Journal of Molecular Evolution

, 66:232

A Candidate Complex Approach to Study Functional Mitochondrial DNA Changes: Sequence Variation and Quaternary Structure Modeling of Drosophila simulans Cytochrome c Oxidase

Authors

  • Richard G. Melvin
    • Ramaciotti Centre for Gene Function Analysis, School of Biotechnology and Biomolecular SciencesUniversity of New South Wales
  • Subhash D. Katewa
    • Ramaciotti Centre for Gene Function Analysis, School of Biotechnology and Biomolecular SciencesUniversity of New South Wales
    • Ramaciotti Centre for Gene Function Analysis, School of Biotechnology and Biomolecular SciencesUniversity of New South Wales
Article

DOI: 10.1007/s00239-008-9078-4

Cite this article as:
Melvin, R.G., Katewa, S.D. & Ballard, J.W.O. J Mol Evol (2008) 66: 232. doi:10.1007/s00239-008-9078-4

Abstract

A problem with studying evolutionary dynamics of mitochondrial (mt) DNA is that classical population genetic techniques cannot identify selected substitutions because of genetic hitchhiking. We circumvented this problem by employing a candidate complex approach to study sequence variation in cytochrome c oxidase (COX) genes within and among three distinct Drosophila simulans mtDNA haplogroups. First, we determined sequence variation in complete coding regions for all COX mtDNA and nuclear loci and their isoforms. Second, we constructed a quaternary structure model of D. simulans COX. Third, we predicted that six of nine amino acid changes in D. simulans mtDNA are likely to be functionally important. Of these seven, genetic crosses can experimentally determine the functional significance of three. Fourth, we identified two single amino acid changes and a deletion of two consecutive amino acids in nuclear encoded COX loci that are likely to influence cytochrome c oxidase activity. These data show that linking population genetics and quaternary structure modeling can lead to functional predictions of specific mtDNA amino acid mutations and validate the candidate complex approach.

Keywords

Genetic hitchhikingCytochrome c oxidaseCandidate complex approach

Introduction

A continuing challenge is to describe the population genetic architectures within species and to identify, and order, the evolutionary forces responsible for the observed subdivision (Avise 1994). Due to its ease of enzymatic amplification with conserved primers, mitochondrial DNA (mtDNA) is frequently used to infer population genetic and biogeographic subdivision within species. A potential problem with studying the evolutionary dynamics of mtDNA is that classical population genetic techniques, such as the sliding window approach, will not identify selected substitutions because the low recombination rate results in genetic hitchhiking (Kaplan et al. 1989; Maynard-Smith and Haigh 1974). In this study we circumvent this problem and employ a candidate complex approach to study sequence variation of mitochondrial electron transport chain complex IV (cytochrome c oxidase; COX) in Drosophila simulans. This study is the first to make predictions about the functional significance of specific mutations based on a quaternary structure model of the Drosophila COX holoenzyme. It builds on the mtDNA sequence data of Ballard (2000) and nuclear DNA sequence data of Ballard et al. (2007b).

Two well-known modes of selection may operate on mutations that cause amino acid replacements (Charlesworth et al. 1993; Kreitman 1983; Nachman 1998). Background selection and accumulation of slightly deleterious mutations are likely to account for most of the observed polymorphism in the linked, rarely recombining mtDNA molecule (Charlesworth 1993). Under background selection rare deleterious mutations are quickly eliminated along with all linked variants, while slightly deleterious mutations remain at an intermediate frequency in the population and some may become fixed. Infrequently, positive selection will cause an advantageous mutation to be swept to high frequency. As the selected mutation increases in frequency linked neutral and slightly deleterious mutations will concomitantly increase in frequency. Positive selection may also occur in the mtDNA if a genotype is associated with a maternally inherited factor that imparts a selective advantage to the host such as the α-proteobacteria Wolbachia. Four strains of Wolbachia are known to infect Drosophila simulans (James and Ballard 2000).

Four of the five protein complexes of the electron transport chain are composed of both mtDNA and nuclear encoded subunits. The exception is complex II, which is assembled from four nuclear encoded subunits. Theory predicts that it is more likely for nuclear encoded mutations to compensate for slightly deleterious mtDNA mutations than vice versa (Rand et al. 2004). The effective population size of nuclear genes is greater than that of mitochondrial encoded genes under the majority of scenarios and the recombination rate in nuclear genes tends to be higher than in mtDNA (Hudson and Turelli 2003). The alternative hypothesis, that mutations in nuclear encoded genes are accommodated by mtDNA mutations (Schmidt et al. 2001), is less likely in Drosophila, where the mutation rate in single-copy nuclear DNA is very similar to that observed in mtDNA (Powell et al. 1986).

