ABSTRACT
The
Escherichia coli
AlkB protein is involved in protecting cells against mutation and cell death
induced specifically by SN
2
-type alkylating agents such as methyl methanesulfonate (MMS). A human cDNA
encoding a polypeptide homologous to
E.coli
AlkB was discovered by searching a database of expressed sequence tags (ESTs)
derived from high throughput cDNA sequencing. The full-length human AlkB homolog (
hABH
) cDNA clone contains a 924 bp open reading frame encoding a 34 kDa protein
which is 52% similar and 23% identical to
E.coli
AlkB. The
hABH
gene, which maps to chromosome 14q24, was ubiquitously expressed in 16 human
tissues examined. When
hABH
was expressed in
E.coli alkB
mutant cells partial rescue of the cells from MMS-induced cell death occurred. Under the conditions used expression of
hABH
in skin fibroblasts was not regulated by treatment with MMS. Our findings show
that the AlkB protein is structurally and functionally conserved from bacteria
to human, but its regulation may have diverged during evolution.
Alkylating agents from both exogenous and endogenous sources contribute to
alkylation on more than a dozen types of DNA lesions, some of which can lead to
mutation and cell death. In mammals, mutations caused by DNA alkylation can
also lead to neoplasia. Based on their cytotoxicity, alkylating agents are
employed as antiviral drugs and are used in chemotherapeutic treatment of
cancers (
1
,
2
).
DNA alkylation damage repair is best studied in
Escherichia coli
, where two defense mechanisms are responsible for protecting the accuracy of
the genetic information against attack by alkylating agents. One mechanism,
which is constitutive, depends on expression of the
ogt
gene, encoding an
O
6
-methylguanine (
O
6
MeG) methyltransferase (MTase) (
3
,
4
), and the
tag
gene, encoding a 3-methyladenine (3MeA) DNA glycosylase (
5
,
6
). The second mechanism is induced upon exposure to a sublethal dose of
alkylating agent and is called the adaptive response (
7
-
9
). Four genes are involved in this system:
ada
,
alkA
,
aidB
and
alkB
.
The
ada
gene encodes a DNA MTase which removes the methyl group from
O
6
MeG and
O
4
-methylthymine (
O
4
MeT) to its active site Cys321 residue. Since
O
6
MeG and
O
4
MeT mispair during DNA replication, repair by the Ogt and Ada MTases protects
E.coli
from alkylation-induced mutation (
8
,
10
). In addition, Ada MTase transfers a methyl group from methylphosphotriester
DNA lesions to a second cysteine active site (Cys69), upon which the Ada
protein is transformed into a transcriptional activator for the four genes of
the adaptive response, including itself.
alkA
, the second Ada-regulated gene, like
tag
, encodes a DNA glycosylase which repairs 3MeA, 3-methylguanine,
O
2
-methylthymine and
O
2
-methylcytosine (
11
,
12
). The AlkA and Tag glycosylases specifically protect cells from alkylation-induced cell death, since 3MeA is a lethal lesion which blocks DNA
replication (
13
,
14
). Eukaryotic homologs of
O
6
MeG MTases have been cloned in yeast (
15
), mouse (
16
) and human (
17
,
18
) and 3MeA glycosylase homologs have been cloned in yeast (
19
-
21
), rat (
22
), mouse (
23
), human (
24
-
26
) and
Arabidopsis thaliana
(
27
). Although some of these eukaryotic enzymes are inducible by alkylating agents,
there is no conclusive evidence suggesting that eukaryotes have the same
adaptive response mechanism to alkylating agents as
E.coli
(
28
,
29
). Unlike the above genes, little is known about the function of the
aidB
product in DNA alkylation damage repair.
The precise biochemical function of the fourth member of the adaptive response,
AlkB, is not clear. However, genetic studies indicate its importance in
protecting cells from alkylating agent-induced DNA damage.
