A Gene Specifying Subunit VI11 of Human Cytochrome c Oxidase Is Localized to Chromosome 11 and Is Expressed in Both Muscle and Non-muscle Tissues*

Subunit VI11 of mammalian cytochrome c oxidase (COX; EC 1.9.3.1) exists in at least two isoforms, be- cause different but related polypeptides have been identified in COX isolated from liver and heart of both beef and pig. We have isolated a full length cDNA specifying subunit VI11 of human COX from a human liver cDNA library. Sequences hybridizing to this cDNA are present at only one site, the COX8 locus, on human chromosome llq12-q13. The deduced human polypeptide is 58% identical with COX VI11 isolated from beef liver, but only 38% identical with COX VI11 isolated from beef heart. Transcriptional analysis shows that an mRNA identical with the isolated cDNA is present in abundant amounts not only in human and monkey liver tissue, but in heart and skeletal muscle as well, tissues not known previously to contain this isoform. Since the only COX VI11 subunit found in human heart agrees 100% with the polypeptide de- duced from this COXVIII cDNA, it may well be that, in distinction to other mammals, only one form of COX VI11 exists in primates. independent

Subunit VI11 of mammalian cytochrome c oxidase (COX; EC 1.9.3.1) exists in at least two isoforms, because different but related polypeptides have been identified in COX isolated from liver and heart of both beef and pig. We have isolated a full length cDNA specifying subunit VI11 of human COX from a human liver cDNA library. Sequences hybridizing to this cDNA are present at only one site, the COX8 locus, on human chromosome llq12-q13. The deduced human polypeptide is 58% identical with COX VI11 isolated from beef liver, but only 38% identical with COX VI11 isolated from beef heart. Transcriptional analysis shows that an mRNA identical with the isolated cDNA is present in abundant amounts not only in human and monkey liver tissue, but in heart and skeletal muscle as well, tissues not known previously to contain this isoform. Since the only COX VI11 subunit found in human heart agrees 100% with the polypeptide deduced from this COXVIII cDNA, it may well be that, in distinction to other mammals, only one form of COX VI11 exists in primates.
Cytochrome c oxidase (COX)' is the terminal enzyme of the respiratory chain, coupling the transfer of electrons from cytochrome c to molecular oxygen with the concomitant production of a proton electrochemical gradient across the inner mitochondrial membrane (1, 2). In addition to three mitochondrially encoded subunits (3), which correspond to the three subunits of the prokaryotic enzyme (4) and which perform the catalytic function, the eukaryotic enzyme contains a number of additional nuclear-encoded smaller subunits, ranging from four in Dictyostelium discoideum ( 5 ) to ten in mammals (6-9). The function of the nuclear-encoded subunits is still unknown, although it has been proposed that they might be involved in the modulation of the catalytic function * This work was supported by National Institutes of Health Research Grants GM26105 and NS11766 and by grants from the Muscular Dystrophy Association and the Aaron Diamond Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted

504823.
to the GenBankTM/EMBL Data Bank with accession number(s) Fellow of the Muscular Dystrophy Association. The abbreviations used are: COX, cytochrome c oxidase; cox, gene coding for COX; COX, genetic locus of cox; bp, base pair(s); kb, kilobase pair(s).
(6-8, 10, 11). This possibility is also in agreement with the results of Bisson and Schiavo (12), who have shown that alternative forms of COX subunits are expressed in different phases of the cell cycle of D. discoideum.
In higher organisms, COX exhibits a tissue-specific plasticity, as indicated by the different kinetic and biochemical properties of the enzyme isolated from different tissues (13,14) and by the presence of human COX deficiency diseases in which the biochemical defect and clinical involvement are limited to one or a few tissues (15-17). At the molecular level, the tissue specificity of COX appears to reside in the nuclearencoded subunits (6-8) and not in the mitochondrial-encoded subunits, as there is no evidence of tissue-specific differences in the mitochondrial genome.
There is strong evidence that subunits VIa, VIIa, and VI11 (nomenclature of Kadenbach et al. (18)) have liver-and heartspecific isoforms in cows, pigs, and rats (6-8). In particular, COX VI11 isolated from mammalian heart differs from the same subunit isolated from liver in its electrophoretic migration, thiol content, and immunological properties (19-22). The amino acid sequences of COX VI11 isolated from beef heart and liver show the presence of two different but related polypeptides which are 60% identical in the sequenced region (23, 24). Amino acid sequences of the N-terminal regions of COX VI11 isolated from pig heart and liver show a similar result (18). As part of an ongoing interest in the molecular basis of the COX deficiency disorders and in the tissuespecific nature of COX activity (25-27), we have isolated a full length cDNA specifying subunit VI11 of human COX and have determined the chromosomal localization of the COXVIII gene.

