Article Text

This article has a correction. Please see:

The imprinted region on human chromosome 7q32 extends to the carboxypeptidase A gene cluster: an imprinted candidate for Silver-Russell syndrome
  1. L Bentley1,
  2. K Nakabayashi2,
  3. D Monk1,
  4. C Beechey3,
  5. J Peters3,
  6. Z Birjandi1,
  7. F E Khayat1,
  8. M Patel2,
  9. M A Preece4,
  10. P Stanier1,
  11. S W Scherer2,
  12. G E Moore1
  1. 1Department of Fetal and Maternal Medicine, Institute of Reproductive and Developmental Biology, Faculty of Medicine, Imperial College, London, UK
  2. 2Department of Genetics, The Hospital for Sick Children, and Department of Molecular and Medical Genetics, University of Toronto, Toronto, ON, Canada
  3. 3Mammalian Genetics Unit, Medical Research Council, Harwell, Didcot, Oxfordshire OX11 0RD, UK
  4. 4Institute of Child Health, University College London, London, UK
  1. Correspondence to:
 L Bentley, Department of Fetal and Maternal Medicine, Institute of Reproductive and Developmental Biology, Faculty of Medicine, Imperial College, Hammersmith Campus, Du Cane Road, London W12 0NN, UK;{at}


Imprinted gene(s) on human chromosome 7q32-qter have been postulated to be involved in intrauterine growth restriction associated with Silver-Russell syndrome (SRS) as 7–10% of patients have mUPD(7). Three imprinted genes, MEST, MESTIT1, and COPG2IT1 on chromosome 7q32, are unlikely to cause SRS since epigenetic and sequence mutation analyses have not shown any changes. One hundred kilobases proximal to MEST lies a group of four carboxypeptidase A (CPA) genes. Since most imprinted genes are found in clusters, this study focuses on analysing these CPAs for imprinting effects based on their proximity to an established imprinted domain. Firstly, a replication timing study across 7q32 showed that an extensive genomic region including the CPAs, MEST, MESTIT1, and COPG2IT1 replicates asynchronously. Subsequently, SNP analysis by sequencing RT-PCR products of CPA1, CPA2, CPA4, and CPA5 indicated preferential expression of CPA4. Pyrosequencing was used as a quantitative approach, which confirmed predominantly preferential expression of the maternal allele and biallelic expression in brain. CPA5 expression levels were too low to allow reliable evaluation of allelic expression, while CPA1 and CPA2 both showed biallelic expression. CPA4 was the only gene from this family in which an imprinting effect was shown despite the location of this family of genes next to an imprinted cluster. As CPA4 has a potential role in cell proliferation and differentiation, two preferentially expressed copies in mUPD patients with SRS syndrome would result in excess expression and could alter the growth profiles of these subjects and give rise to intrauterine growth restriction.

  • Silver-Russell syndrome (SRS)
  • allele specific replication timing
  • carboxypeptidase A (CPA)
  • genomic imprinting

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

In mammals, a subset of autosomal genes are expressed according to their parental origin. This phenomenon is known as genomic imprinting and is thought to have evolved over at least 100 million years ago.1 Imprinted gene expression has been shown to be critical for early mammalian embryonic development, growth, and behaviour.

Dysfunctional imprinted gene(s) on human chromosome 7 have been postulated to be involved in the growth restriction associated with Silver Russell syndrome (SRS) as 7–10% of patients have maternal uniparental disomy (mUPD(7)).2 SRS is a common cause of growth failure in children with an incidence of approximately 1:3000 births, similar to that observed for Turner syndrome.3 Features include intrauterine growth restriction (IUGR), persistent short stature, characteristic facial features, asymmetry, and fifth finger clinodactyly.4–6 Most cases of SRS are sporadic with no established Mendelian inheritance pattern.

