Overlapping patterns of IGF2 and H19 expression during human development: biallelic IGF2 expression correlates with a lack of H19 expression

A subset of mammalian genes are expressed from only one allele in a manner dependent on the parental origin (Surani, 1991). To date, four genes have been shown to be parentally imprinted in the mouse: Igf2r (Barlow et al., 1991) and H19 (Bartolomei et al., 1991) are preferentially expressed from the maternal allele while Snrpn (Leff et al., 1992) and Igf2 (DeChiara et al., 1991) are primarily expressed from the paternal allele. In humans, IGF2 (Ohlsson et al., 1993; Giannoukakis et al., 1993; Rainier et al., 1993; Ogawa et al., 1993) and H19 (Rainier et al., 1993) have been shown to be parentally imprinted in the same direction as in the mouse. The IGF2 and H19 loci are of particular interest since, in addition to their reciprocal imprinting patterns, they are closely linked on chromosome 11 in humans. In addition, the spatiotemporal patterns of expression of these genes appear very similar during mouse embryogenesis (Lee et al., 1990; Poirier et al., 1991). These reciprocal patterns of imprinting appear to have been lost during the generation of a subtype of Wilms’ tumours since both parental alleles of IGF2 (Rainier et al., 1993; Ogawa et al., 1993) and H19 (Rainier et al., 1993) were expressed. Biallelic expression of Igf2 has, however, been observed in the choroid plexus and leptomeninges of the mouse during normal development (DeChiara et al., 1991).

IGF-II gene expression patterns appear to be evolutionarily conserved during mammalian development although notable exceptions have been reported. In man (Ohlsson et al., 1989a; Brice et al., 1989), mouse (Lee et al., 1990) and pig (unpublished observation), IGF-II expression appears to be first activated in trophectodermal derivatives after postimplantation. In the mouse, Igf2 expression first appears in the embryo proper at the late primitive streak/neural plate stage suggesting a role for IGF-II in the formation of the heart, derivatives of the foregut and cranial aspects of neural crest derivatives (Lee et al., 1990). During later stages of intrauterine development, a more extensive expression pattern is found, with organspecific similarities among rodents (Stylianopoulou et al., 1988a; Bondy et al., 1990; Lee et al., 1990; Ayer-LeLievre et al., 1991; DeChiara et al., 1991) and man (Han et al., 1988; Brice et al., 1989). Some species differences may exist, however, since no expression could be detected in rat adrenal cortex and pancreas and comparatively low levels of Igf2 transcripts were found in rat kidney while corresponding human embryonic tissues express IGF2 at high levels (Stylianopoulou et al., 1988a; Han et al., 1988; Brice et al., 1989). Considerable interspecies differences can be found during postnatal regulation of IGF2 gene transcription: In the rat, its expression declines after birth in all tissues except for choroid plexus and leptomeninges (Stylianopoulou et al., 1988b), while postnatal expression has been demonstrated in several other human tissues (reviewed by Rechler and Nissley, 1990). The liver is the likely source of serum IGF-II in adult humans (Jansen et al., 1990) produced from IGF2 transcripts derived from the human-specific P1 promoter (de Pagter-Holthuizen et al., 1988; Sussenbach, 1989). Three other promoter regions with counterparts in mouse and rat have been identified in the human IGF2 gene (Jansen et al., 1990). These have been referred to as ‘fetal promoters’, although adult tissues such as the pregnant and nonpregnant endometrium express P2-, P3- and P4-derived IGF2 transcripts (Glaser et al., 1992). Although considerably less information is available on the expression patterns of the H19 gene in comparison with IGF2, it has been noted that these two genes show strong similarities in their temporal and spatial patterns of expression (Lee et al., 1990; Poirier et al., 1991). This is of particular interest since these two loci appear to express opposite functions (DeChiara et al., 1991; Hao et al., 1993).

The purpose of this study was to examine the extent of overlap in the IGF2 and H19 expression patterns during some aspects of human development. Although these loci are coexpressed in a large variety of cell types within several tissues, the choroid plexus/leptomeninges and the ciliary anlage of the retina were notable exceptions, since they expressed the IGF2 but not H19 locus at detectable levels. Moreover, we show that IGF2 is monoallelically expressed in all tissues examined except for the choroid plexus/leptomeninges during early postnatal development. These results are discussed with respect to possible shared transcriptional control mechanisms between IGF2 and H19.

