With the present study, we establish the feasibility of peptide ligand insertions into the Ad41s fiber knob domain resulting in molecules that retain both fiber/capsid integrity and the potency of the inserted peptide to bind its cellular receptor. Our study represents, to the best of our knowledge, the first report on internal and functional ligand insertions into an adenovirus fiber other than the Ad5 fiber and into loops different from the HI loop. A previous effort for functional insertion of the CDCRGDCFC ligand into the HI loop of the adenovirus serotype 3 fiber (in the context of the long Ad5 shaft and tail domains) was unsuccessful (10
), demonstrating structural constraints of adenovirus fiber molecules for internal peptide insertions. With our data, we could also demonstrate that functional insertion of a ligand into loops of short shafted adenovirus fibers is feasible.
The realization of selective adenovirus cell entry into target cells represents a critical step towards the implementation of adenoviral therapeutics in gene transfer, vaccination, and oncolysis. This goal requires the development, in parallel, of two modifications to the tropism of the most commonly used adenovirus serotype, i.e., serotype 5. First, cell entry into normal, nontargeted cells that express the Ad5 receptor CAR, most prominently hepatocytes, needs to be abolished. Second, targeting ligands need to be incorporated into the virus capsid without a loss of ligand function or deleterious effects to the virus structure. Recently, the replacement of the Ad5 fiber with the short fiber of subgroup F adenovirus Ad40 or Ad41 has been shown to achieve substantially reduced transduction of the liver and other organs after intravenous injection into mice and rats (25
). Therefore, the short fibers of Ad40 and Ad41 represent promising “detargeted” platforms for retargeting of adenovirus cell entry by insertion of peptide ligands.
Our study shows that peptide insertion into the EG, HI, or IJ loop, but not into the AB or DE loop, of the Ad41s fiber knob retains the fiber's essential properties of trimerization and incorporation into adenovirus capsids. Notably, these results demonstrate that it is possible to insert peptides into the Ad41s fiber knob in such a way that they face the top (IJ loop) or the side (EG and HI loops) of the fiber molecule (Fig. ). In this regard, it is important that adenoviruses can bind to their cellular receptors either via residues in the side (Ad5) (19
) or via residues in the top (Ad37/Ad19p) (11
) of their fiber molecules. The inability of F5/41s with a DE loop insertion to form trimers can be explained by direct steric hindrance of the inserted amino acids, because this loop is located at the interface of fiber monomers. Also, according to the crystal structure of the short Ad41 fiber knob (36
), which was published after we designed and produced our constructs, the DE region contains an alpha-helix, not a flexible loop. In contrast, the AB loop is located on the fiber surface. The AB loop of the Ad5 fiber is involved in CAR binding but is structurally different in the Ad41s fiber knob (33
). We hypothesize that the insertion of peptides into the AB loop of the Ad41s fiber knob interferes with folding of the fiber monomer and thus, indirectly, with fiber trimerization. We found that C-terminal fusion of peptide ligands to the Ad41s fiber knob via differently sized linkers (8 or 16 amino acids) does not interfere with fiber trimerization. This result is in line with previous studies that show the feasibility of C-terminal extensions of the Ad5, Ad3, and Ad40s fiber knobs (25
All of our recombinant fibers, with the exception of F41sCLL-RGD, contained the Ad5 fiber tail and the Ad41 short fiber shaft. Thereby, we sought to guarantee the incorporation of recombinant fibers into Ad5 capsids but also to retain a short-shafted fiber. To detect fiber molecules in Western blots, we used the monoclonal anti-fiber antibody clone 4D2 (16
), which binds to a highly conserved N-terminal motif (17
) of adenovirus fiber molecules that is also found in the Ad41 short fiber. Although we were unable to detect trimeric F41sCLL-RGD in Western blots of cell lysates after transient transfection, we observed a weak band for this fiber in Western blots of purified AdβGF41sCLL-RGD. Together with the transduction data for this virus, our results indicate that the F41sCLL-RGD fiber can trimerize and is incorporated into virus particles, but with less efficiency than fibers with the Ad5 fiber tail. These results were not expected in light of the sequence homology between the Ad41 short fiber and Ad5 fiber tail domains and considering our data that show no interference of the CLL-RGD fusion with trimerization of the F5/41s fiber. Moreover, previous reports showed the incorporation of a fiber chimera with Ad40s tail and shaft domains fused to the Ad5 knob into the Ad5 capsid (25
) and more efficient incorporation of the Ad41 short fiber than the Ad41 long fiber into Ad5 particles (34
). Our results argue for the conservation of the Ad5 fiber tail in pseudotyped, Ad5-derived adenoviruses.
