Abstract
Caenorhabditis elegans senses environmental signals through ciliated sensory neurons located primarily in sensory organs in the head and tail. Cilia function as sensory receptors, and mutants with defective sensory cilia have impaired sensory perception1, 2. Cilia are membrane-bound microtubule-based structures and in C. elegans are only found at the dendritic endings of sensory neurons3. Here we show that mutations that cause defects in sensory cilia or their support cells, or in sensory signal transduction, extend lifespan. Our findings imply that sensory perception regulates the lifespan of this animal, and suggest that in nature, its lifespan may be regulated by environmental cues.
We determined the lifespans of eleven mutants (in nine genes) with various defects in sensory cilia, including the absence of cilia (daf-19), deletion of the middle and distal segments (che-2, che-13, osm-1, osm-5 and osm-6) and reduced or irregular cilial segments (che-3, che-11 and daf-10). One of these genes, osm-6, has been cloned. osm-6 is expressed only in ciliated neurons4, and encodes a homologue of the Chlamydomonas reinhardtii intraflagellar transport particle, required for assembly of the flagellum, a cilium-like structure5. We found that all of these mutants were long lived (Fig. 1; Table 1). Thus, the function of sensory cilia is required for C. elegans to age normally.
Figure 1: Animals with defects in sensory cilia and sensory transduction live longer than wild type.
The fraction of animals remaining alive is plotted against animal age.
High resolution image and legend (41K)![Table 1 - Adult lifespans at 20|[thinsp]||[deg]|C](/common/images/table_thumb.gif)
Sensory cilia are present in 60 of the 302 neurons present in the C. elegans hermaphrodite3. The amphids, a pair of lateral sensilla in the head, are the principal sensory organs. Each amphid contains the ciliated endings of twelve sensory neurons, plus a sheath and a socket cell, which form a pore to the exterior. The phasmids are similar but smaller sensory organs in the tail. Four other classes of cuticular sensilla are also located in the head. We determined the lifespans of five mutants in four genes thought to affect only a subset of sensilla. daf-6(e1377) mutants, in which the amphid and phasmid pores are closed because of a defect in the sheath cells6, were long lived (Fig. 1; Table 1). mec-8(e398) mutants, in which amphid sensory cilia partially fail to fasiculate1, as well as two osm-3 mutants, in which only the cilia of the phasmid neurons and 16 out of the 24 amphid neurons are defective1, were also long lived (Fig. 1; Table 1). osm-3 encodes a kinesin-like protein that is expressed in the 20 ciliated sensory neurons that are affected in the mutant, and in 6 other ciliated sensory neurons that are unaffected in the mutant7. The remaining mutant, a mec-1 mutant that has a relatively weak defect in amphid cilia fasciculation1, was not long lived (Fig. 1; Table 1). Finally, to test the involvement of the amphids directly, we ablated the two amphid sheath cells with a laser microbeam and found that this treatment also extended lifespan (Fig. 1; Table 1). Together, these findings suggest that at least some of the neurons whose activity influences lifespan are amphid sensory neurons.
In vertebrate olfactory and visual systems, cyclic nucleotide-gated channels are required to transduce receptor activity into electrical activity8. In C. elegans, the cyclic nucleotide-gated channel 
-subunit TAX-4 and 
-subunit TAX-2 function at the ciliated endings of a subset of amphid neurons to mediate several sensory behaviours9, 10. tax-2 and tax-4 are thought to function directly in sensory transduction9; mutations in these genes do not affect the structure of the sensory cilia10. We tested two tax-4 alleles and found that both extended lifespan (Fig. 1; Table 1). We also tested three tax-2 alleles, and found that they had little or no effect on lifespan (Fig. 1; Table 1). TAX-4 is able to form functional channels in the absence of TAX-2 (ref. 11). These findings suggest that a TAX-4-dependent channel that does not require TAX-2 may operate within sensory cilia to regulate lifespan.
We then considered whether cilium structure mutants had any other obvious defects that might influence lifespan. We analysed mutants in five genes, che-2, daf-6, daf-10, osm-3 and osm-5. The feeding behaviours of these mutants were normal. The percentage of animals pumping (a pharyngeal behaviour that draws bacteria into the mouth and intestine12) was similar to wild type, as was the rate of pumping in individual animals (Table 2). The mutants were less likely to stray from the bacterial lawns (their food) than were wild-type animals (Table 2); this preference for food may be caused by their altered sensory ability2. The mutants had a normal appearance and length of postembryonic development, and they had either normal or slightly elevated brood sizes (Table 2). Because these mutants appeared normal in all of these respects, it is possible that their longevity is caused directly by altered sensory perception.
