ã
Oncogene (1999) 18, 3788 ± 3792
1999 Stockton Press All rights reserved 0950 ± 9232/99 $12.00
http://www.stockton-press.co.uk/onc
SHORT REPORT
p53-Dependent growth arrest and altered p53-immunoreactivity following
metabolic labelling with 32P ortho-phosphate in human ®broblasts
Jane A Bond1, Katherine Webley1, Fiona S Wyllie1, Christopher J Jones1, Ashley Craig2,
Ted Hupp2 and David Wynford-Thomas*,1
1
Cancer Research Campaign Laboratories, Department of Pathology, University of Wales College of Medicine, Cardi€ CF4 4XN,
UK; 2Department of Molecular and Cellular Pathology, Ninewells Hospital Medical School, University of Dundee,
Dundee DD1 9SY, UK
The tumour suppressor gene p53 plays a major role in
the cellular response to DNA damage, mediating growth
arrest and/or apoptosis. Phosphorylation of the protein
occurs at numerous sites in vivo and is likely to be a
major mechanism for modulation of its activity as a
transcriptional transactivator. Not surprisingly, therefore, p53 has been intensively studied by 32P metabolic
labelling. Here we show however, using normal human
®broblasts, that typical labelling conditions induce (i) a
p53-dependent inhibition of DNA synthesis and (ii) an
increase in the cellular content of p53 protein detectable
by the phosphorylation-sensitive antibody DO-1 but not
by antibody DO-12. These data demonstrate for the ®rst
time that 32P labelling is sucient to induce a
biologically-signi®cant, p53-mediated cellular response
and strongly suggest that it perturbs the phosphorylation
state of p53 which it is being used to measure. This
highlights the need to re-evaluate earlier data by nonradioactive approaches using phospho-speci®c antibodies.
Keywords: p53; metabolic labelling; DNA damage;
phosphorylation; cell cycle
The tumour suppressor protein p53 mediates growth
arrest and/or apoptosis in response to a wide range of
cellular `stress' signals, including in particular radiation-induced DNA damage (Cox and Lane, 1995).
Although changes in abundance of the protein were
initially thought to be the principal regulatory
mechanism, it is now clear that post-translational
modi®cation, leading to activation of existing molecules, may be of greater signi®cance (Hupp et al., 1995;
Blaydes and Hupp, 1998; Woo et al., 1998; Waterman
et al., 1998). Multiple biochemical functions of p53
may be regulated in this way, of which the most wellcharacterized is undoubtedly its ability to transactivate
a wide range of DNA damage response genes, some of
which play a direct role in growth arrest (Ko and
Prives, 1996).
Although other modi®cations may be important, e.g.
glycosylation (Shaw et al., 1996) and acetylation (Gu
and Roeder, 1997), a major mechanism of post-
translational regulation of p53 activity is via phosphorylation (Meek, 1994; Mundt et al., 1997; Blaydes and
Hupp, 1998; Woo et al., 1998; Waterman et al., 1998).
The p53 protein contains at least 15 potential
phosphorylation sites, clustered mainly in the N- and
C-termini, and is potentially phosphorylatable by at
least a dozen protein kinases. More recently there has
been increasing evidence that the phosphorylation state
may also be modulated by site-speci®c phosphatases
(Waterman et al., 1998).
Not surprisingly, therefore, p53 represents one of
the most intensely studied cellular proteins with
respect to phosphorylation status. Such studies have
inevitably relied in the past on metabolic labelling
with 32P-orthophosphate, which, given the relatively
low abundance of p53 protein, requires millicurie
quantities of isotope, and typically several hours of
labelling time (Milne et al., 1995). This raises the
concern that the very presence of the label may induce
a DNA damage response, thereby distorting the
phosphorylation state of the p53 being studied.
Indeed, for two other isotopes, [3H]thymidine and
[35S]methionine, good evidence now exists that the p53
damage response pathway is signi®cantly activated by
standard labelling conditions (Dover et al., 1994;
Yeargin and Haas, 1995), a dilemma which has been
likened (Cox and Lane, 1995; Yeargin and Haas,
1995) to the classic `Schrodinger cat' problem in
nuclear physics, in which the system cannot be studied
without perturbing it.
