ABSTRACT
The DNA repair proteins XRCC1 and DNA ligase III are physically associated in
human cells and directly interact
in vitro
and
in vivo
. Here, we demonstrate that XRCC1 is additionally associated with DNA polymerase-
[beta]
in human cells and that these polypeptides also directly interact. We also
present data suggesting that poly (ADP-ribose) polymerase can interact with XRCC1. Finally, we demonstrate that
DNA ligase III shares with poly (ADP-ribose) polymerase the novel function of a molecular DNA nick-sensor, and that the DNA ligase can inhibit activity of the latter
polypeptide
in vitro
. Taken together, these data suggest that the activity of the four polypeptides
described above may be co-ordinated in human cells within a single multiprotein complex.
Cellular processes of DNA repair are fundamental to the maintenance of genetic
integrity and survival, and their inactivation by mutation can dramatically
increase individual susceptibility to cancer (
1
-
4
). The use of rodent cell mutants for biochemical studies and for gene cloning has greatly enhanced our understanding of mammalian DNA repair processes in recent years. Two of these mutants,
denoted EM9 and EM-C11, are Chinese hamster ovary cell lines that are unable to efficiently
rejoin DNA single strand breaks resulting from exposure to agents that induce
DNA base-damage, and consequently are hypersensitive to these compounds (
5
-
8
). The human gene that fully corrects the repair defect in EM9 and EM-C11 has been cloned and is denoted
XRCC1
(
6
). Consistent with its proposed role in single strand break rejoining, XRCC1
polypeptide is physically associated in mammalian cells with DNA ligase III (
9
,
10
). On the basis of the cellular phenotype of EM9 and EM-C11, we have proposed that XRCC1 is involved in DNA base excision repair
in mammalian cells, at a step subsequent to enzymatic incision of the
phosphodiester bond at the damaged nucleoside but prior to or at DNA ligation (
9
,
10
). The importance of this polypeptide to mammals is exemplified by the
observation that mice in which the
XRCC1
locus has been removed by gene targeting are inviable (
11
). The biochemical function of XRCC1 is unknown, but the polypeptide most likely
fulfils a novel role since it does not exhibit strong homology with any other
known protein. Levels of DNA ligase III polypeptide are reduced in EM9 and EM-C11, suggesting that XRCC1 is required for physical stability of the DNA
ligase (
9
,
10
,
12
). However, physical stabilisation of this protein is unlikely to reflect the
primary role of XRCC1. Rather, XRCC1 is most likely required for some aspect of
DNA ligase III activity, and/or may interact and affect other components of
BER.
The N-terminus of DNA ligase III possesses a putative zinc finger motif that
exhibits homology with a `nick sensing' zinc finger present in the nuclear
protein, poly (ADP-ribose) polymerase (PARP) (
13
,
14
). It is not obvious why DNA ligase III would require a zinc finger since no
other known DNA ligase has one, indicating that they are not required for DNA
ligation,
per se
. The zinc finger motif present in PARP confers upon the protein an ability to
recognise and rapidly bind DNA containing strand breaks, including those
arising during BER (
15
-
18
). Following DNA binding, PARP catalyses the transfer of ADP-ribose from nicotinamide adenine dinucleotide (NAD
+
) to itself and to a number of other nuclear proteins. This process results in
the synthesis in large negatively charged poly (ADP-ribose) polymers which facilitate dissociation of the modified PARP
protein from DNA, allowing other enzymes to access and repair the DNA strand
breaks (
16
,
17
for recent reviews). The significance of this cyclical DNA binding by PARP is
unclear, as is the cellular role of the enzyme. Nevertheless, irrespective of
the precise role of PARP, the presence in DNA ligase III of a functional PARP-like `nick sensing' zinc finger would suggest that the activities of these
polypeptides are co-ordinated during BER.
In this study, we have begun to address two questions concerning the XRCC1
complex: namely, the identity of any additional polypeptides that associate
with XRCC1 and whether the putative zinc finger motif in DNA ligase III does
confer `nick-sensing' activity upon this polypeptide.
