ABSTRACT
We have determined the nucleotide sequence of a 7.5 kb full-size
gypsy
element from
Drosophila subobscura
strain H-271. Comparative analyses were carried out on the sequence and molecular
structure of
gypsy
elements of
D.subobscura
(
gypsyDs
),
D.melanogaster
(
gypsyDm
) and
D.virilis
(
gypsyDv
). The three elements show a structure that maintains a common mechanism of expression. ORF1 and ORF2 show typical motifs of
gag
and
pol
genes respectively in the three
gypsy
elements and could encode functional proteins necessary for intracellular expansion. In the three ORF1 proteins an arginine-rich region was found which could constitute a RNA binding motif. The main
differences among the
gypsy
elements are found in ORF3 (
env
-like gene);
gypsyDm
encodes functional
env
proteins, whereas
gypsyDs
and
gypsyDv
ORF3s lack some motifs essential for functionality of this protein. On the basis of these results, while
gypsyDm
is the first insect retrovirus described,
gypsyDs
and
gypsyDv
could constitute degenerate forms of these retroviruses. In this context, we have
found some evidence that
gypsyDm
could have recently infected some
D.subobscura
strains. Comparative analyses of divergence and phylogenetic relationships of
gypsy
elements indicate that the
gypsy
elements belonging to species of different subgenera (
gypsyDs
and
gypsyDv
) are closer than
gypsy
elements of species belonging to the same subgenus (
gypsyDs
and
gypsyDm
). These data are congruent with horizontal transfer of
gypsy
elements among different
Drosophila
spp.
Gypsy
-like elements constitute a group of retrotransposons which includes
elements from heterogeneous species:
gypsy
,
17.6
,
297
and
412
, from
Drosophila melanogaster
(
1
-
5
),
tom
from
D.ananassae
(
6
),
Ulysses
from
D.virilis
(
7
),
micropia
from
D.hydei
(
8
),
TED
from the lepidopteran
Trichoplusia ni
(
9
,
10
),
del1
and
IFG7
from plants (
11
-
13
),
Ty3
and
Tf1
from yeast (
14
-
16
),
Ctf1
from fungi (
17
) and
SURL
from a marine invertebrate (
18
). According to phylogenetic trees based on reverse transcriptase amino acid
sequences, this group of retroelements clusters with vertebrate retroviruses
and caulimoviruses (
12
,
13
,
19
,
20
). They also have a
pol
gene functional domain order which is similar to that of retroviruses [PR
(protease), RT (reverse transcriptase), RH (RNase H) and EN (endonuclease)] and
some of the
Drosophila
elements (
gypsy
,
17.6
,
tom
.and
297
) exhibit a third ORF of similar size and location to the
env
protein of retroviruses (
2
,
3
,
5
,
6
,
21
). Early functional analyses from ORF3 of
gypsy
-like elements seemed to indicate that these ORF3 encode non-functional
env
proteins and consequently
gypsy
-like sequences lacked the ability to be infectious particles.
Nevertheless, it has recently been established that
gypsy
ORF3 from
D.melanogaster
encodes a fully functional protein which accomplishes the typical functions of
a retroviral envelope protein and, consequently, the authors consider the
gypsy
element as the first insect retrovirus described (
22
,
23
). Supporting this hypothesis, it has been found that
gypsy
sequences can be horizontally transferred by feeding (
24
). We wonder if this remarkable infectious ability of the
gypsy
element of
D.melanogaster
(referred to in this paper as
gypsyDm
) is shared by other members of the
gypsy
family.
