Although a great deal is known about the structure and molecular pathology of the human haemoglobin genes it is still not clear how their differential expression during normal development is regulated. As well as being of considerable interest to developmental geneticists, this problem has important practical implications. Variability in the expression of the foetal globin genes plays a major role in modifying the clinical course of some of the common genetic disorders of adult haemoglobin production. If it were possible to prevent the switching off of foetal haemoglobin production after the neonatal period, or to reactivate it even partially, we would have an extremely valuable approach to the management of these conditions, which are globally the commonest single gene disorders.

Here I shall summarize what has been learnt from the experimental systems that are being used to study the regulation of the developmental changes in globin gene expression. It will be possible to touch on only those areas that seem to be of particular promise for future work. Several recent reviews cover human haemoglobin genetics and the developmental biology of haemoglobin in more detail (Wood & Weatherall, 1983; Collins & Weissman, 1984; Orkin & Kazazian, 1984; Weatherall, Higgs, Wood & Clegg, 1984; Weatherall & Wainscoat, 1985); original references to much of the experimental work described here will be found in these articles.

The structures of the different haemoglobins that are synthesized during embryonic, foetal and adult life are summarized in Fig. 1. They are all tetramers consisting of two pairs of unlike peptide chains, each associated with a haem molecule. Normal adults have a major haemoglobin, haemoglobin A (α2β2), and a minor component called haemoglobin A22 δ2). The main haemoglobin in foetal life is haemoglobin F, which has α chains combined with γ chains (α2 γ2)-It is a mixture of two different molecular forms that differ only by one amino acid in their γ chains, glycine or alanine at position 136; the γ chains that make up these two types of foetal haemoglobin are thus referred to as Gγ and Aγ. In embryonic life there is yet another series of haemoglobins in which the α chains are replaced by g chains and the γ and β chains by ε chains.

As shown in Fig. 1 the globin genes are organized in two families, an α-like gene cluster on chromosome 16 and a β-like cluster on chromosome 11. Within each complex the genes, together with several inactive pseudogenes, are all in the same 5’ to 3’ orientation and are arranged in the order in which they are expressed at different stages of development. However, comparison with other vertebrate species suggests that it is unlikely that there is any general relationship between gene order and developmental expression.

The β-like genes are distributed over approximately 60 kb (103 bases) and are arranged in the order 5’-ε-Gγ—Aγ— Ψβ’.– ε–β –3’ The α-like genes form a smaller cluster on chromosome 16, in the order 5’-ς -ς Ψ Ψ αα2–α1–3’The Ψ β, Ψ ς and tpagenes are pseudogenes. The position of the introns is shown in Fig. 1. The 5’ flanking regions of each of the genes contain two regions of homology. One, the ATA box, is 20–30base-pairs (bp) upstream from the RNA initiation site; the other, the CCAAT box, is 70—90 bp upstream from this site. Each a gene is located within a region of homology approximately 4 kb long, interrupted by two small non-homologous regions. The exons and the first introns of the two a globin genes have identical sequences. The two g genes are also highly homologous. Like the α1 and α2 genes, the Gγ and Aγ genes appear to be virtually identical, reflecting a process of gene matching during evolution. In fact, the Gγ and Aγ genes on one chromosome are identical in the region 5’ to the centre of the large intron, yet show greater divergence in a 3’ direction. At the boundary between the conserved and divergent regions there is a block of simple sequence, which may be a ‘hotspot’ for the initiation of recombination events that lead to unidirectional gene conversion.

Several classes of repetitive sequences have been identified in the εγ δβ globin gene cluster. There are single Alu repeat sequences upstream from the γ globin genes and from the ft genes, and inverted pairs of Alu sequences upstream from the ε and <5 genes and downstream from the β globin gene. The three inverted pairs are orientated tail to tail with about 800 bp of non-repetitive DNA between them. The second major class of repeat sequences belongs to the Kpn family. One copy lies downstream from the globin gene; another between the ε and y genes. The latter region, over 6 kb in length, has been sequenced and at the end near the y globin gene has strong homology with the retrovirus long-terminal repeat (see Collins & Weissman, 1984).

The embryonic haemoglobins are synthesized mainly during the period when erythropoiesis is confined to the yolk sac. Throughout the rest of foetal life the liver and spleen are the main sources of red cell production, although the marrow starts to produce red cells during the second trimester and becomes the major erythropoietic site during later foetal life.

Recently, the patterns of globin chain production at very early stages of embryonic development, during the transition from yolk sac (primitive) to hepatic (definitive) erythropoiesis, have been analysed (Peschle et al. 1984, 1985). During the 4th to 5th week çand ε chains and very small quantities of γ chains are synthesized. During the 6th to 7th week a, ç, E, Gγ and Aγ chains are produced by the remaining primitive erythroblasts, and α, ε, Gγ and Aγ chains by the definitive line. By the 7th to 8th week ε and ç chain synthesis is no longer detectable and the main globin chains synthesized are α, Gγ and Aγ; β chain production is just detectable at this time and gradually increases, so that at about 10 weeks it constitutes about 10% of total non α chain production. Thus there appears to be a slight asynchrony in the switch from S’to a compared with εto γ chain production; the g-+ α chain transition is completed slightly earlier.

From the 10th to the 33rd week of gestation the main globin chains produced are α, Gγ> Aγ and β. Assessment of the output of the two linked α globin genes by mRNA analysis suggests that they are expressed in the ratio α2/α1 of 1·5—3·0/1 throughout foetal life (Liebhaber & Kan, 1981; Orkin & Goff, 1981); this does not change during development and is the same as that observed in normal adults. The relative rates of G γ and A γ chain production are also constant throughout foetal life at a G γ /A γ ratio of approximately 3/1 (Nute, Pataryas & Stamatoyannopoulos, 1973). Between the 32nd and 36th week of gestation the relative rate of α chain synthesis increases and that of y chain production declines, so that at birth β chain synthesis constitutes approximately 50% of non-α chain synthesis. After birth the level of y chain production declines steadily and that of α chain production increases ; at the end of the first year y chain synthesis reaches the low level characteristic of adult life. During the first few months of life the G γ /A γ ratio changes from 3/1 to 2β (Schroeder et al. 1972). Delta chain production has been observed as early as 32 weeks gestation; ô chain activation lags behind that of β chains, and the adult β/ô chain synthesis ratio is only reached at about 4–6 months after birth.

