Nature Structural Biology
9, 507 - 511 (2002)
Published online: 10 June 2002; | doi:10.1038/nsb812
Structure of human phosphatidylcholine transfer protein in complex with its ligandSteven L. Roderick1, Wayne W. Chan2, Diana S. Agate1, Laurence R. Olsen1, Matt W. Vetting1, K.R. Rajashankar3, 4
& David E. Cohen1, 21 Department of Biochemistry, Marion Bessin Liver Research Center, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, New York 10461, USA. 2 Department of Medicine, Marion Bessin Liver Research Center, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, New York 10461, USA. 3 Brookhaven National Laboratory, National Synchrotron Light Source, Building 725A-X9, Upton, New York 11973, USA. 4 Present address: Argonne National Laboratory, APS/SGX-CAT, Building 438/A002 9700, South Cass Avenue, Argonne, Illinois 60439, USA.
Correspondence should be addressed to Steven L. Roderick roderick@aecom.yu.edu or David E. Cohen dcohen@aecom.yu.edu.Phosphatidylcholines (PtdChos) comprise the most common phospholipid class in eukaryotic cells. In mammalian cells, these insoluble molecules are transferred between membranes by a highly specific phosphatidylcholine transfer protein (PC-TP) belonging to the steroidogenic acute regulatory protein related transfer (START) domain superfamily of hydrophobic ligand-binding proteins. The crystal structures of human PC-TP in complex with dilinoleoyl-PtdCho or palmitoyl-linoleoyl-PtdCho reveal that a single well-ordered PtdCho molecule occupies a centrally located tunnel. The positively charged choline headgroup of the lipid engages in cation− interactions within a cage formed by the faces of three aromatic residues. These binding determinants and those for the phosphoryl group may be exposed to the lipid headgroup at the membrane−water interface by a conformational change involving the amphipathic C-terminal helix and an  -loop. The structures presented here provide a basis for rationalizing the specificity of PC-TP for PtdCho and may identify common features used by START proteins to bind their hydrophobic ligands.
Phosphatidylcholines (PtdChos) are the most common class of phospholipids in the majority of eukaryotic cell membranes. The aqueous monomeric solubilities of PtdChos are extremely low ( 10-10 M)1, and their spontaneous transfer rates between membranes are negligible. Mammalian phosphatidylcholine transfer proteins (PC-TPs) promote the rapid intermembrane exchange of phosphatidylcholines but no other phospholipid class2,
3,
4,
5,
6,
7,
8. PC-TP, phosphatidylinositol transfer protein (PI-TP) and sterol carrier protein 2 (SCP2) account for most of the phospholipid exchange activity present within the cytosol8. Although the biological function of PC-TP remains uncertain, its highest level of expression in humans is found in the liver7. Therefore, PC-TP may shuttle PtdChos from their site of synthesis in the endoplasmic reticulum to the inner monolayer of the plasma membrane, replenishing PtdChos that are removed from the outer monolayer as extracellular apolipoprotein A-I is converted to nascent high density lipoprotein particles9.
Recent sequence database analyses have led to the classification of PC-TP as a protein containing a steroidogenic acute regulatory protein (StAR)-related transfer (START) domain10,
11. The START domain binds hydrophobic ligands and frequently appears in multidomain proteins, where it regulates the activity of associated domains, including pleckstrin homology, p122-RhoGAP and thioesterase domains. START proteins function in a wide variety of biological processes and have been identified in at least 30 human proteins12,
13. The putative START domain encompasses almost the entire 214-residue amino acid sequence of PC-TP and is not fused to an additional domain. Therefore, mammalian PC-TPs represent the minimal START protein structure.
The crystal structure of the unliganded START domain of human MLN64, a homolog of StAR thought to function in steroidogenesis, has been reported together with a model of its complex with a single molecule of cholesterol14. Here we present the first structures of a START domain in complex with its ligand. These structures explain the exquisite specificity of PC-TP for PtdChos, offer insight into the lipid transfer mechanism of PC-TPs and provide a prototypical example for understanding how START domain proteins bind their natural ligands.
