A Yeast Sigma Composite Element, TY3, Has Properties of a Retrotransposon*

Sigma is a 340- or 341-base pair repetitive element which is located almost exclusively within 19 base pairs of the 5’ ends of various tRNA genes in the Saccharomyces cerevisiae genome. Although most sigma elements characterized to date are isolated insertions, a few of the elements occur relatively closely spaced. One of these pairs is a direct repeat of the sigma element separated by an internal domain 4.7 kilobase pairs in length. Not only does this structure resemble a composite transposable element, but regions within the sigma elements and intervening domain are homologous to conserved regions in retroviruses and retrotransposons of yeast and other organisms. Two features suggest that the sigma elements and intervening DNA transposed in a concerted event: only one of the two sigma elements is associated with a tRNA gene, and only the outside ends of the two elements are flanked by the 5-base pair direct repeats that usually flank individual sigma insertions. Examination of gen- omic DNA from five laboratory strains indicates that the 4.7 kilobase pair internal domain is present in one to four copies per haploid genome and that the genomic location of this domain differs from strain to strain. In addition, Northern blot analysis showed the presence of a 5.2 kilobase poly(A) transcript which hybridizes to both sigma and internal domain-specific probes. The existence of this composite element may suggest new ways to consider the mechanisms by which retrotrans- posons select their targets.

Sigma is a 340-or 341-base pair repetitive element which is located almost exclusively within 19 base pairs of the 5' ends of various tRNA genes in the Saccharomyces cerevisiae genome. Although most sigma elements characterized to date are isolated insertions, a few of the elements occur relatively closely spaced. One of these pairs is a direct repeat of the sigma element separated by an internal domain 4.7 kilobase pairs in length. Not only does this structure resemble a composite transposable element, but regions within the sigma elements and intervening domain are homologous to conserved regions in retroviruses and retrotransposons of yeast and other organisms. Two features suggest that the sigma elements and intervening DNA transposed in a concerted event: only one of the two sigma elements is associated with a tRNA gene, and only the outside ends of the two elements are flanked by the 5-base pair direct repeats that usually flank individual sigma insertions. Examination of genomic DNA from five laboratory strains indicates that the 4.7 kilobase pair internal domain is present in one to four copies per haploid genome and that the genomic location of this domain differs from strain to strain. In addition, Northern blot analysis showed the presence of a 5.2 kilobase poly(A) transcript which hybridizes to both sigma and internal domain-specific probes. The existence of this composite element may suggest new ways to consider the mechanisms by which retrotransposons select their targets.
Transposable elements are DNA sequences which are capable of movement from one position to another in the genome. Depending upon their sites of insertion, they can modify both the expression and arrangement of host genes (for reviews, see Kleckner, 1981;Finnegan, 1985;Roeder and Fink, 1983;Williamson, 1983;and Varmus, 1983). One class of transposable elements, which includes the retroviruses of birds and mammals and the retrovirus-like (retrotransposon) * This work was supported by National Institutes of Health Grant GM 33281 (to S . B. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

to the GenBankTM/EMBL Data Bank with accession number(s)
The nucleotide sequence(s) reported in this paper has been submitted 503500.
$Supported by a President's Fellowship from the University of California.

Supported by National Institutes of Health Predoctoral Training
1 To whom reprint requests should be addressed. elements of yeast and fruit flies, transposes through an RNAmediated process. This common mode of transposition is reflected in the similar structures of the integrated elements. Retrotransposons are flanked by 4-6-bp' direct repeats of target sequence. Each element is a composite consisting of an internal region of several kbp, flanked by LTRs of several hundred bp each. The ends of these LTRs are inverted repeats, up to 10 bp in length, which terminate with the conserved dinucleotides TG-CA. The LTRs contain initiation, termination, and polyadenylation signals which direct transcription of the internal coding sequences. Hormone-response sequences and/or enhancer sequences (Laimins et al., 1983;Errede et al., 1985;Roeder et al., 1985) which regulate this transcription have also been identified, both in the LTRs and in the promoter-proximal portion of the internal domain. The retrotransposon gene products that have been identified include proteins with homology to the retroviral gag-and polencoded proteins (Saigo et al., 1984;Mount and Rubin, 1985;Clare and Farabaugh, 1985;Hauber et al., 1985;and Warmington et al., 1985).
Two closely related sets of yeast retrotransposons have been described, Tyl, or Class I (Cameron et al., 1979;Farabaugh and Fink, 1980;Gafner and Philippsen, 1980), and Ty2, or Class I1 (Kingsman et al., 1981). They are present in a total of 30-35 copies per haploid genome in standard laboratory strains. Tyl and Ty2 elements consist of an internal domain, called epsilon, which is 5.3 kbp in length, and two flanking direct repeats, called delta elements, which are 334-338 bp in length. Ty insertion in the 5"flanking sequence of a gene can inactivate that gene (Chaleff and Fink, 1980;Roeder and Fink, 1980;Eibel and Philippsen, 1984) or increase its expression, an increase which is correlated with the expression of Ty itself (Errede et al., 1980;Williamson et al., 1981). Ty elements contain blocks of sequence (Roeder et al., 1985;Errede et al., 1985) with homology to the SV40 enhancer core sequence (Laimins et al., 1983) that overlap blocks of sequence with homology to mating-type control regions (Strathern et al., 1981;Miller et aL, 1985;Siliciano and Tatchell, 1986).
Regions containing these elements mediate at least some of the observed regulation of Ty transcription. The expression of T y in yeast has also been correlated with levels of reverse transcriptase activity Mellor et al., 1985). Boeke et al. (1985) recently showed that increased transcription of a particular Ty element resulted in increased levels of Ty transposition. In addition, when a yeast intron was introduced into a marked Ty, it was spliced out during transposition, thereby demonstrating an RNA intermediate.

