Intracellular targeting of telomeric retrotransposon Gag proteins of distantly related Drosophila species

Abstract

The retrotransposons that maintain telomeres in Drosophila melanogaster have unique features that are shared across all Drosophila species but are not found in other retrotransposons. Comparative analysis of these features provides insight into their importance for telomere maintenance in Drosophila. Gag proteins encoded by HeT-A^mel and TART^mel are efficiently and cooperatively targeted to telomeres in interphase nuclei, a behavior that may facilitate telomere-specific transposition. Drosophila virilis, separated from D. melanogaster by 60 MY, has telomeres maintained by HeT-A^vir and TART^vir. The Gag proteins from HeT-A^mel and HeT-A^vir have only 16% amino acid identity, yet several of their functional features are conserved. Using transient transfection of cultured cells from both species, we show that the telomere association of HeT-A^vir Gag is indistinguishable from that of HeT-A^mel Gag. Deletion derivatives show that organization of localization signals within the two proteins is strikingly similar. Gag proteins of TART^mel and TART^vir are only 13% identical. In contrast to HeT-A, surprisingly, TART^vir Gag does not localize to the nucleus, although TART^vir is a major component of D. virilis telomeres, and localization signals in the protein have much the same organization as in TART^mel Gag. Thus, the mechanism of telomere targeting of TART^vir differs, at least in a minor way, from that of TART^mel. Our findings suggest that, despite dramatic rates of protein evolution, protein and cellular determinants that correctly localize these Gag proteins have been conserved throughout the 60 MY separating these species.

Drosophila melanogaster has unusual telomeres formed by retrotransposons (reviewed in ref. 1). Because most species have telomerase-maintained telomeres, it is important to determine how D. melanogaster has come to use a variant mechanism for serving the same telomere functions. To learn more about the evolution of the retrotransposon telomere we have studied Drosophila virilis, separated from D. melanogaster by 60 MY (2–4).

D. melanogaster telomeres are composed of two non-LTR retrotransposons, HeT-A and TART, with few copies of Tahre (5), an element that appears to combine parts of both HeT-A and TART. We have found homologs of both HeT-A and TART in D. virilis (Fig. 1). Because retrotransposons have rapidly changing sequences, we used low-stringency hybridization with the most conserved part of the D. melanogaster telomere sequences (2), the TART pol gene, to clone a fragment of D. virilis DNA. This homologous probe was used to screen a genomic library and obtain large clones containing tandem arrays of TART^vir elements. One TART^vir was adjacent to a HeT-A^vir element. This HeT-A^vir sequence was used to obtain clones with more elements in tandem arrays. We have not detected a Tahre homolog, although we have found one copy of Uvir, a novel element apparently formed by replacement of the HeT-A Gag gene with sequence encoding reverse transcriptase. The Uvir element was found within a tandem array of HeT-A^vir elements.

Fig. 1.

Comparison of telomere retrotransposons in D. melanogaster and D. virilis, showing conserved features and nucleotide identities of specific regions. Diagrams of HeT-A and TART homologs, drawn approximately to scale. Solid bars on the right indicate phylogenic relationships. %, percentage of nucleotide identity of sequence between arrowheads; for HeT-A, both the Gag coding region and the entire element are compared; for TART, only the Gag and Pol regions are similar enough to calculate % identity: dashed lines under HeT-A indicate irregular A-rich repeats: striped lines under TART^mel indicate position of perfect nonterminal repeats (PNTR); dotted lines under TART^vir genes indicate concentrations of polyglutamine repeats. X, extra domain of pol coding region. (D. melanogaster has multiple subfamilies of TART. TART^mel represents a consensus of these subfamilies). Note that the noncoding regions of HeT-A are more highly conserved than the coding regions.

HeT-A and TART elements have undergone significant sequence divergence between D. melanogaster and D. virilis; nevertheless, in both species, these elements transpose specifically forming head-to-tail arrays on chromosome ends. Furthermore, in analyses of the relationship of these sequences to other elements from the same phylogenetic clade, the resulting tree shows that these elements are true homologs of HeT-A and TART. Although there may be other subfamilies of these elements in the D. virilis genome, we have not found any in the sequence now in the database.

