Seven geographical isolates of C. reinhardtii were employed in this study; their strain numbers, mating types, origins of isolation, and strain abbreviations are presented in Table 1. To access levels of genetic diversity we sequenced the complete mitochondrial genome and portions of 7 single-copy nuclear genes from each of the 7 isolates. A genetic map of the C. reinhardtii mitochondrial genome is shown in Figure 1, and partial genetic maps of the 7 nuclear genes are shown in Figure 2. We sequenced the entire mtDNA in order to employ both intergenic regions and synonymous sites in our calculations of π_silent – previous studies on genetic diversity in mitochondrial genomes, due to a paucity of intraspecific sequence data, have tended to use only synonymous sites for estimating π_silent. Moreover, whole mtDNA sequences from C. reinhardtii allow for the comparison of synonymous-site nucleotide diversity (π_syn) in the standard mitochondrial protein-coding genes to that of rtl, a mitochondrial open reading frame (ORF) in the C. reinhardtii mtDNA coding for a putative reverse-transcriptase-like protein 15. It has been suggested that synonymous sites in rtl are under less selective constraints than those of the standard mtDNA protein-coding genes and that they may be more appropriate for estimating the neutral mutation rate in the mitochondrial compartment 7. For the nuclear loci, we sequenced mostly introns rather than exons because it is believed that in the C. reinhardtii nuclear genome intronic sites are more neutrally evolving than synonymous sites and may give more reliable estimates of the neutral mutation rate 6. Sequences for two of the nuclear loci from the 7 isolates have been previously reported 161718 allowing us to confirm both our strain assignations and sequencing methods.

We were able to obtain the complete mtDNA sequence from an 8^th strain of C. reinhardtii (CC-503) by collecting and assembling mtDNA sequences that were generated from the C. reinhardtii nuclear-genome sequencing project 819. Both C. reinhardtii CC-503 and C. reinhardtii CC-277 (one of the 7 isolates described in Table 1) are cell-wall-less mutants recovered from the same "Ebersold-Levine" wild-type background of C. reinhardtii, but they have been separated for at least 35 years 20. The mtDNA sequence of C. reinhardtii CC-503 is identical to that of C. reinhardtii CC-277; and when we downloaded the sequences of the 7 nuclear loci for C. reinhardtii CC-503, they too were identical to those of C. reinhardtii CC-277. Therefore, for the purpose of this study we will be considering C. reinhardtii MA-1 as synonymous with C. reinhardtii CC-277 and CC-503.

Prior to this study a complete mtDNA sequence for C. reinhardtii was already available (Genbank accession number NC_001638); this sequence, which resulted from the accumulated efforts of multiple parties, mostly came from C. reinhardtii CC-277, or in some cases from strains having the same genetic background as C. reinhardtii CC-277. The mtDNA sequence of C. reinhardtii CC-277 presented here differs at 46 positions relative to NC_001638; because 44 of these 46 differences are also present in the mtDNA of the other six C. reinhardtii isolates described here and because the C. reinhardtii CC-277 mtDNA sequence from this study was shown to be identical to the C. reinhardtii CC-503 mitochondrial genome, we feel that our version of the C. reinhardtii CC-277 mtDNA is currently the most accurate and that the discrepancies between our sequence and NC_001638 are the result of sequencing errors in the latter.

It is important to note that our annotation of the C. reinhardtii mitochondrial genome (Figure 1) does not contain the so-called rRNA-coding modules L2b and L3a. In previous studies each these modules was presumed to code for a non-core region of the large-subunit (LSU) rRNA 21. However, because sequence homologs of L2b and L3a have not been identified in the mtDNA of close relatives to C. reinhardtii 72223, or in any other genome, we have classified these regions as intergenic DNA, and have treated them as such for all genetic analyses.

