INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Ciliates are unicellular eukaryotes in which germ line and somatic functions are assumed by two kinds of nuclei coexisting in the same cytoplasm. The diploid micronuclei are transcriptionally silent during vegetative growth; their main function is to provide gametic nuclei upon meiosis. Vegetative transcription takes place in the polygenomic, somatic macronuclei that divide amitotically and are lost soon after meiosis. Following fertilization, both kinds of nuclei differentiate from copies of the diploid zygotic nucleus. Macronuclear development involves extensive rearrangements of the germ line genome: chromosomes are fragmented into shorter, acentric molecules, and tens of thousands of internally eliminated sequences (IESs) are removed from coding and noncoding sequences, in a reproducible and often highly precise manner. In addition, the genome is amplified to the final ploidy level, ~1,000n in Paramecium aurelia complex species (for general reviews, see references 5 and 34; for a review of IESs, see reference 19).

All known Paramecium IESs are short (26 to 882 bp), AT-rich, single-copy elements flanked by 5'-TA-3' direct repeats in the germ line sequence. Excision invariably leaves one of the TA repeats in the rearranged macronuclear sequence, suggesting that the same basic mechanism is used for all IESs. Although the excision mechanism is currently unknown, reaction products and intermediates have been identified in other ciliate species. In Euplotes crassus, the structure of free circles generated by the excision of IESs with very similar characteristics (TA IESs) is best explained by a mechanism involving staggered double-strand breaks at both ends (17). In the more closely related Tetrahymena thermophila, the mapping of transient DNA breaks suggests a transposition-like mechanism, in which a staggered double-strand break at one end of the element is followed by a transesterification step that joins the flanking sequences together, releasing the IES as a linear molecule (37). Tetrahymena IESs, however, differ from those of Paramecium and Euplotes in that they lack invariant TA boundaries.

The analysis of some 44 kb of Paramecium germ line sequences has revealed a high density of IESs, one every 1,300 bp on average, with no significant difference between coding and noncoding sequences (7, 22, 38, 41-43). Although most of the sequences analyzed (35 kb) belong to one multigene family encoding alternative surface antigens, the density of IESs is higher in the remainder of the sequence sample. Thus, the total number of different IESs in the haploid genome is likely to be greater than 50,000. One of the most puzzling questions raised by this massive excision program is how such a large number of different elements can be specifically recognized and excised during macronuclear development. Indeed, no absolutely conserved sequence motif, other than the TA repeats, has been identified within or outside IESs.

A partial answer came from a statistical analysis of the sequence of IES ends, which showed that base composition is not random over at least six nucleotides immediately internal to each of the TA repeats. The sequence of preferred bases forms a degenerate inverted repeat consensus with striking similarity to the ends of the inverted repeats of Tc1/mariner transposons (18). Several large, multicopy elements present in the germ line genome of E. crassus are clearly members of this transposon family; like IESs, they are precisely excised between TA direct repeats during macronuclear development and generate free circles with a similar structure (14). These similarities have led to the hypothesis that present-day IESs are mutated and internally deleted remnants of ancient transposon insertions; the observed consensus could reflect the conservation of minimal cis-acting determinants required to direct excision (19).

The functional significance of the Paramecium IES consensus is supported by the fact that mutations in the TA repeats or at other positions can abolish excision (6, 24, 25). However, the comparison of natural allelic variants of IESs located in noncoding sequences shows that excision boundaries can be displaced over evolutionary time, implying that mutations can also result in the recruitment of novel TA boundaries in adjacent sequences (6). The degenerate consensus may therefore reflect a process of convergent evolution adapting IES ends to mechanistic constraints of the excision machinery. It should be emphasized that only a small fraction of IES ends conform exactly to the consensus, while many perfect matches are found in sequences that are not excised during development. Thus, it is likely that other, unidentified determinants are involved in IES recognition.

Recent evidence suggests that some of these determinants are epigenetic in nature, i.e., that the excision pattern is not entirely determined by the germ line sequence itself (reviews in references 27 and 28). Sequence-specific maternal effects were first revealed by the selective transformation of the vegetative macronucleus with a plasmid containing a 222-bp IES located in the gene for surface antigen G. Following the induction of autogamy (a self-fertilization sexual process), the maternal macronucleus progressively stops replicating its DNA and is eventually lost when vegetative growth resumes; its genetic material, including the transforming plasmid, is not transmitted to sexual progeny. Surprisingly, the presence of the IES in the maternal macronucleus was shown to inhibit specifically the excision of the homologous IES from the germ line genome in the developing macronucleus (8). The efficiency of this transnuclear effect, as measured by the fraction of G-gene copies retaining the IES in the new macronucleus, was further shown to increase with the copy number of the IES in the maternal macronucleus and with the length of IES flanking sequences present on the plasmid. The retention of the IES on the endogenous macronuclear chromosomes has an even stronger effect during the following sexual cycle, resulting in an ever-increasing fraction of IES-retaining copies in subsequent sexual generations. Thus, it is possible to establish stable cell lines in which this particular IES is never excised, although the germ line genome remains entirely wild type.

