regA, a Volvox gene that plays a central role in germ-soma differentiation, encodes a novel regulatory protein

Volvox carteri, a simple multicellular green alga, is an attractive model for studying the genetic control of germ-soma differentiation, because each spheroid (or ‘individual’) contains only two cell types: mortal somatic cells and immortal germ cells (Kirk, 1998; Schmitt et al., 1992). A typical asexual spheroid contains 16 large asexual reproductive cells, or ‘gonidia’, just internal to a surface monolayer of small, biflagellate, Chlamydomonas-like somatic cells (Fig. 1A). Gonidia are immotile and specialized for reproduction; each will divide to produce a juvenile containing a new cohort of gonidia and somatic cells (Starr, 1969, 1970). Somatic cells, in contrast, are specialized for motility but have no reproductive potential; once differentiated, they never divide, and they eventually undergo programmed death (Pommerville and Kochert, 1981, 1982).

The regA locus plays a central role in establishing this germsoma division of labor by acting as a negative regulator of reproductive functions in somatic cells. Whereas regA⁺ somatic cells can not be forced to divide, regA⁻ somatic cells exhibit all of the reproductive potentials of the species (Huskey and Griffin, 1979; Kirk et al., 1987; Starr, 1970). In the asexual phase of the life cycle, regA⁻ mutants exhibit the ‘somatic regeneration’ (or ‘Reg’) phenotype, in which somatic cells appear to differentiate normally, but then redifferentiate as gonidia and divide to produce juveniles of like phenotype; in the sexual phase regA⁻ somatic cells produce offspring containing eggs or sperm (Starr, 1970). All of the >100 Reg mutants that have been analyzed to date (Huskey and Griffin, 1979; Kirk et al., 1987; D. Kirk, M. Kirk and K. A. Stamer, unpublished observations) have lesions that map to regA, indicating that this is the only locus that can mutate to yield a Reg phenotype. No effect of regA mutations on development of ‘true’ gonidia can be detected, indicating that the gene is not expressed in gonidia. Molecular studies reinforce the view that regA acts as a negative regulator of reproductive functions in somatic cells, because regA⁻ somatic cells accumulate transcripts that are normally accumulated only by developing gonidia (Tam and Kirk, 1991b). We conclude that regA is a master regulatory gene that plays a central role in germ-soma differentiation by acting in somatic cells to prevent expression of functions required for germ cell development.

Prior efforts to clone the regA locus by performing a chromosome walk from two different linked polymorphisms (Adams et al., 1990; Harper et al., 1987) did not succeed (D. L. Kirk and J. K. Miller, unpublished). Here we report success in using the cold-inducible Volvox transposon, Jordan, to tag and recover the regA locus, and we show that regA encodes a nuclear protein with a combination of features similar to ones found in various transcription factors.

Except where otherwise specified, V. carteri cultures were maintained in standard Volvox medium (SVM) at 32°C, under standardized conditions (Kirk and Kirk, 1983, 1985). SVM lacking the usual urea or ammonium chloride (N-SVM) was used to select for ability to reduce nitrate.

EVE (Adams et al., 1990), is a subclone of HK10, the standard female lab strain of V. carteri (Starr, 1969, 1970). CRH22 (Miller et al., 1993), is a morphologically wild-type, Nit⁻ (reduced nitrogen requiring) derivative of EVE. HB11A is a multiply-marked derivative of EVE used in previous mapping studies (Adams et al., 1990); it carries a bleomycin induced regA mutation, and a stable nitA mutation that causes a splicing defect (Gruber et al., 1996). Progeny analyzed in Fig. 2C came from crosses between HB11A and PM1, a genetically distant, wild-type male (Adams et al., 1990). Strain 153-68, a Reg⁻, Nit⁻ member of that progeny cohort, was the recipient in all regA transformation attempts; the mutant regA locus it inherited from HB11A has a 281 bp deletion in exon 6 (Figs 3C and 6) that deletes 93.67 codons, causes a frame shift that generates a stop codon 12 nt down stream, and should result in a mutant regA product only 291 residues long.

