WILLIAM L. TREPICCHIO, MICHAEL A. DYER, AND MARGARET H. BARON*. Department of Cellular and Developmental Biology, The Biological Laboratories, ...
Vol. 13, No. 12
MOLECULAR AND CELLULAR BIOLOGY, Dec. 1993, p. 7457-7468
0270-7306/93/127457-12$02.00/0 Copyright X 1993, American Society for Microbiology
Developmental Regulation of the Human Embryonic 13-Like Globin Gene Is Mediated by Synergistic Interactions among Multiple Tissueand Stage-Specific Elements WILLIAM L. TREPICCHIO, MICHAEL A. DYER, AND MARGARET H. BARON* Department of Cellular and Developmental Biology, The Biological Laboratories, Harvard University, 16 Divinity Avenue, Cambridge, Massachusetts 02138 Received 15 July 1993/Returned for modification 2 September 1993/Accepted 8 September 1993
The stage-specific regulation of mammalian embryonic globin genes has been an experimentally elusive problem, in part because of the developmentally early timing of their expression. We have carried out a systematic analysis of truncation and internal deletion mutations within the 5'-flanking region of the human embryonic 1M-like globin gene (£) in erythroid and nonerythroid cell lines. Within a 670-bp region upstream from the constitutive promoter are multiple positive and negative control elements. Of these, a positive regulatory element (E-PRE II) which is active only in embryonic erythroid cells is of particular interest. Remarkably, although it is inactive on its own, in the presence of other sequences located further upstream, it confers tissue- and developmental stage-specific expression on a constitutive e-globin or heterologous promoter. The activity of E-PRE II is also modulated by another positive regulatory domain located further downstream to direct erythroid cell-specific, but little or no embryonic stage-specific, transcription. A nuclear factor highly enriched in embryonic erythroid cells binds specifically within a 19-bp region of r-PRE II. Nuclei from adult erythroid cells also contain a factor that binds to this region but forms a complex of faster electrophoretic mobility. We speculate that interactions between E-PRE II and other upstream control elements play an important role in the developmental regulation of the human embryonic n-like globin gene.
promoters of cis-linked genes (25). In transgenic mice (45, 50) or in differentiating embryoid bodies (35), the human e-globin gene, linked in cis with LCR sequences, is expressed in a developmentally appropriate manner which is independent of the presence of other globin genes within the
Erythropoiesis is a complex process that involves changes in the site of red cell production, changes in red cell morphology and size, and switches in globin gene expression (53). In humans, "hemoglobin switching" occurs at two distinct developmental stages: (i) the embryonic-to-fetal switch occurs very early in gestation and involves a change in expression of both the a- and 3-globin clusters; (ii) the fetal-to-adult switch, involving only the 13 cluster, occurs around the time of birth. Relatively little is known about erythroid differentiation during very early stages of mammalian development, but it is believed to begin in the blood islands of the embryonic yolk sac and involves the activation of embryonic a- and A-like globin genes in large nucleated ("primitive") erythroid cells. As the site of erythropoiesis begins to shift to the fetal liver, embryonic globin gene expression is gradually down-regulated and transcription of the fetal globin genes begins (33, 61). In humans, this hemoglobin switch takes place during week 5 or 6 of gestation (43). Thus, the embryonic globin genes are activated at an early stage of human development and are expressed for only a few weeks before they are turned off again. A major advance in our understanding of globin gene regulation came with the identification of the locus control region (LCR) (reviewed in references 11 and 41). The human 13-globin LCR is located within the 6- to 18-kb region upstream of the embryonic 13-like globin gene (E) and contains a series of four erythroid cell-specific DNase I-hypersensitive (HS) sites that are present at all stages of development (16, 56). A DNA fragment containing these HS sites confers erythroid cell-specific, high-level expression on the * Corresponding author. Electronic mail address: husc4.harvard.edu.
locus. Relatively little is known about the regulation of the earliest-expressed human 1-like globin gene (£). The e-globin gene promoter contains several of the same DNA elements known to be critical components of the constitutive promoters of fetal and adult 1-globin genes (TATA, CCAAT, and CACCC). This promoter also contains potential binding sites for the erythroid proteins GATA-1 and NF-E2 (an AP-1-like complex) (for reviews, see references 13 and 42). Although several studies have mapped protein-DNA binding upstream of the e-globin start site (23, 26, 27, 44, 62), only the CACCC (40, 62) and GATA (23) elements have been subjected to analysis by mutagenesis and formally shown to play a role in promoter function. Even in the case of the CACCC element, which at least in vitro can bind SP1, TEF-2, and at least two erythroid cell-specific factors (reviewed in reference 58), it is not known which, if any, of these proteins actually regulates globin gene expression. A silencer element has been identified upstream of the human a-globin gene promoter (8, 27). Deletion of this region results in the continued expression of the human embryonic A-like globin gene during late fetal and adult development in transgenic mice (46), suggesting that in vivo it may play a role in silencing the embryonic 13-globin gene at these later developmental stages. It is not known how the e-globin gene is activated early in development. We have demonstrated that heterokaryon formation between embryonic erythroid and adult erythroid
mbaron@ 7457
7458
TREPICCHIO ET AL.
or nonerythroid cells results in the trans activation of human or mouse embryonic A-like globin gene expression in the
nuclei of the previously nonexpressing cells (3-5). The rapid reprogramming of globin gene expression in these multinucleated heterokaryons must be mediated by both erythroid cell- and developmental stage-specific trans-acting regulatory molecules. As a first step in identifying some of the critical components of this regulatory network, we have exploited the results of the cell fusion studies in established erythroid lines. We have prepared a series of truncation and internal deletion mutations within the first 849 bp of the human embryonic P-like globin 5'-flanking region and have analyzed their transcriptional behavior in some of the same erythroid and nonerythroid cell lines used earlier (4, 5). Although the most rigorous demonstration of cell type and stage specificity will require analysis with transgenic mice, the advantage of beginning with a cultured cell system is that it permits more rapid screening of a large number of mutated constructs. A subset of the most informative mutations can then be tested in developing transgenic animals. Comparison of the expression patterns of these mutated sequences revealed the presence of multiple positive and negative control elements within a 670-bp region upstream of the constitutive promoter. An embryonic erythroid cellspecific positive regulatory element (here termed e-PRE II) is of particular interest. Its developmental stage specificity is observed only in the presence of other sequences, located further upstream, with which it interacts synergistically. Although neither region is active on its own in monomeric or multimerized form, when present together within the intact upstream control region (where they are separated by approximately 200 bp of intervening sequence) or when ligated in tandem, they activate a constitutive e-globin or heterologous promoter in a manner that mimics the tissue and developmental stage specificity of the normal e-globin gene. The erythroid cell (but not stage) specificity of e-PRE II is also reconstituted in the presence of another element located further downstream. A nuclear factor highly enriched in embryonic erythroid cells binds specifically within a 19-bp region of e-PRE II which contains a novel, evolutionarily conserved, 6-bp dyad repeat. Adult erythroid nuclei also contain a factor that binds to this region, but the complex formed migrates more rapidly during nondenaturing electrophoresis. We propose that in vivo, high-level expression of the e-globin promoter in yolk sac erythroblasts results from synergistic-cooperative interactions among the multiple positive regulatory elements, of which e-PRE II is likely to play an important role in the regulation of this gene during de-
vel6pment.
MATERILS AND METHODS Plasmid constructions. LCRe(-179)/CAT contains c-globin 5' upstream sequences from the BamHI site at -179 to the StuI site at +44 which was modified by addition of an 8-bp XhoI linker. A 1.8-kb KpnI-HindIII fragment containing this e-globin promoter region ligated to the chloramphenicol acetyltransferase (CAT) gene was excised from plasmid AMM-1 (a gift from Alan Michelson) and ligated with KAnI-EcoRI-digested pUC19 and a 2.5-kb "1xLAR" (15) derivative in which an internal HindIII site in the truncated LCR was destroyed by blunt ligation (a gift from Tariq
Enver). LCRe(-849)/CAT is similar to LCRe(-179)/CAT but contains e-globin sequences from -849 (ClaI site filled in with the Klenow fragment of DNA polymerase I) to the XhoI-
MOL. CELL. BIOL.