Studies of the telomere retrotransposons in D. melanogaster show that HeT-A and TART both have unusual features not seen in other non-LTR retrotransposons (Fig. 1). Both elements have unusually long 3′ UTRs. HeT-A's 3′ UTR has an irregular pattern of A-rich repeats. HeT-A does not encode reverse transcriptase. TART has a large perfect nonterminal repeat (PNTR, sequence at the 5′ end that is perfectly repeated near, but not at, the 3′ end). An important feature of HeT-^mel and TART^mel is the interactive targeting of their Gag proteins to chromosome ends in interphase nuclei, providing a potential explanation for the telomere-specific transpositions of these elements. This targeting depends on sequences in several parts of the Gag proteins (6). It also depends on cell-type-specific contributions of host cells. For example, polyploid larval cells do not support nuclear localization of HeT-A^mel Gag, which, instead, accumulates in cell-type-specific cytoplasmic locations (7).

The telomere transposons of D. virilis have conserved many of these unusual features (Fig. 1). At the level of resolution in Fig. 1, HeT-A^vir is nearly identical to HeT-A^mel; TART^vir, on the other hand, differs from TART^mel in several ways. TART^vir lacks the PNTRs in the 5′ and 3′ UTRs of TART^mel, has a much shorter 3′ UTR, and its pol gene encodes an extra domain of unknown function, the X domain [seen in the closely related Drosophila americana but not in TART from other species (2)]. This description is based on two complete and two partial TART^vir elements from two phage clones. As mentioned, repeated searches in the database of D. virilis sequences have not revealed other subfamilies of TART^vir

The extensive sequence divergence between D. melanogaster and D. virilis is at odds with the important function of these elements in Drosophila species. In the studies reported here, we investigate how this extensive divergence has affected function by studying the intracellular targeting of HeT-A and TART Gags in D. virilis and D. melanogaster cells.

Results

The D. virilis WR Dv-1 Cell Line Has HeT-A and TART Elements Similar to Those in D. virilis Flies.

Our early studies on the telomere retrotransposons in D. virilis were on flies (2, 3). For the transfection experiments reported here we used the cultured D. virilis cell line, WR Dv-1. All of the D. melanogaster cell lines that we have studied have multiple HeT-A and TART elements. Genomic hybridization shows that WR Dv-1 cells also have HeT-A and TART elements. As expected from the low level of sequence homology between these elements in the two species of flies, neither HeT-A nor TART gag probes give detectable cross-hybridization with DNA of cells from the other species [see supporting information (SI) Fig. 6].

HeT-A^vir Gag Is Targeted to Telomeres in Both D. virilis and D. melanogaster Cells.

One of the remarkable properties of the HeT-A^mel element is the specificity with which its Gag protein localizes to telomeres in interphase cells (Fig. 2A). This localization has not been seen for Gag proteins from nontelomeric retrotransposons (8) and is likely to reflect HeT-A's role in telomere maintenance. To see whether this property has been conserved in HeT-A^vir, we transiently transfected D. virilis cells with a construct expressing HeT-A^vir Gag protein fused to GFP. As in our experiments with D. melanogaster, the Gag protein efficiently entered the nucleus and formed many small dots (Het dots), strongly suggesting that telomere targeting of this protein has been conserved in D. virilis (Fig. 2B).

Fig. 2.

Confocal images of cells transiently transfected with HeT-A Gag-GFP constructs. In each experiment, transfected cells have Gag-GFP localized to dots in the nucleus. The number of HeT-A dots is smaller in D. melanogaster SL2 cells (A and C) than in WR Dv-1 cells (B and D), whether the cell is expressing HeT-A^mel Gag or HeT-A^vir Gag. DNA in all cells is stained with DAPI (blue). (Left) Summation of all planes taken with the GFP filter and merged with the DAPI image. (Right) Transmission image merged with DAPI.

