The mitochondrial reading frame RF3 is a functional gene in Saccharomyces uvarum.

The yeast mitochondrial genome contains three reading frames, RF1, RF2, and RF3, which are related to the group I maturases, though they are not intronic sequences. In the Saccharomyces cerevisiae strain D273-10B/A, these reading frames are interrupted by G + C-rich sequences (GC clusters) which break the frames. In the present work we described a Saccharomyces uvarum strain which possesses a RF3 continuous sequence devoid of GC clusters. Moreover, our results strongly suggest that in the same strain RF2 and RF1 are also continuous sequences. As all three RFs belong to transcription units which are highly expressed, it is most reasonable to assume that RF1, RF2, and RF3 are functional genes. Furthermore, we have discovered a rule which seems to explain the transposition of GC clusters, considered as mobile elements, in the mitochondrial genome.

In Saccharomyces cerevisiae, the mitochondrial genes encoding 21 S ribosomal RNA, cytochrome b, and subunit 1 of cytochrome c oxidase are split by intronic sequences. In the largest mitochondrial genomes 13 introns have been identified, 10 of them containing an open reading frame encoding the maturases (Jacq et al., 1982). These 13 intronic sequences can be arranged into two distinct groups (Michel et al., 1982;Michel and Dujon, 1983). Members of the same group share distinctive nucleotide stretches and may be folded up into similar secondary structures (Michel and Dujon, 1983;Davies et al., 1982;Waring and Davies, 1984). Two short conserved oligopeptide sequences which are characteristic of the open reading frame of group I introns were called P1-P2 by Michel et al. (1982), decapeptides in Waring et al. (1982), and LAGLI-DADG by Hensgens et al. (1983). Genetic studies and biochemical analyses have disclosed or suggested four different functions for the intron-encoded proteins (Kotylak et al., 1985). ( a ) Some maturases are essential for the correct splicing of mitochondrial intervening sequences (Banroques et al., 1986). ( b ) The maturases of group I1 introns show a significant homology with reverse transcriptase (Michel and Lang, 1985) and are supposed to play a role in the deletion of intervening sequences from the genes (Gargouri et al., 1983). (c) The intron-encoded protein in the 21 S rRNA intron is necessary * This investigation was supported by the Centre National de la Recherche Scientifique (ATP 06931, Organisation et expression du genome) and the Ministhre de la Recherche et de la Technologie (MRT 85T 0698). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordame with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequencefs) reported in this paper has been submitted to the GenBankTM/EMBL Data Bunk with accession number(s) 502738.
for the duplicative transposition of this intron (Colleaux et al., 1986). ( d ) A maturase was shown recently to be involved in the induction of genetic recombination between homologous exons (Kotylak et al., 1985).
Besides the intron-encoded proteins, three open reading frames related to the group I maturases exist which are not intron sequences (Coruzzi et al., 1981;Michel, 1984;Seraphin et al., 1985). They were named RF1, RF2, and RF3. They are part of three multigenic transcription units, and each of them is, respectively, located downstream of a gene which is not mosaic: oxil (subunit 2 of the cytochrome oxidase); oxi2 (subunit 3 of the cytochrome oxidase); and oli2 (subunit 6 of the mitochondrial ATPase) (Coruzzi et al., 1981;Thalenfeld et al., 1983;Michel, 1984;Simon and Faye, 1984b).