Candidate Jordan-tagged Reg mutants were isolated by microscopic screening of flask cultures of CRH22 after 7-10 days growth at 24°C. Prospective mutants were cultured at 32°C in individual wells of multi-well plates and examined at intervals to determine the stability of the mutant phenotype. No more than one mutant was saved from any flask. This procedure yielded 36 Reg mutants, seven of which were highly revertible, and 28 of which were subsequently found to exhibit readily detected RFLPs at the regA locus.

Methods for isolation and Southern blot analysis of Volvox DNA were as previously described (Miller et al., 1993). The first fragment representing part of the regA locus was cloned as follows. The novel 2.7-kb Jordan-containing SalI restriction fragment (Fig. 2A) of strain 22Reg5 (Fig. 1B) was gel-purified, cloned in plasmid pJOE533 (a suicide vector, Josef Altenbuchner, personal communication), and used as the source of a genomic fragment flanking the site of Jordan insertion that was then subcloned in Bluescript KS⁺ (Stratagene), yielding plasmid pVR2-B.

About 10⁴ pfu from an unamplified library of randomly sheared EVE DNA fragments in λ Dash II (Stratagene) were plated on XL1-Blue MRA (P2) bacteria, and screened with a regA probe, resulting in recovery of 17 distinct genomic clones, one of which (λ7/1, Fig. 3B) was successfully used for phenotypic rescue of Reg strain 153-68.

A plasmid, pVcRegA1, containing the 15.3-kb insert of λ7/1 was prepared as follows. First pSK⁻ was digested with NotI and XhoI, and a three-partner ligation was performed with it, a 2.7-kb NotI/ClaI fragment derived from the left end of the λ7/1 insert, and a 5.8-kb ClaI/XhoI fragment derived from the right end of the λ7/1 insert. The resulting plasmid was reopened by ClaI digestion, treated with alkali, and ligated to a 6.8-kb ClaI central fragment of the λ7/1 insert. Cloning sites and the orientation of the central ClaI fragment were analyzed by sequencing, and the presence of a complete regA gene was verified by transformation rescue of strain 153-68.

Genomic clones were tested for ability to rescue the Reg phenotype via biolistic co-transformation (Schiedlmeier et al., 1994), with the modified V. carteri nitA gene carried on plasmid pVcNR15 (Gruber et al., 1996) used as the selectable marker, and strain 153-68 as the DNA recipient. Following bombardment, samples were cultured in small aliquots in N-SVM in multiwell plates and monitored at intervals for the appearance of green, growing Nit⁺ survivors. Survivors were transferred to well plates or tubes containing fresh N-SVM and scored with respect to their Reg phenotypes. No more than one survivor was saved per well. Putative transformants were subjected to PCR amplification of the region in which strain 153-68 has a deletion, using the primers pR1 (5′GTCACATTCG-GGTGCAGTCG3′) and pR4 (5′GCCACCAGCTCTTCCAGGTT3′) which generate 1038 bp and 757 bp products with the wild-type and mutant templates respectively.

The genomic sequence of regA was obtained from plasmid subclones of λ7/1, using dideoxynucleotide chain-termination sequencing (Sanger et al., 1977) with either the T7 Sequenase Quick Denature plasmid sequencing kit (Amersham Life Science) with α-³⁵S-dATP (Amersham), or the Thermo Sequenase radiolabled-terminator, cycle-sequencing kit (Amersham) with [α-³³P]ddNTPs. All regions of regA present on the λ7/1 insert, including all non-coding regions, were sequenced in both directions to a 99.9% level of accuracy. cDNA clones and RT-PCR products were also sequenced by the dideoxy chain-termination method, either manually (using a kit from United States Biochemical), or with an ABI Prism automated sequencer.

