Museum of Zoology, University of Michigan, 1109 Geddes Avenue, Ann Arbor, Michigan 48109, USA; and 2Department of Biological Sciences, University of Denver, Denver, Colorado 80208, USA The development of the polymerase chain reaction (PCR) and dramatic improvements in technologies for DNA sequencing over the last decade have pro- vided new opportunities in avian systematics (Min- dell 1997) and the study of population genetic struc- 3 Present address: Department of Biology, Boston University, 5 Cummington Street, Boston, Massachu- setts 02215, USA. E-mail: msoren@bu.edu ture (e.g. Quinn 1992). For reasons associated with its rapid rate of evolution, haploidy, and maternal in- heritance (e.g. Moore 1995, Zhang and Hewitt 1996), recent work has focused on mitochondrial DNA (mtDNA). Direct sequencing of PCR products am- plified from extracts of genomic DNA has circum- vented the need for purified mtDNA. Almost any material is now a workable source of DNA for PCR, including blood, feathers, eggshells, feces, and other tissues from live birds; skin, feathers, cartilage, and bone from museum specimens; and even subfossil bones (e.g. Cooper et al. 1992, Cooper 1994). PCR amplification of mtDNA from total genomic DNA, however, is associated with a significant prob- lem: the potential for nuclear sequences to be ampli- fied instead of or in addition to the targeted mtDNA sequence. The transposition of mtDNA sequences to the nucleus has been documented in a wide variety of taxa from fungi to insects to vertebrates, and mounting evidence suggests that it is a common if not ubiquitous phenomenon (see Zhang and Hewitt 1996). Lopez et al. (1994) suggested the term "Numt" to refer to nuclear sequences of mitochon- drial origin, and we will use the term here. Dating to transposition events from thousands to tens of millions of years ago (e.g. Collura and Stewart 1995, Sorenson and Fleischer 1996), Numts and mi- tochondrial sequences diverge following transposi- tion, evolving at different rates and under different constraints (Arctander 1995, Zischler et al. 1995). If mistaken for mtDNA, Numts may introduce signifi- cant error in phylogenetic analyses (see below). If co- amplified with mtDNA, Numts greatly complicate the accurate determination of mitochondrial se- quences. Nonetheless, Numts present interesting op- portunities both in the study of molecular evolution and in phylogenetic analysis (see Zhang and Hewitt 1996, Quinn 1997). Several recent studies using ge- nomic DNA extracted from avian blood samples have encountered nuclear homologues of mitochon- drial genes (Quinn 1992, Arctander 1995, Sorenson and Fleischer 1996). Because red blood cells in birds (and reptiles) are nucleated and relatively depauper- ate in mtDNA, nuclear "contamination" is particu- larly problematic with blood samples. We briefly re- view published examples of this phenomenon in birds, provide a new example showing that the prob- lem is not limited to blood samples, and provide sug- gestions for avoiding, recognizing, and working with Numts. Quinn and White (1987) provided the first exam- ple of a Numt in birds, demonstrating through DNA hybridization experiments that sequences homolo- gous to mitochondrial genes occur in the nuclear ge- nome of Snow Geese (Anser caerulescens). In addition, Quinn (1992) found that blood and tissue samples from Snow Geese yielded control-region sequences falling into two divergent groups. Because blood had been collected from one population and liver tissue from another, the mistaken conclusion that these populations were differentiated in their mtDNAs might have been made. Careful evaluation, however, revealed faint shadow bands (representing the mtDNA copy) in sequences from blood and led to the design of specific primers that allowed independent amplification of either the mitochondrial or Numt se- quence. Arctander (1995) and Sorenson and Fleischer (1996) also amplified nuclear sequences from arian blood samples, but in these cases mitochondrial and nuclear copies were approximately equally repre- sented in the PCR product, resulting in sequences