Topoisomerase Ii Inhibitor‐related Acute Myeloid Leukaemia

In the late 1980s, we and colleagues described unique cases of treatment-related acute myeloid leukaemia (AML) in patients who had received intensive chemotherapy, including the epipodophyllotoxins, for acute lymphoblastic leukaemia (ALL) or solid tumours ( Ratain et al, 1987 ; DeVore et al, 1989 ; Pui et al, 1989 , 1990; Albain et al, 1990 ; Pedersen-Bjergaard et al, 1990 ; Prieto et al, 1990 ). A causative role of the epipodophyllotoxins was suspected but could not be established unequivocally. Thus, the terms secondary AML and therapy-related AML were used to describe these cases. By 1991, the leukaemogenic potential of the epipodophyllotoxins was firmly established ( Pui et al, 1991 ), leading to the term epipodophyllotoxin-related AML ( Pui, 1991; Whitlock et al, 1991 ). It has since been recognized that other topoisomerase II inhibitors, such as the anthracyclines (doxorubicin, 4-epi-doxorubicin), mitoxantrone, dactinomycin and dioxypiperazine derivatives (e.g. bimolane), can also induce AML with similar features, albeit at a lower frequency ( Andersson et al, 1990 ; Pui, 1991; Pedersen-Bjergaard et al, 1992 ; Xue et al, 1992 ; Sandoval et al, 1993 ; Pedersen-Bjergaard Smith et al, 1996 ). Collectively, these cases have been classified as topoisomerase II inhibitor-related leukaemia ( Ratain Pedersen-Bjergaard et al, 1993 ; Sandoval et al, 1993 ; Smith et al, 1995 ). We review here the pathogenesis, clinical and biological characteristics, risk factors and treatment responsiveness of this complication. Acute myeloid leukaemia after topoisomerase II inhibitor therapy can be distinguished from other therapy-related tumours by its unique molecular signature: balanced translocations involving the MLL ( mixed- lineage leukaemia or myeloid/ lymphoid leukaemia) gene on chromosome 11, band q23 ( DeVore et al, 1989 ; Pui et al, 1989 , 1991; Pedersen-Bjergaard et al, 1990 ; Ratain Felix et al, 1993 ; Super et al, 1993 ). The MLL gene, also referred to as ALL-1, HRX and HTRX1, encompasses more than 30 exons spread over approximately 100 kb of genomic DNA and is highly conserved among species ( Cimino et al, 1991 ; Ziemin-van der Poel et al, 1991 ; Gu et al, 1992a ; Downing Rasio et al, 1996 ). The homologue in Drosophila, the HRX or HTRX (trithorax) gene, is critical in segmental development ( Tkachuk et al, 1992 ; Parry et al, 1993 ). MLL is thought to act as a transcription factor, as it displays zinc finger and AT-hook domains, as well as regions of homology to DNA methyltransferases, all of which suggest a direct interaction between MLL and DNA ( Zeleznik-Le et al, 1994 ; Downing 11) more commonly associated with the former disease and the t(11;19) and t(9;11) with the latter ( Ziemin-van der Poel et al, 1991 ; Thirman et al, 1993 ; Downing Pratt 1998) and by its requirement for normal cell proliferation and differentiation during haematopoiesis ( Fidanza et al, 1996 ; Hess et al, 1997 ). Moreover, all chimeric mice with the MLL/AF9 transgene developed myeloid leukaemia at 4–12 months after transduction ( Corral et al, 1996 ). This delayed onset has led to speculation that mutagenic events subsequent to MLL rearrangement are needed to induce a leukaemic phenotype. Whatever the mechanism, MLL rearrangement (or at least MLL/AF9 fusion) appears to be a pivotal event in the genesis of therapy-related AML, rather than merely a genetic epiphenomenon. Much research has focused on the characteristics of MLL fusion genes. First, MLL gene fusion always results in an in frame transcript, with 5′MLL in frame with one of a variety of 3′ partner genes ( Downing Rubnitz et al, 1996 ; DiMartino Prasad et al, 1995 ), although most evidence indicates that interference with normal MLL function exerts the greatest influence. Whether that role resides in a gain-of-function, dominant-negative effect of the fusion gene or a loss of wild-type MLL function is not completely resolved. Sequencing of the breakpoints of MLL and its partner genes in leukaemic blasts cells has provided clues to underlying fusion mechanisms. Most such breakpoints occur in exons 5–11 and involve sequence-specific recombinogenic mechanisms. Rearrangements have been localized within Alu repeats ( Djabali et al, 1992 ; Gu et al, 1994 ; Schichman et al, 1994 ; So et al, 1997 ; Super et al, 1997 ), at conserved topoisomerase II cleavage recognition sites ( Negrini et al, 1993 ; Gu et al, 1994 ; Domer et al, 1995 ; Atlas et al, 1998 ), adjacent