Because repetitive screenings of existing V. carteri cDNA libraries with regA genomic probes yielded no bona fide regA cDNAs, a new library was constructed in Lambda ZAPII (Stratagene) using poly(A)⁺ RNA isolated from young juvenile spheroids of EVE. regA cDNAs were present in this library at a frequency of approx. 3×10⁻⁴, but the longest insert recovered in four rounds of screening was only approx. 2.4 kb long, approx. one third as long as the regA transcript detected on northern blots. (Subsequently it was found that the region upstream of the 5′ end of the longest cDNA was 63% G+C, and extremely difficult to copy with reverse transcriptase.) Hence RT-PCR amplification was used to complete the analysis of the regA transcript.

The RNA used as template in RT-PCRs designed to complete the sequencing of the regA transcript was isolated as previously described (Kirk and Kirk, 1985) from young juvenile spheroids. First-strand cDNAs were prepared using either oligo(T), random-sequence deoxynucleotide hexamers, or (in a few cases) regA-specific oligomers as primers for the reverse transcription. Each PCR used a pair of gene-specific primers (Gibco BRL), and cycling conditions were empirically established for each primer pair. The final step in the RT-PCR analysis of the cDNA involved the use of two nested pairs of gene-specific primers to perform ‘circular first-strand cDNA-mediated rapid amplification of cDNA ends’ (cRACE) by the method of Maruyama et al. (1995), and thereby define the 5′ end of the transcript.

First strand cDNAs for semi-quantitative RT-PCR were generated (as described above) from total RNA samples isolated from either whole spheroids, purified somatic cells, or purified gonidia/embryos harvested from synchronous cultures at intervals throughout the life cycle. Two primer pairs were used for PCRs: A 237 bp regA-specific product was generated using pR62 (5′GAACTCC-CGAAAACCTT-G3′), which corresponds to a region of the regA sense strand in exon 5, and pR2 (5′CGACTGCACCCGAATGTGAC3′), which corresponds to a portion of the antisense strand in exon 6, while a 355 bp S18-specific control product was generated with primers c38-1 (5′ATGGGCTCTCTGGTCCACGGCG3′) and c38-2 (5′CGGATC-TTCTTCAGGCGCTCCA3′). (The C38 RNA, which encodes the S-18 ribosomal protein, is present in approximately constant abundance in both cell types; Tam and Kirk, 1991a.) Each RT product used as a template was diluted to the point that the amount of PCR product produced with each primer pair was proportional to the amount of template used. Aliquots of six regA-specific and six S18-specific PCR products generated at the same time from a single RT product were electrophoresed through a 1% agarose gel, stained with ethidium bromide, and visualized by UV transillumination. Digital images of each gel were obtained with a CCD camera (Cosmicar/Pentax), and the intensity of the bands was quantitated with the NIH Image1.61 software. The relative abundance of regA message at each time point was then calculated by dividing the mean product yield of the six regA PCRs by the mean product yield of the six control (S18) PCRs.

The deduced RegA peptide sequence was analyzed by the SAPS (Brendel et al., 1992), PSORT (Nakai and Kanehisa, 1992), BLAST (Altschul et al., 1990) and PHDsec (Rost and Sander, 1993a, b) computer programs.

One copy of the coding sequence for the hemagglutinin (HA) nonapeptide epitope (Atassi and Webster, 1983; Field et al., 1988) was inserted into the plasmid pVcRegA1, using a KpnI site at the end of exon 6 (Fig. 3C) and a pair of oligonucleotides designed to have Volvox codon bias (Schmitt et al., 1992). The resulting plasmid carrying an HA-tagged regA gene (pVcRegA2) was used to co-transform strain 153-68 and generate strain A2-12, a Nit⁺ Reg⁺ co-transformant that was then used for western blot and immunocytological analyses.

Extracts of A2-12 for western blots were prepared as previously described (Huber et al., 1995), and SDS-PAGE and immunoblotting were carried out according to standard procedures (Sambrook et al., 1989). Immunoblots were reacted for 1 hour at room temperature with a 1:120 dilution of a commercial anti-HA antibody (Boehringer Mannheim), and immunoreactivity was visualized using horseradish peroxidase-linked sheep-anti-mouse antibody (Amersham, 1:1000) and the ECL detection system (Amersham).

