This study presents targeted, drug carrying filamentous bacteriophages (phages) as a drug delivery platform for targeting cancer cells. Our phages represent a modular targeted drug-carrying platform of nanometric dimensions (particle diameter 8 nm, length of a few hundred nm) where targeting moieties, conjugated drugs and drug release mechanisms may be exchanged at will. Specifically, we have generated engineered phages that carried either the drug hygromycin covalently linked to the phage coat, or the drug doxorubicin linked through a cathepsin-B cleavable peptide that was engineered into the major coat protein of the phage. As targeting moieties we used three IgG antibodies; trastuzumab and chFRP5 that target ErbB2 and cetuximab that target the EGFR. When target cells were treated with the targeted drug-carrying phages, selective cell killing could be demonstrated with a potentiation factor of up to several thousand over the free drug.
In our study we used EGFR and ErbB2 as targets; both are very well characterized in the field of targeted anti cancer therapy. In fact, two of the three antibodies we used as targeting moieties are used clinically to treat cancer patients (trastuzumab and cetuximab). All three antibodies we used were shown before to facilitate the delivery of cytotoxic payloads to target cells and tumor models [36
]. In addition, antibody-displaying filamentous phages have been shown to undergo internalization into target cells, which laid the foundation for proposing to use such "internalizing phages" as gene delivery vehicles [26
The drug carrying capacity of the platform is a key issue in its potency. With antibody-drug conjugates, the amount of cytotoxic drug that can be conjugated to the antibody is usually limited by the conjugation chemistry that, if pushed to the upper limit, may reduce its capacity to bind antigen. As a result, such conjugates carry no more than 8 drug molecules per mAb [17
]. Recently more elaborate drug conjugation schemes, such as the use of dendrimers and branched linkers to increase carrying capacity were devised to maximize the drug payload per targeting molecule that binds a target site [44
]. Our phages carry as much as 104
drug molecules/phage which maximizes the intracellular drug load upon internalization of the platform into the target cells.
Considering the linkers used to attach the drug to the targeting moiety, an ideal linker should be stable in the serum and readily degraded within the intracellular milieu. Some examples from the field of antibody-drug conjugates are acid-labile linkers and enzyme-cleavable linkers [4
]. We chose to evaluate two approaches; direct covalent conjugation of the drug to the carrier and conjugation of the drug through an engineered cathepsin-B cleavage site. Our results (Fig. ) show that both approaches are viable, but the results may vary with different drugs and/or targeting antibodies. Phages that were used to deliver covalently linked hygromycin to ErbB2 expressing cells (chFRP5 IgG as the targeting moiety, Fig. ) could cause specific cell growth inhibition. When phages were used to deliver doxorubicin to either erbB2 expressing cells (trastuzumab as the targeting moiety, Fig. ) or EGFR expressing cells (cetuximab as the targeting moiety, Fig. ), phages to which the drug was linked covalently were inefficient in inhibiting cell growth, while phages that carried the cathepsin-B releasable drug were more efficient, suggesting that with this particular phage-drug combination, an engineered drug release mechanism is necessary to maximized potency. We could not link hygromycin to DFK-displaying phages, since we found that hygromycin with a single amino-acid adduct (as is the product of cathepsin-B release drug in our system) is inactive as a drug (data not shown).
Early antibody-drug conjugates were comprised of a monoclonal antibody covalently linked to several molecules of a clinically used anti-cancer drug. The linker connecting the antibody and the drug was either non-cleavable or cleavable upon entry into the cell. In the early development phase of antibody-drug conjugates, it was believed that the tumor specificity of anti-cancer drugs could be improved merely by linking these drugs directly to antibodies via amide bonds [17
]. In most cases, the conjugates lacked cytotoxic potency and were less potent than the un-conjugated drugs [47
]. Only in the past few years the critical parameters for optimization have been identified and have begun to be addressed. These include low drug potency, inefficient drug release from the mAb and difficulties in releasing drugs in their active state [44
]. On the basis of this much research has been focused on designing new linker technology. The use of peptides which are susceptible to enzymatic cleavage, as conditionally stable linkers for drugs to mAbs. The peptides are designed for high serum stability and rapid enzymatic hydrolysis, once the mAb-drug conjugate is internalized into lysosomes of target cells. For example cathepsin-B sensitive peptides. Cathepsin-B is a cysteine protease found in all mammalian cell lysosomes. The cathepsin-B cleavable di-peptide Phe-Lys was used for conjugating doxorubicin to BR96 mAb which were previously conjugated through a hydrazone labile linker [48
]. The resulting immunoconjugate showed levels of immunological specificity that had been unobtainable using the corresponding hydrazone-based conjugates.