Recombinant human XRCC1 protein containing a C-terminal decahistidine tag (XRCC1-His) was purified by immobilised metal-chelate affinity chromatography (IMAC) from
Escherichia coli
as described (
10
). Recombinant human PARP was expressed in baculovirus and purified as described
(
19
). Recombinant DNA ligase III was expressed in
E.coli
from the construct pET16BHL3, which was generated by inserting a
Cel
lII-
Dra
I fragment isolated from cDNA clone HGS473238 (
13
) into the
Cel
lII-
Hin
dIII site of pET16B (Novagen). To facilitate ligation, recessed 3' termini produced by
Hin
dIII were converted to blunt ends with Klenow fragment. The recombinant DNA
ligase (His-DNA ligase III) possessed an N-terminal decahistidine tag to facilitate purification by IMAC (as
described for XRCC1) and was fully active as judged by its activity on defined
oligonucleotide substrates when compared with DNA ligase III partially purified
from HeLa nuclear extract (unpublished observations). DNA ligase III
polypeptide lacking an intact zinc finger motif (His-DNA ligase III
Zn-
) was expressed from the construct pET16BHL3
[delta]158-170
and purified by IMAC as described for XRCC1. The construct pET16BHL3
[delta]158-170
was generated from pET16BHL3 by the removal of 39 base pairs (encoding 13 amino
acids) with
Xma
I and
Kpn
I restriction enzymes and subsequent treatment with T4 DNA polymerase and T4 DNA
ligase. Truncated DNA ligase III polypeptide comprised of the first 242 amino
acids (His-DNA ligase III
1-242
), including the putative zinc finger motif, was expressed in
E.coli
from the construct pET16BHL3
1-242
. His-DNA ligase III
1-242
was purified by IMAC under denaturing conditions (6 M guanidine-HCl), since this polypeptide was largely insoluble in
E.coli
(unpublished observations). Following purification, His-DNA ligase III
1-242
was renatured by step-wise dialysis in 50 mM Tris-HCl, 0.1 M NaCl, 10% glycerol, 1 mM DTT and decreasing
concentrations of guanidine-HCl, with ZnCl
2
included in all but the final step. The construct pET16BHL31-242
was generated from pET16BHL3 by digestion with
Hin
dIII and subsequent religation.That recombinant polypeptides of the expected
sizes were expressed and purified was confirmed by SDS-PAGE (data not shown), and protein concentrations determined by a
coomassie blue assay procedure (BioRad).
Escherichia coli
harbouring a human DNA polymerase-[beta] expression construct were kindly provided by Dr Sam Wilson, and
were grown at 30oC to an OD 600 ~0.5 after which expression of the DNA polymerase was induced by
incubation at 42oC. Aliquots (8 ml) of the culture were sampled prior to and 4 h following
induction, and pelleted cells resuspended in 0.5 ml (pre-induction) or 1.5 ml (post-induction) 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5% glycerol, 1 mM DTT, 1 mM PMSF. After freeze-thawing, cell suspensions were sonicated (6* 15 s bursts on ice; 30 s intervals on ice) and
insoluble material pelleted (microfuge, 15 min). Protein concentrations were
typically 1.5-2.0 mg/ml.
TL-25 anti-DNA ligase III polyclonal antibodies were raised against His-DNA ligase III, and along with anti-DNA ligase I monoclonal antibody TL-5, was a gift from Tomas Lindahl. Anti-DNA polymerase-[beta] polyclonal antibodies were kind
gifts from Sam Wilson. Anti-PARP monoclonal antibodies were raised against recombinant PARP (S.
Aoufouchi and C. Milstein, unpublished data), and p450 anti-DNA-PK catalytic subunit (DNA-PK
cs
) polyclonal antibodies (31-4) were raised against purified human polypeptide and kindly provided by
Steve Jackson. Anti-RPA (p70) and anti-PCNA (PC10) monoclonal antibodies were kind gifts from Rick Wood,
and anti-XRCC1 monoclonal antibody 33-2-5 has been described previously (
10
).
Oligonucleotides were synthesised by Zeneca Pharmaceuticals and purified by
HPLC. To generate oligonucleotide duplexes, complementary oligonucleotides were
mixed in equal molar ratios, heated to 70oC for 10 min, and allowed to cool slowly to room temperature to allow
annealing. NaCl was added to 50 mM and oligonucleotide duplexes stored at -20oC. Radiolabelled oligonucleotides (2 [mu]g) were prepared, prior to annealing with unlabelled
complementary oligonucleotides, by incubation with T4 polynucleotide kinase (25
U) and [[gamma]-
32
P]ATP (40 [mu]Ci) in the presence of 10 [mu]M `cold' ATP. Unincorporated nucleotides were removed by sephadex G50
spin-column chromatography.
