In Caenorhabditis elegans, EFL-1 (E2F), DPL-1 (DP) and LIN-35(pRb) act coordinately in somatic tissues to inhibit ectopic cell division,probably by repressing the expression of target genes. EFL-1, DPL-1 and LIN-35 are also present in the germline, but do not always act together. Strong loss-of-function mutations in either efl-1 or dpl-1 cause defects in oogenesis that result in sterility, while lin-35 mutants are fertile with reduced broods. Microarray-based expression profiling of dissected gonads from efl-1, dpl-1 and lin-35 mutants reveals that EFL-1 and DPL-1 promote expression of an extensively overlapping set of target genes, consistent with the expectation that these two proteins function as a heterodimer. Regulatory regions upstream of many of these target genes have a canonical E2F-binding site, suggesting that their regulation by EFL-1/DPL-1 is direct. Many EFL-1/DPL-1 responsive genes encode proteins required for oogenesis and early embryogenesis, rather than cell cycle components. By contrast, LIN-35 appears to function primarily as a repressor of gene expression in the germline, and the genes that it acts on are for the most part distinct from those regulated by EFL-1 and/or DPL-1. Thus, in vivo, C. elegans E2F directly promotes oogenesis and embryogenesis through the activation of a tissue-specific transcriptional program that does not require LIN-35.
In mammals, the retinoblastoma tumor suppressor (pRB) plays a crucial role in controlling cellular proliferation and differentiation, in large part by regulating the activity of a transcription factor called E2F. E2F is typically a heterodimer composed of one E2F and one DP subunit. In the canonical model for Rb pathway function, pRB binds to E2F in the G0 or early G1 phase of the cell cycle, preventing activation and/or promoting repression of target genes (reviewed by Trimarchi and Lees, 2002). Phosphorylation of pRB by cyclinD/cdk4 releases E2F, which is then free to activate genes important for DNA synthesis, promoting progression from G1 into S phase. However, the roles of pRB and E2F are more diverse than this model indicates, as these proteins modulate apoptosis, senescence and differentiation, as well as cell cycle progression (reviewed by Dimova and Dyson, 2005; Cobrinik, 2005).
Greatly complicating the understanding of this crucial pathway is the fact that each of these proteins is a member of a gene family. The mouse and human genomes each encode three pRB-like `pocket' proteins, eight E2F-like proteins and two DP-like proteins (reviewed by Dimova and Dyson, 2005). Family members can exhibit both redundant and independent functions. For instance, loss of pRB is compensated by the activity of another pocket protein, p107, in mouse embryo fibroblasts(Sage et al., 2003). p107 and the third pocket protein p130 have independent functions as well, as they typically bind to a different set of E2Fs than does pRB (reviewed by Trimarchi and Lees, 2002). E2F proteins are functionally divided into those that activate transcription(E2F1-3) and those that repress transcription (E2F4-8). Activator E2Fs are primarily bound by pRB, repressors E2F4 and E2F5 are bound by p107 and p130,while E2F6-8 lack any pocket protein-binding domain. Moreover, deletion mutations in any of six endogenous E2F genes in mice do not display the same spectrum of defects, suggesting that each protein acts at distinct times and places in development and organ homeostasis (reviewed by Dimova and Dyson, 2005). In vivo, different E2Fs appear to be able to both promote and hinder tumor formation, and the mechanisms underlying these diverse outcomes are largely unknown.
Recent microarray and chromatin immunoprecipitation experiments in tissue culture cells have identified E2F target genes that expand the role of E2F/DP beyond the regulation of G1/S phase transition, including genes that act in mitosis, DNA repair and recombination(Ishida et al., 2001; Muller et al., 2001; Ren et al., 2002). In Drosophila S2 cells, the sole activator E2F, dE2F1, primarily activates cell cycle-regulated genes, while the repressor dE2F2 targets developmentally regulated genes independent of the cell cycle(Dimova et al., 2003). Overall, these and other analyses have led to the concept that E2Fs activate genes that promote the cell cycle and repress genes required for differentiation. However, most of these analyses have been performed in cell culture, and how tissue-specificity and developmental context impact the regulation of E2F target genes is just beginning to be addressed.
The hermaphrodite nematode Caenorhabditis elegans provides both a streamlined pathway and the opportunity for in-vivo analysis. Its genome encodes a single pRB-like protein (LIN-35), a single DP-like protein (DPL-1)and three E2F-like proteins (EFL-1, EFL-2 and F49E12.6). Of the three E2F-like proteins in C. elegans, a phenotype has been attributed only to efl-1, which is most closely related to the mammalian repressor E2F4(Ceol and Horvitz, 2001; Page et al., 2001). RNA-mediated interference studies of efl-2 and F49E12.6 have not identified any apparent phenotype (Boxem and van den Heuvel, 2002; Ceol and Horvitz, 2001; Kamath et al., 2003; Rual et al.,2004; Sonnischen et al., 2005).
The C. elegans pRB/E2F pathway is required for multiple aspects of somatic development. lin-35, efl-1 and dpl-1 are all components of the SynMuv B pathway, which inhibits ectopic vulval development redundantly with a second pathway, the SynMuv A pathway(Lu and Horvitz, 1998; Ceol and Horvitz, 2001). In addition to Rb and E2F orthologs, the SynMuv B pathway includes components of a histone deacetylase complex that probably acts to repress transcription(Lu and Horvitz, 1998; Korenjak et al., 2004). LIN-35, EFL-1 and DPL-1 proteins have been demonstrated to physically associate in vitro (Ceol and Horvitz,2001). Together, the common phenotype, physical association and similarity to mammalian proteins support the argument that LIN-35, EFL-1 and DPL-1 act in concert to repress gene expression. Additionally, lin-35,efl-1 and dpl-1 function together in G1 cell cycle control in other tissues that exhibit postembryonic cell division, including the intestine and ventral cord (Boxem and van der Heuvel, 2002). In all of the above cases, lin-35,efl-1 and dpl-1 act together to negatively regulate cell division. Notably, lin-35 acts redundantly with other genes such as ubc-18 and xnp-1 during pharynx and somatic gonad development, respectively, without regulating cell proliferation(Fay et al., 2004; Bender et al., 2004).
