In vitro anthracycline cross-resistance pattern in childhood acute lymphoblastic leukaemia.

Daunorubicin (DNR) is a major front-line drug in the treatment of childhood acute lymphoblastic leukaemia (ALL). Previously, we showed that in vitro resistance to DNR at diagnosis is related to a poor long-term clinical outcome in childhood ALL and that relapsed ALL samples are more resistant to DNR than untreated ALL samples. In cell line studies, idarubicin (IDR), aclarubicin (ACR) and mitoxantrone (MIT) showed a (partial) lack of cross-resistance to the conventional anthracyclines DNR and doxorubicin (DOX), but clinical studies in childhood ALL have been inconclusive about the suggested lack of cross-resistance. In the present study we determined the in vitro cross-resistance pattern between DNR, DOX, IDR, ACR and MIT in 48 untreated and 39 relapsed samples from children with ALL using the MTT assay. The relapsed ALL group was about twice as resistant to DNR, DOX, IDR, ACR and MTT as the untreated ALL group. Thus, resistance developed to all five drugs. We found a significant cross-resistance between DNR, DOX, IDR, ACR and MIT, although in some individual cases in vitro anthracycline cross-resistance was less pronounced. We conclude that IDR, ACR and MIT cannot circumvent in vitro resistance to DNR in childhood ALL. Clinical studies may still prove whether IDR, ACR or MIT has a more favourable toxicity profile than DNR.

Two-thirds of children with newly diagnosed acute lymphoblastic leukaemia (ALL) can now be cured with combination chemotherapy, but chemotherapy fails in the remaining third, mainly because the leukaemia relapses (Niemeyer et al.. 1991). Despite intensive salvage chemotherapy. children with relapsed ALL still have a poor prognosis, and only one-third of them can be cured (Henze et al., 1991). Anthracychnes, such as daunorubicin (DNR) and doxorubicin (DOX). are commonly used in combination with several other classes of drugs in the treatment of childhood ALL. However, their clinical use is limited by cardiotoxicity and the development of drug resistance (Weiss, 1992). Previously, we showed that samples from children with relapsed ALL are more resistant to DNR than samples from children with untreated ALL (Pieters et al.. 1992: Klumper et al.. 1993. Anthracycline analogues lacking cross-resistance may circumvent resistance to DNR or DOX and may improve chemotherapy for relapsed childhood ALL. Many analogues have been developed since DNR and DOX were discovered in the early 1960s, but only a few have reached clinical trials (Muggia and Green, 1991). Idarubicin (IDR), aclarubicin (ACR) and the structurally closely related mitoxantrone (MIT) show a (partial) lack of cross-resistance to DNR and DOX in different cell lines (Hill et al.. 1985;Coley et al.. 1989;Erttmann et al.. 1991). Clinical studies with anthracycline analogues in childhood ALL are inconclusive about the suggested lack of cross-resistance. since no randomised comparative studies have been reported to date. In the present study, we determined the in vitro cross-resistance pattern between DNR. DOX. IDR, ACR and MIT within an uniform group of fresh samples obtained from 48 children with untreated ALL and 39 children with relapsed ALL.

Drugs
We tested the following drugs: DNR (Polyfarma, The Netherlands), DOX  DNR, DOX and IDR were dissolved in distilled water. ACR was dissolved in ethanol. MIT was obtained in soluble form. All drugs were further diluted with RPMI-1640 (Dutch modification. Gibco, Uxbridge. UK) and stored at -20'C in stock solutions of 50gml-1' (MIT). l00 gml-' (DNR, IDR) and 400 g ml-l (DOX. ACR). Microculture plates were prepared with serial 4-fold drug dilutions derived from these stock solutions, and the plates were stored at -20°C to facilitate large-scale testing. We used in vitro concentration ranges that covered clinical plasma concentrations (Table I). Anthracycines can be safely stored at -20'C up to 9 months without decomposition (Scott et al.. 1986). and without loss of in vitro anti-leukaemic efficacy of all drugs tested (data not shown).
