Anti-tumour effects of an antibody-carboxypeptidase G2 conjugate in combination with phenol mustard prodrugs.

ADEPT is an antibody-based targeting strategy for the treatment of cancer. We have developed two new prodrugs, 4-[N,N-bis(2-chloroethyl)amino]-phenoxycarbonyl-L- glutamic acid (PGP) and (S)-2-[N-[4-[N,N-bis(2-chloroethyl)amino]- phenoxycarbonyl]amino]-4-(5-tetrazoyl)butyric acid (PTP), which are cleaved by the bacterial enzyme CPG2 to release the 4-[N,N-bis(2-chloroethyl)amino] phenol drug. In vitro, both prodrugs are approximately 100- to 200-fold less potent than the parent drug (1 h IC50 = 1.4 microM) in LoVo colorectal tumour cells. These prodrugs have been evaluated for utility in ADEPT when used in combination with a conjugate of CPG2 and the F(ab')2 fragment of the anti-CEA monoclonal antibody, A5B7. The conjugate was shown to localise specifically to established LoVo tumour xenografts growing in nude mice and optimal tumour-normal tissue ratios were achieved after 72 h. Administration of either prodrug, at doses which cause 6-8% body weight loss, 72 h after administration of the A5B7-CPG2 conjugate to the LoVo tumour-bearing mice resulted in tumour regressions and growth delays of 14-28 days. The PTP prodrug in combination with a high dose of conjugate (10 mg kg-1) gave the best anti-tumour activity despite being a 10-fold worse substrate for CPG2 than PGP. Prodrug alone, active drug alone or prodrug in combination with a non-specific conjugate had minimal anti-tumour activity in this tumour model.

The first step involves the administration of a tumourselective antibody linked to an enzyme. Following tumour loclisation of the conjugate and clearance of conjugate from blood and normal tissues, the second step involves administration of an inactive prodrug which is converted by the targeted enzyme at the tumour site into a potent cytotoxic drug. ADEPT has a number of potential advantages over other forms of antibody-based targeted therapy. Each enzyme molecule can potentially convert a large number of prodrug molecules into drug molecules at the tumour site. Potency limitations of directly targeting a cytotoxic drug molecule to a tumour with an antibody caused by the fact that only a limited number of drug molecules can be directly attached to an antibody without compromising tumour targeting (Thorpe, 1985) are thus overcome by ADEPT. A second advantage is that a small molecular weight cytotoxic drug is generated locally at the tumour site and outside the cell. The drug should be capable of diffiusion to reach tumour cells not directly targeted by the conjugate either owing to antigenic heterogeneity (Woodruff, 1983) or because the high molecular weight conjugate has failed to diffuse away from the tumour vasculature and reach tumour cells distant from the blood supply (Jain, 1989).
We have previously reported on an ADEPT system which utilises the bacterial enzyme carboxypeptidase G2 (CPG2) (Bagshawe et al., 1988;Springer et al., 1991a). CPG2 has no known mammalian equivalent and catalyses the hydrolytic cleavage of reduced and non-reduced folates to pteroates and L-glutamate (Minton et al., 1983). The amide bond in the glutamate derivatives of benzoic acid mustards are also cleaved by CPG2 to release the benzoic acid mustard drug. A series of studies has shown that the combination of these benzoic acid mustard-glutamate prodrugs and a conjugate of CPG2 linked to either the anti-$-hCG antibody, W14 (Bagshawe et al., 1988;Springer et al., 1991a) or the anti-CEA antibody, A5B7 (Blakey et al., 1993) results in signficant anti-tumour activity in a number of tumour models. Clinical trials are ongoing with the 44(2chloroethyl)2-mesyloxyethyl)amino]-benzoyl-L-glutamic acid prodrug (CMDA) and an F(ab%A5B7-CPG2 conjugate in patients with advanced colorectal cancer Bagshawe, 1993).
