Report

Similarity of the C. elegans Developmental Timing Protein LIN-42 to Circadian Rhythm Proteins

See allHide authors and affiliations

Science  05 Nov 1999:
Vol. 286, Issue 5442, pp. 1141-1146
DOI: 10.1126/science.286.5442.1141

Abstract

The Caenorhabditis elegans heterochronic genes control the relative timing and sequence of many events during postembryonic development, including the terminal differentiation of the lateral hypodermis, which occurs during the final (fourth) molt. Inactivation of the heterochronic gene lin-42 causes hypodermal terminal differentiation to occur precociously, during the third molt. LIN-42 most closely resembles the Period family of proteins fromDrosophila and other organisms, proteins that function in another type of biological timing mechanism: the timing of circadian rhythms. Per mRNA levels oscillate with an approximately 24-hour periodicity. lin-42 mRNA levels also oscillate, but with a faster rhythm; the oscillation occurs relative to the approximately 6-hour molting cycles of postembryonic development.

The timing of distinct biological processes within a cell or organism is carefully controlled. One class of temporal regulators, exemplified by the C. elegans heterochronic genes, times the onset of developmental events. These genes control the relative timing of diverse stage-specific events during postembryonic development such as dauer larva formation, vulva formation, and the terminal differentiation of the hypodermis (1). In other organisms, members of this general heterochronic gene class include theTeopod genes and Glossy15 of maize,hasty of Arabidopsis, and anachronismof Drosophila (2). Mutations in these genes either advance or retard expression of certain developmental programs.

A second class of temporal regulators comprises the “clock” components that regulate circadian rhythms, the approximately 24-hour periodicity of biological processes such as sleep-wake cycles in humans (3). Genes that control circadian rhythms have been cloned from several organisms, including Drosophila,Neurospora, and mouse. Although shared motifs have been observed between certain circadian rhythm proteins, shared motifs between these proteins and the timing factors that control developmental progression have not been reported previously. Here we demonstrate that the C. elegans heterochronic genelin-42 encodes a protein with similarity to theDrosophila period protein (PER) and its counterparts in other organisms. This discovery provides a molecular connection between these two types of cellular timekeepers.

Caenorhabditis elegans proceeds through four postembryonic larval stages (L1 to L4) on the way to adulthood. During the fourth (final) molt, the lateral hypodermal seam cells terminally differentiate; they exit the cell cycle and secrete a morphologically distinct adult cuticle. Mutations in the heterochronic genes alter the timing of this event. Loss-of-function mutations in the heterochronic genes lin-14, lin-28, and lin-42 cause precocious phenotypes in which seam cell terminal differentiation is executed prematurely, whereas mutation of lin-4 orlin-29 retards this differentiation event (1,4). These latter mutants undergo additional rounds of ecdysis during which the hypodermis remains undifferentiated. These genes form a negative regulatory pathway that restricts seam cell terminal differentiation to the fourth molt (5). The transcription factor LIN-29 is the most direct trigger of the switch to the adult hypodermal program; the remaining genes act upstream, ensuring that lin-29 activity is correctly timed. We cloned and characterized lin-42 to better understand its role in this developmental timing mechanism.

