AU7761298A

AU7761298A - Carbohydrate-deficient glycoprotein syndrome type i

Info

Publication number: AU7761298A
Application number: AU77612/98A
Authority: AU
Inventors: Gert Matthijs
Original assignee: Genzyme Ltd
Current assignee: Euroapi UK Ltd
Priority date: 1997-04-30
Filing date: 1998-04-30
Publication date: 1998-11-24
Also published as: WO1998049324A2; EP0981632A2; WO1998049324A3; CA2289109A1

Description

WO 98/49324 PCT/EP98/02593 -1 "Carbohydrate-deficient glycoprotein syndrome type I" This invention relates to carbohydrate-deficient glycoprotein syndrome Type I (CDG1), or Jaeken disease; more particularly, it relates to the identification of the molecular defect associated therewith, to a novel gene and to diagnostic and other applications of these findings. CDG 1 is the prototype of a class of genetic multi-system disorders characterised by defective glycosylation of glycoconjugates, (see Jaeken, J. et al, Pediatr. Res., 14, 179, 1980; Jaeken, J., & Carchon, H., J. Inher. Metab. Dis., 16, 813-820, 1993; and Jaeken, J., metal, J. Med. Genet., 34, 73-76, 1997). It is mostly a severe disorder which presents neonatally. There is a severe encephalopathy with axial hypotonia, abnormal eye movements and pronounced psychomotor retardation, as well as a peripheral neuropathy, cerebellar hypoplasia and retinitis pigmentosa. The patients show a peculiar distribution of subcutaneous fat, nipple retraction and hypogonadism. There is a 20% lethality in the first years of life due to severe infections, liver insufficiency or cardiomyopathy, (see Jaeken & Carchon, 1993, ig gil; Jaeken, et al, 1997, loc cit; and Stibler, H., et al, J. Neurol. Neurosurg. Psychiatry, 57, 552-556, 1994). CDG1 is inherited in an autosomal recessive manner and its locus has been mapped to chromosome 16p, (see Martinsson, T., et al, Hum. Mol. Genet., 3, 2037-2042, 1994; and Matthijs, G., et al, Genomics, 35, 597-599, 1996). Most patients show a deficiency of phosphomannomutase (PMM), (see Van Schaftingen, E., & Jaeken, J., FEBS Lett., 377, 318-320, 1995), an enzyme necessary for the synthesis of GDP-mannose. This gene (to be termed PMM1 in future), which is on chromosome 22q13, (see Matthijs, G., et al, Genomics, 40, 41-47, 1997), has previously been cloned. The basis of the present WO 98/49324 PCT/EP98/02593 -2 invention is the surprising identification of a second human PMM gene, PMM2, which is located on chromosome l6p13 and which encodes a protein having 66% identity to PMM1. Eleven different missense mutations in PMM2 were found in sixteen CDG1 patients of different geographical origin and having a documented phosphomannomutase deficiency. The 5 present results give conclusive support to the biochemical finding that phosphomannomutase deficiency is the basis for CDG l. The first evidence for a glycosylation defect in CDGl was reported in 1984 by the demonstration of a sialic acid deficiency of serum transferrin, (see Jaeken, J., et al, Clin. 0 Chem. Acta.. 144, 245-247, 1984). Japanese investigators subsequently provided evidence for a very early defect in the glycosylation pathway, (see Wada, Y., et al, Biochem. Biophys. Res. Commun., 189, 832-836, 1992; and Yamashita, K., et al, J. Biol. Chem., 268, 5783-5789, 1993), and, in 1995, Van Schaftingen and Jaeken (loc cit) identified a deficiency of phosphomannornutase activity. PMM is a cytoplasmic enzyme that isomerizes 5 mannose 6-phosphate into mannose I-phosphate, which is then converted to GDP-mannose (EC 5.4.2.8). The GDP-mannose is required for the synthesis of dolichol-P oligosaccharides. The phosphomannornutase deficiency has now been found in more than fifty CDG I patients of different geographical origin. Moreover, intermediate activities were found in parents of CDG I patients, (see Van Schaftingen & Jaeken, 1995, loc cit), which 0 strongly supports the hypothesis that the gene encoding phosphomannomutase is a major candidate gene for CDG 1. A search through Genbank had previously revealed the availability of several human cDNA's having similarity to Candida or yeast phosphomannomutase and led to the isolation of PMM1 WO 98/49324 PCT/EP98/02593 -3 (see Matthijs, e al, 1997, log it). However, PMMI could be unambiguously assigned to chromosome 22q 13, excluding that this gene could harbour the primary defect in the majority of CDGI patients (see Matthijs, et al, 1997, loc cit). Recent biochemical investigations on PMMI and phosphomannomutases from rat and human liver provided evidence for the 5 existence in mammalian systems of a second phosphomannomutase having different kinetic and antigenic properties. What may be regarded as the starting-point for the present invention was the identification in the public database of a more recently released cDNA sequence having a high similarity to PMMI (cDNA clone 364509 from the Soares foetal heart library, see Lennon, G., et al, Genomics, 33, 151-152, 1996) and the isolation of a 0 corresponding 2.3 kb cDNA from a library from Mo cells (see accompanying Fig. 1-la). This clone probably represents the full-length mRNA or PMM2. The putative open reading frame is 738 bp long and predicts a protein of 246 amino-acids. Upon alignment with PMM1, the identity in the coding region is 65% at the nucleotide level 5 and 66% at the amino acid level (see accompanying Fig. l-lb). With yeast phosphomannomutase (SEC53, see Kepes, F., & Schekmon, R., J. Biol. Chem., 263, 9155 9161, 1988), the identity is 57% at both levels. Thus, the genes encoding phosphomannomutase are extremely conserved from yeast to man. A slightly higher degree of identity of PMM2 with SEC53, as compared to PMMI, suggests that PMM2 remained .0 closer to the ancestral gene (see accompanying Fig. l-lb). Two EcoRI fragments of respectively )12 kb and 0.8 kb were identified on a Southern blot with human total genomic DNA using a probe derived from the 5' end of the cDNA. The larger band is derived from chromosome 16 and the 0.8 kb fragment comes from a processed WO 98/49324 PCT/EP98/02593 -4 pseudogene on chromosome 18. The cDNA probe did not cross-hybridize to the PMM1 gene on chromosome 22. Several BAC clones were isolated using the same probe and further characterised by Southern blot analysis: clones 41M16, 27D21 and 342N15 contained the )12 kb EcoRI fragment. They were mapped to chromosome 1 6 p in FISH experiments 5 on human metaphase chromosomes (see accompanying Fig. 1-2a). Taking advantage of the phosphomannomutase activity measurements to detect carriers in CDGl families, the candidate region has now been refined from 13 cM to less than I cM, between D16S406 and D16S404. Evidence that PMM2 maps to this region was obtained by hybridizing the same cDNA probe as above to DNA from the YACs in a contig that spans this region (see 0 Doggett, N.A., et al, Nature, 377, Supplement, 335-365, 1995). Only YAC 909F5 gave a positive hybridization signal on Southern blot. This YAC is also positive for D16S406, suggesting that the gene is located in the vicinity of this marker. This is consistent with the genetic data, (see Martinsson, et al, 1994, loc cit; and Matthijs, et al, 1996, log cit). 5 The genomic structure of the PMM2 gene has now been determined (see accompanying Fig. 1-2b). Eight exons were identified, spanning approximately 20 kb of genomic DNA. The genomic structure of the PMMl and PMM2 genes is very similar, as all splice sites are conserved. 0 Northern blot analysis showed highest expression of the PMM2 gene in pancreas and liver, two organs having an important production of secreted proteins (see accompanying Fig. 1 2c). Most other tissues also express the gene, with only minor differences between PMMI and PMM2. PMM2 is weakly expressed in brain, as opposed to PMMI (see accompanying Fig. 1-2c). It is not clear at present how these levels of expression would correlate to the WO 98/49324 PCT/EP98/02593 -5 clinical picture, in which brain is one of the most severely affected organs. For mutation analysis by single-strand conformation polymorphism (SSCP), the cDNA sequence has been arbitrarily divided into four fragments. cDNA was available from thirty 5 three unrelated patients having a confirmed phosphomannomutase deficiency, (see Van Schaftingen & Jaeken, 1995, loc il), originating from eleven different countries. Examples of the SSCP results for fragments 2 and 4 are shown (see accompanying Fig. 1-3a). In total, eleven different missense mutations have been identified. No Linkage PMM Values Mutation 1 Mutation 2 Origin Nucl. AA Nucl. AA 3 Yes 0.33 (Ly) 425G/A R141H 647A/T N2161 Sicily 4 Yes 2.75 (Ly) 385G/A V129M 425G/A R141H Sicily 5 Yes 0.15 (F) 357C/A FI19L 425G/A R14lH Netherlands 9 Yes 0.05 (Le) 391C/G P131A 425G/A R141H France 5 12 Yes 0.0 (F) 425G/A R141H 710C/T T237M Spain 15 Yes 0.10 (F) 425G/A Rl41H 691G/A V231M Belgium 16 NI 0.13 (Le) 317A/G Y106C 425G/A R141H Belgium 17 Yes 0.10 (F) 338C/T PIl3L 425G/A R141H Belgium 27 Yes 0.0 (F) 357C/A F119L 425G/A R141H USA 0 30 NI 0.09 (Le) 0.50 (F) 323C/T A108V 425G/A R141H France 31 NI 0.17 (F) 357C/A Fl19L 425G/A R141H Germany 35 NI 0.13 (F) 357C/A Fl19L 425G/A R141H U.K. 38 Yes 0.0 (F) 338C/T Pl13L 425G/A R141H Germany 40 NI 0.19 (F) 338C/T PIl13L 425G/A R141H France 5 41 NI 0.19 (F) 357C/A FI19L 425G/A R141H Netherlands 42 NI 0.43 (F) 425G/A R141H [484C/T R162W UK WO 98/49324 PCT/EP98/02593 -6 Table 1-1 Prevalence of the different mutations in the PMiM2 gene identified in CDG1 patients from different geographical origins 5 (1) "Yes" denotes that segregation of the defect within the family is compatible with a genetic defect on chromosome 16p13. NI: not informative (mostly single cases without siblings). (2) Enzymatic activities of phosphomannomutase are expressed as a % of controls. 0 Control activities were (means - SD) 2.5 t 0.5 mU per mg protein (n= 12) in leukocytes (Le), 9.5 ± 1.3 (n=6) in lymphoblasts (Ly) and 3.8 * 0.9 (n=8) in fibroblasts (F). (3) Nucleotides and corresponding amino acid changes. Only those patients in which 5 both mutations were identified have been listed. In Table 1-1 above, there are included data on sixteen patients in which both mutations were identified. Eight of the mutations affect residues that are strictly conserved among 0 PMM's and at least seven of them (Tyrl06-Cys; Prol l3-Leu; Phel19-Leu; Prol31-Ala; Argl4l-His; Asn216-Ile, Thr237-Met) result in the replacement by a residue of markedly different size or hydrophobicity as compared to the normal enzyme (see accompanying Fig. 1-1b). Val231-Met results in the replacement of a hydrophobic residue by a bulkier one, as do two other mutations (Alal08-Val and Val129-Met) which concern semi-conserved 15 positions. In the last case (Argl62-Trp), a basic residue is replaced by tryptophan. It is likely that these mutations result in proteins having lower enzymatic activity and/or stability. Some of the mutations (R141H and Fl19L) are more frequent than others found, 30 respectively, on 16 and 5 out of 32 CDG1 chromosomes) From the remaining seventeen patients, seven were heterozygous for one of the PMM2 mutations reported in Table 1-1. No mutation was found in ten patients using this approach. These observations suggest that either a frequent mutation was missed or that other mutations are located outside the WO 98/49324 PCT/EP98/02593 -7 regions which have been analysed. Since mutation analysis was done by SSCP starting from cDNA, splice mutations, mutations leading to unstable transcripts, mutations in the regulatory regions or major deletions would not be identified. The availability of the flanking intronic sequences will allow further analysis of the gene. 5 The mutations detected in exons 4 and 5 were confirmed on genomic DNA by SSCP analysis after PCR amplification using intronic primers. This approach was used to follow the parental origin of the mutations in the informative families. For instance, in family 38, PI13L is of maternal origin and R14lH is of paternal origin. The unaffected sib is a 0 carrier of the PI 13L mutation, which fits with the linkage data (see accompanying Fig. 1 3b). It is expected to find more mutations in the cohort of thirty-three patients having a documented phosphomannomutase deficiency. However, it should be recalled that genetic 5 heterogeneity exists for CDGI, (see Matthijs, et al, 1996, loc it), and the syndrome may require a new sub-classification on the basis of the phosphomannonutase results. It is proposed to classify the cases having a phosphomannomutase deficiency as CDG Type 1A (CDGlA). Z0 It remains to be investigated which consequences of the PMM deficiency are most detrimental to the cell and the organs. The mere accumulation of abnormally processed proteins might affect the function of the endoplasmic reticulum. Abnormal function as a result of deficient glycosylation has been shown for a large number of circulating proteins, including transport proteins, enzymes and complement factors, (see Jaeken, et al, 1997, loc WO 98/49324 PCT/EP98/02593 -8 git), but the function of many other (membranous and intracellular) glycoproteins is most probably also impaired. Moreover, GDP-mannose and dolichol-P-mannose are essential for the biosynthesis of GPI anchors and thus membrane-bound enzymatic and receptor functions, cell-to-cell signalling and cell-adhesion might also be affected. The identification 5 of the molecular basis for the disease will allow the further unravelling of the pathogenesis of CDG1 and related syndromes. Referring to the accompanying illustrative Figures: Figure 1-1 - Sequence of the PMM2 cDNA and alignment of the predicted PMM2 protein 0 with known phosphomannomutases. (a) Sequence of a cDNA clone, obtained after screening a library from a human leukemia T-cell line. Uppercase letters represent coding sequence, lowercase letters are for the untranslated sequences. The deduced amino acid sequence is shown below the nucleic acid sequence. The exon boundaries are indicated by vertical lines. The sizes of 5 the coding exons range from 66 to 116 bp. All exon/intron boundaries have been sequenced and show consensus splice sites. In the 3'UTR., the polyadenylation signal (ATTAAA) is underlined. (The cDNA sequence has Genbank accession No. U85773.) (b) Comparison of the amino acid sequences of human phosphomannomutases (PMM1, PMM2, translated from cDNA sequence), Saccharomyces cerevisiae SEC53 (sacc) and 0 Candida albicans phosphomannomutase PMMCANAL (cana) (sequences obtained from Genbank). Identical or closely related amino acid residues are boxed. The predicted PMM2 protein is shorter at its amino- and carboxy-terminal ends, and has an internal deletion of two aminoacids (Gly-Asp) after codon 56, when compared to PMMl. This Gly-Asp is also absent in yeast and Candida albicans PMM (see Matthijs, t al, 1997, lc WO 98/49324 PCT/EP98/02593 -9 cit). The positions of the mutations identified in this study have been indicated. Figure 1-2 - Localization, genomic organisation and expression of PMM2. (a) High resolution mapping of human PMM2 by FISH. The BAC clone 41M16 was used as a biotin-labelled probe. The chromosomes 16 are indicated by arrows. Similar results were obtained with BACs 27D21 and 342N15. The G/Q banding pattern was generated from the DAPI counterstaining. (b) The HindIII (H) map of the region has been constructed by (partial) digestions of cosmids 428D1, 408C7, 422F4 and 404H6, isolated from a chromosome 16-specific ) arrayed cosmid library (see Longmire, J.L., et al, Genetic Analysis : Techniques and Applications, 10, 69-76, 1993). The 3' end of the gene is not present in cosmids and has been characterised from the overlapping BAC clones. The position of an intragenic EcoRI site (E) is indicated, together with the location of the eight exons that divide the coding region of the gene. The approximate size of the introns was obtained from estimations of 5 the length of restriction fragments and PCR products. Intron 5 is the smallest intron at 0.5 kb, while intron 4 spans 4.5 kb. The size of intron 7 could not be determined with certainty. (c) Northern blot analysis of human tissues. A commercial multiple tissue Northern blot with approximately 2 pg of poly A* RNA from different tissues was hybridized with D the PMM2 (upper panel), PMMI (middle panel) and a /-actin probe (lower panel). A commercial Northern blot, containing poly A' RNA from different regions of the brain, showed weak signals after prolonged exposure, but did not reveal any regional difference in expression.