Samples of A2-12 spheroids for immunocytologic examination were harvested by filtration, disrupted mechanically, attached to 0.1% polyethylenimine-treated coverslips, fixed for 15 minutes with freshly prepared 3.7% EM grade formaldehyde (Ted Pella Inc.), 0.1% Triton X-100, and 1 mM dithiothreitol (DTT) in phosphate-buffered saline (PBS), extracted for 1 hour with 1% NP40, 1% bovine serum albumin and 1 mM DTT in PBS to remove chlorophyll, exposed for 1 hour to ‘blocking solution’ (1% BSA and 0.1% Tween 20 in PBS), then overnight to primary antibody (1:10 dilution of anti-HA monoclonal 12CA5, provided by L. Ellis, Washington Univ.) and 4 hours to a mixture of secondary antibodies (2 μg/ml each of Cy™2-labeled goat anti-mouse IgG, and Cy™3-labeled goat anti-rabbit IgG, Amersham). After each of the above steps specimens were washed extensively with 0.1% BSA and 0.1% Tween 20 in PBS. Coverslips were mounted with 1 mg/ml phenylendiamine in 50% glycerol, examined by epifluorescence with a Zeiss Photomikroskop I, and photographed on Hypertech film (Microfluor, Ltd., Stony Brook, NY).

Transposition of the mobile Volvox element Jordan can be stimulated by several types of stress (Miller et al., 1993). During cultivation at 22-24°C (approx. the lower limit for growth of V. carteri) transposition rates of approx. 10⁻⁴ per gonidium per generation are observed, which is >30-fold higher than the rate at the usual culture temperature of 32°C (S. M. Miller, unpublished). We recovered 36 independent Reg mutants (Fig. 1B) from cultures of CRH22, a morphologically wild-type strain, maintained at 22-24°C and observed that one of them (22Reg5) exhibited a novel, well-resolved Jordan-containing genomic DNA fragment (Fig. 2A). DNA flanking the Jordan insertion site in this novel fragment was cloned to produce plasmid pVR2-B. When used as a hybridization probe, the pVR2-B insert detected RFLPs in 11 of 23 independent cold-induced Reg mutants that were analyzed on DNA blots (Fig. 2B) and also detected a 281 bp deletion in a bleomycin-induced Reg mutant that co-segregated perfectly with the Reg phenotype among the approx. 50 sexual progeny analyzed (representative data, Fig. 2C).

Two of 17 overlapping clones recovered when pVR2-B was used to probe a lambda library of randomly sheared V. carteri DNA fragments are shown diagrammatically in Fig. 3B. When various fragments of these clones were used as probes, they detected RFLPs similar to those illustrated in Fig. 2 in 28 of the 36 Reg mutants that had been isolated following low-temperature growth of CRH22. The distribution of these RFLPs (Fig. 3A) provided the first estimate of the boundaries of the regA locus.

Strain 153-68 (Fig. 4A), which carries a non-revertible mutation of the nitA (nitrate-reductase encoding) locus, as well as the 281 bp deletion at the regA locus (Fig. 3C), was used as a recipient to test the capacity of putative regA genomic clones to achieve mutant rescue via biolistic co-transformation with plasmid pVcNR15 carrying the Volvox nitA gene as the selectable marker (Schiedlmeier et al., 1994). When clone λ7/1 (Fig. 3) was used for cotransformation, approx. 30% of all Nit⁺ transformants that were recovered had wild-type (regA⁺) morphology (Fig. 4B), and Southern blot analysis of four such wild-type strains revealed that in each case the putative regA transformant had stably incorporated the transgene (Fig. 4C). In contrast, no amelioration of the Reg phenotype was exhibited by any of the 130 Nit+ transformants recovered following co-bombardment of the same recipient strain with clone λ4/2, the insert of which is only 300 bp shorter at the 5′ end than the λ7/1 insert is (Fig. 3). This suggested that the 5′ end of the regA functional unit might be located near the left end of the λ7/1 insert.