Overall structure of PC-TP
Three structures of human PC-TP in complex with phosphatidylcholines were solved corresponding to two different PtdCho molecular species. Selenomethionyl PC-TP in complex with DLPC (1,2-dilinoleoyl-sn-glycerol-3-phosphorylcholine; 18:2(9,12)-18:2(9,12)) was determined first to 2.9 Å resolution by MAD and subsequently to 2.4 Å resolution from an alternate crystal form by molecular replacement. In addition, a binary complex with PLPC (1-palmitoyl,2-linoleoyl-sn-glycerol-3-phosphorylcholine; 16:0−18:2(9,12)) was solved to 2.8 Å resolution.
The overall structure of PC-TP consists of a curved antiparallel  -sheet, which consists of nine strands, and four  -helices (Fig. 1). Helix 1 (residues 9−22) rests against the back of this -sheet, and helices 2 and 3 are inserted between strands 3 and 4.  -loops are inserted between 5−6 (1) and 7−8 (2). The long amphipathic C-terminal helix, 4 (residues 184−209), is kinked at Pro 197 and lays over the top of a tunnel that accommodates a single molecule of PtdCho. The walls of this tunnel are formed by the curved -sheet, 2, 3, 4 and loop 1.
 | | Figure 1. Overall structure of human PC-TP. | | |  | a, Stereo view of DLPC with density contoured at 1.0  from a 2Fo - Fc map before fitting the lipid. b, Stereo view ribbon diagram of PC-TP in complex with DLPC. The central antiparallel -sheet is dark blue. c, Stereo view of the polypeptide chain conformation of PC-TP (residues 8−210). The secondary structure nomenclature used for human PC-TP follows that used to describe the START domain of human MLN64 (ref. 14). The -helix identifiers and residue ranges for human PC-TP are 1 (9−22), 2 (64−74), 3 (75−82) and 4 (184−209). The -strand identifiers and residue ranges are 1 (31−36), 2 (39−46), 3 (51−61), 4 (84−93), 5 (96−104), 6 (111−123), 7 (130−138), 8 (150−162) and 9 (168−178). The -loop identifiers and residue ranges are 1 (105−110) and 2 (139−149). Panel (a) was prepared using SETOR41. All others were prepared using MolScript42 or BobScript and Raster3D43,
44.
 Full Figure and legend (89K) |
| |
The structure of PC-TP in complex with DLPC determined from form II crystals revealed unexpectedly the presence of a disulfide bond joining two conserved Cys residues — Cys 63 donated from a loop joining 3−2 and Cys 207 of the C-terminal helix that overlays the lipid binding tunnel. This bond is not observed in either of the other two crystals studied here and its presence induces only slight local changes in structure. Our measurements of the specific activities of the C63A and C63S mutants of human PC-TP, both of which are incapable of forming a disulfide linkage, were 78% and 45% of the wild type protein, respectively. In addition, the activity of bovine liver PC-TP is invariant in the presence of 0− 200 mM of dithiothreitol (DTT)15. As a result, the relevance of this observed disulfide linkage to the function of PC-TP in vivo is doubtful.
The overall fold of PC-TP is similar to the START domain of human MLN64 (ref. 14) and to birch pollen allergen Bet v 1 (ref. 16), a protein of unknown function that has been implicated in the stress/pathogen response of plants10. A superposition of the structure of PC-TP onto these proteins yields an r.m.s. deviation of 1.75 Å for 126 C positions of MLN64 (Fig. 2) and 1.76 Å for 86 C positions of Bet v 1. As noted by Yoder and co-workers17, the overall fold of MLN64 is similar to that of the -isoform of rat phosphatidylinositol transfer protein (PITP), although this protein is not classified as a member of the START protein superfamily. This similarity extends to PC-TP as demonstrated by an r.m.s. deviation between PC-TP and PITP of 1.87 Å for 72 C positions.