WO 98/49324 PCT/EP98/02593 -10 Figure 1-3 - Representative SSCP patterns for mutations identified in CDG 1 patients. (a) SSCP analysis of cDNA fragments. Fragments 2 (lanes 1-4) and 4 (lanes 5-8) were amplified from cDNA. The underlying mutations have been identified by sequencing. In all cases, only one mutation is present in the fragment under investigation. The nucleotide changes and the corresponding amino acid substitutions are shown. (b) Inheritance pattern of the PMM2 mutations in a representative family. PCR amplification of exon 4 (upper panel) and exon 5 (lower panel) on genomic DNA from the patient, parents and sib, was followed by SSCP analysis. The bands associated with the mutation are indicated by an arrowhead. The bars represent the inheritance of the disease chromosome (black), the normal paternal (white) or the normal maternal chromosome (hatched), as deduced from the results obtained with the polymorphic markers D16S406, D16S404, D16S407, D16S414, D16S497, D16S519 and D16S500. These are ordered from telomere to centromere, top to bottom, and the PMM2 gene is located between D16S406 and D16S404. The PIl3L mutation is of maternal origin in this family. The R141H mutation is inherited from the father. The unaffected sib also inherited the maternal disease allele, as shown with polymorphic markers that flank the gene, and is a carrier of the Pll3L mutation. By way of further exemplification of the present invention: Isolation of a full-length cDNA A probe, specific for human PMM2, was generated by PCR from cDNA derived from a human epidermoid carcinoma cell line (BB49; mouth tumour; kindly provided by F. Brasseur, Brussels) and its identity confirmed by sequencing. The primers 5'- WO 98/49324 PCT/EP98/02593 -11 CCCAGCGCTCTGCCTCTTCGA-3'/5'-ACGTITAACATCCCATTCGG-3' (nucleotides 15 to 383) were derived from the partial cDNA sequence present in the IMAGE clone 364509, available through Genbank (Accession Nos. AA022583 and AA022584). This cDNA clone was also kindly provided by UK HGMP Resource Centre. 5 The cDNA library screened to obtain the full-length cDNA was provided by J.C. Renauld (Brussels); it is a human T-cell leukemia (Mo cells-ATCC CRL8066 (see Chen, I.S.Y., et al, Nature, 305, 502-505, 1983)) library inserted between the NotI and BstXI sites of pcDNAI/Amp (Invitrogen). Approximately 180,000 colonies were transferred on 10 GeneScreen* filters (Amersham) and hybridized overnight at 65'C in I M NaCl, 1% SDS, 10% dextran sulfate, 20 pg/ml salmon sperm DNA with the cDNA probe (EcoR1-Noti fragment from clone 364509) labelled with [a- 3 P]dCTP by random priming. The filters washed twice for 5 minutes at room temperature in 2 x SSC and twice for 30 minutes at 65'C in 2 x SSC containing 1 % SDS. Positive clones were submitted to a secondary 15 screening. Two types of clones were obtained with inserts of approximately 2.1 and 2.3 kb. The latter was sequenced with universal and specific primers, and the sequence is shown in accompanying Fig. I-la. IMAGE clone 364509 is a partial cDNA clone, derived from the 5' end of the PMM2 mRNA, down to codon 143. Most probably, the downstream adenine-rich region (nucleotide positions 491 to 515) has allowed spurious 20 oligo-dT priming during cDNA synthesis. Mapping of the gene to chromosome 16 The probe, generated by PCR (see above), was used for Southern blot analysis of a genornic mapping'panel. 5 pg of each cell-line from the Coriell Mapping panel 2 (NIGMS, WO 98/49324 PCT/EP98/02593 -12 Camden, NJ) were digested with EcoRI. Filters were hybridized overnight in 50% formamide and 5 x SSPE, 10 x Denhardt's solution, 2% SDS, and 100 pig/ml heparin, and washed at high stringency in 0.1 x SSPE, 0.1% SDS at 65 0 C for 30 minutes. Contrarily to what is stated by NIGMS, the human chromosome 14 cell line GM10478 also contains 5 material derived from human chromosome 16 (p13. 1-q22. 1). The same probe was used on a membrane containing DNA from YACs that have previously been assigned to the CDG1 candidate region (D16S406 tp D16S407), (see Doggett, et al, 1995, lc git). 0 BAC and cosmid clones High density human BAC colony DNA membranes (#96055, Research Genetics) were screened with the PMM2 cDNA-probe, and the positive clones were purchased from Research Genetics. Chromosome 16-specific cosmids were identified on an arrayed cosmid 5 library, provided by N. Doggett (LANL, Los Alamos). All clones were further characterised by restriction digestion with EcoRI and HindIII, followed by Southern hybridizations, as described. The genomic structure of the PMM2 gene was determined by Southern hybridizations of the cosmid clones the oligonucleotide probes derived from the cDNA sequence. These oligonucleotides were typically 20 bp in length and were 5' !0 end labelled using [y- 32 P]ATP and T4-PNK and hybridizations were done in 6 x SSPE, 5 x Denhardt's, 0.5% SDS and 200 ptg/ml heparin. Filters were washed in 2 x SSPE, 0.1% SDS at 42'C for 5 minutes. A detailed HindIII map of cosmids 428D1, 408C7, 422F4 and 404H6 was generated by partial digestions. 4 pg of purified cosmid DNA was digested to completion with NotI (20 WO 98/49324 PCT/EP98/02593 -13 U) in 20 pl and further digested with HindIllI (5 U) in 100 pl. Aliquots of 25 pl were taken at 2, 5, 10 and 60 minutes and the reaction was stopped by adding 1 pl 0.1 M EDTA. Samples were analysed by field-inversion gel electrophoresis (FIGE) and blotted. Hybridizations were performed with 3 2 P-labelled T3 and T7 oligonucleotides under standard conditions. Gene-specific primers have also been used for genomic PCR on cosmids, BACs and total human DNA. Standard PCR conditions were (50 pl) 50 ng DNA, 200 LM dNTP, 25 pmol each primer, 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl 2 , 0.01% gelatin and I U AmpliTaq polymerase (Perkin Elmer). The Long Expand PCR kit was used, as prescribed by the manufacturer (Boehringer Mannheim). Fluorescently-labelled primers (FITC, fluorescein-isothiocyanate) were used for cycle sequencing on cosmid DNA, using Amersham's Thermo-sequenase kit and 2 to 6 pg of DNA. Fluorescence in situ hybridizations (FISH) DNA from the different BAC clones, purified after alkaline lysis of the bacterial hosts, was labelled with biotin by nick-labelling, and used for FISH on human metaphase chromosomes, as previous described, (see Matthijs, et al, 1997, loc cit). D Northern blot analysis The probe for PMM2 and a cDNA probe derived from the previously isolated PMM1 gene were labelled by incorporation of [a_- 2 P]dCTP by random primed labelling and hybridized overnight to a commercial multiple tissue Northern blot (Human MTN blot #7760-1, Clontech) and a human brain Northern blot (Human Brain II #7755-1, Clontech) as WO 98/49324 PCT/EP98/02593 -14 described. Blots were washed at high stringency (20 minutes at 65'C in 0. 1 x SSPE, 0.1% SDS). A -actin probe (Clontech) was used as a control. Single-strand conformation polymorphism (SSCP) analysis and sequencing 5 Total RNA was isolated from cultured fibroblasts and lymphoblasts from patients using the phenol-chloroform procedure (Trizol, Life Technologies). Starting from 5 pg total RNA, cDNA was synthesised with MMLV-RT and oligo-dT primers, as prescribed (MMLV-RT, Life Technologies). Typically, 2 pl of this product was used for PCR-amplification. Based on the available sequence information, primers were designed for the RT-PCR amplification of four cDNA fragments, suitable for SSCP analysis. The following primers were used: for fragment 1, PMM16-IAF (5'-FITC-GTGCGGCTAGAAACTGGGAC-3'; -20 to -1) and PMM16-lB (5'-TCGTATTTTTCAACCACATC-3'; 184 to 194); for fragment 2, PMM16-2AF (5'-FITC-TGATGTGGTTGAAAAATACG-3'; 185 to 195) and PMM16-2B (5'-TCAATGCGTTCTTCTTGGCTGC-3'; 410 to 431); for fragment 3, 5 PMM16-3AF (5'-FITC-GGTACTTTCATTGAATTCC-3'; 349 to 368) and PMM16-3B(5' TCTCTTGTCCCATCCATCAG-3'; 551 to 570); for fragment 4, PMM16-4AF (5'-FITC GAGATACTGTCTGCGACATG-3'; 567 to 586) and PMM16-4B (5' GGAAGTCCAGACTGCACATG-3'; 870 to 889). Intronic primers were used for SSCP analysis of exons 4 and 5: for exon 4, PMM16-int3F(5'-CTGGGTTTGCTATGAAGCTG 0 3') and PMM16-int4RF (5'FITC- ACCATGTGACACTACGCTATG-3') and for exon 5, PMM16-int5R (5'-GTGTTGGGATTACAGGCATG-3') and the exonic primer PMM16 3AF (see above). 10 to 15 y1 of the PCR products were mixed with an equal volume of formamide, denatured for 5 minutes at 95'C, loaded onto a non-denaturing polyacrylamide gel (0.5 x WO 98/49324 PCT/EP98/02593 -15 Hyrolink MDE, J.T. Baker, in 0.6 x TBE (10 x TBE is IM Tris, 0.82 M boric acid, 10 mM EDTA)) and electrophoresed at 4'C for 10 hours at 400V. The gels were directly scanned on a Fluorlinager (Vistra) and the signals were analysed with the ImagequaNT software (Molecular Dynamics). 5 The RT-PCR fragments were sequenced by cycle-sequencing. Prior to sequencing, the PCR-fragments were purified using Qiaquick-PCR purification kit (Qiagen) and typically 50 to 100 ng was used with I pmol of fluorescently-labelled primer, for 15 to 22 cycles. The above is the subject of GB 97 08851.2. The work reflected therein has since been 0 pursued and will now be further disclosed in three parts. The first may be given the sub-heading "Prenatal diagnosis in CDGl families: beware of heterogeneity" and may be outlined as follows: 5 Carbohydrate-deficient glycoprotein syndrome type I (CDG1) is an autosomal recessive metabolic disorder with severe psychomotor retardation and a high mortality rate in early childhood. Most patients have a deficiency of phosphomannomutase, due to mutations in PMM2, a gene located on chromosome 16pl3. Over a period of 18 months we offered prenatal diagnosis to eight families. In six cases and prior to the identification of the gene, 0 the diagnosis was based on linkage analysis and phosphomannomutase measurements. Subsequently direct mutation analysis has been used in two families. It shown here that phosphomannomutase activities are strongly reduced in cultured amniocytes and trophoblasts of affected foetuses. We refrained from offering prenatal testing in two other families, because either the disease did not link to chromosome 16 and/or normal WO 98/49324 PCT/EP98/02593 -16 phosphomannomutase activities were measured in fibroblasts from the proband. This confirms earlier suggestions of heterogeneity for CDG1. Introduction 5 Carbohydrate-deficient glycoprotein syndrome type I (CDGl) or Jaeken disease (OMIM 212065) is a genetic multi-system syndrome. At birth, the patients have a peculiar abnormal distribution of subcutaneous fat and nipple retraction. There is encephalopathy with axial hypotonia, abnormal eye movements, internal strabismus, pronounced psychomotor retardation without regression, as well as peripheral neuropathy, cerebellar .0 hypoplasia and retinitis pigmentosa. Severe infections, liver insufficiency or cardiomyopathy lead to 20% leathality in the first years of life [1,2]. The disease is associated with defective glycosylation of glycoconjugates [3-6]. The diagnosis of CDG1 has clasically been based on the abnormal pattern which is observed .5 after iso-electric focusing (IEF) of serum transferrins [3,4]. Unfortunately, the transferrin assay does not reveal an aberrant pattern in amniotic fluid or in foetal blood [7,8]. Since the CDG1 locus has been mapped to chromosome 16p [9], prenatal diagnosis has become possible by linkage analysis. However, a major concern has arisen with the identification of genetic heterogeneity in CDGl. It was initially shown that the disease did not link to 20 chromosome 16pl3 in a family with two affected siblings [10], and we describe another case in the present report. Obviously, a prenatal test with linked genetic markers on chromosome 16p is of no value in such cases. In 1995, Van Schaftingen and Jaeken identified a deficiency of phosphomannomutase WO 98/49324 PCT/EP98/02593 -17 activity in patients with CDG 1 [11]. We have recently identified PMM2, a phosphomannomutase gene on chromosome 16pl3 [12]. Mutations in this gene in CDG1 patients provide substantial evidence that it is the gene for CDG1. The phosphomannomutase deficiency has now been found in more than 50 CDG 1 patients from 5 various geographical origin [13], and mutations have been identified in a majority of these patients [14] submitted. Thus a direct test is now available for prenatal diagnosis. Here we report on prenatal diagnosis in eight CDGl families, thereby showing that phosphomannomutase activities are reduced in amniocytes and trophoblasts of affected 0 foetuses. We would like to spread a cautionary note on the use of linked genetic markers or phosphomannomutase measurements in the prenatal diagnosis of CDG 1 in cases in which the diagnosis has not been confirmed by phosphomannomutase measurements or mutation analysis in the proband or in the parents. 5 Materials and Methods Linkage Analysis Polymorphic CA-repeats in the interval between D16S405 and D16S406 were amplified from genomic DNA from lymphocytes, fibroblasts, or cultured amniocytes [9, 10]. Markers D16S3020 and D16S3087 have only recently been mapped to the interval between 0 D16S513 and D16S404 [15]. Phosphomnannonutase Activities Phosphomannomutase was assayed spectrophotometrically [13]. In this report, the enzymatic activities are expressed as a percentage of the mean values in controls. Control WO 98/49324 PCT/EP98/02593 -18 activities were (means ± SD) 2.46 + 0.48 mU/mg protein (n= 12) in leukocytes; 3.77 ± 0.86 (n=8) in fibroblasts; and 6.7 + 1.0 (n=3) in amniocytes. Mutation Analysis of PMM2 5 Single-strand conformation polymorphism (SSCP) analysis was used to identify mutations in the PMM2 gene, and the underlying base substitutions were identified by sequencing, as described [12]. A complete set of primers for the analysis of PMM2 will be described elsewhere [14]. 0 Figure 2-1 Schematic presentation of three CDG1 ftunilies with marker data and phosphomannomutase (PMM) results The polymorphic markers are ordered from telomeric to centromeric, top to bottom. The dark chromosome is associated with the disease. In family 18, the disease is not linked to chromosome 16. Enzymatic activities are expressed as a % mean values in controls. 5 NA= no phosphomannomutase data. Figure 2-2 Direct mutation analysis of the PMM2 gene PCR amplification of exon 4 (left panel) and exon 5 (right panel) of the PMM2 gene, that harbour the A108V and R141H mutations respectively (see [12]), on genomic DNA from 0 the patient, parents and foetus, was followed by SSCP analysis. The bands associated with the mutation are indicated with an arrowhead. The A108V mutation is of maternal origin in this family. The R141H mutation is inherited from the father. The foetus inherited the paternal disease allele, and is a carrier of the R141H mutation.