A pSK-based plasmid, pVcRegA1, carrying the λ7/1 insert was prepared and found to be capable of rescuing Reg mutants even more efficiently than λ7/1 (data not shown). Studies using variants of this plasmid carrying regA genes lacking one or more introns are in progress.

Previous studies (Huskey and Griffin, 1979; Tam and Kirk, 1991a,b; Tam et al., 1991) led to the conclusion that regA first acts while presumptive somatic cells are in the early stages of cytodifferentiation, many hours before the Reg phenotype can be detected microscopically. When developmental northern blots were probed with cloned regA fragments, an approx. 7 kb transcript was detected in very young somatic cells (Fig. 5A), but never in gonidia (data not shown). However, the amount of regA transcript often was so near the limit of detection on RNA blots that expression patterns analyzed this way were not very reproducible. Therefore, semi-quantitative RT-PCR was used to examine temporal and spatial patterns of accumulation of the regA transcript. These studies indicated that the transcript appears in very young somatic cells immediately after the end of embryogenesis, and then is present in fluctuating (but generally increasing) abundance in somatic cells throughout most of the life cycle (Fig. 5B). In contrast, the transcript was not detected by RT-PCR in gonidia.

Repetitive screenings of a somatic-cell cDNA library produced several categories of regA cDNA clones that all terminated downstream of a canonical algal polyadenylation signal (TGTAA; Schmitt et al., 1992), but differed in their 5′ ends. But because the longest cDNA thus recovered was only approx. one third the length of the transcript detected on northern blots, RT-PCR (using primers based on the genomic sequence) was used to amplify and sequence the rest of the transcript. The 5′ end of the transcription unit was then identified by the ‘cRACE’ method of Maruyama et al. (1995).

Comparison of the cDNA sequence with the genomic sequence revealed that the 12,477 nt-long regA transcription unit commences 23 nt downstream of a putative TATA box, and only 70 nt downstream of the left end of the λ7/1 insert (the genomic clone that rescues the Reg phenotype of mutants with high efficiency; Fig. 3B). The transcription unit comprises eight exons and seven introns, with four of the introns being located in the 5′ UTR (Fig. 3C). The mature regA transcript as depicted is 6,725 nt long and consists of a 940-nt 5′ UTR, a 3,147-nt coding region, and a 2,638-nt 3′ UTR that contains a canonical green-algal polyadenylation signal (UGUAA) 13 nt upstream of the polyadenylation site.

Although all objective evidence presently indicates that the transcription unit has the structure depicted in Fig. 3B, we cannot rule out the possibility that one or more alternative exon 7-intron 7 (E7-I7) splice sites may be used. Although in each of several experiments with independent RNA samples the only RT-PCR product recovered so far has had the E7-I7 structure illustrated, reverse transcription of this region has consistently been exceptionally difficult. Whereas the intron depicted exhibits the same deviation from the canonical border sequence (5′GC….AG3′) as three previously described volvocalean introns (Fabry et al., 1993; Harper and Mages, 1988; Kropat et al., 1995), multiple in-frame, canonical potential splice sites are present downstream, in an open reading frame potentially capable of encoding a peptide region very similar in amino acid composition to the rest of the predicted protein. Studies to resolve this issue are in progress.

The regA transcript shown in Fig. 3B is predicted to encode a 1,049-residue, approx. 111 kDa polypeptide that is unusually hydrophilic, and that has a net charge of −47 and a predicted pI of approx. 4.8 (Fig. 6). Computer-aided evaluation of this sequence revealed the following features. RegA is in the 99% quantile of proteins in the database with respect to exceptionally high content of glutamine, alanine and proline, and in the 5% quantile with respect to exceptionally low content of major hydrophobic amino acids. There are 113 ‘multiplets’ (homo-oligopeptides, such as P₂, A₇, Q₁₂, etc.; all overlined in Fig. 6), seven of which (at residues 246-50, 604-11, 661-7, 789-95, 798-802, 847-58, and 936-45) are unusually long. No N-terminal signal sequence, potential transmembrane segments, or mitochondrial- or chloroplast-localization signals were detected; however, six nuclear localization signals of the SV40 large T antigen type (one each at residues 461-4, 911-14, and 1016-19 and three overlapping ones at residues 798-803), plus one nuclear localization signal of the Xenopus nucleoplasmin type (at residues 618-34) and two potentially significant helix-loop-helix domains (at residues 639-708 and 781-826) were detected.

