Enhanced Expression of P16ink4a Is Associated with a Poor Prognosis In

Home , Cyclin A

Leukemia (1999) 13, 181–189  1999 Stockton Press All rights reserved 0887-6924/99 $12.00 http://www.stockton-press.co.uk/leu Enhanced expression of p16ink4a is associated with a poor prognosis in childhood acute lymphoblastic leukemia Y Mekki1, R Catallo1, Y Bertrand1,2, AM Manel1,3, P Ffrench3, N Baghdassarian1, P Duhaut4, PA Bryon1,3 and M Ffrench1,3

1Laboratoire de Cytologie Analytique, Universite´ Claude Bernard, MESRT JE 1879; 3Laboratoire Central d’He´matologie et de Cytoge´ne´tique, Hoˆpital E Herriot; 2Service d’He´matologie Pe´diatrique, Hoˆpital Debrousse; and 4RECIF, Universite´ Claude Bernard, Lyon, France

The tumor suppressor gene p16ink4a is homozygously deleted p16ink4a overexpression leads to a G1 arrest in the presence of in numerous T as well as in some B lineage acute lymphoblas- a functional pRb.5 When present in high levels, the inhibitor tic leukemia (ALL). We therefore analyzed the clinical and bio- ink4a competes with D cyclins for binding to CDKs, which are logical implications of this feature by studying p16 6 ink4a expression in 58 cases of childhood ALL. mRNA and protein sequestered in inactive protein complexes. p16 mRNA is, were significantly correlated and both appeared more highly at least in part, negatively controlled by pRb. However, it has expressed in B than in T lineage ALLs: 13 out of the 15 T cell also been shown that p16ink4a mRNA and protein accumulate ALLs did not show any p16ink4a expression. The main result of in senescent fibroblasts, and an apparent overexpression ink4a this study is the strong prognostic value of p16 expression. could be caused by an increase in the number of When stratifying the patients in three groups according to 7 p16ink4a expression, we observed in univariate analysis: (1) the population doublings. shortest disease-free survival for patients presenting a high The INK4a gene was mapped to chromosome 9p21, and a p16ink4a level; (2) contrasting with the good prognosis in the high frequency of homozygous deletions or rearrangements of group of patients expressing p16ink4a at low level; (3) while the p16ink4a locus was reported in numerous human cancer cases without any expression of the inhibitor were associated cell lines, including leukemic cell lines.8 Homozygous INK4a = with a medium course of the disease (P 0.0165). This prog- deletions are less frequent in primary leukemic cells than in nostic value was confirmed by the multivariate analysis show- ink4a cell lines, but may occur without any detectable cytogenetic ing therapeutic regimen and p16 protein expression as the 9 only variables retained in the model. A specific metabolic pro- changes in 9p. This deletion is observed more frequently in file related to cellular survival and proliferation was observed T acute lymphoblastic leukemia (ALL) than in B lineage in each of the three p16ink4a expression groups. Among the cell ALL.10–15 The comparison of clinical features at presentation cycle-related proteins we analyzed, only p21waf1 bcl-2 and CDK4 in p16ink4a−/− and in p16ink4a germ line ALL cases showed a ink4a were significantly and positively correlated to p16 . Further- greater leukemic cell mass and a higher white blood cell more, CDK6 was also strongly expressed in the group of cases 13 ink4a with high p16ink4a level. An enhancement of p16ink4a, p21waf1 and count in the first group. However, it is unlikely that p16 bcl-2 was previously described in prolonged cellular survival, gene deletion could represent a clinically prognostic factor in while aging cells showed a decrease in CDK4 expression. The childhood ALL.14,15 concomitant high expression of the oncogenic protein CDK4 Results of treatment in childhood ALL remain incompletely (and of CDK6), we observed, may amplify the leukemic advan- satisfactory because of relapses observed even with high-dose tage of prolonged lifespan blast cells by favoring cell pro- chemotherapy. In order to better understand the regulation of gression through G1 phase. These data suggest that at least two mechanisms may be associated in the oncogenesis of very mitotic activity of lymphoblastic cells in childhood ALL and its aggressive ALLs, ie deregulation of cell multiplication and relationship to treatment outcome, we turned to the analysis of prolonged blast lifespan. p16ink4a mRNA and protein expression and their relationships Keywords: childhood acute lymphoblastic leukemia; p16ink4a; to the other cell cycle regulatory or related proteins. We show p21waf1; cyclin; CDK; prognosis the poor influence of an enhanced expression of p16ink4a on prognosis and the possible linkage of a p16ink4a high level to prolonged blast survival. Introduction

