Primary Active Transport Systems

Chapter 6

Primary Active Transport Systems We saw in Chapter 5 that counter- and cotransport of many metabolites into and out of the living cell are brought about by coupling of such flows with the flow of sodium or of hydrogen ions. By these means, the electrochemical gradient of the pumped metabolite is coupled to the existing gradient of the sodium ion or proton. But how is the electrochemical gradient of the sodium or proton itself established? In this chapter we shall see that these ions are pumped out of the cell or organelle by primary active transport systems that use metabolic energy directly. We shall also see that the source of metabolic energy in the different primary active transporters may be the splitting of adenosine triphosphate (ATP), the harnessing of the flow of electrons in an oxidation
reduction reaction, or the absorption of the energy of a photon in the light-driven pumps. We shall start our discussion by considering the sodium pump of animal cells and continue with a consideration of several other, closely related cation pumps. Then we discuss the ion pumps of bacterial cells, mitochondria, chloroplasts, and the light-driven pumps, and finally, the multidrug resistance (MDR) transporters.

6.1 THE SODIUM PUMP OF THE PLASMA MEMBRANE 6.1.1 The Function of the Sodium Pump We all know that human red blood cells can be withdrawn from the body, stored under refrigeration for several weeks, and then transfused into a patient where they will function adequately in maintaining life processes in the recipient (i.e., if the blood donor and recipient have compatible blood types). After many years of careful exploration, medical researchers worked out the best methods for storing blood. Blood cells freshly drawn from the body have a sodium ion concentration that is about one-tenth that of the extracellular medium, the blood plasma. In contrast, their potassium ion concentration is more than 10-fold that of the plasma. In cold storage, the cells lose potassium and gain sodium ions, until cellular and extracellular levels of these ions become almost equal, at about 15 mM for potassium and about Channels, Carriers, and Pumps. DOI: http://dx.doi.org/10.1016/B978-0-12-416579-3.00006-X © 2015 Elsevier Inc. All rights reserved.

247

248 | Channels, Carriers, and Pumps

Cation content in mequiv/5 mM of hemoglobin

120 mM for sodium (Figure 6.1, points at time zero). If the cells are then incubated in the presence of a metabolizable substrate, sodium is ejected from the cell and potassium taken up until the ion concentrations normally present in the cell are restored (Figure 6.1, lines marked “potassium added”). Thus cells for transfusion must be stored with an adequate supply of metabolizable substrate and then incubated before transfusion takes place. Now, the subsequent movement of both sodium and potassium ions is against their concentration gradients. Hence, both cannot be driven by the mutual coupling of their flows. Something other than an antiport or symport is required here. We find that the system needs the direct consumption of metabolically produced ATP for the pumping of both ions. The system is thus a primary active transporter, the so-called sodium pump. A major step forward in the study of this ion pump came from an experiment performed by the Hungarian biochemist, George Gardos. If red blood cells are swollen osmotically (see Section 2.7), pores open up in the plasma membrane. The cells lose their internal hemoglobin and, of course, their red color. They are then termed “ghosts.” During the process of swelling of the cells, as the hemoglobin leaves, normally nonpermeable substances can be induced to enter the cell through the pores. Gardos loaded red cells in this way with ATP and arranged the ion concentrations of the cells so that sodium was 120

100

Na

80 Potassium added

60

Control

40 K

20 Time of addition

0

2 4 6 Hours at 37°C

8

FIGURE 6.1 Sodium and potassium pumping across red cell membranes. Ion movements (sodium efflux and potassium influx) from human red blood cells stored at 2 C for 9 days and then incubated at 37 C with a metabolizable substrate (inosine). The circles show data for sodium; the triangles, for potassium. Filled symbols indicate incubations in the presence of external potassium (at 21 mM); empty symbols, in absence (controls). When fresh, these cells contained about 100 mM potassium and 40 mM sodium ions. Taken, with permission, from R. L. Post et al. (1957). Biochim. Biophys. Acta 25, 118
128.

Primary Active Transport Systems Chapter 6

|

249

at the same level at the two faces of the membrane, as was potassium. Gardos showed, by direct measurement, that the cells hydrolyzed the ATP as they pumped sodium out of, and pumped potassium into, the cell. ATP added outside the cell could not be used; only that present at the intracellular face of the cell membrane could. Ian Glynn showed that pumping out of sodium at an appreciable rate requires the presence of potassium ions outside the cell. (Evidence for this is shown in Figure 6.1.) Careful quantitative stoichiometric studies showed that three sodium ions are pumped out of the cell and two potassium ions are pumped in for each molecule of ATP that is split. Finally it was shown that the split products of the ATP, namely, adenosine diphosphate (ADP) and inorganic phosphate ion (Pi), are liberated at the intracellular face of the membrane. Pumping requires that sodium and ATP be present inside the cell, potassium outside, and under these conditions hydrolysis of ATP brings about the coupled ion pumping. Now this hydrolysis of ATP is the typical action of an ATPase enzyme. The finding that the pump splits ATP in the presence of sodium and potassium ions meant that cell membranes must contain an ATPase that is activated by sodium and potassium. Sure enough, the Danish physiologist Jens Christian Skou soon found such an enzyme, originally isolating it from crab nerves (although it was later shown to be present in nearly all animal cell plasma membranes). This enzyme is the sodium-, potassium-activated ATPase (the Na1, K1-ATPase, also known as NKA). An important advance was made by Hans Schatzmann, who showed that the enzyme could be totally, and specifically, inhibited by the drug ouabain, found in the seeds of the plant Strophanthus. Ouabain is one of the glycosides, used clinically today in the treatment of heart conditions. (Another of this family is digitalin, found in the foxglove plant Digitalis.) The identification of this powerful inhibitor meant that experiments could be easily conducted to study the role of the pump enzyme, since its action could always be blocked by ouabain. One such experiment was performed by Robert Post. He showed that cell membranes, isolated from dog kidney, became phosphorylated when they were incubated with ATP (radioactively labeled in the terminal phosphate residue) in the presence of sodium ions. This transfer of labeled phosphate from ATP to the membrane proteins was inhibited by ouabain and hence a function of the Na1,K1-ATPase. The addition of potassium to the phosphorylated membranes released from them their covalently bound phosphate, the process again being inhibited by ouabain. This meant that the ATP splitting must be taking place in two steps, each step being activated by one of the two cations—phosphorylation by sodium, dephosphorylation by potassium—to give an overall hydrolysis of ATP, activated by the combination of Na and K. Post found that treatment of kidney membranes with the sulfhydryl reagent N-ethylmaleimide (NEM), an inhibitor of this ATPase, enabled the two steps to be functionally dissected. As Figure 6.2 shows, phosphorylated membranes in the absence of NEM were, of course, dephosphorylated by potassium

250 | Channels, Carriers, and Pumps 100

32P-intermediate

(%)

X + X K

ADP

Add ADP or K+ 10

ADP K+

Native

NEM-blocked

1 0

5 Seconds after unlabeled Mg-ATP

10

FIGURE 6.2 Dephosphorylation of phosphorylated Na1,K1-ATPase. (Blue) Dephosphorylation of the native Na1,K1-ATPase enzyme from guinea pig kidneys at 0 C. The enzyme was first phosphorylated from Na1 and radioactively labeled ATP, and then the breakdown of the phospho intermediate visualized by adding excess unlabeled ATP. At the arrow, ADP, K1 or nothing was added. (Red) Dephosphorylation of the NEM-blocked enzyme. As above, but the enzyme was first treated with N-ethylmaleimide (which, this experiment suggested, blocks the interconversion of two conformations of the enzyme). Modified, with permission, from Post et al. (1969). J Gen Physiol 54, S306
S326. Reproduced by copyright permission of The Rockefeller University Press.

(Figure 6.2, blue). But if NEM was added (Figure 6.2, red), potassium failed to activate dephosphorylation, and only by adding ADP could the phosphate be released from the enzyme (and transferred back to reform ATP!). Further, NEM blocked the transformation of the enzyme to the potassiumsensitive form. Post hypothesized that two conformations of the enzyme must exist. The first conformation to form is ADP-sensitive, the second is potassiumsensitive, while NEM paralyzes the transformation of the first to the second. These two forms of the phosphorylated pump enzyme are now known as E1P and E2P (P symbolizing the covalently bound phosphate residue). Whereas E1P is ADP-sensitive, E2P is potassium-sensitive. Identification of these two forms enabled Post, together with Albers, to propose a scheme for linking the biochemical events of ATP splitting and the transport events of sodium
potassium exchange. A more updated version of their scheme, known as the Albers
Post model, is depicted in Figure 6.3. In this scheme, E1 and E2 are two forms of the pump enzyme in which the cation-binding sites face the intracellular surface and the extracellular surface of the membrane, respectively. In the presence of intracellular sodium ions (three in number) and ATP, E1 is phosphorylated, releasing ADP and yielding E1PNa. This then undergoes a conformation change to yield E2PNa, the form in which the cation-binding sites face the extracellular surface. The three sodium ions dissociate from this form to yield E2P and two

Primary Active Transport Systems Chapter 6 E1P.Na

E2P.Na

E1ATP.Na

E2P

E1

E2P.K

E1.K

E2.K

|

251

FIGURE 6.3 Kinetic scheme for Na1,K1-ATPase. The Albers
Post scheme for the action of the sodium
potassium pump. The symbol E is the pump enzyme; Na and K, the cations; ATP and P, the phospholigands. Symbols 1 and 2 represent the conformations with sodium-binding sites facing the cytoplasm and the extracellular phases, respectively. The single-headed arrows represent conformational changes of the enzyme. The double-headed arrows represent pumpligand association/dissociation equilibria.

potassium ions from the extracellular surface take their place, yielding E2PK. This form now undergoes dephosphorylation to E2K, finally interconverting to E1K. The potassium ions, now bound to a site that is accessible from the intracellular surface, dissociate and release free E1 once more, to restart the whole cycle. In a single cycle, three sodium ions are transferred from the intracellular to the extracellular medium, two potassium ions are transferred in the opposite direction, while a molecule of ATP is split to ADP and Pi. The movements of sodium and potassium and the splitting of ATP are strictly coupled. We will discuss in section 6.2.2 the details of how the coupling is controlled. This has been a crucial problem for later investigators to solve. The two forms, E1 and E2—whether phosphorylated or not—differ in conformation. Thus the two behave differently toward proteolytic digestion. Peter Leth Jørgensen from Denmark showed that the former is split in two places by chymotrypsin, the latter only in a third site, coinciding with neither of the first two. Steven Karlish showed that the two forms differed with respect to fluorescent properties, a finding that has enabled investigators to undertake many studies on the rates of interconversion between the two conformations. We now know that the scheme of Figure 6.3 is a simplification of the true picture since a major, intermediate form that is bound to potassium has its cation-binding and cation-releasing sites accessible at neither side of the membrane. The potassium in this form is said to be “occluded” within the enzyme. Sodium can also be “occluded” in E1PNa, the left-hand form of the two sodium-bound forms shown in Figure 6.3. Interestingly, the Na1,K1-ATPase recognizes an analog of phosphate, the anion vanadate, at the site on the enzyme that normally binds phosphate. This vanadate binds to the enzyme relatively well, but its dissociation from the enzyme, under the influence of potassium, is exceedingly slow. Hence, once bound, it remains bound to the enzyme, blocking the phosphate-binding

