Letters to the Editor 1124 most of them the study population does not reach 200 cases. S Pigullo1, R Haupt2, C Dufour1, P Di Michele1, MG Valsecchi3, G Basso4, C Rizzari5, A Biondi5, M Lanciotti1 Interestingly, studies with larger cohorts give results similar to 1 ours. In a study of 197 patients (163 Caucasians and 34 blacks) Hematology Unit, Department of Pediatric Hemato- Chen et al5 found a higher prevalence of the double null GSTT1 Oncology, G Gasline Children’s Hospital, Genova, Italy; 2Epidemiology and Biostatistics Section, Scientific Directorate, and GSTM1 genotype among black children. In a large series of 8 G Gaslini Children’s Hospital, Genova, Italy; 651 (616 Caucasians and 35 black) ALL patients Davies et al 3Section of Medical Statistics, University of Milano-Bicocca, found GSTT1 and GSTM1 frequencies similar to those of Monza, Italy; controls, but they do not confirm the data on black children. In 4Pediatric Onco-Hematology Clinic, University of Padova, contrast a Canadian study on 177 cases showed an increased Padova, Italy; frequency of GSTM1 null genotype in children with ALL.9 5Centro Ricerca Tettamanti, Pediatric Clinic, University of A smaller number of studies have been reported on the Milano-Bicocca, Monza, Italy possible influence of GSTP1 genotype on childhood ALL. All of E-mail: [email protected] them had negative results; only one on 278 cases reported a role of GSTP1 in the susceptibility to childhood ALL in individuals carrying the variants Val105/Ala114 but did not shown clear results on the role of each variant.10 In these studies, positive References results were obtained in a French-Canadian population, suggesting that the influence of GSTs on the risk of leukemia 1 Salinas AE, Wong MG. Glutathione : a review. Curr Med could differ among populations, or a different exposure could be Chem 1999; 6: 279–309. involved in leukemogenesis. 2 Rebbeck TR. Molecular epidemiology of the human Glutathione S- genotypes GSTM1 and GSTT1 in cancer susceptibility. Several factors must be considered in the design of a reliable Cancer Epid Biomark Prevent 1997; 6: 733–743. case–control study. A large sample size with an adequate power 3 Zimniak P, Nanduri B, Pikula S, Bandorowicz-Pikula J, Singhal SS, is one of the most important factors. The choice of the control Srivastava SK et al. Naturally occurring human Glutathione population is also crucial because the possible different S-transferase GSTP1 isoforms with isoleucin and valin in position exposure to environmental toxicants should be considered. All 105 differ in enzymatic properties. Eur J Biochem 1994; 224: these factors were taken in account in our study and, in order to 893–899. 4 Ye Z, Song H. Glutathione s-transferase polymorphisms rule out geographical differences in genetic backgrounds and in (GSTM1, GSTP1 and GSTT1) and the risk of acute leukaemia: a environmental exposure to carcinogens, the normal controls systematic review and meta-analysis. Eur J Cancer 2005; 41: were matched not only for gender and age, but also for 980–989. geographical area of residence. This is the first study that takes 5 Chen CL, Liu Q, Pui CH, Rivera GK, Sandlund JT, Ribeiro R et al. into consideration also the residence area as a criterion for the Higher frequency of Glutathione S-transferase deletions in black selection of controls. children with acute lymphoblastic leukemia. Blood 1997; 89: 8 1701–1707. In addition this is the second study after that of Davies et al 6 Harries LW, Stubbins MJ, Forman D, Howard GC, Wolf CR. that considered the possible influence of genotype on the Identification of genetic polymorphisms at the Glutathione patients’ clinical characteristics, confirming the lack of associa- S-transferase Pi locus and association with susceptibility to tion between ALL subtype and GST polymorphism. bladder, testicular and prostate cancer. Carcinogenesis 1997; 18: In conclusion, this study represents one of the largest studies 641–644. on GSTT1 and GSTM1 genotypes, and the largest study on 7 Sala S, Lanciotti M, Valsecchi MG, Di Michele P, Dufour C, Haupt R et al. Genotypes of the glutathione S-transferase super- GSTP1 genotypes in children with ALL. Our data do not support family do not correlate with outcome of childhood acute a role of GST genotypes in genetic susceptibility to ALL. Our lymphoblastic leukaemia. Leukemia 2003; 17: 981–983. findings, although negative, contribute to give a definitive 8 Davies SM, Bhatia S, Ross JA, Kiffmeyer WR, Gaynon PS, Radloff answer to the question on the role of GST genotype in childhood GA et al. Glutathione S-transferase genotype, genetic susceptibility, ALL etiology at least in the Caucasian population. and outcome of therapy in childhood acute lymphoblastic leukemia. Blood 2002; 100: 67–71. 9 Krajinovic M, Labuda D, Richer C, Karimi S, Sinnett D. Suscept- Acknowledgements ibility to childhood acute lymphoblastic leukemia: influence of CYP1A1, CYP2D6, GSTM1 and GSTT1 genetic polymorphisms. Blood 1999; 93: 1496–1501. This work was supported by Compagnia di S. Paolo; Fondazione 10 Krajinovic M, Labuda D, Sinnett D. Glutathione S-transferase P1 CARIGE, ERG S.p.A. We thank Dr Anna Capurro for critical genetic polymorphisms and susceptibility to childhood acute reading of the manuscript. lymphoblastic leukaemia. Pharmacogenetics 2002; 12: 655–658.