A similar parental control of developmental IES excision has also been reported in T. thermophila for two different elements (2), raising the prospect that the study of epigenetic regulation will bring valuable insight into basic mechanisms of recognition and excision of ciliate germ line-specific elements in general. We have tested the generality and sequence specificity of the maternal inhibition of excision by analyzing the effects of macronuclear transformation with 13 different Paramecium IESs. We show that five of them can be maintained in the zygotic macronucleus in response to maternal transformation. The effect is completely sequence specific, and each IES appears to inhibit its own excision with a different efficiency relative to maternal copy number.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Cell lines and cultivation. Paramecium tetraurelia 51 and d4.2 are well-characterized homozygous stocks carrying the A⁵¹ and A²⁹ alleles of the gene for surface antigen A, respectively (40). Strain d4.2 was obtained by repeated back-crosses with strain 51 and is therefore largely isogenic; this study confirmed that it carries the same allele of the gene for surface antigen G (G⁵¹) by restriction mapping analysis and the sequencing of over 2,800 bp from both strains, including the central and most variable region of the coding sequence. Cells were grown in a wheat grass powder (Pines International Co., Lawrence, Kans.) infusion medium bacterized the day before use with Klebsiella pneumoniae and supplemented with 0.8 mg of -sitosterol (Merck, Darmstadt, Germany) per liter, at 27°C. Basic methods of cell culture have been described previously (39).

Recombinant  phages and plasmids. The 51Gmic and 51Amic phages, containing the micronuclear G⁵¹ and A⁵¹ genes, were isolated from the library of micronuclear DNA from strain 51 constructed by C. J. Steele et al. (41) in the GEM11 vector, by using IES 51G4404 (29) and a 210-bp ScaI fragment from the central part of the coding sequence (31) as probes, respectively. The 51Gmic insert was entirely sequenced (12,326 bp). The G⁵¹ coding sequence was identified by comparison with known alleles of the G gene from Paramecium primaurelia (33); it is 88% identical with G¹⁵⁶. IES excision boundaries were confirmed by sequencing the relevant parts of the macronuclear version from the macronuclear phage 51Gmac, which was cloned from a library of total DNA from strain d4.2 in the EMBL3 vector (29). Plasmids p51A-712mic and p51A-712mac are SalI-SpeI fragments of 51Amic and the macronuclear clone SA1 (9), respectively, subcloned into vector pUC18. Plasmid p29A6649mic was obtained by PCR amplification of micronuclear DNA from strain d4.2 with primers 6435A and 6649-3' (see below).

Microinjections. Young cells (less than eight divisions after autogamy) were injected in mineral water (Volvic Co., Volvic, France) containing 0.2% bovine serum albumin, under an oil film (Nujol), while being visualized with a phase-contrast inverted microscope (Axiovert 35M; Zeiss). CsCl-purified  phage DNA was cut within the vector arms with SmaI, leaving 3,380 and 728 bp of  sequences on either side of the insert for 51Gmic and 51Amic and 3,362 and 728 bp for 51Gmac. Column-purified (Qiagen) plasmid DNA was cut within the vector sequence with XmnI. Restriction digests were extracted with phenol, filtered on a 0.22-µm-pore-size Ultrafree-MC filter (Millipore), and precipitated with ethanol. Approximately 5 pl of a 5-mg/ml solution in water was delivered into the macronucleus. Injected DNA molecules replicate autonomously at stable copy numbers in the vegetative macronucleus, forming linear monomers and multimers to which telomeres are added (11, 12, 15, 16).

Dot blot analyses. For each sample, 100 cells were pipetted from depression slide cultures and transferred to 400 µl of 0.4 N NaOH-50 mM EDTA. The lysates were incubated for 30 min at 68°C and loaded on a Hybond N⁺ membrane (Amersham, Little Chalfont, United Kingdom) with a dot blot apparatus. Genomic DNA samples were treated in the same way. The membrane was kept wet with 0.4 N NaOH for 15 min, washed in 2× SSC (1× SSC is 150 mM NaCl-15 mM sodium citrate), and subsequently treated as a Southern blot.

