Báo cáo Y học: Modeling the three-dimensional structure of H+-ATPase of Neurospora crassa Proposal for a proton pathway from the analysis of internal cavities pptx

Modeling the three-dimensional structure of H+-ATPaseProposal for a proton pathway from the analysis of internal cavities Olivier Radresa1, Koji Ogata2, Shoshana Wodak2, Jean-Marie Ruyss

Modeling the three-dimensional structure of H+-ATPase

Proposal for a proton pathway from the analysis of internal cavities

Olivier Radresa1, Koji Ogata2, Shoshana Wodak2, Jean-Marie Ruysschaert1and Erik Goormaghtigh1

1 Service de Structure et Fonction des Membranes Biologiques, Universite´ Libre de Bruxelles, Bruxelles, Belgium; 2 Unite´ de Conformation des Macromole´cules Biologiques, Universite´ Libre de Bruxelles, Bruxelles, Belgium

Homology modeling in combination with transmembrane

topology predictions are used to build the atomic model of

Neurospora crassaplasma membrane H+-ATPase, using as

template the 2.6 A˚ crystal structure of rabbit sarcoplasmic

reticulum Ca2+-ATPase [Toyoshima, C., Nakasako, M.,

Nomura, H & Ogawa, H (2000) Nature 405, 647–655]

Comparison of the two calcium-binding sites in the crystal

structure of Ca2+-ATPase with the equivalent region in the

H+-ATPase model shows that the latter is devoid of most of

the negatively charged groups required to bind the cations,

suggesting a different role for this region Using the built

model, a pathway for proton transport is then proposed

from computed locations of internal polar cavities, large

enough to contain at least one water molecule As a control,

the same approach is applied to the high-resolution crystal

structure of halorhodopsin and the proton pump

bacterio-rhodopsin This revealed a striking correspondence between

the positions of internal polar cavities, those of crystallo-graphic water molecules and, in the case of bacteriorho-dopsin, the residues mediating proton translocation In our

H+-ATPase model, most of these cavities are in contact with residues previously shown to affect coupling of proton translocation to ATP hydrolysis A string of six polar cavi-ties identified in the cytoplasmic domain, the most accurate part of the model, suggests a proton entry path starting close

to the phosphorylation site Strikingly, members of the halo-acid dehalogenase superfamily, which are close structural homologs of this domain but do not share the same function, display only one polar cavity in the vicinity of the conserved catalytic Asp residue

Keywords: neurospora; P-ATPase; homology model; cavity;

H+

The 3D structures have been determined for only a limited

number of membrane proteins Growing large, well

ordered, 2D or 3D crystals of membrane proteins remains

indeed a major limiting step for X-ray or electron

crystal-lography Alternative approaches for obtaining structural

information are therefore very useful One such approach is

the homology modeling technique whereby a known 3D

structure of a related protein is used as a template for

building an atomic model from the amino acid sequence of

the target protein Although validity of the resulting model

requires experimental confirmation, it can provide useful

insights into the structure–function relationship in related

enzymes

The plasma membrane H+-ATPase of the fungi

Neuro-spora crassa(referenced hereafter as PMA1_NEUCR) is a

member of the large and ubiquitous P-type ATPase family This family currently counts almost 200 members involved

in the transport of a variety of ionic substrates including charged amino phospholipids [1] PMA1_NEUCR compri-ses a cytoplasmic catalytic site responsible for MgATP hydrolysis that is anchored in the membrane by 10 transmembrane segments As other P-type ATPases, PMA1_NEUCR is fully active as a monomeric 100 kDa polypeptide chain, it transports ions outside the cell in an electrogenic way using energy from MgATP hydrolysis and its catalytic cycle is characterized by the formation

of a covalent enzyme-aspartyl phosphate intermediate [2,3,42,43]

The 3D structures of PMA1_NEUCR and of another P-type ATPase, the Ca2+-ATPase of rabbit sarcoplasmic reticulum (referenced hereafter as ATC1_RABIT), have been determined at 8 A˚ resolution [4,5] At this resolution, the electron density map is accurate enough to depict the packing and tilt angle of each of the transmembrane segments Comparison of the two low-resolution models, which are believed to represent different conformational states, revealed that they displayed strikingly similar pack-ing of their respective 10 transmembrane segments whereas their cytoplasmic domains appeared too different to allow direct superimposition [6,7]

Recently, the resolution of the structure of ATC1_ RABIT was increased to 2.6 A˚ providing the first structure

at near atomic resolution for a P-type ATPase [8] This latter

Correspondence to E Goormaghtigh, Universite´ Libre de Bruxelles,

Campus Plaine CP 206/2, B 1050, Bruxelles, Belgium.

