The a subunit, a membrane protein from the E. coli F1FO ATP synthase has been examined by Fourier analysis of hydrophobicity and of amino-acid residue variation. The amino-acid sequences of homologous subunits from Vibrio alginolyticus, Saccharomyces cerevisiae, Neurospora crassa, AspergiUus nidulans, Schizosaccharomyces pombe and Candida parapsilosis were used in the variability analysis. By Fourier analysis of sequence variation, two transmembrane helices are predicted to have one face in contact with membrane lipids, while the other spans are predicted to be more shielded from the lipids by protein. By Fourier analysis of hydrophobicity, six amphipathic t~-helical segments are predicted in extra-membrane regions, including the region from Glu-196 to Asn-214. Fourier analysis of sequence variation in the b- and the c-subunits of the Escherichia coli F1F0 ATP synthase indicates that the single transmembrane span of the b-subunit and the C-terminal span of the c subunit each have a face in contact with membrane lipids. On the basis of this analysis topographical models for the a- and c-subunits and for the F 0 complex are proposed
Biochimica et Biophysica Acta, 1140 (1992) 199-207 © 1992 Elsevier Science Publishers B.V All rights reserved 0005-2728/92/$05.00 199 BBABIO 43728 Prediction of transmembrane topology of F proteins from Escherichia coli FIF ATP synthase using variational and hydrophobic moment analyses Steven B Vik and Nguyen N Dao Department of Biological Sciences, Southern Methodist University, Dallas, TX (USA) (Received 10 April 1992) Key words: Alpha helix; a-Helical periodicity; Fourier transform; Hydropathy analysis; ATP synthase, FIF0-; Proton channel The a subunit, a membrane protein from the E coli F1FO ATP synthase has been examined by Fourier analysis of hydrophobicity and of amino-acid residue variation The amino-acid sequences of homologous subunits from Vibrio alginolyticus, Saccharomyces cerevisiae, Neurospora crassa, AspergiUus nidulans, Schizosaccharomyces pombe and Candida parapsilosis were used in the variability analysis By Fourier analysis of sequence variation, two transmembrane helices are predicted to have one face in contact with membrane lipids, while the other spans are predicted to be more shielded from the lipids by protein By Fourier analysis of hydrophobicity, six amphipathic t~-helical segments are predicted in extra-membrane regions, including the region from Glu-196 to Asn-214 Fourier analysis of sequence variation in the b- and the c-subunits of the Escherichia coli F1F0 ATP synthase indicates that the single transmembrane span of the b-subunit and the C-terminal span of the c subunit each have a face in contact with membrane lipids On the basis of this analysis topographical models for the a- and c-subunits and for the F complex are proposed Introduction The F t F A T P synthase from Escherichia coli is a prototype of the A T P synthases found in mitochondria and chloroplasts [1,2] It comprises an F~ complex, which contains the catalytic subunits and an F complex, which conducts protons across the membrane T h r e e different subunits, a, b and c form F with a stoichiometry of : : - [3] The n u m b e r and relative location of the transmembrane spans of the F subunits is still at issue Based on a variety of evidence [4,5], the c subunit is thought to contain two m e m b r a n e spanning a-helices connected by a tight turn that faces the cytoplasm The b subunit contains a single hydrophobic region that is long enough to span a m e m b r a n e as an a-helix and that is located at the extreme amino-terminus The bulk of the protein extends into the cytoplasm and is thought to interact with F subunits [6-9] The arrangement of the largest subunit, a, is less certain Models have been Correspondence to: S.B Vik, Department of Biological Sciences, Southern Methodist University, Dallas, TX 75275-0376, USA Abbreviations: TID, 3-trifluoromethyl-m-(iodophenyl)diazirine;GES, hydrophobicity scale of Engleman et al [16] offered ranging from four to eight transmembrane spans [6,10-15] based on considerations of hydrophobicity and charge and on alkaline phosphatase gene fusion experiments The low solubility of this protein has hindered efforts to learn more about its arrangement in the m e m b r a n e and within the F complex Two different models for the organization of a, b and c subunits in the F complex have been offered Based primarily upon the labeling of all three subunits by the hydrophobic reagent 3-trifluormethyl-m(iodophenyl)diazirine (TID), H o p p e and Sebald [4] have proposed an oligomer of c subunits adjacent to an a-b complex Alternatively, Cox et al [11] have proposed that an a-b complex is surrounded by a ring of c subunits, shielding them from the lipids We have applied Fourier analysis of hydrophobicity and of variation within a group of related proteins to detect a-helical periodicity within F subunits The resulting hydrophobic moments can identify transmembrane helices that have segments with a polar face They also can identify extramembrane segments that are located at a polar-nonpolar surface The variational moments can