Carboxyl-Terminal Truncations Alter the Activity of the Human α-Galactosidase A

Fabry disease is an X-linked inborn error of glycolipid metabolism caused by deficiency of the human lysosomal enzyme, α-galactosidase A (αGal), leading to strokes, myocardial infarctions, and terminal renal failure, often leading to death in the fourth or fifth decade of life. The enzyme is responsible for the hydrolysis of terminal α-galactoside linkages in various glycolipids. Enzyme replacement therapy (ERT) has been approved for the treatment of Fabry disease, but adverse reactions, including immune reactions, make it desirable to generate improved methods for ERT. One approach to circumvent these adverse reactions is the development of derivatives of the enzyme with more activity per mg. It was previously reported that carboxyl-terminal deletions of 2 to 10 amino acids led to increased activity of about 2 to 6-fold. However, this data was qualitative or semi-quantitative and relied on comparison of the amounts of mRNA present in Northern blots with αGal enzyme activity using a transient expression system in COS-1 cells. Here we follow up on this report by constructing and purifying mutant enzymes with deletions of 2, 4, 6, 8, and 10 C-terminal amino acids (Δ2, Δ4, Δ6, Δ8, Δ10) for unambiguous quantitative enzyme assays. The results reported here show that the k cat/K m approximately doubles with deletions of 2, 4, 6 and 10 amino acids (0.8 to 1.7-fold effect) while a deletion of 8 amino acids decreases the k cat/K m (7.2-fold effect). These results indicate that the mutated enzymes with increased activity constructed here would be expected to have a greater therapeutic effect on a per mg basis, and could therefore reduce the likelihood of adverse infusion related reactions in Fabry patients receiving ERT treatment. These results also illustrate the principle that in vitro mutagenesis can be used to generate αGal derivatives with improved enzyme activity.


Introduction
Mutations in the αGal gene result in the sphingolipidosis named Fabry disease [1]. The enzymatic defect is inherited as an X-linked disorder and is associated with a progressive deposition of the glycosphingolipids, including globotriaosylceramide, galabioasylceramide, and blood authors and the funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The corresponding author has read the journal's policy and the authors of this manuscript have the following competing interests: The corresponding author (DHC) is an officer of ShelbyBioTech LLC and Shelbyzyme LLC. The work reported here was supported in part by a research grant from ShelbyBioTech LLC and in part by the other sources listed in the funding section above. The City University of New York and Shelbyzyme LLC have patents related to Fabry disease and the corresponding author has a financial interest in these patents. The university, these corporations, and these funding agencies had no role in study design, data collection and analysis, decision to publish, or preparation of this manuscript. The university (Research Foundation of the City University of New York) was assigned ownership of the related U.S. Provisional Application No. 62/011,399, "Purification of Recombinant Human α-Galactosidase A with Cterminal Truncation with Increased Activity". The university delegated all academic responsibilities to the authors and has no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The university has submitted past applications to the U.S. Patent Office that have resulted in other issued or pending patents related to Fabry disease but these are not directly related to the pending manuscript (PONE-D- ). This does not alter the authors' adherence to all PLOS ONE policies on data sharing and materials. recombinant human αGal preparations used therapeutically are produced in human and CHO cells and these have distinct glycosylation patterns and differ in levels of sialic acid and mannose-6-phosphate [24]. The recombinant αGal produced in insect cells [49,50] and in P. pastoris [53] contain variable levels of mostly complex and high mannose side chains, respectively. Glycoproteins produced in P. pastoris typically contain from 6 to 14 mannose units (Man 6-GlcNac 2 to Man 14 GlcNac 2 ) that sometimes produces a Gaussian-like distribution of oligomannosides that may center near Man 12 GlcNac 2 to Man 13 GlcNac 2 [59].
These carbohydrate moieties serve a structural and functional role. For example, it has been demonstrated that glycosylation, particularly at Asn-215, is required for enzyme solubility [26]. Also, uptake of the enzyme by cells in vivo is affected by terminal mannose-6-phosphate residues on the enzyme [27], and the 10-12 sialic acid residues on the plasma form of the enzyme accounts for the prolonged circulatory half-life of the enzyme compared to the tissue form with only one or two sialic acid residues [60]. The identification of these multiple forms as derivatives of the same protein in purified enzyme preparations can conveniently be monitored by treatment with specific N-glycosidases or by Western blots.