Fax: +32 26505382, Tel.: +32 26505386,

E-mail: egoor@ulb.ac.be

Abbreviations: HAD, haloacid dehalogenase; TM, M,

trans-membrane segment; PSP, phosphoserine phosphatase.

Enzymes: PMA1_NEUCR, Neurospora crassa plasma-membrane

H+-ATPase; ATC1_RABIT, Oryctolagus cuniculus (rabbit) Ca2+

-ATPase of sarcoplasmic reticulum (splice isoform SERCA1a).

(Received 27 May 2002, revised 23 August 2002,

accepted 6 September 2002)

structure obtained in presence of two buried calcium ions is

believed to represent an open conformational state

analog-ous to the previously determined 8 A˚ resolution

PMA1_NEUCR structure

This, together with the striking similarities between the

8 A˚ electron density maps, suggests that the 2.6 A˚ structure

of ATC1_RABIT would be a valid template for building an

atomic model of PMA1_NEUCR Recently, partial models

comprising the first six transmembrane segments and a

portion of the cytoplasmic loop responsible for ATP

hydrolysis were built for plant and yeast H+-ATPases on

the basis of the ACT1_RABIT crystal structure [9] From

these models is was proposed that proton transport in the

H+-ATPases is mediated through specific binding of a

hydronium ion at a site structurally equivalent to the second

calcium binding site in ACT1_RABIT

In this work, we combine homology modeling techniques

and transmembrane topology predictions to build a model

of PMA1_NEUCR that comprises all 10 transmembrane

helices This model is then used to propose an alternative

hypothesis for the proton transport pathway This

hypo-thesis is based on the assumption that as in the case of the

well-known bacteriorhodopsin proton pump, H+ ions

would be the relevant transported species, with their

transport mediated by one or more acidic side chains of

the protein and a network of interacting buried or partially

buried water molecules [10] To find pathways consistent

with this hypothesis, the PMA1_NEUCR model is used to

compute the positions of internal polar cavities that are

large enough to contain at least one water molecule

Analyses of X-ray structures of soluble proteins have indeed

shown that such cavities usually harbor bound water

molecules [11–13] Furthermore, control calculations

repor-ted here, in which the same approach is applied to the

highest resolution structures of the proton pump

bacterio-rhodopsin and halobacterio-rhodopsin, reveal a good

correspon-dence between the positions of identified polar cavities,

water molecules and residues believed to mediate proton

transport in the proton pump

Calculations performed on our H+-ATPase model

identify a string of internal polar cavities, tracing a

well-defined pathway connecting the phosphorylation site in the

extracellular domain to the intracellular side of the

mole-cule Most of these cavities are in contact with residues

previously shown to affect coupling of proton translocation

to ATP hydrolysis Some are also in contact with residues

whose role in proton transport has as yet not been analyzed

This pertains in particular to residues in the cytoplasmic domain which might be involved in the pathway of proton entry

In the current absence of detailed 3D data, these suggestions could be tested by mutatagensis experiments Furthermore, the approach might be a useful preliminary tool for identifying putative proton pathways in other proton-transporting enzymes for which structural data are available

M A T E R I A L S A N D M E T H O D S

Building the 3D model of PMA1_NEUCR

In our approach, the model of PMA1_NEUCR was built using the ATC1_RABIT crystal structure as template by combining transmembrane topology predictions with stand-ard homology modeling techniques This was necessary as sequence similarity between the template and target proteins

is rather low for much of their C-terminal portion Transmembrane topology predictions The transmem-brane topology of PMA1_NEUCR was predicted from the amino acid sequence by averaging the results of five different predictive algorithms: DAS [14], PHDHTM [15],

HMMTOP[16],TMHMM[17] andTMPRED[18]

To assess the accuracy of the predictions, the same algorithms were applied to the ATC1_RABIT protein and the results were compared with the topology defined in the 2.6 A˚ structure (1EUL.pdb) (Table 1), taking into account the position of aromatic residues at the boundaries of the transmembrane segments It was verified that none of these algorithms included this 3D structure in its learning set This

is certainly important for theDASandTMPREDalgorithms which rely directly on a database of known structures The predictions made for PMA1_NEUCR were com-pared with experimental data on the insertion into micro-somes of recombinant peptides from helices M3, M5, M7, M8 and M10 [19,20] (Table 2)