Summary statistics of the nucleotide diversity in C. reinhardtii are shown in Table 2. Two measures of nucleotide diversity were used to calculate variation within the C. reinhardtii mitochondrial and nuclear genomes: π, which is the average number of pair-wise nucleotide differences per site between sequences in a sample 24, and θ_W, which is based on the number of polymorphic sites in a sample of sequences but is independent of their frequency 25. With respect to both measures, the nuclear compartment shows significantly more silent-site nucleotide diversity than the mitochondrial compartment: net values of π_silent for the nuclear DNA and mtDNA were 31.96 × 10^-3 and 8.54 × 10^-3, respectively. Net values of θ_W at silent sites are slightly higher at 33.02 × 10^-3 for the nuclear compartment and 9.18 10^-3 for the mitochondrial compartment. In all cases, silent sites in the various nuclear loci show more diversity than the silent sites in the mitochondrial compartment. The only exception to this is the nuclear gene CBLP, which has less synonymous-site diversity (π_syn = 2.77 × 10^-3) than that of the mitochondrial protein-coding regions. Within the mitochondrial compartment, diversity at intergenic and synonymous sites is similar (8.92 × 10^-3 vs. 8.52 × 10^-3), as is the diversity of protein-coding regions and regions coding for structural RNAs (2.06 × 10^-3 vs. 2.42 × 10^-3). The mitochondrial gene rtl, which encodes a putative reverse transcriptase, shows more diversity than the other mitochondrial protein-coding genes when all 3 codon sites are considered (3.07 × 10^-3 vs. 2.06 × 10^-3) and slightly less diversity when looking only at synonymous sites (7.88 × 10^-3 vs. 8.52 × 10^-3); however, it is unlikely that these observations are statistically significant.

For both the nuclear and mitochondrial compartments, insertions and deletions (indels) represent a large proportion of the observed polymorphisms (Table 2). In our alignments of the nuclear loci from the 7 different strains of C. reinhardtii, 36% of mismatched nucleotides result from indels. The nuclear indels range from 1–31 nucleotides (nt) in length and have an average size of 4.5 nt. In the mitochondrial compartment indels represent 20% of the mismatched nucleotides. The mitochondrial indels range from 1–6 nt in length and have an average size of 2.5 nt. It is important to note that our estimates of nucleotide diversity shown in Table 2 are derived from sites in the alignment where all seven strains of C. reinhardtii have a nucleotide; therefore, sites corresponding to indels were removed from the alignment. If our methods for calculating π are modified to include indels (by counting each gap in the alignment as a nucleotide change) the overall values of π_silent in the nuclear and mitochondrial compartments become 49.27 × 10^-3 (± 4.89 × 10^-3) and 10.93 × 10^-3 (± 1.96 × 10^-3), respectively.

Two statistical tests were performed on the mitochondrial and nuclear datasets to examine for traces of selection: Tajima's D-test, which compares the average number of nucleotide differences between pairs of sequences to the total number of segregating sites 24, and the McDonald-Kreitman test, which compares the ratio of nonsynonymous to synonymous differences observed within a species to that observed between species 26. Tajima's D is slightly negative in all cases pertaining to the mitochondrial compartment and in most cases pertaining to the nuclear compartment, but it is slightly positive for a few of the nuclear loci (the exons of MAT3 and PDK, and the introns of CBLP, PETC, PDK, and ACTIN) (Table 2). In no case is Tajima's D-test statistically significant. The McDonald-Kreitman test was performed by comparing the ratio of nonsynonymous to synonymous polymorphisms within C. reinhardtii to the ratio of nonsynonymous to synonymous fixed differences between C. reinhardtii and Chlamydomonas incerta (one of the closest known non-interfertile relatives of C. reinhardtii 27) (Table 3) – this was done for all of the protein-coding regions surveyed in this study. Overall, no significant departures from neutral expectations were detected for any of the mitochondrial or nuclear loci, and in no case is the McDonald-Kreitman test statistically significant.