The organization of the RF2 and RF3 sequences, in the strains so far studied, is very peculiar (Fig. 1). Each of them is composed of four fairly large open reading frames which overlap within G+C-rich sequences (GC clusters (de Zamaroczy and Bernardi, 1986)). Furthermore, a shift of +1 or -1 base was found between each pair of consecutive reading frames. A unique GC cluster is present in the RF1 sequence that could also introduce a -1 frameshift since some uncertainty resides in the sequence published by Coruzzi et al. (1981). The odd structure of the RF2 and RF3 sequences suggests that they could be silent genes inactivated by the insertion of GC clusters and that the loss of their functions could be compensated for by other coding sequences located elsewhere in the mitochondrial genome. However, if they are functional genes we have to think of some mechanisms (acting on transcription or translation) able to introduce changes in the reading frames at the level of the GC clusters so as to correct the frameshifts. In the present work, we describe a yeast strain which possesses a continuous RF3 sequence, i.e. it is not interrupted by GC clusters. Moreover, our results strongly suggest that in the same strain RF2 and RF1 are also continuous sequences devoid of GC clusters. As all three RFs belong to transcription units which are highly expressed (Simon and Faye, 198413;Thalenfeld et al., 1983;Coruzzi et al., 1981) it seems most reasonable to assume that RF1, RF2, and RF3 are functional genes. Furthermore, we have discovered a rule which appears to explain the transposition of GC clusters in the mitochondrial genome.
The rapid yeast DNA procedure used was adapted from the method described by Davis et al. (1980).
Cloning and Sequencing-Preparation of plasmid DNA, restriction analysis, and cloning in Escherichia coli were carried out as described arrows represent open reading frames. The position of GC clusters is shown in each RF. +1 (or -1) indicates a +1 (or -1) frameshift introduced in the reading frames by the GC clusters. If some mechanism is able to correct these frameshifts, then translation products (corresponding to the junction of the reading frames represented by arrows) of about 500 amino acid residues should be formed, either for RF1, RF2, or RF3. PI and P2 are the two dodecapeptides generally encountered in the maturases of group I (Michel, 1984). The question marks mean that it is not certain whether the RF1 GC cluster introduces a frameshift because of some uncertainty in the published sequence (Coruzzi et al., 1981;Michel, 1984).
in Maniatis et al. (1982). Plasmid p602-IV is derived from pBR328, bearing the EcoRI restriction fragment RR8 of the mitochondrial genome of strain D273-10B/A at its EcoRI site (Seraphin et al., 1985). Plasmid pN1 contains the RR8 fragment of strain NCYC74. Plasmid pN-RF2 harbors the HindIII-EcoRI fragment of strain NCYC74 which includes the major part of the RF2 sequence (Thalenfeld and Tzagoloff, 1980;Michel, 1984). The bacteriophage M13K07 (Pharmacia P-L Biochemicals) was used as a helper to produce singlestranded DNA from pTZ18R-derived recombinants.
For DNA sequencing, the fragment RR8 of strain NCYC74 was cloned in phage M13tg130 (Amersham Corp.) in both orientations; two recombinants, 311 and 312, containing the fragment RR8 with opposite polarity were isolated (Messing, 1983). To generate a series of deletions in this fragment, 311 and 312 were partially digestedwith the restriction endonuclease DraI, the enzyme was then inactivated, and T4 DNA ligase was added. After transfection of E. coli a set of subclones with overlapping sizes was selected by analyzing on agarose gel slabs either single-stranded DNA preparations in the presence of 0.2% SDS or S1 nuclease-resistant hybrids between the singlestranded DNA preparation and the cloned insert in the opposite orientation (Poncz et al., 1982). The XbaI-EcoRI fragment was deleted in the recombinant pN-RF2 to determine the HindIII-XbaI DNA sequence of RF2. The Sanger dideoxy sequencing technique was used (Sanger et al., 1977).
Southern and Northern Transfers-Southern transfers were carried out as described by Maniatis et al. (1982). Mitochondrial RNA of yeast cells grown in 1% yeast extract, 1% Bactopeptone, 2% raffinose, 0.1% glucose was extracted as described (Faye and Simon, 1983). RNA transfers to nitrocellulose filter and RNA-DNA hybridization were performed according to the Southern method as modified by Thomas (1980). SI Mapping"S1 nuclease protection experiments were carried out according to Sharp et al. (1980).