identify transmembrane helices that interact with m e m b r a n e lipids on one face and with other transmembrane helices on the opposite face 200 Similarly, they can also identify surface helices in globular proteins or extra-membrane segments of membrane proteins Materials and Methods 3.5 2.5 A "~ 1.5 ~ I).5 "%/Iso ~'i ~ "~ 0.5 2s0 -1.5 Hydropathy profiles were generated using the hydrophobicity scale (GES) of Engelman et al [16] and a window of 21 residues Potential amphipathic ahelices were detected using the GES scale and a window of 11 residues Hydrophobic moment power spectra/z(o~) [17] were calculated for stretches of 11 amino acids, according to -2.5 Residue n u m b e r 3.5 2.5 B :o 1.5 o)(tx) = [kL_l( hk hk ) cos(ko)) ]2+ [k~-~Sl(hk hk) sin( ko) ) ] ~ 0.5 ~ 0.5 -1.5 where n = 11, h k is the hydrophobicity of the k th residue and h is the mean hydrophobicity of the sequence of n residues: n n ~,k=l / The a-helical amphipathic index, ~h,(g(o~)), measures how much of the power spectrum /z(to) resides in the vicinity of 100°, i.e., 3.6 residues per turn for an a-helix, as compared to the entire spectrum ,'120 3~J9° ~(o)) do) I r180 8oJ0 /x(o)) do) Fourier analysis of the variability of amino-acid sequences was performed as described by Komiya et al [18], using a window of 11 or 21 residues The calculations are formally the same as above, except that variability, which is defined as the number of different amino-acid residues that appear at a single position among a group of aligned sequences, replaces hydrophobicity This analysis is based on the observation that surface residues are less highly conserved than are buried residues [19] In the case of a family of related transmembrane proteins, such as the reaction centers of photosynthetic bacteria, it has been found that residues in contact with the membrane lipids tend to be more variable than those in contact with protein [20] For this reason, surface transmembrane a-helices can be detected by a periodicity in the variability profile consistent with an a-helix For variability analysis, the alignment of sequences of a-subunits from E coli and Vibrio alginolyticus was that of Krumholz et al [21] and fungal a-subunits were aligned essentially according to Gu61in et al [22] Bacterial b-subunits were aligned by inspection The alignment of c-sub- i,,r , v rflllllq~llllE[]]lll¢lllfl] J ~ " II]]llllmllllllglllllllflllLtlll~J~][~ [~I~[E~]'~[[~ '!]i]Tll]![:,ii~ii -2,5 Residue n u m b e r Fig Hydropathy profiles of (A), a-subunit from Escherichia coil and (B), a-subunit (ATP 6) from Saccharomyces cerevisiae using a window of 21 residues and the GES [16] hydrophobicity parameters Underlined regions correspond to the aligned sequences shown in Fig units was according to Sebaid and Hoppe [4], with the following additional sequences included: I~ alginolyticus [21], Propionigenium modestum [23], Bacillus megaterium [24], Synechococcus [25], Anabaena [26], Marchantia chloroplast [27], sunflower mitochondria [28], rice mitochondria [29], sugar beet mitochondria [30], petunia mitochondria [31], pea mitochondria [32], tobacco mitochondria [33] and maize mitochondria [34] Secondary structure predictions were carried out using Chou-Fasman parameters [35] with a window of residues Results Hydropathy analysis of the a-subunit A hydropathy profile of the E coli a-subunit is shown in Fig 1A Five prominent hydrophobic regions with average hydrophobicity values of greater than 1.5 can be seen The first three are well-resolved, while the last two are nearly contiguous In addition, there are two minor peaks of value about 0.7, one following the first major peak and the second following the third major peak Previous authors have interpreted such results as indicating 5, or transmembrane spans For comparison, the hydropathy profile of the yeast, S cerevisiae., mitochondrial a-subunit ( A T P 6) is shown in Fig lB The peaks corresponding to the minor peaks seen in the E coli a-subunit profile are somewhat different The first one is no longer hydrophobic, while the second one is a major peak with an average 201 TABLE I Comparison of potential span IV sequences Average hydrophobicity is calculated from the GES parameters [16] Sequences were obtained from the indicated references: maize mitochondria [66] S cerevisiae mitochondria [4], Drosophila yakuba [67] bovine mitochondria [68], E coli [10], pea chloroplast [69] and B megaterium [25] Source Sequence FFSFLLPAGVPLPLAPFLVLL S cerecisiae FFSLFVPAGTPLPLVPLLVII Drosophila yakuba MFAHLVPQGTPAILMPFMVCI Bovine mitochondria SLAHFLPQGTPTPLIPMLVII E coli FTKELTLQPFNHWAFIPVNLI Pea chloroplast GLAYFGKYIQPTPILLPINIL B megaterium FSAYTKDYFKPMAFLFPLKII - 2.08 2.03 1.63 1.33 0.72 0.59 0.11 10 20 28 29 I D L S C L N L T T F S L Y T I I V L L V I T S L Y T L 38FWT I N I D S M F F S V V L G L L F L V LFRSVAK Span n S.cerevisiae E.coli 9OKNWGLYFPM %GKS KLIAPLA Span I n l S.cerevisiaellSIPYSFALSAHLVFI E.coli 147DV N V TLSMA Average hydrophobicity Maize mitochondria Span I S.cerevisiae E.coli 1o I FTLFMFIFI LT I FVWVFLMN 20 28 A N L I S M I P Y LMDLLP I 10 20 V IWLGNTI S I KMKG L GV I SLSI F I L I LFY Span IV 10 