Escherichia coli alkB
mutant cells are extremely sensitive to MMS-induced mutation and cell death (
30
), suggesting that the AlkB pathway is extremely effective in defending against
alkylation toxicity. At least two lines of evidence indicate that AlkB acts in
repairing rather than preventing DNA alkylation damage: MMS-treated [lambda] phage survive better in wild-type cells than in
alkB
-
cells (
30
) and wild-type and
alkB
mutant
E.coli
genomes are alkylated to roughly the same extent when exposed to dimethyl
sulfate (
31
). In
E.coli
the AlkB repair pathway is probably independent of the AlkA glycosylase repair
pathway, based on the following observations: (i) the sensitivity of an
alkB alkA
double mutant to MMS and
N
-methyl-
N
'-nitrosoguanidine (MNNG) is at the level predicted from the additive
sensitivity of each of the single mutants (
32
); (ii) purified
E.coli
AlkB does not reveal any glycosylase activity (
33
); (iii) wild-type
alkB
cannot complement an
alkA
mutant (
31
). Although the 27 kDa
E.coli
AlkB protein has been purified to homogeneity and a number of studies have been
attempted to define its function, its mechanism of action remains unknown (
33
,
34
). Expression of
E.coli
AlkB in alkylation-resistant and alkylation-sensitive human cell lines rescues the cells from MMS-induced cell death (
31
), indicating that
E.coli
AlkB may function independently in mammalian cells and that mammalian cells
might have a similar AlkB repair pathway. To date, no eukaryotic AlkB homolog
has been identified. Identification of the eukaryotic AlkB homolog would
represent an important step towards understanding the function of AlkB and may
help elucidate a new pathway for eukaryotic alkylation damage repair. In
addition, a number of DNA repair genes are involved in human neoplasia, e.g.
the XP excision repair genes involved in xeroderma pigmentosum and Cockayne's
syndrome (
35
) and mismatch repair genes involved in hereditary non-polyposis colon cancer (
36
-
40
). Identification of new human DNA repair genes may lead to the discovery and
identification of new disease genes.
In order to identify potential human AlkB homologs we searched a human cDNA
database generated by high throughput cDNA sequencing and analysis using the
expressed sequence tag (EST) method (
41
). One EST, derived from a human synovial sarcoma cDNA library, was found to be
homologous to the C-terminal half of
E.coli
AlkB. The full-length hABH protein shares significant sequence homology with
E.coli
AlkB and is able to partially rescue an
E.coli alkB
mutant from MMS-induced cell death, suggesting that the AlkB protein is structurally and
functionally conserved during evolution, like many other DNA repair enzymes.
Escherichia coli
strains HK81 (
thr-1 leu-6 proA2 his-4 argE3 thi-1 lacY1 galK2 ara-14 xyl-5 mtl-1 tsx-33 strA31 supE44 nalA
) and HK82 (as HK81 but
alkB22
), kindly provided by Dr Leona Samson (Harvard School of Public Health, Boston,
MA), were used as the hosts of wild-type and mutant
alkB
respectively. Human skin fibroblast cell line CCD-965SK was purchased from the American Type Culture Collection. Plasmid
vector pQE-9 was purchased from Qiagen. MMS was purchased from Aldrich Chemical Co.
Inc. The cDNA libraries were constructed using the Stratagene Uni-ZAP XR cDNA kit.
A database containing ~500 000 human ESTs has been generated through the combined efforts of The
Institute for Genomic Research and Human Genome Sciences Inc. using high
throughput automated DNA sequence analysis of randomly selected human cDNA
clones (
41
,
42
). Sequence homology comparisons of each EST were performed against the GenBank
database using the blastn and tblastn algorithms (
43
). ESTs having homology with previously identified sequences (
P
<= 0.01) were given a tentative name based on the name of the sequence to
which it was homologous. A specific homology search using the known
E.coli
AlkB amino acid sequence against this human EST database revealed one EST
having >34% homology.
The partial
hABH
cDNA (~1.3 kb in length) identified within the human EST database was radiolabeled
and hybridized to 20 nylon membranes containing 600 000 individual plaques from
a human synovial sarcoma Uni-ZAP II cDNA library. Hybridization and washing were under the conditions
described by Maniatis
et al
. (
44
). Three positive clones, each ~0.6 kb longer than the original partial cDNA clone, were identified.
