Prodrug applications have been utilized in cancer treatment for improving drug delivery, minimizing toxicity at non-target sites, maximizing therapeutic effect and increasing oral bioavailability [1,2,3]. Amino acid monoester prodrugs of antiviral agents like acyclovir and ganciclovir have been reported to improve their oral bioavailability by facilitating carrier-mediated transport [4,5,6,7]. Amino acid monoester prodrugs of the anti-cancer drug gemcitabine have also exhibited enhanced affinity to the oligopeptide transporter (PEPT1) [8]. The PEPT1 transporter has been extensively examined as a delivery site to improve oral drug absorption and the pro-moieties targeting this transporter have been investigated to improve their binding affinity [9,10,11].

The anticancer agent 5-fluoro-2′-deoxyuridine (floxuridine) has been clinically used in the treatment of colon carcinoma and colorectal cancer metastases to the liver. The mechanical action of 5-fluorouracil (5-FU) and floxuridine is well understood [12]. Floxuridine has a specific activity in DNA and has less cytotoxicity in RNA and, thus, this specificity of floxuridine minimizes adverse effects unlike 5-FU [13,14]. It has been reported that the potency and the growth inhibition of floxuridine is 10- to 100-fold better than those of 5-FU [15,16]. Hence, the protection of the glycosidic bond in floxuridine will produce better clinical outputs. Floxuridine, however, is rapidly converted to 5-FU by thymidine phosphorylase, which is ubiquitously observed in many tissues, including the liver [17]. Therefore, the protection of glycosidic bond of floxuridine leads to improve therapeutic efficiency. A variety of enzymes have been tested as the possible target enzyme to activate prodrugs because those prodrugs have to be activated in order to reveal their therapeutic effects. Irinotecan (CPT-11) is an anti-cancer prodrug of active metabolite SN-38 and has been approved for the treatment of colon cancer. CPT-11 has been investigated for specific activation of irinotecan at tumor sites by carboxylesterase to minimize toxicity in non-tumor sites [3,18,19,20,21]. The progression of technology can make it possible to examine the crystal structure of carboxylesterase and to design more efficient prodrugs activated by specific enzymes [18,22,23]. Highly expressed enzymes have been studied as possible enzyme targets in certain disease stages. The enzyme targeting for prodrug activation in various diseases has also been carried out. It has been reported that endopeptidase cathepsins have been upregulated in tumor cells and are expressed more than those in non-cancer cells [24,25,26]. Furthermore, the signaling pathways related to upregulation of cathepsins have been gradually recognized [27,28]. It is well documented that cathepsins B, D, and L are the most abundant proteases and the increased expression of those cathepsins plays a part in the invasion and metastasis in the tumor growth [29,30,31,32,33,34]. Therefore, cathepsin B, D, and L have a potential to be good targets for prodrug activation. Doxorubicin and daunorubicin prodrugs conjugated with peptides and amino acids show the less toxicity than their parent drugs, doxorubicin and daunorubicin, and are activated by hydrolytic enzymes like the cathepsins [35,36,37,38,39,40]. Capecitabine is a prodrug of 5-FU like floxuridine and is used for the treatment of breast and colorectal cancers. This prodrug, which requires a few activation steps by enzymes to show its therapeutic effect, has been investigated in its activation, distribution, and pharmacokinetics [41,42,43]. However, the avoidance of adverse effects in CPT-11 and capecitabine still has not been fully successful. The understanding of prodrug activation in tumor cells is necessary and leads to the design and production of more advanced prodrugs. Yet, the activation of prodrugs neither has been examined nor has been given adequate attention even though this is one of the most important steps in prodrug approaches.

In this report, the feasibility of cathepsin D targeting for the activation of amino acid/dipeptide monoester prodrugs of floxuridine and the capability of inhibiting cancer cell growth for those activated prodrugs are described. Also, those 5′-dipeptide monoester prodrugs of floxuridine are compared with the 5′-mono amino acid monoester prodrugs of floxuridine. The enzymatic activation of these prodrugs was evaluated to determine the effects of amino acid/dipeptide promoieties and esterification site for enzyme-mediated activation. The activation of translocated prodrugs in tumor cells, the enzymes involved in the prodrug activation, and the ability of tumor growth inhibition for tested prodrugs were discussed.

