Characterization of Stage Progression in Chronic Myeloid Leukemia by DNA Microarray with Puri®Ed Hematopoietic Stem Cells

Home , CP1 (classification), CP4 (classification), CP5 (classification), CP6 (classification), CP7 (classification)

Characterization of stage progression in chronic myeloid leukemia by DNA microarray with puri®ed hematopoietic stem cells

Ken Ohmine1,2, Jun Ota1, Masuzu Ueda1,2, Shu-ichi Ueno1,3, Koji Yoshida1, Yoshihiro Yamashita1, Keita Kirito2, Shigehiko Imagawa4, Yuichi Nakamura5, Kenji Saito5, Miyuki Akutsu6, Kinuko Mitani5, Yasuhiko Kano6, Norio Komatsu2, Keiya Ozawa2 and Hiroyuki Mano*,1

1Division of Functional Genomics, Jichi Medical School, Kawachi-gun, Tochigi 329-0498, Japan; 2Division of Hematology, Jichi Medical School, Kawachi-gun, Tochigi 329-0498, Japan; 3Division of Cardiology, Jichi Medical School, Kawachi-gun, Tochigi 329-0498, Japan; 4Department of Hematology, University of Tsukuba School of Medicine, Tsukuba, Ibaraki 305-8575, Japan; 5Department of Hematology, Dokkyo University School of Medicine, Mibu, Tochigi 321-0293, Japan; 6Tochigi Cancer Center, Utsunomiya, Tochigi 320-0834, Japan