Three of the C. reinhardtii strains (PA-1, MA-2, and FL) have introns in their mtDNA (Figure 1). C. reinhardtii MA-2 has a single intron, inserted into cob; C. reinhardtii FL has 2 introns, one in the L5-rRNA-coding module (the L5-intron) and one in the L7-rRNA-coding module (the L7-intron); and C. reinhardtii PA-2 has 3 introns, two in cox1 and the L5-intron (note: the DNA sequence of the L5-intron in C. reinhardtii PA-2 is identical to that of C. reinhardtii FL). Of these introns only that of cob in C. reinhardtii MA-2 has been previously described 13. Like the intron of cob in C. reinhardtii MA-2, each of the 4 introns presented here has an ORF for which the deduced amino acid sequence shows similarity to a LAGLIDADG-type endonuclease. RT-PCR experiments confirm that all five introns, including their ORFs, are spliced-out in mature transcripts. Secondary-structure modelling suggests that the two introns in cox1 are group I introns belonging to subgroup D. Our analyses of the L5- and L7-introns suggest that they lack the core sequence and potential secondary structure necessary to be classified as either group I or group II introns; thus, at the present time they are considered highly-degenerate "unclassified" introns. A 35-nt duplicated portion of the L5-rRNA-coding module is found within the 5' end of the L5-intron; RT-PCR experiments validate that this segment is in fact a component of the intron. The insertion sites of the L5- and L7-introns within the C. reinhardtii mtDNA and the nature of the repeat found within the L5-intron are described in Figure 3A, B, and 3C, respectively. The insertion sites of the L5- and L7-introns in context to the C. reinhardtii LSU rRNA sequence are shown in Figure 3D.

Before we interpret our data on nucleotide diversity in relation to the mutational-burden hypothesis, let us first try to account for the values of π_silent that we observed. Overall, in C. reinhardtii we found 3.7-fold more nucleotide diversity at silent sites in the nuclear compartment than in the mitochondrial compartment; and when indels are taken into consideration, π_silent appears to be 4.5 times greater in the nuclear DNA compared to the mtDNA. Assuming that π_silent is equal to 2N_gμ, we can discuss our findings on nucleotide diversity in relation to μ and N_g.

In a recent study that compared silent-site substitution rates in the mitochondrial and nuclear genomes between C. reinhardtii and C. incerta, it was concluded that the mutation rate in the nuclear compartment of these taxa is approximately the same as that in the mitochondrial compartment 67. If is similar in both the nuclear DNA and mtDNA of C. reinhardtii, then it appears that differences in N_g would have to explain the disparity in nucleotide diversity that we observe between these genomes.

In order to arrive at the values of π_silent observed in this study, N_g would have to be higher for the nuclear genome than for the mitochondrial genome (again, assuming equal mutation rates); there are a few reasons why this might be the case. For the haploid alga C. reinhardtii, nuclear genes are inherited biparentally and mitochondrial genes are inherited uniparentally 28. As mentioned earlier, this is thought to reduce N_g in the mitochondrial compartment by approximately one-half relative to that in the nucleus 5. Thus, for C. reinhardtii we might expect a value of π_silent in the mitochondrial compartment to be around one-half of what is observed in the nuclear compartment. Uniparental inheritance also implies that the mtDNA of C. reinhardtii has less opportunity for recombination than the nuclear DNA, which may make the mitochondrial genome as a whole more susceptible to the effects of selective sweeps and purifying selection, both of which can reduce N_g, resulting in an even smaller than expected value of π_silent in the mitochondrial compartment 2930. One way to detect the influences of selection is with the McDonald-Kreitman test, where positive selection is inferred if the test returns a value for the neutrality index (NI) < 1, and purifying selection is indicated by NI > 1 293031. Although the results of the MacDonald-Kreitman test for C. reinhardtii versus C. incerta showed no significant departure from neutral expectations, values of NI were < 1 for rtl and the concatenated sequence of the standard mitochondrial protein-coding regions (Table 3), which might be an indication of positive selection. Moreover, Tajima's D test returned negative values for all of the mtDNA regions that were examined (Table 2) – negative values of Tajima's D test can be an indication of a recent selective sweep of a linked mutation 32 – but again these findings were not statistically significant.