Two potential limitations of these ADEPT systems incorporating benzoic acid mustard prodrug are firstly that the drugs generated are not very potent (Springer et al., 1991b;Blakey et al., 1993) and secondly that the drugs have relatively long chemical (60 min) (Springer et al.,199 lb) and biological (1.6 h) half-lives  which potentially permit escape of the drug from the tumour site into the periphery resulting in enhanced non-specific toxicity.
In this paper we describe the evaluation of two new prodrugs which are cleaved by CPG2 to release a potent and highly reactive phenol mustard drug. When compared with the original benzoic acid mustard prodrugs described previously (Bagshawe et al., 1988, Springer et al., 1991aBlakey et al., 1993) these prodrugs when used in combination with the F(ab%A5B7-CPG2 conjugate result in improved anti-tumour activity in a colorectal tumour xenograft model.

Matedak and methods
The prodrugs used in these studies are 4-tN,N-bis(2-chloroethyl)aminoJ-phenoxycarbonyl-L-glutamic acid (phenol glutamate prodrug; PGP) and (S)-2{NN4-[N,N-bis(2-chloroethyl)amino]-phenoxycarbonylJamino]-4-(5-tetrazoyl)butyric acid (phenol tetrazole prodrug; PTP). The structure of these prodrugs and the corresponding phenol drug are shown in Figure 1. Their synthesis will be described elsewhere ( Methotrexate was obtained from Sigma, Poole, Dorset. Carboxypeptidase G2 from Pseudomonas sp strain RS16 was cloned into Escherichia coli (Minton et al., 1983) and prodneed as described by Sherwood et al. (1985). A5B7 antibody (IgGI) which reacts with the carcinoembryonic antigen (CEA) was kindly supplied by the Department of Medical Oncology, CRC Laboratones, Charing Cross Hospital (Harwood et al., 1986). MOPC-21 (IgGl) control antibody which has no known tissue reactivity was generated from the hybridoma cell line P3X63Ag8 acquired from the European Collection of Animal Cell Cultures (ECACC no. 85011401). The F(ab% fragment of A5B7 and MOPC were prepared by papain digestion. For A5B7 and MOPC enzyme-antibody ratios of 1:70 for 24 h at 3TC and 1:20 for 4 h at 3TC were used respectively. The F(ab% fragments were conjugated to CPG2 as described previously by Melton et al. (1993). Both conjugates had sp. act. of 150-200 U CPG2 mg'I protein and predominantly consisted of one F(ab% fragment of A5B7 conjugated to one CPG2 molecule. The molkular weight of these conjugates is thus approximately 180 kDa. One unit of CPG2 activity corresponds to the amount of enzyme required to hydrolyse 1 pmnol of methotrexate min' ml-'reacion mixture at 3TC (Sherwood et al., 1985).
The LoVo colorectal tumour cell ne was obtained from the European Collection of Animal Cell Cultures (ECACC no. 87060101).

E;nzyme kinetics
The K. and k,. of the prodrugs for CPG2 was determined using a method based on the literature CPG2 assay method for methotrexate (Sherwood et al., 1985). The absorbances of prodrug and corresponding drug were scnned from 200 nm to 350 nm using a spectrophotometer (Perkin Elmer Lambda 2) and the wavelength where the maximal absorbance difference (due to cleavage of the carbamate linkage) tween prodrug and drug was determined. For PGP and PTP this was 258 and 259nm respectively and the A2m, for PGP= 16.1 x 103M-'cmand theAs2m for PTP = 15.3 x 103m-1 cm-'. The Km and V.. were determined by measuring the initial rate of conversion of prodrug to drug at these wavelengths using a range of prodrug concentrations (1-100IM) and CPG2 enzyme concentrations (0.05-1 U).
The k,, was calculated from the V.= by dividing by the amount of CPG2 in the reaction mixture.