In this study we employ a candidate complex approach to study sequence variation in mitochondrial and nuclear encoded subunits of COX in D. simulans. COX is the final protein complex in the mitochondrial electron transport chain and is hypothesized to be the rate-limiting step in oxidative phosphorylation (Villani et al. 1998). In Drosophila, COX is composed of 12 protein subunits produced by three genes located in the mitochondrial genome and 9 nuclear genes that produce proteins that are imported into the mitochondrion (Das et al. 2004). The three mtDNA encoded subunits form the catalytic core of COX, and nuclear encoded subunits regulate activity and stabilize the COX dimer (Capaldi 1990). A limitation of this study is that selection may also be operating outside of the region surveyed.

The human commensal D. simulans has three mtDNA haplogroups (siI, -II, and -III) that are globally nonrandomly distributed. This nonrandom biogeographic distribution, in the face of human transport, implies that the mitotypes are locally adapted and/or under strong selection. This hypothesis is supported by studies showing that the distinct mtDNA types have differential fitness in perturbation cage studies, key life history traits, and cold coma recovery (Ballard and James 2004; Ballard et al. 2007b; James and Ballard 2003). Further, mitochondria have unequal fitness following microinjection and in studies of mitochondrial bioenergetics (de Stordeur 1997; Katewa and Ballard 2007). An alternate explanation is that the current biogeographic distribution is a result of a series of random colonization events.

Here we sequence the nine nuclear encoded COX genes, and their isoforms, from four fly lines that harbor siI mtDNA. We then perform a battery of statistical analyses on 15 lines to determine the genetic variation within and among mitotypes. A quaternary structure model of Drosophila COX is then constructed by homology to the bovine crystal structure (Tsukihara et al. 2003). We then utilize the model to test predictions on the functional significance of specific amino acid changes in the mtDNA and nuclear genomes, a subset of which may be investigated in future studies. To illustrate the application of the model we employ it to examine the experimentally determined COX activity data from the three mtDNA mitotypes and the structural and functional consequences of the tenured (Mandal et al. 2005) and levy (Liu et al. 2007) mutations in D. melanogaster.

Materials and Methods

Fly Lines

The D. simulans flies used in this study were collected in Honolulu (Hawaii) or in Nairobi (Kenya) during November 2004. The mtDNA haplogroup and Wolbachia infection status of each line were determined by PCR and sequencing (Ballard 2000; James and Ballard 2000). The flies from Hawaii were numbered 1-HW01, 1-HW05, 1-HW07, and 1-HW11. The flies from Kenya were 2-KY15, 2-KY17, 2-KY18, 2-KY21, 3-KY10, 3-KY12, 3-KY14, and 3-KY20 (Ballard et al. 2007b). The 1/2/3 refers to the mtDNA haplogroup, HW designates Hawaii, and KY Kenya. The two final numbers designate the specific line number. All lines from Hawaii were infected with the wHa strain of Wolbachia. Lines 3-KY10 and 3-KY20 were infected with the wMa strain. All other lines were uninfected.

Sequence Variation

We determined the nucleotide variation in the three mitochondrial and nine nuclear (including four isoforms) COX loci from the four Hawaii fly lines. Heterozygous nucleotide positions and insertion/deletions were identified using the default “call secondary peaks” option of Sequencher 4.5 (Gene Codes, Ann Arbor, MI, USA). Sequence chromatograms of both strands were checked manually to confirm putative heterozygous and insertion/deletion sites.

Eleven additional sequences were included in the statistical analyses. DNA sequences of COX loci from eight Kenya lines are presented by Ballard et al. (2007b). Further, one additional line of each mtDNA type was obtained through NCBI Blast searches against the D. simulans genome (1-NC48, 2-MD106, and 3-MD199: NCBI Genome Project 12463; http://www.ncbi.nlm.nih.gov/projects/genome/seq/BlastGen/BlastGen.cgi?pid=12463).

Nucleotide variation of mitochondrial and nuclear encoded COX loci was compared within and among mtDNA haplogroups. Eleven of the nuclear encoded genes occur on autosomes, and two (COX 6A-3 and COX 6B) occur on the X chromosome. The number of alleles and synonymous and nonsynonymous nucleotide substitutions was calculated for each gene using DnaSP 4.0 (Rozas and Rozas 1999). Nucleotide diversity (π) and the neutral parameter (θ) were based on the number of segregating sites. Ks and Ka were calculated per synonymous and nonsynonymous site, respectively. To test for consistency of the observed mutation patterns with a neutral model of molecular evolution, we employed Tajima’s D (Tajima 1989) and Fu and Li’s F* (Fu and Li 1993). The assumption of these tests that the sample is taken from a single randomly mating population is violated in this study because the lines were collected in Hawaii, Kenya, Madagascar, and New Caledonia. No sympatric populations are known to harbor all mtDNA haplogroups.