gypsy
sequences homologous to that of
D.melanogaster
are widespread among
Drosophila
species, also occurring in the
copia
and
1731
retrotransposons (
25
-
27
). The distribution of
gypsy
elements is consistent with the hypothesis that these sequences are of ancient
origin. They were probably present in a common ancestor before early
Drosophila
radiation and subsequently transmitted vertically. However, the complex
phylogenetic relationships of
gypsy
-like sequences among
Drosophila
and distant, unrelated species suggests that horizontal transfer of
gypsy
sequences between major taxonomic groups has contributed to the generation of
the present
gypsy
family. As cited above, recent data strongly support consideration of the
gypsyDm
element as an infectious particle (
22
-
24
). If this ability is a general feature of
gypsy
elements, it is obvious that the spreading of
gypsy
sequences throughout populations and/or species could be faster than that expected of
true retrotransposons. From this point of view it would seem interesting to
study the evolutionary history of
gypsy
sequences amongst
Drosophila
species. To date, only the
gypsy
element from
D.virilis
(in this paper
gypsyDv
) has been sequenced, which showed a similar structure to
gypsyDm
(
28
). In a previous paper we presented preliminary data on the sequence and
molecular structure of
gypsy
elements from
D.subobscura
(
gypsyDs
), a species belonging to the
obscura
group. We found that the
gypsyDs
sequence of
D.subobscura
is closer to that of
gypsyDv
than that of
gypsyDm
, which is not consistent with the phylogenetic relationships between these
species (
29
). In the present paper we report a comparative analysis of the sequence and
molecular structures of the
gypsyDs
,
gypsyDm
and
gypsyDv
elements. From this analysis it could be deduced that while
gypsyDm
shows the potential ability to be an infectious particle,
gypsyDs
and
gypsyDv
have lost this ability. The
gypsyDs
and
gypsyDv
elements are probably degenerate forms of ancestral infective particles. From
molecular characterization of PCR products we found
gypsy
elements identical to those of
gypsyDm
in different
D.subobscura
natural populations, co-existing with
gypsyDs
elements. All these data indicate that
D.subobscura
natural populations could have been invaded by infectious
gypsyDm
elements.
Gypsy
-homologous sequences were isolated from a genomic DNA library, kindly
provided by R. González-Duarte and G. Marfany, constructed by partial
Mbo
I digestion and ligation of
D.subobscura
strain H271 DNA into the [lambda] EMBL4 phage vector. A 7.0 kb
Xho
I-
Xho
I fragment of
gypsyDm
was used as a probe and we followed the procedures of Kaiser and Murray and
Benton and Davis (
30
,
31
). Twenty eight positive clones were identified, of which at least three
contained full-length elements. From these, clone DsE1 was selected for further analyses.
Cloning of restriction fragments of this clone was carried out in pUC18/19 by
standard procedures (
32
,
33
).
The nucleotide sequence of clone DsE1 was determined for both strands by the
dideoxynucleotide chain termination method (
34
,
35
) using [
35
S]dATP and the T7 DNA polymerase sequencing kit (Pharmacia). Gaps in the
sequence were filled in with nested deletions, using synthetic oligonucleotides
as primers for the sequencing reactions.
In situ
hybridization was performed on larval salivary gland chromo- somes of
D.subobscura
strain H271 following the procedures described in Terol
et al.
(
36
). As a probe we used a
3
H-labelled 3 kb
Pst
I-
Pst
I fragment of genomic DNA from the DsE1 clone.
mRNA from adult flies of
D.subobscura
was extracted using the guanidinium thiocyanate method (
37
). The poly(A)
+
fraction was purified by affinity chromatography through oligo(dT)-cellulose spun columns (Pharmacia), according to the manufacturer's protocol.
Electrophoresis of a 3 [mu]g RNA sample and Northern blots were carried out as described in Sambrook
et al.
(
33
) for nylon filters. Hybridization was performed at 50oC in 7% SDS, 5* SSC, 50 mM phosphate buffer, 50% formamide, 2% blocking reagent, 0.1% laurylsarcosine and 50 [mu]g/ml yeast tRNA. The 6.9 kb
Xho
I-
Xho
I fragment of
gypsyDs
was used as the probe. Washing conditions were 1* SSC, 0.1% SDS at 50oC. Detection was performed using the DIG chemiluminescent method,
following the manufacturer's protocol (Boehringer-Mannheim).
Multiple alignments of
gypsy
sequences from different species were performed using the CLUSTAL program (
38
). Sequence analyses and phylogenetic constructions were determined using the
PHYLIP 3.4 package programs.
