Fabry disease: Identification of 50 novel α-galactosidase A mutations causing the classic phenotype and three-dimensional structural analysis of 29 missense mutations

Fabry disease, an X-linked recessive inborn error of glycosphingolipid catabolism, results from the deficient activity of the lysosomal exoglycohydrolase, α-galactosidase A (EC 3.2.1.22; α-Gal A). The molecular lesions in the α-Gal A gene causing the classic phenotype of Fabry disease in 66 unrelated families were determined. In 49 families, 50 new mutations were identified, including: 29 missense mutations (N34K, T41I, D93V, R112S, L166G, G171D, M187T, S201Y, S201F, D234E, W236R, D264Y, M267R, V269M, G271S, G271V, S276G, Q283P, A285P, A285D, M290I, P293T, Q312H, Q321R, G328V, E338K, A348P, E358A, Q386P); nine nonsense mutations (C56X, E79X, K127X, Y151X, Y173X, L177X, W262X, Q306X, E338X); five splicing defects (IVS4-1G > A, IVS5-2A > G, IVS5 + 3A > G, IVS5 + 4A > G, IVS6-1G > C); four small deletions (18delA, 457delGAC, 567delG, 1096delACCAT); one small insertion (996insC); one 3.1 kilobase Alu-Alu deletion (which included exon 2); and one complex mutation (K374R, 1124delGAG). In 18 families, 17 previously reported mutations were identified, with R112C occurring in two families. In two classically affected families, affected males were identified with two mutations: one with two novel mutations, D264Y and V269M and the other with one novel (Q312H) and one previously reported (A143T) mutation. Transient expression of the individual mutations revealed that D264Y and Q312H were localised in the endoplasmic reticulum and had no detectable or markedly reduced activity, whereas V269M and A143T were localised in lysosomes and had approximately 10 per cent and approximately 35 per cent of expressed wild-type activity, respectively. Structural analyses based on the enzyme's three-dimensional structure predicted the effect of the 29 novel missense mutations on the mutant glycoprotein's structure. Of note, three novel mutations (approximately 10 per cent) were predicted not to significantly alter the glycoprotein's structure; however, they were disease causing. These studies further define the molecular heterogeneity of the α-Gal A mutations in classical Fabry disease, permit precise heterozygote detection and prenatal diagnosis, and provide insights into the structural alterations of the mutant enzymes that cause the classic phenotype.


Introduction
Fabryd isease is an X-linked recessive inborne rror of glycosphingolipid catabolism resulting from the deficienta ctivity of the lysosomal exoglycohydrolase, a -galactosidase A( EC 3.2.1.22; a -Gal A). 1 The enzymatic defect causes the progressive accumulation of globotriaosylceramide (GL-3) and related glycosphingolipids with terminal a -linked galactosyl moieties in the plasma and in tissuel ysosomes throughout the body.I nc lassicallya ffected males whoh avel ittle,i fa ny, with cardiac and/or renal disease and lack the major classical manifestations, including angiokeratoma, acroparesthesias, hypohidrosis and oculara bnormalities. [2][3][4][5] To date,avarietyo fm utations have been identifiedw hich cause the classic phenotype,i ncluding missense,n onsense and splice-sitem utations, as well as partial gene rearrangements, including small and large intragenic deletions and insertions 1,6,7 (the Human Gene MutationD atabase). Most mutations have been private (ie unique to one or af ew families), witht he exception of certain mutations found in unrelated individuals that occurreda tC pG dinucleotides, known hot spots for mutation. 8,9 Of note,non-coding sequence variants have been identifiedi n a -Gal A alleles from normal individuals and patients with Fabryd isease 10,11 (theS NP database), with the exception of D313Y ,a ne xonic sequence variant that encodes an enzyme with 60 -70per cent of wild-type activity,but does not cause disease in males or females with this variant. 12,13 Mutation detection in Fabrydisease is important for several reasons. First, heterozygote detection by enzyme assayo f carriersfor this X-linked recessive disease is unreliable because obligate carriersc an have normal activity due to random X-chromosomal inactivation. 14 -16 Secondly,safe and effective treatment for Fabryd isease by a -Gal Ar eplacement therapy has recently become available worldwide. 17 -20 Because affected maless hould be treated early to prevent the serious complications of thed isease,a nd becausec arriersm ay be asymptomatic or have very mild manifestations, 1,21,22 it is important to identify affected males and carriersf or medical monitoring and early treatment. 23 Thirdly,r ecent in vitro studies 24 have demonstrated that certain a -Gal Am issense mutations result in mutantp roteins that are misfolded and degradedi nt he endoplasmic reticulum (ER). Some of these misfoldedmutant proteins, particularlythose that have residual enzymatic activity ( . 1p er cent), have been rescued by pharmaceutical chaperones, such as galactose and deoxygalactonojirimycin, which arer eversible competitivei nhibitors of a -Gal A. 25 -28 In fact, intravenous administration of galactose (1 g/kg) to a5 6-year-old male with the cardiac variantp henotype resulted in marked improvement of his cardiac manifestations. 29 Thus, it is of interest to predict which of the missense mutations that cause the classic Fabryp henotype might be rescuable based on the enzyme's3 Ds tructure. 30,31 These mutations can then be studied in vitro to determine their rescuability.
In this paper,5 0n ew and 17 previously reported a -Gal A mutations were identifiedi n6 6u nrelated classically affected families.T wo unrelated patients with the classic phenotype had mutant alleles withtwo missense mutations, each of which was expressed in vitro .I na ddition, the structural alterations resulting from the 29 novelmissense mutations were predicted based on the 3D structure of the wild-type enzyme homodimeric glycoprotein.S everal of these mutations resulted from misfolding and arec andidates to determine their rescuability by pharmacological chaperones.