In the case of ATC1_RABIT, combining the predictions from the different algorithms yielded accurate predictions for nine out of 10 transmembrane segments This is consistent with previous reports where it has been shown that secondary structure predictions tend to be improved upon averaging the results from different methods [21] The average transmembrane topology for PMA1_NEU

CR, presented in Table 2, contains 10 transmembrane Table 1 Topological predictions and average topological model for ATC1_RABIT Comparison with the topology of the crystal structure.

segments in agreement with the 8 A˚ electron density map

and with experimental data on recombinant peptides It is

worth noting at this point that these data do not provide

information on the precise boundaries of transmembrane

segments but rather on the regions initiating or impairing

insertion of a segment in the membrane Therefore, in the

case of M8 where the length of the M8 recombinant

peptide is probably shorter than that of the M8 helix in

the native protein [20], the TMHMM predictions for M8

were used

Sequence alignment The sequences of Neurospora crassa

H+-ATPase (PMA1_NEUCR) and rabbit SERCA1a/

Ca2+-ATPase (ATCI_RABIT) were retrieved from the

Swissprot database [22]

Obtaining a correct sequence alignment is the cornerstone

of success in all homology modeling procedures Here two

different algorithms were used to align the two sequences

CLUSTALW [23] was used to generate a global alignment

This alignment showed 21.6% amino acid identity In

addition theSIMalgorithm [24], was used to generate local

alignments, where short segments of both sequences were

optimally aligned The results from the local alignment

obtained withSIMwere used to manually adjust the global

alignment at the boundaries of the gapped segments In

both alignment methods, the Blosum62 substitution matrix

was used and the open gap and extension penalties were 12–

2 forCLUSTALWand 12–4 forSIM

PHDsec [25] was used to generate secondary structure

prediction for PMA1_NEUCR sequence and the DSSP

algorithm [63] was used to produce secondary structure

assignments from the coordinates of the template

The domain displaying the highest sequence similarity

between the target and template proteins extends from the

N-terminus of ATC1_RABIT up to transmembrane helix

M6 This domain includes the large cytoplasmic loop

containing the active site responsible for MgATP hydrolysis

which is located between M4 and M5 This cytoplasmic

loop seems to be shorter in PMA1_NEUCR than in

ATC1_RABIT In order to superimpose the strictly

conserved ÔVKGAP777Õ and ÔDPPR537Õ motives, ClustalW

introduces three large gaps in the PMA1_NEUCR

sequence, representing deletions of 14, 17 and 45 residues

in ATC1_RABIT The corresponding parts of the 3D

model were, however, close enough to be linked by short

connecting loops using the loop modeling procedure, as described below These loops are located in the periphery of the structure and do not interfere with the core domain Aside from this gapped cytoplasmic loop, sequence identity in this domain exceeded 30% making this part of the alignment more straightforward We could furthermore verify that this sequence alignment was consistent with the alignment between the secondary structures predicted solely from the sequence of PMA1_NEUCR and those assigned

in the equivalent regions of the ATC1_RABIT template (Fig 1)

The C-terminal domain of both enzymes displayed a much lower level of sequence similarity, a common feature

in eukaryotic P-type ATPases [26] In addition, several lines

of evidences suggest that PMA1_NEUCR contains an additional cytosolic regulatory domain at the C-terminus This domain is located following the last transmembrane segment M10

Consequently, the raw global alignment produced by

CLUSTALWfor the C-terminal domain had to be revised, but without the help from local alignments produced bySIMfor this region, as those concerned very short segments separ-ated by large gaps Information on the predicted trans-membrane topology was therefore used as a guide to align the sequence from M6 onwards Despite the low level of sequence identity, this topology-based alignment was con-sistent with theCLUSTALW alignment up to segment M7, suggesting that the loops connecting, respectively, M5 and M6, and M6 and M7, would be equally short in both enzymes The region encompassing M8, M9 and M10 were aligned manually by aligning the corresponding transmem-brane segments of both enzymes while maximizing sequence identity

As mentioned above, the prediction for the M8 segment is consistent with available results on recombinant peptides The position of M10 is consistent with results on the tryptic cleavage of the additional C-terminal regulatory domain in PMA1_NEUCR, which showed that the 897–920 segment had a cytoplasmic location [27] In plant H+-ATPases, where an equivalent domain is also located after M10, this domain was suggested to interact directly with the enzyme active site [28] In the ATC1_RABIT template, the active site is located in the large cytoplasmic loop between M4 and M5, almost 45 A˚ away from the end of M10 These considerations impose some constraints on the

Table 2 Topological predictions and average topological model for PMA1_NEUCR Comparison with available data on recombinant peptides.