Radiolabeled oligonucleotide duplexes (0.2 pmol; 10 ng) were incubated with 50-150-fold excess, by weight (0.5-1.5 [mu]g), of either supercoiled plasmid DNA or 12mer
oligonucleotide duplex (
Eco
RI linker; New England Biolabs) and 5 pmol of either His-DNA ligase III (0.5 [mu]g), His-DNA ligase III
Zn-
(0.5 [mu]g) or His-DNA ligase III
1-242
(0.18 [mu]g) on ice in 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM DTT, 0.1 mg/ml BSA. After 10 min, loading buffer was added and samples subjected to non-denaturing polyacrylamide gel electrophoresis (5% gels, BioRad Mini
Protean II apparatus) at 20 mA for 2-4 h in pre-chilled 1* TBE, unless otherwise stated. Following electrophoresis,
gels were fixed (10% acetic acid, 10 min), dried, and subject to
autoradiography.
Recombinant human PARP (0.1 [mu]g; 1 pmol) was incubated in the absence or presence of sonicated calf thymus
DNA (40 ng) either with BSA (0.35 [mu]g) and recombinant His-DNA ligase III (0.5 [mu]g; 4.5 pmol), BSA (0.35 [mu]g) and His-XRCC1 (0.35 [mu]g; 4.5 pmol), or with XRCC1-DNA ligase III complex 0.85 [mu]g; 4.5 pmol, pre-formed on ice for 10 min) at room
temperature in 50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM DTT, 30 [mu]M ZnCl
2
, 2 mM MgCl
2
and [adenylate-
32
P]NAD+ (4 [mu]Ci; 8 nmol). After 10 min, reactions were stopped by the addition of SDS-PAGE loading buffer, subjected to SDS-PAGE, and analysed by autoradiography. DNA strand breaks were
limiting for PARP activity under the conditions described (results not shown).
ATP was omitted from reactions to prevent ATP-dependent DNA ligase activity.
Anti-XRCC1, PARP, DNA ligase III and RPA antibodies were incubated with 300 [mu]g HeLa crude nuclear extract (CNE; ref.
20
) in 25 mM Tris-HCl, pH 7.5, 0.2 M KCl, 5% glycerol, 1 mM DTT (300 [mu]l total) for 30 min on ice, after which 30 [mu]l protein-A-coupled Sepharose beads (BioRad) were added and
incubation continued for a further 30 min with frequent mixing. After removing
50 [mu]l of the final suspension for samples of CNE prior to immunoprecipitation
(`CNE'), the protein-A-Sepharose beads were pelleted gently in a microfuge (1 s: 3000
r.p.m.) and the immunoprecipitate washed extensively (5* 500 [mu]l) with wash buffer (25 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5% glycerol, 1 mM DTT). The
immunoprecipitate was washed finally with 1* 100 [mu]l wash buffer, which was kept for analysis (`wash' samples), and the
immunoprecipitate (`IP') resuspended in a further 100 [mu]l of wash buffer. Aliquots of `CNE' samples (5 [mu]l; ~5 [mu]g total nuclear protein), `wash' samples (10 [mu]l) and `IP' samples (10 [mu]l; protein precipitated from ~15 [mu]g total nuclear protein) were subject to SDS-PAGE, electroblotted, and immunoblotted
with appropriate antibodies. DNA ligase III was detected by immunoblotting, or
as radiolabelled polypeptides following adenylylation in the presence of [[alpha]-
32
P]ATP as previously described (
10
). Unless indicated, ethidium bromide (10-50 [mu]g/ml) was present in the initial immunoprecipitation reactions to
minimise DNA-protein interactions.
XRCC1-His (6 [mu]g) or His-DNA ligase III (10 [mu]g) was mixed with
E.coli
cell extract (75 [mu]g) lacking (pre-induction; see above) or containing (post-induction; see above) human DNA polymerase-[beta] in sonication buffer (see above) containing
imidazole (10 mM) and fresh PMSF (1 mM) in a final volume of 220 [mu]l. After 30 min on ice, 25 [mu]l bed volume of NTA-agarose was added to bind histidine-tagged proteins and incubation continued on ice with
frequent mixing. After 25 min, a 60 [mu]l sample of the suspension was removed and stored on ice (`mix' samples).