Here we investigate the function of lin-35, efl-1 and dpl-1 in the germline. Null phenotypes of efl-1, dpl-1 and lin-35 reveal distinct requirements for these proteins in regulating germline development. While EFL-1 and DPL-1 are essential for fertility,LIN-35 is dispensable. Global gene expression profiling of dissected gonads from lin-35, efl-1 and dpl-1 mutants reveals an extensively overlapping gene expression program for EFL-1 and DPL-1. Loss of efl-1 and dpl-1 decreased expression of a common set of genes that promote specific aspects of a developmental program -oogenesis and early embryogenesis - rather than cell cycle. We identified an over-represented sequence in the 5′ regions of these genes that closely resembles the mammalian E2F-binding site. LIN-35 plays only a minor role in the regulation of these genes and instead appears chiefly to repress a distinct set of genes with diverse functions in the germline. Our results demonstrate that, in vivo, E2F can directly initiate a developmental program by activating genes that promote differentiation with only minor effects on the cell cycle.
MATERIALS AND METHODS
Strains and maintenance
Nematode strain maintenance was as described(Sulston and Hodgkin, 1988). All strains were grown at 20°C. The following variants were used: N2 (wild type), lin-35(n745) I (Lu and Horvitz, 1998), unc-4(e120); dpl-1(n3316) II(Ceol and Horvitz, 2001), efl-1(n3639) V (Coel and Horvitz, 2001), rme-2(b1005) IV(Grant and Hirsh, 1999) and AZ212 (H2B:GFP) (Praitis et al.,2001). We examined two different isolates of lin-35(n745)and found that both isolates exhibited sterility at 25°C in our hands, a phenotype not previously reported.
Gonad dissection and microarray analysis
Wild-type and mutant worms were staged by bleaching gravid adults to collect eggs, which were then hatched in S-basal solution in the absence of food. Starved L1 larvae were cultured with food (bacterial strain OP50) and harvested 70-72 hours later, the first point at which efl-1(n3639)mutants could be distinguished from siblings. Adult worms were placed in dissection buffer (M9 with 0.1% levamisole and 0.001% Tween 20) on a coverslip. For each strain, a pair of 30 1/2-gauge needles was used to extrude approximately 50 gonad arms, excising each just proximal to the spermatheca. Dissected gonads were carefully isolated from carcasses and transferred into an Eppendorf tube. Total RNA from each sample (∼100 ng) was isolated using Trizol (Invitrogen) and amplified with T7 RNA polymerase using one round of linear amplification (Baugh et al.,2001). Three independent dpl-1 and lin-35 mutant samples and four independent efl-1 mutant samples were collected.
Fluorescence-labeled cDNA probe for DNA microarray hybridization was prepared from 3 μg of amplified RNA as described(DeRisi et al., 1997). lin-35 and efl-1 mutant cDNA was labeled with Cy3 and compared to wild-type (N2) cDNA labeled with Cy5. Cy3-labeled unc-4;dpl-1 mutant cDNA was compared to unc-4cy5-labeled cDNA. Caenorhabditis elegans whole genome microarrays were used for hybridization as described(Jiang et al., 2001). Each slide was scanned using an Axon scanner (Molecular Devices, Sunnyvale, CA),and the expression levels for each gene in each channel were collected using GenePix 3.0 software. Cy5/Cy3 ratios were calculated and normalized by setting the overall median of ratios to one. All data have been deposited in GEO under Accession Number GSE5071.
For each set of mutant data, the repeats were averaged, and a Z test[Z=(observed-expected)/SE] was performed in Excel. A moderate correction for multiple testing (∼17,600 genes) was performed by multiplying the calculated P-value by 10,000. After this correction, all genes with up- or downregulation greater than twofold, P<0.05 in any given mutant were selected. The hypergeometric probability test(http://elegans.uky.edu/MA/progs/overlap_stats.html)was used to calculate the significance of overlap of gene groups. We determined whether transcripts of Group I-IV genes are bound by GLD-1, based on a minimum criteria of >1.5× enrichment in GLD-1 immunoprecipitated samples compared to control immunoprecipitations (P<0.01). Genes in Groups III and IV had very little overlap with GLD-1-bound transcripts(1/42 and 1/84, respectively), while genes in Groups I and II had a significantly enriched overlap with candidate GLD-1 targets (26/74 and 5/43,respectively).
Regulatory motif analysis
To identify candidate regulatory sequences in the 5′ noncoding regions of target genes, the online program MEME (Mutiple Em for Motif Elicitation) http://meme.sdsc.edu/meme/intro.html)was applied to sequences upstream taken from the start codon of each target gene to the neighboring gene, up to 1 kb. Each group of genes (Groups I-IV)was examined separately, as well as two control sets of genes, one with oogenic germline-enriched expression not regulated by EFL-1/DPL-1, and one that does not show germline-enriched expression. Manual examination was performed by taking all variations of the MEME-derived consensus motif [e.g. TTC(G/C)CGC(C/G)] and searching through each 5′ regulatory sequence for an exact match. Manual examination of upstream regions of those Group I genes in which MEME failed to find the E2F consensus motif did not uncover any additional instances of the motif. The motif sequence logo was created using the online program(http://weblogo.berkeley.edu/logo.cgi).
