The efficacy of the anthracycline prodrug daunorubicin-GA3 in human ovarian cancer xenografts.

The prodrug N-[4-(daunorubicin-N-carbonyl-oxymethyl)phenyl] O-beta-glucuronyl carbamate (DNR-GA3) was synthesized for specific activation by human beta-glucuronidase, released in necrotic areas of tumour lesions. In vitro, DNR-GA3 was 18 times less toxic than daunorubicin (DNR) and the prodrug was completely activated to the parent drug by human beta-glucuronidase. The maximum tolerated dose of DNR-GA3 in nude mice bearing s.c. human ovarian cancer xenografts was 6-10 times higher than that of DNR. The prodrug was cleared more rapidly from the circulation (elimination t1/2 = 20 min) than the parent drug (elimination t1/2 = 720 min). The anti-tumour effects of DNR-GA3 and DNR were investigated in four different human ovarian cancer xenografts OVCAR-3, FMa, A2780 and MRI-H-207 at a mean tumour size between 100 and 200 mm3. In three out of four of these tumour lines, the prodrug given i.v. at the maximum tolerated dose ranging from 150 to 250 mg kg(-1) resulted in a maximum tumour growth inhibition from 82% to 95%. The standard treatment with DNR at a dose of 8 mg kg(-1) given i.v. weekly x 2 resulted only in a maximum tumour growth inhibition from 40% to 47%. Tumour line FMa did not respond to DNR, nor to DNR-GA3. Treatment with DNR-GA3 was also given to mice with larger tumours that would contain more necrosis (mean size 300-950 mm3). The specific growth delay by DNR-GA3 was extended from 2.1 to 4.4 in OVCAR-3 xenografts and from 4.4 to 6.0 in MRI-H-207 xenografts. Our data indicate that DNR-GA3 is more effective than DNR and may be especially of use for treatment of tumours with areas of necrosis.

. In human ovarian cancer xenografts. N-L-leucyl-DOX was shown to be more effective than DOX (Boven et al. 1992). Clinical studies on N-L-leucyl-DOX have yet to be completed.
Human >-glucuronidase is another enzyme in which levels are elevated in tumour tissue when compared with normal tissues (Connors and Whisson, 1966). Albin et al (1993) have shown that the concentration of S-glucuronidase was six times higher in breast cancer tissue of patients than in peritumoral tissue. when measuring the enzyme activity of tissue homogenates. Bosslet et al (1995) and Schumacher et al (1996) have shown by enzyme histochemistry that S-glucuronidase was expressed in a wide range of tumour types and was particularly localized in necrotic areas. The enzyme can only be detected in very low concentrations in the circulation (Fishman. 1970). It is hypothesized that anthracycline-glucuronide prodrugs may be selectively activated in tumour tissue on the basis of high fr.glucuronidase levels released by necrotic cells. We have developed such a glucuronide prodrug of DNR: N-[4-(daunorubicin-N-carbonyl-oxymethyl) phenyll O->-glucuronyl carbamate (DNR-GA3) (R. G. G. Leenders. submitted: Figure 1).
In the present studies. we compared the antiproliferative effects of DNR-GA3 and DNR in vitro and their respective elimination half-life times (t,,:) from the circulation of mice. After determination of the maximum tolerated dose (MTD) of DNR-GA3 in tumour-bearing mice. the efficacy of the prodrug was compared with that of DNR in four human ovarian cancer xenografts. Large tumours have more necrosis than small tumours and were expected to contain more extracellular S-glucuronidase. Therefore. special attention was paid to the influence of the tumour size on drug effects.
In vitro antiproliferative effects The in xitro antiproliferatixe effects of drug. prodrug and spacer xxere determined wxith the use of OVCAR-3 cells as prexiously described (Houba et al. 1996a H-207 tumour lines are undifferentiated carcinomas wxith xolume doubling times of 2.0 and 3.5 davs respectixely. Tumours from prexvious recipients were transferred bV implanting tissue fraLtments w ith a diameter of 2-3 mm into both flanks of 8to 1 0-wxeek old mice. Upon growxth. tumours wxere measured by the same obserx er. The tumour xolume >-as calculated b\-the equation length x wxidth x thickness x 0.5. and expressed in mm'.