A BLAST search (Altschul et al., 1990) did not identify any protein in the database with a highly significant level of overall sequence similarity to RegA. However, three glutamine-rich regions, two alanine-rich regions and two proline-rich regions of RegA (boldfaced in Fig. 6) were identified as having significant similarity (sum probabilities of 10⁻¹⁰ to 10⁻²²) to various sequence elements found in a diverse array of proteins – primarily, but not exclusively, transcription factors – that have been implicated in developmental regulation in other model organisms (see Discussion).

Strain A2-12 carrying a transgene encoding the RegA protein tagged with the influenza-virus hemagglutinin (HA) epitope was generated by co-transformation of strain 153-68 with a suitably modified plasmid. Anti-HA-stained western blots of stage-specific extracts of A2-12 revealed that HA-tagged RegA protein can first be detected by this method approx. 4 hours after the lights come back on after the completion of embryogenesis (Fig. 7A). In contrast, the regA transcript can be detected several hours earlier, at the very beginning of the light period (Fig. 5 and other data not shown). Whether this represents a difference in the threshold sensitivity of the RNA and protein detection methods, or reflects a delay in the translation of the regA message is currently unknown. After its initial appearance, HA-tagged RegA protein accumulates progressively for several hours, and then it remains moderately abundant throughout the life of the somatic cells, but it is never detected in gonidia (data not shown).

Indirect immunofluorescence examination of A2-12 with anti-HA antibody clearly reveals, as predicted by PSORT analysis of the sequence, that RegA is a nuclear protein (Fig. 7B-D). No nuclear staining with the anti-HA antibody could be detected in gonidia of strain A2-12, or in somatic cell nuclei of a control strain (EVE) not carrying the HA-tagged transgene.

Ever since Richard Starr described the first mutant with a ‘fertile somatic cell,’ or ‘somatic regeneration’ phenotype more than a quarter century ago (Starr, 1970), the regA locus, to which all such mutations map, has fascinated Volvox developmental biologists. It has long been appreciated that this locus must act as a master regulatory gene that suppresses all aspects of reproductive behavior in the somatic cells of V. carteri, because whereas regA⁺ somatic cells have no reproductive potentials, regA⁻ somatic cells redifferentiate as germ cells that exhibit all of the asexual and sexual reproductive potentials of the species. The central role played by regA in programming the germ–soma dichotomy of Volvox carteri motivated our efforts to clone the locus in order to begin an analysis of its regulation and function.

The present study confirms the prediction made at the time that we initially described the transposon Jordan (Miller et al., 1993), namely that it has properties that should make it highly suitable for tagging and recovering Volvox genes of developmental interest. Of particular importance is the fact that Jordan can be induced to transpose at a substantially elevated rate by any one of a variety of mild environmental stresses (Miller et al., 1993; S. M. Miller, unpublished observations). The stress used here (cultivation at reduced temperature) is clearly the simplest and most convenient for routine use. Of the 36 Reg mutants that were isolated following cultivation of a wild-type strain at 24°C, 28 (80%) have readily detected RFLPs at the regA locus (Fig. 3), and of these at least 24 have lesions consistent with the hypothesis that they have been inactivated by a Jordan insertion (data not shown).