RESULTS
Southern Blots-The mitochondrial DNA of 20 laboratory yeast strains from different sources ( Table I) was analyzed in order to know whether RF1, RF2, or RF3 sequences devoid of GC clusters could be found. The total DNA from each strain, extracted by a rapid procedure, was digested with appropriate restriction endonucleases. The DNA fragments obtained were then separated on an agarose gel, transferred onto a nitrocellulose filter using the method described by Southern (1975), and then hybridized with probes labeled by nick translation. The hybridization pattern from each strain was then compared to that of strain D273-10B/A which was used as a reference pattern since the nucleotide sequence of the RF1, RF2, and RF3 regions was mainly determined in this strain (Coruzzi et al., 1981;Michel, 1984;Seraphin et al., 1985). A.
The RF2 region was analyzed by digesting the DNA preparations with EcoRI, PuuII, and MboI (Fig. 3). The fragments obtained were probed with the plasmid pN-RF2 (cf. "Materials and Methods"). Only strain MH41-7B (lane 20) displays the pattern A which is characteristic of strain D273-10B/A (lane I ) . In the strains showing pattern B, the band corresponding t o t h e 468-bp-long fragment of strain D273-10B/A is slightly larger indicating that a second GC cluster might be present in this fragment. S. uuarum (lane 5, pattern C) is the sole strain in which bands corresponding to the 698-and 468-bplong fragments show a small reduction in size suggesting the absence of GC clusters in its RF2 open reading frame. The RF2 sequence is absent in half of the strains studied (pattern E). In fact, it is known that the RF2 segment is missing in some strains of S. cereuisiae (Thalenfeld et al., 1983;Fox and Reif, 1983).
Concerning the RF3 region, the DNA preparations were

~-~
The abbreviation used is: bp, base pair(s).  digested either with EcoRI alone or with both EcoRI and HpaII. The probe used was the plasmid pN1. Eleven strains do not possess the RF3 sequence. We have previously sequenced the mitochondrial DNA of one of them, strain JM6, in the region immediately downstream of the oli2 gene and in that overlapping the deleted RF3 sequence (Seraphin et al., 1985). Seven strains share the pattern of D273-10B/A (Fig.   4, lane I ) . There is no HpaII site in the RR8 EcoRI fragment of S. uuarum (lane 5 ) , which is about 100 bp shorter than the corresponding RR8 fragment of D273-10B/A. Thus, it seems that the RF3 open reading frame of S. uuarum does not contain any GC clusters. The mitochondrial genome of strain 4870-6B (lane 8) possesses a RR8 EcoRI fragment which is about 1.8 kilobases long (data not shown), whereas the pN1 probe reveals only three EcoRI-HpaII fragments. Table I1 shows that the strain polymorphism, with respect to the presence or absence of the RF2 and RF3 sequences in the mitochondrial genome and the presence or absence of GC clusters in the three regions studied is rather high, since 11 different arrangements are observed. At least three among the seven strains which share the same arrangement (arrangement VII) were obtained from the Yeast Genetic Stock Center (Berkeley). It is known that many mutant and sepegant strains from this collection have been derived from the same progenitor strain (Mortimer and Johnston, 1986). This could explain why arrangement VI1 was observed more frequently on our Southern blots.

Genes in Yeast
S. uuarum appears to be the bona fide candidate strain we were looking for; it seems to have no GC clusters in its RF1, RF2, and RF3 reading frames.
However, this strain does contain numerous GC clusters elsewhere in its mitochondrial genome; whereas 200 GC clusters are scattered throughout the mitochondrial genome of S. cereuisiae (de Zamaroczy and Bernardi, 1986), their number is only slightly smaller in S. uuarum .
DNA Sequencing-To demonstrate definitively the absence of GC clusters in the RF3 reading frame of S. uuarum and to know how its sequence is organized, the RR8 restriction fragment from this strain was cloned in the vector M13tg130 and was sequenced according to the strategy described under "Materials and Methods." The 1673-bp-long RR8 fragment does not contain any of the three G+C-rich clusters which punctuate the RR8 sequence of D273-10B/A. The comparison of the two RR8 sequences shows that in strain D273-10B/A the three GC clusters have been inserted at three AG sites.