Predictive algorithms

Recombinant peptides

M3 291–312 289–313 297–314 292–314 292–310 292–313 292–314

M5 691–714 689–713 697–721 691–713 688–713 691–715 688–713

M7 757–774 755–779 758–775 757–775 755–774 756–775 755–779

M9 828–843 822–846 826–843 829–849 827–847 826–846 827–847 M10 854–877 854–878 859–876 860–878 860–877 856–878 854–878

PMA1_NEUCR model in this region In particular, the

regulatory domain of PMA1_NEUCR should be able to

reach across the active site If we believe that it adopts an

a-helical structure as indicated by the prediction (Fig 1),

then the corresponding helix would have to be of at least 30

residues long This implies in turn that the M10 segment of

PMA1_NEUCR would end before residue 890 It is worth

noting that the M9 and M10 helices of this enzyme were

detected by all algorithms with a high level of confidence

and were accordingly superimposed to their structural

equivalents in ATC1_RABIT

The final sequence alignment is shown in Fig 1 with a

comparison between the secondary structure predictions for

PMA1_NEUCR sequence and the assigned secondary

structure of the 3D model Although this secondary

structure prediction was not used in the alignment part of

the modeling procedure, it is presented here to show that

in the cytosolic domains, the secondary structure

predic-tion and the secondary structure resulting from the

modeling procedure are remarkably consistent Position

of PMA1_NEUCR and ATC1_RABIT transmembrane segments are indicated as shaded colored boxes

Model building Using the sequence alignment displayed in Fig 1 and the 2.6 A˚ resolution X-ray structure of ATC1_RABIT (RSCB-pdb code: 1EUL) [8], a first model

of PMA1_NEUCR was generated with the PromodII [29] package ofDEEPVIEW3.7 [30] Reconstruction of the loops

in gap regions was achieved with the loop database module

ofDEEPVIEW Model quality assessment and refinement The quality of the model was assessed using the WHATIF

v.4.99 [31] andPROCHECK[32] validation suites The 2.6 A˚ resolution structure of ATC1_RABIT was analyzed with the same suites as a control The results showed that the bonded geometry and the stereochemistry of both structures

Fig 1 Sequence alignment of PMA1_NEUCR and ATC1_RABIT PMA1_NEUCR PHD: secondary structure prediction of PMA1_NEUCR using the PHDsec algorithm PMA1_NEUCR DSSP: secondary structure of the model structure assigned with the DSSP algorithm (arrows: b-sheets; helices: a-helices) Shaded colored box: transmembrane domains of ATC1_RABIT and PMA1_NEUCR Shaded orange box: PMA1_NEUCR transmembrane segments built with the topology-based alignment Numbers below the sequences refer to the residues in direct contact with the cavities 1 to 9 in PMA1_NEUCR; italic font is used when a mutation has been reported Residues appearing in white over a red background are identical Residues appearing in red over a white background are homologous.

were of similar quality, indicating that care was taken in

both the homology modeling and X-ray refinement

proce-dures to optimize these parameters

However, significant differences were observed in the

number of close nonbonded contacts Those were higher in

the constructed model of PMA1_NEUCR than for the

crystal structure of the rabbit enzyme To relieve this strain

the model was subjected to two short energy minimization

runs with GROMOS 96 [33] using the GROMOS 43B1 force

field These runs involved 200 steps of steepest descent

followed by 300 steps of conjugate gradient optimizations

This led to a significant drop in the unfavorable nonbonded

contacts, while producing only very minor displacements of

the atomic coordinates

Identification of internal cavities

Internal cavities were identified from the atomic

coordi-nates of the PMA1_NEUCR model, ATC1_RABIT

(1EUL.pdb), bacteriorhodopsin (1C3W.pdb),

halorhodop-sin (1E12.pdb),L-2-haloacid dehalogenase (1JUD.pdb) and

phosphoserine phosphatase (1F5S.pdb) using the surface

module of DEEPVIEW 3.7 and the program SURVOL [34]