The intra-cellular control of cell proliferation is mainly Materials and methods effected by a group of serine-threonine kinases associated, when active, with cyclins and called cyclin-dependent kinases Patients (CDK). CDK4 (or CDK6) associated with cyclin D (D1, D2 or D3), their regulatory subunits, are involved in G1 pro- Fifty-eight cases of childhood ALL were studied from January gression.1,2 These kinases phosphorylate and inactivate the 1994 to July 1997. Thirty patients were male and 28 were retinoblastoma susceptibility gene product (pRb), enabling female, with ages ranging from 0.24 to 17.95 years (median: cells to continue the cell cycle. The kinasic activity is mainly 6.6). Cell samples were obtained, at diagnosis, from bone regulated by: (1) the association with cyclins; (2) several steps marrow aspiration in all cases. Selection for evaluation was of phosphorylation and dephosphorylation; and (3) binding of based on the availability of cell sample; consequently, specific inhibitors called CDK inhibitors (CKI).3 p16ink4a patients having high initial blast-cell counts are numerous in inhibits the kinasic activity of the CDK4 (or CDK6)/cyclin D this study (WBC over 100 G/l in 26% cases). The ALLs were enzymes and represents a major CKI for the G1 phase.4 classified according to the French–American–British (FAB) recommendations16 after cytological examination and cyto- chemical assays (peroxidase and naphthyl acetate esterase) as Correspondence: M Ffrench, Laboratoire de Cytologie Analytique, L1 (50 cases), L2 (five cases), L3 (one case) and unclassified Faculte´ de Me´decine Rockefeller, 8 Av Rockefeller, 69 373 Lyon Cedex 08, France; Fax: 04 72 11 73 05 (one case). All cell samples contained at least 80% blast cells. The first and second authors contributed equally to this work. Mononuclear cells were isolated by Ficoll–Hypaque density Received 24 August 1998; accepted 2 November 1998 gradient (d = 1.077) centrifugation. In all cases, immunolog- p16ink4a and prognosis of ALL Y Mekki et al 182 ical phenotype, cytogenetics and cell cycle variables were gel(7.5% SDS polyacrylamide gel for pRb) and transferred to determined on cell suspension at diagnosis and a specimen nitrocellulose membrane in a semi-dry system as previously of a neoplastic tissue was frozen at −80°C pending protein described.17 Upon completion of transfer, Ponceau S staining and mRNA studies. was performed to verify that an equal amount of protein was Immunological tumor cell phenotype was determined as present in all lanes.20 To ensure independence of expected previously described.17 Lymphoblast populations were div- variations between different ALL groups from differences ided into four categories: (1) early early pre-B (CD10−) ALL resulting from gel and blot preparations, we ran a sample of (six cases), (2) early pre-B (CD10+) (33 cases) and pre-B-ALL Reh-6 cell line as an internal control. For p16ink4a, the HeLa (IgM+) (three cases), (3) B ALL (one case) and (4) T-ALL (15 cell line was used as internal control. cases). The membranes were probed with the following antibodies: Chromosome analysis of bone marrow was performed in mouse monoclonal anti-p16ink4a (1:2000), anti-pRb (1:500), all cases by short-term cultures without phytohemaglutinin. mouse monoclonal IgE anti-cyclin A (1:1000) from Phar- Banding analysis was based on RHG banding. 20 banded Mingen (San Diego, CA, USA); mouse monoclonal anti- metaphases were karyotyped per patient. A t(9;22)(q34;q11) p34cdc2 (1:500), anti-cyclin B (1:500), anti-cyclin E (1:250), translocation was observed in two cases and a anti-p53 (1:500) and rabbit polyclonal anti-cdk2 (1:500), anti- t(4;11)(q21;q23) translocation in two other cases. cdk4 (1:2000), anti-cdk6 (1:1000), anti-cyclin D3 (1:250), Patients were treated according to the EORTC 58881 CLCG anti-p27Kip1 (1:1000) from Santa Cruz Biotechnology (Santa trial18 in the same institution (Debrousse Hospital, Lyon, Cruz, CA, USA); mouse monoclonal anti-p21Waf1 (1;250) from France). Patients were stratified according to their risk factor. Amersham (Little Chalfont, UK), anti-Bcl2 (1:500) from Dako- A so-called very high risk group (VHR) was defined by the patts (Copenhagen, Denmark). The membranes were then presence of at least one of the following criteria regardless of incubated with biotinylated rabbit anti-mouse immuno- the initial blast cell count: (1) blast cell count over 1 G/1 at globulin antibody (Dakopatts), biotinylated goat anti-rabbit the end of the glucocorticoid pre-treatment; (2) non-achieve- immunoglobulin antibody (Sigma) or biotinylated rat anti- ment of complete remission at the end of induction; (3) poor mouse IgE antibody (Southern Biotechnology Associates, cytogenetic features, ie t(9;22), t(4;11), t(1;19) or near hap- Birmingham, AL, USA). In order to confirm the similarity of loidy. A resistance to glucocorticoid was found in 16 cases. specimen quality a labeling of ␤-actin (monoclonal antibody Thirty-five patients were treated with the standard risk sched- from Serotec) was realized. The final step was incubation with ule, 20 with the VHR protocol. Two patients received another streptavidin-biotinylated horseradish peroxidase complex treatment: the unique L3 ALL of this series, and a patient (Amersham) and development with an enhanced chemo- treated in another hospital. These two patients are alive, in luminescence (ECL) detection system (Amersham). complete remission. Follow-up could not be obtained for one The total intensity of the bands was measured by densito- patient, a Jehovah’s Witness who did not accept treatment. metry with correction for the background gray labelling of the Complete remission was obtained in all the 57 remaining lanes with a Quantimet 570 image analyzer (Leica, Cam- treated cases. A relapse occurred in eight of these cases, two bridge, UK) (Centre Commun de Quantime`trie, Claude of them presented a poor cytogenetic feature (one t(9;22) and Bernard University Lyon I). Numerical values of sample sig- one t(4;11)). These medullary relapses were followed by death nals were given as arbitrary units (AU). A.U. were calculated in six cases. Two children died in complete remission. by dividing the numerical value of a sample signal by that of the signal from internal positive control on the same blot. In addition, cdk1 and pRb expression were also analyzed Flow cytometry methods qualitatively, ie the presence of slower migrating supplemen- tary bands, corresponding to phosphorylated isoforms, was A flow cytometry study of DNA content was carried out on recorded. cell suspensions for 57 patients in accordance with the 19 method already published. Briefly, ethanol-fixed cells were Reverse transcription (RT)-PCR assay stained with propidium iodide after ribonuclease A III (Sigma, St Louis, MO, USA) treatment. The stained cells were ana- Ink4a mRNA expression was studied in 42 out of the 58 cases lyzed on a FACscan flow cytometer (Becton Dickinson, San for which we had enough cell material. Jose, CA, USA). The discrimination between G1 doublets and Total RNA was extracted from ALL cell samples by a guani- G2 cells was achieved by correlated measurements of pulse dinium–thiocyanate–phenol/chloroform procedure using RNA area and peak of red fluorescence signal. For each sample, Plus(Bioprobe Systems, Montreuil, France). Semi-quantitative 10 000 cells were analyzed. The percentages of cells in the RT-PCR of Ink4a was performed in accordance with the cell cycle phases G0/G1, S and G2+M were calculated ‘Primer-Dropping’ method described by Wong et al21 with co- according to the R-FIT model (software provided by the amplification of GAPDH as internal control in each sample. manufacturer). Ploidy was determined by flow cytometry and Primers for GAPDH were dropped into the reactions starting controlled by cytogenetics (in one case, ploidy was determ- from the 16th-cycle, during the denaturation step, and the ined only by cytogenetics). Lymphoblasts were diploid in 33 PCR amplification was continued for 25 cycles. The number cases, hyperploid in 23 cases, and hypoploid in two cases. of cycles was defined on previous experiments in order to remain in independent exponential amplification for both templates. Negative controls were realized in order to detect Protein studies contaminations, without addition of reverse transcriptase and with water instead of RNA. To ensure independence of p16ink4a was analyzed in all 58 cases and compared to the expected variations between different ALL groups from differ- expression of other cell cycle regulatory proteins. ences resulting in reverse transcription and amplification, we Cell samples were lysed and 70 ␮g of proteins were separ- ran a sample of HeLa cell line as an internal control in each ated electrophoretically in a 12% SDS-polyacrylamide set of RT-PCR. p16ink4a and prognosis of ALL Y Mekki et al 183 Aliquots of PCR reaction products were electrophoresed highly in one case (0.86 AU). A highly expressed protein was through 2% agarose gels containing 0.5 ␮g/ml ethidium bro- associated with a high expression of mRNA (Figure 1). mide. Gels were illuminated with UV light, photographed Among the 58 cases studied for protein expression, 43 ALLs using a numeric Kodak Camera (Eastman Kodak Company, were B lineage ALLs and 15 T cell ALLs. The median and Rochester, NY, USA) and quantified using the EDAS software extreme values of the proteins studied, biological data concern- from Kodak. Numerical values of sample signals were given ing cell proliferation (S and S+G2/M fractions), and tumor burden as AU. A.u. were calculated by dividing the numerical value (WBC) according to immunological phenotype are indicated in of a sample signal by that of the signal from Hela cell-positive Table 1. p16ink4a expression varied significantly according to control on the same RT-PCR set (each divided by the value of immunological subtypes (P Ͻ 0.0001) (Figure 2). p16ink4a protein the coresponding GAPDH internal control). was not observed in 13 out of the 15 T cell ALLs, and in 10 out The sequences of human ink4a and GAPDH primers were as of 43 B lineage ALLs (10 early pre-B ALLs). p16ink4a was described by Yamada et al22 and Wong et al.21,23 Primer expressed in all the six early early pre-B ALLs and reached the sequences for Ink4a were 5′-AGCATGGAGCCTTCGGCTGACT- highest levels in this group. Among the proteins studied, beside 3′ (exon 1 ␣); 5′-GACCTTCCGCGGCATCTATGC-3′ (exon 2) p16ink4a,p21waf1 and Bcl2 were significantly more strongly and for GAPDH 5′-CGGAGTCAACGGATTTGGTCGTAT-3′;5′- expressed in B lineage than in T-ALLs (P Ͻ 0.001). p21waf1 was AGCCTTCTCCATGGTGGTGAAGAC-3′. The length of PCR not found in three T-ALLs and was very low in the 12 remaining products expected were 430 and 306 bp, respectively. T cell cases (Table 1); CDK4 was also higher in B lineage ALLs compared to T cell ALLs (P = 0.025). Inversely to p16ink4a,CDK1 and CDK6 were significantly higher in T than in B lineage ALLs Ͻ Ͻ Statistical methods (P 0.002 and P 0.02, respectively) (Figure 2).