The NTA-agarose in the remaining reaction mixture was gently pelleted by
centrifugation for 1 s in a microfuge at ~3000 r.p.m. The supernatant containing non-adsorbed material was removed and the NTA-agarose beads washed with 4* 200 [mu]l sonication buffer containing 25 mM imidazole and
finally 1* 50 [mu]l sonication buffer containing 25 mM imidazole. The supernatant from
the final wash was retained for analysis (`final wash'). XRCC1-His, His-DNA ligase III and other bound polypeptides were finally eluted from
the NTA-agarose beads by 2* 25 [mu]l washes with sonication buffer containing 250 mM imidazole.
Aliquots of 5 or 30 [mu]l of `mix' samples, 25 [mu]l of the final wash samples and 25 [mu]l of the combined NTA-agarose eluates were fractionated through 12% SDS-polyacrylamide gels and visualised by coomassie blue
staining or immunoblotting, as indicated.
XRCC1, PARP, DNA ligase III and DNA polymerase-[beta] open reading frames were subcloned into pACTII and/or pAS1CYH2
vectors (Clontech) to generate chimaeric open reading frames additionally
encoding the GAL4 activation domain or the GAL4 DNA binding domain,
respectively. These constructs, denoted pAS-XRCC1, pAS-PARP, pAS-Polb, pACT-XRCC1, pACT-PARP and pACT-Lig3, were introduced into
Saccharomyces cerevisiae
(Y190) either singly or pair-wise and transformants selected on minimal media plates containing
tryptophan (for pACTII
+
, pAS1CYH2
-
cells) or leucine (for pACTII
-
, pAS1CYH2
+
cells), or lacking both (for pACTII
+
, pAS1CYH2
+
cells), as appropriate. Those transformants expressing polypeptides that
interacted in this system were identified by the presence of [beta]-galactosidase activity and by their ability to grow on plates
lacking histidine (which additionally contained 50 mM 3-aminotriazole to prevent growth resulting from leaky expression of the
endogenous histidine gene). To detect [beta]-galactosidase activity, minimal media plates containing independent
transformants and, in later experiments, suspensions of individual
transformants spotted onto YEPD plates, were lifted onto filter paper (3MM, Whatman), freeze-thawed for 10 s in liquid nitrogen, and incubated at 30oC for 1-16 h on filter paper (3MM) soaked in PBS additionally
containing 2 mM MgSO
4
and 2 mg/ml 5-bromo- 4-chloro-3-inodoyl-[beta]-D-galactopyranoside (X-Gal, Boehringer
Mannheim). To select for loss of pAS constructs, Y190 transformants harbouring appropriate pACT and pAS constructs were grown for multiple generations on YEPD, to facilitate spontaneous loss of plasmid DNA, and replated onto minimal media plates containing
tryptophan and cyclohexamide (2.5 [mu]g/ml).
We have previously reported that XRCC1 is directly associated in human cells
with DNA ligase III (
9
,
10
,
13
). To identify other possible polypeptide components of the XRCC1-DNA ligase III complex immunoprecipitates were isolated from HeLa nuclear
extract using anti-XRCC1 antibodies, under conditions in which protein-DNA interactions were minimised (see Materials and Methods), and
examined for the presence of other polypeptides involved in DNA repair. One
polypeptide considered as a possible candidate for associating with XRCC1-DNA ligase III was PARP, since this protein has been implicated in BER.
Consistent with this notion, we observed significant co-immunoprecipitation of PARP with XRCC1 and DNA ligase III by anti-XRCC1 antibodies (Fig.
1
, lane 3). Additional experiments revealed that XRCC1, DNA ligase III and PARP
were also co-precipitated by anti-DNA ligase III antibodies (lane 6) and anti-PARP antibodies (lane 4), but not by control anti RPA
antibodies (lane 5; PARP and DNA ligase III only), suggesting that the
polypeptides can associate as a single complex. A second polypeptide considered
candidate for associating with the XRCC1 complex was DNA polymerase-[beta] (pol [beta]) since this polypeptide is involved in BER and its activity
has been reported to be affected by XRCC1
in vitro
(Y. Kubota and T. Lindahl, personal communication). Furthermore, pol [beta] had previously been observed to co-immunoprecipitate with PARP from human nuclear extract (S.