In both programs the computed cavities were delimited by

the molecular surface computed with a probe size of 1.4 A˚

R E S U L T S A N D D I S C U S S I O N

Comparison of the ion binding sites in ATC1_RABIT

with the equivalent region in PMA1_NEUCR

In ATC1_RABIT and other mammalians P-type ATPases

[35,36], several amino acids involved in cation binding were

identified by site-directed mutagenesis along

transmem-brane segments M4, M5, M6 and M8 The crystal structure

of ATC1_RABIT reveals how these residues assemble to

form a binding pocket surrounding two Ca2+ ions A

comparison of this binding pocket with the corresponding

domain of our 3D model of PMA1_NEUCR is shown in

Fig 2

Ion binding site I Five residues form the first Ca2+

binding site The calcium ion is bound to the side-chain

oxygen of Asn768 and Glu771 on M5, of Thr799 and

Asp800 on M6 and Glu908 on M8 All the oxygens are

arranged in roughly the same plane except for the side chain

of Glu771, which is located below

In our PMA1_NEUCR model, these residues are replaced by Ser699, Leu702 on M5, by Ala729 and the conserved Asp730 on M6, and Ala814 on M8 With this residue constellation, the ion-binding site is most probably lost The possibility remains, however, that the presence of Ser and Asp residues in this region would allow it to participate in proton transport Alanine-scanning mutagen-esis along M5 [37] showed that Ser699 is probably involved

in proton transport, although it is not essential to the coupling mechanism Substitution of Leu702 to Ala resulted

in an enzyme with a normal coupling ratio Replacement of Asp730 a conserved residue in segment M6 by Asn or Val led to a poorly folded enzyme, arrested in the endoplasmic reticulum [41] Nevertheless, double mutants in which both charged Arg695 and Asp730 were inversed or replaced by Ala reverted to a fully functional enzyme These observa-tions indicate that these residues are linked by a salt bridge, making a direct participation in proton transport unlikely [39] In plant H+-ATPases (Arabidhopsis thaliana), how-ever, the conserved Asp residue does not seem to be involved in a salt bridge and might hence play a role in proton transport [40] Nevertheless, among the investigated residues of yeast and fungi H+-ATPases corresponding to calcium binding site 1 of ATC1_RABIT, only Ser699 seems

to play a role, albeit a nonessential one, in proton transfer Ion binding site II The second calcium-binding site of ATC1_RABIT is formed by six residues, nearly all of which are located on M4 This site is formed by main-chain carbonyl oxygen of Val304, Ala305, and Ile307 on M4; by side-chain oxygens of Glu309 on M4 and of Asn796 and Asp800 on M6 In our PMA1_NEUCR model, these residues correspond to Ile331, Ile332, Val334, and Val336

on M4 and Ala726 and Asp730 on M6

Alanine-scanning mutagenesis along segment M4 of yeast H+-ATPase showed that replacement of Ile331 and Val334 had little or no effect on ATP-dependent proton transport [41], not inconsistent with the fact that the mutations do not change the nature of the backbone Replacement of Ile332 or Val336 resulted, however, in a coupled mutant enzyme displaying altered kinetics consis-tent with a slow down of the E1P–E2P transition step believed to be coupled to the charge transfer reaction Finally, Asp730 corresponding to Asp800 in ATC1_RABIT was shown to be involved in a salt bridge with Arg695 precluding direct participation in proton transport, as discussed above

Fig 2 Comparison of ATC1_RABIT (A) and PMA1_NEUCR (B) ion-binding sites regions Calcium ions in ATC1_RABIT are labeled according to the binding site numbers A possible conformation for Arg695 and Asp730 making a salt bridge between M5 and M6 is indicated by the dotted line.

In conclusion, comparisons of the residues involved in

ATC1_RABIT calcium binding sites with their structural

equivalents in PMA1_NEUCR indicate clearly that this

region is not conserved in the latter enzyme However, as it

appears from the mutagenesis studies described above, the

possibility that some residues in this region are involved in

proton transfer can not be ruled out Only three residues

(Ser699, Ile332 and Val336) out of the 10 residues equivalent

to the cation binding residues in ATC1_RABIT might play

a role in this process, but probably not an essential one

Identification of a putative proton pathway

in Neurospora crassa H+-ATPase

The chemiosmotic model for PMA1_NEUCR In the

P-type proton pumps, the origin of the transported proton

as well as the proton entry pathway is still unknown The

so-called Ôchemiosmotic modelÕ for PMA1_NEUCR, based

largely on biochemical data, makes an interesting proposal

concerning the initiation site for proton transport [49] It

stipulates that this transport is initiated by the lysis of a

water molecule in the cytoplasmic phosphorylation site,

implying that the proton pathway would begin close to the

phosphorylated aspartate (Asp378)

