Redirecting Retroviral Tropism by Insertion of Short, Nondisruptive Peptide Ligands into Envelope

Abstract

A potentially powerful approach for in vivo gene delivery is to target retrovirus to specific cells through interactions between cell surface receptors and appropriately modified viral envelope proteins. Previously, relatively large (>100 residues) protein ligands to cell surface receptors have been inserted at or near the N terminus of retroviral envelope proteins. Although viral tropism could be altered, the chimeric envelope proteins lacked full activity, and coexpression of wild-type envelope was required for production of transducing virus. Here we analyze more than 40 derivatives of ecotropic Moloney murine leukemia virus (MLV) envelope, containing insertions of short RGD-containing peptides, which are ligands for integrin receptors. In many cases pseudotyped viruses containing only the chimeric envelope protein could transduce human cells. The precise location, size, and flanking sequences of the ligand affected transduction specificity and efficiency. We conclude that retroviral tropism can be rationally reengineered by insertion of short peptide ligands and without the need to coexpress wild-type envelope.

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 (RGD₁₃ or RGD₂₁, respectively; Table 1). The core of the RGD₁₃ ligand is a six-amino-acid peptide, GRGDSP, which represents an RGD consensus sequence. The core of the RGD₂₁ ligand is a 14-amino-acid sequence, QGATFALRGDNPQG, derived from the mouse laminin protein (3). Both the RGD₁₃ and RGD₂₁ 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 RGD₁₃ ligand, 16 chimeric envelope derivatives containing the RGD₂₁ ligand, and 5 chimeric envelope derivatives containing an RGE₂₁ ligand, a control nonbinding peptide (3, 10, 12, 23), were constructed.

TABLE 1.

Description of RGD viruses^a

Envelope no. (sequence)	Location (aa)b	No. of insert(s)	Nucleotides deleted from envelope
RGD13 (CAAA-GRGDSP-TRC)
1	1	1
2	1	2
3	1	4
4	38	1
5	38	3
6	38	1	5990-6082
7	68	1
8	68	2
9	68	1	6082-6191
10	120	1
11	120	2	6238-6281
12	120	3
13	185	1
14	230	1
15	230	2
16	235	1
17	235	4
18	310	1
19	310	2
20	321	1
21	321	2
22	382	1
23	382	2
24	382	3
25	388	1
26	388	2
RGD21 (CAAA-     QGATFALRGDNPQG-TRC)
1	1	1
2	38	1
3	38	1	5990-6082
4	68	1
5	68	1	6082-6191
6	120	1
R	120	1	6238-6281
8	185	1
9	230	1
10	235	1
11	310	1
12	321	1
13	382	1
14	388	1
15	168	1, 1
16	1230	1, 1
RGE21 (CAAA-     QGATFALRGENPQG-TRC)
1	1	1
2	38	1	5990-6082
3	68	1
4	68	1	6082-6191
5	230	1

^aThe sequences of the RGD₁₃, RGD₂₁, and RGE₂₁ ligands are shown. For the chimeric envelope derivatives RGD₁₃ 1 to 26, RGD₂₁ 1 to 16, and RGE₂₁ 1 to 5, the position of the ligand insertion, the number of inserts, and any additional modifications are indicated. For the construction of chimeric envelope derivatives, the extracellular domain (gp70) of ecotropic MLV envelope was linearized at random locations by partial digestion with blunt-end restriction endonucleases in the the presence of 50 to 400 ng of ethidium bromide per ml. The 13-amino-acid RGD sequence (CAAAGRGDSPTRC; RGD₁₃) was derived by annealing two oligonucleotides, RGD₁₃-A (TGCGCGGCCGCTGGCCGTGGCGATTCTCCCACGCGTTGT) and RGD₁₃-B (ACAACGCGTGGGAGAATCGCCACGGCCAGCGGCCGCGCA). The annealed sequence was ligated into the linearized envelope plasmid, and subclones were screened for insert position and orientation. The resultant chimeric envelope derivatives were cloned into the envelope expression vector, pCEE (14). The RGD₁₃ 1 to 3 chimeric envelope derivatives were constructed by insertion of an NaeI linker at the C terminus of the signal sequence of wild-type envelope, and the annealed RGD oligonucleotides were cloned into the NaeI site. Chimeric envelope derivatives with the 21 amino acid RGD sequence (CAAAQGATFALRGDNPQGTRC; RGD₂₁) were constructed by restriction endonuclease digestion of RGD₁₃ envelopes with NotI and MluI and insertion of the RGD₂₁ annealed oligonucleotides RGD₂₁-A(GGCCGCTCAAGGCGCAACGTTCGCGCTCAGAGGCGATAATCCACAGGGGA) and RGD₂₁-B (CGCGTCCCCTGTGGATTATCGCCTCTGAGCGCGAACGTTGCGCCTTGAGC). The RGD₂₁ envelope derivatives were cloned into an expression plasmid that contained a zeocin selection marker (Invitrogen). RGE₂₁ was constructed analogous to RGD₂₁. Chimeric envelope derivatives expressing two RGD sequences, RGD₂₁-15 and RGD₂₁-16, were constructed by removal of the BstEII/ClaI fragment of RGD₂₁-1 and insertion of the BstEII/ClaI region from RGD₂₁-4 and RGD₂₁-9, respectively.