RME-2 and MEX-5 localization was performed as described(Kelly et al., 2002). Briefly,gonads were dissected as described above, fixed in 3.7% paraformaldehyde, and mounted on a slide. The slides were frozen on a dry ice block and the coverslip cracked off before storing the slides in -20°C methanol. The slides were washed three times in phosphate-buffered saline (PBS), blocked in PBS containing 0.1% Tween 20 and 0.5 mg/ml BSA, and then incubated at 4°C overnight with α-RME-2 (1:100; gift from B. Grant), α-MEX-5(undiluted; gift from J. Priess) and α-LIN-3 (1:50; Santa Cruz Biotechnology, Santa Cruz, CA). The samples were then stained with DAPI,washed and incubated at room temperature for 3 hours with a fluorescent secondary antibody (Molecular Probes, Carlsbad, CA). After further washing in PBS, the slides were mounted with anti-fade solution and viewed using a Zeiss Axioplan 2 imaging epifluorescence microscope.
Group I genes with oogenic germline-enriched expression that were reported to have embryonic lethal RNAi phenotypes (25) were tested for PIE-1:GFP mis-localization. Of these, ten gave embryonic lethality in our hands:T22F3.3, T05G5.7, spn-4, rnr-2, rme-2, R05H5.3, puf-3,pos-1, C28C12.2 and cyb-3. mex-5 and mex-6 were used as controls. PCR using gene-specific primers containing T7 sequences was followed by in-vitro transcription using T7 RNA polymerase. The resulting dsRNA was ethanol precipitated with Pellet Paint (Novagen) and resuspended to a concentration of ∼2 mg/ml. pie-1::GFP worms(Reese et al., 2000) were synchronized and grown until the L4 stage. Worms were washed in M9 and soaking buffer (0.25× Mg++-free M9, 3 mmol/l spermidine, 0.05%gelatin) twice before soaking. Approximately 30 worms in 2 μl soaking buffer were added to 2 μl dsRNA for each gene, incubated for 24 hours at 20°C, and then plated to NGM seeded with OP50 bacteria. For each gene giving embryonic lethality, embryos released from the uterus of ∼10 mothers were examined using a Zeiss Axioplan 2 microscope. Cross-RNAi between mex-5 and mex-6 transcripts probably occurred, resulting in PIE-1 mis-localization (Schubert et al.,2000).
The C. elegans germline
Each lobe of the bi-lobed hermaphrodite germline is arranged along a distal-to-proximal axis, with germ cells progressing from an undifferentiated mitotic state to fully differentiated gametes along this axis[Fig. 1; reviewed by Hubbard and Greenstein (Hubbard and Greenstein,2005)]. During the fourth larval stage, spermatogenesis occurs in the proximal germline. Upon the onset of adulthood, oogenesis then initiates in the proximal germline. Early stage oocytes enlarge substantially by incorporating cytoplasm from the central core of the germline and taking up yolk proteins that are synthesized by the intestine. Oocyte cytoplasm contains many maternally provided mRNAs and proteins that are necessary for proper embryonic development. The most proximal oocyte undergoes maturation and ovulation, becoming fertilized when it comes into contact with sperm in the spermatheca. Fertilization triggers rapid eggshell formation and completion of meiosis. The newly fertized embryos pass from the spermatheca into the uterus,where they are held briefly before being laid.
dpl-1 and efl-1, but not lin-35, are required for fertility
Immunolocalization studies show that DPL-1 protein is present in all germline nuclei except sperm (Ceol and Horvitz, 2001). EFL-1 protein is also present in germline nuclei,but its expression is detected only in the medial region, when nuclei are in the pachytene stage of meiosis I (Page et al., 2001) (Fig. 1). Partial loss-of-function point mutations in either dpl-1 or efl-1 result in temperature-sensitive maternal-effect embryonic lethality due to defects in embryonic asymmetry(Page et al., 2001). Additionally, Page et al. (Page et al.,2001) demonstrated using genetic mosaic analysis that loss of efl-1 activity in the germline, but not the soma, was sufficient to cause this embryonic lethality. To identify the primary defect upon complete loss of dpl-1 and efl-1 activity, we examined animals bearing probable null alleles in dpl-1 or efl-1. The dpl-1(n3316) allele contains a deletion that removes the first half of the protein, including the EFL-1-binding domain. The efl-1(n3639)allele harbors an early stop mutation, truncating the protein upstream of the predicted DP and pRB-binding domains (Ceol and Horvitz, 2001).
The dpl-1(n3316) and efl-1(n3639) mutant phenotypes are very similar: germ cells of all stages are present and the gonad has essentially normal morphology, but adults display nearly 100% penetrant sterility (i.e. do not lay fertilized embryos) (Coel and Horvitz, 2001)(Fig. 2C,D). Most oocytes either do not enter or do not exit the spermatheca normally, and over time the proximal gonad in efl-1 and dpl-1 mutants fills up with degenerating oocytes that become endomitotic. The occasional oocyte that does pass through the spermatheca frequently fails to develop a complete eggshell,and does not undergo normal cell division. The efl-1(n3639) and dpl-1(n3316) phenotypes persist even when exogenous sperm is provided from wild-type males, indicating that the phenotypes are not due solely to defective spermatogenesis.
