Designed Human Serum Hyaluronidase 1 Variant, HYAL1ΔL, Exhibits Activity up to pH 5.9

Hyaluronidases from diverse species and sources have different pH optima. Distinct mechanisms with regard to dynamic structural changes, which control hyaluronidase activity at varying pH, are unknown. Human serum hyaluronidase 1 (HYAL1) is active solely below pH 5.1. Here we report the design of a HYAL1 variant that degrades hyaluronan up to pH 5.9. Besides highly conserved residues in close proximity of the active site of most hyaluronidases, we identified a bulky loop formation located at the end of the substrate binding crevice of HYAL1 to be crucial for substrate hydrolysis. The stretch between cysteine residues 207 and 221, which normally contains 13 amino acids, could be replaced by a tetrapeptide sequence of alternating glycine serine residues, thereby yielding an active enzyme with an extended binding cleft. This variant exhibited hyaluronan degradation at elevated pH. This is indicative for appropriate substrate binding and proper positioning being decisively affected by sites far off from the active center.

Hyaluronan (HA), 3 a linear polysaccharide found in the extracellular matrix of most tissues and body fluids of vertebrates, is enzymatically degraded by hyaluronidases (1). Mammalian-type hyaluronidases are grouped into EC 3.2.1.35 (2,3) or the glycoside hydrolase family 56 (4). Members of this enzyme family hydrolyze the 1,4-␤-glycosidic linkage between N-acetyl-D-glucosamine and D-glucuronate within HA polymers (5). In mammalians, hyaluronidases have been found in testis, liver lysosomes, and serum. They are involved in controlling HA levels and are thus implicated in various diseases related to defects of HA metabolism (6).
The crystal structures of hyaluronidase from bee (7), wasp (8), and only recently that of human serum hyaluronidase 1 (HYAL1) (9) have been deciphered. In addition to the N-terminal catalytic domain of the insect enzymes, which resembles a distorted (␤/␣) 8 barrel, HYAL1 contains yet another domain. HA hydrolysis is achieved by a pair of acidic amino acids via a retaining double displacement mechanism and a substrate-assisted catalysis, in which the carbonyl oxygen of the N-acetyl group of the cleaved HA subunit acts as the catalytic nucleophile (7).
Mammalian-type hyaluronidases display different pH optima. HYAL1 (10) and hyaluronidase 2 (HYAL2) (11) exhibit highest activities at acidic conditions, whereas the hyaluronidase found in Xenopus laevis kidney is only active at neutral pH (12). Bee venom hyaluronidase (13), as well as sperm hyaluronidase, PH20 (SPAM1) (14), are capable of degrading HA over a broad pH range. Up to three PH20 isoforms with greatly different pH optima could be found in protein preparations from bovine testis (15). Extensive analysis of hyaluronidase structures did not bring forward any insights as to what residues or regions of the enzymes specify a specific pH optimum.
Profiles of pH-dependent activities can be assigned by computing the electrostatic interactions of the enzyme, which are primarily determined by the ionization states of its amino acid side chains. The pK a values of titratable groups of the enzyme reflect pH-dependent properties such as stability, enzymatic interaction, and substrate interactions (16). Here we present computational and experimental data on the replacement of a loop region located at the end of the substrate binding groove yielding a variant hyaluronidase with an altered pH profile.