Although there has been extensive debate about the intercellular distribution of different haemoglobins during development it is now believed that the transition from embryonic to foetal and foetal to adult haemoglobin production occurs within the same erythrocyte populations. This conclusion is also consistent with recent studies of the patterns of γ and α chain production in red cell colonies grown from foetal and neonatal blood. It is also clear that the type of globin chains produced at different stages of development is not related to the site of erythropoiesis; both çand ε chains are synthesized in both primitive and definitive cell lines (Peschle et al. 1984, 1985) and the switchover from γ to β chain production occurs synchronously throughout the liver, spleen and bone marrow during the later stages of foetal development (Wood & Weatherall, 1973; Wood et al. 1979). Furthermore, the transition from y to α chain synthesis is related closely to gestational age and not to birth ; premature infants continue to synthesize relatively high levels of y chains until about 40 weeks gestation (Bard, 1975).

Thus the various developmental haemoglobin transitions occur within the same cell populations, are synchronized between the changing sites of erythropoiesis during development, and are closely related to the gestational age of the foetus. These changes in haemoglobin constitution are associated with other developmental modifications of the red cell, particularly the switching on of one of the carbonic anhydrase isozymes and several alterations in surface antigens.

Normal adults produce small amounts of haemoglobin F, which range from 0·3 to 0·8% of the total haemoglobin. Analysis of the intercellular distribution suggests that this is confined to a small population of adult red cells, which for this reason are called F cells, although they also include large amounts of haemoglobin A. The relative proportion of F cells is remarkably constant in different individuals and appears to be genetically determined, though how many genes are involved is not clear. The relative number of F cells increases during rapid regeneration of the marrow after periods of transient aplasia. Studies of disorders such as polycythaemia vera or chronic myeloid leukaemia suggest that F cells are not clonally derived but arise from the same stem cell pool as adult haemoglobin-containing red cells; the apparent restriction of γ chain production to a small proportion of adult red cells may be an artefact of the methods used to assess the intercellular distribution of foetal and adult haemoglobin (see Wood & Weatherall, 1983).

Because so little is known about the developmental regulation of gene expression, and the lack of good experimental models with which to investigate this problem, our current approaches to the analysis of gene switching are, of necessity, indirect. Areas of study that are providing some information on this question are summarized in Table 2.

Changes in globin gene structure during development

Several aspects of the structure of the globin gene complexes have been studied in an attempt to understand the mechanism of the differential expression of their constituent loci during development. Comparisons of the primary sequences of the individual genes and their flanking regions have shown that, in general, they rapidly lose homology upstream and downstream of the transcribed sequences except in the case of those that have diverged recently. However, interspecies comparisons have not identified any sequences either 5’ or 3’ to the structural genes that might be candidates for regulation of their differential expression during development. I shall return to some very recent studies of globin gene mutations that are relevant to this question in a later section. There is no evidence that there are any major rearrangements of the globin gene clusters at different stages of human development.

The globin genes are hypomethylated in tissues in which they are expressed and are differentially methylated at different stages of development. It appears that at all stages of ontogeny the β-like globin genes show a strong correlation between their methylation state and expression. Similarly, there is a strong tissue and agedependent relationship between their differential sensitivity to nuclease digestion. DNase I hypersensitive sites have been found 5’ to the Gγ, Aγ, ô and β genes in foetal liver haemopoietic tissue but only 5’ to the δ and β genes in adult haemopoietic tissues. These changes are presumably due to alterations in chromatin structure, both around the cluster reflecting its potential expression in erythroid cells, and within a cluster as each gene is activated at different times during development (Collins & Weissman, 1984).

5-Azacytidine, a cytidine analogue that is incorporated into DNA but cannot be methylated, appears to be able to activate β gene expression in adult animals and in humans (DeSimone, Heller, Hall & Zwiers, 1982; Ley et al. 1982). Demethylation of the β chain genes was observed in erythroid cells after azacytidine treatment of humans and experimental animals, although this was also true of the ε genes in the case of humans, yet the latter were not expressed. These studies also suggest that hypomethylation may be necessary for expression. I shall return to this question later.

Mutations associated with persistent y chain synthesis in adult life

The mutations that are characterized by persistent foetal haemoglobin production are summarized in Table 3. The most important are the P thalassaemias and P chain haemoglobin disorders such as sickle cell anaemia, and a family of conditions that are characterized by persistent foetal haemoglobin synthesis without any major haematological abnormalities, hereditary persistence of foetal haemoglobin (HPFH). Gene-analysis studies of HPFH have shown that the condition can be divided into deletion and non-deletion forms. More recently it has been found that the latter group can be subdivided into conditions in which the genetic determinants are linked to the β globin gene cluster and those in which they segregate independently from the cluster.

Sickle cell anaemia and P thalassaemia

The factors that are involved in the production of elevated levels of foetal haemoglobin in the blood of individuals with these conditions are extremely complex. Haemoglobin F protects against sickling. In β thalassaemia, cells that produce y chains are at an advantage since the latter combine with excess α chains; red cell precursor destruction in this disorder results from the deleterious effect of excess a chains that accumulate due to defective β chain synthesis. Thus in both of these disorders red cell precursors or mature red cells that have the capacity for producing γ chains undergo intense selection, either in the marrow or in the peripheral blood. On the other hand, it is equally clear that genetic factors are also involved in haemoglobin F production in these conditions. In some individuals with sickle cell anaemia or β thalassaemia, in whom unusually high levels of haemoglobin F afford protection from the effects of the disease, it is possible to find normal or heterozygous family members with increased levels of haemoglobin F. Thus it appears that a gene (or genes) for heterocellular HPFH (see below) is segregating in these families. This is not always the case, however.

