Exosome biogenesis may occur in a delayed manner, as when previously synthesized intralumenal vesicles (ILVs) are released as exosomes upon fusion of endosomes with the cell surface, or in an immediate fashion, which occurs when these vesicles are synthesized at endosomal patches that are contiguous with the plasma membrane. These endosomal patches of the cell surface are transient entities that form as a consequence of endosome-plasma membrane fusion, but are subsequently reinternalized from the cell surface. As a result, the Trojan exosome hypothesis predicts that retroviruses should also bud into endosomes and from the cell surface. The empirical evidence that retroviruses bud at both the cell surface and at endosomal membranes includes examples from all major classes of exogenous vertebrate retroviruses, including the alpharetroviruses [avian myeloblastosis virus (1) and avian leukosis virus (2)], betaretroviruses (mouse mammary tumor virus; ref. 3), gammaretroviruses (murine leukemia virus; refs. 4 and 5), deltaretroviruses [human T-lymphotrophic virus (HTLV-1) (6) and bovine leukemia virus (BLV); ref. 7], lentiviruses [simian immunodeficiency virus , HIV , and Maedi-Visna virus ], and spumaviruses . These citations represent a small portion of the available evidence, as dual budding has been observed for other characterized and uncharacterized exogenous retroviruses as well as for endogenous retroelements that encode the intracisternal A particles .

In the main text we discussed the Trojan exosome hypothesis and its ability to reconcile a wide variety of data into a single mechanistic model of retroviral biogenesis and transmission. Here we discuss its implications for the initial genesis of retroviruses. It is widely believed that retroviruses evolved from the more ancient LTR-containing retrotransposons . LTR retrotransposons, such as the Ty elements of yeast, express a genomic RNA and proteins similar to the retroviral Gag and Gag-Pol proteins (Fig. 2a). These proteins and the genomic RNA assemble into a cytoplasmic replication particle similar to that of retroviruses, reverse transcribe their RNA genome, and insert a DNA copy of their genome into the host cell chromosome (Fig. 2b). However, they differ from retroviruses in that (i) their replication cycle is entirely cytoplasmic, (ii) they lack an Env gene, and (iii) they are not infectious .

The evolution of retroviruses from LTR retrotransposons has been modeled by Temin . This model of retroviral evolution and its derivatives assumes that the transduction of a nascent Env gene is the critical event in retroviral evolution and focuses on the molecular genetics involved in gene transduction. However, this perspective does not take into account the complex cell biology of vesicle biogenesis, transport, and fusion reactions . In essence, this model proposes the transduction of a gene that (i) encodes a membrane protein, (ii) captures the replication capsid of the LTR retroelement, (iii) directs vesicle formation around the replication capsid, (iv) mediates release of a membrane-bound replication particle from the cell, (v) directs binding of the encapsulated particle to a neighboring cell, and (vi) catalyzes fusion between the "viral" membrane and the membrane of the neighboring cell. Given that none of these properties is inherently advantageous without the others, it is extremely unlikely that a functional Env gene could be generated by transduction of a cellular gene. Moreover, the possibility that retroviruses may be the oldest enveloped viruses hampers any model of retroviral evolution that rests on the transduction of an Env gene from another virus.

The Trojan exosome hypothesis offers a relatively simple model of retroviral genesis in which a simple mutation in the Gag gene of an LTR retrotransposon would convert it into a retrovirus by directing its Gag protein, and the replication particle it forms, into ILVs of a host that participates in exosome exchange (Fig. 3a). Such a mutation might be as simple as the addition of an N-terminal myristoylation signal or a peptide-targeting motif and would provide the replication particle with a membrane, the ability to leave the cell, bind to other cells, and fuse with their membranes through the pathway of exosome exchange. Because this "Trojan exosome" acquired additional mutations in subsequent rounds of replication, the significant advantages of horizontal transmission would favor additional Gag mutants that increased the efficiency of Gag targeting into ILVs. Other mutations that enhanced its binding to neighboring cells, such as the transduction of an exosomal adhesion molecule, would be favored, explaining the existence of Env-type genes in most retroviruses. Moreover, once these genes were selected, they could form the requisite genetic context for subsequent gene transductions that joined them with fusogenic peptides to generate modern Env genes, which greatly enhance the kinetics of membrane fusion between retroviral particles and their host cells, or added genes that amplified the exosome biogenesis pathway.

