ca 2ϩ with concomitant loss of ligand binding activity. Although these studies provide insight into the effect of pH on ligand binding activity and receptor conformation, they do not address the identity of the amino acids that sense the environmental pH.
The results of site-directed mutagenesis reported here identify three residues, including a histidine conserved in other endocytic receptors, that determine the pH dependence of ligand binding to RHL-1. Loss of ligand binding activity at low pH is closely correlated with decreased affinity for Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes were purchased from New England BioLabs. Gal 34 -BSA was obtained from E-Y Laboratories and was iodinated by the chloramine-T method (11). Na 125 I was purchased from Amersham Pharmacia Biotech. Galactose-Sepharose was prepared by the divinyl sulfone method (12). Immulon-4 polystyrene wells were from Dynex Technologies. DNA sequencing was performed using the Sequenase II kit from Amersham Pharmacia Biotech.
Mutagenesis-Mutagenesis was performed by insertion of doublestranded synthetic oligonucleotides (GenoSys and Applied Biosystems, United Kingdom) at appropriate restriction sites of a cDNA for RHL-1 as described previously (13). For expression, mutant constructs of RHL-1 were transferred to the pT5T vector (14) and transformed into Escherichia coli strain C41(DE3), a modified version of strain BL21(DE3) (15). Procedures for the mutagenesis of a cDNA for rat mannose-binding protein and construction of modified expression vectors have been described (16).
Protein Production-Mutant versions of QPDWG were produced and isolated as described previously (17). Mutant versions of RHL-1 were prepared as described in Ref. 13, except that RHL-1 CRD production was induced in the transformed C41(DE3) strain with a final concentration of 0.7 mM isopropyl-␤-D-thiogalactoside when the cells had grown to OD 600 of 0.6. For fluorescence measurements, the mutant versions of QPDWG were further dialyzed to ensure the absence of Ca 2ϩ . The proteins were dialyzed at 4°C against three changes of 500 ml of 50 mM Tris-Cl, pH 7.8, 100 mM NaCl made with distilled water from which metal ions had been removed by passage over a Chelex-100 column.
Solid Phase Binding Assays-Each assay was performed in duplicate at least twice. Ca 2ϩ -free BSA and incubation buffer were prepared by filtration through Chelex ion-exchange syringe filters (Bio-Rad). For the pH dependence assay, Immulon-4 wells were coated with protein and incubated at 4°C overnight. After the protein solution was removed, the wells were rinsed twice with cold loading buffer (1.25 M NaCl, 25 mM Tris-Cl, pH 7.8, 25 CaCl 2 ) and blocked with 5% BSA in loading buffer for 2 h at 4°C. The wells were rinsed twice with cold 1.25 mM NaCl, incubated for 10 min at 4°C in the same solution and then rinsed a third time. Aliquots (50 l) of 125 I-Gal 34 -BSA (1 g/ml) in 2 ϫ incubation buffer (10 mM NaCl, 2 mM CaCl 2 , 0.5% BSA) were mixed with equal volumes of 2 ϫ pH buffer (50 mM sodium acetate in the pH 4.5-6.5 range, 50 mM MES in the pH 6 -7.5 range, and 50 mM Tris-Cl in the pH 6 -8 range, with all solutions containing 2.5 M NaCl and 4% BSA) and transferred to the Immulon-4 wells. Incubation was for 2 h at 4°C, which was sufficient to achieve equilibrium binding. The wells were emptied, rinsed three times with cold high-Ca 2ϩ loading buffer (1.25 M NaCl, 25 mM Tris-Cl, pH 7.8, 250 mM CaCl 2 ), dried, and counted in a Wallac Wizard 1470 ␥-counter.
For the determination of Ca 2ϩ dependence of neoglycoprotein ligand binding, Immulon-4 wells were coated with protein and incubated at 4°C overnight. One set of wells was prepared as described above and rinsed twice with cold high pH Ca 2ϩ -free rinse buffer (1.25 M NaCl, 25 mM Tris-Cl, pH 7.8) and a duplicate set of wells was rinsed with low pH Ca 2ϩ -free rinse buffer (1.25 M NaCl, 25 mM sodium acetate, pH 5.4). All wells were incubated for 10 min at 4°C and then rinsed a third time in the appropriate Ca 2ϩ -free rinse buffer. Aliquots (50 l) of 125 I-Gal 34 -BSA (1 g/ml) in high pH, Ca 2ϩ -free incubation buffer (1.25 M NaCl, 25 mM Tris-Cl, pH 7.8, 5% BSA) or in low pH, Ca 2ϩ -free incubation buffer (1.25 M NaCl, 25 mM sodium acetate, pH 5.4, 5% BSA) were added to equal volumes of serial 3:2 dilutions of CaCl 2 and then transferred to the appropriate Immulon-4 wells. Incubation was for 2 h at 4°C. The wells were emptied, rinsed three times with cold high-Ca 2ϩ loading buffer, dried, and counted as above.