Expression and mutation status of candidate kinases in multiple myeloma

Leukemia (2007) 21, 1124–1127. doi:10.1038/sj.leu.2404612; chain locus is found in 15% of patients. FGFR3 mutations, published online 8 March 2007 activating the receptor, are found in both primary patient samples and cell lines, supporting the role of FGFR3 signaling in the pathogenesis of the disease. In recent years, several other Kinase hyperactivity often results in deregulation of cellular kinases have also been reported to be somatically mutated in pathways involved in proliferation and survival. In multiple hematologic and solid organ malignancy. Furthermore, mutated myeloma (MM), for example, the translocation of FGFR3, a kinases are ideal targets for therapeutics. These observations receptor (RTK), to the immunoglobulin heavy have led us (and others) to attempt kinome-wide sequencing in

Leukemia Letters to the Editor 1125 the hope of identifying other, as yet unidentified, somatic IDs on Affymetrix U133 Plus 2 corresponding to known kinases. mutations in MM by focusing on RTKs. Of note, major Analysis of these kinases in healthy controls, MGUS, smoulder- sequencing efforts are currently being launched that may in ing MM (SMM), MM and MM cell lines revealed significant part be duplicative. Subsequently, we report here our data to differential expression in 185 unique kinases (Supplementary date derived from the sequencing of the kinase domains of 31 Information). Supervised hierarchical clustering indicated by the candidate and one kinase-regulatory gene CKS1B. heatmap in Figure 2b grouped the normals, cell lines and patient Although we did not find recurrent mutation in genes other samples as separate clusters. When normal plasma cells were than FGFR3, we have noted non-synonymous sequence compared with the malignant counterparts of MGUS, SMM or differences in ROR2, EGFR, GUCY2F and EPHB4 that are not MM, 64% (65/102) of kinases showed at least 2.84-fold higher present in dbSNP or pooled DNA from ethnically diverse expression in the disease state whereas 36% of the kinases (37/ populations that may now be further studied in larger cohorts of 102) showed diminished expression. From this list of upregu- MM patients for any pharmacogenetic significance. lated RTKs, seven RTKs were identified for sequencing: JAK3, Our general approach is detailed in Figure 1. Genomic DNA MET, KIT, CSK, ABL1, FGFR3 and ROR2. from 32 MM cell lines, and from CD138 þ purified marrow ROR2 expression showed a pattern reminiscent of the gene plasma cells from 24 MM patients (including 12 hyperdiploid), ‘spikes’ observed in translocated FGFR3, albeit at a level 10-fold eight monoclonal gammopathy of undetermined significance less than FGFR3 (Figure 3). Like FGFR3 mutations, ROR2 allelic (MGUS) and five plasma cell leukemia (total of 37 samples) mutations cause skeletal malformations in that inactivating were used in the study. Patient samples were obtained from mutations result in autosomal dominant type B3 Mayo Clinic (Scottsdale, AR, USA) and Princess Margaret and autosomal recessive .4 ROR2 is a Hospital (Toronto, ON, Canada) after informed consent. When negative regulator of Wnt signaling5 and its expression is linked necessary, a whole genome amplification method (GenomiPhi, in patients to the absence of bone disease (not shown). The GE Health Care, Piscata-way, NJ, USA Pharmacia) was used to parallel skeletal phenotype of germline mutations in both FGFR3 amplify a limited amount of patients’ genomic DNA. and ROR2, together with the spiked and differential pattern of PCR amplification of genomic DNA was performed in 96-well gene expression noted above, led us to consider ROR2 as strong plates using primers that have been previously described1,2 or candidate for kinase mutation screening. which were designed using the Primer3 program through the As a positive control, sequencing of