In vitro drug resistance and prognostic impact of p16INK4A/P15INK4B deletions in childhood T-cell acute lymphoblastic leukaemia

Authors


N. L. Ramakers-van Woerden, University Hospital Vrije Universiteit, Department of Paediatric Hematology/Oncology, PO Box 7057, 1007 MB Amsterdam, the Netherlands. E-mail: ramakersvanwoerden@azvu.nl

Abstract

p16 gene deletions are present in about 70% of primary paediatric T-cell acute lymphoblastic leukaemia (T-ALL) and 20% of common/precursor B-cell ALL cases. It is not clear what the impact of the frequent p16 deletions is within the subgroup of T-lineage ALL. We studied the relationship between p16/p19ARF deletions, using fluorescence in situ hybridization, and in vitro drug resistance and prognosis in childhood T-ALL at diagnosis. The cellular drug resistance was measured with the methyl thiazol tetrazoliumbromide assay using a panel of drugs and the thymidylate synthase inhibition assay for methotrexate. There was a complete overlap of individual LC50 values of p16 gene homozygously deleted and p16 germ-line cases for most of the nine classes of drugs tested. The only difference was for dexamethasone: the p16-deleted group was more sensitive than the germ-line p16 group (P = 0·030). The homozygously deleted p16 T-ALL patients (n = 34) treated with the modern multiagent chemotherapy schemes of the Dutch Childhood Leukaemia Study Group ALL-VII/-VIII or Co-operative ALL-92/-97 protocols have a significantly lower 5-year disease-free survival (DFS) than germ-line p16 T-ALL (n = 25) (65·1 ± 9·1% vs. 95·5 ± 4·4%, Plog rank= 0·021). Hence, this study identifies a subpopulation of primary childhood T-ALL that appears to have an extremely high DFS. However, the observed differences in outcome do not seem to be related to intrinsic resistance for the tested drugs.

Notwithstanding great improvements in the treatment of children with acute lymphoblastic leukaemia (ALL), about 20–25% still die of disease- or treatment-related complications. Hence, clinical and biological features of prognostic importance are sought in an attempt to further tailor the therapy according to risk (Pui, 1997). The presence of specific chromosomal abnormalities in the leukaemic cells has proved to be a valuable independent prognostic factor and there is evidence that these aberrations may reflect differences in drug sensitivity (Whitehead et al, 1992; Kaspers et al, 1995; Kersey, 1997; Schrappe et al, 1998). The in vitro cellular drug resistance profile itself has proven to be a strong and independent prognostic factor (Pieters et al, 1991; Hongo et al, 1997; Kaspers et al, 1997).

T-lineage leukaemia accounts for about 10–15% of childhood ALL and is associated with a number of unfavourable characteristics at presentation, such as hyperleucocytosis, older age and mediastinal mass. Children with T-ALL have often been reported to have a worse outcome than B-lineage ALL (Dowell et al, 1987; Crist et al, 1988; Shuster et al, 1990; Uckun et al, 1998), although contemporary multiagent chemotherapy schemes have made considerable progress in the outcome of T-ALL (Gaynon et al, 1988; Garand et al, 1990; Schorin et al, 1994; Uckun et al, 1997, 1998). T-ALL is a heterogeneous disease and it remains important to search for factors that discern ‘poor’ from ‘good’ cases. Non-random chromosomal translocations, frequently involving the T-cell receptor loci, are present in up to 50% of cases. However, the most common genetic abnormality in T-ALL is probably the deletion of the tumour suppressor gene p16 (also known as p16INK4a/MTS1α/INK4a/CDKN2/CDKI) at 9p21. Deletions of the 9p21 region are commonly observed in a wide range of cancers and often span a relatively large region, accompanied by co-deletion of the neighbouring genes p15 (p15/INK4b/MTS2), the methylthioadenosine phosphorylase (MTAP) gene, and interferon α, β and ω genes (Diaz et al, 1990; Einhorn et al, 1990; Olopade et al, 1992; Einhorn & Heyman, 1993; Batova et al, 1996). However, in childhood ALL the p16 site appears to be the target and is deleted in about 70% of T-ALL and 20% of B-lineage ALL cases (Hebert et al, 1994; Fizzotti et al, 1995; Okuda et al, 1995; Quesnel et al, 1995; Takeuchi et al, 1995; Rubnitz et al, 1997). The p16 gene encodes two unrelated proteins transcribed through different promoters: p16INK4a and p19ARF (mouse p19ARF11/human p14ARF/MTS1β) (Duro et al, 1995; Mao et al, 1995; Quelle et al, 1995; Stone et al, 1995). Recent studies suggest that the p19ARF gene might be an important target in p16/MTS locus rearrangements in T-ALL (Gardie et al, 1998). p16INK4a protein is an inhibitor of cyclin-dependent-kinase 4 (CDK4), thus blocking the cell cycle at the G1 phase through the activation of Rb protein (Serrano et al, 1993). p19ARF protein interacts with Mdm-2, resulting in a reversal of the inhibitory binding of Mdm-2 to the p53 protein (Kamijo et al, 1998; Honda & Yasuda, 1999; Tao & Levine, 1999); thus, deletion of p19ARF facilitates the degradation of p53. Therefore, this locus is involved in the upstream regulation of both Rb and p53 proteins, and INK4a/ARF rearrangements have been described in association with chemoresistant, aggressive tumours in various model systems (Stone et al, 1996; Matsumura et al, 1997; Schmitt et al, 1999).

