This study was undertaken to investigate the expression patterns of three proteins (PGP, LRP and MRP1) that confer in vitro multidrug resistance, including anthracyclines, in APL based on reports by Head et al (1995) and Avvisati et al (1990) , who showed the major role of DNR in affecting the outcome of APL. It included 23 adult APL cases at onset who were diagnosed, managed and followed up at a single institution. A second point was to evaluate onset vs. relapse MDR phenotype in paired samples and to compare possible increases in the proteins' expression with the amount of cytotoxic drugs delivered to the patients. Similar to 96 de novo ANLL cases evaluated at onset with FAB cytotypes other than M3 ( Michieli et al, 1999 ), the majority of the APL cases had low amounts of PGP and LRP. However, in APL, the evidence of PGP or MRP-1 expression exceeding the negative controls was a much rarer event. In this series, no APL case at onset had PGP levels higher than the negative controls, whereas LRP was elevated in only one out of 23 cases. Low amounts of MRP1 were found in less than one-third of our APL patients, and no case had an MRP1 expression that exceeded the negative controls. In accordance with these data, a reduced cellular DNR accumulation was only observed in the case showing LRP overexpression. This is a relevant point as it confirmed the immunophenotypic finding but suggests that, besides PGP, LRP and MRP1, other proteins involved in anthracycline transport (such as MRP2, 3, 4 or 5) are unlikely to be overexpressed and active in APL cases at onset. Concerning PGP expression in APL, the data are in line with the work of several authors who used different reagents and techniques for PGP evaluation ( Paietta et al, 1994; Drach et al, 1995 ; Guerci et al, 1995 ; Del Poeta et al, 1997 ). As in our work, Paietta et al (1994) showed a low PGP activity in 11 APL cases. Up to now, only sporadic data have been available on LRP expression in APL. Filipitis et al (1998) reported one out of nine LRP-overexpressing cases, while List et al (1996) did not find any APL-positive cases in four samples analysed. Concerning ANLL with a FAB cytotype other than M3, several authors have reported a higher PGP and LRP overexpression compared with APLs ( Zhou et al, 1995 ; Filipitis et al, 1997 ; Leith et al, 1997 , 1999; Michieli et al, 1997 ; Willman, 1997; Damiani et al, 1998 ; Legrand et al, 1999 ). Few, and sometimes conflicting, data are available on the relationship between PGP, LRP or MRP expression and cytogenetics in ANLL other than M3 ( Del Poeta et al, 1996 , 1997; List et al, 1996 ; Leith et al, 1997, 1999; Michieli et al, 1999 ). A marginal non-significant association between poor-prognosis cytogenetics and PGP overexpression has been reported ( Del Poeta et al, 1996 , 1997; Leith et al, 1997 , 1999), as well as a trend towards a lower frequency of high PGP and LRP expression in patients with a good-prognosis karyotype ( List et al, 1996 ; Michieli et al, 1999 ). At present, no data are available on MDR phenotype in M3 cases at relapse. By comparing the immunophenotypic MDR features of the seven patients who ultimately relapsed vs. those of the patients who achieved and are maintaining CR, we failed to identify any particular functional or immunophenotypical feature useful in distinguishing high- from low-risk relapsing patients. The case that initially showed LRP overexpression associated with low cellular drug accumulation and relapsed at 8 months into CR was only a sporadic finding. Serial analysis undertaken in the seven patients who ultimately relapsed revealed overexpression of PGP in one out of seven at first relapse; however, the acquisition of an MDR phenotype appeared to become increasingly frequent with subsequent relapses (three out of four cases overexpressed PGP at the second relapse). Notably, the only case that substantially increased the baseline PGP expression at first relapse was the only one who received DNR instead of idarubicin or mitoxantrone as first-line therapy. This observation, although occasional and requiring confirmation, suggests that, compared with idarubicin or mitoxantrone, DNR, the toxicity and transport of which has the strongest relationship with PGP action, may also have the highest activity in stimulating PGP expression ( Berman & McBride, 1992 ; Michieli et al, 1993 ). In addition, it must be stressed that the patient who showed an increased PGP expression at first relapse received the highest dose of arabinosyl cytosine (AC), which has been shown to have a high activity in inducing PGP expression ( Grimaz et al, 1998 ). The suggestion that promyelocytic blasts may require multiple courses of drug exposure to become PGP overexpressing was also supported by Su et al (1994) and Chaudhary & Roninson (1993) who carried out experiments on HL60 cell lines. In fact, whereas Su et al (1994) showed that PGP-overexpressing sublines could be obtained after multiple 18-h drug pulsing, Chaudhary & Roninson (1993) failed to obtain PGP-overexpressing HL60 sublines after a single transient high-dose drug exposure. As for PGP, no changes in onset vs. first relapse LRP expression were found. MRP1 expression did not vary significantly during follow-up of the seven relapsing patients. In the literature, the onset vs. relapse data are still limited and sometimes conflicting even in ANLL without a restriction according to the FAB cytotype. Based on non-paired samples, a number of authors have concluded that PGP, LRP and MRP1 expression was lower at onset than in relapsed ANLL patients ( Schneider et al, 1995 ; Hart et al, 1997 ; Nuessler et al, 1997 ; Legrand et al, 1999 ). In paired samples, Hart et al (1997) showed no significant difference in PGP, LRP and MRP1 expression in eight ANLL cases, while Schneider et al (1995) showed an increase in MRP1 expression at relapse in more than 80% of 17 acute leukaemic samples evaluated.
In conclusion, this work is consistent with the view that low PGP, LRP and MRP1 expression and a lack of defects in cell drug accumulation may account for the high sensitivity of APL to anthracyclines. It suggests that the analysis of the MDR phenotype at onset is unlikely to help in screening high- from low-risk APL relapsing patients. Because of the small series, data on PGP, LRP and MRP1 expression in paired onset vs. relapse samples are preliminary and require confirmation. Our data suggest that no relevant changes in PGP, LRP or MRP1 expression are likely to occur at first relapse. In contrast, relevant increases in baseline PGP expression are likely to develop later in the patient's history as a consequence of the delivery of additional chemotherapy courses.