Genetic Loci Modulating Fitness and Life Span in Caenorhabditis elegans: Categorical Trait Interval Mapping in CL2a x Bergerac-BO Recombinant-Inbred Worms

Corresponding author: Robert J. Shmookler Reis, Research-151, 4300 W. 7th St., Little Rock, AR 72205., reisrobertjs{at}uams.edu (E-mail)

Communicating editor: J. B. WALSH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Quantitative trait loci (QTL) can implicate an unbiased sampling of genes underlying a complex, polygenic phenotype. QTL affecting longevity in Caenorhabditis elegans were mapped using a CL2a x Bergerac-BO recombinant-inbred population. Genotypes were compared at 30 transposon-specific markers for two paired sample sets totaling 171 young controls and 172 longevity-selected worms (the last-surviving 1%) from a synchronously aged population. A third sample set, totaling 161 worms from an independent culture, was analyzed for confirmation of loci. At least six highly significant QTL affecting life span were detected both by single-marker (²) analysis and by two interval-mapping procedures—one intended for nonparametric traits and another developed specifically for mapping of categorical traits. These life-span QTL were located on chromosomes I (near the hP4 locus), III (near stP127), IV (near stP44), V (a cluster of three peaks, near stP192, stP23, and stP6), and X (two distinct peaks, near stP129 and stP2). Epistatic effects on longevity were also analyzed by Fisher's exact test, which indicated a significant life-span interaction between markers on chromosomes V (stP128) and III (stP127). Several further interactions were significant in the initial unselected population; two of these, between distal loci on chromosome V, were completely eliminated in the long-lived subset. Allelic longevity effects for two QTL, on chromosomes IV and V, were confirmed in backcrossed congenic lines and were highly significant in two very different environments—growth on solid agar medium and in liquid suspension culture.

MANY genes affecting longevity have been discovered or implicated on the basis of their mutant phenotypes. With few exceptions, long-lived mutant alleles identified thus far—in yeast, Drosophila, or Caenorhabditis elegans—result from null or severely hypomorphic mutations that eliminate or abate that gene's normal effect (e.g., LUCKINBILL et al. 1990 ; VAN VOORHIES 1992 ; KENYON et al. 1993 ; MORRIS et al. 1996 ; GEMS et al. 1998 ; Y. LIN et al. 1998 ; S. J. LIN et al. 2000 ; OGG and RUVKUN 1998 ; FELKAI et al. 1999 ; KIRCHMAN et al. 1999 ; PARADIS et al. 1999 ; PARK et al. 1999 ; VAJO et al. 1999 ; ROGINA et al. 2000 ; CLANCY et al. 2001 ; FABRIZIO et al. 2001 ; TATAR et al. 2001 ). This implies that the wild-type gene's function is detrimental, either directly or indirectly, to longevity.

An alternative approach to dissecting the genetics of life span, common in the analysis of other complex traits, is to map preexisting gene polymorphisms that may underlie natural phenotypic variation. Positioning of quantitative trait loci (QTL) should lead to the identification of genes distinct from those implicated through mutation studies, on the basis of the following arguments. The few mutagenesis studies in which longevity was screened directly have yielded exclusively alleles of age-1 (KLASS 1983 ; DUHON et al. 1996 ) and a single new gene of modest effect, age-2 (YANG and WILSON 1999 ). Other investigators elected instead to screen putative longevity-associated traits such as thermotolerance (e.g., LITHGOW 2000 ), but in so doing, constrain gene discovery to loci affecting those traits. Thus, mutagenesis studies must choose between a laborious screen with very low mutant yield and a facile screen giving a biased outcome due to substitution of a surrogate trait during primary screening. A second concern is that mutations occur in a context of gene sequences that are already highly evolved and thus unlikely to benefit from random alterations. The overwhelming majority of such mutations are thus deleterious to any function for which a gene has been subjected to significant natural selection. In choosing to screen only for changes enhancing longevity or stress resistance, investigators eliminate a plethora of "uninteresting" mutations that globally decrease fitness. At the same time, however, they restrict their attention to genes exhibiting "antagonistic pleiotropy"—quite possibly a minor fraction of all genes contributing to longevity. Only genes with a second, pleiotropic function antagonistic to longevity will enhance survival when mutated, because the mutation impairs (as expected) the primary function of that gene as impacted by selection.

The C. elegans genome contains at least eight transposon classes (MOERMAN and WATERSTON 1984 ; DREYFUS and EMMONS 1991 ); the most abundant of these is the Tc1 family, composed of 27–32 elements in most C. elegans strains, but >500 copies in strain Bergerac-BO (EGILMEZ et al. 1995 ). We constructed a heterogeneous population of many distinct recombinant-inbred worms, derived from a cross between Bergerac-BO (high Tc1 copy number) and CL2a (a low Tc1 copy strain). Data were analyzed by three mapping procedures—single-marker ² analysis, nonparametric interval mapping (KRUGLYAK and LANDER 1995 ), and categorical trait interval mapping—all of which produced consistent localization of six to eight QTL significantly affecting life span. Five QTL were coincident with longevity loci mapped previously for other interstrain crosses (EBERT et al. 1993 , EBERT et al. 1996 ; SHOOK et al. 1996 ; AYYADEVARA et al. 2001 ), serving to confirm those QTL and allowing estimation of the total number of comparable loci through recapture statistics. Multiple congenic lines were created by backcrossing the Bergerac-BO allele of each QTL-spanning region (on chromosomes IV and V) into a CL2a background. After 20 generations the resulting congenic lines were made homozygous for the introgressed regions and their life spans were assessed, demonstrating retention of the expected longevity-modifying phenotype in each case. Consistent results were also obtained for reciprocal congenic lines, constructed by repeated backcrossing into a Bergerac-BO background.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strains:
Bergerac-BO (RW7000) and CL2a (DR1345) strains of C. elegans were provided by the Caenorhabditis Genetics Center (St. Paul, Minnesota), funded by the National Institutes of Health National Center for Research Resources. These strains have been described previously (EGILMEZ et al. 1995 ). Worms are grown at 20° on 100-mm plates of solidified agar containing nematode growth medium, seeded with a lawn of Escherichia coli strain OP50 (BRENNER 1974 ).

Recombinant-inbred populations:
A cross was initiated using one Bergerac-BO hermaphrodite and three CL2a males on each of 10 plates. From those plates exhibiting 50% males in the F₁ generation, 1300 F₂ hermaphrodites were picked at the fourth larval stage (L4) and carried to the F₇ generation by self-fertilization. At each generation during the construction of recombinant-inbred populations (F₂ through F₇), eggs were recovered from day 4 (posthatch) hermaphrodites lysed in alkaline hypochlorite (BRENNER 1974 ), rinsed, and allowed to hatch overnight in S buffer.

Mass aging:
Culture conditions were as described previously (EBERT et al. 1993 ; AYYADEVARA et al. 2001 ). A recombinant-inbred population of 10⁶ F₇ worms at the L1 larval stage was placed in 500 ml liquid survival medium and shaken orbitally at 20°. Aging cohorts were grown en masse in the presence of 200 µM each of 5-fluoro-2'-deoxyuridine (FUdR; Sigma, St. Louis) and uridine 2'- and 3'-monophosphate (UMP 2', 3' mixed isomers, Sigma), a procedure shown to inhibit embryonic and larval development with little or no effect on adult longevity (MITCHELL et al. 1979 ; EBERT et al. 1993 , EBERT et al. 1996 ). For unselected controls, 200 worms were picked on day 5 (all ages are posthatch), placed individually into 0.5-ml tubes containing lysis mix and proteinase K (WILLIAMS et al. 1992 ), and stored at -70°. On day 35 posthatch, the last 1% of surviving worms were separated from carcasses by centrifugation on a step gradient of 60% sucrose overlaid with 40% Percoll (Sigma), as described by FABIAN and JOHNSON 1994 . The recovery of live worms by this method was 80-90%. A random sample of 200 worms was picked from this group for genotyping.

