ABSTRACT
In an effort to understand how the heme biosynthetic pathway is uniquely
regulated in erythroid cells, we examined the structure of the gene encoding
murine
[delta]
-aminolevulinate dehydratase (ALAD; EC4.2.1.24), which is the second enzyme
of the pathway. The gene contains two first exons, named 1A and 1B, which are
alternatively spliced to exon 2, where the coding region begins. Each first
exon has its own promoter. The promoter driving exon 1A expression is TATA-less and contains many GC boxes. In contrast, the exon 1B promoter bears
regulatory sequences similar to those found for
[beta]
-globin and other erythroid-specific genes. Tissue distribution studies reveal that ALAD mRNA
containing exon 1A is ubiquitous, whereas mRNA containing exon 1B is found only
in erythroid tissues. This finding, together with our further observation that GATA-1 mRNA levels increase 3-fold during maturation of murine erythroid progenitor cells, may help explain simultaneous
3-fold increases in exon 1B expression. The unexpected result that exon 1A
expression also increases 3-fold during CFU-E maturation may be attributable to the action of NF-E2, since there is a potential binding site in a position
analogous to the NF-E2 site in the locus control region of the
[beta]-globin gene cluster.
The level of heme synthesized in erythroid cells is orders of magnitude greater
than that in other cells (
1
) and, under normal conditions, is stoichiometrically matched with the levels of
globin chain synthesis. It is becoming apparent that the genetic mechanisms responsible for this considerable quantitative difference are often distinctive for each of the eight heme biosynthetic enzymes. For
example, the polypeptide chain of the first enzyme in the heme pathway, [delta]-aminolevulinate synthase (ALAS) is encoded by two divergent genes
in chicken (
2
) and in humans, one ALAS gene, ALAS-N, is located on an autosome (
3
) and the other, ALAS-E, is X-linked (
4
). The autosomal gene is expressed in all tissues and has regulatory protein
binding sites similar to those for other housekeeping genes (
5
) while the X-linked one is expressed only in erythroid cells and has some of the same
regulatory sites found for other erythroid-specific genes; namely, GATA-1 and NF-E2 (
6
,
7
). In murine erythroleukemia (MEL) cells, it has been shown that there is an inverse relationship between ALAS-N and ALAS-E expression: ALAS-E mRNA increases and ALAS-N mRNA decreases upon DMSO-induction (
8
).
In contrast, all mRNAs for the third enzyme in the pathway, porphobilinogen
deaminase (PBG-D), arise from the same locus (
9
) but there is differential splicing according to tissue type, leading to
housekeeping and erythroid-specific isozymes. A typical TATA-less promoter is found upstream of the housekeeping variant (
10
) whereas GATA-1 and NF-E2 sites are found upstream of the erythroid-specific version (
11
,
12
). Amongst expression patterns for other heme pathway enzymes thus far examined,
those for uroporphyrinogen decarboxylase (
13
), coprophyrinogen oxidase (
14
), protoporphyrinogen oxidase (
15
) and ferrochelatase (
16
), the fifth, sixth, seventh and eighth enzymes, respectively, appear to have messengers and polypeptide structures alike in all tissues. Although it is unclear what mechanisms are
responsible for augmented expression of uroporphyrinogen decarboxylase (
17
) and coprophyrinogen oxidase (
14
) in erythroid cells, mechanisms for differential tissue expression of ferrochelatase include a multipurpose promoter region, which includes not
only binding sites for housekeeping transcriptional factors, but also ones for
GATA-1 and NF-E2 (
18
) and a downstream erythroid repressor (
19
).
In this report, we show how the regulation of the second enzyme in the heme pathway, [delta]-aminolevulinate dehydratase (ALAD), illustrates yet another way in
which differential expression in erythroid and non-erythroid cells can be accomplished. We find that expression of ALAD exons
is alike in all tissues except for two untranslated first exons, 1A and 1B.
These are differentially spliced to the second exon where the translation start
signal is located. Consequently, while ALAD enzyme is identical in all tissues,
ALAD mRNA occurs in housekeeping (1A) and erythroid-specific (1B) forms. In both man and mouse, the promoter region upstream
of exon 1B contains GATA-1 sites, a result which may help explain observed increases in ALAD-1B mRNA during erythropoiesis. In contrast, while the promoter
region immediately 5' of exon 1A resembles that of other housekeeping genes, the region 2.3 kb
upstream of exon 1A bears a site which may be bound by the erythroid regulatory
protein, NF-E2. This site, if it functions in a manner analogous to the enhancer NF-E2 binding site found 50 kb upstream of the human [beta]-globin gene (
20
), could help account for our finding that ALAD-1A mRNA also increases during erythropoiesis.