The model of retroviral genesis predicted by the Trojan exosome hypothesis is attractive at several levels. First, it satisfies the basic requirements of explaining how retroviruses could evolve from an LTR retrotransposon. Second, it explains the selection of retroviruses and their unique properties by the accretion of simple, immediately advantageous mutations that exploit a preexisting cellular pathway, which is a paradigm of viral evolution. Third, it does not rely on unlikely mutations or preexisting enveloped viruses, and thus, is simpler than the alternative model. Fourth, it explains why retroviruses possess unique host cell lipids and proteins that closely resemble those of exosomes, as well as their ability to infect cells independently of their Env genes or host cell viral receptors (see text). This point is particularly significant, because the alternative model fails entirely to predict or explain the exosomal nature of host cell proteins and lipids that are present in retroviral particles or ability of retroviruses to infect cells by receptor-independent or Env-independent mechanisms. Fifth, and finally, it blurs the line between LTR retrotransposons and retroviruses by redefining retroviral Env genes as extremely important but nonessential facilitators of infection. In addition to explaining the horizontal transmission of what appears to be an LTR retrotransposon , this model predicts multiple origins for retroviruses, fluctuating interconversion of LTR retrotransposons and retroviruses, and calls into question the common practice of differentiating between retroviruses and LTR retrotransposons solely on the basis of an Env-type gene in the proviral genome .

This model of retroviral evolution assumes a preexisting pathway of exosome exchange. Our current knowledge of this pathway’s origins is limited to conjecture, but even on this basis there is reason to believe that it is ancient. The origins of exosome exchange appear to lie in two originally unrelated and ancient cell biological processes, the selective degradation of membrane proteins and the repair of plasma membrane wounds. Organisms as primitive as yeast selectively degrade membrane proteins of the late secretory and endocytic organelles by targeting them into endosomal ILVs and destroying ILV components in degradative vacuoles . Primitive unicellular organisms also have to repair tears in their plasma membrane, a process that involves the fusion of late endosomes and lysosomes with the plasma membrane . The consequences of this multiorganellar fusion event are (i) the reformation of a contiguous cell membrane, (ii) the presence of endo/lysosomal membrane patches at the cell surface, and (iii) the release of any ILVs and soluble proteins present in the endosomes into the extracellular milieu. Thus, eukaryotic cells that expressed both the ILV synthesis and plasma membrane wound repair pathways would generate exosomes by accident, without any initial benefit.

Viewed from this perspective, the final stage in the genesis of exosome exchange requires nothing more than the capacity for exosome resorption. In the context of a unicellular organism, such a capacity would have the advantage of allowing the cells to conserve numerous functional, fully formed proteins and lipids, as well as the retention of nonfunctional molecules that can be degraded, with their building blocks used for either biosynthetic or bioenergetic purposes. In the context of a multicellular organism, the capacity for exosome exchange has the added advantage of delivering numerous molecules from one cell to another, which has obvious implications for cell-cell communication and development. In this context, it is interesting to note that wingless-mediated signaling appears to involve exosome exchange . Thus, although empirical evidence on the origins of exosomes is lacking, there is a reasonable basis for their existence as early as the rise of metazoans, which roughly coincides with the genesis of retroviruses.

As a mechanistic model of retroviral biogenesis and transmission, the Trojan exosome hypothesis predicts that retroviral tropism in vivo will be influenced by a variety of factors, including (i) the ability of a cell to transmit the virus within the body and between individuals, (ii) the sum of the interactions between proteins on the virus surface and the cell surface, and (iii) the effect that these interactions have on Env-dependent and -independent fusion between the viral and cellular membranes. Thus, the Trojan exosome hypothesis does not predict high efficiency infection by exosome exchange of all, or even many, cells. Rather, it predicts that a particular retroviral isolate will possess a multifactorial, graded tropism that reflects (i) its selective history, including its past reliance on Env-dependent and Env-independent transmission mechanisms, (ii) the host cell proteins on the retrovirus surface, (iii) the surface proteins of the host cell, and (iv) the competency of the target cell for exosome exchange. In the example of HIV, the Env protein contributes high efficiency infection of CD4+, coreceptor-expressing cells , whereas MHC/peptide complexes on the viral surface appear to target retroviruses into particular subsets of CD4⁺ cells , and integrins and other adhesion molecules on the surface of these "viral exosomes" facilitates targeting to other cell types that possess the appropriate complementary ligands and capacity for exosome uptake .

