Enhanced Fitness Conferred by Naturally Occurring Variation in the Circadian Clock

See allHide authors and affiliations

Science  07 Nov 2003:
Vol. 302, Issue 5647, pp. 1049-1053
DOI: 10.1126/science.1082971


Natural variation in clock parameters is necessary for the circadian clock to contribute to organismal fitness over a broad geographic range. Considerable variation is evident in the period, phase, and amplitude of 150 Arabidopsis accessions, and the period length is correlated with the day length at the latitude of origin, implying the adaptive significance of correctly regulated circadian timing. Quantitative trait loci analysis of recombinant inbred lines indicates that multiple loci interact to determine period, phase, and amplitude. The loss-of-function analysis of each member of the ARABIDOPSIS PSEUDO-RESPONSE REGULATOR family suggests that they are candidates for clock quantitative trait loci.

The circadian clock optimizes the relationship, or phase angle, between biological activities and the dawn and dusk, thereby allowing specific biological activities to occur at precise times of day, or phases (1). Hallmarks of the circadian clock are that it persists without environmental time cues and it maintains a period of about 24 hours, which is theorized to enhance fitness (2). Indeed, an intact circadian clock confers greater fitness in Arabidopsis thaliana, Drosophila melanogaster, and Tamias striatus (chipmunk) (35). In Synechococcus elongatus (a cyanobacterium), free-running period affects fitness (6). The Drosophila clock component PERIOD displays a robust latitudinal cline in the length of a Thr-Gly repeat, where the southern variant better maintains a 24-hour period at higher temperatures (7). Apparently, natural variation of the circadian clock can contribute to variations in fitness within specific environments (8).

We surveyed clock-mediated leaf movement for natural variation of three circadian parameters—period, phase, and amplitude— in a global collection of 150 A. thaliana accessions. The period is the time required to progress through one circadian cycle, the phase is the time when the leaves are maximally pointing upward, and the amplitude is the vertical distance that a leaf travels. Circadian parameters were evaluated by fast Fourier transform nonlinear least squares (9). Figure 1A illustrates two Arabidopsis accessions that display both period and phase differences. As a result of these differences, the maximal leaf position occurs at two distinct times of day in conditions of continuous light and temperature. This collection of accessions shows considerable variation in period (22.0 to 28.5 hours), phase (10.1 to 20.4 hours), and amplitude (14.13 to 113.90 arbitrary units) (Fig. 1, C, E, and G; table S1). One accession, Sierra Alhambra (Sah-0), displayed no circadian clock–mediated leaf movement (statistically and visually; table S1).

Fig. 1.

Natural variation in the Arabidopsis circadian clock. (A) Average leaf-movement rhythms (n = 6) for two Arabidopsis accessions, Fl-1 (period: 25.57 hours; phase: 17.63 hours) and Trinity (period: 23.79 hours; phase: 10.10 hours). (B) Average leaf-movement rhythms (n = 6) for the Col-Ler RIL parental lines, Col (period: 24.40 hours; phase: 16.02 hours) and Ler (period: 23.16 hours; phase: 16.57 hours). (C and D) Circadian period for (C) 150 Arabidopsis accessions and (D) 76 Col-Ler RILs. (E and F) Phase plotted against the strength of the rhythm for (E) 150 Arabidopsis accessions and (F) 76 Col-Ler RILs. Phase is normalized to period length and plotted on a 24-hour circadian clock face. Strength of the rhythm, expressed as relative amplitude error (RAE), is plotted along the radius with the strongest rhythms (RAE = 0) at the outer edge of the circle and weakest rhythms (RAE = 1) at the center. (G and H) Amplitude of the leaf movement rhythmsfor (G) 150 Arabidopsis accessions and (H) 76 Col-Ler RILs. Data represent at least two independent experiments.

Because the primary function of a circadian oscillator is to synchronize an organism with its specific environment, circadian parameters may show environmental dependence. We observed a positive relationship between the period of clock-mediated leaf movement and the recorded latitude of the Arabidopsis accessions, particularly at high (>50°) latitudes (fig. S1D). In Arabidopsis, latitudinal clines have been observed in sensitivity to light (10) and in rosette and seed size (11). The relationship between latitude and day length is curvilinear; accordingly, we calculated day length at the latitude of origin and observed a highly significant correlation between day length and period (Fig. 2A). There were no other significant correlations of any clock parameter (period, phase, or amplitude) with day length or with the latitude, longitude, or altitude of the accessions (fig. S1).

Fig. 2.