Genotype determination:
Single worms in lysis mix, both young control and age selected, were thawed and heated to 60° for 60 min, followed by 95° for 15 min. Worm genotypes were ascertained at 30 Tc1-based markers (WILLIAMS et al. 1992 ), using PCR and electrophoresis conditions as described previously (AYYADEVARA et al. 2001 ). Each multiplex PCR contained six locus-specific primers and a single, common Tc1-specific primer (the latter end labeled by polynucleotide kinase with [-³²P]ATP), for a total of 30 markers in five reactions. PCR amplification, commencing with 3–6% of single-worm DNA lysate in 10 µl, was performed in a hot-air thermal cycler (Idaho Technology, Idaho Falls, ID) for 30 cycles, each composed of denaturation (94°, 10 sec), annealing (58°, 60 sec), and extension (72°, 30 sec). These cycles were preceded by initial denaturation (45 sec at 94°) and followed by a final extension (15 min at 72°). Most multiplex reactions included a shared Tc1 locus as a positive control, and all were negative when run with CL2a DNA.

Data analysis:
The number of Tc1+ alleles was counted separately for unselected young and age-selected subgroups (40–50 worms/group), at each marker locus tested in a single experiment by PCR amplification and gel analysis, and was expressed as a percentage of total worms assessed. Means, standard deviations, and standard errors of means were then calculated from the four "batches" or subgroups for each age class and marker locus, to ensure that the processing groups do not differ significantly. Data from the first two processing batches were combined as young and age-selected C1 groups, each n = 86; the remaining two processing batches (totaling 85 young and 86 aged worms) comprised C2 groups. A third data set, C3, consisted of 80 young and 81 age-selected worms from an independent worm expansion and aging cohort. In this set, initial genotypes were heavily biased toward the CL2a allele at all markers (allele ratios of 0.84–0.95, presumably due to inadvertent selection during breeding), thereby compromising QTL mapping power. This experiment was therefore used only to confirm peaks observed in the C1/C2 data sets. The ratios of age-selected to young-unselected allele frequencies were calculated, and the significance of differences in these ratios was assessed by a ² test. An expanded genetic map for the 30 marker loci was obtained using MapMaker/EXP (LANDER et al. 1987 ), representing recombination accumulated over multiple intercross generations.

Association between genotypes and life span at each locus was first assessed by single-marker analysis, wherein the significance of shifts in allelic proportions was determined by the ² test. We then employed interval mapping, using a nonparametric algorithm that requires no assumptions about trait distribution (KRUGLYAK and LANDER 1995 ), and a newly developed Bayesian maximum-likelihood procedure designed for categorical traits (A. GALECKI, R. A. MILLER, S. AYYADEVARA and R. J. SHMOOKLER REIS, unpublished results). The latter method, categorical trait interval mapping (CTIM), utilizes logistic regression in a manner analogous to case-control statistics. It takes as its underlying model an example of a latent class model in which a categorical (here, dichotomous) phenotype is treated as a function of both the unknown (unobservable) genotype at a putative marker and the known (observable) genotype at flanking markers. Under this model each probability defining the joint distribution of phenotype data and unobservable genotypes (complete data) is decomposed into its Bayesian factors as the product of two underlying probabilities: Prob(phenotype|unobservable genotype) x Prob (unobservable genotype). Component probabilities are then modeled using logistic regression and Haldane's mapping function, respectively. This can also be considered as an elaboration of generalized linear models, with a composite link function to odds ratios, and binomial distribution (MCCULLAGH and NELDER 1989 ). The methodology (PREISSER et al. 2000 ; GALECKI et al. 2001 ) utilizes maximum-likelihood calculations implemented with SAS module PROC NLIN, with a separate SAS macro allowing permutation test calculations. Details are available at http://www-personal.umich.edu/~agalecki/ via a link to "CTIM: a set of SAS macros for categorical trait interval mapping."

Threshold values for single-marker statistics and LOD scores:
Significance of single-marker allele ratios or QTL peaks can be evaluated in terms of the expected incidence of false positives (type I errors). In single-marker analysis, the ² threshold corresponding to an overall -value of 0.05 (i.e., a 5% chance of obtaining at least as strong an association of marker to trait, purely by chance, anywhere in the genome) was estimated empirically (CHURCHILL and DOERGE 1994 ) by determining the ² statistics for each marker over 1000 permutations of the trait category assigned to each genotype. Because multilocus analysis involves multiple comparisons, thresholds for significance (-values) should be determined for full-genome scans; the Z-score false-positive threshold for the nonparametric interval mapping was determined from computer simulations, adjusted for genome size, cross type, and marker coverage (KRUGLYAK and LANDER 1995 ), whereas the CTIM thresholds were determined empirically from 1000 permutations (see above).

Backcrossing QTL-spanning regions from BO into CL2a:
The BO alleles of life-span QTL regions on chromosomes IV and V were introduced into the CL2a genetic background by marker-based selection during 20 generations of backcrossing. We first mated CL2a males to Bergerac-BO hermaphrodites and then crossed CL2a hermaphrodites to males of the F₁ progeny (and subsequent backcross progeny)—thereby ensuring the propagation of successful crosses only. DNA extraction and PCR-based marker selection were performed as described previously (AYYADEVARA et al. 2001 ). For the chromosome IV QTL, a BO-derived region spanning markers stP13 and stP35 was introgressed into the CL2a background. Locus-specific primers were used, corresponding to sequences adjoining two Tc1 insertion sites: stP13 (5'-CCCACAACCTTTTGCTACAAC-3') and stP35 (5'-GCAGTCTCTAATAGAGCTGC-3'). For the QTL on chromosome V, the backcrossed lines were selected for retention of the chromosomal region between stP192 and stP18. Locus-specific primers used for selection were adjacent to stP192 (5'-GCACGCTGAGAGTAAGTGC-3') and stP18 (5'-TTGAACTTCTCCCACTCCTC-3'). PCR amplification was performed with these two primers and the Tc1-transposon primer (5'-GAACACTGTGGTGAAGTTTC-3'). The reaction buffer contained 10 mM Tris pH 8.3, 50 mM KCl, 1.5 mM MgCl₂, 200 µM of each dNTP, and 0.5 µM of each primer. Polymerase chain reaction in a hot-air thermocycler (Idaho Technology) entailed 36 PCR cycles—each consisting of 0 sec at 94°, 30 sec at 58°, and 30 sec at 72°—followed by a 7-min final extension at 72°. PCR products were electrophoresed on a 4% agarose gel (3:1, Nusieve, SeaPlaque; FMC, Rockland, ME) and stained with ethidium bromide.

Survivals:
Procedures for assessing survival in liquid cultures and on agar were described previously (EBERT et al. 1993 , EBERT et al. 1996 ). In brief, 50 L4 larvae were selected from mixed cultures and placed in each 60-mm petri dish. For liquid survivals, each dish contained 3 ml of survival medium [S buffer (BRENNER 1974 ) containing 10⁹ bacteria (E. coli strain OP50) and 10 µg/ml cholesterol]. For agar survivals, each 60-mm dish contained 5 ml of NGM agar (BRENNER 1974 ), including 10 µg/ml cholesterol in solidified 1.7% agar, spotted with a suspension of OP50 bacteria near the center of each dish. Worms were maintained in humidified containers at 20° and were manually transferred to fresh plates daily while fertile (through the ninth day posthatch) and every 2–3 days thereafter. Each survival population began with 100 worms in two dishes for each strain or line, and each survival was performed at least twice in independent replicate experiments. Differences between survivals were evaluated for significance by Gehan's Wilcoxon test, a nonparametric procedure based on the log-rank test. Interaction between strain and environment was assessed using Cox's proportional hazards model.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A cross was constructed between two strains initially isolated from the wild: CL2a and Bergerac-BO, which are strains of low and high Tc1 copy number, respectively (see EGILMEZ et al. 1995 , for strain details). After random mating in the F₁ generation, 1300 F₂ hermaphrodites were picked as preadult (L4) larvae and agitated to favor self-fertilization (and hence homozygosity) through the F₇ generation. A population of approximately 10⁶ F₇ worms was synchronously aged, and 171 young unselected and 172 age-selected worms—at day 35, when <1% remained alive—were analyzed at 30 Tc1 insertion markers distributed across the six chromosomes. These data generated both a Tc1-marker linkage map and approximate locations of at least six life-span QTL in the C. elegans genome.