The tropism of HIV for CD4⁺ cells, and particularly those that also express an appropriate coreceptor, is a critical aspect of HIV pathogenesis and transmission . However, the Env-dependent nature of HIV’s tropism for CD4⁺ cells has obscured the fact that many retroviruses preferentially infect cells of the immune system. A review of the literature shows that nearly all characterized retroviruses replicate in peripheral blood leukocytes (Table 1). Moreover, the example of HTLV-1 demonstrates that the retroviral tropism for immune cells does not reflect a universal affinity of retroviral Env proteins for the CD4 molecule .

The concept that retroviral tropism will be affected by the status of MHC/peptide complexes on the retroviral surface and TCR receptor affinity on T cells has important implications for infectious synergy between retroviruses and other infectious agents. In brief, simultaneous infection of an antigen-presenting cell, such as a macrophage, by a retrovirus and another pathogen will result in the production of "viral exosomes" that carry MHC/peptide complexes loaded with peptides encoded by the other pathogen. In the same way that MHC/peptide complexes containing HIV peptides appear to target HIV into HIV-tropic T cells, viral exosomes carrying peptides from the secondary infectious agent will target the retrovirus into T cell subpopulations that are specific for peptides encoded by the secondary infectious agent. Such a model is consistent with the empirical data on HIV, which induces infectious synergy when combined with other pathogens and targets subsets of CD4+ T cells in vivo prior to its general elimination of the CD4+ T cell population.

Alloimmunization of humans is not currently associated with deleterious effects on the immunized individuals . However, safety of the alloimmunization procedure does not mean that alloimmunization of entire populations to combat HIV and HTLV-1 has no drawbacks. The most obvious complication is that alloimmunization will increase the probability of blood and tissue rejection. In addition, alloimmunization has the potential to induce maternal-fetal incompatibility, as can occur when Rh- mothers are exposed to the Rh antigen and subsequently conceive Rh+ fetuses. There are medical procedures that can neutralize blood and fetal incompatibilities, but some societies may still consider these potential costs of alloimmunization as greater than the potential benefit, particularly societies that currently have relatively low rates of HIV and HTLV-1 infection. However, those societies that currently suffer most from HIV and HTLV-1 infection may reach the opposite conclusion, due in part to the sparse application of transfusion and transplantation procedures, the current high risk of HIV and HTLV-1 infection by blood and tissue transplantation, and the relatively small burden of reduced fertility for some couples.

In addition, the efficacy of alloimmunization in preventing retroviral infection will vary considerably among different members of a society. Some of the variables affecting its efficacy include the allotype of the individual that is immunized, the alloantigens in the allovaccine, and the allotypes of the individuals that subsequently share tissue, whether by sex or other means. As a result, even the most successful alloimmunization program is unlikely to provide complete prophylactic immunity for all alloimmunized individuals. However, alloimmunization programs that included significant testing and educational components that alerted individuals to relative risks might minimize this complication.

It should be noted that many of the costs associated with alloimmunization seem particularly small when one considers their application to other animals. Retroviruses pose significant risks to a number of domestic, farm, and wild animal populations. Given that many of the medical and ethical consideration that might prevent alloimmunization of human populations are either minimal or nonexistent when applied to animals, there is little reason to delay the application of alloimmunization to stop the spread of animal retroviruses.

Vaccines comprised of whole killed viral particles have the potential to induce both alloimmunity and antiviral immunity. Moreover, there is empirical evidence that such vaccines can induce prophylactic protection against retroviral infections . However, it is difficult to predict whether whole killed retrovirus vaccines are inherently more or less effective than strict allovaccines, which lack retroviral antigens. The evolution of alloimmunity under retroviral selective pressure suggests that the simultaneous exposure to both sets of antigens might generate a distinct type of response to retroviral antigens that would enhance the development of prophylactic protection against retroviral infection. Alternatively, the presence of retroviral antigens in these vaccines might induce responses that will ultimately potentiate infection and/or pathogenesis, as predicted by the Trojan exosome hypothesis and documented for recombinant vaccines based on retroviral envelope proteins . Clearly, empirical studies are needed to resolve this question and will have to control for the alloantigenic status of the virus particles and/or infected cells used in the vaccine as well the alloantigenic status of the trial participants.