A Sigma Composite Element in Yeast
Resolution of sequence differences between the two delta termini of a single Ty in the course of transposition was observed by Boeke et al. (1985) and is consistent with a retrovirus-like mechanism of transposition. In addition to delta, two other short, repetitive elements, sigma (del Rey et al., 1982;Sandmeyer and Olson, 1982) and tau (Genbauffe et al., 1984), are found in the yeast genome. Isolated delta, sigma, and tau elements are present in 15-30 copies each in the yeast haploid genome. Sigma and tau elements are related to delta elements by sequence similarities at their termini and by 6-11 short (5-10 bp) similar sequences spaced between the termini (Genbauffe et al., 1984). Sigma, tau, and delta elements are all flanked by direct 5-bp repeats of their insertion sites; however, no de novo transposition of any of these isolated elements has ever been demonstrated. Unlike delta elements, sigma and tau elements have not been found associated with large repetitive coding domains, such as epsilon. This has encouraged speculation that sigma and tau elements are not active transposable elements. On the other hand, the possibility remains that an unidentified subset of these elements might exist as part of a larger, mobile genetic element.
Sigma elements are distinguished from delta and tau elements by their extremely close association with tRNA genes (Brodeur et al., 1983). A survey of the position and orientation of sigma elements in one strain of yeast has shown that 15 of 17 sigma elements are found starting at -16 to -19 bp upstream of tRNA genes.' While most sigma elements are separated in the genome, a few are found in proximity to one another (Brodeur et al., 1983). We have investigated the properties of one of these pairs of sigma elements, and we report here that, together with a 4.7-kbp intervening DNA sequence, it possesses the structural features of a retrotransposon described above. In addition to these properties, almost full-length transcripts from this element are present in the poly(A) RNA. These observations suggest that this sequence constitutes a retrotransposon with sigma termini. If this element is a member of an uncharacterized subset of sigma elements which are transpositionally active, then this subset has retained the unusual tRNA-gene association which characterizes the isolated elements previously described. The present studies should aid in the construction of an appropriate selection system in which the position-specific transposition specified by the sigma element could be observed.
Enzymes and Chemicals -Restriction endonucleases (with the exception of NdeI), DNA polymerase I (large fragment), and calf alka-S. B. Sandmeyer, V. W. Bilanchone, D. J. Clark, P. Morcos, G. F. Carle, and G. M. Brodeur, unpublished data. line phosphatase were purchased from Boehringer Mannheim. NdeI and T4 DNA ligase were obtained from New England Biolabs. Polynucleotide kinase, S1 nuclease, and T4 RNA ligase were obtained from P-L Biochemicals. Sequenase was obtained from U. S. Biochemical Corporation. Radionuclides, including [y-32P]ATP (5000 Ci/ mmol), [a-3ZP]dATP (3000 Ci/mmol), [(u-~'P]GTP (3000 Ci/mmol) and cytidine [5'-32P] 3',5'-bisphosphate (3000 Ci/mmol), were obtained from Amersham Corp. The hexanucleotide primer mixture for randomly primed synthesis of radioactive probes was obtained from P-L Biochemicals. Total bakers' yeast tRNA (type X-S), used as a probe for tRNA genes, and yeast a-factor were obtained from Sigma. Oligo(dT)-cellulose was obtained from Collaborative Research. Plasmid pIBI20 was purchased from International Biotechnologies, Inc. The synthetic oligonucleotide, 5' GTAACCTTTAACGCTACCAAG 3', was synthesized by Louise Schmidt, University of California, San Diego, PON Facility. It is complementary to the Ty3 transcript in the left-most region of the internal domain. The set of RNA markers was obtained from Bethesda Research Laboratories.
Manipulations of Recombinant Clones-Rapid isolation of analytical amounts of plasmid DNA was by the alkaline lysis procedure of Birnboim and Doly (1979). Large scale preparation of DNA was carried out using a modification of the boiling method (Holmes and Quigley, 1981). Supercoiled plasmid DNA was further purified by centrifugation to equilibrium in a cesium chloride density gradient.
The 6.5-kbp HindIII/EcoRI fragment from lambda clone 112, which contained the two sigma elements, was subcloned by standard procedures into the HindIII and EcoRI sites of the pIBI2O polylinker. The recombinant plasmid is referred to as pSBS12. Single-stranded pSBS12, of the same sense as the Ty3 transcript, was produced by infecting NM522 cells transformed with pSBS12 with the singlestranded helper phage M13K07 (IBI). The 112 lambda clone was identified originally as a clone containing two sigma elements associated with tRNA genes (Brodeur et al., 1983); the present studies indicate unambiguously that only one tRNA is present on this HindIII/EcoRI fragment (see "Results"). Restriction sites in the yeast insert were mapped by digestion with various combinations of enzymes including HindIII, EcoRI, AccI, NdeI, PstI, Sen, and XhoI. Digestions were carried out under standard conditions specified by suppliers' instructions. The cloning of pPM15s, a 204-bp internal fragment from the sigma element associated with SUQ5, into the EcoRI site of pBR322, was described previously (Brodeur et al., 1983).
The XhoI/TuqI 339-bp fragment, containing almost the complete SUP2 sigma element (Sandmeyer and Olson, 1982), was cloned in this study into the EcoRI and HindIII sites of pSP64 yielding the plasmid, pSBS6. The Ty probes were derived from T y l (S13) (Cameron et al., 1979) from R. Davis, Stanford University. and flanking DNA in pSBS12 was by the chemical modification Sequence Analysis-Sequence analysis of the two sigma elements method of Maxam and Gilbert (1980). The sequencing strategy is shown in Fig. 1. Sequence was compiled, edited, and translated using the DNA sequence analysis programs of A. Goldin (California Institute of Technology) and G. Gutman (University of California, Irvine).
Comparisons with the GenBank Data Base were made through the core library programs of BIONET (Intelligenetics, Inc.). The homologous sequence ladder used for size determinations in the 5"mapping experiment was primed with the synthetic oligonucleotide described above and synthesized by the dideoxy-chain termination method (Sanger et el., 1977) on a double-stranded pSBS12 DNA template with reverse transcriptase. The sequence ladder used for 3' mapping was produced by chemical modification and degradation of the homologous end-labeled DNA.
DNA and RNA Hybridization Analysis-DNA and RNA hybridization probes were prepared from restriction fragments by randomly primed synthesis in the presence of [LY-~'P]~ATP, as described by Feinberg and Vogelstein (1983;. The tRNA probe was prepared by reacting the 3' ends of yeast tRNA with [5'-32P]cytidine 3',5'bisphosphate in the presence of T4 RNA ligase under conditions specified in a modification of the procedure of Bruce and Uhlenbeck (1978;Pirtle et al., 1980). Unincorporated label was separated from tRNA by chromatography over Sephadex G-50. RNA size markers were made by radiolabeling the Bethesda Research Laboratory RNA ladder in the same manner. The synthetic oligonucleotide was endlabeled by reacting it with [y3'P]ATP in the presence of polynucleotide kinase.