Human total RNA isolation was modified from the RNAzol B procedure of Biotecx
Laboratories Inc. Briefly, 2 g tissue were mixed with 40 ml RNAzol B solution
and homogenized on a Polytron PT3000 (Brinkmann) at 22 000 r.p.m. for 2-3 min. Then 5 ml chloroform was added to the homogenate and the mixture
was vortexed for 15 s and incubated on ice for 5 min. The suspension was
centrifuged at 12 000
g
at 4oC for 15 min and the upper phase homogenate was mixed with an equal volume
of isopropanol and incubated at 4oC for 5 min to precipitate RNA. The samples were centrifuged again at 12
000
g
for 15 min and the pellet was resuspended in 2 ml DEPC-treated water and extracted twice with an equal volume of
phenol/chloroform (1:1 v/v, equilibrated with DEPC-water). The RNA was ethanol precipitated and resuspended in water. An
aliquot of 15 mg total RNA was electrophoresed in a 1% agarose-formaldehyde gel (
44
), transferred to a nylon membrane and hybridized with the 1.9 kb
hABH
cDNA fragment. The blot was washed three times with 0.1* SSC and 0.1% SDS at 50oC for 1 h and subjected to autoradiography.
The 1 kb open reading frame (ORF) coding for the hABH protein was amplified by
the polymerase chain reaction (PCR) using the oligonucleotides 5'-gcgc
Functional complementation of hABH in
E.coli alkB
mutant cells was measured by the cell survival rate on MMS plates. The wild-type and mutant
E.coli
cells harboring pRSV
alkB
(
E.coli alkB
; see
31
), pQE-9 vector or pQE9hABH were grown in LB/ampicillin medium to log phase.
Cells were diluted with M9 medium and plated on LB/ampicillin plates containing
0, 0.02, 0.035 or 0.05% MMS and the plates incubated at 37oC overnight. The cell survival rate at each dose was calculated by the
number of colonies at that dose divided by the number of colonies on non-MMS plates.
Genomic clones of
hABH
were isolated from an lFIX-2 library (Stratagene) using standard procedures (
44
). A genomic clone containing a portion of the
hABH
gene was nick-translated using digoxigenin-dUTP (Boehringer Mannheim) and fluorescence
in situ
hybridization was carried out as detailed in Johnson
et al
. (
45
). Individual metaphase chromosome spreads were counter stained with DAPI and
color digital images, containing both the DAPI and gene signals, were recorded
using a triple bandpass filter set (Chroma Technology Inc., Brattleburo, VT) in
combination with a charged couple device camera (Photometrics Inc., Tucson, AZ)
and variable excitation wavelength filters (
46
). The same chromosome spreads were then stained using conventional g-banding procedures and images were again recorded. Chromosomes from the
DAPI/gene images of the same spreads were aligned with the g-band images using the ISEE software package (Inovision Corp., Durham, NC),
allowing assignment of precise g-band positions.
As most of the DNA repair proteins are conserved during evolution, we reasoned
that human AlkB might be homologous to
E.coli
AlkB. To identify a potential human AlkB homolog the amino acid sequence of
E.coli
AlkB (216 residues) was compared with a human cDNA database containing ~500 000 human ESTs using the blastn and tblastn algorithms (
43
). A 249 bp EST derived from a human synovial sarcoma cDNA library was
identified which had 34% identity and 59% similarity at the amino acid level to
a region within the C-terminal portion of
E.coli
AlkB (data not shown). This EST clone was discovered as part of a joint
collaboration between scientists at The Institute for Genomic Research and
Human Genome Sciences. The cDNA clone from which this EST was derived had a 1.3
kb cDNA insert. There was no similarity between the two genes at the nucleotide
level.
Sequencing of the 1.3 kb cDNA insert revealed that it contained an ORF truncated
at the 5'-end. The translated polypeptide sequence from this ORF shares
homology with the C-terminal region of
E.coli
AlkB. We therefore searched for the full-length cDNA of this gene by screening the human Uni-ZAP XR II synovial sarcoma cDNA library from which the initial EST
was obtained. As described in Materials and Methods, 600 000 plaques were
screened and three positive clones, each ~600 bp longer at the 5'-end than the original clone, were obtained.