252 | Channels, Carriers, and Pumps

center, and inhibiting the enzyme from taking part in any further cycle of phosphorylation and dephosphorylation. It seems that all those cationactivated ATPases that demonstrate an enzyme intermediate, with phosphate bound to the enzyme, can be inhibited by vanadate in such a nearly irreversible reaction. (We will see in a later section of this chapter that members of another class of membrane pumps, the ABC transporters are also blocked by vanadate.) This family of enzymes, of which we discuss a few other members in the following sections, is referred to as the vanadate-inhibitable ATPases, or following Peter Pedersen and Ernesto Carafoli’s suggestion: “PATPases”—since they form a phosphorylated intermediate during their action (Box 6.1). Just as we saw for the carrier (Section 4.2) and for the anti- and symporters (Sections 5.2.1 and 5.3.2), one can study the kinetics of the sodium and potassium transport on this primary transporter. One can measure a Michaelis constant Km for each of the cationic substrates, sodium and potassium, and for the phosphate derivatives, ATP, ADP, and Pi, where binding is studied for each of

BOX 6.1 Additional Modes of Action of the Sodium Pump Other major features of the pump enzyme that can be inferred directly from Figure 6.3 concern the ability of the enzyme to catalyze the exchange of intracellular and extracellular sodium, and of intracellular and extracellular potassium. Both these reactions involve the reversal of partial reactions of the pump enzyme. Consider Figure 6.3 carefully: If the pathways involving sodium are reversible, then when no potassium is present to bind to the enzyme, sodium ions will be exchanged across the membrane (as Patricio Garrahan and Ian Glynn showed in the experiment depicted as the lower curve in Figure 6.4). This reaction needs the presence of ATP and ADP (the latter to accept P from E1PNa) but results in no overall splitting of the ATP. Similarly, potassium ions (when little or no sodium is present) can be exchanged in a reaction that is speeded up by ATP and Pi but results in no hydrolysis. Part of the upper curve in Figure 6.4 reflects this flux. Some other properties of the system do not follow from Figure 6.3 and, indeed, show that the scheme is an oversimplification. There is a slow, but definite, exchange of potassium ions in the absence of any phosphate or possibility of phosphorylation, and there is also a net transport of potassium ions in such conditions. Finally, there is a slow hydrolysis of ATP in the complete absence of potassium ions, accompanied by the transfer of sodium from intracellular to extracellular surface. This means that there is a route for the (slow) breakdown of E2P in the absence of potassium ions (as can be seen, indeed, in Figure 6.2). All these reactions are true “slippages” of the pump enzyme. (See Section 5.2.2 for “slippage” on co- and counter-transporters.) (Continued )

Primary Active Transport Systems Chapter 6

|

253

BOX 6.1 (Continued)

Flux as fraction of its maximal value

1.0

Ouabain-sensitive K influx

0.8

0.6

0.4

0.2 Ouabain-sensitive Na influx 0 0 1 2 3 4 5 10 15 External K concentration (mM)

20

FIGURE 6.4 Pump-mediated cation exchanges across red cell membrane. Fresh human red blood cells were incubated at 37 C in media containing labeled potassium (upper curve) or sodium (lower curve) and the ouabain-sensitive component of the cation influxes measured as a function of the external potassium concentration. The upper curve represents sodium
potassium exchange by the pump; the lower curve, sodium
sodium exchange. Taken, with kind permission, from P. J. Garrahan and I. M. Glynn (1967). J. Physiol. (London) 192, 189
216.

the two forms E1 and E2. (Note that ATP, ADP, and Pi can bind to each form, E1 and E2, although their binding sites exist only at the intracellular face of the membrane.) The Na1,K1-ATPase is quite clearly asymmetric. The form E1 has a high affinity for sodium and for ATP and ADP, but a low affinity for potassium and Pi. In contrast, E2 has a high affinity for potassium and Pi, but a low affinity for sodium, ATP, and ADP. The high affinity of E1 for sodium and low affinity for potassium enable the pump to bind sodium in preference to potassium at the intracellular face (where the binding sites of E1 are accessible) and this maximizes the rate of pumping of sodium out of the cell. Similarly, the high affinity of E2 for potassium and low affinity for sodium enables the pump to bind potassium preferentially at the extracellular face and maximizes the rate of inward pumping of potassium.

254 | Channels, Carriers, and Pumps

How does the splitting of ATP drive the active transport of the ions? It does so as a result of the strict coupling that occurs between the conformational interchanges of E1 and E2, and the splitting of ATP. Coupling ensures that a complete cycle of ion transport and ATP splitting takes place only when (i) the sodium ions from the intracellular face become bound to E1, (ii) the conformation alters to E2, (iii) at the extracellular face potassium exchanges for the sodium ions, and finally (iv) the conformation returns to E1, bringing potassium to the intracellular face. Each time that a net breakdown of ATP and ADP and Pi occurs, the exchange of sodium and potassium has to occur. The ATP is a “high-energy” compound in that the equilibrium between ATP and ADP 1 Pi is well over on the side of the splitting of ATP at the usual levels of ATP, ADP, and Pi in the cell. Thus, a reaction leading to hydrolysis of ATP tends to occur. If this reaction occurs on the pump, the exchange of sodium with potassium will, under physiological conditions, be coupled with it, in the direction of sodium efflux and potassium influx (Figure 6.3). But this overall reaction, like any chemical reaction, is, in principle, reversible. If the resources of ATP in the cell are low in comparison with the concentrations of ADP and Pi, the reaction of ATP hydrolysis might not be so favorable, energetically. If, in addition, extracellular levels of sodium are high and those of potassium intracellularly are high, a reversal of pumping will occur. Sodium will enter the cell and potassium will leave it, on the pump. The direction of travel around the scheme of Figure 6.3 will be reversed and ATP synthesis will occur. Thus it is ATP splitting, driving the interconversion of E1 and E2 that brings about pumping. The change of affinity for the cations during this interconversion speeds pumping but does not drive it. We can be somewhat more precise about how these changes of affinity for the cations are brought about. We have shown in Section 4.6.2, where we considered the asymmetry of carriers, that a transporter can exist in two conformational states that differ in free energy. The binding of a substrate will appear to be of low affinity when it binds to the form of lower free energy. This enables us to understand the affinity changes that the sodium
potassium pump undergoes. The change of affinity for potassium is brought about without any hydrolysis of the ATP. (Indeed, ADP is almost as good as ATP at doing this.) ATP binds strongly to the E1 form of the enzyme, but poorly to the E2 form. The mere binding of ATP stabilizes the E1 form to which potassium ions are therefore less strongly bound. For sodium ions, Figure 6.3 reveals that sodium binds to E1 together with ATP, but to E2 where this form is bound to Pi. Thus the two forms to which sodium binds have a different free energy, differing by the free energy of ATP hydrolysis, and E2P is the form of low free energy. This form, therefore, has a low affinity for sodium ions. (Box 6.2 discusses the electrogenic nature of the Na1,K1-ATPase.)

Primary Active Transport Systems Chapter 6

|

255

BOX 6.2 Electrogenicity of the Na1,K1-ATPase Since the Na1,K1-ATPase pump extrudes three sodium ions for every two potassium ions that it pumps inwards, there is a net efflux of one positive charge during each turn of the pump cycle. The pump is, therefore, electrogenic. The actual amount of current flow is given by the product of (1) the number of pumps in the cell membrane surface, (2) the rate of turnover of the pump (the number of cycles it performs per unit time), and (3) the charge on the electron (1.60 3 10219 C). For pumps from many cell membranes, the maximum velocity of pumping is some 175 ATP molecules split per second at 37 C. In the human red blood cell about 200 pumps are present in the 137 μm2 surface area of the cell. This gives rise to a maximum current flow of 200 3 175 3 1.6 3 10219 or 5.6 3 10215 A, or a current of 2.8 3 10215 A when the pump is working at physiologic levels of ATP, potassium, and sodium at the inner face and of potassium and sodium at the extracellular face. The resistance of the red cell membrane is approximately 1.1 3 1011 Ω. Thus a voltage of 20.3 mV (volts 5 [amperes 3 ohms]) will be established by the pump in physiological conditions. In comparison with the prevailing membrane potential of 29 mV for the red cell, this is a very small contribution to the overall potential. In general, such seems to be the case for the sodium pump in many cell membranes. In squid axon, for instance, the pump seems to contribute only some 5 mV for a total transmembrane potential of some 60 mV. Only in very specialized cells is the electrogenic contribution of the sodium pump significant. (Full details of such measurements can be found in the book “Electrogenic Transport” edited by M. Blaustein and M. Lieberman, Raven Press, New York, 1984.)

6.2 THE CALCIUM PUMP OF SARCOPLASMIC RETICULUM The contraction and relaxation of muscle are controlled by the cytoplasmic concentration of free calcium ions. Calcium is stored within a cytoplasmic organelle, the sarcoplasmic reticulum (SR). In response to a signal transmitted to the tubular membrane system of the muscle across the nerve muscle end plate (see Section 3.2), calcium is rapidly released from the SR and binds to a protein, troponin, in an event that initiates contraction of the muscle (Figure 6.5). Relaxation of the muscle follows when the cytoplasmic calcium concentration is lowered again, under the action of the calcium pump of the SR, a system studied originally by Wilhelm Hasselbach and Anne-Marie Weber. These researchers demonstrated the presence of an ATPase in the SR membrane that very effectively pumps calcium from the cytoplasm into the lumen of the SR vesicle. The SR membrane is packed tightly with molecules of this Ca21-ATPase, which constitutes its major protein. This high concentration of the pump enzyme makes the SR a very handy object for obtaining pure preparations of a cation pump, ensuring that the calcium pump from the SR membrane has been an object of intense research over the years.

256 | Channels, Carriers, and Pumps Tubules

Tubules

SR

SR

(A)

(B)

FIGURE 6.5 Calcium controls the relaxation
contraction cycle of muscle. The upper half of the figure shows schemes for the tubular membrane system of muscle with attached sarcoplasmic reticulum (SR) vesicles (indicated by ellipses). The lower half of the figure shows the actomyosin bridge complexes of striated muscle. On the left (A) the muscle is relaxed, with calcium (solid circles) sequestered within the SR. (B) The cell membrane has depolarized, and calcium has left the SR and become attached to the muscle proteins, bringing about the contraction step. Relaxation back to (A) occurs when the SR can once again pump the calcium ions back into itself on the ATP-driven pump.