Southern blots of different
D.subobscura
strains (H-271, Finland; SN, Sweden; TÜ, Germany; DIE, Switzerland; CJ, USA; PC, Canada; MA, Madeira; RA,
Canary Islands; MAR, Morocco) were performed using standard procedures. These strains came from natural
populations maintained for some years in the laboratory. A sample of 5 [mu]g genomic DNA of different
D.subobscura
strains quantified by spectrofluorimetry were digested with
Xho
I,
Eco
RI or
Kpn
I. Hybridizations were performed at 65oC with 0.02% SDS, 5* SSC, 0.5% blocking reagent, 0.1% lauroylsarcosine using two probes, the
Xho
I-
Xho
I fragments of
gypsyDs
and
gypsyDm
. Washes were at 65oC in 0.5* SSC, 0.1% SDS when the
gypsyDs
probe was used and at 45oC in 1* SSC, 0.1%SDS when the
gypsyDm
probe was used. DIG colour detection was performed using Boehringer Mannheim
standard procedures.
PCR analyses of the
gypsyDm
sequences were carried out on the genomic DNA of the TÜ and MAR strains. Specific oligonucleotides were designed following the
canonical
gypsyDm
sequence (
3
) and selected from the most variable regions in comparison with the
gypsyDs
sequence in order to selectively amplify
gypsyDm
sequences in the
D.subobscura
genome. The 5' position and the sequence of the primers used in the reactions were as
follows:
The genomic library of
D.subobscura
strain H271 was screened under moderate stringency conditions using the 7.0 kb
Xho
I-
Xho
I fragment of the
gypsyDm
element as probe. Of the 28 positive clones obtained, 16 were studied by
restriction analysis, which revealed that at least three of them contained
putative full-length
gypsy
elements with a very similar restriction pattern. Clone DsE1 was selected arbitrarily for
further study.
The nucleotide sequence of the
gypsy
element present in this clone was fully determined in both strands (EMBL
accession no. X72390). The chromosomal location of the
gypsyDs
present in clone DsE1 was determined by
in situ
hybridization of larval polytene chromosomes. The signal was found at the
border of the 35AB puff in the E chromosome using as probes both an adjacent
genomic DNA fragment (Fig.
1
A) and the 6.9 kb
Xho
I-
Xho
I fragment of
gypsyDs
(Fig.
1
B). The target sequence found at the insertion site is
The enhancer region, located between the 5' LTR and ORF1, shows a heterogeneous sequence but a similar structure in
the three elements, in which a conserved repeat of 12 bp is interspersed in AT-rich regions with variable number and position (Fig.
2
B). In
D.melanogaster
these 12 bp repeats constitute the binding site for a protein with zinc finger
domains that acts as a transcription activator (
39
). The number of repeats varies, being 13 in
gypsyDm
, six in
gypsyDv
and nine in
gypsyDs
. The existence of these repeats in the other
gypsy
elements could indicate that they have the same function in these species. A
poly(A) tract followed by an imperfect palindrome can be found 5' of this regulatory region in
gypsyDm
and
gypsyDs
.
This sequence in
gypsyDm
binds a repressor protein, probably encoded by
su(f)
(
40
).
The products of ORF1 and ORF2 are necessary for both intracellular and
extracellular retroelement movements. Assuming that the three
gypsy
elements are transpositionally active (
3
,
28
,
29
), the essential motifs of these proteins must be conserved. This is exactly
what happens with the ORF2 products, in which the PR, RT, RH and EN domains can
be observed (Fig.
3
B) These domains are homologous to those present in the
pol
gene of retroviruses, being in the same order.
The 5' region of the ORF1 sequences is the most variable, due to deletions or
changes in
gypsyDs
and
gypsyDv
that do not produce frame shifting. Proteins with
gag
functions have RNA binding domains in order to recognize RNA and specifically
direct it to the capsid. Although no RNA binding domains could be found in the
putative
gypsyDm
and
gypsyDv
ORF1 proteins (
20
), a more detailed analysis looking for different types of motifs (
41
) revealed that an arginine-rich region near the C-terminus can act as an ARM RNA binding motif in the three
gypsy
elements (Fig.
3
A). Furthermore, the proline content (found at conserved positions) of ORF1 is
high in
gypsy
elements, being 5.1% in
gypsyDm
and
gypsyDv
and 4.6% in
gypsyDs
.