with ambiguities wherever the two copies differed. Interestingly, the pattern of phylogenetic relation- ships among Numts and mtDNAs differed greatly in these two studies. Arctander found that Numt cy- tochrome-b sequences from several tapaculos (Scy- talopus spp. and Myornis senilis) were more closely re- lated to each other than to any of the mtDNA se- quences of these species, suggesting a single ancient transposition event in their common ancestor. In contrast, Sorenson and Fleischer found that each of six different Numts derived from the mtDNA control region in diving ducks (Aythyini) was a close rela- tive of the ratDNA sequence of the species in which it was found, suggesting six recent, independent transposition events. The implication of these con- trasting results is that the magnitude of the error in- troduced by an unrecognized Numt in a phyloge- netic analysis is unpredictable. Recently, we found an apparent Numt homolo- gous to the mitochondrial cytochrome oxidase sub- unit I gene (COI) in the Wandering Whistling-Duck (Dendrocygna arcuata). In the process of sequencing the mitochondrial genome of this species (Mindell et al. unpubl. data), we used primers L6615 (5'- CCTCTGTAAAAAGGACTACAGCC-3' [Miranda et al. 1997]; L and H numbers designate the location of the 3' base in the light or heavy strand, respectively, of the published chicken mtDNA sequence [Desjar- dins and Morais 1990]) and H7032 (5'-TTGCCAGCT- AGTGGGGGGTA-3' [Miranda et al. 1997]) to ampli- fy a 416-bp fragment including parts of COl (387 bp) and the flanking transfer RNA (29 bp). We obtained a single unambiguous sequence from the PCR prod- uct but it did not perfectly match overlapping se- quences amplified from the same sample with other primer pairs. In addition, the inferred sequence of amino acids included substitutions at seven sites oth- erwise conserved in vertebrates (Table 1). Amplifi- cation of a larger fragment with primers L6615 and H7539 (5'-GATGTAAAGTAGGCTCGGGTGTCTAC- 3' [Miranda et al. 1997]) yielded a different unam- biguous sequence that matched overlapping se- quences for D. arcuata and that did not include the unusual amino-acid substitutions. To explore this problem further, we sequenced six other whistling-ducks and the White-backed Duck (Thalassornis leuconotus; sequences deposited in Gen- bank, accession numbers U97731 to U97739), using primers L6615 and H7032 or H7338 (5'-CCGAA- GAATCAGAAKARRTGTTG-3', degenerate sites as follows: K = G or T, R = A or G). None of these se- quences had unusual amino-acid substitutions. The sequence we initially obtained for D. arcuata is not the result of contamination from, for example, a bac- terial or invertebrate source. Phylogenetically, it is a whistling-duck COl sequence, closer to several pri- TABLE 1. Inferred amino-acid sequence for the putative D. arcuata Numt compared with that of seven Den- drocygna (all seven species identical; see Fig. 1 for species names) and Thalassornis leuconotus mtDNAs and comparison of H7032 primer sequence with D. arcuata mtDNA. Amino-acid substitutions at sites conserved across vertebrates a are marked with an asterisk. Taxon Dendrocygna spp. D. arcuata Numt Thalassornis Dendrocygna spp. D. arcuata Numt Thalassornis Dendrocygna spp. D. arcuata Numt Thalassornis H7032 Primer D. arcuata mtDNA (H-strand) COI amino-acid sequence VTFINRWLFSTNHKDIGTLYLIFGAWAGMIGTALSLLIRAELGQPGTLLG DDQIYNVIVTAHAFVMIFFMVMPIMIGGFGNWLVPLMIGAPDMAFPRMNN ........................ L ................... LLW... MSFWLLPPSFLLLLASSTVEAGAGTGWTV Primer sequence TTGCCAGCTAGTGGGGGGTA ..T..T .... AA..T..A.. Including two fish (Cyprinus carpio, GenBank accession number X61010; Crossostoma lacustre, M91245), an amphibian (Xenopus laevis, M10217), four mammals (Homo sapiens, J01415; Mus musculus, 101420; Ornitorhynchus anatinus, X83427; Didelphis virginiana, Z29573), a turtle (Chrysemys picta), Alligator mississippiensis, and several birds (Mindell et al. unpubl. data). marily Australasian taxa than to D. autumnalis, D. ar- borea, or Thalassornis (Table 2, Fig. 1). That the se- quence codes for nine amino-acid substitutions