Linkage map:
An expanded genetic map based on unselected F₇ (CL2a x BO) genotypes was calculated using MapMaker software (LANDER et al. 1987 ). The uncorrected genetic distances within the linkage map appeared expanded by an average of 4.4-fold (range, 2.5- to 9.6-fold) over the standard genetic map, due to the accrual of recombinations over multiple generations. This expansion is evident in the maps of Fig 1 as the difference between the upper and lower x-axis scales. Similar map expansions have been reported previously for recombinant-inbred (RI) lines and populations (DIXON 1993 ; EBERT et al. 1993 , EBERT et al. 1996 ; AYYADEVARA et al. 2001 ).

View larger version (44K): In this window In a new window Download PPT slide   Figure 1. CTIM and NpIM of life-span QTL (1sq1–lsqXc) determined by analysis of genotypes for unselected young adults and extreme (highest 1%) survivors in an F7 recombinant-inbred population generated from a Bergerac-BO x CL2a cross. Green symbols and lines show interval-mapping LOD scores generated by CTIM, a Bayesian interval mapping procedure for categorical traits (A. GALECKI, S. AYYADEVARA, R. A. MILLER and R. J. SHMOOKLER REIS, unpublished results) based on generalized linear models. Blue symbols and lines indicate Z scores generated by NpIM. Diamonds indicate data from sample set C1 (each n = 86), while triangles indicate data from set C2 (young n = 85, long-lived n = 86). Squares represent data from another, independent recombinant-inbred expansion and synchronous aging cohort at F7 (C3; young n = 80, aged n = 81). LOD scores for this data set only are shown at twice their actual values ("x2") because of the very low power in this experiment, due to bias in initial allele ratios. Each bottom x-axis indicates the uncorrected genetic distance in map units, whereas each top x-axis shows positions in centimorgans, scaled to the standard C. elegans genetic map. Each left y-axis indicates the LOD scale for CTIM (green symbols), and each right y-axis indicates Z values for NpIM (blue symbols). Dotted horizontal arrows (green for CTIM, blue for NpIM) indicate thresholds of genome-wide significance at P = 0.01, i.e., LOD is 3.5, on the basis of 1000 CTIM permutations of trait value with respect to genotype, and Z = 4.1, on the basis of NpIM simulations adjusted for C. elegans genome size and recombinant-inbred populations (KRUGLYAK and LANDER 1995 ). Green double arrows indicate 95% confidence intervals for CTIM peak positions; those for NpIM are not shown, but are listed in Table 2.

Loci affecting Darwinian fitness or segregation distortion:
If alleles at a functional locus have differing effects on the bearer's fitness, the allele conferring greater fitness will increase in the population. Because this effect is compounded over multiple generations, even relatively small allelic differences in fitness can markedly alter the allele ratios at nearby markers within the F₇ population prior to age selection. For markers on autosomes, not linked to any gene impacting fitness, the initial Bergerac-BO (Tc1+) allele frequencies are expected to approximate 50%. Due to the nonreciprocal nature of the cross, the expected value is 67% for similarly neutral markers on the X chromosome. There were significant deviations from these expected frequencies for 23 of the 30 markers. Among young unselected worms, the Bergerac-BO allele was enriched over the CL2a allele for markers surveyed on all chromosomes except X. A BO allele favoring fitness can be localized near markers with the highest BO allele frequencies, such as stP124 (chromosome I), stP101 (II), stP17 (III), stP35 (IV), stP23, and stP128 (V), whereas a CL2a (Tc1^-) allele favoring fitness was localized near stP33 on the X chromosome (Table 1).

Mapping loci that affect longevity:
QTL modifying life span were mapped by single-marker analysis (² tests comparing allele frequencies) and by interval mapping using algorithms for either nonparametric analysis (KRUGLYAK and LANDER 1995 ) or categorical trait analysis developed for binary or categorical phenotypes such as membership in an extreme-trait group (A. GALECKI, R. MILLER and R. J. SHMOOKLER REIS, unpublished results). At least six significant life-span QTL were detected by all three procedures (see Fig 1 and Table 1). Multiple adjoining peaks on chromosome V may represent one or several QTL—resolvable only by partitioning this region through recombination during the construction of congenic lines.

Single-marker analysis:
As discussed above, the allele frequency at any marker—defined as the proportion of total alleles tested that are of the BO type (Tc1+)—may deviate from its expected value due to intergeneration selection affecting the young control population and its progenitors. The allele frequency in the young control population, however, provides an internal baseline for assessing the effects of allelic selection on the aging population. Genetic contributions to life span are revealed by shifts in allele frequency at nearby markers, between the control and long-lived groups, with the magnitude of the shift indicating proximity to the underlying QTL and the strength of the QTL. Differences between the ratios of age-selected vs. young marker allele frequencies were evaluated by ² tests, with significance assessed both by reference to the ² distribution and empirically by iterated permutation of the trait-class designation (young or long lived) with respect to marker class (BO or CL2a) as indicated in Table 1, columns 7 and 8. Markers that produced peak ²-statistic values for each QTL are indicated in boldface type in Table 1. Significant enrichment of the CL2a allele was observed in the age-selected population on chromosomes I (at markers hP4 and tcbn2), III (stP127), IV (all three markers, reaching 4.5-fold near stP13), V (six markers, reaching 6.7-fold at stP23), and X (stP129, stP2). The BO allele was weakly associated with extreme longevity at one or more loci on chromosome II (e.g., maP1), but the allelic enrichment was not significant.

The standardized effect (2a/_P = s/i) at each marker across the QTL intervals was estimated as described previously (FALCONER and MACKAY 1996 ; AYYADEVARA et al. 2001 ). Peak values ranged from 0.17 to 0.33 (Table 2), nominally equivalent to 1.1–2.1 days of life extension. As QTL-to-marker distances increase, effects measured at markers are increasingly underestimated.

Interval mapping:
Quite similar results were obtained by nonparametric interval mapping (NpIM) and CTIM, each performed on the same replicate, random samples from the young and age-selected populations (Fig 1). Peak positions are essentially coincident for the two analytical methods, but CTIM peaks (Fig 1, green symbols/lines) are consistently somewhat sharper and thus have smaller 95% confidence intervals than those generated by NpIM (Fig 1, blue symbols/lines)—consistent with the contention that CTIM represents a more appropriate statistical model. Locations and heights of peak maxima differ more between the two sets of samples; this was particularly apparent on chromosome IV, where the first set (C1) peaked at 10–12 cM (lsq4a) while the C2 samples produced peaks at 35–40 cM (lsq4b) by either mapping procedure. A third, completely independent pair of samples (C3), was taken from a population expansion in which initial allele frequencies were strongly biased toward the CL2a genotype, thereby substantially reducing power for QTL mapping. CTIM scans based on these genotypes, none of which led to significant peaks, are indicated as light green squares (Fig 1), after multiplying their LOD scores by 2 to facilitate identification of peaks. Estimates of the location, peak height, significance, and effect of each QTL are given in Table 2 for each group, employing both separate and combined samples. The significance threshold for CTIM false-positive incidence over the entire genome was determined empirically by conducting 1000 permutations of trait category with respect to the observed genotypes, each time reanalyzing the data for QTL by CTIM. The P = 0.01 genome-wide threshold occurs at LOD 3.5, whereas the equivalent (P = 0.01) threshold for NpIM—determined from computer simulations (KRUGLYAK and LANDER 1995 ) and adjusted for the C. elegans genome size and for recombinant-inbred populations—is Z = 4.1. Seven or more QTL peaks—one each on chromosomes I, III, and IV, and at least two each on V and X—exceeded both of these stringent criteria for both the C1 and C2 data sets (Fig 1). Of these, all but lsqXb were also supported by data from the low-power C3 experiment.

QTL peaks observed by either interval-mapping procedure were entirely consistent with single-marker data (Table 1), but provided additional information on QTL position and effect size. Significant life-span QTL peaks were located by both NpIM and CTIM to peaks on chromosomes I (between stP124 and hP4), III (near stP127), IV [between stP13 and stP44 (for replicates C1 and C3) or between stP44 and stP35 (C2)], V (multiple peaks representing one to five QTL; C1 and C3 agree on a major peak between stP23 and bP1), and X (two distinct peaks at stP129–stP72 and close to stP2). QTL peaks to the left of stP129 on the X chromosome were inconsistent among groups, in contrast to previous observations (AYYADEVARA et al. 2001 ).