Patient specimens
Peripheral blood wasc ollected from the probands of 66 unrelated families with the classic phenotype of Fabryd isease. The a -Gal Aa ctivity wasd etermined in the plasmaa nd/or lymphocytes as previously described. 32 GenomicD NA was extractedu singt he Puregene isolation kit according to manufacturer'si nstructions( Gentra Systems, Minneapolis, MN,U SA). All specimens were obtained with informed consent and the approval of the Institutional Review Board of the Mount Sinai School of Medicineo fN ew Yo rk University.

Mutation analysis
Mutationa nalysis wasp erformed as previously described. 33 Briefly,e ach of the a -Gal A exons and adjacent flanking and/ or intronic sequences wasa mplified by means of polymerase chain reaction (PCR) from genomic DNA. Each amplicon wast hen analysed by denaturing high-performance liquid chromatography, and abnormally runningf ragments were sequenced usinga nA BI Prism 3700 CapillaryA rray Sequencer with the ABI Prisme BigDye e Te rminator Ready ReactionM ix (Perkin-Elmer-Cetus, Norwalk, CT,U SA). Each mutationw as confirmed by repeat PCR amplification and sequencing of the opposite strand, and/or by co-segregation of the lesion and disease phenotype in other members of each family.I na ddition, 100 normal chromosomes were examined to rule out ap olymorphism for each missense mutation.

Southernb lot analysis
Genomic DNA (10 m g), extracted from normalc ontrols and the proband, wasd igested with 100 units (U) of HindIII and PvuII for 16 hours. After electrophoresis on a1perc ent agarose gel, DNA wast ransferred onto aH ybond-N þ membrane (Amersham GE Healthcare,P iscataway, NJ,U SA) in 0.4MN aOH solution using standard procedures. The filter wasultraviolet-irradiated and then hybridised with arandomly primed 32 P-labelled PCR-generated probe spanninge xons 1 and 2o ft he a -Gal A gene. After overnight hybridisation using PerfectHyb Plus hybridisation buffer (Sigma Aldrich), membranes were washed in diluted standard saline citrate buffers with0.1 per cent sodium dodecyl sulphate.Bands were visualiseda fter exposure and analysed on aM olecular Dynamics STORM 860 phosphoimager (GE Healthcare).

Microsatellite studies
The twop robands with the R112C missense mutationw ere studied to determine if they were related or if the mutations occurred independently.T heir genomic DNAs were haplotyped with the microsatellitem arkersc lose to the a -Gal A locus, including DXS1231,D XS8020, DXS8034, DXS8089, DXS8100, DXS8063 and DXS8096.F orwardp rimersw ere fluorescent dye-labelled (Invitrogen Life Te chnologies

Conservationo fm issense mutations
Each of the missense lesionsw as analysed to determine the relative conservation of the substituted aminoa cid by comparison with four mammalian and 22 non-mammalian eukaryotic a -Gal Ao rthologues and eight a -N-acetylgalactosaminidase ( a -Gal B) orthologues in the GenBank database. These searches were performed using the MacVector program (Oxford Molecular Group). Highly conservedr esidues were defined as those that were present in three of the four (75 per cent)mammalian orthologues in at least 17 (77 per cent) of the eukaryotico rthologues and in six (75 per cent) of the eukaryotic orthologues of the related a -Gal B gene. 34