Fabry disease patients with adverse reactions to the infusions are currently treated with antihistamines and antipyretics and the initial immune response has been manageable to date [61,62], but it can be anticipated that life-long treatment required for these patients will lead to unacceptable levels of neutralizing antibodies. In this context it is reasonable to devise approaches to circumvent these adverse reactions and the development of derivatives of the enzyme with more activity per mg is a logical approach. Miyamura and coworkers [63] reported that carboxyl-terminal deletions of 2 to 10 amino acids of αGal led to an increase in activity of about 4 to 6-fold as compared to wild type (WT). However, this data was qualitative or semi-quantitative and relied on comparison of the amounts of mRNA present in Northern blots to αGal enzyme activity during transient infection of COS-1 cells. Here we use a P. pastoris expression system for the construction and purification of mutant enzymes with C-terminal deletions. The quantitative results reported here with purified enzymes reveal that C-terminal deletions results in an increase (Δ2, Δ4, Δ6, and Δ10) or decrease (Δ8) in enzyme activity.
Four frozen MGY cultures of 4 ml each were used to inoculate four 100 ml MGY cultures in 1-liter baffled flasks and grown at 250 rpm and 30°C until the OD600 reached 2 to 6. The cultivation was divided into three phases, the glycerol batch, glycerol-fed batch, and methanolfed batch. The glycerol batch phase was initiated with 400 ml of inoculum shake-flask culture added to 4 L of the fermentation medium containing 4% glycerol and an initial value of 100% dissolved oxygen until a spike was observed indicating complete consumption of glycerol. Next, the glycerol-fed batch phase was initiated and a 50% w/v glycerol feed rate of 18.15 ml/h/ liter initial fermentation volume and maintained until a cell yield of 180 to 220 g/liter wet cells was achieved. At this point the glycerol feed was terminated manually and a methanol-fed batch phase was initiated by starting a 100% methanol feed containing 12 ml PTM1 trace salts per liter. Methanol was initially fed at 3.6 ml/h/liter of initial fermentation volume, then increased to 7.3 ml/h/liter and finally increased to 10.9 ml/h/liter of initial fermentation volume for the remainder of the fermentation. Dissolved oxygen spikes were used during the glycerolfed batch phase and methanol-fed batch phase and to monitor substrate levels. A dissolved oxygen level of 40%, pH of 6, and temperature of 25°C were maintained by an ADI 1030 regulator. Sampling was performed at the end of each phase and at lease twice daily and analyzed for cell wet weight and increased αGal activity over time. Cultivation was terminated once a plateau in αGal activity was observed.

Construction of strains
Plasmid pMS118 [48] contains the αGal cDNA cloned as an EcoRI fragment to the EcoRI site of plasmid pUC9. PCR primers (Fig. 1, 2) were used with plasmid pMS118 DNA and the PCR system (Roche, No. 11732641001) according to the vendor's instructions. This generated cDNAs with a 5 0 extension containing an XhoI site, Kex2 and Ste13 yeast signal cleavage sites, a 3 0 end with an XbaI site, and a deletion of C-terminal amino acids to generate Δ2to Δ10 mutants ( Fig. 1, 2). The PCR products were ligated to pCR-XL-TOPO to generate Δ2to Δ10 plasmids (Table 1). These plasmids were used for electroporation [53] into E. coli strain TOP10 or TOP10F 0 (Table 1).
DNA sequence analysis using the universal M13 primers (Table 2) confirmed the expected insert for these pCR-XL-TOPO derivatives. The modified cDNAs were excised from pCR-XL-TOPO plasmids using XhoI and XbaI endonucleases and ligated into expression plasmid pPIC-ZαA treated with the same two restriction enzymes to generate expression plasmids ( Table 1) that were subsequently used for electroporation [53] into strain TOP10 or TOP10F 0 ( Table 1). The nucleotide sequence of mutant cDNAs in pPICZαA derivatives was analyzed (Genewiz) using 5 0 AOX, 3 0 AOX and α-factor sequencing primers (Table 2). These pPICZαA derivatives were used for electroporation of P. pastoris strain X-33 to generate yeast expression strains (Table 1). by 12-hour interval soaks. Elution was discontinued when the absorbance at 280 nm and enzyme assays showed negligible presence of protein and αGal activity. No substantial difference in recovered enzyme was observed between purifications carried out with modified elution buffer I versus sugar saturated elution buffer I (data not shown). The Con A pool was subjected to diafiltration using a 50 kDa pore size hollow fiber filter (Spectrum Labs, No. M25S-300-01N) against binding buffer (25 mM citrate-phosphate buffer, pH 4.8 containing 0.1 M NaCl).
