Antibodies against F,-ATPase a-Subunit Recognize Mitochondrial Chaperones EVIDENCE FOR AN EVOLUTIONARY RELATIONSHIP BETWEEN CHAPERONIN AND ATPase PROTEIN FAMILIES*

Antibodies raised against two synthetic peptides from rat liver F,-ATPase a-subunit sequence recognized two main heat-shock proteins from Drosophila (p71 and pS6) and rat liver (p74 and pS4) cells. One of the antisera showed a 20-fold higher reactivity toward Escherichia coli GroEL chaperonin than toward the a-subunit puri- fied from Drosophila. Indirect immunofluorescence microscopy and subcellular fractionation experiments lo- cated both mammalian heat-shock proteins in the mitochondria. The recent findings of functional homology between chaperonins and a-subunits, together with the unsuspected immunological reactivity of two mitochondrial molecular chaperones toward antisera de- rived from two different sequence motifs of the a-sub-unit, strongly argue in favor of the existence of an evolutionary relationship between chaperonins and a-subunits. The complete sequence alignment of F-type ATPase a-subunits and chaperonins revealed the exist- ence of eleven most conserved regions (“30% of each protein sequence) with an overall amino acid identity of 20 2% and similarity of 39 4%. A search of protein data bases with three different consensus sequences

The mitochondrial FIFO-ATP synthase is a multisubunit protein complex of the inner mitochondrial membrane that couples the H+ electrochemical gradient generated from the respiratory chain to the synthesis of ATP (for reviews, see Amzel and Pedersen (1983) and Carafoli (1987a and1987b). The overall structure and function of mitochondrial ATP synthase is closely related to the bacterial and chloroplast counterparts. Its catalytic activity has been located to the watersoluble F, portion. It consists of five different subunits with the stoichiometry ratio of a,P,y&. Of the two major subunits of the F, complex, the p subunit is considered to be the catalytic protein, whereas a regulatory role has been ascribed to the a-subunit (Maggio et al., 1988;Moradi-Ameli et al., 1989). Besides, both a-and P-subunits of the F,-ATPase complex are evolutionary related protein families with the A and B subunits from V-type ATPases (Gogarten et al., 1989;Sudhoff et al., 1989;Nelson, 1992).
Recent reports from our and other laboratories have provided evidence that the a-subunit of the F,-ATPase complex is implicated in cellular processes different from those related to the synthesis ofATP. The existence of at least two highly conserved amino acid stretches among the chaperonin protein family and various a-subunits  led us to suggest a putative chaperonin role for the a-protein . This hypothesis has been confirmed by the reconstitution of an active hybrid ATPase from 0-less Rhodospirillum rubrum chromatophores and p-subunits from plant chloroplasts, possible only in the presence of trace amounts of the a-subunit (Avni et al., 1991). Furthermore, in the absence of a-subunit, the p-subunit binds the chromatophore without reconstituting an active hybrid ATPase (Avital and Gromet-Elhanan, 19911, thus suggesting that the a-subunit is able to promote the correct folding/assembly of the P-subunit. More recently, Yuan and Douglas (1992) have shown that yeast a-subunit deletion mutants exhibit, both in vivo and in vitro, delayed kinetics of mitochondrial protein import and processing. This has been noticed in the case of several mitochondrial precursor proteins related or unrelated to the ATP synthase complex and has provided further evidence that the a-subunit is involved in the import of mitochondrial precursors, another functional activity in which molecular chaperones play a fundamental role.
The existence of structural ) and functional (Ami et al., 1991; Yuan and Douglas, 1992) similarities between chaperonin and a-subunit protein families raises the possibility of an evolutionary relationship between them. Here, we provide evidence for the existence of an extensive amino acid similarity between chaperonins and a-subunits. The results of immunological experiments and complete sequence alignment strongly support the evolutionary relatedness of chaperonins and F,-ATPase a-subunits. The extension of these observations to the other major subunits of V-and F-type ATPases allows us to predict several amino acid residues in chaperonin sequences that could be involved in ATP binding.

Synthetic Peptides and Antibody
Production-Peptides DR-1 (NH,-VGLKAPGIIPRI-COOH) and DR-2 (NH,-YLHSRLLERAAKM-COOH), corresponding to the sequence regions of the F,-ATPase a-subunit from rat liver mitochondria comprising residues 162 to 173 and 333 to 345 of the preprotein, respectively, were obtained from Bio-Synthesis, Inc. (Denton, TX) as free peptides. Conjugation of the peptides to carrier keyhole limpet hemocyanin was carried out by the carbodiimide method. For antibody production, New Zealand rabbits were intradermically administered with a priming dose of 500 pg of the conjugated peptide. Boosting doses of 250 pg of conjugated peptide were administered subcutaneously every 2 weeks. The titer of the antibodies was tested by its ability to recognize increasing quantities of peptide by dot blot procedure. Titration curves of the antibodies showed a linear response in the peptide range of 0.1-100 pg (data not shown). For affinity purification of antipeptide antibodies, 25 mg of keyhole limpet hemocyanin were conjugated to 2.5 g of cyanogen bromide-activated Sepharose 4B by standard procedures. The column-unbound anti-DR-1 or anti-DR-2 antibodies were used in these studies. Both preinmune rabbit serum and the anti-keyhole limpet hemocyanin IgG serum fraction did not recognize any protein in SDS-PAGE'-fractionated proteins from Drosophila cells or from rat liver (data not shown).