Leukaemic samples Bone marrow (BM) and or peripheral blood (PB) samples were obtained, with informed consent, from 48 children with newly diagnosed ALL and 39 children with relapsed ALL. All children from the relapsed ALL group had previously been exposed to DNR and or DOX as part of multidrug front-line chemotherapy. None of the children with relapsed ALL had been tested before at initial diagnosis. Samples were processed within 24 h after collection.
Mononuclear cells were isolated by Ficoll density-gradient centrifugation (Ficoll Paque, density 1.077g ml-'; Pharmacia, Sweden) and washed twice in RPMI-1640 containing 0.1% bovine serum albumin. Representative in vitro drug resistance data can be generated if more than 70% ALL cells are present after a 4 day cell culture. since in vitro drug resistance will be overestimated if more than 30% contaminating nonmalignant cells are present (Kaspers et al.. 1994). To increase the number of evaluable cell cultures, the leukaemic cell population was enriched in 10 87 samples by removing contamiinating non-malignant cells using monoclonal antibodies linked to magnetic beads (Dynabeads M450. Dynal. Norway). We incubated cell suspensions (50 x 106 cells ml-') for 30 mmn at 37'C with one or a combination of the following mouse monoclonal antibodies (ITK, The Netherlands) directed against myeloid cells (CD13 and CD1 5, dilution 1:50 AMedian LC< in ig ml-'of the untreated ALL group. bLC50 of each drug relative to IDR. cMaximum tolerable dose in mg m-after one i.v. bolus. dMTD of each drug relative to IDR. or 1:100). monocytes (CD14. dilution 1:100) or T lymphocytes in case of non-T-lineage ALL samples (CD2. dilution 1:100). Cell suspensions were washed three times with RPMI and 10% fetal calf serum. Magnetic beads. coated with sheep anti-mouse immunoglobulin G. were added to the cell-antibody suspension (ten beads to one target cell) for 30 min at 37°C. The contaminating normal cells. linked through antibodies to the beads, were separated from the leukaemic cells by magnetic force. The mean percentage of leukaemic cells of the ten samples treated with beads increased from 75% to 89% using this method. The in vitro chemosensitivity of leukaemic cells was not influenced by treatment with monoclonal antibodies linked to beads (Kaspers et al.. 1994).
In vitro chemosensitivitv In vitro chemosensitivity did not differ between BM and PB samples (Kaspers et al.. 1991). All samples were freshly cultured with the exception of one cryopreserved sample. The MIT assay was performed as described before (Pieters et al.. 1990). Briefly, microculture plates containing 96 wells with 20 ji1 frozen aliquots of a drug were thawed just before testing and 80 Li of leukaemic cell suspension (2 x 106m1') was added. Leukaemic cells were cultured for 4 days in the absence or presence of six concentrations of each drug in duplicate.
May-Gru5nwald Giemsa-counterstained cytospins of control cells were made and showed that all samples contained ) 80% leukaemic cells at onset of the cell culture and ) 70% leukaemic cells after a 4 day cell culture. After 4 days. we added 10 jii of 5 mg ml-' MIT to each well and the microculture plates were incubated for another 6 h. The tetrazolium salt MTT is reduced to dark-coloured formazan crystals by viable cells only. Formazan crystals were dissolved with 100 pi of acidified isopropanol. The optical density (OD) was measured at 565 nm with an EL-312 microplate reader (Biotek Instruments. Winooski, USA). The OD is linearly related to the number of viable cells (Kaspers et al., 1991). We calculated the leukaemic cell survival (LCS) from the following equation: We used the LC«. the drug concentration lethal to 50% of the leukaemic cells. as parameter of in vitro chemosensitivity. Comparable in vitro drug resistance data can be obtained by repeated testing of samples. LCio values fall within the range of one dilution step (data not shown).

Statistics
The LC,0 values were non-parametrically distributed. Therefore, differences in the LC* distribution between untreated and relapsed childhood ALL samples were tested using the two-tailed Mann-Whitney L'-test. Spearman rank correlation coefficients (rho) were calculated to determine the cross-resistance patterns between the drugs tested.

