Two enhancers and one silencer located in the introns of regA control somatic cell differentiation in Volvox carteri

Abstract

The regA gene plays a central role in germ-soma differentiation of Volvox carteri by suppressing all reproductive functions in somatic cells. Here we show that the minimal promoter of regA consists of only 42 bp immediately upstream of the transcription start site, and that it contains no discernible regulatory elements. However, introns 3 and 5 are both required for regA expression in somatic cells, and intron 7 is essential for silencing regA in gonidia (asexual reproductive cells). A regA gene lacking intron 7 rescues the normal phenotype of mutant somatic cells, but also results in gonidia that reproduce only weakly and soon die out. The same phenotype is observed when a regA gene containing intron 7 is placed under control of a constitutive promoter, suggesting that the silencing activity of intron 7 is promoter specific. Intron 7 is unusual in that it contains a potential ORF that is in frame with exons 7 and 8, and some transcripts are produced in which intron 7 is retained. However, a regulatory role for the intron 7 translation product can be ruled out, because a construct in which intron 7 must be translated, and one in which it cannot be translated, both result in wild-type development of both cell types. Furthermore, intron 7 is unable to act in trans to silence regA, but is able to exert its normal effect when placed in a different location within the gene. Therefore, it appears that intron 7 functions in gonidia as a classical cell-type-specific and promoter-specific enhancer, of the inhibitory type that is often referred to as a silencer.

Volvox carteri is a multicellular green alga with a complete division of labor between somatic cells and asexual reproductive cells (gonidia) (Fig. 1A). Gonidia are immotile and specialized for reproduction (Starr 1969, 1970). In contrast, somatic cells are specialized for motility and phototaxis, but have no reproductive potential. Once differentiated, they never divide, and they eventually undergo senescence and programmed cell death (Pommerville and Kochert 1981, 1982).

Figure 1.

Mutant phenotype, map, and promoter analysis of the regA gene. (A) An adult wild-type Volvox carteri spheroid containing a surface monolayer of tiny, terminally differentiated somatic cells (arrow) and 16 juvenile spheroids that are about ready to hatch out and become free-swimming individuals; ∼20× magnification. Each juvenile contains a surface monolayer of small somatic cells and ∼16 large asexual reproductive cells called gonidia. (B) A representative spheroid of strain 153–68, the non-revertible Reg/Nit (regA⁻/nitA⁻) mutant used as the DNA recipient in all cotransformation studies reported here; ∼20× magnification. This spheroid is about the same age as the wild-type spheroid shown to its left, but its somatic cells (arrow) have begun to grow and redifferentiate rather than remaining small and terminally differentiated; they will all become fully functional gonidia. (C) The recombinant plasmid, pVcRegA1, containing regA in pBlueScript SK(−). The vector portion is shown in grey. Also shown are cloning sites NotI and XhoI at the 5′ and 3′ ends, respectively, and the EcoRI site used for cloning of promoter deletions. Filled segments symbolize exons, open segments the promoter, introns, 3′ UTR and the 3′ region downstream of regA. (D) Map of the regA locus. Open boxes symbolize noncoding exons, solid boxes represent coding exons, and lines between exons represent introns. The transcription start point (+1), start codon (ATG), stop codon (TGA), and poly(A) signal (TGTAA) are positioned according to scale. The 70-nt sequence upstream of +1 and the extent of six deletions (ΔPr1 to ΔPr6), each with the number of deleted nucleotides in parentheses, are shown below the map. Symbols + or − at the bottom, right indicate whether a construct with that deletion did or did not provide cotransformation rescue of strain 153–68, restoring it to the wild-type morphology depicted in A. The denominators of the quotients to the right of these symbols indicate the total number of Nit⁺ transformants recovered in each case, and the numerators indicate the number of these Nit⁺ strains that were rescued with respect to the Reg phenotype. The start of the minimal promoter defined by these deletion constructs is indicated by a broken line at T42. Putative TATA and initiator (Inr) elements within this promoter are in bold letters

Genetic analyses identified the regA gene as a master gene that plays a key role in establishing the germ-soma dichotomy by suppressing all aspects of reproductive activity in somatic cells. In the first published discussion of V. carteri developmental genetics, it was reported that whereas wild-type somatic cells lack any reproductive potential whatsoever, the fertile somatic cells of what is now called a regA, or Reg, mutant (Fig. 1B) are able to participate in both asexual and sexual reproduction (Starr 1970). Such studies led to the working hypothesis that regA encodes a negative regulator that is active in somatic cells — but not in gonidia — and that represses genes required for reproductive development (Kirk 1988). Evidence consistent with that hypothesis came when regA was cloned, characterized, and found to encode a protein with features of a transcriptional repressor that appears early in somatic cell development, is present in somatic cell nuclei through most of the life cycle, but can never be detected (at either the RNA or the protein level) in gonidia (Kirk et al. 1999). Further support for such a hypothesis came when 16 genes that are expressed abundantly in both gonidia and Reg somatic cells, but not in wild-type somatic cells (and therefore are putative targets of regA regulation), were shown to be nuclear genes encoding chloroplast proteins. This study led to the newer, more specific hypothesis that regA blocks reproductive activity in somatic cells by preventing their growth (Meissner et al. 1999).

All such prior studies of regA have led to the rather obvious question that we address here: What is the molecular basis for the cell-type-specific expression of regA?