These three new clones were identical based on their restriction enzyme
digestion pattern and 5'- and 3'-end sequence information. The nucleotide sequence of
this cDNA clone reveals a 1953 bp insert containing a 924 bp ORF extending from
bp 200 to 1124 and a poly(A) tail at the 3'-end (Fig.
2
). The polypeptide encoded by this ORF (Fig.
2
) showed significant homology to
E.coli
AlkB (Fig.
3
). Therefore, this clone probably contains the full-length
hABH
cDNA. There are two putative translation initiation codons at the beginning of
this ORF, one at bp 200 (immediately after an in-frame upstream stop codon) and one at bp 224. The ATG at bp 224 is likely
to be the actual translation initiation site, since it lies in an optimal
translation start consensus (
47
) and the homology to AlkB starts immediately thereafter (Fig.
3
), whereas the ATG at position 200 is preceded by an unfavored sequence (
47
) and there is no sequence homology to AlkB in the region from bp 200 to 230
(Fig.
3
). Assuming that translation of
hABH
starts at the ATG at bp 224 and ends at the first in-frame stop codon at bp 1124, hABH contains 299 amino acid residues
encoding a 34.040 kDa protein. This agrees with the protein size of
in vitro
translated hABH (data not shown). Two consensus AATAAA polyadenylation signals
(
48
) located at bp 1868 and 1912, which are 61 and 19 bp respectively upstream of
the poly(A) tail, were observed. The amino acid sequences of hABH and
E.coli
AlkB are conserved throughout the coding region, with 52% similarity and 23%
identity (Fig.
3
). A stretch of 104 amino acids within the middle of the protein sequence (boxed
in Fig.
3
) reveals the best homology, with 34% identity. The nucleotide sequence of
hABH
cDNA has been deposited in GenBank with the accession no. X91992.
To determine the expression pattern of human AlkB in different tissues we
carried out an RNA blot analysis on 16 normal adult human tissues, including
brain, kidney, small intestine, testis, pancreas, prostate, heart, liver, lung,
thymus, spleen, placenta, colon, ovary, leukocyte and muscle. As shown in
Figure
4
,
hABH
RNA is ubiquitously present in all these tissues as a 2.1 kb message,
suggesting that
hABH
is probably a housekeeping gene that plays a fundamental role in most human
tissues.
To examine whether hABH plays a similar role to
E.coli
AlkB in defending cells against alkylating agent-induced cell death we expressed an N-terminal tagged hABH in
E.coli alkB
mutant cells and tested for its ability to protect the
E.coli alkB
mutant from MMS-induced cell death. The bacterial expression construct pQE9hABH (Fig.
1
) was prepared as described in Materials and Methods. As it is not certain which
ATG is the actual translation initiation codon, we inserted the entire ORF from
the first ATG at bp 200 to the first in-frame stop codon at bp 1124 into the pQE-9 vector. The hABH expressed from this plasmid contains 307 amino
acid residues from its coding region fused with 14 amino acids from the vector,
including six histidines at the N-terminus. The His tag will facilitate protein purification by nickel
affinity column chromatography.
The plasmid pQE9hABH was transformed into
E.coli alkB
mutant strain HK82 and the MMS resistance of the transformants was examined by
a plate assay (see Materials and Methods). As a control, plasmid vector pQE-9 was transformed into wild-type HK81 and
alkB
mutant HK82 cells. Six independent HK82/pQE9hABH transformants were tested,
each of which showed an increased survival rate of ~10-fold (Fig.
5
).
Escherichia coli alkB
can fully complement HK82 under the same conditions (Fig.
5
). This result suggests that hABH plays a similar role to
E.coli
AlkB in protecting cells against alkylating agent-induced cell death. It also indicates that the AlkB alkylation damage repair pathway has been conserved during evolution.
