Good to CU

Article metrics

A powerful arm of the cellular defence against microbial invaders has been characterized. APOBEC3G, a protein that can fight off HIV, works by introducing 'typographical errors' during viral replication.

HIV encodes a protein that allows the virus to multiply in otherwise resistant human cells1,2,3. This protein, Vif — for 'viral infectivity factor' — works by overcoming a cellular protein, APOBEC3G, whose task is to inhibit the replication of HIV and other retroviruses4. As they describe in Cell5 and on pages 94 and 99 of this issue6,7, three groups have now discovered how APOBEC3G hampers viral replication. The authors find that, as the retroviral RNA genome is copied into DNA in host cells, this protein changes 'C's to 'U's in the DNA. APOBEC3G therefore represents a remarkable innate cellular defence mechanism that drastically alters the chemical composition and coding capacity of replicating retroviruses.

Retroviruses, which include HIV and murine leukaemia virus (MLV), have genomes that are made up of single-stranded RNA. For these viruses to replicate in a host cell, a double-stranded DNA copy of the RNA genome must be made, by the process of reverse transcription. This happens in two steps. First, a DNA strand that has a sequence complementary to the RNA is produced (this is called 'minus-strand' DNA). Then the RNA is removed and a second (plus-strand) DNA molecule, complementary to the first (and therefore matching the sequence of the original RNA), is synthesized. The resulting double-stranded DNA is incorporated into the host-cell genome, and, when switched on, is used to generate the proteins and RNA needed to form new virus particles (virions).

It has been known for more than a decade that the Vif protein is required for HIV to replicate in some human cell types (termed 'non-permissive' cells), but not others ('permissive' cells)1,2,3. Non-permissive cells express APOBEC3G (also known as CEM15), and permissive cells become non-permissive upon expression of this protein4. These and other data indicate that APOBEC3G blocks viral replication, and that Vif works by overcoming this block. However, the precise nature of the replication block has not been clear — and it is unusual in that it is manifest after the virus has entered a new target cell. So, Vif-deficient viruses produced by non-permissive cells are released at normal levels, but are somehow impaired in their ability to form stable, productive replication intermediates when they enter new target cells3,8.

Clues to how this might happen have come from several observations. First, APOBEC3G can be packaged into HIV-1 virions4. Second, APOBEC3G is related to a family of enzymes that catalyse the deamination of cytosine bases (C's) in DNA and RNA, thereby producing uracils (U's). Finally, APOBEC3G can mutate the DNA of the bacterium Escherichia coli9. Thus it was suggested that APOBEC3G might inhibit HIV replication by altering the coding capacity of viral RNA or DNA4.

Harris et al.5, Mangeat et al.6 and Zhang et al.7 have now shown that APOBEC3G can indeed alter the sequences of the HIV and other retroviral genomes, producing an exceptionally high frequency of guanine-to-adenosine (G-to-A) substitutions in the plus (protein-coding) strand of the viral DNA. As shown in detail in Fig. 1, this observation implies that APOBEC3G changes C's to U's in the DNA minus strand during replication (because, when the plus strand is made using the minus strand as a template, G's are inserted opposite C's, but A's opposite U's). Importantly, G-to-A hypermutation is suppressed by the presence of Vif5,6,7,10, and APOBEC3G has the required deaminase activity5,7 and can act on single-stranded DNA substrates5. Deamination by APOBEC3G is quite promiscuous, but exhibits some sequence selectivity, with a preference for C/T–C–C sequences5,6,7.

Figure 1: Editing HIV: how the cellular APOBEC3G protein might inhibit HIV replication5,6,7,10.

From left, APOBEC3G has been packaged, together with the viral genome, in a viral particle. Then, when the retrovirus enters a new cell, the viral reverse transcriptase copies the genomic RNA into a double-stranded DNA. Replication begins at a tRNA primer (purple), and creates first a DNA minus (−) strand that is a complementary copy of the RNA, and then a DNA plus (+) strand that is a complementary copy of the minus strand. However, APOBEC3G deaminates 2′-deoxycytidines (C) in the minus strand, producing 2′-deoxyuridines (U). Plus-strand synthesis then converts these C-to-U changes into G-to-A mutations, because C pairs with G, whereas U pairs with A. C-to-U conversion can diminish HIV-1 replication in several ways: deglycosylation of uracil can lead to strand breakage and destruction; the presence of U's can reduce plus-strand synthesis11; and G-to-A hypermutations in the viral genome can impair all subsequent viral functions. The viral Vif protein blocks the effects of APOBEC3G through an as yet undefined mechanism.

Conversion of C's to U's in the genomic DNA minus strand apparently blocks viral replication in several ways (Fig. 1). First, the presence of uracils can target the minus strand for destruction. Uracil is not normally found in DNA and can therefore be excised by host DNA-repair enzymes. This could eventually lead to strand breakage and destruction — explaining the apparent instability of viral reverse transcripts. Second, uracils in the minus strand can impair the initiation of plus-strand synthesis during HIV-1 reverse transcription11. Finally, even when complete, double-stranded DNA copies of the viral genome are made, they contain many G-to-A mutations, which result in amino-acid changes and aberrant 'stop' signals in the encoded proteins, and generally reduce viral fitness at every subsequent stage in replication6. The biased G-to-A mutation profile might also explain the overall A-richness of the HIV-1 genome, as well as the non-lethal hypermutation sometimes observed during viral replication5,6,7,10.