Previously, we established that a cloned genomic fragment that provides efficient phenotypic rescue of a non-revertible Reg mutant has a rather unusual structure in that it contains only ∼70 bp of DNA upstream of the transcription start site, has four introns within the 5′ UTR (as well as three in the coding region), and a rather long 3′ UTR (Kirk et al. 1999). We now have performed systematic deletions of portions of the 3′ and 5′ regions, and each of the introns, to determine which of these regions may contain cis-regulatory elements that are required for normal cell-type-specific expression of regA. Here we show that introns 3 and 5 contain enhancers that are required for expression of regA in somatic cells, whereas intron 7 contains a silencer that is required to prevent regA expression in gonidia.

Results

The minimal regA promoter is only 42 bp long

Biolistic cobombardment with a selectable marker was used to test the capacity of various regA deletion constructs to achieve phenotypic rescue of a Reg strain (153–68; Fig. 1B), which carries non-revertible mutations of both the regA and nitA (nitrate reductase-encoding) loci. The selectable marker was the nitA gene carried on the plasmid pVcNR15, which is able to restore Nit⁻ recipients to Nit⁺ (able to grow on nitrate as a sole source of nitrogen). Transforming plasmids integrate into the Volvox genome at random locations, frequently in the form of tandem repeats (Babinger et al. 2001). The frequency with which Nit⁺ transformants recovered in such an experiment exhibit cotransformation of an unselected marker such as regA (and hence transformation rescue of the mutant phenotype) typically ranges between 30% and 80% (Schiedlmeier et al. 1994; Kirk et al. 1999).

The starting point for making a set of six nested 5′-deletion mutants was plasmid pVcRegA1, which contains the native regA locus (Fig. 1C,D), and which was used as the positive control for all cotransformation experiments reported in this work. Four of the 5′-deletion clones, ΔPr1 through ΔPr4, all of which retained at least 42 bp upstream of the transcription start point (+1), yielded an average of 35% cotransformants with wild-type morphology (Fig. 1D). In contrast, no rescue of the Reg phenotype was achieved with clones bearing deletions that extended any further downstream (ΔPr5 and ΔPr6; Fig. 1D). This defines the 42-bp sequence upstream of +1 as the minimal promoter of the regA gene. As indicated in Figure 1D, this sequence includes a plausible TATA box (TAATTGA) starting at position −28, and an initiator element (CACTCAT) overlapping the transcription start site at an A, but it contains no obvious regulatory elements. Therefore, other noncoding regions of the regA region were systematically screened for regulatory sequences.

There are no regulatory elements downstream of the 3′ UTR

A set of five clones with nested deletions covering the 2-kb region immediately downstream of the poly(A) site had no effect on the capacity of the cloned regA gene to achieve mutant rescue (data not shown). Therefore, we concluded that the cis-regulatory elements controlling regA expression must be located within the transcription unit.

Enhancer elements are present in introns 3 and 5

It has been shown that a variety of volvocalean genes, such as nitA of V. carteri, and CAH1, ODA6, and RBCS2 of the closely related unicellular alga Chlamydomonas reinhardtii, carry regulatory elements in their introns (Gruber et al. 1996; Kang and Mitchell 1998; Lumbreras et al. 1998; Kucho et al. 1999). Therefore, we screened the seven introns of regA for such elements. Individual introns were deleted (in frame) from plasmid pVcRegA1 by use of overlap extension PCR (Horton et al. 1989). All seven deletion constructs (Fig. 2A) were tested for their ability to rescue the Reg mutant phenotype. Constructs lacking introns 1, 2, 4, or 6 (ΔI₁, ΔI₂, ΔI₄, and ΔI₆), all rescued the Reg phenotype of strain 153–68 with an efficiency in the 60% range (Fig. 2A), and PCR analysis revealed that in each tested case the transgene had been stably incorporated into the transformant genome (data not shown). In contrast, although PCR analysis revealed that more than half of the Nit⁺ transformants recovered after cobombardment with the regA constructs lacking introns 3 and 5 (ΔI₃ or ΔI₅) had stably integrated the regA transgene, no amelioration of the Reg phenotype was observed with either of these derivatives (Fig. 2A). This result suggested that deletion of either intron 3 or intron 5 prevents regA expression. All attempts to detect transcripts from stably integrated copies of the ΔI₃ or the ΔI₅ transgenes by RT–PCR failed. It was therefore concluded that introns 3 and 5 must both contain positive regulatory elements (enhancers) that are required for regA gene transcription.

Figure 2.

Intron deletions and modifications that were tested for their ability to achieve cotransformation rescue. (A) Maps of constructs with deletions of intron 1 to 7 (ΔI₁–ΔI₇), which led to the fusion of neighboring exons as shown. Symbols and quotients to the right of these maps are as defined in Fig. 1D. The term fruitless denotes spheroids containing wild-type somatic cells and nonviable gonidia (see text). (B) The role of a 937-bp EcoRV fragment of intron 5 in marker rescue was tested by deletion (ΔRV), inversion (RVinv), and transposition into the singular XhoI site of intron 3 in a ΔI₅ background (I₃RVΔI₅). Symbols and quotients to the right of these maps are as defined in Fig. 1D.