In C. elegans, signals from the reproductive system influence lifespan13; the germ line produces signals that shorten lifespan, and the somatic gonad appears to produce counterbalancing signals that lengthen lifespan13. We ablated the germline precursors, Z2 and Z3, in five mutants, daf-6, daf-10, osm-3, osm-5 and tax-4, and found that their lifespans were extended (Fig. 2; Table 3). Thus, germline signalling appeared normal. We then removed the entire gonad by killing the somatic gonad precursors, Z1 and Z4. This treatment prevents the development of the germ line, as well as the somatic gonad, and has no effect on wild-type lifespan13, 14. Three of the mutants, daf-6, osm-3 and tax-4, behaved like wild type; that is, killing the entire gonad had no effect on their lifespans (Fig. 2; Table 3). In contrast, in daf-10 and osm-5 mutants, killing the entire gonad lengthened lifespan to the same extent as did killing only the germ line (Fig. 2; Table 3). Thus, in these mutants, somatic gonad signalling appeared to be silenced. daf-10 and osm-5 mutations affect most of the ciliated sensory neurons, whereas daf-6, osm-3 and tax-4 mutations affect only subsets of these neurons1, 9. Therefore, it is possible that somatic gonad signalling requires the activity of sensory neurons that are affected in daf-10 and osm-5 but not in daf-6, osm-3 or tax-4 mutants. Alternatively, impaired function of a large number of neurons could be required to silence somatic gonad signalling.
Figure 2: Survival curves of germline- and gonad-ablated animals, and double mutants.
The fraction of animals remaining alive is plotted against animal age. The deaths of the last two surviving osm-5(p813) gonad-ablated animals are not shown. These animals died when they were 95 and 121 days old.
High resolution image and legend (69K)The lifespan of C. elegans is thought to be regulated hormonally. Mutations in daf-2, which encodes an insulin/IGF-1 receptor homologue, and in downstream signalling components, cause animals to live more than twice as long as wild type14, 15 (reviewed in ref. 16). This longevity requires DAF-16, a forkhead-family transcription factor14, 17, 18, 19, 20. daf-16 also acts in pathways that appear to regulate lifespan independently of daf-2, such as the germline signalling pathway13.
To investigate the role of daf-16, we constructed double mutants between a daf-16 null allele, mu86 (ref. 20), and mutations in five genes, osm-3, osm-5, daf-10, daf-19 and tax-4. Loss of daf-16 markedly reduced but, in most cases, did not completely eliminate lifespan extension (Fig. 2; Table 1). Thus, most of this lifespan extension requires daf-16 activity, but a fraction is daf-16 independent. daf-16 could act downstream of a sensory signal to regulate lifespan, or it could act in a parallel pathway to provide an activity that these animals simply require for their longevity.
We also determined the lifespans of daf-2 double mutants. This method has been used to show that the PI3 kinase age-1 is likely to function in a pathway with daf-2, as the double mutant has a lifespan that is similar to those of the single mutants18. In contrast, the longevity produced by germline ablation appears to be daf-2 independent, because ablating the germ line in daf-2 mutants causes an additional doubling of lifespan13. We constructed daf-2(e1370) double mutants with daf-10 and osm-3 mutations. The double mutants did not live longer than the daf-2(e1370) single mutant; instead, unexpectedly, their lifespans were slightly shorter than those of daf-2(e1370) (Fig. 2, Table 1). A trivial explanation, that the cilium-structure mutations somehow limit the potential lifespan of the animal, seems unlikely, as germline-ablated daf-10 mutants lived much longer than daf-2; daf-10 double mutants (Fig. 2, Table 3). Thus, we infer that the cilium-structure mutants do not act exclusively in a daf-2-independent pathway. To reconcile these findings, we propose the model that sensory neurons do act in a daf-2-dependent pathway, but that they also affect the animal in another way. Specifically, we suggest that an environmental signal perceived by sensory cilia regulates a DAF-2 ligand. In cilium-structure mutants, the level of this ligand falls and lifespan is extended. In addition, we propose that the cilium-structure mutants produce a second, weaker signal that shortens lifespan. In this model, if a daf-2 mutation completely mimics the lifespan-extending effect of the cilium-structure mutants, but does not completely block the effect of the lifespan-shortening signal, the double mutants would have a lifespan that is shorter than that of the daf-2 single mutant.
Certain long-lived daf-2 mutants, such as the tyrosine kinase-domain mutant e1370 (ref. 15), also appear to lack somatic gonad signalling13. In contrast, the daf-2(e1368) mutant, which has an altered DAF-2 ligand-binding domain15, has normal somatic gonad signalling13. This suggests that DAF-2 can respond to two ligands, one of which is somatic-gonad dependent (and can still bind to the e1368 receptor), and one of which is not (and cannot activate either mutant receptor)13. Perhaps the cilium-structure mutations that block somatic gonad signalling affect both DAF-2 ligands, whereas the cilium-structure mutations that do not block somatic gonad signalling affect only the gonad-independent ligand.