Another approach to this problem has been developed recently. Scattered throughout the P globin gene cluster there are a number of restriction fragment length polymorphisms (RFLPs), which can be used as genetic markers for following mutations of the β globin genes (Antonarakis, Boehm, Giardina & Kazazian, 1982). It has been found that particular arrangements of these RFLPs (haplotypes) may be associated with an unusually high production of haemoglobin F in individuals with sickle cell anaemia or β thalassaemia (Wainscoat et al. 1985a). This suggests that there may be a genetic determinant within or linked to the β globin gene cluster, which, since these haplotypes are not associated with increased haemoglobin F production in symptomless heterozygotes, results in an unusually high level of y chain production in states of increased erythropoiesis. The only clue to the nature of this determinant is the recent observation that an alteration in the relative amount of G γ to A γ chain production in individuals with sickle cell anaemia, and possibly an increased capacity for γ chain synthesis, may be associated with a single base change, C → T, at position —158 in the G γ globin gene (Gilman & Huisman, 1984). I shall return to the significance of this finding in a later section.

The ôp thalassaemias and deletion forms of HPFH

These conditions are all characterized by long deletions of the γ <5β globin gene cluster. Their rather daunting nomenclature is explained in the legend to Fig. 2. They constitute a spectrum of disorders in which absent γ chain production is compensated by persistent γ chain synthesis. If compensation is more or less complete the condition is haematologically normal and is called HPFH; if there is less efficient γ chain synthesis, and hence unbalanced globin chain production, the condition is called <5/8 thalassaemia. However, in all these disorders there is an absolute increase in y chain production in adult life that cannot be accounted for by cell selection. This suggests that the deletions that cause them must be responsible for the high output of γ chains.

The different deletions that produce δ β thalassaemia and HPFH are summarized in Fig. 2. A question of major interest is whether a comparison of their site and size can explain the difference in phenotype between HPFH and δ β thalassaemia and hence tell us anything about the position of putative regulatory regions in the y<5/8 gene cluster. A comparison of particular interest is the 5’ extent of the deletions that cause either (<5β)° thalassaemia or (<5β)° HPFH in Black populations (Fig. 2). These deletions end within lkb of each other in the Alu repeat region 5’ to the δ globin gene. The deletion that causes HPFH ends in the middle of the upstream Alu repeat while that which causes δ β thalassaemia ends 1 kb downstream from the latter in the other Alu repeat (see Collins & Weissman, 1984). Is this region involved in the regulation of γ and β chain synthesis during development? While this may be the case, the fact is that both these deletions cause considerably elevated levels of γ chain production in adult life ; the difference between them is only a matter of degree. Furthermore, a Greek family has recently been described in which homozygotes have the clinical picture of a mild form of β thalassaemia and heterozygotes have either normal or marginally elevated haemoglobin F levels, and yet this condition results from a deletion that removes the entire region occupied by the Alu repeat sequences (Wainscoat et al. 1985b).

Another interpretation of these different phenotypes is that they do not depend directly on the region of DNA that is deleted but rather on the particular sequences that are brought into apposition to the β globin gene complex by the deletion.

Perhaps, it has been argued, in some forms of HPFH sequences are brought in from 3’ to the β globin gene complex that act as cis enhancers, thus allowing expression of the foetal genes in adult life (Collins & Weissman, 1984). However, as shown in Fig. 2, the 3’ ends of all these deletions are different; do they all contain rather similar enhancer sequences? This seems unlikely, and perhaps the most attractive hypothesis to explain the phenotypic variability of these deletions is that the γ β globin gene cluster is organized into two chromatin domains, one surrounding the foetal genes, the other flanking the δ and β genes, both with distinct 5’ and 3’ borders. Interference with any of these domain boundaries may prevent activation of the adult domain and leave the foetal genes unrepressed. This hypothesis was discussed in detail by Bernards & Flavell (1980). Additional deletion mutations of the γ δ β globin gene cluster are being mapped in an attempt to clarify these issues. However, because they all cause such a major disruption of the gene complex, their study may be of limited value for providing information about gene control during normal development.

Non-deletion HPFH

Because these conditions are not associated with major disruptions of the β globin gene cluster they are of much greater potential interest for analysing the regulation of gene expression during development. As mentioned earlier, the genetic determinants for some of these conditions map within or near the β gene complex while others segregate independently from the β globin gene markers and hence must be at a considerable distance away on chromosome 11 or even on another chromosome.

Some well-defined forms of non-deletion HPFH in which the genetic determinants are within the β globin gene cluster are summarized in Fig. 3. In the G γ β+ variety of HPFH, which has been found exclusively in Black populations, heterozygotes produce approximately 20% haemoglobin F of the Gγ type and there is β chain production in cis. Sequence analyses of the Aγ gene and of the/1/w repeat region 5’ to the <5 gene have shown no abnormality (Jones, Goodbourn, Old & Weatherall, 1985 ; Dr Oliver Smithies, personal communication). However, a single point mutation (C-→ -G) has been identified 202 base-pairs 5’ to the CAP site of the Gγ gene (Collins et al. 1984). A related disorder, called Greek HPFH, in which heterozygotes produce approximately 15 % foetal haemoglobin in adult life has also been analysed at the molecular level. The sequence of the A γ globin genes showed no abnormalities except for a single change, G → A at position —117, i.e. 117 bases 5’ to the CAP site (Collins et al. 1985; Gelinas et al. 1985). This change has been found in two unrelated heterozygotes. In a similar disorder observed in Italy a single base change, G → T, has been found at position —196 in the A γ gene (Giglioni et al. 1984). This has also been observed in a Chinese individual with a similar phenotype (Dr G. Stamatoyannopoulos, personal communication).

In the British form of non-deletion HPFH (Weatherall et al. 1975a) homozygotes have about 20% haemoglobinF, which is mainly of the A γ variety; heterozygotes have between 3 and 10% haemoglobin F. We have studied several heterozygotes for this condition at birth and followed their pattern of haemoglobin F production during the first year of life (Wood et al. 1982). The G γ /A γ chain production ratio is normal at birth but the rate of decline of haemoglobin F production is retarded and the adult pattern of predominantly A γ chain synthesis appears during the first few months of life. These observations suggest that the primary defect in this condition affects a regulatory region involved in the neonatal suppression of A γ chain production; when the G γ and A γ loci are fully activated in foetal life the expression of the genes is normal. Recently, we have sequenced the G γ and A γ genes from an individual homozygous for this condition; both genes are normal except for a T—>C change at position —198 in the A γ gene (Tate et al. unpublished data).

