High Spontaneous Rate of Gene Duplication in Caenorhabditis elegans

SUMMARY

Gene and genome duplications are the primary source of new genes and novel functions and have played a pivotal role in the evolution of genomic and organismal complexity [1, 2]. The spontaneous rate of gene duplication is a critical parameter for understanding the evolutionary dynamics of gene duplicates; yet few direct empirical estimates exist and differ widely. The presence of a large population of recently derived gene duplicates in sequenced genomes suggests a high rate of spontaneous origin, also evidenced by population-genomic studies reporting rampant copy-number polymorphism at the intraspecific level [3–6]. An analysis of long-term mutation-accumulation lines of Caenorhabditis elegans for gene copy-number changes using array Comparative Genomic Hybridization yields the first direct estimate of the genome-wide rate of gene duplication in a multicellular eukaryote. The gene duplication rate in C. elegans is quite high, on the order of 10⁻⁷ duplications/gene/generation. This rate is two orders of magnitude greater than the spontaneous rate of point mutation per nucleotide site in this species and also greatly exceeds an earlier estimate derived from the frequency distribution of extant gene duplicates in the sequenced C. elegans genome.

RESULTS

Most of the recent progress in elucidating the role of gene duplications in the history of life has been the result of analyses of whole genomes using comparative genomics. Although genomes can provide a rich record of the history of gene duplications in a particular lineage, the population-genetic dynamics and selection pressures on duplicated genes remain poorly understood. The spontaneous gene duplication rate shapes the natural variance in gene copy-number and is an important parameter for understanding the early evolutionary dynamics of novel genes [7, 8]. Ultimately, the frequency of gene copy-number polymorphisms in genomes as well as their rate of fixation is determined by a combination of the spontaneous duplication rate and the probabilities of preservation or elimination of these changes by evolutionary forces such as natural selection, genetic drift, and various mutations [8, 9].

Estimates of the spontaneous rate of gene duplication come primarily from three sources: (i) calculations based on the abundance of very recent gene duplications in sequenced genomes [2, 10], (ii) calculations assuming mutation-selection balance where the fitness consequences of the duplication are known [11], and (iii) direct measurements on individual loci where gene copy-number differences result in a distinct phenotype or genotype [12–16]. With method (i), Lynch and Conery [17] utilized the distribution of synonymous site divergence between duplicate genes in several sequenced genomes to estimate a duplication rate of 0.1×10⁻⁸/gene/yr in D. melanogaster, 0.4×10⁻⁸/gene/yr in S. cerevisiae and 1.6×10⁻⁸/gene/yr in C. elegans, among others. Translating these rate estimates into duplications/gene/generation requires knowledge of the number of generations/year. For C. elegans, the rate of gene duplication was calculated to be similar to the synonymous substitution rate, and because the frequency of base substitutions in C. elegans has been estimated to be 2×10⁻⁹/site/generation in long term mutation-accumulation experiments (MA henceforth) [18], the gene duplication rate per generation based on the genomic data would then be on the order of 10⁻⁹ duplications/gene/generation. Method (ii) estimates the rate of gene duplication using the frequency of gene duplications in a population and population-genetic theory of mutation-selection balance. Using this approach, the rate of new gene duplications in the X-linked human dystrophin gene leading to Duchenne Muscular Dystrophy (DMD) was estimated to be ~10⁻⁵ duplications/gene/generation [11]. Direct empirical measures of the gene duplication rate based on method (iii) generally yield much higher values than those generated from those based on extant duplicates in sequenced genomes. For example, reports of locus-specific duplication rates in bacteria, Drosophila, and humans range from 10⁻³ to 10⁻⁷/gene/generation [11,12–16,19]. These estimates are based on a handful of loci and may not be representative of all duplicated loci in these genomes. The discrepancy between the genome sequence estimates and empirical measures is particularly stark in yeast. Bioinformatic analyses of the sequenced yeast genome suggested that the rate of gene duplication in yeast is half that of the per nucleotide base substitution rate [2]. However, whole-genome sequencing of S. cerevisiae MA strains has now revealed that the duplication rate per locus is ten thousand-fold higher than the base substitution rate [20]. The five orders of magnitude discrepancy in the rate of spontaneous gene duplication in preceding studies is likely due to a combination of the use of different gene loci, species, and approaches of quantification.