Demonstration of QTL effect on life span in near-isogenic backcrossed lines:
To test the effects of life-span QTL on chromosomes IV and V, we created congenic lines, near-isogenic constructs in which the Bergerac-BO allele was repeatedly backcrossed into a CL2a background. The BO-derived chromosomal segment was monitored by allele-specific PCR at flanking markers during 20 generations of backcrossing, preserving two or three independent lineages for each QTL. The resulting congenic heterozygotes were self-fertilized and their progeny screened by PCR for homozygous markers. Homozygous congenic lines were thus established and tested for survival in two distinct environments: on solidified agar plates and in liquid suspension cultures. Two lines representing lsq4b(BO) and three representing lsq5a(BO), each in a CL2a background, had median longevities reduced by 2–3.5 days (14–25%) and maximum life spans reduced by 2–8 days, relative to CL2a parents (each P < 0.02, taken separately, by Gehan's Wilcoxon test). Representative survival data are shown in Fig 2 and summarized in Table 3. Duplicates of each survival gave nearly identical results, and survivals conducted on congenic lines backcrossed instead into the Bergerac-BO parental background, although for technical reasons limited to three generations of backcrossing, demonstrated very similar allelic effects with the CL2a alleles, here extending median longevity by 2–3 days (data not shown).

View larger version (38K): In this window In a new window Download PPT slide   Figure 2. Survivals in liquid culture and on agar, at 20°, of congenic lines for lsq4b(BO) and lsq5a(BO) in a CL2a background. One comparison (of two replicates) is shown for each QTL. Ages are shown on the x-axis as days from the L4 stage, at which they were selected for synchronous survival; posthatch survival times would be 2.5 days greater. (A and B) Chromosome IV initial congenic lines SR700 and SR701 differ in life span from CL2a (intraexperimental P < 0.001 for each comparison by Gehan's Wilcoxon test), whether assessed on agar medium (A) or in liquid survival medium (B). (C and D) Chromosome V initial congenic lines SR702, SR703, and SR704 differ in life span from CL2a by Gehan's Wilcoxon test [P < 0.01 for all assays on agar medium (C) or P < 0.02 for assays in liquid survival medium (D)].

View this table: In this window In a new window   Table 3. Survival data for Lsq4b and Lsq5a (BO CL2a) congenic lines

The introgressed regions in these initial congenic lines span 7–12 cM (10–16 Mb). Although such spans may contain several hundred genes each, extensive backcrossing into a uniform background reduces the likelihood of polygenic or interactive effects within any congenic interval.

Gene x environment and gene x gene interactions:
Gene x environment interactions were tested using Cox's proportional hazards model and the liquid and agar survival data for lsq4b or lsq5a shown in Fig 2 and summarized in Table 3. Although the age and environment terms were highly significant, as expected, the interaction term was not significant for either QTL.

Epistatic (gene x gene) interactions affecting a categorical trait are readily detected by the consequent distortion in diallele frequencies from the values predicted on the assumption of independence. Interactions among more than two genes are much more difficult to evaluate, due to a marked decrease in power as the number of potential combinations rises. We tested 136 diallele matrices by Fisher's exact test, pairing each of the 8 markers that showed peak association to longevity QTL with every marker but itself, in a panel of 18 markers. The panel was composed of these 8 peak markers plus 10 others, selected from the 30 markers used in mapping by the sole criterion of providing even coverage across the six chromosomes. In either a young or a long-lived population, considered separately, if two markers act independently (without interaction), the frequency of any diallele combination should equal the product of corresponding single-allele frequencies at the two loci. A total of 33 diallele frequencies were found to differ significantly from this expectation, with probability to allow for 136 interactions tested (see Fig 3). Of these, 20 are likely to reflect linkage between two markers at apparent genetic distances of <50 cM. This leaves 12 nontrivial interactions of genome-wide significance within the young group (i.e., affecting fitness), 3 of which occur in both young and long-lived worms, and one—between stP127(III) and stP128(V)— that is specific to just the long-lived subgroup and hence affects longevity. These nontrivial interactions consist of three interchromosomal associations—stP127(III) to stP35(IV), stP127(III) to stP128(V), and stP6(V) to stP129(X)—and 10 associations between distal markers on chromosome V, separated by at least 80 map units (Fig 3A) and thus corresponding to a recombinant fraction of 0.46 by the Kosambi mapping function (LYNCH and WALSH 1998 ). Indeed, 3 such interactions involve markers with apparent genetic distances of 145–167 cM and Kosambi recombinant fractions >0.49—equivalent to unlinked loci. Each produced a genome-wide P value of <0.001 in the young group, but P > 0.3 for long-lived worms.

View larger version (26K): In this window In a new window Download PPT slide   Figure 3. Schematic diagrams of significant gene x gene interactions in two crosses. (A) Gene x gene interactions observed in the present cross, CL2a x Bergerac-BO. (B) Gene x gene interactions observed in a previous cross, RC301 x Bergerac-BO (AYYADEVARA et al. 2001 ). Markers were tested for pairwise interactions by Fisher's exact test on 2 x 2 diallele tables. The indicated P values are for genome-wide false positives, after Bonferroni adjustment for multiple comparisons. Straight lines represent chromosomes. Markers interacting between chromosomes are indicated in boldface type. Markers within a shaded oval share a common interaction with other marker groups (on chromosome V only). Thick solid arrows, interaction specific to age-selected worms (shown as dashed arrows if the genome-wide false-positive level is >0.05); thin solid arrows, interchromosomal fitness interactions (significant in both young control and longevity-selected worm populations), omitting apparent interactions among proximal markers. Thin dotted arrows indicate interactions that were highly significant (P < 0.001) in young control worms but absent from age-selected worms (reverting to P 0.3); note that in both crosses, identical interactions within chromosome V are seen in young worms but not in age-selected worms.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

QTL associations with life span were tested after seven generations of inbreeding, during which each lineage approached homozygosity while recombination accrued to expand the apparent genetic map. Through the use of recombinant-inbred animals, map expansion, and selective genotyping of trait extremes (LANDER and BOTSTEIN 1989 ; CAREY and WILLIAMSON 1991 ; DARVASI and SOLLER 1992 ) defined within a survival, we generated data sets with improved power to discover and resolve QTL affecting life span. Long-lived individuals were sampled from the last 1% of an initial population of 10⁶ recombinant-inbred worms and were compared to a similar random sample taken from the initial group prior to longevity selection. From the numbers of worms taken at each generation to produce the next, we know that the expanded population represents >2600 essentially independent genotypes (calculation not shown). The immense problem of intergroup variance in longevity, which plagues conventional mapping studies (see EBERT et al. 1993 ; SHOOK et al. 1996 ), was thus avoided, whereas information extracted for QTL analysis was equivalent to complete genotyping of a population containing >2000 individuals (LANDER and BOTSTEIN 1989 ; DARVASI and SOLLER 1992 ).

Despite these considerable advantages, use of this experimental design has been hampered by the lack of an interval-mapping tool appropriate to the analysis of categorical traits (or categorical samples drawn from continuous traits) in interstrain cross protocols. Nonparametric interval mapping could have been utilized (KRUGLYAK and LANDER 1995 ; AYYADEVARA et al. 2001 ), but the consequences for power and precision of mapping had not been hitherto assessed. We here used a parametric interval-mapping tool suitable for categorical responses, which we term categorical trait interval mapping. It should be noted that the two categories are not strictly exclusive, since each initial young population includes the corresponding age-selected subpopulation. However, since the latter group makes up 1% of the former, the error introduced by treating these as separate categories is quite small. We observed consistently sharper peaks by CTIM than by NpIM on all chromosomes (see Fig 1), presumably due to our use of a more appropriate statistical model for this experimental design, producing a better fit to the underlying distribution.