(A) Period is positively correlated with day length at the geographic site of origin among 150 Arabidopsis accessions. Day lengths at each latitude were calculated for the vernal equinox (21 March) and at monthly intervals through the summer solstice (21 June, the data presented). In each case, the correlation of day length with period was highly significant (P < 0.001). (B) Period and phase are negatively correlated in the Col-Ler RIL. (C) Period and phase are positively correlated among different clock mutants (regression line with positive slope) but are negatively correlated among individualswithin a genotype. The cca1-1 was previously described (30). lhy-20 and ztl-4 are previously uncharacterized T-DNA insertion alleles. ztl-4 (Salk 012440) has the T-DNA inserted in the second exon after the first nucleotide of codon 570. lhy-20 (Salk 031092) has the T-DNA inserted within the third intron, after nucleotide 622 (relative to the translational start). lhy-20 [period (±SD): 20.30 ± 0.64 hours; phase (±SD): 8.40 ± 1.46 hours], solid squares; cca1-1 (period: 22.30 ± 0.33 hours; phase: 9.29 ± 2.34 hours), open triangles; Col isogenic wild type for aprr9-1 (period: 24.49 ± 0.41 hours; phase: 16.13 ± 1.83 hours), solid circles; Col parent for lhy-20 and ztl-4 (period: 25.11 ± 0.57 hours; phase: 14.75 ± 2.76 hours), open circles; aprr9-1 (period: 24.59 ± 0.58 hours; phase: 19.39 ± 1.43 hours), x; ztl-4 (period: 27.64 ± 1.09 hours; phase: 26.92 ± 3.52 hours), solid triangles.

The striking correlation of period with day length suggests the adaptive significance of circadian timing, despite the recent spread and lack of genetic structure in Arabidopsis populations (12). However, this observation was made in plants that were grown at 22°C, which may be warmer than the temperature that northern accessions normally encounter. Similar period lengthening has been observed in higher latitude strains of Drosophila auraria (13). Why would an organism have a circadian clock with a period other than 24 hours? At higher latitudes, plants must continue to accurately entrain to the 24-hour day, despite the sharp increase in day length during the spring. In accordance with Aschoff's rule, pacemakers with periods of >24 hours enhance the ability to track dawn (13). Therefore, periods deviating from 24 hours should enhance seasonal acuity, particularly at high latitudes.

How might natural selection act on period length when the period is normally entrained to a 24-hour cycle? One might suspect that the entrained phase, not the period in constant environment, would be of adaptive significance. However, we observed no significant correlation between phase and latitude or phase and day length (fig. S1, B and E). Flowering in Arabidopsis is photoperiodic and the transition from vegetative to reproductive growth is accelerated by long days. We suspect that the lengthened period serves to delay the onset of flowering until later in the season; this could prove advantageous in avoiding late-spring cold weather, which is more common at higher latitudes. Delayed flowering also may reduce herbivore damage if herbivore activity were high or if Arabidopsis were one of the few species available as attractive forage in early spring (14).

To identify the loci that are responsible for natural genetic variation in circadian clock phenotypes, we took advantage of the Columbia (Col)-Landsberg-erecta (Ler) recombinant inbred lines (RILs) (15). The Col and Ler parental lines displayed only slight circadian differences (Fig. 1B; tables S1 and S2). Nonetheless, the 76 Col-Ler RILs (table S2) exhibited circadian variation for period, phase, and amplitude (Fig. 1, D, F, and H) as great as that observed among the accessions (Fig. 1, C, E, and G). Transgressive segregation, the emergence of extreme phenotypes in hybrid populations that exceed the variation found in parental lines, is common in plants (16) and has been noted in Arabidopsis for both flowering time (17) and circadian rhythms (18).

Circadian period and phase were inversely correlated in the Col-Ler RILs (Fig. 2B), although not among the collection of accessions (fig. S2). We detected no correlation between the period and amplitude or between the phase and amplitude for either population (fig. S2, C to F). The relationship between the period and phase among RILs suggests that the segregation of the quantitative trait loci (QTL) that lengthen the period is correlated with the segregation of the QTL that advance the phase. A positive correlation between the period and phase has been observed in circadian mutants under entraining conditions (19, 20) and in wild-type plants grown in altered light-dark cycles (21). Similarly, we observed a positive correlation between the period and phase in the population means of a number of clock mutants during continuous light (Fig. 2C), but we also observed an inverse relationship between the phase and period among individual seedlings of a given genotype, including wild-type and mutant plants with altered period and phase (Fig. 2C). The slopes of these regressions were not significantly different (F5,60 = 1.08, P > 0.35), but the midpoints of most regressions were significantly different (analysis of covariance; F5,65 = 96.55, P < 0.0001).