We used Comparative Genome Hybridization (CGH) to measure the spontaneous gene duplication and deletion rate in C. elegans using experimental evolution lines that were generated during a long-term MA experiment (Figure 1) by enforcing single-worm bottlenecks each generation to greatly reduce the efficacy of natural selection [21]. Under these conditions, nearly all mutations are able to accumulate in the genome largely independent of their fitness consequences, which enables an estimation of the rate of spontaneous mutations. Analyses of ten C. elegans MA lines (bottlenecked for an average of 432 generations) with NimbleGen CGH microarrays detected 14 duplicated and 11 deleted segments that were unique to particular MA lines (Tables 1 and 2, respectively). These duplications and deletions were verified by quantitative PCR (Tables S1 and S2). The 14 duplicated segments involved the complete and partial duplication [22] of 11 and 19 loci, respectively. The C. elegans genome contains approximately 20,400 protein coding genes (excluding alternative splice forms), so the probability that any given gene is duplicated at least partially is 30/(20,400 × 432 × 10) = 3.4 × 10⁻⁷/gene/generation. The eleven deleted segments resulted in complete or partial deletions of 19 ORFs and a deletion rate of 2.2 × 10⁻⁷/gene/generation.

Figure 1. Nimblegen CGH array duplication and deletion.

Each spot is a log₂ ratio of the fluorescence of the experimental DNA and the control DNA, arranged in linear order according to position on the sequenced chromosome. a, Duplication on Chromosome III of MA line 78. The DNA sequence of the breakpoint of this tandem duplication event is shown in Supplemental Figure 1C. b, Deletion on Chromosome II in MA line 18. c, Adjacent deletion and duplication on Chromosome III of MA line 99.

Table 1.

Characterization of 14 duplication events detected in ten mutation accumulation lines of C. elegans using CGH microarray analysis.

MA Line ID	Bottleneck Generations	Chromosome	Start Position	Stop Position	Length of Duplication (bp)	No. of ORFs (complete, partial)
2	438	V	18,507,783	18,519,661	11,878*	3(1,2)
18	464	V	10,445,133	10,455,580	10,448	3(1,2)
18	464	V	17,847,927	17,858,066	10,140	1(0,1)
29	468	IV	17,482,852	17,490,972	8,121	2(1,1)
29	468	X	12,763,189	12,767,835	4,647	2(1,1)
41	438	–	–	–	–	–
63	425	V	4,893	18,375	13,483	2(2,0)
63	425	X	3,559,284	3,567,765	8,482	2(0,2)
78	428	I	6,682,405	6,688,767	6,361*	2(0,2)
78	428	III	9,135,580	9,145,930	10,351*	5(4,1)
78	428	X	17,694,155	17,696,571	2,417	1(0,1)
83	385	IV	11,695,251	11,700,130	4,880	2(1,1)
84	465	–	–	–	–	–
94	367	III	813,463	819,305	5,843*	2(0,2)
99	464	I	10,716,364	10,721,038	4,675	2(0,2)
99	464	III	12,190,163	12,194,367	4,205	1(0,1)

Quantitative PCR results confirming these duplications are presented in Supplemental Table 1. Duplication lengths with an asterisk are based on the DNA sequence of duplication breakpoints shown in Supplemental Figures 1A through D. Other length estimates are minimum estimates based on the location of probes included in the duplicated region. The numbers of ORFs were based on Wormbase sequence version WS219.

Table 2.

Characterization of 11 deletion events detected in ten mutation accumulation lines of C. elegans using CGH microarray analysis.