Aoufouchi, unpublished observations). Consistent with these observations, we
observed significant immunoprecipitation of pol [beta] with anti-XRCC1 antibodies (Fig.
1
, lane 3), and complementary experiments revealed that XRCC1 and PARP were
immunoprecipitated by anti-pol [beta] antibodies (results not shown).
Yeast two-hybrid analysis was employed to identify the interactions responsible for
co-immunoprecipitation of pol [beta] with the XRCC1. Neither XRCC1, DNA ligase III, pol [beta] nor PARP polypeptides were themselves able to activate
transcription from GAL4 responsive promoters as indicated by the failure of
appropriate Y190 transformants to grow in the presence of 3-aminotriazole (3-AT), indicative of histidine auxotrophy, and by the absence of [beta]-galactosidase activity (Fig.
2
A-D, top panel, and results not shown). However, Y190 cells harbouring both
XRCC1 and DNA ligase III fusion proteins exhibited histidine prototrophy
(compare Fig.
2
A and B; `s-xrcc1: t-lig3') and contained significant [beta]-galactosidase activity (Fig.
2
C), consistent with the known interaction of these polypeptides (
9
,
10
). In addition, primary Y190 transformants co-expressing XRCC1 and pol [beta] fusion proteins also resulted in histidine prototrophy (compare
Fig.
2
A and B; `s-polb: t-xrcc1') and [beta]-galactosidase activity (Fig.
2
C), indicating that XRCC1 also interacts with DNA polymerase-[beta]. However, we failed to detect any interaction between pol [beta] and either PARP or DNA ligase III in this assay (Fig.
2
; `s-polb: t-parp' and `s-polb: t-lig3'). To address the specificity of the interaction
between XRCC1 and pol [beta] in this assay, we examined XRCC1 for its ability to interact with a
number of control polypeptides, including itself. None of the resulting two-hybrid transformants resulted in histidine prototrophy or [beta]-galactosidase activity, confirming that the interaction of
XRCC1 with pol [beta] was specific (Fig.
2
; `s-xrcc1: t-xrcc1', and results not shown).
To dissect the protein interactions responsible for co-immunoprecipitation of PARP with the XRCC1 complex we again utilised the
yeast two-hybrid assay. As described above, PARP and pol [beta] did not interact in this system indicating that the co-immunoprecipitation of PARP with the XRCC1 complex was not
mediated through these polypeptides. PARP and DNA ligase III similarly failed
to interact in the two-hybrid assay, as indicated by the absence of histidine prototrophy and
significant [beta]-galactosidase activity in Y190 transformants harbouring the appropriate constructs (Fig.
2
; `s-parp: t-lig3'). However, PARP and XRCC1 did appear to interact, though in
contrast to the interaction of XRCC1 with DNA ligase III and pol [beta], the interaction between XRCC1 and PARP was not observed in primary
transformants unless selected for on plates containing 3-AT. Nevertheless, histidine prototrophic clones selected in this manner
contained significant [beta]-galactosidase activity (Fig.
2
C; `s-xrcc1: t-parp'), and subsequent removal of the PAS-XRCC1 plasmid from such clones (see Materials and Methods)
resulted in concurrent loss of both phenotypes (results not shown). This
indicates that [beta]-galactosidase activity and histidine prototrophy were dependent on
the presence of the XRCC1 construct and suggests that the apparent interaction
was genuine.
We have previously reported that human XRCC1 and DNA ligase III polypeptides
physically interact
in vitro
and that the two proteins are tightly associated in mammalian cells (
9
,
10
). In this report, we demonstrate that XRCC1 also interacts with DNA polymerase-[beta],
in vitro
and
in vivo
, and additionally report a possible interaction between XRCC1 and PARP.
Finally, we report that DNA ligase III possesses a novel `nick sensing' zinc
finger motif of the type present in PARP, suggesting that this polypeptide may
fulfil a unique role unusual to DNA ligases.