Cell Culture and Heat-shock Deatment to Embryonic Drosophila Melanogaster CelLsSL2 cells were plated in M3 medium supplemented with 10% fetal calf serum and grown at semiconfluence in 60-mm plates at 24 "C. For the heat-shock treatment, plates were placed in a culture chamber at 37 "C. After a 60-min period of hyperthermic exposure, the remaining plates were placed back in a culture chamber at 24 "C, for an additional 60-min recovery period. At the indicated time intervals before, during, and aRer the heat-shock treatment, cells were removed and resuspended in 600 pI of a 20 m~ phosphate buffer containing 150 m~ NaCI, 1% Triton X-100, pH 7.0, and the following protease inhibitors: 2 p g / d leupeptine, 1% aprotinin, and 1 m~ phenylmethylsulfonyl fluoride. Resuspended cells were freezethawed three times and centrifuged for 10 min at 100,000 x g .
Zndirect Immunofluorescence Microscopy-Norma1 rat kidney (NRK) and liver (Clone 9) cells were plated in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and grown at semiconfluence on coverslips. Coverslips were fixed in methanol at -20 "C and then incubated with a 1500 dilution of the IgG fraction from a rabbit anti-(rat liver mitochondrial F,-ATPase) serum (Valcarce et al., 1988) or the IgG fraction of anti-DR-1 (1:50) or anti-DR-2 (1:50) serum for 30 min at 37 "C. Then, they were washed with PBS and hrther incubated for 20 min at 37 "C with a 1:lOO dilution of a goat anti-rabbit IgG fluorescein-conjugate (TAG0 Immunochemicals). Next, they were washed with PBS and mounted with GELVATOLrM. Fluorescence microscopy was performed with a Zeiss Axiovert 35 microscope.
Isolation of Rat Liuer Cells, Heat-shock Deatment, and Zsolation of Rat Liuer Subcellular Organelles-Albino Wistar rats fed on standard laboratory chow were used for the experiments. Isolated liver cells were prepared (Cuezva et al., 1986). Briefly, the rats were anesthetized with Nembutal (40 mgkg body weight) and the liver perfused with Ca2+-free Krebs-Henseleit buffer containing 0.5 m~ EGTA for 10 min. After this initial perfusion, the liver was perfused for another 20 min with normal Krebs-Henseleit buffer containing 0.05% (w/v) collagenase. Washed isolated liver cells were incubated ( 5 4 x IO6 cells) in 4 ml of Krebs-Henseleit buffer containing 2.5% (w/v) bovine serum albumin, 5 m The abbreviations used are: PAGE, polyacrylamide gel electrophoresis; NRK, normal rat kidney; PBS, phosphate-buffered saline. glucose, 2 m~ glutamine, and 5% (w/v) of a mixture of essential amino acids, at 37 or 44 "C (heat-shock treated), with continuous shaking (100 strokedmin) in 25-ml Erlenmeyer flasks with a mixture of 95% 0, and 5% CO, as gas phase. For measurement of rates of protein synthesis, 100 pCi of [36S]methionine (1,200 Cilmmol) were added to the incubation mixture. 1 h after incubation liver cells were pelleted, washed twice with PBS, and resuspended in mitochondrial isolation medium (Valcarce et al., 1988). Mitochondria were isolated in sucrose gradients as described . Mitochondrial pellets were resuspended in the same medium and processed as above described for SL2 cells. Cytosolic and microsomal fractions were obtained by centrifugation at 105,000 x g for 1 h at 4 "C from postmitochondrial supernatants. Rates of cellular protein synthesis under unstressed and heat-shocked experimental conditions were determined by the incorporation of radioactivity in 5 p l of the cellular homogenates .
Western Blot Determination of Immunoreactive Proteins-Aliquots of cellular (SL2 cells), or mitochondrial, cytosolic and microsomal (rat liver) proteins of the resulting supernatants were fractionated in SDS, 12.5% polyacrylamide gels, transferred to polyvinylidene difluoride membranes (Millipore), and probed with anti-DR-1 and anti-DR-2 serum as described (Valcarce et al., 1988;Izquierdo et al., 1990). Quantitation of the stain deposition on the membranes was carried out by laser densitometric scanning. Parallel gels were stained with Coomassie Brilliant Blue R-250. Protein concentrations were determined with the Bio-Rad protein assay.
Affinity Purification of Rat Liuer Proteins with Antipeptide Antibodies-IgGs from anti-DR-1 and DR-2 sera were purified using the Bio-Rad Econo-Pac serum IgG purification kit, following the supplier's instructions. Purified IgGs were coupled to Afli-Gel Hz (Bio-Rad) agarose matrix by the Fc region as indicated in the immunoaffinity kit instruction manual (Catalog number 153-6060) provided by the supplier. Coupling of IgGs to the agarose matrix was within the 70440% range. For sample preparation, rat liver hornogenates prepared in PBS, pH 7.0, were further polytroned (30 s, setting 6) and centrifuged at 3,000 rpm. Triton X-100 (1% final concentration) was added to the resulting supernatant, freeze-thawed three times, and further centrifuged at 45,000 rpm for 1 h. The final supernatant was diluted 1:2 and filtered (Millipore, 0.45 pm). The filtered sample was incubated overnight with IgG-immobilized matrices under continuous shaking. After elution of the nonspecifically absorbed proteins (25 ml of PBS containing NaC10.5 M (once) and 25 ml of PBS containing NaClO.01 M (twice)), immunoabsorbed proteins were eluted from the column with 2.8 M MgCb. Eluted fractions were dialyzed and lyophilized for further analysis by SDS-PAGE.
Sequences Alignment, Secondary Structure Prediction, and Data Base Search-The alignment of sequences belonging to the same protein family was performed by hierarchical clustering of the individual sequences based on the pairwise similarity scores, as described by Sneath and Sokal(1973) and implemented in the PILEUP program of the University of Wisconsin Genetics Computer Group Software Package (Deveraw et al., 1984). The identification of discrete amino acid sequences common in chaperonins and a-subunit protein families was achieved by MOTIF program (Smith et al., 1990). The alignment of sequences belonging to different protein families with low overall similarity (i.e. chaperonins versus a-F,-ATPases) was performed by visual heat-shock proteins in cultured embryonic Dronophila cells (SL2). I'rptltlr I)R-1 ~~H~-~~~I ,~P~~l l l ' l~l -~'~~~~l l I rorrr.spnnr1.; t n t h r wqucncr of t h r Fl-ATl'asr ~r-suhunit from rat liver mitochondria and comprises rrsidurs 162-173 of thr prrprntrin. Antlpcptidr :lntll)ndws wrrr raised in rahhits. S I 2 (rmhryonic I). rnr/nno,ws/rrJ crlls were plated in MR mrdium supplrmrntrd wlth 10'; frtal calf s w u m a n d Lvown a t srmiconfluencr at 24 "C. Hrat-shock trratmmt: fi0-mm platrs with SL-2 crlls wrrr placrd In a culturr chamhrr at 37 ( ' at the inittntlon of t h r heat-shock treatmrnt. Aftrr a 60-min heat-shock period, thr rrmaining platrs wrrr transfrrrrd to a culturr chamhrr at 24 ('. A . rc-prrwntntIvr The antiserum recognizes two protrins, a constitutive p56 and an inducihlr p T 1 . Ihth prntrins incrrasr in amount dunng thv hrat-qhwk t r r a t m m t . C, quantitation by Iasrr drnsitomrtric scanning of fivr diffrrent rxprrimrnts. similar to thosr drpictrd in 13. For rnch t h t . t h r m m n internrated volumr of t h r 0-min Ip561 and 45-min (pT1) samplrs was taken as rrfrrrncc. Thc rrsults shown arc' mcwns 5 S F : O p r n a n d ~( J V P~ nrrowhrnds indicate the initiation and end of thr hrat-shock period. D. Coomassir Rlur stained grls llnnrs 1 I and Ui.strm tllots ~l r r n r~ 2 1 w1tt1 anti-DRI srnlm of purifird F,-ATPasr cr-suhunit from Drosophiln and GroEI, chaprronin from E. coli.

Structural Identities between Chaperonins and a-F,-ATPases
inspection and detailed comparison of the sequence alignmrnt for each individual protrin family. For prrdicting the distrihution of srcondary structurr clements in thr diffrrrnt protrin familirs. Chou-Fasman and Gamier-Os~~t.horpe-Rohson's algorithms were applied on each individual srqucnce implrmented in the PEPTIDESTRUCTURE program of the UWGCGSI' (1)rvrraux P/ nl., 1984). The most rrprrsrnted conformational state for rvery rrsidur in the primary structure alignment of each protrin family was plottrd. Thr srnrch of consensus srqurnces hetween chaprrnnins and n-suhunits in thr SwissProt data hasr was prrformed by the algorithm rrportrd hy Wilbur and Lipman 119733). implcmentrd in the WORDSEARCHIMASK option of the program of the UWGC(X1' (Drvrraux I,/ nl.. 198.11.

RESULTS
Immunological Evidrncr for thr Existencc. of Structural Relatadnrss hrfnrrpn Chaperonins and n-Srrhr~nits-Experimental evidence for the existence of structural relatedness hetween chaperonin and n-suhunit protein families was obtained with antibodies raised against a synthetic peptide (DR-1) derived from r a t liver F,-ATPase a-subunit sequence, corresponding to the most conserved sequence element among hoth protein families (Luis rt al., 1990; see also rrgion V in Fig. 4). Analysis by Western blotting of the immunoreactive proteins in heatshocked cultured embryonic Drosophila cells (SL2) revealed that the antibodies exclusively recognized two heat-shock proteins ( Fig. 1R): a constitutively expressed p56 that showed a 2-fold progressive increase during hyperthermic exposure of the cells (Fig. 1, R and C ) and an inducihlr p71 whosr rxprrssion could solely he detected 45 min after thc initiation of hcatshock treatment ( Fig. 1, TI and C). Interestingly. although thc antiserum recognized the purified F,-ATPase tr-suhunit. from Drosophila (Fig. lD), the immunorractivity of this protrin toward the antiserum was 20-fold lower than that of thc. purifird E. coli GroEL chaperonin (Fig. ID). On the othrr hand, thc purified a-subunit from Drosophila (Santarbn r / 01.. 19931 showed an apparent electrophoretic mipation on STIS-T'AGE of 52 kDa (Fig. lD), significantly faster than that of thr immunoreactive p56, thus supporting t,he idca that thr constitutivr p56 is not the n-subunit or an n-suhunit isnfnrm. hut rathrr a Drosophila chaperonin homologue.