WO 98/49324 PCT/EP98/02593 -19 Results The proband in family 15 (Figure 2-la) has a phosphomannomutase deficiency with a value below 0.1 mU/mg protein in fibroblasts (HS in Table I in Van Schaftingen and Jaeken [11]. DNA analysis with linked genetic markers on cultured amniocytes revealed that the 5 foetus was not affected but was a heterozygotus carrier. Phosphomannomutase measurements on these amniocytes showed intermediate values, compatible with a carrier status. A healthy baby was born, as expected. In family 37, a crossover within the CDG I region between D16S414 and D16S407 was 10 observed on the maternal chromosome 16 in the patient (Figure 2-lb). Based on published data, the test with the genetic markers was inconclusive. However, the phosphomannomutase measurements showed that the foetus was affected (less than 2% of the normal values) and these data were considered conclusive. This is consistent with more recent mapping data: the PMM2 gene is located between D16S406 and D16S3078 [16] (this 15 location is more telomeric than we had previously inferred from association data [10], and is in accordance with the data from the Scandinavian group [9, 17]. The pregnancy was terminated. Foetal tissue was obtained and assayed for phosphomannomutase: the activity was ( 0.05 mU/mg protein in muscle and liver. 20 In a third family with a phosphomannomutase deficiency in the proband and evidence for linkage to chromosome 16p13 (family 1 in Matthijs e a [10]) the proband and foetus have inherited a different paternal chromosome (not shown). However, direct phosphomannomutase measurements on cultured amniocytes revealed low values (20% of mean control values), which could either be interpreted as a (partial) deficiency or a WO 98/49324 PCT/EP98/02593 -20 heterozygotic value. As the genetic data are not compatible with a disease state, the latter was withheld. The low values were attributed to the poor growth of amniocytes. The diagnosis in this case thus mostly relied on the genetic data. A healthy baby was born. In a fourth family, we refrained from offering a prenatal test because fibroblasts from the proband showed normal phosphomannomutase values. The marker data also suggested that linkage to chromosome l6p13 was unlikely (family 18, Figure 2-ic). The absence of linkage to the CDG I locus was confirmed after the birth of a second affected child in this family. A prenatal diagnosis was offered to three other families with a known phosphomannomutase deficiency (families 2, 9 and 32, data not shown), and in all cases the biochemical and genetic data were consistent. Healthy babies were born (one is predicted to be a carrier). By now, PMM2 mutations have been identified in these families 5 [14]. More recently, prenatal diagnosis was based on the direct mutation analysis of the PMM2 gene in families in which the mutations had been identified. In family 30, the proband had the A108V and the common R141H mutations, as shown by SSCP analysis and confirmed 0 by sequencing [12]. SSCP analysis of DNA isolated from cultured trophoblasts revealed that the foetus had inherited the paternal R141H mutation but not the maternal A108V mutation (Figure 2-2). A phosphomannomutase value of 1.2 mU/mg protein was compatible with the carrier status. In family 45, no material was available from the deceased proband. However, phosphomannomutase measurements on leukocytes of the WO 98/49324 PCT/EP98/02593 -21 parents revealed that they were carriers, which was confirmed by mutation analysis (data not shown): the father carries the F119L mutation and the mother the R141H mutation, and by deduction, the proband had the F119L/R141H genotype. Both mutations were found in the foetus. Phosphomannomutase values of 0.35 mU/mg protein in cultured amniocytes 5 confirmed the affected status. In family 44, in which the proband did not show a phosphomannomutase deficiency, prenatal testing was not possible. Discussion An unfortunate result in prenatal diagnosis was obtained when, a few years ago, Clayton 0 and collaborators used the transferrin assay on foetal blood [7]. Apparently, the IEF pattern is normal in foetal blood and soon after birth, an observation that remains unexplained to date. In 1994, Martinsson and collaborators mapped to a region of the CDGI candidate region .5 to chromosome l6p13, 13 cM between markers D16S406 and D16S405 [9]. Thus an indirect genetic test became available, based on the analysis of linked markers. However, the observation of genetic heterogeneity for CDG1 [10] also jeopardised this assay. A major breakthrough for CDGI was the identification of a phosphomannomutase 20 deficiency in patients [11]. In this report, we show that phosphomannomutase is deficient in (cultured) trophoblasts and amniocytes from affected foetuses. In eight families, linkage or mutation analysis was combined with phosphomannomutase measurements of foetal cells. The results suggest that low to intermediate values should be interpreted with caution. Low values were obtained in poorly growing amniocytes in family 1, even though WO 98/49324 PCT/IEP98/02593 -22 linkage analysis predicted a normal genotype. Note also that there are important variations in the phosphomannomutase values obtained in different cell types from patients with the same genotype. For instance, in fibroblasts from the proband in family 37 (see Figure 2 1), activities of up to 27% of normal values were obtained in two independent assays. 5 However, in the amniocytes from the affected foetus in the same family, phosphomannomutase values were virtually zero, whereas both patients share the same genotype. It might well be that the phenotype is partly rescued in cells after prolonged culture of the fibroblasts; however, in no instance have we measured in fibroblasts from patients values above 30% of the normal values. 10 The problem of interpretation of biochemical data or the risk of recombination between the markers and the mutation is solved by direct mutation analysis, provided that the paternal and maternal mutation have been identified. We presented the results for two such cases. Taken together, we predicted two affected foetuses and three carriers of either a paternal 15 or maternal allele, while the remaining three foetuses did not inherit a disease chromosome. Given the sample size, this is consistent with the autosomal recessive inheritance of the disease. It is also now clear that some patients with the characteristically abnormal IEF pattern of serum transferrins, do not have a phosphomannomutase deficiency ([13] and the present report). Most likely, however, in all families with a phosphomannomutase 20 deficiency in the proband, there is linkage to the CDG1 locus on chromosome 16pl3. In conclusion, the present results show that prenatal diagnosis for CDGI can now reliably be made. Our recommendations for the prenatal diagnosis of CDG I are as follows. One should first determine the phosphomannomutase activity in fibroblasts, leukocytes or WO 98/49324 PCT/EP98/02593 -23 lymphocytes of the proband, or if not possible, in leukocytes of the parents, and look for mutations in the PMM2 gene. If no phosphomannomutase deficiency is found, no prenatal diagnosis should be offered at this stage. If a phosphomannomutase deficiency is found and the mutations are identified, the prenatal diagnosis should primarily be based on the 5 detection of mutations. If the mutations are not found, phosphomannomutase measurements should be combined with linkage analysis. References (Second Section) 1. Jaeken J., Carchon H.: The carbohydrate-deficient glycoprotein syndromes: an .0 overview. J Inher Metab Dis 1993; 16: 813-820. 2. Jaeken J., Matthijs G., Barone R., Carchon H.: Syndrome of the month: carbohydrate-deficient glycoprotein (CDG) syndrome type I. J Med Gener 1997; 34: 73-76. 3. Jaeken J. et al: Sialic acid-deficient serum and cerebrospinal fluid transferrin in a newly recognised genetic syndrome. Clin Chem Acta 1984; 144: 245-247. 15 4. Stibler H., Jaeken J.: Carbohydrate deficient serum transferrin in a new systemic hereditary syndrome. Arch Dis Child 1990; 65: 107-111. 5. Wada Y. et al: Structure of serum transferrin in carbohydrate-deficient glycoprotein syndrome. Biochem Biophys Res Coninun 1992; 189: 832-836. 6. Yamashita K. et al: Sugar chains of serum transferrin from patients with 20 carbohydrate deficient glycoprotein syndrome. J Biol Chen 1993; 268: 5783-5789. 7. Clayton P. et al: Carbohydrate deficient glycoprotein syndrome: normal glycosylation in the fetus. Lancer 1993; 341: 956. 8. Stibler H., Skovby F.: Failure to diagnose carbohydrate-deficient glycoprotein deficient syndrome prenatally. Pediatr Neural 1994; 11: 71.

WO 98/49324 PCT/EP98/02593 -24 9. Martinsson T. et a]: Linkage of a locus for carbohydrate-deficient glycoprotein syndrome type I (CDG1) to chromosome 16p and linkage disequilibrium to micro-satellite marker D16S406. Hum Mol Genet 1994; 3: 2037-2042. 10. Matthijs G. et al: Evidence for genetic heterogeneity in the carbohydrate-deficient 5 glycoprotein syndrome type I (CDG1). Genomics 1996; 35: 597-599. 11. Van Schaftingen E., Jaeken J.: Phosphomannomutase deficiency is a cause of carbohydrate-deficient glycoprotein syndrome type I. FEBS Lert 1995; 377: 318-320. 12. Matthijs G. et al: Mutations in a phosphomannomutase gene, PMM2, on chromosome 16 in carbohydrate-deficient glycoprotein type I syndrome (Jaeken syndrome). 0 Nature Gener 1997; 16: 88-92. 13. Jaeken J. et al: Phosphomannomutase deficiency is the main cause of carbohydrate deficient glycoprotein syndrome with type I isoelectrofocusing pattern of serum sialotransferrins. J Inher Metab Dis 1997; 20: 447-449. 14. Matthijs G. et al: Lack of homozygotes for the most frequent disease allele in 5 carbohydrate-deficient glycoprotein syndrome type IA. Am. J. Hum. Genet. (unpublished). 15. Dib C. et al: A comprehensive genetic map of the human genome based on 5264 microsatellites. Nature 1996; 380 Supplement. 16. Schollen E. et al: Comparative analysis of the phosphomannomutase genes PMM1, 20 PMM2 and PMM2<p. The sequence variation in the processed pseudogene is a reflection of the mutations found in the functional gene. Hum. Mol. Gener. (unpublished). 17. Bjursell C. et al: Fine mapping of the gene for carbohydrate-deficient glycoprotein syndrome, type I (CDGl): linkage disequilibrium and founder effect in Scandinavian families. Genomics 1997; 39: 247-253.