Results
Anti-leukaemic activitY In general. for each drug steep dose-response curves were obtained. In vitro chemosensitivity was not influenced by cell culture efficiency: we found no significant correlations between the LC,0 values of each drug and the control leukaemic cell survival (rho = -0.01 to 0.06. P> 0.50) or between the LC,0 values and the OD per 10' viable control cells (rho =-0.01 to 0.19, P> 0.08). The in vitro chemosensitivity varied 40to 300fold among all patients: LC_% values (jag ml-') of DNR ranged from 0.012 to 1.294, of DOX from 0.023 to 1.374. of MIT from 0.003 to > 1. of IDR from 0.002 to 0.363 and of ACR from 0.037 to 1.531.
IDR was in vitro the most active anti-leukaemic drug. Intra-patient comparisons showed that in general the lowest LCO values were found for IDR. followed by MIT, DNR.
ACR and DOX. In Table II the median LC_% values of the untreated ALL samples are ranked and the anti-leukaemic activity is given relative to IDR; for example. a 3.3-fold higher DNR concentration is required compared with IDR to obtain an equal in vitro anti-leukaemic response. However. Table I1 shows that an increase in in vitro anti-leukaemic activity corresponds to an increase in the clinical toxicity as defined by the maximum tolerable dose (MTD) after one intravenous bolus of each drug; for example compared with DNR. IDR is in vitro about 3-fold more active, but the equitoxic dose of IDR is about 3-fold lower.
U-ntreated vs relapsed childhood ALL The control leukaemic cell survival did not differ (P = 0.66) between untreated (median 75%. range 35-138%) and  (Figure 1). We calculated the resistance ratios, i.e. the ratio of the median LC50 values from the relapsed and untreated ALL group; the resistance ratios ranged from 1.5 to 2.7 (Table III). We found a significant (P<0.001) correlation between the LC50 values of all drugs tested in 87 childhood ALL samples (Table IV). Figure 2 shows the cross-resistance pattern in childhood ALL between the front-line drug DNR vs DOX, IDR, ACR and MIT. Separate analyses of the untreated (n = 48) and relapsed (n = 39) ALL groups gave comparable Spearman rank correlation coefficients. A strong correlation was found between DNR, DOX, IDR and MIT (rho = 0.75-0.84), and to a lesser extent between ACR and the other four drugs (rho = 0.50-0.57). We have shown previously that resistance to DNR at initial diagnosis is correlated with a poorer long-term clinical outcome in childhood ALL (Pieters et al., 1991). Moreover, resistance to DNR and several other drugs such as prednisolone and L-asparaginase. might explain the poor prognosis of relapsed childhood ALL (Klumper et al.. 1993). Analogues lacking cross-resistance to DNR can theoretically circumvent DNR resistance, which may improve chemotherapy in relapsed childhood ALL. These analogues, such as IDR. ACR and MIT, have been identified in several cell line studies (Hill et al.. 1985;Coley et al., 1989;Erttmann et al., 1991). However, other cell line studies have reported contrasting in vitro resistance patterns, for example a full cross-  Scott et al., 1986: Gupta et al.. 1988). There is a need for such comparative studies using patient samples.
In the present study, we showed that samples from children with relapsed ALL were 2-fold more resistant not only to DNR and DOX. but also to IDR, ACR and MIT. compared with the untreated childhood ALL group. Moreover, the in vitro cytotoxicities of DNR. DOX, IDR. ACR and MIT were closely correlated. Thus, a pronounced cross-resistance developed to all five drugs suggesting that IDR. ACR and MIT cannot circumvent DNR resistance in childhood ALL. However, our results do not exclude the possibility that some individual children with relapsed ALL might benefit from ACR after front-line therapy including DNR, since we found a less pronounced although still significant cross-resistance between ACR and the other drugs tested. Although relapsed ALL samples were more resistant than untreated ALL samples to all five drugs tested, the LC^O   values of both groups showed considerable overlap. This suggests that DNR resistance in relapsed ALL may already be present at first diagnosis and that some children with relapsed ALL remained chemosensitive to DNR. Cell lines often express a 10-to more than 100-fold drug-induced resistance, whereas we found relative low resistance ratios for the anthracyclines in childhood relapsed ALL, ranging from 1.5 to 2.7. However, these resistance ratios were based upon the ratio of the median LC_% of the relapsed compared with the untreated ALL group; inter-patient chemosensitivities differed over 100-fold. Although