ORF3 corresponds in position and size to the
env
gene of retroviruses and, as has been found in
gypsyDm
, it acts as a true envelope protein (
22
,
23
). As in retroviruses, it has been established that
gypsyDm
Env protein is translated from a 2.1 kb mRNA produced by differential splicing.
The splicing donor and acceptor sites have been thoroughly determined in
gypsyDm
(
22
). Looking for splicing sites in
gypsyDs
and
gypsyDv
shows that a donor site located in the 5' leader region is identical in the three
gypsy
elements. However, the acceptor site located between ORF2 and ORF3 shows some
variation. In
gypsyDs
we have found an acceptor site consensus sequence, but it was not so evident in
gypsyDv
. Two potential acceptor sites in these
gypsy
elements are shown in Figure
3
C. As in
gypsyDm
, in
gypsyDs
the splicing generates an ATG, underlined in Figure
3
C, but translation cannot occur from this ATG because a number of stop codons
appear downstream. A second ATG can generate a 5' truncated homologous envelope protein. In
gypsyDv
, independent of the splicing acceptor site, a TAA stop codon is present
immediately adjacent to the ATG, which also generates a 5' truncated protein in this element.
gypsyDm
ORF3 shows the structural motifs typical of envelope proteins, such as a signal
peptide at the N-terminus, a dibasic cleavage site in the middle of the sequence and a
transmembrane domain located near the C-terminus. Furthermore, a number of cysteines and three putative
N
-linked glycosylation sites involved in post-translational modification have been found in
gypsyDm
(
22
). Surprisingly, multiple alignments of the putative products of
gypsyDm
,
gypsyDs
and
gypsyDv
ORF3 show that although the sequence is well conserved among them,
gypsyDs
and
gypsyDv
lack some of the envelope motifs, while others are conserved (Fig.
3
D). Firstly,
gypsyDs
and
gypsyDv
lack the signal peptide sequence, because the start codon is located downstream in both cases (Fig.
3
D). Furthermore, the sequence that corresponds to the start site in
gypsyDm
is variable among
gypsy
elements,
gypsyDs
and
gypsyDv
having lost the start codon in such a way that ORF3 in these elements starts
from a downstream ATG that is not in phase. This could result in totally
different proteins, but the reading frame is recovered in
gypsyDs
and
gypsyDv
by independent and single insertion events (indicated in Fig.
3
D by empty circles) that have occurred at different locations. Second, the
dibasic cleavage site,
N
-linked glycosylation sites and cysteines are highly conserved among
gypsy
elements. Finally,
gypsyDs
lacks the transmembrane domain, because a single deletion in the upstream
region has produced a frame-shift in this element that provokes early protein termination. In summary,
gypsyDs
and
gypsyDv
lack some essential motifs necessary to produce functional envelope proteins.
Table
1
presents the per cent identity at the nucleic and amino acid levels obtained
from general comparisons between
gypsy
sequences corresponding to the three ORFs. The major differences in the per
cent identity between the nucleic and amino acid sequences are found in ORF2,
which encodes a putative protein with PR, RT, RH and EN domains. This result is
in accord with the important function of this protein in the replicative cycle
of retrotransposons. Nevertheless, the most important conclusion from this
analysis is that similarity is greater between
gypsy
elements of species belonging to different subgenera (
D.virilis
and
D.subobscura
) than those belonging to the same subgenus (
D.melanogaster
and
D.subobscura
). These data are not in accord with a strictly vertical transmission of
gypsy
sequences in the genus
Drosophila
.
A more detailed analysis was carried out in order to determine whether
gypsyDs
is functionally active (Table
2
). These studies show that in all cases the number of silent substitutions per
effectively synonymous site (
D
S
) is much higher than the number of replacement substitutions (
D
R
), indicating that
gypsyDs
has been subjected to functional constraints. Small differences can be observed
among the ORFs. For instance,
D
T
(number of total substitutions) and
D
R
are smaller when ORF2s are compared, while
D
S
is greater in comparison with those of other ORFs. This implies major
functional constraints on ORF2. In contrast, ORF1 shows more divergence.