to V(D)J recombination signal sequences ( Gu et al, 1992b , 1994; Negrini et al, 1993 ; Height et al, 1994 ) and at a DNase I hypersensitive site ( Strissel et al, 1998 ). Interestingly, the MLL breakpoints found in therapy-related and infant leukaemias generally cluster in the telomeric end of the breakpoint cluster region, whereas those in de novo leukaemias tend to cluster in the centromeric end of the breakpoint cluster region ( Domer et al, 1995 ; Broeker et al, 1996 ; Cimino et al, 1997 ; Stanulla et al, 1997a ). This finding has lent further credence to the notion that infant and therapy-related leukaemias with MLL rearrangements arise through a common mechanism. Infant leukaemias may arise in utero ( Gill Super et al, 1994 ; Megonigal et al, 1998a ) after maternal exposure to environmental xenobiotics (dietary or otherwise) that interact with topoisomerase II ( Ross et al, 1994 ). Scaffold attachment regions, which may play a role in facilitating the access of the cellular recombination machinery to the breakpoint cluster region of MLL, have also been localized to this critical region of MLL ( Broeker et al, 1996 ). Because of the potent inhibition of religation of double-strand DNA breaks induced by etoposide and other topoisomerase II-active agents, it is plausible that cells would attempt DNA recombination through alternative mechanisms (Fig 1), many of which are not completely defined in mammalian cells. )rearrangement], a leukaemogenic process may result. Host factors that may increase susceptibility to etoposide-induced AML include low-thiopurine methyltransferase activity (TPMT), low-glutathione transferase activity (GST), younger age, deficient folate status, decreased DNA repair proficiency, mutant topoisomerase (topo) II or increased oxidative metabolic activation through cytochrome P450 3A4 (CYP3A4). In vitro studies have shown that etoposide increases the frequency of non-homologous V(D)J recombinase-mediated DNA rearrangements ( Chen et al, 1996a , b), as well as the frequency of direct MLL breaks ( Aplan et al, 1996 ; Stanulla et al, 1997a , b). Southern blot analysis of enzyme-digested DNA from etoposide-treated cell lines and from peripheral blood cells after in vivo treatment with etoposide showed frequent rearrangement of MLL ( Aplan et al, 1996 ; Stanulla et al, 1997a ), but not of other genes ( Aplan et al, 1996 ). Importantly, MLL rearrangement can also occur after treatment with anticancer drugs that do not interact potently with topoisomerase II, such as cytarabine and methotrexate ( Stanulla et al, 1997b ), or even in cells that have not been exposed to anticancer drugs ( Stanulla et al, 1998 ). Finally, MLL fusions or duplications have been detected at both transcriptional ( Marcucci et al, 1998 ; Schnittger et al, 1998 ; Uckun et al, 1998 ) and genomic ( Schnittger et al, 1998 ) levels in haematopoietic cells with a normal karyotype. In summary, the available evidence indicates that MLL is very prone to rearrangement in haematopoietic cells; that the precise location of the rearrangement is determined by specific nucleotide sequences that direct recombinogenic processes; and that MLL rearrangement, although by itself not sufficient for leukaemogenesis, is very likely to be a necessary step in the transformation of myeloid or lymphoid progenitors. In contrast to the myeloid leukaemias induced by alkylating agents, topoisomerase II inhibitor-related AML has a short latency period (median 30–34 months), lacks a myelodysplastic phase and is characterized by a predominance of myelomonoblastic and monoblastic morphologic features, as well as balanced chromosomal translocations, generally involving the 11q23 region (over 70% of cases; Table I) ( Pui et al, 1989 , 1991; Pedersen-Bjergaard Smith et al, 1995 ; Archimbaud et al, 1998 ). The most common reciprocal translocations are the t(9;11)(p21;q23) and the t(11;19)(q23;p13) ( Pui et al, 1995a ; Secker-Walker et al, 1998 ), although balanced translocations affecting regions other than 11q23 have been recognized in topoisomerase II inhibitor-related leukaemias ( Table II) ( Pedersen-Bjergaard et al, 1993 , 1994; Sandoval et al, 1993 ; Pui et al, 1995a ; Archimbaud et al, 1998 ; Nishiyama et al, 1999 ). The relation of specific translocations to the type of topoisomerase II inhibitor, age or race is still uncertain. In general, 11q23 translocations predominate in children treated with epipodophyllotoxins ( Pui et al, 1991 , 1995a; Winick et al, 1993 ); the t(15;17) is frequent