S.cerevisiaeI48GWVFFSLFVPAGTPLPLVPLLVI E.coli 172GG F T KEL T L Q P FNHWA S [a~ll V S.cerevisia~ E.coli Iff?LG S N I LAGH 2IlL F G NMYAGE 28 L IGG 20 I ETL F P V N L L E GV 10 L LMV l L AGLT L F L I AGLL 20 F N FML PWWSQWI 28 S S 28 I NL F LN Sp,m Vl 10 20 S.cerevisiae2251 L A M I L E F A I G l I QSYVWT LTASYLK E.coli Z3SNV PWA I FH I L I l TLQA F F MV L T VYL 28 S Fig Predicted transmembrane-spanning regions of the S cerevisiae a-subunit and corresponding regions in the E coli a-subunit Identical residues are shown in boldface hydrophobicity per residue of 2.0 In addition, the second and third major peaks are contiguous, due to a shorter polar region T h e six predicted t r a n s m e m b r a n e spans in the yeast protein are shown in Fig In each case, spans of 28 residues are shown, although it is expected that the actual lengths may vary from 22 to 30 residues C o r r e s p o n d i n g sequences from the E coli protein are also shown, with four to eight identical residues f o u n d between each pair O n e of the least similar pairs, that o f span IV, also contains the E coli span with the lowest hydrophobicity In Table I, this region of the a-subunit from several diverse organisms is c o m p a r e d A p a t t e r n of low sequence similarity can be seen t h r o u g h o u t the series, but the average hydrophobicity per residue varies from 2.08 to - 1 Therefore, based on h y d r o p a t h y data alone, the identification of this region as a t r a n s m e m b r a n e span is somewhat uncertain, assuming that all a-subunits have the same n u m b e r of t r a n s m e m b r a n e spans A m p h i p a t h i c a-helices in the a-subunit T h e Fourier analyses of the hydrophobicities o f the E coli and yeast a-subunits are presented in Fig T h e amphipathic index ~ h , ( / z ( w ) ) is plotted as a function of the amino-acid residue number A value of > 2.0 is indicative that an a-helix is amphipathic [36] Segments with potential hydrophobic m o m e n t s within the predicted t r a n s m e m b r a n e spans are listed in Table II Most o f the t r a n s m e m b r a n e spans contain regions with hydrophobic m o m e n t s and there is considerable similarity between the two proteins In each case, there are no hydrophobic m o m e n t s found in span III In each pair of spans II, IV, V and VI, there are hydrophobic m o m e n t s at similar locations In general, the polar faces of these predicted helices contain several polar but u n c h a r g e d residues M a n y potential amphipathic a-helices are f o u n d in the the e x t r a - m e m b r a n e regions and these are listed in Table III A m o n g these segments, there is m u c h less TABLE II Transmembrane segments of the a subunit with high hydrophobic moment Segments that appear to be homologous are listed on the same line The numbers in parentheses indicate the positions of the residues within the transmembrane spans ~ is the a-helical amphipathic index as defined in Materials and Methods Polar face residues are listed in the order in which they appear in the sequence E coli S cerevisiae Span I (8-19) residues 45- 56 ~ 2.31 polar face SSG II (4-14) II (16-28) IV (9-19) V (10-26) 99-109 111-123 180-190 220-236 1.71 2.52 1.99 2.64 KPTL AVND QNHF APSQ VI (15-27) 252-264 2.92 QFTY span residues ~ polar face I (18-31) II (4-15) II (16-31) IV (4-18) V (9-19) VI (6-16) VI (15-26) 46- 59 93-104 105-120 151-165 195-205 230-240 240-251 2.41 3.58 3.57 2.33 2.05 1.97 2.14 TTSN LPTM IANS SPTP HMAT EGQ QSWTT 202 A ~u 5(:I 100 150 200 250 Residue number ~2 50 100 150 200 250 Residue number Fig Amphipathic a-helical index plots for (A), the E cell a-subunit and (B), the S cerevisiae a-subunit using a window of 11 residues Peaks indicate segments with hydrophobic moments, if a-helical correspondence between the E coli and yeast proteins, consistent with the more limited sequence similarity in these regions One exception is the region between spans IV and V, EGVSLLSKPVSLGLRLFGN, (E coli resdiues 196-214) This is a highly conserved region among all a-subunits and contains an arginine residue (Arg-210) shown by mutagenesis to be essential for proton translocation [37,38] Near the amino-terminus of the E coli a-subunit is a segment of 13 residues, QDYIGHHLNNLQL (9-21), predicted to be an amphipathic a-helix In the region between spans I and II, the E coli a-subunit has three segments, LFRSVAKKATSG (59-70), QTAIELVIGF (75-84) and KDMYHG (91-96) with predicted amphi- pathic a-helices, separated by two probable turns: VPGK (70-73) and VNGSV (86-90) Between spans II and III is a segment of 16 residues predicted to form an amphipathic a-helix: LPIDLLPYIAEHVLGL (121-136) In addition, a second segment, which would be contiguous with the transmembrane span, could also form an amphipathic helix, PALRVVPSADV (137147), with arginine and aspartic acid on the polar face The segment connecting span III and IV, MKGIGGFTKE, while amphipathic if projected as an a-helix, contains a region GIGG likely to be a turn Variational moment analysis of a-subunits Fourier analysis of sequence variation is plotted in Fig In Fig 4A, five fungal sequences (Refs 22, 40-42 and Lang, B.F., personal