Chronic myeloid leukemia (CML) is characterized by the Introduction clonal expansion of hematopoietic stem cells (HSCs). Without eective treatment, individuals in the indolent, The expression of the p210BCR-ABL fusion protein in chronic phase (CP) of CML undergo blast crisis (BC), hematopoietic stem cells (HSCs) plays an essential role the prognosis for which is poor. It is therefore important in the pathogenesis of chronic myeloid leukemia to clarify the mechanism underlying stage progression in (CML) (Era and Witte, 2000), resulting in the CML. DNA microarray is a versatile tool for such a expansion of these cells without impairment of their purpose. However, simple comparison of bone marrow ability to dierentiate (Kantarjian et al., 1993). The mononuclear cells from individuals at dierent disease numbers of all types of blood cells are thus increased in stages is likely to result in the identi®cation of pseudo- both peripheral blood and bone marrow (BM) of positive genes whose change in expression only re¯ects individuals with CML during the so-called `chronic the dierent proportions of leukemic blasts in bone phase (CP)' of the disease. Treatment of aected marrow. We have therefore compared with DNA individuals with interferon (IFN)-a or hydroxyurea microarray the expression pro®les of 3456 genes in the (HU), or both, reduces the number of blood cells but it puri®ed HSC-like fractions that had been isolated from is still dicult that such treatment totally eradicates the 13 CML patients and healthy volunteers. Interestingly, BCR ± ABL ± expressing cells (Silver et al., 1999). The expression of the gene for PIASy, a potential inhibitor of eventual loss of the ability of the malignant clones of STAT (signal transducer and activator of transcription) CML patients to dierentiate results in progression to proteins, was down-regulated in association with stage the terminal `blast crisis (BC)' stage of the disease, progression in CML. Furthermore, forced expression of which resembles de novo acute myeloid leukemia PIASy has induced apoptosis in a CML cell line. These (AML). Some patients exhibit an intermediate `accel- data suggest that microarray analysis with background- erated phase (AP)' between the CP and BC stage of matched samples is an ecient approach to identify CML. Transformation to the BC stage renders CML molecular events underlying the stage progression in refractory to treatment including allogeneic BM CML. Oncogene (2001) 20, 8249 ± 8257. transplantation. The prognosis of patients in BC is thus poor, with a median survival time of only 3 ± 4 Keywords: chronic myeloid leukemia; DNA micro- months. Improvement in the overall prognosis for array; AC133; PIASy individuals with CML therefore requires both clari®cation of the mechanism of stage progression as well as the ability to predict the probability that a given patient will undergo transformation to BC at a given time. DNA microarray analysis allows characterization of the expression pro®les of thousands of genes at a time. It should therefore be a suitable approach with which to identify genes that might prove informative for stage diagnosis of CML or that contribute to stage *Correspondence: H Mano, Division of Functional Genomics, Jichi progression (Duggan et al., 1999). However, a simple Medical School, 3311-1 Yakushiji, Kawachi-gun, Tochigi 329-0498, comparison of BM mononuclear cells (MNCs) among Japan; E-mail: [email protected] Received 18 June 2001; revised 28 September 2001; accepted 9 dierent patients often produces a large number of October 2001 pseudo-positive results, given that the proportion of DNA microarray analysis of CML K Ohmine et al 8250 leukemic blasts in BM varies markedly among such expression analysis. BM MNCs of every patient were individuals. If, for example, total MNCs from the BM proved to carry either Ph1 chromosome (judged by of patients containing 5 or 90% identical leukemic karyotype analysis) or the fusion transcript encoding blasts are compared, DNA microarray analysis would p210BCR-ABL (judged by reverse-transcription (RT ± mistakenly suggest that the two individuals suer from PCR)). The healthy control sample from the Blast distinct disorders. This `population shift' eect is one Bank (a mixture of specimens from healthy volunteers) of the main causes of pseudo-positive results in high- was also included in the study. throughput screening. Eective elimination of such We ®rst tested whether puri®ed AC133+ cells still pseudo-positive data requires isolation of the leukemic contain the transformed CML clones. Fluorescence in blasts and comparison of the transcriptomes of these situ hybridization (FISH) analysis detected the BCR- puri®ed cells. The character or background of the ABL fusion gene in 76.0 and 63.9% of BM MNCs and samples subjected to microarray analysis should be AC133+ cells, respectively, prepared from the same matched as closely as possible, so that they dier only patient. Karyotype analysis of this individual exhibited in the parameter under investigation, such as whether Ph1 chromosome in 20 out of 20 BM MNCs examined. the patient has undergone transformation to BC or In another patient, 70.7% of AC133+ cells were positive not. We have proposed that such a microarray for the presence of the fusion gene. Considering the approach with highly puri®ed specimens be termed moderate detection eciency of FISH analysis, these background-matched population (BAMP) screening data proved that the Blast Bank samples from CML (Miyazato et al., 2001). patients are mainly composed of malignant CML clones. HSCs or pluripotent immature cells that are early Biotin-labeled cRNA was then synthesized from each descendents of HSCs are the blood cells that are sample and subjected to hybridization with oligonu- aected in many types of leukemia, including CML, cleotide microarrays representing 3456 human genes. AML and myelodysplastic syndrome (MDS). Thus, the Through our microarray analyses, we had noticed that HSC-like fraction would be expected to be a suitable neither b-actin nor glyceraldehyde 3-phosphate dehy- target for BAMP screening of leukemic disorders. drogenase gene was a reliable internal control, at least, Puri®cation and analysis of the HSC-like fraction when used a single control gene. The digitized data should allow direct evaluation and comparison of the were, therefore, normalized on the basis of the average characteristics of leukemic blasts irrespective of the spot intensity for each hybridization. Hierarchical proportion of blasts in BM. Given that AC133 is one clustering analysis of the data resulted in generation of the most speci®c cell surface markers available for of a dendrogram, or `gene tree', in which genes with HSCs (Hin et al., 1997), we have puri®ed AC133- similar expression pro®les are positioned near each positive cell fractions from individuals with a wide other (Figure 1a). The expression level of each gene range of hematologic disorders and have stored these was assigned a pseudocolor, with high expression fractions for our large-scale genomics project, termed represented by red and low expression by green. The `Blast Bank'. Furthermore, DNA microarray analysis gene tree revealed that between *25 and 33% of the with the Blast Bank samples has resulted in the genes studied were transcriptionally active in the identi®cation of molecular markers that dierentiate puri®ed blasts of individuals with CML. Several MDS from de novo AML (Miyazato et al., 2001). clusters of genes were active only in the blasts With the use of Blast Bank samples puri®ed from corresponding to a speci®c disease stage or to individuals with CML, we have now identi®ed sets of subgroups of patients at a speci®c stage. genes whose expression is dependent on disease stage. Identi®cation of stage-speci®c genes is potentially of Transcriptome pro®ling of the Blast Bank samples also clinical importance, given that it should facilitate stage revealed that the blasts of one CP patient exhibited a determination and the design of an optimal therapeutic transcriptome similar to that of blasts from BC strategy for each patient. To identify such genes, we patients. Consistent with this observation, control of calculated the average expression intensity of each gene the white blood cell (WBC) count in the peripheral for the CP, AP, and BC groups and used these data to blood of this CP patient was dicult, despite sustained generate another dendrogram, `average tree' (Figure treatment with HU. 1b). This average tree has revealed the presence of clusters of genes that are expressed speci®cally in either healthy control, CP, AP, or BC specimens. The genes in these clusters are thus candidates for the CML Results and discussion stage-speci®c molecular markers. Transcriptomes of CML blasts Genes specific to CP Our Blast Bank is rapidly expanding and currently comprises 192 specimens of AC133+ blasts, including From the average tree shown in Figure 1b, we selected 24 samples derived from individuals with CML. At the several groups of genes whose expression pro®les were timing of microarray experiment, we had a total of 13 suggestive of biological relevance. The expression of a (CML samples (derived from seven patients in CP, two total of 39 genes (at least in terms of mean expression patients in AP, and four patients in BC (all in myeloid intensity) was speci®c to CP. The detailed expression crisis)) in our Bank, all of which were subjected to pro®les of each of these genes are shown in Figure 2a.