Fluorescence Measurements-Measurements were performed on a Jasco FP777 fluorescence spectrophotometer with a 170 W lamp using a 1-cm square quartz cuvette. Emission spectra were recorded for 30 s using excitation at 295 nm (10 nm slit width) and measuring emission at 340 nm (5 nm slit width). For the determination of Ca 2ϩ affinity, aliquots of CaCl 2 were added to tubes containing 1.5 ml of a 1 M protein solution in 50 mM sodium acetate, pH 5.5-6.2, 50 mM MES, pH 6.2-7.5, or 50 mM Tris-HCl, pH 6.7-7.8, with all solutions containing 100 mM NaCl. All buffers were made with distilled water passed over a Chelex-100 column. The tubes were allowed to equilibrate at 37°C and the temperature of the measurements was maintained at 37°C. Background intensity in samples without protein was subtracted from the measurements.
Data Analysis-Using a nonlinear least-squares fitting program (SigmaPlot, Jandel Scientific), data from the pH dependence assay were fitted to Equation 1, where pH B is the pH value at which half-maximal binding is observed. Data from the Ca 2ϩ affinity binding assays and the fluorescence exper-iments were fitted to Equation 2, Fraction maximal binding (or fluorescence) ϭ ͓Ca 2ϩ ͔ 2 /͑K Ca where K Ca is the concentration of Ca 2ϩ at which half-maximal binding is observed. Values reported are the mean Ϯ S.D. for at least two assays. Binding experiments were each performed in duplicate.
Molecular Modeling-The coordinates of QPDWG modified to contain histidine at residue 202 (7) were used for modeling RHL-1. Residues from the RHL-1 sequence were inserted at corresponding positions using the standard rotomer library in Insight II (BioSym). No attempt was made to minimize the structure.

RESULTS
pH-dependent Binding of Ligands to C-type CRDs-A solidphase binding assay was used to analyze the pH dependence of ligand binding to the isolated CRD of RHL-1 ( Fig. 1). At a Ca 2ϩ concentration near to that found in extracellular fluids (1 mM), ligand binding to the isolated CRD decreases as the pH goes from extracellular, pH 7.3, to endosomal, pH 5.4. The pattern of binding is similar to that of the intact receptor (18), indicating that residues which confer pH sensitivity are located within the CRD portion of RHL-1. In contrast, QPDWG, the galactosebinding mutant CRD derived from serum mannose-binding protein, shows substantially less change in ligand binding over the pH range 7.3 to 5.4, although ligand binding activity is lost at lower pH values. Similar behavior is observed for unmodified serum mannose-binding protein, which is not directly involved in receptor-mediated endocytosis and therefore does not require the ability to release ligand at endosomal pH (data not shown). Points of half-maximal ligand binding, designated pH B , are 7.1 Ϯ 0.1 for the CRD from RHL-1 and 5.0 Ϯ 0.3 for QPDWG (averages for three different experiments).
The CRDs of QPDWG and RHL-1 are 31% identical in overall amino acid sequence and are based on the same protein fold (5). The difference in the behavior of these two CRDs suggests that pH sensitivity of ligand binding in the range between pH 5.4 and 7.3 is conferred by specific amino acid residues found in RHL-1 but not in QPDWG. Side chains that might be involved in establishing the appropriate pH sensitivity for cycles of ligand binding and release during endocytosis were highlighted by comparing the amino acid sequence of RHL-1 to the sequences of QPDWG and two other C-type lectins that have pH-dependent ligand-binding profiles similar to RHL-1: the chicken hepatic lectin (10) and the macrophage asialoglycoprotein receptor (pH B of 6.3). 2 On the assumption that pH sensitivity of ligand binding results from the interaction of protons with amino acid side chains, candidate residues for controlling pH sensitivity in RHL-1 were identified as residues conserved in the macrophage asialoglycoprotein receptor but not found in QPDWG. Comparison of the sequences shown in Fig. 2 reveals nine potentially protonatable residues that meet this criterion.