tinkered StuI site at +44. A 0.9-kb ClaI-XhoI fragment from pSPHeRIXhoCAT (a gift from Alan Michelson) was ligated into SmaI-XhoI-digested LCRs(-179)/CAT from which the globin sequences had been excised. To make 3' (internal deletion) constructs, c-globin sequences from the ClaI site at -849 to the BamHI site at -179 were excised from pUCHeX (2a) and subcloned into pSP73 (Promega). Exonuclease III digestion (29) was initiated at the BamHI site and allowed to proceed toward the 5' terminus of the e-globin fragment, yielding a series of nested deletions. Deletion endpoints were located by doublestranded DNA sequencing (48). Selected mutant fragments were excised by digestion with SmaI and PvuII and were subcloned in the correct orientation into the SmaI site of LCRs(- 179)/CAT. In the course of sequencing these mutants, we found numerous discrepancies between our sequences and the sequence in the GenBank data base (2). Therefore, fragments of the 5'-flanking region of the human e-globin gene were subcloned from the phage genomic clone XHeG1 (19) (a gift from Nick Proudfoot and Edward Fritsch) into pBluescript (pBSK) phagemid vectors (Stratagene) to allow for single-stranded sequencing of both DNA strands. Our plasmid and phage sequences were in agreement (55) and are available upon request. The numbering of the constructs used in this study is based on these sequences. To make 5' deletion (truncation) constructs, the ClaI-XhoI fragment containing e-globin sequences from -849 to +44 was excised from LCRe(-849)/CAT and subcloned into pBSK+ (Stratagene). Exonuclease III digestion was initiated at the ClaI site and allowed to proceed downstream. Deletion endpoints were identified by DNA sequencing. Selected mutant fragments were excised by digestion with BamHI and were subcloned in the correct orientation into a BglIIdigested derivative of LCRs(-179)/CAT containing a BglII linker inserted at the SmaI site [LCRe(-179)BglII/CAT]. Regulatory elements identified by analysis of the 5' and 3' deletion mutants were isolated by polymerase chain reaction (PCR) amplification and were subcloned into BglII-digested LCRe(-179)BglII/CAT or into BamHI-digested pBLCAT2 (36) containing the herpes simplex virus thymidine kinase (tk) promoter from positions -105 to +51. Oligonucleotides used to amplify specific sequences were synthesized on an Applied Biosystems 380B DNA synthesizer. Sense (s)- and antisense (as)-strand oligonucleotides contained BglII and BamHI sites, respectively (underlined sequences), to facilitate cloning and multimerization. The sequences are as follows: -770 to -751 (s), 5'-AAAAfiAj.jCCCAGTGAG AAGTATAAGCA-3'; -499 to -480 (s), 5'-AAAAGAI TGAAACTAAGGTACAGAAGTT-3'; -420 to -403 (s), 5'AAAAGAI3lCITATT CTTTTCCTTGG-3'; -682 to -701
(as), 5'-AAAGLGATCCTGACCACAGGGGACACTGGC-3'; -419 to -438 (as), 5'-AAAGGATCCAAAGCTGTTAAA ACAGAATC-3'; and -297 to -314 (as), 5'-AAAGGATC CAAGAAAGCCTCATATAAA-3'. Sequences from -419 to -499 and -770 to -682 were subcloned in the appropriate orientation into the BglII site of LCRe(-179)BglII/CAT. Cell culture and induction. K562, MEL, GM979, and HeLa cells were maintained and induced as described previously (4, 5). Some experiments were done with other K562 sublines, from P. Rowley (lines LA4 and RA6 [47]). All K562
sublines gave very similar results. Transfections and enzymatic assays. MEL and K562 cells were transfected by using a modification of the method of Stuve and Myers (54). Briefly, cells plated on poly-L-lysinecoated 10-cm-diameter dishes (4) were treated with a DNA-
VOL. 13, 1993
DEVELOPMENTAL REGULATION OF THE HUMAN e-GLOBIN GENE
DEAE-dextran mixture for 40 min at room temperature. After aspiration of this mixture, cells were incubated for 2 to 3 h at 370C in medium containing 50 jiM chloroquine. MEL and GM979 cells were plated 1 day prior to transfection in medium containing 12% fetal calf serum. GM979 cells did not adhere well to the treated dishes and were transfected by a modification of the foregoing procedure in which all incubations were performed with cells suspended in 1 ml of medium, and changes of medium were carried out by gentle centrifugation. For all erythroid cell transfections, 7.5 jtg of the appropriate CAT construct was mixed with 2 jig of an internal reference plasmid, pCMV3-gal (37). HeLa cells were transfected by calcium phosphate coprecipitation (24) using 18 jig of CAT construct and 2 jig of pCMVP-gal. Enzymatic CAT assays (49) were performed with equal amounts of protein from all cell lysates. Incubations were for 3 h for the erythroid cells and 1 to 3 h for HeLa cells. P-Galactosidase activities (52) were used to correct for differences in transfection efficiencies among samples. For all four cell lines, each construct was tested in duplicate or in triplicate in each of three to six independent transfection experiments. To control for variations in DNA samples, at least two different preparations of each plasmid construct were tested. In our experiments, dimethyl sulfoxide (but not butyrate) dramatically increased MEL cell transfection efficiency; this effect has been observed by others (54). For GM979 cells, butyrate treatment (but not dimethyl sulfoxide alone) was required for efficient transfection, and hemin increased the transfection efficiency of K562 cells. Butyrate treatment of HeLa cells neither activated endogenous human globins nor had any effect on the activity of the embryonic stage-specific element. Nuclear extracts, EMSA, and methylation interference. Nuclear extracts were prepared from 1 to 3 liters of induced erythroid or nonerythroid cells by the method of Dignam et al. (10) except that the following protease inhibitors were added to buffers A, B, and C: pepstatin A, 2 jig/ml; leupeptin, 1 ,ug/ml; and aprotinin, 5 ,ug/ml. Binding reaction mixtures contained 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 10% glycerol, 0.5 mM dithiothreitol, and 3 ,ug of poly(dI-dC). For binding reactions (18, 21), nuclear extracts (2 to 4 ,ug) were mixed in a 20-pl volume with 0.1 to 0.5 ng of Klenow polymerase end-labeled oligonucleotide at room temperature for 30 min. The reactions were analyzed by electrophoretic gel shift mobility assay (EMSA) in a nondenaturing 4% polyacrylamide gel at 4°C in 1 x Tris-borate-EDTA-5% glycerol. For methylation interference (1), DNA was labeled with T4 polynucleotide kinase, methylated, and incubated with nuclear extract. Bound and free DNAs were separated by EMSA, cleaved with piperidine, and analyzed on a 15% denaturing polyacrylamide gel. Oligonucleotides for protein binding. Sense- and antisensestrand oligonucleotides contained one to five additional terminal nucleotides to generate complete BamHI and BglII sites at the 5' and 3' ends, respectively, of the doublestranded oligonucleotide. The endpoints of human e-globin sequence included in the oligonucleotides summarized in Fig. 6C are as follows: oligonucleotide 1, -464 to -415; oligonucleotide 2, -464 to -445; oligonucleotide 3, -450 to -430; and oligonucleotide 4, -438 to -415. For oligonucleotides 1, 2, and 4, filling in of the 5' end with the Klenow fragment of DNA polymerase was required for all e-globin sequences to be present in double-stranded form. Oligonucleotide 4 contains a T-to-C transversion (as the result of incorporation of nucleotides to complete a BamHI site) that does not affect DNA binding. The sequences of the sense (s)
internal (3) deletions
-4Ban*i
Clal 849
12
-179
5 m
HS IV-I
+1
exolXl
7459
Xhol
HimdlIll I
+44
7-1
EcoRI
CAT
Spro t xoll e truncation (5') deletions
FIG. 1. General structure of the truncation and internal deletion constructs used in this study. The basic construct used in these experiments contained the constitutive promoter (epro) from -179 to +44, ligated to a CAT reporter gene. A IpLCR cassette was inserted downstream from the chimeric e-CAT gene. The orientation of this cassette, which in these constructs is reversed, is irrevelent to its activity (unpublished observations). Truncation deletions were generated by exonuclease III (exoIII) digestion beginning at the ClaI site at -849 and proceeding toward the transcriptional start site. Internal deletions were produced by exonuclease III digestion beginning at the BamHI site at -179 and proceeding upstream. and antisense (as) strands of the octamer-containing oligonucleotide used as nonspecific competitor are as follows: octa (s), 5'-GATCTCAATGCAAATATCTG-3'; and octa (as), 5'-GATCCAGATAATTGCATTGA-3'. RESULTS The 5' upstream region of the human embryonic IP-like globin gene (e) contains multiple tissue-specific regulatory elements. For the experiments reported here, the u-globin constitutive promoter (8) from -179 to +44 was linked in cis to a CAT reporter gene (Fig. 1). Because the activities of some erythroid cell- and/or stage-specific elements might be apparent only in the presence of the LCR enhancers (reviewed in reference 11) and might escape detection in its absence (23), we included a "micro LCR" (jiLCR) immediately downstream from the 8-CAT fusion gene. The jLLCR is a truncated version of the P-LCR which contains all four of the super-HS sites (15) and allows correct tissue- and stagespecific regulation of a human embryonic 3-globin transgene in mice (45). The validity of the present approach is underscored by recent studies of Jane et al. (31), who observed correct developmental regulation of LCR-linked human fetal and adult P-globin genes in the absence of chromosomal integration (i.e., in a transient expression system). Under the conditions of our assay, essentially no u-globin transcription can be detected in the absence of an enhancer (P-LCR or simian virus 40; data not shown), consistent with the observations of others (8, 23, 40). Sequences required for erythroid cell- and stage-specific expression of the human e-globin gene are contained within the region upstream to -849 (numbering is based on our sequence [55]; see Materials and Methods for details) (32, 51). A series of truncation and internal deletions was generated by exonuclease III digestion downstream from -849 or upstream from -179, respectively (Fig. 1). The various deletion constructs were analyzed for tissue- and developmental stage-specific expression by comparison of their activities in several erythroid and nonerythroid cultured cell lines (4, 5): mouse (adult) erythroleukemia cells (MEL cells), GM979 cells (a mouse erythroleukemia cell line that expresses embryonic as well as adult P-globin genes), K562 cells (a human embryonic/fetal erythroid line), and HeLa
7460
MOL. CELL. BiOL.