These interesting new forms of HPFH are summarized in Fig. 3. They appear to result from a series of single base changes clustered 5’ to the Gγ or Aγ genes, upstream from the CCAT and ATA boxes. In view of the developmental history of foetal haemoglobin production in the British variety, it is possible that these mutations alter DNA/protein interactions that ate involved in the neonatal suppression of γ gene synthesis.

Finally, there are non-deletion HPFH-like conditions associated with relatively low levels of haemoglobin F production in heterozygotes. Although some of them may be caused by determinants that map within the α globin gene cluster (Old et al. 1982) several families have now been reported in which this is not the case (Gianni et al. 1983). This is the first evidence for the existence of genes that influence globin chain synthesis that are not linked to the globin gene cluster. Currently, linkage studies are being carried out to attempt to determine the chromosomal location of these putative regulatory regions. Recently, we have established a linkage for a form of HPFH of this type to a restriction fragment length polymorphism defined by a mini-satellite probe (Jeffreys, Wilson & Thein, 1985).

Haemopoietic cell transplantation between developmental stages

The pattern of switching between foetal and adult haemoglobin is similar in sheep and man, and thus it has been possible to perform interdevelopmental stage haemopoietic cell transfer studies in this species (Wood et al. 1985). The rationale for these experiments was as follows. Foetal haemopoietic cells, obtained from liver and bone marrow, can be transplanted into irradiated lambs in which the switch from foetal to adult haemoglobin synthesis is already complete. If the transplanted cells switch over to adult haemoglobin synthesis immediately, it would imply that switching is determined mainly by the microenvironment of the erythroid progenitors in the bone marrow. If, on the other hand, switching occurs in the donor cells at about the same time as it would had the cells remained in the foetus, this would point to the existence of an intrinsic regulatory mechanism or ‘developmental clock’ within the foetal haemopoietic stem cells. Finally, if the transplanted cells continue to produce foetal haemoglobin indefinitely, it suggests that gene switching is under the control of a regulatory mechanism that is only present at a particular time during foetal development and hence had been bypassed by the transplantation.

Given the technical difficulty of these studies, the results of a large number of experiments are now reasonably consistent (Wood et al. 1985). Foetal haemopoietic cells transplanted into newborn animals continue to synthesize foetal haemoglobin and then gradually switch over to adult haemoglobin production. The timing of the transition is related to the gestational age of the foetus from which the donor cells were obtained, although it may be accelerated very slightly in the recipient. In the converse experiment, adult bone marrow cells transplanted into a foetus synthesize predominantly adult haemoglobin, implying that once the switch has occurred it is irreversible. To date, the results of these transplant experiments are compatible with a ‘developmental clock mechanism’ for the regulation of foetal globin gene expression.

Gene expression in neoplastic cell lines

The notion that analysis of gene expression in haematological neoplasms might provide some useful models for studying the developmental genetics of haemoglobin is not new. It has been known for some time that infants with juvenile chronic myeloid leukaemia (JCML) revert to a pattern of red cell protein production that is very similar to that observed late in foetal life (Weatherall, Edwards & Donohoe, 1968; Weatherall et al. 1915b). Their haemoglobin consists predominantly of Hb F with a marked reduction of Hb A2 and carbonic anhydrase, another protein that is switched on during the later part of foetal development. Unfortunately it has not been possible to establish JCML cells in culture.

There are, however, several established cell lines that are of potential interest for studying the developmental genetics of haemoglobin. The main stimulus to these studies came from the observation that a mouse erythroleukaemic cell line (MEL), first established by Charlotte Friend, when induced by dimethyl sulphoxide, haem or other agents, undergoes terminal erythroid differentiation and synthesizes haemoglobin, in this case of the adult variety (Marks & Rifkind, 1978). The human cell line K562 was originally derived from a patient with transforming chronic granulocytic leukaemia (CGL) (Lozzio & Lozzio, 1975). Although there are slight variations between different K562 lines, most of them synthesize predominantly ε and ç chains with smaller amounts of a and y chains when induced with haem or other agents (Rutherford, Clegg & Weatherall, 1979). No β chain production has been found in these cells. The fi globin genes are intact, have a normal structure and are expressed normally when cloned and transferred to COS cells. Curiously, of the two a genes, only al is expressed in K562 cells (D. R. Higgs, unpublished observation).

Another line, in this case derived from a patient with erythroleukaemia and called Human Erythroleukaemia Line (HEL), after induction, produces mainly Aγ and Gγ chains with some abut no β chains (Martin & Papayannopoulou, 1982). Another human leukaemia cell population that synthesized only haemoglobin F was also derived from an adult patient with transforming CGL, but a permanent line could not be established (Potter et al. 1984).

Thus it appears that some adult-derived leukaemia cell lines can be induced to express their embryonic and, or, foetal globin genes. There does not appear to be any structural abnormality of the later-developmental globin genes, which are not expressed in these cells; in a sense they appear to be ‘frozen’ at an early developmental stage, similar to JCML cells described earlier. Hence they offer a useful model for chromosome or gene transfer experiments for defining the possible role of irans-acting regulatory factors that might be involved in the expression of globin genes during development.

Intact chromosome or gene transfer experiments

A number of experiments have been carried out that have asked whether there is any evidence for developmental-stage-specific trans regulation of the γôβ globin gene complex. The interpretation of these studies is based on the assumption that some of the mouse or human malignant cell lines that can be induced to produce adult or foetal haemoglobin, described above, are ‘fixed’ at specific developmental stages. It has been found that if chromosome 11, cosmids containing the human β, y and E genes, or plasmids containing the genes alone, are inserted into mouse erythroleukaemia (MEL) cells there is a significant increase in β globin gene expression after induction of haemoglobin synthesis, whereas there is no increase in the expression of the y or E genes (Willing, Nienhuis & Anderson, 1979; Wright, De Boer, Grosveld & Flavell, 1983). Similarly, when intact human chromosomes 16 are transferred into MEL cells there is expression of the agiobin genes but not of the embryonic Ç globin genes (Zeitlin & Weatherall, 1984). Furthermore, if chromosome 16 of K562 cells, in which the embryonic globin genes are active, is transferred into MEL cells, no ç gene expression is observed (Anagnou et al. 1985).