The Con A pool was applied to an immobilized-D-galactose gel column (Thio-Gal, Pierce No. 20372) pre-equilibrated with binding buffer. The column was washed with 5 column volumes of binding buffer and αGal was eluted with elution buffer II (25 mM citrate-phosphate buffer, pH 5.5, 0.1 M NaCl, 0.1 M D-galactose) over 1.5 column volume blocks separated by 12 Introduction of a C-terminal deletion of 2 amino acids into αGal. The strategy shown here for the Δ2 mutant was used to generate all five deletion mutations (Fig. 2). Plasmid pMS118 [48] contains the αGal cDNA cloned as an EcoRI fragment to the EcoRI site of plasmid pUC9. Primers AAF and AARD2 ( Fig. 2) were used as PCR primers for plasmid pMS118 DNA to generate cDNAs with a 5 0 extension containing an XhoI site, Kex2 and Ste13 yeast signal cleavage sites, a 3 0 end with an introduced XbaI site, and a deletion of C-terminal amino acids to generate the Δ2 mutant. Primer AAF anneals to the cDNA at the sequences encoding the N-terminal sequences of αGal and primer AARD2 anneals to the C-terminal sequences of αGal. Primer AARD2 anneals 12 nucleotides from the 3 0 end of the cDNA and introduces a stop codon (UAA) after the aspartate codon three amino acids from the C-terminal end of the coding sequences of αGal resulting in a deletion of the two C-terminal amino acids (Leu-Leu) of the human enzyme (right panel). Cloning to the XhoI and XbaI sites of plasmid pPICZαA generates a protein fusion with the yeast signal peptide coding sequences in the vector. This signal peptide is removed by the Kex2 and Ste13 yeast signal peptidases through cleavage immediately upstream of the leucine corresponding the first amino acid of the mature form of αGal (left panel). This strategy was generalized to create the other deletion mutants using the primers in Fig. 2. In the left panel, the N-terminal peptide LDNGLAR was identified in mass spectrometric analysis while EALDNGLAR was not (Fig. 5).  Primers and αGal cDNA used to generate Δ 2, Δ 4, Δ 6, Δ 8 and Δ 10 mutant cDNAs. DNA primers AAF, AARD2, AARD4, AARD6, AARD8 and AARD10 corresponding to the Δ2, Δ4, Δ6, Δ8 and Δ10 mutants were annealed to the cDNA of mature αGal contained within pMS118 to generate 3 0 end truncated PCR products for carboxy-terminal deleted enzymes. Primer AAF and primers AARD2 to AARD10 (indicated above) were annealed to the 5 0 and 3 0 ends of the cDNA, respectively. Primer AAF contains an XhoI site (indicated above) and partially encodes for a yeast signal peptide (see Fig. 1) to produce a fusion protein targeted for secretion from P. pastoris. Primers AARD2 to AARD10 were used to introduce an XbaI site (indicated above) and a premature UAA stop codon via an antisense ATT triplet immediately downstream of nucleotides complementary to αGal (bold font) to produce cDNAs encoding for Δ2, Δ4, Δ6, Δ8, Δ10 mutants. The boxed LDNGLAR and SHINPTGTVLLQLENTMQM protein sequences (indicated above) are peptide fragments that were identified through mass spectrometry of the Δ6 mutant (Fig. 5 hour soaks. Fractions were assayed for enzyme activity and protein concentration and a peak tube with high specific activity was chosen as the sample to be used in a substrate saturation curve.

Western blot analysis
Western blot analysis was performed using an anti-αGal polyclonal antibody produced in chicken (Pierce/ThermoSci #PA1-9528) and horseradish peroxidase-conjugated anti-Chicken IgY antibody (Sigma #A9046). After SDS-PAGE (2 μg of samples loaded), the gel was incubated with a nitrocellulose membrane (Whatman, No. 10402594) for 15 minutes at room temperature in Transfer Buffer (48 mM Tris, 39 mM glycine, 20% MeOH, pH 9.2) and the proteins were then transferred to the nitrocellulose membrane using a Bio-Rad Trans Blot SD Semi-Dry Transfer Cell. The membrane was blocked with 8% (w/v) non-fat dried milk in PBST [10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl and 0.2% Tween 20 (pH 7.4)] at room temperature for 20 minutes. The membrane was then treated with primary antibody diluted in a milk/blot solution [1% (w/v) non-fat dried PBST] for 2 h at room temperature with mild shaking. After rinsing with PBST solution, the membrane was treated for 1 h at room temperature with secondary antibody diluted in the milk/blot solution. Protein bands were visualized on Kodak BioMax XAR film (VWR #lB1651454) with a Konica SRX-101A processor.