Hill et al. (1989) have argued that cell lines expressing low levels of drug-induced resistance would be more suitable for the study of clinically relevant resistance mechanisms, our results based on patient samples suggest that cell lines with both low and high levels of drug-induced resistance could be used to investigate drug resistance. These large inter-patient variations probably reflect the clinical heterogeneity of tumour samples, which hampers the extrapolation of results from single cell line studies to clinically relevant results. No anthracycine resistance mechanisms or strategies to modulate anthracycine resistance with major clinical relevance in childhood ALL have been identified so far. P-glycoprotein expression has been most extensively studied in childhood ALL. but most studies could not detect significant differences in P-glycoprotein expression between untreated and relapsed childhood ALL samples (Ubezio et al., 1989: Tawa et al.. 1990: Mizuno et al., 1991Gekeler et al.. 1992: Pieters et al.. 1992: Brophy et al., 1994, whereas the latter samples were significantly more resistant to anthracycines in the present study. Moreover, we showed previously that neither verapamil nor cyclosporin A could modulate in vitro DNR resistance in childhood ALL (Pieters et al., 1992). Although multiple factors are likely to cause drug resistance, short-term cell culture drug resistance assays, such as the MTT assay, measure the end point of all possible resistance mechanisms, i.e. leukaemic cell kill, which is shown to be of predictive value in childhood ALL (Pieters et al., 1991).
Our results suggest that IDR, ACR and MIT do not have a higher therapeutic index than DNR since their in vitro anti-leukaemic activity was paralleled by their toxicity on normal cells, represented by the MTD. We found that IDR was the most active anti-leukaemic drug in vitro, a fact that is well known (Fields and Koeller, 1991). One clinical study reported a higher (statistically not significant) complete remission rate with IDR (75%) than DNR (59%) when both are used in combination chemotherapy in relapsed childhood ALL (Feig et al.. 1992). However, this study used increasing IDR doses to determine its MTD in combination chemotherapy, while the dose intensity of DNR was given below the MTD level. as shown by a significant higher toxicity in the group of children treated with IDR. Although our study suggests that the therapeutic index may not differ between IDR and DNR. we did not take into account other advantages of IDR compared with DNR, such as oral administration (Erttmann et al.. 1988;Pui et al.. 1988) and penetration of the cerebrospinal fluid (Reid et al.. 1990). In contrast to childhood ALL. a large randomised study in adult acute myeloid leukaemia (AML) reported a superior response rate for IDR compared with DNR in remission induction chemotherapy (Berman et al.. 1991).
Although our results suggest that IDR. ACR and MIT cannot circumvent DNR resistance in childhood ALL. clinical studies with IDR. ACR and MIT showed marked anti-leukaemic responses in children with relapsed ALL previously treated with DNR or DOX (Vietti et al.. 1983: Starling et al.. 1985: Ungerleider et al.. 1985: Fengler et al.. 1987Madon et al.. 1987: Tan et al.. 1987: Erttmann et al.. 1988Pui et al.. 1988;Giona et al.. 1990: Graham et al., 1991. However, the response rate for DNR and DOX in such selected patient groups is not known and might well be similar to that of the other three drugs mentioned. Moreover. a second complete remission rate up to 90% can be achieved with combination chemotherapy in relapsed childhood ALL (Henze et al.. 1991). Thus, a response to IDR. ACR or MIT in relapsed childhood ALL is not conclusive for a lack of cross-resistance to DNR and or DOX. Phase I II single-agent studies are difficult to compare as small. highly selective patient groups are usually tested, in contrast to our study comparing five drugs within a uniform patient group.
In summary, we showed that the relapsed ALL group was about twice as resistant to DNR, DOX, IDR. ACR and MIT as the untreated ALL group. Significant cross-resistance was observed between DNR, DOX. IDR, ACR and MIT, indicating that IDR, ACR and MIT cannot circumvent in vitro resistance to the conventional compounds DNR and DOX in childhood ALL samples. These results suggest that IDR, ACR and MIT are unlikely to enhance the antileukaemic response by replacing DNR in combination chemotherapy of relapsed childhood ALL, but these results, based upon cellular drug resistance proffles, do not exclude possible pharmacokinetic advantages of these analogues.