XRCC1, DNA ligase III and PARP were co-immunoprecipitated from HeLa nuclear extract by antibodies specific for
each of the three polypeptides, suggesting that the three polypeptides can
associate within a single complex. The observed interaction between XRCC1 and
pol [beta] suggests that this polypeptide is also a component of the complex, though
the possibility that pol [beta] is in a separate complex with XRCC1 has not been excluded. The notion of
co-ordinating polypeptides involved in BER as a multiprotein complex is
attractive, since the sequential activity of PARP, pol [beta] and DNA ligase III could fulfil the remaining steps of BER following DNA
glycosylase and AP endonuclease activity. However, one caveat to the
constitutive presence of PARP within such a complex is the relative abundance
of this protein compared with XRCC1, pol [beta] and DNA ligase III. Since PARP is very abundant, much of this polypeptide
cannot be constitutively bound to XRCC1 complex. It is possible, therefore,
that XRCC1 interacts only with a specific form of PARP, such as that already
bound to DNA, for example. The observed requirement for selection on media
containing 3-AT for apparent interaction between XRCC1 and PARP in the two-hybrid system could reflect such complexity, since specific mutation
or very high levels of expression of one or both polypeptides may be required
for significant interaction in the absence of DNA damage. Clearly, however,
further work is required to confirm that PARP associates with the XRCC1 complex
and that it can interact directly with the latter polypeptide.
The biochemical function of XRCC1 is unclear, but its interaction with pol [beta]
in vitro
and
in vivo
is consistent with a role for the polypeptide in BER, since an essential role
for the DNA polymerase in this process has already been demonstrated (
21
,
22
). One possible function of XRCC1 in this process is as a molecular chaperone,
guiding XRCC1 complex to sites of repair by binding to DNA intermediates of
BER, for example. Alternatively, XRCC1 may possess an as yet unidentified
enzymatic activity, such as modulating the activity of pol [beta], PARP or DNA ligase III in some way, possibilities that are currently
under investigation. Modulation of DNA ligase III activity is the most likely
of these possibilities, since XRCC1 is more intimately related to this
polypeptide than to pol [beta] or PARP. This notion is suggested by the comparative dependency of the
three polypeptides on XRCC1 for physical stability, since cellular levels of
the DNA ligase are reduced 3-6-fold in XRCC1 mutants whereas levels of pol [beta] and PARP are unaffected (
10
; K. Caldecott, unpublished observations).
Concerning the role of the `nick sensing' zinc finger in DNA ligase III, it is
possible that this motif can target the polypeptide, and therefore the XRCC1
complex, to DNA strand breaks arising during BER. It remains to be determined
whether DNA ligase III and PARP compete
in vitro
or
in vivo
for binding DNA strand breaks, or whether there are subtle differences in
specificity and/or affinity for nicked DNA intermediates which prevent direct
competition. The observed ability of DNA ligase III to inhibit PARP activity on
sonicated calf-thymus DNA is consistent with the former notion, though that the mechanism
of this inhibition involves direct competition for DNA breaks requires
confirmation. It is also possible that,
in vivo
, the respective `nick-sensing' activities of these polypeptides are co-ordinated during BER, such that unwanted interference of each
polypeptide with the other does not occur.
A multiprotein complex that can conduct BER
in vitro
was recently identified in testis extract, and contains at least pol [beta] and DNA ligase I (
23
). The relationship between this complex and the putative XRCC1 complex
described here is unclear at present, though the failure of DNA ligase I to co-immunoprecipitate with XRCC1 (Fig.
1
; ref.
10
) suggests that they are different. It is conceivable that multiple complexes
may be involved in BER since the involvement of multiple BER pathways has been
proposed in eukaryotic cells, on the basis of the apparent involvement in this
process of multiple DNA ligases and DNA polymerases (
24
-
26
).Ultimately, biochemical purification of the XRCC1 complex and detailed
analysis of its component polypeptides is required to fully characterise this
structure and to clarify its role in the cellular response to genomic insult.
We are indebted to Steve Jackson, Rick Wood and Tomas Lindahl for providing
antibodies, and to Sam Wilson for providing pol-[beta] antibodies and cDNA. We also thank John Hickman for support and
encouragement. This work was supported by Zeneca Pharmaceuticals and the
Medical Research Council (G9603130).
REFERENCES
Return