In vitro expression studies
The full-length wild-type a -Gal A cDNAwas cloned into the pAsc8v ector. 35  All constructs were confirmed by re-sequencing and plasmid preparations were made usingt he Qiagen plasmid midi kit (Valencia, CA, USA). The wild-type and each mutant construct were individually transfected into COS-7 cells and analysed for intracellular a -Gal Aa ctivity and subcellular localisation, as previously described. 13 Structural analysis of a -Gal Am issense mutations The 3.25 A˚X-raystructure of human a -Gal A 30 wasthe basis of the structural analysis. Side-chain positions were compared with an independently derived human a -Gal Am odel based on the chicken a -Gal BX -rays tructure. 36 Mutations were modelleda nd visualisedi nt he program O 37 and energy minimised with the CNS program. 38

Mutation detection
Ta ble 1s ummarises the 50 novela nd 17 previously reported mutations detected in 66 unrelated patients withc lassic Fabryd isease.P CR amplificationo ft he a -Gal A exons and adjacent intronic or flankings equences from genomic DNA, and electrophoresis of the amplicons, did not reveal any gene rearrangements ( . 50 base pairs[ bp]), except one in which exon 2did not amplify.Southernblot analysisindicated alarge deletion of , 3kilobases (kb), which included exon 2.
To determine the precise breakpoints of the deletion,s everal random primer pairsi ni ntrons 1a nd 2w ere used to amplify the proband'sg enomic DNA. Using as ense primer (GCTA-ATGGCAAGACCCTG) located at g2765 in intron 1a nd an antisensep rimer( AAATCCCCCAGTTCTGCTGAGCTA) at g7218 in intron 2, an approximately 4.5 kb PCR fragment wase xpected; however, these primersa mplified a1 .3 kb fragment. Sequencing of this PCRf ragment identifiedt he breakpoints in Alu repetitive sequences at g3260 and at g6410, resulting in an Alu-Alu rearrangement that deleted 3,152bp, including the entire exon 2( Figure 1).
In the remaining 65 unrelated probands, sequencing the respective a -Gal A amplicons detected single mutations in each, witht he exception of twop robands -o ne having two novelm issense alterations (D264Y and V269M) and the other having one novela nd one previously reported mutation (Q312Ha nd A143T,r espectively) in their respective a -Gal A alleles (Table 1) The previously reported lesions included seven missense mutations (A31V,R 112C,Y 134S, A143T,P 259R, G328A and L414S), sixn onsense mutations (Y152X, Q157X, Q221X,W 226X, R227X and R342X) and four small deletions( 26delA, 1031delTC,1 209delAAGa nd 1235delCT). The R112C substitution wasi dentifiedi nt wo probands that were found to be unrelated from microsatellites tudies. Of note,t he R112S,R 112C,Y 152X, R227X and R342X mutations occurreda tC pG dinucleotides, known mutational hot spots. 9 Although previously reported sequence variants were detected in a -Gal A alleles from normal individuals and patients with Fabryd isease 10 -13 (the SNP Database),n o new non-pathological sequence variants were detected in this study.