Enzyme and protein assays
Activity of αGal was assayed using the synthetic substrate, 4-methylumbelliferyl-α-D-galactopyranoside (MUG) as described [53] with modifications to a microtiter plate format (below). Enzyme activity is measured in units/ml where one unit is defined as the amount of enzyme required to convert 1 nmole of MUG to 4-methylumbelliferone in one hour at 37°C. An aliquot of 3 μl was added to 27 μl of enzyme assay buffer (5 mM MUG in 40 mM sodium acetate buffer, pH 4.5). This mixture was incubated at 37°C and 10 μl aliquots were taken at two time points and added to 290 μl of 0.1 M diethylamine in a microtiter plate to stop the reaction. Typically time points were chosen as 1-4 minutes and values that were proportional to time were considered valid. The fluorescence of each sample was measured at an excitation wavelength of 365 nm and an emission wavelength of 450 nm using a Tecan Infinite F200 microtiter plate reader. A standard curve of 10 μl of 0 -0.5 nmol 4-methylumbelliferone dissolved in MeOH in 290 μl of 0.1 M diethylamine was used to quantitate MUG cleavage at specific time intervals. Analysis of the effects of MeOH indicated no effect on the 4-methylumbelliferone standard curve. For samples containing higher protein concentrations, the BioRad DC Protein Assay (No. 500-0116) with a standard curve of (0.2 -1.5) mg/ml was used according to the manufacturer's specifications. For dilute samples of purified αGal, a more sensitive fluorescence-based fluorescamine assay [68] with a standard curve containing lower protein concentrations of (4.0-160) μg/ml was used. Briefly, 150 μl of 0.05 M sodium phosphate buffer and 50 μl of 1.08 mM fluorescamine dissolved in acetone were added to an aliquot of 50 μl of the sample and standards, mixed and incubated for 12 minutes. The fluorescence of each sample was measured at an excitation wavelength of 400 nm and an emission wavelength of 460 nm. Bovine serum albumin (Bio-Rad No. 500-0112) was used as the standard in both assays. Absorbance and fluorescence measurements were conducted on a Tecan Infinite F200 microplate reader using 96well plates.

Mass spectrometry of a purified mutant enzyme
The Δ6 mutant was selected for mass spectrometry analysis conducted at the Rockefeller University Proteomics Resource Center in collaboration with M.T. Mark. SDS-PAGE gel slices were washed, de-stained, reduced using 10 mM dithiothreitol, alkylated using 100 mM iodoacetamide, and digested using trypsin. Peptides were then extracted from the gel two times, dried, and re-suspended in a 5% acetonitrile and 2% formic acid mixture. One third of each sample was loaded onto a C18 PepMap1000 micro-precolumn (300 μm I.D., 5 mm length, 5 μm beads, Thermo Scientific) at a flow-rate of 5 μl/min, and subsequently onto an analytical C18 column (75 μm I.D., 3 μm beads, Nikkyo Technos Co.) at a flow rate of 300 nl/min. The gradient was 40 min long in the range 5 to 45% B (buffer A was 0.1% formic acid in water, and buffer B was 0.1% formic acid in acetonitrile). Eluted peptides were applied by electrospray directly into the LTQ-Orbitrap XL mass spectrometer from Thermo Scientific, operating in a 300 to 1800 m/z mass range. Tandem mass spectrometry was performed by collision induced dissociation using nitrogen as a collision gas. The resulting spectra were analyzed using Mascot and Proteome Discoverer 1.3 (Thermo Scientific) to identify the peptides in the sample.

Thermostability and pH optimum of WT and mutant αGal
Purified enzyme samples were diluted in 25 mM citrate-phosphate buffer, pH 5.5, 0.1 M NaCl, 0.01 M D-galactose. Samples of 50 μl were incubated in triplicate at 50°C, 30°C and 40°C. Aliquots of 3 μl were removed for enzyme assays every 15 minutes for two hours. Samples were assayed in 0.02 M citrate buffer, pH 3.0-pH 6.5, containing 2 mM MUG.

Characterization of kinetic properties
Substrate saturation curves for αGal have been reported using MUG at concentrations up to 2 mM, 5 mM, and 10 mM (in the presence of 0.1% BSA and 0.67% EtOH [24]). We noted that under our experimental conditions MUG is fully soluble at 2 mM, partially soluble at 5 mM, and chemically oversaturated at higher concentrations. Other investigators reported the use of sonication or detergent treatment to increase the solubility of MUG (e.g., [69]) but we avoided this approach in order to avoid potential artifacts due to the use of these techniques. Substrate saturation curves using 2 mM and 5 mM MUG as the highest concentrations were carried out and the kinetic parameters for αGal were calculated separately obtaining similar values. The values reported here (Table 3a) were obtained using a substrate saturation curve of 0.3 to 2 mM MUG since this is the highest concentration that is fully soluble under our experimental conditions. The K m and V max values were calculated using Lineweaver-Burk and non-linear regression through the program Sigma-Plot (Systat Software, San Jose, CA).