Aberrant ovulation and the endomitotic (Emo) phenotype can be caused by defects in either the oocyte or the somatic gonad(McCarter et al., 1997; McCarter et al., 1999). To test the site of action of dpl-1 and efl-1, we injected dsRNA corresponding to dpl-1 or efl-1 into wild-type and rrf-1 mutants, which can carry out RNAi in the germline but not the soma (Sijen et al., 2001). Relative to wild-type controls, rrf-1;efl-1(RNAi) and rrf-1;dpl-1(RNAi) animals displayed a decreased incidence of the Emo phenotype. This observation suggests that efl-1 and dpl-1 do play a role in the soma, possibly in the somatic gonad, to regulate ovulation (data not shown). However, rrf-1;efl-1(RNAi) and rrf-1;dpl-1(RNAi)animals were still completely sterile and collected partial, torn oocytes in the uterus, indicating that defects in ovulation, though more subtle, still occurred. This observation suggests a role for efl-1 and dpl-1 in the germline to promote fertility. In sum, the null phenotype for both efl-1 and dpl-1 in the germline is more severe than in previously studied point mutants(Page et al., 2001).
The lin-35(n745) mutation truncates the 961-amino acid LIN-35 protein after the first 150 amino acids, removing the A and B pocket domains,and probably acting as a null (Lu and Horvitz, 1998). The germline of lin-35(n745) mutants generally appears morphologically normal(Fig. 2B). Unlike dpl-1(n3316) and efl-1(n3639) mutants, lin-35(n745)animals are fertile and do not display obvious defects in oogenesis or ovulation. However, the brood size of lin-35(n745) mutants is 41% of wild type (132±13 versus 320±42; Student's t test: P<<0.001).
EFL-1 and DPL-1 regulate a common set of genes and LIN-35 regulates a distinct set of genes
Because EFL-1, DPL-1 and LIN-35 are predicted to act as transcriptional regulators, we used whole-genome C. elegans DNA microarrays to identify candidate target genes. To focus on the role of EFL-1, DPL-1 and LIN-35 in the germline, we isolated gonads away from the rest of the animal. For each mutant or control sample, we collected ∼50 dissected adult gonads, carefully discarding the spermatheca and uterus as well as the rest of the carcass. The only somatic cells retained were the distal tip cell and gonadal sheath cells, ensuring that the vast majority of the isolated RNA derived from germ cells. Because the germline morphology of efl-1(n3639),dpl-1(n3316) and lin-35(n745) mutants are visually similar to wild type, non-specific effects on gene expression should be minimal.
We used linear amplification to increase the amount of RNA for all samples,and reverse transcribed the RNA to labeled cDNA. Using microarrays containing∼90% of predicted C. elegans genes(Jiang et al., 2001), we performed at least three hybridizations directly comparing each mutant to a control. The resulting data were analyzed using both an average fold-difference and a Z-test with correction for multiple testing(see Materials and methods). We have focused on 272 genes whose expression is either up- or downregulated in efl-1, dpl-1 or lin-35mutants relative to controls, exceeding a twofold difference at P<0.05 (see Table S1 in the supplementary material). Throughout this paper, we will use the term `downregulated' to refer to target genes with lower expression in dpl-1, efl-1 and/or lin-35 mutants relative to controls, and `upregulated' to refer to genes with higher expression in any mutant relative to controls.
We first examined the overlap and independence between the different mutant datasets (Fig. 3). Our results show that DPL-1 and EFL-1 both function primarily to promote gene expression. At our twofold cutoff, 116 genes were downregulated and 30 genes were upregulated in dpl-1 mutants. Similarly, in efl-1 mutants,114 genes were downregulated and 20 genes were upregulated. Remarkably, the two mutants showed downregulation of 70% of the same target genes (95/135; P<<0.001) (Fig. 3A). This extensive overlap is consistent with these two proteins acting as a heterodimer to regulate transcription. Only a subset (36) of the 95 genes downregulated in efl-1 and dpl-1 mutants were also downregulated in lin-35 mutants. Instead, distinct genes were primarily upregulated in lin-35 mutants; strikingly few of these were also regulated in either efl-1 or dpl-1 mutants(Fig. 3B). The above observations suggest that EFL-1/DPL-1 can function without LIN-35 to regulate expression of many target genes and, conversely, that LIN-35 often acts independently of EFL-1/DPL-1. When LIN-35 does act on genes regulated by EFL-1/DPL-1, it cooperates with, rather than antagonizes, EFL-1/DPL-1(Fig. 3A).
Examining the overlap among preselected gene groups can exclude genes that come close but fail to meet selection criteria, partially disguising general trends. We therefore used hierarchical clustering to place the 272 genes into groups with similar regulation (Eisen et al., 1998) (Fig. 4). From this analysis, we defined four expression groups that encompass 248 of the 272 genes; the remaining 24 genes showed variable expression and were not included in a group(Fig. 4; see Table S1 in the supplementary material). Group I comprised 75 genes with decreased levels in dpl-1(n3316) and efl-1(n3639) mutants compared with controls and were termed `downregulated in efl-1 and dpl-1'. Sixty-five out of 75 genes in this group were regulated more than twofold(P<0.05) by both EFL-1 and DPL-1. Seven of the remaining ten genes just missed the twofold cutoff in one mutant and the remaining three lacked data for one mutant. Thus these genes are almost uniformly under the control of both EFL-1 and DPL-1, suggesting that they are targets of an EFL-1/DPL-1 heterodimer. Only 16 out of the 75 genes in Group I were significantly downregulated in lin-35 mutants, although many more were moderately affected.
