¹Department of Haematology, Royal Liverpool University Hospital, Level 2 Ducan Building, Prescot Street, Liverpool, Merseyside L7 8XP, UK

Correspondence: AR Pettitt, E-mail: andrew.pettitt@rlbuht.nhs.uk

Received 7 March 2005; Revised 31 May 2005; Accepted 3 June 2005; Published online 15 August 2005.

Abstract

p53 is the most frequently inactivated gene in human cancers, reflecting its pivotal role in maintaining genomic integrity. The present study was conducted to explore the possibility that tumour cells with no intrinsic defects of the p53 pathway might nevertheless acquire p53 dysfunction through extrinsic suppression of the pathway by microenvironmental factors. Neoplastic cells from patients with chronic lymphocytic leukaemia (CLL) were cultured in the presence or absence of basic fibroblast growth factor (bFGF) and exposed to ionizing radiation (IR) to induce p53 accumulation. bFGF is greatly increased in the plasma of CLL patients and can suppress p53 activation in some experimental models. IR induced a marked increase in p53 levels in 28 samples from 24 patients. bFGF inhibited IR-induced p53 accumulation to some extent in most of these samples and by more than 50% in seven samples from seven patients. Suppression of p53 activation by bFGF was frequently but not always accompanied by upregulation of the p53-inhibitory protein MDM2 and/or phosphorylation of MDM2 at serine 166, and was associated with impaired transcriptional activation of the p53 target gene p21. These observations provide the first demonstration in human cancer cells that the p53 pathway can be suppressed by factors in the tumour-cell microenvironment.

p53 is a tumour suppressor protein that accumulates in response to DNA damage and coordinates the cellular response to such damage by inducing cell-cycle arrest and DNA repair or apoptosis (Brooks and Gu, 2003; Oren, 2003; Xu, 2003). In doing so, it fulfils a pivotal role in protecting cells from oncogenic mutation in the face of genotoxic stress (Lane, 1992). Such stress may be physiological or deliberately induced. Thus, many anti-cancer agents exert their anti-neoplastic effects at least in part through the induction of p53-mediated apoptosis (Lowe et al., 1993; Sax and El-Deiry, 2003). It therefore follows that tumour cells with p53 dysfunction are intrinsically resistant to killing by such agents and also susceptible to clonal evolution and acquisition of drug resistance through secondary mechanisms.

The clinical importance of p53 is reflected in the fact that it is the most frequently mutated gene in human cancers (Bourdon et al., 2003). In haemic malignancies, p53 is mutated in only a minority of cases (Wattel et al., 1994) but may be functionally impaired through alternative mechanisms. For example, a proportion of cases of chronic lymphocytic leukaemia (CLL) have mutations of ATM, a kinase required for p53 activation in response to double-strand DNA breaks (Pettitt et al., 2001; Stankovic et al., 2002). Such cases display impaired radiation-induced p53 responses and have a poor clinical outcome resembling that of p53-mutant cases (Lin et al., 2002).

As an alternative to intrinsic defects of p53 or proteins that regulate p53, it is intriguing to speculate that functional impairment of the p53 pathway might arise through extrinsic suppression by factors in the tumour-cell microenvironment. This idea is plausible since a number of signalling proteins that are activated following ligation of cell-surface receptors can directly regulate p53 or its inhibitory partner, MDM2 (Fuchs et al., 1998; Buschmann et al., 2001; Mayo and Donner, 2001; She et al., 2001 2002; Ashcroft et al., 2002; Ogawara et al., 2002; Zhu et al., 2002). There is some evidence to suggest that basic fibroblast growth factor (bFGF) might fulfil such a role. For example, bFGF has been reported to lower baseline p53 levels in human endothelial cells (Ashton et al., 2004) and increase MDM2 levels and inhibit cisplatin-induced apoptosis and transcriptional activation of p21 and bax in mouse fibroblasts (Shaulian et al., 1997). On the other hand, bFGF upregulated MDM2 but had no effect on p53 or p21 levels in human ovarian granulosa cells transfected with a temperature-sensitive mutant p53 (Hosokawa et al., 1998). This inconsistency between experimental models illustrates that the regulation of p53 varies markedly between different cell types (Oren, 2003), and highlights the need to establish whether or not bFGF can suppress p53 activation in unmanipulated human cancer cells.