rel- ative to other Dendrocygna, strongly suggests that it is no longer evolving under the same selective con- straints and supports the conclusion that it is a Numt. Although we have not determined the flank- ing sequence of this presumed Numt, the H7032 primer matches D. arcuata mtDNA very poorly (Table 1), suggesting that this primer preferentially ampli- fied a nuclear sequence that it matched somewhat better. It is possible that this apparent Numt actually represents an intra-mitochondrial duplication of a part of the COI gene, but this has not been detected TABLE 2. Corrected percent genetic distance (Ki- mura 1980) between D. arcuata Numt and mtDNA COI sequences and other Dendrocygnini mtDNA COI sequences. D. arcuata D. arcuata mtDNA Numt D. arcuata mtDNA -- 11.6 D. javanica 5.8 13.1 D. guttata 6.9 12.8 D. bicolor 5.8 12.5 D. eytoni 8.0 13.1 D. autumnalis 11.6 14.7 D. arborea 12.3 14.7 Thalassornis leuconotus 16.1 17.0 in sequencing over half of the mitochondrial genome of this species. We find this example significant because we ini- tially obtained a single unambiguous sequence with DNA extracted from muscle tissue, presumably with a ratio of mitochondrial to nuclear genome copies of 1,000s to 1 (Robin and Wong 1988). Had our intent been to infer the phylogeny of Dendrocygna based on this single 416-bp fragment, the unusual amino-acid sequence would have been the only indication of a problem. Although quite divergent from D. arcuata mtDNA, this apparent Numt has sustained no mu- tations introducing stop codohs or reading frame shifts. In phylogenetic analysis, substituting the D. arcuata Numt for its mtDNA sequence leads to a very different conclusion about the relationships of this species (Fig. 1). Based on our experiences with Numts in birds and published observations, we make the following com- ments and practical suggestions for dealing with this phenomenon. Similar but less detailed recommen- dations were made by Zhang and Hewitt (1996) and Quinn (1997). Avoiding Numts.--Because of their relatively high ratio of nuclear to mitochondrial genome copies, avi- an blood samples are particularly likely to yield Numts and should be avoided for mtDNA sequenc- ing. Although more expensive and time consuming, preparations of purified mtDNA are the least likely to yield Numts, followed by extractions of total DNA from mtDNA-rich tissues. Neither, however, guar- antees that Numts will not be amplified (e.g. Collura and Stewart 1995). Numts may have high copy num- ber in the nucleus (e.g. Lopez et al. 1994), and PCR primers may favor a Numt over the mtDNA sequence (see Smith et al. 1992, Zhang and Hewitt 1996). In our experience, tissues that might seem relatively poor sources of DNA, such as hardened feather quills, work well for amplification of mtDNA and are less problematic than blood samples with respect to nuclear contamination. Primer design is a critical issue in comparative se- quencing studies. Because Numts may evolve more slowly than mtDNA following transposition (Arc- tander 1995, Collura and Stewart 1995, Zischler et al. 1995, Sorenson and Fleischer 1996), they diverge less from the ancestral sequence and may be more similar to the sequences of related taxa. As a result, primers based on sequences from species other than the study taxa and so-called "universal" primers may be particularly prone to amplification of Numts. In ad- dition, primers designed to accommodate a number of taxa by using a consensus rule to determine the nucleotide at each position will tend to approximate ancestral sequences and as such may favor Numts. One solution is to use primers targeted to highly conserved sequences that vary little (or not at all) among all birds, such as conserved blocks in the 12S and 16S rRNA genes and the anticodon stem and loop of some tRNA genes. For protein-coding genes, however, few 20- to 25-base stretches of highly con- served sequence exist, because most 3rd positions are free to vary. In contrast to Zhang and Hewitt (1996), we suggest that primers incorporating degen- erate sites will more consistently amplify mtDNA from