Genetic mapping using a combination of visible markers and restriction fragment length polymorphism (RFLP) markers placedlin-42 on the left arm of linkage group II (LGII) betweenveP2 and nP48, nearer to veP2 (Fig. 1A) (6). Because of the paucity of genetic or molecular markers in this region, we chose a transposon tagging approach to identify the lin-42 locus within the interval from veP2 to nP48(veP2-nP48). We searched for lin-42 alleles induced by insertion of the Tc1 transposon (7).lin-42 alleles can be isolated efficiently among mutations that restore adult cuticle synthesis in a lin-4 mutant background (8), where seam cell terminal differentiation would otherwise not occur. We identified 34 mutations that restore adult cuticle to lin-4 mutants, presumably as a result of Tc1-mediated gene disruptions. To identify candidate lin-42mutations among these, we performed a secondary screen for animals resembling previously analyzed lin-42 lin-4 double mutants. For example, lin-42 mutations do not suppress the vulvaless phenotype of lin-4 mutants or their inability to form dauer larvae (8). Mutants meeting these criteria were outcrossed to separate the new mutation from lin-4. Two of the outcrossed mutants contain new lin-42 alleles based on the following properties of the new mutations: (i) they cause precocious adult cuticle synthesis during the third molt, (ii) they show weak linkage to lin-4 [lin-42 and lin-4are approximately 12 centimorgans (cM) apart on LG II], and (iii) they fail to complement an existing lin-42 allele,ve16. We sequenced genomic DNA flanking the Tc1 element responsible for the lin-42 lesion in the ve27allele and used it to search the C. elegans genome sequence. In agreement with the genetic map location of lin-42, the sequence identified cosmid F47F6, which maps to theveP2-nP48 interval on LG II, nearer to veP2 (Fig. 1A).

Figure 1

Analysis of the lin-42 locus and alleles. (A) The top line represents the genetic map of the left arm of LGII. The placement of mab-9 between veP1 andveP2 is predicted from map distances. Cosmids identified by sequences flanking the lin-42 allele-specific Tc1,veP3, are indicated below the genetic map. The region corresponding to the lin-42 locus is expanded and shown relative to restriction sites Pst I (P), Eco RI (E), Eag I (Ea), and Sac II (S). Beneath the DNA line is the predicted structure of thelin-42 coding region. Exons are shown by open boxes.lin-42 is trans-spliced to SL1 20 nt 5′ to the translation initiation codon. @, summary of the transformation rescue experiments: + indicates that >90% of the L3 molt seam cells were wild type and thus lacked adult alae; − indicates that >90% of the seam cells were mutant and synthesized adult alae precociously. (B) Diagrams of lin-42::gfp fusions and a summary of their rescuing ability. The gfp open reading frame is indicated by the solid rectangle. unc-54 3′ UTR is indicated by the hatched rectangle. *, the size given for these constructs does not include gfp or unc-54 sequences. (C) The predicted amino acid sequence of the longest lin-42 open reading frame (35). (D) The effect oflin-42 lesions on the predicted lin-42 protein product. The locations of the lesions are indicated with respect to the first nucleotide of the cDNA (after the trans-spliced leader), which occurs 20 nt 5′ to the ATG. Triangles indicate a deletion of the subsequent base pairs. Arrows indicate a replacement with the indicated base pairs. AA, amino acid. (E) Domain organizations of CLOCK, PER, and LIN-42. The direct repeats of the PAS domain are indicated by black boxes labeled A and B. The gray areas flanking the direct repeats indicate the extents of the PAS domains. CLOCK contains a bHLH domain in its NH2-terminus (hatch marks) and a glutamine repeat (Q) in its COOH-terminus (36), and PER contains a glycine-threonine repeat in its COOH-terminus (TG). LIN-42 contains a basic region COOH-terminal to its PAS domain (dark shading).

Transformation with cosmid F20E9, which extensively overlaps F47F6 (Fig. 1A) (9), rescued the precocious phenotype oflin-42 mutants. The lin-42–rescuing activity was further delimited by the injection of F20E9 subclones and a polymerase chain reaction (PCR) fragment from the region. The smallest rescuing fragment was an 8.9-kb Sac II fragment that spanned the ve27Tc1 insertion site and contained a single large open reading frame based on GeneFinder predictions (9) (Fig. 1A). The Tc1 element was inserted into the fourth of five predicted exons, and it eliminated the COOH-terminal half of the 453–amino acid predicted protein.

To confirm that this gene corresponds to lin-42, we determined the DNA sequence of the predicted exons from four otherlin-42 alleles. Each of these mutations results in a COOH-terminal truncation of the predicted lin-42protein product (Fig. 1D). mg152 produces the most severe truncation, yielding a protein containing only the NH2-terminal 66 amino acids, and is likely to be a null mutation.