MEL cell lines, M18b (
21
) and 270-2 (
22
) were cultured with 10% fetal bovine serum (Hyclone) in Minimum Essential
Medium-[alpha] (Life Technologies) and induced to differentiate with 1.5 or 2%
DMSO, respectively.
CFU-E were enriched by thiamphenicol treatment of BALB/cByJ mice (Jackson
Laboratory) and purified by centrifugal elutriation and Percoll density
gradients as previously described (
23
-
25
). CFU-E were cultured at 37oC under 5% CO
2
, in media containing 30% fetal bovine serum and, as indicated recombinant human erythropoietin (gift of Genetics Institute). Morphologically, CFU-E purity at the outset was 85-90% in Wright-stained cytospins (Shandon) and 80% of these formed 32-cell erythroid colonies after 40 h (
26
).
We used an SM/J mouse genomic library (gift of Steven Weaver, University of
Chicago) constructed by partial
Mbo
I digestion of genomic DNA, followed by size-selection and ligation into the
Bam
HI site of [lambda]L47.1. Phage from this library were grown on
E.coli
LE392 and screened according to Maniatis
et al
. (
27
), using both ALAD3 and ALAD7 rat cDNA probes (
28
). Hybridizing plaques were purified by three successive rounds of plating and
rescreening. Recombinant [lambda] DNA was purified by the plate-lysate method of Maniatis
et al
. (
27
). Gel-purified restriction fragments of [lambda]mALAD-1 and [lambda]mALAD-2 were subcloned into pIBI30, pIBI31
(International Biotechnologies), pEMBL18 (gift from S. Lazarowitz, Carnegie
Institution, Baltimore, MD) or pBluescriptIIKS
-
(Stratagene). Standard procedures were used for all subsequent nucleic acid
manipulations (
27
). Both strands were used to determine nucleotide sequences using Sequenase (US
Biochemical) chain termination reaction as suggested by the manufacturer.
RNA was harvested either by the method of Ullrich
et al
. (
29
) or Chomczynski and Sacchi (
30
) using one additional phenol/CHCl
3
(1:1) extraction. RNA from yolk sacs was derived from 9.5 day post coitum F2
embryos of BALB/c *C57Bl/6J matings. Fetal liver RNA from 14.5 day embryos was provided by
Dave Bodine (NIH). S1 nuclease assays (
31
,
32
) were performed following overnight hybridization of 50 [mu]g total RNA and 30 fmol
32
P end-labeled DNA probe. Non-hybridized probe and single-stranded carrier DNA were digested at 30oC with 300 U/ml S1 nuclease for 90 min.
Sph
I-
Sty
I (690 bp) and
Sca
I-
Bam
HI (420 bp) genomic ALAD fragments (Fig.
1
B) were used to protect exon 1A- and exon 1B-containing transcripts, respectively.
For RNase protection assays (
33
,
34
), an ALAD exon 1A-specific probe was prepared by subcloning a 160 bp
Rsa
I-
Ava
II fragment from [lambda]mALAD-2 (delimited by vertical arrows in Fig.
3
) into pIBI31. The orientation of the cloned fragment was determined by
nucleotide sequencing. An exon 1B-specific probe was constructed by subcloning a 138 bp
Xba
I-
Pst
I fragment from an A-PCR clone into pIBI31. In later studies, mild Exonuclease III digestion
was used to remove the poly(dG) tail from the 5'-end of the exon 1B probe. Removal was verified by nucleotide
sequence analysis. Plasmid DNA, purified by CsCl gradient centrifugation and
linearized with
Eco
RI, was used to prepare radiolabeled antisense riboprobes with T7 RNA polymerase (Life Technologies).
Each RNase protection assay utilized 5 or 20 [mu]g of test RNA and enough phenol-extracted, ethanol-precipitated
E.coli
tRNA to make the total RNA mass equal to 30 [mu]g, plus 50 000 d.p.m. of radiolabeled antisense riboprobe. RNaseA and
RNaseT1 both overdigested the RNA even at the lowest concentrations tested
(data not shown), whereas 15 U/ml RNaseT2 left protected fragments intact. We
suppose that the successful use of T2 reflects the elevated GC content of ALAD
exon riboprobes and a cleavage specificity for A residues.