The analysis was based on non-parametrical methods: Spear- man rank correlation coefficients, and Kruskall and Wallis test. Variable distribution was summarized by box plot rep- Cell proliferation and p16ink4 expression resentation showing (1) the median, (2) the 25th and the 75th percentile corresponding to each extremity of the box, (3) the smallest and the largest observed value that is not outlier No correlation could be observed between p16ink4a protein (or (drawn as ‘whiskers’), and (4) the outlier (circle) and extreme mRNA) expression and the percentage of cells in S or S+G2+M (star) values. phases, nor with CDK1 or cyclin A, two cell cycle regualtory The prognostic value of the biological variables studied proteins known to be linked with cell proliferation.25,26 p16ink4a were analyzed by univariate analysis. We chose as cut-off expression did not significantly vary according to the phos- point the median of the distribution for most variables (CDK4, phorylation levels of CDK1. CDK6, p21waf1, bcl-2). For WBC, three groups of patients were defined: (1) under 10 G/l, (2) over 100 G/l, and (3) between these two values. For p16ink4a protein, three groups of patients were also considered: (1) with p16ink4a equal 0, (2) with p16ink4a superior to the median value, 0.210 AU, obtained in cases expressing the protein, and (3) with p16ink4a between these two values. For the mRNA, three groups were also defined: (1) ink4a mRNA equal to 0, (2) ink4a mRNA over the 0.75 quartile value, and (3) between these two values. Sur- vival curves were estimated by the Kaplan–Meier method. The significance of a survival difference observed between groups was assessed with the log-rank test and the Tarone and Ware test. A multivariate analysis was performed using the Cox model with the forward stepwise procedure. The analysis was performed with SPSS Statistical Software (SPSS, Chicago, IL, USA).