communication) have been analyzed, in which the pairwise sequence identities are about 40-70% The alignment of these sequences is shown in Fig Two of the predicted transmembrane spans, II and VI, are associated with peaks of value greater than 2.0, indicative of an a-helix with a single face in contact with membrane lipids In each span, the most polar residues are conserved: N-110, E-233 and Q-240 The other four transmembrane spans are associated with regions having values of 1.0 or less Similar results, shown in Fig 4B were obtained from a variability analysis of the a-subunit from E coli and V alginolyticus, the only a-subunit known with greater than 40% identity with the E coli protein [23] Again, spans II and VI have values greater than 2.0, indicating a face in contact with membrane lipids In the regions of spans I, III, IV and V and at the amino-terminus there are a large number of peaks against a broad background To resolve these peaks, the calculations were redone with a window of 11 residues This allows detection of a-helical periodicity that is exhibited by segments shorter than a transmembrane helix These results are presented in Fig 4C TABLE III Extramembrane segments in the a subunit with high hydrophobic moment Loop I-II refers to the extramembrane segment connecting the putative spans I and II ~ is the a-helical amphipathic index as defined in Materials and Methods have is the average hydrophobicity per residue using the GES parameters [16] E coli S cerevisiae Region residues ~ h ave region N-terminus Loop I-II Loop I-If Loop I-II Loop II-III Loop II-III Loop III-IV Loop IV-V - 21 59- 73 75- 88 86- 96 121-136 135-147 168-179 191-215 2.68 2.36 2.51 1.96 2.96 2.27 2.51 2.82 - 1.60 - 0.72 0.38 - 1.39 0.24 - 0.28 - 0.62 0.13 N-terminus Loop I-II Loop III-IV Loop IV-V Loop V-VI residues ~ h ave - 17 73- 86 2.89 3.08 - 1.22 - 1.27 136-148 166-190 208-225 2.08 3.48 2.82 0.02 0.13 2.22 203 Using a window of 11 resdiues, only spans II and VI have broad peaks with values greater than 2.0 In other regions, several narrow peaks can be identified, which correspond to segments of approx 11 or fewer residues with a-helical periodicity in sequence variation The previously identified amphipathic helix at the aminoterminus, Q D Y I G H H L N N L Q (9-20) exhibits a sharp peak in the variability plot, indicating a conserved helical face, shown in projection in Fig 6A Examination of the sequence reveals that the conserved face contains most of the nonpolar residues, including the two histidines, while most of the polar residues vary The carboxy-terminal half of span I, L G L L F V L F R S V (52-63), shows a very high variational moment (3.92), in which the two polar residues arginine and serine vary The region connecting spans III and IV, I K M K G I G G F T K (166-177) has a high variational moment that corresponds closely to the region (168-179) previously shown to posess a hydrophobic moment The carboxy-terminal half of span IV, A F I P V N L I L E G (187-197), also has a very high variational moment (3.61), but this region does not contain a significant hydrophobic moment Finally, a region (200-211) with a high hydrophobic moment containing the essential ' A 50 1O0 150 200 250 Residue n u m b e r Arg-210, thought to connect spans IV and V, also has a significant variational moment (3.17), shown in projection in Fig 6B Analysis o f b- and c-subunits Among bacterial F1F0 A T P synthases, some are thought to contain two identical b-subunits [10,2325,43,44], while others have one each of two similar subunits b and b' [26,27,45] Each b- or b'-subunit is predicted to contain a single transmembrane spanning region near its amino-terminus For variability analysis, the sequences of six different b-subunits, from bacteria with a single type of b-subunit, were used as shown in Fig The ~ value for this segment of twenty one residues is 2.80, indicating an a-helix with a face in contact with membrane lipids Examination of the sequences via helical wheel projection revealed that the conserved face contained two polar residues, Gin-10 and Lys-23 The sequences of 24 c-subunits were examined for variational moments and the results are presented in Fig The two anticipated transmembrane helices show quite different results At the carboxy-terminus, including the conserved acidic residue (Asp-61 in E coli), is a broad region with values of 2-3, indicating an a-helical face in contact with membrane lipids The maxima centered near residues 55 and 63 might indicate two a-helical regions, separated by a proline or glycine in most species [4] However, at the aminoterminus, no a-helical segments can be found with a conserved face The most highly conserved residues correspond to positions 14, 20, 23, 25, 27 and 29, which are always alanine or glycine in the 24 cases examined here, with a single exception These positions would appear on opposite faces in an a-helical projection, as shown in Fig 9A Discussion 50 100 150 200 250 Residue n u m b e r • 50 100 150 200 250 Residue number Fig a-helical index plots of variability for (A) the S cerevisiae a-subunit and (B) the E coil a subunit using a