QTL mapping results can be tested and positioned within absolute boundaries by backcrossing either allele of a QTL-containing region into the parental strain of contrasting genotype. We have thus far introgressed four longevity QTL—two on chromosomes IV and V, from the present cross, and two on chromosomes III and X from an RC301 x Bergerac-BO cross (AYYADEVARA et al. 2001 )—into the background of one parental strain over 20 backcross generations. In each case, we found the expected effect of QTL on trait: 1.5–4 days alteration in median and 1–8 days in maximal life span (Fig 2; see also AYYADEVARA et al. 2001 ; A. VERTINO, S. AYYADEVARA, J. J. THADEN and R. J. SHMOOKLER REIS, unpublished results).

Evidence of fitness selection:
Significant deviations in the initial allele frequencies (of young unselected worms), from their expected frequencies, were observed at several marker loci. Such deviation from expected Mendelian segregation ratios has been observed in many species and is referred to as segregation distortion or transmission ratio distortion (GARNER et al. 1991 ; LYTTLE 1991 ; VONGS et al. 1993 ; XU et al. 1997 ). Segregation distortion effects may be caused by cumulative fitness selection favoring one allele in either the individual or the gametes, compounded over multiple generations. In this CL2a x Bergerac-BO cross, the BO (Tc1+) allele was represented at more than the expected frequency—0.5 for autosomes and 0.67 for the X chromosome due to a nonreciprocal cross design—at most or all markers on five of the six chromosomes. This may appear paradoxical in view of the relatively small brood size and delayed egg laying of Bergerac-BO worms (JOHNSON and WOOD 1982 ). It should be noted, however, that we recovered eggs at each generation from hermaphrodites lysed in alkaline hypochlorite. Worms were lysed at day 4 posthatch, 1 day after the onset of laying by wild-type, Bristol-N2 worms, by which point Bergerac-BO has higher fecundity (JOHNSON and WOOD 1982 ).

Loci displaying lower- or higher-than-expected BO (Tc1+) allele frequencies suggest linkage to a dimorphic locus affecting reproductive or gametic fitness or affecting the resistance of eggs to alkaline hypochlorite. The latter is a rather unusual criterion for "fitness," which must be considered in interpreting these results. The BO allele was significantly enriched at 18 of the 30 marker loci, whereas the CL2a allele was enriched at 4 loci on the X chromosome, from its expected 33% frequency to 49–77%. Genes with allelic effects on "fitness" were thus localized on chromosomes I (near stP124), II (stP101), III (stP17), IV (stP35), V (maxima near stP23 and stP128), and X (stP33; see Table 1). Most of these loci were positioned almost identically in a previous RC301 x Bergerac-BO cross (AYYADEVARA et al. 2001 ) near markers stP35–stP44 on chromosome IV, stP192–stP23 and stP108 on chromosome V, and stP33 on the X chromosome.

Loci affecting nematode longevity:
Significant associations between individual markers and longevity were established by ² tests (single-marker analysis) comparing allele frequencies in age-selected vs. young control groups (Table 1). We determined maximum-likelihood positions of these QTL by both nonparametric interval mapping (KRUGLYAK and LANDER 1995 ) and categorical trait interval mapping, developed for this experimental design (A. GALECKI, S. AYYADEVARA, R. MILLER and R. J. SHMOOKLER REIS, unpublished results). QTL likelihood maxima generated by these two algorithms were consistent with each other and also with the results of single-marker ² analysis, although the precise location of the maximum-likelihood peak sometimes differed between sample groups [e.g., compare the C1 and C2 curves for chromosome IV (Fig 1) by either interval mapping protocol]. Confirmation of the QTL locations requires the construction and fine-map analysis of congenic lines and derived recombinant sublines. It is instructive that examination of congenic recombinants spanning the QTL interval on chromosome IV confirmed the existence of a longevity QTL only between markers stP44 and stP35 (as indicated for samples C2), but not within the stP13–stP44 interval implicated by samples C1 and C3 (A. VERTINO, S. AYYADEVARA, R. AYYADEVARA, J. J. THADEN and R. J. SHMOOKLER REIS, unpublished results). This illustrates that interval mapping can be sensitive to variation in sampling, as has been previously reported (CAREY and WILLIAMSON 1991 ; DARVASI and SOLLER 1992 ; EBERT et al. 1993 ); the observed sampling effects were consistent for the two interval-mapping procedures.

Comparison to previous genetic mapping of longevity QTL in C. elegans:
Because QTL mapping can reveal only loci that are dimorphic between parental strains, multiple pairings of strains are necessary to identify all influential loci. To date, we have performed three independent interstrain crosses (one in duplicate) for QTL mapping of longevity genes in C. elegans: Bristol-N2 x Bergerac-BO (EBERT et al. 1993 , EBERT et al. 1996 ; AYYADEVARA et al. 2001 ), RC301 x Bergerac-BO (AYYADEVARA et al. 2001 ), and CL2a x Bergerac-BO (this study). A fourth study employed Bristol-N2 and DH424, a strain closely related to Bergerac-BO (EGILMEZ et al. 1995 ), and hence was substantially redundant with the first cross in terms of QTL mapping; this cross is ignored for purposes of this discussion.

Lsq1 on chromosome I was coincident and significant (genome-wide P < 0.01) in all three crosses. Chromosome II harbors two longevity QTL: lsq2a on the left (stP100–stP196), observed in the RC301 cross, and lsq2b nearer the center (peaking close to age-1, stP50–stP198) in the Bristol-N2 cross. On chromosome III, a highly significant QTL (Z scores of 7.4 and 16, respectively) appeared in the RC301 and CL2a crosses. The lsq4a peak was apparent in the present CL2a cross for the C1 data set and constituted the major chromosome IV peak in the RC301 cross. However, a distinct lsq4b peak was seen to the right of lsq4a, as the sole chromosome IV peak implicated in the Bristol-N2 cross and for the C2 and C3 data sets from the CL2a cross and as a lesser peak in the RC301 mapping. Testing of recombinants derived from chromosome IV congenic lines, in a CL2a background, supports only the existence of lsq4b. QTL mapping on chromosome V was complex, indicating at least two longevity loci in the present cross (lsq5a and lsq5b; see Fig 1), of the three that were significant in the RC301 cross (adding lsq5c at the right extreme). Multiple QTL also reside on the X chromosome: lsqXa was evident in the RC301 and Bristol-N2 crosses, but lsqXb and lsqXc were more prominent in the present CL2a cross (Fig 1).

Among long-lived worms in these three crosses, the Bergerac-BO allele was favored only three times whereas the low-Tc1 strain was enriched a total of 17 times. This bias is consistent with the median life span of Bergerac-BO that is shorter than that of any of the other parental strains (17 vs. 19–20 days) and may reflect the cumulative effect of Bergerac's mutator status over much of its evolutionary history (EGILMEZ et al. 1995 ). From their positions and allelic effects in three interstrain crosses, the longevity-determining alleles can be ranked in the three parental strains, as illustrated in Fig 4. Differences in allelic effect or "strength" among these strains undoubtedly contribute to the wide variation in peak height observed for a given QTL in different crosses.

View larger version (26K): In this window In a new window Download PPT slide   Figure 4. Ordering of allele effects in four C. elegans strains as indicated by the observation of QTL in three interstrain cross-mapping studies. Alleles conferring increased longevity, among F7 RI cross progeny, are shown above each chromosome line as greater than (>) the shorter-lived alleles at that locus. If a cross between two strains produced no evidence of a longevity QTL at a given locus, the corresponding alleles are shown as equal. QTL locations are shown as marker intervals below each chromosome line.