The calcium pump of the SR has been a favorite object for attempts at crystallization of pump protein, and the direct visualization of the structure of the pump by electron microscopy. Indeed, calcium ATPase was the first member of the P-type ATPases for which a molecular structure could be obtained, as we shall see in section 6.2.1 that follows. The calcium pump-enzyme is very similar in its properties to the Na1, 1 K -ATPase that we discussed in Section 6.1. It can exist in two major conformations, and just as in the Na1,K1-ATPase, the two forms give different fluorescent signals. Thus rates of interconversion of the two conformations can be measured. The absolute rates are rather similar to those of the Na1, K1-ATPase. Calcium binds to the E1 form with high affinity, as this is the form in which calcium-binding sites face the cytoplasmic surface of the SR vesicle. See Box 6.3 for more on these binding sites. The E2 form binds calcium with far lower affinity. However, ATP binds to the E1 form with high affinity, Pi with low affinity, while ATP binds to E2 with a low affinity, and Pi with high. The enzyme is phosphorylated when calcium binds to Ei, following addition of ATP (see the curve labeled EP in Figure 6.6). There is, however, no equivalent of the potassium ion to activate the dephosphorylation of the pump enzyme, a step that occurs spontaneously at a rapid rate

Primary Active Transport Systems Chapter 6

|

257

BOX 6.3 Site-Directed Mutagenesis for Delineation of the Calcium-Binding Groups of Ca21-ATPase The cloning of the gene for the SR calcium ATPase has enabled Giuseppe Inesi and David MacLennan, with their colleagues, to use site-directed mutagenesis to determine what amino acid side chains are present at the calcium-binding sites of the enzyme. Complementary DNA cloned from the rabbit fast-twitch muscle Ca21-ATPase gene was transfected into a line of mouse kidney cells grown in cell culture. These cells then expressed the ATPase in their microsomal membrane fraction at levels which could reach 50-fold that of the control, nontransfected cells. Mutations were introduced into the DNA at known positions in the sequence. In all, over 24 mutations were introduced (one at a time) into the enzyme structure. For each mutation, MacLennan and colleagues measured its effect on the ability of calcium to stimulate phosphorylation of the enzyme by ATP. (See Figure 6.2 for phosphorylation of the sodium pump and Figure 6.6 for phosphorylation of the calcium pump.) Mutations at six positions in the sequence (at Glu-309, Glu-771, Asn796, Thr799, Asp-800, and Glu-908) prevented calcium from stimulating phosphorylation of the enzyme by ATP, but did not prevent phosphorylation of the enzyme by phosphate in the absence of calcium. This latter observation was an important control in that it showed that the site-directed mutations were not upsetting the whole structure of the enzyme and preventing it from undergoing any phosphorylation reaction. These six sites are in regions of the enzyme that the hydropathy plot suggests are in hydrophobic regions. We cannot be sure that all six of these residues are indeed at the calcium-binding sites, but the evidence is suggestive and the technique capable of much further application. In another study, also using site-directed mutagenesis, MacLennan and colleagues showed that the socalled stalk region of the molecule is not involved in calcium binding. Thirty of the carboxyl groups in this region of the molecule were replaced without significantly affecting the ability of calcium to be transported into SR vesicles. The recent structural studies (section 6.2.1) confirm these early observations.

(the curve labeled Pi in Figure 6.6). Figure 6.6 also shows the rate of uptake of calcium ions into the SR vesicles (the curve marked Ca21). Note that the slope of this curve is twice that of the rate of formation of Pi, i.e., of the rate of splitting of ATP. It follows that the stoichiometry of calcium ions pumped to ATP molecules split is 2:1. There is more to be learned from a careful study of this important experiment. Consider the intercept to zero time of the curves for Ca21 and Pi uptake. Almost twice as much calcium as phosphate is bound very rapidly to the enzyme. This is calcium that is not taken up within the vesicles, but is rather bound to the enzyme itself. This is “occluded” calcium (see Section 6.1.1 for a similar finding for the sodium pump). Clearly, two ions of calcium are occluded per molecule of the calcium pump. (Interestingly, detailed studies by Inesi of this occlusion show that of the two calcium ions that are bound to the pump at the cytoplasmic

258 | Channels, Carriers, and Pumps 30

n mol/mg protein

Ca2+ 20

Pi 10 EP 0 0

200

400

600

MSEC FIGURE 6.6 Phosphorylation and calcium uptake by the SR Ca21-ATPase. Formation of the phosphoenzyme (filled triangles), phosphate liberation (open circles), and calcium uptake (filled circles) were measured with SR membranes at 25 C. At zero time 100 μM ATP was added in the presence of KCl and calcium ions. From the ratio of the slopes of the upper two curves, the stoichiometry of phosphate liberated and Ca21 can be seen to be 1:2. Taken, with kind permission, from G. Inesi et al. (1982). Ann. N. Y. Acad. Sci. 402, 515
532.

face, the one that is bound first is the first to be released at the extracellular surface after the protein’s conformation change has occurred. Thus the calcium ions seem to “glide” through the protein interior!) Vanadate, as in the Na1,K1-ATPase, binds the calcium pump of SR membranes in place of phosphate to form a stable complex, inhibiting further action of the enzyme. There is only one polypeptide chain in the enzyme, in contrast to the Na1,K1-ATPase. The calcium pump, as we saw, occludes the calcium as the ion is transferred across the membrane, just as the Na1,K1ATPase binds its substrates—sodium and potassium. The strong analogy between the calcium pump and the sodium
potassium pump points to a common mechanism of action. Of course, both pumps are driven by the act of ATP splitting, and the reaction of splitting is strongly favored at the prevailing concentrations of ATP, ADP, and Pi in the cell. Hence, the cycling of the calcium pump in the direction of removal of calcium from the cytoplasm is favored. This pump is, again, reversible, and ATP synthesis from ADP and Pi will occur when levels of calcium are such as to favor this direction of cycling. It is the strict coupling of calcium transport and ATP splitting that drives transport. The fact that the calcium pump changes its affinity from high to low as it alters conformation from E1 (with binding sites facing the cytoplasm) to E2 (with binding sites facing the lumen of the vesicle) speeds up the transport of calcium but is not the force that drives it.

Primary Active Transport Systems Chapter 6

|

259

FIGURE 6.7 Scheme for the pumping cycle of the calcium ATPase, SERCA1a. The boxed intermediates are those for which a structure is available. Taken, with permission, from Toyoshima and Cornelius. New crystal structures of PII-type ATPases: excitement continues. Curr Opin Struct Biol. 2013;23:507
14. For structures in PDB click on 3W5A, 3W5C, 3W5D, 3AR4, 3AR5.

6.2.1 Structural Studies on the Calcium ATPase (SERCA1a) Just as in the case of the secondary active transporters that we discussed in Chapter 5, so for the primary active transporters, the advent of detailed structural information has transformed the way we can study and understand the molecular movements in transport. Since the first crystallization and determination of the structure of the E1.2Ca21 form of the molecule by Toyoshima and his colleagues in 2000, more than 20 crystal structures have been determined in his and other laboratories. These are sufficient to provide at least one example from nearly every intermediate step along the pumping cycle that was depicted (for the related Na1,K1-ATPase) in Figure 6.3. Consider Figure 6.7 that shows many of the pump cycle intermediates with known structures (the symbols within parentheses show the PDB (Protein Data Bank—http://www.rcsb.org/pdb/home/home.do) references for these solved structures. A trick used to fix these different states into a form that could be crystallized was to co-crystallize the ATPase with inhibitors of transport such as thapsigargin (TG in the figure), adenosine 50 -(beta,gamma-methylene)triphosphate (AMPPCP), or the phosphate analogues, magnesium, aluminum or 22 2 beryllium fluorides (MgF22 4 ; AlF4 ; BeF3 ).

260 | Channels, Carriers, and Pumps

FIGURE 6.8 Four intermediates in the pumping cycle of the Ca21-ATPase. Taken, with permission, from Toyoshima C. Structural aspects of ion pumping by Ca21-ATPase of sarcoplasmic reticulum. Arch Biochem Biophys 2008; 476: 3
11.

A secondary structure model (α-helices-loops and β-strands) of four of these structures is shown in Figure 6.8, Part A and a cartoon view in Figure 6.8, Part B. The four structures shown are representative of the four canonical states of the pump cycle: two E1 forms that interact with calcium ions at the cytoplasmic face and two E2 forms, interacting with Ca at the luminal (or, in other pumps, the external) face. Of these forms, two are analogues of phosphorylated intermediates (E1P 5 E1AlF4.ADP and E2P 5 E2BeF3), two are dephospho forms. The ATPase is seen to be made of a stalk that is held within the membrane and a head that is composed of three domains A, P, and N. Domain A is the actuator (for examples of actuators in the human-scale world, see http://en.wikipedia. org/wiki/Actuator), P is the domain that receives the terminal phosphate group from ATP, while N is the part that binds the nucleotide (ATP or ADP). In Part B of Figure 6.8, the calcium ions are depicted as green spheres, the protons that take their place during the cycle as white spheres (in E2). In Part A, the protein chains are colored from blue at the N-terminal end to red at the C-terminal. (See the legend to Figure 6.7 for the links to PDB.) Almost all parts of the molecule undergo coordinated changes in conformation as the pump cycle takes place. Only the membrane-embedded helices M7 to M10 remain largely unchanged, providing a membrane anchor to the rest of the protein. Helices 1, 4, and 5 undergo large bending and then flexing movements through the cycle and helix 6 rotates. During the cycle, domain A rotates and then returns by some 100 , N approaches A and finally moves away from it, while P buckles and then straightens again.

Primary Active Transport Systems Chapter 6

|

261

Let us start with the ground state E2 in the lower left of the cycle. It still has two protons bound to it (their position in the molecule can only be inferred), P is unbroken and N is closely apposed to A at its amino acid sequence TGES (a sequence that is highly conserved among the cation ATPases). First, a magnesium ion enters, displacing the water molecules that occupy the sites to which calcium ions will subsequently bind. (We discuss the details of this step in the following paragraphs.) Binding of calcium ions from the cytoplasmic face provides the energy to bring about the conformation change to E1  2Ca21, which opens up a gap between the N and A domains, allowing ATP access. The binding of ATP to N then brings N closer to the phosphorylation site at D351 on P. The structure E1  ATP is formed; ATP is still a high-energy molecule. The terminal phosphate of the ATP is transferred to the aspartyl residue (D351), and a spontaneous conformational change brings the molecules to the E2P state, in which the TGES sequence of the A domain approaches the phosphorylated aspartyl residue. The dissociation of the 2 Ca ions at the luminal face releases energy for the transformation to E2P. Subsequently, two protons enter the empty sites and the E2P is hydrolyzed, returning the molecule again to E2. Finally a reverse conformational change returns the molecule to the starting state E1 and a new cycle can take place. The structural information is now so detailed that one can follow the serial, cooperative binding of the two calcium ions that the pump carries across the membrane. Look at Figure 6.9, that shows in the center the state E1  Mg21, an intermediate that forms between E2 and E1  2Ca21 which we saw in Figure 6.8. (The crystal structure of E1  Mg21was determined following the molecule’s crystallization from Mg21-containing solutions.

FIGURE 6.9 In the center: E1  Mg21, an intermediate form in the E2 to E1 transition. Taken, with permission, from Toyoshima C., et al. Crystal structures of the calcium pump and sarcolipin in the Mg21-bound E1 state. Nature. 2013;495:260
4.

262 | Channels, Carriers, and Pumps

The structure was stabilized and its resolution improved by the addition of TNPAMP (20 ,30 -O-(2,4,6-trinitrophenyl)adenosine 50 -monophosphate), a ligand that cross-links the N and P domains.) In Figure 6.9, the magnesium ion is seen as a green sphere (within a pink circle) bound between the stalk helices of E1  Mg21, the two calcium ions being depicted as the brown spheres (in pink circles) in E1  2Ca21. A magnesium ion is small enough to enter E2 and does so, bringing about the rotation of domain A and opening up the gap between A and N, sufficient to allow a single calcium ion to enter and displace the magnesium. It binds at its first binding site, site I. This binding then opens the structure further allowing the second calcium to enter and bind at site II. Note how transmembrane helices M1 and M2 take part in the conformation changes, the two aligning themselves parallel to the anchoring helices M7 through M10 by the time E1  2Ca21 is reached. These re-dispositions of the cation-binding sites can be seen clearly in Figure 6.10, the three subfigures each being a helix-cylinder and stick cartoon of the corresponding figure in Figure 6.9. In E2, the vestibule enclosed by helices M5, M6L (for lumen), M4L, and M8 is filled with water molecules. Magnesium (the cyan sphere) enters, binds to residues N796, E771, and N768 and coordinates the two remaining water molecules. Hence, E1  Mg21is formed. Note how M6L has rotated to bring N796 closer to M4L and D800 to where it will bind to the first calcium ion (the two calcium ions being depicted as blue spheres). This can now enter and bind, displacing the magnesium ion. As it binds, it pulls M4L toward M6L, rotating it in the process and forming the binding site for the second calcium so that E1  2Ca21 is achieved. (The gold-shaded edge-on cylinder in (C) designates the position that M4L occupied in (B), emphasizing the conformation change that occurs.) Binding of the first calcium ion enables the binding of the second. This accounts for the cooperative nature of the association between calcium and the ATPase.

FIGURE 6.10 The figure is designed to delineate the actual amino-acid residues that will, in turn, bind to the cations. The red spheres depict water molecules, the cyan a magnesium ion and the numbered blue spheres, the calcium ions bound at sites I and II. The empty circles in (B) mark the positions that will later be filled by the calcium ions. Taken with permission from Toyoshima and Cornelius (adapted). New crystal structures of PII-type ATPases: excitement continues. Curr Opin Struct Biol. 2013;23:507
14.

Primary Active Transport Systems Chapter 6

|

263

While analyzing their data on the E1  Mg21crystals, Toyoshima and his colleagues found additional matter surrounded by helices M2, M6, and M9, over and above what they expected from the known amino acid composition of the Ca21-ATPase. They surmised that this was perhaps sarcolipin (SLN), a 31-residue polypeptide known to be involved in the regulation of the pump’s activity. Further analysis showed this to be correct. SLN’s position in the molecule is depicted as the pink cylinder in Figure 6.11. Binding of sarcolipin in this way stabilizes the E1 state, thereby decreasing the operative affinity of SERCA for calcium. In so doing it can slow the pumping cycle, inhibiting the enzyme’s action. An analogous molecule, phospholamban (PLN), involved in contraction of the heart, has been more fully studied than

FIGURE 6.11 Sarcolipin (SLN) bound to the E1  Mg21conformation of SERCA1a. SLN is depicted as the pink cylinder, the magnesium ion as a green sphere. Taken, with permission, from Toyoshima and Cornelius. New crystal structures of PII-type ATPases: excitement continues. Curr Opin Struct Biol. 2013;23:507
14.

264 | Channels, Carriers, and Pumps

SLN. When PLN is phosphorylated it can no longer bind tightly to the heart’s Ca21-ATPase, relieving its inhibition of the pump. Before these crystal structures were determined, we could only speculate as to what might be the conformation changes that transport kinetics required us to postulate. Now, we can describe them in a detailed molecular picture.

6.2.2 Structural Studies on the Na1,K1-ATPase The sodium pump (also known as Na1,K1-ATPase) that we learned about in detail earlier in this chapter has been much more difficult to crystallize than its calcium homologue. This is largely because it is more complex, being made of three subunits, the α-chain (ATP1A1), the β-chain (ATP1B1), and an additional FXYD chain, an example of which is phospholemman (FXYD1). Adding to the complexity is the fact that there are numerous versions of each of these subunits, often present in the same tissue, making difficult the preparation of pure material needed for crystallization, and the protein itself is never present in a tissue in the high amounts found for the calcium pump in the muscle’s SR. Thus, instead of 20 different crystal forms having been studied by X-ray diffraction, as in the calcium pump, only 4 are currently (2014) available for the sodium pump. This is enough, however, to elucidate many aspects of its function (Figure 6.12, (i) and (ii)). Note the N, P, and A domains that are quite similar to those of the calcium pump shown in Figures 6.8
6.10, although the structure is somewhat less compact. Clearly seen is the FXYD subunit bound largely to the transmembrane helices and the β-chain, much of which is present bound at the extracellular face of the α-chain, although the transmembrane helix of the β-chain binds to transmembrane helices M7 and M10. In (ii), the ouabain molecule is seen lying bound within the transmembrane stalk region to helices M1, M2, and M5, seen more clearly (as OBN) in the stick model of Figure 6.13. This close apposition between the potassium ions and ouabain explains their well-known interactions. Thus, when the pump is bound to potassium ions, its affinity for ouabain is substantially decreased. From Figure 6.13 it is easy to see that K1 ions bound within the transmembrane stalk region will tighten the structure, making it more difficult for ouabain to bind. This binding of ouabain will, however, block the release of already-bound potassium, keeping the pump in its potassium-bound, E1 form. Conversely, the high affinity binding of ouabain to the pump in the absence of potassium, inhibits potassium binding, stabilizing the pump in the E2P form.

6.2.2.1 A Comparison of the E2 and E1 Conformations of the N1,K1-ATPase The successful structure determination of the sodium-bound E1 conformation of Na1,K1-ATPase in a form bound to ADP and to aluminum fluoride (which

˚ resolution structure of Na1,K1-ATPase in a FIGURE 6.12 Subfigure (i) shows the 2.4 A helix-loop depiction, while (ii) shows part of the molecule in the state bound to the cardiac drug ouabain. (The cytoplasmic face is above, the extracellular below.) Na1,K1-ATPase, taken from the Protein Data Bank under accession numbers 2ZXE and 3A3Y, respectively. Left: FXYD is the regulator molecule, a phospholemman homologue, almost always found together with native Na1,K1-ATPase. “beta” is the β chain. N, P, and A are the Nucleotide binding, Phosphorylation, and Actuator domains, respectively (compare with SERCA1a). Right: the Na1,K1-ATPase bound to ouabain (the view is rotated from that on the left so as to show the ouabain molecule, which is bound within the transmembrane section of the molecule. The two K symbols show the approximate positions of bound potassium ions.

FIGURE 6.13 OBN marks the bound ouabain molecule. Note how the β-chain, signified by β, is bound both from the extracellular face and within the transmembrane stalk. Reprinted, with permission, from Ogawa, Shinoda, Cornelius and Toyoshima. Crystal structure of the sodiumpotassium pump (Na1,K1-ATPase) with bound potassium and ouabain. Proc Natl Acad Sci USA. 2009;106:13742
7.

266 | Channels, Carriers, and Pumps

1 FIGURE 6.14 (A) is the ribbon model structure of the E1UAlF2 4 UADPU3Na form while (B) is a similar model of the E2  MgF422  2K1 form. Figure prepared by Flemming Cornelius and reproduced by his kind permission. Indicated are the α and β chains and the N, A, and P domains (see text). The alpha unit is colored with N-terminus blue changing to orange at the C-terminus. Beta is orange, and FXYD is red (lies behind). In the ion-binding sites, ions are circled with red and numbered. AlF and MgF are red. ADP is in ball and stick, and Mg21-ions are in green. Sugars attached to the beta ectodomain are cyan. The left panel is the shark E2  Pi  2K1 structure (PDB 2ZXE) with FXYD10, the right panel is the pig kidney E1BP  ADP  3Na1 structure (PDB 3WGV) with FXYD2 (the gamma subunit). The three sodium ions are depicted as I, II, and III half-way cross the membrane-spanning portion in (A); the two potassium ions are I and II, somewhat closer to the cytoplasmic face, in (B). Note a major feature: the A domain rotates by some 180 around an axis perpendicular to the membrane plane as the conformation changes from E1 to E2.

stabilizes the molecule by taking the place of inorganic phosphate) has clarified the nature of the sodium-binding sites and led to the understanding of how the E1 form can bind three sodium ions while the E2 form binds just two potassium. The structure can be visualized in the Protein Data Bank at 3WGV. Look first at Figure 6.14 that compares the atomic model for the 1 E1UA1F2 form (a) while (b) is a similar model of the 4 UADPU3Na 1 22 E2  MgF4  2K form: Note how similar are the two conformations but note also that the A domain of the molecule (see earlier in the main text for a description of the three domains A, N, and P) rotates through 100 as the conformation

Primary Active Transport Systems Chapter 6

|

267

FIGURE 6.15 Higher detail versions of the models depicted in Figure 6.14. (A) depicts views of the models parallel to the membrane, (B) perpendicular from the cytoplasmic face. (E1 conformation above, E2 below.) In B, below, the E1 conformation is the fainter yellow in the background—as seen on its own above—while the cyan coloured parts are E2 itself. The red spheres are water molecules. (C) illustrates successive snapshots of the three sodium ions entering into their binding sites. Note that the orientation of the molecule is seen in a different view in (C)— helix 4 is here to the left of 5. Taken, with permission, from Kanai R, Ogawa H, Vilsen B, Cornelius F, Toyoshima C., Crystal structure of a Na1-bound Na1,K1-ATPase preceding the E1P state. Nature. 2013;502:201
6.

changes between E1 and E2. This is associated with a major conformation change in several transmembrane helices (especially 1, 4, 5, and 6). Now look at Figure 6.15 where, in (A) and (B), the model structures of the two conformations are shown in higher detail. Cartoons (A) show the model looking from within the membrane at the cation-binding sites formed by helices 6, 5, and 4. The E1 conformation is shown above and E2 below. Cartoons (B) are as (A) but now looking down from the cytoplasmic face. Note from (A) in this figure that the different disposition of helix 5 (well separated into its cytoplasmic (5C) and extracellular (5E) halves in E1, but not in E2) enables an additional binding site for sodium to be formed (the III site), which is simply not available in E2. Note also that the two binding sites for Na (labeled I and II) in E1 are much closer together than are the corresponding binding sites for K in E2. The region that allows the two sodium ions to bind in E1 cannot accommodate two (bigger) potassium ions nor the two Ca ions that are bound in this general area in the calcium ATPase. Part (B) of the figure shows the details of the amino acid residues that contribute to the ion binding in the two conformations E1 and E2.