MA Line ID	Bottleneck Generations	Chromosome	Start Position	Stop Position	Length of Deletion (bp)	No. of ORFs (complete, partial)
2	438	–	–	–	–	–
18	464	II	5,779,858	5,784,774	4,917	1(0,1)
29	468	X	12,759,841	12,761,557	7,717	1(0,1)
41	438	–	–	–	–	–
63	425	V	1	3,147	3,147	1(1,0)
78	428	V	7,382,127	7,384,417	2,290*	2(0,2)
78	428	X	12,111	12,925	815	0
78	428	X	17,698,889	17,718,629	19,741	5(3,2)
83	385	II	184	4,901	4,718	1(1,0)
83	385	IV	8,582,021	8,613,791	31,771	5(5,1)
83	385	IV	15,187,709	15,187,923	215	0
84	465	X	6,449,100	6,451,323	2,224*	1(1,0)
94	367	–	–	–	–	–
99	464	III	12,186,190	12,189,700	3,511	1(0,1)

Quantitative PCR results confirming these deletions are presented in Supplemental Table 2. Duplication lengths with an asterisk are based on the DNA sequence of deletion breakpoints shown in Supplemental Figures 1E and F. Other length estimates are minimum estimates based on the location of probes included in the deleted region. The numbers of ORFs were based on Wormbase sequence version WS219.

If only complete duplicates are taken into consideration, the average duplication rate per gene becomes 1.2 × 10⁻⁷/gene/generation (Bootstrap confidence 95% interval = 0.6 – 2.1 × 10⁻⁷/gene/generation). Both of these estimates of the gene duplication rate in C. elegans are quite high, about two orders of magnitude greater than the spontaneous rate of point mutation per nucleotide in this species (~ 10⁻⁹/site/generation) [18]. Additionally, our empirically determined rate of spontaneous gene duplication for experimental C. elegans MA lines is two orders of magnitude higher than that determined from analyses based solely on the frequency distribution of extant duplicates of varying evolutionary ages in the sequenced N2 genome [2]. Our direct gene duplication rate estimates may in fact be downwardly biased for two reasons, namely (i) that small duplications are likely to go undetected because the number of adjacent microarray probes signaling gene copy-number changes may not be sufficient for detection, and (ii) these CGH DNA microarrays are restricted to unique probes only and duplications of genes in recently duplicated regions, for instance by unequal crossing over, may not be detected. The genome-wide duplication and deletion rate reported here does not add much to the overall mutation rate per genome. The base substitution rate per genome in C. elegans is ≈ 0.1/genome/generation [18] and if we count each duplication and deletion as an independent mutation, then the duplication/deletion rate per genome/generation is 0.007, and 0.011 when the calculation is based on copy-number changes in individual ORFs.

If the duplication and deletion rates are homogeneous across MA lines, the number of copy-number changes per line is expected to be Poisson distributed. Two potential sources of bias in estimating the rate of gene duplication and deletion from MA experiments is that these rates might be subject to change, either due to mutations in recombination and repair genes or due to fitness-dependent differences in the rates [23]. These two sources of bias would result in a larger variance in gene copy-number changes than expected under the Poisson distribution. Nevertheless, the ratio of the variance to the mean in the number of gene duplications and deletions across different MA lines is close to random expectations (F-value = 1.13; p > 0.25) suggesting the lack of a significant contribution from these two sources.

The duplication lengths ranged from 2.4–13.9 kb with a median duplication size of 7 kb. Deletions ranged in length from 0.8–31.7 kb with a median value of 3.5 kb. The difference in the length distributions of duplications and deletions are marginally significant (Wilcoxon two-sample test, p = 0.05). However, small deletions are more likely to be detected relative to small duplications and this may have influenced the difference in the median length of duplication and deletions. The median duplicon size of 7 kb in this data set is significantly greater than the median duplication size of 1.4 kb [24] for extant evolutionarily young gene duplicates with low synonymous divergence in the sequenced genome of the N2 laboratory strain of C. elegans (Wilcoxon two-sample test; p < 0.0001). This discrepancy can be due to either one or a combination of three possibilities, namely, (i) duplications are contracting in length due to internal deletions subsequent to their origin, (ii) there is purifying selection against larger duplicates, and/or (iii) CGH arrays are biased in favor of detecting larger duplications.