Homozygous deletion of p16 in childhood ALL has been reported to be associated with high-risk features and an unfavourable outcome (Fizzotti et al, 1995; Okuda et al, 1995; Takeuchi et al, 1995; Heyman et al, 1996; Kees et al, 1997; Rubnitz et al, 1997; Zhou et al, 1997). However, most of these studies on the prognostic impact of p16 deletions have been performed on childhood ALL in general and are therefore biased by an overrepresentation of (low-risk) B-lineage ALL in the p16 germ-line group and T-ALL with associated high-risk presenting features in the p16-deleted group. p16/p15 inactivation was shown to be an independent adverse prognostic factor in a multivariate analysis of event-free survival (EFS) that included immunophenotype as a variable (Heyman et al, 1996). p16 deletion status was also of prognostic significance in separate analyses of B-precursor and T-cell phenotypes in an extended Nordic study (unpublished oberservations). In a comprehensive unbiased study of childhood T-ALL cases, Diccianni et al (1997) showed that p16/p15 deletions were associated with shortened survival after relapse but had no prognostic impact at initial diagnosis. Hence, the impact of the frequent p16 deletions is not clear within the subgroup of primary T-lineage ALL.

In the present study, we analysed a group of T-ALL patients at diagnosis for the presence of p16/p19ARF deletions and subsequently related these findings to prognosis and in vitro drug sensitivity. We aimed to identify a genotypical subgroup of ALL with prognostic significance and analyse the cellular resistance profile thereof, in order to rationally design new therapeutic strategies. We have confirmed that p16 homozygously deleted childhood T-ALL cases have a significantly poorer disease-free survival (DFS). The p16/p19ARF homozygously deleted cases might be expected to show a significant drug resistance, owing to inactivation of the p16/Rb and p19ARF/p53 pathways. However, the observed differences in outcome do not seem to be explained by intrinsic resistance for the drugs tested.

Patients and methods

Patients and leukaemic cell samples Freshly obtained cells from the bone marrow or peripheral blood of 122 children with newly diagnosed, untreated T-cell ALL from the Dutch Childhood Leukaemia Study Group (DCLSG) and the German Co-operative ALL (COALL) study group were used for this study.

Immunophenotyping was performed at the reference laboratories of the participating groups; patient characteristics [gender, age, white blood cell count (WBC) at diagnosis] and follow-up were collected by study centres. The T-lineage immunophenotype was defined as terminal deoxynucleotidyl transferase (TdT)+/cytoplasmic CD3+/CD7+. Patients with T-cell non-Hodgkin's lymphoma, defined as those patients with a bone marrow blast cell percentage lower than 25, were excluded. Leukaemic blast cells were isolated within 48 h of sampling using density gradient centrifugation (Lymphoprep, 1·077 g/ml, Nycomed Pharma, Oslo, Norway; at 480 g for 15 min). After washing, the cells were resuspended in Roswell Park Memorial Institute (RPMI)-1640 medium (Dutch modification; Gibco BRL, Breda, The Netherlands) containing 20% fetal calf serum (Gibco BRL) and other supplements (Pieters et al, 1990). When necessary, contaminating normal cells were removed by monoclonal antibodies linked to magnetic beads, as described previously (Kaspers et al, 1994). All samples used for the methyl thiazol tetrazoliumbromide (MTT) assay contained more than 80% leukaemic cells, as determined using cytospin preparations stained with May–Grünwald–Giemsa (Merck, Darmstadt, Germany). Cytospin preparations were made with the same cell suspension (50 μl of 0·5 × 106/ml per cytospin), stored after desiccation at −20°C.

Treatment Dutch children with ALL at initial diagnosis were treated according to DCSLG protocols ALL-VII or ALL-VIII based upon Berlin–Frankfurt–Münster (BFM) regimens, as described previously (Kaspers et al, 1997; Van Der Does-Van Den Berg et al, 1998; Kamps et al, 1999). The T-ALL patients in this study (n = 31) were treated either with medium-risk group (MRG) therapy [ALL-VII risk group (RG) or equivalent ALL-VIII-MRG] or with high-risk group (HRG) therapy [ALL-VII experimental group (EG) or equivalent ALL-VIII-HR].