WO 98/49324 PCT/EP98/02593 -25 The second may be given the sub-heading "Comparative analysis of the phosphomannomutase genes PMMI, PMM2 and PMM2<p: The sequence variation in the processed pseudogene is a reflection of the mutations found in the functional gene" and may be outlined as follows: 5 The search for the carbohydrate-deficient glycoprotein syndrome type I (CDGI) gene has revealed the existence of a family of phosphomannomutase (PMM) genes in humans. Two expressed PMM genes, PMMI and PMM2 are located on chromosome bands 22q13 and l6p13 respectively, and a processed pseudogene PMM24 is located on chromosome 18p. o Mutations in PMM2 are the cause of CDG type IA whereas no disorder has been associated with defects in PMM] as yet. Here, we describe the genomic organization of these paralogous genes. There is a 65% identity of the coding sequence, and all intron/exon boundaries have been conserved. The processed pseudogene is more closely related to PMM2. Remarkably, several base substitutions in PMM2 that are associated with disease, 5 are also present at the corresponding positions in the pseudogene. Thus, mutations that occur at a slow rate in the active gene in the population have also accumulated in the pseudogene. Introduction 0 Carbohydrate-deficient glycoprotein syndrome type I (CDGl) is the paradigm of a group of genetic multi-system disorders characterized by a deficiency in the glycosylation pathway (see (1) for a recent review). CDG1 is an autosomal recessive disorder, with a major disease locus located on chromosome band 16pl3. Van Schaftingen and Jaeken (2) have previously identified a deficiency of phosphomannomutase activity in CDGI patients. The WO 98/49324 PCT/EP98/02593 -26 enzyme isomerizes mannose 6-phosphate into mannose 1-phosphate, which is then converted into GDP-mannose. The latter is required for the synthesis of dolichol-P oligosaccharides in the endoplasmic reticulum. The deficiency has now been confirmed in the vast majority of CDG1 cases (3). 5 The search for the CDG 1 gene led to the discovery of two genes in the human genome that encode active phosphomannomutases. Their identification was based on their sequence similarity to yeast phosphomannomutase or SEC53 (4). The first human phosphomannomutase or PMM gene, PMMI is localized on chromosome band 22q13 and could therefore not harbor the primary defect in CDG1 patients (5). More recently, PMM2 10 was mapped to the CDGI candidate region on chromosome 16p13 (6). Mutations in PMM2 have been identified in CDG I patients with a documented phosphomannomutase deficiency, providing evidence that this gene is the CDGl gene (6). The genomic structure of PMMI and PMM2 has now been determined and compared, and the data are presented here. In the course of the experiments, a processed pseudogene 15 PMM2 was identified. The availability of the flanking sequences has also allowed an extensive analysis of PMM2 in CDG I patients (Matthijs et al., submitted). Interestingly, PMM2 carries a number of mutations at positions corresponding to those where mutations were found in PMM2 in patients. 20 Results The mapping of the CDG1 syndrome and precise localization of the PMM2 gene The CDG1 gene had been localized to chromosome band 16p13 by linkage analysis (7,8). Critical cross-overs in patients and in carriers have allowed us to reduce the candidate region to a region of less than I cM between markers D16S406 and D16S404 (figure 3-1 WO 98/49324 PCT/EP98/02593 -27 and unpublished data). The carriers were identified by measuring phosphomannomutase activities in leucocytes, as described by Van Schaftingen and Jaeken (2). The same critical region has been delineated by linkage disequilibrium studies in Scandinavian CDGl families (9). 5 Figure 3-2 presents the final mapping data on the PMM2 gene and our contribution to the physical map of chromosome 16 (10). A YAC contig covering this minimal region was constructed, starting from the physical map which had been constructed for chromosome 16 at the Center for Human Genome Studies (Los Alamos National Laboratory, Los 0 Alamos; see (11)). As shown in figure 3-2, the YAC contig spans the region of interest, be it that only a single YAC, 802G4, spans the gap between markers D16S406 and D16S3087, and is positive for marker s54A6. The order of the markers on the Los Alamos map was accepted, but markers D16S513 and D16S406 might be switched. There is conflicting data in the literature about the order of these markers and their position relative 5 to D16S495 and D16S502 ((9) and data from the Whitehead Institute and from Los Alamos). YAC 802G4 was unstable in our hands, and a new screening of the CEPH MEGA YAC library with marker D16S3020 did not identify any other YAC clones to bridge the region. Therefore, a BAC library was screened with markers s52C5, D16S3087, s54A6, 0 D16S3020, D16S406 and D16S502, and several BACs could be positioned on the map and linked to the YAC contig. Overlaps between different BACs (as indicated in figure 3-2) were identified by hybridization of Southern blots of the BACs with end fragments, generated by vectorette PCR. No BACs were identified for D16S502. As soon as the PMM2 cDNA was available (6), it was used as a probe for hybridization of Southern blots WO 98/49324 PCT/EP98/02593 -28 of the YACs: positive signals were obtained for YAC clones 802G4 and 909F5. These results were confirmed by PCR on YAC DNA with primers specific for exon 4 of PMM2. This result localizes the gene between D16S3020 and D16S502/D16S406. Marker D16S3020 is close to (within 20 kb) and upstream (5') of the PMM2 gene, because 5 positive signals were obtained by PCR with the BAC clones and several of the cosmids that contain the 5' end of the gene (see below); the chromosomal orientation of the gene has not been determined. On the high-resolution somatic cell hybrid panel of chromosome 16 (references 11-13), the gene was mapped to the interval between CY197/CY196 and CY198. The BAC clones isolated with the PMM2 cDNA and localized on 16p by FISH 0 analysis (41MI6, 27D21, 342N15) (6), have been integrated into the map (figure 3-2). PMMI, PMM2 and a processed pseudogeiie onx chromosome 18 PMM1 has been unambiguously localized to 22ql3 (5). Hybridizations at low stringency with a PMM1 cDNA probe on a chromosomal mapping panel did not reveal any other 5 signals. PMM2 was identified after a cDNA sequence, similar to but different from PMM1 was identified among the IMAGE data in Genbank (6). Several BAC clones were isolated using a PMM2 cDNA probe. However, the results of Southern blot analysis of the BAC clones could only be reconciled if a different chromosomal origin of several of the clones was invoked. 0 On a genomic Southern blot, the samd cDNA probe identified two EcoRI fragments of )12 kb and 0.8 kb (not shown). The larger band was also observed with DNA isolated from the cell-lines GM10567, which contains chromosome 16 as the sole human chromosome in a mouse background, and GM10487, known to contain chromosome 14 and material derived from chromosome 16p. The 0.8 kb fragment originates from chromosome 18, WO 98/49324 PCT/EP98/02593 -29 because it was only observed with DNA from GM1l010, a cell-line with human chromosome 18 in a hamster background (not shown). FISH analysis revealed that BAC clones 27D21, 41M16 and 342N15, which contain the )12 kb EcoRI fragment, mapped to the candidate region on the short arm of chromosome 16 (16pl3), and contained (part of) 5 the PMM2 gene, while BAC clones 147L 11, 1 10G4, 322Cl and 311 N20, containing the 0.8 kb EcoRI fragment (figure 3-3A), were derived from the short arm of chromosome 18 (18p) (figure 3-3B). Cloning and sequencing of the 0.8 kb EcoRI fragment revealed a sequence, closely related 10 to the PMM2 cDNA sequence, indicating that it represented a processed pseudogene on chromosome 18. Additional sequence of the processed pseudogene was obtained by cycle sequencing on BAC DNA with specific primers (figure 3-4). The overall sequence similarly between the coding region of PMM2 and the corresponding regions of the intronless, processed pseudogene (PMM2') is 88%. When compared to PMM2, several 15 base substitutions and single base insertions or deletions are present, suggesting that this processed pseudogene has been inactivated by mutations (figure 3-4). The open reading frame is disrupted by an insertion at nucleotide 837 in PMM2V which results in a frameshift and leads to a stop at what would be codon 143 of the pseudogene. The region upstream of PMM2V has no apparent characteristics of a promoter region, but contains two 20 short repeats and an Alu sequence (not shown). We did not find repetitive elements typically flanking retrotransposed genes (14). Downstream of the position corresponding to the stopcodon in the PMM2 transcript, the sequence homology is weak, which suggests that only a partial cDNA has been transposed.

WO 98/49324 PCT/EP98/02593 -30 Genomic structure of PMM1 and PMM2 To determine the genomic structure of PMMI and PMM2, cosmids were isolated by hybridization of arrayed chromosome-specific cosmid libraries from chromosome 22 (reference 15) and chromosome 16 (reference 16) with the respective full length cDNA 5 probes. Southern blot analysis of EcoRI, HindIII and double digests of cosmid DNA, hybridized with several probes derived from the cDNA's or from genomic PCR products, and with exon-specific oligonulceotides, and partial HindIll digestions (in the case of the chromosome 16 cosmids) were combined to determine the genomic structure of both genes. D Figure 3-5 shows the genomic structure of PMM] and PMM2 with a detailed EcoRI and HindIll map. The position of the different cosmids is indicated. For PMM2, the overlapping BAC clones have been included. All exon-intron boundaries and partial intron sequences were obtained by cycle sequencing on cosmid or BAC DNA. Both PMM] and PMM2 are composed of 8 exons, and the major 5 difference resides in exon 8 in which the 3' untranslated region is contained (540 bp for PMMI and 1599 bp for PMM2). The intron/exon junctions conform with the eukaryotic consensus sequences for splice donors and acceptors (figure 3-6). The intron lengths have been determined by PCR. They vary from 160 bp (intron 3) to 4.9 kb (intron 5) in PMM] and from 0.5 kb (intron 5) and ) 4 kb (intron 7) in PMM2. The ) exact size of the last intron in PMM2 has not been precisely determined: none of the available cosmids contained exon 7 plus exon 8, and (long) PCR on genomic or BAC DNA with primers in exon 7 and 8 has not been successful. The PMM1 gene thus spans approximately 13 kb of genomic DNA, whereas the size of PMM2 is at least 17 kb.

WO 98/49324 PCT/IEP98/02593 -31 DISCUSSION We describe a novel family of genes in the human genome, encoding phosphomannomutases and represented by PMMJ on 22q13 and PMM2 on 16q13. A 5 processed pseudogene is present on chromosome 18p. PMM2 is the CDG1 gene, as mutations in this gene have been identified in patients with the syndrome (6). There is currently no disease associated with defects in PMMI. A YAC contig and partial BAC contig were constructed across the CDGI minimal region, 0 which had been reduced to an interval on the genomic map of less than 1 cM between markers D16S406 and D16S404. The positional cloning approach and the physical mapping effort were suspended due to the availability of a candidate gene, i.e. a cDNA for PMM2 (6). PMM2 has now been precisely mapped to the candidate region for CDG1, within 20 kb of D16S3020. PMM1 has previously been isolated by a similar approach and 5 mapped to bin 16.2 on chromosome 22 (reference 5). In functional assays, both proteins display phosphomannomutase activity, be it that the purified enzymes have different kinetics and a distinct substrate specificity ((17); Pirard and Van Schaftingen, unpublished data). The differences indicate that PMMI and PMM2 may 0 have distinct functions. At the amino acid level, the proteins are strongly conserved. An alignment was published in Matthijs et al. (6). PMM2 is shorter than PMMl and yeast and Candida albicans PMM, and the difference is mainly due to a deletion of 7 aminoacids in the N-terminal part of the protein. An insertion of 2 aminoacids, after position 63 and between conserved domains, is unique to PMM I. These variations do not coincide with WO 98/49324 PCT/EP98/02593 -32 exon boundaries. At present, nothing is known about functional domains in the proteins. For comparison, five monomeric phosphoglucomutase (PGM) isoforms are known, all with a different substrate specificity and thermostability (18,19). However, the sequence similarity between the PGM proteins is low. 5 Comparison of the genomic structure of PMMI and PMM2 indicates that these genes have arisen by gene duplication. There are 78 differences at the aminoacid level between PMMI and PMM2, not including the 4 mutational events that have led to difference in length of the proteins. At the DNA level, the silent sites vary in over 50% of the positions, suggesting that the ancestral gene was duplicated before the mammalian radiation that occurred 75 to 110 million years ago. Indeed, we have cloned the PMMI and PMM2, the orthologues of PMMI and PMM2 in the mouse, and these genes are located on syntenic chromosomal regions in the human and the mouse genome (unpublished data). 5 The identification of PMM2 on l6pl3 and PMMI on 22q13 has prompted us to look for other possible paralogous genes in these regions. There are now at least three genes on 16p13 that have homologues on 22q12-ql3: CBP (CREB-binding protein (OMIM 600140)), mutated in the Rubinstein-Taybi syndrome (OMIM 180849) and located on l 6 p1 3 (20-22) has a functional homologue, p300, on 22ql3 (23), and the HMOX-l (heme oxygenase-1, 0 OMIM 141250) and HMOX-2 (OMIM 141252) genes are on 22q12 and 16pl3.3 respectively (24). It is thus very likely that these chromosomal regions are paralogous regions that have arisen by duplication. In view of the theory of S. Ohno, in which it is hypothesized that the genome of higher WO 98/49324 PCT/EP98/02593 -33 organisms has arisen by a round or two of tetraploidization (25), one now wonders whether up to four linkage groups or paralogous regions exist for 16p13. However, thus far, no other members of the PMM family have been identified in humans nor have other proteins related to the HMOX-l and -2 genes been described. At least 2 proteins, related to 5 CBP/p300 are known (26,27) but the corresponding genes have, to our knowledge, not been located in the human genome. On the other hand, the MYHI I and MYH9 genes on 16pl3.13 and 22qll-ql3 may also be paralogous genes and related MYH genes are clustered on 14q12 and 17p13, whereas members of the somatostatin receptor family have been mapped to 22q13 (SSTR3), l6pl3 (SSTR5), 14q13 (SSTRI), 17q24 (SSTR2) and D 20qll.2 (SSTR4). Thus, regions on 14q and on 17p to 17q are candidate regions in the context of Ohno's suggestions. The gene on chromosome 18 is a processed pseudogene of PMM2. It is closely related to the PMM2 cDNA sequence and the absence of homology in the 5' upstream region suggests 5 that it has arisen by retrotransposition. The 3' tail of the processed pseudogene has not been fully sequenced, so it is not known whether the poly A stretch, characteristic for processed pseudogenes. is present. The presence of a stopcodon at position 143 implies that this gene cannot be translated into a functional enzyme. Most likely, this processed pseudogene has never been actively transcribed. The fact that the pseudogene has 0 accumulated mutations at silent and replacement sites at the same rate fits with this assumption: there is a 10.5 % sequence divergence at all sites, and a 10.8 % divergence at silent sites. If a UEP (unit evolutionary period = time needed for a 1 % divergence) of 2.1 million years is accepted for synonymous or silent sites, one can estimate that this processed pseudogene arose approximately 23 million years ago (28). It is therefore unique WO 98/49324 PCT/EP98/02593 -34 to humans (and possibly primates). Of the 22 missense mutations and 2 polymorphisms thus far identified in CDG1 patients ((6); Matthijs et al., submitted), 7 were also present at the corresponding positions in the processed pseudogene. This is an important observation, because these mutations might 5 interfere with certain mutation detection strategies, e.g. dot blot analysis, and probes need to be designed carefully. In case of the frequent R141H (CGC to CAC) mutation, and of the A108V (GCG to GTG), R123G (CGA to CAA) and T237M (ACG to ATG) mutations, the corresponding base is unchanged, but a mutation occurred in the same CpG dinucleotide, probably caused by a C to T transition in the opposite strand. Since the 10 mutations in the PMM2 gene are single base pair changes, without variations in the neighbouring sequences as in the pseudogene, they can not be explained by gene conversion. It rather indicates that the inactive pseudogene accumulated the same mutations that arose independently in the active PMM2 gene where they cause disease. The C to T and G to A transitions prevail, which is in accordance with the propensity of 5 15 methylcytosine to undergo deamination to form thymine (29). In the case of the frequent R141H mutation, we infer from this comparison that all carrier chromosomes must originate from a single mutation event, because the equally probable R141C mutation seen in the pseudogene has not been observed among CDG I patients. Similar inferences can be made for other mutations. 20 In conclusion, the precise mapping of the PMM] and PMM2 genes allows these genes to be integrated in the physical maps of the respective chromosomes. Their gene structure has been determined, and strongly supports their origin by duplication. The latter suggestion sheds light on a common origin of the chromosomal regions of 16pI3 and 22q13 in human.