The method of Loh
et al
. (
35
) was used with the following modifications. Poly(A)
+
RNA from mouse liver or DMSO-induced MEL cells was used as template for AMV reverse transcriptase (Life
Sciences) with oligo(dT) as a primer. After alkali treatment, poly(dG) tails
were added to the cDNA using terminal deoxynucleotidyl-transferase (Life Technologies). Tailed cDNA (8 ng) was used in A-PCR with 1 [mu]M ALAD oligonucleotide 25 (oligo 25) and 1 [mu]M of a 1:9 mixture of oligo 2 and oligo 3 anchors. The oligo
25 sequence, spanning residues 368 to 387 in the translated murine ALAD
sequence (
36
), was ATGGGAGGTGTAGGGGCACA; anchor oligo 3 (
35
) was TGCGGCCGCGGATCCGAATTC; and anchor oligo 2 was the same as oligo 3 plus a
poly(C)
14
tail.
Taq
polymerase amplification (Perkin-Elmer Cetus) was carried out for 25 cycles: denaturing at 94oC for 1 min, annealing at 55oC for 1.5 min and elongating at 72oC for 2.5 min. Products were electrophoretically
fractionated on a low-melt agarose gel (FMC). Melted gel slices were subjected to a second round
of amplification using oligo 3 and ALAD oligo 24 primers. The oligo 24
sequence, spanning murine ALAD 319-338, was GGGAAGGTCTTCCTCAGCAG. A-PCR products were ligated into
Bam
HI- and
Eco
RV-digested pBluescriptIIKS (Stratagene) and plasmid DNA from clones of
transformed
E.coli
HB101 cells analyzed by dideoxynucleotide sequencing.
Poly(A)
+
RNA from CFU-E cultured for various times was subjected to electrophoresis in 1%
agarose, 2.2 M formaldehyde gels (
27
); transferred to GeneScreen Plus (DuPont/NEN); and hybridized with
32
P-dCTP-oligolabeled probes (Pharmacia) prepared from mouse ALAD subcloned
fragments. Blots were melted and rehybridized to radioactive [beta]-actin cDNA (
37
) and autoradiograph band intensities were quantitated using a PhosphorImager and
ImageQuant software (Molecular Dynamics, Sunnyvale, CA 94086).
Plaque lifts (
27
) were prepared from a human chromosome 9-specific library (ATCC #57781) and probed with a 350 bp
Eco
RI-
Bam
HI fragment from the mouse ALAD exon 1B promoter (delimited by downward arrows
in Fig.
4
). The
Eco
RI site was generated by PCR amplification with a primer that mismatched the
genomic sequence by one base. Plaques were picked and rescreened until a
purified plaque was isolated. Southern blot analysis revealed a 1.2 kb
Sst
I-
Xba
I fragment that contained the human ALAD exon 1B and its promoter and this
fragment was subjected to nucleotide sequence analysis (
38
).
Approximately 900 000 recombinant plaques were screened using radiolabeled rat
cDNA inserts from clones ALAD3 (
39
) and ALAD7 (
28
). Three positively hybridizing plaques were isolated. All three recombinant
DNAs had identical restriction enzyme maps and were named [lambda]mALAD1 (Fig.
1
A). A 3.1 kb
Bam
HI fragment from the 5'-end of [lambda]mALAD1 (marked by asterisks in Fig.
1
A) was used to rescreen an additional 820 000 plaques from the same library. One
new clone, out of six acquired, contained a 17 kb insert extending ~11 kb upstream from the 5'-end of [lambda]mALAD1 and was named [lambda]mALAD2 (Fig.
1
A).
cDNA hybridization, subcloning and nucleotide sequence analysis of [lambda]mALAD1 and [lambda]mALAD2 were used to identify exons 2 through 12 (Fig.
1
B). Exon lengths ranged in size from 37 to 187 bp and all splice sites matched
the AG/GT rule (
40
,
41
). The coding sequence in exons 2 through 12 (data not shown) corresponds
exactly to the sequence of mouse reticulocyte ALAD cDNA (
36
). Analysis of genomic sequence further upstream from the end of the cDNA
homology revealed that exon 2 was flanked by a splice acceptor site (Fig.