Irnrnunorractirv Strc,ss-rcspnnair~r Pro/rins A r r I~~c~n l i z r r l in thr Mitochondria ofMarnrnalian ~,'rll.s-Antihndirs ncainst t h r synthetic peptide were used to charactrrizr. at thr crllulnr level, the suhcellular compartmrnt(s) whrrr thr immunnrractive stress-responsive proteins arr localized. lndircct immunofluorescence microscopy of the mammalian ccllular linrs Clone 9 and NRK with anti-DR-1 antihodics rrvralrd a punctuated intracellular distrihution and morpholoq (Fig. 2,  Clone 9 ( A ) and NRK ( B ) cells grown to semiconfluence were incubated with a 1500 dilution of the IgG fraction from a rabbit anti-(rat liver mitochondria F,-ATPase) (left panel) or 1 5 0 dilution of the IgG fraction from anti-DR-1 serum (right panel). They were further processed as described under "Experimental Procedures" using as secondary antibody a goat anti-rabbit IgG fluorescein conjugate. Preimmune rabbit sera did not reveal any cellular immunoreactive materials. Bars, 15 p m . C, cytosol, mitochondria, and microsomes were obtained from rat liver homogenates. 50 pg of protein from each subcellular compartment were fractionated in 12.5% SDS-PAGE and either stained with Coomassie Brilliant Blue R-250 (left panel) or processed for Western blotting with anti-DR-1 serum (right panel). Bars indicate the electrophoretic migration of mitochondrial proteins recognized by this antiserum. D , heatshock treatment to isolated rat liver cells; cells were incubated for 1 h at 37 or 44 "C. After subcellular fractionation, 20 pg of protein were processed by SDS-PAGE and Westem-blotted with anti-DR-1 serum. The panel shows the resulting blot for mitochondrial proteins under stressed (44 "C) and nonstressed (37 "C) conditions. No immunoreactive proteins were detected in microsomal and cytosolic fractions (not shown).
mostly to the mitochondria.
Mitochondrial localization of the immunoreactive proteins in mammalian cells was further assessed by subcellular fractionation experiments of rat liver post-nuclear supernatants. Western blotting of cytosolic, mitochondrial, and microsomal proteins revealed the presence of a strong immunoreactive 74-kDa protein in the mitochondrial compartment, whereas negligible or no immunoreactivity was detected in cytosolic or microsomal proteins (Fig. 2C). Besides, the antibodies recognized three additional mitochondrial proteins with lower intensity ( Fig. 2C; apparent molecular masses on SDS-PAGE of -160, 57, and 54 kDa). However, a 1-h 44 "C heat-shock treatment of isolated rat liver cells, a treatment that promotes a 16-fold reduction in the rate of cellular protein synthesis (not shown), revealed that only p74 and p54 displayed a n up-regulated stress-responsive nature (Fig. 20 ). Interestingly, the heat-shock-promoted induction of mitochondrial p54 was much more pronounced than that of p74 (Fig. 30). In contrast, the heat-shock treatment did not promote the accumulation of any stress-responsive protein recognizable by this antiserum in either the cytosolic or microsomal compartment of rat liver cells (not shown).
Antibodies Raised against Another Synthetic Peptide Derived from Rat Liver a-Subunit Sequence Recognized the Same Set of Mammalian Mitochondrial Proteins-The results described above (Figs. 1 and 2) were consistent with the conservation of common antigenic epitopes in mitochondrial stress-responsive proteins of the 60-and 70-kDa family, both in insect and mammalian cellular types. Recognition of mitochondrial (Fig. 2) and bacterial chaperonin (Fig. 1C) by this antibody is not surprising due to the high amino acid identity existing between the peptide sequence of a-subunits and chaperonins . On the other hand, recognition of a stress-responsive 70-kDa member in the mitochondrial compartment, and not of other 70-kDa members present in the cytosol and in the endoplasmic reticulum (for review see Gething and Sambrook (1992)), reveals a striking immunological relatedness between mitochondrial molecular chaperones, despite their apparent low sequence similarity (Ellis and van der Vies, 1991;. Furthermore, to confirm the structural relatedness between the three protein families, antibodies were raised against another synthetic peptide (DR-2) derived from rat liver F,-ATPase a-subunit sequence, corresponding to a conserved sequence element found between chaperonin and a-subunit protein families ; see also region VZZI in Fig. 4).
In contrast to DR-1 antibodies, anti-DR-2 antibodies did not recognize blotted proteins either from Drosophila or mammalian cells (not shown). However, they reacted with proteins contained in soluble cell extracts as shown by affinity chromatography experiments performed with rat liver extracts passed on antibody columns. The pattern of specifically immunoabsorbed proteins obtained with both antibodies was similar, showing an enrichment in two liver proteins, p74 and p58, although with different efficiency; p58 and p74 being preferentially enriched with anti-DR-1 and anti-DR-2 antibodies, respectively (Fig. 3, right panel). Finally, in agreement with the immunofluorescence microscopy localization studies performed with antibodies to DR-1 (Fig. 21, Clone 9 and NRK cells stained with anti-DR-2 antibodies showed strong specific staining of mitochondria (Fig. 3, left panel ).
All together, the results described above indicate the existence of antigenic epitopes common to a-F,-ATPases and mitochondrial stress-regulated proteins of the 60-and 70-kDa family (Figs. 1-3). The partial sequence  and functional (Avni et al., 1991;Yuan and Douglas, 1992) similarities existing between chaperonins and a-subunit protein families, together with the results reported above, raise a most interesting question as to whether these two protein families shared a common ancestor. This, and further questions, are addressed in the following.
Sequence Identity between Chaperonin and F,-ATPase a-Subunit Protein Families-The alignment of chaperonins and a-subunit primary structures is shown in Fig. 4. Selected sequences included in each protein family a t least a prokaryotic, Complete sequence alignment between chaperonins and F,-Al"ase a-subunits. The selected chaperonins used in the alignment carboxylase subunit binding protein (RSBP). The sequences of F,-ATPase a-subunits are from rat liver (Rat), bovine heart (Bovine), yeast (Yeast), and maize (Maize) mitochondria and from E. coli. Sequence alignment between chaperonin and a-subunits was done by manual intervention using previously identified conserved amino acid sequences (regions Vand VZZZ)  as reliable anchor points for the alignment. The MOTIF program was used for the search of common amino acid sequences among different protein families. Numbers between slashes indicate the amino acid position in each sequence. Amino acid identities in the alignment are shown in black bores with white letters. Relevant amino acid similarities are indicated by gray boxes. Dots indicate gaps introduced for optimizing the alignment. Brackets above chaperonin sequences labeled with Roman numerals (I-=) indicate the length and position of the 11 conserved regions. Brackets below a-subunit sequences, labeled with capital letters (AX), indicate regions from which consensus amino acid sequences for both protein families were obtained and further used for searching the SwissProt protein data base (see Table 11). Consewative amino acids were considered as follows: Gly =Ala = Ser; Ala = Val; Val = Ile = Leu = Met; Ile = Leu = Met = Phe = Tyr = T r p ; Lys = Arg = His; Asp = Glu = Gln = Asn; Ser = Thr and Pro = Gly. yeast, plant, and mammalian species. Previously identified conserved amino acid sequences  were used as anchor points to perform this alignment. Additional amino acid sequences, conserved in both protein families, were identified by the MOTIF program (Smith et al., 1990). The MOTIF program proved also to be helpful for the identification of putative sequence duplications among the proteins belonging to the same family. Two of these duplications are shown in Fig. 4 for the chaperonins. Interestingly, duplications reinforced sequence homology between chaperonins and a-subunits. The optimization of the alignment required the introduction of only nine gaps in the chaperonin sequences.
The alignment allowed the identification of eleven conserved regions (Fig. 4, see roman numerals) arranged exactly in the same order within all sequences, spanning more or less the complete sequences and comprising 158 residues (-30% of each protein sequence) with an overall amino acid identity of 20% and similarity of 39% (Table I). The comparison of aligned sequences revealed 20 invariant and up to 44 identical plus highly conserved residues between both protein families, most of them clustered in the 11 regions defined in Fig. 4. The comparison of PILEUP alignment of chaperonins with PILEUP alignment of a-subunits (data not shown) pointed out that the 11 conserved regions between chaperonins and a-subunits (Fig.  4) corresponded to 11 most conserved regions within each protein family (Table I). Interestingly, whereas chaperonins showed an apparent duplication of -80 residues at the NH,terminal end of their sequences (Fig. 4), the a-subunits showed three insertions of -90, 9, and 20 residues in the central part of their sequences (see below and Fig. 5).
In spite of relatively low sequence identity between the two protein families, the homology between chaperonin and a-subunits shown in Fig. 4 was further supported by the results obtained by two independent approaches. First, a mean consensus secondary structure was obtained for each protein family (see "Experimental Procedures"). The alignment of both predictions according to Fig. 4 reveals that 14 a-helices, 9 p-sheets, and 18 turns are located at equivalent positions within their sequences (data not shown), suggesting that the conservation of secondary structure elements between these two protein families represents 90% of their residues if the insertions of chaperonin and a-subunits are not considered and 70% if considered. Second, a search (WORDSEARCH) of the protein data base (SwissProt) with three different consensus sequences derived from three different regions of the alignment (see Fig. 4, A X ) was carried out. Query sequences were designed (Table  11) from regions of the alignment (Fig. 4, A -C ) with either high (Table 11, A and B) or low (Table 11, C) stringency (percent of k e d residues in the sequence) always avoiding gaps. The results obtained from the search provided in the three cases a significant percent of the known a-subunit and chaperonin sequences, most remarkably in the absence of sequences for other proteins (Table 11). Thus, it is reasonable to suggest that the existence of such a similar amino acid arrangement along a significant length of the protein backbones is not the result of mere coincidence but rather of an evolutionary relationship between these two protein families (see below).

Identification of Putative Chaperonin Amino Acid Residues Involved in Nucleotide
Binding-In order to infer putative residues in chaperonin sequences responsible for adenine-nucle-between Chaperonins and a-F,-ATPases otide interaction, we carried out a multiple sequence alignment of the catalytic (A and P) and noncatalytic (B and a) subunits of V-and F-type ATPases with chaperonins. This approach is possible (Baron et al., 1991;Blundell et al., 1987;Hirst and Sternberg, 1991;Rost et al., 1993), because the a-subunit protein family is evolutionary related to the rest of the major subunits from V-and F-type ATPases (Gogarten et al., 1989;Siidhoff et al., 1989). Fig. 5 shows only the portion of this alignment (regions VI-Xin Fig. 4) that contains amino acid residues involved in adenine nucleotide binding in the P-subunits (for a recent review, see Penefsky and Cross (1991)).
This alignment reveals that the molecular footprint of chaperonins is present in the sequences of the four protein families of V-and F-type ATPases (note white letters on black background in Fig. 5). The similarity is reinforced in those regions shown to contain highly conserved residues between chaperonins and a-subunits. Quantitation of sequence similarity between the different ATPase subunits and chaperonins in various of the conserved regions, including those shown in Fig. 5, reveals that in most of them, the F-type ATPase a-subunits are the most closely related proteins to chaperonins (Table 111).
However, regions VI and I X show more or less the same degree of sequence conservation in the four ATPase protein families when compared with chaperonins (Table III), although residue conservation is not always found at the same position (Fig. 5). This is the main reason why the WORDSEARCWMASK search (Table 11) was not able to pick up these other homologous subunits. As mentioned previously for the a-subunit, all the AT-Pase sequences present three-amino acid insertions between regions VI and VII, VI11 and E, and M and X (Fig. 5). Interestingly, the insertion between regions VI and VII, and between regions M and X, contains 14 and 4 residues, respectively, that form the nucleotide binding site for the 6-subunit of the F,-ATPase complex (indicated with asterisks in Fig. 5). However, 11 residues (indicated by circles in Fig. 51, corresponding also to the P-subunit nucleotide binding site, are located in regions of amino acid Similarity between ATPases and chaperonins. To illustrate this point, the inset in Fig. 5 shows a three-dimensional model of the nucleotide binding site for the P-subunit of the F,-ATPase complex adapted from Vogel and Cross (19911, in which we indicate, according to the alignment shown in Fig. 5, the chaperonin residues of this active site that are identical (white letters in a black background), conserved (gray background), or unrelated to the P-subunit ones. Interestingly, chaperonins show in this model 6 identical and 3 conservative residues in the considered catalytic nucleotide binding site (inset of Fig. 5 on the left) of the P-F-type ATPase (Vogel and Cross, 1991;Penefsky and Cross, 1991). However, the regulatory nucleotide binding site (inset on the right) and the P-loop (Saraste et al., 1990) are missing in chaperonins, because these amino acid residues are located in the three insertions characterizing V-and F-type ATPases (residues indicated with x in Fig. 5). Fig. 5 also shows that the 70-kDa protein of V-type ATPases (A subunits) contains the highest number of.identica1 residues in the nucleotide binding sites (69% identity, 73% similarity), as expected from their sequence and functional homology with @-subunits of F-type ATPases (Pedersen and Carafoli, 1987a;Nelson and Taiz, 1989). F-type ATPase a-subunits show 48% identity (66% similarity) with respect to the active site residues. V-type ATPase B subunits (60 kDa) show 38% identity (52% similarity).
It should be pointed out that amino acid residues Gly'= and G~u~~~ in the sequence of the mature P-F,-ATPase protein (Walker et al., 1985; see inset in Fig. 5) were not present in the original model derived by Vogel and Cross (1991) and have been introduced in our approach as a result of their invariable ap-  Fig. 4 The amino acid length and exact location in the sequence of conserved regions (I-XI) are indicated by brackets in Fig. 4 Regions V and VI11 are those previously reported (Luis et al., 1990). Asterisks denote the regions i n which a gap is introduced as a result of the alignment. The percentage of identity of the conserved regions within each protein family was calculated considering the consensus sequence (invariant + plus highly conserved amino acid residues) obtained by the PRETTY algorithm, applied to a PILEUP alignment of 27 different chaperonin and 31 different a-subunit sequences (data not shown). The percentage of identity and similarity (in parentheses) between the two protein families was done by aligning (according to Fig. 4) the two independent consensus and calculating the number of shared identical and invariant plus highly conserved residues, respectively. The percent similarity among the complete protein sequences compared is calculated by the BESTFIT program: 50 t 5% for chaperonins, 69 4% F,-ATPase a-subunit, and only 18 f 2% among F,-ATPase a-subunits and the chaperonins. pearance in all sequences considered (equivalent residues are those indicated by the first two closed circles in Fig. 5). Interestingly, a mutant P-subunit protein bearing only two amino acid changes (GlylS3 + Val and .--) Arg) revealed a lower affinity for a nucleotide analogue compared with the wild type protein (Garboczi et al., 19901, strongly suggesting the involvement of GlyIffi in the nucleotide binding site of P-subunits and by analogy in other ATPase subunits and chaperonin sequences (Fig. 5). Additionally conserved amino acid residues appearing in all or most of the sequences considered are: TG (316, 317); SRV (445-447); acidic residue (323, 349, 350, and 419) and hydrophobic residue (374 and 387) whose structural andlor functional significance is presently unknown. The sequence of the mitochondrial chaperonin hsp 60 (first sequence in Fig. 5) is used as a reference for numbering of these residues.