Mutations in RHL-1 That Alter pH Sensitivity-As an initial test of the role of specific residues in defining pH dependence of ligand binding to RHL-1, each of the nine residues identified in the preceding section was changed to a non-dissociating amino acid of a similar size. Changing Arg 273 to a leucine residue results in a CRD that can no longer be purified by affinity chromatography on galactose-Sepharose, suggesting that this mutation causes misfolding of the domain. The remaining eight mutants containing single site changes could be purified on galactose-Sepharose, indicating that the basic fold of the CRD is unaffected by these mutations. In the solid-phase binding assay, five of the mutants, with changes at positions His 203 , Asp 224 , Asp 227 , Glu 229 , and Asp 260 , have profiles indistinguishable from wild-type (Table I), suggesting that side chains of these residues are unlikely to play a role in establishing pH sensitivity of ligand binding.
Of the three single site mutations that do alter pH dependence of ligand binding, two cause a decrease in the pH B value. Representative results for mutants H256Q and D266N are shown in Fig. 3A. For both of these mutants, the pH that supports half-maximal ligand binding is approximately 1 unit lower than for wild-type CRD. The pH profile and pH B value for a double mutant containing both of these changes are very similar to those of the single mutants, indicating that His 256 and Asp 266 do not function in an additive fashion. The ability of the singly and doubly mutant CRDs to maintain ligand binding at endosomal pH also suggests that the role of His 256 and Asp 266 is to destabilize the CRD so that physiological release of ligand can take place.
A model for RHL-1 was generated based on the crystal structure of the CRD from a previously described variant of QPDWG that contains the histidine corresponding to His 256 in RHL-1. Insertion of the additional RHL-1 side chains that have been mutated reveals that the five residues which appear not to be involved in establishing pH dependence of ligand binding are at least 16 Å distant from the sugar-binding site adjacent to Ca 2ϩ site 2. In contrast, His 256 and Asp 266 are predicted to lie within 7 Å of this Ca 2ϩ (Fig. 4). The fact that the His 256 and Asp 266 side chains are close to each other is consistent with the idea that they work together to affect pH dependence and thus that mutations at these positions produce similar and non-additive effects. The proximity of these side chains to the sugar-binding site suggests that partial protonation of one of these residues might have a direct, local effect on the binding site, causing loss of ligand binding activity.
The remaining single mutant that displays a change in the pH dependence profile is R270L, for which the pH B is shifted to a higher value. This change is accompanied by a substantial loss in affinity for Ca 2ϩ , so pH dependence was measured in the presence of 5 mM Ca 2ϩ (Fig. 3B). Mutation of the adjacent residue Arg 269 to leucine results in essentially no change in pH dependence of ligand binding, suggesting that the effect of the Arg 270 mutation results from a specific effect of this arginine side chain. Combination of the H256Q and D266N mutations with the R270L change results in partial compensation between the shifts seen with the separate mutations ( Table I). The loss in absolute affinity for Ca 2ϩ makes it difficult to interpret the effects of the change at position 270. However, as shown in Fig. 4, the arginine side chain is likely to be positioned close to His 256 and Asp 266 and within 8 Å of Ca 2ϩ site 2. It thus forms part of a cluster of amino acid side chains that affect pH sensitivity.
Modulation of Ligand Binding by Ca 2ϩ -It has previously been demonstrated that pH-dependent binding of saccharide ligands to the chicken hepatic lectin is correlated with pH-dependent changes in affinity of the C-type CRD for Ca 2ϩ (10). The ability to generate mutant domains with altered pH dependence makes it possible to examine the relationship between pH and Ca 2ϩ binding in more detail. As a starting point, the Ca 2ϩ dependence of galactose-BSA binding to the wild-type CRD of RHL-1 and to QPDWG at high and low pH values are compared in Fig. 5. In all cases, binding is dependent on the second power of the concentration of Ca 2ϩ , reflecting the presence of two Ca 2ϩ sites in the CRD (10). The Ca 2ϩ dependence can be characterized by an apparent K Ca , which represent the Ca 2ϩ concentration that supports half-maximal ligand binding. Like the chicken hepatic lectin, RHL-1 shows substantially reduced affinity for Ca 2ϩ at endosomal pH. In contrast, QP-DWG is relatively insensitive to a reduction in pH from 7.8 to 5.4.