Human embryos from first trimester pregnancies were obtained from therapeutic terminations carried out at the Huddinge university hospital and the Karolinska hospital, with the permission of the local ethical committees. All specimens were, so far as could be assessed, morphologically and physiologically normal. Gestational age was assessed by ultrasound analysis, crown-rump lengths and patient information. In addition, the postmortem tissues of a newborn human patient who died 16 weeks following birth at 26 weeks of gestation were analyzed. The likely cause of death was a fulminant bacterial infection. Since the choroid plexus epithelium cannot be separated from the leptomeningal component, the sample of this specimen is referred to as choroid plexus/leptomeninges. The morphologically normal liver samples were obtained from 9 and 18 months old children undergoing operations for hepatoblastoma. The tissue specimens were either snap-frozen for DNA/RNA extraction or subjected to standard histological procedures.

Genomic DNA was prepared by repeated phenol/chloroform extractions of tissue cells lysed with 0.5% SDS and proteinase K (200 μg/ml final concentration). Total cellular RNA was prepared as has been described previously (Chomczynski and Sacchi, 1987).

The polymorphic (C-A)_n repeats of IGF2 were PCR-amplified as has been described (Ohlsson et al., 1993). The P1 promoter probe of IGF2 was generated by PCR-amplifying the clone pALP7 (kindly provided by Dr P. Schofield, Cambridge) using the following primers: CAATCTGCACCTTTCCTGAG and CTCACATACCTCAGCTC-CAG. These primers gave a product including exon 1 and about 285 bases of 5′ flanking sequence which was subcloned into the pCRII vector (Stratagene) to generate the pP1TE clone. Exon 4 and parts of exons 3 and 5 of H19 were PCR-amplified according to Zhang and Tycko (1992). The resulting fragment was subcloned into pCRII (Stratagene) to generate the pH19VT clone. The identity of each clone was verified by sequencing both ends using the Sequenase kit (USB).

³⁵S-labeled antisense IGF2 and H19 riboprobes for in situ hybridization analysis were generated from a 572 bp HinfI-PstI human cDNA IGF2 insert, cloned into pGem-3 (pHIGF2; Ohlsson et al., 1989b) and a 863 bp human H19 fragment (see above) inserted into the pCRII vector (pH19VT). Riboprobes from these templates were transcribed from supercoiled plasmids, yielding probes with similar specific activity (approximately 250 Ci/mmol) and size (600–700 bases). The exon-specific riboprobes (derived from the genomic clone hINS-3 (kind gift of Dr G. Bell) (see Fig. 4A) were generated from linearized Bluescript templates: P2, a PstI/SacI fragment of 340 bp, linearized with EcoRI; P3, an EcoRI/SalI fragment of 425 bp, linearized with EcoRI; P4, a SmaI/PstI fragment of 330 bp, linearized with SalI. Antisense RNA probe of the P1 promoter clone was generated from a HindIII-linearized template.

For RNase protection analysis, antisense ³²P-labeled P1-P4 promoter-specific riboprobes were generated from linearized subclones of exons 4, 5 and 6, as described above (specific activity approximately 900 Ci/mmol). An antisense ³²P-labeled riboprobe (specific activity of approximately 90 Ci/mmol) diagnostic for the polymorphic region of the exon nine of IGF2 was generated from the HindIII-linearized pPA1 plasmid (T7 polymerase) (Ohlsson et al., 1993). The simultaneous detection of IGF2/H19 transcripts by RNase protection used antisense riboprobes from XhoI-linearized pHIGF2 (covering bases 131 to 260 in human IGF2 sequence; accession number XO 7868 in GenBank) and Sacl-linearized pH19VT (covering bases 3036 to 3376 in human H19 sequence; accession number M32053 in GenBank) vectors, respectively. The specific activities of these probes were approximately 250 Ci/mmol.