The PtdCho-binding site
The lipid-binding tunnel of PC-TP accommodates one molecule of PtdCho (DLPC van der Waals volume 787 Å3) within a solvent accessible volume of 882 Å3 and reduces the bulk solvent accessible surface area of the lipid by >99% (Fig. 3a). The tunnel extends to bulk solvent through two narrow portals 3−5 Å in diameter. The overall composition of the tunnel is partly hydrophilic, because it contains at least five ordered water molecules, the phenolic hydroxyls of eight tyrosine residues and the side chains of four additional hydrophilic residues. The tunnel of PC-TP is substantially larger than that of MLN64, which binds a smaller cholesterol molecule. This is partly the result of the closer approach made by the C-terminal helix 4 and loop 1 to the curved -sheet (Fig. 2).
| | |
The position and conformation of DLPC and PLPC as bound to PC-TP are quite similar, despite the differences in their sn-1 acyl chain length and saturation (18:2(9,12) versus 16:0). Both acyl chains adopt a C-shaped conformation and show some disorder toward their termini (Fig. 3b). The C-shape of the unsaturated sn-1 palmitoyl chain of PLPC seems unlikely in a less confining cavity than the lipid binding pocket of PC-TP. Of the 28 residues that make contact with the PtdCho ligand, 15 are aliphatic, 11 are aromatic and 2 are hydrophilic (Table 1). Residues are contributed by 12 of the 15 secondary structural elements of the protein, including 8 of the 9 -strands of the curved -sheet, loop 1 and helices 2, 3 and 4. Residues of the curved -sheet or the C-terminal helix 4 form all but four contacts to the ligand. The sn-1 acyl chain of the PtdCho interacts with the side chain groups of six residues donated by strands 1−5, a single residue from loop 1 and five additional residues from the C-terminal helix. An ester oxygen hydrogen bonds to a water molecule. The sn-2 acyl chain contacts residues from 2−3, 2, 8−9 and 4. Although PC-TP is known to preferentially bind PtdChos with sn-1 palmitoyl plus sn-2 polyunsaturated acyl chains8,
18, the structures of PC-TP determined here do not provide a convincing rationale for this acyl chain specificity.
| | |
The phosphorylcholine head group of the lipid interacts with the hydrophilic side chain groups of Tyr 72, Gln 157 and Arg 78. Arg 78 interacts directly with the phosphoryl group and forms a salt bridge with Asp 82, as predicted by Tsujishita and Hurley14 from their analysis of the structure of MLN64, its amino acid sequence alignment with PC-TP and the zwitterionic nature of PtdCho. However, the carboxyl group of Asp 82 is located 9 Å from the positively charged quaternary amine of the lipid. As noted by these investigators, this Arg-Asp pair is found in the sequence of many START proteins. The direct interaction of this Arg residue with the phosphoryl group of PtdCho suggests that detection of this Arg-Asp pair in the sequence of orphan START proteins could point to a preferred ligand type that bears a formal negative charge if this protein−ligand interaction is conserved.
The methyl groups of the lipid quaternary amine contact the side chains of Val 103, Tyr 116 and Tyr 175 and insert into a three-walled aromatic cage formed by the ring faces of Trp 101, Tyr 114 and Tyr 155. The use of aromatic residues to form alkylamine binding pockets using such cation− interactions19,
20 has been observed in the substrate-binding pockets of trimethylamine dehydrogenase21 and acetylcholinesterase22 and, most recently, in complexes of Lys 9-dimethylated or -trimethylated histone H3 tail peptides bound to the heterochromatin-associated protein 1 (HP1) chromodomain23,
24. However, these interactions have not been found in the structure of PITP in complex with sn-1, 2-dioleoyl PtdCho17. The structural similarity of the quaternary amine binding pockets of PC-TP and HP1 is a striking example of evolutionary convergence to form a three-walled aromatic cage capable of binding quaternary amines through cation− interactions (Fig. 3c).
Mechanism of PtdCho transfer
PC-TP catalyzes PtdCho one-for-one exchange, as well as net transfer of PtdCho between membranes in vitro25,
26. One-for-one exchange could promote the redistribution of PtdChos with differing acyl chain compositions in vivo, whereas net transfer is presumably required for the transport of PtdChos from their point of synthesis in the endoplasmic reticulum to the PtdCho-depleted inner monolayer of the plasma membrane for eventual incorporation into HDL particles9. In both cases, PC-TP is thought to accelerate the rate-limiting step of desorption27.