D. melanogaster studies have shown that the Het dots in that species are associated with telomeres, because the dots colocalize with the telomere-associated protein, HOAP (9). We do not have useful markers for D. virilis telomeres; however, the similarity of the localization of HeT-A^vir Gag with that of HeT-A^mel Gag suggests that HeT-A^vir Gag is also targeted to telomeres. The conclusion that Het dots are associated with telomeres is supported by analysis of the number of nuclear dots produced by the two proteins in both D. virilis and D. melanogaster cells. The analysis was done by confocal microscopy because the fluorescence microscope used in the other experiments captures only a fraction of the dots in each optical section. The multiple planes obtained with the confocal microscope revealed more dots and showed that some of the telomeres were under others and therefore obscured in a nonconfocal image. The number of nuclear dots should be related to the number of telomeres. Therefore, if both proteins give similar numbers of dots in the same cell type, we can assume that both are at the telomeres.

D. virilis cells have more telomeres than D. melanogaster cells. D. virilis has six chromosomes (24 ends in a diploid cell before replication), whereas D. melanogaster has four chromosomes (16 ends in a diploid cell). As expected from nonconfocal pictures, confocal pictures show that the two Gags form similar Het dots in nuclei of either cell line; however, the number of dots depends on the species of the transfected cell and not the species of the HeT-A Gag. In both cell types, some dots are bigger than others; probably because of failure to resolve closely spaced or fused dots. Fused dots might be expected because the Gag proteins in these experiments are overexpressed and have a tendency to form aggregates. Nevertheless, we found that confocal images of both HeT-A Gag proteins showed substantially more dots in D. virilis cells (Fig. 2B) than in D. melanogaster cells (Fig. 2A).

HeT-A is telomere-specific in D. virilis, as it is in D. melanogaster, so Het-dot targeting of its Gag protein is not surprising. However, we were surprised to find that HeT-A^vir Gag formed Het dots when it was expressed in D. melanogaster cells (Fig. 2C) and that HeT-A^mel formed Het dots in D. virilis cells (Fig. 2D). Thus, despite their very different amino acid sequences, both proteins interacted appropriately with cellular targeting proteins in the other species. We therefore conclude that the two HeT-A Gag proteins are targeted to telomeres in both homologous and heterologous cells.

HeT-A^vir Does Not Form Cytoplasmic Het-Bodies in Either D. virilis or D. melanogaster.

In addition to forming nuclear Het dots, HeT-A^mel Gag forms a characteristic cytoplasmic structure, a Het body, in some transfected cells. Het bodies are round or oval with smooth edges. Typically, there is only one per cell and there is no association with a known cytoplasmic organelle to suggest function for the bodies. These bodies may indicate saturation of some part of the system for transporting Gag into the nucleus, because Het bodies are seen only in cells that have formed nuclear dots and are larger in cells with more Gag protein. We have not seen similar structures from Gags of other elements, including TART. HeT-A^mel Gag forms Het bodies in both D. melanogaster (8) and D. virilis cells (Fig. 3A). Surprisingly, HeT-A^vir Gag does not form Het bodies in either cell line. This is the most obvious difference in the behavior of the two proteins. The two proteins appear to express at similar levels in both cells, and therefore it is possible that the formation of the Het bodies is a particularity of HeT^mel Gag.

Fig. 3.

WR Dv-1 cells expressing HeT-A^mel Gag-GFP or TART^vir Gag-GFP. (Left) GFP+DAPI. (Right) DIC+DAPI. (A) Daughters of a cell expressing HeT-A^mel Gag-GFP show nuclear dots and also very large cytoplasmic Het bodies. It appears that the Het bodies did not segregate at cell division, but both remained in the lower cell. These are examples of very large Het bodies, and they are overexposed to reveal the nuclear dots. (B) Cells expressing TART^vir-GFP have no detectable nuclear label.

TART^vir Gag Does Not Appear to Enter Nuclei in D. virilis Cells, Although TART^mel Gag Moves Efficiently into Nuclei in Both D. melanogaster and D. virilis Cells.

In contrast to HeT-A^vir Gag, TART^vir Gag does not behave like its D. melanogaster homolog. In studies of D. melanogaster, TART^mel Gag moved efficiently into the nucleus but formed loose clusters that did not preferentially associate with telomeres unless coexpressed with HeT-A^mel Gag (9). Surprisingly, the localization of TART^vir Gag is very different. TART^vir Gag made characteristic clusters that spread throughout the cytoplasm but did not seem to enter the nucleus (Fig. 3B). The difference between the two Gags is determined by the proteins themselves and not the cells in which they are expressed. TART^vir Gag clusters were cytoplasmic in both D. virilis (Fig. 3B) and D. melanogaster cells, whereas TART^mel Gag clusters were nuclear in both cell types (data not shown).