Cytotoxicity Studies
The colorectal tumour cell line LoVo (CEA positive and A5B7 reactive) was incubated with prodrug, prodrug phls CPG2 (1.0 U per well) or drug, in 96-well (2500 cells per well) microtitre plates for 1 h. The cells were then washed and incubated for a further 3 days at 3TC. Trichloroacetic acid (TCA) was then added and the amount of cellular protein adhering to the plates was assessed by the addition of SRB dye according to Skehan et al. (1990). Potency of the compounds is expressed as the concentration required to inhibit cell growth by 50% (IC,4).

Tumour localisation studies
The F(ab%A5B7-CPG2 conjugate was radioiodinated with carrier-fiee iodine-125 using the lodogen reagent, following the manufacturer's rm ed method. In vitro retention of >70% immunoreactivity after radioiodination was confirmed by binding to LoVo cells using the method of Lindmo et al. (1984). Approximately 10 1g of conjugate containing 10 pCi'I was injected i.v. into athymic nude mice [nu/nu:Alpk(outbred)] bearing established LoVo tumour xenografts (1 x I07 LoVo tumour cells injected s.c. 7 days previously). Following injection of the conjugate, groups of three mice were killed 4, 24, 72 and 168 h later. The tumour, a sample of blood and a range of other tissues were removed, weighed and counted in a gamma counter.

Toxicity studies
Groups of five female Alderley Park mice were injected with prodrug (three doses at 1 h intervals) i.p. 3 days after injection of F(ab%A5B7-CPG2 conjugate (500 or 200 U CPG2 kg-' i.v.) Body weight and condition of the mice was monitored daily. A dose of prodrug was established which caused 15% body weight loss (defined as the MTD in these Phe musvd -DC Blakey et at studies) on at least 1 day after treatment. Body weight loss was generally maximal at days 2-3 and mice generally regained body weight by day 7 after treatment with prodrug.
Preliminary studies were carried out with groups of two mice to establish the approximate MTD thus minimising the number of animals required to obtain an accurate MTD. The toxicity of conjugate and either PGP or PTP prodrug in normal and tumour-bearing athymic nude mice were not significantly different.
Anti-twnour studies Groups of 8 -10 female athymic nude mice were injected s.c. with 1 x 10' LoVo tumour cells. When the tumours were 4 -5 mm in diameter F(ab')2A5B7-CPG2 conjugate (500-2000 U CPG2 enzyme activity Kg-') or phosphate-buffered saline (PBS), was injected i.v. Seventy-two hours later prodrug was injected i.p. (three doses at 1 h intervals). The length of the tumours in two directions was then measured three times a week and the tumour volume calculated using the formula: where D is the larger diameter and d is the smaller diameter of the tumour.
Tumour volume was expressed relative to the tumour volume at the time of initiation of the prodrug arm of the therapy. At this stage tumours measured 7-8 mm in diameter and had a calculated weight (assuming a density of 1.0) of approximately 0.2-0.3 g. The anti-tumour activity was compared with control groups given PBS instead of either conjugate or prodrug. Other groups of tumour-bearing mice received F(ab')2MOPC-CPG2 control conjugate followed by prodrug or they were given prodrug or phenol mustard drug alone at the same time as the prodrug was administered in the combination arm of the study. Toxicity was monitored throughout the studies by measuring body weight and monitoring the condition of the animals. Statistical significance of the anti-tumour effects was judged using the analysis of variance (one-way) test (Armitage and Berry. 1987).