A further consideration is that the mutation rate in the C. reinhardtii mitochondrial compartment may be slightly lower than that in the nuclear compartment. When Popescu and Lee 7 estimated μ to be similar for both the nuclear and mitochondrial genomes of C. reinhardtii, they used the synonymous substitution rate in rtl (which was about double that of the standard mitochondrial protein-coding regions) as an estimate of the neutral mutation rate in the mitochondrial compartment. In contrast, we found π_syn in rtl to be similar to that of the standard mitochondrial protein-coding regions (7.88 × 10^-3 vs. 8.52 × 10^-3), and although this could be an artefact of a small sample size, it might suggest that Popescu and Lee overestimated μ in the C. reinhardtii mtDNA.

Our estimates of π_silent in the nuclear and mitochondrial genomes of C. reinhardtii are approximately 32 × 10^-3 and 8.5 × 10^-3, respectively; and with indels factored in, these values become ~50 × 10^-3 for the nuclear DNA and about 11 × 10^-3 for the mtDNA. The only other estimates of nucleotide diversity in the C. reinhardtii nuclear and mitochondrial genomes that we could find were that of Lynch and Connery 1, who, using 11 kb of mostly noncoding nuclear DNA from C. reinhardtii strains MA-1 and MN, found π_silent in the nuclear compartment to be ~40 × 10^-3 (they estimated N_eμ to be ~20 × 10^-3), which is in agreement with our observations. With respect to other green algae, we are unaware of any reported estimates of within-population silent-site variation in either nuclear or mitochondrial genomes.

There is an overall lack of data on π_silent for unicellular species; however, the data that are available are relatively consistent with our results for C. reinhardtii. Among unicellular fungi, average values of π_silent in the nuclear and mitochondrial compartments (based on data for 3 different species) were estimated to be 50 × 10^-3 and 12 × 10^-3, respectively 13, which is comparable to what we observed in C. reinhardtii. Furthermore, the genera Paramecium and Trypanosoma appear to have ratios of π_{silent(mitochondrion)}/π_{silent(nucleus)} of approximately 0.4 3, which is within the range of our estimates for C. reinhardtii (0.2–0.25). All studies deriving measures of π_silent from the mitochondrial genome of unicellular eukaryotes have used 3 or fewer loci for their estimates and have tended to focus only on synonymous sites, which makes our analysis one of the most comprehensive for any mitochondrial genome from a unicellular eukaryote to date.

It is also of interest to compare the nucleotide diversity of C. reinhardtii to that of multicellular species. Our estimation of nucleotide diversity in the nuclear genome of C. reinhardtii is much greater than that observed for animals: ~9 times larger than the average estimate for mammals (3.6 × 10^-3), and 2 times greater than the average for invertebrates, which is believed to be < 14.8 × 10^-3 1. Our approximation of π_silent for the mitochondrial genome of C. reinhardtii is much smaller than that of animals: 0.2 times the average for mammals (40 × 10^-3) and 0.1–0.6 times the average for invertebrates (11 × 10^-3– 67 × 10^-3) 2. In comparison to land plants, C. reinhardtii has twice the amount of silent-site nucleotide variation in its nuclear genome (π_silent in land plants is estimated to be ~15 × 10–3), and although we are unaware of any reliable assessments of π_silent in land plant mtDNA, it is purported to be < 0.4 × 10^-3 3, which is less than 0.05 times what we observed for the C. reinhardtii mtDNA.

At first glance, our estimates of nucleotide diversity in the mitochondrial and nuclear genomes of C. reinhardtii appear contrary to what would be expected under the mutational-burden hypothesis. We found π_silent to be ~4 times greater in the nuclear compartment than in the mitochondrial compartment, but based on the streamlined nature of the C. reinhardtii mitochondrial genome in relation to its noncompact nuclear genome, one might expect the mutational-burden hypothesis to predict a greater value of π_silent for the mtDNA. However, before we conclude that our findings are in opposition to the mutation-burden hypothesis, we must first consider what the actual "encumbrance" of noncoding DNA is for the mitochondrial and nuclear genomes in C. reinhardtii.