the FGFR3 kinase domain Canadian Bioinformatics Resource web portal (cbr-rbc.nrc- of MM cell lines confirmed the known K650E FGFR3 mutation cnrc.gc.ca/index_e.php). Aliquots of PCR products were se- in OPM1 and OPM2 (Figure 2c). Interestingly, however, we also quenced on 96-well plates without purification using nested identified a novel heterozygous non-synonymous mutation sequencing primers. Sequence files generated from ABI 3730XL S433C (AAG to GAG) in KMS12PE, a cell line derived from DNA analyzer were analyzed using SeqScape software (Applied pleural effusion of an MM patient (Figure 2c). This mutation is Biosystems, Foster City, CA, USA) (ABI) and sequence changes not present in the bone marrow-derived cell line from the same were visually and manually inspected. Mutations were con- patient, KMS12BM (Figure 2c), suggesting that this mutation firmed by reverse direction and repeat sequencing. In 17 of the appeared later during disease progression. Sequencing of this genes sequenced, we covered 100% of the kinase domain exon in 37 patient samples did not identify additional mutations whereas 15 of these kinases had their kinase regions partially in this codon or in this region. However, in the same set of sequenced between 55 and 85%. JAK2 was only sequenced patient samples, we identified a novel mutation in an MGUS within the mutation hot spot observed in lymphoproliferative patient that was not seen in peripheral blood control lympho- disorders whereas CSK1B was fully sequenced. cytes from the same individual. This heterozygous non-synon- Initially, we focused on 24 RTKs (Figure 2a) implicated in ymous mutation, GCC to AGC (G197S), located in the other malignancies and one kinase regulatory subunit (CKS1B) immunoglobulin 2 domain, has not been previously reported. that is often amplified in MM. More importantly, this mutation is the first evidence that FGFR3 In an attempt to narrow the search, we then employed gene point mutation can occur early in the development of MM. expression profile data to look for other promising kinase Our broader kinome sequencing effort generated B2.0 candidates to screen for somatic mutations. From a reference list megabases of sequence. Analysis of the sequences generated of 518 kinases in the , we identified 1098 probe identified several single nucleotide polymorphisms (SNP) (Supplementary Information) located in the coding exons of the sequenced genes including sequence differences that have Schema not been reported or cataloged as SNP, perhaps owing to rarity DNA extracted of these alleles in the general population. Alternatively, these 32 cell lines Genome wide DNA amplification may represent rare sequence alterations in MM. We also 37 patient CD138+ cells identified four genes ROR2, EGFR, GUCY2F and EPHB4 that showed single nucleotide change (Figure 2c) when compared to their known coding sequence that were more suggestive of Compiled list of 25 candidate kinases DNA Sequence of Kinase domains from literature acquired mutation. EGFR, GUCY2F and EPHB4 each showed single nucleotide changes (Supplementary Table) in at least one sample when 4 candidate kinases identified compared to their known coding sequence, but the significance RNA extracted and of these codon differences is undetermined. For example, a differential kinase G917R change in EGFR sequenced from cell line MM1 was not expression examined DNA Sequence of 7 in CD138+ cells from patients, differentially expressed kinase seen in dbSNP or in ethnic pool DNA, but Western and normal individuals domains fluorescence activated cell sorting analyses (not shown) showed and cell lines no expression of the EGFR in this cell line, indicating that the Figure 1 The general schema and workflow are schematically consequence of this amino-acid change is neither functionally illustrated. nor physiologically relevant. Similarly, single nucleotide differ-