tissue samples. Primers with degenerate sites that accommodate alternative nucleotides at 3rd po- sitions will be less likely to preferentially amplify one sequence over another, allowing the usually high ratio of mitochondrial to nuclear copies rather than asymmetry in primer matching to determine the PCR product. This assumes that variable sites in taxa already sequenced will predict the likely variation in additional species, an assumption that is probably reasonable for 3rd positions in a region coding a con- served sequence of amino acids, for example. See Kwok et al. (1994) for suggestions on the design of degenerate primers. Ideally, the groundwork for a sequencing project should include a careful evaluation of primer se- quences in relation to published sequences for a tax- onomic range somewhat broader than that repre- sented in the study and sequencing of flanking regions for a sample of the study taxa so that specific primers with appropriate degenerate sites can be de- signed for the study. If possible, this initial work should be done with purified mtDNA (see Dowling et al. 1996). Another potential means of avoiding Numts is to amplify the entire mtDNA or large portions of it us- ing protocols for extended or long-PCR (e.g. Cheng et al. 1994). Although large mtDNA fragments may be transposed to the nucleus (e.g. Lopez et al. 1994), most published examples involve sequences less than 4 kB in length (Zhang and Hewitt 1996), such that primers separated by 5 to 16 kB may amplify only mtDNA. Even if the entire mitochondrial ge- nome is transferred to the nucleus, a pair of primers facing away from each other that amplify the circular mtDNA will not amplify a linear Numt, unless the Numt begins and ends precisely in the region not amplified by the primers, or unless the Numt is re- peated tandemly in the nucleus (e.g. Lopez et al. 1994). Once obtained, long products can serve as templates for re-amplifications using internal prim- ers for the gene of interest. This technique has been used to eliminate apparent Numts where amplifica- tion of smaller fragments yielded sequences with nu- merous ambiguities (Sorenson unpubl. data). This approach does not always work, and we speculate that heteroduplex formation between Numts and in- complete mtDNA extension products (as in "jump- ing PCR"; Pbo et al. 1990) results in the incorpo- ration of some Numt sequence in the final PCR prod- uct. Recognizing Numts.--In our experience, Numts of- ten coamplify with the mtDNA copy, producing am- biguous sequences in affected species. A careful ex- amination of sequencing gels and electropherograms for positions with two bands or two peaks should be routine. Particularly when found at corresponding positions of complimentary strands, the temptation to simply "call" the stronger base and disregard the weaker as an artifact should be resisted. A compar- ison of sequences from tissue sources that vary in the ratio of mitochondrial to nuclear genome copies (e.g. avian blood versus muscle tissue) may provide a strong inference that a Numt is responsible for such ambiguities (e.g. Quinn 1992, Sorenson and Fleischer 1996). Other lines of evidence, such as unusual amino- acid substitutions, stop codons, and length muta- tions in protein-coding regions may lead to the con- clusion that a "clean" sequence is a Numt. Sequences of rRNA and tRNA genes should be reconciled with appropriate secondary structure models and checked for changes incompatible with that struc- ture. Mismatches in overlapping sequences from a given individual should also raise suspicions. In- deed, sequencing broadly overlapping PCR products is a good precaution against Numts because a low copy number Numt is unlikely to be preferentially amplified by two different primer pairs. Finally, sig- nificant disagreement for a particular taxon in ge- netic distances or phylogenetic relationships derived from different gene fragments may indicate that one is a Numt. Numts also may yield significantly short- er branches in phylogenetic analyses as a result of their slower rate of evolution (Sorenson and Fleischer A Short Communications /ol D. javanica !