Another significant finding obtained from this study is that the divergence
between
gypsyDs
and
gypsyDv
is less than that expected in all ORFs, suggesting that genetic distance does
not correlate with phylogenetic distance between the species.
Table 1
To obtain data about the relationships between
gypsy
sequences we have constructed trees using different phylogenetic methods
(neighbour-joining; Fitch and Margoliash, assuming variable mutation rate; maximum
parsimony) using the amino acid sequences corresponding to ORF2 in all cases.
As an outgroup we have used the RT-homologous ORF of
412
, a
D.melanogaster
retrotransposon that belongs to the
gypsy
group. The topology in the trees obtained by different methods agree in those
where
gypsyDs
is grouped with
gypsyDv
instead of
gypsyDm
. As an example, Figure
4
A shows the results obtained with the neighbour-joining method. In order to obtain a conventional phylogeny we have
considered a 250 nt sequence of exon 8 of the
Antennapedia
(
Ant
) gene from the same species of
Drosophila
. As an outgroup we used the homologous sequence of the aquatic crustacean
Artemia franciscana
. Figure
4
B shows that
AntDs
and
AntDm
are grouped together rather than with
AntDv
, which is consistent with the conventional phylogeny and different from the
topology obtained for
gypsy
sequences.
Table 2
Comparative analysis of the
gypsy
nucleotide sequences corresponding to the three ORFs
Previous data about the distribution of sequences homologous to the
gypsyDm
element among 25
D.subobscura
strains from natural populations over its distribution area indicate that five
of the strains analysed showed striking hybridization patterns with strong
signal intensities and different banding patterns (
29
). One of the hypotheses to explain this hybridization pattern is that
gypsy
elements in these strains are closely related to those of
gypsyDm
. In order to test this hypothesis we performed Southern blot analysis using the
Xho
I-
Xho
I fragment of
gypsyDs
and
gypsyDm
sequences as probes.
The most simple explanation to account for the higher level of sequence
similarity between
gypsyDv
and
gypsyDs
than between
gypsyDm
and
gypsyDs
and
gypsyDm
and
gypsyDv
is that horizontal transmission of
gypsy
elements between
D.subobscura
and
D.virilis
genomes occurred 17-24 million years ago. A more detailed study of the distribution of
gypsy
elements in the
obscura
and
virilis
subgroups will have to be done in order to determine the donor and acceptor
species. The phylogenetic trees, based on
gypsy
sequences compared with that of the
Antennapedia
gene, support a close phylogenetic relationship between
gypsyDv
and
gypsyDs
, while
Antennapedia
sequences cluster according to the phylogenetic relationships proposed for the
three species (
44
,
45
). Furthermore,
gypsyDv
and
gypsyDs
share some LTR and ORF structural characteristics (i.e. variability in the 5' region of ORF1 and lack of the potential signal peptide in ORF3) which
indicate that these sequences are evolutionarily closer than to that of
D.melanogaster
.
The occurrence of horizontal transfer of eukaryotic transposable elements among
phylogenetically related or distant species could be an extended phenomenon (
46
-
49
). From the inconsistencies found in the phylogenetic analysis horizontal gene
transfer may be suspected, but various additional supporting lines of evidence
are usually necessary to firmly establish this hypothesis, as occurs in the
evolution of the
P
transposable element (
47
,
49
-
52
). In the case of
gypsy
elements horizontal transfer was earlier proposed to explain the close
relationship between
gypsy
elements from
D.melanogaster
and
Ty3
from yeast (
53
). A more extended distribution of
gypsy
group elements among different taxa strongly supports lateral transfer events
in their evolutionary history.
gypsy
group elements have been detected in insects (
gypsy
,
412
,
17.6
,
297
,
micropia
and
Ulysses
from
Drosophila
and
mag
and
TED
from the lepidoptera
Bombyx
mori
and
Trichoplusia ni
respectively), yeast (
TY3
from
Saccharomyzes cerevisiae
and
Tf1
from
Schizosaccharomyces pombe
), fungi (
Cft1
from
Cladosporium
fulvum
), marine invertebrate (
SURL
from the sea urchin
Tripneustes
gratilla
) and plants (
del
from
Lilium
henryi
and
IFG7
from
Pinus
spp.) (
7
-
9
,
15
-
18
,
54
,
55
). Phylogenetic trees based on RT and RH amino acid sequence domains show clear
inconsistencies, i.e.