The RF3 open reading frame of strain NCYC74 is continuous and has the potential to code for a 476-amino acid-long protein, with a calculated molecular weight of 58,870 (Fig. 5 ) . Except for the three GC clusters, only one nucleotide transition was observed between strains D273-10B/A and NCYC74, Yeast Mitochondria  Total DNA of 20 strains was digested with the HpaII and PuuII restriction enzymes, electrophoresed on 1.0% agarose gel, and then transferred to a nitrocellulose filter. The 32P-labeled probe used was the mitochondrial DNA of the rho-strain DS302 (Coruzzi et al., 1981). The number marking each lane refers to the name of the strain as it is written in Table I, whereas the letters refer to the restriction maps depicted in the louler part of the figure. The sizes of the PuuII-HpaII restriction fragments from D273-10B/A (lane I ) , as known from the sequence data, are indicated in the margin. Concerning the maps B-F, only the restriction fragments differing in size from those of D273-10BIA are indicated. The small black boxes and the small open boxes represent GC clusters whose existence is certain or assumed, respectively. H , HpaII restriction sites; P, PuuII restriction sites; OXZI, gene of the subunit I1 of cytochrome oxidase; RFI, RF1 gene. namely a G for an A changing a glycine codon for a glutamic codon, respectively. This confirms that S. cerevisiae and S.
uvarum are phylogenetically related and should belong to a single species. In fact, the mitochondria of S. uvarum can be transferred by cytoduction (Lancashire and Mattoon, 1979) to S. cerevisiae and have no effect on growth and respiration (data not shown).
Part of the RF2 sequence of strain NCYC74 located between sites Hind111 and XbaI as well as the sequence of a segment overlapping the third GC cluster was determined (Fig. 6). Both sequences confirm that the three GC clusters of the D273-10B/A RF2 sequence are absent in strain NCYC74 and indicate that this reading frame is continuous in the latter strain, at least in the DNA segments studied. The target site of the three GC clusters of RF2 in strain D273-10B/A is also an AG sequence (taking into account that the last two GC clusters are oriented in a direction opposite to that of the first one (Michel, 1984)). Except for the GC clusters, the two RF2 fragments (the total length of which is The DNA preparations were digested with EcoRI, PuuII, and MboI. The probe used was the HindIII-EcoRI fragment from strain NCYC74 which contains the major part of the RF2 sequence (plasmid pN-RF2). Except for D273-10B/A the presence of the GC cluster in the 1162-bp-long EcoRI-MboI restriction fragment is not proved. The RF2 sequence is missing in strains 7, 8, 9, 11, 12, 14, 16, 17, 18, 19. E, EcoRI; P, PuuII; M , MboI. 600 bp) we have sequenced in NCYC74 are identical to the corresponding sequence of D273-10B/A.
Northern Blots-We have previously shown that the RF3 open reading frame belongs to a long transcription unit which contains cox1 (oxi3), aapl, oli2, and RF3 (Simon and Faye, 1984b). The primary RNA produced from this transcription unit, in strain D273-10B/A, undergoes several processing events; in particular two cleavages occur at the sites AATAA-TATTCTT which liberate a polycistronic RNA bearing aapl, oli2, and RF3 (Simon and Faye, 1984b). This RNA is further processed at its 5'-end and gives rise to a new 566-b shorter RNA. Both RNA species are abundant. Because the RF3 open reading frame was continuous in strain NCYC74, we were curious to know how the RF3 region is transcribed. The mitochondrial RNAs of two isonuclear strains containing either the mitochondrial genome of strain D273-10B/A or that of strain NCYC74 were each purified, electrophoresed on a 1.1% agarose gel, transferred to a nitrocellulose filter, and then probed with the plasmid pN1. The autoradiogram is shown in Fig. 7. The estimated sizes of the probed RNAs are 4200 and 3600 bases in strain D273-10B/A and 4100 and 3500 bases in strain NCYC74. The small size differences of these major RNAs may be accounted for by the presence of GC clusters in the RF3 sequence of strain D273-10B/A. Minor transcripts, both larger and shorter, are also visible on the autoradiogram.