Most imprinted genes are found in clusters.7 To date, there are three imprinted regions distributed along the length of human chromosome 7. These are 7p11.3-p13 including growth factor receptor protein 10 (GRB10),8–11 7q21.3 including the paternally expressed gene 10 (PEG10) and epsilon-sarcoglycan (SGCE),12 and an imprinted gene cluster at 7q32 including the mesoderm specific transcript (MEST), MEST intronic transcript 1 (MESTIT1), and non-clathrin coat protein gene intronic transcript 1 (COPG2IT1).13–15

The human MEST gene is imprinted in fetal tissues and localised to chromosome 7q32.2. Two isoforms of the gene are transcribed using alternative first exons. Isoform 1 is paternally expressed in fetal tissues,13 while isoform 2 is biallelically expressed apart from in the placenta and possibly kidney where it is also paternally expressed.14,16 The non-clathrin coat protein gene (COPG2) is located immediately telomeric to MEST in a tail to tail orientation and appears to be biallelically expressed.15 However, COPG2IT1 is a transcript located within intron 20 of COPG2 that is paternally expressed in all fetal tissues examined.15 An additional untranslated antisense transcript has been identified for MEST, namely MEST intronic transcript 1 (MESTIT1).14,16 All of these transcripts provide evidence that there is an extensive imprinted cluster at 7q32. Even though most imprinted genes are found in clusters in the genome, this does not exclude biallelic expression of individual genes within such a cluster, such as the testis specific protein A14 (TSGA14), which is located directly centromeric to MEST and is biallelically expressed in human fetal tissues.17

Several lines of evidence have highlighted the region around MEST and instigated the search for imprinted genes in this region as candidates for SRS. Not only are 10% of SRS patients mUPD(7), but a single SRS patient has been reported with segmental mUPD(7)q31-qter.18 The only known imprinted gene within this region that appears to contribute directly to growth is MEST. A mouse knock out of Mest with a deletion of exons 3 to 9 show pre- and postnatal growth restriction but no asymmetry (one of the commonly found phenotypic characteristics of SRS patients) when paternally inherited.19 Several groups provided genetic evidence against a role of MEST in SRS, reporting the lack of sequence and epigenetic mutations in non-mUPD SRS patients.20,21 This suggests that other imprinted genes flanking MEST at 7q32 may play a role in the pathogenesis of SRS. The connection between imprinting, human 7q, and SRS is strengthened by the fact that the human region is orthologous to mouse proximal chromosome 6. A region in proximal 6, 6A3.2 to 6B3, defined by the breakpoints of the reciprocal translocations T(4;6)77H and T(6;13)6Ad has been shown to be responsible for growth restriction when maternally duplicated.22

The transcripts proximal to MEST and TSGA14 belong to the carboxypeptidase A (CPA) cluster of genes. The functions of the protein products of the carboxypeptidase gene family range from the digestion of food to the selective biosynthesis of neuroendocrine peptides.23 Metallo-CPAs are zinc containing exopeptidases, divided into either the N/E or A/B subfamilies based on the overall domain structure and amino acid sequence similarities. All members of the A/B subfamily contain about a 90 amino acid long N-terminal “pro” region that functions as a chaperone to assist with the folding of the active carboxypeptidase (CP) domain. Four members of the A/B subfamily, CPA1, CPA2, CPA4, and CPA5, were found to be clustered within an interval of 120 kb on human chromosome 7q32.24CPA1 and CPA2 (together with CPB1 on chromosome 3q24) are referred to as pancreatic carboxypeptidases. Although CPA1 and CPA2 are expressed most highly in pancreas, expression of the genes in other tissues was also reported.25CPA4 was identified as a gene upregulated in NaBu-treated PC-3 cells (an androgen independent prostate cancer cell line),26 suggesting the possibility that CPA4 protein may have functions in cell proliferation and differentiation. Although the enzymatic properties of CPA4 have not been reported, the deduced amino acid sequence encoded by the CPA4 gene shows a high degree of similarity to other CPA proteins. CPA5 was identified as a new member of the CPA gene cluster on 7q32.24 The gene has been shown to be strongly expressed in testis germ cells. The active site of CPA5 was predicted to cleave substrates with C-terminal hydrophobic residues, as do CPA1 and 2. The four CPA genes are only 100 kb proximal to the cluster of imprinted genes (MEST, MESTIT1, and COPG2IT1) on 7q32.13–16

As a part of systematic screening for additional imprinted genes in the vicinity of the MEST/COPG2 locus, we have examined the imprinting status of the four CPA genes.