Organization of Functional Regions Within the Gag Proteins Resembles That of Their Homologs in the Other Species.

Deletion derivatives of the D. melanogaster Gag proteins have been used to identify regions of the protein capable of activities such as nuclear localization or clustering of the protein. These studies showed that HeT-A^mel Gag and TART^mel Gag had a similar organization of these regions (6). To compare the internal organization of HeT-A^vir and TART^vir Gag proteins with that of their D. melanogaster homologs, we constructed and expressed deletion derivatives (Fig. 4;. see C and F for maps). Intracellular localizations of these deletion proteins showed striking similarities to those of analogous regions of the D. melanogaster homologs. D. virilis deletion proteins had the same localizations in both D. virilis and D. melanogaster cells, showing their ability to interact with cells of either species.

Fig. 4.

WR Dv-1 cells expressing deletion derivatives of HeT-A^vir Gag-GFP or TART^vir Gag-GFP. Deletion derivatives show that both ends of HeT-A^vir Gag enter the nucleus, although their distribution within the nucleus differs. Only the N-terminal end of TART^vir Gag enters the nucleus. The C-terminal end forms cytoplasmic clusters. (A) HeT-A^vir Gag, HV (1–462). GFP; lamin (red); merged GFP, lamin and DAPI; merged DIC and DAPI. (B) HV (463–907), GFP; merged DIC and DAPI. (C) Diagrams of HeT-A^mel and HeT-A^vir Gag proteins showing MHR (single gray box), zinc knuckles (three white boxes), and glutamine-rich region (black dots below line). Gray bars below each protein show sequence in each deletion derivative with first and last amino acids. (D) TART^vir Gag, TV (1–440) GFP; lamin (red); merged GFP, lamin (red) and DAPI; merged DIC and DAPI. (E) TV(441–1039) GFP; merged DIC and DAPI. Note that lamin staining in A and D is exterior to the GFP signal, indicating that the Gag-GFP fusion is in the nucleus. (F) Diagrams of TART^mel and TART^vir Gag proteins with deletion derivatives, marked as in C. Transient transfection of the construct TV (1–703) is shown in Fig. 5 C and D.

HeT-A^vir Gag contains nuclear localization signals in both N-terminal and C-terminal ends as does the orthologous HeT-A^mel protein, HM (1–534). However there were some differences in localization. N-terminal HV (1–462) was largely concentrated just inside the nuclear membrane (Fig. 4A), whereas HM (1–534) was more evenly distributed in the nucleus. C-terminal HV (463–907) efficiently entered the nucleus (Fig. 4B), where it formed many clusters. This localization resembled that of the analogous HM (589–921).

The N-terminal derivative of TART^vir Gag, TV (1–440), entered the nucleus, spreading without visible clusters (Fig. 4D), whereas the C-terminal fragment of TART^vir Gag, TV(441–1039), remained in the cytoplasm (Fig. 4E). This behavior was like that of the analogous TART^mel fragments. However, the C-terminal end of TART^vir Gag has an activity not seen for TART^mel Gag; it determines the cytoplasmic localization of intact TART^vir Gag. A deletion derivative lacking the very last ≈200 aa, TV (1–703), entered the nucleus and spread evenly (Fig. 5 C and D). This C-terminal region may act as a nuclear export or a cytoplasmic retention signal or, alternatively, affect the conformation of the full-length protein to block the nuclear localization signal at the N terminus.

Fig. 5.

WR Dv-1 cells expressing cotransfected Gag-CFP and Gag-YFP proteins. Colocalization shows that HeT-A^mel Gag interacts with TART^mel Gag to move TART^mel Gag into Het dots (A), but HeT-A^vir Gag does not affect localization of TART^vir Gag (B). The deletion protein TV (1–703) interacts with HeT-A^vir Gag to interfere with its formation of Het-dots (C), but TV (1–703) does not affect the localization of HeT-A^mel Gag (D).