Recently, chromosome 11 derived from the human cell line HEL has been transferred into MEL cells (Papayannopoulou et al. 1985). As mentioned earlier, HEL cells synthesize embryonic and foetal non-a chains but do not produce adult ft globin chains. After transfer, the globin genes were activated, suggesting that they are transcriptionally competent and thus that they may be responding to a positive trans-acting element within the MEL environment. Presumably they fail to express in the HEL environment because of the absence of this factor or the presence of a trans-acting inhibitor of ft globin gene expression.

The results of these experiments suggest that, if it is assumed that MEL cells are ‘adult’ in character, they lack the appropriate stage-specific developmental trans regulatory factors that are required for the expression of the embryonic or foetal globin genes but produce traws-acting factors capable of supporting a and β gene expression. However, other data have not all been consistent with this interpretation. The agiobin genes, when part of chromosome 16, show induction, as does the globin gene on chromosome 11, but when a globin genes are introduced into MEL cells as part of a cosmid or plasmid they are expressed at a high level, independent of induction. This behaviour contrasts with that of the ft globin genes, which still show induction dependence when introduced in either a cosmid or plasmid. Nevertheless, this is a promising approach to the definition of trans regulators and it should be possible to expand these studies provided that a source of embryonic or foetal recipient cells can be obtained for similar transfer experiments.

There is an extensive literature on the differential expression of the y and P globin genes in clonal colonies derived from BFU-Es and CFU-Es, erythroid progenitor cells which can be defined by their size and time of appearance in culture in vitro. This work is summarized in the published accounts of three recent conferences on gene switching (Stamatoyannopoulos & Nienhuis, 1979, 1981, 1983).

The expression of the foetal and adult globin genes in colonies reflects the stage of maturation of the individuals from which the progenitors were obtained. Colonies derived from foetuses produce predominantly y chains, whereas those obtained from adults synthesize mainly β chains although there is always a higher proportion of γ chains produced in adult-derived colonies than is present in adult red cells. Both y and γ chain genes are expressed in the same colonies and there is a continuum in the relative proportion of γ and βchain production in BFU-Es obtained from newborns at different gestational ages. The latter argues against there being specific stem cell populations that are programmed for foetal or adult globin chain synthesis.

A variety of modifications of experimental conditions, particularly those that affect the growth of colonies, can change the relative expression of γ and β globin genes in erythroid colonies. This may be because the relative expression of y and β genes is related to the maturity of the progenitors. At least in some species there appears to be asynchrony of γ and gene expression during colony maturation, with γ genes expressed earlier than β genes. Recent experiments in which the absoute amount of γ and β chain accumulation in colonies has been estimated suggest that this may not always be the case, particularly in human BFU-Es. However, when absolute levels of haemoglobin accumulated per cell are calculated, the amount in vitro is always considerably below that obtained in vivo. Thus the attention that has been paid to the higher proportion of haemoglobin F in adult BFU-Es has not yet been demonstrated to have any bearing on globin gene regulation in vivo.

Perhaps the most interesting observation relating to the regulation of gene switching that has arisen from studies of erythroid colonies is that the expression of the γ genes in both adult BFU-Es and those from individuals with various genetic conditions associated with persistent γ chain production, including deletion HPFH, can be ‘switched off’ by a factor that is present in foetal sheep sera. This has led to the development of a model in which switching occurs when the appropriate receptors for this putative inhibitory factor are expressed during the later stages of foetal development.

Several other models of the regulation of the y and βglobin genes have been derived from analyses of the pattern of globin gene expression in colonies. For example, it has been suggested that programming may reflect a ‘decision’ by early progenitors to move to terminal differentiation in which γchain production is more likely, rather than to go through further divisions and differentiation steps that make βgene expression more probable. This stochastic model of differentiation has been extended to encompass the regulation of foetal and adult globin gene expression during in vivo development. However, the in vitro colony model has not provided any clear insights into how these different pathways of differentiation might be mediated or regulated.

In vivo modification of y globin gene expression

As mentioned earlier, the administration of agents such as 5-azacytidine, which cause hypomethylation of genes, is associated with a modest increase in y chain production in patients with sickle cell anaemia or thalassaemia (Ley et al. 1982; Charache et al. 1983). Since this effect can be obtained with cytotoxic agents that do not cause hypomethylation of DNA it has been suggested that the effect on y chain production results from perturbations of the patterns of erythroid maturation (Letvin et al. 1985). It is possible that both mechanisms play a role.

Clearly, it is impossible to synthesize the diverse information outlined in this review and provide a coherent model for the regulation of globin gene expression during normal human development. One of the great difficulties in this field is uncertainty about whether the mutations that are associated with persistent y chain production, or for that matter the experimental models that have been used to study the differential expression of the foetal and adult globin genes, have any real relevance to the normal globin gene switching mechanism. They may have, but only with respect to a limited part of what must be an extremely complex multi-step regulatory system.

The consistent changes in chromatin and the methylation state of the /J globin gene cluster that are associated with activation of the different gene loci at various stages of development provide an anatomical explanation for the activity of these loci but tell us nothing about how these changes are mediated. However, the gene or chromosome transfer experiments suggest that there may be developmental-stagespecific trans factors involved in the regulation of these genes, presumably by interacting with chromatin. This is a very promising area for further work although, since all these experiments rely on the expression of genes in neoplastic cell lines, the results have to be interpreted with particular caution. Equally interesting is the possibility that the ‘upstream’ mutations, which have been found recently in some varieties of non-deletion HPFH, will provide a clue as to the site of interactions between chromatin and regulatory proteins. Thus at least we have some indications of what might be the most productive area of investigation for trying to characterize the mechanisms of regulation at the chromosomal level. Similarly, recent successes with the expression of ‘foreign’ globin genes in transgenic mice point the way to how we might learn more about the tissue specificity of globin gene expression (Chada et al. 1985).