Unlike weak loss of daf-2 function, which increases adult lifespan, more severe loss of daf-2 function causes juvenile animals to become dauer larvae21, a stress-resistant, developmentally arrested larval form normally induced by food limitation, high temperature and crowding22. The ability of daf-2 mutants to become dauer larvae requires the activity of daf-16 (ref. 21). Sensory cilia are also involved in the regulation of dauer larva formation1, 6, and cilium-structure and tax-4 mutations are known to interact genetically with other mutations in genes that affect dauer larva formation6, 21, 23, 24. Some mutations have been shown to cause dauer formation at relatively high temperature, 27 °C (refs 16, 25), including cilium-structure mutants (J. Thomas, personal communication). We confirmed this, and found that tax-4 mutants also formed dauer larvae constitutively at 27 °C. This dauer-formation phenotype, like that of daf-2, was daf-16 dependent in each of three mutants, osm-5(p813), daf-10(m79) and tax-4(p678) that we tested (Table 4). This suggests that, during dauer formation, sensory input regulates the activity of daf-16, possibly by changing the level of daf-2 activity. This 27 °C dauer phenotype was not observed in mutants with defects in sheath cells or cilia fasciculation (Table 4); thus, the roles of sensory structures in dauer formation and adult lifespan can be uncoupled.
In summary, our findings suggest that perception of an environmental signal may regulate the lifespan of C. elegans. What might this signal be? The amphid neurons mediate responses to temperature, volatile and soluble odorants, and mechanical stimuli. It is unlikely that the signal is only temperature, because ttx-1 mutants, which lack the sensory endings of the amphid finger cells that mediate thermotaxis1, have normal lifespans (Fig. 1; Table 1). In addition, the amphid finger cells are not affected by osm-3 mutations1, which do affect lifespan. The signal affecting lifespan may be a substance that the animal can smell or taste. For example, it may be a pheromone such as dauer pheromone, which reflects population density22, or a compound that originates from organic material and reflects food availability.
Our findings indicate that the ageing process of C. elegans is quite plastic and can be affected by signals from sources that would seem to be particularly relevant to its survival and reproductive success: its environment and its reproductive system. Moreover, sensory perception may affect reproductive signalling. Under laboratory conditions, the germ lines and somatic gonads of normal animals appear to produce counterbalancing signals13. In this situation, perturbations that retard the development or activity of the germ line but not the somatic gonad might lengthen lifespan, thus, possibly increasing the chances that the animal is still youthful enough to bear progeny when its germ line matures. Perhaps other environmental conditions mimic daf-10 and osm-5 mutations, in which germline signals are no longer counterbalanced by somatic gonad signals. Under these conditions, the ageing process could become much more sensitive to general perturbations affecting reproduction.
Methods
Lifespan analysis
Lifespan assays were conducted at 20 °C as described13, 26; at the L4 moult, animals were transferred to plates containing 16 
M 5-fluoro-2'-deoxyuridine (FUDR, Sigma), which kills their progeny as embryos. Control experiments indicated that FUDR does not significantly affect lifespan. We used the L4 moult as t = 0 for lifespan analysis. Strains were grown at 20 °C for at least two generations before lifespan determination. A major cause of censoring was explosion, which generally occurred around day 10 of adulthood. In general, the pattern of censoring was not different among strains, except for daf-19 and osm-5 mutants, which were more likely to crawl off the plate. In addition, we were unable to estimate the lifespan of the che-14(e1960) mutants, because almost all animals exploded. We used Statview 5.0.1 (SAS) software to carry out statistical analysis, and to determine means and percentiles. Censored animals were incorporated into the data set as described27.
Laser ablations
L1 larvae were mounted on 2%-agarose pads containing 1–4 M NaN3 as an anaesthetic within an hour of hatching. The amphid sheath cells (AMshL and AMshR) or gonad precursors were killed with a laser microbeam as described13. Unablated control animals were treated identically but no cells were killed. In separate controls, the AMshL and AMshR cell deaths were confirmed by the absence of DiI (1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate, Molecular Probes) dye filling in amphid neurons4.
Strain construction
In double-mutant construction, cilium-structure mutations were assayed by the inability of the amphid and phasmid neurons to fill with the dye DiI. daf-2(e1370) and tax-4(p678) were assayed by the presence of dauers at 25 °C and 27 °C, respectively. daf-16(mu86) and daf-16(+) were distinguished by PCR using allele-specific primers20. daf-16(mu86) is not Dyf and did not suppress the Dyf defect of daf-10(m79), daf-19(m86), osm-3(p802) or osm-5(p813) mutants.
Other assays
For dauer assays, animals born at 20 °C were transferred onto a new well-seeded plate as eggs and incubated at 27 °C for 48 h. All other assays were conducted at 20 °C. To measure brood size, two (or in some cases one) L4 animals were cultured together and transferred to new plates every day or every 12 h, and their progeny counted. A set of reserve animals was cultured and transferred in parallel. In the case of osm-5 (but not the other mutants we examined), animals whose progeny we were counting sometimes crawled off the plates. When this happened, the plate was replaced with another containing age-matched reserve animals, and their subsequent progeny were then counted. We used F-tests to compare the variance of the mean brood sizes and found no statistical difference when comparing wild type with any group (P > 0.05 in all cases). Pharyngeal pumping was assayed using a dissecting microscope. To determine the length of postembryonic development, larvae that hatched within an hour were monitored every 2 h as they neared adulthood and were scored as adults when they had undergone the final moult and a vulva could be observed.