Estimation of the total number of life-span QTL in C. elegans:
From the numbers and locations of QTL mapped using three different interstrain crosses, we estimated the total number of QTL that strongly influence nematode life span. We considered 7 QTL positioned in the present cross (ignoring the left-most peak or "shoulder" on chromosome V; see Fig 2), 8 identified in the RC301 x Bergerac-BO cross (AYYADEVARA et al. 2001 ), and 4 supported by nonparametric interval mapping in the Bristol-N2 x Bergerac-BO cross (AYYADEVARA et al. 2001 ) from the 5 loci that were significant by single-marker analysis (EBERT et al. 1993 ). Of these 13 total QTL, 8 were indicated in more than one cross, and 2 (on chromosomes I and IV) were mapped to similar positions in all three crosses. Recapture statistics (FELLER 1968 ) can be used to estimate the total number of life-span QTL (n') of comparable significance, as , where n₁ is the number of QTL identified in a given cross, n₂ is the number of QTL identified in a second cross, and k is the number of QTL common to both crosses. Considering the three possible pairings among these three crosses, we estimate the number of comparable QTL that govern the life span of C. elegans to be 12–24, whether or not uncertain peaks (e.g., adjoining larger QTL peaks, as on chromosome V) are included. Incomplete map coverage (especially on chromosomes I, III, and IV) and failure to resolve closely linked QTL (as on chromosome V) may lead to underestimation, whereas variability in QTL peak height may cause overestimation of total QTL number. The true number could be as low as 13 (the number actually mapped to date), but is unlikely to exceed 36.

Interactions:
Most "interactions" observed in the initial young control group involved markers within a chromosome—predominantly chromosome V—and thus could reflect marker interdependence due to linkage. In the three most extreme cases, however, the interacting loci were clearly too far apart to support this interpretation: >130 cM in the expanded map, corresponding to recombinant fractions >47%. We therefore interpret these as genuine instances of epistasis, occurring between genes located on one chromosome. Just two significant interchromosomal interactions were detected for "fitness," stP6(V):stP129(X) and stP127(III): stP35(IV).

Only one epistatic interaction was unique to the age-selected subpopulation between stP127(III) and stP128 (V) (P < 0.05; thick solid arrow in Fig 3A), although additional interchromosomal epistases were suggested by the disappearance, among aged worms, of strong interactions seen in the young controls, i.e., genome-wide P values of <0.001 in the young group, reverting to P > 0.3 for long-lived worms. Such reversals occurred between markers at the right end of chromosome V (stP108, stP105, and stP128) and either the center markers (bP1 and stP6) or the left end (stP3, stP192), indicated by dotted arrows in Fig 3A. Since age-selected worms began life with the same allele distribution as the young controls, this reversal of a diallele bias among late-surviving worms may suggest antagonistic pleiotropy, wherein the same allele combinations that confer reproductive fitness are associated with decreased life span. It is noteworthy that, on chromosome V, essentially identical gene x gene interactions were observed in our previous study (see Fig 3B), whereas all other interactions differed between these crosses.

Epistasis tends to be underestimated when assessed in QTL mapping, because only those QTL that are polymorphic between the parents are detectable in a mapping experiment, and interactive partners also must be polymorphic between the same two parents for interactions to be observed. Thus, the scarcity of epistatic interactions involving markers on different chromosomes may reflect only the limited instances of epistasis among the (possibly minor) subset of longevity-affecting QTL that are visible in that cross.

QTL mapping was not performed here under distinct environmental conditions, as had been the case for a previous cross (EBERT et al. 1993 ). However, subsequent survivals were conducted in two environments, liquid culture and growth on solidified agar medium, for lsq4b and lsq5a congenic lines and their parental controls [ Table 3, Fig 2, and near-identical replicate experiments (not shown)]. Although these data suggest a genotype x environment interaction at lsq5a for median longevity (e.g., see Fig 2C and Fig D), no interaction was significant for either QTL when analyzed in a Cox proportional hazards model (data not shown). This does not necessarily imply that such interactions do not exist, but certainly indicates that much larger studies would be necessary to detect them. Our results contrast with those reported by the Mackay group (VIEIRA et al. 2000 ) in an elegant experimental design comparing longevities of Drosophila imagoes in five environments. Those data imply that genotype x environment interactions impacting life span are common in Drosophila and, along with sex x genotype interactions, account for the preponderance of genetic variance for longevity within their cross populations.

Antagonistic pleiotropy:
As noted above, antagonistic pleiotropy—the proposed existence of allelic differences in "Darwinian fitness" that confer opposite effects on postreproductive survival—is suggested by the presence of strong gene x gene interactions only in the young control group. In general, however, the significant QTL determinants of fitness (as defined by our experimental protocol) did not co-localize with the major peaks for longevity QTL. Examination of Table 1 shows that, of seven significant peak markers for each trait, only one (stP23, chromosome V) appeared to be coincident for both. Thus, while antagonistic pleiotropy remains a plausible explanation for the evolution of some longevity-regulating genes, it may not account for the majority of such loci in C. elegans.

Identifying candidate genes in QTL regions:
Since the C. elegans genome sequence is known (C. ELEGANS SEQUENCING CONSORTIUM 1998), it is possible to identify potential candidate genes for some of the QTL identified in the present and earlier crosses. Although this is largely speculative at present, given the hundreds of genes within each implicated interval, it may be of interest to briefly mention a few functional candidates lying near QTL maximum-likelihood positions (see Fig 1). On chromosome I, rad-8, sod-2, and daf-16 all lie close to the fitness maximum observed near marker stP124, but rather far from the longevity peak at hP4. Of these, daf-16 is of particular interest in the context of longevity determination since it appears to mediate the bulk of dauer-pathway effects on life span and other phenotypes. One branch of the bifurcate dauer-formation pathway has striking structural and functional homology to the IGF-1 response pathway of mammals (LIN et al. 1997 ; OGG et al. 1997 ). Some other members of this pathway (see TISSENBAUM and RUVKUN 1998 ; WOLKOW et al. 2000 ) are age-1 (chromosome II), daf-2 (III), akt-1 (V), akt-2 (X), and daf-12 (X); only the last three are close to QTL peaks implicated in this study. Clk-1, clk-2, and gro-1 (III), genes implicated in adult and especially total survival time (LAKOWSKI and HEKIMI 1996 , LAKOWSKI and HEKIMI 1998 ; EWBANK et al. 1997 ), are possible candidates in the vicinity of lsq3 (Fig 1). Another candidate close to the lsq3 peak is mev-1, encoding a subunit of succinate dehydrogenase cytochrome b; mutations to mev-1 render worms short lived and oxygen sensitive, perhaps by heightened superoxide production (ISHII et al. 1998 ). An active copy of sir-2 lies near the chromosome IV peaks; this gene is involved in the modulation of yeast longevity by energy supply (LIN et al. 2000 ). Mutations of spe-26, close to sir-2, appear to increase C. elegans longevity through attenuation of sperm production (VAN VOORHIES 1992 ), an effect also noted for spe-10 (CYPSER and JOHNSON 1999 ) situated near lsq5a, but not for other spermatogenesis mutations.

To define the precise span of each QTL region on the C. elegans physical map, we constructed congenic lines in which a chromosomal segment from one parental strain is isolated in the genetic background of another. Recombinants between the QTL-flanking markers are then identified and assessed for both phenotype and fine-map genotype (AYYADEVARA et al. 2000 ). This procedure has already led to confirmation and absolute physical positioning of three longevity QTL (A. VERTINO, S. AYYADEVARA, R. AYYADEVARA, J. J. THADEN and R. J. SHMOOKLER REIS, unpublished results). Subsequent narrowing of the implicated intervals then proceeds by using interposed markers to progressively partition them, eventually reducing each interval to a small number of genes that could be tested for allelic differences in expression level or coding sequence and ultimately for the predicted phenotypic effects.

Caveats and prospects for gene identification:
Because QTL mapping often utilizes transposable elements as markers, especially in C. elegans (EBERT et al. 1993 , EBERT et al. 1996 ; SHOOK et al. 1996 ; SHOOK and JOHNSON 1999 ; AYYADEVARA et al. 2001 ) and Drosophila melanogaster (LEIPS and MACKAY 2000 ; PASYUKOVA et al. 2000 ), at least some of the variable loci thus discovered may have arisen by transposon insertion, within or adjacent to a gene. That scenario is consistent with the mutator status of Bergerac-BO and with the observation that its median survival at 20° is 3 days shorter than that of CL2a or Bristol-N2 (data not shown). Although the interpretation of such loci would be altered, this mode of mutagenesis greatly facilitates gene identification, since the inserted transposon would provide both an efficient means of gene identification (effectively transposon-tagging mutagenesis) and the potential to readily obtain revertants of the mutation—arising by spontaneous germ-line excision—which could then be tested for phenotypic reversion.