/F  D. guttata r_-' l D. bicolor  I O. eytoni I '  D. arcuata (Numt) 14/11 ] D. autumnalis 85 I D. arborea Thalassornis leuconotus [Auk, Vol. 115 B 3/10 85 1/6 1/51 D. arcuata (mtDNA) 1/21 D. javanica I D. guttata D. bicolor D. eytoni D. autumnalis D. arborea Thalassornis leuconotus c 3/7 80 1/. D. arcuata (mtDNA) D. javanica 1/5 9 D. guttata . bicolor D. eytoni D. arcuata (Numt) D. autumnalis D. arborea Thalassornis leuconotus FIG. 1. Alternative phylogenetic hypotheses for Dendrocygna using (A) the putative D. arcuata Numt, (B) the D. arcuata mtDNA sequence, and (C) both. A single most parsimonious tree was found in each analysis using equal weights for all characters in PAUP 3.1 (Swofford 1993). Branches are drawn proportional to the 1996). Note, however, that in the absence of selective constraints, Numts eventually will accumulate sub- stitutions at sites conserved in mtDNAs (Arctander 1995; Fig. 1), resulting in a greater perceived rate of change for older Numts. Such changes in addition to length mutations and recombination eventually will make older Numts unavailable to PCR primers. Of course, phenomena such as mitochondrial het- eroplasmy (Mundy et al. 1996), duplication events within the mitochondrial genome (Moritz and Brown 1986), and cross-contamination of samples or reagents may account for PCR products including more than one sequence. Numts, however, have been the most common of these phenomena. Working with Numts.--When Numts and mtDNA are co-amplified, a number of techniques may be used to separate the two copies. First, the PCR prod- uct can be cloned and several clones sequenced to de- termine the different sequences included in the PCR product (e.g. Arctander 1995). Second, primers in- ternal to the original PCR primers can be designed to discriminate between the two sequences (e.g. Quinn 1992, Sorenson and Fleischer 1996). Consid- erations of primer design are exactly opposite to those discussed above in that a primer that maxi- mally discriminates between the two sequences is needed. Ideally, the 3' base of the new primer(s) should be located where Numt and mtDNA differ by a transversion (see Kwok et al. 1990) and near ad- ditional differences. This approach will be more cost effective than cloning for studies of intraspecific variation. Third, the two copies can be separated by digesting either the genomic DNA prior to PCR or PCR product with a restriction enzyme that cuts only one of the copies between the PCR primers. Choice of an enzyme is based on ambiguous positions in ini- tial sequencing. The digested PCR product is run out on an agarose gel to separate the fragments, which can then be excised, purified, and sequenced (e.g. So- renson and Fleischer 1996). Although very effective, this technique is limited by the locations of differ- ences between the two copies. Once two clean se- quences are obtained, identifying them, respectively, as Numt and mtDNA may be based on a number of potential criteria discussed above. Alternatively, the expressed mtDNA copy can be obtained using re- verse-transcription of mRNA followed by PCR am- plification of the resulting cDNA (Collura et al. 1996). This approach is powerful in that it automat- ically identifies the functional mitochondrial copy, but it is limited to primer pairs within a single gene. The relative instability of mRNAs also may limit the samples to which this technique is applied. To the extent that Numts provide a historical record of ancestral mtDNA sequences, they also provide op- portunities for phylogenetic analyses. Numts may be ideal outgroups for extant mtDNA haplotypes (Quinn 1992, Zischler et al. 1995), their presence or absence among species can provide evidence of relationships (e.g. Sorenson and Fleischer 1996), and they provide a means to compare mechanisms and rates of molecular evolution between the mitochondrial and nuclear ge- nomes (Sorenson and Fleischer 1996, Zhang and Hewitt 1996). For these reasons, researchers may want to dis- cover Numt sequences intentionally. A more direct but technically demanding