It has been known for some time that bFGF is increased in the urine of patients with a wide range of neoplastic diseases (Nguyen et al., 1994). Among the leukaemias and myelodysplastic syndromes, the highest plasma levels of bFGF are found in CLL (Aguayo et al., 2000). It is likely that the malignant cells are the main source of the growth factor in vivo. Thus, CLL cells constitutively produce and secrete bFGF ex vivo (Menzel et al., 1996; Krejci et al., 2001; Kay et al., 2002), a three-way correlation exists between high intracellular levels of bFGF, high serum levels of bFGF and advanced clinical stage (Menzel et al., 1996; Bairey et al., 2001; Kay et al., 2002; Molica et al., 2002; Gora-Tybor et al., 2003), and serum concentrations of bFGF decrease following successful therapy (Gora-Tybor et al., 2002). Despite producing endogenous bFGF, CLL cells are nevertheless responsive to exogenously added bFGF when cultured ex vivo. For example, bFGF has been reported to inhibit the killing of CLL cells by fludarabine (Menzel et al., 1996), a process that is partly p53 dependent (Pettitt, 2003).

In the light of these considerations, we sought to establish whether bFGF suppressed the p53 pathway in cultured CLL cells. In order to activate p53, the cells were exposed to ionizing radiation (IR). In contrast to genotoxic drugs, IR induces DNA breaks directly and is known to activate p53 in an ATM-dependent fashion in CLL cells (Pettitt et al., 2001; Stankovic et al., 2002). p53 levels were measured by flow cytometry, using the median fluorescence intensity (MFI) values for quantification. A fixed time-point of 6 h was employed in order to standardize the analysis. Only those CLL samples displaying a more than 50% increase in p53 levels were analysed further owing to the difficulty of observing an inhibitory effect in the absence of significant protein accumulation. Among the 37 CLL samples examined, 28 samples from 24 patients fulfilled this criterion.

Incubation of CLL cells with bFGF resulted in a varying amount of inhibition of IR-induced p53 accumulation (Figure 1). In all, 12 samples from 10 patients showed >25% inhibition, while in seven samples from seven patients, bFGF inhibited IR-induced p53 accumulation by >50%. bFGF increased the amount of p53 accumulation induced by IR in only two cases (samples 21 and 22). In contrast to its effect in irradiated cells, bFGF had no consistent effect on p53 levels in nonirradiated cells (data not shown). Three samples from two patients (samples 2a, b and 4b) were tested on more than one occasion and consistent results obtained (data not shown). Furthermore, in three patients, more than one sample was obtained during the course of the disease (2a and b; 4a and b; 16a, b and c). In each case, the separate samples from each patient gave similar results (Figure 1), although it should be noted that two successive samples from one patient taken at an interval of 15 months (samples 2a and b) showed an increase in p53 suppression that corresponded with disease progression. Western blotting experiments using a different p53 antibody were also performed, using nine of the ten samples most sensitive to the p53-suppressive effect of bFGF as determined by FACS analysis (sample 8 was not available). In each sample, bFGF produced clear inhibition of IR-induced p53 accumulation, indicating that the observation did not depend on the method or antibody used (Figure 2).

Figure 1.

Effect of bFGF on IR-induced p53 accumulation as detected by flow cytometry. CLL cells (2 10⁶/ml in RPMI 1640 + 1% BSA) were incubated at 37°C for 5–10 min in the presence or absence of 100 ng/ml bFGF (Gibco). Cell suspensions were then exposed to -irradiation (5 Gy over 1.5 min) from a ¹³⁷Cs source, and cultured at 37°C for 6 h. For FACS analysis, CLL cells were fixed for 30 min at 20°C in 2% paraformaldehyde and permeabilized overnight at -20°C in 80% ethanol. The cells were then incubated with the anti-p53 antibody DO-1 (Oncogene Research, Cambridge, MA, USA) followed by a goat anti-mouse antibody conjugated to fluorescein isothiocyanate (Becton Dickinson, San Jose, CA, USA). Analysis was performed using a Becton Dickinson FACScan and CellQuest software. Note that a value of >100% inhibition indicates that bFGF reduced p53 levels to below those of nonirradiated cells, and that a negative value indicates that bFGF increased, rather than reduced, IR-induced p53 accumulation. Overall, bFGF inhibited p53 accumulation by a median of 14.7% (95% confidence interval12.7%)

Full figure and legend (15K)

Figure 2.