The relationship between K Ca and pH over this range is demonstrated in greater detail for RHL-1 in Fig. 6. The value of log K Ca shows an approximately linear dependence on pH. The point at which this line crosses the physiological Ca 2ϩ concentration range represents the pH at which half-maximal binding of ligand will be achieved and thus corresponds to the pH B value. The behavior of the H256Q/D266N mutant of RHL-1 contrasts with the wild-type in that K Ca values at all pH values are lower. This shift has the effect of reducing the pH for half-maximal binding at physiological pH. For the receptor to function during endocytosis, the position of the line in Fig. 6 must be such that it passes through the physiological Ca 2ϩ concentration range between the pH value at the cell surface and the value in endosomes. As noted above, the effect of His 256 and Asp 266 is to decrease affinity for Ca 2ϩ over the entire pH range.
While it was not practical to measure K Ca values at all pH values for all of the mutant CRDs, the approximately linear relationship between pH and log K Ca makes it possible to gain an appreciation for the effects of the mutations by measuring K Ca at the two extreme pH values of 5.4 and 7.8. A compilation of such values for various mutant CRDs from RHL-1 is pre-2 H. Dobbyn and S. Wragg, unpublished observations.

FIG. 2. Aligned amino acid sequences from CRDs of RHL-1 and other C-type lectins.
Regions of secondary structure in QPDWG are indicated above the sequence. L denotes loop, S denotes ␤-strand, and H denotes ␣-helix. Residues coordinating the two Ca 2ϩ are denoted 1 and 2 above the sequence. The numbering corresponds to the RHL-1 sequence. Amino acids with potentially titratable side chains that are present in RHL-1 and the macrophage asialoglycoprotein receptor (RHL-M) but not in QPDWG are shaded. Positions at which amino acids from RHL-1 were introduced into QPDWG are boxed. Residues in QPDWG that have been replaced are 157-162 (EVTEGQ), 202 (Thr), 212 (Ile), and 216 -218 (ASH). Within this region, the macrophage asialoglycoprotein receptor shows 78% sequence identity with RHL-1, while the chicken hepatic lectin (CHL) is more distantly related (42% sequence identity).
sented in Table I. The results confirm that decreases in pH B are accompanied by an overall increase in apparent affinity for Ca 2ϩ throughout the pH range and that increases in pH B are correlated with weaker binding of Ca 2ϩ . Thus, the effect of the mutations that decrease the pH B is to increase Ca 2ϩ affinity so that the domain continues to bind ligand at physiological Ca 2ϩ concentrations.
Mutations in QPDWG Designed to Alter pH Sensitivity-The roles of residues His 256 , Asp 260 , and Asp 266 in RHL were further investigated by incorporating these amino acids at equivalent positions into the relatively pH insensitive CRD of QP-DWG. The results, which are summarized in Table II, confirm the ability of His 256 to raise pH B . Incorporation of this residue into QPDWG increases the pH B by more than half of a pH unit. In contrast, incorporation of residues equivalent to Asp 260 or Asp 266 alone has no significant effect on pH B , although there is a slight increase in pH B when these residues are incorporated in the presence of the histidine residue. Although these effects are largely consistent with the results for RHL-1, the highest pH B obtained, in the presence of both His 256 and Asp 266 , is still roughly one pH unit lower than for wild-type RHL. The effect of pH on K Ca was also investigated for the QPDWG mutants, since changes on Ca 2ϩ affinity provide a means of determining how accurately these mutants mimic the behavior  The model was constructed based on the x-ray crystal structure of QPDWG mutant T202H complexed with Nacetylgalactosamine (Protein Data Bank code 1BCH). Carbon atoms are shaded black, oxygen white, and nitrogen gray. Large black sphere represents Ca 2ϩ . Four side chains, including Asn 264 , that form axial coordination bonds to Ca 2ϩ 2 (dotted lined) are shown. These four residues also form hydrogen bonds with the bound sugar ligand, which would be located above the plane of the page, directly above the Ca 2ϩ . Side chains of residues that alter the pH dependence of the CRD are also shown. The distance between Asn 264 and His 256 , indicated by the dashed line, is 2.97 Å. This figure was prepared with Molscript (19). of a CRD involved in endocytosis. As shown in Table II, the K Ca of QPDWG determined from the Ca 2ϩ dependence of neoglycoprotein binding is relatively insensitive to pH, as it changes by a factor of roughly 2.5 over the range of 5.4 to 7.8. These results suggest that shifts in the pH B values either up or down are not related to large changes in the relative K Ca values at pH 5.4 and 7.8, but represent an overall loss of affinity for Ca 2ϩ at all pH values.