EXPERIMENTAL PROCEDURES
Molecular Modeling-Tertiary structure models were generated (17) with Deep View/Swiss-Pdb Viewer version 3.7 (18) using the three-dimensional coordinates of bee hyaluronidase (Protein Data Bank (PDB) accession 1fcv) and HYAL1. (PDB accession 2pe4). Images were generated with the aid of the open source PyMOL 0.99rc6 (19).
Mutagenesis of HYAL1-Coding regions of wild-type (accession number NM_007312) and mutated HYAL1 were inserted into the plasmid pT7TS (P. A. Krieg, University of Arizona) using restriction sites EcoRV and BcuI, respectively. Fourprimer PCR technology was applied to substitute the native amino acid sequence between cysteine residues at positions 207 and 221. Flanking primers together with the primer pair 5Ј-TGGGCAGGAGCCACTACCGCAGTCAGGGAAGCC-3Ј and 5Ј-GACTGCGGTAGTGGCTCCTGCCCATCAGGC-ATC-3 were used to replace the original loop by the sequence of the 4 amino acid residues, glycine serine glycine serine, resulting in a cDNA encoding HYAL1 ⌬L .
Hyaluronidase Expression in X. laevis Oocytes-Frog surgery, oocyte preparation, and cRNA injection was accomplished as described previously (20). The coding regions of HYAL1 and HYAL1 ⌬L in pT7TS vector linearized with BamHI were transcribed in vitro. 50 ng of capped cRNA were injected into freshly prepared oocytes. The cells were kept in culture medium O-R2 containing antibiotics at 16°C for 2 days.
pK a Calculations and Electrostatic Potential Calculations-Titration curves for all titratable groups were calculated using the WHAT IF pK a calculation package, as described previously (21), except using a single uniform protein dielectric constant of 8. Calculations for HYAL1 and HYAL1 ⌬L are available (using the pKD program available from the Protein Analysis and Design Group, with accession numbers 2pe4_apo.pka.pdb and hyal_deltaL_apo.pka.pdb). Titration curves and pK a values were analyzed with pKaTool (16).
Electrostatic potential maps at pH 3.8 and pH 5.5 were calculated by linearly scaling the optimized potentials for liquid simulations (OPLS) charges (22) for each titratable group by the fractional degree of protonation of that group at the desired pH value, which were directly derived from the titration curves calculated in the pK a calculations. The hydrogen bond network in each structure was optimized as described previously (23), and the electrostatic potential was solved using DelPhi 2 (24) using a 65-cubed grid, a protein dielectric constant of 8, a solvent dielectric of 80, an ionic strength of 0.144 M, and an ion exclusion layer of 2.0 Å. Electrostatic potential maps were visualized using PyMOL (19).
Hyaluronidase Activity Measurement and Evaluation-After cRNA injection and culture, healthy oocytes were pooled and homogenized prior to enzymatic analysis. As described previously (12,25), the amount of lysate and the time of hydrolysis were titrated to specify the protein amount and assay duration, at which cleavage rates were in the linear range appropriate for subsequent comparative analyses of the respective pH profiles. Hydrolysis rates were quantified by determining the relative median length along with the amount of the remaining hyaluronan applying histogrammatic characterization of the digitized blots using ImageJ (NIH). A paired, two-tailed Student's t test was applied for statistical evaluation.

RESULTS
HYAL1 was initially expressed in Escherichia coli. The expressed recombinant protein was exclusively found as an insoluble protein aggregate, and no enzymatic activity could be obtained following solubilization and refolding. After separation by SDS-PAGE, recombinant HYAL1 was used to raise polyclonal antibodies in rabbits, which were applicable for Western analysis.
Bee hyaluronidase has been co-crystallized in the presence of HA tetrasaccharides (PDB accession code: 1fcv) (7). Due to the polymeric structure of HA, substrate binding is mediated alongside one-half of the globular protein, opposite to the side where the N and C termini are located. In HYAL1, across from the active center and located at the end of the substrate binding groove, a loop, which is positioned between 2 cysteine residues at positions 207 and 221, leads to a tapering geometry thereof. Hence, we altered the native sequence between the two cys-teines both in silico (Fig. 1) as well as in reality (Table 1) by replacing it either with one of the respective sequences of the bee venom (HYAL1 ⌬beeh ) or with that of the testicular (human PH20) hyaluronidase (HYAL1 ⌬PH20 ) or by mutating the potential N-glycosylation site at position 218 from a threonine to an alanine HYAL1 ⌬glyc . Employing a protein structure modeling FIGURE 1. Structural models. On the basis of the three-dimensional structures of HYAL1 and bee hyaluronidase, models for wild-type hyaluronidase (HYAL1 WT ) and the loop variants (HYAL1 ⌬glyc , mutated potential glycosylation site (arrowhead) within the loop; HYAL1 ⌬beeh , native sequence replaced with loop sequence of that of bee venom hyaluronidase; HYAL1 ⌬PH20 , replaced with sequence of human testicular hyaluronidase, PH20; and HYAL1 ⌬L ) with a bound HA tetrasaccharide (stick mode) were generated. For detailed sequence information for these models, see Table 1. The active center and the loop region are highlighted in color.