Prodrug approaches have been investigated to improve oral absorption and therapeutic effects of poorly permeant drugs [44]. Amino acid/dipeptide monoester prodrugs of antiviral and anticancer drugs such as gemcitabine, acyclovir, 2-bromo-5,6-dichloro-1-(β-D-ribofuranosyl)benzimidazole (BDCRB) and floxuridine have been synthesized and characterized in our previous reports [8,45,46,47,48,49,50,51]. The synthesis methods and characterization of those floxuridine prodrugs have been reported [49,50].

Lysosomal proteases called cathepsins may be attractive target enzymes because of high expression in tumor cells. It has been reported that lysosomal cathepsins, especially cathepsin B and D, are upregulated in tumors and redistributed to the outside of lysosomes such as cytosol and plasma membrane [28,52,53,54,55,56,57,58]. Thus, cathepsin B and D in tumors might be good target enzymes even though they are lysosomal enzymes. It has been reported that cathepsin D is present in all tissues but the expression levels significantly vary in the type of tissues and their physiological state [59]. The location and expression level of enzymes are important decisive factors to select possible target enzymes for prodrug activation. For this purpose, enzymes which are located in the plasma membrane and cytosol, as well as highly expressed enzymes in lysosomes, were selected as candidates of prodrug-activation enzymes.

The identification and selection of the specific enzymes for prodrug activation along with designing prodrugs have been investigated in order to minimize adverse effects and improve therapeutic index [60,61,62,63,64]. One of the most common tools for performing high-throughput expression measurements would be a microarray analysis to select target candidates. The Affymetrix GeneChip (GeneChip Human Genome U133 Plus 2.0) was used to obtain the gene expression profile of top 50 enzymes in specific locations, cytosol and plasma membrane as well as highly expressed lysosomal enzymes, of Capan-2 cell. Cathepsin D, puromycin-sensitive aminopeptidase, TPP2, and DPP3 were chosen as candidates of target enzymes due to their locations and high expression levels of gene in Capan-2 cell (Table 1). The experiments concerning identifying activation enzymes for floxuridine prodrugs were performed at 37 °C in Capan-2 cell homogenates in pH 7.4 phosphate buffer with the presence and the absence of selective enzyme inhibitors.

Table 2 displays the estimated half-lives (t_1/2) obtained from the initial rates of hydrolysis by plotting the logarithm of remaining prodrugs as a function of time. All prodrugs showed shorter half-lives in cell homogenates than in pH 7.4 phosphate buffer suggesting enzyme catalyzed hydrolysis. No degradation of prodrugs was confirmed in lower than pH 6.0 phosphate buffer (data not shown). Amino acid monoester prodrugs of floxuridine and their parent drug were observed from 5′-O-L-valyl-L-phenylalanylfloxuridine, 5′-O-L-phenylalanyl-L-tyrosylfloxuridine and 5′-O-L-glycyl-L-leucyl-floxuridine. However, only floxuridine was observed from the rest of tested prodrugs. The stabilities of 5′-O-L-leucyl-L-glycylfloxuridine and 5′-O-L-valyl-L-phenylalanylfloxuridine were 1.4- to 8.3-fold and 0.8- to 1.7- fold better with the presence of enzyme inhibitors, respectively. However, significant differences among them could not be observed. Only 5′-O-L-glycyl-L-leucylfloxuridine exhibited no change in its stability in all conditions. This result is indicating that there are many enzymes to activate 5′-O-L-glycyl-L-leucylfloxuridine or/and selected enzyme inhibitors did not effectively inhibit enzymatic activation of this prodrug. 5′-O-L-phenylalanyl-L-glycylfloxuridine and 5′-O-L-phenyl-alanylfloxuridine exhibited 2.4- to 7.0-fold and 5.1- to 20.2-fold longer half lives with the presence of enzyme inhibitors. This suggests that those targeted enzymes associate to those prodrug activations. However, activation enzymes for those two prodrugs could not be specified. The half life of 5′-O-L-phenylalanyl-L-tyrosylfloxuridine in Capan-2 cell homogenate with the presence of pepstatin A was significantly prolonged, while the half lives of this prodrug with other enzyme inhibitors were not improved. Thus, the activation enzyme of 5′-O-L-phenylalanyl-L-tyrosylfloxuridine would be specific and cathepsin D might be the one of major activation enzymes for this prodrug activation.