COX Quaternary Structure Modeling and Functional Predictions

We constructed a D. simulans COX quaternary structure model from the inferred amino acid sequence of each COX gene. Amino acid sequences were inferred from nucleotide sequences using DnaSP 4.0 (Rozas and Rozas 1999) and aligned using T-Coffee (Notredame et al. 2000). Mitochondrial targeting sequences were identified using Mitoprot (Claros and Vincens 1996) and trimmed before modeling. The tertiary structure of each COX protein subunit was modeled by homology to the crystal structure of bovine COX (PDB ID 1v54) using SwissModel (Schwede et al. 2003). The three-dimensional quaternary structure of D. simulans COX was visualized and amino acid polymorphism highlighted using PyMol software (Delano Scientific, Palo Alto, CA, USA).

We used the quaternary structure model of D. simulans COX to predict whether a specific amino acid change, or insertion deletion event, is likely to affect COX activity and plausibly have phenotypic consequences. An amino acid change was considered likely to affect COX activity if physical or chemical properties were changed and/or the site was in close contact with another subunit. Distance between amino acid residues was determined using Deep View/Swiss-PDB Viewer (Guex and Peitsch 1997). Amino acids ≤ 4 Å apart were considered to be in contact (Martin et al. 1997; Schmidt et al. 2005).

Results and Discussion

Background selection and accumulation of slightly deleterious mutations are likely to account for most of the observed polymorphism in mtDNA (Charlesworth et al. 1993; Nachman 1998). In rare cases positive selection will cause an advantageous mutation to be swept to high frequency. Both of these forms of selection, in addition to selection for changes that optimize mitonuclear interactions, have the potential to influence the evolutionary dynamics of nuclear encoded subunits of the electron transport chain (Rand et al. 2004; Schmidt et al. 2001). To investigate the pattern of amino acid mutations in a composite OXPHOS complex, we used a candidate complex approach to study sequence variation in mtDNA and nuclear encoded COX genes among three distinct D. simulans mtDNA haplogroups.

Sequence Variation

For mtDNA, there is subdivision in COX loci as has been reported previously (Ballard 2000). Overall, sequencing studies of 15 lines identified a total of 9 nonsynonymous changes and 118 synonymous changes in the three mtDNA COX loci (Tables 1 and 2). Of these nine amino acid changes, zero are shared, three are fixed in siI, zero polymorphic in siI, three are fixed in siII, three polymorphic in siII, and zero fixed or polymorphic in siIII (Table 1). For the nuclear DNA there is no evidence of subdivision associated with the mtDNA. In the 13 nuclear encoded genes, 23 nonsynonymous changes, 1 two-amino acid deletion, and 101 synonymous mutations were detected. Of these 23 amino acid changes, 4 are shared among mtDNA haplogroups, 0 are fixed in siI, 4 polymorphic in siI, 0 are fixed in siII, 8 polymorphic in siII, and 0 fixed and 7 polymorphic in siIII (Table 1).
Table 1

Nucleotide polymorphisms in Drosophila simulans mitochondrial and nuclear cytochrome c oxidase genes from 15 isofemale lines

Location

Subunit (flybase)

L (bp)

Nal

Coding region

Noncoding region

Poly. sites

Nucleotide diversity

Neutrality test

Poly. sites

Nucleotide diversity

Neutrality test

Syna

Non-syna

π (x E-2)b

θ (x E-2)b

Ks (x E-2)b

Ka (x E-2)b

Tajima’s Db

Fu & Li’s F*b

Silenta

π (x E-2)b

θ (x E-2)b

Tajima’s Db

Fu & Li’s F*b

mtDNA

I

1540

6

71

3

2.27

1.55

19.89

0.44

2.05*

1.88*

     

2

1

0

0.03

0.04

18.35

0.26

−0.61

−0.48

     

3

5

1

0.16

0.19

22.64

0.53

−1.15

−1.18

     

1

0

0

0

N/A

18.35

0.51

N/A

N/A

     
 

63

2

           

II

685

5

21

6 (5)

1.81

1.24

17.63

0.91

1.96

1.80*

     

1

0

0

0

N/A

19.62

0.93

N/A

N/A

     

2

0

1

0.06

0.07

16.76

1.25

−0.82

−0.77

     

2

1

0

0.06

0.07

16.85

0.56

−0.82

−0.48

     
 

20

5 (4)

           

III

789

5

28

1

1.61

1.16

18.08

1.08

1.70

1.46

     

1

0

0

0

N/A

16.48

1.07

N/A

N/A

     

3

2

1

0.15

0.18

19.89

1.10

−1.05

−1.05

     

1

0

0

0

N/A

17.59

1.07

N/A

N/A

     
 

26

0

           

Nuclear

IV–1 (CG10664)