Expression and subcellular location of the double missense mutations
To determine the functional effects of each substitution in the twoaffected males whose a -Gal A alleles had twomissense lesions, expression studies in COS-7 cells were carried out ( Ta ble 2). The V269M allele had approximately 10 per cent of the mean expressed wild-type activity and the mutant enzyme protein waslocalised immunohistologically to thelysosomes, whereas the D264Y allele had no detectable activity and the mutant protein remained in the ER. The A143T allele had approximately35per cent of expressed wild-type activity and localised to the lysosome, whereas the Q312H allele had approximately5per cent of expressed wild-type activity and was detected predominantly in the ER. Constructs withthe double  T he threonine side-chain points into as mall pocket away from the actives ite.A lthough there is room for the side-chain of the substituted isoleucine,h ydrogen bonding between threonine-41 and histidine-225 would be lost, mostl ikely resulting in protein misfolding. D93V: Aspartate-93 is on the rimo ft he actives ite and provides the negativec harge environment surroundingt he galactosyl residue that will be cleaved. Although it is not one of the activea spartate residues, it interacts with the hydroxyl group at the C6 of the terminalg alactose and orientates the substrate for cleavage.T he substitution of a valine would destroy the interaction with the C6 hydroxyl group and,p resumably,w ould prevent proper binding and orientation of the substrate for cleavage. R112S: Arginine-112 is located on the a 2-helix of the N-terminal b / a -barrel, some distance from the active-site pocket. Its side-chain points toward the disulphide bondbetween cysteine-52 and cysteine-94, possibly stabilisingt he bond. As ubstitution to serine is predicted to destabilise both the disulphide bonda nd the pocket that the arginine occupies,l eading to protein misfolding. L166G: Leucine-166 is on the b 4-strand of the N-terminal b / a -barrel that lines the edge of thea ctive-site pocket. As ubstitutiono faglycine is predicted to enlarge the active-site pocket, probably causing lowers pecificity and efficiency because it would be more difficult to orientate the substrate.G lycine is also morefl exible,a dding to the inability of the activep ocket to stabilise the substrate. G171D: Glycine-171 is located at the end of the b 4-strand along the edge of the actives ite and next to thee nzymatically active aspartate-170.G lycine is as mall and flexible aminoa cid suit-     edge of the b / a -barrel and fills al arge cavity close to the dimer interface where there is ap ocket. This cavity has hydrophobic residues along one side towardst he coreo ft he protein and hydrophilic residues on the side closest to the solvent. An arginines ubstitution with ap ositivec harge on its side-chain could position its charge near the hydrophilic residues, whereas theh ydrophobic middle of the side-chain could interact with the hydrophobic side of the cavity.A rginine would not occupyt he large pocket of the tryptophan, however, leadingtodestabilisation of the enzyme.T ryptophan residues are probably instrumentali ne arly establishment of ap rotein's hydrophobic core during protein folding, so that this substitutionw ould also interferew ithf olding. D264Y: Aspartate-264 is on the b 7-strand of the b / a -barrel and has its side-chain in the actives ite.Achange to tyrosinew ould constrict the actives ite and remove the negativec harge that assists in orientating the substrate.T his substitutionw ould cause either misfolding or markedly impair substrate binding. M267R: Methionine-267 is on al oop between the b 7-strand and a 7-helixt hat aligns the entrance to the actives ite and has its side-chain pointing into the actives ite.Achange to arginine would constrict the actives ite and add an additional positivecharge.Itw ould also interfere with lysine-168, which appearst oh elp to alignt he substrate in the activep ocket. Thus, this substitution would markedly interfere with substrate binding. V269M: Va line-269i so naloop between the b 7-strand and a 7-helixt hat surrounds the entrance to the actives ite,b ut is slightly more removedf romt he actives ite than methionine-267.I ti si nasmall hydrophobic pocket and as ubstitution with methionine would cause somec onstriction of this area leadingt om isfolding and impairment of substrate binding. G271V/S: Glycine-271 is located in aturn between the b 7-strand and a 7-helixi nt he N-terminal b / a barrel. Thep hi/psi angles ared isallowedf or theo ther 19 aminoa cids. Glycine-271 is in ab uried hydrophobic area surrounded by polar residues. As ide-chain of either valine or serine could be accommodated at this position, but the rigidity of this turnr equires ar esidue with the flexibility of glycine. Therefore, these changes would lead to misfolding. S276G: Serine-276 