As with all other immune responses, APOBEC3G must successfully distinguish between 'self' and 'non-self'. So it is remarkable that, as two of the groups show, the protein can inhibit the infectivity of retroviruses that differ markedly in sequence, including MLV5,6, simian immunodeficiency virus, equine infectious anaemia virus (EIAV), and even engineered retroviruses — derived from HIV-1 — that contain little of the original genome6. Presumably, these genomes are specifically targeted because APOBEC3G is incorporated into virus particles, and/or because only virus-specific intermediates are selected for deamination. It will therefore be important to learn how APOBEC3G is packaged into virions, and how it selects its substrates. Intriguingly, another member of this enzyme family, AID (activation-induced cytidine deaminase), which is involved in antibody-gene diversification, also targets single-stranded DNA. It requires transcription of the targeted gene into RNA12,13, and the presence of an RNA-degrading enzyme13. By analogy, APOBEC3G might target single-stranded minus DNA after reverse transcription and removal of the RNA template by the RNA-degrading activity of the reverse transcriptase enzyme. Alternatively, substrate targeting could be mediated by a cofactor, as is the case for another family member14.

Another issue, which could have therapeutic implications, is how Vif overcomes the antiviral activity of APOBEC3G. Possibilities include downregulation of APOBEC3G protein levels, exclusion of the protein from progeny virions, and inactivation of incorporated APOBEC3G molecules4. Moreover, do viruses such as MLV and EIAV, which do not have the Vif gene, avoid the effects of APOBEC3G by expressing other proteins with Vif-like activities? Or do they simply replicate only in host cells that lack APOBEC3G?

It is becoming increasingly apparent that cells have evolved a remarkable number of mechanisms to defend themselves from microbial invaders, and that microbes have discovered clever ways to circumvent those defences. Retroviruses are particularly good tools with which to study innate cellular immune systems, because the battle between cells and retroviruses is an ancient and intense one — as exemplified by the fact that retroviral elements make up an astonishing 8% of the human genome15. Other cellular defences against retroviruses include ZAP, a protein that targets viral messenger RNAs for destruction16, and Fv-1 (and related cellular proteins), which inhibits early stages of retroviral replication17,18. The molecular description of APOBEC3G activity5,6,7,10 has revealed yet another arm of the innate immune system, in which cells actually edit the genomes of invading retroviruses.


  1. 1

    Fisher, A. G. et al. Science 237, 888–893 (1987).

  2. 2

    Strebel, K. et al. Nature 328, 728–730 (1987).

  3. 3

    Gabuzda, D. H. et al. J. Virol. 66, 6489–6495 (1992).

  4. 4

    Sheehy, A. M., Gaddis, N. C., Choi, J. D. & Malim, M. H. Nature 418, 646–650 (2002).

  5. 5

    Harris, R. S. et al. Cell 113, 803–809 (2003).

  6. 6

    Mangeat, B. et al. Nature 424, 99–102 (2003).

  7. 7

    Zhang, H. et al. Nature 424, 94–98 (2003).

  8. 8

    von Schwedler, U., Song, J., Aiken, C. & Trono, D. J. Virol. 67, 4945–4955 (1993).

  9. 9

    Harris, R. S., Petersen-Mahrt, S. K. & Neuberger, M. S. Mol. Cell 10, 1247–1253 (2002).

  10. 10

    Lecossier, D., Bouchonnet, F., Clavel, F. & Hance, A. J. Science 300, 1112 (2003).

  11. 11

    Klarmann, G. J., Chen, X., North, T. W. & Preston, B. D. J. Biol. Chem. 278, 7902–7909 (2003).

  12. 12

    Chaudhuri, J. et al. Nature 422, 726–730 (2003).

  13. 13

    Bransteitter, R., Pham, P., Scharff, M. D. & Goodman, M. F. Proc. Natl Acad. Sci. USA 100, 4102–4107 (2003).

  14. 14

    Mehta, A., Banerjee, S. & Driscoll, D. M. J. Biol. Chem. 271, 28294–28299 (1996).

  15. 15

    International Human Genome Sequencing Consortium Nature 409, 860–921 (2001).

  16. 16

    Gao, G., Guo, X. & Goff, S. P. Science 297, 1703–1706 (2002).

  17. 17

    Best, S., Le Tissier, P., Towers, G. & Stoye, J. P. Nature 382, 826–829 (1996).

  18. 18

    Hatziioannou, T., Cowan, S., Goff, S. P., Bieniasz, P. D. & Towers, G. J. EMBO J. 22, 385–394 (2003).

Download references

Author information

Correspondence to Wesley I. Sundquist.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Gu, Y., Sundquist, W. Good to CU. Nature 424, 21–22 (2003) doi:10.1038/424021a

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.