Northern blot analysis

Multiple tissue northern blots containing adult human poly (A)+ RNAs (2 μg/lane) were purchased from Clontech. The cDNA fragments used as probes for CPA4 and CPA5 correspond to nt 1440 to 1693 of GenBank acc NM_016352, and nt 1315 to 1420 of NM_080385, respectively. Probes were labelled with [α-32P] dCTP using random priming. Hybridisation and washing were performed following the protocols described by Sambrook and Russell.27

Analysis of replication timing

Replication timing was analysed in S phase cells labelled by BrdU incorporation in a normal lymphoblastoid cell line according to the methods described by Hitchins et al.28 Methods used for probe preparation and FISH were based on those described by Harper et al29 and the protocols recommended by Vysis (Vysis, Richmond, UK). Minimums of 200 S phase nuclei were scored for one of three replication patterns: two singlets (single signals representing an unreplicated allele), a singlet plus a doublet (two signals less than two signal widths apart representing a replicated allele), or two doublets. The number of singlet/doublet nuclei were represented as a percentage of the total number of nuclei counted. This reflects the percentage asynchrony of a given genomic region.

Fetal tissue and maternal DNA samples for imprinting analysis

Fetal tissues were acquired following termination of pregnancy at Queen Charlotte’s and Chelsea Hospital (QCCH), London. A paired maternal peripheral blood sample was obtained for DNA extraction at the same time. Informed consent was obtained for each sample set. Local ethics approval for obtaining fetal and maternal samples for the study of growth related disorders was granted by the Research Ethics Committee of the Royal Postgraduate Medical School (96/4955). Fetal pancreas tissue was obtained from the MRC Tissue Bank at Hammersmith Hospital. All tissues were washed in sterile PBS, snap frozen in liquid nitrogen, and stored at −70°C. Genomic DNA was extracted from the fetal placenta and maternal blood using the standard phenol-chloroform technique.

PCR amplification and sequence analysis

PCR for genotyping and RT-PCR for expression analysis were performed according to standard protocols in 25 μl reactions supplemented with 1.5 mol/l betaine to increase specificity.27 PCR and RT-PCR products were purified using microCLEAN (Microzone Ltd). The purified products were sequenced directly using Big Dye terminator kit and resolved on a 3100 Genetic Analyser (Applied Biosystems).

Imprinting analysis in human fetal tissues

Imprinting analysis was carried out by single nucleotide polymorphism (SNP) analysis that requires heterozygous fetal genotypes to differentiate between expression from each of the parental genomes. A homozygous maternal genotype for the same SNP allows identification of the paternal allele since it is the only allele which could have been contributed by the mother. The other fetal allele is therefore paternal in origin. Analysis of fetal RNA by RT, RT-PCR, and DNA sequencing allows us to determine whether both or just one of these alleles is expressed.


A C/G polymorphism in the third coding exon (nt173 of BC005279) was found by EST comparison ( Fetal DNA for which matched pancreatic cDNA was available was screened for informative heterozygotes by sequencing genomic PCR products (table 1) with the forward primer used for amplification. RT-PCR products were amplified for informative sample sets. Allele expression was evaluated by sequencing products with the forward RT-PCR primer.

Table 1

Primer sequences, conditions, and expected product sizes for human genomic and cDNA PCR amplification used for SNP analysis of CP genes


A C/T polymorphism for SNP imprinting analysis was identified in exon 7 (nt631 of U19977) by screening the DNA sequence of exons of a set of normal DNA samples. Fetal DNA for which matched pancreatic cDNA was available was screened for heterozygote informativity by sequencing genomic PCR products (table 1) with a nested forward primer (gtgtgaggttgacatgtatc 5′-3′) and reverse PCR primer. RT-PCR products were amplified for informative sample sets. Allele expression was evaluated by sequencing with nested primer caagctacggcactttggac (5′-3′).


A G/T polymorphism in exon 9 (nt 914 of AF095719) was identified using the Celera database. Fetal DNA samples, for which matched maternal DNA and a range of fetal tissues were available, were screened for heterozygote informativity by sequencing PCR products amplified from genomic DNA (table 1) using a nested forward primer (caacccttgctccgaagtg 5′-3′). RT-PCR products were amplified for informative sample sets in as many tissue types as were available. Allele expression was evaluated by sequencing with nested forward primer gaggtgaaatcagtggtag (5′-3′) and the reverse, RT-PCR primer.