The major steps of the model are as follows: The first

step is a covalent phosphoryl-transfer reaction from the

MgATP molecule to the strictly conserved Asp residue

(Asp378), as shown by radioactively labeled ATP [42,43]

The next step is dephosphorylation of the

aspartyl-phosphate group with subsequent release of aspartyl-phosphate

at the cytoplasmic side of the enzyme This reaction

involves a water molecule whose oxygen atom promotes

disruption of the covalent bond between the conserved

Asp378 residue and Pi [44] The released protons have

been proposed to be withdrawn by functional residues

acting as general bases on their way to the transport

reaction [49]

Internal polar cavities as loci of proton transport in

membrane proteins In monomeric and globular proteins,

internal polar cavities often contain buried water molecules

which interact both with other protein groups and with one

another [11–13] Such cavities could therefore represent loci

where water molecules could form H-bond networks that

foster proton conduction [46,48] On the basis of this

hypothesis, we set out to identify a possible pathway for

proton transport in PMA1_NEUCR by identifying the

position of internal polar cavities in the built molecular

model

The role of water molecules as a crucial determinant of the

proton conduction network has been amply documented in

bacteriorhodopsin [45] To lend support to our hypothesis,

we therefore applied the same approach to the

high-resolu-tion structure of bacteriorhodopsin and the closely related

structure of halorhodopsin, which are the two highest

resolution structures for any transmembrane protein to date

Position of internal polar cavities in the high-resolution

structures of halo- and bacteriorhodopsin Using the

procedure described in the Methods section, we identified

the internal polar cavities in the 1.55 A˚ resolution structure

of bacteriorhodopsin (1C3W.pdb) and the 1.8 A˚ resolution

structure of halorhodopsin (1E12.pdb) In both structures,

essentially all the buried crystallographic water molecules are located in internal polar cavities, as illustrated in Fig 3A,B [46,47] In the case of the bacteriorhodopsin proton pump, eight such cavities have been identified The

15 polar residues that line these cavities are Thr5, Arg7, Thr46, Tyr57, Tyr79, Arg82, Tyr83, Asp96, Ser193, Glu194, Glu204, Tyr205, Asp212, and Lys216 Of these, 10 residues (Thr46, Tyr57, Arg82, Asp85, Asp96, Glu194, Glu204, Glu205, Asp212 and Lys216) are believed to be involved in the network of the hydrogen-bonded residues and water molecules that define the proton pathway [48] Thus, in this case, determining the position of internal polar cavities in the 3D structure enable to outline the pathway for proton transfer Identifying such cavities in the model of PMA1_NEUCR, a membrane protein for which the proton pathway has not as yet been delineated might provide useful information about this pathway

Positions of internal polar cavities in the PMA1_NEUCR model Applying the procedure of cavity identification to the PMA1_NEUCR model yields a total of 21 internal cavities, whose volume varies from 14 to 93 A˚3 Most of them are located along the longitudinal axis of the protein as seen in Fig 4 Along this axis, two main groups of cavities can be distinguished The first is located just below the cytoplasmic site of MgATP hydrolysis The second group, located in the membrane domain, almost connects the region homologous to the ion-binding site to the extra-cytoplasmic moiety of the molecule

The positions of the internal polar cavities hence suggest that the proton translocation pathway might begin close to the phosphorylation site in the large cytoplasmic loop, in agreement with the semi-empirical chemiosmotic model for PMA1_NEUCR [49]

In order to verify this suggestion, we listed all the residues lining these cavities and compared them to those shown to affect the proton transport reaction by mutagenesis experi-ments The cavity-lining residues number 54 in all, of which

22 are identical to those in ATC1_RABIT (see alignment of Fig 1)

Table 3 lists the 54 residues, together with the cavity they contact, and the effect of mutations on the kinetic properties and the coupling ratio between ATPase activity and proton transport These residues are numbered on the sequence alignment of Fig 1, in italic fonts when a mutation has been reported or in regular font when they have not yet been investigated

(a) Internal polar cavities in the transmembrane domain In the transmembrane domain, the first small cavity (cavity 7) is located between Val289 and Ile293 in the N-terminal end of M3, and Trp756 and Gly757 in the cytoplasmic loop connecting M6 and M7 As shown by mutagenesis in yeast, mutation of Val289 to Phe resulted in

an altered phenotype that seems not to be directly related to the charge translocation step [50] In ATC1_RABIT, however, the homologous residues were identified as possible determinants for control of the gateway to the