Kinetic parameters were also determined using the colorimetric substrate, para-nitrophenyl-α-D-galactopyranoside (PNPαGal) [70]. Purified enzymes were diluted to approximately 20,000 units/mL as determined by fluorescent MUG assay. These diluted samples were then added at a proportion of 1:9 citrate-phosphate buffer (0.1 M) containing 7 -50 mM PNPαGal. Aliquots of 20 μl of the enzymatic reaction were removed at 15 minute intervals to terminate the reaction over the course of an hour and added to 320 μl of borate buffer (pH 9.8) in a microplate [71]. Product formation was monitored by absorbance at 400 nm. Linear reaction velocities were observed for all measurements. A standard curve of 0-150 μM p-nitrophenylate in borate buffer (pH 9.8) [71] was used to quantitate product formation. K m and V max parameters were determined through non-linear regression using Sigma-Plot (Systat Software, San Jose, CA).

Protein structure analysis
The crystal structure of αGal (PDB 1R47) was viewed and analyzed in PyMOL (Delano Scientific). The MSLDKLL and QMSLKDLL peptides corresponding to the last 7 or 8 C-terminal amino acids of αGal were built in PyRosetta [72] and visualized in PyMOL [73]. Interatomic distances were measured using the PyMOL wizard distance command.
A homology model of the coffee bean α-galactosidase was generated on the Phyre2 server [74]. The primary sequence of coffee bean α-galactosidase (GenBank No. AAA33022.1) was set as the query. The crystal structure of rice α-galactosidase (73% sequence identity to coffee α-galactosidase, PDB# 1UAS) was set as the template. Superposition of the coffee homolog and human crystal structure of αGal (PDB# 1R47) was conducted in PyMOL [73]. Primary sequence alignments were carried out in ClustalOmega (EMBL-EBI).

Purification of WT and mutant αGal
The WT and mutant enzymes were obtained from a 7 L bioreactor and purified (Table 4, Fig. 3) using Con A and Thio-Gal tandem affinity chromatography. This two column purification is simpler and faster than our previous purification methods that used three [50] or four [49,53] columns and the yield, degree of purity, and final specific activities were similar for all three methods. The non-glycosylated form of αGal (41.8 kDa) is isolated from cells as multiple glycosylated species with a predominant band of about 50 kDa and multiple higher molecular weight forms that differ in extent of glycosylation ( Fig. 3; See Introduction). We have previously demonstrated that high molecular weight glycoforms produced in insect cells and P. pastoris can be identified as derivatives of αGal rather than contaminants and these glycoforms are converted to a single band on SDS gels of about 41.8 kDa with endoglycosidase treatment [49,50,53]. In this report we also use a Western blot (Fig. 4) to confirm that the high molecular weight forms seen on SDS gels for the WT and deletion mutants are all glycoforms of αGal. In some cases lower molecular weight species present in purified enzyme preparations can be identified as αGal fragments in Western blots (e.g., Fig. 4, lane 2). We quantitated the distribution of glycoforms in (Fig. 3, S1 Fig, online supplement) and there is no obvious correlation between the glycosylation pattern and catalytic activity. It is well established that glycosylation affects enzyme stability and enzyme uptake (above) but to our knowledge there is no evidence that the glycosylation pattern affects the catalytic properties of this enzyme.

Mass spectrometry of a purified mutant enzyme
Two possible amino terminal amino acids, glutamate or leucine, could be produced in P. pastoris depending upon the selection of the signal peptidase, Kex2 or Ste13 (Fig. 1). Due to the fact that potential improper amino terminal processing may have an effect on kinetics, we selected one of the purified mutant enzymes (Δ6) for mass spectrometry analysis in order to identify the amino terminal sequence of this enzyme. This analysis also made it possible to provide independent verification of the expected changes in the C-terminal amino acid sequence predicted by in vitro mutagenesis (Fig. 1, 2).