Nevertheless this has apparently not deterred many
groups from continuing to use 32P, perhaps because no
direct evidence of this artefact has so far been provided
for this isotope, and because there are theoretical
grounds for anticipating that the much longer path
length (and hence lower linear energy transfer)
characteristic of the beta particles emitted by 32P may
make cellular damage through auto-irradiation less
likely than with 3H and 35S.
To provide objective evidence to resolve this
uncertainty we describe here an analysis of p53mediated responses to typical 32P-labelling conditions
applied to human diploid ®broblasts.
P-labelling induces a p53-dependent growth arrest
32
*Correspondence: D Wynford-Thomas
Received 24 September 1998; revised 22 January 1999; accepted 25
January 1999
Normal human diploid ®broblasts (HDF), strain
HCA2 in monolayer culture, were washed and then
incubated for 1 h in [32P]orthophosphate (Amersham
Code PBS11; 10 mCi/ml) at a ®nal activity of 0.7 mCi/
32
P-induced activation of p53
JA Bond et al
ml in phosphate-free DMEM (containing 10% FCS),
giving a total dose representative of typical metabolic
labelling conditions (Milne et al., 1995). Following
extensive washing they were then re-fed with standard
medium and replicate dishes analysed at intervals up to
18 h thereafter. Un-irradiated controls were processed
identically except that non-radioactive orthophosphate
was included in place of 32P.
Exposure to 32P resulted in a fall in the proportion of
cells undergoing nuclear DNA synthesis (S phase), as
detected by a 1 h labelling with BrdU, from 441% in
untreated controls to 14% by 9 h and to just 2% by
18 h. In unirradiated controls only a small initial drop
was observed at 8 h ascribable to perturbation
following washing and/or re-feeding.
To demonstrate that the above induction of growth
arrest by 32P-labelling was dependent on intact p53
function, we initially compared the responses of pooled
clones of HDF expressing a dominant-negative human
p53 mutant (ala143) (Kern et al., 1992) (HCA2-scx) with
corresponding controls expressing a neo gene only
(HCA2-neo). These were derived by retroviral infection
of early-passage stocks with vectors psi-CRIP-SCX
(expressing p53 ala143) and psi-CRIP-neo (Wyllie et al.,
1993) respectively, followed by G418 selection and
pooling of 450 clones from each. Loss of wt p53
function in the pooled clones expressing mutant p53
has previously been con®rmed by showing loss of the
G1/S arrest and p21WAF1 induction responses to DNA
damage induced by bleomycin (Wyllie et al., 1995;
Bond et al., 1995).
Once again, exposure of HCA2-neo cells to 32P
resulted in a fall in BrdU nuclear labelling index (LI)
to 53% by 18 h, whereas the value in unirradiated
controls was still 35% at this time point. In contrast, in
HCA2-scx, irradiated cells showed a nearly identical
BrdU LI to that of unirradiated cells at 9 h and only a
slightly lower value (27% compared to 35%) at 18 h
(Figure 1a).
As further con®rmation, we repeated the above
experiment using pooled HDF clones expressing
human papillomavirus HPV 16 E6 as an alternative
means of abrogating wild-type p53 function, again
veri®ed by lack of p21WAF1 induction in response to
DNA damage (data not shown). In this experiment
the time course was also extended to 36 h. As with
HCA2-scx, exposure of HCA2-E6 to 32P resulted in
signi®cantly less inhibition of DNA synthesis than in
neo controls. Although in this case, partial inhibition
was observed, BrdU L1 nevertheless remained above
20% in 32P-treated E6-expressing cells up to the
latest time point studied (36 h) at which time it had
fallen to below 4% in the 32P-treated neo controls
(Figure 1b). (It is not known why the neo controls
showed a slower decline in LI in this case than in
Figure 1a.)