MATERIALS AND METHODS
Enzymes and Reagents-Restriction enzymes were from Boehringer Mannheim, New England Biolabs, Bethesda Research Laboratories, and IBI; the oligonucleotide "random-primed" labeling kit was from Boehringer Mannheim; the Sequenase sequencing kit was from United States Biochemicals; sequencing primer was from New England Biolabs; hybridization reagents were from Sigma; [cP~'P]~NTPs (800 Ci/mmol) were from Du Pont-New England Nuclear.
Isolation and Characterization of cDNA Clones-Using a nicktranslated (28) cDNA-encoding subunit VI11 of rat COX isolated from a rat hepatoma cDNA library (29) as a probe, we screened 60,000 recombinants from a X g t l l human liver cDNA library (a kind gift of Dr. G. Ricca). Hybridization was at 42 "C, and posthybridization washing was a t 50 'C, as described (25). We obtained one positive clone (XhCOX8.31) with an insert of about 500 bp. The phage DNA was isolated by the plate-lysate method (30); the insert was excised from the phage vector by digestion with EcoRI and was subcloned into appropriately digested M13 (31) vectors prior to DNA sequencing (32) with the Sequenase kit (United States Biochemicals). The se- quences were visualized by autoradiography of thin, denaturing 5% acrylamide gels (33); the entire cDNA was sequenced on both strands ( Fig. 1). Using the random-prime labeled (34) EcoRI insert of XhCOX8.31 as a probe at low stringency, we isolated one coxVZZZ clone from B human Xgtll fetal muscle cDNA library (XHCOX8.21) and one clone from a human X g t l l adult heart cDNA library (Clontech; XHCOX8.51); these isolated clones were isolated and analyzed in the same manner as was the clone obtained from the liver cDNA library.
RNA Analyses-Total RNA was prepared from human and monkey (Mucaca jascicularis) tissues by the guanidinium isothiocyanate/cesium chloride method (35). Northern blot and S1 nuclease protection experiments were performed as described (24). For S I analysis, the EcoRI insert of XhCOX8.21 was subcloned into the EcoRI site of Ml3mpll and homogeneously labeled single-stranded probes (36) were used in both sense (control) and anti-sense (test) orientations, as described (25).
Chromosomal Assignment-The origin and the characterization of the somatic cell hybrids between Chinese hamster V79/380-6 cells and human diploid leukocytes, lymphoblastoid cells, or fibroblasts have been summarized elsewhere (37). The hybrids used for this study were derived from seven independent fusion experiments. Their chromosomal content was determined at the time of DNA extraction by cytogenetic analysis as well as by gene marker studies. The various hybrids used for regional mapping of COX8 on chromosome 11 have been described (38). Restriction enzyme digestion, electrophoresis, Southern transfer, and blot hybridization were carried out by standard methodology with modifications as described (38). Probe pCOX8.2 contains the EcoRI insert of XhCOX8.21 inserted into the EcoRI si e of Bluescribe M13+ (Stratagene). A 280-bp BamHI-EcoRI(1inker) subfragment from the 3' end of the insert of pCOX8.21 (see Fig. l), labeled by random oligomer priming (34), gave singlecopy signals with little background and was used for all mapping experiments. For the in situ hybridization to human metaphase chromosomes, probe pCOX8.21 was labeled by nick translation (28) using [3H]dATP, [3H]dCTP, and [3H]dTTP to a specific activity of 2 X lo7 cpm/pg. Hybridization, washes, and staining were carried out as described (39) with modifications (40).