Results

Expression of p16ink4a in ALL according to immunological phenotype

A strong positive correlation was found between p16ink4a mRNA and protein expression (42 cases, r = 0.80, P Ͻ 0.00001) and this is in agreement with the previously described transcriptional regulation of the human ink4a gene.7,24 Protein and mRNA were Figure 1 Correlation between ink4a mRNA analyzed by RT-PCR ink4a not found in 12 cases (seven T cell and ﬁve B lineage ALLs) out and p16 protein analyzed by Western blot. (a) Example of RT-PCR of the 42 cases studied at both expression levels. The protein with co-ampliﬁcation of ink4a and GAPDH and (b) Western blot of p16ink4a with the internal control, ie HeLa cells, and with nine samples was never observed when the corresponding mRNA was not of different immunological phenotype of ALL. (c) Correlation between found. In three B lineage ALL cases, the protein was at 0 while mRNA and protein expression on the 42 cases analyzed (Spearman mRNA was weakly expressed in two cases (0.184 and 0.09) and test, r = 0.80. p16ink4a and prognosis of ALL Y Mekki et al 184 Table 1 Expression of cell proliferation regulatory proteins, WBC and cell-cycle variables (S and S+G2/M) according to immunological phenotype

Early early pre- Early pre-B and B-ALL T- cell ALL P value B-ALL pre-B ALL M (N) M, EV (N) M, EV (N) M, EV (N)

p16ink4a 0.40,0.03–0.93 (6) 0.08,0–0.75 (36) 0.01 (1) 0,0–0.13 (15) Ͻ0.0001 p16mRNA 0.62,0.37–0.93 (5) 0.59,0–1.4 (28) ND 0,0–0.51 (9) Ͻ0.0001 p21waf1 0.41,0.2–0.7 (6) 0.15,0–2.7 (36) 0.04 (1) 0.01,0–0.12 (15) Ͻ0.001 p27kip1 4.8,0.07–10.9 (6) 1.8, 0.02–18.4 (36) 1.83 (1) 1.3,0.35–3.4 (15) NS cdk1 0.47,0.14–0.69 (6) 0.32,0.04–0.83 (36) 0.44 (1) 0.7,0.11–0.97 (15) Ͻ0.002 cdk2 0.35,0.2–0.58 (6) 0.23,0–0.88 (30) ND 0.28,0.07–0.82 (15) NS cdk4 0.88,0.57–1.5 (6) 0.49,0–1.1 (36) 0.39 (1) 0.25,0.05–1.2 (15) Ͻ0.025 cdk6 0.46,0.24–0.66 (6) 0.37,0.04–0.97 (36) 0.18 (1) 0.62,0.06–1 (15) Ͻ0.02 cyclin A 0.07,0.01–0.11 (6) 0.035,0–1.2 (30) ND 0.03,0–0.28 (15) NS cyclin E 0.37,0.17–0.79 (6) 0.24,0–0.69 (36) 0.17 (1) 0.27,0.05–1.2 (15) NS cyclin D3 0.56,0.3–0.91 (5) 0.14,0.003–0.9 (30) ND 0.12,0.05–0.63 (15) NS pRb 0.22,0.05–0.48 (6) 0.09,0–0.5 (30) ND 0.01,0–0.22 (15) NS p53 0.19,0.14–0.44 (5) 0.1,0.01–0.45 (26) ND 0.1, 0.01–0.4 (15) NS bcl-2 0.65,0.58–0.71 (6) 0.55,0.01–1.3 (36) 0.09 (1) 0.3,0.04–1 (15) Ͻ0.001 WBC 84.2,1.8–675 (6) 15.8,1.2–649 (36) 18.3 (1) 97,5.8–583 (15) Ͻ0.025 S 4.9,1.1–11.3 (6) 4,0.9–20.1 (35) 13.5 (1) 6.6,0.9–29.4 (15) NS S+G2/M 6.8,2.5–12.7 (6) 7,1.9–22.1 (34) 15 (1) 11,1.5–29.6 (15) NS

Protein expression is given in arbitrary units (AU), cell cycle data as the percentage of cells in S or S+G2/M phases, WBC level in G/l. Differences between groups are analyzed by the Kruskall and Wallis test. WBC, white blood cells; M, median; EV, extreme values; N, number of cases.

Figure 2 Box plot representation of cell cycle regulatory proteins showing significant differences according to immunological phenotype in ALL. p16ink4a, p21wafl, bcl-2, CDK1, CDK4, CDK6. (1: early early pre-B; 2: early pre-B and pre-B, 3: B; 4: T cell ALL; N: number of cases.) p16ink4a and prognosis of ALL Y Mekki et al 185 population was restricted to the B lineage ALL (43 cases, P Ͻ 0.025) or to the VHR treatment group (20 cases, P Ͻ 0.03). Although not significant, the ink4a mRNA high level group was also associated with the poorest prognosis (42 cases, P = 0.118). p16ink4a protein expression was included in a multivariate Cox regression analysis with known prognostic factors: therapeutic regimens (which take into account poor cytogenetic features, ie t(9;22), t(4;11), t(1;19)), corticoid sensitivity, WBC level, ploidy and CDK4. On this series of 57 patients, in univariate analysis, DFS was shorter with VHR treatment (P = 0.015), corticoid resistance (P = 0.015), hyperleukocytosis (P = 0.04), diploidy (P = 0.085) and high CDK4 level (P = 0.058) (Table 2). The multivariate analysis using the forward stepwise method was performed on the 57 patients. The kind of treatment and p16ink4a protein expression appeared to be the best prognostic predictors and were the only variables retained in the model (Table 3).