Anti-tumour activity of anthracyclines in vivo
First. the MTD of the prodrug gixen i.x-. >-as determined. At the MTD. a mean rexersible loss was required of approximately 10c of the initial wxeight x ithin 2 wxeeks after the first injection. Deaths occurnc wxithin 2 weeks after the final injection wxere considered as toxic deaths. The MTD of DNR in tumour-bearing mice wxas considered to be 10 mg kgi.x. wxeekly x 2. as higher doses induced ascites (Boxven et al. 1996). The MTD of DNR-GA3 gixven once or weekly x 2 xas determined in non-tumour-bearing mice first and adjusted in tumour-bearing mice. After defininc the MTD for the prodrug. treatment experiments xxere carried out. At the start (day 0 mice x ere grouped to obtain similarities in the mean tumour x-olume. For small tumours. the mean xolume ranged from 119 to 194 mm and for large tumours from 337 to 953 mm . Control and treatment groups consisted of six animals each. DNR xxas gixen in a dose of 8-10 mg kg-; i.X xxeekly x 2 to mice xxwith small tumours only-. DNR-GAxx was studied at the MTD i.x. once or wxeekly x 2. Mice wxere xxeighed tx-ice per xxeek and tumours wxere measured on the same dayvs.  Figure 2 In vitro antproliferative effects in OVCAR-3 celis exposed to various concentratons of DNR or DNR-GA3. Cell growth was measured after 72 h by suphorhodamine B staining and was expressed as te percentage of growth in control cells. (-) DNR; (*) DNR-GA3; (U) DNR-GA3 in the presence of an excess of human -giucuronKiase. Bars, ± s.d.
Differences in efficacy between treatment groups were expressed as the percentage of maximum growth inhibition (GI). The relative tumour volume was expressed by the formula VT/VO where VT is the volume on any given day and V0 is the volume on day 0. The ratio between the mean of the relative volumes of treated tumours and that of control tumours x 100% (T/C%) was assessed on each day of measurement and used to calculate the GI (GI = 100% -T/C%). The maximum GI was scaled as follows: GI .50% was defined as not sensitive. 50% < GI < 75% was defined as sensitive, and GI >75% was defined as very sensitive (Boven et al. 1988). The GI range from 40% to 50% was called borderline sensitive. The efficacy of the treatment was also expressed by calculating the days for each tumour to double twice in volume (TD1,4). If a tumour did not reach two volume-doubling times, this volume was extrapolated from the last two available measurements. Differences in mean TD,I 4 between groups were evaluated with Student's t-test. In addition. differences in efficacy between the treatment groups of small vs large tumours were expressed as the specific growth delay (SGD: Boven et al. 1988). The SGD was calculated according to the following formula: SGD = (TDI,, treated -TDI__ control)/TD01 control

In vitro antiproliferative effects
The antiproliferative effects of DNR and DNR-GA3 were determined by measuring the growth of OVCAR-3 cells with the Time after administration (h) Fgure 3 Pharmacoknetics of (U) DNR-GA3 10 mg kg-l or (-) DNR 10 mg kg-', given i.v. to BALB/c mice. At different time points after injection, plasma was analysed for DNR-GA3 and DNR content by reversed-phase HPLC as descrbed in Materials and mettods. Bars, ± s.d.
sulphorhodamine B assay. DNR (IC, = 2 gm) was 18 times more toxic than the prodrug (IC, = 35 )1M) when cells were exposed to drugs for 24 h. Incubation of cells with DNR-GA3 in the presence of excess human $-glucuronidase resulted in an increase of the antiproliferative effects reaching the same IC,O as for DNR ( Figure   2). This indicates that the relatively non-toxic prodrug was completely activated to the toxic drug by the enzyme. Decomposition of the carbamate spacer will liberate 4aminobenzyl alcohol. When OVCAR-3 cells were incubated with 4-aminobenzyl alcohol alone. no toxicity was observed at concentrations up to 100 gm (data not shown).

Kinetics of DNR-GA3 and DNR
The pharmacokinetics of the prodrug and the drug were determined in BALB/c mice (Figure 3). DNR cleared slowly from the blood with a terminal tV,: of 720 min (n-=3) and was detectable for more than 24 h in the circulation. DNR-GA3 cleared rapidly with a terminal t1,2 of 20 min (n=3). At 4 h. the DNR-GA3 concentration was under the detection limit of 0.01 Jm. After the i.v. administration of DNR-GA3, no DNR was detectable in the plasma of the mice.