^baa, amino acid.

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. 1 and 2), 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 (RGD₁₃-4,5 and RGD₂₁-2,3), the VRA (RGD₁₃-8,12 and RGD₂₁-5), and the C-terminal region (RGD₁₃-19,23,34 and RGD₂₁-15,16) did not. Several of these latter RGD viruses also failed to transduce human cells (RGD₁₃-12,19,23,24 and RGD₂₁-5,15,16), whereas for others (RGD₁₃-4,5,8 and RGD₂₁-2,3) the defect was mouse cell specific. Second, most RGD₂₁ viruses transduced NIH 3T3 cells with efficiencies comparable to those of the equivalent RGD₁₃ viruses, and none of the RGD₂₁ viruses transduced NIH 3T3 cells with greater efficiency than the equivalent RGD₁₃ virus.

FIG. 1.

Transduction of NIH 3T3 cells and A375 human melanoma cells by RGD₁₃ viruses. NIH 3T3 and A375 human melanoma cells were infected with an RGD virus and then selected with G418 for 2 weeks, fixed, and stained with Giemsa, and the colonies were counted. The amphotropic virus, Amph, was generated by expressing the amphotropic envelope, pCAA, and the ecotropic virus, Eco, was generated by expressing the wild-type ecotropic envelope, pCEE. Note the log scale. The pCAA expression vector was generated by removing the amphotropic envelope gene from the full-length infectious clone(17) and insertion into pCEE after removal of the ecotropic envelope. The packaging construct, LAPNL, was generated by removal of the VSV-G envelope from LGRNL (30) and insertion of the secreted alkaline phosphatase gene (SEAP; Tropix). Pseudotyped virus producer cell lines were generated by cotransfection of Anjou 65 cells with LAPNL and a plasmid expressing a chimeric envelope derivative using Dotap (Boehringer), followed by selection in zeocin (200 μg/ml) for 2 weeks. RGD₁₃ required cotranfection of a zeocin expression plasmid (Invitrogen).

FIG. 2.

Transduction of NIH 3T3 cells and A375 human melanoma cells by RGD₂₁ viruses (as described for Fig. 1).

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. 1 and 2). Transduction occurred when the RGD peptide was inserted at the N terminus (RGD₁₃-1-3 and RGD₂₁-1), within the N-terminal region (RGD₁₃-4-6 and RGD₂₁-2,3), within the VRA region (RGD₁₃-7,8,10,11 and RGD₂₁-4,6,7), and upstream of the PRR (RGD₁₃-14,15 and RGD₂₁-8,9). 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 (RGD₁₃-4,5,8 and RGD₂₁-2,3), 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. 3). 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. 4).

FIG. 3.

Requirement of the RGD sequence for transduction of human cells. NIH 3T3 (A) and A375 human melanoma (B) cells were infected with an RGD₂₁ or RGE₂₁ virus, and transduction was analyzed as described in the legend to Fig. 1.

FIG. 4.

Antibodies to integrin receptors block transduction of human cells. NIH 3T3 (A) and A375 human melanoma (B) cells were pretreated with polyclonal antibodies to β₁, β₃, and α_ν integrin receptors (Santa Cruz Biotechnology), infected with RGD₂₁-1, RGD₂₁-4, or RGD₂₁-9 viruses, and transduction was analyzed as described in the legend to Fig. 1. The three antibodies were diluted 1:100 in Dulbecco modified Eagle medium and incubated with cells for 4 h. Cells were then incubated with pseudotyped virus for 6 h.

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.