TREPICCHIO ET AL.
B
A -849
-19
-795
5,
1
.
+44
-849
+44
1
2
(HeLa)
-775
-775
I E-PRE V
-736
-7361
| sPRE IV
_
-639
-639 .
5
1~
14
'
15
-503
,
-503
6
-464
-464
1
.
Nonerythroid
-795
L
7
1
7
-382
-382
Embryonic Erythroid
-308
-308
|
9
_
-179>R 2
0
9
= K562
-179
L-4~
= GM979
3'
10
0
6
4
C
D
+44
-179
-849
2
4
6
Relative Acetylation
Relative Acetylation 1
5,
-795 -775 i_
.E
I
S0 :
S-PREy
-639
| -503
*--
I
5
1
.
E-PRE IV
:~
c
0
-736
1
-
0
6
a)
464
.
L
1
CO
7
-382
a) t+1
-308
Adult Erythroid
9
-179
cc
(MEL)
T 0
2
4
6
3,
8
Relative Acetylation -849
-795
-775
-736
-639
-503
-464
-382
-308
-179
Deletion Endpoint FIG. 2. Analysis of human e-globin 5' deletion mutants in GM979 and K562 (embryonic erythroid) cells (A), HeLa (nonerythroid) cells (B), and MEL (adult erythroid) cells (C) and summary of results for all cell lines (D). Values graphed represent averages of three to six separate experiments ± standard deviations, using at least two independent DNA preparations. Activities are expressed relative to the activity of the constitutive 179-bp promoter. The key constructs that led to the identification of e-PRE V and IV are indicated by solid bars to the right of the histograms. The results for all cell lines are superimposed in the summary (D), which allows a direct comparison of expression patterns for these constructs.
cells. For the purposes of this study, the erythroid lines are referred to as embryonic (GM979 or K562) or adult (MEL) erythroid cells. To identify putative tissue- and stage-specific regulatory elements between -849 and -179, a nested series of 5' deletions (Fig. 1) was first analyzed in the two embryonic erythroid cell lines (Fig. 2A). Progressive deletion from -849 to -775 (Fig. 2A) resulted in a very modest (approximately twofold) increase in expression for both cell lines, possibly suggesting the presence of a weak negative regulatory element between -849 and -775. Compared with the -179 promoter construct, the 5' deletion to -775 showed an increase in activity of about fivefold (GM979 cells) or two- to threefold (K562 cells). Upon further deletion beyond -775, activity levels dropped below that of the constitutive promoter, consistent with the loss of a positive regulatory element. It is noteworthy that the positive element between -775 and -736 (here termed e-positive regulatory element V [e-PRE VJ; see Fig. 6 and Table 1) is apparently active even in the presence of the silencer (-294 to -251), which can function in embryonic erythroid cells (8, 27). In HeLa cells (Fig. 2B), the positive element between -775 and -736
TABLE 1. Summary of regulatory elements upstream of the human embryonic P-globin promoter Positions
-304 -323 -419 -446 -736 -682 -729 -775
to to to to to to to to
-179 -304 -392 -419 -639 -640 -682 -736
Type -a
+ -
+ + -
+
+1,d
Name
e-NRE Ia s-PRE I e-NRE IIF e-PRE II e-PRE IV e-NRE III s-PRE III e-PRE V
Tissue
Stage
specific
No Yes ND Yes Yes ND No Yes
specific
NDb No ND
Embryonic No ND No
?
Contains the silencer element first identified by Cao et al. (8) and later mapped between positions -294 and -251 by Gutman et al. (27). For uniformity of terminology, we refer to it here as e-NRE I. b ND, not determined (it is difficult to assign tissue- or stage-specific regulatory properties to these negative control regions solely on the basis of a
the deletion analysis presented here; see text). ' May be identical to a second, recently reported silencer (59). d May contain overlapping or adjacent positive and negative elements else may play a dual role during development.
or
DEVELOPMENTAL REGULATION OF THE HUMAN e-GLOBIN GENE
VOL. 13, 1993 A
+44
-1 71 -304
-849
-323 a
L-A
-392
-446 L-47
3'
Nonerythroid
1
-304.
-8,49
11
(HeLa)
2
+1
8323
3
*3,92
-ENRE 11
L,4
-419
Ic-PRE 11
"45
-523
1
I IS-PRE
34
L
-419
-179 +44
3' E-NRE
_I
2
_
1
B
0
1I
7461
I~~~~~~~~~~~
-446
!
I
tL-
-523
a
6
47
-560
-560
~H
1
-596
-596 9
-640
4
L~1c D
-682
"4 "4122 11
-7
-729
0
-640
,----i
Embryonic Erythroid = K562 _ = GM979 2
4
8
6
4 10
.
I S-NRE I -PRE
-682 !
11
-729 A
.5'
10
I
"412
'
5, 0
12
-1 2
4
6
8
10
12
Relative Acetylation
Relative Acetylation
C
III
III
D
-179 +444
3' ~L -323
-3,92
+ -
-PRE
L-143 ,
4
.,
,
..
5
-446 cI 0
Li
-523
V
t
-560
0
-K
:.:7:71-B.
-596
LJ~~~L-
9
IA
-682 *
-729 A
I
| 11
E
| 12
IE-NRE
III
£-PRE
III
c1
5.
2
4
6
8
10
12
S-PRE III
S-PRE 11
E-PRE
-179
Relative Acetylation
FIG. 3. Analysis of human e-globin 3' deletion mutants in GM979 and K562 (embryonic erythroid) cells (A), HeLa (nonerythroid) cells (B), and MEL (adult erythroid) cells (C) and summary of results for e-PRE I, II, and III in all cell lines (D). Values graphed represent averages of three to six separate experiments standard deviations, using at least two independent DNA preparations. Activities are expressed relative to the activity of the constitutive 179-bp promoter. The key constructs that led to the identification of e-PRE I, II, and III are indicated by solid bars to the right of the histograms; for greater clarity, only these latter results are summarized in panel D. The key constructs that led to the identification of s-NRE I, II, and III are indicated by cross-hatched bars to the right of the histograms.
failed to stimulate transcription over constitutive levels and is therefore likely to be erythroid cell specific. Analysis of the 5' deletion series was next carried out in MEL cells, which express only adult mouse P-globin, to examine the potential stage specificity of the -775 to -736 positive element. The increase in activity observed in the adult (MEL) cells with progressive deletion beyond -775 suggests the presence of a negative regulatory domain between -775 and -736 (Fig. 2C). Further deletion resulted in decreased expression of the reporter gene, indicating the presence of a second erythroid cell-specific positive control region between -736 and -639 (termed s-PRE IV) which also shows weak activity in the embryonic erythroid cells. A reciprocal pattern of expression was seen in the region of -775 to -736 (s-PRE V) for the adult compared with the embryonic erythroid cells: whereas deletion from -775 to -736 decreased the activity of this region in the embryonic cells (Fig. 2A), the same deletion resulted in increased activity in the adult erythroid cells, to as high as eight times that of the constitutive promoter (see -639 construct in Fig. 2C). In summary, we have identified two tissue-specific regulatory domains (e-PRE V and e-PRE IV) that show
positive activity in embryonic erythroid cells (Fig. 2D). The region between -775 and -736 (E-PRE V) activates e-globin gene expression in embryonic erythroid cells and suppresses its expression in adult erythroid cells, suggesting that it may play a dual role in the developmental regulation of the embryonic P-globin gene. e-PRE IV functions in both embryonic and adult erythroid cells. A stage-specific positive regulatory element is contained within the 5' upstream region of the human embryonic "-like globin gene (E). To examine the behavior of positive elements in the absence of the silencer at -294 to -251 (8, 27), we generated a series of internal (3') deletion mutants (Fig. 1). Analysis of this set of deletion mutants in the embryonic erythroid cells (Fig. 3A) revealed the presence of two additional positive elements with boundaries between -446 and -419 (ac-PRE II) and between -323 and -304 (e-PRE I). The presence of these elements was not predicted from the 5' deletion analysis (e.g., see deletion constructs -308 through -639 in Fig. 2A) but became apparent only after deletion of sequences located immediately downstream (summarized in Table 1). These sequences, which are likely
MOL. CELL. BIOL.