This may be as far as we can go in the immediate future. The central question remains, however. How is the differential expression of the globin genes during development actually timed? All we know at the moment is that it is related fairly closely to gestational age and that it is not tissue dependent. The only experimental data relating to this question, derived from the sheep transplant model, suggest that there may be a ‘developmental clock’ built into the haemopoietic stem cell. Here we have a serious conceptual difficulty because there is no obvious model with which to analyse time-related events; none of the forms of HPFH is, strictly speaking, a heterochronic mutation. That is, these conditions are not characterized by a change in the time of globin gene switching; in the only form of non-deletion HPFH that has been studied during the period of switching the timing of the transition from foetal to adult haemoglobin production was completely normal; there was a delay in the rate of decline of foetal haemoglobin production suggesting that the mutation involved the mechanism of adult suppression of y chain production (Wood et al. 1982; Tate et al. unpublished data).

One of the main difficulties in designing experiments to ask questions about the timing of events during development is the lack of any clear concept of how the process might be mediated. One possibility, that should be amenable to investigation, is that differential globin gene expression is related to the number of haemopoietic stem cell divisions. A clock based on such a mechanism is feasible; in the mouse it has been estimated that there may only be a limited number of haemopoietic stem cell divisions during foetal development. However, our preliminary experiments in sheep suggest that the foetal haemopoietic stem cell is extremely resistant to perturbation by agents such as busulphan, and that chronic hypertransfusion of the foetus, which might be expected to reduce the number of stem cell divisions, has very little effect on the timing of the transition from foetal to adult haemoglobin (W. G. Wood & C. Bunch, unpublished observations).

Clearly, our current understanding of the developmental genetics of haemoglobin is at an extremely rudimentary stage. However, it is apparent that there are several promising areas for further work and that the globin gene model may still have something to offer to developmental biology.