Oncogene DNA microarray analysis of CML K Ohmine et al 8251 2 expression were treated with IFN-a after sampling of AC133+ cells; two of them (CP6 and CP7) responded well to the treatment, resulting in good control of peripheral blood cell number, whereas patient CP5 underwent BC transformation within a few months of treatment onset. The genes for type 1 and type 2 receptors for IFN-a, IFN-b, and IFN-o were not signi®cantly expressed in any of the samples analysed in the present study. Expression of the genes for IFN-g receptor 2 (IFNGR2) and for IFN-g, however, was increased in both patients CP6 and CP7 (Figure 2a). The biological relevance of the overexpression of the genes for the components of this IFN-g autocrine loop remains to be determined. In addition to IFNGR2, several other genes, including those for LAGE-1 (Lethe et al., 1998) and B melanoma antigen (BAGE) (Boel et al., 1995), were selectively expressed in CP patients 6 and 7 (Figure 2a). Both LAGE-1 and BAGE were identi®ed as a result of screening for melanoma-speci®c antigens that are Figure 1 Transcriptomes of CML blasts. (a) Hierarchical recognized by cytotoxic T cells. They are known to clustering of 3456 genes based on their expression pro®les in be expressed in solid tumors but not, with the Blast Bank samples derived from 13 individuals with CML and in a pooled healthy control sample (CTRL). Each column represents exception of the testis, in normal tissues. Expression a single gene on the microarray, and each row a separate patient of these two genes in hematopoietic lineages has not sample. The ¯uorescence intensity of every gene was normalized previously been demonstrated. As with IFIT-2, the relative to the mean ¯uorescence value of all spots in each expression of LAGE-1 and BAGE did not appear to hybridization, and the normalized value is shown color-coded as indicated on the left. The position of a cluster of CP-speci®c genes correlate with the administration of IFN-a. It remains is indicated by the arrow. Gray indicates blank spots. (b) Mean unknown whether the loss of LAGE-1 or BAGE expression values for each gene were calculated for CP, AP, and expression contributes to the escape from immuno- BC stages, and used to generate a dendrogram. The presence of logical surveillance of blast cells in CML and to the clusters of genes whose expression is speci®c to healthy control, transformation from CP to AP or BC stages. CP, AP, or BC samples is evident

Genes up-regulated in AP or BC The gene showing the highest level of CP-speci®c The activated BCR ± ABL tyrosine kinase plays an expression was that for the interferon-induced protein important role in the pathogenesis of CP in CML. with tetratricopeptide repeats-2 (IFIT-2). The abun- However, the molecular events that trigger the dance of IFIT-2 transcripts was especially high in ®ve transformation to AP or BC stages remain unclear. out of the seven samples of CP blasts (CP3 to CP7), It is possible that both transcriptional activation and but was low or undetectable in the healthy control transcriptional repression contribute to this transfor- sample and samples corresponding to the other stages mation process. From the average tree shown in of CML. Figure 1b, we identi®ed 17 genes that appeared not IFIT-2 was originally identi®ed by dierential to be expressed in the AC133+ blasts of healthy screening in an attempt to isolate IFN-inducible genes control and CP individuals but were transcriptionally (Wathelet et al., 19888). Although most individuals in active in those of AP or BC patients. The expression CP of CML are treated with IFN, the increased pro®les of these genes, including those for DNA expression of IFIT-2 in the CP blast samples was not damage ± binding protein 1 (DDB1) (Chu and attributable to an eect of IFN treatment. Four (CP4 Chang, 1988) and NADH-ubiquinone oxidoreductase to CP7) of the ®ve patients exhibiting high expression 1b subcomplex 8 (NDUFB8) (Loeen et al., 1998), of IFIT-2 were diagnosed with CML on the basis of are shown in Figure 2b. the BM aspirates that provided the samples for this DDB1 functions in the repair of DNA damage, and study, and therefore had not been previously treated loss of its activity is responsible for pathogenesis of for CML. The patient CP3 had once been treated with complementation group E of xeroderma pigmentosum. IFN-a, but the administration of this cytokine was A reduction in the stability of DNA as well as various terminated as a result of progression of mental types of chromosomal anomaly are often apparent in depression, a severe side eect of IFN-a treatment, blasts from individuals in the advanced stages of CML. before the analysed blast sample was obtained. The induction of DDB1 expression now detected in the It remains to be determined whether the expression blasts of AP and BC patients may therefore re¯ect an of IFIT-2 is related to the blast sensitivity to IFN-a. accumulation of DNA damage in these cells. NDUFB8 Three (CP5 to CP7) of the ®ve patients with high IFIT- is a component of a large enzyme complex that

Oncogene DNA microarray analysis of CML K Ohmine et al 8252 Genes down-regulated in AP or BC We next identi®ed nine genes whose expression was reduced in the AC133+ cells of AP or BC patients compared with that in AC133+ cells of healthy individuals or CP patients (Figure 2c). One such gene was that for the protein inhibitor of activated STAT (signal transducer and activator of transcription) y (PIASy), which belongs to the PIAS family of proteins that also includes PIAS1, PIAS3, PIASxa, and PIASxb (Shuai, 2000). PIAS1 and PIAS3 bind to and inhibit the activity of STAT1 and STAT3, respectively. Recently, Liu et al. (2001) have revealed that PIASy is also a transcriptional repressor of STAT1. STAT proteins are often hyper-phosphorylated and activated in aected cells from individuals with a variety of leukemias (Gouilleux-Gruart et al., 1997). Introduction of the BCR ± ABL tyrosine kinase into hematopoietic cell lines results in the activation of STAT proteins (Carlesso et al., 1996), indicating that these proteins mediate BCR ± ABL signaling. PIASy may therefore counteract the transforming potential of activated STAT proteins in the leukemic blasts of CP patients, and the subsequent disappearance of this protein may result in deregulation of cell growth and stage progression. Another interesting gene in this group may be that for death-associated protein kinase 3 (DAPK3), or ZIP kinase (Kawai et al., 1998). DAPK3 homo-dimerizes through its leucine-zipper domain, and subsequent activation of DAPK3 induces apoptotic cell death. Loss of DAPK3 expression may therefore renders the leukemic blasts resistant to physiological apoptotic stimuli, and this may be an important step for the further transformation into AP or BC stages.