The C-terminal region contains the largest cluster of polyglutamine repeats in TART^vir Gag. Polyglutamine repeats are a distinguishing feature of both D. virilis Gag proteins and have no counterparts in their D. melanogaster homologs. These repeats are also found in other D. virilis proteins. Their functions are not known, but such repeats have been implicated in protein–protein interactions. It is interesting that removal of ≈200 aa containing a significant repeat causes the protein to spread evenly in the nucleus rather than clustering in the cytoplasm. We have not determined whether polyglutamine is responsible for either the clustering or the cytoplasmic localization, but it is the most obvious sequence difference from the equivalent region of TART^mel Gag.

TART^mel Gag Interacts with HeT-A^mel Gag in Both D. melanogaster and D. virilis Cells, Whereas TART^vir Gag and HeT-A^vir Gag Interact Only if TART^vir Gag Is Forced to Enter the Nucleus.

Although TART^mel Gag moves into the nucleus very efficiently, it does not localize to telomeres unless coexpressed with HeT-A^mel Gag. In D. melanogaster, HeT-A^mel Gag moved TART^mel Gag into telomeric Het dots (9), showing that the two proteins interact. This interaction and telomere localization also occurred in D. virilis cells (Fig. 5A), suggesting that either the two proteins interact without intervention of other cellular components or that necessary cellular components have been functionally conserved for the 60 MY that separate the two Drosophila species.

The interaction between the two D. melanogaster Gags suggested that the two D. virilis Gags might interact and that this interaction might be necessary to move TART^vir Gag into the nucleus, as well as localizing it to telomeres. However, this does not seem to be the case. When TART^vir Gag was cotransfected with HeT-A^vir Gag, each of the proteins had the same localization seen in single transfections, showing no evidence that the proteins interact (Fig. 5B). There was also no interaction between HeT-A^vir Gag and TART^vir Gag when coexpressed in D. melanogaster cells.

Comparisons of the Gag proteins show that the most conserved amino acid sequences of both HeT-A Gag and TART Gag are in the region resembling the major homology region (MHR)-zinc knuckle region of retroviral Gag proteins (SI figs. 7 and 8). This region has been shown to be necessary for interaction between HeT-A^mel Gag and TART^mel Gag (6). The conservation of this region suggested that the two D. virilis Gag proteins might interact if TART^vir Gag could be moved into the nucleus. To accomplish this, we coexpressed HeT-A^vir Gag with a deletion derivative of TART^vir Gag, TV (1–703), containing the putative interaction region but lacking the C-terminal region that prevents entry into the nucleus.

TV (1–703) entered the nucleus but showed no tendency to form dots, although it has the MHR-zinc knuckle region and the equivalent TM (1–793) does form clusters. TV (1–703) localization was not affected by coexpression with either HeT-A^vir Gag or HeT-A^mel Gag (Fig. 5 C and D). However, there was evidence that TV (1–703) interacted with HeT-A^vir Gag, because in the coexpression HeT-A^vir Gag was inhibited from forming nuclear dots. This interaction between the two D. virilis proteins was seen in both D. virilis and D. melanogaster cells.

Importantly, TV (1–703) did not interact with HeT-A^mel Gag (Fig. 5D). HeT-A^mel Gag formed characteristic Het dots, whereas the TART^vir deletion remained dispersed in the nucleus. In cells with cytoplasmic Het bodies, there was no detectable TART^vir associated with this structure. This result argues that, although the putative interaction region does show amino acid sequence conservation, changes have occurred that allow HeT-A and TART Gags to interact within a species but not between species.

Discussion

The Telomeric Targeting of HeT-A Gag Is Conserved Despite a Very High Level of Amino Acid Sequence Difference.

HeT-A and TART are remarkable because their coding sequences show a close relationship to members of the jockey clade of non-LTR retrotransposons in Drosophila (4, 10), yet their transposition targets are the opposite of those of other members of the clade. HeT-A and TART never transpose into the euchromatic chromosome arms, yet other elements are found at many sites in these arms. Conversely, other elements never transpose into the long HeT-A/TART arrays that make up the telomeres (11). The basis for telomere-specific transposition is not known; however, the unique telomere targeting of the HeT-A Gag protein in D. melanogaster seems to offer at least a partial explanation. The work reported here demonstrates that this telomere targeting of HeT-A Gag is conserved in the distantly related D. virilis, despite significant differences in amino acid sequence.