The spontaneous duplications and deletions in the ten MA lines were spread across all six chromosomes in the C. elegans genome (Figure 2a). Four duplications appear to be coupled with adjacent deletions and two of these are located at the ends of chromosomes. In addition, four duplications appear to involve more than a single copy addition, usually resulting in three to four copies, but in one case, perhaps as many as eight copies according to the qPCR results. Using divergent primers at the end of duplicons, we sequenced the breakpoints associated with four duplications and two deletions (Figures S1a-f). We were not successful in sequencing the coupled and high copy-number duplications using this strategy which is only expected to yield results when the duplicated segments are adjacent and there are no further rearrangements associated with the copy-number change. The breakpoints indicate direct tandem duplications with little or no sequence identity at the ends of the duplicons (Figures S1a-d). Moreover, in some instances, several additional nucleotides have been inserted at the breakpoint (Figures S1a,i,j). One deletion appears to have been the result of unequal crossing-over (Figure S1e).

Figure 2. Chromosomal distribution of spontaneous duplications and deletions.

The horizontal lines represent the six chromosomes comprising the C. elegans genome. a, Location of 14 duplications and 11 deletions across ten mutation accumulation (MA) lines derived from a single hermaphrodite of a N2 laboratory isolate of C. elegans. Black shaded rectangles above and below the line denote the location of duplications and deletions, respectively. b, Location of inferred duplications and deletions in the N2 laboratory isolate of C. elegans that was the source of reference DNA in the CGH microarray experiments.

In addition to the copy-number changes unique to individual MA lines, we also observed six copy-number differences that are shared among all the MA lines. These comprise five duplications and one deletion ranging from 634 to 19,358 bp (Tables 3 and S3, Figures 2b and S1g-j). These differences represent copy-number changes between different N2 laboratory isolates of C. elegans, specifically the N2 laboratory strain that was used as a source of DNA in our CGH microarray experiments and the N2 laboratory strain that served as the ancestral stock for all the experimental MA lines established by Vassilieva and Lynch [21]. The deletion in the common N2 ancestor of all the MA lines was recently described as a common deletion found in strains that were subjected to mutagenesis with ethyl methanesulfonate and may in fact have been present in the genetic background of these strains prior to mutagenesis [25].

Table 3.

Characterization of duplication and deletion events detected in the common N2 ancestor of all MA lines and the reference strain of N2 used for hybridization against ten mutation accumulation lines of C. elegans for CGH microarray analysis.

Chromosome	Start Position	Stop Position	Length of Indel (bp)	No. of ORFs (complete, partial)
Duplications:
Va	2,995,387	2,999,015	3,628*	2(0,2)
Vb	18,706,963	18,726,320	19,358	3(2,1)
Vb	19,428,007	19,431,266	3,260*	1(0,1)
Xb	86,369	87,002	634	1(0,1)
Xb	7,510,066	7,523,734	13,668*	1(0,1)
Deletion:
Vc	1,645,712	1,647,498	1,786*	1(0,1)

Quantitative PCR results confirming these duplications and deletions are presented in Supplemental Table 3. Duplication lengths with an asterisk are based on the DNA sequence of duplication and deletion breakpoints shown in Supplemental Figures 1G through J. Other length estimates are minimum estimates based on the location of probes included in the duplicated region. The numbers of ORFs were based on Wormbase sequence version WS219.

^{a, c}correspond to a duplication and deletion event in the common N2 ancestor of all MA lines.

^bcorresponds to duplication events in the N2 reference strain used for the CGH microarray analysis.