This concept was further supported by functional analysis of three constructs bearing three intron 5 variants that are depicted in Figure 2B. If a central 937-bp EcoRV fragment was deleted from the 1816-bp intron 5, no regA mutant rescue was observed among 21 Nit⁺ transformants. However, if this same EcoRV fragment was reintegrated in reverse orientation, mutant rescue was observed at the same level as with the wild-type construct (Fig. 2B). Furthermore, a construct from which intron 5 had been deleted regained the capacity to rescue the Reg phenotype with high efficiency when the intron 5 EcoRV fragment was inserted into intron 3 (Fig. 2B). This sort of ability to function in both orientations and in different locations is a classical feature of enhancer elements (Banerji et al. 1981). However, database searches using BLAST (Altschul et al. 1990), PLACE (Higo et al. 1999), and TRRD (Kolchanov et al. 2000) failed to identify any previously defined enhancer motif in the 937-bp fragment that had been inverted, or in the rest of intron 5, or in intron 3.

A construct lacking intron 7 causes a lethal fruitless phenotype

Nine of 19 Nit⁺ transformants recovered following cobombardment of strain 153–68 with the regA derivative lacking intron 7 (ΔI₇; Fig. 2A), resulted in a completely novel phenotype that we have termed fruitless. The somatic cells of the fruitless strains exhibit all of the features of phenotypic rescue that are seen following transformation with the control regA plasmid. That is, at a developmental age when the somatic cells of the recipient Reg strain have resorbed their eyespots and flagella and have begun to grow rapidly and redifferentiate as reproductive cells (Fig. 1B), the somatic cells of fruitless strains exhibit no tendency to grow or redifferentiate. They remain tiny, with typical wild-type somatic features such as eyespots and flagella (Fig. 3A,B). The novel aspect of the fruitless strains was that their gonidia grew and reproduced extremely slowly at first, but then invariably died out after no more than two or three asexual generations. At a developmental stage by which the gonidia should have completed their growth, developed typical gonidial morphology, and divided, most gonidia of a fruitless strain were small, dense, misshapen, and clearly moribund (Fig. 3A). We were unable to recover enough cells from these mutants before they died to perform any molecular analyses.

Figure 3.

The fruitless phenotype and evidence that it involves expression of regA in gonidia. (A) A fruitless transformant; ∼20× magnification. Three generations are included. The outermost sphere contains the somatic cells of the first generation. Most of the spheroids within it are second generation individuals, but their differences in size attest to the fact that they developed asynchronously (compare the synchronous development depicted in Fig. 1A). Within these second generation spheroids, only one gonidium has grown enough to divide and produce a third generation offspring (arrow, bottom, left). The rest of the gonidia are small, pycnotic, and moribund. (B) Three somatic cells of the first generation individual seen in A viewed at ∼1000× magnification. Eyespots, a typical marker of differentiated somatic cells, are indicated by arrowheads. At the same age, somatic cells of the DNA-recipient strain, 153–68, would have resorbed their eyespots, grown manyfold, and would have divided to produce progeny spheroids. (C) Construct P_TubRegA that was used to obtain constitutive regA expression. It contains the coding region of regA, minus intron 5, fused to the promoter (P_Tub) and 3′ UTR of β2-tubulin. Parts of the pBlueScript SK(−) vector are symbolized by broken lines and tubulin gene elements by grey rectangles. Transcription start site (+1), start codon (ATG), stop codon (TGA), and poly(A) signal (TGTAA) are shown. Exons (solid boxes) and introns (lines) are numbered as in Fig. 1. A notch indicates the 281-bp region that is deleted in the strain 153–68 regA mutant allele (Kirk et al. 1999). Arrows indicate the regions complementary to the oligonucleotides used as primers for RT–PCR as discussed under D. The expected sizes of RT–PCR products are indicated. (D) RT–PCR analysis of P_TubRegA transformants that had a phenotype such as shown in A and B. Cell-cycle-arrested gonidia and somatic cells were isolated from eight independent transformants and used for RNA isolation and RT–PCR. Oligonucleotide pregA served as primer for first-strand synthesis. In subsequent PCR, the primer pair pregB/pregC were used to amplify a transformant-specific 1622-nt transcript extending from the tubulin 5′ UTR to regA exon 7 (left two lanes). The absence of intron 6 rules out a potential artifact arising from inadvertent amplification of a genomic DNA contaminant. The pregLE6/pregRE6 primer pair amplifies products (right two lanes) templated by transcripts of both the endogenous mutant allele (469 bp) and the transgene (750 bp) in the somatic cell sample, but only the product templated by the transgene transcript in the gonidial sample.

Constitutive expression of regA also causes a fruitless phenotype

The foregoing result led to the working hypothesis that the fruitless phenotype was a consequence of the fact that in the absence of intron 7 regA is expressed at some level in gonidia, in which its product, RegA, functions as it normally does in somatic cells; that is to say, it represses chloroplast biogenesis, which leads eventually to a failure of growth and reproduction (Meissner et al. 1999). This hypothesis led to the prediction that any alternative means of eliciting regA expression in gonidia should result in the fruitless phenotype—even in the presence of intron 7. To test this prediction, we developed a construct, P_TubRegA (Fig. 3C), in which intron 5 of the regA gene was deleted (to abolish its usual expression pattern; see above) and then the rest of the regA coding region was fused to a constitutive promoter and the 3′ UTR of the V. carteri β2-tubulin gene, which is known to be expressed in both gonidia and somatic cells (Mages et al. 1995).