Maximum tolerated dose (MTD) and toxicity For DNR. a dose of 10 mg kg-1 i.v. weekly x 2 studied in OVCAR-3-bearing mice was too toxic because five out of six mice suffered from ascites and rapid death between 16 and 92 days after the first injection. DNR 8 mg kg-' i.v. weekly x 2 was well tolerated in subsequent treatment experiments.   In non-tumour-bearing nude mice. the MTD of DNR-GA3 was 250 mg kg-' i.v. A higher dose of prodrug was considered to be too toxic because this resulted in >15% weight loss and several toxic deaths (data not shown). In OVCAR-3-bearing mice. we studied DNR-GA3 in a range of 100-250 mg kg-' i.v. Although weight loss at doses of 200-250 mg kg-' i.v. varied slightly between experiments. a single dose of 250 mg kg-' was defined as the MTD. This dose resulted in a maximum weight loss of 10.7% for mice with small and 7.6% for mice with large OVCAR-3 tumours. In the group with the small tumours, one out of six mice died within 14 days (Table 1). If DNR-GA3 was given weekly x 2. doses of 150 mg kg-' and 200 mg kg-' were well tolerated. No toxic deaths occurred and the weight loss was not more than 6.4%.
While experiments were in progress, it was found that in FMa-. A2780and MRI-H-207-bearing mice the weight loss from DNR-GA3 varied and required adjustment of the dose. The dose of 250 mg kg-' was too toxic for mice bearing small FMa tumours because the animals developed ascites. and five out of six died between day 16 and day 36. In mice bearing large FMa tumours. a dose of 200 mg kg-' caused ascites in two out of six animals.
Smaller doses of DNR-GA3 were not studied in FMa as this dose of prodrug was ineffective. In A2780-bearing mice, a dose of 200 mg kg-' was too toxic. This resulted in >15% weight loss and Britsh Journal of Cancer (1998) 78(12) slirhtly. but not significantlv. less toxic in mice I MRI-H-207 tumours and was considered as the MTI Anti-tumour activity of DNR and DNR-GA3 in vivo The anti-tumour effects of DNR were different among the four human ovarian cancer xenografts (Table 2. Figure 4). The OVCAR-3. A'780 and MRI-H-207 tumour lines were borderline sensitive with maximum GI Xalues of 47%7. 41 9% and 40%5 respectively. whereas the FMa tumour line was not sensitive to DNR. At equitoxic doses. the molar amount of DNR-GA3 that could be administered was six-(A2780) to tenfold (OVCAR-3) higher than that of the parent drug. The FMa tumour line wAas not sensitive to DNR-GA3. In three out of four xenogorafts (OVCAR-3. A2780 and MRI-H-207) that were sensitive to DNR. DNR-GA3 induced a maximum GI of approximately 90%7. which was considerably higher than that of DNR (Table 2. Figure 4). The better antitumour effect of DNR-GA3 was also demonstrated in a further 30 35 increase in two tumour v-olume-doubling times. wAhich was sianificant for OVCAR-3 and MRI-H-207 xenografts (P <0.02. Table 2).
DNR-GA3 treatment in large tumours appeared to result in a ays of treatment better inhibition of growth than the same treatment in small mg kg-' on tumours in two out of three tumour lines with sensitivity to DNR. .m.
The SGD increased in OVCAR-3 tumours from 2.1 to 4.4. and in MRI-H-207 tumours from 4.4 to 6.0.
nall and large Dose dependency ing mice was 1.4% and 9.7% To determine whether a higher dose of DNR-GA3 was more ivelv. In mice effective in the treatment of tumour-bearing mice than a lower g-' resulted in dose. mice bearing OVCAR-3 xenoggrafts were injected with this dose was 150 mg kg-' DNR-GA3 weekly x 2 or 200 mg kg-' DNR-GA3 bearinga small weekly x 2. Control groups were treated with DNR weekly x 2. or D.
receixed no treatment. Both prodrug doses were more effectiVe Anthracycline prodrug therapy 1605 than DNR (P<0.01). Also. the higher dose of 2 x 200mg kg-' DNR-GA3 was slightly. but not significantly. more effective than the lower dose of 2 x 150 mg kg-' DNR-GA3 ( Figure 5). Similar data were obtained for the MRI-H-207 tumour line in which 200 mg kg-' DNR-GA3 on day 0 was more effective than the lower dose of 100 mg kg-' DNR-GA3 on day 0 (P < 0.002) (Figure 4 and Table 2).