is on the outer edge of the b / a -barrel at the start of the a 7-helix neart he dimer interface.I ts side-chain is involved in extensiveh ydrogen bonding with the backbone and side-chain nitrogens of glutamine-279a nd witht he backbone carbonyl oxygens of phenylalanine-273 and leucine-275. Becauset his hydrogen bonding network stabilises this parto ft he enzyme,achange to glycine would disrupt proper folding. Q283P: Glutamine-283 is partoft he a 7-helix in the N-terminal b / a -barrel. Ap rolinea tt his position would disrupt proper folding of the enzyme. A285P/D: Alanine-285 is parto ft he a 7-helix in the b / a -barrel and lies buried in the interface between the N-and C-terminald omains. The prolines ubstitution would disrupt the helix, interfering with the proper folding of the enzyme.T he helix also is beside the C-terminal domain, and this interaction would be disturbed.
Substitution of an aspartate residue,w hich is negatively charged, would be extremely unstable and would disrupt folding because therea re no surroundingr esidues to counter this charge. M290I: Methionine-290i sa tt he end of the a 7-helix, where it occupies al arge hydrophobic pocket. An isoleucine at this position would not occupyt he same volume and would destabilise this section of the enzyme,t hus disruptingp roper folding of the enzyme. P293T: Proline-293 occursj ust before the b 8-strand of the N-terminal b / a barrel and is buried in acentral portion of the enzyme some distance from the actives ite.P rolines often provide rigidity to the protein that promotes proper folding. Athreonine substitution at this positions hould be tolerated, but there will probably be acosttothe folding dynamics of the protein. Any protein that does forms hould have enzymatic activity,b ut there mayb e little,i fa ny,s table enzyme. Q312H: The last a -helix ( a 8) of the N-terminal b / a barrel is actually formed from twohelices separated by twor esidues (glutamine-312 and aspartate-313) that aren ot in the more extended conformation.T he glutamine-312side-chain is exposed to solvent on one side and the side-chain of tryptophan-81 on the other.T he structure should be able to accommodate the substituted histidine, whose side-chain could even makeamore favourable interaction with the side-chain of tryptophan-81. This mutationi s also distant from the actives ite. Structurally,i ti sd ifficult to determine whyt his mutation would be deleterious; however, the pKa of histidine suggests that it would be protonated in the lysosomea nd thus have ap ositivec harge,w hich could interfere with theo rganisation in this area of the protein.
Q321R: Glutamine-321o ccursi nt he last a -helix ( a 8) of the N-terminal b / a -barrel. Its side-chain is mostly exposed, but it hydrogen bonds with the side-chain of threonine-39. An arginine substitution would add ap ositivec harge to this area, and its longer side-chain would prevent the interaction with threonine-39, possibly destabilising this area and preventing proper folding. G328V: Glycine-328 is located in the loop between the a 8-helixo ft he N-terminald omain and the b 9-strand of the C-terminald omain. Glycine is ideal for fitting into tight areas of aprotein's structurebecause it has no side -chain, and this is the case for glycine-328.W hen changed to valine, there is no room for the bulkier side-chain -and the substitution would cause somed isruption in the wayt hat the N-and C-termini were packed together.M ost likely,t he enzyme would not fold properly and would be degraded. E338K: Glutamate-338 is located in the b 10-strand of the C-terminal domainand is hydrogen bonded with tryptophan-340 and arginine-356. Arginine-356 makesastrong salt link with aspartate-244 that helps to stabilise the interaction between the N-and C-terminal domains. The substitutiono f lysine would competew ith arginine-356 and destabilise the enzyme'ss tructure,l eading to misfolding. A348P: Alanine-348 is at the start of the b 11-strand in the C-terminal domain. Ap roline substitution would be difficult to accommodate and would disrupt the preferred secondarys tructure.B ecause it occursr ight after al oop,h owever, one would predict that a small amount of protein would fold properly and be functional. E358A:Glutamate-358 is in aloop after the b 11-strand of the C-terminald omain; it is completely solvent-exposed and an alanine substitution could easily fit into this position. Glutamate-358 does hydrogen bondt ot ryptophan-236 and lysine-240, however, stabilising this loop region,which is near the dimer interface.A lanine would not support these interactions and lead to misfolding. Q386P :Glutamine-386 is located in the b 13-strandofthe C-terminal domain. Aproline replacement does not contain an amide hydrogen, so it will not maintain the hydrogen bonding of the b -sheets. This mutation would interferew itht he proper folding of the C-terminal domain.