When P_TubRegA was tested in mutant-rescue experiments, 8 of the 23 Nit⁺ transformants recovered exhibited the fruitless phenotype described above with wild-type somatic cells and small, dense gonidia that soon ceased dividing. Gonidia and somatic cells harvested from the eight transformants with fruitless morphology yielded sufficient RNA for RT–PCR tests. As shown in Figure 3D, both the gonidial and the somatic samples yielded a 1622-bp RT–PCR product with the pregB/pregC primer pair, and a 750-bp product with the pregLE6/pregRE6 primer pair. These are the sizes of the RT–PCR products that should have been produced in the presence of a transcript of the P_TubRegA transgene (see Fig. 3C). In contrast, only the somatic cell sample generated a product of the size (469 bp) that should have been produced in the presence of a transcript of the endogenous regA gene of the host strain, 153–68, which has a 281-bp deletion in the region that is amplified by use of the pregLE6/pregRE6 primer pair (Fig. 3C,D). Thus, these data provide clear evidence that the regA transgene driven by the tubulin promoter is transcribed in gonidia, whereas the native regA gene under the control of its own promoter is not. Furthermore, these results also are fully consistent with the twin hypotheses that (1) expression of regA in gonidia extinguishes the reproductive potentials of these cells, and (2) that intron 7 normally prevents regA from being expressed in gonidia.

A spliced transcript retaining intron 7 is present at a low level

The above observations provided strong motivation for exploring possible mechanisms by which intron 7 might act in the gonidia to silence regA. In this context, we were intrigued by our earlier observation that the 1194-bp intron 7 is extremely unusual, in that it constitutes an ORF that is fully in frame with the ORFs of the two flanking exons (Kirk et al. 1999), raising the question of whether intron 7 is ever retained in a mature transcript. This possibility was reinforced by the knowledge that intron 7 of regA (like three other volvocalean introns; Harper and Mages 1988; Fabry et al., 1993; Kropat et al. 1995) has a GC, instead of the canonical GT, at the donor splice site, which could lead to inefficient splicing. However, our earlier Northern-blot analysis had revealed the presence of only one kind of processed transcript, a 7-kb component lacking intron 7 (Kirk et al. 1999). Therefore, we decided to use a more sensitive RT–PCR method to determine whether we could detect any processed transcript that retained intron 7. To that end, we designed two primers, pregE and pregF, that are complementary to the ends of exons 6 and 8 (Fig. 4A). RT–PCR with these primers was predicted to yield a 310-bp product with an RNA template from which both exon 6 and exon 7 had been spliced out, a 2845-bp product with an unprocessed nuclear RNA (or a DNA) template, and a 1504-bp product with an RNA molecule from which intron 6 had been completely removed, but intron 7 had been retained (Fig. 4A).

Figure 4.

RT–PCR analysis of intron 7 splicing. (A) Maps of the exon 6 to exon 8 region of regA, showing the sizes of the RT–PCR products to be expected by use of pregD as a primer for first strand synthesis and the primer pair pregE/pregF for subsequent PCR if the template RNA lacks intron 7 (top), or retains it (bottom). (B) RT–PCR analysis of regA transcripts from cells of wild-type and mutant (153–68) spheroids. (St) Size standard; (lane 1) wild-type embryos in early cleavage; (lane 2) 1-day-old wild-type gonidia; (lane 3) 1-day-old wild-type somatic cells; and (lane 4) 2-day-old wild-type somatic cells; and (lane 5) 2-day-old Reg mutant (153–68) spheroids. The 310-bp band is the major processed transcript that lacks intron 7, whereas the 1504-bp band contains the complete intron 7 sequence, as was subsequently confirmed by sequence analysis.

The results of this experiment were unambiguous (Fig. 4B): The major product templated by RNA from somatic cells of both wild-type and Reg spheroids is the 310-bp product. However, lesser amounts of the 1504-bp product are also present in each of these samples, indicating the presence of a minor RNA component that lacks intron 6, but retains intron 7. However, it is important to note that no product of either size was templated by RNA from either gonidia or embryos.

Translation of intron 7 has no discernible effect on germ-soma differentiation

The demonstration that a processed regA transcript that retains intron 7 is present as a minor RNA component in wild-type somatic cells immediately raised the question of whether translation of the intron 7 ORF at some level is required to obtain normal cell-type-specific expression of regA (either expression of regA in somatic cells or—more likely—silencing of regA in gonidia).

This question was addressed by testing the ability of five different types of regA variants produced by in vitro mutagenesis (Fig. 5A) to rescue the Reg phenotype as follows: (1) In CCI₇, the unconventional GC donor splice site was replaced with a CC, making it impossible to remove intron 7 by splicing. (2) In GCI₇ 5′Stop, intron 7 was left in a spliceable form, but a stop codon was introduced near its 5′ end, making it possible to generate two kinds of mRNAs, but precluding translation of most of intron 7. (3) In CCI₇ 5′Stop, the two types of mutations described above were combined, precluding the translation of both intron 7 and exon 8. (4) In GCI₇ 3′Stop, intron 7 was left in a spliceable form, but a stop codon was introduced near its 3′ end, making it possible to generate two kinds of processed transcripts, and to translate most of intron 7 when it was present. (5) In CCI₇ 3′Stop, the loss of the intron 7 donor splice site was combined with the stop codon near the end of intron 7, assuring that most of intron 7 would be—but exon 8 would not be—translated.

Figure 5.