Results
In vitro properties ofprodrugs The PGP and PTP prodrugs both contain a carbamate linkage between the phenol mustard drug and either a glutamic acid moiety in the case of PGP or an analogue of glutamic acid containing a tetrazole unit in the case of PTP. They were both good substrates for CPG2 and were cleaved to release the phenol mustard drug. The K. and k, for PGP and PTP are shown in Table I. PGP has a low K, which is 5to 10-fold lower than either PTP or methotrexate (the standard substrate for CPG2). Consequently, although both PGP and PTP have similar kr values for CPG2, the turnover number (k/.K^,) for PGP is approximately 10-fold higher than for PTP. This demonstrates that PGP is more efficiently converted to drug by CPG2 than is PTP. The cytotoxicity of PGP and PTP in vitro in LoVo colorectal tumour cells and the corresponding phenol mustard drug released following cleavage by CPG2 are shown in Figure 2a and b. PGP and PTP had IC50 values of 254 and 175 lM respectively (mean value of at least six separate studies) after a 1 h incubation with LoVo colorectal tumour cells. Both prodrugs were some 100to 200-fold less cytotoxic than the corresponding phenol mustard drug (IC50 = 1.4 pM). Addition of I U of CPG2 to either PGP or PTP for 1 h in vitro resulted in an equivalent cytotoxicity in LoVo cells to the active drug (Figure 2a and b) thus confirming the ability of CPG2 to catalyse the release of active drug from either prodrug.
Localisation of F(ab')2A5B7-CPG2 to LoVo tumour xenografts The ability of the F(ab')2A5B7-CPG2 conjugate to localise to LoVo tumour xenografts in nude mice was evaluated by radioiodinating the conjugate and measuring the tumour, blood and normal tissue levels after injection into nude mice bearing established (approximately 0.2-0.4 g) LoVo tumour xenografts. LoVo tumour cells express CEA and in vivo, as judged by immunohistology, approximately 60% of the LoVo cells react with A5B7 in the tumour xenografts (results not shown). The ability of the conjugate to localise to these LoVo tumour xenografts is shown in Figure 3. After 24 h approximately 2.5% of the injected dose of conjugate was present per g of tumour but at this time point there was more conjugate (3.2%) present in the blood. After 72 h approximately 1% of the injected dose was present per g of tumour and this level now exceeded the amount present in blood and normal tissues by a factor of 3 and 10to 50-fold respectively. Levels of conjugate in all normal tissues examined (liver, kidney, stomach, lung, skin) were less than Al 01-4 --d DC pbkedi-et OVA~~~~~~~~~~~~~~~D C Blakey et af prodrug following administration of either conjugate dose. Blood enzyme levels were determined 72 h following administration of either 500 or 2000 U kg-' conjugate and were found to be 0.05 and 0.2 U ml-' plasma respectively.
The anti-tumour effects of the PGP and PIP prodrugs in LoVo tumour xenografts are shown in Figure 4a and b and summansed in Table III. Both the 500 U and 2000 U kg-' conjugate dose levels were evaluated in combination with a prodrug dose which represented half the MTD (Table HI)  only small amounts of conjugate remained in the tumour, blood and normal tissues. The level of conjugate and the tumour-blood ratio at 72 h did not vary over a conjugate dose range of 100-2000 U kg-' F(ab')A5B7-CPG2 conjugate (results not shown). Based on these data a time interval of 72 h between conjugate and prodrug administration was chosen for the anti-tumour studies.
Anti-tmnour activity A prodrug dosing regimen of three doses hourly over a 2 h time period was selected since both the PGP and PTP prodrugs have relatively short biological half-lives in mice of 20-30 min (results not shown) and it was thought that this dosing regimen might optimise exposure of prodrug to enzyme at the tumour site. Two conjugate dose levels were evaluated (500 and 2000 U CPG2 kg-1 corresponding to approximately 2.5 and 10 mg kg-' of total protein). The dose of prodrug, in combination with these conjugate dose levels, that caused 15% body weight loss was established, and the results are shown in Table II. Based on body weight loss, the PGP prodrug was 7to 8-fold more toxic than the PTP    'T/C, the volume of the treated tumour/volume of the control tumour 14 days after prodrug administration. bGrowth delay is the time it takes treated tumours to increase their volume by 4-fold minus the time it takes control tumours to increase their tumour volume 4-fold. cData are mean values of two separate studies. 'Data are mean values of three separate studies with standard deviations in parentheses. of 14-28 days compared with control tumours. The most effective regumen was the FTP prodrug combination with the 200 U kg-' conjugate dose. FP prodrug was significantly (P<0.05) more active in combination with the 2000 U kg-I conjugate dose than with the 500 U kgconjugate dose. The activity of PGP with either conjugate dose was not significntly different.