Figure 3 Disease-free survival curves in the 57 ALL patients stratified according to p16ink4a expression; numerical values are given in Relationships between enhanced p16ink4a expression and ink4a Ͼ ink4a = Table 3; (a) p16 0 and below the median value; (b) p16 0; biological characteristics of ALLs (c) p16ink4a above the median value. In order to understand the reasons for the prognostic value of ink4a ink4a p16 , we sought out the existence of distinctive clinical or p16 expression and disease-free survival (DFS) in ink4a ALLS metabolic features associated with each of the three p16 sub- groups presented in Table 4. Hyperleukocytosis and diploidy (two factors of poor prognosis) were more frequently observed Complete remission was obtained in all the 57 cases treated, ink4a and a relapse occurred in eight of these cases. when p16 was not found than in cases expressing this protein (P = 0.002 and P = 0.0075, respectively). No difference in The three groups of patients, as previously defined according ink4a to p16ink4a expression, were considered for univariate DFS analy- p16 expression was observed according to glucocorticoid sis. A significant difference was observed for the DFS according sensitivity or according to the kind of treatment. No peculiar to the level of p16ink4a (57 cases, P = 0.0165) (Figure 3 and cytogenetic characteristics could be retained for the poor prognostic group: we may point out that the two patients with a Table 2). The shortest DFS was obtained for the enhanced ink4a ink4a t(4;11)(q21;q23) translocation highly expressed p16 .One p16 expression group (16 patients, five relapses), contrasting ink4a with the good evolution of patients expressing the CKI at low patient presenting a t(9;22) translocation also expressed p16 level (18 patients, no relapse); the absence of p16ink4a expression at a high level, while the other patient showing this translocation belongs to the low p16ink4a expression group. No correlations was associated with a medium evolution (23 cases, three ink4a relapses). This prognostic value was retained when the patient could be found between p16 and pRb protein expression or between p16ink4a expression and pRb phosphorylation level. Among all proteins analyzed in the whole group of patients, ink4a waf1 Table 2 Univariate disease-free survival analysis for all patients few proteins were correlated to p16 expression: p21 (r = 0.66), CDK4 (r = 0.51) and bcl2 (r = 0.55) (P Ͻ 0.0001). Factor No. of Estimated No. P value Enhanced expression of p16ink4a was associated with a high level patients % DFS at risk of these three proteins (Figure 4, Table 4). These significant positive correlations were maintained when the analysis was restric- Corticoid sensitivity ted to the 43 B lineage ALLs. Furthermore, comparison between No 16 68.7 11 0.015 the three groups defined by the level of p16ink4a expression Yes 41 92.6 38 showed a significant difference not only for p21waf1,CDK4and Treatment bcl-2, but also for CDK6, which was significantly under VHR 20 70 14 0.015 ink4a Standard and 37 94.6 35 expressed when p16 was present at a low level (Figure 4, other Table 4). In univariate analysis, p21waf1,CDK6andBcl2 WBC Ͻ10 19 94.7 18 у10 and Ͻ100 23 91.3 21 0.04 у Table 3 Multivariate proportional hazards regression analysis of 100 15 66.6 10 disease-free survival of all patients Ploidy diploid 33 78.8 26 0.085 aneuploid 24 95.8 23 Prognostic factors Odds 95% CI Survival p16ink4a (AU) ratio significance level =0238720 Ͼ0 and р0.210 18 100 18 0.0165 p16ink4 4.99 1.39–17.64 0.0128 Ͼ0.210 16 68.7 11 Treatment 5.75 1.15–28.79 0.0329 CDK4 (AU) Ploidy — — 0.38 р0.460 29 93.1 27 0.058 Corticoid sensitivity — — 0.65 Ͼ0.460 28 78.5 22 WBC — — 0.71 CDK4 — — 0.78 DFS, disease-free survival; VHR, very high risk; WBC, white blood cells; No, number; AU, arbitrary units. 95% CI, confidence interval at 95%. p16ink4a and prognosis of ALL Y Mekki et al 186 Table 4 Clinical and biological data of ALL according to p16ink4a expression.

p16ink4a = 0 p16ink4a low level p16ink4a high level P value

Sex: F/M (N) 13/10 7/12 10/6 NS WBC: (N) Ͻ10 3 10 6 Ͼ10 and Ͻ100 11 8 5 Ͻ0.016 Ͼ100 9 1 5 Immunology: (N) B lineage 10 17 16 Ͻ0.0001 T lineage 13 2 0 Ploidy: D/An 19/4 7/12 7/9 Ͻ0.006 Treatment: (N) Standard 10 4 6 NS VHR 13 12 10 Other 2 None 1 CS: yes/no (N) 14/9 16/2 11/5 NS p21waf1(M, EV) 0.02, 0–0.59 0.12,0–1.3 0.35,0.08–2.7 Ͻ0.00001 bcl-2 (M, EV) 0.31,0.01–1.07 0.55,0.09–1.33 0.66,0.39–0.99 Ͻ0.0004 cdk4 (M, EV) 0.27,0–1.19 0.39,0.01–1.05 0.78,0.15–1.48 Ͻ0.0006 cdk6 (M, EV) 0.53,0.04–1 0.23,0.11–0.88 0.58,0.2–0.97 Ͻ0.003

The cut-off for determining low and high level of expression is the median in cases expressing p16ink4a (0.210 AU). We only report here the proteins (median (M), extreme values (EV)) whose expression showed a signiﬁcant difference between groups (expression is given in arbitrary units). Differences between groups are analyzed by ␹2 test for dichotomic variables and by Kruskall and Wallis test for continu- ous variables. F, female; M, male; WBC, white blood cells; D, diploid; An, Aneuploid; VHR, very high risk; CS, corticoid sensitivity; N, number of cases.

Figure 4 Box plot representation of cell cycle regulatory proteins showing signiﬁcant differences according to p16ink4a expression groups. (1: p16ink4a = 0; 2: p16ink4a Ͼ 0 and below the median value; 3: p16ink4a above the median value).