C/T and G/T SNPs were identified in exon 9 (nts1007 and 1014 of AF384667) using the Celera database and used to analyse isoform 1 of CPA5. Fetal DNA, for which matched maternal DNA and a range of fetal tissues were available, was screened for heterozygote informativity by sequencing genomic PCR products (table 1) using the nested primer agtgaacttcatcacagccc (5′-3′). RT-PCR products were amplified for informative sample sets in as many tissue types as were available. Allele expression was evaluated by sequencing products with nested forward (agtgaacttcatcacagccc 5′-3′) and reverse (aatgtactcgatccc 5′-3′) primers.

For isoform 3, a SNP was analysed which lies on exon 11 at nt1132 of AF384667 or nt 78 of the reverse compliment of AI191876. Fetal DNA, for which matched maternal DNA and a range of fetal tissues were available, was screened for heterozygote informativity by sequencing genomic PCR products (table 1) using nested primer aggatgcctctgtaccttgc (5′-3′). RT-PCR products were amplified for informative sample sets in as many tissues as were available. Allele expression was evaluated by sequencing these products with a nested reverse primer (cggaccatcatggagcacac 5′-3′).

Expression analysis of CPA4 using the pyrosequencing system

Pyrosequencing was used as a quick and accurate method of genotyping fetal and maternal genomic DNA samples and subsequently to quantify allelic expression in RT-PCR products amplified from cDNA generated from fetal tissues of informative subjects. Amplification of fetal and maternal genomic DNA was carried out according to standard PCR protocols using caacccttgctccgaagtgt (5′-3′) and 5′ biotinylated reverse primer tccctataaagctgcagagc (5′-3′) to genotype subjects for the same SNP in exon 9 that was used for analysis by DNA sequencing. The forward pyrosequencing primer aaacatgggaatttcaag (5′-3′) for genomic DNA was designed using SNP Primer Design from Pyrosequencing AB version 1.0.1. For RT-PCR 5′ biotinylated primer caacccttgctccgaagtgt (5′-3′) and reverse primer ctagctggatagacagtggt (5′-3′) flanking exons 10 and 11 were used with a reverse sequencing primer caggtcgatgaagc (5′-3′). Pyrosequencing was carried out according to manufacturers standard protocols (PyroSequencing AB).


Localisation, genomic organisation, and expression of CPA genes

The four CPA genes, CPA1, CPA2, CPA4, and CPA5, are clustered in an interval of 120 kb on 7q32 (CPA3 is not part of this cluster and has been localised to chromosome 3q24). The distance between the CPA genes and a cluster of imprinted genes (MEST, MESTIT1, and COPG2IT1) on chromosome 7 is approximately 100 kb (fig 1A).

Figure 1

Gene structure of CPA genes on human chromosome 7q32, (A) Seventeen known genes/transcripts including three imprinted genes (MEST, MESTIT1, and COPG2IT1) in the 1.5 Mb interval between D7S530 and D7S649. A common fragile site, FRA7H,41 is shown in a hatched box. The arrowheads show the transcriptional orientation of the genes. BAC/PAC clones for replication timing analysis are shown as horizontal bars at the bottom. The percentage asynchrony for each clone is shown in parentheses. B277J23 and B156M7 are BAC clones from the CITB library; the others are PAC clones from the RPCI-1 and -3 libraries.42 BAC/PAC end positions were determined by end sequences and are represented by vertical bars while those determined by PCR based mapping are shown by open circles. (B) A map for the four CPA genes (CPA1, CPA2, CPA4, and CPA5). Spidey (, an mRNA to genomic alignment program, was used to determine the exon-intron structures. The mRNA sequences used are NM_001868 (CPA1), NM_001869 (CPA2), NM_016352 (CPA4), and AY155230 (CPA5). The genomic DNA sequences used are AC024085 (for CPA2 and CPA4) and AC007938 (for CPA5 and CPA1). CPA1 consists of 10 coding exons; the other genes consist of 11 coding exons. Two CpG islands (shown by grey circles) were detected by GrailEXP ( one is in intron 10 of CPA2 (the ratio for observed/expected GC, 0.90; the percent GC, 54.83%); the other is 5′ upstream of exon 1 of CPA4 (the ratio for observed/expected GC, 0.71; the percent GC, 51.47%). (C) Exons and functional domains in the deduced amino acid sequences encoded by the four CPA genes. 5′ and 3′ UTRs are shown in open boxes. The coding regions are shown in light grey. The signal peptide sequences are shown in black boxes. Carboxypeptide activation peptides and zinc carboxypeptidase domains are shown in dark grey and mid grey boxes, respectively. The repetitive sequence (AluSp) detected by RepeatMasker ( in the 3′ UTR of CPA4 is shown by a bar with arrowheads (nt 1822 to 2110 of NM_016352).