in Japanese and Italian children treated with epipodophyllotoxins for Langerhans cell histiocytosis ( Horibe et al, 1993 ; Matsuzaki et al, 1994 ; Haupt et al, 1997 ; Kudo et al, 1998 ) and in Chinese treated for psoriasis with bimolane ( Xue et al, 1992 ); and the t(8;21), t(3;21), inv(16) and t(8;16) are more commonly associated with anthracyclines than with other genotoxic agents ( Quesnel et al, 1993 ; Sandoval et al, 1993 ; Pedersen-Bjergaard Pedersen-Bjergaard et al, 1994 ; Dissing et al, 1998 ). Treatment with epipodophyllotoxins or anthracyclines is only rarely associated with the development of Philadelphia chromosome — or t(4;11)-positive leukaemias (generally ALL) ( Pui, 1992; Pedersen-Bjergaard et al, 1997 ). Of interest, the molecular characteristics of the t(15;17) and the inv(16) differ between de novo and therapy-related leukaemias ( Naoe et al, 1997 ; Dissing et al, 1998 ). In epipodophyllotoxin-related acute promyelocytic leukaemia with the t(15;17), the PML breakpoints were located within intron 6, whereas those of the RARA gene were found in a restricted region of intron 2. In de novo cases, the PML breakpoints clustered within intron 3 and around intron 6, whereas those of the RARA gene were widely distributed within intron 2 ( Naoe et al, 1997 ). Similar differences were noted for MYH11 breakpoints in cases of de novo and therapy-related AML with the inv(16) ( Dissing et al, 1998 ). Risk estimates for the development of topoisomerase II inhibitor-related AML vary widely among studies. These discrepancies have been attributed to different cumulative drug doses, different treatment schedules, co-administration of other antineoplastic agents and different patient populations (hence different host factors) ( Smith et al, 1995 ). There are conflicting data as to whether there is a dose–response relationship in the development of topoisomerase II inhibitor-related AML ( Pedersen-Bjergaard et al, 1991 ; Hawkins et al, 1992 ; Bajorin et al, 1993 ; Nichols et al, 1993 ; Bokemeyer Bokemeyer et al, 1995 ; Boshoff et al, 1995 ). In a case–control study of children treated for a variety of cancers, Hawkins et al (1992 ) showed a positive correlation between the relative risk of AML development and the cumulative dose of epipodophyllotoxins. Patients who received these agents at a cumulative dose exceeding 1200 mg/m2 had a relative risk of 17.1, compared with the risk among patients receiving lower cumulative doses. Most of the data supporting a dose–response relationship come from studies of patients with germ cell tumours ( Table III). In general, patients who received etoposide at a cumulative dose of less than 2000 mg/m2 had a negligible risk (< 1%) of AML development. However, with rare exceptions, the risk generally does not exceed 5% in patients treated for solid tumours, even among those who received high cumulative doses of topoisomerase II inhibitors ( Kollmannsberger et al, 1998 ; Smith et al, 1999 ). In contrast, several authors have reported a greater than 5% rate of AML induction in patients treated for ALL or non-Hodgkin's lymphoma ( Pui et al, 1991 , 1995b; Verdeguer et al, 1992 ; Sugita et al, 1993 ; Winick et al, 1993 ; Amylon et al, 1999 ). Reasons for this discrepancy are not obvious, but probably include factors (treatment intensity and types of co-administered agents, for example) that differ between treatments for solid tumours and those for leukaemias. Given identical cumulative doses of epipodophyllotoxins, patients with lymphoid malignancies who were treated with these agents on a twice-weekly or weekly schedule had a greater risk of AML development than patients who received the agents every other week ( Table IV) ( Pui et al, 1991 , 1995a). Thus, the schedule of administration of epipodophyllotoxins has an important effect on the risk of therapy-related AML. Interestingly, daily administration of epipodophyllotoxins for 4 or 5 days, as typically specified for patients with solid tumours, has been associated with a relatively low risk of AML induction ( Bajorin et al, 1993 ; Nichols et al, 1993 ; Bokemeyer et al, 1995 ; Kollmannsberger et al, 1998 ; Smith et al, 1999 ), most probably because of the lower cumulative doses of epipodophyllotoxins, as well as the use of different concurrent or preceding chemotherapy. Whether prolonged exposure to relatively small doses of oral etoposide (e.g. 50–100 mg/m2/day) is associated with a decreased risk of this complication is uncertain. To date, only a few patients treated