The virus transduction experiments of our study demonstrate that tropism modification of Ad5 by insertion of a peptide ligand into short-shafted adenovirus fibers, in this case the F5/41s fiber, is feasible. This is important especially in consideration of the negative charge of Ad5 hexon molecules in the virus capsid and the potentially resulting repulsion from cell surfaces. In melanoma cells, all modified RGD-containing fibers that were incorporated into virus particles mediated enhanced adenovirus cell entry relative to that of the RGD-less control virus. In SK-MEL-28 cells, all viruses with C-terminal RGD showed similar transduction efficacies, and AdβGF5/41sCLL-RGD was superior to AdβGF5/41sCL-RGD and AdβGF41sCLL-RGD in primary melanoma cells and HUVEC. Different expression levels of different integrins might be responsible for these intercellular variations. Still, a long linker might be advantageous for C-terminally fused peptide ligands. The lower transduction efficiency for the virus with the Ad41s fiber tail might be explained by reduced fiber incorporation into Ad5 virions (as discussed above). Importantly, we show that internal insertion of RGD peptides into different loops of the Ad41s fiber knob results in transduction efficiencies that are superior to those of C-terminal RGD fusions. Either better peptide positioning or a superior peptide conformation (loop-like versus linear) might be responsible for these results. In this regard, it is important that (i) the C terminus of the Ad41s fiber, like the C terminus of the Ad5 fiber, points towards the virus capsid rather than to the outside (36
); and (ii) the CDCRGDCFC peptide, which was originally selected from a cyclic peptide phage library (20
), is not able to circularize per se in the reducing milieu of the cytoplasm and nucleoplasm where viruses are produced. However, the peptide might be forced into a circular conformation when incorporated into a loop of the Ad41s fiber knob. The highest transduction efficiencies in all cell types were obtained for the virus that contained the RGD peptide in the HI loop, followed by the virus with the EG loop insertion. Note that both loops are located on the side of the Ad41s fiber knob, indicating that this might be a favorable locale for insertion of targeting ligands.
The CDCRGDCFC peptide mediated adenovirus cell entry at any of the positions at the tip of the EG, HI, or IJ loop or at the C terminus of the Ad41s fiber. However, our transduction results with AdβGF5/41s-2xRGD and AdβGF5/41s-loopRGD clearly show that the precise peptide position within a loop is critical for its ability for tropism modification. Even a slight change of the RGD peptide position in the IJ loop, as represented by AdβGF5/41sIJ-2xRGD, ablates its ability to redirect adenovirus transduction. Also, AdβGF5/41sIJ-loopRGD did not show an enhanced transduction efficacy compared to AdβGF5/41s. For this virus, in addition to the length of the whole insert, the amino acids that directly flank the RGD peptide are different from the CDCRGDCFC peptide insert in AdβGF5/41sIJ-RGD. However, this loop insert mediated efficient cell entry when inserted into the HI loop of the Ad5 fiber (8
). Our transduction results for AdβGF5ts*/41sIJ-RGD indicate that the mutant long fiber shaft of Ad5 is not compatible with adenovirus tropism modification by ligand peptide insertion into the Ad41s fiber knob, at least for peptide insertions into the IJ loop. The rationale for developing this construct was to improve virus transduction by extending the distance between the peptide ligand and the virus surface. By this means, we sought to reduce potentially harmful repulsive effects resulting from the negative charge of the Ad5 fiber capsid. In light of our results for the long fiber shaft virus, it is even more important that peptide-mediated adenoviral transduction in the context of the short Ad41 fiber shaft is feasible. From our data, we conclude that the length and structure of the fiber shaft critically influence the ability of peptide ligands inserted into the fiber knob to interact with their cellular receptors. Interestingly, a recent publication indicates that mutation of the KKTK motif reduces transduction efficacy mediated by a ligand inserted into the HI loop of an Ad5 fiber (7
). Further investigations are warranted to clarify the influence of the KKTK mutation on both virus detargeting and ligand-mediated retargeting of viruses with the Ad41s fiber knob.