The objective of the experiments described herein was a feasibility study of applying targeted drug delivery as an anti cancer tool. The system was designed on three key components: 1) a targeting moiety, exemplified here by various monoclonal tumor specific antibodies complexed via the ZZ domain [34
]. 2) A high-capacity drug carrier, exemplified here by the filamentous phage, with its 3000 copies of major coat protein, each amendable to drug conjugation. 3) A drug linked directly or through a labile linker that is subject to controlled release, exemplified here by hygromycin conjugate directly or by doxorubicin that was linked through a cleavable peptide expressed on all copies of phage major coat protein. In the case of covalently-linked hygromycin, we postulate that a partial non selective release in the lysosomes post internalization would eventually lead to a specific killing of target cell. Several features led us to use hygromycin as the model drug: The first was the simplicity in which hygromycin can be conjugated to the phages through simple EDC chemistry. Hygromycin has two primary amino groups, one for phage conjugations and the other for drug or analyte conjugation (such as FITC as we have report previously) [34
]. Another important feature at this stage is the high drug solubility in water. With this chemistry, a carrying capacity in excess of 104
drug molecules/phage was previously reported by us [34
The second example was a controllable release mechanism that was genetically engineered into the phage major coat protein g8p (p8). We mutated the N-terminus of p8 to express a cathepsin-B cleavage peptide with the sequence of DFK [48
]. Aspartate (D) was added to the sequence FK for the creation of two options for chemical conjugation; through the α-amine or through the carboxyl side chain. In addition to this mutation, the native aspartic side chains were mutated to non carboxyl side chains (asp 11 (Fig. ) was retained since it is buried within the phage coat an inaccessible to conjugation [49
]). The native lys8 was mutated to glu7 in the newly mutated p8 and used as internal control for drug conjugation and to maintaining balanced number of charged residues that are important for phage solubilisation. Indeed, from Fig. one may appreciate that there was partial release of doxorubicin upon cathepsin-B treatment, since doxorubicin molecules linked to glu7 were not released. Since two of the native carboxyl residues asp4 and asp5 were deleted it is important to note that by this genetic modification in the structure of the major coat protein-8 we have reduced the potential drug capacity by more then 60%.
Doxorubicin was used as a model drug; primary by two reasons; the first is its reporter properties, fluorescence as well as specific emission in the wavelength of 480 nm. This property was helpful for monitoring of drug release. The second is the relative tolerance for conjugation of linkers into the single primary amine located to the aminoglycoside ring tailored specifically for the solubilization of this highly hydrophobic drug. Doxorubicin was conjugated to phages through EDC chemistry, resulting with reddish solution. The releasing experiment with commercial cathepsin-B led to a complete release of all connected doxorubicin molecules. Each DFK phage release about ~3500 doxorubicin molecules as we measured by a specific reading at 480 nm with a reference of a calibration curve of free doxorubicin. This results correlates with the maximal theoretical capacity of this phages. Using HPLC analysis, we found that the release was limited to the DFK phage only, while doxorubicin linked covalently to the wild-type phage coat was not released. Further MALDI-TOF MS analysis showed that the released moiety was doxorubicin with aspartate linked to it. Such adducts are common when labile linkers are used do deliver drugs, and in the case of doxorubicin, do not seem to inactivate it. This is similar to the released drug of the "non-cleavable" antibody-drug conjugates where upon degradation in the lysosomal compartment, the drug remains covalently linked to a single amino acid, either lysine for maytansinoid conjugates or cysteine for auristatin conjugates [50
Internalization of the phages, unconjugated or armed with drug could be demonstrated by fluorescence confocal microscopy (Fig. ). The result of the target cell killing assays showed that the soluble drug hygromycin connected through a non cleavable stable amide bond, directly to phage coat carboxyl residues, could achieve impressive potency improvement, in factor of >1000 over the free drug (Fig ) (5 × 1010 phages carrying 104 drug molecules/phage, carry 0.43 μg free drug, which inhibits cell growth as well as ~1 mg of free drug). This occurred although poor drug release within the cells could be expected.
The interpretation of the results of doxorubicin-carrying phages is more complex, since the goal of this system was a proof of new concept for the construction of cleavable linkers by genetic engineering instead of conventional organic chemical linkage. Our results show the DFK peptide to be specifically cleaved by cathepsin-B, specifically at the engineered site (Fig. ) and within the target cells. Further, doxorubicin-carrying DFK phages (Fig. and , white bars) are more potent in comparison the phages that carry covalently-linked doxorubicin (Fig. and , grey bars) even though the latter carry 10000 drug molecules/phage while the former carry 3500 drug molecules. Moreover, phages to which doxorubicin was covalently linked, although they were target-specific, inhibited cell growth less efficiently than non-specific, human IgG linked DFK phages, further demonstrating the contribution of the engineered drug-release mechanism to the potency of the platform. As for the limited specificity of doxorubicin-carrying DFK phages, we have already observed that coating phages with a high density of hydrophobic molecules limits the solubility of the phages [34
] and cases them to become "sticky", that is, to bind non-specifically to bacteria and to cells (unpublished results). The results shown in Fig. and suggest that this may also be the case with doxorubicin-carrying phages, since comparably levels of cell-killing could be observed when that phages were linked to the targeting antibodies, or to the irrelevant control, normal human IgG. Such non-specific killing could not be observed with hygromycin-armed phages (Fig ). Doxorubicin is known as a drug of which doses are limited by unwanted toxicity to non-tumor tissues [53
]. Doxorubicin and other anthracyclines are amphiphilic molecules known to interact with cell membranes [54
], which may cause non-specific binding. Large polymer-doxorubicin conjugates were reported as having limited solubility [55
], which may also affect target specificity. One may concluded that drug delivery platforms that carry the drug on the outside will be limited to highly water soluble drugs that do not bind non-discriminately to cells. However, a remedy to this limitation may be found in linking the hydrophobic drug to the phage through a solubility-enhancing linker, as we have recently reported [34
]. We have shown that the potency and the target specificity of anti bacterial chloramphenicol-armed phages was substantially improved when this hydrophobic drug was linked to the phage coat through small hydrophilic molecules that served as solubility-enhancing linkers. On the other hand, phages that were directly armed with the hydrophobic drug chloramphenicol were less specific [34
]. An additional advantage of such an added hydrophilic coat is that it reduces both the immunogenicity and antigenicity of the drug-carrying phages upon injection into mice (unpublished data).