way to detect Numts is to use purified mitochondrial DNA as a probe against DNA isolated on a Southern blot. Total DNA extracted from blood or other tissues is restriction-endonuclease digested, size fractionated in an agarose gel, and transferred to a nitrocellulose or nylon membrane. In the absence of Numts, hy- bridization of a purified mitochondrial probe with the blot should reveal bands corresponding in posi- tion to those obtained from digesting purified mt- DNA alone. If Numts are present, additional bands may be detected. Even if the Numt sequence retains all of the same restriction sites as the mtDNA, flank- ing regions should generate bands of unique size. Unique bands also may correspond to restriction fragments spanning the junction between tandemly repeated Numt copies (e.g. Lopez et al. 1994). By including total DNA extracts from tissue types with different mitochondrial to nuclear DNA ratios, the nuclear origin of anomalous bands can be veri- fied by tissue-specific variation in the relative inten- sity of the putative nuclear and mitochondrial bands. Lopez et al. (1994) and Quinn and White (1987) used information from Southern blots to complement anomalous PCR-based sequencing results. In each case, the Numt was present in multiple copies ar- ranged in a tandem array. Note that such tandem re- peats result in multiple nuclear targets for PCR primers, potentially offsetting the high ratio of mi- tochondrial to nuclear genome copies in most tis- sues. This approach provides a broader survey of the mitochondrial genome than the small regions usu- ally studied with PCR techniques. One potential dif- number of steps using ACCTRAN character optimization. Support indices (the number of additional steps required in a tree without the node in question; Bremer 1988) and minimum branch length are displayed above each node. Bootstrap percentages are shown below nodes present in more than 50% of 1,000 replicates. For A: tree length = 190, CI = 0.69, RI = 0.49. For B: tree length = 172, CI = 0.70, RI = 0.50. For C: tree length = 201, CI = 0.68, RI = 0.47. A sister relationship between D. arcuata and D. javanica is consistent with analyses based on complete 12S rDNA sequences (Sorenson and Johnson unpubl. data). ficulty is the high ratio of mitochondrial to nuclear genome copies in most tissue extracts, which might lead to large differences in the signal intensity be- tween Numt versus mitochondrial targets. Avian blood is the least biased tissue in this respect and hence has the best chance of yielding Numts. Acknowledgments.--The initial sequences for D. ar- cuata were obtained as part of a mitochondrial ge- nome project directed by David Mindell. Subsequent laboratory analyses were supported by NSF Grant IBN-9412399 to Robert B. Payne. T. W. Quinn was supported by NSF Grant DEB-9629462 during prep- aration of this manuscript. We thank Mike Lubbock, Les Christidis, and Woody Martin for supplying whistling-duck samples; Derek Dimcheff for assis- tance with lab work; Carey Krajewski, David Min- dell, Robert Payne, and two anonymous reviewers for comments; and Alan Cooper for extensive dis- cussion of Numts. LITERATURE CITED ARCTANDER, P. 1995. Comparison of a mitochondri- al gene and a corresponding nuclear pseudo- gene. Proceedings of the Royal Society of Lon- don Series B 262:13-19. BREMER, K. 1988. The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42:795-803. CHENG, S., R. HIGUCHI, AND M. STONEKING. 1994. Complete mitochondrial genome amplification. Nature Genetics 7:350-351. COLLURA, R. V., M. R. AUERBACH, AND C.-B. STEW- ART. 1996. A quick, direct method that can dif- ferentiate expressed mitochondrial genes from their nuclear pseudogenes. Current Biology 6: 1337-1339. COLLURA, R. V., AND C.-g. STEWART. 1995. Insertions and duplications of mtDNA in the nuclear ge- nomes of Old World monkeys and hominids. Na- ture 378:485-489. COOPER, A. 1994. DNA from museum specimens. Pages 149-165 in Ancient DNA: Recovery and analysis of genetic material from paleontologi- cal, archaeological, museum, medical, and fo- rensic specimens (B. Herrmann and S. Herrm- ann, Eds.). Springer-Verlag, New York. COOPER, A., G. K. CHAMBERS, C. MOURER-CHAUVIRE, g. VON HAESELAER, g. C. WILSON, AND S. P..Bo. 1992. Independent origins of New Zea- land moas and kiwis. Proceedings of the Nation- al Academy of Sciences USA 89:8741-8744. DESJARDINS, P., AND g. MORAIS. 