TREPICCHIO ET AL.
7462
A
200 bp
to contain negative regulatory elements, are here designated +1
£-PRE III,IV
449
e-NRE I and r-NRE II (Fig. 3A and Table 1). Neither a-PRE I nor e-PRE II was active in HeLa cells (Fig. 3B), and they are therefore likely to be erythroid cell specific (see Fig. 3D for comparison of all four cell lines). However, in contrast with the element at -323 to -304 (W-PRE I), which was active in MEL cells, the element at -446 to -419 (e-PRE II) was not (Fig. 3C). Therefore, r-PRE II is both tissue and (embryonic) stage specific. We conclude that the region between -849 and -179 contains at least four tissue-specific positive elements (-775 to -736, e-PRE V; -736 to -639, e-PRE IV; -446 to -419, s-PRE II; and -323 to -304, e-PRE I) that are active in embryonic erythroid cells (Fig. 2D and 3D; see Table 1 and Fig. 5 for summaries). Of these elements, at least one (e-PRE II; Fig. 3D) and possibly a second (-775 to -736, e-PRE V; Fig. 2D), is also embryonic stage specific. Interestingly, e-PRE II was active even in the absence of e-PRE I located further downstream (the -419 deletion does not contain e-PRE I; Fig. 3A) but was not detected in the absence of upstream sequences containing e-PRE IV and V (Fig. 2A). Conversely, the activity of s-PRE V was not apparent from the 3' deletion data (Fig. 3A), suggesting that it must interact with at least one other positive element (e.g., e-PRE I and/or II) located downstream (see below). Finally, the activities of e-PRE I and e-PRE II (Fig. 3A and C) were not apparent from the 5' deletion data (Fig. 2A and C), reflecting the presence of the downstream silencer and possibly the absence of other positive regulatory domains further upstream (e.g., e-PRE V). Analysis of the internal deletion mutants in adult erythroid (MEL) and nonerythroid (HeLa) cells revealed yet another positive element (termed e-PRE III) positioned between -729 and -682 (Fig. 3B and C). Essentially no activity was
-179
8-globin £ gV1e
E-PRE
E-PRE
V
11
E-PRE
-499/-419 -770/4682 -499/-294 419/294 -7701-419
m EM
__~
- PCR fragments
B .170r -19+44
LEGEND:
M_
1
M
HeLa
_2~~~~~~~~~~
K562 1(embryonic
= =
__
I_
(nonerythroid)
IEl = MEL (adult erythroid) GM979
erythroid)
3
~ 4
_
_~~~~~~~~~~~~~~~~~~~~~
~~~~~~~6_
2
_7~~~~~~~
-
_ I
=
3
~~~~~~9
_
_
1 EI~i0 D
_
1
1
12
12
0
609
C
LEGEND:
5
10
15
20
25
Relative Acetylation
F = HeLe (nonerythrold)
= MEL (adult erythrold) M * = K562
(embryonic erythrold)
subcloned for these experiments are shown below. For simplicity, putative negative regulatory regions (Table 1 and Fig. 6) have been omitted. (B) Different combinations of positive regulatory elements interact synergistically in embryonic erythroid cells. The various constructs were analyzed in embryonic erythroid (GM979 and K562), adult erythroid (MEL), and nonerythroid (HeLa) cells. e-PRE II is not active on its own but interacts synergistically with e-PRE V and/or e-PRE III (both of which are contained within the -770 to -682 fragment, which also shows no activity in isolation). DNA sequences between -848 and -682 (containing e-PRE III) are active in MEL and HeLa cells, but not in the embryonic erythroid cells, in the absence of e-PRE II (see Fig. 3B and C). The 180-bp fragment designated 3y3' is from the 3' untranslated region of the human fetal Gy-globin gene; this fragment is not known to contain any regulatory sequences. Values graphed represent averages of two to five separate experiments standard deviations, using at least two independent DNA preparations. Activities are expressed relative to the activity of the constitutive 179-bp promoter. (C) Positive regulatory domains from the upstream region of the human e-globin gene can confer tissue- and stage-specific expression on a heterologous promoter. The -105 to +51 fragment from the tk gene was used as a heterologous promoter. The combination of e-PRE II and the 88-bp upstream region is active in the embryonic erythroid K562 cells but shows little or no activity in the adult erythroid (MEL) or nonerythroid (HeLa) cells. Both tissue- and developmental stage-specific expression are therefore reconstituted for this combination of regulatory regions. In contrast, e-PRE I and II confer tissue- but not developmental stage-specific activity on the tk promoter. Values graphed represent averages of three to five standard deviations, using at least two separate experiments independent DNA preparations. Activities are expressed relative to the activity of the tk promoter. ±
1
2
.105 .51
I
9*h7A CAT_
2
FIG. 4. Interactions
among
positive e-globin regulatory ele-
ments. (A) Diagram of DNA fragments used in this analysis. The
positive regulatory elements identified in this study are represented schematically in the diagram at the top, and DNA fragments
±
DEVELOPMENTAL REGULATION OF THE HUMAN 8-GLOBIN GENE
VOL. 13, 1993
constitutive promoter -775/-736 -7361439
4461-419
-323/-304
+1_ -9-2 -162 OL
-27 |~
GATA CACCC CCAAT ATA
III
F D 482/-640 s-NRE III
-106 -80
419/-392 E-NRE11
-3041-179 E-NRE
FIG. 5. Schematic summary of the embryonic 1-like globin (£globin) gene upstream regulatory region. The positions of the TATA (ATA), CCAAT, CACCC, and GATA sites of the constitutive promoter (shaded bar), as well as the silencer element, are shown. Numbering corresponds to our DNA sequence (see Materials and Methods) and refers to distance from the cap site. Positions are marked for the center of each constitutive promoter element. Indicated here are approximate 5' and 3' boundaries of positive (solid rectangles) and negative (lightly shaded lower boxes) regulatory regions shown in the present work to be active in embryonic erythroid cells. Curved arrows represent interactions between s-PRE II and sequences located upstream (contributing to both tissue- and developmental stage-specific expression) and downstream (contributing to tissue-specific expression). The crosshatched bar represents the boundaries of the region (-770 to -682) shown in the experiments of Fig. 3 and 4 to reconstitute tissue- and developmental stage specificity when linked to e-PRE II. For simplicity, s-PRE III (active in adult erythroid and nonerythroid cells but not in embryonic erythroid cells) is not shown. The figure is not drawn to scale.