For the identification of heterozygous alleles of IGF2, ³²P-labeled antisense RNA probes (100,000 cts/minute) generated from pPA1 template were annealed to 5 μl of PCR incubation mixture (corresponding to approximately 50 ng of IGF2 exon nine DNA) at 30 °C for 15 hours using the hybridization solution of the RNase protection kit (Ambion). For a corresponding RNase protection analysis of cellular RNA, the same amount of radiolabeled probe was annealed to 10 μg of total cellular RNA at 45 °C for 15 hours. The comparison of IGF2 and H19 transcript levels was performed as above by annealing 200,000 cts/minute of IGF2 and H19 RNA probes, generated from pHIGF2 and pH19VT templates, respectively, to 1.5 μg of total cellular RNA. The promoter usage was similarly analyzed by annealing 500,000 cts/minute of exon-specific, antisense ³²P-labeled riboprobes with 5 μg of total cellular RNA. Following removal of excess probe by RNase degradation according to the manufacturers protocol, the denatured samples were analyzed on 6% acrylamide-urea sequencing gels.

Serial 5 μm sections from formaldehyde-fixed and paraffin-embedded human embryos and fetuses were used for in situ hybridizations. No anomalies were seen and since hybridization patterns were concordant in all specimens they were accepted for evaluation of normal development. In situ hybridization analyses were performed as has been described previously (Ohlsson et al., 1989b). Sections were hybridized to ³⁵S-labeled riboprobes at 56°C overnight and washed stringently prior to RNase treatment as described previously (Ohlsson et al., 1989b). Following application of Ilford K5 photographic emulsion (diluted 1:1 in 2% glycerol) and exposure for 3–16 days at 4 °C, sections were counterstained with Mayer’s hematoxylin and mounted.

It has previously been noted that there is a strong similarity between the IGF2 and H19 gene expression patterns during murine embryogenesis (Lee et al., 1990; Poirier et al., 1991). To examine this issue in some detail during parts of human embryonic and fetal development, we analyzed the expression of H19 and IGF2 in adjacent sections of embryos/fetuses. Fig. 1 shows that there is a strikingly similar pattern of IGF2 and H19 expression in several human embryonic and early fetal specimens, with the particular exceptions of the choroid plexus, leptomeninges and the ciliary portion of the embryonic retina, which all express IGF2 but not H19. The lack of H19 expression in the choroid plexus appears to persist into at least early postnatal development (Fig. 1K). In contrast, H19 is comparatively more active than IGF2 (although showing the same spatial pattern of expression) in the epithelium of the small intestine (Fig. 1S,T), myocardium, epithelium of urinary collecting system and respiratory epithelium (data not shown). We conclude that the spatiotemporal patterns of IGF2 and H19 expression are very similar in the examined specimens, with the ciliary body, choroid plexus and leptomeninges as the most prominent exceptions.

Since IGF2 and H19 expression patterns did not overlap in the choroid plexus/leptomeninges which expresses both parental alleles of Igf2 in the mouse (DeChiara et al., 1991), we next addressed if IGF2 is biallelically expressed in the corresponding human tissues. RNase protection analysis of PCR-amplified polymorphic (C-A)_n repeats of exon nine of IGF2 (Ohlsson et al., 1993) showed that an early postnatal patient carried heterozygous IGF2 alleles (Fig. 2). When the same RNase protection approach was employed for the expressed IGF2 transcripts and compared with the polymorphisms of the PCR-amplified DNA, it was evident that IGF2 was monoallelically expressed (albeit at varying levels) in tissues such as the paraganglion, the pancreas, the thyroid, the thymus, the adrenals, the myocardium and the muscle of psoas (Fig. 2). We also conclude that the same parental allele, in all likelihood the paternally derived, was expressed in all of these tissues. In contrast, the choroid plexus/leptomeninges tissue expressed both parental alleles (Fig. 2). We conclude that human IGF2 is biallelically expressed in a tissue-specific manner in a similar or identical way to that of the mouse (DeChiara et al., 1991). To determine whether or not the choroid plexus/leptomeninges of the newborn patient also displayed the lack of H19 expression demonstrated for other specimens of earlier human development (Fig. 1), we examined the relative levels of IGF2 and H19 transcripts. Fig. 3 shows a simultaneous RNase protection analysis using probes encompassing part of exon 5 of H19 (see Zhang and Tycko, 1992) and the coding sequence of IGF2, respectively, of the same samples analysed above for the allelic usage of IGF2. The results again show that there is a correlation between the levels of H19 and IGF2 expression in all tissue samples analyzed except for the choroid plexus/leptomeninges which did not express H19 at a detectable level. Moreover, in situ hybridization analysis of an adjacent portion of the choroid plexus/leptomeninges sample used in the RNase protection analysis above shows that, while IGF2 is abundantly expressed in this tissue (Fig. 1J), we failed to detect a reliable H19 signal (Fig. 1K) except in some of the invading blood vessels (data not shown). Biallelic expression of IGF2 in the human choroid plexus/leptomeninges correlates, therefore, with a lack of H19 expression.