We confirmed the lin-42 gene structure by reverse transcription–PCR (RT-PCR) (10). Sequence analysis of RT-PCR products spanning overlapping portions of the five exons revealed that the GeneFinder prediction is accurate (Fig. 1).

Database searches reveal that the lin-42 protein is most similar to members of the PERIOD (PER) family of circadian rhythm proteins from insects and mammals. The most striking region of similarity includes a protein interaction domain, the PAS domain (11), which has recently come to be viewed as a signature feature of circadian rhythm proteins, including the insect and mammalian PER proteins, the WHITE COLLAR proteins ofNeurospora, and the CLOCK and BMAL proteins of mice and their Drosophila counterparts, dCLOCK and CYC (3). The PER PAS domain is an approximately 260–amino acid region containing two divergent hydrophobic direct repeats of about 50 amino acids, known as the PASA and PASB repeats (Fig. 2) (11,12). The region of highest similarity between LIN-42 and PER encompasses the PASB repeat (Fig. 2) and includes the cytoplasmic localization domain (CLD) of PER (13). The two proteins are 30% identical and 45% similar throughout this 139–amino acid region. The percent identity between LIN-42 and other PER family members in this region is also about 30% (29% for cockroach PER; 28% for human and mouse PER1). By comparison, human PER1 (hPER1) and Drosophila PER share 39% identity in this region. The similarity between vertebrate and invertebrate PER proteins is lower in the PASA repeat. Drosophila PER and LIN-42 each share 20% identity with hPER1 over the 50–amino acid PASA repeat (Fig. 2).

Figure 2

Sequence alignments comparing LIN-42 to the PER-related proteins from mammals, Drosophila, and cockroach. Amino acids identical in at least 50% of the proteins are indicated in red, and similarities are indicated in blue (35). The dPER/LIN42 line indicates conservation with emphasis on LIN-42 and dPER. Red letters indicate amino acids that are conserved among all six proteins. Black letters indicate identities, and single dots indicate similarities between LIN-42 and dPER. The PASA and PASB repeats (12) are indicated (arrows). Periplanta americana sequence is unavailable for the PASA region. When the GT repeat of Drosophila PER is omitted from BLAST or tFASTA searches, LIN-42 is the closest match to PER in the C. elegans genome. GenBank accession numbers are as follows: human PER1_hum, AF022991; mouse PER1_mus, AF022991; mouse PER2_mus, AF035830; mouse PER3_mus, AF050182;Drosophila PER_dros, P07663; American cockroachPeriplanta americana PER_peri, U12772; and LIN-42,AF183400.

The PAS domain was originally named for its presence in PER, the aryl hydrocarbon receptor nuclear translocator (ARNT), and theDrosophila single-minded protein (SIM) (11). Basic helix-loop-helix (bHLH) motifs are found NH2-terminal to the PAS domain in several PAS proteins, including ARNT, SIM, CLOCK, and BMAL (3). Drosophila PER and LIN-42 both lack bHLH domains in their NH2-termini. However, a basic region COOH-terminal to the LIN-42 PAS domain (Fig. 1C) is followed by stretches of predicted alpha helix based on PHDsec (14). Among all of the circadian and noncircadian PAS domain–containing proteins identified to date, the LIN-42 PAS domain most closely resembles that of the PER family. Similarly, BLAST and tFASTA algorithms identify LIN-42 as the closest match to Drosophila PER in the essentially complete C. elegans sequence database (Fig. 2).

In addition to encoding a PAS domain protein, another hallmark ofper is that its mRNA levels oscillate with a 24-hour periodicity (15). To test whether mRNA oscillation is also a feature of lin-42 expression, we examined the accumulation of lin-42 mRNA during development at 25°C and found thatlin-42 mRNA levels cycle during each larval stage (Fig. 3, A and B). We performed RT-PCR on cDNA pools derived from synchronized worm populations sampled at 2-hour intervals during postembryonic development (16, 17). In these experiments, the quantity of lin-42 cDNA amplification product was measured relative to that of ama-1amplified from the same pools as a control for the amount of mRNA originally present in the sample. ama-1 encodes the large subunit of RNA polymerase II, and its mRNA abundance is essentially constant throughout postembryonic development (16, 18) (Fig. 3D).