seen in the embryonic erythroid lines (Fig. 3A and D). This element is not tissue specific but is expressed in both adult erythroid and nonerythroid cells. Because its activity was revealed by deletion of the region between -682 and -640, it is likely to be located upstream of a negative regulatory element (e-NRE III). To rule out the possibility that the results presented in Fig. 3 reflect differences in the distances between 5' regulatory sequences and the ,uLCR located further downstream, the -179 constitutive promoter construct and -848/-419 3' deletion construct (containing s-PRE II, III, IV, and V) were redesigned so that the ,iLCR was positioned immediately upstream of the -179 and -849 endpoints, respectively (see Materials and Methods). Both the absolute and relative activities of these 5'-LCR constructs were indistinguishable from those of the corresponding 3'-LCR constructs (data not shown). These studies have identified at least three negative regulatory domains upstream of the e-globin promoter (Table 1). We note that it is difficult to define criteria by which to assign tissue or stage specificity to these regions. Their deletion unmasked the activities of other positively acting regions that are tissue and/or stage specific. However, the negative domains themselves may or may not show such specificity, and additional studies will be required to clarify this point. An embryonic stage-specific element and a second element cooperate to confer correct developmental regulation of the e-globin gene. Our interest in understanding how the e-globin gene is activated during embryonic development prompted us to examine the stage specificity of e-PRE II in greater detail. To determine whether this element was sufficient to confer embryonic stage-specific expression on the constitutive promoter, the region between -419 and -499 containing e-PRE II was amplified by PCR and subcloned in its native orientation directly upstream of -179 (Fig. 4A; construct 2 of Fig. 4B). This 80-bp fragment failed to up-regulate expression from the constitutive promoter in any of the four cell
7463
lines examined (Fig. 4B). Numerous studies (e.g., references 22 and 63) have demonstrated that for some regulatory regions, individual sequence elements may function poorly (or not at all) in isolation but that transcriptional activation is regained upon multimerization of a single element or upon combination with other elements. Four copies of the 80-bp fragment (construct 3) gave at best only a threefold stimulation of activity compared with the constitutive promoter, and this effect was neither tissue nor stage specific. Taken together with the 3' deletion analysis (Fig. 3), these results suggested that additional sequences, between -499 and -849, are required for developmentally appropriate expression. In fact, a larger fragment extending from -770 to -419 was sufficient to activate the constitutive promoter in an erythroid cell- and embryonic stage-specific manner (construct 8). Specificity was reconstituted when the 80-bp fragment containing s-PRE II was ligated in tandem with an 88-bp fragment containing sequences between -770 and -682 (construct 6). The 88-bp fragment did not activate expression on its own, either as a monomer (construct 4) or as a tetramer (construct 5). Activation by combining the 88-bp fragment with the 80-bp fragment containing e-PRE II was orientation dependent (compare constructs 6 and 7) and specific, as ligation of a 180-by fragment from the 3' untranslated region of the human y-globin gene in place of the 88-bp fragment (construct 12) did not result in stimulation. (The junction sequences are identical for constructs 6 and 12, so a novel regulatory element was not artificially created by ligation.) These results indicate that the region containing e-PRE II interacts synergistically with sequences located further upstream to confer tissue- and stage-specific expression on the e-globin constitutive promoter. We suspect that this interaction involves sequences within e-PRE V but cannot exclude other sequences (e.g., e-PRE III) within the 88-bp fragment. Furthermore, this interaction is likely to be modulated by additional regulatory sequences, as the magnitude of the effect seen with construct 6, which lacks sequences between -682 and -499, was lower than that with construct 8, which contains these sequences. Tissue-specific activation was also seen when the s-PRE II-containing region (-499 to -419) was extended downstream to -294 to include s-PRE I (as well as one or more putative negative regulatory elements; construct 9). However, a DNA fragment extending from -419 to -294 was inactive on its own (construct 10). Therefore, the region containing e-PRE II can also interact independently with sequences located further downstream (containing e-PRE I) to confer tissue-specific (but not stage-specific) expression on the constitutive s-globin promoter. The e-globin upstream regulatory elements that are active in embryonic erythroid cells are diagrammed in Fig. 5. Combinations of fragments containing the e-PREs can confer erythroid cell and/or embryonic stage specificity on a heterologous promoter. To determine whether stage-specific regulation could be conferred on a nonerythroid promoter by combinations of e-PRE-containing fragments, the tk promoter from - 105 to +51 (37) was substituted for the e-globin promoter. Both the -770/-682+-499/-419 fragment (construct 1; Fig. 4C) and the -499/-294 fragment (construct 2; Fig. 4C) activated expression of the tk promoter by four- to sixfold in the embryonic erythroid K562 cells but decreased tk promoter activity in the nonerythroid HeLa cells by approximately 70%. The combination of sequences containing e-PRE II and s-PRE V, but not e-PRE II and s-PRE I, was stage specific: in the adult erythroid MEL cells, no significant activation was observed for construct 1, but a
TREPICCHIO ET AL.
7464
MOL. CELL. BIOL.
A Erythroid Embryonic (GM979) 1
2
3
4
5
6
Adult (MEL)
Nonerythroid (Hela)
8
10 11 12 13 14 15 16
7
9
B
c _4.
*
]euF
B
Erythroid Embryonic (GM979)
1
2
3
Embryonic (K562)
Adult
(MEL) 5 6
4
7
8
9
gill-, * A X,* "3*-4B _,m
-4-- C
C E-PRE 11
element identified
by deletion -455
-465
-445
-I
F.
-415
-425
-435
-
-
-405 ---
CTCCTTAAGAGAGCTAGAACTGGGTGAGATTCTGTTTTAACAGCTTTATTTTCTTTTCCTT 1EMSA 2 3
4
HI.
(A) Erythroid cell specifically to e-PRE In this experiment, 0.1 ng of oligonucleotide 4, labeled by filling in with the Klenow fragment of DNA polymerase I, was incubated with 2 jig of protein from GM979, MEL, or HeLa cell nuclear extracts in the absence or presence of unlabeled competitor oligonucleotide. The labeled arrows indicate protein complexes formed. Complex A was formed by protein from embryonic erythroid cells (lane 1) and was also seen at much lower concentration in the nonerythroid HeLa cells (lane 11). When the extract used for EMSA was from adult erythroid cells, a complex of slightly faster mobility (complex B) was observed. Complex C is nonspecific and was formed by nonerythroid extracts and, in some experiments, was observed at much lower levels for erythroid extracts. Lanes 1, 6, and 11, no unlabeled competitor. All other samples were incubated FIG.
6.
EMSA of
nuclear extracts
are
protein binding
enriched in
a
to e-PRE
factor that binds
modest activation (threefold) was observed for construct 2. Thus, relative stage and/or tissue specificity could be maintained even in the presence of a nonerythroid (tk) promoter that contains TATA and CCAAT elements but not the CACCC, GATA, or NF-E2/AP-1-like elements required in some combination for functional erythroid promoter activity (58). A nuclear factor highly enriched in embryonic erythroid cells binds specifically within a 19-bp region of E-PRE II. To identify proteins that might mediate the activity of s-PRE II, we examined the DNA-binding properties of nuclear extracts from embryonic erythroid (GM979 and K562), adult erythroid (MEL), and nonerythroid (HeLa) cells, using an EMSA (Fig. 6). The oligonucleotide probes used in these experiments are summarized schematically in Fig. 6C. Oligonucleotide 1 includes the region identified by deletion analysis to contain e-PRE II. The three overlapping oligonucleotides, 2, 3, and 4, span the sequence of oligonucleotide 1. Only oligonucleotides 1 and 4 formed specific protein complexes when incubated with extracts from embryonic erythroid cells (Fig. 6A and B and data not shown). Complex A was formed by incubation of embryonic erythroid extracts with oligonucleotide 4 (GM979 [Fig. 6A and B, lanes 1] and K562 [Fig. 6B, lane 7]). Binding was specifically competed for when challenged with an excess of unlabeled oligonucleotide 4 (Fig. 6A, lane 4; Fig. 6B, lanes 2 and 8) or oligonucleotide 1 (Fig. 6A, lane 5) but was not affected by the presence of a nonspecific oligonucleotide containing an octamer motif, ATTlGCAT (Fig. 6A, lane 3; Fig. 6B, lanes 3 and 9). Complex A was also observed for nonerythroid cell (HeLa) extracts, but at much lower levels (Fig. 6A; compare lanes 1 and 11). Interestingly, oligonucleotide 4 could compete effectively for binding only if its termini were filled in by treatment with Klenow DNA polymerase (Fig. 6A; compare lanes 2 and 4,
with a 50-fold molar excess of unlabeled competitor oligonucleotide as follows: lanes 2, 7, and 12, oligonucleotide 4, ends not filled in (see text); lanes 3, 8, and 13, octamer motif-containing oligonucleotide; lanes 4, 9, and 14, oligonucleotide 4, ends filled in; lanes 5, 10, and 15, oligonucleotide 1, ends not filled in. Lane 16 contained oligonucleotide probe but no protein and provided a negative control for binding. F, free radiolabeled probe. (B) Embryonic and adult erythroid nuclear extracts form complexes with different mobilities. To improve the resolution between the complexes formed by embryonic (complex A) and adult (complex B) erythroid nuclear extracts, the gel was run at 200 V for 8 h (instead of the usual 4 h). In this experiment, 0.5 ng of oligonucleotide 4 probe was incubated with 4 pg of protein from GM979, MEL, or K562 extracts. The specific complexes A and B formed by the embryonic erythroid (GM979) and adult erythroid (MEL) nuclear extracts, respectively, now show clearly distinct mobilities (compare lanes 1 and 4). K562 extracts (lane 7) produced a complex of mobility identical to that formed by GM979 nuclear extracts (lane 1). Lanes 1, 4, and 7 contain no cold competitor. A 20-fold molar excess of cold competitor oligonucleotide was included in the other samples as follows: lanes 2, 5, and 8, oligonucleotide 4 (specific competitor), ends filled in; lanes 3, 6, and 9, nonspecific oligonucleotide containing an octamer motif. (C) Oligonucleotides used in the EMSA analyses. The regulatory element defined by the 3' deletion analysis is represented by the shaded bar above the DNA sequence (-446 to -419). The oligonucleotide probes used in the EMSA experiments are shown below the sequence (our numbering; see Materials and Methods). Oligonucleotides 1 and 4 formed specific protein complexes by this assay. Asterisks mark nucleotides in probe 4 that must be present in double-stranded form for binding to occur (see
text).