608

12

8

3

0.46

0.62

6.52

1.53

−1.01

−1.10

4

1.78

2.02

−0.38

0.61

4

3

3

0.59

0.53

6.86

1.40

0.76

0.79

2

1.31

1.57

−0.97

−0.95

4

2

1

0.26

0.26

6.12

1.59

−0.17

−0.18

3

2.30

2.36

−0.17

−0.18

5

5

1

0.48

0.53

6.57

1.59

−0.67

−0.69

2

1.64

1.57

0.24

0.24

 

0

0

      

0

    

IV–2 (CG10396)

531

4

4

0

0.1

0.23

9.33

2.21

−1.82*

−2.52*

     

2

1

0

0.08

0.09

9.27

2.21

−0.82

−0.77

     

2

2

0

0.15

0.18

9.43

2.21

−0.97

−0.95

     

2

1

0

0.08

0.09

9.28

2.21

−0.82

−0.77

     
 

0

0

           

Va (CG14724)

450

6

5

3

0.37

0.56

9.61

0.06

−1.28

−1.78

     

2

1

2

0.27

0.32

9.28

0.12

−1.05

−1.05

     

5

4

1

0.49

0.53

10.18

0.06

−0.56

−0.58

     

3

2

0

0.26

0.24

9.3

0.00

0.59

0.50

     
 

0

0

           

Vb–1 (CG11015)

1160

8

4

3

0.41

0.57

7.27

0.22

−0.93

−1.23

88

2.87

3.91

−1.12

−1.37

4

1

3

0.32

0.45

7.70

0.15

−1.43

−1.62

52

2.99

3.09

−0.18

0.01

4

2

1

0.44

0.40

6.21

0.29

0.70

0.70

30

1.80

1.97

−0.62

−0.67

3

1

1

0.24

0.24

7.67

0.24

−0.05

0.04

42

2.27

2.55

0.71

−0.68

 

0

0

      

0

    

Vb–2 (CG11043)

524

9

16

3

0.96

1.33

8.31

0.69

−1.17

−1.40

4

1.45

2.73

−1.61

−1.83

2

2

0

0.22

0.24

7.64

0.58

−0.71

−0.60

0

0.00

0.00

N/A

N/A

3

8

0

0.71

0.83

8.25

0.63

−0.99,

−0.88

3

2.37

2.44

−0.17

−0.18

4

14

3

1.88

2.01

9.00

0.87

0.67

−0.67

2

1.70

1.85

−0.71

−0.60

 

0

0

      

0

    

VIa–1 (CG17280)

498

5

4

0

0.33

0.39

8.27

1.21

−0.56

−0.59

8

1.18

1.74

−1.25

−1.70

3

3

0

0.51

0.50

8.86

1.21

0.17

0.15

2

0.68

0.74

−0.71

−0.60

3

2

0

0.31

0.33

7.90

1.21

−0.71

−0.60

1

0.34

0.37

−0.61

−0.48

3

2

0

0.25

0.29

8.09

1.21

−0.97

−0.95

7

2.16

2.27

−0.33

−0.35

 

0

0

      

0

    

VIa–2 (CG14077)

1041

12

26

7

0.88

1.26

13.30

2.89

–1.31

–1.53

23

4.04

4.62

–0.54

0.06

3

9

0

0.55

0.58

13.69

2.83

–0.49

–0.49

5

1.46

1.60

–0.80

–0.75

5

20

6

1.41

1.47

14.32

2.95

–0.29

–0.31

17

4.53

5.08

–0.80

–0.78

5

13

2

0.75

0.85

12.88

2.83

–0.81

–0.69

20

5.21

5.90

–0.87

–0.87

 

0

0

      

0

    

VIa–3 (CG30093)

285

8

7

1

0.67

0.91

14.64

0.50

–1.02

–1.22

     

3

3

0

0.64

0.57

15.09

0.47

1.09

0.97

     

4

2

3

0.70

0.77

14.35

058

–0.78

–0.72

     

3

4

0

0.63

0.67

14.44

0.47

–0.41

–0.42

     
 

0

0

           

Vib (CG14235)

535

2

1

0

0.05

0.11

1.87

0

–1.16

–1.51

1

0.06

0.13

–1.15

–1.48

1

0

0

0.00

N/A

1.74

0

N/A

N/A

0

0

N/A

N/A

N/A

1

0

0

0.00

N/A

1.74

0

N/A

N/A

0

0

N/A

N/A

N/A

2

1

0

0.14

0.17

2.09

0

–0.82

–0.77

1

0.17

0.20

–0.82

–0.77

 

0

0

      

0

    

Vic (CG14028)

384

3

4

0

0.67

0.57

4.92

0.59

0.57

0.48

3

0.98

0.67

1.52

1.37

2

0

0

0.43

0.47

8.22

0.59

−0.71

−0.60

0

0

N/A

N/A

N/A

1

0

0

0

N/A

3.27

0.59

N/A

N/A

0

0

N/A

N/A

N/A

1

0

0

0

N/A

3.27

0.59

N/A

N/A

0

0

N/A

N/A

N/A

 

2

0

      