The high expression levels of cathepsins in tumors have been reported [65,66,67,68]. It has been investigated if the protein levels of cathepsin B and D would associate with pathological parameters in order to assess the difference in protein levels of cathepsins as possible tumor biomarkers [34,69,70,71]. Cathepsin B, one of the highest gene expressions in enzymes of Capan-2 cell, is a cysteine endopeptidase and reportedly has similar activity to carboxypeptidase and peptidyl-dipeptidase [72,73,74]. On the other hand, cathepsin D has been reported that it preferentially hydrolyzes the peptide bond of bulky hydrophobic amino acids and the peptide bond between Phe-Phe of Bz-Arg-Gly-Phe-Phe-Pro-4MbNA [75,76,77]. The metabolic stability of floxuridine prodrugs was assessed using human cathepsin B and D in pH 7.4 phosphate buffer. Figure 1 shows the stability of floxuridine prodrugs against cathepsin B and D compared with the absence of those endopeptidases, which representing chemical stability of prodrugs. All tested floxuridine prodrugs with exception of 5′-O-L-phenylalanyl-L-glycyl-floxuridine and 5′-O-L-phenylalanyl-L-tyrosylfloxuridine did not exhibit differences in their half lives. Only 5′-O-L-phenylalanyl-L-glycylfloxuridine and 5′-O-L-phenylalanyl-L-tyrosylfloxuridine showed significant degradation by cathepsin B and D. The production of amino acid mono ester prodrug of floxuridine was clearly observed from 5′-O-L-phenylalanyl-L-tyrosyl-floxuridine. However, the enzyme specificity between cathepsin B and D in prodrug activation for those two prodrugs was not observed. As Offermann et al. have reported, aromatic amino acid incorporated prodrugs would be possible substrates for cathepsin B and D but the result showed that 5′-O-L-valyl-L-phenylalanyl-floxuridine was not a good substrate for cathepsins despite the incorporation of phenylalanine in the dipeptide promoiety [76]. The differences in the half lives of amino acid monoester prodrug of floxuridine, 5′-O-L-phenylalanylfloxuridine, was not observed with cathepsin enzymes. Thus, this result indicates that not only N-terminus aromatic amino acid but also dipeptide are required to be a substrate for cathepsins.

In stability studies of test compounds in cell homogenates, the location of enzymes in living cells is not considered and the inaccessible enzymes in vivo condition may be able to degrade those analytes in this in vitro setting. Therefore, the result of prodrug stabilities in cell homogenate may be different to one in vivo, where all enzymes are arranged in all specific locations. The experiments of prodrug activation concerning the location of enzymes in cells were performed at 37 °C in Capan-2 cell homogenates and in cytosolic enzyme extracts. For the comparison purpose, the same experiments were performed at 37 °C in Caco-2 cell homogenates and in cytosolic enzyme extracts. All tested prodrugs except 5′-O-L-phenylalanyl-L-tyrosylfloxuridine and 5′-O-L-glycyl-L-leucylfloxuridine showed 2.0- to 7.4-fold longer half lives in cytosolic extract than ones in Capan-2 cell homogenate in pH 7.4 phosphate buffer (Figure 2). On the other hand, all tested prodrugs except 5′-O-L-leucyl-L-glycylfloxuridine exhibited shorter half lives in cytosolic extract than ones in Caco-2 cell homogenate in pH 7.4 phosphate buffer. All tested prodrugs in Caco-2 cells did not exhibit significant difference between in whole cell homogenate and in the cytosolic extract. The difference between them was minimal (1.4- to 2.5 fold). Floxuridine and amino acid mono ester prodrugs of floxuridine were observed from 5′-O-L-valyl-L-phenylalanylfloxuridine, 5′-O-L-phenylalanyl-L-tyrosylfloxuridine and 5′-O-L-glycyl-L-leucylfloxuridine but only floxuridine was detected from 5′-O-L-leucyl-L-glycyl-floxuridine, 5′-O-L-isoleucyl-L-glycylfloxuridine, 5′-O-L-phenylalanyl-L-glycylfloxuridine, and 5′-O-L-phenylalanylfloxuridine. In Capan-2 cells, the stabilities of 5′-O-L-phenylalanyl-L-glycylfloxuridine and 5′-O-L-phenylalanylfloxuridine were improved 7.4- and 7.3-fold, while the ones of 5′-O-L- phenylalanyl-L-tyrosylfloxuridine and 5′-O-L-glycyl-L-leucylfloxuridine were declined 3.9- and 6.0-fold, respectively. The latter result suggests: (1) there is the same amount of protein but there is difference in the population of enzymes between cell homogenate and cytosolic fraction; (2) there are more enzymes in cytosol extract to activate 5′-O-L-phenylalanyl-L-tyrosylfloxuridine and 5′-O-L-glycyl-L-leucylfloxuridine than ones to activate other prodrugs; and (3) the redistribution/trafficking of upregulated lysosomal enzymes to outside of lysosomal compartments [53,54,55,78]. Those results indicate that the prodrug stabilities in vitro might mislead prodrug stabilities in vivo due to the difference in cellular enzymatic population and location.