If PGP or PTP were adminisred in the absence of conjugate or the active phenol mustard drug was administered (3 x 2mg kg-') at a dose which caused 7-8% body weight loss, growth delays of less than 5 days were seen (Figure 4).
Similarly, if the F(ab%A5B7-CPG2 conjugate was replaed with a control F(ab)2MOPC-CPG2 conjugate which does not bind to LoVo cells, growth delays with either PGP or PTP were seen of leIs than 5 days (Figure 4). The plasma enzyme levels of the F(ab%MOPC-CPG2 conjugate were the same as those with the specific F(ab)A5B7-CPG2 conjugate at the time of prodrug administration (72 h) in these therapy studies. Thus to achieve tumour regrssions and growth delays>5 days, the combination of both the specific conjugate and prodrug was required.
The major finding to emerge from these studies is that we have been able to develop two new prodrugs of a potent phenol mustar drug which in combination with an F(ab%A5B7-CPG2 conjugate result in pronounced antitumour activity in a colorectal tumour xenograft model. The enzyme used in this ADEPT system, CPG2, is a bacterial carboxypptidasewith specificity for cleavage of amie bonds with a C-terminal glutamic acid residue. Previously it has been demonstrated that CPG2 is capable of cleaving amide bonds in both methotrexate (Sherwood et al., 1985) and benzoic acid musard-glutamate prodrugs (Bagshawe et al., 1988;Springer et al., 1991a). Surprisingly, both PGP and PIP are very good substrates for CPG2 despite the fact that they contain a carbamate linkage (Table I).
The phenol mustard drug liberated from PGP andFTP by CPG2 is some 50to 100-fold more potent than the benzoic acid mustard drug liberated from the CMDA prodrug (Springer et al., 1991b;Blakey et al., 1993) which is currently in chnical trials combination with the F(ab%A5B7-CPG2 conjugate (Bagshawe et al., , 1995Bagshawe, 1993). The increase potency along with retention of good enzyme kinetics should mean less conjugate is required at the tumour site for the conversion of sufficient prodrug to drug to result in anti-tumour activity. In addition, the drug generated has a very short chemical half-life of approximately 5 min in buffer at 3TC (RJ Dowell et at., unpublished data) when compared with drug generated from CMDA which has a half-life of approximately 60 min (Springer et al., 1991b). This reduction in half-life should have the advantage that drug generated within the tumour will be less able to escape into the periphery and cause toxicity.
Attachment of either the acid or tetrazole glutamic acid residue via the arbamatelink to the phenol mustard reduces it cytotoxic potency by greater than 100-fold ( Figure 2). Sine the chemical rctivity and thus intrinsic alkylating activity of the mustard arms is only reduced by approximately 6-fold in both PGP andPTP (RI Dowell et al., unpublished data; DH Davies et al., unpublshed data) it seems likely that the major reason for the drease in cytotoxicity is that the anionic groups in theglutamate and tetrazole residues decrease the rate of uptake of the prodrug by the cells. However, we have no direct data to confirm this hypothesis.
The localisation of the F(ab%A5B7-CPG2 conjugate to LoVo tumour xenografts (Figure 3) is similar to that reported previously in LS174T colorectal tumours (Blakey et al., 1993). The opfimal time for prodrug administration is a balance between reaining sufficient conjugate at the tumour site to activate sufficient drug for tumour cell ki vs ensuring that the level of enzyme in the blood and normal issues does not cause excessive toxicity. We have been able to define a time interval when this criteria is met as judged by the anti-tumour results. In other instan, including the combination of CMDA prodrug and F(ab%A5B7-CPG2 conjugate, the use of a claring system has been used to improve anti-tumour activity (Sharma et al., 1990. The claring systms accelerate the blood cleance of the conjugate, enabling prodrug to be administered when tumour enyme levels are higher. The major disadvantage of using a clearing system is that it adds an additional level of complexity to the ADEPT approach.