SI Mapping-The fact that apart from the inserted GC clusters, the RF3 coding sequences are identical in strains D273-10B/A and NCYC74 is intriguing; if RF3 is functional in strain D273-10B/A we have to suppose that some mechanism(s) eliminates the phase discontinuities introduced by the GC clusters. One possibility is that a cut and splice process would excise the G+C-rich sequences from the RF3 transcript and would restore the continuity of the frame. To test this possibility we performed S1-mapping experiments by hybridizing total mitochondrial RNA of strain D273-10B/A to selected restriction DNA fragments labeled either at their 3'or 5'-ends and overlapping either the first or the third GC clusters. Each resulting DNA-RNA hybrid was then subjected The probe used was the EcoRI-EcoRI fragment from strain NCYC74 which includes the RF3 sequence. The autoradiograms of 7 out of 20 strains studied are shown. Strains 4,6,7,9,11,12,14,16,18,and 19 lack the RF3 sequence (pattern B). The dotted fragment drawn on the restriction map B corresponds to the DNA sequence localized immediately downstream of the EcoRI-EcoRI fragment of strains such as D273-10B/A that contain the RF3 sequence (Skaphin et al.,  1985). The HpaII sites on restriction map D were not accurately located. E , EcoRI; H , HpaII.
to S1 nuclease trimming, and the length of the nucleaseresistant fragments produced was analyzed on polyacrylamide-urea gel slabs. The probes used are depicted in Fig. 8. Using as a probe the RsaI-AuaII fragment 5'-labeled a t its AuaII end (Probe 1, Fig. 8), an S1 nuclease-protected fragment was detected ending just at the TAGT sequence of cluster I. The 3'-end-labeled AhaIII-Aha111 probe overlapping cluster I (probe 3, Fig. 8) located a cleavage near the 3'-end of the AAGGAG sequence (see also Fig. 9). Similar results were obtained for cluster I11 with the RsaI-HinfI probe 5'-labeled at its HinfI end (probe 2, Fig. 8) and with the 3'-end-labeled AhaIII-PstI probe (Probe 4, Fig. 8). This suggests that the DNA probes were protected by RNA species retaining the GC-rich sequences either a t their 5'-or 3'-end (cf. Fig. 8,  lower part). These results are unexpected. We rather anticipated that probes hybridized to spliced RNA molecules would be cleaved by S1 nuclease on both sides of the excised GCrich RNA sequences or in the loop of their hairpin structure (de Zamaroczy and Bernardi, 1986). In any case, signals corresponding to size-protected probes were in large excess over those produced by the hybridization to "cleaved" RNA molecules with probes 1, 2, and 4.  FIG. 5. Nucleotide sequence of the RR8 fragment of strain NCYC74. The last 27 amino acid residues of oli2 and the sequence of the RF3 putative protein are indicated by the single-letter amino acid code, using the codon recognition rule of yeast mitochondria where TGA is used for tryptophan, CTN for threonine, and ATA for methionine. The comparison of the RR8 sequences of strains NCYC74 and D273-10B/A shows that in this latter strain the three GC clusters have been inserted at three AG sites. These sites are underlined and marked with the symbol '7. Boxes enclose the two conserved dodecapeptidic sequences PI and P2. The glutamic residue substituting the glycine residue of strain D273-10B/A is enclosed in an open circlq the nucleotide changes of the corresponding codon is underlined.