The structure of human PC-TP offers insights into the mechanism of binding of PtdCho. PC-TP binds a single molecule of PtdCho in a tunnel joined to bulk solvent through portals that are too narrow to permit binding or release of PtdCho. As a result, a substantial conformational change of the protein must accompany ligand binding or release. For several reasons, we believe that the major conformational change involves the C-terminal 4 helix and perhaps loop 1. These secondary structural elements interact exclusively with the acyl chains of bound PtdCho and make no contacts with the phosphorylcholine headgroup (Table 1). Instead, the headgroup-binding pocket is formed by residues donated from the curved -sheet that interact with the trimethylammonium moiety and those from helices 2 and 3 that contact the phosphoryl group. Hence, a conformational change of the C-terminal helix or 1 from their observed positions lining the tunnel would not necessarily disrupt any element of the phosphorylcholine-binding pocket and could expose the necessary groups for interaction with the lipid headgroup at the membrane−water interface. In addition, this conformational change could promote the interaction of the displaced C-terminal helix with the membrane. We have shown that the C-terminal helix is essential for activity because removal of the last five residues of human PC-TP decreases activity by 50% and removal of the last 10 residues abolishes activity altogether and markedly decreases the level of binding to the membrane28. Moreover, the -helical content of peptides designed to the C-terminal helical portion of human PC-TP increases in the presence of small unilamellar vesicles28.
Such a conformational change of the C-terminal helix or 1 could also occur in other START proteins, as has been proposed for MLN64 on the basis of the location of disease-linked mutations of StAR that map to helix 4 of MLN64 and the generally hydrophobic surface surrounding loop 1, which could interact with a membrane14. One residue of the three-walled aromatic cup that binds the quaternary amine of PtdCho, Trp 101, is equivalent to MLN64 Asp 332, a residue that has been proposed to interact with the 3-hydroxyl group of its cholesterol ligand14. Trp 101 is also equivalent to Glu 169 of StAR, a residue that is mutated to Gly or Lys in the sequence of StAR from two individuals suffering from congenital lipoid adrenal hyperplasia29. These important residues may mark the general location of a binding pocket within the tunnel of START proteins that is exposed on conformational change to phospholipid headgroups or other hydrophilic moieties oriented at the membrane−water interface.
Methods
Crystal preparation and data measurement.
Human PC-TP was overexpressed, bound with synthetic DLPC or PLPC from small unilamellar vesicles in crude cytosolic extracts, purified and then crystallized by the hanging drop vapor diffusion method as described30. The crystallization solutions contained 3.4−3.8 M sodium formate, 0.1 M sodium acetate, pH 5.7, with 5 mM DTT added for SeMet-PC-TP. Two distinct but macroscopically indistinguishable crystal forms were obtained from both native and selenomethionyl protein preparations. Form I crystals belong to space group P4212, with a = 134.7, c = 82.7 Å, and contain two copies of the monomeric protein in the asymmetric unit, related by a translation of approximately (½,½,½). Form II crystals belong to space group I422, with similar unit cell parameters, but contain just one molecule in the asymmetric unit. Prior to X-ray data measurement, crystals were introduced into a synthetic liquor containing 20% (w/v) PEG 400 as a cryoprotectant and then cooled to 117 K. X-ray diffraction data were measured at either the National Synchrotron Light Source beamline X9A using a marCCD detector or in-house using CuK radiation from a Rigaku RU-H3R X-ray generator equipped with Osmic Blue optics and an R-Axis IV++ image plate detector. The data were reduced with HKL31 (Table 2). Crystals of PC-TP diffracted to varying resolutions and invariably showed anisotropy, with the weakest diffraction in the c-direction.
| | |
Primary phasing and structure refinement.
Primary phases for a form I selenomethionyl containing crystal were determined by the method of MAD using data measured at three wavelengths (Table 2). The amino acid sequence of human PC-TP contains seven methionine residues. Using SnB v2.1 (ref. 32), 10 selenium positions were located and arranged as five identical sites in each of two molecules in the asymmetric unit. The 2.9 Å resolution electron density map was calculated using MAD phases from MLPHARE33, yielding a figure-of-merit of 0.39, and subsequently solvent flattened with RESOLVE34. This map supported model building of a single polypeptide chain using O35. Because higher resolution data were measured from a form II crystal and these crystals contained just one molecule in the asymmetric unit, a preliminary model corresponding to the form I crystal was located in the form II unit cell by molecular replacement for subsequent model building and refinement. Simulated annealing and least squares refinements were carried out in CNS36, setting aside 5% of the X-ray diffraction data for cross-validation37. The refinement of the original PC-TP and SeMet-PC-TP form I crystals was resumed at 2.9 Å using the newly refined form II crystal coordinates while maintaining the same set of reflections for cross-validation. Finally, molecular replacement was used to solve the structure of a form I crystal of PC-TP in complex with PLPC. The three atomic models of PC-TP lack residues 1−7 and 211−214 for which there was weak or absent electron density. A Ramachandran plot places all but two residues in the most favored or additionally allowed regions38,
39.