Figure 1 32P induces p53-dependent growth arrest in human ®broblasts. (a) HCA2 ®broblasts expressing a control (neo) gene
(squares) or sister clones expressing a dominant-negative (ala143) mutant of p53 (circles) were fed either 32P (solid symbols) or
control (open symbols) media for 1 h. The proportion of cells undergoing DNA synthesis was analysed at intervals over the
following 18 h by labelling with bromo-deoxyuridine (BrdU) (Boehringer) (10 mM) for 1 h prior to ®xation and
immunocytochemical detection of incorporated nuclear BrdU (Bond et al., 1996). Note that in both cases a small shift in
labelling index was observed in both irradiated and mock-treated cells compared to the value in totally `undisturbed' cultures (large
diamonds: open for neo; shaded for mp53). (b) Repeat of (a) comparing response of control (neo) ®broblasts (squares) to sister
clones expressing HPV E6 (circles). Note that although in this extended time-course 32P induced partial inhibition even in cells
expressing E6, there is still a highly signi®cant di€erence in LI between these and the neo controls at 36 h (P50.01). All labelling
indices are means+s.e. based on a sample size of at least 1000 nuclei. Some SEs are obscured by data symbols and for clarity have
been omitted for the 0 ± 8 h time points in (a)
3789
32
P-induced activation of p53
JA Bond et al
3790
Changes in immunoreactivity of p53 protein following 32P
labelling
It is well established that activation of p53 transcription factor function by DNA damage is often (Kastan
et al., 1991; Lu and Lane, 1993) although not always
(Hupp and Lane, 1994; Lutzker and Levine, 1996)
accompanied by an increase in the steady state level of
the protein. Having shown a p53-dependent induction
of growth arrest following 32P exposure, we therefore
next tested its e€ect on p53 protein levels.
Using a monoclonal antibody, DO-12, directed
against an epitope in the central core domain of p53
(Vojtesek et al., 1995; Blaydes and Hupp, 1998) no
change in the abundance of p53 protein could be
detected in lysates prepared from HCA2 cells 4 h after
labelling with 32P compared to unirradiated, but
otherwise identically treated controls. In contrast,
using monoclonal antibody DO-1 (Vojtesek et al.,
1992) which binds to an N-terminal epitope (amino
acids 20 ± 25) on p53 containing a known phosphorylation site (serine-20) (Stephen et al., 1995), a clear
increase in the amount of immunodetectable protein
was observed in three independent lysates (Figure 2).
This was estimated by densitometry to represent a 3 ± 4fold increase over the signal seen in mock-treated cells.
To provide a direct comparison with previous
literature, the above experiment was repeated using
[35S]methionine in place of 32P. Again typical metabolic
labelling conditions were used consisting of exposure
for 1 h to 35S-methionine/cysteine mixture (Redivue
Pro-mix; Amersham Code AGQ0080) at a ®nal activity
of 0.7 mCi/ml in methionine- and cysteine-free
DMEM. Essentially identical results were obtained on
Western analysis to those seen following exposure to
32
P (Figure 2).
Phosphospeci®city of DO-1 antibody
Several lines of evidence indicate that the changes in
p53 immunoreactivity observed in Figure 2 re¯ect the
Figure 2 Immunoblot analysis of p53 protein following 32P or
35
S-labelling. (a) Lysates were prepared from HCA2 ®broblasts
4 h after either mock treatment (lanes 1, 3, 5 and 7) or metabolic
labelling with 32P-orthophosphate (lanes 2 and 6) or 35Smethionine/cysteine (lanes 4 and 8). Lysis was carried out for
15 min at 08C in 1% NP40 in a bu€er containing 25 mM HEPES
(pH7.6), 5 mM DTT, 0.4 M KC1, 5 mM EDTA, 10 mM NaF,
2 mg/ml Perfabloc (Boehringer-Mannheim), 20 mg/ml leupeptin,
1 mg/ml aprotinin, 2 mg/ml pepstatin, 10 mg/ml trypsin inhibitor
and 1 mM benzamidine. Lysates (5 mg protein per lane) were
electrophoresed on a 10% SDS-polyacrylamide gel and blotted to
PVDF (Immobilon-P; Millipore, UK). Western blot analysis was
performed using antibodies DO-1 (lanes 1 ± 4) or DO-12 (lanes
5 ± 8), followed by detection using ECL (Amersham, UK).