In HeT-A^mel Gag, the determinants of telomere targeting are not confined to a specific region of the protein but are distributed along the protein (6). Therefore, conservation of telomere targeting in HeT-A^vir Gag must respond to strong selective pressure over each of these multiple determinants. It might be supposed that the divergence in the amino acid sequence is driven by need to coevolve with cellular components acting in the several steps by which Gag moves from the ribosome to the telomeric Het dot. However, our results argue that this is not a major factor in the sequence changes; HeT-A Gag from either species shows appropriate targeting in cells of both species. In this case, conservation of protein structure may be more important than conservation of amino acid sequence.

Studies on D. melanogaster show that homomultimerization of HeT-A Gag depends on the MHR-zinc knuckle region of the protein to form Het dots. However, these studies give no hint to which interactions are involved in telomere association (6). The only D. melanogaster protein known to remain predominantly telomere-associated in interphase nuclei is HOAP (9, 12). Het dots are telomere-associated only in interphase; colocalization of Het dots with clusters of HOAP provides strong evidence that the Het dots are at telomeres. However these studies provide two kinds of evidence that HeT-A Gag does not bind HOAP (7). First, although both Het dots and HOAP clusters vary in size, the size of a Het dot frequently does not correlate with the associated HOAP cluster. If the proteins bind, we would expect their amounts to vary together. Second, when cells were centrifuged onto slides to break nuclei and spread the DNA, Het dots and HOAP still associated closely, showing that the association can withstand the force that broke the nuclei. However, this association does not appear to be due to direct contact between Gag and HOAP. The two proteins remain close together but can be slightly separated by the force of centrifugation, suggesting both contact something else. The HOAP gene is one of the most rapidly changing genes known (13), and no D. virilis equivalent has been reported. Thus, we cannot currently test for an association of HeT-A^vir Gag with HOAP from D. virilis. However, the observation that the HeT-A^vir protein forms Het dots that are indistinguishable in number and distribution from those formed by HeT-A^mel Gag in the same cell line strongly suggests that HeT-A^vir Gag localizes at the telomeres in both cell types.

The Mechanism of TART Localization Is Less Conserved.

HeT-A and TART have been found together on telomeres in all Drosophila species and stocks that have been studied. These two non-LTR elements have closely related Gag proteins but are different in ways that suggest that they evolved from different ancestors. We have suggested TART became a telomere-specific element when it acquired a Gag protein that allowed it to be targeted to telomeres and that the most likely source of this protein is transfer from a telomere-specific HeT-A (14), although it could have been inherited from a common ancestor (15). We suppose that TART acquired the new Gag well before the separation of D. melanogaster and D. virilis, because both elements have been found at telomeres in all Drosophila studied. HeT-A and TART are interdependent. TART^mel Gag needs HeT-A^mel Gag for correct localization to telomeres in D. melanogaster cells, whereas HeT-A elements do not encode a reverse transcriptase and might therefore depend on TART for this activity. As expected from their apparent relationship, the Gag proteins from both TART^vir and HeT-A^vir are about equally diverged from their D. melanogaster homologs (4). However, unlike HeT-A^vir, TART^vir has differences from its homolog that can be seen by inspecting its sequence organization (see Fig. 1).

The work presented here shows additional differences between TART^vir and TART^mel. TART^mel Gag moves efficiently into the nucleus, whereas TART^vir Gag remains in clusters throughout the cytoplasm. We believe that TART^vir Gag proteins in our experiment are fully functional. They were expressed from two full-length gag genes, one from each of the two phage clones isolated in our screen. It seems unlikely that both genes would be defective. Neither protein was seen to enter the nucleus. Nor was HeT-A^vir Gag able to transport TART^vir Gag into the nucleus. We note that, in neither of these transfection experiments, could we be certain that no TART^vir Gag entered the nucleus. However, in equivalent experiments with D. melanogaster elements TART^mel Gag is efficiently moved into the nucleus, showing a significant difference from the localization of TART^vir Gag. This difference is inherent in the Gag proteins because the behavior is not changed by expression in cells of the other species.