RESULTS
Isolation of cDNA Clones Encoding Subunit VIII of Human COX-Using a partial length cDNA clone specifying rat COX subunit VI11 (29), we screened a human adult liver cDNA library at low stringency and isolated one hybridizing clone, designated XhCOX8.31. The EcoRI insert of the clone contains ( Fig. 2  ation signal (42) 14 bp upstream from a 7-bp-long poly(A) tail. The deduced polypeptide appears to represent the precursor to subunit VI11 of human COX, as the 44 C-terminal amino acids are 58% and 63% identical with the liver isoform of subunit VIII of beef (24) and rat (29) COX, respectively (Fig. 2B). However, the polypeptide is only 38% identical with the heart isoform of subunit VI11 of beef COX (23). The nucleotide sequence is 79% identical with that of rat liver COZVIII (29), with a remarkably high conservation of nucleotide identity in the 3"untranslated region (81%). Based on these sequence identities, the deduced polypeptide of XhCOX8.31 seems to represent the liver isoform of subunit VI11 of human COX. The human COX VI11 sequence was so diverged from that of the analogous yeast subunit VIIa (43, 44) that we could not align the two sequences with certainty; the identity with yeast VIIa is only about 20%. In spite of the large divergence of the amino acid sequence among human "liver-type" VIII, beef heart VIII, and yeast VIIa, all three subunits exhibit strikingly similar secondary structure, as inferred from their hydropathy profiles (Fig. 3).

(-25)
The 25 N-terminal amino acids, which are absent in the published protein sequences of beef and pig subunit VI11 (18,23,24), are rich in basic and hydroxylated amino acids and are devoid of acidic residues; they appear, therefore, to represent the mitochondrial presequence which targets the COX VI11 precursor polypeptide to mitochondria, and which is cleaved after import (for a review, see Ref. 45). We note, however, that yeast VIIa lacks a presequence (43). On the other hand, yeast VIIa contains four amino acids at its C terminus that do not appear in the mature polypeptide (43), and the mature protein is also 10 amino acids longer at its C terminus than human VIII. Whether a mitochondrial targeting signal is contained in this C-terminal extension or is embodied in the rest of the mature polypeptide is currently unknown. However, another COX subunit, rat VIc (29,46), also does not contain a cleavable presequence. Presumably in this case the targeting signal is part of the mature protein.
In order to isolate the putative human heart isoform of COX VIII, we used the EcoRI insert of XhCOX8.31 as a probe at low stringency and screened about 60,000 plaque-forming units each of human cDNA libraries derived from fetal muscle and adult heart tissue. Two hybridizing clones were isolated a nearly full length clone (XhCOX8.21; 447-bp cDNA insert) from the fetal muscle library and a partial length clone from the adult heart library (XhCOX8.51; 318-bp insert). The nucleotide sequences of the two clones, isolated from libraries extracted from tissues not known to contain the liver isoform of subunit VIII, were the same as that of the liver clone.
Gene Mapping of Subunit VIII of Human Cytochrome c Oxidase-In genomic Southern hybridization analysis using DNA digested with BglII, we detected a single hybridizing human band approximately 21 kb in size and two Chinese hamster bands of 10.5 and 6.8 kb (Fig. 4A); in a panel of Chinese hamster X human somatic cell hybrids, only hybrids containing human chromosome 11 showed the 21-kb human BglII fragment (Fig. 4A, lanes 2 and 4 ) . After digestion with PstI, a single human fragment was detected, while digestion with HindIII generated two fragments; in a panel of 12 Chinese hamster x human somatic cell hybrids whose DNA was digested with PstI or HindIII, the human-specific coxVIII restriction fragment was concordant with human chromosome 11 (not shown). All other chromosomes were excluded by two or more discordant hybrids (Table I).
In a regional mapping panel, the human coxVIII signal was present in hybrids that contained region llpll.2->qter and was absent in three independent hybrids containing the short arm of chromosome 11. The result assigns the coxVIII gene to the long arm of chromosome 11 (Fig. 4B). We designate this locus COX8. In order to confirm this regional assignment and more precisely define the COX8 localization, we performed in situ hybridization to human metaphase chromosomes. Initially 66 randomly chosen cells were analyzed for localization of grains on chromosome 11. Among the total 233 grains (3.5 grains/cell), 15 grains (6%) were located at the specific site 11q12-ql3. Of the 66 cells, 14 (21%) had label at this site. In addition, 38 cells were selected for the presence of a grain anywhere on chromosome 11. In these cells, 39 grains were found on chromosome 11, of which 18 (46%) were located at the specific site 11q12-ql3 (Fig. 4C). These results demonstrate localization of COX8 on the proximal long arm of chromosome 11 and suggest the existence of only a single gene for COX VI11 in the human genome.