expression did not have a significant impact on prognosis. On oncogenic activity. On the contrary, numerous data support the the contrary, a high level of CDK4 was associated with a short notion that p16ink4a is a tumor suppressor.30,31 It is then surprising DFS for all 57 cases (P = 0.058) as well as for B lineage ALL that our results show a significant unfavorable course for child- (P Ͻ 0.04) (Figure 5 and Table 2). This prognostic value was not hood ALL presenting an enhanced p16ink4a expression. However, retained in a multivariate Cox model analysis with other known analysis of prognosis according to the presence of homozygous prognostic factors. deletion failed to demonstrate a worse evolution for p16ink4a−/− cases in children.14,15 Furthermore, Volm et al,32 in an immuno- cytochemistry study showed that the absence of p16ink4a Discussion expression seemed to be associated (while not statistically significant) with a prolonged relapse-free disease. The interest of Several proteins involved in cell cycle regulation, such as D-type these results is reinforced by recent findings in breast cancer cyclins27,28 as well as their catalytic sub-unit, CDK429 present an showing an identical poor prognostic value of p16ink4a high p16ink4a and prognosis of ALL Y Mekki et al 187 CDK4 (and of CDK6) while, in the context of senescence, a downregulation of CDK4 is expected.38 This increase in kinases regulating G1 phase may favor the progression through G1 phase and finally cell multiplication. Such a situation may be compared to the elevated p16ink4a expression in SV40-transformed primary mouse fibroblasts approaching their finite lifespan in culture. In these cells there were no apparent differences in the status of cyclin D1–CDK4 complexes irrespective of their age or pRb function. Thus, the observed accumulation of p16ink4a does not occur as a simple consequence of growth arrest associated with senescence, since it is apparent in fibroblasts that escape senescence as a result of T-Ag function.39 Indeed, the sole prolonged cell life did not seem sufficient to explain the implication of p16ink4a as a prognostic factor. Bcl-2 did not show significant statistical prognostic value in this series. A similar result, for bcl-2, was demonstrated recently on a large Figure 5 Disease-free survival curves in the 57 ALL patients strati- series of childhood ALLs.40 Other explanations than a simple fied according to CDK4 expression; numerical values are given in prolonged cell life could be proposed to explain our results. A Table 3; (a) CDK4 below; (b) CDK4 above the median value. high expression of p21waf1 leads to a decrease in chemotherapy sensitivity41 and this effect might be raised by an increase in p16ink4a expression.42Furthermore, a concomitant high expression.33 We present a differential analysis according to the expression of the oncogenic protein, CDK4, and of CDK6 might p16ink4a mRNA and protein expression, and both approaches amplify the leukemic advantage by sequestering p16ink4a or other gave concordant results: the shortest DFS was obtained for the CKIs, and thus favoring the progression through G1 phase. We enhanced p16ink4a expression group, which contrasts with the have shown the negative influence of a high CDK4 expression good evolution of patients expressing the CKI at low levels. This on event-free survival, and Volm et al,32 with another method- finding did not imply an oncogenic role for p16ink4a, but the ology, had the same result. Therefore, the bad prognosis of an increase in protein expression might reflect the convergence of enhanced expression of p16ink4a might be related both to a pro- several unfavorable metabolic conditions. longed leukemic cell lifespan and to a growth advantage. In order to better understand this metabolic context, we stud- Nevertheless, we can point to the very good prognosis asso- ied, besides p16ink4a, cell cycle variables (S and S+G2+M) ciated with the expression of p16ink4a at a low level. In these through the analysis of DNA content, the expression of other ALLs, we also found a low and moderate expression of CDK6 CKIs (p21waf1 and p27kip1), of CDKs (CDK1, CDK2, CDK4, and CDK4, respectively. This metabolic context may favor the CDK6), of cyclins (cyclin D3, E, A), and of cell cycle-related p16ink4a tumor suppressor capability. Conversely, we found that proteins (pRb, p53 and bcl-2). No correlation was found the absence of p16ink4a expression is associated with a signifi- between the percentage of cells in S or S+G2+M and p16ink4a cantly more frequent diploidy and hyperleukocytosis (two factors expression. As previously shown in another study,17 CDK1 and of poor prognosis) and a medium course of the disease. So far, CDK6 are significantly more expressed in T cell ALLs. On the no association has been found between the presence of homo- contrary, p16ink4a is significantly higher in B lineage than in T zygous deletion of Ink4a gene and prognosis in childhood cell ALLs, reaching the highest values in the least differentiated ALLs.14,15 This result may be explained by the prognostic hetero- (early early pre-B) cases. p16ink4a mRNA and protein are not geneity of the group with Ink4a expression (which includes all expressed in most T cell ALLs, as expected, because of the cases with low and enhanced p16ink4a expression): both groups, known frequency of homozygous deletion in this form of ALL. with or without homozygous deletion, may present then, an Among all proteins studied, only p21waf1, bcl-2 and CDK4 were intermediary disease-free survival. This interpretation is sup- statistically correlated to p16ink4a expression and were signifi- ported by the poor