Variants of intron 7 tested in transformation–rescue experiments. (A) Nucleotide and peptide sequences near the 5′ and 3′ ends of wild-type (wt) and modified versions of intron 7. Exon sequences are shown in lower case and intron sequences in upper case letters. Nucleotide numbering is according to Fig. 1D. (CCI₇) A single-nucleotide was changed (GC to CC) at the 5′ splice site of intron 7 to prevent its splicing. (GCI₇ 5′Stop) A single-nucleotide was changed to introduce a stop codon near the 5′ end of intron 7. (CCI₇ 5′Stop) An inactivated 5′ splice site as in CCI₇ was combined with a stop codon as in GCI₇ 5′Stop. (GCI₇ 3′Stop) A stop-codon oligonucleotide was inserted into an EcoRV site near the 3′ end of intron 7. (CCI₇ 3′Stop) An inactivated 5′ splice site as in CCI₇ was combined with a stop codon as in GCI₇ 3′Stop. Donor and recipient splice motifs are underlined (wt), substitutions and additions are shown in bold lettering. Symbols and quotients to the right are as defined in Fig. 1D. (B) Map of the exon 6 to 8 region of regA with the G to C substitution that inactivates the donor site of intron 7 (CCI₇). To exclude PCR amplification of the transcript from endogenous regA, primer pregG corresponding to a position within the region that has been deleted in the mutant allele (notch) was used for RT–PCR. The expected size of the amplification product (including intron 7) is shown. (C) RT–PCR analysis of RNA extracted from cells of four independent CCI₇ transformants in either the first or second day of the 2-day asexual life cycle. (St) Size standard; (lane 1) 1-day-old gonidia of strain CCI₇ #5; (lane 2) 1-day-old somatic cells of strain CCI₇ #5; (lane 3) 2-day-old somatic cells of strain CCI₇ #5; (lane 4) 2-day-old somatic cells of strain CCI₇ #6; (lane 5) 2-day-old somatic cells of strain CCI₇ #7; (lane 6) 2-day-old somatic cells of strain CCI₇ #8;. (lane 7) control, 2-day-old spheroids of strain 153–68 (the DNA recipient). (D) Construct P_TubI₇ used to obtain expression of regA intron 7. The recombinant plasmid contains the β2-tubulin promoter (P_Tub) and 3′ UTR as described for P_TubRegA (Fig. 3C) fused to the sequence of intron 7, in frame, to allow potential translation. The CCI₇ splice site variant (A) was used to inhibit potential splicing and to allow accumulation of intron 7 RNA. Nucleotide and peptide sequences near the 5′ and 3′ ends of intron 7 are shown below. Start and stop codons from β2-tubulin are shown in small letters, intron sequences are shown as in A.

To our great surprise, the results (Fig. 5A) clearly indicated that the translation of intron 7 is neither necessary for, nor detrimental to, the differentiation of either cell type. Spheroids with normal, wild-type somatic cells and gonidia were recovered with similarly high frequencies when the Reg recipient was transformed with constructs carrying either the wild-type regA gene (pVcRegA1), the version of the gene in which intron 7 cannot be removed by splicing, and must be translated (CCI₇), or two versions of the gene in which intron 7 could be removed by splicing, but could not be fully translated if it were not removed (GCI₇ 5′Stop and GCI₇ 3′Stop). In striking contrast, no rescue of the Reg phenotype was obtained with either of the two constructs that contained a stop codon that precluded translation of exon 8 (CCI₇ 5′Stop and CCI₇ 3′Stop).

RT–PCR analysis was used to confirm that construct CCI₇ did, in fact, generate a mature RNA that contained all of intron 7 (Fig. 5B,C). The observation that this construct rescued the Reg phenotype about as well as the wild-type gene leads to the astonishing conclusion that two versions of the RegA polypeptide that differ by the presence of nearly 400 residues—near the middle of the polypeptide—have indistinguishable biological activities!

Intron 7 RNA has no detectable trans-regulatory effect on regA gene expression

Having been unable to detect any regulatory role for the peptide product of intron 7, we wished to determine whether the RNA product of this intron might be involved in regA regulation. To that end, we used our standard cotransformation system to test the effects of a construct, P_TubI₇, in which intron 7 had been placed under the control of the β2-tubulin promoter (Fig. 5D). Eleven of the twenty-nine Nit⁺ transformants that were recovered when Reg strain 153–68 was cobombarded with the nitA, ΔI₇ and P_TubI₇ constructs exhibited the fruitless phenotype that had first been observed in spheroids transformed with ΔI₇ in the absence of P_TubI₇. No transformants were recovered in which gonidial development was modified in any discernible way that could be attributed to the presence of the P_TubI₇ construct. Thus, we found no evidence that the intron 7 transcript could act in trans to silence regA in gonidia. Nor did we find any evidence that it could affect somatic cell development in any way.

Intron 7 acts in a position-independent manner

To determine whether intron 7, like intron 5, can exert its effect from more than one position (thereby meeting one of the criteria of a classical enhancer; Banerji et al. 1981), we used cotransformation to test the behavior of two constructs in which intron 7 had been placed where intron 6 is normally located (Fig. 6A). In one of these (GCI₇ΔI₆; Fig. 6A), the GC splice site of intron 7 was left intact, so that the intron could be spliced out in transformants, whereas in the other construct (CCI₇ΔI₆) this splice site had been changed to CC, and thereby inactivated.

Figure 6.