Clustering of patients according to gene expression profiles The dendrograms of the expression of all genes analysed Figure 2 Identi®cation of CML stage-speci®c genes. (a) Expres- (Figure 1) as well as the expression patterns of CP- sion pro®les of CP-speci®c genes. Each curve corresponds to a single gene and is colored according to the expression level of the speci®c genes (Figure 2a) both demonstrated that a set gene in the healthy control blasts. Curves corresponding to the of genes was transcriptionally active only in CP patients genes for IFIT-2, IFNGR-2, LAGE-1, and BAGE are indicated 6 and 7 (Figure 1a, arrow), suggesting the existence of by arrows. (b) Pro®les of genes whose mean expression level is subgroups within CP patients. To evaluate statistically speci®cally increased in AP or BC blasts. Note that the expression the heterogeneity among CP patients, we performed of genes for DDB1 and NDUFB8 is also increased in patient CP1. (c) Pro®les of genes whose expression level is reduced in AP two-way clustering analysis (Alon et al., 1999) on the or BC blasts relative to that in healthy control or CP blasts. The gene tree shown in Figure 1a. Such analysis generates a pro®les of the genes for PIASy and DAPK3 are indicated. Gene dendrogram that groups patients according to the names and accession numbers as well as expression intensity data similarity of their gene expression pro®les (Figure 3). in the ®gures are available upon request The patients CP6 and CP7 were indeed clustered in the same branch of the dendrogram. Patients CP4 and CP5 were also grouped in the same branch. Two additional contributes to electron transport in mitochondria. clusters of patients were of interest: one comprising two Induction of NDUFB8 expression and a consequent AP patients (#1 and #2) and one BC patient (#1), and increase in the rate of the mitochondrial respiratory one containing one BC patient (#3) and CP patient 1. reaction in transformed blasts may re¯ect an increased Indeed, transcription of the AP- or BC-speci®c genes catabolic activity in these cells. Both DDB1 and was increased in patient CP1, whereas none of the CP- NDUFB8 were also expressed in the patient CP1, speci®c genes was expressed at a substantial level in this raising the possibility that this individual may have patient (Figure 2a,b). The AC133+ blasts of this been in the process of transformation to the AP or BC individual thus exhibit a transcriptome most similar to stage. that of BC blasts.

Oncogene DNA microarray analysis of CML K Ohmine et al 8253 What is the biological relevance of this classi®ca- Given the possible involvement of PIASy in the tion of CML patients based on the results of STAT-mediated signalings, we next tried to clarify the microarray analysis? Patient CP1 was admitted to biological signi®cance of the change in PIASy expres- hospital with markedly increased WBC (240 300 cells/ sion level during the course of CML progression. If ml) and platelet (1 751 000 cells/ml) counts in addition PIASy protein has a growth suppressor activity, then to anemia (hemoglobin concentration, 7.7 g/dl). the decrease of its expression may allow a deregulated Treatment with HU resulted in a decrease in WBC growth of leukemic blasts in the advanced stages of and platelet counts (7300 and 712 000 cells/ml, respec- CML. If this is the case, forced expression of PIASy in tively) without the correction of anemia. However, the CML cells from the BC stage would induce the initial attempt to reduce the dosage of HU (from apoptosis of, or restore the dierentiation ability of, 1000 to 500 mg/day) resulted in a rapid rebound in the CML blasts. We thus examined the eect of PIASy WBC count (from 7300 to 150 200 cells/ml). The expression in a human cell line, KCL22, that is derived second attempt at dosage reduction (from 1500 to from a patient in BC of CML. 1000 mg/day) again failed to maintain the WBC When characterizing cell clones stably expressing an count within the normal range; the count increased exogenous gene, it is often dicult to determine whether from 9160 to 21 030 cells/ml. This clinical history the phenotype of the transformants arises from the reveals the highly proliferative nature of the leukemic expression of the incorporated gene or from clonal blasts of this patient, which is an unusual character- variations within transfected cells. To overcome such istic for the CP stage of CML. The similarity of the ambiguity, here we have constructed a novel retroviral transcriptome of the blasts from patient CP1 to that plasmid (pMX-tetOFF), by which a gene of interest can of blasts from BC patients suggests that this be conditionally expressed in polyclonal cell population individual may undergo transformation into AP or through a single round of retroviral infection. The BC stage in a relatively short period. group of J-K. Yee has developed a b-estradiol- dependent tetracycline repressor (tTAER) (Iida et al., 1996). Only under the presence of b-estradiol and Construction of a novel retroviral vector for conditional absence of tetracycline, this protein becomes active and expression of an exogenous gene To con®rm the changes in the expression level of the CML stage-speci®c genes, we performed quantitative `real-time' PCR. The abundance of IFIT-2 and PIASy transcripts was determined relative to that of b-actin mRNA in an extended number of Blast Bank samples (one healthy volunteer, nine patients in CP, ®ve patients in AP and seven patients in BC). Consistent with the results obtained with microarray analysis, the expression of IFIT-2 was marked only in the blasts from CP patients, but was below the level of detection in the blasts from other individuals (Figure 4a). Similarly, expression of PIASy was greatly reduced in the blasts of AP and BC patients compared with that in blasts from healthy volunteers and CP patients (Figure 4b). The discrepancy of the PIASy expression levels in the control sample between the array data (Figure 2c) and real-time PCR (Figure 4b) may re¯ect the dierent methods for normalization; array data is normalized on the basis of mean expression value for each hybridization while the data in real-time PCR is on the expression level of b-actin.