Figure 3

lin-42 and tim-1 mRNA levels during postembryonic development. (A) Graph showing the relative abundance over time of the lin-42 transcripts, expressed as a ratio of the amount of their amplified products to that of the ama-1 transcript for each RT-PCR reaction (17). The tick marks on the x axis indicate 2-hour intervals of postembryonic development at 25°C after L1 larval arrest. Error bars are ±1 SE. Variances were computed empirically with the use of a large sample approximation. (B) Autoradiograms of RT-PCR products from lin-42, ama-1, andtim-1 amplifications analyzed by Southern blotting. Data are representative of experiments graphed in (A) and (C). (C) Graph of tim-1 mRNA levels relative toama-1 levels for each RT-PCR reaction. Details are as in (A). The signals for lin-42 and tim-1 depend on specific activities of the probes used and exact washing conditions (16) and therefore cannot be directly compared. (D) Top panel, graph of an experiment as in (A), except that the animals were reared at 20°C and the tick marks on thex axis indicate 3-hour intervals of postembryonic development. Bottom panel, representative autoradiograms of the data graphed and the results of a Northern blot of 10 μg of total RNA probed with ama-1 sequences. The Northern blot lanes are shown beneath the samples to which they correspond. (E) Results of an experiment as in (D), except thatlin-42(mg152) animals were used. lin-42(mg152)development is delayed relative to that of the wild type.

The lin-42 and ama-1 signals were quantified (17), and the relative levels oflin-42:ama-1 mRNA are graphed in Fig. 3A with respect to developmental time. Relative to ama-1, lin-42mRNA levels peak during the intermolt periods, approximately 3 hours before each molt, and decline dramatically during ecdysis. After the L4-to-adult molt, lin-42 mRNA levels remain low. Examination of synchronized worms by Northern (RNA) analysis also reveals lower levels of lin-42 mRNA in animals undergoing ecdysis as compared to levels in intermolt animals (19).

The observed oscillations of lin-42 mRNA levels do not correspond to absolute time from hatching; rather, they are synchronized to the molting cycles as is demonstrated by the lengthening of the period of oscillation when worms are grown at a lower temperature (20°C versus 25°C) (Fig. 3D). These experiments indicate that lin-42 mRNA levels oscillate relative to the execution of molting cycles. C. elegans undergoes four rounds of ecdysis, at approximately 6-hour intervals (at 25°C), as they develop from the newly hatched L1 larva to the adult. Thelin-42 mRNA expression pattern suggests a possible role forlin-42 in promoting or coordinating aspects of the molting cycle. However, lin-42 mutants do not exhibit obvious molting defects until the execution of the final molt (8). At this stage, lin-42 mutants often have difficulty shedding the L4 cuticle. This observation suggests that lin-42functions in at least this ecdysis event, and perhaps another factor supplies this function during the earlier molts. During each molt cycle, a new worm cuticle is synthesized, composed mainly of collagens. Expression levels of six collagen genes have been found to oscillate relative to the molting cycles (16), raising the possibility that lin-42 cycling could function in, or be synchronized with, collagen gene regulation.

In Drosophila, PER is regulated in part through interaction with a second circadian rhythm protein, encoded by timeless(3). The accumulation of timeless mRNA cycles with periodicity indistinguishable from that of per(20). The existence of a worm gene with similarity to fly and mammalian timeless has been noted (21, 22). We sequenced cDNAs corresponding to this gene, tim-1, and found that the worm protein shares 23 and 37% identity withDrosophila and mouse timeless proteins, respectively, across their entire lengths (Fig. 4). BLAST and tFASTA searches show thattim-1 is the best match to TIMELESS in the worm genome. Thus, a second component of the fly circadian regulation pathway has also been conserved in C. elegans. We asked whether the temporal pattern of tim-1 mRNA accumulation parallels that of lin-42. Unlike coordinated per andtim expression in flies, the worm timelessexpression pattern was distinguishable from that of lin-42in the RT-PCR assay, and it did not oscillate with each molt (Fig. 3, B and C). In general, tim-1 mRNA levels increase during late postembryonic development and, unlike lin-42 mRNA,tim-1 mRNA is most abundant in adults.