The German children with T-ALL at initial diagnosis (n = 44) were treated according to the COALL-92 (Janka-Schaub et al, 1996) or, when diagnosed after 1 September 1997, according to the COALL-97 protocol (Janka-Schaub et al, 1999). For the COALL-92 and -97 studies, all T-ALL patients in the present study were included in the high-risk treatment group.

Patients that were elegible for the latter protocols but received a different treatment, or who were not in the above-mentioned trials, were excluded from the analysis of outcome.

Fluorescence in situ hybridization (FISH) Dual-coloured FISH experiments for the presence of p16/p15 genes were performed with the P1 1063 probe covering the region from exon 2 of p15 to distal to exon 2 of p16, spanning the coding sequence of p19ARF (Kamb et al, 1994). The P1 164 probe containing the complete coding sequence of the ETO gene was used as a control (Sacchi et al, 1995). Nick translation was used to label the probes with biotin-16-dUTP/digoxygenin-11-dUTP (Boeringher Mannheim, Amsterdam, The Netherlands) (Rigby et al, 1977).

FISH analysis was performed on cytospins as described before (Arnoldus et al, 1990). Briefly, slides were pretreated with RNAse and pepsin solutions, followed by post-fixation with acid-free formaldehyde and denaturation. The probes were denatured and preannealed in the presence of an excess of Human-COT-1 DNA. Hybridization of the probes to the slides was allowed to proceed overnight at 37°C. The biotin hybridization signal (P1 1063) was visualized using fluorescein avidin (Vector Laboratories, Berlingame, USA) and biotinylated anti-avidin D sandwich detection (affinity purified, Vector Laboratories). The digoxigenin hybridization signal (P1 164) was detected using anti-digoxigenin-rhodamine (Boehringer Mannheim) and donkey anti-sheep Texas-red (Jackson ImmunoResearch Laboratories, Westgrove, USA). The cells were counterstained with 4,6-diamidino-2-phenyl-indole (DAPI). A minimum of 200 nuclei were blindly scored by two independent observers. The cut-off values for the detection levels of deletion were determined by the FISH results of the above-mentioned probes on non-leukaemic samples (bone marrow samples of children without leukaemia and peripheral blood samples of healthy adult volunteers): the mean of the scores for homozygous or hemizygous deletions plus 3x standard deviation, 5% and 18%, respectively, for P1 1063.

Southern blot analysis DNA was extracted and Southern blot analysis (digestion, transfer and hybridization) was performed as described previously (Einhorn et al, 1990). The signal intensity was assessed by visual inspection and measured by scanning densitometry using an Ultroscan XL (LKB, Bromma, Sweden). Three different probes were used for detection of the Ink4 locus: (i) cDNA containing both p15ink4b exons; (ii) cDNA comprising the p16ink4a exon coding sequence; (iii) the E1β exon from the p16ARF gene (Heyman et al, 1996). The E1β exon probe (736 bp) was obtained from polymerase chain reactions (PCRs) using the following primers: 5′-TCCCAGTCTGCAGTTAAGG-3′ and 5′-GTCTAAGTCGTTGTAACCCG-3′ (Mao et al, 1995). Two probes were used as controls: (i) the cDNA Bcl-1 probe located on chromosome 11 band q13 (Meeker et al, 1989) for the smaller fragments, and (ii) a 3-kb Eco R1 fragment, p105–789Rb from chomosome 5-D5S78 (Leppert et al, 1987), for the larger fragments.

In vitro drug resistance assayIn vitro drug cytotoxicity was determined in the MTT assay as described previously (Pieters et al, 1990; Kaspers et al, 1991). Briefly, 100 μl of cell suspension was cultured for 4 d in the absence (i.e. control) or presence of six duplicate concentrations of each drug in 96-well round-bottomed microculture plates. The range of drug concentrations used was based on earlier studies and aimed at obtaining an LC50 value both for highly sensitive and resistant cases. The following drugs and ranges of concentrations were used: prednisolone disodium phosphate (0·06–250 μg/ml), dexamethasone sodium phosphate (0·0002–6·0 μg/ml), vincristine (0·05–50 μg/ml), l-asparaginase (0·003–10 IU/ml), daunorubicin (0·002–2·0 μg/ml), doxorubicin (0·008–8·0 μg/ml), mitoxantrone (0·001–1·0 μg/ml), 6-mercaptopurine (15·6–500 μg/ml), 6-thioguanine (1·6–50 μg/ml), cytarabine (0·002–2·5 μg/ml), teniposide (0·003–8·0 μg/ml), etoposide (0·05–50 μg/ml) and 4-hydroperoxy-ifosphamide (0·10–100 μg/ml; kindly provided by Asta Medica, Frankfurt am Main, Germany).