3

    

VIIa (CG18193)

373

13

18

4

1.50

2.15

10.19

0.78

−1.25

−1.62

7

2.71

3.24

−0.57

−0.19

4

9

2

1.52

1.68

10.03

0.75

−0.65

−0.69

3

2.11

2.53

−1.05

−1.05

5

11

2

1.87

2.13

11.07

0.92

−0.89

−0.85

3

3.79

4.14

−0.56

−0.58

4

7

1

1.08

1.22

9.51

0.67

−0.81

−0.85

3

2.07

2.48

−1.05

−1.05

 

0

0

      

0

    

VIIc (CG2249)

454

3

2

0

0.34

0.33

9.07

0

0.10

−0.32

2

0.40

0.58

−0.91

−0.61

2

1

0

0.30

0.24

8.95

0

1.22

1.16

0

0

N/A

N/A

N/A

3

2

0

0.51

0.55

9.15

0

−0.71

−0.60

13

2.93

2.77

0.58

0.79

2

1

0

0.34

0.28

9.15

0

1.63

1.28

7

1.50

1.49

0.04

0.38

 

0

0

           

VIII (CG7181)

287

2

2

0

0.34

0.30

0.1

0

0.30

0.87

3

1.56

1.11

1.20

1.25

1

0

0

0

N/A

0

0

N/A

N/A

1

0.72

0.58

1.22

1.16

3

0

0

0.40

0.47

0.81

0

−0.97

−0.95

3

1.93

1.74

0.70

0.70

2

2

0

0.59

0.47

1.62

0

1.46

1.43

3

2.17

1.74

1.57

1.58

 

0

0

      

0

    

Note. The data set includes five lines with siI, five with siII, and five with siIII mtDNA. Three mitochondrial and 13 nuclear encoded genes are included. L, length in nucleotides; Nal, number of alleles; Poly. sites, polymorphic sites; Syn, synonymous substitutions; Nonsys, nonsynonymous substitutions; N/A, not available. Nucleotide diversity (π) and the neutral parameter (θ) were based on the number of segregating sites. Ks and Ka were calculated per synonymous and nonsynonymous site, respectively. To test whether the observed mutation patterns are consistent with a neutral model of molecular evolution, we employ Tajima’s D and Fu and Li’s F* (for synonymous sites); *significant at p < 0.05. We used D. melanogaster as an outgroup

aResults are given as follows: first row, total; second row, within siI; third row, within siII; fourth row, within siIII; fifth row, fixed among mtDNA types. Complex mutations at the first (343A > T) and second (344A > C) codon positions cause the predicted Asn > Ser replacement in siI flies

bResults are given as follows: first row, total; second row, within siI; third row, within siII; fourth row, within siIII

Table 2

Number of nonsynonymous and synonymous nucleotide changes in COX genes located in mtDNA, autosomes, and the X chromosome

Genomic location

Site type

Nonsynonymous

Synonymous

mtDNA

Total

9

118

Fixed between haplogroups

6

109

Segregating within haplogroup

3

9

Autosomes

Total

22

93

Fixed between haplogroups

0

2

Segregating within haplogroup

22

91

X chromosome

Total

1

8

Fixed between haplogroups

0

0

Segregating within haplogroup

1

8

To investigate the pattern of mutations across mtDNA, across autosomal DNA, and on the X chromosome, a contingency test was used to test the null hypothesis that the proportion of nonsynonymous substitutions is independent of whether the substitutions occur in the three genome locations (Table 2). In the mtDNA, the number of nonsynonymous substitutions is lower than expected or the number of synonymous substitutions is elevated (χ22 = 7.89, G2p = 0.02 for comparison of mtDNA, autosomal DNA, and X chromosome, and Fisher’s exact test p = 1.0 for comparison of the autosomal and X chromosome data). To investigate this result further a Fisher’s exact test was used to test the null hypothesis that the proportion of nonsynonymous substitutions is independent of whether the substitutions are fixed between or segregating within mtDNA haplogroups. If we assume that synonymous mutations are neutral or very slightly deleterious, the combined results suggest that there is an excess of nonsynonymous mutations that are segregating within haplogroups and they are being quickly removed from the population (Fisher’s exact p = 0.04). Dean and Ballard (2005) came to a similar conclusion using maximum likelihood models.

To investigate the pattern of synonymous changes further we employed Tajima’s D and Fu and Li’s F* (Table 1). For the three adjacent mitochondrial encoded genes combined, Tajima’s D and Fu and Li’s F* are significantly positive (1.83, p < 0.02, and 2.00, p < 0.05, respectively). This result indicates an excess of old mutations among but not within mtDNA types. In the nuclear genes Tajima’s D and Fu and Li’s F* suggest that only subunit 4 isoform 2 shows a significant negative departure from a strictly neutral model of molecular evolution (Table 1). However, this result does not provide a compelling case for strong selection on this locus for two reasons. First, the sample is not taken from a single randomly mating population. Second, Tajima’s D and Fu and Li’s F* tend to be slightly negative for the majority of loci consistent with an expanding population (Depaulis et al. 2003). Twelve nuclear loci do not differ significantly from a neutral model of molecular evolution and have generally negative values of Tajima’s D and Fu and Li’s F* (mean Tajima’s D = −0.73, mean Fu and Li’s F* = −0.91).