To locate the genes responsible for the observed natural variation, we mapped the QTL in the Col-Ler RIL population (Fig. 3). QTL studies have proven useful in confirming known clock loci as well as identifying previously unknown clock loci in Arabidopsis (18) and the mouse (22). Five QTL [two periods (TAU), one phase (PHI), and two amplitudes (AMP)] were significant (P < 0.05), as determined by the 1000 permutation test, and were numbered on the basis of chromosomal location: TAU1A, TAU1B, PHI5A, AMP2A, and AMP3B (Fig. 3A). Significant QTL account for 25.15% (TAU1A, 13.48%; TAU1B, 11.67%), 12.68% (PHI5B), and 42.72% (AMP2A, 33.56%; AMP3A, 9.16%) of the variation observed for the period, phase, and amplitude, respectively. Also, 10 suggestive QTL overlapped with regions of the genome that contained known circadian-associated genes (Fig. 3). By testing pairwise interactions between QTL, we detected no evidence for epistasis in our model (23). Period and phase QTL overlapped on the bottom of chromosome V, whereas other period QTL did not overlap with phase QTL (Fig. 3A). QTL analysis of the Col-Ler RIL population is consistent with complementary gene action, in which segregation of positive and negative alleles at multiple loci underlies transgressive segregation of the period, phase, and amplitude traits in the Arabidopsis circadian clock (Fig. 3B).

Fig. 3.

QTL map for circadian loci in Col-Ler RILs. (A) Likelihood of odds (LOD) scores for circadian period (TAU; red), phase (PHI; blue), and amplitude (AMP; green) plotted against the Arabidopsis chromosomes I, II, III, and V. Confidence levels 0.100 (green), 0.050 (pink), 0.025 (orange), and 0.010 (dark blue) were established with 1000 permutations with QTL cartographer software package. (B) Additive effects (+, Col; –, Ler) for period (red), phase (blue), and amplitude (green). (C) Map positions of genes relevant to the Arabidopsis circadian clock. cM, centiMorgans.

Period QTL, ANDANTE (AND), and ANOTHER ANDANTE (AAN) have been identified at the top of chromosome V (18) in the region where we observed TAU5A. ARABIDOPSIS PSEUDO-RESPONSE REGULATOR 7 (APRR7), a member of the TIMING OF CAB EXPRESSION 1 (TOC1)/APRR family (24), maps to this region (Fig. 3C). Accumulating evidence implicates the APRR genes in Arabidopsis clock function. TOC1/APRR1 was initially identified on the basis of the short period of a loss-of-function allele (25). mRNA abundance for each APRR is under clock control, with peak expression at distinct times of day (24). Overexpression of APRR9 shortens the period (26), overexpression of APRR1 lengthens the period (27) or yields arrhythmicity (28), and overexpression of APRR5 reduces the amplitude for multiple rhythms (29). To confirm that the APRRs function in the circadian clock and may account for natural variation between Col and Ler, we identified loss-of-function transferred DNA (T-DNA) insertion alleles for APRR3, APRR5, APRR7, and APRR9 (fig. S3). All experiments with T-DNA insertion alleles were performed against isogenic siblings to account for variation resulting from transformation. Three aprr7 alleles each lengthen the period of clock-mediated leaf movement by 1.5 to 2 hours without affecting the phase (Fig. 4, A and B; table S3). Two aprr5 alleles shorten the period by 1.5 to 2 hours, again without altering the phase (Fig. 4, C and D; table S3). Similarly, two appr3 alleles shorten the period but do not affect phase (Fig. 4, E and F; table S3). In contrast, one aprr9 allele confers a wild-type period but a phase that lags by 4 to 5 hours (Fig. 4, G and H; table S3). Each allele affects either the period or phase, but not both, which is consistent with our observations that the period and phase are not correlated among the accessions; together, these observations show that the period and phase are under different genetic controls.

Fig. 4.

T-DNA disruption of the APRRs results in defectsin clock-mediated leaf movement. Clock-mediated leaf movement (A, C, and E) and scatter plot of period versus RAE (B, D, and F) for [(A) and (B)] aprr7-3 (solid circles) and APRR7-3 (open squares); [(C) and (D)] aprr5-1 (solid circles) and APRR5-1 (open squares); [(E) and (F)] aprr3-1 (solid circles) and APRR3-1 (open squares). Clock-mediated leaf movement (G) and scatter plot of period versus circadian phase (H) of aprr9-1 (solid circles) and APRR9-1 (open squares).

We present evidence that the period of the circadian clock in Arabidopsis displays a great deal of environmentally dependent natural variation. Our observation of a latitudinal cline in the period of the Arabidopsis circadian clock is consistent with a primary role of the circadian clock in the synchronization of an organism with its periodic surroundings. We also demonstrate transgressive segregation of clock parameters in hybrids derived from two commonly studied accessions with very similar clock parameters, which would facilitate the exploitation of new ecological niches or competition in new environments (16). Loci such as the APRR family may act as primary sources of natural variation, allowing modest complementary positive and negative effects to modulate the circadian period and phase to enhance fitness in local environments.

Supporting Online Material

Materialsand Methods

Figs. S1 to S3

Tables S1 to S3


References and Notes

Stay Connected to Science

Navigate This Article