prognostic value of Ink4a homozygous cantly higher in B lineage than in T cell ALLs. deletion observed in a subgroup of standard risk adult ALLs Bcl-2, p21waf1 and p16ink4a are implicated in prolonged cell where very aggressive ALLs were excluded.43 survival and in senescence. Overexpression of the anti-apoptotic The three biological situations involving p16ink4a absent or bcl-2 oncoprotein has been shown to be associated with a pro- expressed at low or high levels, may be taken into consideration longed survival of B lineage ALL cells and may contribute to when attempting to understand the strong prognostic value of leukemogenesis by prolonging the lifespan of progenitor cells this protein expression in a multivariate survival analysis. We unable to differentiate further.34 p21waf1 also accumulates in hypothesize that the differential expressions of p16ink4a may near-senescent cells,35 and it was originally identified via this reflect the different mechanisms involved in the neoplastic pro- phenomenon.36 This up-regulation of p21waf1 by senescence is cess, ie deregulation of cell multiplication and prolonged blast independent from p53.36 Similarly, p16ink4a mRNA and cellular lifespan, and perhaps may participate in them. protein levels gradually rise in near-senescent and senescent human diploid fibroblasts. An apparent overexpression of p16 in pRb-negative cell lines is caused by the loss of repression by pRb but also by an increase in the number of cell population Acknowledgements doubling.7 This CKI was shown to be complexed, in senescent human fibroblasts, to both CDK4 and CDK6, inhibiting their kin- This work was supported in part by the Hospices Civils de Lyon asic activity on pRb and taking part in the G1 accumulation and the Centre National de Recherche Scientifique, the ‘Ligues of aging cells.37 These data suggest that ALLs with enhanced contre le Cancer du Rhoˆne, de Saoˆne et Loire, de la Droˆme, de expression of p16ink4a show a near-senescent phenotype. How- l’Arde`che’, and the ‘Association pour la Recherche sur le ever, we also observed, in these cases, an increased level of Cancer’. p16ink4a and prognosis of ALL Y Mekki et al 188 References of protein transferred to nitocellulose after separation by SDS-Page. Anal Biochem 1986; 156:341–347. 1 Matsushime H, Ewen ME, Strom DK, Kato JY, Hanks SK, Roussel MF, 21 Wong H, Anderson WD, Cheng T, Riabowol KT. Monitoring mRNA Sherr CJ. Identification and properties of an atypical catalytic subunit expression by polymerase chain reaction: the ‘primer-dropping’ (p34aPSK-13/cdk4) for mammalian D type G1 cyclins. Cell 1992; 71: method. Anal Biochem 1994; 223: 251–258. 323–334. 22 Yamada H, Ochi K, Nakada S, Takahara S, Nemoto T, Sekikawa T, 2 Meyerson M, Harlow E. Identification of G1 kinase activity for cdk6, Horiguchi Y. Interferon modulates the messenger RNA of G1-con- a novel cyclin D partner. MolCellBiol1994; 14: 2077–2086. trolling genes to suppress the G1-to-S transition in Daudi cells. Mol 3 Sherr CJ, Roberts J. Inhibitors of mammalian G1 cyclin-dependent Cell Biochem 1995; 152: 149–158. kinases. Genes Develop 1995; 9: 1149–1163. 23 Wong H, Riabowol K. Differential CDK-inhibitor gene expression in 4 Serrano M, Hannon GJ, Beach D. A new regulatory motif in cell- aging human diploid fibroblasts. Exp Gerontol 1996; 31:311–325. cycle control causing specific inhibition of cyclin D/CDK4. Nature 24 Li Y, Nichols MA, Shay JW, Xiong Y. Transcriptional repression of the 1993; 366: 704–707. D-type cyclin-dependent kinase inhibitor p16 by the retinoblastoma 5 Medema RH, Herrera R, Lam F, Weinberg RA. Growth suppression susceptibility gene product pRb. Cancer Res 1994;54: 6078–6082. ink4a by p16 requires functional retinoblastoma protein. Proc Natl 25 Wolowiec D, Deviller P, Simonin D, Souchier C, Rimokh R, Ben- Acad Sci USA 1995; 92: 6289–6293. chaib M, Bryon PA, Ffrench M. Cdk1 is a marker of proliferation in 6 Parry D, Bates S, Mann DJ, Peters G. Lack of cyclin D–Cdk com- ink4a/MTS1 human lymphoid cells. Int J Cancer 1995; 61: 381–388. plexes in Rb-negative cells correlates with high levels of p16 26 Paterlini P, Suberville AM, Zindy F, Melle J, Sonnier M, Marie JP, tumor suppressor gene product. EMBO J 1995; 14: 503–511. Dreyfus F, Brećhot C. Cyclin A expression in human hematological 7 Hara E, Smith R, Parry D, Tahara H, Stone S, Peters G. Regulation CDKN2 malignancies: a new marker of cell proliferation. Cancer Res 1993; of p16 expression and its implication for cell immortalization 53: 235–238. and senescence. MolCellBiol1996; 16: 859–867. 27 Wang TC, Cardiff RD, Zukerberg L, Lee E, Arnold A, Schmidt EV. 8 Nobori T, Miura K, Wu DJ, Lois A, Takabayashi K, Carson DA. Mammary hyperplasia and carcinoma in MMTV-cyclin D1 trans- Deletions of the cyclin-dependent kinase-4 inhibitor gene in multiple genic mice. Nature 1994; 369: 669–671. human cancers. Nature 1994; 368: 753–756. 28 Hinds PW, Dowdy SF, Eaton EN, Arnold A, Weinberg RA. Function 9 Stranks G, Height SE, Mitchell P, Jadayel D, Yuille MAR, De Lord of a human cyclin gene as an oncogene. Proc Natl Acad Sci USA C, Clutterbuck RD. Treleaven JG, Powles RL, Nackheva E, Oscier 1994; 91: 709–713. DG, Karpas A, Lenoir GM, Smith SD, Millar JL, Catovsky D, Dyer 29 Haas K, Staller P, Geisen C, Bartek J, Eilers M, Mo¨roÿ T. Mutual MJS. Deletions and rearrangement of CDKN2 in lymphoid malig- requirement of CDK4 and Myc in malignant transformation: evidence nancy. Blood 1995; 85: 893–901. for cyclin D1/CDK4 and p16ink4a as upstream regulators of Myc. 10 Hebert J, Cayuela JM, Berkeley J, Sigaux F. Candidate tumor-sup- Oncogene 1997; 15:179–192. pressor genes MTS1 (p16ink4a) and MTS2 (p15ink4b) display frequent 30 Serrano M, Gomez-Lahoz E, DePinho RA, Beach D, Bar-Sagi D. Inhi- homozygous deletions in primary cells from T- but not from B-cell bition of ras-induced proliferation and cellular transformation by lineage acute lymphoblastic leukemias. Blood 1994; 84: 4038–4044. p16ink4a. Science 1995; 267: 249–252. 