Quaternary Structure Modeling and Functional Predictions

Quaternary structure modeling suggests that six of the nine mtDNA mutations will influence COX activity (Table 3). Surprisingly, no nuclear encoded mutations interact with any polymorphic mtDNA mutations and only 2 of the 23 amino acid polymorphisms and the heterozygous deletion are predicted to affect COX activity (Table 4). The quaternary structure model of D. simulans COX is shown in Fig. S1 and that of the mtDNA encoded subunits in Figs. 1a–c. The Pymol project file for the whole model may be downloaded at http://billb.babs.unsw.edu.au/2007cox.htm. The PDB files for each subunit may be obtained from the corresponding author.
Table 3

Structural and functional amino acid changes in Drosophila simulans cytochrome c oxidase mtDNA subunits that are fixed in one mitotype (siI, -II, or -III) or polymorphic within a mitotype

 

Subunit

mtDNA position

Amino acid change

Structural

Distance (Å)a

Interaction

Affect function

Fixed

    SiI

COI

2827

Val(451) → Ile

No

>4

Unlikely

    siI

COI

2882

Tyr(468) → Phe

Yes, polar → nonpolar

<4

COX8

Likely

    SiI

COII

3431

Asn(114) → Ser

Yes

<4

Cytochrome c

Likely

3432b

    SiII

COII

3473

Thr(129) → Ser

No

>4

Unlikely

    SiII

COII

3477

Thr(130) → Ile

Yes, polar → nonpolar

<4

COX4, cytochrome c

Likely

    SiII

COII

3581

Val(165) → Ile

Yes

>4

Unlikely

Polymorphicc

    2-KY17

COIII

5092

Ser(116) → Leu

Yes, polar → nonpolar

>4

Cytochrome c

Likely

    2-MD106

COI

1558

Val(28) → Ile

Yes

<4

COX7C

Likely

    2-MD106

COII

3441

Phe(118) → Tyr

Yes, nonpolar → polar

<4

Cytochrome c

Likely

aDistance to nearest amino acid residue in an adjacent COX subunit

bComplex mutations at the first (3431A > T) and second (3432A > C) codon positions cause the predicted Asn > Ser replacement in siI flies

cPolymorphic sites are listed by line in which they occur. The number preceding the line name indicates the D. simulans mtDNA type

Table 4

Polymorphic amino acid changes in structural domains of the nuclear encoded Drosophila simulans cytochrome c oxidase subunits and their predicted likelihood of affecting COX structure or function

Line(s)a

COX subunit (ID)

Amino acid pos.b

Amino acid change

Affect structure

Distance (Å)c

Interaction

Affect function

1-HW01d

5B-1 (CG11015)

28

Met→Leu

Yes

<4

COIII

Likely

1-HW07

7A (CG18193)

28

Ile→Phe

Yes

<4

COIII

Likely

1-HW01

7A (CG18193)

85 & 86

ΔTrp & ΔVal

Yes

<4

COI & COIII

Likely

1-HW07

4-1 (CG106644)

88

Ala→Thr

No

>4

Unlikely

1-HW11

2-KY18

2-KY21

3-KY10

3-MD199

1-NC48

4-1 (CG106644)

120

Ala→Ser

No

<4

COI

Unlikely

1-HW01

4-1 (CG106644)

163

Met→Ile

No

>4

Unlikely

1-HW07

1-HW11

1-NC48

5B-1 (CG11015)

19

Ala→Val

No

>4

Unlikley

2-KY18

6A-3 (CG3093)

47

Ala→Gly

No

<4

COIII

Unlikley

2-KY15

2-MD106

7A (CG18193)

45

Thr→Ala

No

>4

Unlikely

aPolymorphic sites are listed by the line(s) in which they occur. Sequences from 15 lines were used in the analysis

bPosition numbers include mitochondrial targeting sequences

cDistance to nearest amino acid residue in an adjacent COX subunit

dThe number preceding the line name indicates the D. simulans mtDNA type

https://static-content.springer.com/image/art%3A10.1007%2Fs00239-008-9078-4/MediaObjects/239_2008_9078_Fig1_HTML.jpg
Fig. 1

Quaternary structure models of cytochrome c oxidase subunits encoded by mitochondrial DNA from Drosophila simulans. Black residues and solid arrows denote amino acids that are fixed in one mitotype (siI, -II or III); dark-gray residues and dotted arrows denote amino acids that are polymorphic within a mitotype. (a ) COI. (b) COII. Dotted octagon indicates the cytochrome c docking region. (c) COIII