Copy number quantification. The copy numbers of phage inserts in transformed clones were determined after hybridization of dot blots with a 3,359-bp SmaI-SacI fragment from the small arm of phage GEM11, which is linked to the inserts in all SmaI-digested phage DNAs, with a Fuji Bas 1000 imager. Measured signals were normalized by rehybridizing the same dot blots with an oligonucleotide specific for macronuclear telomeric repeats, as previously described (8). Probes specific for IESs 51A2591 and 51G4404 were used to check relative variations of IES copy numbers. The normalized  probe signals were then translated for all clones into copy numbers per haploid genome (phg), by using a G-gene probe to compare the signals obtained for clones transformed with G-gene phages to those obtained for reference samples (genomic DNA from clones transformed with G-gene plasmids) loaded on the same dot blots, for which copy numbers had previously been determined (8). The reference value of one copy phg was arbitrarily defined as the average copy number of the G gene in uninjected clones. Plasmid copy numbers were determined on Southern blots of preautogamous DNA samples digested with PvuII after hybridization with a 322-bp PvuII fragment from pUC18, by using the same reference samples.

Autogamy. Autogamy was induced by starving the cells after they had reached the appropriate clonal age (30 vegetative divisions) and assessed by staining with a 15:1 (vol/vol) mix of carmine red (0.5% in 45% acetic acid) and fast green (1% in ethanol). For karyonidal analyses, cells were isolated from depressions showing 100% autogamous cells. After the first cellular division, one of the two karyonides was isolated and cultivated. For mass autogamies, about 20 autogamous cells (i.e., 40 karyonidal clones) were transferred to bacterized medium and grown collectively.

Genomic DNA extraction. Cultures of exponentially growing cells (400 ml) at 10³ cells/ml were centrifuged. After being washed in 10 mM Tris (pH 7.0), the cell pellets were resuspended in 0.5 volume of the same buffer and quickly added to 3 volumes of lysis solution (0.44 M EDTA [pH 9.0]-1% sodium dodecyl sulfate [SDS]-0.5% N-laurylsarcosine [Sigma]-1 mg of proteinase K [Merck] per ml) at 55°C. The lysates were incubated at 55°C overnight, gently extracted once with phenol, and dialyzed twice against TE (10 mM Tris-HCl-1 mM EDTA, pH 8.0) containing 25% ethanol and once against TE.

Southern blotting. DNA restriction and electrophoresis were carried out according to standard procedures (36). DNA was transferred from agarose gels to Hybond N⁺ membranes (Amersham) in 0.4 N NaOH after depurination in 0.25 N HCl. Hybridization was carried out according to the procedure described in reference 3 in 7% SDS-0.5 M sodium phosphate-1% bovine serum albumin-1 mM EDTA (pH 7.2) at 63°C. Probes were labeled with a random priming kit (Boehringer, Mannheim, Germany) to a specific activity of 3 × 10⁹ cpm/µg. Membranes were washed for 30 min in 0.2× SSC-0.5% SDS at 60°C prior to autoradiography or image plate quantification.

PCR amplification and sequencing. Macronuclear DNA was amplified with primers located about 25 bp away from each deletion junction, from total genomic DNA of caryonidal clones, or from  phage DNA. PCRs were carried out in capped 0.5-ml Sigma polypropylene tubes with reaction mixtures (25 µl) containing 100 ng of total genomic DNA or 1 ng of cloned DNA, 1× PCR buffer (Promega), 80 µM (each) deoxynucleoside triphosphate, 2 µM (each) oligonucleotide, and 0.8 U of Tfl polymerase. Amplifications were run for 20 cycles (92°C, 1 min; 63°C, 1 min 15 s; 74°C, 30 s) in a Perkin-Elmer Cetus thermocycler. Micronuclear DNA was amplified from total genomic DNA with one primer located within an IES and the other in the macronuclear sequence. PCRs were run for 30 cycles under the same conditions as for macronuclear amplification, except that the annealing temperature was 55°C and the extension time was increased to 1 min. PCR products were cloned into plasmid pCRScript (Stratagene) and sequenced with the Sequenase sequencing kit (U.S. Biochemical Corp.). At least two clones arising from PCR products of two different karyonidal clones were sequenced to check for PCR-induced substitutions. The sequenced portions of the micronuclear A²⁹ allele were amplified with primer pairs 1835A/2591-3' and 6435A/6649-3'. The same primer pairs were used to check the micronuclear sequence in some postautogamous karyonidal clones.

Nucleotide sequence accession number. The complete sequence of the 51Gmic insert was submitted to the EMBL database (AJ010441).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Experimental design. We chose to test IESs from two surface antigen genes, one of which contains the previously studied 222-bp IES (8). The A and G genes are related, nonessential members of a family of at least 11 genes showing mutually exclusive expression, so that selection against macronuclear genomes containing nonfunctional forms of these genes is not expected to bias the results. Recombinant  phages containing the germ line versions of these genes were isolated from a library of micronuclear DNA from strain 51 (Fig. 1). The 12.4-kb insert from 51Amic contains all but the last 500 bp of the A-gene coding sequence, as well as about 2.9 kb of 5' flanking sequences. Nine IESs (51A-712 through 51A6649) have previously been identified, seven of them located within the coding sequence and two in the upstream region (41). Sequencing of the 12.3-kb insert from 51Gmic showed that it contains the complete micronuclear G gene and revealed the presence of six IESs, four of which are studied here (51G1413, 51G1832, 51G2832, and 51G4404, the previously studied 222-bp IES; IESs are named according to the positions of their insertion sites in the macronuclear sequence, relative to the translation start).