Leukemia Letters to the Editor 1126 Normals MGUS, SMM, MM, Cell Lines a % Scanned b

0 40

60 80 20 100 ACVR2 FLT3

EPHA 4 FGFR3 BTK GUCY2F CSK EPHB6 MAP3K9 ABL1 ACK1 ILK SRC ROR2 MET CKS1B MAP3K14 FGFR3 FGR EPHB4 PTK2 ROS1 KIT IGF1R NTRK3 JAK3 EGFR ABL1 MET KDR EPHA3 KIT FES MLK4 JAK3 CSK PIK3CA ROR2

c FGFR3 FGFR3 FGFR3 MGUS862 GGC/AGC OPM1 G197S KMS12 PE AAG/GAG TCC/TGC K650E S433C MGUS 862 PB GGC/GGC OPM2 G197 AAG/GAG KMS12 BM K650E TCC/TCC MM552 S433 GGC/GGC G197

EGFR EPHB4 GUYC2F

KMS12 JJN3 H929 GGA/GGA AGG/AGG CTC/CTC G917G R564R L100 MM1 KMS12 GGA/CGA KMS26 CTC/CTC G917R AGG/AAG R564K L100

RPMI 8226 OCI-MY5 L363 GGA/GGA AGG/AGG GTC/(Y chr) G917G R564R V100

Figure 2 (a) Sequence screening of 31 genes in a panel of 32 myeloma cell lines. The graph shows the percentage of kinase domains screened in the 31 kinases. (b) Heat map showing the groups of normals, primary myeloma samples and cell lines when expression of kinase genes was used for cluster analysis. (c) Nucleotide sequence changes in FGFR3, EGFR, EPHB4 and GUCY2F showing changes in the boxed sequence.

ences were detected in GUCY2F (L100V) from the L363 cell line impossible to determine the exact nature of these observations. and EPHB4 gene (R564K) from the KMS26 cell line, but the The functional importance of these changes can only be inferred unavailability of control DNA from these cell lines made it from amino-acid alignment of homologous sequences in other

