The phylogeny of nine species of the Drosophila obscura group inferred by the banding homologies of chromosomal regions

The phylogeny of nine species of the group inferred by the banding homologies of chromosomal regions. 11. The phylogenetic relationships among nine species of Drosophila belonging to the ohscura group were investigated by establishing (according to their banding similarities) the homologous chromosome segments of element E (equivalent to chromosome 0 of D. suhohscura). The phylogenetic relationships were based on the existence of segments in different triads of species. which could only be produced by overlapping inversions. This permitted the ordering of the species belonging to each triad. Drosophila ohscura. D. ambiguu and D. tristis were found to be very closely related and thus forming a cluster in which D. ambigua occupies an intermediate position between the other two species. Drosophila ohscura seems to be the species more directly linked to three other separate lineages, that of D. subsilvestris, the two African species (D. microlabis and D. kitumensis), and the suhohscura cluster. The species from this last cluster may be ordered as follows: D. subobscura - D. madeirensis - D. guanche. It is not clear which species of this triad is the direct link to D. obscura. These results completely agree with those produced in an independent study, where element B was considered for the same nine species. Furthermore, the present study clarifies some ambiguities concerning the phylogenetic relationships which remained obscure due to the conservative nature of chromosome B.

In the previous paper of this series (BREHM et al. 1990) we have reported the results of a study concerning the homologies of segments belonging to element B (according to the terminology of MULLER 1940) for nine species of the obscura group. From these data, phylogenetic relationships among these species were inferred. In the present study we pursue further the examination of the phylogenetic relations with data concerning element E for the same nine species. Thus, we are in a position to compare phylogenies derived from independent chromosomes and discuss their agreement.
The ultimate aim of these studies is to produce phylogenetic trees that will depict accurately the sequence of natural events that have occurred. Trees based on several studies of chromosomal data (there are five long chromosomal elements that can be studied independently) will be compared with DNA sequence data gathered from the same species. A consensus tree, eventually produced, could be considered as a good approximation to the "natural" tree. It can then be used to replace invented trees in simulation procedures in order to test the per-formance of different genetic distance estimators and/or algorithms for tree construction when electrophoretic and molecular data are utilized. The adoption of the most efficient distance estimators and algorithms is a valuable information for those using electrophoretic or, possibly, other kinds of data to detect phylogenetic relationships.
Until now, some studies have appeared in which banding similarities are used in order to ascertain homologies between sections or regions of salivary gland chromosomes of Drosophilu (for the D. melanogaster subgroup, see LEMEUNIER andASHBURNER 1976, 1984, for the Hawaiian picture wing DI-osophilu, see CARSON and KANESHIRO 1976, for the repleta group WASSERMAN 1982, and for the virilis group, see THROCKMORTON 1982), but these generally concern species that are evolutionary closely related. It is indeed difficult, but by no ways impossible (STALKER 1972), to deal with Drosophilu species further apart. Regarding the obscura group of species, phylogenies based on overlapping inversions have been constructed for the pseudoobscuru cluster (DOBZHANSKY 1970), the hifasciutu cluster (YA\IAGLC.HI 1973). and for the subgroup ufliiiis (Mii.Lt:R 1977). Finally. K R i v m s and LOIIKAS ( 19x4) and € 3~~3 1~ and K K I V B .~S (1990a). homologized the \pecies from the ~irhohscrri~u cluster. However. a serious work trying to link the different clusters froin the subgroup o h s c~~r i u using such a ical approach ha\ not been conducted until now.
Attempts to establish a phylogeny for the ohscxrrr group of species have been made using diverse techniques. L.AKOL NR I et al. (1976). LOL K A S et al.  (1988) constructed phylogenetic trees using electrophorctic data. Most of the times the patterns depicted by such trees are not in agreement with each other. The first t\\o works appear to be the ones more consistent either with morphological characters or chromosome banding pattern data. In the tree depicted by Cariou et al.. the o/7.wiri.c1 cluster seems to occupy a central position in the group, from which the s i i h~h s~~~~i .~~.
the African. and the p s e w tlooh.\c~irin clusters may have departed. which is supported by the data we present here.
Mitochondria1 DNA also has been used to track phylogenctic relationships. The technique. while having a wide range of applications between populations of the same species (AL' ISE et al. 1979a) or very closely related species (AVISE et al. 1979b). has severe limitations. L. ATOKRE et al. (1986) could not find the European population of D . .sithohsc,ura from which the colonizers of Chile (and the rest of the New World) may have originated and, attempting to do a phylogeny for the ob.wrru group. clustered the species in evident opposition to data obtained froni morphological characters. chronio-soma1 homologies. and electrophoretic markers (LATORRI. et al. 1988).
More recently GODOAKD et al. (1990). using DNA-DNA hybridization. made ;I phylogeny for 5 specie5 from the ~h ,~c . i i / .~~ group. but none in common to the nine studied by us.
The present study a n d the data in BKEHV et al. (1990) prove that using an especially rich photographic material chromosomal homologies can be established lor most part. if not for the entire length. of the chroniosonic cven if the hpecies involved are not \o closcl) related.
M ii t e r i a 1 and met hods nomorphic for the O,, gene arrangement). D . mu-. deiiw7si.s and D . guanche strains originated, respectively, from the Madeira and Canary Islands.
One strain o f D . kitrmensis and another of D. micr-olnhis both originated from Kenya (CARIOU et al. 1988). '4 number of European strains of D. ohscui-o were investigated but the gene arrangement depicted in the photographs is found in a strain from Switzerland. One strain of D. tristis, one from D. unihigrtu, and one from D. sirhsil~~estris. all from Switzerland. were also used.
Details on culture of strains, salivary gland preparations and photography of slides are as the ones, described in B w t i M et al. (1990).
Except for the species belonging to the suhohscwa cluster ( D . sirhohscur-a, D. niudeirensis and, D . g u u i~h e ) the species studied do not hybridize.