Discussion
Mutation analysis of the a -Gal A gene in 66 unrelated probands with Fabryd isease identified5 0n ew mutations, demonstrating the extensivem olecular genetic heterogeneity underlying this lysosomal storage disease.O ft he new mutations, several were notable,i ncluding a3 .1 kb deletion (only the fifthl arge deletion detected in this 'Alu-rich' gene [Human Gene Mutation Database]), an allele with twoa djacent base substitutions (L166G)and twoalleles,each with two missense mutations (D264Y/V269M, and Q312H/A143T).
Although all types of mutations have been found to cause Fabrydisease, 1 there have been relatively few large gene deletions, considering the fact that the a -Gal A geneisan Alu-rich gene with about one Alu per kb.Previously,only four deletions over 1kbwere reported amongthe over 400 mutations causing this disease,and only one of these four wasdue to Alu -Alu recombination. 1 Thelarge deletion reported here most probably resulted from unequal,but homologous, recombination between the highly homologousand similarly orientated Alu sequences in introns 1and 2, therebyresulting in the approximately 3.1 kb loss, including all of exon 2 (Fig. 1).
The L166G mutation is unusual,inthat it involves twobase changes, ad ouble transversion of CT to GG at cDNA positions 496 and 497. As ingle event most likely resulted in the other complex mutation that gave the missense mutation K374R,l ocated 3bpu pstream of aG AG deletion.
The D264Yand V269Mmutations were found in aclassically affected male.Both mutations were located in exon 6. Structural analysispredicted that V269M would constrict the activesite, but that the protein would fold properly and retain some, albeit reduced, activity.Bycontrast, the aspartate at position 264 lines the active-site pocket, and the change to tyrosine predicts a marked constriction of the activesite.Also,the negativecharge that probably assists in orientating the substrate would be lost, markedly altering folding and enzyme activity.These structural predictions were confirmed by in vitro expression assays which demonstrated that the V269M allele had about 10 per cent of the mean wild-type expressed activity,which wasdetectable in the lysosomes (Figure 2), whereas the D264Y allele expressed no detectable enzyme activity,and the enzyme protein wasdetected immunologically in the ER.
Most mutations causing Fabrydisease areprivate,occurring in as ingle or few families; 1,7 however, several mutations at CpG dinucleotides, known mutational hot spots, occur more often in unrelated families. These includeR 112S,R 112C, A143T,Y 152X, R227X and R342X in unrelated classically affected families. Here,the R112S mutationisfirstreported in ac lassically affected male. In addition, other residues are encoded at CpG dinucleotides, including T39, R49, R118, C142, D153, R220, R301, D315, V316, R356, R363, I367 and A368. Mutations have been reported at all of these CpG sites, except at codons 39, 118, 315, 316, 367 and 368. Four of the mutations reported here (8 per cent)were detected only in the affected proband and were not present in either of the probands' parents.T hree of these de novo mutations occurred at CpG sites(R112C,Y152X and Q157X). One other de novo mutation occurred at S201F.
In this paper,wehavemapped the novelmissense mutations onto the native, properly folded enzyme to better understand their 3D locations and proximityt ot he actives ite. It is easy to understand whym utations at residues aligning the active-site pocket, such as aspartate-93, aspartate-264 or methionine-267,p erturb enzyme function becauset hey are critical in binding and orientating the substrate for hydrolysis. 30 Other missense mutations interfere with the proper folding of the enzyme,l eading to retention in the ER. The enzyme exists as ah omodimer,b ut therei sn oi ndication of functional cooperativity between the twos ubunits. Tw oo f the mutations occur near the dimer interface (aspartate-234 and tryptophan-236), although they do not interactw ith the other subunit. The structural analyses predict that three of the substitutions (M187T,Q312H and A348P) would be tolerated or have residual activity.T hese mutations all resulted in the classic phenotype,h owever,a nd only four mutations had significantr esidual activity in the plasma or leukocytes. The structural studies are based on the properly foldedenzyme and cannot provide all of theinformation needed to determine the folding mechanism of this complex glycoprotein. When the structural analyses arec oupledw itho ther in vitro expression studies, however, they can give ac learer insight into genotype-phenotype correlations. The mutations that are predicted to be accommodated by structural analysis might be functional if allowedt of old and arec andidates for studies with pharmacological chaperones.
In summary, theses tudies identify 50 additional mutations causing Fabryd isease,b ringing the total number of reported mutations to over 450. Them utations continue to be found in individual families and only af ew have been found in other unrelated families -m osto ft heseo ccurring at CpG dinucleotides, which arem utational hot spots.T he studies reported here predict most of the structural alterations resulting from the amino acid substitutions. Most disrupt the natives tructure and cause protein misfolding. Those located around the actives ite interfere with orientation and/or binding with the substrate,therebyaltering enzyme function. Afew located on or near the dimer interface alter proper dimerisation of the glycoprotein. None of the mutations described here replaced anyo ft he three asparagines at N-glycosylation sites, but N34K presumably could affectg lycosylation at N192,w here asparagines-34 interactsw ith the side-chains of asparagines-192. Several of the normally substituted amino acids were predicted to be structurally accommodated and thereforet olerated (ie T41I, M187T,G 271V,G 271S,P 293T, Q312H and A348P); however, these replacements all caused severe loss of enzyme function or stability becauset hey resulted in the classic phenotype.T hus, predicting the phenotype based on the structural alterations of the mutations mayu nderestimate the severity of the substitution on the protein's ability to fold into afunctional configuration. Clearly, most of the novelm issense mutations described here altered folding and presumably ledt ot he glycopolypeptide's aggregation/retention in the ER and subsequent proteosomal degradation in the cytosol, consistent witht he classic phenotype of Fabryd isease.