Group II genes (43 genes) generally had lower expression in efl-1,dpl-1 and lin-35 mutants than in control gonads. Twenty-three genes were significantly regulated in all three mutants, and 36 were regulated in at least two mutants, so we call Group II genes `commonly downregulated'. Conversely, genes in Group III (42 genes) had higher expression in one or more mutants relative to controls. However, only five were significantly upregulated in all three mutants and only 17 were upregulated in at least two mutants, so we termed this group `variably upregulated'. Group IV contains 88 genes with significantly increased levels only in the lin-35 mutant relative to controls; we referred to Group IV genes as `upregulated in lin-35'. From these data, we conclude that efl-1 and dpl-1 mutants show strong similarity in gene regulation over all four groups, even if a given gene did not surpass the selection criteria in one mutant, reinforcing the notion that EFL-1 and DPL-1 are likely to act as a heterodimer and regulate common targets. lin-35 shared common targets with efl-1 and dpl-1 for one of the four groups (Group II),but showed distinct regulation in the other three groups. Most notably, lin-35 had minimal effect on Group I genes regulated by efl-1 and dpl-1, and instead regulated expression of a large set of genes not affected by loss of efl-1 or dpl-1 (Group IV).
A consensus E2F-binding site is present upstream of EFL-1/DPL-1-activated genes
To identify candidate regulatory sequence motifs for each group, we examined the sequence upstream of each open reading frame (maximum 1 kb) using the MEME program, which identifies over-represented sequences de novo(Bailey and Elkan, 1994). The most significant hit among any of the groups was the sequence TTC(G/C)CGC(C/G)(P<<0.001), found upstream of 58 out of 75 Group I(downregulated in efl-1 and dpl-1) genes(Fig. 5A; see Table S1 in the supplementary material). This sequence bears striking similarity to the known mammalian E2F consensus site TTTCGCGC(Chittenden et al., 1991),suggesting that most Group I genes are likely to be direct targets of the EFL-1/DPL-1 transcription factor. The TTCGCGCC motif was present within the first 500 base pairs upstream of the translation start in 48 genes. Multiple copies of the motif were present in 28 genes, and the duplicate sites were often juxtaposed (Fig. 5B). MEME did not identify the E2F consensus motif in Groups II, III or IV, in a control set of oogenic germline-enriched genes not responsive to EFL-1, DPL-1 or LIN-35 (n=38), or in a control set of genes that did not show germline-enriched expression (n=51).
Our ability to find an E2F consensus site in Group I genes but not the other three groups suggests that most genes in Groups II-IV are not regulated by direct binding of an E2F heterodimer, but are either indirect or downstream. To determine whether at least some genes in Groups II, III and IV might be direct targets, we examined their 5′ regulatory sequence manually and identified 13 Group II, 8 Group III and 30 Group IV genes with a C. elegans consensus E2F site (see Table S1 in the supplementary material). Thus, it is possible that the regulatory regions upstream of certain genes in these groups are directly bound by EFL-1/DPL-1.
MEME detected a different motif, TTTTCCAG, in the regulatory regions of Group IV `upregulated in lin-35' genes (P<<0.001; Fig. 5C). This consensus sequence was present in 58/88 regulatory sequences, but its location was not biased toward the translation start site(Fig. 5D). We could find no clear match to any known transcription factor consensus sequence for this motif in any database. No significant motifs were identified in Group II or Group III genes.
The germline expression pattern of downregulated genes differs from upregulated genes
We examined the spatial expression patterns of genes in groups I-IV using data from an online in-situ hybridization database (Y. Kohara, personal communication; http://nematode.lab.nig.ac.jp),which has images of expression patterns available for 151 of the 248 genes(see Table S1 in the supplementary material). Images are available for 54 Group I `downregulated in efl-1 and dpl-1' genes, and 47 of these show staining in the germline. Notably, 37 genes have undetectable levels in the distal germline with an abrupt increase in the medial germline that persists proximally, often into embryos(Fig. 6A-D). Onset of expression coincides with the appearance of EFL-1, which is detectable only in the mid-pachytene region of the medial germline(Page et al., 2001) (see Fig. 1). Fewer Group II`commonly downregulated' genes show detectable expression in the germline based on in-situ data (12/26); however, 10 of 12 were expressed in a medial/proximal restricted expression pattern, similar to Group I genes. Thus,germline expression of most genes downregulated in the mutants (Groups I and II) is generally restricted to the medial and proximal gonad, coincident with peak EFL-1 protein levels.
Conversely, most Group III and IV genes are broadly expressed in the wild-type germline. Out of 46 Group IV `upregulated in lin-35' genes with in-situ images, 35 have detectable germline expression, with 32 visible in both the distal and proximal germline and three restricted to the medial/proximal germline (Fig. 6E-H). Only 6/16 Group III genes show detectable germline expression, but four of these have broad expression and two show medial/proximal restricted expression. Given that our dissected gonad microarray data indicates that these genes are normally at lower levels in the wild-type germline than in the lin-35 mutant germline, this broad germline expression was unexpected. One possibility is that LIN-35 acts broadly to decrease, but not abrogate, expression of these genes, because elevated levels are slightly detrimental to germline function. Another possibility is that LIN-35 acts to repress expression of these genes in a specific region of the germline, but that this effect is not substantial enough to be visible by in-situ hybridization methods, which are difficult to quantify. Finally, we could be detecting LIN-35-mediated repression of these germline-expressed genes in the few somatic cells that are included in our analysis (distal tip and sheath cells) rather than in germ cells, which would be difficult to distinguish by in situ. This last possibility is consistent with the observation that lin-35 prevents ectopic germline gene expression in the soma (Wang et al.,2005).