WO 98/49324 PCT/EP98/02593 -35 Also, the identification of a processed pseudogene is interesting from the point of view of molecular evolution, in that it seems to have acted as a sink for mutations since its creation by retrotransposition. Mutations at the corresponding nucleotides in PMM2<p in the human genome and in PMM2 in patients may inadvertently interfere with molecular diagnosis of 5 CDG1. MATERIALS AND METHODS YAC analysis 0 YAC clones were obtained from "mega" YAC libraries constructed at CEPH. The selected YAC clones were grown on ura- trp- plates and individual colonies were picked, grown in selective AHC medium, and analyzed by pulsed-field gel electrophoresis (PFGE) to check the size of the insert. Yeast DNA was prepared in agarose plugs (200 ld) according to Ragoussis (30). The STS content of the YACs was checked by PCR. For PCR analysis, [5 a quarter of a plug was dissolved in 1 ml H 2 0 and 5 s.I of this solution was used in 25 pI reaction volume, under standard PCR conditions. A PFGE gel with YACs 820D10, 948A8, 931F5, 912C5, 905F10, 909F5, 802G4, 944F4, 951F4, 925B10 and 936B6 was blotted and the filter was hybridized with the PMM2 cDNA probe under standard conditions (31). The cDNA probe contained the entire coding, and 20 part of the 3' untranslated region (to nucleotide 1077). Isolation of cosmid and BAC clones The cosmid clones for PMM1 were isolated and kindly provided by M.L. Budarf (5). Cosmids for PMM2 were identified by hybridization of a chromosome 16 specific arrayed WO 98/49324 PCT/EP98/02593 -36 cosmid library (16). BAC clones for markers s52C5, D16S3087 and D16S406 and PMM2 were obtained by hybridization of "high density human BAC colony DNA membranes" (Research Genetics). The probe for s52C5 was generated by PCR from the amplicon of s52C5 and random primed labeling. For the polymorphic markers D16S406 and D16S3087, 5 the specific PCR primers were used as probes for oligonucleotide hybridization. The primer sequences were obtained from GDB. For PMM2, the insert of the PMM2 cDNA clone was purified and labeled. The BAC clones for marker s54A6 were isolated by PCR analysis of BAC DNA pools (Research Genetics). 0 BAC ends were rescued by vectorette PCR. Vectorette libraries were constructed as described by Riley et al. (32) with minor modifications. In brief, 200 ng DNA was digested with 5 U Rsal and ligated to 6 pmol Rsal vectorette cassette (top strand: 5' CAAGGAGAGGACGCTGTCTGTCGAAGG-3'; bottom strand: 5' CTCTCCCTTCTCGAATCGTAACCGTTC-3'). The end fragments were then amplified 5 by PCR with the universal vectorette primer (5' CGAATCGTAACCGTTCGTACGAGAATCGCT-3') and the SP6 or T7 primer. The purified PCR product was used as a probe in Southern hybridization. Analysis of BACs and cosmid clones 0 BAC clones for PMM2 and cosmids for PMMI and PMM2, were analyzed by hybridization of Southern blots of EcoRI and Hindlll digestions. cDNA or genomic probes were labeled with - 32 P-dCTP by random primed labeling and used for hybridization overnight in 50% formamide, 5X SSPE, loX Denhardt's solution, 2% SDS and 100sg/ml heparin. Filters were washed for 30 min in 0.1 X SSPE and 0.1 % SDS at 62'C. Oligonucleotide probes, WO 98/49324 PCT/EP98/02593 -37 derived from the cDNA sequence and used for the determination of the genomic structure, were typically 20 bp long and 5'-labeled using y- 32 P-ATP and T4 PNK according to established protocols (31). Hybridizations with oligonucleotide probes were done in 6X SSPE, 5X Denhardt's solution, 0.5% SDS and 200 pg/ml heparin. Filters were washed in 5 2X SSPE, 0.1 % SDS at 45'C for 5 min. A detailed Hindlll map of cosmids 428D1, 408C7, 422F4 and 404H6 was generated by partial digestion. Four pg of purified cosmid DNA was digested to completion with Notd (20U) in 2 0

/,

1 and further digested with Hindlll (5U) in 100pl. Aliquots of 25[l were taken at 2, 5, 10 and 60 minutes. The reaction was stopped by adding iu 1 0. 1 M EDTA. Samples were analyzed by field-inversion gel D electrophoresis (FIGE), blotted and hybridized with T3 and T7 oligonucleotides. Gene-specific primers have been used for genomic PCR on cosmids, BACs and total human DNA under standard conditions. The Long Expand PCR kit was used as prescribed by the manufacturer (Boehringer Mannheim). 5 Fluorescently labeled primers (FITC, fluorescein-isothiocyanate) were used for cycle sequencing of cosmid DNA using Amersham's Thermo-sequenase kit and 2 to 6 pg of

DNA.

WO 98/49324 PCT/EP98/02593 -38 REFERENCES (Third Section) 1. Jaeken, J., Matthijs, G., Barone, R. and Carchon, H. (1997). Syndrome of the month: Carbohydrate-deficient glycoprotein (CDG) syndrome type 1. J. Med. Genet., 34, 73-76. 5 2. Van Schaftingen, E. and Jaeken, J. (1995) Phosphomannomutase deficiency is a cause of carbohydrate-deficient glycoprotein syndrome type 1. FEBS Lett., 377, 318-320. 3. Jaeken, J., Artigas, J., Barone, R., Fiumara, A., de Koning, T.J., Poll-Ths, B.T., de Rijk-van Andel, J.F., Hoffmann, G., Mayatepek, E., Pineda, M., Vilaseca, M.A., Saudubray, J.M., Schl~iter, B., Wevers, R. and Van Schaftingen, E. (1997). Phosphomannoim utase deficiency is the main cause of carbohydrate-deficient glycoprotein syndrome with type I isoelectrofocusing pattern of serum sialotransferrins. J. Inher. Metab. Dis., in press. 5 4. Kepes, F. and Schekman, R. (1988) The Yeast SEC53 gene encodes phosphomannomutase. J. Biol. Chem., 263, 9155-9161. 5. Matthijs, G., Schollen, E., Pirard, M., Budarf, M.L., Van Schaftingen, E. and 0 Cassiman, J-J. (1997). PMM (PMM1), the human homologue of SEC53 or yeast phosphomannomutase, is localized on chromosome 22ql3. Genonics, 40, 41-47. 6. Matthijs, G., Schollen, E., Pardon, E., Veiga-da-Cunha, M., Jaeken, J., Cassiman, J.

J. and Van Schaftingen, E. (1997) Mutations in a phosphomannomutase gene, PMM2, on WO 98/49324 PCT/EP98/02593 -39 chromosome 16 in carbohydrate-deficient glycoprotein type I syndrome (Jaeken syndrome). Nature Genet., 16, 88-92. 7. Martinsson, T., Bjursell, C., Stibler, H., Kristiansson, B., Skovby, F., Jaeken, J., 5 Blennow, G., Strb5mme,P., Hanefeld, F. and Walistr6m, J. (1994) Linkage of a locus for carbohydrate-deficient glycoprotein syndrome type I (CDG 1) to chromosome 16p, and linkage disequilibrium to microsatellite marker Dl 6S406. Hum. Mol. Genet., 3, 2037-2042. 8. Matthijs, G., Legius, E., Schollen, E., Vandenberk, P., Jaeken, J., Barone, R., 10 Fiumara, A., Visser, G, Lambert, M. and Cassiman, J.-J. (1996) Evidence for genetic heterogeneity in the carbohydrate-deficient glycoprotein syndrome type I (CDG1). Genonics, 35, :597-599. 9. Bjursell, C., Stibler, H., Wahlstr6m, J., Kristiansson, B., Skovby, F., Str6mme, P., 15 Blennow, G. and Martinsson, T. (1997) Fine mapping of the gene for Carbohydrate Deficient glycoprotein syndrome, type I (CDG 1): linkage disequilibrium and founder effect in scandinavian families. Genomics, 39, 247-253. 10. Doggett, N.A., Breuning, M.H. and Callen, D.F. (1996) Report of the fourth 20 international workshop on human chromosome 16 mapping 1995. Cyogenet. Cell Genet., 72, 271-293. 11. Doggett, N.A., Goodwin, L.A., Tesmer, J.G., Meincke, L.J., Bruce, D.C., Clark, L.M., Altherr, M.R., Ford, A.A., Chi, H.-C., Marrone, B.L., Longmire, J.L., Lane, WO 98/49324 PCT/EP98/02593 -40 S.A., Whitmore, S.A., Lowenstein, M.G., Sutherland, R.D., Mundt, M.O., Knill, E.H., Bruno, W.J., Macken, C.A., Torney, D.C., Wu, J.-R., Griffith, J., Sutherland, G.R., Deaven, L.L., Callen, D.F. and Moyzis, R.K. (1995) An integrated physical map of human chromosome 16. Nature, 377 (supp), 335-350. 5 12. Callen, D.F., Baker, E., Eyre, H.J. and Lane, S.A. (1990). An expanded mouse human hybrid cell panel for mapping human chromosome 16. Ann. Gener., 33,190-195. 13. Callen, D.F., Doggett, N.A., Stallings, R.L., Chen, L.Z., Whitmore, S.A., Lane, 0 S.A., Nancarrow, J.K., Apostolou, S., Thompson, A.D., Lapsys, N.M. et al. (1992) High-resolution cytogenetic-based physical map of human chromosome 16. Genonics, 13, 1178-1185. 14. Maestre, J., Tchenio, T., Dhellin, 0. and Heldmann, T. (1995) mRNA retroposition 5 in human cells: processed pseudogene formation. EMBO, 14, 6333-6338. 15. Budarf, M.L., Eckman, B., Michaud, D., McDonald, T., Gavigan, S., Buetow, K.H., Tatsumura, Y., Liu, Z., Hilliard, C., Driscoll, D., Goldmuntz, E., Meese, E., Zwarthoff, E.C., Williams, S., McDermid, H., Dumanski, J.P., Biegel, J., Bell, C.J. and 0 Emanuel, B.S. (1996) Regional localization of over 300 loci on human chromosome 22 using a somatic cell hybrid mapping panel. Genonics, 35, 275-288. 16. Longmire, J.L., Brown, N.C., Meincke, L.J., Campbell, M.L., Albright, K.L., Fawcett, J.J., Campbell, E.W., Moyzis, R.K., Hildebrandt, C.E., Evans, G.A. et al.

WO 98/49324 PCT/EP98/02593 -41 (1993) Construction and characterization of partial digest DNA libraries made from flow sorted human chromosome 16. Genetic Analysis: Techniques and Applications, 10, 69-76. 17. Pirard, M., Collet, J.-F., Matthijs, G. and Van Schaftingen, E. (1997) Comparison 5 of PMM1 with the phosphomannomutases expressed in rat liver and in human cells. FEBS Lett., 411, 251-254. 18. Harris, H. and Hopkinson, D.A. (1976) Handbook ofenzyme electrophoresis in human generics. North Holland, Amsterdam. 0 19. Edwards, Y.H., Putt, W., Fox, M. and Ives, J.H. (1995) A novel human phosphoglucomutase (PGM5) maps to the centromeric region of chromosome 9. Genomics, 30, 350-353. 5 20. Petrij, F., Giles, R.H., Dauwerse, H.G., Saris, J.J., Hennekam, R.C.M., Masuno, M., Tommerup, N., van Ommen, G.-J. B., Goodman, R.H., Peters, D.J.M. and Breuning, M.H. (1995) Rubinstein-Taybi syndrome caused by mutations in the transcriptional co-activator CBP. Nature, 376, 348-35 1. 0 21. Chen, X.-N. and Korenberg, J.R. (1995) Localization of human CREBBP (CREB binding protein) to 16p 13.3 by fluorescence in situ hybridization. Cyrogenet. Cell Genet., 71, 56-57. 22. Wydner. K. L., Bhattacharya, S., Eckner, R., Lawrence, J. B. and Livingston, D.M.