Intron rearrangements tested in transformation–rescue experiments. (A) Maps of the wild-type regA gene (wt), a variant in which intron 7 with an intact splice site replaces intron 6 (GCI₇ΔI₆), and a similar variant in which the intron 7 splice site has been inactivated (CCI₇ΔI₆). Note that the ORF of intron 7 is in frame with exon 6 in both of these constructs. Symbols and quotients to the right are as defined in Fig. 1D. (B) Double and triple intron deletions tested in transformation rescue experiments. Symbols and quotients to the right are as defined in Fig. 1D. No fruitless transformants were detected, despite the absence of intron 7 in each construct.

The GCI₇ΔI₆ construct provided phenotypic rescue of the Reg recipient that was nearly as efficient as that obtained with the wild-type gene; 14 of 24 Nit⁺ transformants were converted to wild type, and no fruitless cotransformants were recovered. This clearly indicated that intron 7 was able to silence regA in gonidia just as well when it was in the intron 6 location as when it was in its usual location.

However, it is noteworthy that the CCI₇ΔI₆ construct, in which splicing of intron 7 had been precluded, provided no phenotypic rescue (Fig. 6A), even though its ORF was in frame with the exons on either side, just as it is in its usual location. Although the RegA protein apparently is fully functional when the peptide encoded by intron 7 is incorporated downstream of the peptide encoded by exon 7, it is nonfunctional when the order of these two peptide regions is reversed (Fig. 6A).

Introns 3 and 5 are required for expression of regA in gonidia, just as they are in somatic cells

To assess possible epistatic relationships among the three cis-regulatory elements that are present in regA introns, we next used cotransformation to evaluate double and triple deletion constructs lacking intron 7 plus one or both of the introns that are required for somatic cell expression of regA. As anticipated, none of these constructs lacking introns 3 and/or 5 was able to rescue the phenotype of the Reg somatic cells (Fig. 6B). However, of greater importance was the fact that no gonidia with the fruitless phenotype could be detected in any of the 78 Nit⁺ transformants that were recovered in these experiments, despite careful microscopic examination. The significance of this will be discussed below.

Discussion

Differential gene expression is an essential prerequisite for cellular differentiation in all multicellular organisms. The regA gene has been shown to play a critical role in V. carteri germ-soma differentiation, by acting as a master regulatory gene that normally is expressed exclusively in somatic cells, and that represses functions required for cellular growth and reproduction (Kirk et al. 1999; Meissner et al. 1999). This immediately raised a question similar to that raised in the case of any other important regulatory locus, how is the master regulator itself regulated in a cell-type-specific manner? Here we show that at least a partial answer to this question is provided by three cell-type-specific cis-regulatory elements that reside within three of the seven regA introns.

Two considerations had led us to suspect a priori that important cis-regulatory elements might reside within the regA introns. The first was our observation that the amount of upstream DNA that had to be included in a regA genomic clone to achieve high-efficiency phenotypic rescue of Reg mutants was surprisingly small. We had shown initially that a genomic clone that included only 70 bp of DNA upstream of the transcription start point was capable of rescuing a Reg mutant via cotransformation (Kirk et al. 1999). Here we have shown that no decrease in transformation efficiency is observed when this upstream region is trimmed down to a 42-bp sequence that contains no obvious regulatory elements (Fig. 1D). The second consideration was prior evidence that introns in both V. carteri and the closely related unicell, C. reinhardtii, contain regulatory information. Introns within the C. reinhardtii genes encoding dynein (Kang and Mitchell 1998) and a rubisco small subunit (Lumbreras et al. 1998) contain classical type enhancer elements that exert strong positive effects on transcription. Introns within the Volvox nitA gene, on the other hand, appear to exert their positive influences on nitA expression at the post-transcriptional level (Gruber et al. 1996), as do many introns of higher-plant genes (Koziel et al. 1996).

We found that introns 3 and 5 of regA must both be present in order to accumulate a detectable amount of regA transcript, or to achieve phenotypic rescue of a Reg mutant. We also found that a central 937-bp fragment of intron 5 that is essential for phenotypic rescue functions equally well in either orientation or in either of two locations (Fig. 2B). Such orientation- and position-independent functioning is, of course, a hallmark of classical enhancer elements (Banerji et al. 1981), and there is presently no reason to suspect that introns 3 and 5 are anything other than classical enhancers.

By classical enhancers, we mean cis-regulatory elements lying outside of the basal promoter region of a locus that bind specific trans-acting factors (transcriptional regulators), and thereby act to initiate, increase, decrease, or prevent transcription of the locus. The use of the term enhancers for such elements is unfortunate, of course, because although the term was deliberately introduced to indicate that such elements act only to stimulate transcription of loci that would otherwise be transcribed at some lower level (Banerji et al. 1981), it is now well established that a single such element can exert both qualitative and quantitative, and both positive and negative effects on transcription of the locus of which it is a part (for review, see Novina and Roy 1996; Arnone and Davidson 1997). As one example of the mechanistic ambiguity of the term enhancer, it has now been established that the stripe 3 and 7 enhancer of the Drosophila even-skipped (eve) locus functions exclusively in a negative manner (i.e., by repression) to establish the 3rd and 7th stripes of eve transcription in the fly embryo. When fused to a reporter gene, this enhancer silences (represses) the gene in the regions anterior to stripe 3, posterior to stripe 7, and between the two stripes, but permits expression within both stripes (Small et al. 1996). A different illustrative example is a well-studied enhancer of sea urchins that acts to silence the endo 16 gene in mesomeres under one set of conditions, but to activate transcription of endo 16 in mesomeres under other conditions (Yuh et al 1998).

