Restriction by APOBEC3 proteins of endogenous retroviruses with an extracellular life cycle: ex vivo effects and in vivo "traces" on the murine IAPE and human HERV-K elements

Abstract

Background: APOBEC3 cytosine deaminases have been demonstrated to restrict infectivity of a series of retroviruses, with different efficiencies depending on the retrovirus. In addition, APOBEC3 proteins can severely restrict the intracellular transposition of a series of retroelements with a strictly intracellular life cycle, including the murine IAP and MusD LTR-retrotransposons.

Results: Here we show that the IAPE element, which is the infectious progenitor of the strictly intracellular IAP elements, and the infectious human endogenous retrovirus HERV-K are restricted by both murine and human APOBEC3 proteins in an ex vivo assay for infectivity, with evidence in most cases of strand-specific G-to-A editing of the proviruses, with the expected signatures. In silico analysis of the naturally occurring genomic copies of the corresponding endogenous elements performed on the mouse and human genomes discloses "traces" of APOBEC3-editing, with the specific signature of the murine APOBEC3 and human APOBEC3G enzymes, respectively, and to a variable extent depending on the family member.

Conclusion: These results indicate that the IAPE and HERV-K elements, which can only replicate via an extracellular infection cycle, have been restricted at the time of their entry, amplification and integration into their target host genomes by definite APOBEC3 proteins, most probably acting in evolution to limit the mutagenic effect of these endogenized extracellular parasites.

Figure 1

Murine and human APOBEC3 proteins inhibit endogenous retroviruses. (A) Rationale of the assay for detection of infection events by endogenous retroviruses in the presence of APOBEC3 proteins. The IAPE-D and HERV-K elements used in the assay are marked with the neo reporter gene – inserted in reverse orientation – and carry their own functional genes, except for the env gene which is supplied in trans, thus allowing only for single rounds of infection. Human 293T cells are co-transfected with the indicated expression vectors for APOBEC3 family members, the supernatants collected 2-days post-transfection to infect HeLa target cells, and infection events detected upon G418 selection. (B) Analysis of the activity of murine and human APOBEC3 proteins on the indicated endogenous retroviruses. Viral titers are given as percentages relative to a control (no apobec: expression vector with a nonfunctional hA3G; 622 and 549 G418^R clones/ml for IAPE-D and HERV-K, respectively). Data are the means ± standard deviations (s.d.) for at least three independent experiments. Bottom: retrotransposition frequency of an active autonomous IAP element marked with a neo indicator gene for retrotransposition [17] in the presence of the corresponding APOBEC3 proteins; the assay was performed by cotransfection of HeLa cells with the marked IAP and APOBEC expression vector as previously described [26]; values are the means ± standard deviations (s.d.) for at least three independent experiments and are given as percentages relative to the control (no apobec; 1.3 × 10^-3 G418^R clones/cell).

Figure 2

APOBEC3 proteins induce specific G-to-A hypermutations. Two-entry tables showing nucleotide substitution preferences in the presence of the indicated APOBEC3 proteins for the IAPE-D and HERV-K integrated proviruses. n, total number of bases sequenced. The adjacent graphs represent the relative frequencies of observed G-to-A mutations as a function of the G neighboring nucleotides (+2 position for the expected mA3 footprint, +1 position for the other APOBEC3s); for the two-entry tables, p-values calculated by a Poisson regression in a log-linear model for the occurrence of the G-to-A versus C-to-T mutations yielded p < 0.03 in all cases (except for hA3DE (p = 0.18) due to the low number of mutations); for the adjacent graphs, p-values calculated by a chi square test were p < 0.01 in all cases (except again for hA3DE, p = 0.7); similar levels of significance (or even higher) were obtained using the Kruskal Wallis test.

Figure 3

Distribution of the nucleotide substitutions in the IAPE-D and HERV-K genomic copies residing in the mouse and human genomes. Endogenous sequences were extracted from the mouse and the human genome databases, aligned and compared to the derived consensus. (A) Upper panels: percentage of substitutions for each nucleotide, for the endogenous IAPE-D and HERV-K elements (with the IAPE-A elements used as a control). Lower panels: two-entry tables showing nucleotide substitutions preferences, with the G-to-A values in bold (higher that the "non-specific" C-to-T value for IAPE-D and HERV-K, and identical in the case of the IAPE-A control). n, total number of nucleotides analyzed. (B) Influence of nucleotides at position -2, -1, +1 and +2 on G-to-A mutations (the mutated G is at position 0). Data represent the percentage of indicated target di- or trinucleotide sequences bearing G-to-A mutations. X represents any nucleotide. P-values calculated for the two-entry tables (in A) by a Poisson regression in a log-linear model for the occurrence of the G-to-A versus C-to-T mutations yielded p < 0.003 for IAPE-D and HERV-K; in B, p-values calculated by a chi square test were p < 0.001; similar levels of significance were obtained using the Kruskal Wallis test. (C) Example of G-to-A mutations present in twenty IAPE-D (upper panel) and HERV-K (lower panel) sequences. GXA trinucleotides and GG dinucleotides are underlined in the consensus sequence of IAPE-D and HERV-K, respectively.

Figure 4

Variability of the number of G-to-A mutations within the endogenous IAPE-D and HERV-K proviruses depending on the element. The total numbers of G-to-A mutations (plain + hatched grey bars) for each IAPE-D (A) and HERV-K (B) proviruses are represented, together with that of the C-to-T (black bars) "none-strand-specific" mutations, given as an internal control (also used for ordering the elements) indicative of the genetic drift-associated age-dependent amount of mutations for each copy (same rank order as the sum of all non-G-to-A base substitutions). The number of G-to-A mutations specifically associated with the mA3 or hA3G APOBEC footprints is indicated with hatched grey. (C) Positioned G-to-A mutations (red bars) specifically associated with the hA3G APOBEC footprint ("GG") for the individual HERV-K elements in (B); yellow bars correspond to deletions in the proviruses (relative to the Phoenix consensus sequence). The data and the image shown in figure 4C were generated using the Hypermut 2.0 software available at the website.