It is also of concern that short-lived alleles might have arisen during the period of laboratory cultivation for the Bergerac-BO mutator strain. If this were the case, however, then the shortest-lived alleles should be unique to just this one strain. On the contrary, at each of the 10 longevity loci currently mapped for all four strains (Fig 4), alleles of one or two of the other three strains conferred equivalent or lesser life span than the BO allele. It is thus unlikely that many of the Bergerac-BO alleles of longevity QTL arose in a benign laboratory environment wherein the only selective pressures may be those favoring rapid and early reproduction; rather, most or all such BO alleles must have appeared much earlier in the strain's natural history and hence would reflect genetic variation that has met the test of Darwinian selection.

	ACKNOWLEDGMENTS

We thank Paula Roberson for advice and assistance in assessing interactions by Cox's proportional hazards model. This work was supported by grant R01 AG091413 from the National Institute on Aging (National Institutes of Health) and by support from the Department of Veterans Affairs to R.J.S.R.

Manuscript received April 5, 2002; Accepted for publication July 11, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

AYYADEVARA, S., J. J. THADEN, and R. J. SHMOOKLER REIS, 2000  Anchor polymerase chain reaction display: a high-throughput method to resolve, score, and isolate dimorphic genetic markers based on interspersed repetitive DNA elements. Anal. Biochem. 284:19-28.[Medline]

AYYADEVARA, S., R. AYYADEVARA, S. HOU, J. J. THADEN, and R. J. SHMOOKLER REIS, 2001  Genetic mapping of quantitative trait loci governing longevity of Caenorhabditis elegans in recombinant-inbred progeny of a Bergerac-BO x RC301 interstrain cross. Genetics 157:655-666.[Abstract/Free Full Text]

BRENNER, S., 1974  The genetics of Caenorhabditis elegans. Genetics 77:71-94.[Abstract/Free Full Text]

CAREY, G. and J. WILLIAMSON, 1991  Linkage analysis of quantitative traits: increased power by using selected samples. Am. J. Hum. Genet. 49:786-796.[Medline]

Genome sequence of the nematode C. elegans: a platform for investigating biology. (1998) Science 282:2012-2018. (errata: Science 283: 35, 2103; 285: 1493).[Abstract/Free Full Text]

CHURCHILL, G. A. and R. W. DOERGE, 1994  Empirical threshold values for quantitative trait mapping. Genetics 138:963-971.[Abstract]

CLANCY, D. J., D. GEMS, L. G. HARSHMAN, S. OLDHAM, and H. STOCKER et al., 2001  Extension of life-span by loss of CHICO, a Drosophila insulin receptor substrate protein. Science 292:104-106.[Abstract/Free Full Text]

CYPSER, J. R. and T. E. JOHNSON, 1999  The spe-10 mutant has longer life and increased stress resistance. Neurobiol. Aging 20:503-512.[Medline]

DARVASI, A. and M. SOLLER, 1992  Selective genotyping for determination of linkage between a marker locus and a quantitative trait locus. Theor. Appl. Genet. 85:353-359.

DIXON, L. K., 1993  Use of recombinant inbred strains to map genes of aging. Genetica 91:151-165.[Medline]

DREYFUS, D. H. and S. W. EMMONS, 1991  A transposon-related palindromic repetitive sequence from C. elegans. Nucleic Acids Res. 19:1871-1877.[Abstract/Free Full Text]

DUHON, S. A., S. MURAKAMI, and T. E. JOHNSON, 1996  Direct isolation of longevity mutants in the nematode Caenorhabditis elegans. Dev. Genet. 18:144-153.[Medline]

EBERT, R. H., II, V. A. CHERKASOVA, R. A. DENNIS, J. H. WU, and S. RUGGLES et al., 1993  Longevity-determining genes in Caenorhabditis elegans: chromosomal mapping of multiple noninteractive loci. Genetics 135:1003-1010.[Abstract]

EBERT, R. H., II, M. A. SHAMMAS, B. H. SOHAL, R. S. SOHAL, and N. K. EGILMEZ et al., 1996  Defining genes that govern longevity in Caenorhabditis elegans. Dev. Genet. 18:131-143.[Medline]

EGILMEZ, N. K., R. H. EBERT, II, and R. J. SHMOOKLER REIS, 1995  Strain evolution in Caenorhabditis elegans: transposable elements as markers of interstrain evolutionary history. J. Mol. Evol. 40:372-381.[Medline]

EWBANK, J. J., T. M. BARNES, B. LAKOWSKI, M. LUSSIER, and H. BUSSEY et al., 1997  Structural and functional conservation of the Caenorhabditis elegans timing gene clk-1.. Science 275:980-983.[Abstract/Free Full Text]

FABIAN, T. J. and T. E. JOHNSON, 1994  Production of age-synchronous mass cultures of Caenorhabditis elegans.. J. Gerontol. 49:B145-B156.

FABRIZIO, P., F. POZZA, S. D. PLETCHER, C. M. GENDRON, and V. D. LONGO, 2001  Regulation of longevity and stress resistance by Sch9 in yeast. Science 292:288-290.[Abstract/Free Full Text]

FALCONER, D. S., and T. F. C. MACKAY, 1996 Introduction to Quantitative Genetics, pp.184–227. Addison Wesley Longman, Essex, England.

FELKAI, S., J. J. EWBANK, J. LEMIEUX, J. C. LABBE, and G. G. BROWN et al., 1999  CLK-1 controls respiration, behavior and aging in the nematode Caenorhabditis elegans.. EMBO J. 18:1783-1792.[Medline]

FELLER, W., 1968 An Introduction to Probability Theory and Its Applications, Vol. I, Ed. 3, pp. 43–47. Wiley, New York.

GALECKI, A. T., T. R. TEN HAVE, and G. MOLENBERGHS, 2001  A simple and fast alternative to the EM algorithm for incomplete categorical data and latent class models. Comput. Stat. Data Anal. 35:265-281.

GARNER, A., A. JAHOOR, J. SCHONDELMAIER, H. SIEDLER, and K. PILLEN, 1991  Construction of an RFLP map of barley. Theor. Appl. Genet. 83:250-256.

GEMS, D., A. J. SUTTON, M. L. SUNDERMEYER, P. S. ALBERT, and K. V. KING et al., 1998  Two pleiotropic classes of daf-2 mutation affect larval arrest, adult behavior, reproduction and longevity in Caenorhabditis elegans.. Genetics 150:129-155.[Abstract/Free Full Text]

ISHII, N., M. FUJII, P. S. HARTMAN, M. TSUDA, and K. YASUDA et al., 1998  A mutation in succinate dehydrogenase cytochrome b causes oxidative stress and ageing in nematodes. Nature 394:694-697.[Medline]

JOHNSON, T. E. and W. B. WOOD, 1982  Genetic analysis of life span in Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 79:6603-6607.[Abstract/Free Full Text]

KENYON, C., J. CHANG, E. GENSCH, A. RUDNER, and R. TABTIANG, 1993  A C. elegans mutant that lives twice as long as wild type. Nature 366:461-464.[Medline]

KIRCHMAN, P. A., S. KIM, C. Y. LAI, and S. M. JAZWINSKI, 1999  Interorganelle signaling is a determinant of longevity in Saccharomyces cerevisiae.. Genetics 152:179-190.[Abstract/Free Full Text]

KLASS, M. R., 1983  A method for the isolation of longevity mutants in the nematode Caenorhabditis elegans and initial results. Mech. Ageing Dev. 22:279-286.[Medline]

KRUGLYAK, L. and E. LANDER, 1995  A nonparametric approach for mapping quantitative trait loci. Genetics 139:1421-1428.[Abstract]

LAKOWSKI, B. and S. HEKIMI, 1996  Determination of life-span in Caenorhabditis elegans by four clock genes. Science 272:1010-1013.[Abstract]

LAKOWSKI, B. and S. HEKIMI, 1998  The genetics of caloric restriction in Caenorhabditis elegans.. Proc. Natl. Acad. Sci. USA 95:13091-13096.[Abstract/Free Full Text]

LANDER, E. S. and D. BOTSTEIN, 1989  Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121:185-199.[Abstract/Free Full Text]