Grail EXP was used to highlight potential CpG islands in the human genomic reference sequence AC024083 (RP11-190G13). No CpG islands were localised; however, two small regions were shown to have a slightly higher CpG dinucleotide content than the surrounding DNA (fig 1B, shown by grey circles). The first was located within intron 10 for CPA2 (102788-103093-PCT GC 54.83) and contained three HpaII (ccmgg) and six Hin6I (cgmcg) restriction sites. The second was located 650 bp upstream of the first exon CPA4 (109418-109648-PCT GC 51.47) and contains only two HpaII restriction sites. Both regions were subjected to RepeatMasker analysis, showing that these regions both align to SINE/Alu elements. The sequence from within intron 10 of CPA2 consists of two copies of SINE/Alu elements, whereas the second region represents only a single element. Preliminary methylation sensitive PCR showed that both regions are methylated in mUPD(7) and pUPD(7) lymphoblast DNA, as is expected for SINE/Alu elements.30

CPA1 has 10 coding exons and the others have 11 coding exons; exon 3 in CPA1 (234 bp) corresponds to two exons (3 and 4) in the others (fig 1C). The introns 3 of CPA2, CPA4, and CPA5 have progressively larger introns containing more repeat elements than one another. CPA4 has a longer 3′UTR containing a repetitive sequence (Alu) (fig 1C). Interestingly, the 3′UTR of mouse Cpa4 is also longer and contains repetitive elements (data not shown). Otherwise, the four genes have similar exon-intron structures. All intron-exon boundaries of the four genes conform to the splicing donor/acceptor consensus sequences (GT/AG). Intron phases are completely conserved among the four genes for all exon-intron boundaries. The putative amino acid sequences encoded by the four genes have a similar domain structure: a signal peptide at the N-terminus, carboxypeptidase activation peptide (pfam02244), and zinc carboxypeptidase domain (pfam00246) (fig 1C). CPA5 has a longer signal peptide than other members (fig 1C31). The similarity of amino acid sequences of two of the four proteins ranges from 51% to 68%.

In addition to 11 coding exons of CPA5, three non-coding exons at the 5′ end have been reported24 (un1 to 3, fig 2). By examining ESTs for the CPA5 gene (Unigene Hs.144699), we found two ESTs indicating the existence of alternative splicing isoforms at the 3′ region of the gene: AW072465 (without exon 10, referred to as isoform 2) and AI191876 (without exons 9 and 10, referred to as isoform 3) (fig 2). Skipping only exon 10 causes a frameshift, and the resulting novel protein sequence after the frameshift did not show significant sequence similarity with any known proteins. Skipping exons 9 and 10 does not shift the reading frame, resulting in a truncated form of the CPA5 protein. By Northern analysis using a multiple human tissue blot (Clontech) with a probe from the 3′UTR, two bands (a major 2.4 kb band and a minor 3.1 kb band) were detected in testis but not other tissues, consistent with the result for mouse Cpa5 by Wei et al24 (data not shown). The expression of CPA5 was detected by RT-PCR in many adult and fetal tissues at various levels (data not shown). In RT-PCR analysis using primers common for the three isoforms (a forward primer on exon 8 and a reverse primer on exon 11), isoforms 1 and 3 were detected in many fetal and adult tissues, but isoform 2 was undetectable in the tissues tested.