Results
From the comparison of a number of homologous segments produced by overlapping inversions between triads of the nine species studied, it became apparent that the arrangement displayed by D. obscrmi occupies an intermediate position. Thus, we decided to use this species as a pivotal extant gene arrangement with which all others are directly or indirectly compared. Of course this species displays an extremely rich inversion polymorphism which is discussed elsewhere (BREHM and KRIMBAS 1990b).
We chose one common gene arrangement, which we present here, in order to proceed to comparisons with the other species studied. This gene arrange-. ment was divided in 13 sections in order to facilitate, identification of segments in the other species. Except for D. ohscura. D. suhohscuru and D .
kitirn7en.si.7 all the other species are considered to be monomorphic for this chromosome. D. ohscuru is; polymorphic for 6 inversions but none of them is useful for determining phylogenetic relationships, with any of the other 8 species studied. D. kitumerrxis is polymorphic for an inversion by which it differs from its closely related species, D . nzicr-olahis. Segment of chromosomes from all species. dis-played in Fig. 3 to S. are compared with segments of D . ohsc.rrru that can be exactly located in the entire sequence of this species, displayed in Fig. 1 and 7 . In ordcr t o avoid doubtful inferences. the ovcrlnpping homologies used to construct the un-rooted phylogenetic tree presented in the Discussion are iiiiitlc of segments for which the banding pat-. tern\ arc clearly identical in ii given triad of species. It seems relatively simple to derive D . amhigua's gene arrangement from that of D. ohscura ( Fig. 1 and 2). Only four inversions are needed for that, two located in the left arm and two overlapping ones in the right arm; these last ones produce the displacement of a segment comprising sections 12A to 13B by inverting it twice subsequently. D. tristis differs only by one inversion from D. ambigua; however, this inversion is an overlapping one on the previously mentioned inversions in the right arm of D. ambigua. This observation permitted us to order the gene arrangements of the three species depicted in the following sequence, according to the principles stated by STURTEVANT and DOBZHANSKY (1936) and DOBZHANSKY (1937): The gene sequences of these two species in relation to the D. obscura's standard gene arrangement are The chromosomal element E of D. guanche compared with that of D. obscura indicates extensive rearrangements. D. guanche (as well as its closely related D. subobscura and D. madeirensis) have an acrocentric chromosome while in all other species examined it is metacentric. The centromere of D. guanche's element E lies at the extreme left part of the chromosome, as depicted in Fig. 3. The homology of segments 3C/4A is not completely convincing, but after consulting a great amount of photographs for this region we are inclined to interpret them as homologous. For other segments we could not find any correspondence between the two species. The remaining segments show fairly good homologies and some could be easily identified with the help of landmarks as, for example, the puff regions of subsections 5BC and 11B.
The situation between the three species, D. suhobscura, D. madeirensis and D. guanche is quite clear because of the presence of overlapping inversions. D. madeirensis has a gene arrangement characterized by inversion O3 (which is fixed in all 13B/lOD--8B/l lCBA/13C 13B/1 ODCB/13C/I 1 ABC/8B-1 OA chromosomes 0, equivalent to the E element), while in D. guanche over the inversion O3 another one is superimposed specific to it (KRIMBAS and LOUKAS 1984). D. subobscura has several arrangements deriving from two basic ones: one bears the combination 3+4 (that is an inversion O4 superimposed (overlapping) on inversion 03), and a second one is Osl, The order of gene arrangements, according to the rule of Sturtevant and Dobzhansky, is where g is an inversion specific to D. guanche.