2
, junction of non-shaded and shaded sequences) and not by any obvious transcriptional
promoter elements. Such findings suggested the existence of at least one more
upstream non-coding exon, which echoed our earlier discovery of a rat ALAD processed
pseudogene whose sequence included 160 nondescript nucleotides upstream of what
would have been the start of the coding region (
42
). We therefore supposed that reported ALAD cDNAs from mice (
36
), rats (
28
) and humans (
43
), are in fact truncated copies of the mRNA.
As depicted in Figure
5
A, a genomic fragment containing ALAD exon 1A protected RNA from brain, liver
and spleen from S1-nuclease digestion, whereas an exon 1B-containing genomic fragment protected RNA only in spleen samples.
Protected fragment sizes ranged from 37 to 100 bases in the case of exon 1A and
from 56 to 87 bases in the case of exon 1B. The prominent smallest bands
correspond to positions +1 in Figures
3
and
4
.
In fetal life (Fig.
6
A), the abundance of exon 1B transcripts correlated with the extent of
erythropoiesis. They were barely evident in 9.5 d yolk sacs, where erythrocytes
are confined to the blood islands, but readily perceptible in 14.5 d fetal
liver, where definitive erythropoiesis is abundant. In adults, exon 1B transcripts are scarce in freshly harvested CFU-E (0 h), but then increase to reach a maximum between 21 and 28 h of
culture (Fig.
6
B), when heme biosynthesis also approaches an apex (
26
). No significant difference could be observed on the level of exon 1B
transcripts in CFU-E grown with or without erythropoietin.
A 350 bp fragment containing the mouse erythroid promoter and the 5' half of exon 1B was used as a probe to isolate the human ALAD clone. Using computer-assisted nucleotide sequence alignments, it was found that only 250 bp of the mouse fragment had
70% of nucleotide sequence similarity with the human DNA (Region 2, Fig.
7
). A comparison between the nucleotide sequences from three species was used to
reveal conserved and thus functionally important regions of ALAD erythroid-specific promoters. Region 2 also contains the longest stretch of identity
(23 bp) among the three species. This stretch contains a double CACCC box,
resembling a potential erythroid kruppel-like factor binding site (
48
). Likewise, two of the three potential GATA-1 sites (asterisks in Fig.
7
; -60 to -23 in mouse and rat ALAD, Fig.
4
) are identical in all three species. The region between these two GATA-1 sites is identical between mouse and rat and shares 79% identity between
mouse and human.
In freshly harvested murine CFU-E, GATA-1 mRNA levels are readily detectable (Fig.
8
). During CFU-E maturation, GATA-1 mRNA levels increase 3-fold relative to actin mRNA levels.
After 21 h of CFU-E maturation, both 1A and 1B mRNA levels increase relative to actin (Table
1
), which disproved our
a priori
hypothesis that only the tissue-specific, exon 1B, would be regulated during erythropoiesis. Likewise, on
two other northern blots probed with ALAD cDNA, increases of ALAD mRNA compared
with actin were quantitated to be 2.8 and 3.6 (data not shown). When equal
masses of RNA were used in RNase protection assays, the greatest amount of exon
1B was at 21 hours (Fig.
6
B). The similarity of results using these two different methods makes it
unlikely that the changes observed in ALAD mRNA levels are due solely changes
in actin levels during CFU-E maturation. Furthermore, a similar increase in both ALAD exons 1A and 1B
occurs during DMSO-induction of MEL cells using an S1-nuclease protection assay (data not shown).
The endproduct of the heme biosynthetic pathway, heme, must attain
stoichiometric levels that match [alpha]- and [beta]-globin production during the substantial increase of
hemoglobin accumulation during erythropoiesis. In other words, regulation of a
pathway must be coordinated with the regulation of single genes. Heme synthesis
increases ~400-fold during the maturation of CFU-E (S.H. Boyer, unpublished) and, since the pathway consists of
eight enzymes (
49
), the genes encoding the heme synthesis enzymes must be regulated in a precise
manner.