The difference between the localization of TART^vir Gag and that of TART^mel Gag is surprising. Sequences of the two proteins have diverged significantly, but the difference is about equal to the difference between HeT-A^vir Gag and HeT-A^mel Gag, and the latter proteins have strikingly similar patterns of localization. Sequence comparisons show that TART^vir Gag is evolving under selection and coevolving with HeT-A^vir Gag in the MHR and zinc knuckle regions (4). This pattern of sequence change suggests that Gag is important to TART^vir and that it acts in partnership with HeT-A^vir Gag. If the partnership resembles the partnership in D. melanogaster, the interaction should be detected in the nucleus. However we were able to introduce significant amounts of TART^vir Gag into the nucleus only by deleting the C-terminal end of the protein, which had been shown to prohibit nuclear import or retention of TART^vir Gag. This deleted protein retained the most conserved region of the protein, the MHR-zinc knuckle region which is involved in the HeT-A^mel/TART^mel interactions. This shorter protein did interact with HeT-A^vir Gag but the interaction prevented HeT-A^vir Gag from localizing to telomeres. The mislocalization may have been the result of defective heteromultimers caused by the deleted TART^vir Gag. Nevertheless the experiment demonstrates an interaction between the two proteins. Importantly, this interaction is species-specific; the deleted protein did not affect localization of HeT-A^mel Gag when the two were coexpressed. This specificity is reminiscent of the selectivity of retroviral Gags in forming heteromultimers (16, 17). We note that the TART^vir Gag derivative used in these experiments contained the MHR-zinc knuckle region important for multimerization of retroviral Gag proteins (18–20) and also the interaction of HeT-A^mel Gag with TART^mel Gag (6).

One possible explanation for the unexpected TART^vir Gag localization might be that localization was studied by overexpressing a tagged protein. Overexpressed TART^mel Gag moves into the nucleus. Perhaps at some point in its journey to the nucleus, TART^vir Gag requires a cell component that is present in limiting amounts, and that component is overwhelmed by the overexpression of Gag. If this explanation is correct, TART^vir Gag localization may be accomplished by only a minor modification of that used by TART^mel Gag. If a complete TART^vir Gag could enter the nucleus, it might interact appropriately with HeT-A^vir Gag. (This discussion is based on the hypothesis that TART Gags must enter the nucleus to enable transposition of this element. If TART^vir can transpose without a nuclear Gag, the cytoplasmic localization seen in these experiments may be appropriate.) The results reported here allow us to conclude only that the mechanism of TART Gag localization has diverged more than that of HeT-A Gag.

We note that TART^vir may not be the only source of enzyme activity for transposition of HeT-A^vir. The U^vir element contains a pol gene flanked by 5′ and 3′ UTR sequences highly similar to HeT-A^vir. This association suggests that U^vir Pol could also supply enzyme activity for HeT-A^vir. We have suggested that U^vir may have been fortuitously formed from the mRNA of a cellular gene (3) and not a bona fide element. Currently available D. virilis sequence is still consistent with this possibility. The putative cellular gene might have been the original source of enzyme activity for HeT-A and, in fact, may still provide this activity in D. virilis.

Despite Significant Amino Acid Sequence Divergence Among the Telomere Retrotransposon Gags, the Organization of Functional Domains Tends to Be Conserved.

Comparison of the amino acid sequences for HeT-A and TART Gags shows the highest similarity in the regions of two motifs, the MHR and the zinc knuckles, that characterize retroviral Gag and are found in the jockey clade of non-LTR retrotransposons to which HeT-A and TART belong (6). The zinc knuckles are located roughly one-third of the length of the protein from the C terminus, and the MHR is slightly N-terminal of this. Sequences before and after these regions are more diverse and lack motifs that would give clues to function. However, study of deletion derivatives of HeT-A^mel Gag and TART^mel Gag identified functions associated with different domains of the sequence and showed that the linear order of these domains in the two proteins was similar.