1. de The, G., Becker, C. & Beard, J. W. (1964) J. Natl. Cancer Inst. 32, 201-235.

2. Podmaniczky, E., Szende, B., Lapis, K. & Ferenc, G. (1976) Int. J. Cancer 18, 536-539.

3. Bevilacqua, G. (1983) Exp. Mol. Pathol. 39, 271-281.

4. Srinivas, R. V., Melsen, L. R. & Compans, R. W. (1982) J. Virol. 42, 1067-1075.

5. Hansen, M., Jelinek, L., Jones, R. S., Stegeman-Olsen, J. & Barklis, E. (1993) J. Virol. 67, 5163-5174.

6. Kiernan, R., Marshall, J., Bowers, R., Doherty, R. & McPhee, D. (1990) AIDS Res. Hum. Retroviruses 6, 743-752.

7. Filice, G., Carnevale, G., Lanzarini, P., Orsolini, P., Soldini, L., Trespi, G. & Cereda, P. M. (1988) Microbiologica 11, 1-5.

8. Ringler, D. J., Wyand, M. S., Walsh, D. G., MacKey, J. J., Chalifoux, L. V., Popovic, M., Minassian, A. A., Sehgal, P. K., Daniel, M. D., Desrosiers, R. C., et al. (1989) Am. J. Pathol. 134, 373-383.

9. Blom, J., Nielsen, C. & Rhodes, J. M. (1993) APMIS 101, 672-680.

10. Raposo, G., Moore, M., Innes, D., Leijendekker, R., Leigh-Brown, A., Benaroch, P. & Geuze, H. (2002) Traffic 3, 718-729.

11. Lee, W. C., McConnell, I. & Blacklaws, B. A. (1996) Vet. Microbiol. 49, 93-104.

12. Khan, A. S., Sears, J. F., Muller, J., Galvin, T. A. & Shahabuddin, M. (1999) J. Clin. Microbiol. 37, 2678-2686.

13. Calarco, P. G. (1975) Biol. Reprod. 12, 448-454.

14. Riccardi, R., Pimpinelli, N., Ficarra, G., Borgognoni, L., Gaglioti, D., Milo, D. & Romagnoli, P. (1990) Hum. Pathol. 21, 897-904.

15. Kuff, E. L., Feenstra, A., Lueders, K., Smith, L., Hawley, R., Hozumi, N. & Shulman, M. (1983) Proc. Natl. Acad. Sci. USA 80, 1992-1996.

16. Temin, H. M. (1980) Cell 21, 599-600.

17. Coffin, J. M., Hughes, S. H. & Varmus, H. E. (1997) in Retroviruses (Cold Spring Harbor Laboratory, Plainview, NY), p. 843.

18. Roth, J. F. (2000) Yeast 16, 785-795.

19. Kirchhausen, T. (2000) Nat. Rev. Mol. Cell Biol. 1, 187-198.

20. Katzmann, D. J., Odorizzi, G. & Emr, S. D. (2002) Nat. Rev. Mol. Cell Biol. 3, 893-905.

21. Syomin, B. V., Leonova, T. Y. & Ilyin, Y. V. (2002) Mol. Genet. Genomics 267, 418-423.

22. Griffiths, D. J. (2001) Genome Biol. 2, REVIEWS1017.

23. Bowen, N. J. & McDonald, J. F. (2001) Genome Res. 11, 1527-1540.

24. Bowen, N. J. & McDonald, J. F. (1999) Genome Res. 9, 924-935.

25. McNeil, P. L. (2002) J. Cell Sci. 115, 873-879.

26. McNeil, P. L. & Terasaki, M. (2001) Nat. Cell Biol. 3, E124-E129.

27. Miyake, K. & McNeil, P. L. (1995) J. Cell Biol. 131, 1737-1745.

28. Jaiswal, J. K., Andrews, N. W. & Simon, S. M. (2002) J. Cell Biol. 159, 625-635.

29. Reddy, A., Caler, E. V. & Andrews, N. W. (2001) Cell 106, 157-169.

30. Rodriguez, A., Webster, P., Ortego, J. & Andrews, N. W. (1997) J. Cell Biol. 137, 93-104.

31. Greco, V., Hannus, M. & Eaton, S. (2001) Cell 106, 633-645.

32. Levy, J. A. (1998) HIV and the Pathogenesis of AIDS (Am. Soc. Microbiol., Washington, DC).

33. Cohen, O. J. & Fauci, A. S. (2001) in Fields Virology, eds. Knipe, D. M. & Howley, P. M. (Lippincott Williams & WIlkins, Philadelphia), Vol. 2, pp. 2043-2094.