window of 21 residues In (C) the E coli a subunit is analyzed using a window of 11 residues Peaks indicate segments that would exhibit a conserved face if a-helical, i.e., a variational moment The use of variational analysis has previously been applied to reaction center proteins from Rhodopseudomonas viridis and Rhodobacter pseudomonas [18,20] and results were confirmed by the high resolution structure determined for the R viridis proteins [46] We have applied a similar analysis to a pair of Gramnegative bacterial a-subunits and to a group of homologous proteins from fungal mitochondria for which high-resolution structural data are unavailable In each case, two segments of 20-30 amino acids of low overall polarity were found to display sequence variation with a period of about 3.6 residues, consistent with the face of an a-helix This suggests that these segments form transmembrane a-helices that interact on one face with membrane lipids Similar analysis of the other two proteins that make up the bacterial F0 complex, the b- and c-subunits, 204 S.cerevisiae A.nidulans I FGL Q S S F I D L S C L N L T T F S L Y T I IVLLVIT FSLNANVLGNI HLS I TNIGLYLS IGLLLTLGYH LSI DTLGNLH I S I TN IGFYLT IGAFFFLV IN FGF Y L F N Y H F D F S N F G F Y L G L S A L IAI SLA LLI TDNLTFSITNYTLYLI IVSLI I I FYS 3 5 , 5 3 3 4 5 [! M F N L L N T Y I T S PLDQFEIRTL MY Q F N F I L S PLDQFEIRDL M F N I L S PLNQFEIRDL M F Z T S PLEQFELNNY MF ¥ S PLDQFELKPL 1131112342 N.crassa S.pombe Cparapsilosis Variation S.~re A.nid Ncr~sa S.~m~ 52SLYTLT LLAA LLSE IINLTPYGSG SII 3255555544 NNN NK I I GSRWLI N NKI IPNNWSI NY NRLVSNSWS AKIVPQKFGI RH NYLGSSRWGV 2323334232 III 101TLFMFIFIANLISMIPYSFA A L F I F I L V N N L I G M V P Y S F A T L F I F I L I N N L I G M V P Y S F A S L F V L I L F S N L L R L I P Y G Y A T I F N F I L I A N L I S M I P Y S F A 2 3 1 2 1 2 ~p~8 VarJ~ion S.~re An~ ~cr~$8 S.~m~ C~a Vari~ion 154LFVPAGTPLP LFVPSGCPLG LLVPAGCPLA LFLPSGTPTP LFVPSGTPLA 1221212123 S.ce~ A.n~ Ncr~sa S.~m~ C.p~a Vari~ion S.~re A.nid Ncr~sa S.~m~ 207NFMLINLFTL NIMTSGILFF NIMTSGIIFF TFMGLNLITF SLMSSSFLGF 3314333332 ~a Vari~ioN SQEAI YDT I MNMTKGQ I GG KNWGLYFPM I F SQEAI YATVHS IVINQLNP TKGQLYFPFI Y I SQESLYATI YS IVTSQINP RNGQ I YFPFI¥ AMEAI YFTMLNLVENQ I HSSKTVSGQSYFPF IW S V I A I Y D T I L N L V N G P I G R KGGY YFPL I F 2 3 2 5 4 1 3 IV LSAHLVF I I SLS I V I W L G N T I L G L Y K H G W V F F S ST6HF ILTFSMSFT I V L G A T F L G L Q R H G L K F F S STSH FVVTFALSFT IVLGAT ILGFQKHGLE FFS TTAQL I FTLGLS I S IL IGAT ILGLQQHKAKVFG I SAQLVAVVS F S L T L W I G N V V L G L Y L H G W G F F A 2 2 4 3 3 2 1 2 4 V LVPLLVI IETL SY IARAI S L G L R L G S N I L A G H L L M V ILAGLTF LLPLLVLIEFI S Y L S R N V S L G L R L A A N ILSGHMLLS ILSGFTY LLPLLVLIEFI S Y L A R N I S L G L R L A A N ILSGHMLLH I LAGFTY SY IARGLSLG I RLGANI IAGHLTMS ILGGL I F LIPLLVLIEFV SYASRAI SLGLRLGANI LSGHLLML ILGSL I I LVPVLVLIEAL 1 3 1 1 2 1 2 1 2 1 2 13121121133 VI A I M I L E F A I G I I Q S Y V W T I L T A S Y L K D T L Y L H VFGFVPLAMIL FLGLIPLAFI I A F S G L E L A I A F I Q A Q V F V V L T C S Y I K D G L D L H FLGLIPLAFI I A F S G L E L G I A F I Q A Q V F V V L T S G Y I K D A L D L H A I S L L E F G I A F I Q A Y V F A I L T C G F I N D S L N L H I IGFLP ITVLV A I T I L E F G I A I I Q A Y V F S I L L S G Y I K D S V E L H VSGI IP ILAVV 3 1 2 2 1 2 2 2 2 4 1 34133123433 Fig Amino-acid sequence alignment of five fungal a subunits The first aligned serine residue has been shown to be the first residue in the processed protein of S cerevisiae and C parapsilosis [22] The predicted transmembrane spans (Fig 1) are designated by a line above the S cerevisiae sequence An alternate alignment of residues 25-52 (S cerevisiae) did not appreciably affect the variational analysis B 3: L conserv_ L ~ H R R P ~ P 0.s 0 Fig a-Helical projections of (A), residues 9-21 and (B), residues 198-215 of the E coil a-subunit The conserved residues indicated are those conserved in E coli and V alginolyticus and in the case of (B) are also generally conserved among the five fungal subunits E coli V alginolyticus PS3 B megaterium B firmus OF4 P modestum 6 21 18 14 TILGQAIAFVLFVLFCMKYVW TLLGQAISFALFVWFCMKYVW TIIYQLLMFIILLALLRKFAW DILFQLVMFLILLALLQKFAF SALYQLLAFSVLLFFLSKFAL NMFWQIINFLILMFFFKKYFQ Fig Alignment of five bacterial b-subunit sequences for variability analysis The numbers indicate the positions of the first residues shown in each case Sequences used are from Escherichia coli [10], Pibrio alginolyticus [23], thermophilic bacterium PS3144], Bacillus rnegaterium [25], Bacillus firrnusOF4 [45] and Propionigenium modesturn [24] 10 20 30 40 50 60 70 80 Residuenumber Fig a-Helical index plots of variability for 24 c-subunits using a window of 21 residues Numbering is that of the E coli protein A Fig a-Helical projection of the amino-terminal residues of the E coli c-subunit (10-27) The inner lines mark the two highly conserved regions Residues marked (*) are those labeled by the hydrepbobic reagent TID [48] 205 showed that these proteins also exhibit a-helical sequence variation in regions that have been identified as transmembranous These results suggest that the two b-subunits, each