Figure 4 Quantitation of IFIT-2 and PIASy transcripts in CML Figure 3 Two-way clustering of the Blast Bank samples. blasts. Complementary DNA prepared from the blasts of CML Hierarchical clustering of the CML patient and healthy control patients and healthy volunteers was subjected to real-time PCR samples was generated on the basis of the clustered gene with primers speci®c for IFIT-2 (a), PIASy (b), or b-actin genes. expression pro®le shown in Figure 1a. Clusters containing CP1 The ratio of the abundance of the target transcripts to that of n and BC3, the AP patients and BC1, and CP patients are shown in b-actin mRNA was calculated as 2 , where n is the Ct value for red, green, and blue, respectively b-actin cDNA minus the Ct value of the target cDNA

Oncogene DNA microarray analysis of CML K Ohmine et al 8254 induces the transcription via the tetracycline repressor- binding sequence (tetO). However, either addition of tetracycline or depletion of b-estradiol results in the blockage of the tetO-mediated transcription. The internal ribosome entry site (ires) elements are known to allow translation of internal cistrons of mRNAs independently from the well-characterized capping-mediated ribosome scanning (Jang et al., 1989). The ires fragment derived from encephalomyo- carditis virus (EMCV) was ligated with the cDNA for blasticidin S-resistant marker protein (BSR) (Izumi et al., 1991). All of the tTAER and ires-BSR fragments and the tetO sequence were inserted into pMX retroviral vector (Onishi et al., 1996) in this order, generating pMX-tetOFF (Figure 5a, top panel). When transduced, the long terminal repeat (LTR)-driven RNA should produce the capping site-mediated translation of tTAER as well as the ires-mediated translation of BSR, conferring the transduced cells the resistance against blasticidin-S. Under the presence of tetracycline or absence of b-estradiol, tTAER cannot drive the transcription from the tetO site (middle panel). Removal of tetracycline and addition of b-estradiol, however, should conditionally induce the expression of an exogenous gene (FLAG-tagged PIASy (PIASy-F), in this case) placed downstream of the tetO site (lower panel). Infection of target cells with the retrovirus generated from pMX-tetOFF containing the PIASy-F cDNA (pMX-tetOFF/PIASy-F) should thus allow blasticidin- S-resistant polyclonal cell population, almost all of which express PIASy-F through a dual-regulation system with tetracycline and b-estradiol. Recombinant retrovirus was generated from pMX-tetOFF or pMX- tetOFF/PIASy-F, and was used to infect KCL22 cells Figure 5 Construction of a dual regulation vector, pMX- which were subsequently cultured under the presence of tetOFF. (a) The cDNA for tTAER, ires-BSR cassette and the blasticidin-S and tetracycline. The drug-resistant mass tetO fragment were inserted in this order into the retrovirus culture of each infection was then transferred to a backbone vector, pMX, to generate pMX-tetOFF. For the medium containing blasticidin-S and b-estradiol, but expression of PIASy-F, the corresponding cDNA was placed downstream to tetO (top panel). Under the presence of without tetracycline. Total cell lysates as well as anti- tetracycline (tet) and absence of b-estradiol (est), tTAER can no FLAG immunoprecipitates were prepared from each longer bind to tetO, and therefore the transcription of PIASy-F is fraction, and were probed with anti-FLAG antibody. suppressed (middle panel). Removal of tetracycline and addition As expected, PIASy-F was readily detected in the cells of b-estradiol, however, induce the dimerization of tTAER and its infected with MX-tetOFF/PIASy-F under the induction binding to the tetO fragment, leading to the activation of transcription for the PIASy-F gene (bottom panel). (b) After condition (Figure 5b). It should be noted, however, that incubation for 24 h under the presence of tetracycline (7)orb- a marginal leaky expression of the protein was estradiol (+), KCL22 cells infected with MX-tetOFF (Mock) or detectable even under the presence of tetracycline. MX-tetOFF/PIASy-F (PIASy) were subjected to solubilization. Total cell lysates (TCL; 10 mg/lane) and the immunoprecipitates with anti-FLAG antibody (aFLAG IP) were then probed with the PIASy as an inducer of apoptosis antibodies to FLAG. The positions of PIASy-F and IgH are indicated by an arrow and an asterisk, respectively, at the right. Does the expression of PIASy aect the growth of The positions of molecular weight standards (610-3) are also KCL22 cells? Surprisingly, as shown in the left panel of shown at the left Figure 6a, growth rate of KCL22 cells was severely retarded only when PIASy-F was induced. In addition, the growth suppression was only marginal in the cells Such rapid suppression of cell growth could be the infected with MX-tetOFF/PIASy-F at a non-induced result of an induction of apoptosis or dierentiation condition. On the contrast, the mock-infected cells into mature blood cells. To clarify the mechanism of grew exponentially irrespective of the culture condi- growth suppression in KCL22, morphology of the cells tions. Concomitant with these growth properties, the maintained at the induction condition for 6 days were viability of KCL22 cells became rapidly impaired in the analysed by Wright-Giemsa staining. Mock-infected PIASy-expressing KCL22 cells, while that of the other cells under the presence of b-estradiol retained their KCL22 culture remained healthy (right panel). parental phenotypes; middle-sized cells with basophilic