DEVELOPMENTAL REGULATION OF THE HUMAN s-GLOBIN GENE
VOL. 13, 1993 CODING MEL
A
IGM9791
TABLE 2. Known or potential protein binding sites
NONCODING
IGM9791
MEL
Positions
C c
r
-395 to -389 e-NRE II -369 to -364 e-NRE II
G A T
A A _45TT
T
T T
A
-435T6
A
Transcription factor
E4TF1, Ets-l' 5'-GGAAGgG-3' AP-2' 5'-CCCCAGGC-3' See Fig. 7 5'-CTGTT(T/A)-3' dyad repeat AP-la/NF-E2b 5'-TTtGTCA-3' Basic helix5'-CAaaTG-3'
loop-helix2
C
T
-445
a
Reference 14. b Reference 39.
C
T A
T A
A G T
G ;
G
A G A
-435
C .w::
-445T A A
G
_
A
_
A A A A
T
-438, -432, and -425 could block the formation of both complexes. The residues at positions -432 and -425 fall within a dyad repeat of the two half-sites 5'-CTGTT(T/A)-3' (Fig. 7B). It is noteworthy that the residue at position -438 is located in the region of oligonucleotide 4 that must be filled in for effective competition of binding, again suggesting that these sequences are important for formation of complexes A and B. Finally, a phylogenetic comparison (28) reveals broad species conservation within e-PRE II, overlapping the dyad repeat.
1T c 4
..1_iqw,
G
B -445
DNA element
-774 to -768 e-PRE V -723 to -717 OPRE IV, III -434 to -423 e-PRE II
T T
-425 c
7465
-435
-425
-41 5
II
TGGG TGAMATTC TMTTTAAC AG C TT TA TTT AC C CAC TC TAAGACAAAATTETC GAAATAAA FIG. 7. (A) Methylation interference analysis of e-PRE II. The coding and noncoding strand of oligonucleotide 1 (-464 to -415) were treated with dimethyl sulfate bound with protein from embryonic (GM979) or adult (MEL) erythroid nuclear extracts. Proteinbound (B) and unbound (free [F]) complexes were isolated by EMSA, cleaved, and analyzed on a sequencing gel. Guanine residues that interfered with binding are marked by dots. Extracts from GM979 and MEL cells gave identical results, which are summarized in panel B. e-PRE II was initially identified by deletion from -446 to -419 and is contained within the oligonucleotide used in this analysis. Methylated residues that interfered with binding are highlighted by boxes. The dyad repeat described in the text is marked by arrows.
lanes 7 and 9, and lanes 12 and 14), suggesting that sequences at the 5' end may be critical for binding. This idea is supported by methylation interference experiments (see below). A complex of faster mobility (complex B) was observed when protein from the adult erythroid (MEL) nuclear extracts was used in this assay (Fig. 6A, lane 6; Fig. 6B, lane 4). In competition experiments, the formation of complex B was prevented by addition of oligonucleotide 4 (Fig. 6A, lane 9; Fig. 6B, lane 5) or oligonucleotide 1 (Fig. 6A, lane 10) to the binding reaction mixture but not by the addition of the octamer-containing oligonucleotide (Fig. 6A, lane 8; Fig. 6B, lane 6) or by oligonucleotide 2 (not shown). These observations suggest that both complexes A and B involve specific binding to s-PRE II. Methylation interference was used to identify guanosine residues involved in essential contacts in complexes A and B. As shown in Fig. 7A, methylated guanosines at positions
Several potential protein binding motifs (Table 2) were found in a systematic search within the 5' upstream region of the e-globin gene examined here. No other known protein binding sites were apparent from a broad survey of such sequences. In particular, none of the elements that we have identified contain binding sites for the two erythroid DNAbinding proteins GATA-1 (reviewed in references 13 and 42) and NF-E4 (20). A number of the protein binding sites identified by EMSA in a recent report (26) are located within the regions that we have designated e-PRE III and V. The functional significance (if any) of these putative binding sites will require more detailed studies.
DISCUSSION Positive regulation of the human embryonic 13-like globin Analysis of de novo embryonic mouse (4) or human (5) globin gene activation in heterokaryons has established that embryonic erythroid cells contain both tissue- and stagespecific trans-acting regulators of globin gene expression. The apparent plasticity of embryonic 3-globin gene expression revealed by cell fusion prompted us to examine the regulation of the human embryonic P-like globin gene (£) in some of these same cell lines (4, 5) in greater detail. In this report, we show that developmental control of this gene is complex and involves, in part, synergistic interactions among multiple positive control elements. The results of this study have revealed cell type- and stage-specific regulatory elements located further upstream from the embryonic P-like globin promoter than appears to be the case for either the human fetal or adult 3-globin genes (for a review, see reference 13). At least four functionally interdependent positive control elements, here termed s-PRE I, II, IV, and V, contribute to the tissue-specific expression of the embryonic ,B-globin gene. In vivo, e-PRE I, II, IV, and V might be involved in either the activation or the transcriptional maintenance of the earliest-expressed human ,B-globin gene. None of these regions is active in isolation, but together they cooperate to confer the regulatory characteristics of the full upstream regulatory region on a constitutive e-globin promoter or on a heterologous (tk) progene.
7466
TREPICCHIO ET AL.
moter. e-PRE II, in combination with one or more of the positive control elements located farther upstream, is functional only in embryonic erythroid cells, not in adult erythroid or nonerythroid cells. This stage specificity is lost, but erythroid specificity is reconstituted, when e-PRE II is paired with e-PRE I. Thus, we have demonstrated a functional dissociation between the tissue- and stage-specific determinants of e-PRE II activity. Surprisingly, an erythroid cell-specific promoter containing a GATA element and a CACCC or NF-E2/AP-1 element (58) is not absolutely required to reconstitute the general tissue- or stage-specific expression patterns directed by the more distal regulatory regions reported here. The activities of these regulatory domains are more reminiscent of an enhancer than of classical upstream promoter elements, as different combinations confer tissue- and/or stage-specific expression on a heterologous promoter. Unlike an enhancer, however, at least for the embryonic stage specificity of e-PRE II, activity is not orientation independent (see constructs 6 and 7 in Fig. SB). Evidence of a positive control region upstream of the silencer (between -535 and -453) has been suggested by in vitro transcription experiments (57), but its tissue and stage specificities have not been examined. e-PRE II is, to our knowledge, the first stage-specific regulatory element identified for a vertebrate embryonic globin gene. The functional synergy between e-PRE II and the 88-bp region upstream could involve s-PRE V or other sequences within this region (Fig. 5). The region containing e-PRE V activates expression in embryonic erythroid cells and suppresses expression in adult erythroid cells. This observation suggests the intriguing possibility that the same regulatory element alternates during development between a positive (embryonic) and a negative (later stages) control function. A precedent for such a model is provided by the oscillation of the E2F site between a positive and negative element during the cell cycle, which depends on the phosphorylation state of the protein Rb in an E2F-Rb complex (60). Whether the activity of e-PRE V reflects a single element with a dual developmental role, or two distinct elements, will require additional studies. Downstream from E-PRE V is another domain, here termed s-PRE III, that is active in adult erythroid and nonerythroid cells but not in embryonic erythroid cells (Table 1). The relevance of e-PRE III to embryonic 0-globin regulation is not yet clear. Suppression of the activity of e-PRE III in embryonic erythroid cells may be mediated by the negative regulatory domain (F-NRE III) located further downstream. Alternatively, it is conceivable that in cooperation with another regulatory element, e-PRE III in fact functions in embryonic erythroblasts. Negative regulation of the human embryonic 1-like globin gene. The results of these studies suggest the presence of multiple negative regulatory domains (Table 1 and Fig. 4) within the upstream region of the human 8-globin gene. From inspection of our 3' deletion data, we tentatively conclude that the region between -419 and -179 contains at least two negative elements. The first (here designated 8-NRE I for uniformity of terminology) includes the silencer (8, 27, 46) and is located between -304 and -179. Deletion of this region reveals the activity of e-PRE I. The second (s-NRE II) is located between -419 and -392; deletion of this region reveals the activity of 8-PRE II. e-NRE II may be identical to a second silencer element recently reported by Watt et al. (59). The negative elements identified in this study may help to explain how the strong effect of the LCR on s-globin expression is overcome during the embryonic-to-fetal switch
MOL. CELL. BIOL.