EFL-1 and DPL-1 are required for normal expression of RME-2 and MEX-5
We independently tested whether two Group I genes, rme-2 and mex-5, were regulated in a manner consistent with the microarray results. rme-2, which encodes the yolk receptor(Grant and Hirsh, 1999), and mex-5, which encodes a CCCH zinc finger protein required for correct embryonic polarity (Schubert et al.,2000), both require efl-1 and dpl-1 activity for their expression, based on our microarray experiments. lin-35 has a moderate effect on mex-5 but not rme-2 expression (see Table S1 in the supplementary material). We performed semi-quantitative RT-PCR of both genes in wild type, efl-1, dpl-1 and lin-35 mutants and saw decreased expression of both rme-2 and mex-5 in efl-1 and dpl-1 mutants relative to wild type, with minimal effects in lin-35 mutants, consistent with our microarray data (see Fig. S1 in the supplementary material). We also performed immunohistochemistry to determine whether localization and expression level of RME-2 and MEX-5 differs in efl-1, dpl-1 and lin-35 mutants compared to wild type (Fig. 7). In wild-type gonads, RME-2 localizes to the oocyte membrane. We saw similar levels and localization in lin-35 mutants, but efl-1 or dpl-1mutants displayed severely reduced RME-2 levels, even at tenfold higher exposure times, consistent with a role for EFL-1 and DPL-1 in promoting expression of rme-2.
MEX-5 showed a slightly different response from RME-2 to loss of efl-1 and dpl-1. In wild type, MEX-5 began to accumulate in the cytoplasm of late pachytene germ cells, increasing markedly in oocytes(Fig. 7B). In lin-35mutants, MEX-5 expression appeared very similar to wild type. In efl-1 and dpl-1 mutants, no MEX-5 expression was evident in the cytoplasm of pachytene nuclei or immature oocytes, but was still present in the most proximal oocytes (Fig. 7H,K). This result suggests that EFL-1 and DPL-1 are important for increasing levels of mex-5 expression, but are not solely responsible for its activation. This finding is consistent with a previous observation that MEX-5 was still present in efl-1(se1ts) or dpl-1(zu355)mutant embryos (Page et al.,2001). Thus, for both Group I genes we tested, we saw decreased expression in both efl-1 and dpl-1 mutants, but not in lin-35 mutants, relative to wild type, consistent with the microarray results.
To test whether the reduced expression of RME-2 and MEX-5 seen in efl-1 and dpl-1 mutants was a downstream, non-specific effect of decreased oogenesis or oocyte maturation, we examined LIN-3. lin-3 is expressed during oogenesis, but is not differentially regulated in lin-35, dpl-1 or efl-1 mutants in our microarray experiments. In wild type, LIN-3 is localized to the oocyte membrane (Fig. 7C), while in lin-3(RNAi) animals this staining is absent (see Fig. S2 in the supplementary material). LIN-3 localization was not altered in lin-35,dpl-1 or efl-1 mutants relative to wild type, suggesting that oogenesis genes not identified in our microarray study are probably still expressed normally in dpl-1 and efl-1 mutants.
Group I genes promote oogenesis and early embryogenesis
Studies in mammalian cells have focused on the role of E2F in regulating genes important for promotion of the cell cycle, particularly those acting in DNA synthesis (reviewed by Trimarchi and Lees, 2002). However, recent global expression studies in mammalian and Drosophila cell culture have found that various E2Fs can activate genes that act at other phases of the cell cycle, as well as repressing genes that act to promote differentiation(Ren et al., 2002; Dimova et al., 2003). The complexity of the Rb/E2F network has made it difficult to perform and interpret in-vivo studies, leaving a gap in our understanding of how accurately the transcriptional properties of E2F in culture reflect its role in vivo.
In our experiments, we found that EFL-1/DPL-1 primarily promotes expression of pro-differentiation genes in the wild-type germline(Table 1). Group I`downregulated in efl-1 and dpl-1' genes include only a few likely to have direct roles in the meiotic and mitotic cell cycle, such as a ribonucleotide reductase subunit rnr-2, and two cyclin B orthologs, cyb-3 and cyb-2.1(Sonneville and Gonczy, 2004)(www.wormbase.org). Four previously studied targets encode proteins with decreased expression that is likely to contribute to the defects in fertilization and eggshell formation seen in dpl-1(n3316) and efl-1(n3639) mutants. These include the yolk receptor RME-2, the Tis11-like oocyte maturation protein OMA-2, the predicted chitin-binding protein CEJ-1, and an EGF-related protein required for fertilization, EGG-2 (Grant and Hirsh,1999; Detwiler et al.,2001; Lee and Schedl,2001; Kadandale et al.,2005). In particular, rme-2 mutants display ovulation defects and have a reduced brood size as a consequence(Grant and Hirsh, 1999). Thus,decreased rme-2 expression is probably a significant component of the dpl-1 and efl-1 phenotypes.
|Oocyte development .||Metabolism/structural .||Cell division .||Embryonic asymmetry .|
|Oocyte development .||Metabolism/structural .||Cell division .||Embryonic asymmetry .|
Other Group I `downregulated in efl-1and dpl-1' genes act to establish polarity and cell fate determination in the early embryo, such as spn-4, mex-5, mex-6 and mex-1. pos-1 is also regulated by EFL-1/DPL-1, although it is not in Group I. SPN-4, POS-1 and MEX-1 are localized to posterior blastomeres in the early embryo and are important for establishing asymmetry in the early cell divisions. Additionally, SPN-4 and POS-1 act together in the posterior blastomeres to negatively regulate translation of the glp-1 mRNA(Ogura et al., 2003). MEX-5 and MEX-6 are localized to anterior blastomeres and also establish embryonic asymmetry (Schubert et al.,2000). Loss of pos-1, mex-5/6 or mex-1 activity results in ectopic localization of the germline-specific PIE-1 protein in somatic cells, either through mis-specification of the germ lineage or through a failure to degrade PIE-1 in somatic blastomeres, and results in embryonic lethality (Schubert et al.,2000; Tabara et al.,1999; Guedes and Priess,1997).