IGF2 transcripts originate from at least four different promoters (Sussenbach, 1989; Holthuizen et al., 1990). To address whether or not the tissue-specific biallelic expression of IGF2 reflected coordinated changes in the pattern of promoter usage, we analyzed P1-P4 promoter-driven transcription. This was performed by annealing ³²P-labeled exonspecific antisense RNA probes (Fig. 4A) to the same RNA samples that were analyzed for allele-specific expression (see above), followed by RNase protection analysis. Fig. 4B shows that the major band(s) for each protected promoter probe were similar or identical to the ones expected from previous studies (Sussenbach, 1989; Holthuizen et al., 1990). Hence, for the P2, P3 and P4 promoter probes, protected transcripts of about 340, 132 and 93-103 bases, respectively, could detected (Fig. 4B). We have on numerous occasions with a large variety of RNA samples observed additional less abundant transcripts such as those shown in Fig. 4B. The possibility that these may represent hitherto unreported IGF2 transcript species is currently analysed. The results show that the three ‘fetal’ promoters (P2, P3 and P4) were active in the choroid plexus, paraganglia and adrenal glands with minor variations in the promoter usage within each tissue sample (Fig. 4B). Conversely, low levels of P1-derived transcripts with the protected size of 116–118 bases, which is the expected (Sussenbach, 1989), could be detected only in choroid plexus/leptomeninges (the lanes depicted liver show promoter usage in a sample from an 9 month infant serving as a marker). Interestingly, in one informative infant liver specimen (18 months postnatally) with an active P1 promoter (data not shown), both IGF2 alleles are expressed (Fig. 4C).

To assess if a cell-type-specific promoter usage exists within the various organs during development, the expression of IGF2 in thin sections of human embryos was analysed by using exon-specific ³⁵S-labeled RNA probes. Although the probes revealed that the IGF2 promoter usage/splicing patterns appear to be more complex than previously realized (Fig. 4B), the different probes nonetheless uncovered an almost identical pattern of IGF2 expression. Typically, here represented by the retroperitoneal organs, the results show that P3 promoterderived transcripts were the most abundant IGF2 mRNA species in all tissues analyzed and were indistinguishable from the described cDNA probe pattern (Fig. 5B,C). The probes specific for the P2 and P4 promoters yielded less distinct but similar, if not identical hybridization patterns as the P3 promoter-specific probe (Fig. 5D,E). P1 promoter-derived transcripts evaded sensitivity of detection since these could not be detected by in situ hybridization analysis, even in infant and adult liver samples (data not shown).

The IGF2 and H19 loci are closely linked and reciprocally imprinted in both mouse and man. In addition, these two loci show striking similarities in their expression patterns during mouse embryogenesis. We show here that the IGF2 and H19 genes are extensively coexpressed during human embryonic/fetal development. The striking similarities in their expression patterns apply, therefore, also to humans. These results are consistent with the proposals that the imprinting processes and the overlapping expression patterns of IGF2 and H19 are coordinatedly controlled (Surani, 1991; Zemel et al., 1992; Brandeis et al., 1993; Bartolomei et al., 1993). Since it has previously been shown that Igf2 is not imprinted in the choroid plexus and leptomeninges in the mouse, we considered it interesting to investigate if these tissues express both parental IGF2 as well as H19 alleles during human prenatal development. Precedence for the biallelic expression of both IGF2 and H19 loci can be found in a subtype of Wilms’ tumour forms, although this lack of imprinting was considered to reflect a pathological condition (Rainer et al., 1993; Ogawa et al., 1993). We report here that, in one informative sample, IGF2 was biallelically expressed in the choroid plexus/leptomeninges but monoallelically expressed in all other internal organs examined. Surprisingly, the H19 locus was not expressed at detectable levels in the choroid plexus/leptomeninges (and the ciliary anlage of the embryonic retina) in embryonic/fetal as well as postnatal tissue samples. This is in contrast to the observations of Zhang and Tycko who showed that human fetal leptomeninges express H19 monoallelically (Zhang and Tycko, 1992). It is probable, however, that the H19 expression in the leptomeninges detected by the more sensitive RT-PCR method, originates from invading blood vessels. Since we present only one informative example of biallelic IGF2 expression in the choroid plexus/leptomeninges, the generality of the allelic IGF2 usage in this tissue could be challenged. However, it is the expected result since a similar observation was previously reported in the mouse (DeChiara et al.,1991).