Leukemia Letters to the Editor 1127 a CC Normals Patients 100

10

1 FGFR3

0.1

0.01

b CC Normals Patients 100

10 1 ROR2 0.1

0.01

Figure 3 Gene expression profiles of FGFR3 (a) and ROR2 (b) are compared in cell lines (CC), normal control plasma cells and in patient samples. A similar spiked appearance of the two RTKs is apparent. organisms. Generally, Leu at amino-acid position 100 in Acknowledgements GUCY2F (L100V) is conserved in most species, but it is substituted for Phe in Bos taurus and Val in Oryzias latipes, Financial support for this research was provided by seed funds suggesting that an amino-acid change in this position may not from the Mclaughlin Center for Molecular Medicine, University of be critical to the function of this . Conversely, Arg at Toronto (AKS), research grant from the Multiple Myeloma position 564 in EPHB4 appears to be highly conserved in EPHB Research Foundation (JOC, AKS), Canadian Institutes for Health gene family of human and other species (data not shown). Research (AKS), by the National Institutes of Health Grants Analysis of these sequence differences in 24 MM, eight MGUS CA55819 (JDS, FZ, BB) and CA97513 (JDS) and by the Fund to and five plasma cell leukaemia patient samples did not show Cure Myeloma and Peninsula Community Foundation (JDS). We similar changes in EGFR, GUCY2F or EPHB4 genes. thank M Bali, V Nadeem, Z Li and E Wei for their technical Two non-synonymous variations were observed in ROR2: assistance, and the patients who consented to the use of their G695R located in the kinase domain (in two cell lines) and anonymized samples for research. H349D (in cell line MM1) mapped to the kringle domain. No similar sequence changes were observed in 428 JO Claudio1, F Zhan2, L Zhuang1, R Khaja3, YX Zhu1, 3 1 1 4 of various ethnicity or reported in dbSNP; however, control K Sivananthan , S Trudel , E Masih-Khan , R Fonseca , PL Bergsagel4, SW Scherer2, J Shaughnessy3 and AK Stewart1,4 genomic DNA was not available from the affected individuals to 1 confirm mutation. For H349D, it is not known whether the Department of Medicine, University Health Network, McLaughlin Centre for Molecular Medicine and Ontario resulting amino-acid change from basic to acidic is critical Cancer Institute, Toronto, Ontario, Canada; to the function of the protein, but His in this region can 2Myeloma Institute for Research and Therapy, University of also be substituted for Asp in the kringle domain 3 of hepatocyte Arkansas for Medical Sciences, Little Rock, AR, USA; 4 growth factor and Caenorhabditis elegans CAM-1, suggesting 3Department of Genetics, Hospital for Sick Children, The that the amino acid His in this position may not be a critical Centre for Applied Genomics, Toronto, Ontario, Canada and amino acid. 4Division of Hematology-Oncology, Department of Medicine, Our study is the first reported attempt at high-throughput Mayo Clinic College of Medicine, Scottsdale, AZ, USA screening for kinase mutations in MM. With the exception E-mail: [email protected] of FGFR3, the sequence differences detected involved single nucleotide changes in no more than two (and usually References one) of the 32 cell lines and 37 MM individuals sequenced. Of note, our methodology would not detect big deletions 1 Bardelli A, Parsons DW, Silliman N, Ptak J, Szabo S, Saha S et al. that span the exons being studied. Nevertheless, our data Mutational analysis of the tyrosine kinome in colorectal cancers. suggest that within the 31 sequenced kinases, none is commonly Science 2003; 300: 949. mutated and activated in MM with the exception of 2 Samuels Y, Wang Z, Bardelli A, Silliman N, Ptak J, Szabo S et al. FGFR3. However, as we focused our study mainly on the High frequency of mutations of the PIK3CA gene in human cancers. kinase domain of selected tyrosine kinases, screening the full Science 2004; 304: 554. coding regions of other kinases may yet uncover a common 3 Oldridge M, Fortuna AM, Maringa M, Propping P, Mansour S, Pollitt C et al. Dominant mutations in ROR2,encoding an orphan receptor mutation with functional significance. Large-scale sequencing tyrosine kinase, cause brachydactyly type B. Nat Genet 2000; 24: efforts are about to begin and can be informed by these 375–378. preliminary results. 4 Afzal AR, Rajab A, Fenske CD, Oldridge M, Elanko N, Ternes- The gene ROR2 exhibits differential expression between Pereira E et al. Recessive Robinow syndrome, allelic to dominant myeloma patients and normal counterparts, is associated with brachydactyly type B, is caused by mutation of ROR2. Nat Genet hereditary skeletal defects similar to the myeloma-associated 2000; 25: 419–422. 5 Billiard J, Way DS, Seestaller-Wehr LM, Moran RA, Mangine A, FGFR3, is associated with less bone disease in patients and Bodine PV. The orphan Ror2 modulates contains point mutations not present in GenBank. The role of canonical Wnt signaling in osteoblastic cells. Mol Endocrinol 2005; ROR2 thus deserves further analysis. 19: 90–101.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia