Evolution of the Creatine Kinases THE CHICKEN ACIDIC TYPE MITOCHONDRIAL CREATINE KINASE GENE AS THE FIRST NONMAMMALIAN GENE*

In both mammals and birds, the creatine kinase (CK) family consists of four types of genes: cytosolic brain type (B-CK); cytosolic muscle type (M-CK); mitochon- drial ubiquitous, acidic type (Mi a -CK); and mitochondrial sarcomeric, basic type (Mi b -CK). We report here the cloning of the chicken Mi a -CK cDNA and its gene. Amino acid sequences of the mature chicken Mi-CK pro- teins show about 90% identity to the homologous mammalian isoforms. The leader peptides, however, which are isoenzyme-specifically conserved among the mammalian Mi-CKs, are quite different in the chicken with amino acid identity values compared with the mammalian leader peptides of 38.5–51.3%. The chicken Mi a -CK gene spans about 7.6 kilobases and contains 9 exons. The region around exon 1 shows a peculiar base composition, with more than 80% GC, and has the characteristics of a CpG island. The upstream sequences lack TATA or CCAAT boxes and display further properties of housekeeping genes. Several tran- scription factor binding sites known from mammalian Mi-CK genes are absent from the chicken gene. Although the promoter structure suggests a ubiquitous range of expression, analysis of Mi a -CK transcripts in chicken tissues shows a restricted pattern and therefore does not fulfill all criteria of a housekeeping enzyme.

A sufficient capacity and balanced regulation of "high energy phosphate" supply and turnover is essential for the proper function of any cell. Large amounts of energy-rich phosphagens can be found in many cells or tissues throughout the animal kingdom. In all vertebrates and also in some invertebrates this phosphagen is phosphorylcreatine (PCr) 1 (1). PCr and ADP are the products of the reversible transfer of ␥-phosphate groups from ATP to creatine, catalyzed by the creatine kinases (CKs). Two fundamental types of CKs can be found in vertebrates: cytosolic and mitochondrial CKs. The subcellular localization, the biochemical and kinetic data, and the loss of flagellar motility in spermatozoa upon inhibition of the CK system and other data (for review, see Ref. 2) led to the suggestion of a metabolic PCr circuit with PCr as a transport and storage form of high energy phosphate. The PCr circuit connects sites of high energy phosphate production (glycolysis and oxidative phosphorylation) with those of high energy phosphate consumption. At the producing end of the circuit, CK is thought to have privileged access to ATP generated either by glycoclysis or by oxidative phosphorylation in the mitochondrial matrix and uses this ATP to generate PCr. At the receiving end, CK is functionally coupled to various ATPases (for instance myosin ATPase of myofibrils), which use the ATP generated in the reverse CK reaction.
In chicken there are five different CK subunits known so far; three are found in the cytosol, two in mitochondria. The cytosolic subunits are called M-CK (muscle) and B a -CK and B b -CK (brain, more acidic and more basic, respectively) and can dimerize with each other (3)(4)(5). They are found soluble in the cytosol, but fractions are also associated, for instance, with the M-line of the sarcomeres (6), the sarcoplasmatic Ca 2ϩ -ATPase (7), or the spermatozoan tail. 2 Whereas in mammals there is just one isoform of B-CK, the two B-CK isoforms of the chicken are derived from a single gene by alternative splicing of the second exon (9,10). Additional heterogeneity of B-CK was shown to be due to alternative initiation of translation (11) or posttranslational phosphorylation (12)(13)(14). In tissues of adult chicken, M-CK is predominantly found in skeletal muscle, but contrary to mammals, it is absent from heart (3). B-CK is expressed in almost all tissues and found enriched in various regions of the brain, retina, heart, gizzard, gut, and sperm (2).
Evidence for two different isoforms of chicken mitochondrial CK (Mi-CK) has been found by comparison of translated cDNA sequences, isolated from a leg muscle cDNA library, to partial amino-terminal protein sequences of Mi-CK purified from brain (15). These two isoforms were termed Mi a -CK (a ϭ more acidic pI, in mammals it was called ubiquitous Mi-CK by other authors) and Mi b -CK (b ϭ more basic pI, sarcomeric Mi-CK in mammals) and are found in vivo exclusively as homodimers and homooctamers (16,17). The Mi-CKs are synthesized in the cytosol as precursor proteins with distinct leader peptides (15,18,19) and get imported into the intermembrane space of mitochondria. Mi b -CK is normally coexpressed with M-CK and is present in skeletal muscle and heart (15), but it also has been found in sperm. 2 Mi a -CK is distributed more ubiquitously and has been detected, like B-CK, in chicken brain, gut, and retina. Hence the expression of the mitochondrial CKs matches to a certain degree the pattern found for the cytosolic CKs, indicating possible common regulatory mechanisms.