WO 98/49324 PCT/EP98/02593 -42 (1995) Localization of human CREB-binding protein gene (CREBBP) to 16pl3.2-pl3.3 by fluorescence in situ hybridization. Genomics, 30, 395-396. 23. Eckner, R., Ewen, M.E., Newsome, D., Gerdes, M., DeCaprio, J.A., Lawrence, 5 J.B. and Livingston, D.M. (1994) Molecular cloning and functional analysis of the adenovirus ElA-associated 300-kD protein (p300) reveals a protein with properties of a transcriptional adaptor. Genes Dev., 8, 869-884. 24. Kutty, R.K., Kutty, G., Rodriguez, I.R., Chader, G. J. and Wiggert, B. (1994) 10 Chromosomal localization of the human oxygenase genes: heme oxygenase-1 (HMOXl) maps to chromosome 22q12 and heme oxygenase-2 (HMOX2) maps to chromosome 16pl3.3. Genonics, 20, 513-516. 25. Ohno, S. (1993) Patterns in genome evolution. Curr. Opinion Genet. Devel., 3, 911 15 914. 26. Barbeau, D., Charbonneau, R., Whalen, S.G., Bayleu, S.T. and Branton, P.E. (1994) Functional interactions within adenovirus EIA protein complexes. Oncogene, 9, 359-373. 20 27. Dallas, P.B., Yaciuk, P. and Moran, E. (1997) Characterization of monoclonal antibodies raised against p300: both p300 and CBP are present in intracellular TBP complexes. J. Virol., 71, 1726-1731. 28. Li, W.-H., Luo, C.-C. and Wu, C.-I. (1983) Evolution of DNA sequences. In WO 98/49324 PCT/EP98/02593 -43 Maclntyre, R.J. (ed), Molecular Evolutionary Genetics. Plenum Press, New York. 29. Cooper, D.N. and Krawczak, M. (1993) Human Gene Mutation. Bios Scientific Publishers, Oxford. 5 30. Ragoussis, J. (1996) Restriction analysis of YACs. Methods Mol. Biol., 54, 69-74. 31. Sambrook, J., Fritsch, E.F. and Maniatis, T.A. (1989) Molecular cloning. A laboratory manual. Second Edition. Cold Spring Harbour Laboratory Press, Cold Spring 0 Harbor. 32. Riley, J., Butler, R., Ogilvie, D., Finniear, R., Jenner, D., Powell, S., Anand, R., Smith, J.C. and Markham, A.F. (1990) A novel, rapid method for the isolation of terminal sequences from yeast artificial chromosome (YAC) clones. Nucleic Acids Res., 18, 2887 5 2890.

WO 98/49324 PCT/EP98/02593 -44 FIGURE LEGENDS (Third Section) Figure 3-1: Delineation of the candidate region for CDG1 by genetic analysis of families with a documented PMMII deficiency in the proband. 5 The genetic map of the CDG1 region is shown, and the polymorphic markers were ordered according to the available maps (10,11), with the telomere to the left, centromere to the right. The previously published candidate region extends from D16S406 to D16S405 and spans 13 cM (7). The regions represented by thin lines were excluded by cross-overs between the marker alleles and the disease in patients or carriers in the different families 10 (ni: not informative). The minimal region is defined by markers D16S406 and D16S404, that are less than I cM apart on the genetic map. Figure 3-2: Physical map of the CDG 1 candidate region on chromosome band l6p13.2. A YAC contig is shown for the region from D16S502 to D16S404 (not to scale). 15 Polymorphic markers and STSs (underlined) are shown, whereby the vertical lines indicate positive hits by PCR. The BAC clones were isolated by screening a BAC library with probes for markers s52C5, D16S3087, s54A6, Dl6S3020 and D16S406. The BAC clones isolated with a PMM2 cDNA are represented in bold. Marker D16S3020 is very closely linked to the PMM2 gene. 20 Figure 3-3: Evidence that sequences related to PMM2 are present on chromosome 18p. A. Southern blot analysis of BAC clones, isolated by hybridization of a PMM2 cDNA probe to "high density human BAC colony DNA membranes" (see Materials and Methods). Purified BAC DNA was digested with EcoRI. The blot was hybridized WO 98/49324 PCT/EP98/02593 -45 with the same PMM2 cDNA probe. Results are shown for BAC clones 41M16, 27D21,147L1l and 110G4. An EcoRI fragment > 12 kb is present in clones 41M16 and 27D21 while clones 147L11 and 110G4 contain a unique 0.8 kb EcoRI fragment. 5 B. FISH analysis with BAC 147L1 1 reveals that the clone is derived from chromosome 18p. Similar results were obtained with BACs 110G4, 322C1 and 311N20. The BAC clones containing the ) 12 kb fragment are derived from chromosome 16p (reference 6). 10 Figure 3-4: Alignment of the PMM2 cDNA sequence with the genomic sequence of the processed pseudogene PMM2,p on chromosome 18. The translation of the reading frame for PMM2 is given above the DNA sequence (single letter codes). The PMM2 sequence is numbered from the ATG codon, the sequence of the pseudogene starts at the upstream EcoRl site. which was used for cloning (not shown). Due 15 to insertions and deletions in the PMM2p sequence, the reading frame is interrupted or altered several times. The open reading frame of the pseudogene ends at the position corresponding to codon 143 in PMM2. The homology was calculated from the alignment shown in the figure. When the sequence difference would lead to a different codon in the putative reading frame of the pseudogene, the corresponding amino acid letter code is 20 represented in bold. The codons in PMM2 that were found mutated in CDG1 patients, and 2 polymorphisms ((6) and Matthijs et al., manuscript submitted) are boxed. An asterisk denotes that the same nucleotide change is seen in the patients and in PMM2p, and an arrowhead indicates the nucleotide of interest. An EcoRI site, present in the coding region of PMM2 (exon 5) and in the processed pseudogene, is underlined.

WO 98/49324 PCT/EP98/02593 -46 Figure 3-5: Comparison of the genomic structure of the PMM1 (top) and PMM2 genes (bottom). A restriction map for EcoRI (E) and Hindill (H) has been established for both genes. Both genes are composed of 8 exons, represented by black boxes. The different cosmid and BAC clones that were used to determine the genomic structure, are schematically represented. Figure 3-6: Alignment of the exon/intron boundaries for PMM1 and PMM2. The coding sequence is represented in uppercase. Values between square brackets represent additional nucleotides present in one gene as compared to the other. The third may be given the sub-heading "Lack of homozygotes for the most frequent disease allele in carbohydrate-deficient glycoprotein syndrome type IA" and may be outlined as follows: Carbohydrate-deficient glycoprotein syndrome type I (CDG 1 or Jaeken syndrome, OMIM 212065) is an autosomal recessive disorder, characterized by a defective glycosylation, Most patients show a deficiency of phosphomannomutase (PMM), the enzyme that converts mannose 6-phosphate to mannose 1-phosphate in the synthesis of GDP-mannose. The ) disease is linked to chromosome l6pl3 and mutations have recently been identified in the PMM2 gene in CDGl patients with a PMM deficiency (CDG type IA). The availability of the genomic sequences of PMM2 allowed us to screen 56 CDG 1 patients from different geographical origin for mutations. By SSCP analysis and by sequencing, we identified 23 different missense mutations and one single base pair deletion. In total, WO 98/49324 PCT/EP98/02593 -47 mutations were found on 99% of the disease chromosomes in CDG type IA patients. The R141H substitution is present on 43 of the 112 disease alleles. However, this mutation was never observed in the homozygous state, suggesting that homozygosity for these alterations is incompatible with life. On the other hand, patients were found homozygous for the D65Y 5 and F119L mutations, which must therefore be mild mutations. One particular genotype, R141H/D188G, which is prevalent in Belgium and the Netherlands, is associated with a severe phenotype and a high mortality. Apart from this, there is only a limited relation between the genotype and the clinical phenotype. 10 Introduction Carbohydrate-deficient glycoprotein (CDG) syndromes are a series of genetic disorders characterized by defective N-glycosylation of serum and cellular proteins (Jaeken et al. 1980; Jaeken and Carchon 1993; Jaeken et al. 1997a; Jaeken and Casaer, 1997). At 15 present, 4 types of CDG have been described on the basis of serum transferrin isoelectrofocusing (IEF). CDG type I (CDG 1) is the most frequent type. It is a severe disorder which presents neonatally. There is a life-threatening liver insufficiency (with an overall 20% mortality in the neonatal period), combined with a severe cerebellar dysfunction and peripheral neuropathy, leading to severe psychomotor retardation. These 20 children also have skeletal deformities and a characteristic deposition of adipose tissue (Jaeken and Carchon 1993; Jaeken et al. 1997a). CDG type II (CDG2), type III (CDG3) and type IV (CDG4) represent only two cases each (Jaeken et al. 1994; Stibler et al. 1993; Stibler et al. 1995). CDG2 is caused by a deficiency of UDP-GlcNAc:a-6-D-mannoside 0-1,2-N-acetylglucosaminyltransferase II (GnT II), located in the Golgi apparatus, and WO 98/49324 PCT/EP98/02593 -48 mutations in the GnT II gene (MGAT2) on 14q21 have been identified (Jaeken et al. 1994; Tan et al. 1996). The causes of CDG3 and CDG4 remain unknown. CDG1 is inherited in an autosomal recessive manner and its locus was mapped to chromosome 16p13 (Martinsson et al. 1994). Linkage to the region between D16S406 and D16S500 was confirmed in 10 of 11 informative families (Matthijs et al. 1996). In one family with two affected siblings, the disease was, however, not linked to chromosome 16p, indicating genetic heterogeneity for CDGl (Matthijs et al. 1996). Biochemical evidence has long suggested a basic defect in the synthesis of the dolichol-P oligosaccharides (synthesis of the asparagine-N-linked oligosaccharides) in the endoplasmic ) reticulum (ER) (Jaeken et al. 1984; Wada et al. 1992). In 1995, Van Schaftingen and Jaeken (1995) identified a deficiency of phosphomannomutase activity in patients with CDG 1. This observation has been confirmed on more than 50 CDG 1 patients from different geographical origin (Jaeken et al. 1997b). We have recently cloned the human phosphomannomutase gene PMM2 and shown that it is the CDGI gene (Matthijs et al. 5 1997a). Another PMM gene, PMMI, could be assigned to chromosome 22ql3 (Matthijs et al. 1997b). Both PMM I and PMM2 have been expressed in E. coli and were found to be active proteins (Pirard et al. 1997, and Van Schaftingen and Pirard, unpublished data). We have previously reported 11 missense mutations in 16 CDGI patients with a documented phosphomannomutase deficiency (CDG type IA) (Matthijs et al. 1997a). These 0 mutations have been identified at the cDNA level, after RT-PCR amplification, followed by SSCP (single-stranded conformation polymorphism) analysis and sequencing. To search for mutations in genomic DNA, the PMM2 intron/exon structure has been determined whereby 8 exons have been identified (Schollen et al. 1998), and primers flanking each translated exon have been designed. We here describe the results of an exhaustive mutation WO 98/49324 PCT/EP98/02593 -49 analysis of the PMM2 gene in patients with a documented phosphomannomutase deficiency. Subjects and Methods 5 Patients Fifty-six patients from 12 countries were included in the study. All except 2 were of Caucasian origin. A diagnosis of CDG I was made in all these patients, based on the clinical manifestations, substantiated by the typical IEF pattern of serum transferrins: there is a strong reduction in the intensity of the normal tetrasialotransferrin band, and a 0 concomitant increase in the disialo- and asialotransferrin concentration. The phosphomannomutase deficiency was documented in most cases (see table 4-2 and Results). The clinical features of CDG I patients have recently been reviewed in Jaeken et al. (1997a) and Jaeken and Casaer (1997). In brief, the neurological picture includes abnormal eye movements, combined with slow head movements in the neonatal period and axial 5 hypotonia with hyporeflexia. Most children present with an alternating strabismus. There is a severe psychomotor retardation and failure to thrive, with ataxia and sometimes deafness. Additional features, presenting after infancy, are hypogonadism, retinitis pigmentosa, joint contractures and stroke-like episodes. Most patients never achieve walking without support, but there is no regression. Other symptoms include: mild facial !0 dysmorphism (with large, somewhat dysplastic ears) and skeletal deformities, and a typical subcutaneous deposition of adipose tissue ("fat pads"). There is a mild to moderate hepatomegaly, and some infants develop pericardial effusion and/or cardiomyopathy. About 20% of the patients have died before the age of 5 as a consequence of liver failure, severe infection, cardiac insufficiency, nephrotic syndrome or status epilepticus.