Expression of
E.coli
AlkB is inducible upon treatment with a low dose of alkylating agents as part of the adaptive response to alkylating agents (
7
,
49
). In addition, it has been shown that eukaryotic
O
6
MeG MTase and 3MeA glycosylase can be induced by MNNG and MMS in certain
organisms and cell types but not in others (
28
,
29
,
50
,
51
). Transcription of human
O
6
MeG MTase MGMT and 3MeA-DNA glycosylase ANPG is preferentially induced in transformed cells by DNA
damaging agents (
50
). To test whether hAlkB is inducible by the DNA alkylating agent MMS we used
human skin fibroblasts, a cell type in which regulation of gene expression by
various DNA damaging agents, including MMS, has been well studied previously (
57
). Human skin fibroblast CCD-965SK cells were treated with MMS at concentrations ranging from 0 to
0.08% for 4 h. Under these conditions cell survival was reduced from 100 to
0.003%. Total RNA was isolated from the cells treated at each dose and
subjected to Northern analysis (Fig.
6
). The results showed that expression of hABH is unaltered upon MMS treatment
under the experimental conditions described above, indicating that expression
of hABH in human skin fibroblasts may be regulated differently from
E.coli alkB
.
Because mutations in other human DNA repair enzyme genes are implicated in
genetic diseases, particularly cancer (
35
), it was important to determine the precise chromosomal location of
hABH
. A genomic DNA fragment encoding part of the
hABH
gene was isolated and used for fluorescence
in situ
hybridization to human metaphase chromosome spreads (
52
). Approximately 50 spreads were analyzed by eye, most of which had a doublet
signal characteristic of genuine hybridization on at least one chromosome 14.
The doublet signal was not detected on any other chromosome. Detailed analysis
of 36 individual spreads, using post-hybridization g-banding and high resolution image analysis, indicated that the
hABH
gene is positioned within the distal portion of band 14q24, just at the border
of band 14q31 (Fig.
7
).
The human AlkB protein described here is quite similar to the previously known
bacterial gene, providing strong evidence for the functional importance and
evolutionary conservation of the DNA alkylation repair pathway. Similarities
between the human and bacterial AlkB proteins includes homology in their
respective primary structures and an apparently strong functional conservation,
indicated by the ability of the human gene to partially rescue bacterial
alkB
mutant cells. The two proteins share large areas of amino acid homology,
particularly within a 102 residue stretch in the middle portion of each protein
(boxed in Fig.
3
). Although the precise biochemical role of AlkB remains unknown, it is
reasonable to suggest that these highly conserved regions may represent core
functional domains. In addition to the areas of highly conserved sequence,
hAlkB is significantly larger and appears to contain five insertions of non-homologous domains (Fig.
3
). These five insertions form two clusters; the first two insertions are
clustered at the N-terminus of the 102 residue highly conserved region and the last three
insertions are clustered at the C-terminus of this region. These regions, like the more conserved areas of
the protein, do not share sequence homology with any other known genes or
functional domains. It is possible that these insertions reflect increased
complexity of DNA alkylation repair pathways in higher eukaryotes. For example,
other proteins may participate in the repair process through interaction with
these regions. At present no other higher eukaryotic AlkB homologs are known,
so sequence comparisons cannot be done. A few putative yeast genes have been
shown to be able to complement
E.coli alkB
mutant cells (
55
). However, the yeast genes do not share primary sequence homology with either
bacterial or human AlkB.
The existence of a human counterpart to bacterial AlkB had been previously
suggested by the observation that
E.coli
AlkB can function independently in humans to protect cells against DNA
alkylation-induced cell death (
31
). Our data suggest that the human protein can function similarly in bacterial
cells. Interestingly, hABH only partially rescued the
E.coli alkB
mutant phenotype (Fig.
5
), which may be due either to inefficient expression of hABH in
E.coli
or to variations in the microenvironment of
E.coli
which are not optimal for hABH activity. Alternatively, the non-homologous regions within the larger human protein might function as
regulatory elements, with maximal functioning of the human protein requiring
interaction with cellular proteins not present in
E.coli
. However, because the precise biochemical function of AlkB is unknown and there
is no assay system available, it is hard to identify the specific role of
different regions within either the human or bacterial proteins. It should be
noted that the
E.coli
alkB
gene can fully complement
E.coli alkB
mutant cells (
31
). As we used His-tagged hAlkB for the complementation experiment, it would be interesting
to see if His-tagged
E.coli
AlkB can still function fully or non-tagged hABH can rescue
E.coli alkB
better than tagged hABH. The lack of full complementation in our experiment may
also be due to the extra eight amino acids at the N-terminus, assuming that native hABH starts at the second methionine (at
the ATG at bp 224).