Recently a comparison of the 26 known CK protein sequences suggested a highly conserved protein family and allowed the construction of an evolutionary tree (20). This tree predicts three gene duplications at the origin of the four CK isoforms, which is in agreement with observations on the gene structures of the four published human CK genes (21)(22)(23)(24)(25). In chicken, only the B-CK gene has been completely analyzed so far, and fragments of the M-CK gene indicate an extraordinarily big size (10).
To gain further insight into the evolution of the CK isoenzyme family, the chicken Mi a -CK protein sequence and the Mi a -CK gene structure were elucidated as the first nonmammalian Mi-CK gene. The chicken Mi a -CK leader peptides were analyzed in order to determine if the functional chicken peptides are isoprotein-specifically conserved. Analysis of the promoter elements of the Mi a -CK gene was carried out to investigate if the expression pattern of Mi a -CK is coupled to B-CK expression, indicating common regulatory mechanisms of these two genes. In order to test if such mechanisms are active in the chicken, the expression patterns in several adult tissues were studied. The structure of the chicken Mi a -CK gene, including the localization of intron-exon boundaries was analyzed, and the structure was compared with the mammalian ubiquitous Mi-CK genes, allowing conclusions on the evolutionary relationship of these genes.

Isolation and Cloning of Mi a -CK cDNA-and Genomic Sequences, RT-PCR
A first fragment of chicken Mi a -CK was generated with the following degenerate oligonucleotides synthesized on a Gene Assembler (Pharmacia LKB Biotech Inc., Uppsala, Sweden): 5Ј-GTCGACGTCGACG-GYGAACGYCAGMGVCGVAGGTACCC-3Ј (derived from the aminoterminal sequence GERQRRRYP with two SalI sites at the 5Ј-end; highly Mi a -CK specific) and 5Ј-GTYTTRTCRTTRTTRTGCCA-3Ј (complementary to amino acids 212-218, WHNNDKT; CK specific). Degeneracy of these oligonucleotides was reduced by analysis of the codon usage of the three available chicken CK sequences (9, 15, 26 -28). Total RNA from chicken brain was reverse transcribed with random hexamer primers. PCR on the obtained cDNA was carried out for 60 cycles (94°C for 30 s, 50°C for 30 s, 72°C for 2 min) as described earlier (29). The fragment was gel-purified, reamplified, subcloned into pTZ19U (U. S. Biochemical Corp.) and fully sequenced.
The whole amplified fragment, labeled with 32 P, was used to screen a chicken brain gt10 cDNA library (Stratagene). A total of 2.5 ϫ 10 5 plaques was screened at a density of 25.000 plaques/plate (100 ϫ 100 mm). Lift-offs were performed with Biodyne type A nylon membranes (1.2 m, Pall Inc., East Hills, NY) according to the manufacturer's conditions and were hybridized as described previously (30) without formamide at 68°C overnight. Washing was done at room temperature with three changes of 2 ϫ SSC, 0.1% SDS, 15 min each.
A genomic EMBL3 library with high molecular weight DNA from chicken liver (Clonetech, Palo Alto, CA), partially digested with MboI, was obtained from Dr. B. Trueb (Maurice E. Mü ller Institute, Berne, Switzerland). A total of 3.6 ϫ 10 6 plaques was screened at a density of 10 5 plaques/plate. Plaque lift-offs were done as described above. Hybridization with the 32 P-labeled fragment of clone UB15-9 (EcoRI/SmaI fragment, nucleotides 170 -487 (see Fig. 1A)) was done as above, but in the presence of formamide at 42°C. Lift-offs were washed with 2 ϫ SSC, 0.1% SDS at 65°C.
Mi b -CK 5Ј-RACE-Heat-denatured poly(A) ϩ -selected RNA from chicken leg muscle was reverse-transcribed using Moloney murine leukemia virus reverse transcriptase (Life Technologies AG, Basel, Switzerland) with the RT-primer. Cycling conditions were as follows: PCR 1 was done as in the Mi a -CK-3Ј-RACE; PCR 2 was done as in the Mi a -CK-5Ј-RACE.
Obtained products were verified with an internal, 32 P-labeled oligonucleotide by Southern blotting, subcloning, and sequencing. Additional Mi a -CK 5Ј-RACE-RT-PCRs were performed with a set of oligonucleotides derived from a 60-bp stretch located in exon 1, with methylmercuric hydroxide-denatured RNA and with other reverse transcriptases.

Restriction Mapping and Southern Blot Analysis
Genomic DNA was prepared from chicken erythrocytes with the proteinase K method (30). Restriction digests of genomic DNA or plasmid DNA were done overnight according to the manufacturer's conditions (Boehringer Mannheim and New England Biolabs). Digests of DNA were analyzed by agarose-gel electrophoresis followed by transfer onto nylon membranes (Pall Inc.). After UV cross-linking, the membranes were hybridized according to Church and Gilbert (32), if oligonucleotides were used, or as described in Sambrook et al. (30) for larger cDNA probes. For the 64-bp probe, Denhardt's solution was left out of the hybridization mixture as suggested in recently published protocols (30).

Sequencing
Sequencing was carried out according to the method of Sanger et al. (35). Cycle sequencing was performed with the fmol kit (Promega); the annealing step in the cycling procedure was done at the T m of the oligonucleotide. Maxam-Gilbert sequencing was done according to the modified method of Som and Tomizawa (36).
All plasmids were grown in Escherichia coli XL-1 blue (Stratagene), and DNA was isolated either by an alkali lysis-type method (30) or by column preparation (Nucleobond AX kit, Machery-Nagel, Oensingen, Switzerland) according to the manufacturer's descriptions.

RNase Protection
Total RNA was prepared from excised tissue specimen from adult chicken according to the method of Chomczynski and Sacchi (37). Two ml of solution D were used per 100 mg of tissue for homogenization with a Polytron mixer.
Plasmids pStM5 and pStM6 were linearized with HindIII/NdeI and BamHI/EcoRI, respectively. Both were transcribed with T7 RNA polymerase (Promega), producing probes of approximately 160 and 220 nucleotides, respectively. [ 32 P]rCTP-labeled cRNA-probes were synthesized as follows. Three l of 5 ϫ transcription buffer (Promega) dithiothreitol (nuclease-free, Sigma) to a concentration of 10 mM; rATP, rGTP, and rUTP (Promega; final concentration, 400 M each); and 60 Ci of [ 32 P]rCTP were mixed in a siliconized Eppendorf tube. Diethyl pyrocarbonate (Sigma) -treated distilled H 2 O was added to give a total volume of 15 l. Linearized plasmid template to a final concentration of 40 to 50 ng/l, RNAsin (20 units, Promega), and T7-RNA polymerase (20 units, Boehringer, Mannheim) were added at the end. Transcription was done for 1 h at 37°C. The cRNAs were purified by a preparative acrylamide gel and recovered from the gel slices by incubation in cracking buffer (0.3 M NaCl, 0.1 mM EDTA, 10 mM Tris, pH 7.5) for 1 h at room temperature and subsequent extraction and precipitation.
The RNase protection was done with the RNase protection kit from Boehringer Mannheim. 10 g of total RNA was used from each tissue with ϳ10 5 cpm of labeled probe, and hybridization was done overnight at 50°C. RNase digestion was done with RNase A and RNase T1 at 37°C for 1 h. One-half of the assays was analyzed on a 4% sequencing gel.

In Situ Hybridization
Tissues were removed from a decapitated chicken and treated as described previously (38,39), with one minor modification. After the xylene steps, the tissues were transferred directly into melted paraplast at 60°C.
In vitro transcription in the presence of [ 35 S]UTP (Ͼ1000 Ci mmol Ϫ1 , Amersham Corp.) was done essentially as described above; 60 Ci of [ 35 S]rUTP were used instead of rCTP. The final volume was 20 l. After the first hour of incubation, another aliquot of RNA polymerase was added, and the mixture was incubated for another hour. The template was digested with DNase I to stop the reaction. All probes were purified by Sephadex G-50 columns, and sizes were controlled on an analytical 4% polyacryamide-urea gel. Some conditions of prehybridization, hybridization, and subsequent washing (38,39) were slightly modified. Probe for hybridization was diluted in hybridization buffer to a concentration of ϳ20,000 cpm l Ϫ1 , and 30 l were applied per slide. Dithiothreitol concentration in the hybridization buffer was increased to 100 mM (40). Stringent washes were carried out at 55°C in 0.5 ϫ SSC, 50% formamide, 10 mM dithiothreitol for 2 h.

RESULTS
Cloning of a Mi a -CK cDNA Fragment-In chicken, the cDNA sequences of the products of three out of the four CK genes are known. In case of the Mi a -CK isoform, only partial amino acid sequences have been published. To amplify a cDNA fragment coding for this isoform, a PCR approach with degenerate oligonucleotides was chosen. As chicken Mi a -CK is highly expressed in adult brain (17), total RNA from chicken brain was used in a RT-PCR. This RT-PCR produced a fragment of 653 bp (Fig.  1A), which was subcloned (clone UB11-83) and sequenced. Comparison to the human, rat, and mouse Mi a -CKs (19,24,41) indicated that we had cloned a chicken Mi a -CK fragment. The levels of amino acid identity were around 89% for all of them.
A chicken brain gt10 cDNA library was screened with this fragment to isolate the full-length cDNA. Out of this screening, one clone, containing sequences coding for Mi a -CK, was purified to homogeneity, subcloned, and sequenced (UB15-9). This clones. The localization of these three clones along the gene locus is indicated. Clone 72 is rearranged at a Sau3AI site in intron 7 as verified by direct sequencing of the clone. BamHI fragments of clone 72 were subcloned and sequenced in part (see sequencing strategy, indicated by arrows). The region of the BamHI sites and of the 3Ј-end of the gene were sequenced directly on the corresponding genomic clones (indicated by thick bars above the lines of the clones). The sequencing strategy is shown by arrows, and intron 1 has not been sequenced in total as indicated. At the ends of the genomic -clones there are BamHI sites that probably are cloning artifacts from a MboI-BamHI fusion. At the 5Ј-end of clone 67 is an additional BamHI site. Because clone 72 contains a upstream HindIII site, which in combination with the HindIII site in clones 64 and 67 gives rise to the same HindIII fragment as the one found in genomic Southern blots, it represents a faithful segment of the chicken Mi a -CK locus. B, BamHI; Sa, Sau3AI; H, HindIII; T, TaqI; , parts sequenced directly on the clones; , 5Ј-UTR with length not exactly known; , leader peptide; , normal exon; , exon 6, conserved in all mammalian and avian CKs; , 3Ј-UTR.