The percentage of growth inhibition for amino acid/dipeptide monoester prodrugs of floxuridine in Capan-2 cell was evaluated via 2,3-bis[2-methoxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilideinner salt (XTT) assays (Figure 3). The growth inhibitory effect of tested floxuridine prodrugs after 4 h drug treatment range from 0.0–91.1% at 4 h, 0.0–84.2% at 8 h, and 41.8–84.8% at 24 h in Capan-2 cell lines. 5′-O-L-leucyl-L-glycylfloxuridine and 5′-O-L-phenylalanylfloxuridine immediately inhibited the growth of tumor cells while the rest of floxuridine prodrugs needed a certain time to be activated for their inhibitory effects. While specific activation enzymes for 5′-O-L-leucyl-L-glycylfloxuridine and 5′-O-L-phenylalanylfloxuridine have not been identified, those enzymes may be highly localized at the plasma membrane and cytosol. Thus, the activation of those prodrugs took place immediately after translocation of prodrugs into the cells and started inhibiting tumor growth. Leucine and phenylalanine are reportedly good substrates for aminopeptidases and carboxylesterase, respectively.

Thus, 5′-O-L-leucyl-L-glycylfloxuridine and 5′-O-L-phenylalanylfloxuridine might be activated by aminopeptidases and carboxylesterases, which are highly expressed at the plasma membrane and cytosol, to inhibit the growth of cancer cells [79,80]. 5′-O-L-phenylalanyl-L-glycyl-floxuridine, which exhibited a shorter half life in cell homogenate, did not exhibit the growth inhibition of cancer cells at 4 h and 8 h, suggesting that 5′-O-L-phenylalanyl-L-glycylfloxuridine might be more stable in vivo cell condition than one in cell homogenate. Those results imply that floxuridine prodrugs have to be activated in order to inhibit tumor growth. The prodrug activation would be initiated at different time periods due to the affinity between promoieties and enzymes and the location and population of prodrug-activation-enzymes.

The sites of prodrug activation in cells, the duration for activating prodrugs, the enzyme population, and the activation enzymes based on promoieties are all different. Therefore, in vitro experiments have to be carefully designed and prodrug approaches should be monitored by not only the amount of delivered prodrug but also prodrug activation and their therapeutic effect.

In summary, it is feasible to design prodrugs which are activated by specific enzymes. The prodrug stabilities in cell homogenate do not necessary reflect their stabilities in vivo. Cathepsin D might be a good candidate as a target enzyme for prodrug activation because of its upregulation and redistribution to other cellular compartments in tumor cells and its substrate specificity. The results of stability studies with the presence of enzyme inhibitors indicate that there are particular enzymes activating 5′-O-L-phenylalanyl-L-tyrosyl-floxuridine and 5′-O-L-phenylalanyl-L-glycylfloxuridine. Cathepsin B and D significantly activated 5′-O-L-phenylalanyl-L-tyrosyfloxuridine and 5′-O-L-phenylalanyl-L-glycyl-floxuridine to produce floxuridine. Our studies demonstrate that cathepsin D can significantly contribute to the activation of 5′-O-L-phenylalanyl-L-tyrosylfloxuridine. For tumors that express large amounts of cathepsins, it is likely that a substantial proportion of 5′-O-L-phenylalanyl-L-tyrosyl-floxuridine is hydrolyzed by cathepsin D. Taken together, our experimental results demonstrate that cathepsin D has the significant ability to activate dipeptide monoester prodrug of floxuridine, 5′-O-L-phenylalanyl-L-tyrosylfloxuridine, and, therefore, has the potential to be a target enzyme for prodrug activation in tumors. 5′-O-L-phenylalanyl-L-tyrosylfloxuridine has the capability of being a cathepsin D-targeted prodrug for enzymatic activation and exhibits the respectable growth inhibition of cancer cells. In conclusion, with the consideration of chemical/enzymatic stability of those prodrugs, 5′-O-L-phenylalanyl-L-tyrosylfloxuridine may be the best candidate for in vivo tumor reduction studies because this prodrug would be stable enough to stay longer in the systemic circulation to reach target sites for the activation by the specific enzymes. With careful evaluations, the prodrug approaches designed to ameliorate the toxicity of anti-cancer drugs and to maximize prodrug activation based on enzyme specificity are feasible.