with a single transmembrane span, are on the periphery of the F complex and have extensive contact with lipids Hydrophobic labeling studies with TID support this location of the b-subunits Hoppe et al [47] found that all of the amino-terminal residues of the b-subunit (1-35) could be labeled to some extent, but that residues Gln-10 and Lys-23 were especially resistant As shown here, these are conserved residues that could be on a conserved face of an a-helix involved in protein-protein interactions The 9-10 c-subunits are thought to have two transmembrane regions separated by a highly conserved turn Variational analysis indicates that the carboxyterminal region would form an a-helix with a conserved face, or possibly two a-helical segments flanking the conserved acidic residue, corresponding to the two peaks in Fig The amino-terminal region does not exhibit a-helical variation, but would have two conserved faces due primarily to six highly conserved glycine or alanine residues if it were a-helical Other evidence points to an a-helical amino-terminus The length of the hydrophobic segment in E coli, 23 consecutive amino-acid residues without a hydrogen-bonding side-chain, is consistent with a transmembrane ahelix, but is too long for an extended structure like a /3-strand Labeling by the lipophilic reagent TID indicates that both hydrophobic regions of the c-subunit are exposed to the lipid phase [47] and the pattern of labeled residues at the amino-terminus is consistent with the face of an a-helix, with one exception Tyr-10 (see Fig 9) Therefore, we propose a model of the c-subunit oligomer in which the amino-terminal a-helix contacts two carboxy-terminal helices on opposite faces, while the carboxy-terminal helix contacts two aminoterminal helices on a single surface In this model, residues Tyr-10, Ala-24 and Asp-61 all face the outside of the oligomer, consistent with TID labeling [47] and mutagenesis [48,49] experiments The Fourier analysis of the sequence variation of a-, b- and c-subunits used different groups of homologous proteins The sequences of c-subunits from a highly diverse group of organisms was used, while the less conserved a- and b-subunits required more closely related sources It was determined empirically that sequences with a pairwise identity of 40-70% were optimal for the detection of a-helical variation This is reasonable, since a single conserved residue per turn of an a-helix (1/3.6), on average, will yield 28% sequence identity Likewise, a single variable residue per turn will yield 72% overall sequence identity Since only one other a-subunit was suitably similar to the E coli protein, a second group of five fungal proteins was also analyzed The methods of locating transmembrane a-helices by computation of average hydrophobicity per residue have been proposed and analyzed by numerous authors [16,50-52] The recently published structure of bacteriorhodopsin [53] confirms that seven of the eight predicted helices of Engelman et al [16] lie within the determined transmembrane regions, but that one is slightly shifted The 27-residue F-helix has an average hydrophobicity of only 0.695 (GES scale), considerably less than a typical transmembrane span This seems to confirm that transmembrane helices, like secondary structure in general, are subject to some nonlocal influences and simple predictive schemes are not perfect In the a-subunit, five transmembrane helices can be predicted with confidence from all organisms examined, but in some cases a sixth span (span IV) must be considered somewhat uncertain (see Table I) In general, it seems likely that the number of transmembrane spans must be conserved among a group of homologous proteins In this case the region in question separates two conserved regions, span II and spans V and VI, suggesting that the number of spans in the intervening region must be the same or differ by an even number The span IV regions of a-subunits from Bacillus species are significantly more polar than the bacteriorhodopsin F-helix, bringing into question its location in the membrane Alternatively, since no complete amino-acid sequence of an a-subunit has ever been reported, the possibility of post-transcriptional or translational modification cannot be ruled out Other proposals for the membrane topology of the E coli a-subunit have been made on the basis of alkaline phosphatase gene fusion experiments In two different studies, largely similar results have been obtained, but have been interpreted in different ways Lewis et al [13] proposed four transmembrane spans similar to the first four here, but also proposed four shorter spans at the carboxy-terminus Bj~rb~ek et al [14] proposed eight spans, six similar to the six de- e oligomer Fig 10 Topological model of the a, b and c-subunits of the E coli F o complex The bold lines connecting spans I and II, III and IV and V and VI represent extramembrane loops on one side of the membrane, while the thin lines represent those on the opposite side In an alternative model, where span IV is omitted, span II1 and one b-subunit would slide closer to span V and the last two spans would