cDNA sequence starts only 170 bp downstream of the ATG initiation codon (see Figs. 1 and 2), reaches to the 3Ј-end at 1383 bp of the mRNA (see Fig. 1A), but contains no poly(A) tail. The RACE-RT-PCR method (31) is a way to amplify unknown cDNA ends, with the advantage of avoiding time-consuming screening procedures. Hence this method was used to verify the 3Ј-end of the chicken Mi a -CK cDNA. In this experiment, a poly(A) tail was found at the expected position. The polyadenylation signal sequence CATAAA was identified 19 bp upstream from the 3Ј-end (see Fig. 2, boxed nucleotides at the end). The 3Ј-UTR is 134 bp long, but it lacks the sequence similarities observed by others (42) Fig. 1 (arrows). The length of intron 1, which was not sequenced in total, was determined by analysis of restriction enzyme digests, and only part of the 1 kb that was sequenced is shown. Capital letters represent exon sequences, whereas lower case letters stand for intron and 3Ј-and 5Ј-flanking sequences of the gene. All intron-exon boundaries were sequenced, and splice sites were identified in all cases (underlined; points refer to nucleotides that do not match the consensus splice sequences). The polyadenylation signal is boxed. In the promoter region, the two GC boxes are underlined twice (A), and the putative AP-2 binding sites are indicated by a single line above the sequence. The two potential E2A contact sites are shown by arrows below the sequence and the MRE-site by an arrow above. The derived amino acid sequence is indicated under the respective exon, and the mitochondrial import signal is underlined. The numbering of the protein sequence starts with the first amino acid of the mature protein. Thus the initiating methionine is amino acid Ϫ39. The nucleotide sequence is not numbered with the exception of the region upstream of the ATG. There are 120 nucleotides/full line, and every 10 nucleotides there are asterisks on top of the figure.
The RACE-RT-PCR method was also applied to amplify these missing 5Ј-terminal sequences. The largest extension obtained from these experiments provided only 83 additional bp compared with the cDNA clone UB15-9 (Fig. 1A). Other attempts, using strongly denaturing agents like methylmercuric hydroxide to melt possible secondary structure elements in the RNA, did not yield any further sequence information. The method was working in our hands, as the missing 5Ј-end of chicken Mi b -CK (15) was cloned at the same time. There we obtained the remaining 57 bp of the coding region, and the 156 bp of the 5Ј-UTR (the complete sequence of the Mi b -CK leader peptide is given in Fig. 5). Thus at the 5Ј-end of chicken Mi a -CK, 200 -300 bp resisted cloning by cDNA-based methods for unknown reasons.
The missing sequences were finally obtained by analysis of genomic clones (see below). They code for the leader peptide of 39 amino acids, necessary for import into mitochondria, and the 5Ј-UTR, which we predict to have length of ϳ100 bp. The whole cDNA without poly(A) tail has therefore a length around 1480 bp.
Reduced Conservation of the Chicken Mi-CK Leader Peptides-As mentioned before, we have also cloned the missing 5Ј-part of the chicken Mi b -CK cDNA, which encodes part of the leader peptide necessary for import into mitochondria. Together with the chicken Mi a -CK sequence, we now have two nonmammalian leader peptides available for analysis of function and evolutionary relationships. Both leaders show the typical tripartite structure of mitochondrial import sequences (43). As can be seen in Fig. 3, the mammalian Mi-CK leader peptides are highly conserved in a isoform-specific manner. The leader peptides of mammalian Mi a -CKs display amino acid identities around 90%, and the same is true for the Mi b -CKs ( Fig. 3, table). If the chicken Mi-CKs are included in this comparison, the values of identity are much lower, and the isoform-specific conservation is no longer striking. The chicken Mi a -CK leader peptide shows identities to the ubiquitous Mi-CKs between 56.4 and 61.5%, and the Mi b -CK peptide is 54 and 59% identical to the mammalian sarcomeric Mi-CK peptides. These values are barely higher than those obtained in interisotype comparisons, which range from 38.5 to 51.3%.