Because four genes downregulated in efl-1 and dpl-1mutants are known to affect PIE-1 localization, we tested several other target genes for their requirement in correctly localizing a PIE-1:GFP fusion protein(Reese et al., 2000). We examined a subset of genes with reported RNAi phenotypes that were embryonic lethal in large-scale screens. Out of ten genes tested, we found one, T22F3.3,with inactivation by RNAi that results in persistence of PIE-1 in somatic blastomeres at a frequency similar to the rate of embryonic lethality (9%)(see Fig. S3 in the supplementary material). T22F3.3 encodes the only obvious glycogen phosphorylase in the C. elegans genome and presumably plays a crucial role in modulating glucose availability. Mutations in human glycogen phosphorylase cause McArdle disease, which impairs muscle function(Dimaur et al., 2002). Alterations in available energy levels could conceivably impact some of the processes required for timely degradation of PIE-1 in the soma.
Although Group II genes share some common expression characteristics with Group I genes as described above, they generally encode different types of proteins. For instance, Group II genes include seven that encode histones H2A,H2B or H4, but none encoding H3 or H1. In human cells, the promoters of several histone H2 genes are also bound by E2F4(Ren et al., 2002). RNAi of several Group II genes produces embryonic lethal phenotypes based on large-scale studies, including those encoding histones, the translation elongation factor EFT-2, and chaperone proteins such as HSP-3 and PDI-2.
The predicted protein functions of the genes in Groups III and IV are heterogeneous, and obvious trends are not readily apparent. Five Group III genes show significant upregulation in all three mutants, including akt-2, a kinase that acts downstream of insulin signaling, toh-2, a metalloprotease, and cdr-4, a cadmium-responsive glutathione-S-transferase. Group IV `upregulated in lin-35' genes include several that encode transcriptional regulatory proteins, such as mes-3, dro-1 and ntl-1, as well as four uncharacterized proteins with histone acetyltransferase domains. Genes encoding the DNA synthesis licensing factor CDT-1, the anti-apoptotic factor CED-9 and the Frizzled receptor MOM-5 are also repressed by LIN-35 in the gonad. Thus, the upregulated genes encode proteins of a wider variety of functions than genes downregulated in efl-1 and dpl-1 mutants.
The connection between the transcriptional activity of E2F and its in-vivo role in differentiation is one of the most poorly understood aspects of E2F function. Using the germline of C. elegans as a model system, we investigated the function of the core pRb/E2F pathway and uncovered an unexpected role for the EFL-1/DPL-1 (E2F) heterodimeric transcription factor in vivo. Identification of genes regulated in efl-1, dpl-1 and lin-35 mutant gonads indicates that EFL-1/DPL-1 probably functions as a transcriptional activator, while LIN-35 has a repressive role. Importantly,genes upregulated by EFL-1 and DPL-1 (Group I) have canonical E2F-binding sites in their 5′ regulatory sequences, and their expression is first detectable in the same region of the germline at which EFL-1 reaches peak levels, indicating that they are probably direct targets of EFL-1/DPL-1. They encode proteins that promote oogenesis, fertilization and early embryogenesis,rather than cell cycle regulators. Thus, even though EFL-1 most closely resembles mammalian repressor E2F4, our results demonstrate that E2F directly promotes development through activation of pro-differentiation genes,suggesting a new mechanism by which E2Fs can function as tumor suppressors in vivo.
Limited compensation by Rb/E2F family members
The streamlined nature of the pRb/E2F pathway in C. elegansreduces the possibility of compensation by other family members. The dpl-1 locus encodes the only DP-related protein in the C. elegans genome; thus, loss of dpl-1 should remove all E2F activity, as most E2Fs require a DP subunit to regulate gene expression. Our results show that dpl-1 does not regulate many genes in addition to those regulated by efl-1. This extensive overlap of dpl-1and efl-1 in regulating germline gene expression strongly suggests that EFL-2 or F49E12.6 do not compensate for loss of EFL-1 and that EFL-1 is the major component of E2F activity in the germline.
lin-35 is the only pocket protein encoded in the worm genome and clearly has a distinct role in regulating a distinct set of genes (Group IV). However, additional processes in which LIN-35 participates could be buffered by the action of other proteins. In somatic tissues, genes of diverse functions have exhibited redundancy with lin-35, including components of the SynMuv A or C pathways, the cell cycle inhibitor cki-1, and a regulator of APC activity, fzr-1(Ferguson and Horvitz, 1989; Ceol and Horvitz, 2004; Boxem and van den Heuvel, 2002; Fay et al., 2002). Additionally, a mutation of lin-35 enhances the meiotic recombination defect of him-17 mutants in the germline(Reddy and Villeneuve, 2004). Thus LIN-35 could have a role in germ cell division or differentiation that is not apparent in lin-35 mutants because of redundancy with other factors.