Yeast genomic DNA was prepared from cells grown to early stationary phase in 1% yeast extract, 2% peptone, and 2% glucose medium, following the protocol described by Boeke et al. (1985). DNA samples for hybridization analyses were digested with restriction endonucleases and fractionated by electrophoresis in 1% agarose gels, buffered in 2.5 mM EDTA, 45 mM borate, and 133 mM Tris-HC1, pH 8.3. Electrophoresis was at 2 V/cm for 14-16 h. The DNA was transferred to nitrocellulose by the method of Southern (1979). Radioactive probes were denatured by boiling for 3 min and chilling immediately on ice. Blots were hybridized in 1% SDS, 0.5 M NaCI, 2.5 mM EDTA, and 0.1 M Na2HP04, pH 7.0, for 14-16 h at 65 "c (DNA probe) or 55 "C (tRNA probe). Size markers were lambda DNA cut with HindIII, and 5' end-labeled with y3'P or hybridized to a radioactive lambda probe. After hybridization, filters were washed four times, for 5 min each, in 10 mM Tris, pH 8.0. The first wash included 0.1% SDS. Filters were exposed to Kodak XAR-5 film at room temperature or at -70 "C with a Cronex Quanta 111 intensifying screen, manufactured by Du Pont.
Yeast RNA was isolated from cultures grown to OD, = 0.7 to 0.8 (approximately 3 X lo7 cells/ml of broth) by vortexing the cells with glass beads in the presence of buffer, chloroform, and phenol as described by Elder et al. (1983). Cells were treated, where noted, with 0.4 p~ a-factor for 30 min before RNA isolation, as described by Stetler and Thorner (1984). Poly(A) RNA was further purified by chromatography over oligo(dT)-cellulose as described by Aviv and Leder (1972). Poly(A) RNA samples for hybridization analyses were denatured by reaction with glyoxal as described (McMaster and Carmichael, 1977) and subjected to electrophoresis in a 1.1% agarose gel in 10 mM Nap, pH 7, at 6 V/cm for 3 h. The RNA was transferred directly to nitrocellulose, hybridized, and washed as described by Thomas (1980). Nitrocellulose filters with bound RNA were preincubated in hybridization buffer for 1-2 h. Radiolabeled probes were denatured as before and added to the hybridization buffer. Hybridization buffer was 4 X SSC (1 X ssc is 0.15 M NaC1,0.015 M sodium citrate, pH 7.0), 10% dextran sulfate, 0.8% Denhardt's solution, 40 mM Na2P0,, pH 8.6, and 40% formamide. Hybridizations were for 14-16 h at 38 "C. Following hybridization, the filters were washed four times in 2 X SSC and 0.1% SDS, at room temperature, two times in 0.1 X SSC and 0.1% SDS, at 38 'C, and once in 2 X SSC at room temperature. The washes were 5 min each, except the washes at 38 "C, which were 15 min each. Hybridization with the 5'-labeled oligonucleotide probe was as described by Berent et al. (1985).

Determination of the 5' Ends of the Ty3 Transcript by Reverse Transcription Extension of a Synthetic Oligonucleotide Primer-
The 5' ends of the Ty3 RNA were determined by reverse transcription from an end-labeled synthetic oligonucleotide primer (5' GTAACCTTTAACGCTACCAAG 3') hybridized to poly(A) RNA (Ghosh et al., 1978). This oligonucleotide is complementary to the Ty3 transcript for the first 21 bases within the internal domain and downstream of the left sigma element. 15 pg of poly(A) RNA was annealed to 2 X lo6 cpm of oligonucleotide primer for 2 h at 37 "C in 5 X reverse transcription buffer (1 X is 50 mM Tris, pH 8.3, 10 mM MgClz, 0.4 mM dithiothreitol, and 80 mM KC1). The synthetic reaction was carried out for 1 h at 37 "C with 6.5 units of reverse transcriptase per reaction volume of 21 p1 in 1 X buffer. Actinomycin D was present at 100 pg/ml and dNTPs were present at 500 p~ in the final reaction. One-third of the precipitated reaction, redissolved in electrophoresis sample buffer, was loaded per lane on a denaturing gel. Electrophoresis conditions were as described by Sanger et al., 1977. Determination of the 3' Ends of the Ty3 Transcript by S1 Nuclease Protection Mapping-The 3' ends of the Ty3 transcripts present in a-factor-treated cells were determined by S1 nuclease protection experiments performed essentially as described (Berk and Sharp, 1978). The end-labeled DNA used for determination of the 3' end of the Ty3 transcript was a 1-kbp AccI restriction fragment which spans the downstream sigma element (see Fig. 1). This fragment was specifically labeled on the 3' end that occurs within the Ty3 internal domain, by extension with Klenow polymerase in the presence of [a-32P]dCTP and cold dGTP, dTTP, and dATP. The degeneracy of the AccI recognition site made it possible to label the internal end of this fragment specifically. 5 X lo5 cpm of denatured, end-labeled DNA was hybridized to 15 pg of poly(A) RNA and digested with 1000 units or 1500 units of S1 nuclease per reaction. The sizes of the endlabeled, protected fragments were determined by electrophoretic migration on denaturing acrylamide gels compared to a sequence ladder of the same strand. The sequence ladder was produced by chemical modification, followed by pyridine degradation (Maxam and Gilbert, 1980) of the end-labeled AccI fragment.

RESULTS
A Repeated Sequence Flanked by Sigma Elements-In a survey of sigma elements from yeast strain AB972, one family of lambda clones, the V family (including clone 112), was identified which contains two sigma elements separated by 4.7 kbp of DNA (Brodeur et al., 1983). In order to facilitate discussion, we will refer to the region between these two sigma elements as the internal domain and the entire structure, including the sigma elements, as the sigma composite element (Ty3). This composite structure is reminiscent of the T y l and Ty2 elements, which are composed of an internal domain flanked by direct repeats of delta elements. Because delta elements have some similarities with sigma elements, a structural study of the composite sequence in lambda clone 112 was undertaken to determine whether it constitutes a sigmaflanked, transposed sequence. The 6.5-kbp HindIIIIEcoRI fragment from the lambda V family, which contains the two sigma elements, 4.7 kbp of intervening DNA and 1.2 kbp of flanking DNA, was subcloned into the HindIIIIEcoRI site of pIBI20. This clone was designated pSBS12.