D. subobscura natural populations contain both O,, and 03+4 gene arrangements but not 03. We may suppose that, originally, 03, the middle gene arrangement of the triad, was sometimes present also in D. suhobscura but has been lost later on, probably according to WALLACE'S rule (1953a, b). This rule predicts that the middle member of a triad tends to be selectively lost in order to preserve coadaptated gene blocks included in the two extreme members of the triad. Taking all this into consideration, the ordering of the three subobscura cluster species may be indicated as follows either: D. guanche -D. madeirensis -D. suhohsrura or, if we take into consideration that D. subohscum once contained also the O3 gene arrangement: From the data gathered it is not possible to know which one of the three species is directly related to D. obscura. This species could be more directly derived either by an 03+g, O3 or 0,, arrangement but not by 03+4. The arrangement 03+4 could not be a link to D. obscura. As shown in Fig. 3, the segment 7AB and part of C is clearly homologized to D . guanche. The inversion 4 (from the arrangement 03+& marked as 212 on the chromosome of this last species, would split the segment 7ABC into two. The breakage points for the other inversions (03, 03+g and OSl) do not lie in any homologized segment Fig. 1 and 2. Homologies of the small and long arms of element E from D. ohscura (OBS), D. unihi~yuu (AMB). and D. tristis (TRI). The Standard gene arrangement of D. ohscur-u is depicted and divided in sections in order to iacilitate the recognition of segments in the other species. In order to get a better visualization of the homologies, the photographs were cut in appropriate places to make them linear. Thus all intervals (nicks. *) between placcs of chroiiiownic photographs do not correspond to genetic material. and thus can not provide a clue for inferring which of them is the link to the standard arrangement of D . ohscum.
In Fig. 3 we indicate the breakage points of these inversions by arrows on the D. guanche chromosome: 111 refers to the inversion specific to D. guanche, 313 to that of the 03, and 212 to that of 04. These breakage points may not coincide with those of MOLTO and MARTINEZ-SEBASTIAN (1986) and M O L T O~~ al. (1987) but have been derived from the comparison of photographs of the gene arrangements mentioned above.
All three species from the suhobscura cluster could have been derived only from D. ohscura (and not from D. arnhigua or D. tristis). The proof for this assertion is segment 4B-5C, which is intact only in D . ohscura and every species of the suhobs u " cluster.
D. guanche could not be derived directly from D. suhsilvestris. In D. guanche, segment 1OABCD is intact, as it is in D. ohscuru, but in D. subsilvestris it is split into two. Besides, in D. guanche, the segment 8BC is not followed by the bands of section 9A. Segment XBC9A is a landmark for the remaining species (it appears unchanged in D. suhsilvestris, the African, and the ohscura cluster).
The gene sequence of D. guanclze chromosome expressed in terms of the standard map of D. obscura is the following: The picture on the small arm of both species is clear, with the exception, perhaps, of subsection 3B. The other arm, on the contrary, presents more difficulties. Homologies of subsections 12C/13A are doubtful and should be simply considered as our interpretation. The other photographs are convincing enough for the remaining segments. Subsection 13B, in spite of being just a small puff, is a well-known landmark. The gene sequence of D .
subsilvestris in relation to that of L?. ohscwu is the following:
D. suhsilvestr-is could not be derived from D . amhiguu or D. tristis but only from D. ohscum. The segment 4B to 5C is carried by D . ohscura and D. suhsilvestris but not by the other two species. It contains one of the breakage points for an inversion in the transition from D . ohscuru to D. amhigua (and consequently to D. tristis). The same segment excludes all hypotheses in which any of the African species is an intermediate step between the two species D. ohscura and D. suhsih~estris.
In Fig. 5 the gene arrangement of another cluster of species is depicted, that of the two African, D. kiturnensis and D. microlahis. The direct comparison is made between D. kiturnensis and D . ohscwa.
Some small segments in this case remain unidentified as far as their correspondence is concerned. D. kiturnensis is polymorphic for an inversion by which it differs from the gene arrangement of D . microlahis, as indicated in Fig. 5.
Segments of the D. ohscura element E were homologized either to D. microlahis or to D . kitumerisis, depending of the quality of photographs. Some of the homologies were arrived at with the help of classical landmarks like the puffs of subsections 5C and 11B. If the small arm is reasonably recognized (with the exception of two small segments and subsection 1 AB, which are not so clear), problems arise with the longer arm. We did not find any correspondence for the entire section 7 and subsections 13BC. However, an important segment, XA-9A, was easily homologized, and in D . suhsilvestris its resemblance with D. kiturnensis is even more clear. This segment includes one breakage point of a big inversion, which makes the difference between D . kiturnensis and D. microlahis, fixing a seriation of this cluster with D. ohscura as