Migration of p53 was veri®ed in each case by reference to a
recombinant p53 marker (not shown). (b) Blots were stained with
India ink to verify equivalence of protein loading
sensitivity of DO-1 binding to the phosphorylation
state of its epitope.
Firstly, a similar di€erential increase in binding to
DO-1 (in this case compared to PAb1620) was
observed following DNA damage induced by UV
irradiation of the melanoma cell line A375 which
expresses high levels of wild-type (wt) p53 (Figure 3a).
Treatment of lysates from undamaged A375 cells with
liver protein phosphatase 2A (PP2A) (308C for 30 min)
resulted in a threefold increase in binding of DO-1 to
p53 on subsequent immunoblot analysis (not shown)
strongly suggesting that the e€ect of UV is mediated by
de-phosphorylation.
Secondly, direct evidence was obtained by in vitro
binding studies (Figure 3b) which showed that DO-1
binding to a synthetic peptide (amino acids 13 ± 27)
spanning its epitope is speci®cally blocked by
phosphorylation at serine-20, and that this is
reversible on treatment with potato acid phosphatase.
In summary, therefore, our data show that exposure
of normal human ®broblasts to 32P at doses typically
used for metabolic labelling leads to a profound
growth arrest with kinetics very similar to those
observed previously in response to the clastogenic
agent bleomycin and hence, as in that case, likely to be
due to activation of G1/S, S phase and G2/M
checkpoints (Wyllie et al., 1995; 1996). The magnitude
of inhibition of DNA synthesis was greatly reduced in
cells expressing either a dominant-negative p53 mutant
or HPV E6, so that, since any additional e€ects of
these two genes are likely to be non-overlapping, we
can reasonably conclude that the e€ects observed here
re¯ect their common property of abrogating wild-type
p53 function. Detailed analysis of the kinetics of
growth arrest showed that, as observed previously
with bleomycin (Wyllie et al., 1995; 1996) there is
partial growth inhibition even in cells in which p53
function is compromised, re¯ecting the operation of
non-p53-dependent cell cycle checkpoints. This was not
further investigated here due to local restrictions on
¯ow cytometric analysis of radio-labelled cells and
since it lay outside the main aim of this work. The data
are sucient, however, to conclude that 32P-labelling
induces growth arrest which is at least partially p53dependent.
These ®ndings do not necessarily imply that 32P
induces any direct modi®cation of p53, since the latter
may simply be needed as a `passive' component of a
DNA damage-activated signalling pathway. Evidence
for a primary alteration of p53 was however obtained
by immunochemical analysis. Using antibody DO-12,
directed against an epitope (amino acids 256 ± 270) in
the central core domain, which is not known to be
subject to phosphorylation in vivo, no apparent change
in the abundance of the protein could be detected by
Western blot analysis, following either 32P- or 35Slabelling. In contrast a reproducible 3 ± 4-fold increase
in signal was seen with antibody DO-1 which
recognizes an N-terminal epitope overlapping the
transcription activation domain of p53 and its binding
site for mdm2. This result could arise if irradiation
induces a preferential production of a p53 species
which is DO-12 negative or, alternatively, by a
modi®cation of the N-terminus resulting in increased
availability of the DO-1 epitope, without necessarily
any change in abundance of the protein. We consider
32
P-induced activation of p53
JA Bond et al
the latter to be the more likely since, unlike the DO-12
site, the DO-1 epitope contains a known phosphorylation site (serine-20) which has been shown to play a
role in regulating binding to mdm2 (Halazonetis et al.,
Figure 3 (a) UV irradiation induces a selective increase in DO-1
binding to wt p53. Lysates were prepared from the human
melanoma cell line A375 16 h after exposure to 10 J/m2 or 0 J/m2
UVC and p53 content analysed by ELISA assay. This was
performed according to Hupp et al. (1995), using 2.7 mg of lysate
protein per well with either PAb1620 or DO-1 as the capture
antibody and polyclonal antibody CM1 as the detection antibody
in both cases (b) Phosphorylation of serine 20 abolishes DO-1
binding to its epitope in vitro. A series of N-terminally
biotinylated peptides corresponding to amino acids 13 ± 27 of
human p53 (spanning the DO-1 epitope) was synthesized with or
without single phosphorylated residues at serine-15, threonine-18
or serine-20. Binding of antibody DO-1 to immobilized peptide
was assessed by ELISA assay. Analysis was repeated following
treatment of the phosphoserine-20 peptide with 0.6 U potato acid
phosphatase
1998; Craig and Hupp, 1998). Furthermore, we have
presented direct evidence that DO-1 selectively
recognizes the unphosphorylated form of its epitope.