Several approaches were pursued to determine if the two sigma elements in pSBS12 were functionally associated or were only coincidentally closely positioned in the genome. First, the yeast DNA insert was digested with various combinations of AccI, NdeI, XhoI, PstI, and Sa& and the fragments were analyzed to deduce a restriction map (Fig. 1). The internal domain between the two sigma elements was then analyzed for possible sequence homology to the epsilon region of Ty. The fragment containing the yeast insert in pSBS12 was isolated by digestion with EcoRI and HindIII followed by electrophoresis in a 1% agarose gel and was then transferred to nitrocellulose by the method of Southern (1979). The DNA was hybridized with a 32P-labeled, 1.6-kbp EcoRIISalI frag- A Sigma Compositt ment from the internal region of Ty1 element S13 (Cameron et al., 1979) and with a 32P-labeled, 2.5-kb SalI fragment, which includes a delta element from the S13 Ty. The second probe includes sequences which are homologous to 5 1 and Ty2. Neither the epsilon nor the epsilon plus delta probes hybridized to the internal domain of the sigma composite. The internal domain also failed to hybridize to a plasmid which contained a copy of the tau element.' Because the internal domain did not show homology to Tyl or Ty2 by the criterion of hybridization under these conditions, the alternative possibility, that the two sigma elements in pSBS12 are part of a repetitive element which is not closely related to Tyl and Ty2, was investigated. Genomic DNA from X2180-1A was digested with EcoRI, fractionated by electrophoresis, and transferred to nitrocellulose by the method of Southern. Restriction fragments that together spanned the entire yeast insert in pSBS12 were labeled and used as hybridization probes (Fig. 1). A sigma-specific probe, the pPM15s insert, was used to visualize the genomic fragments with sigma similarity. The autoradiograph of these hybridizations is shown in Fig. 2. Probes derived from fragments located outside the sigma elements hybridized only to a fragment 9.7 kbp in length (probes a and h, lanes 2 and 8). Interestingly, all of these internal probes hybridized to both the DNA fragment 9.7 kbp in length, which hybridized to probe a, and to another fragment 6.5 kbp in length, which is not represented in the probe a pattern. The patterns generated by the sigma-specific probe and probe b are more complex but also have a band at both 9.7 and 6.5 kbp, suggesting that the internal domain is associated with sigma elements on the 6.5-kbp fragment as well.
These results indicate that the region between the two sigma elements in the pSBS12 insert is repeated in the genome, whereas the regions outside the sigma repeats are represented only once. This hybridization pattern is consistent with a repeated domain, approximately 5 kbp in length, which terminates near or within flanking sigma elements. cereuisiue. The genomic DNA from these strains was prepared as described under "Experimental Procedures" and digested with EcoRI. The restriction fragments were separated by electrophoresis and transferred to nitrocellulose. The various DNAs were each hybridized with probe e (as defined in Fig.  1) from the internal domain. The rather striking result is shown in Fig. 3: the number of fragments which hybridize to the internal domain sequence ranges from one to four in the five different genomes tested, and the sizes of the hybridizing fragments from different strains are dissimilar. Because neither probe hybridized to any fragments smaller than 6.5 kbp, the pattern is consistent with elements of 5.4 kbp which do not have internal EcoRI sites. If EcoRI sites are absent in the genomic copies of the pSBS12 composite sequence, then these different numbers of bands correspond to different numbers of copies of sequences closely related, or identical to, the ' *

A Sigma
Composite Element in Yeast 1417 pSBS12 internal domain. The size variation observed for hybridizing fragments suggests that most of these copies are in different genomic locations in different strains. Although restriction site polymorphisms could also explain these observations, hybridization of the composite probe to separated chromosomal DNAs has shown that, at least in some strains with two and four hybridizing fragments, the hybridizing sequences are located on different chromosomes?
Coordinate Insertion of Sigma Elements and the Internal Domain-The genomic Southern blot shown in Fig. 2 places the ends of the repeated region of pSBS12 within or near the sigma elements. We hypothesized that this composite structure could be explained by the coordinate transposition of the sigma elements and internal domain into this genomic locus. We therefore examined the postulated end points of the composite element for structural evidence of a transposition event. Of the 17 sigma elements we have sequenced from AB972, all but two are within 19 bp of the 5' end of a tRNA gene'; thus, a tRNA gene is one of the most conserved features of a sigma insertion site. We examined the regions surrounding the two sigma elements in pSBS12 for the expected tRNA gene target(s) by hybridization with a 32P-labeled total yeast tRNA probe. In order to assess each sigma-containing region separately, the pSBS12 plasmid was digested with HindIII and AccI, and a combination of HindIII, AccI, and XhoI, and subjected to electrophoresis on a 1% agarose gel. Fig. 4A shows the pattern produced by the ethidium-stained DNA. This DNA was transferred to nitrocellulose and hybridized to the tRNA probe. The autoradiograph of the filter is shown in Fig. 4B. The tRNA probe hybridized only to the AccI fragment (lune 2) and to the smaller 700-bp XhoI/AccI fragment containing the HindIII site proximal-flanking sequence and sigma (lune 3). This result is consistent with a single transposition which inserted two sigma elements and intervening DNA near one tRNA gene target.
Sigma insertions are flanked by a 5-bp direct repeat of the target sequence (Sandmeyer and Olson, 1982). The regions of DNA surrounding the two sigma elements in pSBS12 were sequenced by the method of Maxam and Gilbert (1980). The strategy is indicated in the restriction map of Fig. 1. The sequence of the two sigma elements in pSBS12 and their immediate flanking regions is shown in Fig. 5A. These sigma elements are shown in the opposite orientation compared to sigma elements previously described (del Rey et al., 1982;Sandmeyer and Olson, 1982). This sequence and the restriction map in Fig. 1 are oriented to correspond to the direction of transcription of the major sigma-related transcripts (left to right, 5'-3') as determined by Northern blot hybridization with single-stranded probes3 (Van Arsdell et al., 1987). In contrast to other cases of sequenced sigma elements (del Rey et al., 1982;Sandmeyer and Olson, 1982), neither of these elements is immediately flanked by 5-bp direct repeats; instead, the 5-bp sequence on the outside end of the right sigma reiterates the 5-bp sequence found on the outside end of the left sigma. These data are consistent with a single staggered cleavage of the insertion site from which direct repeats of the insertion site sequence on the outside ends of the sigma composite element resulted.