WO 98/49324 PCT/EP98/02593 -50 Salient features in some patients are listed in table 4-2. The blood samples and/or fibroblast or lynphoblast cultures from patients were provided to us with a request for enzymatic assays and molecular diagnosis, and the referring physicians and the families have been informed about the results. Amniocytes were 5 analyzed in the context of prenatal diagnosis. Single-strand conformation polymorphism (SSCP) analysis and sequencing DNA was isolated from fresh blood, or fibroblast or lymphoblast cultures from patients using a high-salt extraction procedure. 0 Based on the available sequence, primers were designed for the PCR amplification of 9 DNA fragments, suitable for SSCP analysis. One primer in each pair was FITC (fluorescein-isothiocyanate) labeled. The primers sequences are given in table 4-1. PCR reactions were typically done in 25 sl, and the cycling conditions were 30" at 95'C, 30" at 50 to 60'C, 30" at 72'C, for 32 cycles. Ten to 15 pAl of the PCR products were mixed .5 with an equal volume of formamide, denatured for 5' at 95'C, loaded onto a non denaturing polyacrylamide gel (0.5X Hydrolink MDE, J.T.Baker, in 0.6x TBE (1OX TBE is IM Tris, 0.82 M boric acid, 10 mM EDTA)) and electrophoresed at 4'C for 10h at 400V. The gels were directly scanned on a Fluorimager (Vistra) and the signals were analyzed with the ImagequaNT software (Molecular Dynamics). 20 The PCR fragments were sequenced by cycle-sequencing or solid-phase sequencing. Prior to cycle sequencing with the Thermosequenase kit (Amersham), the PCR-fragments were purified using the Qiaquick-PCR purification kit (Qiagen) and typically 50 to 100 ng was used with 1 pmol of fluorescently labeled primer, for 15 to 22 cycles. Solid-phase sequencing using a biotinylated template and streptavidin-coated DynaBeads (Dynal) was WO 98/49324 PCT/EP98/02593 -51 done according to established procedures. Results 5 Fifty-six patients, and their affected siblings, were included in this series. Included are 3 pairs of patients of which each pair has at least one common ancestor (table 4-2). The diagnosis of CDG 1 with phosphomannomutase deficiency was confirmed biochemically in 46 patients for whom fibroblasts, lymphoblasts or fresh leukocytes were available. In 2 of these patients, a partial deficiency of phosphomannomutase was found in fibroblasts 0 (patients 47 (SAO) and 48 (GSS)), and in one family (family 4, patients FG and FP, both affected) intermediate values were measured in lymphoblasts. From the remaining 10 patients for whom the diagnosis has previously been made on clinical grounds, no cells were available for the enzymatic assay. In 7 cases, DNA from the patient was available for analysis (patients RJM, KS, NP, SE, PG, RW and RMan), while in the 3 remaining 5 cases (patients HE, MR and AV), the mutation screening needed to be done on DNA from the parents. In 2 such cases (patients HE and AV), the clinical diagnosis was substantiated by the measurement of intermediate phosphomannomutase activities in fresh leukocytes from the parents. 0 The mutation screening was done by a non-radioactive SSCP analysis of fragments amplified from genomic DNA. The PCR fragments encompassed the individual exons, and the flanking sequences. Whenever a fragment revealed an aberrant SSCP pattern, the corresponding exon was sequenced.

WO 98/49324 PCT/EP98/02593 -52 With this approach, 20 different mutations were identified by SSCP in 56 patients (by deduction from the parents' genotypes, when applicable). For those patients in which no or only one mutation was identified by SSCP analysis, the entire coding region was sequenced. Four mutations were not revealed by SSCP analysis under the described 5 conditions, but were identified after sequencing. These mutations are 131 T- C (V44A), 303 C-G (NIOIK), 623 G-C (G208A) and 713 G-C (R238P) in exons 2, 4, 7 and 8 respectively. Also, the mutations 484 C-T (R162W) and 523 G-C (G175R) in exon 6 were not easily identifiable due to the presence of 2 frequent polymorphisms in the flanking regions of the exon. In total, mutations were found on Il l of 112 disease chromosomes 0 (99%). In one Japanese patient, heterozygous for one mutation at the genomic level, a second mutation could not be found. Southern blot analysis has excluded major rearrangements of the gene in this patient (data not shown). The mutational spectrum thus consists of 23 missense mutations and only one single base pair deletion (table 4-3 and figure 4-1). This deletion of a G at position 324 in exon 4 causes a frameshift and 5 premature stop (figure 4-2). Forty-three patients were heterozygous for the R141H mutation, but not a single patient was found homozygous for this mutation. Given the frequency of the R141H mutation (the allele frequency by counting is 40% in the 54 Caucasians in this study), 8 homozygotes are 0 expected in this sample on the basis of the equation of Hardy-Weinberg (X 2 =23.4, d. f. = 1, p <0.00001). The most frequent genotype is the R14lH/FI 19L combination (14 patients). One patient was homozygous for the relatively frequent F I19L mutation (figure 4-3), which is in accordance with Hardy-Weinberg equilibrium. and one was homozygous for the D65Y mutation. The latter is a rare mutation, and the disease allele was probably inherited from WO 98/49324 PCT/EP98/02593 -53 a common ancestor of the parents. In 2 patients, 3 different aminoacid substitutions were identified. In patients SN (and the affected sibling SJ), the A233T and T237R mutations are syntenic, as deduced from the 5 parental genotype. In patient LS, the T237M and A233T mutations are probably on the same chromosome: paternal DNA was not available for analysis but the mother carried only the T237R mutation. Discussion 0 Phosphonannoinutase deficiency and mutations in the PMV12 gene The localization of the PMM2 gene, encoding an active phosphonannomutase, on chromosome 16 and the identification of mutations in this gene that segregate with the disease (Matthijs et al. 1997a) gave conclusive support to the biochemical evidence that 5 phosphomannomutase deficiency is the basis for CDG type IA (Van Schaftingen and Jaeken, 1995; Jaeken et al. 1997b). This conclusion is now strengthened by the fact that in almost all cases with a documented phosphomannomutase deficiency, we found mutations in the PMM2 gene. In one case, a second mutation was not found, despite the fact that all exons were scrutinized by sequencing. The other mutation is probably in a regulatory 0 region of the gene. Recent work showed that phosphomannomutase deficiency was found in a majority, but not in all patients with CDGl. In these patients, there is no reason to search for mutations in the PMM2 gene. Accordingly, in the family in which linkage to chromosome 16 could be WO 98/49324 PCT/EP98/02593 -54 excluded (Matthijs et al. 1996), no phosphomannomutase deficiency was found in the propositus (unpublished results). Due to this heterogeneity, phosphomannomutase assays are still useful for diagnosis. The only drawback of these measurements is that in some cases, the activity found in fibroblasts or in lymphoblasts represents still up to about 30 % of the control value (see, for instance, patients 2, 25 and 39), whereas in other samples derived from the same patients or from other patients with the same genotype, the activity was less than 5% of the control activity. This problem has not been encountered in assays performed with fresh material such as leukocytes or liver. In the latter cases, a profound deficiency has always been observed. We have at present no explanation for these discordant findings. One possibility is that PMM 1 or the mutant forms of PMM2 become overexpressed after several passages in culture. Phosphomannomutase measurements are also useful for the identification of carriers. If cells from an affected child are not available, indirect evidence can be obtained from the phosphomannomutase activities in leukocytes from the parents. This has proven worthwhile in 2 cases with an urgent request for prenatal diagnosis, in which we derived information on the phosphomannomutase deficiency from the parents, before initiating the molecular analysis. In view of the genetic heterogeneity, prenatal testing should only be offered in families with a documented phosphomannomutase deficiency and mutations in PMM2. Thus far, we have combined enzymatic measurements and mutation analysis for prenatal diagnosis (Matthijs et al, 1998). The molecular analysis is the most dependable test for prenatal diagnosis.

