Various strategies have been explored to target lentiviral vector delivery to specific cells in vitro
and in vivo
including approaches involving engineered versions of the Sindbis virus E2 glycoprotein bearing either a Staphylococcus aureus
protein A domain [4
], or single chain antibody fragments [15
]. In a related targeting strategy, the cell recognition function and the fusion function mediated by the Sindbis virus E2 glycoprotein are carried out by separate proteins that are anchored in the vector's membrane [17
]. A drawback with these approaches is that they involve glycoproteins derived from laboratory-adapted strains of Sindbis virus necessitating the introduction of mutations to abrogate high efficiency binding to cell surface receptors such as heparan sulfate. Unfortunately, background transduction levels in the absence of the ligand or with cells lacking the corresponding receptors were substantial due to the leakiness of the mutations that were introduced to abolish cell binding [6
Alternative approaches for lentiviral vector targeting are emerging. For example, in a recent report Funke et al. [35
] have shown that the measles virus hemagglutinin (H) and fusion protein (F) are capable of pseudotyping of HIV-1 vectors. Moreover, engineered H proteins displaying cell-specific ligands or single chain antibodies resulted in specific transduction of target cells. While the background transduction levels in the absence of the ligand were low, the concentration of such pseudotypes by centrifugation has been challenging.
In this study, we investigated the efficiency and selectivity of a cell targeting approach involving lentiviral vectors pseudotyped with the Sindbis virus strain TR339 glycoproteins [30
] plus a separate cell binding domain. The Sindbis virus TR339 strain was originally derived from the AR339 strain which represents the prototype alphavirus and whose sequence differs markedly from that of the common Sindbis virus laboratory strains that have emerged after selection for efficient growth in cell culture [36
]. Such laboratory strains bear mutations that facilitiate binding of the virus to heparan sulfate [30
], a property that is clearly not desirable in the context of targeting stategies. Our data indicate that the titers of SCF-containing vectors bearing the Sindbis virus TR339 glycoproteins were up to 3.2 ± 0.04 × 105
TU/ml on c-kit-expressing 293 cells, while the titers of lentivirus particles lacking SCF were some 80 fold lower (Fig. ). These background transduction levels are comparable to those reported previously by Morizono et al. [6
] using multiply mutated versions of a Sindbis virus laboratory strain-derived E2 protein. However, it is likely that the unmodified E2 protein of the TR339 strain retains considerably more of the non-attachment activity of this protein than the highly modified version used by Morizono et al. Therefore, it is possible that use of the wild type E2 may improve the efficiency of pseudotype particle production. However, a direct comparison of infectivity per particle of the different vectors will be required to answer this question.
We also explored a targeting approach involving vector particles containing a membrane-proximal fusion domain (VSV-GS) derived from the VSV G glycoprotein [29
] plus a separate cell binding domain. The results reported earlier by Jeetendra et al. [29
] showed that the membrane-proximal 42 amino acids (residues 421 to 461) of the VSV-G protein ectodomain together with the TM region and the cytoplasmic tail were able to potentiate the membrane fusion activity of heterologous viral fusion proteins when the two proteins were co-expressed. We investigated the efficiency of the VSV-GS domain to target SCF-displaying lentiviral vector particles to c-kit expressing cells. The results presented in Figs. and show that specific transduction of c-kit-expressing 293 and MO7-e cells was achieved and that transduction efficiencies using SCF-bearing vectors were 40 fold higher (1.4 ± 0.2 × 104
) compared to vectors lacking SCF (3.5 ± 1.5 × 102
). Overall, the titers obtained with VSV-GS-containing particles were low, possibly because VSV-GS mostly promoted hemifusion events [29
]. It was evident that MO7-e cells were transduced less efficiently than 293-c-kit cells although c-kit levels were higher in MO7-e cells as judged by FACS (data not shown). This may be due to the fact that the transduction conditions used for MO7-e cells were not optimal. This view is enforced by the observation that VSV-G pseudotypes on MO7-e cells did relatively poorly compared to 293-c-kit cells (data not shown).
It was interesting to note that lentiviral vectors bearing SCF, but lacking the VSV-GS fusion domain, were capable of transducing 293-c-kit cells but not 293T cells (Fig. ). This may indicate that SCF provided some cryptic fusion function. Furthermore, the transduction experiments involving 293T cells presented in Fig. revealed that transduction efficiencies of vector particles containing SCF increased in a dose-dependent manner. This may have resulted from the binding of vector particles to 293T cells through nonspecific SB-Env interactions and augmented fusion triggered by SCF.
Young and collaborators have shown that alpharetroviral and gammaretroviral vectors carrying unmodified ALV envelope glycoproteins and bearing specific ligand proteins or single chain antibody as a bridge were capable of targeting specific cells in vitro
]. An attractive feature of this bridge targeting system is that the background transduction levels were low. These findings promoted us to investigate the capacity of lentiviral vectors bearing the ALV-A and ALV-B Env glycoproteins to target specific cells in vitro
using soluble TVA and TVB bridge proteins bearing cell-specific ligands. Fig. shows that HIV-1-based vectors were efficiently pseudotyped using both the ALV-A and ALV-B Env glycoproteins as judged from transduction experiments involving 293-DK7 cells that express a hybrid ALV-A/B receptor [34
]. The titers obtained were 1.57 ± 0.47 × 108
TU/ml for pseudotypes bearing the ALV-A Env protein after a 250-fold concentration and 1.20 ± 0.30 × 107
TU/ml for ALV-B pseudotypes (Table ). The ALV-A pseudotype titers obtained compare favorably to those reported previously by Lewis et al. 2001 [37
]. It was interesting to note, however, that the titers reported by Lewis et al. involving an unmodified version of the ALV-A Env protein were considerably lower, less than 104
TU/ml, while ALV-A glycoproteins bearing a truncated cytoplasmic tail or a chimeric cytoplasmic tail were some 5-fold higher. Our work involved unmodified ALV-A Env and ALV-B Env proteins bearing intact cytoplasmic tails. Yet the titers were consistently well above 105
TU/ml for both ALV-A and ALV-B pseudotypes (Table ). The reason for this discrepancy is not clear; it may be due to differences in the TVA-expressing 293 cell lines that were used for vector titration.
Boerger et al. [24
] have previously shown that retroviral vectors preloaded with TVB-EGF bridge proteins were relatively thermostable and could be generated directly from vector-producing cells. Our findings show that the same approach is feasible as far as TVB-Epo-containing lentiviral vector particles are concerned (Fig. ). Interestingly, the same approach was applicable to TVA-Epo-containing lentiviral vectors as well (Fig. ). This was possibly facilitated by metastable fusion-incompetent ALV-A Env/TVA-Epo complexes that were unable to complete the fusion reaction during vector production at neutral pH [38
An attractive feature of the bridge strategy involving ALV-A Env pseudotyped vectors is that background levels were low (Fig. ). However, the same kind of tight specificity was not seen with ALV-B-pseudotyped lentiviral vector particles. Such particles displayed residual transduction in the presence of the heterologous TVA-Epo bridge protein but not in the presence of unmodified TVA alone. The reason for this difference in the binding specificities is not clear at present.