Significantly more PTP than PGP prodrug could be administered to mice 72 h after conjugate administration.
Since both prodrugs release the same active drug, have similar cytotoxicties in vitro ( Figure 2) and have similar pharmacokinetics in mice (unpublished data) it seems likely that this is due to the fact that PTP is approximately a 10-fold kss effective substrate for CPG2 than PGP as judged by the turnover number (k,,,/K.). Residual enzyme levels in the blood and normal tiss is likely to convert less PTP prodrug to drug than is the case with the PGP prodrug and consequently more FTP prodrug can be ani . While PGP produces similr anti-tumour activity at both the 500 and 2000 U kgconjugate dose, the anti-tumour activity of PTP was signiiantly better with the 2000 compared with the 500 U kg-' conjugate dose in two separate stuies. These data support the suggeston, based on computer modelling, that for optimal seectivity a prodrug with relatively poor enzyme kinetics may be more favourable for ADEPT (Yuan et al., 1991). The rationale for this is that such a prodrug would optimise the usage of conjugate localsed at the tumour site. A prodrug which is a very good substrate for the enzyme may be rlipidly tuned over in the periphery by even small quantities of residual enzyme, thus reducing prodrug levels that can reach the tumour to be converted by targeted enzyme.
In the absence of a cearing system the original CMDA prodrug in combination with F(ab%A5B7-CPG2 in either LoVo (unpubished data) or LS174T colorectal tumour xenografts (Blakey et al., 1993) only gives 8-to 10-day growth delays and little evidence of tumour regreison at doses of prodrug which cause similar toxicity to those seen in the therapy studies reported here. Thus by increasing the potency and reactivity of the drug released and altering the enzyme kinetics of the prodrug for CPG2 we have been able to improve the original CMDA prodrug in terms of antitumour actity in colorectal xenografts using a two-step ADEPT approach.
Previously, Wallace and Senter (1991) reported an ADEPT system which incorporated a different prodrug of the phenol mustard drug used in these studies. The prodrug wasp[NVN-Bis(2chothyl)aminonyl phosphate (POMP) and the enzyme used to cleave the phosphate residue to release the drug was alkaline phosphatase. In a lung tumour xenograft model (H2981) the combination of antibody-alkalne phosphatase conjugate and POMP rulted 10to 15-day growth delays with little evidence of tumour regressions. The activity of this system was probablylimited by endogenous alkaline phosphatase causing conversion of prodrug to drug and so enhancing toxicity. Since CPG2 is a bacterial enzyme and no active drug has been detected following the administration of the CMDA prodrug to patients in the absence of conjugate (Bagshawe, 1993;Bagshawe et al., 1995) the presence of endogenous enzyme is unlikely to be a problem with the ADEPT system described in these studies. No phenol mustard drug was detected after administration of eitherFTP orPGP to mice in the absence of conjugate (unpublished data).
The anti-tumour effects ofPNP and TPin combination with F(ab%2A5B7-CPG2 were dependent on targeting of the conjugate to the tumour. A control conjugate prepared with the MOPC antibody which does not bind to LoVo tumour cells when used in combination withPGP or PTP resultedin little anti-tumour activty. This was despite the fact that the plasma enzymeevels of the two conjugates were the same at Ph" dudt o-F M DC Blakey et at 1i8c 72 h. Similarly. prodrug alone or the phenol mustard drug resulted in little anti-tumour activity. In conclusion, we have developed a new ADEPT system which incorporates prodrugs of phenol mustard drug and the conjugate F(ab')2A5B7-CPG2. This ADEPT system results in tumour regression and significant tumour growth delays in a colorectal tumour xenograft model demonstrating its potential for treatment of colorectal cancer.