Figure 2

Three splicing isoforms for CPA5 at its 3′ end. The 11 coding exons for CPA5 (GenBank acc AF384667) are shown at the bottom. Full length cDNA sequences for IMAGE clones 5171382 (from brain, GenBank acc AY155229) and 5270617 (from testis, GenBank acc AY155230) contain additional three untranslated exons (un1 to 3) besides the 11 coding exons. The mRNA sequences containing all 11 coding exons are referred as “isoform 1”. Full length cDNA sequence for IMAGE: 2567523 (GenBank acc AY155228) skips exon 10 (referred to as isoform 2). EST AI191876 (from fetal heart) skips exons 9 and 10 (referred to as isoform 3). UTRs are shown in open boxes. Coding regions are shaded in light grey boxes. The regions corresponding to signal sequence, carboxypeptidase activation peptide (pfam02244), and zinc carboxypeptidase.domain (pfam00246) are shaded in black, dark grey, and mid grey, respectively. The absence of exon 10 in isoform 2 is deduced to cause a frameshift at amino acid 347 and creates 57 amino acids of new sequence showing no similarity with zinc carboxypeptidase.domain (shown in a hatched box).

A probe from 3′UTR of CPA4 (corresponding to nt 1440 to 1693 of GenBank acc NM_016352) was used to analyse a multiple tissue Northern blot for human adult tissues (Clontech, 7760-1). No signals were detected, a result consistent with that of Huang et al.26 However, using RT-PCR the expression was detected in a wide variety of tissues including placenta, testis, uterus, heart, brain, intestine, kidney, skeletal, and muscle (data not shown). CPA1 and CPA2 expression was evaluated by RT-PCR in fetal tissues. CPA1 was expressed exclusively in pancreatic tissue while CPA2 was ubiquitously expressed.

Imprinting analysis of CPA genes in human fetal tissue

Maternal and paternal alleles of imprinted genes replicate at slightly different rates.32 Generally, genes replicate on the homologous chromosomes in a synchronous manner with 10% of nuclei showing background asynchronous hybridisation patterns. In contrast, imprinted genes showed 25–40% asynchronous replication. This approach can therefore be used to assess imprinting across a large region. In a study by Bonora et al,31 asynchronous replication was shown for genomic clones containing the genes MEST and COPG2. In our study, replication timing was assessed using 11 PAC and BAC clones across a genomic region of approximately 1.5 Mb extending from NRF1 to beyond FRA7H. A PAC containing the cystic fibrosis transmembrane conductance regulator (CFTR) gene (PAC clone RPCI-1 22F22) was used as the synchronous control where, as expected, only 10% of nuclei showed the background level of asynchronous replication patterns. Each clone analysed representing a region centromeric to and including MEST displayed percentage asynchronies of 25% and above (a minimum of 200 nuclei were counted per probe). The PAC clone RPCI-3 502B1, spanning the fragile site FRA7H, showed 29.7% asynchronous replication. Apart from imprinted regions, fragile sites also replicate asynchronously, but not in a parental origin dependent way as is seen in imprinted regions.33 Replication timing of the BAC clone B277J23 and PAC clone 477N18 from RPCI-3, which lie between the imprinted asynchronous region and the fragile site, most likely corresponds to the telomeric end of the imprinted domain including MEST.

Allelic expression of individual CPA genes were analysed in fetal tissues by single nucleotide polymorphism (SNP) analysis. Fetal pancreas was the only tissue in which expression of CPA1 could be detected by RT-PCR. Three out of nine fetal DNA samples were heterozygous for the SNP to be analysed. Sequence analysis showed expression of both parental alleles (fig 3B).

Figure 3

Sequence electropherograms showing SNP analysis of CPA1, CPA2, and CPA4. (A) Sequence electropherograms across the polymorphic site (shown by arrows) in exon 7 of CPA2 shown for heterozygous fetal and maternal DNA and corresponding cDNA from various tissues as labelled. Expression of both alleles shows biallelic expression for this in all tissues analysed. (B) Sequence electropherograms across the polymorphic site in exon 3 of CPA1 shown for heterozygous fetal DNA and corresponding cDNA from pancreas tissue showing expression of both parental alleles. (C) Pyrosequencing pyrograms for CPA4 of fetal and maternal DNA and cDNA of various tissues available for F71. Pyrosequencing of genomic DNA show informativity with a heterozygous fetus and homozygous maternal sample. The G allele is therefore maternal in origin. cDNA alleles are expressed as percentages A and C alleles since a reverse primer was used for this section of the analysis. Taken that a 30% or less contribution of one allele indicates preferential expression, brain is classified as biallelically expressed while eye, heart, intestine, kidney, and lung are preferentially expressed from the maternal allele for this fetus. PCR primers and conditions are shown in table 1.