As described earlier, the basic premise of the mutation-burden hypothesis is that noncoding DNA magnifies the target site for deleterious mutations, thereby, increasing the susceptibility of a genome to degenerative changes. The mutational disadvantage of noncoding DNA, therefore, is critically dependent on: 1) the number of nucleotides that are associated with gene function (n), and 2) the per-nucleotide mutation rate (μ). These two terms can be combined to define the overall mutational disadvantage (s), where s = nμ 333. It is predicted that the threshold in a genome below which noncoding DNA can proliferate is 2N_gs < 1, or alternatively 2N_gμ < 1/n 333. Although it is difficult to estimate n for intergenic regions, values of n for intronic regions can be predicted with reasonable confidence. For the spliceosomal introns of eukaryotic genomes, n is believed to be around 25 33, which gives a threshold for intron colonization in the nuclear compartment of 2N_gμ < 0.04. Because mitochondrial introns are self-splicing and do not rely on a spliceosome for excision they have more nucleotides that are critical for proper splicing; thus, we can conservatively say that for mitochondrial introns n is between 75–100 (see 34 for a review on mitochondrial-intron folding), giving a threshold for intron proliferation in the mitochondrial compartment of ~2N_gμ < 0.01. Based on the mutational-burden hypothesis, our estimates of 2N_gμ (i.e. π_silent) in the nuclear and mitochondrial compartments of C. reinhardtii, whether including or excluding the influence of indels, lie too close to the predicted thresholds for intron proliferation in these genomes to accurately forecast intron abundance.

We do not know what the encumbrance of intergenic regions is for either the nuclear or the mitochondrial genome of C. reinhardtii, but if it is (or was at some point in the past) substantially higher for the mtDNA than the nuclear DNA, it would indicate that the mitochondrial compartment is a less permissive environment for the proliferation of intergenic DNA. And although this is highly speculative, there is one reason why this might be the case. In the C. reinhardtii mitochondrial genome, mature-RNA transcripts are generated by precise endonucleolytic cleavage of long polycistronic precursor-messenger RNAs; it is believed that these immature transcripts are cleaved in regions of the RNA corresponding to intergenic sites in the mitochondrial genome, and that processing is critically dependent on the primary sequence and the secondary structure of these regions 9. This implies that the intergenic sites in the C. reinhardtii mitochondrial genome may have a large mutational burden associated with them, perhaps large enough to impose a barrier on the amplitude of intergenic mtDNA. Moreover, the polycistronic nature of the C. reinhardtii mitochondrial transcripts suggest that the regulatory elements within the intergenic mtDNA carry the increased burden of being responsible for many genes – a burden not typically associated with the monocystronic gene regulation of nuclear DNA 35.

If the burden of intronic and intergenic DNA is higher in the mtDNA of green algae and other protists, then we might expect to find very low values of π_silent in the mitochondrial genomes of species from these groups that have an abundance of intronic and intergenic sequences – for examples see references 123637.

It is worth noting that the C. reinhardtii chloroplast genome when compared to its mitochondrial counterpart has a similarly low density of introns but a substantially greater proportion of intergenic DNA 38. Using 1500 nt of chloroplast DNA (composed of the petA gene and a single intergenic region) we found no nucleotide polymorphisms in any of the 7 geographical isolates of C. reinhardtii employed in this study (our unpublished data). For a genome rich in intergenic DNA this value is consistent with what one might expect under the mutational-burden hypothesis.

An unforeseen consequence of this study is the discovery of 4 previously unreported C. reinhardtii mitochondrial introns, two of which (the L5- and L7-introns) are unusual. Although a detailed description of these introns is beyond the scope of this paper, they each contain a characteristic that is notable: 1) the L5-intron carries a 35 nt portion of the L5-rRNA-coding module within the 5 -end of its DNA sequence – this is an unprecedented feature for a mitochondrial intron. And 2) the insertion site of the L7-intron in relation to the C. reinhardtii LSU-rRNA secondary-structure model of Boer and Gray 21 corresponds to domain III (Figure 3); we believe this to be the first example (for any taxon) of an intron found in domain III of the LSU rRNA.