View larger version (14K): [in this window] [in a new window]   FIG. 1.   Maps of the macronuclear A51 gene (51Amac) and of the inserts of phages 51Amic, 51Gmic, and 51Gmac. Coding sequences are represented by thick arrows in the macronuclear sequences. IESs are shown as black boxes in the micronuclear sequences. Double-headed arrows indicate paralogous IES pairs (IESs occurring at homologous positions in the two genes). Open boxes at the ends of the phage inserts symbolize the vector arms. The positions and lengths of probes a, b, c, d, and e are shown. Restriction sites: M, MnlI; P, PstI (only sites relevant to Fig. 2 and 3 are shown).

Gene-specific maternal effects of A- and G-gene micronuclear sequences. DNA from each of the  phages was microinjected into the macronucleus of vegetative cells from strain 51. Injected cells were cultured separately, and the copy numbers of phage inserts were precisely measured after about 20 divisions by a dot blot procedure (see Materials and Methods). Selected clones maintaining between 0.1 and 10 copies phg were then allowed to undergo autogamy. To examine the new macronuclear genome of sexual progeny, about 20 autogamous cells for each clone were refed and grown collectively for DNA extraction. Postautogamous DNA samples were analyzed on a Southern blot after restriction with MnlI. There are many sites for this enzyme in the A- and G-gene coding sequences but none within IESs; the retention of any IES can thus be detected by a size increase of the corresponding MnlI fragment.

View larger version (65K): [in this window] [in a new window]   FIG. 2.   Southern blot analysis of the macronuclear genome of sexual progeny of clones transformed with 51Gmac, 51Gmic, and 51Amic. Total DNA from mass-autogamy samples was digested with MnlI. Because the mic/mac ploidy ratio is ~1/250, only macronuclear DNA is visible. For each sample, the copy number of the phage insert in the maternal macronucleus is indicated by a vertical bar above the lane. Lane W1 is control DNA from uninjected cells. Lanes W2 and W3 contain the same DNA as in lane W1, mixed with a small amount of 51Amic and 51Gmic DNA, respectively, prior to digestion with MnlI. Symbols on the right indicate the positions of IES-retaining and wild-type macronuclear fragments. The same blot was hybridized successively with probes a, b, c, and d (Fig. 1).

View larger version (49K): [in this window] [in a new window]   FIG. 3.   Southern blot analysis of the excision of IESs 51G4404 and 51G2832 in sexual progeny of clones transformed with 51Gmic. Samples G1 to G6 and W1 to W3 are the same as in Fig. 2. Each of the mass-autogamy samples G2 to G6 is followed by a series of karyonidal samples representing individual clones isolated from the same mass autogamy. Vertical bars above the lanes indicate the copy numbers of phage inserts in maternal macronuclei. W0 is a control mass autogamy from an uninjected cell; 0a and 0b are karyonidal clones from this mass autogamy. All samples were digested with PstI. The same blot was hybridized successively with probes d and e (see Fig. 1). In addition to the PstI fragment containing IES 51G2832, probe e cross-hybridizes with the central PstI fragment from the G gene (upper band in lower panel), due to the repeated structure of this region of the coding sequence.

Karyonidal variability of maternal inhibition of excision. Many of the postautogamous DNA samples, which were prepared from mass autogamies of transformed clones, contained both rearranged and IES-retaining molecules. This heterogeneity could mean that IES excision was inhibited only on a fraction of the ~1,000 copies of the genome in each new macronucleus or that only a fraction of the ~40 new macronuclei analyzed in each sample were affected. To address this question, a number of autogamous cells were isolated from the same populations that were used for the mass-autogamy samples. In each autogamous cell, two new macronuclei develop independently from mitotic products of the zygotic nucleus; during the first vegetative division, they segregate without dividing to the two daughter cells, which are called karyonides. To prepare postautogamous DNA samples corresponding to single events of macronuclear development, only one of the two karyonides from each isolated cell was cultured.