It has previously been shown that the source of the leukaemic cells (i.e. bone marrow or peripheral blood) and cryopreservation (5 out of 95 samples tested in the present study) does not affect the drug resistance profile (Kaspers et al, 1991).

In situ TS inhibition assay (TSIA) Methotrexate (MTX) is not cytotoxic to ALL cells in the MTT assay, presumably because of rescue by the hypoxanthine and thymidine released from dying cells. Hence, the TSIA assay was used as a measure of in vitro resistance for MTX (Rots et al, 1999a). Inhibition of TS was determined in whole cells by measuring the TS-catalysed conversion of [3H]-dUMP to dTMP and 3H2O, as described previously (Rots et al, 1999a). Briefly, 0·1 × 106 cells were incubated in 150 μl of culture medium with and without MTX (gift of Pharmachemie, Haarlem, the Netherlands) in five concentrations (0·156–40 μmol/l), for an exposure of 3 h followed by an 18-h drug-free period (short-term exposure) or for 21 h (continuous exposure). After 4 h, [5-3H]-2′-deoxycytidine (Moravek Biochemicals, Brea, CA, USA; final concentration of 1 μmol/l, 92·5 GBq/mmol) was added as a precursor for [3H]-dUMP. Data are expressed as the concentration of MTX necessary to inhibit 50% of the TS activity (TSI50).

MTX accumulation and polyglutamylation Ten million leukaemic cells were incubated for 24 h with 1 μmol/l [3′,5′,7–3H]-MTX (Moravek Biochemicals; final specific activity 74 GBq/mmol) in 5 ml of culture medium, as described previously (Rots et al, 1999b). After washing, samples were counted for radioactivity, cell number and viability. The remaining suspension was centrifuged and the cell pellet used for extraction of polyglutamates. The polyglutamates were separated by high performance liquid chromatography (HPLC) using an anion-exchange column, as described previously (Braakhuis et al, 1993; Rots et al, 1999b). The data are expressed as pmol of MTX-Glun/109 cells.

Statistics The distribution of clinical and biological variables for patients with and without the p15/p16 gene deletions were compared using the Mann–Whitney U (MWU) or χ2 test. LC50 values were compared between homozygously deleted, hemizygously deleted and non-deleted samples for the p15/p16 gene using the MWU test.

We retrospectively evaluated the prognostic impact of the p15/p16 gene deletion in patients treated with Dutch DCSLG ALL-VII and -VIII and German COALL-92 and -97 protocols. Median follow-up time was calculated as the median survival time from time of study entry for those patients who remained alive and disease-free. Complete remission (CR) was defined as less than 5% leukaemic blasts in the bone marrow after counting 200 nucleated cells, with no signs of extramedullary leukaemia. The failure to achieve a CR at the end of the induction treatment, a non-response, was considered an event censored at time-point zero. Early death was defined as death before completion of induction therapy, counted as an event for EFS censored at d 0. A relapse was defined by the reoccurrence of leukaemic cells in the peripheral blood, bone marrow, spinal fluid or other extramedullary sites of a patient who initially achieved a CR. The DFS was defined as the time from first diagnosis to a leukaemia-related event, that is non-response or relapse. EFS was defined as the time from first diagnosis to an event: relapse, death by any cause or secondary malignancy. Estimates of the probability (P) of DFS and EFS were calculated according to the Kaplan–Meier method and compared using the log-rank test of equality. The analyses were two-tailed at a significance level of 5%.

Results

Detection of p16 gene deletions

FISH for the presence of the p15/p16 gene was performed on 122 diagnostic samples from T-ALL patients. Twenty-four samples (19·7%) were not evaluable either owing to inadequate hybridization or poor morphology. The majority of these failures were cytospins stored for longer than 6 years at −20°C. Of the 98 samples with FISH results, 31 (32%) had two alleles present of the p15/p16 locus, as detected by the P1 1063 probe (p16+/+). The remaining 67 (68%) cases showed deletions of p16: 46 with homozygous deletions (p16−/–), i.e. no allele detectable (68·7% of the deletions), seven with hemizygous deletions (p16+/–), i.e. one allele detectable (10·4% of the deletions), and 14 with mixed populations (zero to two alleles present; 20·9% of the deletions).