The localization at 11q12-13 places COX8 in the vicinity of loci for C1 complement inhibitor, which, when defective, is responsible for hereditary angioedema; the human homolog of the INT2 integration site for the murine mammary tumor virus oncogene, which is a member of the fibroblast growth factor family; the BCLl gene that is rearranged in translocations in chronic B lymphocytic leukemia; the human homolog of the SEA avian erythroblastosis oncogene; and a cluster of pepsinogen genes (47).
Transcriptional Analysis of coxVIII-We performed Northern analysis of COXVIII transcription in several monkey and human tissues, including heart and skeletal muscle, tissues which, in other mammals, are devoid of the liver isoform polypeptide. A single hybridizing band was present in all tissues, with a size (about 500 nucleotides) comparable to that of the insert of XhCOX8.31 (Fig. 5, bottom). This result indicates that this coxVIII cDNA is essentially full length. Although equal amounts of RNA were loaded, the intensity of the hybridizing signal varied among tissues, with the strongest signal present in heart and skeletal muscle. An essentially identical pattern was obtained when the cDNA encoding a non-tissue-specific human COX subunit, COX Va (261, was used as a probe (Fig. 5, top). Thus, the difference in hybridizing signal among tissues seems to correlate with the steady state level of transcription of COX subunit genes in  lanes 1-4), hybridized with coxVIII probe. The cell hybrids in lanes 1 and 3 are negative, and those in lanes 2 and 4 are positive for the human fragment. R, regional assignment of COX8 to the long arm of chromosome 11. Rodent X human somatic cell hybrids with defined regions of human chromosome 11 (represented by vertical bars A-D) were analyzed for hybridization with coxVIII probe on Southern blots containing DNA digested with Hind111 or PstI. Only hybrids with region C were positive; hybrids with regions A, R, and D were negative. The localization of COX8 based on these experiments is shown by the bracket on the right. C, localization of silver grains with respect to standard G-bands on chromosome 11 after in situ hybridization of tritium-labeled COXVIII probe to normal human metaphase spreads. Dots indicate the sites of label observed in 38 cells selected for the presence of grains on chromosome 11. Crosses indicate localization of grains in 66 unselected cells. Based on these results, we assigned COX8 to either band llq12 or the proximal part of band llq13.

TABLE I Correlation of human COX8 sequences with human chromosomes in rodent X human somatic cell hybrids
The numbers of hybrids that are concordant (+/+ or -I-) and discordant (+/-or -/+) with the human COX8 sequence are given for each chromosome. Hybrids in which a particular chromosome was structurally rearranged or present in fewer than 10% of cells have been excluded. The human coxVZZZ gene appears to be highly diverged COX VIII, which migrates a t about 500 nucleotides (8, bottom).
from the analogous gene in other mammals, as our cDNA did not hybridize to any mRNA species in either bovine heart or liver, even a t conditions of very low stringency (data not shown).
Since the level of sensitivity of Northern analysis does not exclude the possibility that the cDNA encoding the liver isoform cross-hybridized to an mRNA encoding a putative COX VI11 heart isoform, we performed S1 nuclease protection of mRNA isolated from four different human tissues, using the coxVZZZ cDNA from XHCOX8.21 as a probe. Only an mRNA species of about 450 nucleotides (i.e. of the same size as the entire probe) was protected in RNA isolated from human liver, brain, heart, and skeletal muscle (Fig. 6). Considering the marked difference between the two beef isoforms at the protein level, the mRNA protected in human heart and muscle (which, according to the limit of resolution of this technique, may contain, at most, single-base differences with the probe cDNA) is almost certainly not a related heterologous message encoding the putative human "heart" isoform, but rather must represent the homologous, "liver" isoform. In other words, the mRNA encoding the COX VI11 polypeptide represented by our cDNA is expressed in human liver, brain, heart and muscle and, based on the Northern analysis described above, is most likely expressed in all tissues of the body.

DISCUSSION
The existence of tissue-specific (heart and liver) isoform polypeptides of subunit VI11 of cytochrome c oxidase has been documented in three mammals, beef, pig, and rat. Thus, it was surprising to find that a gene specifying a human livertype COX VI11 polypeptide was actively transcribed in all examined tissues in primates, including humans. Moreover, this transcript is almost certainly translated and assembled into the COX holoenzyme in human heart, because the only form of subunit VI11 that has been isolated from human heart has an amino acid sequence that agrees 100% with the polypeptide deduced from our coxVZZZ cDNA (49). Thus, all the data presently available at the chromosomal, transcriptional, and translational level indicate that only one gene specifying subunit VI11 of COX is active in primates and that, in distinction to what has been observed in other mammals, human COX does not appear to express alternative isoforms of this subunit in different tissues.