Oncogene DNA microarray analysis of CML K Ohmine et al 8255 cytoplasm, high nuclear to cytoplasm ratio, and ®ne highly useful system to identify molecular markers for nuclear structure (Figure 6b, left panel). In contrast, the various stages of CML as well as to provide insight many cells expressing PIASy-F have undergone into the molecular mechanism of transformation. apoptotic changes at day 6; shrinkage of cells, condensation and fragmentation of nuclei (right panel). We could not ®nd in these cells the commitment to dierentiation into any speci®c cell lineages. Materials and methods To con®rm the apoptotic change found in the PIASy-expressing cells, the cells were stained with Preparation of Blast Bank samples Annexin-V (BD Biosciences) and propidium iodide BM aspirates were obtained from subjects with written (PI), and subjected to ¯ow cytometry analysis to detect informed consent. MNCs were puri®ed from the specimens the translocation of phosphatidylserine (PS) from the by Ficoll-Hypaque (Amersham Pharmacia Biotech) density inner part to the outer layer of cell membrane, a gradient centrifugation, labeled with anti-AC133 MicroBeads characteristic feature of apoptosis. As shown in Figure (Miltenyi Biotec, Auburn, CA, USA), and subjected to 6c, forced expression of PIASy-F resulted in an chromatography on a miniMACS magnetic cell separation column (Miltenyi Biotec). The resulting AC133+ cell fractions increase of apoptosis fraction (PS-positive and PI- were divided into portions and stored at 7808C as the Blast negative cells) from 6.3% in the uninduced state to Bank samples. AC133+ cells were also puri®ed from BM 17.0% in the induced state. MNCs of healthy volunteers and mixed for use as a `healthy control' sample. The purity of each sample was con®rmed by Wright-Giemsa staining. When a sucient number of BAMP screening with Blast Bank samples from AC133+ cells was obtained, the purity was also evaluated CML patients by ¯ow cytometric analysis with antibodies to AC133, to Analysis of the Blast Bank samples has provided data CD34, and to CD38 (BD Biosciences, San Jose, CA, USA). of clinical relevance. Previous analysis with MNCs, rather than with AC133+ cells, revealed that CD34 was RNA preparation and DNA microarray analysis one of the genes whose expression was most increased Total RNA was extracted from Blast Bank samples by the in the BC stage (A Miyazato, personal communica- acid guanidinium method and was subjected to ampli®cation tion). However, this result appears to be pseudo- as previously described (Van Gelder et al., 1990). All samples positive, given that analysis of Blast Bank samples were ampli®ed twice, and the correlation coecient of DNA demonstrated almost equal expression of CD34 during microarray data obtained with RNA after the ®rst and the the course of CML (data not shown). The present second rounds of ampli®cation was calculated to be 0.93. One study has identi®ed potential molecular markers for microgram of the ampli®ed cRNA was converted to double- each stage of CML. Although DDB1 was identi®ed as stranded cDNA which was then used to prepare biotin- a marker for AP or BC, it was also highly expressed in labeled cRNA with the use of the ExpressChip labeling patient CP1. If this latter patient undergoes transfor- system (Mergen, San Leandro, CA, USA), and was allowed to hybridize with microarrays (HO-1*3; Mergen) that mation to AP/BC, the expression level of DDB1 (and contain oligonucleotides corresponding to a total of 3456 possibly that of NDUFB8) might prove useful as a human genes (the gene list on the arrays is available at http:// predictor for such transformation. Such predictors may www.mergen-ltd.com/). The microarrays were then incubated facilitate both the design of treatment strategies consecutively with streptavidin, with antibodies to streptavi- targeted to individual CML patients as well as din, and with Cy3-conjugated secondary antibodies (all from determination of the timing of BM transplantation. Mergen). Detection of hybridization signals and statistical More importantly, our present data has proposed a analysis of the digitized data were performed with a GMS novel mechanism for stage progression in CML; the 418 array scanner (Aymetrix, Santa Clara, CA, USA) and change in PIASy expression level. In the AC133+ cells GeneSpring 3.2.2 software (Silicon Genetics, Redwood, CA, of normal individuals and CP patients, the gene for USA), respectively. In the hierarchical clustering analysis, similarity was measured by the standard correlation with a PIASy was highly expressed while it became decreased separation ratio of 0.5. along with the stage progression toward BC. Experi- ments with the KCL22 CML cell line has clearly indicated the growth suppressive nature of PIASy. Since Real-time PCR analysis PIASy has been recently demonstrated to function as a Portions of unampli®ed cDNA were subjected to PCR with STAT1 inhibitor, induction of apoptosis may be SYBR Green PCR Core Reagents (PE Applied Biosystems, attributable to the PIASy regulation of STAT proteins. Foster City, CA, USA). The incorporation of the SYBR However, we could not detect activation/phosphoryla- Green dye into the PCR products was monitored in real time tion of STAT1 in KCL22 cells irrespective of PIASy with an ABI PRISM 7700 sequence detection system (PE expression (data not shown). It may be, therefore, still Applied Biosystems), thereby allowing determination of the threshold cycle (C ) at which exponential ampli®cation of an open question through which mechanism PIASy t PCR products begins. The Ct values for cDNAs correspond- blocks cell growth of a BC-derived cell line. Tyrosine ing to the b-actin gene and target genes were used to phosphorylation of p210BCR-ABL in KCL22 cells was not calculate the abundance of the target transcripts relative to aected by the induction of PIASy (data not shown). that of b-actin mRNA. The oligonucleotide primers for PCR In conclusion, our microarray analysis with the were as follows: b-actin cDNA, 5'-CCATCATGAAGTGT- puri®ed Blast Bank samples was demonstrated to be a GACGTGG-3' and 5'-GTCCGCCTAGAAGCATTTGCG-