The sequence analysis corroborated and extended the conclusions from the tRNA hybridization shown in Fig. 4B. A 72-bp sequence with perfect homology to tRNAcy" (Holness and Atfield, 1976) is found extending from position -16 to position -87 (overlined in Fig. 5A). The alignment of this sequence with the tRNA sequence is shown in Fig. 5B. This coding sequence does not have an intervening sequence and is perfectly homologous to the coding region sequenced by Genbauffe et al. (1984). Absence of homology in the flanking sequence indicates that this tRNACY" gene is not allelic with the one they described. Because probe b (Fig. 1) actually contains this gene in addition to the sigma element, a plausible explanation of the additional band in the probe b genomic hybridization pattern (Fig. 2, lane 3, arrow) is the presence of a t least one additional genomic copy of an homologous tRNACy* gene. The sigma-flanking sequences shown in Fig.  5A were also inspected for conserved patterns in tRNA genes (Galli et al., 1981). compared to National Institutes of Health GenBank yeast tRNA sequences for homology, and searched for potential intrastrand pairing by BIONET core programs (Intelligenetics). Although there were regions of short homology to tRNA-coding sequences in and around the sigma elements, no evidence was discovered for any additional tRNA genes in these regions.
The properties described above characterize a repeated element which terminates with direct repeats of the sigma element and which is organizationally homologous to the previously described yeast transposable elements, Tyl and 5 2 . Because of this organizational homology and because of properties this composite element shares with eukaryotic transposable elements, we refer to this element as 5 3 .

~R N A~Y~ GCUCGUAUGGCGCAGDGGDAGCGCAGCAGAYUGCAAAICUGUUGGDCCUUAGTYCGAUCCUGAGUGCGAGCUCCA
" " " " " " " " " " " " " " " " " " " " " " " " flanked by sigma elements is similar to the organization of the retrotransposon class of transposable elements. The sequence similarity between the termini of sigma elements and retrotransposons has already been noted. In Fig. 5C we compare the terminal repeat-internal domain junction sequences of Ty3 to analogous junction regions which are conserved among retrovirus-like elements and retroviruses. The Tyl (Farabaugh and Fink, 1980;Elder et al., 1983), Ty2 (Warmington et al., 1985), copia (Levis et al., 1980), and SNV (Shimotohno et al., 1980) sequences are aligned to allow maximum correspondence of bases. Starting at the right-hand terminus of the left sigma, the 8 base pairs which form the terminal inverted repeats are homologous to copia and SNV termini, with the exception of the first base. The left ends of the internal domains of all five elements have some sequence homology. This homology is greatest among Tyl, Ty2, and 5 3 elements. The 8 base pairs which are identical among the three yeast elements perfectly complement a region at the 3' end of tRNA? (Venegas et al., 1982; previously noted by Eibel et al., 1980, for Tyl). In retrovirus RNA this region is the minus strand primer-binding site and is typically complementary to a tRNA. The left-most 2 bases of the internal domain are not part of the region with tRNA homology. Thus if Ty3 is replicated from a tRNA primer, it is predicted, as has been shown for retroviruses such as SNV, to lose its terminal dinucleotide sequences upon integration. At the 3' end of the internal domain of retroviruses is an oligopurine sequence postulated to play a role in plus-strand priming (see Varmus, 1983 for review). Similarly, at the right end of the transcript strand of the internal domain flanked by sigma elements is a sequence of 14 purine residues. This sequence of purines is followed by the sigma terminal repeat in the inverted orientation. Unlike the other elements shown here, the sigma termini are perfect inverted repeats of each other. This right-hand end is also identical to the Tyl-and Ty2terminal regions at 6 out of 8 positions.
The identity of the two sigma sequences in 5 3 is consistent with the proposed retrotransposon analogy. The RNA intermediate, in retrovirus transposition, contains only a portion of the proviral LTR at each end, the R and U5 regions at the 5' end, and the R and U3 regions at the 3' end. The complete terminal repeats present in the proviral DNA are regenerated by a mechanism of redundant copying of the U5 and U3 regions (for discussion see Varmus, 1983). One consequence of this mechanism of replication is that the LTRs of a single provirus are highly conserved. In the case of yeast Ty elements, for example, the delta elements flanking a particular Ty are often identical or differ by only a few nucleotides (see Williamson, 1983, for a comparison of delta sequences). Although sigma elements are generally more conserved than delta elements, the two sigma sequences present at the ends of this Ty3 are not only identical, but they have in common differences with previously published sigma element sequences. The features described above, suggest that Ty3 is, or was, a member of the retrovirus-like class of transposable elements.
Composite Element Transcription-Poly(A) RNA was prepared from X2180-1A (MATa), X2180-1B (MATa), and ySBS500 (MATalcr) cells. Poly(A) RNA was also prepared prepared from cells grown to mid-log phase. RNA was denatured by glyoxylation, fractionated by electrophoresis, and transferred to nitrocellulose as described by Thomas (1980). 10 pg of RNA was loaded per lane. The nitrocellulose-bound RNA was hybridized to sigma or Ty3 probes. Sizes were determined based on migration relative to end-labeled size marker RNAs. The sizes of these RNA markers are 9.5, 7.5 Nitrocellulose-bound RNA was hybridized to the 1600-bp AccI fragment from the Ty3 internal domain (Fig. 1,probef) labeled by random primer synthesis. C, control hybridization of poly(A) RNA to a URA3 sequence probe. Lanes 1-4 are the same as in B. The nitrocellulose blot which was exposed to give the autoradiograph shown in B was subsequently hybridized to a 1.1-kb Hind111 fragment containing the URA3 gene (Bach et al., 1979) labeled by random primer synthesis. Only the region of exposure showing the URA3-hybridizing transcript is shown here. from all three cell types treated with a-factor as described under "Experimental Procedures." Poly(A) RNA was denatured by glyoxylation, fractionated by electrophoresis, and transferred to nitrocellulose as described by Thomas (1980). The filter-bound RNA was hybridized to a sigma-specific probe, an internal Ty3 probe, and a synthetic DNA oligonucleotide with homology to the internal region of Ty3 (Fig. 6). Hybridization to the sigma-specific probe is shown in Fig. 6A,  lanes 1-4. The major sigma-hybridizing transcripts in MATa, and MATa cells (lunes 1 and 3, respectively) are 6.0, 5.2, 3.1, 0.96, 0.88, and 0.46 kb in length. The effect on transcript end of internal sequence; and 5"inverted terminal repeat from LTR. Line 1 shows the 5 3 internal domain junctions with the left sigma and right sigma. Lines 2-5 show conserved junction regions of Ty912, a 5 1 (class I) element, Tyl-17, a 5 2 (class 11) element, copia, and SNV, respectively. Except for Ty3, these alignments have been considered previously (Varmus, 1983;Warmington et aL, 1985).The bases shown in parentheses are present at the ends of the full-length linear form of SNV but are deleted during proviral integration; the analogous nucleotides occurring in Ty3, i.e. the nucleotides between the minus strand primer-binding site and the LTR, are also shown in parentheses.