The discovery of 4 new optional introns in the C. reinhardtii mtDNA does not alter the notion that in C. reinhardtii the mitochondrial compartment is significantly more compact than the nuclear compartment: the mitochondrial genome of C. reinhardtii PA-1, the isolate with the most introns, is still ~67% coding compared to < 20% for the nuclear genome.

All of the C. reinhardtii strains employed in this study were obtained from the Chlamydomonas Center at Duke University in July of 2006, with the exception to C. reinhardtii CC-277, which was obtained from the same source in 1991. Clonal cultures of each strain were prepared from a single vegetative colony recovered on agar medium 20. For each of the seven strains, total genomic DNA was extracted using the DNeasy Plant Mini Kit (Qiagen, Germantown, MD), and total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen).

To verify that the 7 C. reinhardtii isolates had been assigned the correct strain numbers we compared our sequences of the ypt4-VI, ypt4-VII, and act-VIII introns for each isolate to those obtained in other studies 161718 from the corresponding isolates. The results confirmed that the 7 C. reinhardtii isolates employed here are the same as those used in previous reports.

A PCR-based approach was used to amplify the mtDNA and the nuclear loci examined in this study. PCR experiments were performed in High Fidelity Platinum SuperMix (Invitrogen, Carlsbad, CA) using total genomic DNA as the template. Reverse-transcriptase (RT) PCR reactions were performed with the SuperScript III One-Step RT-PCR System (Invitrogen) following the manufacturer's protocol. PCR and RT-PCR products were purified using the QIAquick PCR Purification Kit (Qiagen). The purified products were sequenced on both strands at the Macrogen sequencing facility, Rockville, MD, USA.

The complete mitochondrial genome of C. reinhardtii strain CC-503 (cw92 mt⁺) was obtained by collecting and assembling mtDNA sequences generated from the C. reinhardtii nuclear genome sequencing project 19. These sequences were acquired by blasting the complete C. reinhardtii mitochondrial genome against the following databases at the United States Department of Energy Joint Genome Institute (DOE JGI): C. reinhardtii v3.1-unplaced reads and C. reinhardtii v3.1-bonus scaffolds. Blast hits showing > 99% similarity to C. reinhardtii mtDNA were downloaded and assembled. All mitochondrial hits were subsequently checked against the draft nuclear genome of C. reinhardtii (v3.0 unmasked assembly) to insure that no nuclear mitochondrial DNA sequences (NUMTs) were collected. Our assembly of the downloaded mtDNA sequences contains over 500 chromatogram reads and gives a complete C. reinhardtii CC-503 mitochondrial genome with 15-fold coverage.

To ensure that the 7 nuclear-encoded genes employed in this study are present only once in the C. reinhardtii nuclear genome we blasted each of the 7 sequences against the C. reinhardtii draft nuclear genome 19. All 7 genes returned a single hit, which is consistent with the hypothesis that these genes are present in single copies in the C. reinhardtii nuclear genome. The mitochondrial sequences obtained from each of the seven strains were also blasted against the C. reinhardtii draft nuclear genome to confirm that they are not NUMTs: the blast results suggest that there are very few copies of mitochondrial sequences in the nuclear compartment and the few that are present are highly degenerate; therefore, we are confident that none of the mtDNA sequences presented in this study are NUMTs. The mitochondrial introns of cox1 and their secondary structures were identified using RNAweasel 34. DnaSP 4.0 39 was used for calculating all measures of genetic diversity. Nucleotide diversity and its standard deviation were calculated using equation 10.6 and 10.3 of Nei 40, respectively. Theta was calculated using equation 3 of Tajima 32. The MacDonald-Kreitman test 26 and Tajima's D test 24 were performed in DnaSP.

Nucleotide-sequence accession numbers for the sequences that were employed in this study are shown in Table 4.