TABLE 1 Sequence and activity information on HYAL1 variants
The mutated N-glycosylation site is indicated in bold, and flanking cysteine residues are indicated in bold and italics. ND, not detectable.

Loop sequence
Relative activity pH 4.5 pH 5.5 approach, this loop region may be replaced without distorting the structural conformation of the enzyme by the shortest possible stretch of 4 amino acids. In the model, the sequence glycine-serine-glycine-serine (GSGS) was introduced, and the resulting protein model was termed HYAL1 ⌬L (Fig. 1). The latter could be also expressed as an active enzyme.
In previous analyses, we showed that highly active hyaluronidase can be reliably expressed by injection of cRNA into X. laevis oocytes (12,20,26). Here we continued to employ this method to express a variety of HYAL1 variants. The activity of the variants was first examined at pH 4.5 and pH 5.5, the former being optimal, the latter being inhibitory for wild-type HYAL1. Apart from HYAL1 ⌬beeh , all variant forms were active at pH 4.5. Only HYAL1 ⌬L hydrolyzed hyaluronan at pH 5.5, albeit at a decreased rate (Table 1).
Next, cell lysates containing equal amounts of wild-type HYAL1 or HYAL1 ⌬L protein (Fig. 2) were analyzed side-byside, and the enzyme activity was assessed at a wide range of different pH values. Both HYAL1 and HYAL1 ⌬L efficiently degraded HA at pH values around 4. As reported earlier, HYAL1 is inactive above pH 5.1, whereas HYAL1 ⌬L exhibited activity up to pH 5.9 (Fig. 3).
The mutated region neither appears to be part of the active center, nor is it conceivable that direct interactions with residues of the active center do occur. To corroborate this assumption, we calculated pK a values for all residues and analyzed the titration curves of the two carboxyl groups of the active site amino acids, aspartate (Asp-129) and glutamate (Glu-131). The active site pK a values appeared relatively insensitive to the dele-tion. In HYAL1, the pK a value of Asp-129 is 6.1, and that of Glu-131 is 9.4. Similarly, the respective values in HYAL1 ⌬L are 6.3 and 9.5. The calculated pK a values are in good agreement with the observed pH activity profiles of both enzymes, although the magnitude of the pH shift for HYAL1 ⌬L is not accurately captured by these calculations. Because the deletion is a substantial modification, it is possible that structural relaxation, which is neglected in the HYAL1 ⌬L model, causes minor conformational rearrangements at the active site that further alters the pK a values of Asp-129 and Glu-131.
The mutated loop is demarcating the substrate binding crevice at one of its ends. Appropriate binding of HA, in particular the site-specific positioning of the N-acetyl group at the active center, is pivotal for efficient hydrolysis of the polymer. Hence not only steric hindrance, but also a modified intermolecular binding capacity, may contribute to the observed change in the FIGURE 2. Western blot analysis of recombinantly expressed HYAL1 enzymes. Homogenized cell lysates of X. laevis oocytes expressing wild-type hyaluronidase 1 (HYAL1 WT ), the loop-mutant (HYAL1 ⌬L ), or uninjected controls were separated on 12% polyacrylamide gel, electroblotted onto nitrocellulose membrane, and immunochemically detected applying our polyclonal antiserum and anti-rabbit-horseradish peroxidase conjugates. Molecular mass is indicated to the left in kDa. pH activity profile of the enzyme. We therefore calculated the surface electrostatic potential for HYAL1 and HYAL1 ⌬L at pH 3.8 and pH 5.5 using calculated pK a values (electrostatic potential, data not shown) (21,27). HYAL1 ⌬L exhibited a more positive surface electrostatic potential at pH 5.5 than wild-type HYAL1. Thus, besides the protonation states of the active site residues, HA association and dissociation may be rate-limiting for HYAL1 activity. Not only does the absence of the loop in HYAL1 ⌬L make it conceivable that the binding crevice is rendered more accessible to the HA polymer (Fig. 1), but also, changes in the electrostatic potential maps suggest that the pK a values in HYAL1 ⌬L , which govern substrate binding and product release, are altered.