WO 98/49324 PCT/IEP98/02593 -55 Characteristics of the mutations Of the 24 different mutations that were identified, 20 were detectable by SSCP, among which the most frequent ones, like the R141H, PIl3L, FlI9L, V231M and D188G. In practice, in only 1 of the 56 patients with mutations (patient MM, and the sibling MY), no 5 aberrant fragment was detected by SSCP analysis, while in the other patients at least one mutation was detected by SSCP alone. Given that the gene contains only 8 exons, the SSCP analysis represents a fairly simple, reliable and cost-effective approach for the molecular analysis of CDGlA patients. 10 Of interest is the fact that the mutations are unequally distributed between the different exons (see figure 4-1): 6 mutations were found in exons 5 and 8 whereas no mutation was found in exon 1. It is unlikely that the first exon would accommodate mutations more easily than other exons, since its degree of conservation is intermediate between those of exon 5 and exon 8 (Matthijs et al. 1997a). It is more probable that the difference in the mutation 15 rate among exons is due to their sequence context and position in the genome. All mutations except one are missense mutations. It is remarkable that in this collection of mutations, only a single frame-shift mutation has been observed. The Gl75R mutation might also affect splicing. Nineteen of the mutations that were found, affect residues that 20 are strictly conserved among PMMs (figure 4-1 and table 4-3). This supports the notion that mutations at these sites are detrimental to the function of the protein. Also, the comparative data in figure 4-1 suggest that the list of mutations presented here is not exhaustive. An alignment of PMM2 with PMM 1, yeast SEC53 and Candida albicans PMM has been published (Matthijs et al. 1997a). Enzymatic studies of some of the mutant WO 98/49324 PCT/EP98/02593 -56 proteins are underway. The R141H mutation is by far the most frequent (Caucasian) one. The D188G mutation seems to be restricted to Belgian (Flemish) and Dutch patients. The V44A mutation is 5 probably of Spanish origin, while the D65Y mutation, homozygous in a Portuguese patient, is also found in a French patient with Portuguese ancestors. The V129M mutation might well be of Italian origin. We have not tried to link the most frequently observed mutations to a common haplotype of the flanking polymorphic markers, but it seems that some mutations are old mutations that have been present in the different populations for 0 centuries. The R14lH mutation is caused by a CGC to CAC transition, but the equally likely CGC to TGC transition (R141C, which is found in a processed pseudogene, derived from PMM2 and located on chromosome 18p; see Schollen et al, 1998) has not been observed in patients. Thus, the R141H mutation may be an old mutation, like the frequent AF508 in cystic fibrosis (Morral et al. 1993). On the other hand, some mutations must 5 have occurred independently on different chromosomes in the different populations. For instance, the R123G mutation, shared by 2 Spanish and 2 Dutch patients, is syntenic with a polymorphism at nucleotide 324 in the coding region, and was also identified in a Japanese and a French patient, but without the polymorphism. The T237R and T237M mutations are caused by C to T transitions on opposite strands in the same CpG 0 dinucleotide. It is known that CpG-dinucleotides are hot spots for mutations. An interesting observation is that the corresponding CpGs have also been mutated in the processed pseudogene (Schollen et al. 1998). A complex genotype has been observed in 2 families. DNA from the parents has been used WO 98/49324 PCT/IEP98/02593 -57 to establish the phase of these mutations. The A233T mutation is respectively syntenic with the T237R and T237M mutations in patients SN/SJ and LS, whereas the T237R is present on the other allele in the latter family. Taken together, the alternative combination of the T237R and T237M mutations with the A233T mutation is remarkable, and the occurrence 5 of 2 different mutations in the same codon on the 2 chromosomes in patient LS is intriguing. Relationship between the mutations and the phenotype Very limited inferences can be made from the genotype-phenotype comparison. There is 10 no clear correlation between the phosphomannomutase activities and the genotype in 6 and 14 patients with the R141H/Pl 13L and R141H/FI 19L combinations respectively. It has previously been remarked that a great variability exists in the clinical expression among affected siblings (Jaeken et al. 1997a). One significant observation is the high mortality in the patients with the D188G/Rl41H genotype: 4 of 5 patients died before the age of 2, 15 while the fifth patient, now 10 years of age, is severely affected. On the other hand, the twin patients, described originally in Jaeken et al. (1980), are now 21 years old and relatively well, and have the R141H/PIl3L genotype. The R123G/PIl3L genotype is observed in a Spanish patient with pubertal development, which is not normally seen for females with CDGI. 20 Even in the most severe cases, the bulk of N-linked carbohydrate chains is present on serum glycoproteins. This would mean that some residual activity is required. The fact that mainly missense mutations are found, suggests that a total lack of PMM2 activity is incompatible with life. The lack of homozygotes for R141H supports this idea (see below). Alternative pathways to bypass phosphomannomutase activity and to generate mannosel- WO 98/49324 PCT/EP98/02593 -58 phosphate have not been described to date. Given the low frequency of CDGI, one would expect this recessive disorder to be most frequently encountered in consanguineous families. In our previous haplotype analysis on 15 patients, we have not encountered homozygosity for a disease-associated haplotype of linked markers (Matthijs et al. 1996). The molecular analysis revealed the significant lack of homozygotes for R141H. Thus, Hardy-Weinberg equilibrium is not reached, probably due to selection against this genotype. Probably, R141H is a severe mutation that is deleterious in the homozygous state, leading to fetal wastage, miscarriage or early death. It might also give rise to a different phenotype. In view of the fundamental role of PMM2 ) in the normal functioning of the cell, we hypothesize that it is lethal soon after conception. There is only 2 ways to address this question. One is to try to find families in which both parents are carriers of R14lH, to prove that the R141H/R14lH genotype is absent in their children, and to look for an association with miscarriages. Given that the gene frequency is very low, this is practically impossible. The alternative is to await the availability of a 5 transgenic mouse model to find out whether the homozygous knock-out is (embryonic) lethal or results in a different phenotype. Most other mutations must be relatively mild, because they are found in association with R141H. In the recent report on haplotype data from the Swedish group, Bjursell et al. 0 (1997) report at least 2 homozygous cases in their population. On the basis of the complete lack of R14l H homozygous patients in the present series, we speculate that the underlying mutation will be a milder mutation. In our series, patients were homozygous for the D65Y and F119L mutations. The D65Y is rare, but the FI9L is a frequent mutation, and the occurrence of one F l 19L/F l 19L patient in a series of 56 reflects Hardy-Weinberg WO 98/49324 PCT/EP98/02593 -59 equilibrium. The observation suggests that these are mild mutations. This might be reflected by the relatively high value of the residual phosphomannomutase activity in fibroblasts of the D65Y/D65Y patient. However, the same D65Y mutation, in combination with the R141H mutation was found in a patient with a severe phenotype, who died at a 5 very young age (patient 2). The Fl 19L/R141H is a particularly frequent genotype, thus the combination of the two most frequent disease mutations is not lethal. It remains to be investigated how these mutations affect the function of the protein. Nothing is currently known about the functional domains of phosphomannomutases. The availability 0 of a plethora of functional mutants will be of great value in the interpretation of structural data. References (Fourth Section) 5 Bjursell C, Stibter H, Wahlstrbm J, Kristiansson B, Skovby F, Str6mme P, Blennow G, Martinsson T (1997) Fine mapping of the gene for carbohydrate deficient glycoprotein syndrome, type I (CDGI): linkage disequilibrium and founder effect in scandinavian families. Genomics 39:247-253 0 Jaeken J, Vanderschueren-Lodeweyckx M, Casaer P, Snoeck L, Corbeel L, Eggermont E, Eeckels R (1980) Familial psychomotor retardation with markedly fluctuating serum proteins, FSH and GH levels, partial TBG deficiency, increased serumarylsulphatase A and increased CSF protein: a new syndrome ? Pediatr Res 14:179 WO 98/49324 PCT/EP98/02593 -60 Jaeken J, Van Eijk HG, van der Heul C, Corbeel L, Eeckels R, Eggermont E (1984) Sialic acid deficient serum and cerebrospinal fluid transferrin in a newly recognized syndrome. Clin Chim Acta 144: 245-247 5 Jaeken J, Carchon H (1993) The carbohydrate-deficient glycoprotein syndromes: an overview. J Inher Metab Dis 16: 813-820 Jaeken J, Schachter H, Carchon H, De Cock P, Coddeville B, Spik G (1994) Carbobydrate deficient glycoprotein syndrome type II: a defidency in Golgi localized N-acetyt 10 glucosaminyltransferase II. Arch Dis Child 71: 123-127 Jaeken J, Matthijs G, Barone R, Carchon H (1997a) Syndrome of the month: Carbohydrate-deficient glycoprotein (CDG) syndrome type I. J Med Genet 34:73-76 15 Jaeken J, Artigas J, Barone R, Fiumara A, de Koning TJ, Poll-Th6 BT, de Rijkvan Andel JF, Hoffmann G, Mayatepek E, Pineda M, Vilaseca MA, Saudubray JM, Schliter B, Wevers R, Van Schaftingen E (1997b) Phosphomannomutase deficiency is the main cause of carbohydrate deficient glycoprotein syndrome with type I isoelectrofocusing pattern of serum sialotransferrins. J Inher Metab Dis 20:447-449 20 Jaeken J, Casaer P. (1997) Carbohydrate-deficient glycoconjugate (CDG) syndromes: a new chapter of neuropaediatrics. Eur J Paediatr Neurol 1:61-66 Martinsson T, Bjursell C, Stibler H, Kristiansson B, Skovby F, Jaeken J, Blennow G, WO 98/49324 PCT/EP98/02593 -61 Stromme P, Hanefeld F and Wahlstrbm J (1994) Linkage of a locus for carbohydrate deficient glycoprotein syndrome type I (CDGI) to chromosome 16p, and linkage disequilibrium to microsatellite marker D16S406. Hum Mol Genet 3:2037-2042 5 Matthijs G, Legius E, Schollen E, Vandenberk P, Jaeken J, Barone R, Fiumara A, Visser G, Lambert M, Cassiman J-J (1996) Evidence for genetic heterogeneity in the carbohydrate-deficient glycoprotein syndrome type I (CDGl). Genomics 35:597-599 Matthijs G, Schollen E, Pardon E, Veiga-da-Cunha M, Jaeken J, Cassiman J-J, Van 0 Schaftingen E (1997a) Mutations in a phosphomannomutase gene, PMM2, on chromosome 16 in carbohydrate-deficient glycoprotein type I syndrome (Jaeken syndrome). Nat Genet 16:88-92 (Erratum published in Nat Genet, 1997, 16:316) Matthijs G, Schollen E, Pirard M, Budarf ML, Van Schaftingen E, Cassiman J-J (1997b) 5 PMM (PMMI), the human homologue of SEC53 or yeast phosphonannomutase, is localized on chromosome 22ql3. Genomics 40: 41-47 Matthijs G, Schollen E, Cassiman JJ, Cormier-Daire V, Jaeken J, Van Schaftingen E (1998) Prenatal diagnosis in CDGI families: beware of heterogeneity. Eur J Hum Genet 0 in press. Morral N, Nunes V, Casals V, Chill6n, Gim6nez J, Bertranpetit J, Estivill X (1993) Microsatellite haplotypes for cystic fibrosis: mutation frameworks and evolutionary tracers. Hum Mol Genet 2:1015-1022 WO 98/49324 PCT/EP98/02593 -62 Pineda M, Pavia C, Vilaseca MA, Ferrer I, Temudo T, Chabas A, Stibler H, Jasken J (1996) Normal pubertal development in a female with carbohydrate deficient glycoprotein syndrome. Arch Dis Childh 74:242-243 5 Pirard M, Collet JF, Matthijs G, Van Schaftingen E (1997) Comparison of PMM1 with the phosphomannomutases expressed in rat liver and in human cells. FEBS Lett 411:251-254 Schollen E, Pardon E, Heykants L, Renard J, Doggett NA, Callen DF, Cassiman JJ, Matthijs G (1997) Comparative analysis of the phosphomannomutase genes PMMl, PMM2 0 and PMM2<p: the sequence variation in the processed pseudogene is a reflection of the mutations found in the functional gene. Hum Mol Genet, in press Stibler H, Westerberg B, Hanefeld F, Hagberg B (1993) Carbohydrate-deficient glycoprotein (CDG) syndrome - a new variant, type III Neuropediatrics 24: 51-52 5 Stibler H, Stephani U. Kutsch U (1995) Carbohydrate-deficient glycoprotein syndrome a fourth subtype. Neuropediatrics 26: 235-237 Tan J, Dunn J, Jaeken J, Schachter H (1996) Mutations in the MGAT2 gene controlling 0 complex N-glycan synthesis cause carbohydrate-deficient glycoprotein syndrome type II, an autosomal recessive disease with defective brain development. Am J Hum Genet 59: 810-817 Van der Knaap MS, Wevers RA, Monnens L, Jakobs C, Jaeken J, Van Wijk JAE (1996) WO 98/49324 PCT/EP98/02593 -63 Congenital nephrotic syndrome: a novel phenotype of type I carbohydrate-deficient glycoprotein syndrome. J Inher Metab Dis 19:787-791 Van Schaftingen E, Jaeken J (1995) Phosphomannomutase deficiency is a cause of 5 carbohydrate-deficient glycoprotein syndrome type 1. FEBS Lett 377: 318-20 Wada Y, Nishikawa A, Okamoto N, Inui K, Tsukamoto H, Okada S, Tanigochi N (1992) Structure of serum transferrin in carbohydrate-deficient glycoprotein syndrome. Biochem Biophys Res Commun 189, 832-836 WO 98/49324 PCT/EP98/02593 64 Uu uU uU u uu 0 0 0 0 0 0 0 0 0 C) CD C) a) t W) V " U < C < < < < rD < < U UUCDCDHCDH u DHHCu< U U u u u < < a 0 DCDu u DI"u U 0 00 uCD <Cu u U Cd) 0 0 UU UU CD Qd) UU (DU H UD<DD U H H C <H0 << U U ru C u 0 - - < < < = uL -l trcli Ncrr r~ ii--.I Ime tot- WO 98/49324 65PCT/EP98/02593 u E UU E 00 cz > ~u V)~ -,: u ou UU U Eu p 21 .0 Z Z-66 Z666Z 4Z 0 0ZZ66z -u - - - - -- - - -N- -'O -f -f < - 00 cn m m m m--m-ON-ON------------------C7 --------- 0----7--C a, C7-, WO 98/49324 66 PCT/EP98/02593

-

0 E 0 c I-> 0~0 -: Cto -) - " 7 C) ) 0~~ ~ ~ z >z ~ > -, 01)) <- ~O < zN ( -- -- -o~ -7 c- , n , - 0 Mo o oc M mm on c mMT 0 "T 'T"t f WO 98/49324 PCTIEP98/02593 67 000 CAN -~ z Un -ct 5 C'4 C'Ar um n. (r0-m C) q C-4 z* E H -4 C l u0 0' S +1 0 Q ) CA4 00 ~-C w~0 CA IO 21 M -, : Irl,~~- if -1V1 r WO 98/49324 68 PCTIEP98/02593 0 0 0' 0' >~ 5 >~~~ ~ ~ ~ a z -IC -D .0 0 __ 00 Q u < 3 U0 U M ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ C 0n r-"tc 0r 0 ) CA.lr r -0C 4 C-4~~~~~~~ ~ ~ ~ ~ ~ ~ M~ V)'00 0,C- - , - -t0 7 7 It WO 98/49324 PCT/EP98/02593 -69 Legends to the figures (Fourth Section) Figure 4-1 Schematic diagram of the PMM2 cDNA, depicting the location of the various mutations found in the gene. The exons are indicated below. Because nothing is known at present about possible domains in the protein, the degree of similarity with the yeast SEC53, Candida albicans PMM and human PMM1 has been represented: dark bands stand for aminoacid residues that are perfectly conserved among the four proteins, while the dotted bands indicate a partial conservation; white residues are not conserved. Only mutation R238P affects a "white" residue. Mutations that have relatively frequently been encountered in this study, are given in boxes with bold lines. Figure 4-2 Identification of a single basepair deletion in exon 2 of PMM2 by SSCP and sequencing in a patient with CDG IA. In A, SSCP analysis shows an aberrant pattern for patient RMan. In B, direct sequencing using the fluorescent primer int4R reveals a heterozygous deletion of the G at position 324, which is the third base of codon 108. Downstream of the deletion site, the signal is a mixture of 2 overlapping sequences. Note that the sequence of the antisense strand is given. The effect of the deletion on the translational frame is shown in C; the frameshift results in a stop, 19 codons downstream of the deletion site (not shown). Figure 4-3 Sequence analysis revealed the occurrence of the F l19L mutation in the homozygous state. The results of the direct sequencing are shown for A) a normal control (homozygous for WO 98/49324 PCT/EP98/02593 -70 the normal sequence); B) the father of patient SN (heterozygous) and C) patient PG (homozygous for the mutation). Codon 119 is boxed.

Claims

1. A purified and isolated DNA molecule characterised in that it comprises a nucleotide sequence encoding human phosphomannomutase 2 (PMM2) protein or a part 5 thereof as shown in accompanying Figure 3-4.

2. A DNA molecule as claimed in claim I wherein it is a cDNA.

3. A DNA molecule as claimed in claim I or claim 2 wherein the nucleotide 10 sequence includes all coding and non-coding regions as shown in accompanying Figure 1 Ia.

4. The use of PCR primers derived from a DNA molecule as claimed in any of claims I to 3 to obtain genetic mutations in PMM2. 15

5. The use of oligonucleotide primers as shown in Table 4-1 to determine by SSCP mutations in the PMM2 gene.

6. A use as claimed in claim 4 or claim 5 wherein the DNA to be analysed is 20 genomic DNA or cDNA.

7. The use of a sequence element comprising at least nine, preferably at least twenty, continuous bases from any position within the PMM2 sequence indicated in claim 1 to analyse via hybridization on solid phase arrays or silicone chips unknown mutations WO 98/49324 PCT/EP98/02593 -72 in the PMM2 gene.

8. A DNA sequence characterised in that it is identified as a mutation in the PMM2 gene as shown in Table 4-2. 5

9. A DNA sequence characterised in that it is identified as a mutation in the PMM2 gene located in an exon as shown in Table 4-3 or in accompanying Figure 4-1.

10. The use of a sequence element derived from DNA as indicated in claim 1 0 or claim 9 on arrays or solid phase capture systems to detect expression of the PMM2 gene in a human cDNA sample.

11. The use of a sequence derived from DNA as indicated in claim 9 in the detection of mutant PMM2 genes. .5

12. The use of oligonucleotides derived from DNA as indicated in claim 9 in detection of mutant PMM2 in samples isolated from the body.

13. A use as claimed in claim 12 wherein the detection comprises PCR and/or 20 the use of solid phase immobilized oligonucleotides to capture a sequence as indicated in claim 9.

14. A use as claimed in claim 12 wherein a patient sample is immobilized on a solid support and an oligonucleotide as indicated in claim 9, labeled to facilitate detection, WO 98/49324 PCT/EP98/02593 -73 is applied so as to detect the presence of a mutation in the PMM2 gene.

15. The use of a sequence element derived from DNA as indicated in claim 8 to detect expression of cDNA for PMM2 expression in a model system or organism.

16. The use of a sequence element derived from DNA as indicated in claim 1 or claim 9 for the preparation of an expression system for PMM2 protein or a mutant thereof. D

17. A use as claimed in claim 16 wherein the expression system is bacterial, mammalian (preferably mammary gland), insect or plant.

18. The use of a protein prepared using an expression system as indicated in claim 16 or claim 17 in an assay for the evaluation of chemical or biological agents which 5 reduce or enhance the catalytic activity of the enzyme.

19. The use of a sequence element derived from DNA as indicated in claim 1 or claim 9, or of a protein prepared using an expression system as indicated in claim 16 or claim 17 as a therapeutic agent. 0

20. A use as claimed in claim 19 wherein the therapeutic agent is delivered either directly or via an ex-vivo approach, preferably in a viral or non-viral system.

21. The use of a sequence element derived from DNA as indicated in claim 1 WO 98/49324 PCT/EP98/02593 -74 or claim 9 in a transgenic model system for the purpose of drug discovery or development, or to find associated expression pathways.

22. The use of a sequence element derived from DNA as indicated in claim 1 5 or claim 9, or of a protein prepared using an expression system as indicated in claim 16 or claim 17 to produce an antibody, either polyclonal or monoclonal, for use in detection of PMM2 or a mutant thereof.

23. An assay for the detection of PMM2 or a mutant thereof characterised in that 0 it comprises using an antibody as indicated in claim 22 in a capture or competition method.

24. A biochemical assay for phosphomannomutase activity to identify potential carrier status characterised in that a value below 30% of normal is taken as an indication of the presence of a mutant gene, preferably the assay being carried out using a sample 5 generated from culture of trophoblasts. aminocytes, fibroblasts or leucocytes.

25. A genetic assay to determine the difference from PMM2/PMM2 pseudogene as shown in accompanying Figure 3-4.