DISCUSSION

The rate of fixation of duplicated genes due to beneficial, neofunctionalizing mutations has been shown to be dependent on the species’ effective population size as well as the rate of duplication [7, 26]. The direct estimates of gene duplication rates are two orders of magnitude greater than the per nucleotide point mutation rate. This may have important consequences for the role of adaptation in the evolution of duplicated genes. Theoretical and empirical work show that the mutation rate is an important determinant of the rate of fixation of adaptive mutations and that less-fit beneficial mutations can be fixed in the population earlier than the fittest mutation if the former are more frequent [27, 28]. For instance, if an adaptation to a novel environment requires an increase in the expression of a particular gene, and the gene duplication rate far exceeds the per nucleotide base substitution rate, advantageous duplications of the locus are more likely to occur and become fixed in populations before beneficial point mutations. This may explain why recent adaptations in natural populations have often involved an increase in gene dosage through gene duplication and amplification rather than regulatory base substitutions [29–31]. Once such adaptive duplications have become common or fixed, they become targets for mutations that increase the genetic repertoire of the organism. Were beneficial base substitutions more frequent than duplications, an increase in expression would more often be achieved by base substitutions rather than gene duplications. Hence, the relative rates of point mutations and duplications can play an important role in the evolutionary potential of genomes.

A large fraction of duplications do not span the coding sequence of genes in their entirety, and others are unlikely to capture the complete array of upstream regulatory sequences. This may predispose gene duplicates to subfunctionalization, as the first step in this process is the loss of an essential feature in one copy [22,24,32,33]. Moreover, failure to capture the full coding sequence or regulatory repertoire of the ancestral copy may predispose the duplicate copy to a different evolutionary trajectory wherein the ancestral copy is likely to retain its original function and the derived copy is more likely to be neofunctionalized, subfunctionalized, or pseudogenized. Indeed, recent analysis suggests that derived gene copies are evolving at faster rates relative to their ancestral counterparts [34,35].

All empirically-derived estimates of the spontaneous duplication/deletion rates, be they locus-specific [12–16,19] or genome-wide [20], are much greater than bioinformatically-derived estimates from extant duplicates in sequenced genomes for a diverse set of organisms across different kingdoms. This strongly suggests that most gene duplications are efficiently purged from the genome by purifying natural selection in their infancy, leaving a surviving observable pool dominated by duplicates with lower rates of loss. In fact, recent population-genetic analyses of gene copy-number polymorphism found an excess of rare duplications suggestive of purifying selection in Drosophila melanogaster [6]. Thus, prior genome-based estimates of the gene duplication rate may only reflect the birth rates of initially neutral or nearly neutral duplications. If this is the case, we predict that the discrepancy between bioinformatically- and empirically-derived estimates of the gene duplication rate will correlate positively with effective population size. In the case of the yeast Saccharomyces cerevisiae, the rate of spontaneous mutation has been measured as 0.7 × 10⁻⁹ substitutions/site/generation [20] and the parameter N_eμ is approximately 0.023 [36], giving an estimated N_e of 3.3 × 10⁷. This estimated N_e for S. cerevisiae is extremely similar to that measured for its close relative, S. paradoxus (≈10⁷) [37]. In the case of S. cerevisiae, with a large effective population size, the discrepancy between the bioinformatic and empirical estimates of the gene duplication rate [2,20] spans five orders of magnitude. In contrast, the discrepancy is only two orders of magnitude in the case of C. elegans, where the effective population size has been estimated as 9 × 10⁴ individuals [38]. However, it is possible that the present level of genetic variation in C. elegans and hence its small effective population size result from the recent evolution of hermaphroditism in this species [39]. For comparison, the estimated effective population size of C. remanei, an obligate outcrosser, is 1.6 × 10⁶ [40].

Most gene duplicates confer a slight penalty on the fitness of the carrier, possibly due to an initial dosage imbalance. Microorganisms and unicellular eukaryotes with their large effective population sizes and greater efficacy of selection may more effectively purge these newly arisen duplicates with their mildly deleterious effects. Conversely, the relatively smaller effective population sizes of many multicellular eukaryotes compromise their ability to efficiently rid their genome of these new entrants.