The novel fruitless phenotype that was observed when Reg mutants were transformed with a regA gene lacking intron 7 (Fig. 3A,B) indicated that this intron is required to silence regA in gonidia. This interpretation was reinforced when the identical phenotype was observed in Reg mutants that had been transformed with a regA gene under the control of a constitutive promoter that caused it to be transcribed in gonidia (Fig. 3C,D). Certain unusual structural features of intron 7 suggested initially that it might exert its effects by some rather novel mechanism. Like a few other Chlamydomonas and Volvox introns that have been described, regA intron 7 has a non-canonical splice junction, which we suspected might cause a splicing defect. This suspicion was confirmed when we detected a minor fraction of the processed regA transcript that retained intron 7 (Fig. 4B). We had shown earlier that intron 7 consisted entirely of an ORF that was perfectly in frame with the exons on either side, and that encoded a polypeptide very similar in amino acid composition to the rest of the RegA protein (Kirk et al. 1999). Thus, we speculated that the translation product of intron 7 might play some important role in regulating regA expression. However, this possibility can now be ruled out by the observation reported here that both cell types differentiate in a completely normal, wild-type manner both in transformants carrying a regA gene in which intron 7 cannot be translated (GCI₇ 5′ Stop; Fig. 5A), and in transformants in which intron 7 must be translated (CCI₇; Fig. 5A).

These results deserve additional consideration. RT–PCR analysis of four independent transformants that had been rescued with the CCI₇ construct confirmed that (as anticipated) only one form of the regA transcript can be detected in the somatic cells of these strains, the transcript containing all of intron 7 (Fig. 5C). But all of these strains were indistinguishable in phenotype from wild-type V. carteri as are the transformants carrying the GCI₇ 5′Stop construct, in which a stop codon precludes translation of most of intron 7. These astonishing results indicate that the RegA protein must be able to function in the same manner in either of two versions that differ in length by the 398 amino acid residues that are encoded by intron 7 and that are located nearly in the middle of the RegA polypeptide. We are unaware of any precedent for such an observation.

These studies appeared to rule out the possibility that intron 7 exerts its silencing effect on regA in gonidia via its polypeptide product. Therefore, we next reconsidered the possibility that it might do so via its RNA product. Two of our earlier observations made it seem to us that this possibility was rather unlikely. The first was that we had never been able to detect any regA transcripts of any kind in wild-type gonidia (see Figs. 4B and 5C, for example). The second was that an experiment reported here established that intron 7 is unable to silence regA in gonidia when it is present in a transgene that is being transcribed under the control of a constitutive promoter (Fig. 3C,D). That is to say, it appears to function in a promoter-specific manner, which is usually taken as a hallmark of transcriptional, rather than post-transcriptional, regulation. Nevertheless, we performed one more experiment to test the hypothesis that intron 7 exerts its silencing effect in gonidia at a post-transcriptional level, via its RNA product. We cotransformed Reg spheroids with a regA construct lacking intron 7, together with a construct in which intron 7 was expressed (by itself) under the control of the constitutive β2-tubulin promoter (Fig. 5D). We found no evidence that when intron 7 was transcribed in this way — in trans to an intron 7-less gene — it had any effect on development of either cell type.

The fact that intron 7 could not exert its effect in trans established that intron 7 is (by definition) a cis-regulatory element and strongly suggested that it exerts its effect at the transcriptional level. Transcriptional control that is mediated by a cis-regulatory element in a promoter-specific manner is a defining feature of classical enhancers, so we next performed an experiment to determine whether intron 7 met another defining feature of an enhancer: position-independent function. We found that it did. Intron 7 is as able to silence regA in gonidia when it is located in the position where intron 6 is normally located as it is when it is in its native location (Fig. 6A). Therefore, we are led to conclude that the unusual structural features of intron 7 are purely coincidental, and that its cell-type-specific regulatory activity is due to the fact that it contains a classical enhancer element, but of the inhibitory type sometimes called a silencer.

However, the transformations we performed with double and triple intron deletion constructs led to an important additional insight about regulation of regA in gonidia. Although a regA construct lacking only intron 7 is expressed in gonidia and causes all transformants to exhibit the fruitless gonidial phenotype, no such effect is seen if either intron 3 or intron 5 has also been deleted (Fig. 6B). This clearly indicates that both intron 3 and intron 5 must be required for transcription of regA in gonidia, just as they are in somatic cells. This, in turn, indicates that all of the trans-acting factors required for the functioning of the intron 3 and intron 5 enhancers must be present in gonidia, just as they are in somatic cells. Under normal conditions, however, intron 7 obviously overrides the enhancer activity of introns 3 and 5 in gonidia. A model consistent with all of these observations is presented in Figure 7.

Figure 7.

A minimal model for cell-type-specific regulation of regA transcription by the cis-regulatory elements located in introns 3, 5, and 7. In somatic cells, introns 3 and 5 bind complementary trans-acting factors, which then cooperate to activate the regA promoter. Introns 3 and 5 bind these trans-acting factors in gondia also, but in gonidia, intron 7 binds one or more gonidium-specific trans-acting factor(s), which enables it to block the effects of introns 3 and 5 and prevent activation of the regA promoter.