and may constitute a mechanism that ensures a high degree of cell type transcriptional specificity. We propose that in vivo, these negative regulatory regions prevent transcription of the e-globin gene in inappropriate cell types or at later stages of development. They may also serve to modulate levels of expression in embryonic erythroblasts. Removal of the silencer by deletion of the region between -467 and -182 of the human e-globin gene produced developmentally aberrant expression of this gene in transgenic mice: the mutant transgene continued to be expressed in adult mice, although at much lower levels than in embryos (46). On the basis of those results, the authors proposed that an additional negative regulatory sequence outside the silencer region may also be required for complete turn-off of the human e-globin gene and that the deleted region (-467 to -182) does not play a major role in e-globin expression in embryonic erythroid cells. Our data suggest the presence both of other negative regulatory elements further upstream and of two positive elements (e-PRE III and IV) that are active in adult erythroid cells (Table 1) and could partially explain the behavior of the e-globin deletion mutant in transgenic mice. In addition, the present studies indicate that deletion of the region between -467 and -182 removes not only the silencer but also two positive elements (F-PRE I and II) and at least one other negative element (Table 1 and Fig. 5), thereby complicating the interpretation of the results for transgenic mice. The relatively high expression observed in embryonic blood cells for a construct containing the -467 to -182 deletion (46) might reflect the combined activities of 8-PRE-III, IV, and V in the presence of the LCR. Some of these elements may also be functionally redundant. In any case, it will be important to examine the individual and combinatorial activities of each of these elements on 8-globin gene expression in the context of the developing animal (i.e., in transgenic mice). Protein binding by e-PRE II. We have identified a nuclear factor, highly enriched in embryonic erythroid cells, that binds specifically within a 19-bp region of e-PRE II. Three guanosine residues make critical contacts with this factor, and two (at -432 and -425) fall within a novel, evolutionarily conserved 6-bp dyad repeat that does not correspond to any previously reported protein recognition site and could potentially bind a dimeric protein. Adult erythroid nuclei also contain a factor that binds to e-PRE II and makes contact with the same set of guanosine residues, but the complex formed migrates more rapidly on nondenaturing gel electrophoresis. Together, these observations suggest that e-PRE II is bound by different proteins in embryonic and adult erythroid cells or else that the same protein is differentially modified in these cells in a way that modulates its activity. Such a modified protein bound to e-PRE II might act as a repressor, or it may be transcriptionally inactive in adult erythroid cells. In either case, we propose that the nuclear factor that forms complex A with e-PRE II in embryonic erythroid cells is a potential developmental regulator of e-globin gene expression. Human hemoglobin switching. In contrast with the human fetal-to-adult hemoglobin switch, which involves competition between those genes for productive interactions with the LCR (for a review, see reference 11), the activation and subsequent silencing of the human embryonic P-like globin gene are autonomous and do not require linkage in cis to the fetal or adult P-globin genes (45, 50). It is worth noting that the activities of individual HS sites have been shown to contribute to the stage-specific expression of the human fetal and adult ,-globin genes in transgenic mice (17), but specific
VOL. 13, 1993
DEVELOPMENTAL REGULATION OF THE HUMAN e-GLOBIN GENE
interactions of individual HS sequences with the human embryonic n-like globin gene(£) have not yet been reported. However, the combination of HS I and II alone appears to be sufficient to drive the correct temporal regulation of the human e-globin gene in transgenic mice (50, 51). Given evidence from naturally occurring human deletion mutations that HS I may be dispensable to hemoglobin switching (12, 34), it is possible that HS II alone is both necessary and sufficient, in combination with the more gene-proximal elements identified here and by others (8, 27, 46, 59), for the developmentally restricted expression of the embryonic 13-globin gene. The results presented here may provide some insight into the difference between the two major developmental switches within the human P-globin locus. In vivo, high-level expression of the s-globin gene in yolk sac erythroid cells probably results from synergistic-cooperative interactions (30) among the multiple positive regulatory elements that we have identified, with one another and presumably with the LCR (HS II), and by the production in yolk sac erythroblasts of one or more stage-specific regulatory proteins (4, 5) that may function through interaction with e-PRE II. Correct temporal control of this gene must therefore reflect a dynamic balance among the negative and positive regulatory elements reported here. Such interactions are likely to be mediated by changes in the relative stoichiometries and/or activities of positive and negative regulatory factors (see discussion in reference 6) in erythroid cells during development. ACKNOWLEDGMENTS We are grateful to Ranjan Sen, William Fixsen, Victor Ambros, and members of our laboratory for thoughtful reading of the manuscript and to Ross Hardison for sharing results prior to publication. We thank Kenzo Hirose for valuable technical assistance, Patrick Hayes for oligonucleotide synthesis, and Steven Cooke for help with DNA sequencing. This work was supported by grants (to M.H.B.) from the National Institutes of Health (RO1 GM42413), from the Lucille P. Markey Charitable Trust (87-24) and from the March of Dimes Birth Defects Foundation (Basil O'Connor Award 5-715). W.L.T. is a postdoctoral fellow of the Cooley's Anemia Foundation. M.A.D. is supported by NIH predoctoral training grant GM 07598. M.H.B. is a Lucille P. Markey Scholar in Biomedical Science. REFERENCES 1. Ausubel, F., R. Brent, R. E. Kingston, D. D. Moore, J. G. Seidman, J. A. Smith, and K. Struhl (ed.). 1990. Current protocols in molecular biology. John Wiley & Sons, New York. 2. Baralle, F. E., C. C. Shoulders, S. Goodbourn, A. Jeffreys, and N. J. Proudfoot. 1980. The 5' flanking region of the human e-globin gene. Nucleic Acids Res. 8:4393-4404. 2a.Baron, M. Unpublished data. 3. Baron, M. H. Reversibility of the differentiated state in somatic cells. Curr. Opin. Cell Biol., in press. 4. Baron, M. H., and T. Maniatis. 1986. Rapid reprogramming of globin gene expression in transient heterokaryons. Cell 46:591602. 5. Baron, M. H., and T. Maniatis. 1991. Regulated expression of human a- and 0-globin genes in transient heterokaryons. Mol. Cell. Biol. 11:1239-1247. 6. Blau, H. M. 1992. Differentiation requires continuous active control. Annu. Rev. Biochem. 61:1213-1230. 7. Bradford, M. M. 1976. A rapid and sensitive method for the quantification of microgram quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72:248-254. 8. Cao, S., P. Gutman, H. Dave, and A. Schechter. 1989. Identification of a transcriptional silencer in the 5'-flanking region of the
9. 10.
11. 12.
13. 14. 15. 16.
17.
18. 19. 20.
21.
22.
7467
human epsilon globin gene. Proc. Natl. Acad. Sci. USA 86: 5306-5309. Choi, O.-R., and J. D. Engel. 1988. Developmental regulation of 13-globin gene switching. Cell 55:17-26. Dignam, J. D., R. M. Lebovitz, and R. G. Roeder. 1983. Accurate transcription initiation by RNA polymerase II in a soluble extract from isolated mammalian nuclei. Nucleic Acids Res. 11:1475-1489. Dillon, N., and F. Grosveld. 1993. Transcriptional regulation of multigene loci: multilevel control. Trends Genet. 9:134-137. Driscoll, M. C., C. S. Dobkin, and B. P. Alter. 1989. Thalassemia due to a de novo mutation deleting the 5' 13-globin gene activation-region hypersensitive sites. Proc. Natl. Acad. Sci. USA 86:7470-7474. Evans, T., G. Felsenfeld, and M. Reitman. 1990. Control of globin gene transcription. Annu. Rev. Cell Biol. 6:95-124. Faisst, S., and S. Meyer. 1992. Compilation of vertebrateencoded transcription factors. Nucleic Acids Res. 20:3-26. Forrester, W., U. Novak, R. Gelinas, and M. Groudine. 1989. Molecular analysis of the human beta-globin locus activation region. Proc. Natl. Acad. Sci. USA 86:5439-5443. Forrester, W. C., C. Thompson, J. T. Elder, and M. Groudine. 1986. A developmentally stable chromatin structure in the human 3-globin gene cluster. Proc. Natl. Acad. Sci. USA 83:1359-1363. Fraser, P., S. Pruzina, M. Antoniou, and F. Grosveld. 1993. Each hypersensitive site of the human 3-globin locus control region confers a different developmental pattern of expression on the globin genes. Genes Dev. 7:106-113. Fried, M., and D. Crothers. 1981. Equilibria and kinetics of lac repressor-operator interactions by polyacrylamide gel electrophoresis. Nucleic Acids Res. 8:6505-6525. Fritsch, E., R. Lawn, and T. Maniatis. 1980. Molecular cloning and characterization of the human beta-like globin gene cluster. Cell 19:959-972. Gallarda, J. L., K. P. Foley, Z. Yang, and J. D. Engel. 1989. The beta-globin stage selector element (SSE) factor is erythroid specific promoter/enhancer binding protein NF-E4. Genes Dev. 3:1845-1859. Garner, M. M., and A. Rezvin. 1981. A gel electrophoresis method for quantifying the binding of proteins to specific DNA regions: application to components of the Escherichia coli lactose operon system. Nucleic Acids Res. 9:3047-3067. Gerster, T., P. Matthias, M. Thali, J. Jiricny, and W. Schaffner. 1987. Cell type-specificity elements of the immunoglobulin heavy chain gene enhancer. EMBO J. 6:1323-1330.