A C/T polymorphism in exon 7 was used to analyse CPA2 for imprinted expression. Four fetal DNA sample sets which included pancreas were tested and two were found to be heterozygous for the SNP. Both of these showed pancreatic expression of both parental alleles by RT-PCR and DNA sequencing. The same SNP was used to analyse allelic expression in sample sets with a wider range of fetal tissues available. Screening for heterozygotes in these fetal DNA samples showed five informative samples out of 22. Between two and five tissues were available per fetus with a total of seven separate tissues (eye, brain, intestine, lung, placenta, tongue, stomach) between them. Analysis by RT-PCR followed by sequencing showed biallelic expression in all tissues (fig 3A).

For CPA4 imprinting was assessed for a SNP in exon 9. Preferential expression of one or other parental allele was seen in 75% of the tissue samples analysed (data not shown). Owing to concerns of the sequencing approach not being sufficiently quantitative to measure preferential expression accurately, this analysis was repeated using pyrosequencing as a means of quantifying the molecular contribution of each parental allele to expression in a particular sample (fig 3C). For the purposes of this study preferential expression is defined as a skewed percentage contribution of one allele of 30% versus 70% of the other. Pyrosequencing information was obtained for nine fetuses in the first trimester with between two and six tissues available per fetus (table 2). Percentage expression for the four brain samples analyses varied from 39/61% to 45/55%, all of which are classified as biallelic according to our definition of preferential expression. Tongue samples were available for F47 and F65, which showed percentage expression of the C allele of 32 and 33% respectively. These, as well as heart samples for F34 and F47, lie within 3 percentage points of the preferential expression cut off. Of the remaining data, 80% of the data may be classified as preferentially expressed according to the definition stated above. It is important to note that since the maternal genotypes were homozygous for the G allele of the SNP, the G allele is maternal in origin for fetuses 50, 58, 66, and 71 (table 2). For F34, F46, F47, F65, and F67, this is not likely to be the case since the T allele is preferentially expressed and therefore according to the pattern of the data, maternal in origin. Each of the allelic expression percentages of tissues, shown in bold, highlights what we believe to be preferential maternal expression.

Table 2

Pyrosequencing results of CPA4 expressed as percentage contribution of each allele to PCR products analysed. Genotyping of fetal (F) and maternal (M) DNA is expressed as percentages of G and T alleles by sequencing with a forward primer. Results showing informativity for fetal and maternal genotypes are highlighted in bold. cDNA allelic expression is shown as a percentage of C and A alleles since the pyrosequencing primer was in the reverse orientation. For the cDNA expression results show those samples where one allele represents 30% or less of the total expression are in bold

For CPA5, 48 cycles of RT-PCR were required to detect expression in fetal tissues. For each isoform, sequence analysis showed both maternal and parental monoallelic expression and biallelic expression including subjects where each of these patterns was found in different tissues. The data available for each of the isoforms of CPA5 do not allow a clear pattern of inheritance to be determined. This may be related to the high number of cycles of RT-PCR required for amplification and is not likely to be of biological significance.


Four CPA genes are localised 100 kb proximal to a cluster of imprinted genes (MEST, MESTIT1, and COPG2IT1) on 7q32.13–15 Despite CPA1, CPA2, CPA4, and CPA5 having similar domain structures and exon-intron structures, expression patterns and levels are not uniform between genes. CPA1, for example, is only expressed in pancreatic tissue by RT-PCR. Expression levels of CPA1, CPA2, and CPA4 are much higher than that of CPA5, which requires 48 cycles of RT-PCR amplification for detection. At this level of amplification CPA5 may not be functional in the tissue types analysed.

It had previously been reported that MEST replicates asynchronously.11 In this study, replication timing was used to assess potential imprinting effects across a large genomic region around 7q32 with a series of 11 PAC and BAC clones mapped across the region. In this way it has been possible to define the telomeric boundary of asynchronous replication to between COPG2IT1 and TSGA13. Definition of synchrony versus asynchrony at this boundary is not as clear as expected since it is influenced by asynchronous replication of the fragile site. The entire region as far as the nuclear respiratory factor 1 (NRF1) gene, in a centromeric direction relative to MEST, replicates asynchronously. Since imprinted genes tend to be found in clusters, the prime candidates for investigation for imprinting effects are the CPA genes at 7q32 owing to their proximity to MEST.