Fourteen randomly chosen patient samples were also blindly analysed using Southern blot analysis, generally confirming the FISH results (Table I, Fig 1). The Southern blot analysis also detected rearranged bands in two patients and, as expected, gave more precise information on the exact exons deleted. The major discrepancy was for patient number 7684, two alleles present with the FISH and hence classified as ‘germ-line p16’, who had a deletion within the locus (0–1 alleles for p16 exon 1 and p15 exon 2) with the outer extremeties present (p16 exon 2 and p15 exon 1) for the Southern blot analysis.

Table I.  Southern blot analysis compared with FISH results for 14 T-ALL patient samples.
 Southern*FISH
Samplep16 exon 2p16 exon 1p19ARF exon 1ßp15 exon 2p15 exon 1P11063
  • *

    Arranged in order from telomeric (left) to centromeric end (right).

  •  †Number of alleles detected by FISH; percentage of blast cell population with 0, 1 or 2 alleles.
     

  • 2, two alleles present judged by intensity of the band; 1, monoallelic deletion judged by intensity of the band; 0, homozygous deletion; 0 − 1, faint band that could be contaminating normal cells or a small proportion of the cells retaining the gene; R, rearranged band.

4807111111 (80%)
49410000–10–10 (92%)
6291222222 (100%)
6714222222 (98%)
7024222222 (81%)
727300–100–10–10 (90%)
7335222222 (100%)
7486000000 (97%)
75470–10–12220 (77%)
768420 + 1R00 − 122 (100%)
78600000 + 1R0 − 10 (95%)
79920–10–100–10–10 (65%)
8010111111 (72%)
82401–21–211–21–21 (86%)
Figure 1.

Leukaemic cell DNA (numbered patients) and normal control (N) digested with HindIII and hybridized with probes from the Ink4 locus and control probes from chromosome 11 (Bcl-1) and 5 (D5S78). The D5S78 serves as a control for the larger and Bcl-1 for the smaller fragments. Patient 7335 and the normal control show two alleles present for all exons. Patient 8010 and 4807 have hemizygous deletion of all Ink4 exons. Patients 7273, 4941 and 7486 show complete loss of the entire locus. Patient 7547 has a partial deletion with retention of the p15 and E1 beta exons and homozygous deletion of p16 exons. In patient 7684, the ‘outer’ parts of the locus (p15 exon 1 and p16 exon 2) are retained, p15 exon 2 and E1 beta are homozygously deleted. One allele of p16 exon 1 is deleted and the other rearranged (indicated). A faint hybridization signal for p15 exon 2 indicating residual normal cells can be seen.

For this retrospective study, a relatively large bank of cryopreserved cytospins was available, hence, the FISH approach was chosen to analyse the p16 status. This allowed us to study a higher fraction of the patients.

P16 deletions and clinical presentation

For the following analyses of the relationship between p16 and in vitro drug resistance and prognosis, we compared the germ line p16+/+ cases with the p16−/– homozygously deleted cases. The hemizygously deleted p16 (n = 7, these have ≤ 5% p16−/– cells) and the mixed population (n = 14, > 5% p16−/– and > 18% p16+/–) cases have been excluded from the p16-deleted group. These cases still have one allele of the p16 gene present and it is unclear what the status of the remaining allele is in these cases (rearranged, hypermethylated).

The distribution of important clinical parameters within the p16 germ-line and homozygously deleted groups is summarized in Table II. Typical of T-ALL cases in general is the fact that both groups presented at a relatively old age and a large proportion of the cases had a high white blood cell count at initial diagnosis (WBC ≥ 50 × 109/l). An over-representation of the male gender, also a T-ALL phenomenon, was clearest in the p16 homozygously deleted group (ratio male:female is 2·07). This ratio was only 1·2 in the germ-line p16 group, but this did not differ significantly from the p16-deleted group.

Table II.  Clinical and biological characteristics of 77 T-ALL patients with and without the p16 gene.
p16 Germ-line n (%)Homozygously deleted n (%)P-value
  1. Percentages expressed excluding the missing cases. P-values determined by χ2 (sex, WBC) and Mann–Whitney U (age) test.

SexMale17 (55%)31 (67%)0·265
Female14 (45%)15 (33%) 
Median age (months, range) 104 (10–200)104 (32–177)0·862
WBC × 109/l< 5011 (35%)15 (33%)0·846
> 5020 (65%)30 (67%) 
Unknown1 

In vitro drug resistance

Two of the 98 samples successfully analysed with FISH were not tested for in vitro drug resistance. Seventy-one of the 96 tested samples were cultured successfully in the MTT assay (74%). The majority (22 out of 25) of the culture failures were as a result of less than 70% leukaemic blast cells in the control wells on d 4; other causes for failure were insufficient cell survival (n = 2) and poor duplicates (n = 1). The assay success rate did not differ between the p16 gene subgroups. The 35 p16 gene homozygously deleted cases and the 24 germ-line p16 cases showed a complete overlap of individual LC50 values for most of the eight classes of drugs tested with the MTT assay (Table III). The only drug for which the two groups differed significantly was dexamethasone (DEX): the p16 homozygously deleted group was a median 22·6-fold more sensitive to DEX than the germ-line p16 group (P = 0·030, Fig 2). In line with this result is the trend for relative sensitivity to prednisolone (PRED) in the p16 homozygously deleted group (27·4-fold difference of median LC50 value compared with germ-line p16, P = 0·061).