The mature form of the enzyme (signal peptide removed; [75]) produced in humans begins with a leucine codon (Fig. 1, 2). Therefore, tandem mass spectrometry following tryptic digestion of the Δ6 αGal purified from P. pastoris could produce tryptic peptides EALDNGLAR or LDNGLAR, depending upon the use of the Kex2 or Ste13 protease sites (Fig. 1, 2). A putative LDNGLAR peak was identified in the MS spectra with an m/z of 379.71, consistent with the (M+2H) 2+ state of this peptide, while no peaks consistent with an EALDNGLAR peptide were found. We cannot eliminate what we consider to be the less likely possibility that the failure to detect the EALDNGLAR peptide may be due to the failure of the peptide to ionize in this MS experiment. Further fragmentation of the m/z = 379.71 associated peptide peak produced an MS/MS spectrum containing 4 of 7 possible y-ions and 4 of 7 possible b-ions from the expected fragmentation pattern of a hypothetical LDNGLAR peptide (Fig. 5a). This result indicates that the Ste13 signal peptidase of P. pastoris generates an enzyme with an amino terminus identical to the enzyme produced in humans. A search for a Δ6 C-terminal tryptic peptide of SHINPTGTVLLQLENTMQM (Fig. 2) yielded a matching MS m/z = 1064.03 peak consistent with its (M+2H) 2+ state. Further fragmentation also produced an MS/MS spectrum containing 4 of 19 possible y-ions and 9 of 19 possible b-ions consistent with the anticipated sequence (Fig. 5b). This result confirms the predicted C-terminal deletion of 6 amino acids and confirms the efficacy of the mutagenesis protocol used to produce this mutant enzymes.
Thus, the purified Δ6 αGal mutant possesses an N-terminal sequence corresponding to the mature form of αGal and a C-terminal sequence truncated by six amino acids. . The minor bands present in the purified fraction are consistent with high molecular weight glycoforms seen previously when WT enzyme was purified from the same P. pastoris expression system [53].   (Fig. 6). The general trend of these profiles agree with previous results [76]. All enzymes show optimal activity near pH 4.5 (Fig. 7) in accord with previous reports for WT αGal [70,[77][78][79], and there is no significant difference in the activity optima of purified WT and mutant αGal.

Kinetic analysis of WT and C-terminal deletion mutants
The values for K m and V max for WT enzyme (Table 3a) are in accord with published values ( Table 5). The range of K m and V max values for the enzymes purified from several sources in various laboratories over a period of more than 30 years (Table 5) are in good agreement and the observed subtle variations are in the range expected. However, more precision is expected for measurements recorded for enzymes purified from the same source in a single laboratory at one given time (Table 3a). Substrate saturation curves (Fig. 8a) and the calculated values for K m , V max , k cat , and k cat /K m using the MUG substrate (Table 3a)    , and 50°C (c) at pH 5.5 as monitored by fluorescent enzyme assay. Initial activities ranged from approximately 300 to 1,900 units/mL for all enzymes assayed. % Activity is normalized against activity at t = 0 mins. Data points for (a) and (c) are the mean of a triplicate measurement with error bars equivalent to ± 1 standard deviation. Data points for (b) are the results of a single measurement. MUG was used as the substrate for enzyme assay.     (Table 3a). The V max data presented for the Δ8 (0.742 ± 0.21) and WT (3.36 ± 0.29) are derived from multiple assays from three and two independent enzyme samples, respectively, and this indicates the reliability of this data and adds strength to the interpretations of the data from the single enzyme preparations used for the other deletion mutants. The effects of the C-terminal deletions on the kinetic properties of the enzyme using the artificial substrate MUG (Table 3a) could be due to alterations in the inherent catalytic mechanism of the enzyme [85]. Alternatively, the altered kinetic properties could be due to changes in the affinity of the enzyme for specific structural components of the artificial substrate, MUG. In this context, it is of interest to measure these kinetic parameters with an alternative substrate such as PNPαGal. The results (Table 3b, Fig. 8b) indicate that there are similar changes in kinetic parameters using PNPαGal as the substrate, including increases (2.2-fold effect; 9.18/4.18 = 2.2) and decreases (3.2-fold effect; 4.18/1.31 = 3.2) in the k cat /K m for the specific C-terminal deletion mutants (Δ10 and Δ8, respectively). Taken together, these results suggest that the C-terminal deletions likely affect some aspect of the inherent catalytic mechanism of the enzyme.

Discussion
This is the first report to establish in a quantitative manner that the C-terminal residues of αGal act as a modulator of catalytic activity. Our results confirm the general results of Miyamura et al. [63] that C-terminal deletions of 2, 4, 6 and 10 amino acids increase the k cat /K m compared to WT. However, our results differ in that we find that a deletion of 8 amino acids results in a decrease of k cat /K m . It should be noted that there are numerous experimental differences between these two reports. For example, our analysis used purified enzymes expressed in P. pastoris and their experiments measured αGal enzyme activity during transient infection of COS-1 cells. There could be differences in αGal mRNA or protein stability between P. pastoris and COS-1 cells, and other proteins in the cytoplasm of P. pastoris or COS-1 cells could interact directly or indirectly with the αGal protein to affect its catalytic activity.