Assay of native and mutant PC-TPs.
A QuikChange mutagenesis kit (Stratagene Cloning Systems) was used to create the C63A and C63S site-directed mutations in a PC-TP cDNA that was cloned into pET11a. The mutations were confirmed by sequencing, and the vectors were transformed into Escherichia coli BL21(DE3). The specific activities of native and mutant PC-TPs in crude cytosolic extracts28 were determined by measuring the rates of protein-mediated transfer of [14C]PtdCho from small unilamellar vesicles to multilamellar vesicles40 and then normalizing to expression levels that were quantified by densitometric analyses of western blots.
Coordinates.
Coordinates for native PC-TP in complex with DLPC, SeMet PC-TP in complex with DLPC and native PC-TP in complex with PLPC have been deposited in the Protein Data Bank (accession codes 1LN1, 1LN2 and 1LN3, respectively).
Received 5 April 2002; Accepted 6 May 2002; Published online: 10 June 2002.
REFERENCES
- Smith, R. & Tanford, C. J. Mol. Biol. 67, 75−83 (1972). | Article | PubMed | ISI | ChemPort |
- Wirtz, K.W. & Zilversmit, D.B. J. Biol. Chem. 243, 3596−3602 (1968). | PubMed | ISI | ChemPort |
- Geijtenbeek, T.B., Smith, A.J., Borst, P. & Wirtz, K.W. Biochem. J. 316, 49−55 (1996). | PubMed | ISI | ChemPort |
- Cohen, D.E. & Green, R.M. Gene 163, 327−328 (1995). | Article | PubMed | ISI | ChemPort |
- van Helvoort, A. et al. Proc. Natl. Acad. Sci. USA 96, 11501−11506 (1999). | Article | PubMed | ChemPort |
- Wu, M.K., Boylan, M.O. & Cohen, D.E. Gene 235, 111−120 (1999). | Article | PubMed | ISI | ChemPort |
- Cohen, D.E., Green, R.M., Wu, M.K. & Beier, D.R. Biochim. Biophys. Acta 1447, 265−270 (1999). | Article | PubMed | ISI | ChemPort |
- Wirtz, K.W. Annu. Rev. Biochem. 60, 73−99 (1991). | Article | PubMed | ISI | ChemPort |
- Baez, J.M., Barbour, S.E. & Cohen, D.E. J. Biol. Chem. 277, 6198-6206 (2002). | Article | PubMed | ISI | ChemPort |
- Iyer, L.M., Koonin, E.V. & Aravind, L. Proteins 43, 134−144 (2001). | Article | PubMed | ISI | ChemPort |
- Ponting, C.P. & Aravind, L. Trends Biochem. Sci. 24, 130−132 (1999). | Article | PubMed | ISI | ChemPort |
- Ponting, C.P., Schultz, J., Milpetz, F. & Bork, P. Nucleic Acids Res. 27 229−232 (1999). | Article | PubMed | ISI | ChemPort |
- Schultz, J., Milpetz, F., Bork, P. & Ponting, C.P. Proc. Natl. Acad. Sci. USA 95, 5857−5864 (1998). | Article | PubMed | ChemPort |
- Tsujishata, Y. & Hurley, J.H. Nature Struct. Biol. 7, 408−414 (2000). | Article | PubMed | ISI | ChemPort |
- de Brouwer, A. Functional studies on the phosphatidylcholine transfer protein. Ph.D thesis, Utrecht Univ. (2002).