11 Quesnel B, Preudhomme C, Philippe N, Vanrumbeke M, Dervite I, 31 Serrano M, Lee H, Chin L, Cordon-Cardo C, Beach D, DePinho RA. Lai JL, Bauters F, Wattel E, Fenaux P. P16 gene homozygous deletions in acute lymphoblastic leukemia. Blood 1995; 85: 657–663. Role of th INK4a locus in tumor suppression and cell mortality. Cell 12 Okuda T, Shurtleff SA, Valentine MB, Raimondi S, Head DR, Behm 1996; 85: 27–37. F, Curcio-Brint AM, Liu Q, Pui CH,Sherr CJ, Beach D, Look AT, 32 Volm M, Kooma¨gi R, Stammler G, Rittgen W, Zintl F, Sauerbrey A. ink4a ink4b Prognostic implications of cyclin-dependent kinases (CDK2,CDK4) Downing JR. Frequent deletion of p16 /MTS1 and p15 /MTS2 ink4a in pediatric acute lymphoblastic leukemia. Blood 1995; 85: 2321– and tumor-suppressor genes (pRb, p16 ) in childhood acute lym- 2330. phoblastic leukemia. Int J Cancer 1997; 74: 508–512. 13 Fizzotti M, Cimino G, Pisegna S, Alimena G, Quartarone C, Mandelli 33 Dublin EA, Patel NK, Gillett CE, Smith P, Peters G, Barnes DM. Reti- F, Pelicci PG, Lo Coco F. Detection of homozygous deletions of the noblastoma and p16 proteins in mammary carcinoma: their relation- cyclin-dependent kinase 4 inhibitor (p16) gene in acute lymphoblas- ship to cyclin D1 and histopathological parameters. IntJCancer tic leukemia and association with adverse prognostic features. Blood 1998; 79: 71–75. 1995; 85: 2685–2690. 34 Campana D, Coustan-Smith E, Manabe A, Buschle M, Raimondi SC, 14 Takeuchi S, Bartram CR, Seriu T, Miller CW, Tobler A, Janssen JWG, Behm FG, Ashmun R, Arico M, Biondi A, Pui CH. Prolonged survival Reiter A, Ludwig WD, Zimmermann M, Schwaller J, Lee E, Miyoshi of B-lineage acute lymphoblastic leukemia cells is accompanied by I, Koeffler HP. Analysis of a family of cyclin-dependent kinase inhibi- overexpression of bcl-2 protein. Blood 1993; 81: 1025–1031. tors: p15/MTS2/INK4b,p16/MTS1/INK4a, and p18 genes in acute 35 Tahara H, Sato E, Noda A, Ide T. Increase in expression level of sdi1/cip1/waf1 lymphoblastic leukemia of childhood. Blood 1995; 86:755–760. p21 with increasing division age in both normal and SV40- 15 Rubnitz JE, Behm FG, Pui CH, Evans WE, Relling MV, Raimondi transformed human fibroblasts. Oncogene 1995;10:835–840. SC, Harrison PL, Sandlund JT, Ribeiro RC, Grosveld G, Downing 36 Noda A, Ning Y, Venable SF, Pereira-Smith OM, Smith JR. Cloning JR. Genetic studies of childhood acute lymphoblastic leukemia with of senescent cell-derived inhibitors of DNA synthesis using an emphasis on p16, MLL,andETV6 gene abnormalities: results of St expression screen. Exp Cell Res 1994;211:90–98. Jude Total Therapy Study XII. Leukemia 1997; 11: 1201–1206. 37 Alcorta DA, Xiong Y, Phelps D, Hannon G, Beach D, Barrett JC. 16 Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DAG, Gral- Involvement of the cyclin-dependent kinase inhibitor p16(INK4a) in nick HR, Sultan C. The French–American–British (FAB) co-operative replicative senescence of normal human fibroblasts. Proc Natl Acad group. The morphological classification of acute lymphoblastic leu- Sci USA 1996; 93: 13742–13747. kaemia: concordance among observers and clinical correlations. Br 38 Lucibello FC, Sewing A, Brusselbach S, Burger C, Muller R. Deregu- JHaematol1981; 47: 553–561. lation of cyclins D1 and E and suppression of CDK2 and CDK4 in 17 Wolowiec D, Mekki Y, Ffrench P, Manel AM, Bertrand Y, Rimokh senescent human fibroblasts. J Cell Sci 1993; 105:123–133. R, Philippe N, Bryon PA, Ffrench M. Differential expression of cell 39 Palmero I, McConnell B, Parry D, Brookes S, Hara E, Bates S, Jat P, proliferation regulatory proteins in B- and T- lineage acute lymphobl- Peters G. Accumulation of p16ink4a in mouse fibroblasts as a function astic leukaemias. Br J Haematol 1996; 95: 518–523. of replicative senescence and not of retinoblastoma gene status. 18 Boilletot A, Behar C, Benoit Y, Casteels-VanDeale M, Chantraine JM, Oncogene 1997; 15:495–503. Delbeke MJ, Gyselinck G, Hainaut H, Lutz P, Maurus R, Otten J, 40 Uckun FM, Yang Z, Sather H, Steinherz P, Nachman J, Bostrom B, Philippe N, Plouvier E, Robert A, Solbu G, Souillet G, Suciu S. Treat- Crotty L, Sarquis M, Ek O, Zeren T, Tubergen D, Reaman G, Gaynon ment of acute lymphoblastic leukaemia in children with the BFM P. Cellular expression of antiapoptotic bcl-2 oncoprotein in newly protocol. A cooperative pilot study. Am J Pediatr Haematol Oncol diagnosed childhood acute lymphoblastic leukemia: a Children’s 1987; 9: 317–323. Cancer Group study. Blood 1997; 89: 3769–3777. 19 Ffrench M, Bryon PA, Fie`re D, Vu van H, Gentilhomme O, Adeleine 41 Fan S, Chang JK, Smith ML, Duba D, Fornace AJ, O’Connor PM. P, Viala JJ. Cell-cycle, protein content and nuclear size in acute Cells lacking Cip1/Waf1 genes exhibit preferential sensitivity to cis- myeloid leukemia. Cytometry 1985; 6:47–53. platin and nitrogen mustard. Oncogene 1997; 14: 2127–2136. 20 Salinovich O, Montelaro R. Reversible staining and peptide mapping 42 Stone S, Dayananth P, Kamb A. Reversible, p16-mediated cell cycle p16ink4a and prognosis of ALL Y Mekki et al 189 arrest as protection from chemotherapy. Cancer Res 1996; 56: Charrin C, Archimbaud E, Magaud JP, Rimokh R. Molecular analysis 3199–3202. of cyclin-dependent kinase inhibitors in human leukemias. Leukemia 43 Hayette S, Thomas X, Bertrand Y, Tigaud I, Callanan M, Thiebaut A, 1997; 11: 1696–1699.