Sequence specificity of the maternal effects of single IESs. To test the specificity of the effect among the three A-gene IESs showing inhibition, plasmids containing only one of them were constructed. IESs 51A-712 and 51A6649 were chosen to make the test stringent, as the absence of any effect on 51A2591, which was the most readily maintained IES in the 51Amic experiment, would be more conclusive. Plasmid p51A-712mic contains a 1.5-kb segment of the germ line sequence upstream of the A gene, including IES 51A-712 but not IES 51A-10; a negative control is provided by plasmid p51A-712mac, which contains the homologous segment from the rearranged macronuclear sequence (see map in Fig. 4). To study IES 51A6649, we used plasmid p29A6649mic, which contains a segment of the micronuclear A²⁹ allele including the IES as well as 301 bp of flanking sequences (242 bp upstream, which also includes IES 51A6435, and 59 bp downstream; see map in Fig. 5). Within this segment, the A²⁹ allele differs from A⁵¹ by only a single substitution within the IES (see below). The maternal effects of these plasmids were tested as in the  phage experiment, except that plasmid copy numbers in the macronucleus of transformed clones were determined by Southern blotting of preautogamous DNA samples (see Materials and Methods).

View larger version (33K): [in this window] [in a new window]   FIG. 4.   Southern blot analysis of the excision of IESs 51A-712, 51A6649, and 51A2591 in sexual progeny of clones transformed with plasmids p51A-712mac, p51A-712mic, and p29A6649mic (see map in Fig. 5). Plasmid copy numbers in the maternal macronucleus are indicated by the vertical bars above the lanes. Mass-autogamy samples were digested with MnlI and BstBI. The same blot was successively hybridized with probes a, b (Fig. 1), and f (see map; M, MnlI; B, BstBI).

View larger version (24K): [in this window] [in a new window]   FIG. 5.   Summary of the results showing the sequence specificity of the inhibition of IES excision by maternal sequences. The maps show the phage and plasmid inserts used. IESs are shown as black boxes. Paralogous IES pairs are indicated by double-headed arrows. The names of the five IESs showing inhibition are boxed. In the lower panel, the plus sign indicates that excision inhibition was observed. nd, not determined.

Quantitative analysis of excision inhibition. As noted above, the efficiency of excision inhibition, relative to the copy number of phage insert in the maternal macronucleus, differs for each IES. To quantify these differences, the relative intensities of the rearranged and IES-retaining versions of each fragment were determined, by using only mass-autogamy samples to average the karyonidal variability. Since the probes used contain only macronuclear sequences, relative intensities give an exact measure of the fraction of rearranged copies in new macronuclei. This fraction was then plotted as a function of the copy number of phage insert in the maternal macronucleus. The fairly regular aspect of the plots obtained for the two G-gene IESs (Fig. 6A) indicates that karyonidal variability is reasonably well averaged in pools of ~40 karyonidal clones and illustrates clearly the quantitative difference in the effects of the 51Gmic insert on the two IESs. Excision of 51G4404 is reduced by 50% with only 0.4 copies phg in the maternal macronucleus, while about 2 copies phg are necessary to achieve a similar inhibition of the excision of 51G2832. The plots obtained for the three A-gene IESs also show significant differences in the efficiency of inhibition by the 51Amic insert (Fig. 6B). Maternal copy numbers of 51Gmic and 51Amic can be directly compared because they were both measured by probing with a  vector sequence which remains linked to the inserts in injected phage DNA. The blocking of 51A2591 is slightly less efficient than that of 51G4404, while blocking efficiencies of 51A6649 and 51G2832 are comparable. Only a weak effect is observed for 51A-712 (~30% inhibition at the highest copy number tested, 5.9 copies phg).

View larger version (23K): [in this window] [in a new window]   FIG. 6.   Quantitative analysis of the excision of different IESs as a function of the copy number of specific sequences in the maternal macronucleus. "% excision in new macronucleus" is the fraction of rearranged copies in the new macronuclear genome, as measured in mass-autogamy samples. (A) Excision of IESs 51G4404 and 51G2832 after transformation of the maternal macronucleus with 51Gmac or 51Gmic. (B) Excision of IESs 51A2591, 51A6649, and 51A-712 after transformation of the maternal macronucleus with 51Amic. (C and D) Filled symbols show the quantification of the maternal effects of IESs present on the endogenous macronuclear chromosomes and are superimposed on the same graphs as in panels A and B, respectively, to compare the inhibition efficiencies of IES-retaining chromosomes with those of the phage inserts.