Oncogene DNA microarray analysis of CML K Ohmine et al 8256 3'; IFIT-2 cDNA, 5'-CTGCGGTATGGCAACTTTCAG-3' and 5'-AGGAATGCCAAGACATGCAA-3'; and PIASy cDNA, 5'-AACTACGGCAAGAGCTACTCGGTG-3' and 5'-GTTCATCTGCAGGTAGAAGACGGC-3'.

Conditional expression of PIASy Into the pMX retrovirus vector plasmid (Onishi et al., 1996), inserted were the cDNA for tTAER (Iida et al., 1996), ires-BSR (Izumi et al., 1991) and tetO fragments (Iida et al., 1996), giving rise to pMX-tetOFF. The coding sequence of human PIASy cDNA (Helix Institute, Chiba, Japan) was PCR-ampli®ed, fused C-terminally to the oligonucleotides encoding the FLAG epitope tag (Eastman Kodak, New Haven, CT, USA), and placed downstream to the tetO fragment in pMX-tetOFF. The resultant plasmid (pMX-tetOFF/PIASy-F) or pMX-tetOFF was transiently introduced into a packaging cell line, PA317 (American Type Culture Collection, Manassas, VA, USA), to produce amphotropic retrovirus. KCL22 cells (American Type Culture Collection) maintained in RPMI1640 medium (Life Technologies, Rockville, MD, USA) with 10% fetal bovine serum (FBS) were infected with the supernatant of PA317 cells for 24 h under the presence of Retronectin (Takara Shuzo, Shiga, Japan). Cells were then harvested and cultured in RPMI1640 / FBS supplemented with 5 mg/ml blasticidin-S (Funakoshi, Tokyo, Japan) and 1 mg/ml tetracycline (Boehringer Mannheim, Mannheim, Germany). To induce the expression of PIASy- F, the culture medium was changed to RPMI / FBS with 2 mM 17b-estradiol (Sigma, St. Louis, MO, USA). Protein analysis was carried out as described previously (Yamashita et al., 2001). Brie¯y, cells were resuspended into the lysis buer (1% Nonidet P-40, 50 mM Tris-HCl, 7.4, 150 mM NaCl, 1 mM NaF, 1 mM Na3VO4, 200 U/ml aprotinin and 1 mM phenylmethylsulfonyl ¯uoride), and insoluble materials were removed by centrifugation. PIASy- F was immunoprecipitated from 1.5 ± 2 mg of the lysates by anti-FLAG antibody (M2; Eastman Kodak) and protein G- Sepharose beads (Pharmacia LKB, Uppsala, Sweden). For immunoblotting, total cell lysates (10 mg/lane) and the immunoprecipitates were separated through SDS ± PAGE and transferred onto polyvinylidene di¯uoride (PVDF) membranes (Immobilon; Millipore, Bedford, MA, USA). The membranes were incubated for 1 h at room temperature in TBST (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween 20) with 4% bovine serum albumin (Fraction V; Sigma). The membranes were then incubated with the antibodies to FLAG for 1 h at room temperature in TBST. Speci®c bindings of the antibodies were visualized by the ECL detection system (Amersham, Arlington Heights, IL, USA) according to the manufacturer's instruction.

Figure 6 PIASy-F as an apoptosis inducer. (a) KCL22 cells infected with MX-tetOFF (MOCK) or MX-tetOFF/PIASy-F Apoptosis of KCL22 cells (PIASy) were cultured in RPMI1640 / FBS supplemented with tetracycline (7)orb-estradiol (+). Total cell number (left panel) KCL22 cells infected with MX-tetOFF or MX-tetOFF/ and the viability as judged by the trypan-blue dye exclusion PIASy-F were cultured for 6 days under the presence of method (right panel) of each fraction were counted at every other 17b-estradiol. Cytospin preparation of each cell fraction was day, and shown as graphs. (b) KCL22 cells infected with MX- stained with Wright-Giemsa solutions. The cells were also tetOFF (Mock) or MX-tetOFF/PIASy-F (PIASy) were incubated stained with Annexin V-FITC and PI by using the Annexin for 6 days with b-estradiol. Cytospin preparation of each fraction V-FITC apoptosis detection kit (BD Biosciences) and was stained with the Wright-Giemsa solutions. Original magni®- analysed with a FACScan processor (BD Biosciences). cation 6400. (c) KCL22 cells infected with MX-tetOFF (Mock) or MX-tetOFF/PIASy-F (PIASy) were incubated for 0 or 6 days with b-estradiol. Cells were then stained with Annexin V-FITC and PI, and subjected to ¯ow cytometry. The percentages of apoptotic cells (Annexin V-positive and PI-negative) as well as Acknowledgments necrotic/late apoptotic cells (Annexin V-positive and PI-positive) We are grateful to A Iida and J-K Yee for the kind gifts of are designated tTAER cDNA and the tetO fragment, and T. Kitamura for