Figure 4

ClustalW alignment of C. elegans, mouse, and Drosophila timeless proteins. The predicted C. elegans timeless protein, TIM-1, is 1353 amino acids long. Amino acids identical between at least two of the proteins are shown in red and similarities are blue (35). The two PER interaction domains in dTIM (13) are underlined. Bracketed numbers represent amino acids not shown; dashes indicate gaps in the alignment; and asterisks indicate identities in all three sequences. GenBank accession numbers are mTIM AF098161, dTIMAF032410, and TIM-1 AF183401. Percent identity or similarity between these proteins was calculated relative to the total length of TIM-1. The two largest gaps in the alignment are due largely to a 47% identical direct repeat in TIM-1 of a lysine-rich region followed by 102 amino acids.

In flies, PER and TIM participate in a negative autoregulatory feedback loop that is largely responsible for their mRNA oscillations. This regulation is dependent on their heterodimerization through the PER PAS domain and subsequent nuclear translocation (13, 23,24). Neither PER nor TIM binds DNA; rather, they are thought to repress their own transcription through interaction and interference with their positive regulators, the bHLH transcription factors dCLOCK and CYC (25). Loss-of-function permutations result in arrhythmic flies and abolish mRNA oscillations (15). In contrast, the cycling of lin-42 mRNA levels occurs in the absence of functional lin-42 protein (Fig. 3E), which indicates the lack of an autoregulatory feedback circuit.

To investigate the expression pattern of LIN-42, we generated transgenic animals bearinglin-42::gfp fusion constructs (Fig. 1B) (26). These constructs rescue the precocious phenotype of lin-42 mutants, which suggests that the fusion protein is functional. Consistent with the lin-42 phenotype, LIN-42::GFP is present in the lateral hypodermis (Fig. 5). LIN-42::GFP also accumulates in the hyp7 syncytium, which comprises the main body hypodermis, and in head and tail hypodermal cells. LIN-42::GFP expression is first detected in late embryonic stage animals, and it remains detectable into the adult stage. Although we have not quantitated LIN-42::GFP signals, the intensity oflin-42::gfp expression appears higher, in general, in animals undergoing ecdysis. In terms of subcellular accumulation, LIN-42::GFP is generally enriched in nuclei relative to cytoplasm (Fig. 5), but there is no developmental time when LIN-42::GFP is observed to be entirely nuclear or entirely cytoplasmic. Cytoplasmic LIN-42::GFP signal is enhanced in lateral seam cells during the molt periods. During the first three molts, this correlates with seam cell divisions and may be related to cell cycle stage, reflecting nuclear release followed by protein turnover and replenishment after division, because there is a period during cytokinesis where LIN-42::GFP is not observed. However, the LIN-42::GFP signal is also enhanced in the seam cell cytoplasm of L4 molt animals, when cell divisions do not occur. More detailed analysis of LIN-42 levels and subcellular localization awaits immunodetection of the endogenous protein.

Figure 5

Expression of lin-42::gfp. Fluorescence micrograph of an L3 molt stage lin-42(mg152); veIs26 animal folded back on itself so that the anterior is at the lower left. The integrated array contains pMJ11 (Fig. 1). Seam cell nuclei are labeled, as is a subset of Hyp7 nuclei. The seam cells are preparing to undergo their final division, with the exception of H1.aa, which does not divide again. High levels of GFP fluorescence are observed in nuclei of Hyp7 and the lateral seam. Cytoplasmic expression can be seen in the seam cells and is most intense in V1.pppp. Scale bar, 10 μm.