LANDER, E. S., P. GREEN, J. ABRAHAMSON, A. BARLOW, and M. J. DALY et al., 1987  Mapmaker: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1:174-181.[Medline]

LEIPS, J. and T. F. C. MACKAY, 2000  Quantitative trait loci for life span in Drosophila melanogaster: interactions with genetic background and larval density. Genetics 155:1773-1788.[Abstract/Free Full Text]

LIN, K., J. B. BORMAN, A. RODAN, and C. KENYON, 1997  Daf-16: an HNF-3/forkhead family member that can function to double the life-span of Caenorhabditis elegans. Science 278:1319-1322.[Abstract/Free Full Text]

LIN, S. J., P. A. DEFOSSEZ, and L. GUARENTE, 2000  Requirement of NAD and SIR2 for life-span extension by calorie restriction in Saccharomyces cerevisiae.. Science 289:2126-2128.[Abstract/Free Full Text]

LIN, Y. J., L. SEROUDE, and S. BENZER, 1998  Extended life-span and stress resistance in the Drosophila mutant methuselah.. Science 282:943-946.[Abstract/Free Full Text]

LITHGOW, G. J., 2000  Stress response and aging in Caenorhabditis elegans.. Results Probl. Cell Differ. 29:131-148.[Medline]

LUCKINBILL, L. S., V. RIHA, S. RHINE, and T. A. GRUDZIEN, 1990  The role of glucose-6-phosphate dehydrogenase in the evolution of longevity in Drosophila melanogaster.. Heredity 65:29-38.

LYNCH, M., and B. WALSH, 1998 Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, MA.

LYTTLE, T. W., 1991  Segregation distorters. Annu. Rev. Genet. 25:511-557.[Medline]

MCCULLAGH, P., and J. A. NELDER, 1989 Generalized Linear Models, Ed. 2. Chapman & Hall, New York.

MITCHELL, D. H., J. W. STILES, J. SANTELLI, and D. RAO SANADI, 1979  Synchronous growth and aging of Caenorhabditis elegans in the presence of fluorodeoxyuridine. J. Gerontol. 34:28-36.

MOERMAN, D. G. and R. H. WATERSTON, 1984  Spontaneous unstable unc-22 IV mutations in C. elegans var. Bergerac. Genetics 108:859-877.[Abstract/Free Full Text]

MORRIS, J. Z., H. A. TISSENBAUM, and G. A. RUVKUN, 1996  A phosphatidylinositol-3-OH kinase family member regulating longevity and diapause in Caenorhabditis elegans.. Nature 382:536-539.[Medline]

OGG, S. and G. A. RUVKUN, 1998  The C. elegans PTEN homolog, DAF-18, acts in the insulin receptor-like metabolic signaling pathway. Mol. Cell 2:887-893.[Medline]

OGG, S., S. PARADIS, S. GOTTLIEB, G. I. PATTERSON, and L. LEE et al., 1997  The fork head transcription factor DAF-16 transduces insulin-like metabolic and longevity signals in C. elegans.. Nature 389:994-999.[Medline]

PARADIS, S., M. AILION, A. TOKER, J. H. THOMAS, and G. RUVKUN, 1999  A PDK1 homolog is necessary and sufficient to transduce AGE-1 PI3 kinase signals that regulate diapause in Caenorhabditis elegans. Genes Dev. 13:1438-1452.[Abstract/Free Full Text]

PARK, P. U., P. A. DEFOSSEZ, and L. GUARENTE, 1999  Effects of mutations in DNA repair genes on formation of ribosomal DNA circles and life span in Saccharomyces cerevisiae.. Mol. Cell. Biol. 19:3848-3856.[Abstract/Free Full Text]

PASYUKOVA, E. G., C. VIEIRA, and T. F. C. MACKAY, 2000  Deficiency mapping of quantitative trait loci affecting longevity in Drosophila melanogaster.. Genetics 156:1129-1146.[Abstract/Free Full Text]

PREISSER, J. S., A. T. GALECKI, K. K. LOHMAN, and L. E. WAGENKNECHT, 2000  Analysis of smoking trends with incomplete repeated responses. J. Am. Stat. Assoc. 95:1021-1031.

ROGINA, B., R. A. REENAN, S. P. NILSEN, and S. L. HELFAND, 2000  Extended life-span conferred by cotransporter gene mutations in Drosophila. Science 290:2137-2140.[Abstract/Free Full Text]

SHOOK, D. R. and T. E. JOHNSON, 1999  Quantitative trait loci affecting survival and fertility-related traits in Caenorhabditis elegans show genotype-environment interactions, pleiotropy and epistasis. Genetics 153:1233-1243.[Abstract/Free Full Text]

SHOOK, D. R., A. BROOKS, and T. E. JOHNSON, 1996  Mapping quantitative trait loci affecting life history traits in the nematode Caenorhabditis elegans.. Genetics 142:801-817.[Abstract]

TATAR, M., A. KOPELMAN, D. EPSTEIN, M.-P. TU, and C.-M. YIN et al., 2001  A mutant Drosophila insulin receptor homologue that extends life-span and impairs neuroendocrine function. Science 292:107-109.[Abstract/Free Full Text]

TISSENBAUM, H. A. and G. A. RUVKUN, 1998  An insulin-like signaling pathway affects both longevity and reproduction in Caenorhabditis elegans. Genetics 148:703-717.[Abstract/Free Full Text]

VAJO, Z., L. M. KING, T. JONASSEN, D. J. WILKIN, and N. HO et al., 1999  Conservation of the Caenorhabditis elegans timing gene clk-1 from yeast to human: a gene required for ubiquinone biosynthesis with potential implications for aging. Mamm. Genome 10:1000-1004.[Medline]

VAN VOORHIES, W. A., 1992  Production of sperm reduces nematode lifespan. Nature 360:456-458.[Medline]

VIEIRA, C., E. G. PASYUKOVA, Z-B. ZENG, J. B. HACKETT, and R. F. LYMAN et al., 2000  Genotype-environment interaction for quantitative trait loci affecting life span in Drosophila melanogaster.. Genetics 154:213-227.[Abstract/Free Full Text]

VONGS, A., T. KAKUTANI, R. A. MARTIENSSEN, and E. J. RICHARDS, 1993  Arabidopsis thaliana DNA methylation mutants. Science 260:1926-1928.[Abstract/Free Full Text]

WILLIAMS, B. D., B. SCHRANK, C. HUYNH, R. SHOWNKEEN, and R. H. WATERSTON, 1992  A genetic mapping system in Caenorhabditis elegans based on polymorphic sequence-tagged sites. Genetics 131:609-624.[Abstract]

WOLKOW, C. A., C. D. KIMURA, M. S. LEE, and G. A. RUVKUN, 2000  Regulation of C. elegans life-span by insulin like signaling in the nervous system. Science 290:147-150.[Abstract/Free Full Text]

XU, Y., L. ZHU, J. XIAO, N. HUANG, and S. R. MCCOUGH, 1997  Chromosomal regions associated with segregation distortion of molecular markers in F₂ backcross, doubled haploid, and recombinant populations of rice (Oryza sativa L.). Mol. Gen. Genet. 253:535-545.[Medline]

YANG, Y. and D. L. WILSON, 1999  Characterization of a life-extending mutation in age-2, a new aging gene in Caenorhabditis elegans.. J. Gerontol. A Biol. Sci. 54:B137-B142.

This article has been cited by other articles:

				A. Doroszuk, L. B. Snoek, E. Fradin, J. Riksen, and J. Kammenga A genome-wide library of CB4856/N2 introgression lines of Caenorhabditis elegans Nucleic Acids Res., June 19, 2009; (2009) gkp528v1. [Abstract] [Full Text] [PDF]

				J. Leips, P. Gilligan, and T. F. C. Mackay Quantitative Trait Loci With Age-Specific Effects on Fecundity in Drosophila melanogaster Genetics, March 1, 2006; 172(3): 1595 - 1605. [Abstract] [Full Text] [PDF]

				T. E. Johnson Genes, Phenes, and Dreams of Immortality: The 2003 Kleemeier Award Lecture J. Gerontol. A Biol. Sci. Med. Sci., June 1, 2005; 60(6): 680 - 687. [Abstract] [Full Text] [PDF]