268 | Channels, Carriers, and Pumps

To take one example of the differences between the two conformations: glutamic acid D808 from helix 6 is now closer to the ion-binding sites in E1 and coordinates directly with sodium in its site I, whereas it binds an intervening water molecule (the red sphere) in E2. The four sections of Figure 6.15C model the sequence of the binding on the three sodium ions and suggest a basis for the known cooperativity of the sodium ions. In the upper part of (C), all binding sites are empty. This represents the situation after the potassium ions have left their binding sites and ATP has bound to the enzyme. Helix 6 is free to rotate. There is a clear passage for Na to enter site III. Now, Ser 775 (that sits on helix 5), reorients and the sodium ion becomes held in place by the resulting reorientations of Asp 804 and 808 whose carboxyls add to the binding energy. These movements allow the formation of the binding site I by movement of Ala 323. When Na binds to its site II, the resulting movement of D804 allows site II to form, and the bound Na at II is locked in place by the movement of D804, V322, and E327. With all this in place, the P domain can bend and accept Mg21 and phosphate can be transferred from ATP, allowing the pump cycle to continue. Solving the structure of the E1 conformation with that of E2 already known has allowed the differential binding of sodium and potassium (and calcium) to be understood and also the cooperative nature of the binding of the sodium ions. Structural studies have pinpointed the subtle differences in molecular structure brought about during the evolution of the P-ATPases that have led to their specific functions and to their specific roles in animal diversity.

6.2.2.2 Functional Role of the β-Chain Although the three-dimensional (3D) structure shows clearly the disposition of the β-chain with respect to the rest of the molecule, this does not by itself fully elucidate the function of this subunit, but the structure did suggest the devising of experiments that further addressed this question. It was originally thought that the β-chain’s role was to guide the effective synthesis of the α-chain on the endoplasmic reticulum and the adoption of that chain’s correct conformation. Recent studies have led to a more nuanced version of this. Apparently, the α-chain and the β-chain mutually control each other’s expression in the cell, ensuring that they are found, almost always, in a 1:1 stoichiometry. Look at Figure 6.16. On the left (A), one sees, in a helix-coil version, the two subunits with the contacts between them delineated, part within the transmembrane stalk, part at the extracellular face. Sachs and his colleagues mutated residues Y38, F42, and Y43 of the transmembrane portion of the β-chain, substituting all three by alanines. Also residue P245, in the extracellular portion was substituted by glycine. Then they transfected one or other of these mutated sequences (or the corresponding wild type sequence), first linking them to yellow fluorescent protein (YFP), into MDCK kidney cells which they grew on monolayers, where they grow to

Primary Active Transport Systems Chapter 6

|

269

FIGURE 6.16 (A) Model of the Na1,K1-ATPase, showing interactions between the α1 and β1 subunits. (B) Immunostaining of the α1 subunit shows that the wild-type (WT) YFP-β2 precisely colocalizes, the YFY/AAA mutant only partially colocalizes, and the P/G mutant does not colocalize with the α1 subunit in the basolateral membrane. Adapted with permission from Tokhtaeva, Sachs, and Vagin. Assembly with the Na1,K1-ATPase α1 Subunit Is Required for Export of β1 and β2 Subunits from the Endoplasmic Reticulum. Biochemistry 2009, 48, 11421
11431. Copyright 2009 American Chemical Society.

confluence. Subfigure (B) shows confocal microscopic pictures of the cells, where the first column records the wild type (WT) β-chains, the second the mutants at the transmembrane-binding residues, and the third the mutants at the extracellular-binding region. The three rows record the results of staining the cell preparations with, top row, an antibody against YFP, hence showing the location of the β-chains; middle row, an antibody against the α-chains; and bottom row, the top and middle rows merged. In such experiments, if the green-stained material and the red-stained material colocalize, the merged frames appear yellow. Thus, in the wild-type set, all of the α-chains and the β-chains colocalize at the cell membrane. In both mutant β-chain sets, much β-chain material remains within the cell (and is there still within the endoplasmic reticulum). In the mutation P245G (right-hand data sets), practically all the β-chain material is within the body of the cell, with only very little that has reached the membrane. Thus, it is the effective binding of the two subunits to each other that ensures their correct localization at the cell membrane. We note that the β-chains of Na1,K1-ATPase are largely at the extracellular face. In associations between cells, such as are shown in Figure 6.16 and in epithelia in the living body, the β-chains bind to each other across the intercellular space. Experiments similar to those just described enabled Shoshani and her colleagues to suggest the model of an epithelium as depicted in Figure 6.17. Part (A) shows a confocal transverse section of a monolayer of MDCK cells (such as we saw in the plane of the

270 | Channels, Carriers, and Pumps

α β

Intercellular space

β

Apical domain

Cell 1 Tight

α β

Basal domain Vectorial transport

Cell 2

α (A)

(B)

(C)

FIGURE 6.17 (A) Confocal section of a monolayer of MDCK cells. The cells’ nuclei are stained red with propidium iodide and the β subunits of the Na1,K1-ATPase are stained green with a specific antibody. (B) Na1,K1-ATPase molecules of two neighboring epithelial cells interact via their β subunits. (C) Na1,K1-ATPase molecules anchored to the lateral membranes pump Na1 ions into the intercellular space. Reprinted, with permission, from: The polarized distribution of Na1,K1-ATPase: role of the interaction between β subunits. Padilla-Benavides T, Rolda´n ML, Larre et al. Mol Biol Cell. 2010;21:2217
25.

monolayer in Figure 6.16). The nuclei are stained with propidium iodide (red) and the β-subunits of the Na1,K1-ATPase are stained with a specific antibody (green). This subunit is localized to the lateral surfaces of cells, but not to the apical (left) or basal (right) sides. Part (B) shows a cartoon of how the β-chain/β-chain can bridge the gap between two cells, binding them together, while (C) shows how this arrangement, as could be present in the epithelium of a living cell, ensures that sodium is pumped into the intercellular space. The tight junctions at the apical face of the epithelial layer prevent the sodium from escaping at that surface, so that sodium that is brought into the cells by cotransport with metabolites from the apical surface is not returned to that face, but is rather expelled into the basal domain. Although the finding that the β-chains are used to bridge cells in an epithelial layer delineates a role for these chains in epithelial cells, Na1,K1-ATPase is found in nearly all cells of the animal body. There is certainly no suggestion that red blood cells bind to one another by β-chain/β-chain bonding and the same must be the case for the very important sodium pumps of nerve cells. Yet most animal cells have sodium pumps that contain β-chains. The more fundamental role of the β-chains, presumably common to all animal cells, lies in the strong interactions between the α- and β-subunits that provide space for the two potassiums to bind in the protein’s E2 form, whereas in the calcium ATPase, lacking the β-chain, there is room only for protons.

6.2.2.3 FXYD Subunits and Regulation The FXYD proteins are a family of (some seven in mammals) small (60- to 95-residue) molecules that are found bound to Na1,K1-ATPase and are

Primary Active Transport Systems Chapter 6

|

271

FIGURE 6.18 Hypothetical cartoon depiction of a structure
function relationship between phospholemman (PLM) and Na pump subunits. Reprinted, with permission, from: Novel regulation of cardiac Na pump via phospholemman. Pavlovic, Fuller, and Shattock. J Mol Cell Cardiol. 2013; 61: 83
93.

involved in its regulation. The family member in heart is FXYD1 (also known as phospholemman, PLM, or by its gene name of ATP1G1). The transmembrane portion of FXYD runs almost parallel to M9 and is seen clearly in Figure 6.12(i), although its extracellular segment, containing the FXYD motif, is obscured within the β-chain, which is sandwiched between α and FXYD by hydrophobic interactions. However, Na1,K1-ATPase containing only the α- and β-chains is active, but the FXYD subunit, which binds to both these subunits, stabilizes the whole structure. The regulatory action of FXYD1 is controlled by phosphorylation of its cytoplasmic domain (which unfortunately is not resolved in the crystal structures, Figure 6.12), whereas most other FXYDs lack phosphorylation motifs. Unphosphorylated PLM interacts closely with the membrane and with the subunits of Na1,K1-ATPase, producing a low affinity low velocity form (Figure 6.18). Phosphorylation alters the association between the pump and PLM by moving the cytosolic arm away from the pump, but not by promoting their dissociation. Phosphorylation or ablation of phospholemman relieves inhibition of the Na1 pump increasing its Vmax and apparent Na1 affinity. Therefore, under stress, phosphorylation of phospholemman allows the heart to reduce its Na1 and Ca21 load and thus prevents lethal arrhythmias. In an additional regulatory role, phospholemman can bind to, and regulate, the heart’s Na1/Ca21 exchanger. The phosphorylation state of PLM is under continuous fine control by the reciprocal action of numerous kinases and phosphatases, themselves affected by hormones (such as adrenaline), adrenergic drugs, and by nitric oxide.