The most likely interpretation of our data, therefore, is
that radio-labelling induces de-phosphorylation of sites
in or around the DO-1 epitope resulting in a relative
increase in its detection. Irrespective of the exact
mechanism, however, the di€erential change in affinity
to these two monoclonals provides clear evidence that
32
P induces a post-translational modi®cation of p53
similar to that seen with 35S.
Our ®ndings agree broadly with previous reports of
growth arrest (or apoptosis), and increased p53 protein
levels following labelling with 3H and 35S-isotopes
(Dover et al., 1994; Yeargin and Haas, 1995). Indeed,
in the former case, the apparent increase in p53 level
was observed using an antibody, DO-7, which
recognizes the same epitope as DO-1 (Stephen et al.,
1995). The dependence on wt p53 was rather less
clearly demonstrated in these cases however, since
unlike the HCA2-neo, -mp53 and -E6 populations used
here, the comparisons were made between nonisogeneic cell types (e.g. di€erent murine T-ALL
lines) which had been cultured for prolonged periods,
hence raising the possibility that other unknown
genetic events were involved in loss of response in
the mutant p53 lines.
The formal demonstration that such p53-dependent
`metabolic labelling artefacts' apply equally to 32P is
important since its very di€erent radio-biological
properties compared to 3H and 35S made this by no
means a foregone conclusion. Despite the higher mean
energy of 32P beta-particles compared to those of 35S,
their much longer pathlength (maximum 8 mm in
aqueous media compared to 0.3 mm (Brady and
Finlan, 1990)) could have resulted in the rate of
energy absorbed from labelled cell proteins by a
relatively small, close-range target such as nuclear
DNA being below biologically signi®cant levels. Indeed
this is of course one reason why 32P is an inecient
isotope for direct-contact, emulsion-based autoradiography (Brady and Finlan, 1990).
The argument that any potential artefactual e€ects
of 32P can be ignored since they will apply equally to
control cells assumes a simple additive relationship
which is not tenable, since in reality they may obscure
the e€ect being studied, or may have a complex
interaction with it. This poses a major diculty in
studying phosphorylation in intact cells, not only of
p53 itself, but also potentially of any protein in a signal
pathway downstream of p53 whose phosphorylation
state might be subject to modulation by p53 activation.
Alternative methods of analysing phosphorylation
exist, although none are as convenient as 32P-labelling.
In the case of p53, the relevant phosphorylations occur
on serine/threonine residues rather than tyrosine, but
although more problematic, generation of phosphoserine- and threonine-speci®c antibodies against relevant
domains of the p53 protein is possible (an early
fortuitous example being provided by PAb421 (Hupp
et al., 1995) (although with `reverse' speci®city- for
native, rather than phosphorylated, peptide). Furthermore, examples of such antibodies have recently been
described with speci®city for phosphorylated amino
acids in the C terminus of p53 (Blaydes and Hupp,
1998; Waterman et al., 1998).
3791
32
3792
P-induced activation of p53
JA Bond et al
One drawback in this approach, however, is that it
relies on sites which have already been de®ned by 32Plabelling, which will inevitably preclude any which are
de-phosphorylated in response to the 32P-induced
`damage response' described here. The ideal approach
therefore is to raise antibodies, not only to sites
recognized from conventional 32P analysis, but also to
potentially cryptic sites generated in vitro by pure
protein kinases.