Relationship between EFL-1, DPL-1 and LIN-35 in the germline
Based on both genetic and biochemical evidence, components of the C. elegans Rb pathway act together to repress gene expression in somatic tissues (reviewed by Kipreos,2005). In both C. elegans and Drosophila, recent work has demonstrated that genes normally expressed in the germline are repressed in the soma by pRB and E2F(Dimova et al., 2003; Wang et al., 2005). Our study is the first in-vivo demonstration of distinct roles for E2F and pRB in regulating gene expression in the germline. We have shown that E2F mainly functions to activate, rather than repress, genes important for oocyte and embryo differentiation (Group I genes), and that the role of LIN-35(pRB) in this process is minimal or dispensable. Possibly, LIN-35 dissociates from the E2F complex in the germline, freeing E2F to activate target gene expression. E2F activity could also be controlled by accumulation of EFL-1 protein at the mid-pachytene stage of meiosis I rather than by association of LIN-35. Notably, in Drosophila, chromatin immunoprecipitation studies demonstrate that both pRB and E2F can be found at the promoters of genes whose expression is not dependent on pRB (Dimova et al., 2003; Stevaux et al.,2005). By analogy, in the C. elegans germline, LIN-35 could be present at the promoters of Group I genes but not be rate-limiting for their expression.
LIN-35 instead acts to downregulate a distinct set of genes (Group IV) that do not require EFL-1 or DPL-1 for their expression. The absence of canonical E2F binding sites from the 5′ regulatory regions of most Group IV genes suggests that LIN-35 is targeted to these sites through the activity of a different DNA-binding factor, one that potentially binds the TTTTCCAG site that we found highly represented among Group IV genes. Several instances of an E2F-independent function of pRB have been described(Sellers et al., 1998; Thomas et al., 2001; Gagrica et al., 2004). Alternatively, LIN-35 could be acting with either EFL-2 or F49E12.6 in a DPL-1-independent manner that does not require an E2F consensus site. Finally,the increased expression of these genes upon loss of LIN-35 activity could be indirect or occurring in the few somatic cells present in our samples. Additional experiments will be necessary to distinguish among these possibilities.
Tissue-specific EFL-1/DPL-1 transcriptional program
Our results demonstrate that EFL-1/DPL-1 can display tissue specificity in both target gene selection and in the manner of gene regulation (activation or repression). In somatic tissues, EFL-1/DPL-1 represses genes such as fkh-6, mat-3 or cye-1 in various cell types to control the timing and nature of cell division (Chang et al., 2004; Garbe et al.,2004; Tilmann and Kimble,2005; Grishok and Sharp,2005). However, in the germline, EFL-1/DPL-1 does not regulate these target genes, nor is it crucial for proliferation of the germline stem cell population. Instead, EFL-1/DPL-1 activates the expression of a distinct set of genes whose protein products participate in oocyte differentiation and embryogenesis. Indeed, many EFL-1/DPL-1 target genes such as rme-2and mex-1 have known roles only in oogenesis and embryogenesis and are likely to be expressed specifically in the maternal germline. The presence of canonical E2F-binding sites in these promoters strongly indicates that their regulation by EFL-1/DPL-1 is direct, although the mechanisms establishing the EFL-1/DPL-1 germline-specific program are unknown. Possibly,unknown germline-specific transcriptional coactivators or chromatin conformation might influence the activity of EFL-1/DPL-1.
Coordination of gene expression with developmental events
Several transcripts expressed in the medial germline are translationally repressed by the mRNA-binding protein GLD-1, including two that we identified as EFL-1/DPL-1 targets (rme-2 and cej-1)(Jones and Schedl, 1995; Lee and Schedl, 2001). We have generated an expanded list of candidate GLD-1 targets by identifying mRNAs that co-immunoprecipitate with GLD-1 and probing microarrays (M.-H. Lee, V.R. and T. Schedl, unpublished). Candidate GLD-1 target transcripts had significantly enriched overlap with Group I and II genes (26/74 and 5/43,respectively; see Materials and methods). Group I transcripts bound by GLD-1 include several with roles in oogenesis and embryogenesis, such as rme-2,cej-1, oma-2 and egg-2, as well as spn-4, cyb-2.1 and cyb-3. Thus, the germline has established a multi-tier mechanism for ensuring that the expression of these target genes is properly coordinated with oocyte development. First, detectable EFL-1 protein is restricted to the medial germline, which limits the significant accumulation of the target transcripts before the pachytene stage of meiosis I. Second, once expressed,many of these transcripts are probably translationally suppressed by GLD-1 until oogenesis. This dual regulation ensures that crucial components of oogenesis are available but held inactive until germ cells are at the appropriate stage of development.
Our results underline the importance of examining gene expression profiles in vivo in order to gain an accurate understanding of the natural function of transcription regulators. The primary role of EFL-1 in the C. elegansgermline is to coordinately activate a cohort of genes required for the final stages of oocyte development and particular aspects of early embryonic development, suggesting that E2F activity in vivo adapts to tissue-specific environments. Future functional studies of target genes will help us to better understand how a transcription factor coordinates the transition from gamete to embryo.
We thank Yuji Kohara and co-workers for permitting us to include images from their database(http://nematode.lab.nig.ac.jp). We also thank T. Schedl and M.-H. Lee for permitting the discussion of unpublished data, and J. Priess, B. Grant, G. Seydoux and H.R. Horvitz for strains and reagents. We are grateful to Michelle Kudron for RNAi injections. We have benefited greatly by comments on the manuscript from I. Greenwald and members of the Reinke laboratory. Some strains were provided by the C. elegans Genetics Center. This work was supported by a grant from the NIH to V.R. (GM065682).