Comparison of the complete cDNA sequence with the genomic sequence revealed two unusual features of the regA transcription unit that may be related: (1) the presumed TATA box is located only 42 bp downstream – and the transcription start site is located only 70 bp downstream – of the end of the λ7/1 insert, and (2) the 940-nt-long 5′ UTR is spread out over >2.3 kb of genomic sequence and is interrupted by four introns.

The high frequency with which a Reg mutant can be restored to a fully wild-type phenotype by λ7/1 DNA (approx. 30%), or by the pVcRegA1 plasmid carrying the same insert (>40%), clearly indicates that this genomic clone must contain all of the elements required for normal cell-type-specific expression of regA, despite its short (70 bp) upstream region. We speculate, therefore, that at least some of the cis-regulatory elements necessary for proper regulation of regA expression may be located within one or more of the four introns that interrupt the 5′ UTR. This speculation is consistent with the fact that the region where disabling mutations have been recovered at the highest density is in the 5′ UTR (Fig. 3A). We are in the process of testing this hypothesis by evaluating the ability of plasmids lacking one or more of the introns in the 5′ UTR region to provide transformation rescue of mutants. Parallel studies of plasmids carrying inserts truncated at the 3′ end will be used to test for possible regulatory elements in the 3′ UTR.

Conventional genetic reasoning leads to the conclusion that the function of the regA gene product, RegA, is to act as a negative regulator of reproductive functions. Therefore, we have long anticipated that regA might encode a transcriptional repressor of genes whose products are required for reproductive development. A BLAST search (Altschul et al., 1990) did not identify any protein in the database that exhibits a significant level of overall sequence similarity to the deduced peptide sequence of RegA. Nevertheless, many proteins were identified that possess sequence elements with strong similarity (sum probability scores 10⁻¹⁰ to 10⁻²²) to certain RegA sequence elements, particularly three glutamine (Q)-rich regions, two alanine (A)-rich regions, and two proline (P)-rich regions. Among the proteins so identified were a few that are involved in important aspects of developmental regulation, but that are not transcription factors. However, the majority of the proteins that have sequence elements with significant similarity to elements in RegA are known or suspected transcription factors, including 19 different types from yeast, slime molds, plants, insects and humans. This set of relationships is hardly surprising, since it has been reported previously that 82% of the proteins in the SwissProt data bank with the most significant polyQ motifs, and 78% of those with the most significant polyP stretches, are known or suspected transcription factors (Gerber et al., 1994). The possibility that RegA might function as a transcription factor, is reinforced by the fact that the PSORT program assigned it a nuclear location with a ‘certainty’ value of 0.96, and that immunocytology confirmed this prediction (Fig. 7B-D).

Further consideration of the amino acid composition of RegA supports the concept that it might function as a repressor: RegA falls in the 99th quantile with respect to its A+Q+P content (34.5% of all residues) and possesses several regions of substantial length (12-50 residues) that are more than 50% A and/or Q and/or P. Abundance of A, Q and/or P in key functional domains is a unifying feature of a group of transcription factors that often have little discernible structural similarity otherwise, but that all function as ‘active’ repressors – that is, factors that can inhibit transcription initiation directly, as opposed to down-regulating the activity of transcriptional activators (Cowell, 1994; Hanna-Rose and Hansen, 1996). Among the most well-known and most studied active repressors are several of the segmentation genes that act to pattern the early Drosophila embryo: namely, Krüppel (Kr), knirps (kni), even-skipped (eve), engrailed (en) and paired (prd). Each such active repressor possesses a discrete ‘repression domain’ that can be attached to the DNA-binding domain of a transcriptional activator (such as yeast GAL4) to convert the latter into an active repressor (Han and Manley, 1993; Hanna-Rose and Hansen, 1996). No two minimal repression domains yet analyzed have significant sequence similarities, but most of them can be categorized in terms of amino acid composition. For example, the repression domains of Kr, kni and en are all A-rich, whereas the prd repression domain is P-rich, the eve repression domain is A+P-rich, the repression domain of DR1 (a general repressor of transcription) is A+Q-rich, and although the Wilms tumor repressor, WT1, is said to fall into none of these categories, and to have a ‘unique’ type of minimal repression domain (Hanna-Rose and Hansen, 1996), it is 35% A+Q+P (Wang et al., 1995), and mutations in P+Q-rich areas flanking its minimal repression domain reduce WT1 repressor activity (Madden et al., 1993).