Ca2+binding sites by site-directed mutagenesis [51,52] The next cavity in the transmembrane domain is located above the ion-binding site region (cavity 8) Important residues in contact with this cavity are Tyr691 and Val336 Replacement of Tyr691 by Ala resulted in an enzyme

defective in the E1-E2 conformational change Here, the

Val336Ala mutant displayed kinetic properties consistent

with a decrease of the transport-linked E1P–E2P transition

step [41] As seen above, Val336 corresponds to Glu309 of

ATC1_RABIT that contributes directly to calcium binding

site 2 (Fig 2)

The following long cavity (cavity 9) is located between

helices M4, M5, M6 and M8 again in the region

corresponding to the ion-binding site A conserved residue,

Tyr694, is making contact with this cavity The Tyr694 to

Ala mutant strongly decreased ATPase activity, while the

Tyr694 to Gly mutant displayed a strong resistance towards

inhibition by vanadate Interestingly, Tyr694Ala mutant

resulted in a presumably uncoupled enzyme, although the

low rate of ATP hydrolysis prevented a detailed analysis of

the coupling ratio In ATC1_RABIT, mutation of the

corresponding residue (Tyr763), clearly led to an uncoupled

enzyme unable to transport Ca2+ions [53] Another residue

in contact with cavity 9 is Ser699 that was found to be involved in proton translocation (see above)

Two other cavities are observed between M6, M8 and M9 that are in the nonhomologous C-terminal domain built using the topology-based alignment (see Fig 1) The position of the side chains and hence of cavities in this domain are therefore considered as less reliable than in other parts of the model It is nonetheless of interest that residue Glu805 corresponding to Glu803 in the yeast H+-ATPase

is found close to cavity number 10, a small internal cavity that appears as nonpolar in our model Nevertheless mutagenic studies in yeast H+-ATPase revealed that substitution at Glu803 by either Gln or Asn seems to increase the rate proton transport [38] In our model, this residue is partially accessible to the solvent and may therefore act as a gate for the possible exit of a water-coordinated proton occurring at the extra-cytoplasmic end

of the proton transfer path

Fig 3 Position of internal polar cavities and crystallographic water molecules in the high resolution crystal structures of bacteriorhodop-sin (A) and halorhodopbacteriorhodop-sin (B) Red spheres represent buried crystallographic water mole-cules; blue dots represent crystallographic water molecules lying outside the surface envelope of the enzymes Cavities are colored

by atom type (red stands for O, blue for N and yellow for S) The backbone position of the ten polar residues lining the proton pathway and contacting the identified internal polar cavities is colored in green.

Fig 4 Position of internal polar cavities in PMA1_NEUCR model (A) and (B) views and numbering of internal polar cavities The view

in (B) is rotated by nearly 90° with respect to (A) (the NH 2 -terminus is omitted for clarity) Cavities are colored by atom type (red stands for O, blue for N and yellow for S) The horizontal black lines delineate the approxi-mate membrane position The catalytic Asp (Asp378) appears in red.

Table 3 List of the residues in contact with the cavities 1 to 9 The effect of a mutation is reported when the data was available (a) Observed but not characterized due to low rate of ATP hydrolysis ND, not determined.

Region Mutation Expression Coupling Kinetic E1–E2 ATPase activity Reference Cavity 1

Cavity 2

L557 K615

– –

– –

– –

– –

Cavity 3

Cavity 4

Cavity 5

Cavity 6

Stalk M4

We thus see that the pathway outlined by the cavities

numbered 7–10 connects the bottom of the phosphorylation

site organized in a Rossman fold to the extracytoplasmic

moiety Furthermore, all these cavities, with the exception of

cavity 7, are lined by one or more residues previously shown

to be involved in the proton translocation step

(b) Internal polar cavities in the cytoplasmic domain

The cavities identified in this domain trace a clear path that

starts in the center of the b-strands forming the Rossman

fold and leads to the membrane domain at the top of M4

and M5 (see Fig 4, cavities numbered 1–6) As could be

expected from the high level of sequence identity around the

phosphorylated aspartate (Asp378 in PMA1_NEUCR),

several cavities are also seen at the same position in

ATC1_RABIT (data not shown) Aside from mutations

directed against a small stretch of amino acids adjacent to

the phosphorylated Asp [54], limited information is

avail-able from mutagenesis studies on residues involved in

proton translocation in the cytoplasmic phosphorylation

site

At the bottom of the cytoplasmic domain, cavity number

6 is in contact with two residues of interest for the proton

translocation step: Met346 and I366 Mutations of either

Met346 or Ile366 to Ala result in a reduced K for MgATP,

an acidic shift of the optimum pH and an increased resistance towards inhibition by vanadate These effects are consistent with a slow down of the transport-linked E1P–