TABLE I1 Strain polymorphism in the RFI, RF2, and RF3 regions
The numbers referring to the name of strains (Table I)   lines localize the DNA segments which have been sequenced. The comparison of the sequence of these segments from strain NCYC74 with the corresponding sequence from strains D273-10B/A (Michel, 1984) shows that in this latter strain the three GC clusters have been inserted a t three AG sites (note that in strain D273-10B/A the last two GC clusters are oriented in a direction opposite to that of the first one). These sites are underlined and marked with the symbol y. The RF2 reading frame is continuous in the sequences A and B. H, HindIII; X , XbaI; S , SpeI; E, EcoRI. . Northern blot analysis of the mitochondrial RNA of strains BS104-1 and BS104-9. Strains BS104-1 and BS104-9 are isonuclear and contain the mitochondria of D273-10B/A and NCYC74, respectively. These strains were built by cytoduction (Lancashire and Mattoon, 1979). First, K5/2 was crossed with either D273-10B/A or NCYC74; two cytoductants, K273-1 and KUVA32,

DISCUSSION
Our studies lead to two new results. First, they demonstrate that the RF3 open reading frame is continuous and transcribed in strain NCYC74 and clearly suggest that RF1 and RF2 are not interrupted by GC clusters in that strain. Consequently, these genes must be functional. Second they bring to light a rule governing the translocation of GC clusters considered as mobile elements. RF1, RF2, and RF3 Are Maturase-like Genes-The fact that on the one hand RF1, RF2, and RF3, which are continuous open reading frames in S. uuarum, are interrupted by GC clusters that break the continuity of the phases in strain D273-10B/A and that, on the other hand, GC clusters apart, these frames may potentially code for 500 amino acid long proteins leaves us perplexed; are these genes nonfunctional in strain D273-10B/A or are there any mechanisms able to correct the frame disorder introduced by the GC clusters? We have tested the possibility that a splicing mechanism could excise the GC clusters. Our S1 nuclease-mapping experiments which were done with probes overlapping RF3I and RF3III GC clusters seem to rule out this possibility. The RNA species observed might then be intermediates in the degradation of the RF3 mRNA. As an alternative explanation the transcriptional or translational machinery could produce slippage and frameshifting when going across the G+C-rich sequences (Benne et al., 1986;Clare and Farabaugh, 1985).
harboring the nucleus of K5/2 and the mitochondria of D273-10B/A or NCYC74, respectively, were isolated. Strains K273-1 and KUVA32 were crossed with GRTF18/2, and two cytoductants BS104-1 and BS104-9 containing the nucleus of GRF18/2 were obtained. The estimated sizes of the major transcripts (in bases) extrapolated from that of the 21 S rRNA (3270 bases) and 15 S rRNA (1680 bases) used as size markers are indicated in the margin. a, BS104-9; b, BS104-1. In the lower part of the figure the major transcripts (4100 and 3500 bases) are localized on a map of the oli2-RF3 mitochondrial region. The wauy lines schematize transcripts. Arrowheads indicate the position of the dodecamer AATAATATTCTT (cj. Simon and Faye, 198413  HpaII; V, AuaII; L, AluI; P, PstI; N , HinfI); b, probes used for S1 mapping. Middle part offigure: autoradiogram (numbers 1-4 refer to the probes). Probes were hybridized to 4.5 or 9 pg of total mtRNA of strain D273-10B/A in an 80% formamide-containing buffer. RNA-DNA hybrids thereof were then subjected to S1 nuclease trimming as described in Simon and Faye (1984a). Controls for S1 nucleaseresistant fragments generated by strain reannealing were performed in a parallel hybridization mixture where E. coli tRNA instead of mtDNA was used. The size of the S1 nuclease-resistant fragments was determined on 6 or 8% acrylamide/urea sequencing gels (Maxam and Gilbert, 1980) run at 75 watts. Truck 1: lanes a and b, C and T sequence ladders of the probe; lane c, S1 nuclease-resistant fragments; lane d, control of strand reannealing. Track 2 lane a, control of strand reannealing; lane b, S1 nuclease-resistant fragments; lanes c and d, T and C sequence ladders of the probe. Track 3 lane a, control of