Retroviral vectors are attractive vehicles for gene delivery. The retrovirus envelope protein provides the basis for specificity and cell entry. The Moloney murine leukemia virus (MLV) ecotropic envelope binds to an amino acid transporter (
1) expressed only in mouse cells and closely related species and enters the cell by a pH-dependent endocytotic mechanism (
15). The host range is determined by regions of variable sequences (VRA and VRB) within the extracellular domain (SU) of envelope.
Previous studies have examined targeting of pseudotyped MLV to human cells by using several strategies including antibody-streptavidin complexes that bridge the virus to target cell receptors (
7,
21), display of variable antibody fragments as antibody-envelope fusion proteins (
4,
22,
24,
31), and insertion of target ligands into wild-type envelope (
5,
11,
13,
16,
25,
26-
28). Typically, these modifications were at the N terminus of envelope, and viruses bearing such modified envelope derivatives were unable to transduce cells. These relatively large N-terminal modifications are believed to prevent envelope from undergoing the conformational change required for fusion and cell entry (
31). The lack of full activity of N-terminally modified envelope derivatives necessitates the coexpression of wild-type envelope for production of transducing virus.
Integrin receptors are heterodimers composed of α and β subunits that play essential roles in cell-cell and cell-extracellular matrix interactions. Integrin receptor ligands contain the tripeptide RGD. RGD peptides as short as six amino acids (GRGDSP) can bind to integrin receptors and inhibit binding of full-length adhesion proteins (
3,
6,
9,
29). Here we show that short RGD peptide ligands inserted at multiple locations within MLV envelope can direct an MLV retrovirus vector to transduce human cells. In addition, we demonstrate that the length and particular position of the inserted ligand can affect viral tropism.
We constructed >40 chimeric envelope derivatives containing in-frame insertions of either a 13- or a 21-amino-acid RGD peptide (RGD
13 or RGD
21, respectively; Table ). The core of the RGD
13 ligand is a six-amino-acid peptide, GRGDSP, which represents an RGD consensus sequence. The core of the RGD
21 ligand is a 14-amino-acid sequence, QGATFALRGDNPQG, derived from the mouse laminin protein (
3). Both the RGD
13 and RGD
21 peptides were flanked by cysteine residues to constrain the sequence within a loop (
3,
12,
29), as cyclization of RGD peptides has been shown to increase affinity for integrin receptors (
20). In some cases, chimeric envelope derivatives with multiple ligands in tandem were also generated. Several of the chimeric envelope derivatives had deletions of envelope sequences, in addition to ligand insertions, as a result of multiple restriction enzyme cleavages. In all, 26 chimeric envelope derivatives containing the RGD
13 ligand, 16 chimeric envelope derivatives containing the RGD
21 ligand, and 5 chimeric envelope derivatives containing an RGE
21 ligand, a control nonbinding peptide (
3,
10,
12,
23), were constructed.
Pseudotyped virus containing chimeric envelope derivatives (RGD viruses) was generated by using a human 293T-cell-based packaging cell line, Anjou 65 (
18). Immunoblotting of purified virions indicated that in all cases tested the chimeric envelope derivatives were incorporated into the virion and correctly processed (data not shown). The RGD viruses were initially tested for their ability to transduce mouse NIH 3T3 cells and, on the basis of these results (Fig. and ), we can draw several conclusions. First, many of the RGD viruses retained their ability to transduce mouse cells, but those bearing insertions within the N terminus (RGD
13-4,5 and RGD
21-2,3), the VRA (RGD
13-8,12 and RGD
21-5), and the C-terminal region (RGD
13-19,23,34 and RGD
21-15,16) did not. Several of these latter RGD viruses also failed to transduce human cells (RGD
13-12,19,23,24 and RGD
21-5,15,16), whereas for others (RGD
13-4,5,8 and RGD
21-2,3) the defect was mouse cell specific. Second, most RGD
21 viruses transduced NIH 3T3 cells with efficiencies comparable to those of the equivalent RGD
13 viruses, and none of the RGD
21 viruses transduced NIH 3T3 cells with greater efficiency than the equivalent RGD
13 virus.
The RGD viruses were next tested for transduction of A375 human melanoma cells, which have been previously used to study integrin receptor binding (
2,
8,
19). As expected, viruses bearing unmodified MLV envelope failed to transduce this human cell line. Significantly, however, many of the RGD viruses were able to transduce A375 human melanoma cells (Fig. and ). Transduction occurred when the RGD peptide was inserted at the N terminus (RGD
13-1-3 and RGD
21-1), within the N-terminal region (RGD
13-4-6 and RGD
21-2,3), within the VRA region (RGD
13-7,8,10,11 and RGD
21-4,6,7), and upstream of the PRR (RGD
13-14,15 and RGD
21-8,9). RGD viruses with insertions in the PRR and C-terminal region failed to transduce human cells. Several of the RGD viruses that transduced human cells failed to transduce NIH 3T3 cells (RGD
13-4,5,8 and RGD
21-2,3), indicating that viral tropism had been switched. In all cases tested, RGD viruses that transduced A375 human melanoma cells also transduced other human and nonhuman cell lines that contained integrin receptors (data not shown).
To examine the basis and specificity of human cell transduction, two experimental approaches were undertaken. First, the RGD21 ligand was replaced with the equivalent RGE21 sequence. Pseudotyped virus expressing an RGE21 chimeric envelope derivative transduced NIH 3T3 host cells with efficiencies comparable to the equivalent RGD21 derivative; in contrast, transduction of A375 human melanoma cells was significantly reduced (Fig. ). Second, we analyzed the effect of antibodies to integrin receptors on the transduction of RGD viruses. NIH 3T3 and A375 human melanoma cells were pretreated with integrin receptor antibodies, and transduction was performed with two of the RGD21 viruses. Transduction of human but not mouse cells was substantially reduced (Fig. ).
We have developed a strategy for altering the host range of ecotropic retrovirus vectors by using chimeric envelope derivatives bearing short peptide ligands. In many instances, pseudotyped virus expressing a chimeric envelope derivative transduced human cells without removal of the N-terminal region or coexpression of wild-type envelope. Ligand insertions at multiple locations within the N-terminal half of envelope enabled transduction of human cells. In contrast, transduction of human cells did not occur with RGD peptide insertions in the PRR or C-terminal region of envelope, although many of these viruses could transduce mouse cells. Some of the RGD viruses bearing insertions at the N terminus or VRA region (RGD13-4,5,8 and RGD21-2,3) transduced human but not mouse cells. Collectively, these observations indicate that the position of the inserted ligand can dictate tropism.
Transduction efficiencies differed among the RGD viruses, indicating that the precise location of the ligand within envelope is important and can be optimized. In general, RGD13 and RGD21 ligands transduced NIH 3T3 cells with comparable efficiencies, suggesting that envelope can accommodate ligands of different sizes, perhaps longer than those examined in this study. Longer ligands may be more disruptive but may also have increased affinity for the target receptor. The accompanying study confirms and extends these conclusions and demonstrates the generality of the experimental approach.