Macronuclear maintained IESs are not copied from maternal sequences. To test whether the maintenance of IESs in the macronuclear genome could be due to the repair of a gap left in the genomic sequence after constitutive excision of the IES, by a mechanism involving the copying of a homologous template from the maternal macronucleus, advantage was taken of the fact that two different alleles of the A gene, A⁵¹ and A²⁹, can be distinguished by a number of substitutions. Partial sequencing of the micronuclear A²⁹ gene from strain d4.2 (see Materials and Methods) showed that the two alleles are ~99% identical in the regions around IESs 51A2591 and 51A6649. The macronucleus of vegetative cells from strain d4.2 was transformed with 51Amic DNA. After induction of autogamy, the new macronuclear genome was analyzed as described for the experiment with strain 51. IES 29A2591 was maintained, with an efficiency similar to that of 51A2591 (data not shown). The macronuclear maintained IES was PCR amplified and cloned (see Materials and Methods). Its sequence was shown to be identical with the A²⁹ germ line sequence, which differs from the A⁵¹ germ line sequence by one substitution and a 12-bp insertion (Fig. 7A). Thus, it can be concluded that the maintained IES originated from the d4.2 germ line genome and had not been converted by the 51Amic sequence which had caused its retention. This extends an earlier result obtained for 51G4404, showing that the engineering of a novel restriction site within the injected IES does not alter the sequence of the IES recovered from the macronucleus of postautogamous progeny (8).

View larger version (19K): [in this window] [in a new window]   FIG. 7.   Internal IES-like segments excised from the macronuclear maintained copies of 51A2591 (A) and 51A6649 (B). Only the sequences of internal segments are shown. Boldface TAs are the excision boundaries of internal segments. Boxed TAs are the excision boundaries of the whole IESs. The sequences of homologous segments of the A29 allele are also shown; differences with the A51 allele are underlined. The 40-bp segment in 29A2591 is not excised in the macronuclear maintained version.

Short IES-like segments are excised from some of the maintained IESs. Macronuclear maintained copies of IES 51A2591 were also cloned and sequenced after PCR amplification, by using postautogamous samples from the first experiment (progeny of strain 51 cells transformed with 51Amic). Quite unexpectedly, the maintained sequence proved to differ from the germ line sequence by the deletion of an internal 28-bp segment (Fig. 7A). The same size difference could be detected in postautogamous DNA samples by comparison with 51Amic DNA on appropriate Southern blots, showing that all macronuclear maintained copies of IES 51A2591 carry the 28-bp deletion (data not shown). To check that the particular stock of strain 51 used did not carry the deletion in its germ line genome, micronuclear DNA was PCR amplified from total DNA samples of karyonidal clones containing the shortened version in their macronuclear genome, by using micronuclear sequence-specific primers located in other IESs. The micronuclear sequence was found to be identical with the published sequence present in 51Amic. Since micronuclei and macronuclei develop from mitotic products of the same zygotic nucleus, the 28-bp deletion must occur during macronuclear development. We also checked that the 28-bp segment was not deleted in the 51Amic molecules replicating in the maternal macronucleus, by using preautogamous DNA samples from transformed clones.

View larger version (31K): [in this window] [in a new window]   FIG. 8.   Sequences of the ends of tested IESs. The left (L) and right (R) end sequences of each IES are aligned with the degenerate consensus established from the general IES sample (18, 19). All sequences are written 5' to 3'. Subscripts in the general consensus indicate the fraction of sequences in the general sample showing the preferred nucleotide at each position. Sequences marked with asterisks are the ends of internal IES-like segments within 51A2591 (51A28) and 51A6649 (51A29).

Characteristics of IESs showing inhibition. Only five of the tested IESs could be maintained in the new macronucleus, following maternal transformation with micronuclear sequences. They share no obvious feature that could clearly distinguish them from the other eight. Four of them are among the largest (222 to 370 bp), while none of the short ones (28 to 52 bp), including the short IES-like segments located within 51A2591 and 51A6649, could be blocked. However, maternal inhibition is not determined only by size, because only one of several 77-bp IESs shows the effect, and the largest IES in the sample (51A4578, 882 bp) does not. There appears to be nothing special about the location of IESs showing inhibition, as they are interspersed with other IESs in coding sequences; one of them (51A-712) is located in noncoding sequences. Since the only significant sequence feature identified in IESs is the degenerate consensus of end sequences, the ends of tested IESs were examined (Fig. 8). Internal IES-like segments, which were always excised from the maintained versions of 51A2591 and 51A6649, were included in the analysis. In IESs not showing inhibition, the degree of conservation of each of the eight positions of the consensus is similar to that observed in the whole sample of IESs from which the consensus was established. In IESs showing inhibition, the fourth position shows a T in 8 sequences of 10, whereas an A is preferred in 70% of sequences in the general sample. Although the number of sequences is limited, this deviation is interesting because all known IESs beginning with the sequence 5'-TATT-3' on both sides have now been shown to be subject to maternal inhibition. However, this sequence feature is not necessary for the effect, since both ends of 51A6649 show the more standard sequence 5'-TACA-3'.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