- Gajhede, M. et al. Nature Struct. Biol. 3, 1040−1045 (1996). | Article | PubMed | ISI | ChemPort |
- Yoder, M.D. et al. J. Biol. Chem. 276, 9246−9252 (2001). | Article | PubMed | ISI | ChemPort |
- de Brouwer, A.P.M. et al. Chem. Phys. Lipids 112, 109−119 (2001). | Article | PubMed | ISI | ChemPort |
- Burley, S.K. & Petsko, G.A. Science 229, 23−28 (1985). | PubMed | ISI | ChemPort |
- Gallivan, J.P. & Dougherty, D.A. Proc. Natl. Acad. Sci. USA 96, 9459−9464 (1999). | Article | PubMed | ChemPort |
- Bellamy, H.D., Lim, L.W., Mathews, F.S. & Dunham, W.R. J. Biol. Chem. 264, 11887−11892 (1989). | PubMed | ISI | ChemPort |
- Sussman, J.L. et al. Science 253, 872−879 (1991). | PubMed | ISI | ChemPort |
- Nielsen, P.R. et al. Nature 416, 103−107 (2002). | Article | PubMed | ISI | ChemPort |
- Jacobs, S.A. & Khorasanizadeh, S. Science 295, 2080−2083 (2002). | Article | PubMed | ISI | ChemPort |
- Wirtz, K.W., Devaux, P.F. & Bienvenue, A. Biochemistry 19, 3395−3399 (1980). | Article | PubMed | ISI | ChemPort |
- Johnson, L.W. & Zilversmit, D.B. Biochim. Biophys. Acta 375, 165−175 (1975). | Article | PubMed | ISI | ChemPort |
- McLean, L.R. & Phillips, M.C. Biochemistry 20, 2893−2900 (1981). | Article | PubMed | ISI | ChemPort |
- Feng, L., Chan, W.W., Roderick, S.L. & Cohen, D.E. Biochemistry 39, 15399−15409 (2000). | Article | PubMed | ISI | ChemPort |
- Bose, H.S., Sugawara, T., Strauss, J.F. III & Miller, W.L N. Engl. J. Med. 335, 1870−1878 (1996). | Article | PubMed | ISI | ChemPort |
- Chan, W.W., Roderick, S.L. & Cohen, D.E. Biochim. Biophys. Acta 1596, 1−5 (2001). | ISI |
- Otwinoski, Z. & Minor, W. Methods Enzymol. 276, 307−326 (1997).
- Weeks, C.M. & Miller, R. J. Appl. Crystallogr. 32, 120−124 (1999). | Article | ISI | ChemPort |
- Collaborative Computing Project, Number 4. Acta Crystallogr. D 50, 760−763 (1994). | Article | ISI |
- Terwilliger, T.C. Acta Crystallogr. D 56, 965−972 (2000). | Article | PubMed | ISI |
- Jones, T.A., Zou, J.-Y., Cowan, S.W. & Kjeldgaard, M. Acta Crystallogr. A 47, 110−119 (1991). | Article | PubMed | ISI |
- Brünger, A.T. et al. Acta Crystallogr. D 54, 905−921 (1998). | Article | PubMed | ISI |
- Brünger, A.T. Nature 355, 472−475 (1992). | Article | ISI |
- Ramachandran, G.N., Ramakrishnan, C. & Sasisekharan, V. J. Mol. Biol. 7, 95−99 (1963). | PubMed | ISI | ChemPort |
- Laskowski, R.A., MacArthur, M.W., Moss, S.D. & Thornton, J.M. J. Appl. Crystallogr. 26, 283−291 (1993). | Article | ISI | ChemPort |
- Feng, L. & Cohen, D.E. J. Lipid Res. 39, 1862−1869 (1998). | PubMed | ISI | ChemPort |
- Evans, S.V. J. Mol. Graph. 11, 134−138 (1993). | Article | PubMed | ISI | ChemPort |
- Kraulis, P.J. J. Appl. Crystallogr. 24, 946−950 (1991). | Article | ISI |
- Bacon, D.J. & Anderson, W.F. J. Mol. Graph. 6, 219−220 (1996). | Article |
- Merrit, E.A. & Murphy, M.E.P. Acta Crystallogr. D 50, 869−873 (1994). | Article | PubMed |
Acknowledgments
This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases and the Alexandrine and Alexander L. Sinsheimer Fund. W.W.C. is the recipient of an American Gastroenterological Association Summer Student Fellowship. M.W.V. was supported by a grant from the National Institute of Allergy and Infectious Diseases. We would like to acknowledge the assistance of M. Sullivan at NSLS beamline X9A of the Brookhaven National Laboratories.
Competing interests statement:
The authors declare that they have no competing financial interests. |