p53-suppressive effect of bFGF as detected by Western blotting. Irradiated and nonirradiated CLL cells were incubated at 37°C for 6 h in the presence or absence of bFGF and lysed in sample buffer (62.5 mM Tris-EDTA pH 6.8, 2% SDS, 5% 2-mercaptoethanol, 10% glycerol and 0.003% bromophenol blue). Proteins were separated by SDS–polyacrylamide gel electrophoresis and analysed by Western blotting using the anti-p53 mouse monoclonal antibody pAb 1801 (Oncogene Research, Cambridge, MA, USA) and a second layer anti-mouse antibody conjugated to horse-radish-peroxidase (Affiniti, Mamhead, UK). Labelled protein bands were visualized using the ECL system (Amersham, Little Chalfont, UK). Membranes were subsequently stripped and re-probed with an antibody to -actin (Sigma) to compare protein loading between lanes

Full figure and legend (95K)

We next considered the possible mechanisms underlying the p53-inhibitory effect of bFGF. We focused on the E3 ubiquitin ligase MDM2 – a key negative regulator of p53 (Oren, 2003) – and selected for study the same nine samples that were examined by Western blotting for p53 accumulation. Membranes were probed for MDM2 and phosphoserine 166 MDM2. Serine 166 is one of two residues in full-length MDM2 that are phosphorylated by AKT. Phosphorylation of both of these residues is required for the translocation of MDM2 from the cytoplasm into the nucleus and consequent inhibition of p53 (Mayo and Donner, 2001). Baseline levels of MDM2 and phosphoserine 166 MDM2 in untreated cells varied between cases but were mostly low or undetectable. IR produced an increase in levels of MDM2 or phosphoserine 166 MDM2 in two of the nine samples (1 and 4b; Figure 3). This was not unexpected given that MDM2 is a transcriptional target of p53 (Oren, 2003). In nonirradiated cells, bFGF increased levels of MDM2 or phosphoserine 166 MDM2 in three samples (1, 3 and 4a; Figure 3). In irradiated cells, bFGF increased levels of MDM2 or phosphoserine 166 MDM2 in six samples (2b, 3, 4a, 5, 7 and 9; Figure 3). The increase in levels of MDM2 or phosphoserine 166 MDM2 induced by bFGF was consistently greater in irradiated cells than in nonirradiated cells. bFGF failed to induce any increase in MDM2 or phosphoserine 166 MDM2 in two samples (4b and 6; Figure 3). The increase in MDM2 protein induced by bFGF was not reflected by similar changes in MDM2 mRNA (Figure 4a). This suggests that bFGF was regulating MDM2 at the post-transcriptional level. In summary, activation of MDM2 may have contributed to the p53 suppressive effect of bFGF in some of the patient samples. The fact that the p53-suppressive effect of bFGF was associated with upregulation of MDM2 in sample 4a but not in sample 4b suggests that mechanistic differences exist not only between cases but also within individual cases tested on separate occasions. These observations are very much in keeping with the idea that the regulation of p53 activation is complex, and that the p53-suppressive effect of bFGF is mediated by more than one mechanism (Oren, 2003).

Figure 3.

Effect of bFGF on MDM2 and phosphoserine 166 MDM2. Irradiated and nonirradiated CLL cells were incubated at 37°C for 6 h in the presence or absence of bFGF and analysed by Western blotting using the mouse monocolonal antibody SPH14 (Santa Cruz, Santa Cruz, CA, USA). Blots were also probed with a rabbit polyclonal antibody to phosphoserine 166 MDM2 (New England Biolabs UK, Hitchin, UK) and a second layer anti-rabbit antibody conjugated to horse-radish peroxidase (Affiniti, Mamhead, UK). Note that in sample 4a, the actin control is the same as that shown in Figure 2. This is because p53 and MDM2 were analysed in this sample in one experiment using a single nylon membrane

Full figure and legend (127K)

Figure 4.