272 | Channels, Carriers, and Pumps

As in the case of the Ca21-ATPases, the structural studies of the Na1, K -ATPase have clarified many aspects of pump function and its control. Some questions do, however, remain. For instance, we still need an explanation of the high velocity of the sodium pump, which operates at an order of magnitude faster than the calcium pump. And another question: How does the actual splitting of the ATP bring about the conformation changes that lead to the opening and closing of the cation gates? Additional structural studies, combined with kinetic and physical studies and molecular dynamic simulations, based on the knowledge of these structures, should lead to solutions of such problems (Box 6.4). 1

BOX 6.4 Stoichiometry of Ion Pumping We saw in Sections 5.2.4 and 5.3.4 how important, physiologically, the stoichiometry of coupling is in the case of counter- and cotransport. Stoichiometric considerations are equally important in the case of the primary transporters, where the concentration ratio of the pumped ion will be determined by the stoichiometry of the pump for its cation substrate(s). Since, however, we have to take into account the change in chemical potential that occurs during the progress of the chemical reaction of ATP hydrolysis, the concept is a little more difficult. But the problem is still easy to handle, and we proceed along the following lines: We know that for the chemical reaction ATP-ADP 1 Pi we can write the equilibrium constant for the reaction KATP-ADP1Pi in terms of the concentrations of ATP, ADP, and Pi at equilibrium as KATP-ADP1Pi 5 ½ADP½Pi =½ATP

(B6.4.1)

If, instead, we consider ATP hydrolysis linked to the transport of a number m of metal ions M from side 1 of the membrane (where the metal is at concentration M1 to side 2 of the membrane (where it is at concentration M2), the reaction can be written as ATP 1 mM1 -ADP 1 Pi 1 mM2 : For this reaction, we can write the equilibrium condition as KATP-ADP1Pi 5 ½ADP½Pi ðM2 Þm =½ATPðM1 Þm

(B6.4.2)

We can do this because the metal is in the same chemical state on the two sides of the membrane. There is thus no change in its standard-state chemical potential (see Section 2.1) and we do not have to include a term for the “equilibrium constant” for any chemical reaction of the metal ion. We can now rewrite Eq. (B6.4.2) so as to emphasize what determines the steady-state distribution of M: ðM1 =M2 Þm 5 ½ADP½Pi =½ATPKATP-ADP1Pi

(B6.4.3)

Clearly, from Eq. (B6.4.3) if the concentration of ATP is high, compared with that of ADP and Pi divided by the equilibrium constant KATP-ADP1Pi the concentration of M2 will be high compared with that of M1, i.e., M will be pumped from side 1 (the cytoplasmic face of the cell or SR membrane) to side 2. A given ratio of (Continued )

Primary Active Transport Systems Chapter 6

|

273

BOX 6.4 (Continued) ATP, ADP, Pi, and the equilibrium constant KATP-ADP1Pi, will create a steeper concentration ratio of the metal ion M, if the stoichiometric coefficient is less. Take, as an example, values of ATP, ADP, and Pi concentrations of 5 mM, 50 μM, and 5 mM, respectively, as representative of those that occur in many living cells. Take KATP-ADP1Pi equal to 106, again a reasonable approximation to its value at 37 C. Substituting in Eq. (B6.4.3), we see that for m 5 1, i.e., a stoichiometry of unity, we would expect to find a metal ion concentration ratio of (106 3 103)/(0.05) at equilibrium, that is, of 2 3 1010. Such enormous concentration ratios are never found in living cells! With the stoichiometric coefficient of 2, as found for the calcium pump in Figure 6.10, m 5 2 and, therefore, the concentration ratio at equilibrium will be the square root of the number just calculated or 1.4 3 105. This is not too far from the value often encountered for the SR vesicle, where cytoplasmic calcium concentrations can approach 30 nM and intravesicular concentrations, 1 mM, a ratio of 3 3 104. Consider, now, the sodium
potassium pump discussed in Section 6.1. Here the stoichiometry of pumping of sodium is 3 and that of potassium, 2, for each ATP split. We need to rewrite Eq. (B6.4.3) in terms of two types of metal ions, M and N, each with its own stoichiometry m and n. The reader can check that with prevailing cation ratios of about sevenfold for sodium (inside low) and 20-fold for potassium (inside high), prevailing levels of ATP, ADP, and Pi are such as to be consistent with the predictions of the stoichiometry of the pump. The reader will realize that the treatment given above is an oversimplification since we have omitted any consideration of the electrical potential. (We saw in Box 6.2 that the sodium pump, at least, is indeed electrogenic.) Insofar as an ion pump is electrogenic, i.e., if it transfers n units of charge across the membrane during a complete cycle of pumping, we have to modify Eq. (B6.4.3) to take the transmembrane potential ψ into account. The complete equation is ln½ðM1 =M2 Þm  1 ðnF =RT ÞΔψ 5 lnð½ADP½Pi =½ATP1ÞKATP-ADP1Pi

6.3 THE CALCIUM PUMP OF THE PLASMA MEMBRANE The sodium
calcium antiporter, discussed in detail in Section 5.2.5, is only one of the systems by which the cytoplasmic concentration of calcium ions in the cell is kept low. The other method is the harnessing of primary active transport by the calcium-activated ATPase of the plasma membrane (PCAM), a system first identified by Hans Schatzmann. Such a primary transporter of calcium ions is present in the plasma membrane of most animal cells. Two well-studied systems are those of red blood cells and the nerve membrane. The system in the human red blood cell is very effective. With cells stored in their own plasma, at a free calcium concentration of 1.3 mM, the intracellular free calcium can be as low as 26 nM, a 50,000-fold concentration ratio. The transmembrane potential in these red cells is about

274 | Channels, Carriers, and Pumps

10 mV, so that efflux of a divalent cation is hindered by the inside negative potential by a factor of 2.25-fold (59 mV is equivalent to a 100-fold concentration ratio for a divalent cation; see Section 2.2). All in all, calcium ions are pumped out against an equivalent gradient of over 105! (Recall from Box 6.4 that splitting of an ATP molecule can cope with this equivalent concentration ratio using a stoichiometry of two. This is, indeed, the measured stoichiometry of the calcium-pump enzyme of the plasma membrane.) In its properties, the calcium pump of the plasma membrane strongly resembles the Na1,K1-ATPase and the SR Ca21-ATPase that we discussed in the previous sections. This system, too, is inhibited by vanadate, forms an isolatable phosphoenzyme, has a high affinity for its substrate at the cytoplasmic face of the membrane, a low affinity at the extracellular surface. Like the calcium pump of the SR, it is not inhibited by ouabain. It is composed of a single polypeptide chain. Its rate of action can be precisely regulated by a cytoplasmic factor. As we discussed in Section 5.2.5, the free calcium ion concentration of the cell is a major target for the transduction of extracellular signals received by the cell. Wiping out the memory of these signals requires that cytoplasmic levels of calcium be restored rapidly and effectively. The calcium pump plays its part in this task, acting in a regulated fashion. As calcium levels rise, its rate of pumping speeds up and does so far more rapidly than would be the case were it to be activated solely by the direct binding of calcium. There is, as yet, no structural model available for PCAM, but there is much data available on its regulation. Consider Figure 6.19A, which depicts the activity of the calcium pump as a function of the concentration of ATP, for the isolated pump enzyme (lower curve, Figure 6.19A) and for the same preparation but in the presence of a cytoplasmic protein named calmodulin (upper curve, Figure 6.19A). Clearly, calmodulin greatly increases the rate of calcium pumping at all ATP concentrations, and is even more effective at the higher concentrations. Calmodulin (http://en.wikipedia.org/wiki/Calmodulin) is a water-soluble cytoplasmic protein, with a molecular mass of 16,700 Da. A single molecule of calmodulin binds four calcium ions with very high affinity. Calmodulin is a member of a large family of calcium-binding proteins, all sharing the property of possessing a particular concatenation of amino acids at their calciumbinding sites, the so-called calcium hand. Calmodulin binds to the plasma membrane calcium pump, but only when the calmodulin has bound to four calcium ions. Binding of calmodulin to the pump modifies its activity in the manner shown in Figure 6.19A. Thus when concentrations of calcium in the cytoplasm rise, following some extracellular signal, some of these incoming calcium ions become bound to calmodulin, its own binding to the Ca21ATPase takes place and the pump enzyme acts more rapidly, increasing the rate at which calcium is expelled from the cell. As the level of calcium drops again, below the levels at which all four calcium ions can bind to calmodulin, calmodulin dissociates from the pump enzyme and pumping slows down

(Ca2+, Mg2+–ATPase activity (mmol liter–1 hr–1)

(A)

8

+Calmodulin

6

4

2

0

0

500

1000 1500 ATP concentration (μM)

2000

(B) 15

ATPase activity (μmol Pi mg protein–1 min–1)

10

5

5

10

15

20

10

15

20

15

10

5

5 2+

Ca

concentration (μM)

FIGURE 6.19 Regulation of the activity of the plasma membrane Ca21-ATPase. (A) How calmodulin affects the affinity of the pump for ATP and also its maximum velocity. The enzyme, from human red blood cell membranes, was assayed at 37 C in the presence of 100 mM NaCl, 10 mM KC1, ouabain, and 50 μM calcium. The upper curve shows data when 150 μM calmodulin was added; the lower curve, in the absence of calmodulin. Taken, with kind permission, from S. Muallem and S. J. D. Karlish (1980). Biochim. Biophys. Acta 597, 631
636. (B) How calmodulin and the phospholipid content affects the affinity of the pump for Ca21. The upper panel shows data where the membrane vesicles were reconstituted with phosphatidylcholine; the lower panel, with phosphatidylserine. The circles are in the absence of calmodulin; the triangles, in its presence. Data at 37 C in 120 mM KCl. Taken, with kind permission, from K. Stieger and S. Lauterbacher (1981). Biochim. Biophys. Acta 641, 270
275.

276 | Channels, Carriers, and Pumps

again. Thus it is the affinity of calmodulin for calcium ions that helps to regulate the rate of calcium pumping and hence the concentration of cytoplasmic calcium. Phosphorylation of calmodulin can control the pump
calmodulin interaction. How does calmodulin have this dramatic effect on the rate of calcium pumping? Figure 6.19B (upper half) shows how the activity of the pump enzyme is affected by the concentration of calcium ions, in the presence and absence of calmodulin. Again, there is a marked stimulation by calmodulin. These data are for the isolated plasma membrane Ca21-ATPase, reconstituted into phospholipid vesicles made from the neutral phospholipid, phosphatidylcholine. If the same experiment is performed using vesicles made with the phospholipid phosphatidylserine, a negatively charged molecule, no effect of calmodulin is found; the enzyme is fully active even in the absence of the activator (lower half of Figure 6.19B). A fully active enzyme is also obtained when the isolated pump enzyme is digested with trypsin, a treatment which splits off a 35,000 molecular weight polypeptide from the enzyme. This polypeptide partially blocks access of ATP and Ca21 to its active sites on the pump enzyme. (See Figure 6.20A for a cartoon depiction of this model.) When the peptide is removed by tryptic digestion, or when the enzyme is in a

FIGURE 6.20 (A) Schematic representation of calmodulin (PMCA) and peptide sequences from the Calmodulin Binding Domain (CBD) of its C-tail. The cartoon shows the PMCA in the autoinhibited form where the CBD interacts with both the first and second intracellular loop. The C-tail of 150 amino acids (a.a) beginning after the last membrane-spanning domain is indicated in red. (B) The sequence of peptide C28 from the C-tail is shown on the bottom. (C) depicts the structural model of the calmodulin/C28 complex taken from the Protein Data Bank (2KNE). A and B taken, with permission, from Juranic N1, Atanasova E, Filoteo AG, Macura S, Prendergast FG, Penniston JT, Strehler EE., Calmodulin wraps around its binding domain in the plasma membrane Ca21 pump anchored by a novel 18-1 motif. J Biol Chem. 2010;285:4015
24.