Analogous deletion derivatives of the D. virilis Gag proteins reveal similarity of organization, both between the two Gags within a species and between homologs in the two species. N-terminal fragments of all four Gags localize to the nuclei, as do C-terminal fragments of the HeT-A Gags. C-terminal fragments of TART Gags remain in the cytoplasm. N-terminal derivatives of all four Gag proteins tend to spread evenly in the nucleus, whereas C-terminal derivatives have a tendency to form clusters; HeT-A Gags form clusters in the nucleus, and TART Gags form clusters in the cytoplasm. This organization indicates that sequence divergence is occurring without much rearrangement of regions of peptide sequence, suggesting that this organization is important for proper function of the telomere Gags. In fact, we believe that the conservation of several small motifs along the proteins might be responsible for the grouping of the Gag proteins of all HeT-A and TART homologs onto the same subclade (bootstrap value 82, see SI Fig. 9) when compared with other Gag proteins from the Jockey clade (10).

Concluding Remarks.

Comparisons of the telomeres in different species of Drosophila show that, although the mechanism of telomere maintenance is largely conserved, both nucleotide sequence and amino acid sequences change rapidly. As mentioned earlier, protein structure may be more important than amino acid sequence in this case. The low percentage of identity among the different homolog sequences was unexpected for genes that are directly related to an essential cellular role as complex as telomere maintenance. This presents an interesting parallel to the rapidly evolving sequences of centromeres (21, 22). Perhaps rapid evolution is a clue to the mechanism by which telomeres are maintained. If the telomere Gag proteins need to interact with rapidly evolving proteins, such as HOAP, to fulfill their telomere function, the importance of the telomere would explain the strong conservation of function, whereas the sequence change would be driven by the need to interact with other rapidly changing proteins.

Materials and Methods

Cell Lines.

The D. melanogaster SL2 cell line and the D. virilis cell line WR Dv-1 were established by I. Schneider (23). Cells were maintained in Schneider media [Sigma (St. Louis, MO) for SL2 and Invitrogen (Carlsbad, CA) for WR Dv-1] supplemented with 10% FBS (GIBCO, Carlsbad, CA) at 25°C. WR Dv-1 cells were from the Drosophila Genomics Resource Center (Indiana University, Bloomington, IN).

Southern Hybridizations.

Genomic DNA was extracted from SL2 and WR Dv-1 cells. Aliquots of DNA were digested with SalI +EcoRI and HindIII. Southern blot hybridization was under low-stringency conditions (24), with sequence of the Gag genes as probes. Probes were labeled with [³²P]dCTP, Ready to Go labeling beads (Amersham, Piscataway, NJ).

Recombinant DNA.

Fusion protein constructs were made by PCR amplification of Gag sequence from phage clones V8, V2, and V3 (see refs. 2 and 3) with specific primers. The 5′ primers had a methionine codon added if needed. The 3′ primers had the stop codon deleted and minimum changes introduced in the sequence to obtain the desired fusion protein. Constructs were cloned in the pPL17 vector that encodes the desired fluorescent protein, GFP, CFP, or YFP under the armadillo promoter (see ref. 8).

Transient Transfections.

Cells were seeded at 2 × 10⁶ cells per ml 12 h before transfection and transfected by using the calcium phosphate method (25) with 5–10 μg of DNA. For WR DV-1 cells, a 2.5-min DMSO shock (10% DMSO in cell culture media) was performed 5–7 h after transfection. Slides were made 48 and 72 h after transfection.

Slide Preparation and Microscopy.

Cell suspensions were placed on concavalin-treated coverslips, fixed 10 min with 4% paraformaldehyde in PBT, washed with PBT and with PBS, and mounted with mowiol/DAPI medium. Fluorescent fusion proteins were analyzed in an Eclipse E1000 fluorescent microscope (Nikon, East Rutherford, NJ) and a confocal microscope TCS-SP2-AOBS (Leica, Bannockburn, IL). Black and white micrographs were colored with Metamorph 6.3r1 software (Molecular Devices, Sunnyvale, CA) and adjusted with Adobe Photoshop CS version 8.0 (Adobe Systems, Mountain View, CA).

Lamin Inmunoblotting.

Transfected cells, fixed as above, were incubated overnight with LaminDmO antibody (DSHB, ADL67–10). Cy3 anti-rabbit antibody was from Jackson ImmunoResearch (West Grove, PA).

Abbreviations

MHR

major homology region

PNTR

perfect nonterminal repeat.