It is of interest that the two recombinant protein therapeutics, Fabrazyme (agalsidase-beta) and Replagal (agalsidase-alfa) contain C-terminal heterogeneity with truncated species lacking either one or two C-terminal residues [24]. Fabrazyme contains 69.7% full length protein with 7.6% Δ1 and 22.8% Δ2, while Replagal contains only 5.7% full length, with 73.2% Δ1 and 21.1% Δ2. These authors attribute the C-terminal heterogeneity to in vivo proteolytic processing of an undefined nature. These commercial enzymes do not differ significantly from the WT in K m and V max ( Table 5) in spite of this degree of protein heterogeneity [24]. The significantly increased V max of some of the mutants with C-terminal deletions suggests the basis for an improved treatment for Fabry disease.
These results also illustrate the principle that in vitro mutagenesis can be used to generate αGal derivatives with improved enzyme activity. The potential for improved catalytic activity for this enzyme is illustrated by the existence of closely related enzymes with 3-fold to 250-fold higher activity (Table 6). A direct comparison of relevant amino acid residues between the human and related enzymes suggests the basis for rational in vitro mutagenesis to improve catalytic activity of the WT human enzyme. In this context it is likely that mutants which show altered catalysis against the MUG substrate used here also have a correspondingly higher rate of catalysis against the natural substrate, but this possibility needs to be tested directly.
Clinical trials for ERT show seroconversion frequencies of more than 50% for males treated with 0.2 mg/kg agalsidase-alfa [9] and 88% for 1.0 mg/kg agalsidase-beta [8,86]. Increasing the dose of administered enzyme in seroconverted patients raised antibody titers in some patients [86,87]. IgG antibody status shows a strong association with serious infusion associated adverse reactions [87]. IgG positive serum from Fabry patients exhibits in vitro neutralization of enzyme activity [88] and lessens targeting to key disease organs in a Fabry mouse model [89]. The disease marker of urinary globotriaosylceramide levels is increased in seropositive patients as compared to seronegative patients [86,88]. It has been proposed that higher doses of administered enzyme could overcome the inhibitory effect of antibodies on treatment effectiveness [86]. By using an enzyme that is more active on a per mg basis, a therapeutic effect equivalent to WT is achievable through administration of a lower dose. Future studies should include examination of these mutant enzymes relative to WT in cultured cells [90] and in mouse models [91,92].
The specificity constant (k cat /K m ) has a maximum possible value determined by the frequency at which enzyme and substrate molecules collide in solution [93]. If every collision results in formation of an enzyme-substrate complex, diffusion theory predicts that k cat /K m will attain a value of 10 8 -10 9 M −1 s −1 [93]. The k cat /K m of WT human αGal is approximately 5.49 × 10 3 M −1 s −1 (Table 6) suggesting the possibility that altered forms of the human enzyme may exist that have higher catalytic activity. A BLAST analysis [94] identified the 33 sequences most closely related to αGal and kinetic parameters have been reported for 6 of these enzymes (Table 6). These enzymes share a high degree of sequence and structural similarities and are all in the same family 27 of glycosyl hydrolases [94]. There is a broad range in the values reported (Table 6) for K m , k cat and k cat /K m . Thus, a detailed structural comparison of these enzymes may permit the identification of key amino acid residues that influence these kinetic parameters.
Truncation of the C-terminus of the coffee bean α-galactosidase ( Fig. 9) showed that deletion of one or two amino acids decreases activity and deleting 3 or more residues abolished activity completely [103]. The results with the coffee bean enzyme contrast those presented here for the human enzyme. Both results however demonstrate that the C-terminus of αGal is critical for enzyme function.
We examined a superposition between the crystal structure of the human enzyme and a homology model of the coffee bean enzyme (see Methods) to explore a structural basis for this biochemical dissimilarity. The N-terminal catalytic (α/β) 8 domain superimposes very well (RMSD of 0.729 Å over 1126 total atoms) while the C-terminal antiparallel β-domain superimposes poorly (RMSD of 2.494 Å over 342 total atoms). A primary sequence alignment shows a comparable trend; the catalytic domain shows high sequence conservation while the C-terminal domain does not. An alignment of the C-terminal ends of the human and coffee bean enzymes is presented (Fig. 9) indicating secondary structure alignment (left) and primary sequence alignment (right). If the C-terminal domain governs a conserved mechanism of enzymatic regulation across the human and coffee homologs, then the deleterious effect of removing C-terminal amino acids from the coffee enzyme is consistent with observations made by Miyamura et al. [63] on the human enzyme. Because of the low sequence homology observed in the C-terminal residues it can be hypothesized that the effect on the catalytic activity due to C-terminus deletions in both the human and coffee bean α-galactosidase is due to disruption of the enzyme's three-dimensional structure. This disruption could have an effect on enzyme dimerization, the ability to bind substrate, or potential interactions with other molecules in the cell.