This study has shown that the developmental excision of five different Paramecium IESs can be inhibited by transformation of the maternal macronucleus with specific micronuclear sequences. Although 12 of the 13 IESs tested belong to the A and G genes, one of those showing inhibition (51A-712) is located ~700 bp upstream of the A-gene transcription start site. This region is apparently noncoding and is clearly upstream of the A-gene promoter, since normal transcription regulation does not require more than ~270 bp of upstream sequences (21, 23). Thus, the effect is not limited to IESs located within surface antigen genes and may concern a sizeable fraction of the estimated ~50,000 different IESs in the genome. The case of paralogous IESs 51G4404 and 51A4404 is interesting because only the former could be inhibited. IESs that have a common ancestor are unlikely to use radically different excision mechanisms, suggesting that the same mechanism can be differently affected depending on the particular sequence. Paralogous IESs often show limited sequence conservation at their very ends (22, 38); the fact that one of the TATT ends of 51G4404 is changed to TATG in 51A4404 can be viewed as supporting the significance of the deviation from the consensus noted above. However, this single criterion is not sufficient to explain the different behaviors of all tested IESs.

The search for possible mechanisms should take into account all basic features of this transnuclear epigenetic phenomenon. The first and most important is sequence specificity. The micronuclear A gene can inhibit the excision of A-gene IESs in the developing macronucleus but has no effect on G-gene IESs and vice versa. Maternal inhibition is caused by the IESs themselves, as shown by macronuclear controls and the effect of 51G4404 without flanking sequences (8). Complete IES specificity was here demonstrated for the effects of 51A-712, 29A6649, and 51G4404; in an independent study, transformation with 51A2591 alone was shown to block the homologous IES (24). Taken together, these results strongly suggest that each IES inhibits the excision of its germ line copy independently from the others. The homology requirement can nevertheless tolerate small differences between the sequence present in the maternal macronucleus and the germ line sequence, such as the introduction of a restriction site within 51G4404 (8), or the allelic differences between A⁵¹ and A²⁹ (one substitution and a 12-bp insertion at the most, in 29A2591). This allowed us to show that macronuclear maintained IESs originate from the germ line genome and are not copied from maternal sequences. Short internal deletions in the maintained versions of IESs 51A2591 and 51A6649 (28 and 29 bp of 370) do not seem to impair the effect either, as these shortened copies were themselves able to inhibit excision of the germ line copies in the following sexual generation. In contrast, removing a 147-bp internal segment from the 222-bp IES 51G4404 has been shown to abolish the effect (8).

A second important feature is copy number dependence. Because of the large variability observed among individual karyonidal clones derived from a single transformed clone, the quantitative relationship is most clearly evidenced by the analysis of mass-autogamy samples, which measures the average effect in ~40 different karyonidal clones for each point. Karyonidal variability could in principle reflect the unequal distribution of IES copies during amitotic divisions of the transformed macronucleus, leading to variable copy numbers at the time of induction of autogamy. However, this is unlikely to account for the extreme variability observed in some cases. For instance, the absence of any effect on 51G2832 in some karyonidal clones from the G4 population cannot be explained by the absence of the 51Gmic insert in the maternal macronuclei of these particular cells during autogamy, because 51G4404 was fully maintained in these clones. Similarly, one karyonide derived from a clone transformed with 51Amic was found to retain 51A6649 but not 51A2591, although the latter is retained on a higher fraction of copies in the mass-autogamy sample (data not shown). Thus, excision inhibition appears to be a stochastic process in single developing macronuclei; the mass-autogamy correlation indicates that its probability is determined by the copy number of the IES in the maternal macronucleus.

Thirdly, for a given maternal copy number, the probability of inhibition varies greatly between different IESs. It should be noted that when the effect is weak (low maternal copy numbers or unefficient IESs), karyonidal variability may not be correctly averaged in pools of ~40 karyonidal clones, which adds to measurement errors inherent in the quantification of faint hybridization signals. Despite these technical limitations, the consistency of the mass-autogamy plots shows that each IES is inhibited with a characteristic efficiency, which does not correlate with any obvious feature such as IES size, position within the gene, or end sequence. The efficiency of 51G4404 was previously shown to increase with the length of flanking sequences present on the maternal molecule (8). It was here shown to be eightfold higher in the context of the 51Gmic insert than on the largest plasmid tested (851 bp) and very close to that of the chromosome-borne IES (50% inhibition with ~0.3 copies phg). This suggests that the efficiencies determined for other IESs, all present on phage inserts of similar sizes (12.3 to 12.4 kb), are also close to their maximum values. Our attempt to determine directly the efficiencies of chromosome-borne IESs after a second autogamy revealed that their effects are more variable than those of phage inserts, and on average only slightly stronger. In spite of this greater variability, the results confirm the relative efficiencies of the different IESs, indicating that they reflect intrinsic properties of the elements. Since the least efficient IES was only partially inhibited at the highest copy number tested, transformation with higher copy numbers might have resulted in the inhibition of more IESs. Thus, it may be misleading to classify IESs into two categories according to whether inhibition was observed.