E2P transition step, suggesting that they may be involved in the transport reaction [41,55]

An apparent pathway by which a proton might be internalized from the cytosol to reach cavity 6 is through cavities 1–5 Indeed, polar residues are lining these cavities, consistent with the idea that they might harbor a water-mediated H-bond network fostering proton transfer However, a potential problem in validating this proposal

is that the fold of PMA1_NEUCR seems very sensitive to mutations directed against residues adjacent to the catalytic Asp378 Nevertheless, most of the residues lining cavities 1–5 are found in the vicinity of the ÔAMTGDGVNDAP640Õ motif, a region that has as yet not been thoroughly investigated The residues identified in this region (Table 3) might thus constitute suitable targets for site-directed mutagenesis

(c) Internal polar cavities in soluble members of the haloacid dehalogenase superfamily (HAD) With little information on the residues lining our proposed entry pathway available from mutagenesis studies, positions of the cavities located in our model were compared with those

Table 3 (Continued).

Region Mutation Expression Coupling Kinetic E1–E2 ATPase activity Reference Cavity 7

of S368F

Cavity 8

with D730 revertant

[37] [39]

Cavity 9

of internal cavities capable of containing water molecules in

the homologous Rossman fold of the soluble members of

the HAD superfamily (Fig 5)

We used the high-resolution structures of theL-2-haloacid

dehalogenase (1JUD) and of phosphoserine phosphatase

(1J5S) that exhibit the same organization of the active site

(Rossman fold) in the cytoplasmic domain of the P-type

ATPases In phosphoserine phosphatase andL-2-haloacid

dehalogenase structures, respectively, 69% and 66% of the

Ca comprising the Rossman fold displayed an root mean

squared deviation of less than 1.85 A˚ with the homologous

domain of PMA1_NEUCR model This makes them close

structural homologues of the cytoplasmic phosphorylation

site in the P-type ATPases as has already been reported

elsewhere [56–59] Internal polar cavities and

crystallogra-phic water molecules were identified in these structures using

the described procedure Quite strikingly, in both cases, the

structures exhibit a single internal polar cavity in the vicinity

of the conserved catalytic Asp This cavity contains two to

three crystallographic water molecules In stark contrast, the

Rossman fold of PMA1_NEUCR contains as many as six

polar cavities, which link the phosphorylated Asp378 all

through the middle of the a/b Rossman fold, down to the

beginning of the transmembrane domain The presence of this string of polar cavities in PMA1_NEUCR Rossman fold, the most accurate part of our model, and its conspicuous absence in the otherwise close structural relatives from the soluble members of the HAD family, point to a possible functional role of these cavities in PMA1_NEUCR

C O N C L U S I O N

In this study, we built a complete atomic model of Neurospora crassa plasma membrane H+-ATPase (PMA1_NEUCR) using the high-resolution 3D structure

of the rabbit sarcoplasmic reticulum Ca2+-ATPase (ATC1_RABIT) as a template and combining transmem-brane topology predictions with classical homology mode-ling techniques in regions of low sequence similarity

In the first part of this work, the comparison of the ion-binding site of ATC1_RABIT with the homologous domain

in PMA1_NEUCR revealed that this domain is not conserved in both enzymes This suggests that although the P-type ATPases are widely assumed to share a common mechanism of action, each group of proteins probably

Fig 5 Position of internal polar cavities in the

Rossman fold of PMA1_NEUCR and two

other soluble members of the HAD superfamily.

(A) Phosphorylation site of PMA1_NEUCR,

(B) phosphorylation site of Phosphoserine

phosphatase, (C) active site of L-2-haloacid

dehalogenase The catalytic Asp in the

Ross-man fold appears in red Red spheres

repre-sent buried crystallographic water molecules;

blue dots represent crystallographic water

molecules lying outside the surface envelope of

the enzymes Cavities are colored by atom

type (red stands for O, blue for N and yellow

for S).