on this schedule have developed AML ( Stine et al, 1997 ; Yagita et al, 1998 ), although these studies have been limited to patients with relapsed or refractory cancers and short follow-up times. In this regard, our in vitro study demonstrated that prolonged exposures to low concentrations of this agent (a schedule mimicking oral dosing) were associated with a lower frequency of non-homologous site-specific DNA recombination, an event predisposing to AML induction, than were brief exposures to high concentrations of etoposide (a schedule commonly used for patients with leukaemia) ( Chen et al, 1996b ). Co-administration of topoisomerase II inhibitors with other agents that interact with DNA, such as alkylating or platinum compounds, could be expected to potentiate the leukaemogenic effects of both the anthracyclines ( Hawkins et al, 1992 ; Pedersen-Bjergaard et al, 1992 , 1993; Sandoval et al, 1993 ; Kushner et al, 1998 ) and the epipodophyllotoxins ( Ratain et al, 1987 ; Hawkins et al, 1992 ; Pedersen-Bjergaard et al, 1993 ; Heyn et al, 1994 ; Kushner et al, 1998 ). However, the mutagenic effects of the anthracyclines, which intercalate with DNA, may not be clinically obvious because of dose-limiting cardiotoxicity ( Marty, 1993; Riggi Sugita et al, 1993 ; Winick et al, 1993 ). treatment and folate have been shown to to both and ( et al, 1994 ). could concentrations and the of DNA as well as nucleotide ( et al, 1996 ; predisposing to non-homologous recombination after etoposide have indicated that host factors can also patients to topoisomerase II inhibitor-related AML. We have reported that patients with AML tend to have lower methyltransferase than do treated who not AML ( et al, 1998 ), a with the that therapy can the leukaemogenic effects of Moreover, Felix et al ) reported a of among patients with therapy-related leukaemias and 11q23 compared with those with de novo leukaemia and the 11q23 We have found a for more or among children with ALL who developed therapy-related AML than among those this complication ( et al, 2000 ). are rare among patients who topoisomerase II inhibitor-related AML ( et al, 1992 ); in it has been that wild-type is for MLL rearrangement ( Megonigal et al, ). age has also been associated with topoisomerase II inhibitor-related AML in a analysis ( et al, 1998 ), although it is whether age is a biological risk or a for the more leukaemogenic regimens of etoposide used in infants and other A of and therapy-related factors could interact with the recombinogenic effects of etoposide (or other topoisomerase II-active to increase the risk of AML development is shown in and by in ( of the role and potential interaction of and therapy-related factors in leukaemogenesis (Fig ). induction in patients with topoisomerase II inhibitor-related AML, of whether have t(8;21), inv(16) or t(15;17) chromosomal are similar to those reported for cases of de novo AML. on a of cases with t(8;21), or t(15;17) would ( et al, 1992 ; Quesnel et al, 1993 ; Dissing et al, 1998 ; Kudo et al, 1998 ). In the study by et al ), of cases with 11q23 most of were treated with and of our cases with 11q23 after induction treatment with regimens that an ( Pui et al, 1995a ). It is that therapy with induced or in of the a in with in cases of de novo AML ( et al, 1994 ). A was also induced by a of and cytarabine in a with etoposide-induced AML ( et al, 1996 ). Despite these of therapy-related AML with 11q23 are and there are few who were treated with ( Pui et al, 1995a ; et al, 1996 ). in one study, the of such patients was only ( et al, 1994 ). The of cases with the t(8;21), inv(16) or t(15;17) treated with only is less because of the of follow-up haematopoietic stem cell is a for patients who AML after treatment with a topoisomerase II In are a treatment in this There are the results of in children with epipodophyllotoxin-related AML. et al ) reported on patients who received and the of the only patients were in for and months after and et al ) and as the patients at and months with a rate of In our of patients were in for to after a rate of ( et al, 1999 ). the for patients with therapy-related AML even after One alternative is to to host haematopoietic cells in the as a of leukaemia ( et al, 1995 ). is to the leukaemia effect by after therapy ( et al, 1997 ). the results of one study that the leukaemia effect may the rate in patients with therapy-related AML, supporting the use of after ( et al, 1998 ). This was in by and from the of by a of from the of and by