Table III.  Relationship between p16 deletion and in vitro drug resistance in 59 cases of childhood T-ALL.
DrugT-ALL germ-line p16 median (25th−75th)T-ALL homozygous deleted p16 median (25th−75th)P-value
  1. *11 germ-line p16 vs. nine homozygous deleted p16 cases, determined by the short-term exposure TS inhibition assay; IC50 values expressed as μmol/l.
     

  2. Median LC50 values for germ-line p16 (n = 24) vs. homozygous deleted p16 patients (n = 35). All LC50's are expressed as μg/ml except for l-asparaginase (IU/ml). P-values were determined using the Mann–Whitney U-test.

Prednisolone37·5 (2·00- > 250)1·37 (0·31–138)0·061
Dexamethasone1·73 (0·07- > 6·0)0·08 (0·04–0·36)0·030
Vincristine0·76 (0·36–3·71)1·56 (0·59–3·49)0·349
l-Asparaginase0·82 (0·01- > 10)0·21 (0·03–1·57)0·435
Daunorubicin0·09 (0·07–0·22)0·10 (0·09–0·19)0·524
6-Thioguanine3·91 (2·80–7·76)5·73 (4·00–8·10)0·360
Cytarabine0·91 (0·26- > 2·50)0·82 (0·44–2·10)0·905
Etoposide1·65 (0·71–2·68)2·35 (0·87–15·8)0·202
Ifosfamide4·92 (2·23–9·14)5·29 (2·50–12·3)0·620
Methotrexate*0·97 (0·50–17·0)5·79 (0·89–39·8)0·413
Figure 2.

The relationship between p16 gene deletion and in vitro drug resistance for DEX (A) and PRED (B) in childhood T-ALL. LC50 values for germ-line p16 (+/+) vs. homozygously deleted p16 (+/+) patients. The horizontal lines show the median values, the circles represent the individual patient samples. LC50's are expressed as μg/ml. P-value determined by the Mann–Whitney U-test.

Nine homozygously deleted p16 samples could be compared to 11 germ-line p16 T-ALL cases with respect to MTX sensitivity: the median values for the MTX IC50 (concentration necessary to inhibit 50% of TS activity) did not differ for either the short exposure or the continuous exposure conditions (PMWU = 0·302 and PMWU = 0·413 respectively). The median total accumulation of MTX polyglutamates (MTX-Glu1−6), as well as the median percentage of the pharmacologically more important long-chain polyglutamates MTX-Glu4−6, also did not differ between deleted and germ-line p16 T-ALL cases (n = 11 vs. 8, PMWU= 0·741, and n = 9 vs. 7, PMWU= 0·710 respectively).

Clinical outcome

The prognostic impact of the p16 gene deletion was retrospectively evaluated in 59 T-ALL patients treated with BFM-based Dutch DCSLG ALL-VII and -VIII protocols (n = 28) and German COALL-92 and -97 protocols (n = 31). The Dutch protocols use the leukaemia cell mass and response to a prednisolone prephase for stratification into treatment arms, hence, a number of the DCSLG T-ALL patients have received so-called ‘medium-risk’ treatment. For the COALL study group, T-ALL is one of the inclusion criteria for receiving the COALL high-risk treatment protocol.

The germ-line p16 group had an estimated 5-year DFS of 95·5 ± 4·4% (one relapse among 25 patients) compared with 65·1 ± 9·1% for the homozygously deleted p16 T-ALL cases (one non-responder and nine relapses among 34 patients) (Plog rank= 0·021, see Fig 3 for survival curve). The median follow-up of patients at risk was 40 months for both groups (range 11–109 and 7–106 months respectively).

Figure 3.

Prognostic value of the homozygous p16 gene deletion in 59 children with T-ALL treated according to COALL-92 and -97, and DCLSG-VII and -VIII protocols. The probability of disease-free survival (pDFS) is shown for 25 children with germ-line p16 (+/+, solid line) and 34 children with homozygously deleted p16 (−/–, dashed line). The number of patients at risk is indicated above the x-axis. Log-rank analysis: P = 0·020.