DISCUSSION
Herein we provide immunological and structural evidence supporting that chaperonins and F,-ATPase a-subunits are homologous proteins. These findings have important implications for explaining the evolutionary origin of ATPases, as well as for the prediction of critical amino acid residues involved in functional activities common in both chaperonin and a-subunit protein families.
Are Chaperonins the Molecular Ancestors of ATPases?-The striking structural identity, 11 conserved regions arranged exactly in the same order within the protein sequences, and functional similarities (Avni et al., 1991;Yuan and Douglas, 1992) displayed by the chaperonin and a-subunit protein families, strongly suggest that chaperonins and a-subunits have evolved from a common ancestor. Consistent with this view, chaperonins should be also evolutionary related to the major subunits of the rest of ATPase protein families (Siidhof et al., 1989;Gogarten et al., 1989), as shown in Fig. 5      versa. Although it could be argued that the higher sequence similarity existing betweiin the a-subunits and the chaperonins (Table 111) is compatible with a phylogenetic scheme in which chaperonins have evolved from the noncatalytic subunit of the eubacterial ATPase, the following lines of evidence seem to support the opposite. First, (i) the lower sequence similarity within chaperonins  than that within any of the a-, 0-, A-, and B-subunits of ATPases (Sudhof et al., 1989;Gogarten et al., 1989;Nelson, 1992) and (ii) the genomic organization of mammalian chaperonins in a single exon (Venner et ul., 19901, whereas the a-and P-subunit genes of mammalian F,-ATPases (Pierce et al., 1992;Ohta et al., 1988) are organized in 10-12 exons, both argue in favor of an ancient origin of the chaperonin protein family. However, the lower sequence similarities within the chaperonins could also be explained by a relaxed selection pressure on the former family. Second, phylogenetic analysis of protein sequences (data not shown) reveals that chaperonins are partitioned into a branch of the dendogram different from the branch that connects catalytic and noncatalytic subunits of ATPases. The pivotal cellular role played by molecular chaperones in assisting protein folding/assembly of other proteins, affecting also the biogenesis of the ATPase families, seems to us to be a third functional argument supporting that ATPase subunits have evolved from a functional chaperonin, once the ancestral duplicated chaperonin gene experienced the insertions that characterized the second nucleotide binding site ofATPase subunits (Fig. 5). Since the evolutionary relationship between chaperonins and a-subunits is close ( Fig. 5; Table III), it seems reasonable to suggest that a-subunits could represent a stable evolutionary intermediate in the divergence ofATPases from the chaperonin ancestor. Therefore, we suggest that functional constraints regarding the chaperonin activity have prevented the a-subunits from diverging at specific sites (Fig. 4) at the same rate as the rest of ATPases. Prediction of Essential Amino Acid Residues in Chaperonin Protein Sequences-The structural and functional similarities between chaperonin and a-subunits has enabled us to suggest that several residues contained in region VI11 of the alignment ( Fig. 4 and Table 11) might be involved in polypeptide substrate recognition in both protein families (Alconada . In this paper, we have mapped putative residues in chaperonin sequences that could be involved in binding of ATP (Fig. 5). It is important to point out that our prediction of the chaperonin residues involved in ATP binding (Fig. 5) includes two residues, G~u~~~ and that we have suggested to be also involved in the polypeptide recognition domain of E. coli GroEL chaperonin (Alconada and Cuezva, 1993). Previous studies suggested that the equivalent Arg in @subunits was involved in ATP binding (Viale and Vallejos, 1985). However, recent findings have demonstrated that mutation of this residue in the P-subunit does not result in an impairment of the ATPase activity, but rather in the stability of the protein itself (Mueller, 1988; Aschenbrenner et el., 19931, a fact that could question the role of this residue in ATP binding. Anyway, the amino acid residues in the active site of a protein are usually found in the interfaces of the different domains that build the protein structure, so it is conceivable that those amino acid residues that define the chaperonin active site (or domain interface) will contain residues affecting both ATP binding and polypeptide recognition. Interestingly, we should stress that while this paper was under revision, two papers were published (Martin et al., 1993;Horwich et al., 1993), illustrating the potential of our approach. Horwich et al. (1993) have shown that Glu458 in E. coli GroEL is an essential residue for chaperonin function. Moreover, Hart1 et al. (1992) have provided evidence that azido-ATP is crosslinked to the highly conserved 14Tr4'' also in E. coli GroEL TABLE I1 Summary of Swissprot wordsearch output file for several consensus chaperonin and a-subunit sequences Location of consensus (query) sequences (AX) is shown by brackets below ATPase sequences of the alignment shown in Fig. 4. Conserved amino number of residues within the sequences that were not forced to match in the search. Conservative replacements of conserved amino acid residues acid residues in query sequences are indicated by one-letter code. Number between conserved amino acid residues in query sequences indicates the in query sequences are shown in parentheses. Random replacement of 1 or 2 of the conserved amino acid residues in consensus sequences provided unrelated sequences to the a-subunit and chaperonin protein families. The E. coli ams (for altered mRNAstabilityt is a 17-kDa polypeptide involved in mRNA turnover (Chanda et al., 1985) which corresponds to an internal fragment