To show peptide-mediated cell binding and entry of tropism-modified adenoviruses, competition experiments with soluble peptide are frequently performed. However, such experiments for RGD-modified viruses are complicated by the fact that RGD peptides can also interfere with binding of the Ad5 penton base to cellular integrins via its RGD motif. This interaction is required for internalization of cell-bound Ad5 particles. We previously reported that preincubation of HUVEC with circularized CDCRGDCFC peptide inhibits transduction by adenoviruses containing the Ad5 capsid (26
). In the present study, the transduction of HUVEC by AdβGF5 was not inhibited by the GRGDS peptide. This result suggests that the interaction with integrins is weaker for this peptide than for the circular peptide and thus insufficient to block virus internalization in the context of the Ad5 fiber-CAR interaction. Interestingly, we observed reduced transduction of HUVEC by AdF5/41s after preincubation of the cells with the GRGDS peptide. We concluded that transduction by this virus was at least partially dependent on the penton-integrin interaction. Similar results were reported for viruses with the Ad40s fiber (25
). In this case, transduction of primary hepatocytes was inhibited by preincubation of cells with a recombinant penton base. Note that penton bases of subgroup F adenoviruses Ad40 and Ad41 do not contain the RGD motif. After GRGDS peptide preincubation, HUVEC transduction by AdβGF5/41sIJ-RGD was reduced to a level similar to that by AdβGF5/41s, demonstrating that transduction was completely RGD dependent. Transduction of HUVEC by AdβGF5/41sHI-RGD was also significantly reduced by GRGDS preincubation. However, this inhibition was less dramatic than that for AdβGF5/41sIJ-RGD. This result could be explained by partially RGD-independent transduction. However, we suggest that the interaction between the HI-RGD and cellular integrins is stronger than that for the IJ insertion site and cannot be completely inhibited by the linear GRGDS peptide. Other groups have reported quantitatively similar or even less inhibition of RGD-modified Ad5 viruses by RGD peptides or penton base protein (9
). Moreover, our result that AdβGF5/41sHI-RGD showed a higher transduction efficacy than AdβGF5/41sIJ-RGD, while incorporating a similar quantity of fiber molecules, is also supportive of this hypothesis.
Our observation that primary normal cells were readily transduced by AdβGF5/41s without peptide ligand insertion is a novel and striking finding. Previous reports demonstrated strongly reduced transduction of healthy organs in vivo and of rat hepatocytes, rat endothelial cells, or primary human hepatocytes in vitro for Ad5-derived viruses with the Ad41s or Ad40s shaft and knob relative to that of viruses with the Ad5 fiber (25
). To date, a receptor for the Ad41s fiber has not been described. Because the transduction of HUVEC by AdβGF5/41s was RGD dependent, we hypothesize that cell entry of Ad5/41s is mediated, at least in part, by binding of the penton base via its RGD motif to cellular integrins. Thus, unwanted transduction by Ad5-derived viruses with the Ad41s fiber shaft and knob might be abolished by additional mutation of the penton base RGD motif.
In summary, we identified functional insertion sites for peptide ligands in the short-shafted Ad41 fiber knob. Internal insertion into both the top and side surfaces of the knob domain was feasible, mediated enhanced adenoviral transduction, and was superior to C-terminal peptide fusion, as shown with the CDCRGDCFC peptide. Hence, we (i) established that peptides inserted into a short-shafted fiber can mediate cell binding and (ii) defined sites for genetic insertion of targeting peptides into the Ad41s fiber knob. We suggest recombinant fibers with Ad41s fiber shaft and knob domains and peptide insertions in the EG, HI, or IJ loop as a novel platform for genetic targeting of adenovirus cell entry. This platform can now be exploited for the insertion of targeting peptides derived, for example, by various display technologies. Furthermore, we report that the precise positioning of a peptide ligand in the knob domain and also the position of the modified knob in the virus capsid, i.e., the length and structure of the fiber shaft, are critical determinants for the receptor binding capacity of the ligand peptide. Therefore, our results open new avenues for the future development of improved recombinant adenoviruses for therapeutic gene transfer, genetic vaccination, and viral oncolysis.