The quaternary structure model of D. simulans COX predicts that COX function may be affected by two of the three fixed amino acid polymorphisms found in siI mtDNA (Table 3). In COI the polar amino acid residue Tyr468 (Fig. 1a) occurs in a transmembrane helix domain and within 4 Å of the Cys47 residue of COX8. Replacement of Tyr468 with a nonpolar phenylalanine residue in siI may affect the function of COX by altering the interactions between COI and COX8. In COII of siI, Asn114 is replaced with a physically smaller serine residue (Fig. 1b). This replacement is likely to affect COX function by altering the structure and electrostatic charge of the cytochrome c docking site. Both polymorphisms are fixed in the five siI lines surveyed in this study and it is therefore not possible to determine their effect on COX activity independently. We predict that the conservative Val451→Ile replacement in siI encoded COI is unlikely to influence COX activity because it does not change the charge of the amino acid or interact with other subunits.

Three fixed amino acid replacement polymorphisms occur in siII mtDNA encoded COII (Fig. 1b). The structural model predicts that one of these replacements, Thr130→Ile, is likely to affect COX activity. Threonine 130 is a polar amino acid that interacts both with COX4 (at polar glutamine residues 151 and 154) and with the conserved Glu127 residue in the cytochrome c docking site. Replacement of threonine with isoleucine likely weakens these interactions. We predict that the conservative replacements Thr129→Ser and Val165→Ile are unlikely to affect COX activity because these replacements preserve chemical properties and do not interact with other subunits.

We identified three mtDNA amino acid polymorphisms within the five siII harboring fly lines. In 2-KY17, the COIII mutation (Ser116→Leu; Fig. 1c, Table 3) is likely to alter structure or electrostatic charge of the cytochrome c docking site and consequently COX activity (Schmidt et al. 2005). In 2-MD106, two amino acid changes occur in the mtDNA (Val28→Ile in COI and Phe118→Tyr in COII). We predict that each amino acid change is likely to affect COX activity. The Val28→Ile replacement is conservative, but it occurs at a site of interaction between COI and COX7C. Subunit COX7C forms part of the cytochrome c docking site and slight changes in structure of the site are expected to influence enzyme activity. Phenylalanine 118 is within the cytochrome c docking site of COII and interacts with conserved COII residues Asp119 and Asn140. Replacement of the nonpolar phenylalanine with a polar tyrosine residue is likely to alter charge at the cytochrome c docking site. As these two replacements in COII are likely to interact with cytochrome c, and flies harboring the mtDNA Val28→Ile change without the Phe118→Tyr replacement have been found in Australia (Ballard 2004), we questioned whether there were compensatory changes in the cytochrome c of 2-MD106 (Schmidt et al. 2001). To investigate this question we conducted NCBI Blast searches of cytochrome c against all the D. simulans genome sequences. Consistent with expectation, in line 2-MD106 the positively charged lysine residue at position 92 of cytochrome c is replaced by an uncharged, polar asparagine residue. Reciprocal introgressions will allow functional testing of each mutation and may permit the order of mitochondrial and nuclear mutations to be determined.

Within the 13 nuclear genes we identified no amino acid differences that were fixed among mtDNA types. Of the 23 amino acid polymorphisms, 8 occur in structural domains and 15 in nonstructural domains (9 polymorphisms appear in mitochondrial targeting domains and 6 in nonstructural domains of COX6A isoform 3). Two of the eight amino acid polymorphisms that occur in structural domains and the heterozygous deletion are predicted to affect COX activity (Table 4). Each of these mutations occurs at a low frequency. In line 1-HW01 a Met28→Leu replacement in COX5B isoform 1 is likely to change close interactions with the polar Ser229 and Lys230 residues of COIII. In mammals COX5B and protein kinase A interact to inhibit COX activity in a cAMP-dependent mechanism (Bender and Kadenbach 2000). In line 1-HW07 there is an Ile28→Phe replacement in COX7A. Although this change conserves a hydrophobic interaction with Leu224 of COIII, the large phenylalanine residue is likely to affect structure of COX7A and its domain associated with cytochrome c docking. The heterozygous deletion in COX7A (ΔTrp85, ΔVal86) alters the structure of an α-helix domain that contacts subunits COI and COIII and is likely to affect assembly or stability of COX or the structure of the cytochrome c docking region (Schmidt et al. 2005). Construction of lines homozygous for presence or absence of the deletion mutation will allow us to test these alternatives.