The germ-line p16 group had an estimated 5-year EFS of 75·6 ± 8·7% compared with 61·0 ± 9·0% for the homozygously deleted p16 T-ALL cases (Plog rank= 0·392). The germ-line p16 patients had six events: one relapse, two early deaths (before completion of induction therapy), two toxic deaths and one secondary malignancy (acute myeloid leukaemia). The p16-deleted patient group had one early death and one non-leukaemia-related death, as well as the events already mentioned for DFS. Overall survival for the germ-line p16 group was 75·6 ± 8·7% and for the p16 homozygously deleted group 72·8 ± 8·4% (Plog rank= 0·961).

In order to examine the prognostic impact of p16 deletion in more homogeneous treatment groups, the survival distributions were adjusted by stratification for the treatment given. This resulted in three groups: COALL high-risk treatment (n = 31), Dutch ALL medium-risk (VII-RG and VIII-MRG, n = 19) and Dutch ALL high-risk (VII-EG and VIII-HRG, n = 9) treatment arms. p16 homozygously deleted patients still had poorer 5-year estimates of DFS than the germ-line p16 group (Plog rank= 0·036).

Discussion

In the present study, we have determined p16 deletion status using FISH analysis for a large group of childhood T-ALL cases at diagnosis, and examined the impact of homozygous p16 deletion on in vitro drug resistance and clinical outcome within this group.

The frequency of p16 deletions in this study, 68·4%, is in agreement with the published data (Hebert et al, 1994; Fizzotti et al, 1995; Ohnishi et al, 1995; Okuda et al, 1995; Diccianni et al, 1997). The majority of these deletions were homozygous, but 7% had hemizygous deletions, i.e. one allele detectable, and 14% presented a mixture of populations with 0–2 alleles. The detection of this latter group is a further refinement offered by the FISH technique. These hemizygous and mixed population groups are too small to be analysed as separate entities and are hence excluded from the data presented. One could also add these cases to the p16+/+ and/or p16−/– groups. It seems most logical to then place the hemizygously deleted cases in the non-deleted group, as point mutations of the remaining allele are rare. However, the mixed populations are more difficult to assign because some cases have a much larger population of p16−/– cells (and would hence seem to belong to the homozygously deleted group), whereas other cases have at least one allele of the p16 locus present in the greater majority of the cells. Various grouping strategies have been analysed, without adding new insights to the results as presented (data not shown).

p16 germ-line and homozygously deleted p16 T-ALL cases did not differ significantly in patient characteristics at presentation, such as WBC or age. We have not analysed mediastinal mass vs. p16 deletions; a previous study of T-ALL patients found no correlation with mediastinal mass or WBC (Diccianni et al, 1997).

The p16/p19ARF homozygously deleted cases did not differ in cellular drug resistance from the germ-line p16 cases, except for the glucocorticoids. Unexpectedly, the p16 homozygously deleted cases were more senstive to DEX than the p16 germ-line cases, with a trend for sensitivity to PRED. p16 and p15 are both cycline D kinase inhibitors (CDK4/6), thus playing a role in the arrest of the cell cycle at the G1 phase through the phosphorylation status of Rb. The integrity of this cell cycle arrest pathway (with p16 present) has been shown to protect cells from chemotherapy in some model systems (Stone et al, 1996; Shapiro et al, 1998). Moreover, Urashima et al (1997) have described a p16−/– B-lineage cell line that displays DEX-triggered apoptosis. They show that reinduction of p16, associated with dephosphorylated Rb protein, confers resistance to DEX-induced apoptosis, supporting our observations that germ-line p16 cases are relatively more resistant to DEX. The p19ARF protein was also deleted in the p16 locus homozygously deleted cases in this study. p19 ARF interacts with Mdm-2 resulting in a reversal of the inhibitory binding of Mdm-2 to the p53 protein (Kamijo et al, 1998; Honda & Yasuda, 1999; Tao & Levine, 1999). Loss of p19ARF will lead to inactivation of p53 which, if anything, leads to drug resistance (Schmitt et al, 1999). The above mechanism cannot clarify the relative sensitivity of the p16-deleted cases to glucocorticoids. Furthermore, glucocorticoid-induced apoptosis is thought to be p53 independent (Clarke et al, 1993; Lotem & Sachs, 1993; Lowe et al, 1993; Brady et al, 1996). p16 alterations have been associated with a sensitivity to anti-metabolites in primary human astrocytic tumour samples; glucocorticoids were not tested (Iwadate et al, 2000). This effect was ascribed to the loss of the G1 check-point leading to an increase in the proportion of cells in S phase. Hence, it is unclear whether effects of p16 alteration on chemosensitivity are as a result of changes in cellular proliferation, direct effects on apoptosis or other molecular mechanisms.