strand reannealing; lane b, S1 nuclease-resistant fragments; lanes c and d, A and G sequence ladders of DNA of known sequence . Truck 4 lanes a,  b, and f , G, A, and T sequence ladders of the probe, respectively; lane c, S1 nuclease-resistant fragments; lane d, probe; lane e, control of strand reannealing. b indicates S1 nuclease-resistant fragments. Hybridization temperatures were 37,44.5,38, and 40 "C for probes 1, 2, 3, and 4, respectively. Lower part of figure: a schematic summary of the S1 mapping result. Probes are drawn as wavy lines. The 32P-   FIG. 9. Insertion sequence of GC clusters. The insertion sequences of GC clusters in the RF3, RF2,21 S rRNA, aI5P sequences, and in ori regions were compared. The sequence outside the brackets are those of the strains without GC clusters. The duplicated AG sequences are underlined. GC clusters are aligned to illustrate their sequence similarity. The sources of the sequences are as follows: 21 S rRNA (Dujon, 1980); ori (de Zamaroczy et al., 1984); RF2 (Michel, 1984); RF3 (Seraphin et al., 1985). The sequence of aI5P of strain KL14-4A (Hensgens et al., 1983) was compared with that of strain 777-3A (Seraphin et al., unpublished results). The numbers indicating the position of GC clusters in aI5P refer to the sequence published by Hensgens et ul., 1983. Note that the sequence between nucleotides 2325 and 2316 is TATTMATT in KL14-4A. Stars mean that the sequences were not determined.
We have shown that some strains do not possess either RF3 or RF2 and RF3. How do such strains compensate for the lack of these genes? The RF1, RF2, and RF3 reading frames contain the P1 and P2 dodecapeptides characteristic of the open reading frame of group I introns. Therefore, we may suppose that the functions of RF1, RF2, and RF3 are related to those of the intronic maturases and that some of them may take the role of the RF2 or RF3 gene products. It has been suggested that the mitochondrial introns could result from the insertion of mobile elements both within and outside the mitochondrial genes (Hensgens et al., 1983). The gene(s) born by this putative transposon (perhaps related to the retrovirus) would have evolved to the present day maturases. This model could explain why maturases may share some properties with the RF1, RF2, and RF3 gene products.
GC Clusters Are Mobile Elements-Since from one strain to another the same mitochondrial sequence is or is not interrupted by GC clusters (Dujon, 1980;Sor and Fukuhara, 1982;de Zamaroczy and Bernardi, 1986) we have to suppose that these G+C-rich sequences are either themselves mobile elements or are remnants from former mobile elements whose terminal repeats would be GC clusters. Our finding that seven GC clusters are optional in the RF1, RF2, or RF3 sequences strongly supports these hypotheses. Furthermore, we have discovered a rule which seems to explain the insertion of GC clusters. Fig. 9 clearly shows that the target site is the AG sequence. This AG sequence appears duplicated on either side of GC clusters (except for RF3II where an additional A is present, although a sequencing error is not excluded, because the palindromic sequence of GC clusters may cause "band compression" during electrophoresis (Frank et al., 1981)). However, we do not know whether one AG is brought about by the GC clusters or whether the AG sequence of the host DNA is duplicated during the insertion process as happens for the bacterial transposable elements (Calos and Miller, 1980). Most of the GC clusters of the a1 and a2 families (about 50 clusters), as described by de Zamaroczy and Bernardi (1986), are delimited by AG sequences which indicates that the insertion of a large number of GC clusters should obey the rule we have discovered.
The actual function of GC clusters has not yet been definitively established though it has been suggested that some of them are elements of replication origins (de Zamaroczy and Bernardi, 1986). Other roles could be hypothesized. However, several GC clusters are optional (Fig. 91, and their presence or absence is apparently not detrimental to the cell. Then the question of how necessary are these optional GC clusters for the life of the cell may be asked.