One role of LIN-42 is to restrict the hypodermal accumulation of LIN-29 to the L4 stage. In lin-42 mutant animals, LIN-29 accumulates early, during the third larval stage (27), indicating that lin-42, likelin-14 and lin-28, exerts its control of seam cell terminal differentiation by restricting LIN-29 accumulation to the proper time. LIN-42 may be a more direct regulator of LIN-29 accumulation than are LIN-14 and LIN-28 because these latter proteins disappear by the end of the L1 and L2 stages, respectively (28,29). lin-14 and lin-28 are each directly regulated by lin-4 through the binding of the regulatory 22-nucleotide (nt) lin-4 RNA to complementary sequences in their 3′ untranslated regions (UTRs) (29, 30). In contrast,lin-42 is not likely to be a direct target oflin-4 RNA. Searches of genomic DNA downstream from thelin-42 stop codon have failed to reveal lin-4complementary sequences such as those found in the lin-14and lin-28 3′UTRs.

It has been proposed that progressive reduction in the levels of LIN-14 and LIN-28 during the first three larval stages specifies the L1-L2-L3 stage transitions in the lateral hypodermis (29). Null mutations in these genes result in precocious terminal differentiation as a consequence of omitting L1 (lin-14) or L2 (lin-28) seam cell division programs, which are replaced by later stage programs. In contrast, cell lineage analysis suggests that lin-42 null mutations do not cause seam cell lineage defects during the first two larval stages (4, 31), which indicates that a unique requirement for LIN-42 action may be limited to the L3 stage. Wild-typelin-42 activity could be required to promote the L3 stage identity of seam cells or to repress the L4 stage identity until the correct time. lin-42 mutations enhance L2-stage defects of a weak lin-14 allele (4), suggesting thatlin-42 may also play an earlier, nonessential role in the specification of seam cell division patterns. Expression of thelin-42::gfp fusion in L1 and L2 stage animals is consistent with this result.

LIN-42 activity may be stage-specifically modified by interaction through its PAS domain with different partner proteins at different developmental times. These interactions could include homotypic binding to other PAS domain–containing proteins [as seen in ARNT interaction with the aryl hydrocarbon receptor (32)] or heterotypic binding to non–PAS domain–containing proteins, as exemplified by the Drosophila PER-TIM interaction (23). TIM-1 is one candidate partner; others include the products of heterochronic genes identified by mutational analysis. However, none of the heterochronic mutants described to date possess phenotypes identical to those of lin-42 mutants, which indicates that if there are LIN-42 binding partners among the products encoded by these genes, the proteins do not require each other for all aspects of their function.

Biological rhythms such as the ultradian rhythm of defecation have been studied in the soil-living nematode C. elegans(33), but circadian rhythms have not been reported. Nevertheless, we have identified lin-42 and tim-1as C. elegans counterparts to the per andtimeless genes, respectively, of flies and vertebrates. The similarity between lin-42 and per extends beyond sequence conservation to the level of gene expression, as demonstrated by their oscillating mRNA levels. These results suggest an evolutionary link between these two timing mechanisms. In addition to these similarities between per and lin-42, we have observed some interesting differences. Their functions have been adapted to control timing in different contexts: PER controls circadian time but LIN-42 controls developmental time. The periods of their message-level oscillations are also altered to reflect this difference, shortened from 24 hours for per to ∼6 hours forlin-42, which correlates with the execution of molts. Finally, in contrast to PER autoregulation, LIN-42 is not required to generate its own mRNA expression pattern. Thus, some features of the circadian timing mechanism appear to have been evolutionarily conserved with LIN-42 while others have diverged. An alternative view is that PER and LIN-42 represent a remarkable example of convergent evolution involving amino acid sequence, cyclical expression pattern, and type of biological process. The challenge for the future is to provide a more comprehensive comparison of the molecular components between these systems by identifying LIN-42 binding partners and the factors responsible for lin-42 cycling.

  • * These authors contributed equally to this work.

  • Present address: Department of Epidemiology, University of North Carolina, Chapel Hill, NC 27599, USA.

  • Present address: Department of Biology, University of Wisconsin, River Falls, WI 54022, USA.

REFERENCES AND NOTES

View Abstract

Stay Connected to Science

Navigate This Article