Table 1
In the liver, a classical regulatory mechanism is in place: the first enzyme of
the heme pathway, ALAS, is rate-limiting and thus, the entire pathway is upregulated when the amount of
ALAS increases. On the other hand, the regulation of the pathway in erythroid
cells is more complex and it has now been demonstrated that the mRNA level for
each of the early heme enzymes increases with different magnitudes during DMSO-induction of MEL cells (
50
). In cases where the nucleotide sequence of the gene promoter has been
reported, there are GATA-1 sites (for ALAS, PBGD, UROD and FC) and NF-E2 sites (for ALAS, PBGD and FC). However, the presence and distance
from the transcription start site are very species-specific. For example, in the case of the PBGD, the human erythroid
promoter contains an NF-E2 site at -160, but the mouse does not in the same position; the murine PBGD
erythroid promoter contains a double CACCC box, whereas the human gene contains
only one (
51
). In this paper, we show that the ALAD gene has a promoter which is utilized
only in erythroid cells and that it contains at least two potential GATA-1 binding sites. Furthermore, there is an NF-E2 site located 2.3 kb upstream from the housekeeping promoter,
possibly in a distal enhancer. This is reminiscent of [beta]-globin gene regulation, where binding of NF-E2 to the LCR (along with other factors) opens up the
chromatin domain, allowing DNase I hypersensitivity (
52
) and interaction with transcriptional activators, primarily GATA-1, which bind to the proximal promoter.
Thus, a common theme develops with regards to the type of upstream regulatory
elements in the promoters of target genes during erythroid differentiation, but
the spatial arrangements of binding sites and therefore, flanking DNA
sequences, are distinct, not only for each gene in the pathway, but also in the
same gene from different species. Our characterization of the ALAD gene
demonstrates that still another genetic structure has evolved which may be
responsive to genetic regulation in an erythroid-specific manner.
While the increase in ALAD mRNA levels during CFU-E maturation measured in this paper was small, it was consistently detected using three different techniques. The relative contribution of exon 1A- and 1B-containing transcripts is difficult to measure precisely, but
estimates may be made from both S1-nuclease protection assays and northern blot analyses, in cases where the
radiolabeled probes had similar specific activities. In spleen RNA (Fig.
5
A), it may be clearly seen that exon 1B transcripts are predominant over exon 1A
transcripts and during CFU-E maturation, estimates are that exon 1B transcripts comprise about 2/3 to
3/4 of all ALAD transcripts. This estimation assumes that 1A and 1B probes
hybridized with equal efficiency and takes into account the length difference
between the two probes.
Measurements of ALAD enzyme levels during CFU-E maturation and calculations based on the known amounts of hemoglobin in BALB/c RBCs,
demonstrate that ALAD is not a rate-limiting enzyme of the heme biosynthetic pathway in erythroid cells (S.H.
Boyer, unpublished). In fact, ALAD levels are usually 20 times higher than
required to participate in the synthesis of the amount of heme eventually
realized in the RBC (S. H. Boyer, unpublished). We hypothesized that ALAD might
be performing another function besides its role as the second enzyme of the
heme biosynthetic pathway and in fact, Guo
et al
. (
53
) reported that ALAD is the 240 kDa proteasome inhibitor, CF-2. Thus, ALAD acts as an inhibitor of protein degradation via the
proteasome. This function may play a critical role in the rapid hemoglobin
accumulation observed during erythropoiesis. A third, speculative function of
ALAD may be that it is a scavenger of lead, since it has been demonstrated to
reversibly bind lead. A better understanding of these multiple roles is needed
to completely understand the regulation of the ALAD gene. These multiple ALAD
functions may be so critical to erythropoiesis that mutations in the gene are
only rarely viable: only four cases of ALAD-deficiencies have been reported in the literature (
54
).
ACKNOWLEDGEMENTS
We thank Andrea Noyes, Ophelia Rogers and Larry Frelin for their excellent
technical assistance, Dave Bodine for his patient coaching of mouse embryo
dissections, Greg Preston for sharing his RNA samples, members of the Kato and
Bishop labs for manuscript reviewing and especially, Dr Samuel H. Boyer for his
constant encouragement and support. This work was supported by grant DK-38052 from the National Institutes of Health.
Present Addresses:
+
Department of Pathology, University of Washington, Seattle, WA 98195, USA and
[sect]
American Cyanamid, Princeton, NJ 08543, USA
Time of culture (h)
Exon 1A
a
Exon 1B
b
0
1.00
1.00
7
1.28
0.84
14
0.71
0.83
21
3.87
3.04
28
1.77
1.76
35
1.72
1.56
REFERENCES
Return