DISCUSSION
The objective of this study was to investigate the potential increase in the therapeutic index of the glucuronide prodrug DNR-GA3 when compared with DNR. In vitro. the prodrug was 18-fold less toxic than DNR. In mice bearing human ovarian cancer xenografts. the MTD of DNR-GA3 was sixto tenfold higher than that of DNR. The prodrug was apparently activated in the tumours by f-glucuronidase and inhibited tumour growth in human ovarian cancer xenografts that were sensitive to the parent drug DNR. In these three xenografts (OVCAR-3. A2780. and MRI-H-207). the inhibitory effect at MTD was better than the tumour growth delay obtained with the parent drug. The different DNR-glucuronide prodrugs synthesized were designed to be rapidly activated in the presence of human >glucuronidase. DNR-GA3 was most rapidly activated in vitro by human o-glucuronidase (Houba et al. 1996a). and was chosen for in vivo analysis. DNR-GA3 is stable in vivo. it is a hydrophilic molecule that hardly passes through the cell membrane into the cell. DNR-GA3 will. therefore. not be activated by intracellular fglucuronidase. Activation in the circulation is also less likely as the plasma levels of P-glucuronidase are very low (Fishman. 1970). Bosslet et al (1995) and Schumacher et al (1996) have demonstrated high levels of f-glucuronidase in necrotic areas in tumours. Therefore. it could be expected that DNR-GA3 will be activated selectively by human S-glucuronidase released from necrotic tumour cells.
The difference in MTD between DNR-GA3 and DNR in mice bearing human ovarian cancer xenografts may be explained by the more rapid clearance of DNR-GA3 (elimination t,:, = 20 min) than that of DNR (elimination ty. = 720 min) from the circulation. Thus far. we have no information on the nature of the dose-limiting toxicity in mice. We observed. however. the formation of ascites at higher doses of DNR-GA3 as also described for DNR (Boven et al. 1996). The variation in the MTD of DNR-GA3 found among the four different human ovarian cancer xenografts may possibly be clarified by differences in activation and leakage of DNR from the tumours into the circulation.
The treatment expenrments showed that the prodrug DNR-GA3 induced better inhibition of growth in three out of four human ovarian cancer xenografts than equitoxic doses of DNR (OVCAR-3. A2780 and MRI-H-207). This observation may be explained by higher local DNR concentrations in the tumour from activated DNR-GA3. Earlier. it has been demonstrated that there is a steep dose-response curve for anthracyclines (Frei and Canellos. 1980). Bosslet et al (1995) have described that s.c. grown LoVo colon cancer xenografts with a diameter larger than 2 mm had necrotic areas. where S-glucuronidase was present in high concentrations. This group has also demonstrated that a glucuronyl-spacer-DOX prodrug showed better therapeutic effects than DOX.
It was hypothesized that large tumours contain more necrosis and. thus. more $-glucuronidase would be available to activate DNR-GA3. Indeed, we calculated a relatively longer increase in two volume-doubling times for large tumours of the OVCAR-3 and the MRI-H-207 tumour lines when compared with the values of the respective small tumours. With respect to the clinic. this finding is of interest because the treatment of patients with large tumour deposits remains a challenge.
The administration of anthracycline prodrugs to be activated at the tumour site may induce an even better growth inhibition when combined with a second approach: antibody-directed enzyme prodrug therapy (ADEPT: Bagshawe et al. 1988). In ADEPT. prodrugs are activated in the tumour by an administered tumourspecific monoclonal antibody-enzyme conjugate. In our point of view. DNR-GA3 is very suitable for ADEPT. We have shown earlier that a conjugate of monoclonal antibody 323/A3 and human f-glucuronidase bound to tumour cells can activate DNR-GA3 in an efficient manner (Haisma et al. 1992: Houba et al. 1996b). If such a tumour-specific conjugate is administered before DNR-GA3 injection. activation could also occur in the nonnecrotic smaller tumour lesions.
In conclusion. our findings suggest that the glucuronidated anthracycline DNR-GA3 may have a better therapeutic index in advanced solid tumours in which anthracychnes are considered for treatment.