Our ®ndings should provide further impetus for the
development of such reagents and for their use in reevaluating the role of phosphorylation in modulation
of p53 function and stability.
Acknowledgements
We are grateful to the Cancer Research Campaign for
grant support, and to Theresa King for manuscript
preparation.
References
Blaydes J and Hupp T. (1998). Oncogene, 17, 1045 ± 1052.
Bond JA, Blaydes JP, Rowson J, Haughton MF, Smith JR,
Wynford-Thomas D and Wyllie FS. (1995). Cancer Res.,
55, 2404 ± 2409.
Bond JA, Haughton M, Blaydes J, Gire V, Wynford-Thomas
D and Wyllie F. (1996). Oncogene, 13, 2097 ± 2104.
Brady MAW and Finlan MF. (1990). In Situ Hybridization:
Principles and Practice, Oxford University Press, UK
Cox LS and Lane DP. (1995). Bio. Essays, 17, 501 ± 508.
Craig AL and Hupp TR. (1998). 9th p53 Workshop, Crete,
p. 102.
Dover R, Jayaram Y, Pate K and Chinery R. (1994). J. Cell.
Sci., 107, 1181 ± 1184.
Gu W and Roeder RG. (1997). Cell, 90, 595 ± 606.
Halazonetis T, Waterman M, Chehab N, Caruso L and
Stavridi E. (1998). 9th p53 Workshop, Crete, p. 52.
Hupp TR and Lane DP. (1994). Cold Spring Harbor Symp.
Quant. Biol., 59, 195 ± 206.
Hupp TR, Sparks A and Lane DP. (1995). Cell, 83, 237 ± 245.
Kastan MB, Onyekwere O, Sidransky D, Vogelstein B and
Craig RW. (1991). Cancer Res., 51, 6304 ± 6311.
Kern SE, Pietenpol JA, Thiagalingam S, Seymour A, Kinzler
KW and Vogelstein B. (1992). Science, 256, 827 ± 830.
Ko LJ and Prives C. (1996). Genes Dev., 10, 1054 ± 1072.
Lu X and Lane DP. (1993). Cell, 75, 765 ± 778.
Lutzker SG and Levine AH. (1996). Nature Med., 2, 802 ±
810.
Meek D. (1994). Semin. Cancer Biol., 5, 203 ± 210.
Milne DM, Campbell LE, Campbell DG and Meek DW.
(1995). J. Biol. Chem., 10, 5511 ± 5518.
Mundt M, Hupp T, Fritsche M, Merkle C, Hansen S, Lane D
and Groner B. (1997). Oncogene, 15, 237 ± 244.
Shaw P, Freeman J, Bovey R and Iggo R. (1996). Oncogene,
12, 921 ± 930.
Stephen CW, Helminen P and Lane DP. (1995). J. Mol. Biol.,
248, 58 ± 78.
Vojtesek B, Bartek J, Midgley CA and Lane DP. (1992). J.
Immunol Methods, 151, 237 ± 244.
Vojtesek B, Dolezalova H, Lauerov L, Svitakova M, Havli P,
Kovarik J, Midgley CA and Lane DP. (1995). Oncogene,
10, 389 ± 393.
Waterman MJ, Stavridi ES, Waterman JL, Halazonetis TD.
(1998). Nature Genet., 19, 175 ± 178.
Woo RA, McLure KG, Lees-Miller SP, Rancourt DE and
Lee PWK. (1998). Nature, 394, 700 ± 704.
Wyllie FS, Lemoine NR, Barton CM, Dawson T, Bond J and
Wynford-Thomas D. (1993). Mol. Carcinogen., 7, 83 ± 88.
Wyllie FS, Haughton MF, Blaydes JP, Schlumberger M and
Wynford-Thomas D. (1995). Oncogene, 10, 49 ± 59.
Wyllie FS, Haughton MF, Bond JA, Rowson JM, Jones CJ
and Wynford-Thomas D. (1996). Oncogene, 12, 1077 ±
1082.
Yeargin J and Haas M. (1995). Curr. Biol., 5, 423 ± 431.