Primary Active Transport Systems Chapter 6

|

277

region of local negative charge, or when calmodulin is bound to it, the effect of the blocking peptide is removed, and full activity is obtained. More recent studies have localized the calmodulin-binding region of the calcium ATPase to a 28 amino acid residue peptide (C28), its sequence being given in Figure 6.20B, while Figure 6.20C shows the structural model of the complex between calmodulin (brown helices) and C28 (cream helix). The four calcium ions bound to calmodulin are shown as small green spheres.

6.4 THE H1, K1-ATPASE OF GASTRIC MUCOSA: THE PROTON PUMP OF THE STOMACH The lining of the stomach secretes into the lumen of the stomach a highly acid secretion, which can reach as high as 0.16 M HC1. Specialized cells in this lining, the parietal cells, have the function of providing this secretion. Proton pumps are present within the cell in the form of intracellular vesicles. As a result of a hormonal signal, these vesicles fuse with the plasma membrane of the cell, the pumps being inserted into the membrane. The cells can develop a pH difference, inside low, of about 6.6 units of pH. This is equivalent to a concentration ratio of 4 3 106! How is this achieved? Data obtained in the laboratory of George Sachs make this clear. Figure 6.21A shows the phosphorylation by ATP of a membrane preparation obtained from the acid-secreting vesicles of gastric mucosa. The filled symbols in Figure 6.21A show phosphorylation when the extravesicular pH is 5.5, while within the vesicle the pH is 5.5, 6.7, or 8.0. At all internal pH values, ATP is split rapidly to phosphorylate the membrane proteins. In contrast, the empty symbols show data for similar vesicles having the same values of internal pH, but an extravesicular pH of 8.0. Clearly, phosphorylation of the membranes is achieved rapidly only when there is an acid pH externally, i.e., when there are sufficient protons at the extravesicular surface. The intravesicular pH does not affect the rate of phosphorylation. In contrast, Figure 6.21B shows that the rate of dephosphorylation is dependent on the concentration of potassium ions and that these potassium ions have to be present inside the vesicles, carried there by the ionophore gramicidin (see Section 2.1). Thus, overall, a hydrolysis of ATP is activated by protons at the cytoplasmic face and potassium ions at the lumen face of the vesicle. The pump is an H1, K1-ATPase and closely homologous with the Na1,K1-ATPase. The H1, K1-ATPase from gastric mucosa has been highly purified and shown to consist of two polypeptide chains, and the gene coding for the longest chain—that binding to and phosphorylated by ATP—has been cloned. The amino acid sequence of this chain is more than 60% identical to the sequence of the long (α) chain of the Na1,K1-ATPase. It is clear that two protons take the place of the three sodium ions pumped by the Na1,K1ATPase, while the potassium ion has the same function in both systems.

278 | Channels, Carriers, and Pumps (A)

p mol mg–1

1500

1000

500

0

50

100

150

200

Time (msec) (B)

2.0 1.8

log % EP

1.6 1.4 1.2 1.0 0.8 20

40

60

80 100 120 140 160 180 200 Time (msec)

FIGURE 6.21 Phosphorylation and dephosphorylation of the gastric H1,K1ATPase. (A) Phosphorylation of the enzyme, prepared from pig stomach at 22 C. The filled symbols represent vesicles prepared with an extravesicular pH of 5.5 and various intracellular pH values. The empty symbols denote vesicles prepared with an extravesicular pH of 8.5 and various intracellular pH values. (B) Dephosphorylation. As above, but phosphorylated membrane vesicles were allowed to dephosphorylate by adding CDTA in order to complex magnesium. The triangles show vesicles to which potassium chloride had been added and also gramicidin, an ionophore, to allow the potassium access to what is the extracellular surface of the membranes. In contrast, the controls (circles) had no potassium added, or potassium added with no gramicidin, so that the potassium could not reach the extracellular surface of the membranes. Taken, with permission, from Stewart et al. (1981). J. Biol. Chem. 256, 2682
2690.

The stoichiometry of pumping is thus two H1 and two K1 transported for every ATP that is split. Researchers agree that the proton and sodium pumps must have strong evolutionary links, the Na1,K1-ATPase being almost certainly the progenitor of this proton pump. (Indeed, as Rhoda Blostein has shown, in the absence of sodium ions the sodium pump can be activated by protons, so that it will split ATP with only protons at the inner face of the

Primary Active Transport Systems Chapter 6

|

279

membrane, and potassium ions at the outer face. Here, the sodium pump is acting as an H1, K1-ATPase!) Why should the gastric proton pump move two protons rather than three? It must be stoichiometric considerations that have forced the evolution of this characteristic. A pH gradient of 6.6 is above the steepest that an ATP-driven pump can achieve if the stoichiometry is two ions pumped (see Box 6.4 for a discussion of stoichiometry). Three ions cannot be pumped simultaneously against a gradient higher than 104-fold. Indeed, it even seems likely that the stoichiometry of the H1, K1-ATPase is variable, so that at the very highest gradients, one rather than two protons are transported for each ATP hydrolyzed. In all respects, the H1, K1-ATPase is a close relative of the other E1E2ATPases. Figure 6.21 shows that it acts through a phosphorylated intermediate. It is inhibited by vanadate, exists in two conformations that interchange during transport, and has strong sequence homologies with the other E1E2ATPases. In the following section, we discuss these homologies in a little more detail.

6.4.1 The P-Type ATPases in the Context of Protein Evolution A sufficient number of solved 3D structures of phosphatases are now known that the P-type ATPases that we have been discussing can be put into the general framework of protein evolution. Burroughs and his colleagues analyzed over 40 phosphoesterases, ATPases, phosphonatases, dehalogenases, and sugar phosphomutases of known structure, adding to these the amino acid sequence data on numerous other such proteins. With all this, they built a database that allowed them to classify these molecules into a number of different, but related, families. The P-type ATPases that we have been discussing fall into the HAD superfamily, named for haloacid dehalogenase, its simplest member. All of the members of the superfamily have in common an aspartate residue that is phosphorylated and dephosphorylated during a cycle of enzyme activity and they share also a Rossmanoid domain, which we immediately discuss. Rossman showed 40 years ago that many phosphodiesterases shared a common fold, a set of α-helix/β-sheet repeated structures concerned in the binding of nucleotides. (Go to http://blogs.oregonstate.edu/psquared/2012/04/ 16/topology-in-2d-and-3d-the-rossmann-fold/ for a depiction of such a Rossman fold.) Members of the HAD superfamily possess a particular Rossmanoid fold, which appears to be the progenitor of their present diversity. Figure 6.22 depicts, in A, a part of the SERCA1a molecule cut to show the P (phosphorylation) domain. Part B is a schematic view of the Rossmanoid fold of SERCA1a, a depiction of the whole molecule being shown in C. It is worth going to the PDB site to have the structural model at hand while this section is being read. Figure 6.22B, the Rossmanoid domain, repays careful study. The strands S1 through S6 are the seven β-sheets, each joined to the next by an α-helix. (This is the basic Rossmanoid fold.)

280 | Channels, Carriers, and Pumps (A)

(C)

(B) ATPase C1 cap insertion K DxD DxxxD

S6 S5

S4

Squiggle

S1

S2 S3.1 S3

FIGURE 6.22 SERCA1a and its Rossmanoid fold. Part A shows a view of the P domain of SERCA1a (the whole molecule is shown in the same view in C). B is a depiction of the P domain as a succession of α-helix, β-sheet (S1 through S6) turns. “Squiggle” is a flexible polypeptide chain that inserted into the basic Rossmanoid fold, as does the “flap,” seen as the green triangle and green polypeptide chain, which forms the N and A domains of the P-type ATPases. A and C from the Protein Data Bank. B is taken, with permission, from Burroughs AM, Allen KN, Dunaway-Mariano D, Aravind L. Evolutionary genomics of the HAD superfamily: understanding the structural adaptations and catalytic diversity in a superfamily of phosphoesterases and allied enzymes. J Mol Biol. 2006;361:1003
34.

The green arrow and green line show that our ATPase is a member of the C1 Cap family since at this point an additional domain, a “flap,” was added to the basic Rossmanoid domain, through the sequences depicted as the short thick blue arrows. In the case of the P-type ATPases, this addition comprises the N and A domains of Figures 6.8 and 6.9 that act as a lid during the phosphorylation/hydrolysis cycle. Note also the “squiggle,” depicted as a pink thread, this being a polypeptide sequence that gives the flexibility needed during the conformation change cycle. (In the figure, residues denoted by DxxD, K, DxD, and T are invariant amino acids in the C1 family of proteins.) Other members of the HAD superfamily have had different types and numbers of Caps added during the course of their evolution. Figure 6.23, taken from the paper by the Burroughs group, shows the breakdown of HAD family members into the subfamilies C0, C1, and C2. Returning to our P-type ATPases, the Burroughs group summarized their conclusions along the following lines: The key aspects of the HAD catalytic

Primary Active Transport Systems Chapter 6

|

281

FIGURE 6.23 Members of the HAD superfamily of phosphate active enzymes. C0, C1, and C2 are the subfamilies with different cap domains, as indicated and as described in the text. Taken, with permission, from Burroughs AM, Allen KN, Dunaway-Mariano D, Aravind L. Evolutionary genomics of the HAD superfamily: understanding the structural adaptations and catalytic diversity in a superfamily of phosphoesterases and allied enzymes. J Mol Biol. 2006;361:1003
34. For the meaning of the symbols and the list of depicted enzymes, see this Burroughs group publication.

mechanism seem to be (i) an alternation between open and closed states and (ii) an initial reaction of phosphorylation favored by solvent exclusion and a subsequent dephosphorylation step, favored by extensive solvent contact. The principal features responsible for this process are the squiggle and the flap. The squiggle, being close to a helical conformation, is a structure that can be, in turn, tightly or loosely wound. This differential winding in turn induces a movement in the flap immediately juxtaposed to the active site, and results in the alternation of the closed and open states. Seeing that there is strict conservation of the squiggle and the flap across the HAD superfamily, these features are likely to be part of a universal functional feature of this superfamily. The conformational changes in the squiggle and flap could provide the minimal apparatus for solvent exclusion and subsequent solvent access at the active site of these enzymes. Natural selection appears to have favored the further emergence of cap modules, leading to more efficient solvent exclusion and acyl phosphate formation. In addition to aiding the basic catalytic mechanism,

282 | Channels, Carriers, and Pumps

the emergence of diverse caps also provided a means of substrate recognition by supplying new surfaces for interaction with substrates, which was not afforded by the ancestral active site alone. The simplest structures add the cap to the flap motif itself, so as to completely seal the active site in the closed state. The flap region became a hotspot for the insertion of the various C1 caps, which appears to suggest intense natural selection for efficient solvent exclusion. The special additions of the P-type ATPases to the generic C1 structure were the transmembrane helices which allowed the closed/open conformation change of the basic Rossmanoid structure to be harnessed to the shielded, coupled movement of ions across the membrane.