Oncogene DNA microarray analysis of CML K Ohmine et al 8257 the pMX vector. This work was supported in part by a Ministry of Education, Science, Sports, and Culture of Grant-in-Aid for Research on the Second-Term Compre- Japan, and by the Science Research Promotion Fund of the hensive 10-Year Strategy for Cancer Societies from the Promotion and Mutual Aid Corporation for Private Ministry of Health and Welfare of Japan, by a Grant-in- Schools of Japan. J Ota is a research resident of the Japan Aid for Scienti®c Research on Priority Areas from the Health Sciences Foundation.

References

Alon U, Barkai N, Notterman DA, Gish K, Ybarra S, Mack Liu B, Gross M, ten Hoeve J and Shuai K. (2001). Proc. Natl. D and Levine AJ. (1999). Proc. Natl. Acad. Sci. USA, 96, Acad.Sci.USA,98, 3203 ± 3207. 6745 ± 6750. Loeen JLCM, Triepels RH, van den Heuvel LP, Schuelke Boel P, Wildmann C, Sensi ML, Brasseur R, Renauld JC, M, Buskens CAF, Smeets RJP, Trijbels JMF and Smeitink Coulie P, Boon T and van der Bruggen P. (1995). JAM. (1998). Biochem. Biophys. Res. Commun., 253, 415 ± Immunity, 2, 167 ± 175. 422. Carlesso N, Frank DA and Grin JD. (1996). J. Exp. Med., Miyazato A, Ueno S, Ohmine K, Ueda M, Yoshida K, 183, 811 ± 820. YamashitaY,KanekoT,MoriM,KiritoK,ToshimaM, Chu G and Chang E. (1988). Science, 242, 564 ± 567. NakamuraY,SaitoK,KanoY,FurusawaS,OzawaK Duggan DJ, Bittner M, Chen Y, Meltzer P and Trent JM. and Mano H. (2001). Blood, 98, 422 ± 427. (1999). Nat. Genet., 21, 10 ± 14. Onishi M, Kinoshita S, Morikawa Y, Shibuya A, Phillips J, Era T and Witte ON. (2000). Proc. Natl. Acad. Sci. USA, 97, Lanier LL, Gorman DM, Nolan GP, Miyajima A and 1737 ± 1742. Kitamura T. (1996). Exp. Hematol., 24, 324 ± 329. Gouilleux-Gruart V, Debierre-Grockiego F, Gouilleux F, Shuai K. (2000). Oncogene, 19, 2638 ± 2644. Capiod JC, Claisse JF, Delobel J and Prin L. (1997). Leuk. Silver RT, Woolf SH, Hehlmann R, Appelbaum FR, Lymphoma, 28, 83 ± 88. Anderson J, Bennett C, Goldman JM, Guilhot F, Hin AH, Miraglia S, Zanjani ED, Almeida-Porada G, Kantarjian HM, Lichtin AE, Talpaz M and Tura S. Ogawa M, Leary AG, Olweus J, Kearney J and Buck (1999). Blood, 94, 1517 ± 1536. DW. (1997). Blood, 90, 5002 ± 5012. VanGelderRN,vonZastrowME,YoolA,DementWC, Iida A, Chen S-T, Friedmann T and Yee J-K. (1996). J. Barchas JD and Eberwine JH. (1990). Proc. Natl. Acad. Virol., 70, 60545 ± 60549. Sci. USA, 87, 1663 ± 1667. Izumi M, Miyazawa H, Kamakura T, Yamaguchi I, Endo T Wathelet MG, Szpirer J, Nols CB, Clauss IM, De Wit L, and Hanaoka F. (1991). Exp. Cell Res., 197, 229 ± 233. Islam MQ, Levan G, Horisberger MA, Content J, Szpirer Jang SK, Davies MV, Kaufman RJ and Wimmer E. (1989). Candet al (1988). Somat. Cell. Mol. Genet., 14, 415 ± 426. J. Virol., 63, 1651 ± 1660. Yamashita Y, Kajigaya S, Yoshida K, Ueno S, Ota J, Kantarjian HM, Deisseroth A, Kurzrock R, Estrov Z and Ohmine K, Ueda M, Miyazato A, Ohya K, Kitamura T, Talpaz M. (1993). Blood, 82, 691 ± 703. Ozawa K and Mano H. (2001). J. Biol. Chem., 276, Kawai T, Matsumoto M, Takeda K, Sanjo H and Akira S. 39012 ± 39020. (1998). Mol. Cell. Biol., 18, 1642 ± 1651. LetheB,LucasS,MichauxL,DeSmetC,GodelaineD, Serrano A, De Plaen E and Boon T. (1998). Int. J. Cancer, 76, 903 ± 908.

Oncogene