The most straightforward expectation of a series of C-terminal deletions is a direct correlation between the extent of the deletion and the effect on enzyme activity. In this sense, the reduced activity of the Δ8 mutant relative to the other C-terminal deletion mutants (Table 3) is unanticipated. However, we note that similar effects have been reported by others who carried out C-terminal deletion studies, including the IN269 mutation for the integrase of HIV [104], the Δ8 and Δ9 mutants of the thymidine kinase of Epstein-Barr virus [105], and the D5 and D10 mutants of the plant vacullar H (+)-pyrophosphatase [106]. Differential proteolysis may also explain why the Δ8 construct does not follow the same trend as the other mutants.
The crystal structure of αGal [94] revealed that the last visible residue of the C-terminus is separated by approximately 45 Å from the active site on the opposite monomer and is too far to have a direct effect on catalysis. However, within the same crystal structure (PDB 1R47) we measured C-terminal end to be only 20-25 Å away from a second ligand binding site for β-D- The C-termini of human and coffee αgalactosidase. The crystal structure of human αGal and a predicted model of the coffee homolog were superimposed. Underlined terminal residues, (MSLKDLL) in humans and (Q) in the coffee bean enzyme, indicate amino acids that could not be modeled due to conformational disorder. The terminal amino acid of the coffee enzyme (glutamine, Q) aligns with (threonine, T) in the human enzyme and is located 9 amino acids (MQMSLKDLL) from the C-terminus of the human enzyme. galactopyranose [85], demarcated by Tyr329 (Fig. 10). Directly visualizing a putative interaction between the C-terminus and the second binding site is not possible because C-terminal disorder limits resolution to the 7th or 8th amino acid from the full length C-terminal end. If these 7 or 8 amino acids were to adopt a fully-extended conformation, they would span a distance of 22-26 Å (see Methods), bringing the C-terminus within potential contact distance of the secondary binding site. A crucial missing detail is the functional significance of the secondary site. It may serve as a site for small molecule chaperoning [85]. It also might participate allosterically in a manner similar to phosphofructokinase-1, that is allosterically activated by ADP [107,108], a product of the enzyme's ATPase function. The ligand that binds the secondary site on αGal is β-D-galactopyranose [85], which is the mutarotated product of the enzyme's glycoside hydrolase function. A dynamic interplay may exist between the C-terminus, the secondary site and its ligand to affect the catalytic activity of αGal through allostery or structural stability of the protein. Further mechanistic studies will be needed to work out the exact relationships between these putative components and their relevance to in vitro catalysis.
Due to the lack of direct contact between the carboxyl-terminal amino acids and the catalytic site of αGal, the explanation for the effect of the deletions of the carboxyl-terminal amino acids is not obvious. Another hypothesis to be tested is that αGal is in a class of enzymes like the E. coli dihydrofolate reductase [109][110][111][112][113] in which tunneling and coupled motion accounts for the effects of mutations distal from the catalytic site on enzyme function.  [94] is displayed in two different perspectives. Distance relationships relative to one out of the two possible C-termini are discussed. The carbon backbone is rendered in a ribbon format. The C-terminus on monomer A is separated by 20 Å to 25 Å from a secondary binding site for β-D-galactopyranose on the same monomer [85] which is marked by Tyr 329 rendered as spheres. The C-terminus on monomer A is also separated by 45 Å to 50 Å from the active site of monomer B which is marked by the α-D-galactopyranoside ligand also rendered as spheres. Conclusions C-terminal truncation mutants of αGal were constructed, expressed and purified from P. pastoris using Con A and Thio-Gal affinity column chromatography. Michaelis-Menten parameters were measured on the purified mutants. Deletion of 2, 4, 6 and 10 amino acids approximately doubles k cat /K m relative to WT (0.8-1.7-fold effect) while deleting 8 amino acids decreases k cat /K m (7.2-fold effect). Mutants with increased activity are proposed as an improved alternative therapy over WT enzyme for Fabry disease patients. These results also illustrate the principle that in vitro mutagenesis can be used to generate αGal derivatives with improved enzyme activity.