SNP analysis in human fetal tissues showed biallelic expression of CPA1 and CPA2. CPA4 showed predominantly preferential expression of the maternally derived allele in a wide range of human fetal tissues except the brain where biallelic expression was found.

Polymorphic imprinting independent of gestational age has been previously reported. For the Wilms’ tumour suppressor gene (WT1), polymorphic imprinting including biallelic and maternal expression has been seen in placentae while lymphocytes and fibroblasts have displayed polymorphic biallelic and paternal expression.34–36 Similarly, polymorphic imprinting has been shown for insulin-like growth factor II (IGF2) in blood cells37 as well as for the insulin-like growth factor II receptor (IGF2R)38 and the serotonin-2A (5-HT2A) receptor gene (HTR2A).39 Throughout these reported studies of unusual imprinting patterns, only isolated samples have been described where one allele is partially repressed resulting in preferential expression. In this study, however, CPA4 SNP analysis has shown a novel pattern of imprinting where the maternally inherited allele is consistently preferentially expressed. Since CPA4 has been suggested to have a role to play in controlling cell proliferation and differentiation, it may be that a finely controlled balance of expression is required for normal growth. If CPA4 were to act in limiting growth, then over-expression in a subject with mUPD(7) could have a growth restriction phenotype such as that seen in SRS. This is in agreement with the conflict theory where maternally expressed genes have a growth restricting function and paternally expressed ones act as growth stimulants.40

For SNP analysis of CPA5, 48 cycles of RT-PCR was required to detect expression of the two transcripts (isoforms 1 and 3) expressed in fetal tissues. This number of cycles of exponential amplification would theoretically result in 1.4 × 1014 amplicons per original copy of template cDNA. According to Sambrook and Russell,27 48 PCR cycles can amplify one target molecule to 10 ng of DNA (based on a 200 ng PCR product) when the amplification efficiency is 65–100%. Since in practice more than 10 ng of product would be required, the primer efficiency would have to be somewhat greater than 65%. It is therefore possible that these erratic imprinting results are the result of amplification of a single copy of gene specific cDNA simply because only a single copy was present in the template used for amplification and not as a result of allele specific expression. These results for CPA5 are therefore not likely to be a true reflection of imprinting expression and are simply an artefact of the limitations of the techniques used.

It is clear that the CPA gene family is differentially imprinted and the gene products have different biological functions. It is interesting, despite biallelic expression for CPA1 and CPA2, that CPA4 is maternally preferentially expressed in a broad spectrum of fetal tissues but not the brain. It is interesting to speculate that a double dose of a gene that is maternally expressed and important in growth and proliferation might cause growth restriction and may be in part accountable for the SRS phenotype. Mutation screening of CPA4’s 11 exons is being carried out on DNA from a large cohort of SRS patients.


The first two authors contributed equally to this work.

Data access. Sequence data from this article have been deposited with the GenBank Data Library under Accession No AY155229 for full length CPA5 cDNA sequence from IMAGE clones 5171382 (brain), AY155230 containing all 11 exons as well as three untranslated exons (isoform 1 from testis, IMAGE: 5270617), and AY155228, a full length cDNA sequence for IMAGE clone 2567523 which skips exon 10 (isoform 2).


Supplementary materials

  • Publisher Correction

    The Ackowledgement list from the artiocle was omitted from the published version. They are shown here:

    We thank Dr Katariina Hannula and Professor Juha Kere forpreliminary allelic expression analysis of CPA genes in lymphoblastoid cell lines. We also thank the Dunhill Medical Trust, the WellcomeTrust (LB, ZB and FK), the MRC (DM, JP, CB) and the Canadian Institutes of Health Research (CIHR) for supporting this research. KN is a Research Fellow of the CIHR and SWS is a Scientist of the CIHR and the International Scholar of the Howard Hughes Medical Institute. All animal studies were carried out under the guidamce issued by the Medical Research Council (MRC) in The Use of Animals forMedical Research (July 1993) and Home Office Project Licence No30/1517.

Linked Articles

  • Correction
    BMJ Publishing Group Ltd