Based on the above considerations, combined with the conclusion drawn from earlier developmental studies that regA encodes a negative regulator of reproduction, we propose that RegA functions as an active transcriptional repressor.

The 18 families of ‘maturation-abundant’ gonidial cDNAs of V. carteri (Tam and Kirk, 1991a) define a set of genes that are candidate targets of RegA regulation. In wild-type spheroids transcripts of these genes begin to accumulate at a low rate in very young somatic cells and gonidia, but then, at about the time that regA is believed to act, their transcripts disappear from somatic cells while they accumulate at accelerating rates in developing gonidia (Tam and Kirk, 1991a). In Reg mutants, by contrast, these genes are never down-regulated in the somatic cells; rather, their transcripts accumulate progressively, much as they do in wild-type gonidia (Tam and Kirk, 1991b).

When three of these genes were subsequently characterized by cDNA sequencing, it was learned that they were nuclear genes encoding three key chloroplast proteins (Choi et al., 1996). Recently we have extended this unexpected finding by sequencing the remaining 15 maturation-abundant gonidial cDNAs and determining that all 13 of the genes in this category that encode polypeptides of known function are nuclear genes encoding important chloroplast proteins (M. Meissner, K. Stark, D. L. Kirk and R. Schmitt, unpublished data). These observations lead to the working hypothesis that the mechanism by which RegA prevents somatic cells from entering the reproductive pathway is by preventing transcription of genes that are required for chloroplast biogenesis and (indirectly) cell growth. Volvox is an obligate photoautotroph, in which growth is photosynthesis-limited. During embryogenesis, the chloroplast of a gonidium is subdivided among its >2,000 progeny cells, so that each somatic cell begins life with a very modest chloroplast endowment. But if unable to add to this endowment, a somatic cell would have very limited growth potential. And if unable to grow, it would also be unable to reproduce.

One of the more attractive features of the above hypothesis is that it facilitates visualizing how this single gene product, RegA, is able to act as a negative regulator of two complete developmental pathways: the pathways leading to asexual and sexual reproduction. In both green algae and higher plants it has been clearly established that the nuclear and chloroplast genes whose products are required for chloroplast biogenesis are subject to coordinate regulation in response to a wide variety of different intrinsic and extrinsic signals (Kuhlemeir et al., 1987; Sheen, 1994). Of particular relevance in the present context is the fact that in various higher plants, regulatory elements capable of effecting cell-type-specific silencing of chloroplast-specific nuclear genes are both ubiquitous and varied in nature (Kuhlemeir et al., 1987; Sheen and Bogorad, 1986; Simpson et al., 1986). Clearly, the ‘photoregulon’ of plants and algae is highly receptive to the imposition of new levels of negative transcriptional control that have adaptive value. Our current hypothesis is that the regA gene plays its central role in Volvox germ-soma differentiation by exploiting this propensity of algal photogenes. The most attractive feature of this hypothesis is that it is testable.

This paper is dedicated to the memory of Richard Starr, in whose laboratory Volvox was first brought into continuous culture, who isolated the V. carteri strains that served as the progenitors of all of the strains used here, who described the first Reg mutant, and who was a cherished friend and colleague of ours until his sudden death on February 3, 1998. We thank Lisa Ellis for providing the anti-HA monoclonal antibody, Ichiro Nishii for advice regarding preparation of V. carteri embryos for immunocytology and Leonard Duncan for extremely useful discussions and advice during late stages of the research and preparation of the manuscript. This work was supported by grants from the NSF (no. MCB-9304447) and USDA (no. 97-3504-4626) to D. K., a DAAD travel grant to K. S., and a grant from the Deutsche Forschungsgemeinschaft (SFB521/B1) to R. S.