34. Freed, E. O. & Martin, M. A. (2001) in Fields Virology, eds. Knipe, D. M. & Howley, P. M. (Lippincott Williams & Wilkins, Philadelphia), Vol. 2, pp. 1971-2041.

35. Douek, D. C., Brenchley, J. M., Betts, M. R., Ambrozak, D. R., Hill, B. J., Okamoto, Y., Casazza, J. P., Kuruppu, J., Kunstman, K., Wolinsky, S., et al. (2002) Nature 417, 95-98.

36. Demoustier, A., Gubler, B., Lambotte, O., De Goer, M. G., Wallon, C., Goujard, C., Delfraissy, J. F. & Taoufik, Y. (2002) AIDS 16, 1749-1754.

37. Hioe, C. E., Bastiani, L., Hildreth, J. E. & Zolla-Pazner, S. (1998) AIDS Res. Hum. Retroviruses 14 (Suppl. 3), S247-S254.

38. Pizzato, M., Blair, E. D., Fling, M., Kopf, J., Tomassetti, A., Weiss, R. A. & Takeuchi, Y. (2001) Gene Ther. 8, 1088-1096.

39. Pizzato, M., Marlow, S. A., Blair, E. D. & Takeuchi, Y. (1999) J. Virol. 73, 8599-8611.

40. Walker, S. J., Pizzato, M., Takeuchi, Y. & Devereux, S. (2002) J. Virol. 76, 6909-6918.

41. Green, P. L. & Chen, I. S. Y. (2001) in Fields Virology, eds. Knipe, D. M. & Howley, P. M. (Lippincott Williams & WIlkins, Philadelphia), Vol. 2, pp. 1941-1969.

42. Richardson, J. H., Edwards, A. J., Cruickshank, J. K., Rudge, P. & Dalgleish, A. G. (1990) J. Virol. 64, 5682-5687.

43. Kiprov, D. D., Sheppard, H. W. & Hanson, C. V. (1994) Science 263, 737-738.

44. Kiprov, D. D., Nachtigall, R. D., Weaver, R. C., Jacobson, A., Main, E. K. & Garovoy, M. R. (1996) Am. J. Reprod. Immunol. 36, 228-234.

45. Pedersen, N. C., Johnson, L., Birch, D. & Theilen, G. H. (1986) Vet. Immunol. Immunopathol. 11, 123-148.

46. Langlois, A. J., Weinhold, K. J., Matthews, T. J., Greenberg, M. L. & Bolognesi, D. P. (1992) AIDS Res. Hum. Retroviruses 8, 1641-1652.

47. Issel, C. J., Horohov, D. W., Lea, D. F., Adams, W. V., Jr., Hagius, S. D., McManus, J. M., Allison, A. C. & Montelaro, R. C. (1992) J. Virol. 66, 3398-3408.

48. Hoover, E. A., Mullins, J. I., Chu, H. J. & Wasmoen, T. L. (1996) AIDS Res. Hum. Retroviruses 12, 379-383.

49. Chan, W. L., Rodgers, A., Hancock, R. D., Taffs, F., Kitchin, P., Farrar, G. & Liew, F. Y. (1992) J. Exp. Med. 176, 1203-1207.

50. Stott, E. J., Chan, W. L., Mills, K. H., Page, M., Taffs, F., Cranage, M., Greenaway, P. & Kitchin, P. (1990) Lancet 336, 1538-1541.

51. Stott, E. J. (1991) Nature 353, 393.

52. Siebelink, K. H., Tijhaar, E., Huisman, R. C., Huisman, W., de Ronde, A., Darby, I. H., Francis, M. J., Rimmelzwaan, G. F. & Osterhaus, A. D. (1995) J. Virol. 69, 3704-3711.

53. Wang, S. Z., Rushlow, K. E., Issel, C. J., Cook, R. F., Cook, S. J., Raabe, M. L., Chong, Y. H., Costa, L. & Montelaro, R. C. (1994) Virology 199, 247-251.