Identification of cis-regulatory elements regulating cell-type-specific transcription of a gene of interest is, of course, only the first step in defining the complete regulatory network at play in the developmental system of interest. The obvious next step is to identify the cell-type-specific transcription factors for which the newly identified cis-regulatory elements serve as targets. Efforts to identify these trans-acting factors are presently being initiated, using now-standard techniques.

Materials and methods

Construction of regA vectors

The complementing genomic regA clone pVcRegA1 (15.3 kb; GenBank accession no. AF106962) has been described previously (Kirk et al. 1999). Deletion and mutagenesis constructs were generated, as described below, with suitable subclones of pVcRegA1. All cloning sites and the orientation of inserts were analyzed by sequencing with an ABI Prism automated sequencer. PCR products used for cloning were verified by full-length sequencing, to exclude PCR errors. All oligonucleotides were obtained from Metabion (Martinsried, Germany).

Constructs ΔPr1 to ΔPr6 were generated with PCR by use of six oligonucleotides complementary to the upstream transcription initiation site for the regA sequence. Each contained an artificial NotI restriction site on the 5′ end. Oligomer pE2RI (5′-ACTTCAGGAATTCTTCTTG-3′), overlapping the EcoRI site in exon 2, was used as the second primer in PCR. Intron deletions were constructed by overlap extension PCR (Horton et al. 1989). That method was also used to generate a nonfunctional intron 7 splice site (CCI₇) and for insertion of a stop codon near the 5′ splice site (GCI₇ 5′Stop and CCI₇ 5′Stop). Downstream stop codons were inserted into intron 7 by use of complementary oligomers pI7Stop1 (5′-TGATAGTAACTGAC CTGATAG-3′) and pI7Stop2 (5′-CTATCAGGTCAGTTACTA TCA-3′) that were ligated to the EcoRV restriction site at position 8220 to generate plasmids GCI₇ 3′Stop and CCI₇ 3′Stop.

The recombinant plasmid P_TubRegA contains, within the pBlueScript SK(−) vector, the constitutive promoter and 3′ UTR of the V. carteri β2-tubulin gene (Mages et al. 1995; GenBank accession no. L24547) fused to a regA coding region from which intron 5 had been deleted. This construct was prepared as follows: First, a fragment that extends from 338 bp upstream to 119 bp downstream of the β2-tubulin transcription start site (and that therefore terminates with the tubulin initiation codon) was amplified by PCR, by use of appropriate primers containing artificial NotI and NcoI restriction sites. Next, this fragment was fused in frame to the PCR-generated regA coding region using the NcoI site. Then a 355-bp PCR product representing the β2-tubulin 3′ UTR, including its polyadenylation signal, was attached downstream of the regA coding region, exploiting a unique XhoI site that is located just downstream of the regA stop codon. The same tubulin gene elements were used to generate recombinant plasmid P_TubI₇ with the splicing-deficient version of regA intron 7.

Other constructs were made by use of appropriate restriction sites of pVcRegA1 and suitable subclones.

Volvox strains, cultivation conditions, and nuclear transformation

V. carteri cultures were maintained in standard Volvox medium (SVM) at 30°C under a 16:8-h light:dark cycle (Kirk and Kirk 1983, 1985). SVM lacking reduced nitrogen (SVMN) was used to select for clones able to reduce nitrate after genetic transformation. Nuclear transformation of Volvox was performed according to Schiedlmeier et al. (1994) with modifications. Cells of strain 153–68 (which carries nonrevertible mutations in both the nitA and regA loci; Kirk et al. 1999) were cobombarded with plasmid pVcNR15 (which carries the selectable marker, nitA; Gruber et al. 1996) and various regA constructs that are described elsewhere in the text. Eqimolar quantities of each construct were used. The cotransformation efficiency observed previously with this system has varied between 30% and 80% (Schiedlmeier et al. 1994; Kirk et al. 1999).

Analysis of Volvox genomic DNA and RT–PCR analysis of the regA transcript

Volvox genomic DNA was prepared as described (Miller and Kirk 1999). Integration of regA constructs after transformation was tested for by PCR, by use of oligonuclotides pregLE6 (5′-TTGGGAGCCGACCTGCCAT-3′) and pregRE6 (5′-AGCGTG ACCTCGCATGTATC-3′) as primers. This PCR system generates a 750-bp product from the wild-type regA allele and a 469-bp product from the regA allele present in strain 153–68, which carries a 281-bp deletion. Production, electrophoretic separation, and visualization of PCR products were as described (Kirk et al. 1999). The Stratagene kilobase DNA ladder was used as a size marker.

Published protocols (Kirk and Kirk 1985; Hallmann and Sumper 1994) were used to prepare RNA. First-strand cDNA was generated by reverse transcription by use of regA-specific oligomers pregA (5′-GCACTGCCGGAAGAGCCG-3′) and pregD (5′-CGACGCTCCTGTCGAGGC-3′), respectively, as primers. The subsequent PCR used gene-specific primers (pregB, 5′-CGCCTCGTGATCCCGGAC-3′; pregC, 5′-GATG GACCGGTACCACTTG-3′; pregE, 5′-CGCAGTTCGAAAAG-3′; pregF, 5′-TGCGACGGTGGCGGCG-3′; pregG, 5′-GCAGC CGAAGCGAGCATC-3′), as shown in Figures 3, 4, and 5. Cycling conditions were empirically established for each primer pair. Expression of P_TubRegA in transformants was analyzed by RT–PCR of purified RNA from somatic cells and gonidia.