The amino acids conserved in all of the leader peptides of either the Mi a -CK or the Mi b -CK are, with one exception, also conserved in the chicken Mi a -and Mi b -CKs. Preliminary expression studies done with the chicken Mi b -CK cDNA have shown that the Mi b -CK leader peptide is functional and that Mi b -CK gets imported into mitochondria of heterologous mammalian CV-1 cells. 3 Isolation of Genomic Mi a -CK Sequences-To investigate the structure and the putative regulatory elements of the chicken Mi a -CK gene, the EcoRI/SmaI cDNA fragment of clone UB15-9 was taken to screen a chicken genomic library. In four rounds of screening, six of the initially 17 clones could be purified to homogeneity. The clones were analyzed by Southern blots using oligonucleotides from different regions of the known cDNA fragment. Clone gMia 72, encompassing 12 kb, was chosen for subcloning and further analysis. This clone contains about 5 kb of sequences upstream of the ATG, including the putative promoter region, and reaches as far toward the 3Ј-end as into intron 7. However, direct sequencing of the -clone revealed a rearrangement at the 3Ј-end (see Fig. 1B). The 3Ј-region of the gene was therefore directly sequenced on two other, not rearranged, -clones (64 and 67 in Fig. 1, thick black bars).
In total, close to 5.3 kb have been sequenced (Fig. 2) including 500 -600 bp of putative promoter region. About 300 bp of promoter sequences further upstream were analyzed, but they are only sequenced on one strand and therefore are not included. The nonsequenced part is located in intron 1. At the 3 S. M. Mü hlebach and P. Kü nzler, unpublished results. ; the fragements were resolved on an agarosegel and transferred onto a nylon membrane. The first blot (A) was hybridized to a 32 P-labeled 319-bp cDNA fragment (nucleotides 169 -488 in Fig. 1B; random prime labeling; specific activity, 10 9 cpm/g) and the second (B) to a 32 P-labeled 63-bp cDNA fragment (nucleotides 88 -151 in Fig. 1B; PCR labeling; specific activity, 4 ϫ 10 9 cpm/g). The blots were exposed to an x-ray film for 2 days. The fragments of the marker (a combination of phage DNA cut with EcoRI and HindIII and phage DNA cut with BglI) are indicated in kb. The hybridization pattern shows that there is a single gene and that the genomic clones contain hybridizing fragments of identical size. 3Ј-end, there are 83 bp of known sequence after the polyadenylation site. The chicken Mi a -CK gene is approximately 7.6 kb long and has nine exons. The defined exon sizes range from 86 -247 bp, but the putative exon 1 may be larger. Intron 1 is 4.2 kb long, of which 1.2 kb were sequenced. All other introns are rather small (Ͻ520 bp), giving rise to a compact gene structure on the 3Ј-side. Analysis of the nucleotide sequences around the intron-exon junctions (underlined in Fig. 2) shows that they correspond well to the AG-GT splice junction rule (44) for intron-exon boundaries. The dots in the splice regions indicate deviations from the rule.
The 5Ј-region of the chicken Mi a -CK gene shows a high GC-content. It is around 80% in the whole BamHI fragment comprising exon 1 (see Fig. 1B), and the same holds true for the entire exon 1, regardless of the size of 5Ј-UTR proposed. The three mammalian Mi a -CKs (19,24,41) also have a higher GC-content in the exon 1 region of the cDNA, but it is 15% below the content found in chicken. Strong RNA secondary structures, as predicted by the Stemloop program in the GCG software package (45,46), are an additional feature of exon 1. Both the high GC content and the strong secondary RNA structure prevented the mapping of the transcription start site of the chicken Mi a -CK gene (see Figs. 1 and 2), although we tried primer extension analysis and RNase-and S1-protection analysis. From the putative transcription factor binding sites found (see below and Fig. 2), a transcription start site about 100 bp upstream of the ATG initiation codon seems most likely. Alternatively an additional, untranslated exon might exist, but in the sequences upstream of the ATG there was no indication of a further splice site. Therefore it is highly probable that exon 1 is in fact the first exon and that transcription initiates about 100 bp upstream of the ATG.
Analysis of the chicken Mi a -CK gene with the Grail software (47) predicts, around exon 1 (see Fig. 1 and 2), a CpG island with a very high CpG score of 0.98 (for definition of this score, see Ref. 48). The island starts at Ϫ587 bp upstream of the ATG and extends to ϩ506 bp downstream. The human Mi a -CK gene also has such a CpG island from Ϫ198 to ϩ236 bp, with a score of 0.73. In both cases, it has to be investigated if these islands are undermethylated, but their presence in the promoter regions is of interest in the context of their regulation. A lot of so-called housekeeping genes (49) or tissue-specific genes with a broad range of expression (50) do have such CpG islands. In addition, these genes usually lack TATAA and CCAAT boxes. The putative promoter region of the chicken Mi a -CK gene fulfills all of these features and is typical for such a gene type. A search for specific potential binding sites for transcription factors upstream from the ATG revealed, among many SP1 sites (14 on the upper strand), two GC boxes that match the consensus defined by Kadonaga et al. (51) and are marked with a double line in Fig. 2. Further potential binding sites were found for the general factor AP2 (single lines above the sequence) matching the consensus sequence 5Ј-CCCMNSSS-3Ј (52) for the ubiquitously expressed products of the E2A gene (5Ј-GCAGGTGGC-3Ј, arrows below the sequence) and for a metal response element (MRE, at Ϫ82 bp, arrow above), which is identical to the MREa element in the mouse metallothionein-I gene promoter (53). Suzuki and co-workers (54) have identified three sequence elements (Mt1, Mt3, and Mt4) common in the 5Ј-flanking regions of nuclear genes coding for mitochondrial proteins. The same three stretches are also found in human sarcomeric and mouse ubiquitous Mi-CK. However, none of these "mitochondria-related" binding sites is found in the sequenced upstream region of the chicken Mi a -CK, although they might be located further upstream. Glucocorticoid and estrogen response elements were reported in mouse and human ubiqui-tous Mi-CK genes in introns or downstream of the gene. In chicken, only some, possibly nonfunctional, half-sites for the glucocorticoid response element are found in intronic regions.