Anagnou
,
N. P.
,
Yuan
,
T. Y.
,
Lim
,
E.
,
Helder
,
J.
,
Wieder
,
S.
,
Glaister
,
D.
,
Marks
,
B.
,
Wang
,
A.
,
Colbert
,
D.
&
Deisseroth
,
A.
(
1985
).
Regulatory factors specific for adult and embryonic globin genes may govern their expression in erythroleukemia cells
.
Blood
65
,
705
712
.
Antonarakis
,
S. E.
,
Boehm
,
C. D.
,
Giardina
,
P. J. V.
&
Kazazian
,
H. H.
(
1982
).
Non random association of polymorphic restriction sites in the β-globin gene complex
.
Proc. natn. Acad. Sei. U.S A
.
79
,
137
141
.
Bard
,
H.
(
1975
).
The postnatal decline of hemoglobin F synthesis in normal full-term infants
.
J. clin. Invest
.
55
,
395
398
.
Bernards
,
R.
&
Flavell
,
R. A.
(
1980
).
Physical mapping of the globin gene deletion in hereditary persistence of foetal haemoglobin (HPFH)
.
Nucl. Acids Res
.
8
,
1521
1534
.
Chada
,
K.
,
Magram
,
J.
,
Raphael
,
K.
,
Radice
,
G.
,
Lacey
,
E.
&
Costantini
,
F.
(
1985
).
Specific expression of a foreign (J-globin gene in erythroid cells of transgenic mice
.
Nature, Lond
.
314
,
377
380
.
Charache
,
S.
,
Dover
,
G.
,
Smith
,
K.
,
Talbot
,
C. C.
,
Moyer
,
M.
&
Boyer
,
S. H.
(
1983
).
Treatment of sickle cell anemia with 5-azacytidine results in increased fetal hemoglobin production and is associated with nonrandom hypomethylation of DNA around the γ-<5-β-globin gene complex
.
Proc. natn. Acad. Sci. U.S A
.
80
,
4842
4846
.
Collins
,
F. S.
,
Boehm
,
C. D.
,
Waber
,
P. G.
,
Stoeckert
,
C. J.
,
Weissman
,
S. M.
,
Forget
,
B. G.
&
Kazazian
,
H.H
. (
1984
).
Concordance of a point mutation 5 ‘to the Gγ globin gene with Gγβ+ hereditary persistence of fetal hemoglobin in the Black population
.
Blood
64
,
1292
1296
.
Collins
,
F. S.
,
Metherall
,
J. E.
,
Yamakawa
,
M.
,
Pan
,
J.
,
Weissman
,
S. M.
&
Forget
,
B. G.
(
1985
).
A point mutation in the Aγ-globin gene promoter in Greek hereditary persistence of fetal haemoglobin
.
Nature, Lond
.
313
,
325
326
.
Collins
,
F. S.
&
Weissman
,
S. M.
(
1984
).
The molecular genetics of human hemoglobin
.
Prog. Nucl. Acid Res. molec. Biol
.
31
,
315
462
.
Desimone
,
J.
,
Heller
,
P.
,
Hall
,
L.
&
Zwiers
,
D.
(
1982
).
5-azacytidine stimulates fetal hemoglobin synthesis in anemic baboons
.
Proc. natn. Acad. Sci. U.S A
.
79
,
4428
4431
.
Gelinas
,
R.
,
Endlich
,
B.
,
Pfeiffer
,
C.
,
Yagi
,
M.
&
Stamatoyannopoulos
,
G.
(
1985
).
G to A substitution in the distal CCAAT box of the Ay-globin gene in Greek hereditary persistance of fetal haemoglobin
.
Nature, Lond
.
313
,
323
325
.
Gianni
,
A. M.
,
Bregni
,
M.
,
Cappellini
,
M. D.
,
Fiorelli
,
G.
,
Taramelli
,
R.
,
Giglioni
,
B.
,
Comi
,
P.
&
Ottolenghi
,
S.
(
1983
).
A gene controlling fetal hemoglobin expression in adults is not linked to the non-a globin cluster
.
EMBOJ
.
2
,
921
925
.
Giglioni
,
B.
,
Casini
,
C.
,
Mantovani
,
R.
,
Merli
,
S.
,
Comí
,
P.
,
Ottolenghi
,
S.
,
Saglio
,
G.
,
Camaschella
,
C.
&
Mazza
,
U.
(
1984
).
A molecular study of a family with Greek hereditary persistance of fetal hemoglobin and β-thalassemia
.
EMBOJ
.
3
,
2641
2645
.
Gilman
,
J. G.
&
Huisman
,
T. H. J.
(
1984
).
A mutation 158 base pairs 5’ to the Gy gene is associated with elevated Gγ production
.
Blood
64
,
62
.
Jeffreys
,
A. J.
,
Wilson
,
V.
&
Thein
,
S. L.
(
1985
).
Hypervariable “minisatellite” regions in human DNA
.
Nature, Lond
.
314
,
67
73
.
Jones
,
R. W.
,
Goodbourn
,
S. E. Y.
,
Old
,
J. M.
&
Weatherall
,
D. J.
(
1985
).
The sequence of the Aγ globin gene in a Gγβ+ type of hereditary persistence of fetal haemoglobin
.
Br.J. Haemat
.
59
,
357
362
.
Letvin
,
N. L.
,
Linch
,
D. C.
,
Beardsley
,
P.
,
Mcintyre
,
K. W.
,
Miller
,
B. A.
&
Nathan
,
D. G.
(
1985
).
Influence of cell cycle phase-specific agents on simian fetal hemoglobin synthesis
.
J. clin. Invest
.
75
,
1999
2005
.
Ley
,
T. J.
,
Desimone
,
J.
,
Anagnou
,
N. P.
,
Keller
,
G. H.
,
Humphries
,
R. K.
,
Turner
,
P. H.
,
Young
,
N. S.
,
Heller
,
P.
&
Nienhuis
,
A. W.
(
1982
).
5-azacytidine selectivity increases y-globin synthesis in a patient with β-thalassemia
.
New Engl. J. Med
.
307
,
1469
1475
.
Llebhaber
,
S. A.
&
Kan
,
Y. W.
(
1982
).
Different rates of mRNA translation balance the expression of the two human <r-globin loci
.
J. biol. Chem
.
257
,
11 852
11 855
.
Lozzio
,
C. B.
&
Lozzio
,
B. B.
(
1975
).
Human chronic myelogenous leukemia cell line with positive Philadelphia chromosome
.
Blood
45
,
321
334
.
Marks
,
P. A.
&
Rifkind
,
R. A.
(
1978
).
Erythroleukemic differentiation
.
A. Rev. Biochem
.
47
,
419
448
.
Martin
,
P.
&
Papayannopoulou
,
T.
(
1982
).
HEL cells: A new human erythroleukemia cell line with spontaneous and induced globin expression
.
Science
216
,
1233
1235
.
Nute
,
P. E.
,
Pataryas
,
H. A.
&
Stamatoyannopoulos
,
G.
(
1973
).
The Gγ and Aγ hemoglobin chains during human fetal development
.
Am. J. hum. Genet
.
35
,
271
276
.
Old
,
J. M.
,
Ayyub
,
H.
,
Wood
,
W. G.
,
Clegg
,
J. B.
&
Weatherall
,
D. J.
(
1982
).
Linkage analysis of non-deletion hereditary persistence of fetal hemoglobin using DNA polymorphisms
.
Science
215
,
981
982
.
Orkin
,
S. H.
&
Goff
,
S. C.
(
1981
).
The duplicated human a-globin genes: their relative expression as measured by RNA analysis
.
Cell
24
,
345
351
.
Orkin
,
S. H.
&
Kazazian
,
H. H.
(
1984
).
The mutation and polymorphism of the human β-globin gene and its surrounding DNA. A
.
Rev. Genet
.
18
,
131
171
.
Papayannopoulou
,
T.
,
Lindsley
,
D.
,
Kurachi
,
S.
,
Lewison
,
K.
,
Hemenway
,
T.
,
Melis
,
M.
,
Anagnou
,