DISCUSSION
Insect hyaluronidases fold into a single domain. This supports the notion that the homologous part of other hyaluronidases is also sufficient for HA hydrolysis. Specific properties of mammalian-type hyaluronidases such as their pH optima may, however, be effectively influenced by the C-terminal extensions. Vertebrate enzymes come in various isoforms, most likely resulting from posttranslational proteolytic cleavage(s). For instance, mature hyaluronidases can be liberated from their glycosylphosphatidylinositol-anchored state by proteolysis close to the C terminus. In some cases, this first processing is followed by a second proteolytic cleavage. In the case of HYAL1, both processing steps occur. A 57-kDa form is the only isoform found in plasma; besides the 57-kDa form, also a 45-kDa form is found in urine (28). A soluble form of PH20 is also well documented (29,30). The processed 53-kDa form has a pH optimum of 4.0, whereas the unprocessed form is most active at neutral pH (15). The conformational differences at the active center and variations regarding substrate positioning to achieve such differences have not been investigated.
Performing (site-directed) mutagenesis and recombinant expression in Xenopus oocytes, we studied the serum enzyme, HYAL1, which as shown by others previously is only active at pH values below 5.1 (10,31,32). Glutamate at position 131 is the proton donor in the proposed catalytic mechanism for hyaluronidases (7), but Glu-131 is unlikely to be able to donate a proton to the substrate without Asp-129 being protonated. The lowest pK a value in the pair Asp-129 and Glu-131 will therefore govern the pH activity profile of the enzyme. Experimental data on the catalytically competent protonation state in xylanase variant Bcx N35D (family 11 glycosidases) supports the conclusion that only the doubly protonated state is catalytically active (33,34). Many residues in the proximity of the active site are conserved within hyaluronidases. In most cases, mutation at these positions rendered HYAL1 inactive, presumably due to steric hindrance and/or incorrect folding. 4 Because HA is a charged polymer, the pH dependence for substrate association as well as product dissociation must not be neglected. To change the gross geometry of the binding cleft, we mutated the bulky epitope at the tapering end of the binding cleft, which folds up between cysteine residues 207 and 221 and contains 13 amino acids. Mutation of the potential glycosylation site (T218A) within the loop region did not greatly alter the activity of the enzyme with respect to pH. Similar results were obtained when the stretch between the two cysteines was replaced with the interjacent sequence of PH20/SPAM1, a hyaluronidase that also exhibits activity at neutral pH. In addition to being exactly 13 amino acids long, it also bears four positive charges (His-210, His-211; Lys-214, Lys-215) when compared with one negative charge (Asp-211) present in the HYAL1 loop. It is therefore conceivable that due to stronger ionic interactions with the hyaluronan polymer at positions a long way off from the active center, substrate binding at elevated pH may actually be enforced. Although active at pH 4.5, HYAL1 ⌬PH20 was found inactive at pH values above 5. Another family member of the mammalian-type hyaluronidases, which exhibits a broad pH profile still being active at neutral pH, is the enzyme secreted into bee venom (13). Its loop sequence is by 2 amino acid moieties shorter and actually contains no charged residues. Unexpectedly, this variant, although exhibiting some activity at pH 3.8, turned out to be inactive at pH 4.5, demonstrating that the mutated region either appears to be essential for direct substrate binding or is involved in proper positioning of the polymer with respect to the catalytic center. We next applied computer-assisted protein modeling to design a GSGS-interposed variant that displays a greatly enlarged binding crevice. This variant was found active, and in addition, it also processed HA at elevated pH. We therefore conclude that raising pH does not induce major conformational or chemical changes at the active center of HYAL1 ⌬L , and HA can still be sufficiently bound and properly positioned there to be efficiently hydrolyzed.
By performing pK a calculations, we were able to qualitatively predict the upshift in the pH activity profile but only inadequately able to predict the magnitude of the pH change. It is likely that the deletion in HYAL1 ⌬L causes minor structural relaxation that affects the packing around Asp-129 and Glu-131, thus causing the pK a values of these residues to shift even more than computed. Because structural rearrangements are not accurately comprehended by homology modeling methods, we cannot take these into account in the pK a calculations. However, it is well documented that even small changes in protein structures can lead to large changes in pK a value calculations (16,35). It is furthermore possible that interactions between the protein and the glycan side chains are affected by secondary sites, which induce long range effects on substrate accommodation at the active site.