There Is Only One Mi a -CK Gene-Genomic Southern blot analysis, using high molecular weight genomic DNA prepared from chicken erythrocytes, was performed to verify the gene structure and to rule out the possibility of pseudogenes as were found for rat and human B-CK genes (21)(22)(23)55).
The same 317-bp fragment used for the screening of the genomic library was used as a probe and produced only one signal per lane on a Southern blot of DNA digested with BamHI, HindIII, and TaqI (Fig. 4A). The observed hybridizing fragments were 8.5, 11, and 2.9 kb long, respectively, and are in perfect agreement with the restriction map derived from the genomic clones as shown in Fig. 1B.
To further support the gene structure reported here, another Southern blot (Fig. 4B) was hybridized with a 64-bp probe, labeled by PCR, from the 5Ј-end of the known cDNA sequences (nucleotides 88 -151 of the cDNA depicted in Fig. 1A; corresponding to part of exon 1). Again the fragments of 1.1, 11, and 2.3 kb, respectively, were the same as those found in the genomic clones. These data therefore indicate that the chicken Mi a -CK gene is a single copy gene. Additionally the 317-and 64-bp probes both hybridize in HindIII-digested DNA to a fragment of the same length of 11 kb. This shows that the organization of the clones correctly represents the genomic locus.
Expression Pattern of the Chicken Mi a -CK in Adult Tissues-In mammals, the homologue of the chicken Mi a -CK is usually called ubiquitous Mi-CK, thus implicating that is actively transcribed in all tissues. The expression of chicken Mi a -CK was investigated by RNase protection experiments on total RNA from various tissues of the adult chicken (see Fig. 5). Probes specific for Mi a -CK and, as positive control, for B-CK, both derived from the 3Ј-end of the corresponding cDNAs (see "Materials and Methods"), were used to test RNA from brain, leg muscle, heart, gut, kidney, and testis. This showed that Mi a -CK is highly expressed together with B-CK in brain and gut but only weakly in testis. There was no detectable expres- Mi a -CK (B) were performed as described under "Materials and Methods." 5 g of total RNA from brain (2), leg muscle (3), heart (4), gut (5), kidney (6), and testis (7) was used per lane. In lane 1, the undigested probe was loaded (specific activity of the B-CK probe Ͼ 4.6 ϫ 10 8 cpm/g; specific activity of the Mi a -CK probe Ͼ 6.3 ϫ 10 8 cpm/g). The autoradiographs for B-CK and for Mi a -CK were exposed for 12 h at room temperature, except for the testis lane of section B, which is from a 24-h exposure. u, undigested; p, protected. sion in other tissues. Usually the level of B-CK expression was higher than that of Mi a -CK, with the exception of gut, where the inverse was true.
Because it is possible that small groups of cells express Mi a -CK even in tissues where no expression in whole tissue RNA was detected, in situ hybridizations were performed. Paraffin sections from spinal cord, liver, gizzard, and gut were hybridized with the same probes but labeled with 35 S. There was no hybridization in liver either for B-or for Mi a -CK mRNA (Fig. 6, A-C), and B-CK was the only CK present in the smooth muscle portion of the gizzard (Fig. 6K).
In gut, the signal of the B-CK probe was localized over the longitudinal and circular smooth muscle tissue. The B-CK probe hybridized to the base of the villi and a diminishing signal was detected toward the luminal region. There is also hybridization in a ring just inside the smooth muscle layer, which represents the muscularis mucosae. The hybridization to the muscle tissue seems stronger than in the villi. Mi a -CK was restricted to villi and showed no expression in the surrounding smooth muscle tissue (Fig. 6, G-I).