The transcriptional unit of IGF2 encompasses some 32 kb (Sussenbach, 1989). Since the P1 and P4 promoters are approximately 24 kb apart (Sussenbach, 1989), it might be informative to discover whether or not these promoters are coordinately expressed. We show here that whenever IGF2 is expressed during human embryonic/fetal development, the P2, P3 and P4 promoters (separated by at least 5 kb) appears to be active although with some variation in relative strength. In contrast, we could detect expression of P1 promoter-derived transcripts in the choroid plexus/leptomeninges alone during early human postnatal development. These observations hint that the usage of the P2, P3 and P4 promoters but not of the P1 promoter can be coordinated irrespective of whether IGF2 is mono- or biallelically expressed. Although a similar correlation could be observed in a human infant liver specimen, such a link remain inconclusive, however, until the allelic usage for each individual promoter has been examined directly.

The mechanism of parental imprinting is poorly understood at present. The nature of the epigenetic mark on one or both parental alleles of imprinted loci has not yet been identified although CpG methylation has been implicated in the process of parental imprinting for both H19 (Ferguson-Smith et al., 1993) and Igf2r (Stöger et al., 1993). This link suffers, however, from the uncertainty of whether allele-specific CpG methylation patterns are the cause or consequence of the allele-specific expression patterns. The inactivation of X-linked genes, for example, precedes CpG methylation (Riggs and Pfeifer, 1992). It is interesting, however, that, while the chromatin structure of both parental IGF2 alleles appears accessible for nucleases (Sasaki et al., 1992), differences both in CpG methylation patterns as well as accessibility of nucleases in the chromatin between the parental H19 alleles in the mouse have been reported (Ferguson-Smith et al., 1993). The establishment of these differences may reflect independent or interdependent events. Based upon circumstantial evidence, it has previously been proposed that the expression and reciprocal imprinting patterns of IGF2 and H19 are coordinated (Surani, 1991; Zemel et al., 1992; Brandeis et al., 1993; Bartolomei et al., 1993). An overriding transcriptional control element for IGF2 is consistent with the apparently coordinated IGF2 promoter usage as discussed above. However, a common transcriptional control of IGF2/H19 would not fit with the lack of H19 expression in the choroid plexus/leptomeninges, which expresses IGF2 biallelically, unless some form of hierarchy is invoked so that IGF2 and H19 are not normally allowed to be expressed from the same chromosome. In such a scenario, biallelic IGF2 expression would not be consistent with simultaneous expression of H19. The locus control region of the β-globin locus is a precedent for a distal element (separated from the transcription units by up to 60 kb) showing developmentally controlled preferences for different transcription units within this locus (Epner et al., 1992). In this analogy, the expression patterns of IGF2 and H19 in the choroid plexus/leptomeninges could be established by a single event such as the repression of the maternal H19 allele (it should not be essential here whether or not the paternal and maternal H19 alleles are repressed by the same molecular mechanism except in the extreme possibility that the allele-specific expression of IGF2 is indirectly governed by the epigenetic imprint of H19). Other possibilities, such as the tissue-specific loss of or inability to recognize epigenetic imprints as well as cell-type-specific enhancer elements which are refractory to epigenetic imprints, cannot be ruled out at this stage. These alternatives may, however, require at least two independent events (activation and repression of both parental alleles of IGF2 and H19, respectively) to establish the IGF2/H19 expression patterns observed in the human choroid plexus/leptomeninges. Whatever the nature of the involved regulatory mechanisms, they should differ (at least in part) between mouse and man, since mouse embryonic choroid plexus/leptomeninges appear to express a parentally imprinted H19 (Svensson et al., unpublished data).

We are very grateful for valuable advice from Dr Gary Franklin and skilful technical assistance from Virpi Töhönen. This work was supported by the Swedish Cancer Research Foundation (RmC; grants 1955 and 3271), Swedish Natural Science Council (NFR; grant nrs B-Bu-4494-301-303) and research funds of the Karolinska Insitute.