have reversed orientation 206 scribed here, but two additional spans, one between I and II and one between II and III The alkaline phosphatase gene fusion technique [54] has been shown to be reliable when tested on proteins with known topology [55], but ambiguities in interpretation sometimes occur [56] Our results show that the two extra spans proposed by Bj~rbaek et al [14] would have segments with a-helical hydrophobic moments, but have low overall hydrophobicity The four spans at the carboxy-terminus predicted by Lewis et al [13] consist of ten to thirteen residues each, which is conceivable if they span the membrane in an extended conformation, e.g., as /3-strands, or if they are a-helical but not extend to the membrane surface, as discussed by Lodish [57] The variational analysis done here indicates that most of those segments should be a-helical and located at the surface of the protein Therefore, it is difficult to reconcile these two sets of results Amphipathic helices have been proposed to be components of ion channels, such as in the acetylcholine receptor [58], but the a-subunit may not conform to that type of ion channel Recent work [53] has shown that the proton channel in bacteriorhodopsin is formed primarily from three of its seven transmembrane helices, but only half of the residues in the channel have polar character (13 of 26) Furthermore, another 25 residues with polar character are found in the transmembrane region outside of the proton channel Therefore, the detection of amphipathic a-helices within membrane spanning regions could identify either ion channel pathways or helix-helix contacts Polar residues involved in helix-helix interactions might be important in stabilizing the tertiary structure of membrane proteins The a-subunit is likely to be more similar to bacteriorhodopsin, since it constitutes at least part of a proton channel The role of retinal as transducer might be played by the c-subunits In the E coli a-subunit, amphipathic helical segments were detected in all of the putative transmembrane spans except III In both the E coli and yeast proteins, span II is predicted to be the most amphipathic helix and spans V and VI contain glutamyl and histidyl residues thought to be essential for proton translocation [59-61] Recent mutagenesis [62] of the E coli a-subunit suggested that residues Asp-124 and Arg-140 are part of the proton channel Therefore, in analogy with the proton channel of bacteriorhodopsin, we propose that these three spans provide most of the residues in the proton channel Other amphipathic segments, such as those found in spans I and IV might be involved in protein-protein interactions On the basis of these results, a model showing the relative location of transmembrane segments is presented in Fig 10 Because of short connecting loops, span V must be close to span VI and span III must be close to span IV (assuming six transmembrane spans) Span V is close to the c-subunits because of the importance of Glu-219 and Arg-210 (in the loop) in proton translocation Spans II and VI are adjacent to span V to form the channel and are exposed to lipid Both b-subunits are also exposed to lipid One b-subunit is located next to span VI because Pro-240 mutations in the a-subunit can partially suppress the effects of Gly-9 -, Asp mutation in the b-subunit [63,64] Of the remaining two transmembrane spans, I is placed more peripheral to the complex than is III on the basis of its hydrophobic moment In the model, span III has two opposite faces in contact with protein and two in contact with lipid, which could account for the lack of both hydrophobic and sequence variational moments The highly conserved region between spans IV and V is proposed to lie parallel to the membrane, near the c-subunit surface This model is intended to be a guide for site-directed mutagenesis and for site-specific chemical labeling Such experiments are currently underway Acknowledgements We thank B Franz Lang for his unpublished DNA sequence of S pombe ATP This work was supported by the Southern Methodist University Research Council and by USPHS Grant GM-40508 References Senior, A.E (1990) Annu Rev Biophys Biophys Chem 19, 7-41 Futai, M., Noumi, T and Maeda, M (1989) Annu Rev Biochem 58, 111-136 Foster, D.L and Fillingame, R.H (1982) J Biol Chem 257, 2009-2015 Hoppe, J and Sebald, W (1984) Biochim Biophys Acta 768, 1-27 Schneider, E and Altendorf, K (1987) Microbiol Rev 51,477497 Senior, A.E (1983) Biochim Biophys Acta 726, 81-95 Hoppe, J., Friedl, P., Schairer, H.U., Sebald, W., Von Meyenburg, K and J~rgensen, B.B (1983) EMBO J 2, 105-110 Perlin, D.S., Cox, D.N and Senior, A.E, (1983) J Biol Chem 258, 9793-9800 Hermolin, J., Gallant, J and Fillingame, R.H (1983) J Biol Chem, 258, 14550-14555 10 Walker, J.E., Saraste, M and Gay, N.J (1984) Biochim Biophys Acta 768, 164-200 11 Cox, G.B., Fimmel, A.L., Gibson, F and Hatch, L (1986) Biochim Biophys Acta 849, 62-69 12 Hennig, J and Herrmann, R.G (1986) Mol Gen Genet 203, 117-128 13 Vik, S.B., Cain, B.D., Chun, K.T and Simoni, R.D (1988) J Biol Chem 263, 6599-6605 14 Lewis, M.L., Chang, J.A and Simoni, R.D (1990) J Biol Chem 