A Sigma Composite Element in Yeast
patterns of treating MATa cells with a-factor is shown in / m e 2. The 5.2-and 0.46-kb transcript bands are increased at least in intensity and two new classes of transcripts, 0.67 and 0.61 kb in length appear. Van Arsdell et al. (1987) observed induction of three sigma-hybridizing transcripts, 5.2,0.65, and 0.45 kb in length in MATa cells treated with a-factor. These may correspond to the 5.2-kb, the 0.67-and 0.61-kb and the 0.46kb transcripts we observed. Alternatively, differences in the transcription patterns among strains may account for the differences in the observed sizes of transcripts. Lane 4 shows that the 6.0-, 5.2-, and 0.46-kb transcripts are present at reduced levels, if at all, in the MATala diploid. The position of the high molecular weight band in lane 4 is slightly but distinctly ahead of the position of the 5.2 kb transcript in lanes 1-3. MATa and MATala cells were also treated with afactor, with no e f f e~t .~ A synthetic DNA oligonucleotide homologous to 21 bp of the left-most region of the 5 3 internal domain was used to probe poly(A) RNA from MATa, MATa, and MATala cells and a-factor-treated cells. Hybridization was visible only in the RNA prepared from MATa cells treated with a-factor (lane 5 ) . The hybridizing transcript in lane 5 co-migrates with the a-factor-sensitive 5.2-kb transcript. Background hybridization to rRNA was observed after long exposures of the filter hybridized to the synthetic DNA oligonucleotide?
An Acd 1600-bp restriction fragment from Ty3 (Fig. 1, fragment f ) was used to monitor poly(A) transcripts with similarity to the 5 3 internal domain (Fig. 6B). Hybridization of the same poly(A) RNA samples as used in Fig. 6A with this probe shows a different result. This internal domain probe hybridizes only to transcripts 5.2 and 3.1 kb in length. Although the apparent increase in the level of the 5.2-kb transcript after a-factor treatment is greater, when measured with the internal probe than when measured with the sigma probe, the pattern revealed by the two probes is qualitatively similar and is consistent with the common identity of the 5.2-kb transcript which hybridizes to sigma and the transcript seen here. As was observed with the sigma probe, this transcript is absent in the diploid cells. The levels of the 3.1-kb transcript visualized with these probes is affected neither by a-factor treatment nor by the presence of MATala. Incubation of poly(A) RNA from treated and untreated cells with radiolabeled fragments from the regions flanking Ty3 (Fig. 1, probes a and h) did not show any hybridization: suggesting that the 5.2-kb transcript begins and ends in, or near, the terminal sigma repeats. A radiolabeled URA3 probe was hybridized to the blot shown in Fig. 6B in order to corroborate the loading of equivalent levels of poly(A) RNA in the different lanes. Hybridization, shown in Fig. 6C for the region of the gel containing the URA3 transcripts, was similar in all four lanes.
Termini of the 5 . 2 " Ty3 Transcript-The 5' end of the 5 3 5.2-kb transcript was mapped by reverse transcription extension of a primer with homology to the left end of the internal domain. The DNA oligonucleotide used as the primer in the reverse transcription analysis hybridized to the 5.2-kb transcript, as shown in Fig. 6A, lane 5. Southern blot analysis showed that this oligonucleotide also hybridizes to both composite element-containing fragments in the X2180 background but not to any other fragments? The results of the reverse transcriptase-mediated extension using this oligonucleotide primer are shown in Fig. 7. The sizes of the extended molecules were determined by comparison to an homologous sequence ladder. Extensions of the oligonucleotide primer upstream from the left end of the internal domain; of these, the band corresponding to position 64 was more intense. In the several reverse transcription experiments performed, a higher molecular weight band was sometimes observed. This band was intense and migrated at a position just below the position of the full length DNA, as indicated by the sequence ladder. Although this band seemed to be dependent on poly(A) RNA, it was not consistently present in any sample. The results of the reverse transcription experiment presented here indicate that a major fraction of the transcripts of one or both Ty3 elements have 5' ends in the upstream sigma element. S1 nuclease protection experiments were performed with heteroduplexes of poly(A) RNA from MATa cells treated with a-factor and DNA synthesized in vitro starting with the 5'end-labeled primer described above.3 These experiments gave results which were consistent with those obtained by primer extension mapping. Protected fragments were observed which correspond to the 5' ends of transcripts both within the lefthand sigma element at the position (plus or minus 2 bp) indicated by the most intense band in the reverse transcrip-  Fig. 6 were the sources of RNA in these experiments. The 3' ends of Ty3 transcripts were determined by S1 nuclease protection. A 1.0-kb AccI fragment was labeled in the 3' end, inside the Ty3 internal domain, as described under "Experimental Procedures." This end-labeled DNA was hybridized to poly(A) RNA from different cells, lams 1-4, and no RNA, law 5, as described for Fig. 7, and the duplexes treated with S1 nuclease. Lune 6 contains the control with no S1 added.