Finally, in spinal cord, Mi a -CK showed a punctuated hybridization pattern, which localized over the cell bodies of neurons of the gray matter; there it is expressed together with B-CK. In other cells, only unspecific hybridization was observed. On the other hand, B-CK hybridizes at a lower level in gray as well as in white matter (Fig. 6, D-F) and seems not to be restricted to neuronal cells. DISCUSSION Recently several reports described Mi-CK amino acid sequences and the corresponding gene structures of mammalian Mi-CK genes, and regulatory and evolutionary aspects of the Mi-CKs were proposed (20,24,25,54). Among the nonmammalian species, the chicken CK isoenzyme family has already been well documented (2, 10, 11, 57-60) with the exception of the fourth gene, Mi a -CK. Here we present the chicken Mi a -CK amino acid sequence derived from the cDNA as well as the structure of the corresponding gene. The analysis of the chicken system and its comparison with the mammalian CK set allows conclusions to be drawn on functional elements of the protein, like the leader segment, on regulatory properties of the chicken Mi a -CK promoter leading to regulated expression, as well as on evolution of the CK isoforms.
An amino acids sequence comparison (20) of the creatine kinase isoenzymes has shown that the proteins can be arranged into six different groups based on their levels of sequence identities. Two groups consist of the cytosolic CKs of the fishes/amphibians (see ref. 20) and will not be further discussed, while the other four groups are formed by the different isoforms of mammals and birds. The nine exons of the Mi a -CKgene give rise to a precursor protein of 417 amino acids, including an amino-terminal mitochondrial import sequence of 39 amino acids. The mature chicken Mi a -CK fits in the comparison into the group of the Mi a -CKs or ubiquitous Mi-CKs. The levels of identities between the ubiquitous Mi-CKs known so far are around 90% for the mature protein as shown in Fig. 3B.
In addition, an isotype-specific conservation of the leader peptides was noted for mammalian Mi-CKs, and hence it was suggested that these peptides might act as isoprotein-specific import sequences or bind to specific import receptors (19). The observed low identities with the chicken leader peptides (Fig.  3B) are not in favor to extend this hypothesis also to nonmammalian isoforms, unless the isoprotein-specifically conserved residues found between chicken and mammalian leader peptides (amino acid positions 10, 11, 18, 23-25, 27, 28, and 30) were sufficient to ensure the specific import (Fig. 3B). In contrast to the leader peptides, the mature protein Mi a -and Mi b -CK sequences are conserved isoform-specifically to a degree of almost 90%, and no segments can be found in the mature proteins where the conservation drops to the low levels of the leader peptides. Hence, the reduced conservation of the leader peptides cannot be explained by the greater phylogenetic distance of the chicken from mammals since leader peptide and mature proteins must have undergone simultaneous evolution. The noted high homologies in the mammalian peptides are therefore a mere fact of their phylogenetic closeness, and the apparent lower evolutionary pressure for conservation of the leader peptides is only observed when the large evolutionary gap, as the one between mammals and birds, is analyzed. Thus, it is unlikely that the conservation of the mammalian Mi-CK leader peptides represent isoform-specific functions. This is supported by the observed import of the chicken Mi b -CK into the mitochondria of mammalian fibroblastic CV-1 cells, where endogenous Mi a -CK, but no Mi b -CK, can be expected.
On the level of the nucleotide sequences, 78% identity is observed if the mature chicken Mi a -CK is compared with any of the mammalian ubiquitous Mi-CKs. As conservation on the protein level with 90% is much higher, nucleotide changes occur therefore mainly at wobble positions. The same holds true for the Mi b -CKs or any of the cytosolic CKs. In case of the UTRs, the situation is different. Cheng et al. (42) have shown that a sequence in the 3Ј-UTR of the rat ubiquitous Mi-CK is involved in regulation of its expression/translation. Whereas this sequence stretch of 72 bp is conserved in the mouse and the human sequence, it cannot be found in chicken Mi a -CK, suggesting that this mechanism of regulation is not active in chicken.
A schematic representation comparing the four known Mi-CK genes is given in Fig. 7. The organization of the chicken gene is similar to the two other known Mi a -CK genes from man and mouse. The localization of intron-exon boundaries in the coding region of the cDNA and the exon-sizes (except number 1) are conserved between these genes. In addition, the chicken gene lacks a noncoding first exon, like the mammalian ubiquitous Mi-CKs but contrary to all other known CK genes. Whereas the mouse and the human genes show a common bipartite gene structure with a clustering into the goups of exons 1-6 and 7-9 (41), in chicken Mi a -CK, exon 1 is separated from the compact rest of the gene by a rather large intron. The sarcomeric Mi-CK gene, although it is with 37 kb the largest Mi-CK gene known so far and in addition has two noncoding first exons, has exon intron boundaries at the same positions of the coding region as the ones noticed in the ubiquitous Mi-CK genes (see Fig. 7). Hence, the two gene types might have evolved by a gene duplication event.
As noted earlier (25) the mitochondrial CK gene structure is different from that of the cytosolic CKs. There is only one exon (exon 6, respectively exon 8 in human sarcomeric Mi-CK) that is conserved through all of the CKs in its size and localization in the coding region (black exon in Fig. 7). Other features, for instance the conserved 5Ј-noncoding exon in case of the cytosolic CK genes, are not found in the Mi-CK genes. Interestingly the region of the conserved exon is also the exon with the highest homology at the amino acid level among the different guanidino kinases (20). This might indicate a critical role of these residues for the structure and function of guanidino kinases, as has already been shown for a tryptophane residue in this region (62).