Ulysses
and
micropia
are closer to the plant
SURL
family than other
Drosophila
gypsy
elements.
Assuming that some horizontal transfer events could have occurred in
gypsy
group evolution, it can be implied for
gypsy
elements between
D.virilis
and
D.subobscura
on the basis of structural analogies and sequence similarity. However, on
considering, as cited above, that
gypsyDv
and
gypsyDs
seem to be old non-infective parasites of the genome of both species, it could alternatively
be proposed that
gypsyDv
and
gypsyDs
derive from a common ancestor present in some ancestral species before
Sophophora
radiation and consequently derived sequences are found among
Sophophora
and
Drosophila
subgenera. The potentially infective
gypsyDm
sequences could be considered recent parasites of the
D.melanogaster
genome. Nevertheless, this hypothesis does not explain the 90% similarity
between
gypsyDs
and
gypsyDv
when genomic sequences from these species are more divergent.
Horizontal transfer requires special conditions to be successful. Among these
has been proposed the existence of a transmitting vector (
47
). If we admit that
gypsyDs
and
gypsyDv
were probably retroviruses in the past, the horizontal transfer event proposed
to explain the similarity existing between these elements could constitute a
simple infection event.
Assuming that
gypsyDv
and
gypsyDs
represent old sequences probably dispersed in the genome of
Drosophila
spp. and that
gypsyDm
represents a new potentially infective element, it is possible that both old
and new sequences co-exist in the genome of a given species. In support of this proposition
highly diverged
gypsyDm
elements have been found in the
f
1
su(f)
strain (
56
). Two subfamilies of
gypsyDm
elements belonging to SS and MS strains of
D.melanogaster
have been described (
57
). The MS strain is characterized by a high frequency of spontaneous mutation,
which has been correlated with
gypsy
transposition (
58
,
59
). Moreover, co-existence of different subtypes correlated with their activity and/or time
of persistence in a given genome has been described in other transposable
element families, i.e.
mariner
,
I
and
P
(
60
-
62
).
In the
D.subobscura
genome the
gypsy
element analysed was obtained from a genomic library constructed with DNA from
strain H271, which is a laboratory strain homozygous for inversions in all of
their five chromosomes (
63
). A survey for the presence of
gypsy
elements in natural populations representative of the dispersion area (North
Africa, Europe and North and South America) had already been carried out (
29
) by Southern analysis. In five of the populations analysed a differential
hybridization pattern was obtained. In the present paper we have sequenced a
PCR-amplified fragment from
D.subobscura
strain TÜ that was nearly identical to the canonical
gypsyDm
element. From these data we can infer that similar elements could be present in
the other four strains. It can be deduced that in these strains sequences co-exist which are homologous to
gypsyDs
(old
gypsy
elements) and to
gypsyDm
(new
gypsy
elements). In this context, it is possible that
gypsyDm
is invading
D.subobscura
populations by means of its infective properties. The five strains cited came
from natural populations, but have been maintained in the laboratory for some
years. Laboratory conditions could possibly increase the likelihood of
infection by
gypsy
viral particles originating from
D.melanogaster
strains.
This work was supported by grants from the Spanish CICYT (no. PB90-426) and FPI programmes. The nucleotide sequence was obtained at `Servicio
de Secuenciación' and the sequence analyses were done at `Servicio de Bioinformática', both at the University of Valencia. We would like to
thank R. González-Duarte and G. Marfany for the
D.subobscura
genomic library and N. Paricio for kindly providing the mRNA and for her help
with the Northern blots.
Per cent identity
ORF1
ORF2
ORF3
DNA
Protein
DNA
Protein
DNA
Protein
gypsyDm
versus
gypsyDs
71.46
74.29
75.97
82.97
74.78
75.06
gypsyDm
versus
gypsyDv
68.90
71.81
75.18
82.57
74.38
74.95
gypsyDs
versus
gypsyDv
90.52
90.35
92.44
94.16
90.47
86.75
REFERENCES
Return