D. microlahis -D. kiturnensis -D. ohscuru
Segment lOCDl 1A can be found in D. kitumcrrsis and in D. ohscuru, but not in D. suhsil\~stris, where it is splitted in two separate segments.
A careful reading of subsections IOAB shows Schematic representation of homologiei between the nine species used in the present study. Segments that are found inverted in two species are indicated by crossed lines or boxes with arrows with inverted directions. When there are more, then one pair of boxes is used between two species. their correspondence is indicated by using dilferent patterns. Sections of each 5egment are numbered according to the photographic map of D . ohscum, and in accordance to the ones depicted in Fig. I to 5 . Numbers 1 to 3 on D. guurrc,br chromosome indicate the specific inversions differentiating the three species of the suhohscrrra cluster (see text). The black circle represents the centromcrc.  (subohsc,ura, ohscura, amhigua, suhsilwsfris). One other clone (12D8) was labeled on the E element in suhohscura and ohsciira, but on the B element in ambigua and suhsil-\sestris. In none of these cases do the authors present the exact localization of the label sites as well as the complete homologous segments of the same chromosome in all species studied. Also FELGER and PINSKER (1987) made a phylogeny for 7 species, all from the ohscura group, using the histone genes' label sites as landmarks, in spite of these genes not being located in homologous chromosomal elements and varying in their number in the species studied. They constructed two phylogenetic trees for the species of the suhohscuru and ohscura clus-. ten, based on the number of inversions needed to pass from one to another species. However, these authors do not present a comparison of chromosome banding homologies to make these numbers reli-able. We do not agree with the data of F E m t : . R and PINSKER (1987) concerning the species from the .suhohsc~rrru cluster (see BREHM and K R l M B A S 1990a:i as well as those concerning the E element, which do not coincide with the data presented here. Felger and Pinsher found a four inversions differcnce between trnrbi,yrrcr and I~. W~I I /~N . lbur between crmhiguos and Iri.\/is. zinc1 eight between oh.sc.ui-u and /risti.s. in opposition to our present results which are four, one and five, respectively. However, it is possible that the differences are due to the use of strains with more complicated gene arrangements.
The present work is the first that presents element E banding sequence homologies for the nine species studied. Based on these homologies and using only qualitative criteria, that is to say, the rule of triads of gene arrangements enunciated by Sturtevant and Dobzhansky, we may be able to arrive at a phylogenetic topology, an unrooted phylogenetic tree, in the following way. It has been noted already that the three species of the subobscura cluster, regarding the inversions Og, O3 and 04, can be arranged in the linear way: guanche Furthermore, based on the inversions of the right arm of D. tristis and the two other species of the obscura cluster we may arrive at the ordering: In case D. kitumensis would lose its polymorphism and get fixed for the alternative gene arrangement to that fixed in D. microlahis, we would also produce an ordering of a third triad.

D. microlahis -D. kiturnensis -D. ohscura
Three more orderings may be added to these, all of them deriving from the data summarized in

guanche,
based on the segments 3C-5B and 4B-5C. To these data one should add that D. subsilvestris. D. obscura, and the two African species share a segment, 8BC9A, which is not found intact in the remaining species studied here.
There is only one way to arrange all these orderings, respecting also this constraint, in an unrooted phylogenetic tree:

MIC
This topological arrangement agrees completely with the one presented in the first paper of this series, dealing with element B, in which three separate lineages (that of the two African species, D. suhsilvestris, and the suhohscura cluster) derived independently from the members of the ohscura cluster. Due to the conservative nature of the B element this tree contained a restricted amount of information. Element E is a much richer source of information, and we can furthermore clarify several features of this topology. Thus the species of the suhohscura cluster are resolved, as well as those of the oh.sc~uru cluster; it is not clear which one ofthe species of the suhohs~ura cluster is directly linked to 13. ohscuru. This last species helds a pivotal place, from which four different lineages are derived. The excellent agreement between two trees. derived from independent sources of data (two independent chromosomes), is a strong indication for the validity of the qualitative method used.