Despite the structural and functional conservation between human and
E.coli
AlkB, the two proteins also differ in ways besides the inserted regions of non-homologous sequence seen in the human protein. First, in
E.coli
the
alkB
gene is in the same operon as the
ada
MTase gene, whereas in the human
alkB
and
MGMT
are located on different chromosomes (we have shown here that
hABH
is on chromosome 14 and Rydberg
et
al.
have shown that human
MGMT
is on chromosome 10; see
18
). Second, expression of
E.coli alkB
is regulated by Ada as part of the adaptive response to alkylating agents (
7
-
9
). In the human no similar regulatory system has been defined, although both
MGMT and ANPG may be slightly induced upon treatment with DNA alkylating agents
in certain cell types (
50
,
51
). Our results show that expression of hABH in skin fibroblasts is not altered
by MMS under the conditions used. Hence, regulation of hABH expression, at
least in this cell type and under this treatment, is regulated differently than
in
E.coli
. Induction of human MGMT and ANPG by MMS and other DNA damaging agents has been
reported in a number of human transformed cell lines, such as a hepatoma cell
line and a glioblastoma cell line (
50
). It will be important to see if hAlkB is inducible in these cell lines.
Cloning AlkB and other homologs to bacterial DNA repair genes in various species
is an important step towards elucidating DNA repair pathways in higher
eukaryotes. However, cloning cross-species homologs with limited sequence homology can be problematical,
particularly when the precise biochemical functions of the protein are unknown.
The functional complementation approach has been successful in cloning a number
of eukaryotic DNA repair gene homologs (see, for example,
15
,
17
,
19
,
24
), but when it was used to screen for a yeast
alkB
homolog none of several genes that showed an ability to partially complement
the
E.coli alkB
mutant contained sequence homology with
alkB
(
55
). It will be important to determine whether these yeast genes are conserved in
higher eukaryotes and whether they play a role in DNA alkylation repair. An
approach commonly used to clone sequence-homologous genes in different species is to perform degenerate PCR based
on the conserved sequence motifs among the known genes; this usually requires
multiple genes to determine these motifs and is therefore unsuitable for
alkB
or the functional homologs defined in yeast. Using a large database of cDNA EST
sequences to identify homologous genes, as was used here to clone
halkB
, provides a valuable and in some cases essential approach to identifying
homologous genes. This relatively new approach has already led to
identification of other DNA repair genes, including three DNA mismatch repair
genes involved in human hereditary non-polyposis colorectal cancer (
39
,
40
).
Finally, as was the case with the above mismatch repair genes, mutations in
hABH
may be involved in human neoplasia. Alterations in chromosome band 14q24, the
site of the structural gene for hABH, have been observed in >90% of reported
cytogenetic analyses of leiomyoma cases and are seen sporadically in several
other cancers (
53
). Mutations in the neighboring band, 14q31, are only very rarely reported. In
addition, one of the loci for autosomal dominant cerebella ataxia is mapped on
chromosome 14q24-31 (
54
). There is no evidence suggesting that these diseases are directly related to a
deficiency in DNA damage repair, however, as DNA alkylation introduces
mutations, it can cause defective proteins in general. Further studies are
needed to define the potential relationship of hABH to any of these human
diseases.
We would like to thank Drs Craig Rosen, Steve Ruben and Guo-Liang Yu for their helpful comments and stimulating discussions. Many
thanks to Ms Julie Nelson for typing the references. The human tissues used in
this study were supplied by the Cooperative Human Tissue Network, which is
funded by the National Cancer Institute. The initial EST in this study was
sequenced at The Institute for Genomic Research. Nucleotide sequencing of the
hABH
cDNA clone and related constructs was carried out at Human Genome Sciences Inc.
sequencing core facility.
REFERENCES
Return