Finally, the observation that internal IES-like segments are excised from the macronuclear maintained versions of 51A2591 and 51A6649 reveals that at least parts of the inhibited IESs are still accessible to the excision machinery. Internal excision events are not linked to the process of maternal inhibition, as they do not occur in the A²⁹ allele of 51A2591, nor in the other three IESs showing inhibition. Moreover, a point mutation at the left end of 51A2591 has been reported to impair excision, allowing Mayer et al. (24) to show that, in the absence of any maternal inhibition, the internal 28-bp segment is still deleted in the macronuclear version of the IES. Similarly, a point mutation in one of the TA boundaries of 51A6649 reveals that the excision of the internal 29-bp segment occurs independently from the maternal inhibition (25). It has been suggested that the 51A2591 28-bp segment could have originated from the secondary insertion of some mobile element into a preexisting IES or ancestral transposon (24). Our observation that allelic differences between 51A2591 and 29A2591 are all located within this segment does not support a more recent origin. Since the homologous segment from the A²⁹ allele does not appear to be capable of independent excision, the data are perhaps more consistent with the idea that functional IES consensus sequences were created by random mutations in the A⁵¹ allele. However, it remains intriguing that the only allelic difference between 51A6649 and 29A6649 is also located within the 29-bp IES-like segment.

Two different types of models have been proposed to account for the sequence specificity of maternal inhibition of IES excision in Paramecium and Tetrahymena spp. (2, 8). In the first type, IES copies in the maternal macronucleus sequester a sequence-specific protein factor that is required for excision in the developing macronucleus. Although formally a possibility, the model would require an unreasonably large number of different factors if, as suggested by the present work, a significant fraction of IESs in the genome can inhibit their own excision with a high specificity. Furthermore, the factor would have to bind the IES itself, since flanking macronuclear sequences have no inhibitory effect; this is difficult to reconcile with the higher efficiency of molecules containing longer flanking sequences.

The second type of model proposes that IES copies are exported from the maternal macronucleus to the developing macronucleus, where they would affect excision by pairing with homologous sequences of the germ line genome. The model easily accounts for the observed sequence specificity; furthermore, the dependence of inhibition efficiency on the length of IES flanking sequences could be explained by differences in pairing efficiency. Although there is no direct evidence, a transfer of nucleic acids between the two nuclei is also likely to be involved in another type of homology-dependent maternal effects on macronuclear development in Paramecium spp., affecting the level of amplification of macronucleus-destined sequences (27, 28, 30). RNA would seem to be a better candidate than DNA for a messenger molecule able to leave one nucleus and enter another. That IES-containing plasmids and phage inserts can be transcribed without the need for a polymerase II promoter is suggested by a recent study of homology-dependent gene silencing in vegetative cells, in which aberrantly sized transcripts could be detected following transformation of the macronucleus with plasmids containing promoterless coding sequences (35). Furthermore, the maternal macronucleus is known to remain fully active in transcription throughout the development of the new macronucleus (1).

In the simplest version of this model, pairing of a maternal transcript with the germ line sequence could directly inhibit IES excision by steric hindrance. This is unlikely because excision of the internal IES-like segments in 51A2591 and 51A6649 would then be expected to be inhibited in the same way. Alternatively, a transient pairing interaction could induce some epigenetic modification of the germ line IES, which in turn would block excision. A highly localized modification, such as the methylation or other chemical modification of specific nucleotides or dinucleotides, could affect the excision of each IES to a different extent, depending on the location of modified sites relative to IES boundaries. This could explain the absence of any inhibitory effect for some IESs and for internal IES-like segments. For a given IES, the efficiency of inhibition would also depend on the concentration of effective transcripts, determined by the copy number of the maternal IES. The fact that 51Amic can occasionally inhibit 51A6649 without affecting the more sensitive 51A2591 suggests that effective transcripts do not necessarily cover the whole phage insert. Maternal transcription level and RNA stability might thus be additional factors that determine the characteristic inhibition efficiency of each IES.

The targeting of epigenetic modifications to specific genomic sequences through pairing interactions has been implicated in many homology-dependent gene silencing phenomena. These effects have been revealed by transformation in a wide range of eukaryotes (13, 20), including P. tetraurelia (35). Both transcriptional and posttranscriptional cases of silencing are frequently associated with epigenetic modifications of the genes; it has recently been proposed that all types of homology-dependent gene silencing in plants are triggered by a common RNA-based mechanism (44). The transnuclear nature of the Paramecium maternal effect is not unique. A dominant silencing effect has been observed in multinucleated cells of Neurospora crassa (4); in several multicellular organisms, the propagation of homology-dependent silencing signals is thought to involve the transfer of RNA molecules across cellular boundaries (10, 32). The novel aspect of the homology-dependent epigenetic effect described here is that it affects developmental genome rearrangements rather than gene expression.