The homozygously deleted p16 T-ALL patients treated with the modern multiagent chemotherapy of the DCSLG and COALL protocols have a significantly lower 5-year DFS than germ-line p16 T-ALL cases. Hence, the nondeleted p16 cases identify a group of T-ALL patients with a high disease-free survival. However, in this study, the event-free and overall survival of the germ-line p16 group did not differ significantly from that of the homozygously deleted group. The events that cause this difference between DFS and EFS are the relatively high number of early and toxic deaths in the germ-line p16 group. Diccianni et al (1997) found that p16 status in T-ALL at diagnosis was not related to relapse rate or EFS in the Paediatric Oncology Group (POG) 9000 and 9400 protocols. In their study, the p16/p15 deletion had a relationship with a poorer outcome only after relapse, when p15 was also 100% co-deleted (at diagnosis: two out of three were p15 co-deleted). Their findings at initial diagnosis are in contrast to ours. Our study includes a larger fraction of T-ALL cases with a low WBC (in Diccianni et al, 1997 almost all patients were high-risk T-ALL with WBC > 50 × 109/l). However, the studies are in agreement in that, if p16 deletion is of prognostic significance, it is associated with a poor outcome. To the best of our knowledge, no other studies have examined the prognostic value of p16 deletions in a large group of T-ALL patients at diagnosis in detail.

It must be noted that the patients in the present study had been treated with a number of different protocols. A further limitation of the present study, in which FISH was used for the detection of the p16/p15 deletions, is that, owing to the nature of the probe used, we were unable to examine differences as a result of deletions of these genes individually or the specific role of p19ARF. We have also not looked for rearrangements, mutations or hypermethylation of the p16 gene; hence, one cannot be sure of the expression of p16 in the ‘germ-line p16’ patients in this study. However, point mutations, hypermethylation and rearrangements of the p16 gene seem to be relatively rare in childhood T-ALL cases (Diccianni et al, 1997).

It remains somewhat puzzling that, in our study, the p16 homozygously deleted cases are significantly more sensitive to DEX and yet this group fares considerably worse in terms of DFS. It must be noted that one is looking at relative glucocorticoid sensitivity within a subgroup of ALL that has been described to be significantly more resistant to glucocorticoids than common/preB-ALL (Kaspers et al, 1995b; Pieters et al, 1998). p16 homozygously deleted cases are still a median 1·2-fold higher for DEX and 3·2-fold higher for PRED, compared with the median LC50's for a large reference group (n = 413) of common/preB ALL cases. One has then to presume that the poor in vivo response of the p16-deleted cases is governed by other mechanisms. p53 inactivation, through mutation or Mdm-2 overexpression, has been associated with a very poor prognosis in some studies (Diccianni et al, 1994; Marks et al, 1997). This might in part be the basis of the poor DFS of the p16/p19ARF homozygously deleted cases.

In conclusion, p16 homozygously deleted T-ALL patients have a significantly poorer 5-year DFS that cannot be explained by intrinsic drug resistance. Moreover, this study suggests that primary childhood T-ALL with germ-line p16 has a very high DFS.

Acknowledgments

This study was supported by the Dutch Cancer Society grant 5U95–92 and the Children's Cancer Fund of Sweden. We wish to thank the paediatric oncological centres participating in the DCLSG, the Netherlands, and the German COALL study group. Board members of the DCLSG were: H. van den Berg, M. V. A. Bruin, J. P. M. Bökkerink, P. J. van Dijken, K. Hählen, W. A. Kamps, F. A. E. Nabben, A. Postma, J. A. Rammeloo, I. M. Risseeuw-Appel, A. Y. N. Schouten-van Meeteren, G. A. M. de Vaan, E. T. van ‘t Veer-Korthof, A. J. P. Veerman, M. van Weel-Sipman and R. S. Weening. Board members of the COALL-92/97 study are: U. Göbel, U. Graubner, R. J. Haas, G. E. Janka-Schaub, N. Jorch, H. Jürgens, H. J. Spaar and K. Winkler. The P1 1063 clone of the CDKN2 gene was developed by J. Weaver-Feldhaus (Myriad Genetics, Salt Lake City, USA) and Dr N. A. Gruis (Department Human Genetics, University of Leiden, The Netherlands), and the P1 164 clone of the entire ETO gene was developed by Dr N. Sacchi (Medical School, University of Milan, Italy), both kindly provided by Prof. A. Hagemeijer (European Concerted Action Co-ordinator, University of Leuven, Belgium). The p16ink4a and p15ink4b probes for the Southern blot experiments were kindly provided by Dr D. Beach. Dr J. F. van Weerden’s assistance with the DCLSG patients' survival data is gratefully acknowledged.

Ancillary