A Sigma Composite Element in Yeast
Samples were electrophoresed in a denaturing acrylamide gel and exposed to x-ray film as described. The end-labeled AccI fragment was modified and degraded by the chemical method and electrophoresed on the same gel in order to obtain the sequence shown in the left four lanes. Lanes labeled C and C are specific for nucleotides at those positions. Lanes labeled A and T are purine and pyrimidine cleavages, respectively. The sequence shown on the left was deduced by electrophoresing samples for different lengths of time. The endlabeled Ace1 fragment (lam A ) was also partially digested with Sau3A ( l a n e S ) and completely with XhoI (lam X ) in order to provide markers corresponding to the Sau3A site inside the sigma element and the XhoI site 14 bp within the downstream end of the element. These samples were separated on the same denaturing gel as those described above.
tion experiment and at a position a t least 150 bp upstream. The position of the upstream end point was not mapped. The three start sites indicated by primed reverse transcription against the MATa a-factor-induced cell RNA template are the same as the three sites observed by Van Arsdell et al. (1987) for the 5' ends of a-factor-induced transcripts from an isolated sigma element. The coincidence of at least a set of 5' end points determined by S1 nuclease and reverse transcription experiments suggests that the 5.2-kb transcript is not spliced between the position of the oligonucleotide primer and the 5' end of the transcript mapped by primer extension. The 3' end of the Ty3 transcript was also mapped by protection from S1 nuclease degradation (Fig. 8). In this case, the labeled DNA was the AccI fragment which spans the downstream sigma element. It was labeled a t its 3' end, within the internal domain sequence. One class of protected fragments extended from the AccI site of labeling to the end of the sigma element plus or minus one or two nucleotides. Four smaller classes of protected fragments migrated close together at positions corresponding to 75-100 bp within the end of the sigma element (see Fig. 5A). Although there is slight contamination of the uncut AccI fragment with DNA migrating a t this position, all four "protected" fragments were absent in samples digested in the absence of RNA and were intensified after digestion with nuclease in the presence of RNA, arguing that these termini represent actual heterogeneity in the 3' ends of the transcripts. Thus, at least some Ty3 transcripts begin in the leftward sigma element and end in or beyond the rightward sigma element.

DISCUSSION
Ty3 Homology to Retroviruses and Retrotransposons-We show here that the sigma element occurs as terminal repeats flanking a larger element; we have named this composite element Ty3. Ty3 has the properties of a transposable element and shares additional features with retroviruses and retrotransposons. These features include: conserved terminal short inverted repeats, a tRNA-primer-binding region, a composite structural organization, and regulatory signals which direct synthesis of an almost full-length, polyadenylated transcript. Ty3 is further distinguished by an absolute insertional specificity for the 5' ends of tRNA genes.
Analysis of the Major Ty3 Transcript"Ty3 is transcribed into an almost full-length transcript, approximately 5.2 kb in length. One class of the Ty3 transcripts have been shown to begin in the upstream sigma element and end in, or just beyond, the downstream sigma element. Primer extension and S1 nuclease experiments indicated, in addition to the 5' ends mapped in the downstream half of the left sigma element, 5' ends further upstream which were not mapped. The 5.2-kb transcripts then contain at least the minimum information required in a retrotransposition intermediate.
The sigma element was examined for sequence motifs associated with transcription initiation and transcription termination and polyadenylation. In yeast, DNA sequences homologous to the canonical TATAA/TAA/T sequence (Breathnach and Chambon, 1981) are observed 60-100 nucleotides upstream of transcription initiation sites (Dobson et al., 1982). The 5' ends of the major Ty3 transcripts mapped by primer extension coincide with DNA sequence beginning 118 and 64 bp from the right end of the sigma element in haploid and diploid cells, respectively. Promoter sequences, TATAAA and TATAATATATAA, occur 85 bp upstream of the haploid transcription initiation site and 68 bp upstream of the diploid transcription initiation site, respectively. Therefore, both of the Ty3 putative transcription initiation sites appear to be preceded by an appropriately positioned DNA promoter consensus sequence. The sequence PyAAG also has been found close to the site of transcription initiation of a number of yeast genes (Burke et al., 1983;Dobson et al., 1982;Russell et al., 1983Russell et al., , 1986. The sequence TAAG occurs twice near the Ty3 transcription initiation site observed in haploid cells: once starting a t position -5 and once starting at position +l. It is interesting to compare the positions of these consensus promoter sequences with the positions of the consensus promoter sequences and transcription initiation sites of Tyl. The transcription initiation sites of Tyl occur 95 and 107 bp from the right end of the delta element (Elder et al., 1983). If delta and sigma elements are aligned by the short regions of homology noted by Genbauffe et al. (1984), then the transcription initiation sites of Tyl transcripts coincide almost exactly with the Ty3 transcript initiation site observed in diploid cells, and the TATAAACATATAA sequences 73 bp upstream of the Tyl transcription initiation site in delta coincides with the TATAATATATAA sequence 68 bp upstream of the 5 3 initiation site in sigma.
Two sequences have been proposed to specify transcription termination in yeast: TAG. . .TAGT(TATG). .TTT (Zaret and Sherman, 1982) and T"TTATA (Henikoff and Cohen, 1984). These sequences are found in the downstream end of the sigma element. The sequence TAG.. .TAGTATG . .
TTT occurs between 85 and 35 bp upstream of the first Ty3 polyadenylation site, and the sequence TTTTATA occurs approximately 35 bp upstream of that polyadenylation site. The polyadenylation sites of the Ty3 transcripts are closely spaced and this kind of heterogeneity has been reported for several yeast genes (Zaret and Sherman, 1982;Bennetzen and Hall, 1982). S1 nuclease protection analysis of a restriction fragment labeled in the pSBS12 Ty3 internal domain revealed five positions of heterology between the transcripts and endlabeled fragment. These positions, which are indicated in Fig.  5A, occur in the second sigma element between 20 and 50 nucleotides downstream from the position of the major transcription initiation site in the upstream sigma element and at the end of the second sigma element. Interpretation of this pattern is complicated by the presence of two Ty3 elements in the strain used in these experiments (Figs. 2 and 3). Determination of the sequence of the downstream sigma element in the second Ty3 showed that the sequence from this Ty3 and the pSBS12 Ty3 are identical in this region, but that they are flanked by different 5-bp direct repeat sequences. We conclude, therefore, that at least four of the fragments generated by protection from S1 nuclease degradation correspond to sites of transcript polyadenylation, while the fifth fragment, which has its unlabeled terminus close to the end of the second sigma element, may correspond to the position of polyadenylation of transcripts from either or both Ty3 elements in this strain, or to the point of heterology between the pSBS12 Ty3 end-labeled fragment and transcript from the nonallelic Ty3.