N. P.
&
Najfeld
,
V.
(
1985
).
Adult and fetal human globin genes are expressed following chromosomal transfer in to MEL cells
.
Proc. natn. Acad. Sci. U.S A
.
82
,
780
784
.
Peschle
,
C.
,
Migliaccio
,
A. R.
,
Migliaccio
,
G.
,
Petrini
,
M.
,
Calandrini
,
M.
,
Russo
,
G.
,
Mastroberardino
,
G.
,
Presta
,
M.
,
Gianni
,
A. M.
,
Comi
,
P.
,
Giglioni
,
B.
&
Ottolenghi
,
S.
(
1984
).
Embryonic—"fetal Hb switch in humans: Studies on erythroid bursts generated by embryonic progenitors from yolk sac and liver
.
Proc. natn. Acad. Sci. USA
.
81
,
2416
2420
.
Peschle
,
C.
,
Mavilio
,
F.
,
Carè
,
A.
,
Migliaccio
,
G.
,
Migliaccio
,
A. R.
,
Samoggia
,
P.
,
Petti
,
S.
,
Guerriero
,
R.
,
Marinucci
,
M.
,
Lazzaro
,
D.
,
Russo
,
G.
&
Mastroberardino
,
G.
(
1985
).
Haemoglobin switching in human embryos: asynchrony of ζ → α and εγ switches in primitive and definitive erythropoietic lineages
.
Nature, Land
.
313
,
235
237
.
Potter
,
C. G.
,
Bunch
,
C.
,
Potter
,
A. C.
,
Buckle
,
V.
,
Derry
,
S.
,
Wood
,
W. G.
&
Weatherall
,
D. J.
(
1984
).
Erythroid differentiation in CGL cells from a patient with blast crisis
.
Leuk. Res
.
8
,
713
721
.
Rutherford
,
T. R.
,
Clegg
,
J. B.
&
Weatherall
,
D. J.
(
1979
).
K562 human leukaemic cells synthesise embryonic haemoglobin in response of haemin
.
Nature, Land
.
280
,
164
165
.
Schroeder
,
W. A.
,
Shelton
,
J. R.
,
Shelton
,
J. B.
,
Apell
,
G.
,
Huisman
,
T. H. J.
&
Bouver
,
N. G.
(
1972
).
World-wide occurrence of nonallelic genes for the γ-chain of human foetal haemoglobin in newborns
.
Nature, new Biol
.
240
,
273
274
.
Stamatoyannopoulos
,
G.
&
Nienhuis
,
A. W.
(
1979
).
Cellular and Molecular Regulation of Hemoglobin Switching
.
New York
:
Grune and Stratton
.
Stamatoyannopoulos
,
G.
&
Nienhuis
,
A. W.
(
1981
).
Hemoglobins in Development and Differentiation
.
New York
:
A. R. Liss
.
Stamatoyannopoulos
,
G.
&
Nienhuis
,
A. W.
(
1983
).
Globin Gene Expression and Hematopoietic Differentiation
.
New York
:
Alan R. Liss
.
Wainscoat
,
J. S.
,
Thein
,
S. L.
,
Higgs
,
D. R.
,
Bell
,
J. L
,
Weatherall
,
D. J.
,
Al-Awamy
,
B.
&
Serjeant
,
G.
(
1985a
).
A genetic marker for elevated levels of haemoglobin F in homozygous sickle cell disease
.
Br.J. Haemat
.
60
,
261
268
.
Wainscoat
,
J. S.
,
Thein
,
S. L.
,
Wood
,
W. G.
,
Weatherall
,
D. J.
,
Tzotos
,
S.
,
Kanavakis
,
E.
,
Metaxatou-Mavromati
,
A.
&
Kattamis
,
C.
(
1985b
).
A novel deletion in the f globin gene complex
.
Ann. N.Y. Acad. Sci
.
445
,
20
27
.
Weatherall
,
D. J.
,
Cartner
,
R.
,
Clegg
,
J. B.
,
Wood
,
W. G.
,
Macrae
,
I.
&
Mackenzie
,
A.
(
1975a
).
A form of hereditary persistence of fetal haemoglobin characterised by uneven cellular distribution of haemoglobin F and the production of haemoglobins A and A2 in homozygotes
.
Br.J. Haemat
.
29
,
205
220
.
Weatherall
,
D. J.
,
Clegg
,
J. B.
,
Wood
,
W. G.
,
Callender
,
S. T.
,
Sheridan
,
B. L.
&
Pritchard
,
J.
(
1975b
).
Foetal erythropoiesis in human leukaemia
.
Nature, Land
.
257
,
710
712
.
Weatherall
,
D. J.
,
Edwards
,
J. A.
&
Donohoe
,
W. T. A.
(
1968
).
Haemoglobin and red cell enzyme changes in juvenile chronic myeloid leukaemia
.
Br. med. J
.
1
,
679
681
.
Weatherall
,
D. J.
,
Higgs
,
D. R.
,
Wood
,
W. G.
&
Clegg
,
J. B.
(
1984
).
Genetic disorders of human haemoglobin as models for analysing gene regulation
.
Phil. Trans. R. Soc. Lond. B
307
,
247
259
.
Weatherall
,
D. J.
&
Wainscoat
,
J. S.
(
1985
).
The molecular pathology of thalassaemia
.
In Recent Advances in Haematology, 4th edn
(ed.
A. V.
Hoffbrand
), pp.
63
88
.
Edinburgh
:
Churchill Livingstone
.
Willing
,
M. C.
,
Nienhuis
,
A. W.
&
Anderson
,
W. F.
(
1979
).
Selective activation of human γ- but not y-globin gene in human fibroblast X mouse erythroleukaemia cell hybrids
.
Nature, Lond
.
277
,
534
538
.
Wood
,
W. G.
&
Bunch
,
C.
(
1983
).
Fetal-to-adult hemopoietic cell transplantation: Is hemoglobin synthesis gestational age-dependent?
In Globin Gene Expression and Hematopoietic Differentiation
(ed.
G.
Stamatoyannopoulos
&
A. W.
Nienhuis
), pp.
511
521
.
New York
:
Alan R. Liss
.
Wood
,
W. G.
,
Bunch
,
C.
,
Kelly
,
S.
,
Gunn
,
Y.
&
Breckon
,
G.
(
1985
).
Haemoglobin switching: evidence for a developmental clock
.
Nature, Lond
.
314
,
320
323
.
Wood
,
W. G.
,
Macrae
,
I. A.
,
Darbre
,
P. D.
,
Clegg
,
J. B.
&
Weatherall
,
D. J.
(
1982
).
The British type of non-deletion HPFH: characterisation of developmental changes in vivo and erythroid growth in vitro
.
Br.J. Haemat
.
50
,
401
414
.
Wood
,
W. G.
,
Nash
,
J.
,
Weatherall
,
D. J.
,
Robinson
,
J. S.
&
Harrison
,
F. A.
(
1979
).
The sheep as an animal model for the switch from fetal to adult hemoglobins
.
In Cellular and Molecular Regulation of Hemoglobin Switching
(ed.
G.
Stamatoyannopoulos
&
A. W.
Nienhuis
), pp.
153
167
.
New York
:
Gruñe and Stratton
.
Wood
,
W. G.
&
Weatherall
,
D. J.
(
1973
).
Haemoglobin synthesis during human foetal development
.
Nature, Lond
.
244
,
162
165
.
Wood
,
W. G.
&
Weatherall
,
D. J.
(
1983
).
Developmental genetics of the human haemoglobin
.
Biochem.J
.
215
,
1
10
.
Wright
,
S.
,
De Boer
,
E.
,
Grosveld
,
F. G.
&
Flavell
,
R. A.
(
1983
).
Regulated expression of the human β-globin gene family in murine erythroleukaemia cells
.
Nature, Lond
.
305
,
333
336
.
Zeitlin
,
H. C.
&
Weatherall
,
D. J.
(
1983
).
Selective expression within the human agiobin gene complex following chromosome-dependentwoid mouse erythroleukaemia cells
.
Molec. Biol. Med
.
1
,
489
500
.