All of the data mentioned so far show that the CKs form a group of evolutionarily related isoenzymes and can further be embedded into the larger family of the guanidino kinases. The four different CK isoforms most probably evolved by three gene duplication events, with the first of these producing a primordial cytosolic and a primordial mitochondrial isoform. The phylogenetic tree, which can be derived from the protein comparison (20), suggests this first duplication to have occurred before the separation of the echinoderms from chordates, which is in agreement with published data on the expression of Mi-CKs in sea urchins (63,64). The significant homology, still found if the protein comparison of the CKs is extended to other guanidino kinases like ArgK or guanidinoacetate kinase, is a strong indication for a common ancestor of the guanidino kinases. ArgK has been suggested to be this ancestor on the basis of data on its dimerization capacity with cytosolic CKs, the nature of its substrate arginine as being part of basic metabolism, and especially from the distribution of arginine and ArgK in the animal kingdom (1). However, creatine has been found in sperm of many nonvertebrate taxa as well (1,65). Taking into account that the recently cloned guanidinoacetate kinase shows higher homology to CKs than to ArgKs (56), it can be FIG. 7. Conservation of the Mi-CK gene structures in mammals and birds. The chicken Mi a -CK gene structure is compared with the three other known Mi-CK genes of human and mouse (24,25,41). The exon sizes are not drawn to scale, but for any given gene, they are proportional in size. The coloring of exons is as mentioned in Fig. 1. The sizes of the genes are indicated below the names in kb; due to its large size, the scale of human sarcomeric Mi-CK is much smaller as indicated by the scale bar. The numbers below the chicken gene refer to the nucleotide positions of the intron exon boundaries in the chicken Mi a -CK cDNA. As the mammalian Mi-CKs have one amino acid more in their exon 1, the corresponding numbers have to be increased by three (153, 352, and so on). The numbers to the right indicate the 3Ј-end of the cDNAs without the poly(A) tail.
suggested that a duplication event has first produced a guanidinoacetate kinase and ArgK and that later CK has evolved from guanidinoacetate kinase.
The promoter region of chicken Mi a -CK is rather similar to those of the human and mouse genes and displays all of the features attributed to housekeeping genes (49) or tissue-specific genes with a broad range of expression (50). It is at present not known whether any of the transcription factor binding sites identified by sequence analysis are actually used for the regulation of the chicken Mi a -CK gene. During rat pregnancy, B-CK as well as ubiquitous Mi-CK (Mi a -CK), are regulated in rat uterus by steroid hormones, and binding sites for steroid receptors are expected. However, in the chicken Mi a -CK gene, no such sites have been identified, and only possibly nonfunctional half-sites for glucocorticoid receptors have been found. On the other hand, it has been shown for rat B-CK that the regulation is independent of the binding of an estrogen receptor to the promoter region (66). The three binding sites Mt1, Mt3, and Mt4 (54) identified in nuclear genes coding for mitochondrial proteins have also been found in the mouse ubiquitous Mi-CK gene. These sites are present neither in the chicken nor the human gene, but in both cases the known 5Ј-located sequences may be to short to contain these sequences. Summarizing the data shown, the Mi a -CK gene displays in part the same regulatory elements found in mouse and human. Some of the additional elements reported especially in mouse are not found in the chicken Mi a -CK gene, probably due to the limited sequence information at the 5Ј-end. The "missing" binding sites might, however, be important to explain the observed restricted expression pattern found for chicken Mi a -CK.
The presence of Mi a -CK in brain tissues is already well documented (2, 60). Our localization in spinal cord shows that Mi a -CK is coexpressed with B-CK in the cell bodies of neural cells of the gray matter but is absent from any other region where only B-CK is found. Hence, neurons seem to rely on a functional PCr shuttle in spinal cord. Whereas other investigators reported minute CK expression in liver, our in situ hybridizations show no Mi a -CK or B-CK expression in general. Either the transcripts are below detection limit or not present at all. The smooth muscle-containing tissues analyzed by in situ hybridization do not express Mi a -CK in their smooth muscle portions, but they display considerable amounts of B-CK. Hence, gut (duodenum) and gizzard smooth muscle in chicken function in the absence of Mi-CK, which is different from vascular and intestinal smooth muscle of guinea pig (8) or smooth muscle from rat (42). For chicken gizzard, the lack of Mi a -CK is due to its peculiar contractile properties (60). Whether this holds true for chicken gut as well is not known. The only portion in gut expressing Mi a -CK is the border region of the villi, which suggests that a functional PCr circuit might be important for cells of the brush border involved in resorption processes. These data show that Mi a -CK expression in tissues of adult chicken is more restricted than that of B-CK. They indicate that the name ubiquitous, given to the mammalian Mi a -CKs, is not justifiable in chicken. The features of the Mi a -CK promoter indicating a housekeeping gene are misleading, and there must be other regulatory elements narrowing its expression. The putative additional regulatory elements will have to determined by future research.