The extent of the linkage of transcription termination and polyadenylation in yeast is unclear (reviewed by Birnstiel et al., 1985). The sequence AATAAA that signals polyadenylation 11-20 nucleotides downstream, frequently within the dinucleotide PYA, which occurs in the genes of metazoans, is not consistently present near the ends of yeast genes. Nonetheless, the sequence, TAAATAAA/G, has been found 28-34 nucleotides upstream of some message polyadenylation sites in yeast and has been proposed to be related to polyadenylation (Bennetzen and Hall, 1982). The sequence TATATAAG, which differs by only one nucleotide from the consensus, occurs approximately 29 bp upstream of the first Ty3 polyadenylation site. Despite the considerable ambiguity regarding signals for transcription termination and processing of the 3' end of the Ty3 transcript, it is clear that these transcripts do fulfill a basic requirement for a retrotransposition intermediate: Ty3 transcripts begin in the left long terminal repeat and terminate within a redundant sequence, downstream of the homologous position in the right long terminal repeat.
Expression of Sigma Elements-Differentia1 expression of the multiple sigma elements is implicit in the pattern of hybridization observed by Northern blot analysis with various probes and with RNA from different cells. There are, depending on the strain, from five to ten transcription products that are easily visualized with sigma-specific DNA probes. The 5.2-, 0.67-, 0.61-, and 0.46-kb transcripts show dramatic increases in levels in MATa cells treated with a-factor. The levels of the 6.0-, 5.2-, and 0.46-kb transcripts are decreased in MATa/a cells. Of the transcripts visualized with the sigma probe, only the 5.2-kb transcript and the 3.1-kb transcript, which is present at relatively constant levels, hybridize to DNA probes specific for the Ty3 internal domain. The number of transcript classes which hybridize to sigma-specific DNA probes contrasts with the larger number of sigma elements. The different flanking sequences of the 34 or so isolated sigma elements should result in distinct transcripts, at least in cases where the transcript is larger than the element itself. Therefore, we tend to support the notion that relatively few independent sigma elements may be actually expressed. If only a subset of this highly conserved family of elements is active, it is possible that the sequences responsible for this differential expression reside outside the sigma element itself. This situation would be at least partially analogous to the regulation of Tyl and Ty2, which is effected, in part, by enhancer and mating-type control sequences within the internal domain (Errede et al., 1985;Roeder et al., 1985).
The reduced levels of some sigma transcripts which we observed in RNA from diploid cells, compared to haploid cells, is consistent with a l a 2 regulation (Strathern et al., 1981;Miller et al., 1985;and reviewed by Herskowitz, 1987), which is observed for Tyl and Ty2 (Elder et al., 1983). Van Arsdell et al. (1987) have attributed the low level of sigma-hybridizing transcripts in untreated MATa strains that they have studied to low levels of a-factor, synthesized by small populations of switched (MATa) cells. These levels of sigma-hybridizing transcripts are apparently higher in the haploid strains which we studied, which leads us to propose that transcription of sigma elements may be under mating-type control. Preliminary experiments show that the sigma transcripts that are present in low levels in MATaja cells, compared to haploid cells, are also reduced in levels in MATa/a cells compared to MATala and MATala cells? This result is consistent with al-a2 regulation. Although blocks of sequence with strict homology to the al-a2 regulation consensus sequence are not found within the sigma element, this is perhaps not surprising, because the comparable sequences in Tyl and Ty2 are found within the internal domain (Errede et al., 1985;Roeder et al., 1985). Thus, the inferred differences in expression among different sigma elements imply a contribution to the regulation of sigma element transcription from sequences outside the element itself.
Ty3 Protein-coding Potential-The RNA genomes of retroviruses code for protein products having functions in viral replication, capsid formation, and integration (reviewed by Varmus, 1983). Tyl and Ty2 elements have two long open reading frames, TYA and TYB, that code for proteins which bind the RNA and have reverse transcriptase activity (Clare and Farabaugh, 1985;Hauber et al., 1985;Warmington et al., 1985;Garfinkel et al., 1985;Mellor et al., 1985;and Adams, et al., 1987). In Ty3 there is an open reading frame which begins 41 bp downstream of the transcription initiation site observed in haploid cells and continues for the 200 bp for which the DNA sequence has been determined. The first ATG translation start codon occurs 76 bp inside the internal domain. It is preceded by ANN and followed immediately by A; both of these are relatively conserved features of eukaryotic translation initiation sites (Kozak, 1983). The sequence CACACA has been found about 10 nucleotides upstream from the ATG start codon of several yeast messages (Dobson et al., 1982) and has been postulated to have a role in translation (Stiles et al., 1981). The sequence ACACCACA is centered 11 bp upstream of the first start codon in Ty3. Although there is no obvious homology between the amino acid sequence encoded in this portion of Ty3 and the TYA or TYB proteins by dotmatrix analysis, projected amino acid sequences encoded by both types of elements are rich in proline in this region (Clare md Farabaugh, 1985). Thus the internal domain of Ty3 may, like those of other retrotransposons, encode proteins which specify replication and integration functions. Conclusions-Several points suggest that sigma and/or Ty3 is capable of transposition: 1) Ty3 has many of the conserved structural features of retroviruses and retrotransposons; 2) the genomic locations of Ty3 elements in five different strains A Sigma Composite Element in Yeast 1423 of yeast are heterogeneous and, therefore, compatible with genomic rearrangements involving transposition; 3) the internal domain and flanking sigma element(s) are transcribed, giving rise to transcripts that are potential RNA intermediates in replication; and 4) the high degree of homology within the sigma element family is consistent with their recent derivation from a limited number of actively transposing elements.
The regulation of Ty3 transcription, and hence transposition, may be more stringent than that of Tyl and Ty2. In fact, the prezygotic pattern of Ty3 expression provides an intriguing parallel to germ-line activation of P elements in Drosophila (Laski et al., 1986). If Ty3 transposition is pheromone-dependent, it is tempting to speculate that the cell has devised a means of avoiding gross genomic variation except when diploidization is imminent.