Effect of bFGF on MDM2 and of p21 mRNA levels. Irradiated and nonirradiated CLL cells from cases 1 and 2 were incubated at 37°C in the presence or absence of bFGF and analysed for MDM2 or p21 transcript levels. Total RNA was extracted from 2 10⁶ frozen cells using TriZol (Invitrogen, Paisley, Scotland, UK), precipitated, re-suspended in 10 l H₂O and incubated with 20 pmol oligod(T)₁₂ at 65°C for 5 min. The samples were then reverse transcribed using Improm-II reverse transcriptase (Promega, Southampton, UK) and the supplied buffer. The resulting first-strand cDNAs were diluted to 100 l with H₂O and 5 l aliquots used in real-time PCR reactions, all of which were carried out in duplicate using an Opticon 2 thermal cycler (MJ Research, Waltham, MA, USA) alongside no-template controls for each set of primers. Each 25 l reaction included 12.5 l of DyNamo (Finnzymes, Espoo, Finland) master mix (includes Tbr polymerase, SYBR Green I, 5 mM Mg²⁺ and dNTPs) and 20 pmol of each relevant oligonucleotide primer. Reactions were incubated at 95°C for 10 min then cycled 45 times at 94°C for 30 s, 61°C for 30 s and 72°C for 30 s. Following each extension step, the fluorescence was measured at 76, 78 and 80°C (after a 1 s incubation at each temperature). The reactions were then incubated at 72°C for 10 min and, to determine melting temperatures of the PCR products, the fluorescence was measured at 1 s intervals while the temperature was increased from 65 to 98°C in 0.3°C steps. The amount of MDM2 or p21 transcript relative to -actin mRNA was calculated from the numbers of cycles taken to reach the same fluorescence intensity at an early time-point during the linear phase of amplification (Ct values). The data were pasted into Excel (Microsoft) spreadsheets and the formulae '2 ^{- (Ct MDM2 – Ct actin)}' or '2 ^{- (Ct p21 – Ct actin)}' applied. The PCR primer sequences used were: 5'-MDM2; GAC TAT TCT CAG CCA TCA ACT TCT AG and 5'-GAA TTG GTT GTC TAC ATA CTG GGC AG; p21, 5'-AGA CCA TGT GGA CCT GTC ACT GTC and 5'-TTC CAG GAC TGC AGG CTT CCT GTG; actin, 5'-CTG GAC TTC GAG CAA GAG AT and 5'-TCG TCA TAC TCC TGC TTG CT. (a) shows the amount of MDM2 mRNA relative to -actin after 4 h culture. (b) shows the amount of p21 mRNA relative to -actin after 24 h culture

Full figure and legend (26K)

Since p53 is a transcription factor, it was of interest to establish whether the suppression of p53 accumulation by bFGF had any downstream effect on gene expression. We focussed on p21, a well-established transcriptional target of p53 in CLL (Pettitt et al., 2001; Stankovic et al., 2002). As expected, p21 mRNA upregulated strongly in response to IR (Figure 4b). bFGF produced marked inhibition of this upregulation, indicating that its effect in suppressing p53 accumulation did indeed have downstream transcriptional consequences.

The functional responsiveness of CLL cells to bFGF demonstrated in the present study is entirely in keeping with previous reports showing that CLL cells express mRNA encoding several high-affinity tyrosine kinase receptors for bFGF (Krejci et al., 2001) and that bFGF inhibits the killing of CLL cells by fludarabine (Menzel et al., 1996). Our findings do, however, need to be reconciled with reports that bFGF binds only weakly to CLL cells and fails to activate MEK and ERK (Krejci et al., 2001), and that CLL cells do not express syndecan-1 (Witzig et al., 1998), an important heparan sulphate proteoglycan (HSPG) that binds bFGF with low affinity and enhances signalling through high-affinity receptors (Yayon et al., 1991). However, other reports suggest that syndecan-1 expression is a distinctive feature of CLL cells (Sebestyen et al., 1997, 1999; Sutcliffe et al., 2000). Furthermore, HSPG other than syndecan-1 can function as low-affinity receptors for bFGF (Deguchi et al., 2002; Dode et al., 2003; Qiao et al., 2003), and bFGF can activate signalling pathways other than MEK/ERK (Pardo et al., 2001).

bFGF is unlikely to be the only microenvironmental factor capable of modulating the p53 pathway in CLL. Similarly, inhibition of p53 activation is unlikely to be the only way in which bFGF can contribute to the pathogenesis of CLL. Nevertheless, the biological consequences of suppressing the ATM-p53 pathway are sufficiently profound that this is likely to be one of the most important actions of bFGF in CLL, and that sensitivity to this effect may be an important determinant of clinical outcome among patients with no intrinsic defects of p53 and ATM.

In conclusion, the present study significantly takes forward our understanding of how p53 can be inactivated in human cancer and strengthens the idea that suppression of p53 may be a key function of bFGF that is shared between cell types but mediated via a number of different mechanisms including, but not restricted to, MDM2 activation.