265, 10541-10550 15 Bj~rb~ek, C., F6rsom, V and Michelsen, O (1990) FEBS Lett 260, 31-34 16 Engelman, D.M., Steitz, T.A and Goldman, A (1986) Annu Rev Biophys Biophys Chem 15, 321-353 207 17 Eisenberg, D., Weiss, R.M and Terwilliger, T.C (1984) Proc Natl Acad Sci USA 81, 140-144 18 Komiya, H., Yeates, T.O., Rees, D.C., Allen, J.P and Feher, G (1988) Proc Natl Acad Sci USA 85, 9012-9016 19 Smith, E.L (1968) Harvey Lect 62, 231-256 20 Rees, D.C., Komiya, H and Yeates, T.O (1989) Annu Rev Biochem 58, 607-633 21 Krumholz, L.R., Esser, U and Simoni, R.D (1990) J Bacteriol 172, 6809-6817 22 Gu61in, E., Gu6rin, M and Velours, J (1991) Eur J Biochem 197, 105-111 23 Krumholz, L.R., Esser, U and Simoni, R.D (1989) Nucleic Acids Res 17, 7993-7994 24 Esser, U., Krumholz, L.R and Simoni, R.D (1990) Nucleic Acids Res 18, 5887 25 Brusilow, W,S.A., Scarpetta, M.A., Hawthorne, C.A and Clark, W.P (1989) J Biol Chem 264, 1528-1533 26 Cozens, A.L and Walker, J.E (1987) EMBO J 5, 217-222 27 McCarn, D.F., Whitaker, R.A., Alam, J., Vrba, J.M and Curtis, S.E (1988) J Bacteriol 170, 3448-3458 28 Ohyama, K., Fukuzawa, H., Kohchi, T., Sano, T., Sano, S., Shirai, H., Umesono, K., Shiki, Y., Takeuchi, M., Chang, Z., Aota, S Inokuchi, H and Ozeki, H (1988) J Mol Biol 203, 281-298 29 Recipon, H (1990) Nucleic Acids Res 18, 1644 30 Kaleikau, E.K., Andr6 C.P and Walbot, V (1990) Nucleic Acids Res 18, 370 31 Xue, Y., Thomas, C.M and Davies, D.R (1989) Nucleic Acids Res 17, 8857 32 Young, E.G., Hanson, M.R and Dierks, P.M (1986) Nucleic Acids Res 14, 7995-8006 33 Morikami, A and Nakamura, K (1987) Nucleic Acids Res 15, 4692 34 Kazama, S., Suzuki, T., Kadowaki, K and Akihama, T (1990) Nucleic Acids Res 18, 5879 35 Dewey, R.E., Schuster, A.M., Levings III, C.S and Timothy, D.H., (1985) Proc Natl Acad Sci USA 82, 1015-1019 36 Chou, P.Y and Fasman, G.D (1974) Biochemistry 13, 222-245 37 Cornette, J.L., Cease, K.B., Margalit, H., Spouge, J.L., Berzofsky, J.A and DeLisi, C (1987) J Mol Biol 195, 659-685 38 Lightowlers, R.N., Howitt, S,M., Hatch, L., Gibson, F and Cox, G.B (1987) Biochim Biophys Acta 894, 399-406 39 Cain, B.D and Simoni, R.D (1989) J Biol Chem 264, 3292-3300 40 Macino, G.A and Tzagoloff, A (1980) Cell 20: 507-517 41 Grisi, E., Brown, T.A., Waring, R.B., Scazzocchio, C and Davies, R.W (1982) Nucleic Acids Res 10, 3531-3539 42 Morelli, G and Macino, G (1984) J Mol Biol 178, 491-507 43 Ohta, S,, Yohda, M., Ishizuka, M., Hirata, H., Hamamoto, T., Otawara-Hamamoto, Y., Matsuda, K and Kagawa, Y (1988) Biochim Biophys Acta 933, 141-155 44 Ivey, D.M and Krulwich, T.A (1991) Mol Gen Genet 229, 292-300 45 Falk, G and Walker, J.E (1988) Biochem J 254, 109-122 46 Deisenhofer, J., Epp, O., Miki, K., Huber, R and Michel, H (1985) Nature 318, 618-624 47 Hoppe, J., Brunner, J and Jcrgensen, B (1984) Biochemistry 23, 5610-5616 48 Miller, M.J., OIdenburg, M and Fillingame, R.H (1990) Proc Natl Acad Sci USA 87, 4900-4904 49 Fillingame, R.H., Oldenburg, M and Fraga, D (1991) J, Biol Chem 266, 20934-20939 50 Kyte, J and Doolittle, R.F (1982) J Mol Biol 157, 105-132 51 Fasman, G.D and Gilbert, W.A (1990) Trends Biochem Sci 15, 89-92 52 Popot, J and De Vitry, C (1990) Annu Rev Biophys Biophys Chem 19, 369-403 53 Henderson, R., Baldwin, J.M., Ceska, T.A., Zemlin, F., Beckmann, E and Downing, K.H (1990) J Mol Biol 213, 899-929 54 Manoil, C and Beckwith, J (1986) Science 233, 1403-1408 55 Yun, C., Van Doren, S.R., Crofts, A.R and Gennis, R.B (1991) J Biol Chem 266, 10967-10973 56 Calamia, J and Manoil, C (1990) Proc Natl Acad Sci USA 87, 4937-4941 57 Lodish, H.F (1988) Trends Biochem Sci 13, 332-334 58 Leonard, R.J., Labarca, C.G, Charnet, P., Davidson, N and Lester, H.A (1988) Science 242, 1578-1581 59 Howitt, S.M., Lightowlers, R.N., Gibson, F and Cox, G.B (1990) Biochim Biophys Acta 1015, 264-268 60 Cain, B.D and Simoni, R.D (1986) J Biol Chem 261, 1004310050 61 Cain, B.D and Simoni, R.D, (1988) J Biol Chem 263, 6606-6612 62 Lightowlers, R.N., Howitt, S.M., Hatch, L., Gibson, F and Cox, G (1988) Biochim Biophys Acta 933, 241-248 63 Kumamoto, C.A and Simoni, R.D (1986) J Biol Chem, 261, 10037-10042 64 Vik, S.B and Simoni, R.D (1987) J Biol Chem 262, 8340-8346 65 Dewey, R.E., Levings, C.S., III and Timothy, D.H (1985) Plant Physiol 79, 914-919 66 Clary, D.O and Wolstenholme, D.R (1985) J Mol Evol 22, 252-271 67 Anderson, S., De Bruijn, M.H.L., Coulson, A.R., Eperon, I.G., Sanger, F and Young, I.G (1982) J Mol Biol 156, 683-717 68 Cozens, A.L., Walker, J.E., Philips, A.L., Huttly, A.K., and Gray, J.C (1986) EMBO J 5, 217-222 ... Secondary structure predictions were carried out using Chou-Fasman parameters [35] with a window of residues Results Hydropathy analysis of the a-subunit A hydropathy profile of the E coli a-subunit... 40-70% The alignment of these sequences is shown in Fig Two of the predicted transmembrane spans, II and VI, are associated with peaks of value greater than 2.0, indicative of an a-helix with a... a-Helical index plots of variability for 24 c-subunits using a window of 21 residues Numbering is that of the E coli protein A Fig a-Helical projection of the amino-terminal residues of the E coli c-subunit