of E. coli GroEl chaperonin (Hemmingsen et al., 1988, Alconada and  Sequence similarity between chaperonins and the catalytic (70 kDa and P) and regulatory (60 kDa and a) subunits of Vand F-type ATPases in various conserved regions, known to contain amino acid residues inuolued in ATP binding in the P-F,-ATPase subunit in ATP binding in @-F,-ATPase subunits (Penefsky and Cross, 1991) and are represented in Fig. 5. Conserved sequence region Vis included herein Conserved regions V-X have been defined in Fig. 4. Conserved sequence regions VI-X are those known to contain amino acid residues involved to point out the higher sequence similarity existing between chaperonins and F,-ATPase a-subunits in other regions of their sequences. Sequence and each ATPase subunit family was done by the ratio of the number of identical plus conserved amino acid residues and the total number of alignment of region V with other ATPase subunits is not shown. Calculation of the percent similarity of the conserved regions between chaperonins residues in that region. Similarity rank a > @,60,70 = a > 8,60,70 a > P,60,70 a = p = 6 0 > 7 0 a > @ > 6 0 > 7 0 sequence. Fig. 4 shows that both G1uG8 and w477 are found among the predicted conserved and identical residues, respectively, in the a-subunit sequences (see asterisks above sequences in Fig. 4). Moreover, it has been shown that the COOHterminal region of the a-subunit is involved in the regulation of the ATPase activity of the F,-ATPase complex (Tozawa et al., 1993), a finding consistent with a similar observation in the bacterial chaperonin (Langer et al., 1992). These findings strongly suggest that site-directed mutagenesis experiments on conserved residues shared by both protein families could provide the information necessary for understanding the conformation and function of both ATPases and chaperonins. A More General Chaperone Role for the a-Subunit?-Our finding of an immunological relationship between a-ATPases with chaperonins ( Fig. 1) is not surprising considering the significant sequence identity existing between these two protein families (Fig. 4). However, what is really striking is that the two developed antibodies recognized also a mitochondrial molecular chaperone of the 70-M)a family (Figs. 2 and 3). Immunological recognition of the 70-kDa hsp by these antibodies points out the existence of certain extent of structural homology, in spite of the low sequence similarity among the main stress-responsive proteins involved in the biogenesis of mitochondria, and suggests that the sequences and/or structural elements recognized by anti-DR-1 and anti-DR-2 antibodies could be involved in a common functional activity of mitochondrial molecular chaperones.
At present, we could not provide a clear explanation for this unsuspected finding. It could be argued that the two conserved elements found among a-subunits and mitochondrial molecular chaperones of the 60-and 70-kDa family are involved in a functional activity such as bindinghydrolysis of ATP. However, this hypothesis seems to be less likely, since (i) the sequence of both synthetic peptides showed no sequence similarity with motifs of other ATPand GTP-binding proteins (Saraste et al., 1990), and (ii) the antibodies recognized exclusively a mitochondrial hsp 70 member, that is, they did not recognize any other protein, even under stressful cell conditions (see "Results"), in the cytosol or in the endoplasmic reticulum ( Fig. 2  and 3), two subcellular locations where closely sequence related hsp 70 members are known to be present. Thus, we speculate with the idea that such conserved elements define a unique structural and/or functional characteristic of mitochondrial molecular chaperones. The characterization of a chaperonin-like activity in the a-subunit (Avni et aE., 1991;Yuan and Douglas, 1992) could help to explain the molecular rearrangements the H+-ATP synthase complex undergoes during its catalytic cycle (Boyer, 1989), although it argues against the original definition of a molecular chaperone (Ellis and van der Vies, 1991). It could be also argued that the chaperonin-like activity of the a-subunit is necessary only for the assembly of the F,-ATPase complex (Yuan and Douglas, 1992), since it has been shown to promote the correct folding of the p-subunit of the complex (Avni et al., 1991) and that additional proteins with apparently similar functional specificity are necessary for the assembly of this oligomer (Ackerman andTzagaloff, 1990a, 1990b). However, the finding that maize mitochondrial chaperonin co-immunoprecipitates with the a-subunit (Prasad et al., 1990) suggests a functional cooperation between these two chaperones. In fact, the experiments of Yuan and Douglas (1992) seem to point in this direction. These authors have shown that yeast deletion mutants of the a-subunit exhibit delayed kinetics of protein import and processing for various mitochondrial precursor proteins. In contrast, the p-subunit deletion mutant did not show an altered protein import, whereas a double-deletion a,p-mutant, although partially restored its import competence, showed remarkably lower import efficiencies than the yeast wild type and a MAS 70 deletion host (a mutant for the outer membrane receptor of mitochondrial precursor proteins) (Yuan , 1992). Taken all together, these findings could suggest a wider substrate specificity for the chaperonin activities of the a-subunit in the biogenesis of mitochondria.

Structural Identities between Chaperonins and a-F,-ATPases
Acknowledgments-We are indebted to Drs. I. Sandoval and J. F. S a n t a r h for critical reading of the manuscript and supply of pure rified E. coli GroEL chaperonin to S. Marco (Centro Nacional de Bio-a-subunit protein, respectively. We also acknowledge the supply of putecnologia, Spain). We are grateful to Margarita Chamom and Dragana Jelenic for expert technical and secretarial assistance, respectively.