,yb1-
23. Gong, Q.-H., J. Stern, and A. Dean. 1991. Transcriptional role
24. 25. 26.
27.
of a conserved GATA-1 site in the human s-globin gene promoter. Mol. Cell. Biol. 11:2558-2566. Graham, F. L., and A. J. van der Eb. 1973. A new technique for the assay of infectivity of human adenovirus 5 DNA. Virology 52:456-467. Grosveld, F., G. B. van Assendelft, D. Greaves, and G. Kollias. 1987. Position-independent, high-level expression of the human beta-globin gene in transgenic mice. Cell 51:975-985. Gumucio, D. L., D. A. Shelton, W. J. Bailey, K. Hayasaka, J. L. Slightom, and M. Goodman. 1993. Phylogenetic footprinting reveals unexpected complexity in trans factor binding upstream from the £ globin gene. Proc. Natl. Acad. Sci. USA 90.6018-6022. Gutman, P. D., S. X. Cao, H. P. G. Dave, M. Mittelman, and A. N. Schechter. 1992. Binding of erythroid and non-erythroid nuclear proteins to the silencer of the human a-globin-encoding gene. Gene 110:197-203.
28. Hardison, R., K.-M. Chao, M.
Adamkiewicz, D. Price, J. Jack-
son, T. Zeigler, N. Stojanovic, and W. Miller. Positive and negative regulatory elements of the rabbit embryonic e-globin gene revealed by an improved multiple alignment program and functional analysis. DNA Sequence, in press. 29. Henikoff, S. 1984. Unidirectional digestion with exonuclease III creates targeted breakpoints for DNA sequencing. Gene 28:351359. 30. Herschlag, D., and F. B. Johnson. 1993. Synergism in transcriptional activation: a kinetic view. Genes Dev. 7:173-179.
7468
TREPICCHIO ET AL.
31. Jane, S. M., P. A. Ney, E. F. Vanin, D. L. Gumucio, and A. W. Nienhuis. 1992. Identification of a stage selector element in the human -y-globin gene promoter that fosters preferential interaction with the 5' HS2 enhancer when in competition with the P-promoter. EMBO J. 11:2961-2969. 32. Kioussis, D., F. Wilson, K. Khazaie, and F. Grosveld. 1985. Differential expression of human globin genes introduced in K562 cells. EMBO J. 4:927-931. 33. Kovach, J. S., P. A. Marks, E. S. Russell, and H. Epler. 1967. Erythroid cell development in fetal mice: ultrastructural characteristics and hemoglobin synthesis. J. Mol. Biol. 25:131-142. 34. Kulozik, A. E., S. Bail, A. Bellan-Koch, C. R. Bartram, E. Kohne, and E. Kleihauer. 1991. The proximal element of the 0-globin locus control region is not functionally required in vivo. J. Clin. Invest. 87:2142-2146. 35. Lindenbaum, M. H., and F. Grosveld. 1990. An in vitro globin gene switching model based on differentiated embryonic stem cells. Genes Dev. 4:2075-2085. 36. Luckow, B., and G. Schutz. 1987. CAT constructions with multiple unique restriction sites for the functional analysis of eukaryotic promoters and regulatory elements. Nucleic Acids Res. 15:5490. 37. MacGregor, G. R., and C. T. Caskey. 1989. Construction of plasmids that express E. coli beta-galactosidase. Nucleic Acids Res. 17:2365. 38. McKnight, S. L., and R. Kingsbury. 1982. Transcriptional control signals of a eukaryotic protein coding gene. Science
217:316-325. 39. Mignotte, V., L. Wall, E. de Boer, F. Grosveld, and P.-H. Romeo. 1989. Two tissue-specific factors bind the erythroid promoter of the human porphobilinogen deaminase gene. Nucleic Acids Res. 17:37-54. 40. Motamed, K., C. Bastiani, Q. Zhang, A. Bailey, and C.-K. J. Shen. 1993. CACC box and enhancer response of the human embryonic F globin promoter. Gene 123:235-240. 41. Orkin, S. H. 1990. Globin gene regulation and switching: circa 1990. Cell 63:665-672. 42. Orkin, S. H. 1992. GATA-binding transcription factors in hematopoietic cells. Blood 80:575-581. 43. Peschle, C., F. Mavilio, A. Care, G. Migliaccio, A. R. Migliaccio, G. Salvo, P. Samoggia, S. Petti, R. Guerriero, M. Marinucci, D. Lazzaro, G. Russo, and G. Mastroberardino. 1985. Haemoglobin switching in human embryos: asynchrony of t-a and e--y-globin switches in primitive and definitive erythropoietic lineages. Nature (London) 313:235-238. 44. Peters, B., N. Merezhinskaya, J. F. X. Diffley, and C. T. Noguchi. 1993. Protein-DNA interactions in the a-globin gene silencer. J. Biol. Chem. 268:3430-3437. 45. Raich, N., T. Enver, B. Nakamoto, B. Josephson, T. Papayannopoulou, and G. Stamatoyannopoulos. 1990. Autonomous developmental control of human embryonic globin gene switching in transgenic mice. Science 250:1147-1149. 46. Raich, N., T. Papayannopoulou, G. Stamatoyannopoulos, and T. Enver. 1992. Demonstration of a human e-globin gene silencer with studies in transgenic mice. Blood 79:861-864.
MOL. CELL. BIOL. 47. Rowley, P. T., B. M. Ohlsson-Wilhelm, L. Wisniewski, C. B. Lozzio, and B. B. Lozzio. 1984. K562 human leukemia cell passages differ in embryonic globin gene expression. Leuk. Res. 8:45-54. 48. Sanger, F., S. Nicklen, and A. R. Coulson. 1977. DNA sequencing with chain-terminating inhibitors. Proc. Natl. Acad. Sci. USA 74:5463-5467. 49. Seed, B., and J.-Y. Sheen. 1988. A simple phase-extraction assay for chloramphenicol acyltransferase activity. Gene 67: 271-277. 50. Shih, D. M., R. J. Wall, and S. G. Shapiro. 1990. Developmentally regulated and erythroid-specific expression of the human embryonic ,-globin gene in transgenic mice. Nucleic Acids Res. 18:5465-5472. 51. Shih, D. M., R. Wall, and S. G. Shapiro. 1993. A 5' control region of the human e-globin gene is sufficient for embryonic specificity in transgenic mice. J. Biol. Chem. 268:3066-3071. 52. Simon, J., and J. Lis. 1987. A germline transformation analysis reveals flexibility in the organization of heat shock consensus elements. Nucleic Acids Res. 15:2971-2988. 53. Stamatoyannopoulos, G., and A. Nienhuis. 1987. The molecular basis of blood diseases. W. B. Saunders Co., Philadelphia. 54. Stuve, L. L., and R. M. Myers. 1990. A directly repeated sequence in the ,B-globin promoter regulates transcription in murine erythroleukemia cells. Mol. Cell. Biol. 10:972-981. 55. Trepicchio, W. L., M. A. Dyer, R. C. Hardison, and M. H. Baron. Submitted for publication. 56. Tuan, D., W. Solomon, Q. Li, and I. M. London. 1985. The P-like globin gene domain in human erythroid cells. Proc. NatI. Acad. Sci. USA 82:6384-6388. 57. Wada-Kiyama, Y., B. Peters, and C. T. Noguchi. 1992. The e-globin gene silencer: characterization by in vitro transcription. J. Biol. Chem. 267:11532-11538. 58. Walters, M., and D. I. K. Martin. 1992. Functional erythroid promoters created by interaction of the transcription factor GATA-1 with CACCC and AP-1/NF-E2 elements. Proc. Natl. Acad. Sci. USA 89:10444-10448. 59. Watt, P., P. Lamb, and N. J. Proudfoot. 1993. Distinct negative regulation of the human embryonic globin genes t and s. Gene Expression 3:61-75. 60. Weintraub, S. J., C. A. Prater, and D. C. Dean. 1992. Retinoblastoma protein switches the E2F site from positive to negative element. Nature (London) 358:259-261. 61. Whitelaw, E., S.-F. Tsai, P. Hogben, and S. H. Orkin. 1990. Regulated expression of globin chains and the erythroid transcription factor GATA-1 during erythropoiesis in the developing mouse. Mol. Cell. Biol. 10:6596-6606. 62. Yu, C.-Y., K. Motamed, J. Chen, A. D. Bailey, and C.-K. J. Shen. 1991. The CACC box upstream of human embryonic a globin gene binds Spl and is a functional promoter element in vitro and in vivo. J. Biol. Chem. 266:8907-8915. 63. Zenke, M., T. Grundstrom, H. Matthes, M. Wintzerith, C. Schatz, A. Wildeman, and P. Chambon. 1986. Multiple sequence motifs are involved in SV40 enhancer function. EMBO J. 5:387-397.