This interpretation was verified in direct measurements of Ca 2ϩ binding, which can be monitored in mannose-binding protein and derivatives by following changes in intrinsic fluorescence (20). Results for QPDWG and the mutant with the highest measured pH B are shown in Fig. 7. As noted in previous experiments with wild-type mannose-binding proteins, the absolute K Ca values measured by this technique differ from those determined indirectly in the neoglycoprotein-binding assay, probably because of the effect of the bound sugar. Nevertheless, this approach provides a convenient way to compare affinity for Ca 2ϩ over a range of pH values. The results shown in Fig. 7 confirm that the mutations introduced into QPDWG result in an overall decrease in affinity for Ca 2ϩ at all pH values without substantially changing the relative affinity at low and high pH, which would correspond to increased slope in a plot of K Ca as a function of pH. Thus, even the mutant containing residues corresponding to His 256 and Asp 266 , which has the highest pH B value at 1 mM Ca 2ϩ , differs substantially from RHL-1 in its Ca 2ϩ -binding properties.
In an attempt to make the behavior of QPDWG more like RHL, similar experiments were undertaken investigating the contribution of residues Arg 270 and Arg 273 . The presence of a proline residue adjacent to Arg 270 in RHL-1 suggests that the local conformation of the polypeptide may differ substantially from QPDWG. For this reason, all three amino acid residues predicted to be in this loop (Arg-Pro-Tyr) were incorporated into QPDWG at positions 216 -218. As indicated in Table II, this change by itself or in the presence of His 256 alters the pH B only slightly when measured at 1 mM Ca 2ϩ . However, the slight compensatory relationship between the His 256 and Arg 270 positions is amplified at 4 mM Ca 2ϩ . These results are again consistent with the findings for RHL, in which removal of arginine at position 270 causes an increase in pH B , since in QPDWG incorporation of the arginine residue decreases the pH B . In contrast, in the presence of both Asp 266 and His 256 , this arginine actually increases the pH B . However, this triple mutant shows substantially reduced affinity for Ca 2ϩ , again suggesting that changes in this region may result in significant rearrangement of the Ca 2ϩ -binding sites. Similarly, mutants of QPDWG containing arginine at the position corresponding to Arg 273 of RHL-1 display greatly decreased affinity for Ca 2ϩ , which could not be quantified using the solid-phase binding assay.
A further important difference between RHL and QPDWG is the arrangement of ligands for the accessory Ca 2ϩ site (site 1). As noted in Fig. 2, at least one of the site 1 ligands, at position 217 in RHL, must be different from QPDWG and two adjacent residues are absent from RHL. The effect of inserting the shorter loop corresponding to residues 216 -219 of RHL, replacing residues 157-168 of QPDWG, was tested by the creation of two additional mutations, resulting in a dramatic loss in affinity for Ca 2ϩ and a concomitant increase in the apparent pH B (Table II). These results suggest that the arrangement of Ca 2ϩ site 1 ligands may contribute to establishing the appropriate affinity for Ca 2ϩ and the pH B of RHL. However, the data also indicate that mutants containing the modified site 1 do not attain all of the Ca 2ϩ -binding characteristics of RHL. DISCUSSION The results reported here provide strong evidence linking pH sensitivity of ligand binding to C-type CRDs with pH-dependent changes in Ca 2ϩ affinity. The presence of a cluster of titratable amino acid side chains near the ligand-binding site at Ca 2ϩ site 2 in RHL-1 suggests that control of binding by pH is probably mediated at least in part by this set of side chains acting as a pH sensor. The pH sensitivity of the domain is ultimately a result of the summed effect of multiple titratable groups in the domain. However, many potentially titratable groups on the surface of the protein can be mutated without affecting the pH dependence of ligand binding to the CRD from RHL-1, indicating that loss of binding activity in the physiological pH range is a result of a local event involving relatively few amino acid side chains.