1990. Sequence and gene organization of the chicken mitochondrial genome. A novel gene order in higher verte- brates. Journal of Molecular Biology 212:599- 634. DOWLING, t. E., C. MORITZ, J. D. PALMER, AND L. H. RISENBERG. 1996. Analysis of fragments and re- striction sites. Pages 249-320 in Molecular sys- tematics, 2nd ed. (D. M. Hillis, C. Moritz, and B. K. Mable, Eds.). Sinauer Associates, Sunderland, Massachusetts. KIMURA, M. 1980. A simple method for estimating evolutionary rate of base substitutions through comparative studies of nucleotide sequences. Journal of Molecular Evolution 17:111-120. KWOK, S., S.-Y. CHANG, J. J. SNINSKY, AND g. WANG. 1994. A guide to the design and use of mis- matched and degenerate primers. PCR Methods and Applications 3:S39-S47. KWOK, S., D. E. KELLOGG, N. MCKINNEY, D. SPASiC, L. GODA, C. LIVENSON, AND J. J. SNINSKY. 1990. Effects of primer-template mismatches on the polymerase chain reaction: Human immunode- ficiency virus type 1 model studies. Nucleic Ac- ids Research 18:999-1005. LOPEZ, J. V., N. YUHKI, R. MASUDA, W. MODI, AND S. J. O'BRIEN. 1994. Numt, a recent transfer and tandem amplification of mitochondrial DNA to the nuclear genome of the domestic cat. Journal of Molecular Evolution 39:174-190. MINDELL, D. P. (Ed.) 1997. Avian molecular evolu- tion and systematics. Academic Press, New York. MIRANDA, H. C., R. S. KENNEDY, AND D. P. MINDELL. 1997. Phylogenetic placement of Mimizuku gur- neyi (Aves: Strigidae) inferred from mitochon- drial DNA. Auk 115:315-323. MOORE, W. S. 1995. Inferring phylogenies from mtDNA variation: Mitochondrial-gene trees ver- sus nuclear~gene trees. Evolution 49:718-726. MORITZ, C., AND W. M. BROWN. 1986. Tandem du- plication of D-loop and ribosomal RNA sequenc- es in lizard mitochondrial DNA. Science 233: 1425-1427. MUNDY, N. I., C. S. WINCHELL, AND D. S. WOODRUFF. 1996. Tandem repeats and heteroplasmy in the mitochondrial DNA control region of the Log- gerhead Shrike (Lanius ludovicianus). Journal of Heredity 87:21-26. P..BO, S., D. M. IRWIN, AND A. C. WILSON. 1990. DNA damage promotes jumping between tem- plates during enzymatic amplification. Journal of Biological Chemistry 265:4718-4721. QUINN, t. W. 1992. The genetic legacy of Mother Goose--phylogeographic patterns of Lesser Snow Goose Chen caerulescens maternal lineages. Molecular Ecology 1:105-117. QUINN, T. W. 1997. Molecular evolution of the mi- tochondrial genome. Pages 3-28 in Avian molec- ular evolution and systematics (D. P. Mindell, Ed.). Academic Press, New York. QUINN, t. W., AND B. N. WHITE. 1987. Analysis of DNA sequence variation. Pages 163-198 in Avian genetics: A population and ecological approach (E Cooke and P. A. Buckley, Eds.). Academic Press, London. ROBIN, E.D., AND R. WONG. 1988. Mitochondrial DNA molecules and virtual number of mito- chondria per cell in mammalian cells. Joumal of Cellular Physiology 136:507-513. SMITH, M. E, W. K. THOMAS, AND J. L. PATTON. 1992. Mitohcondrial DNA-like sequences in the nucle- ar genome of an akodontine rodent. Molecular Biology and Evolution 9:204-215. SORENSON, g. D., AND R. C. FLEISCHER. 1996. Mul- tiple independent transpositions of mitochon- drial DNA control region sequences to the nu- cleus. Proceedings of the National Academy of Sciences USA 93:15239-15243. SWOFFORD, D. L. 1993. PAUP: Phylogenetic analysis using parsiomony, version 3.1. Illinois Natural History Survey, Champaign. ZHANG, D.-X., AND G. M. HEWITT. 1996. Nuclear in- tegrations: Challenges for mitochondrial DNA markers. Trends in Ecology and Evolution 11: 247-251. ZISCHLER, H., H. GEISERT, g. VON HAESELER, AND S. PAABO. 1995. A nuclear "fossil" of the mito- chondrial D-loop and the origin of modern hu- mans. Nature 378:489-492. Received 9 December 1996, accepted 4 June 1997. Associate Editor: R. M. Zink