At least two different mechanisms could account for such a difference among mammals. Based on comparison of the nucleotide sequence of a partial length cDNA encoding the heart-specific isoform of bovine COX VII12 with that of our cDNA encoding the human liver-type COX VIII, it is likely that the mammalian heart-and liver-specific isoform genes specifying this subunit arose via a gene duplication event, as both cDNAs share regions of significant sequence identity in their respective 3"untranslated regions (not shown). Thus, the expression of only a single COX VI11 polypeptide in humans and monkeys, but of two different but related isoform polypeptides in other mammals (cows, pigs, and rats), suggests either that the putative gene duplication event that gave rise to the two coxVZZZ isoform genes occurred after the mammalian radiation (or more specifically, after the divergence of primates from other mammals), or that the duplication event occurred prior to the mammalian radiation, but that in primates the heart isoform gene may have been silenced in a manner similar to the silencing of the &globin gene in Old World monkeys (50). The high degree of amino acid divergence between beef heart and liver COX VI11 (58%) relative to the high degree of nucleotide and amino acid sequence R. Capaldi, personal communication.
identity between human and bovine COX IV, Va, and Vb (82%, 95%, and 85%, respectively (26)) favors early rather than late divergence. Moreover, gene duplication prior to the mammalian radiation is supported by the presence of liverand heart-specific isoforms of subunit VI11 in chickens (51).
In that case, there may indeed be a second heart-type isoform gene in the human genome. The fact that coxVZZZ sequences hybridized to a single site on chromosome 11 does not preclude this possibility, as either the locus may contain a multigene family, as is the case with the 0-globins, or the two isoform genes may be so highly diverged that they do not crosshybridize, in which case a heart isoform gene located elsewhere would not be detected in our chromosome mapping experiments using the liver-type cDNA.
Our data, however, do not exclude the possibility that, besides the liver-type COX VI11 polypeptide, a second as yet unidentified COX VI11 isoform is present in human heart. At the transcriptional level, significant sequence divergence between liver and heart isoform genes could prevent detection of a heart isoform mRNA in both the Northern and S1 protection experiments; at the protein level, it is possible (although unlikely) that another isoform besides the liver isoform is present in human heart, but was not detected using the specific column chromatography and high performance liquid chromatography procedures employed to isolate human heart COX VI11 (49).
If a heart-type COX VI11 isoform exists and is preferentially assembled into the COX holoenzyme in muscle tissues, as observed in other mammals, the presence of a liver-type mRNA in primate muscle tissues could reflect a low level of "constitutive" expression of this isoform in heart and muscle. This is presumably the case for a truly tissue-specific COX subunit, rat COX VIa. Schlerf et al. (52) isolated cDNAs encoding the heart-and liver-specific isoforms of rat COX VIa; in Northern analyses with these genes, the liver-specific rat coxVZu probe hybridized to mRNA derived from liver, kidney, heart, and skeletal muscle. On the other hand, the muscle-specific coxVZa probe hybridized only to mRNA isolated from heart and muscle tissue. Thus, transcription of the liver-type gene was apparently not tissue-specific, while the heart-type isoform gene showed a tissue-specific transcription pattern. A problem with this interpretation of our results is that in the case of rat coxVZa, the "inappropriate" isoform (i.e. liver mRNA in heart or muscle) was usually expressed at a lower level than was the "appropriate" isoform (i.e. liver mRNA in liver); in the case of human coxVZZZ, the intensity of the hybridization signal in heart relative to both the signal in other tissues and to the coxVa signal indicated that the liver-type coxVZZZ mRNA is a major transcription product in heart. This result, along with the protein analysis data (49), seems to exclude the possibility that transcription of livertype COX VI11 reflects merely a constitutive low level expression of the liver-type polypeptide in human muscle tissues. However, the possibility also exists that events following transcription ( e g . at the level of translation, transport to the mitochondria, or assembly in the enzyme complex) could reduce substantially the appearance of the liver-type polypeptide in the holoenzyme in heart and muscle.
In order to further investigate the existence of tissuespecific isoform polypeptides of subunit VI11 of human COX, we are using both antibodies to bovine heart COX VI11 and a cDNA clone encoding this bovine subunit (kind gifts of R. Capaldi, University of Oregon) to search for the corresponding heart-type isoform polypeptide and cDNA in humans.