One scenario consistent with the evidence is that the protonation state of the histidine side chain at position 256 determines the affinity for Ca 2ϩ binding at site 2, perhaps because it makes direct hydrogen bond interactions with certain of the Ca 2ϩ site 2 ligands. The exact interactions of this histidine residue in the QPDWG model background are known to vary depending on the presence of other side chains in this region (7), so it is not yet possible to propose a specific mechanism for linking the pH sensor to Ca 2ϩ site 2. However, it is possible that a hydrogen bond forms between His 256 and one of the site 2 Ca 2ϩ ligands, corresponding to Asn 264 (Fig. 4), so that protonation of the imidazole group would directly affect the position of the amide group and thus alter affinity for Ca 2ϩ . In any case, the divergent effects of changing residues Asp 266 and Arg 270 could result from the influence of these side chains on the pK a of His 256 . The presence of an aspartate side chain would reduce the pK a of the imidazole group while the arginine side chain would increase the pK a .
Even in the absence of His 256 , ligand binding activity of the CRD from RHL-1 is lost at low pH (below 5.4). This loss of activity, under nonphysiological conditions, may reflect more general unfolding of the domain as multiple aspartate and glutamate side chains start to become protonated. This interpretation suggests that the increased steepness of the pH curves at low pH values (Fig. 3) reflects the involvement of multiple protons in bringing about this unfolding. It is also possible that carboxylate coordination ligands for the two Ca 2ϩ become sufficiently protonated at this pH to result in release of Ca 2ϩ and hence loss of ligand binding activity.
Examination of the CRDs from multiple C-type lectins with differing pH profiles provides additional evidence for the importance of the residue corresponding to His 256 in establishing pH sensitivity in the physiological pH range. Although the CRD of RHL-2/3, the minor subunit of the hepatic asialoglycoprotein receptor, is 59% identical to RHL-1, it binds Ca 2ϩ more weakly and does not show a loss of affinity as the pH decreases from 7.8 to 5.4. Interestingly, this subunit differs at positions corresponding to His 256 , Asp 266 , and Arg 270 (Fig. 2). These residues are also absent from almost all other CRDs in the C-type lectin family, which correlates with the fact that most of these proteins are not known to undergo pH-dependent release of ligand under physiological conditions.
CRDs that do contain a histidine side chain at the position corresponding to His 256 of RHL-1 include the chicken hepatic lectin and the macrophage asialoglycoprotein receptor. The pH dependence of ligand binding to the chicken hepatic lectin is closely similar to RHL-1, although the amino acid residues corresponding to Asp 266 and Arg 270 are not conserved (Fig. 2), suggesting that, while the histidine side chain may remain the  7. pH dependence of Ca 2؉ affinity for QPDWG and mutant QPDWGHD. K Ca values were derived from fluorescence measurements, using three different sets of buffers as indicated in Fig. 1. QPDWGϩHD denotes QPDWG containing histidine and aspartic acid at positions corresponding to His 256 and Asp 266 of RHL-1. Lines were fitted to the data based on the approximately linear relationship between log K Ca and pH. Shaded band represents the physiological Ca 2ϩ concentration range. primary pH sensor, the effective pK a of this side chain is established in different ways. The absence of the critical histidine residue from the primary ligand-binding CRD of the macrophage mannose receptor (CRD-4) correlates with the fact that ligand release from this CRD occurs at lower pH than from RHL-1 (21).
Other endocytic receptors share with the C-type lectins the presence of divalent cations at critical sites in the ligandbinding domains. The Mn 2ϩ residue that forms part of the ligand-binding site in the cation-dependent mannose 6-phosphate receptor is in close proximity to a histidine residue that has been suggested as a possible pH-dependent regulator of sugar binding (22). Many other endocytic receptors, such as the low density lipoprotein receptor, contain homologous complement-like repeats that are folded around a Ca 2ϩ . Several of these domains have been characterized structurally and it has been suggested that loss of ligand binding activity at endosomal pH may be mediated by release of the Ca 2ϩ (23). However, examination of the sequences of the Ca 2ϩ -containing complement-like repeats fails to reveal a conserved histidine residue near to the Ca 2ϩ site. In different receptors, other mechanisms must be responsible for determining the pH-sensitive changes in ligand binding activity that are essential to the endocytic processes. In addition, the concentration of Ca 2ϩ in endosomes appears to be reduced compared with the extracellular environment, providing an alternative route to destabilization of Ca 2ϩ -dependent receptor-ligand complexes (24).