Conclusion

Data collected here suggest that slightly deleterious mutations have accumulated in the siI and the siII mitochondrial lineages. These data are, however, not consistent with the biogeographic, life-history, and population cage data, which show that the rank order of fitness of the three mtDNA types is siII > siIII > siI (Ballard and James 2004; Ballard et al. 2007b; James and Ballard 2003). There are at least four alternate explanations for this result. First, specific mtDNA mutations may confer a selective advantage in the environment in which they occur. Consistent with this hypothesis D. simulans flies from Kenya that harbor siII mtDNA are better able to tolerate cold coma and are found at higher latitudes (Ballard 2004; Ballard et al. 2007b; Katewa and Ballard 2007). Second, one or both of the mutations that are fixed in siII harboring flies are actually advantageous and cause a change in complex IV density and inferred COX activity. This hypothesis can be tested using reciprocal introgressions and concomitantly analyzing COX activity and mtDNA density. Third, mutations outside of the region studied have a significant impact on the bioenergetic efficiency of the mitochondrial molecule. This hypothesis can be tested in the three other OXPHOS complexes that have both mitochondrial and nuclear encoded subunits. Currently, however, no other OXPHOS complexes have been completely modeled. Fourth, the presence of Wolbachia may directly or indirectly affect mitochondrial efficiency of D. simulans in nature. Four strains of maternally inherited Wolbachia are known to infect D. simulans (James and Ballard 2000), and these may protect slightly deleterious mutations from being lost and/or influence organismal fitness (Dean 2006). This explanation is least likely, at least in the African populations, because the Wolbachia infection level is low (Ballard 2003).

The prediction from population genetics is that slightly deleterious mutations are accumulating within each of the three D. simulans mtDNA types. Consistent with this expectation the siIII lineage, which has no fixed or polymorphic amino acid mutations, has the highest observed COX activity (Ballard et al. 2007b; Melvin and Ballard 2006; unpublished). The siII lineage had two fixed amino acid mutations and three polymorphic mutations that are predicted to influence COX activity. One of the fixed and all polymorphic mutations occur in amino acids that are close contact sites. Flies harboring this mitotype have intermediate observed COX activity with high variance (Ballard et al. 2007b). Flies harboring siI mtDNA have two fixed mutations that are predicted to influence COX activity and the lowest observed COX activity (Melvin and Ballard 2006; unpublished). Both of these sites occur in amino acids that are in close contact with nuclear encoded subunits.

There are at least three potential explanations for the result that there is subdivision in mitochondrial but not nuclear COX genes. First, as described above, the genome is well adapted to the environment from which the flies were collected. Second, there may not have been sufficient time for compensatory mutations to accumulate in the nuclear genome. Ballard (2007a) estimates that the three mitochondrial lineages diverged about 1.4 my ago. Finally, complex IV may tolerate more amino acid variation than we expect. This explanation is less likely because the data suggest that slightly deleterious mutations are quickly removed from the mtDNA coded subunits.

Nuclear encoded COX subunits are hypothesized to function in regulation of COX activity but little is known about the functional domains involved (Das et al. 2004; Ludwig et al. 2001). Comparative structural analysis of COX mutations may be used to determine the functional domains of the COX subunits. We employed the D. simulans COX quaternary structure model to examine the structural and functional consequences of the tenured (Mandal et al. 2005) and levy (Liu et al. 2007) mutations in D. melanogaster. The 56-base pair deletion of the tenured mutation creates a null allele of COX5A. COX5A interacts with ATP and is hypothesized to regulate COX activity by a phosphorylation-dependent mechanism (Das et al. 2004). The model predicts that COX activity will be altered in the mutant due to loss of the interaction between COX5A and the mitochondrial matrix domain of COII. The levy mutation is a G-to-A transition at the 3′ splice-junction of the single COX6A intron. The mutation results in missplicing of the COX6A transcript, leading to a shift of reading frame and translation of a truncated 54-amino acid protein (Fig. S3). The model predicts that COX activity will be reduced in the mutant because part of the cytochrome c docking site is absent from the levy mutant. Reduced activity of COX was demonstrated for the levy mutant (Liu et al. 2007).

Here we show that linking population genetics and quaternary structure modeling can lead to functional predictions of specific mtDNA amino acid mutations. Further, we illustrate how a subset of our functional predictions can then be tested experimentally. Outside of Drosophila the direct impact of mtDNA on organismal fitness has been measured in a variety of organisms including humans (Ruiz-Pesini et al.), mice (Roubertoux et al. 2003; Takeda et al. 2000), Tigriopus (Ellison and Burton 2006), and copepods (Schizas et al. 2001), but the influence of specific amino acid changes in the mitochondrial genome has been assessed using a variety of techniques with varying experimental rigor.

Acknowledgments

We thank M. Kreitman for comments on the manuscript. I. Ricafuente assisted with DNA sequencing. This work was supported by Grants NSF DEB-0444766 and NIH R01 GM067862-01.

Supplementary material

239_2008_9078_MOESM1_ESM.doc (770 kb)
MOESM 1 (DOC 770 kb)

Copyright information

© Springer Science+Business Media, LLC 2008