6.5 THE ROTARY ATPASES There is another great family of cation-activated ATPases. This is the family of rotary ATPases, that includes the F0F1 ATPases, found in the membranes of the mitochondria and chloroplasts and in bacterial membranes, the A0A1 ATPases found in the archaebacteria, and the V0V1 ATPases present in the membranes of the acidic vesicles of eukaryotic cells. The E1E2-ATPases of the previous sections of this chapter have, as their usual role, the hydrolysis of metabolically produced ATP, coupled to the pumping of the cations. In contrast, the majority of the ATPases of the present section usually act to synthesize ATP, where synthesis is coupled to the movement of protons (or in some specialized systems, sodium ions) down their electrochemical gradients. In the mitochondrion, the proton gradients are established by metabolism of energy-producing fuels, such as the burning of glucose. This results in coupling of proton movements to the passage of electrons down the electron chain of the oxidation
reduction reactions of the cell. In the chloroplast, trapping of sunlight leads to the building up of a transmembrane proton gradient, which is then used by the chloroplast to produce ATP. These mitochondrial, bacterial, and chloroplast ATPases are often termed “ATP synthases,” an appellation that emphasizes their special role in the cell. However, they certainly can act as ATPases and proton pumps since the reactions that they undergo—like all other chemical reactions in the organism—are reversible. Indeed, one member of the rotary ATPases, the V0V1 ATPases of the vacuolar membranes, is normally poised in the direction of ATP hydrolysis that is linked to proton pumping. A major distinction between rotary ATPases and E1E2-ATPases is that only for the latter enzymes are phosphorylated intermediates formed during ATP hydrolysis. Since no phosphointermediates are formed, vanadate, too, does not combine with the rotary ATPases and they are therefore not inhibited by it. There seems to be little that links the two families of ATPases, either in an evolutionary sense or in the details of their structure and mechanism of action.

Primary Active Transport Systems Chapter 6

|

283

6.5.1 Structure of the Rotary ATPases The rotary ATPases have a far more complex structure than do the E1E2ATPases. We saw that the latter are composed of one, two, or at most three polypeptide chains. A rotary ATPase can be built of no less than 20 chains! The whole ATPase is composed of two readily separable portions, F0 and F1 (or A0 and A1 in the case of the archaebacteria. The following discussion refers to the rotary ATPases in general and we use the F notation to refer to both F-type and A-type rotary ATPases). Further, F0 acts as a molecular turbine, driven by the flow of protons through the complex itself. It rotates within the membrane. This rotation is sensed by F1, and the energy in the rotation leads to the synthesis of ATP from ADP and Pi. The two portions, F1 and F0 are thus mechanically coupled together to form an integral F0F1, ensuring that the proton flow and ATP synthesis or hydrolysis are mechanically linked together. The overall mechanism is astonishingly efficient, with 100% of the energy of proton flow being converted to ATP synthesis. The chemical reagent dicyclohexylcarbodiimide (DCCD) inhibits proton flow in both F0 and the intact F0F1-ATPase, but not ATP synthesis (or hydrolysis) by the isolated F1 portion. Now look carefully at Figure 6.24. Row (A) depicts models of the eukaryote systems, the F0F1 synthase of the mitochondrion and the V1V0 vacuolar ATPase. Row (B) depicts the equivalent bacterial systems and (C) that from the archaebacteria. The structures appear very complex and indeed they are, but let us try to understand them and how they work. We use the eukaryotic F-type ATP synthase (top row on the left in Figure 6.24) as our example. The terminology conventionally used for the other rotary ATPases will be clear from the figure. First, notice that part of the structure is embedded in the cell membrane. This is the F0 portion, functionally an intrinsic membrane protein. It consists of a ring of 9 to 15 (in different systems, just 10 in the mitochondrium) membrane-spanning subunits, symbolized by c, that anchor the whole F0F1 firmly in the membrane. Each subunit is a hairpin of α-helices, with one arm of each hairpin being composed of hydrophobic amino acid residues. Thus the whole ring presents a hydrophobic surface to the membrane lipids within which the F0F1 is embedded. The F0 and F1 portions of the whole enzyme can be physically separated. An early pioneer in the understanding of these systems, Ephraim Racker, showed that when intact membranes are treated with low ionic strength buffers, the F0/F1 connection is broken, leading to solubilization of F1, while F0 remains bound to the membrane. The isolated F1 portion acts as an ATPase. Note that attached to the ring of c-subunits is subunit a. This is the stator and remains stationary when the c-subunit ring rotates during the pump’s operation. One, two, or three peripheral stalks (symbolized by b bound to d and F) link the stator a (The stator is the stationary part of a rotary

284 | Channels, Carriers, and Pumps

FIGURE 6.24 Schematic representations of rotary ATPases across different taxonomic systems. (A) Eukaryotes contain two types of rotary ATPases: mitochondrial F0F1 ATP synthase (left) and vacuolar V1VO ATPase (right). (B) Many eubacteria have only one rotary ATPase, either bacterial F-type (left) or bacterial A/V-type ATPase/synthase (right), which can operate in either direction. (C) Archaea contain A-type ATPases/synthases that are structurally indistinguishable from bacterial A/V-types. (D) Comparison to a power generator in a hydroelectric power plant consisting of a turbine (MT) and an electrical generator (MG). From A. G. Stewart, M. Sobti, R. P. Harvey et al. Bioarchitecture, 3; 2013, 2
12. Rotary ATPases: Models, machine elements and technical specifications; Landes Bioscience, 2013. Copyright and all rights reserved. Material from this publication has been used with the permission of Landes Bioscience.

system—from the Latin root for the word “stand”) to the head groups of F1. These stalks ensure that the stator and the head groups of F1 remain as a connected stationary group when the c-ring rotates. The flow of protons, symbolized by the H1 and the arrow in Figure 6.24, through the F0 portion, drives the rotation of the c-subunit ring in stepwise fashion, each proton that flows through the system driving the rotation by close to one subunit, operating as a ratchet. The F1 portion is formed of nine polypeptides, of five different types, the whole F1 comprising three copies of the α-chain, three of the β-chain, and

Primary Active Transport Systems Chapter 6

|

285

one each of the γ, δ, and ε subunits; the structure of F1 is written as α3β3γδε. The β-chain has been shown to bear the ATP-binding sites, each β-subunit being capable of binding a pair of nucleotide molecules. The genes coding for all of the five types of polypeptide have been cloned and sequenced. The α and β chains and from very widely different organisms share a good deal of sequence similarity. Consider now the water-powered mechanical turbine system depicted on the right of the bottom row of Figure 6.24. Here, water flowing through the rotating lower portion (M1) falls on to the vanes that bring about the rotation of this lower portion. The circle of vanes is bound to the central axis that is itself bound to a circular magnet (M0). This magnet rotates with respect to an outer ring of coils that is fixed to the stator on the left of M1. Rotation of the magnet with respect to the coil induces an electric current so that the energy in the gravity-driven flow of falling water is converted into electrical energy. This is the analogy that we must keep in mind while considering the mechanism of action of the rotary ATPases. The c-ring system of F1 rotates, driven by a flow of protons. This rotation is conveyed to the γ subunit that is fixed to the rotating ring. As γ rotates with respect to F1, the energy of its rotation, its torque (the tendency of a force to rotate an object about an axis), is converted into the chemical energy of ATP formation. The early evidence for the structure of the rotary ATPases came from electron microscope studies such as those depicted in Figure 6.25. Electron micrographs of F1 showed that this can aggregate as a hexagonal array of subunits (Figure 6.25A). Detailed analysis of such electron micrographs, including side-on views of the protein, showed F1 to be composed of a double array of trimers, the α- and β-chain triplets being arranged in two separate layers (Figure 6.25B). Attached to this is a central portion that contains the γ, δ, and ε chains. The F0F1-molecules can be seen in the electron micrograph to associate side by side to form “strings” (Figure 6.25C) whose dimensions give the width of the F0 portion. The whole F0F1-ATPase thus has the structure diagrammed in Figure 6.24, where the F0 portion is transmembranal, while F1 with its major trigonally arranged α- and β-units situated in a plane above the membrane, is attached to F0 through the γ, δ, and ε chains. This structure of the F0F1-ATPase seems to be a very ancient one. Figure 6.26 shows electron micrographs of F1-ATPases from a wide range of organisms, including the mitochondria of beef heart, the chloroplasts from spinach, four different eubacteria, and an archaebacteria. The eubacteria split off from the archaebacteria two to three billion years ago. Thus the structure of the F0F1ATPases seems to have been a marvelously persistent one. For almost all of the components of the F0F1-ATPases, high-resolution structures have now been determined. Figure 6.27 shows, as an example, for the F0F1 of the mitochondrion, the 3D representations of the subunits, and how these pack into the known structure of the envelope of the whole complex as this has been determined from electron microscope studies.

(B)

A

B

C

6 nm

(A)

D

α

α β

10.9 nm E

F

G

β

H

J

K

12.3 nm

I

L

5n

m

α

α γδε

?

β

β α

M

N

O

10.9 nm

P

(C)

FIGURE 6.25 Structure of the F0F1-ATPases. (A) Molecular projections of the chloroplast F1ATPase, as determined by computer-aided classification and aligning of 3,300 electron microscopic images. The subunit structure is clearly seen, (B) Three-dimensional model of the F1ATPase in side (above) and top (below) view. Dimensions refer to the beef heart system, (C) Electron micrographs in side view of the F0F1-ATPases from chloroplasts and mitochondria. Figures (A) and (C) taken, with kind permission, from E. Boekema, M. van Heel, and P. Graber (1988) in “The Ion Pumps” (W. D. Stein, ed.), pp. 75
80. Alan R. Liss, New York. Figure (B) taken, with kind permission, from G. Schafer et al. (1988) in “The Ion Pumps” (W. D. Stein, ed.), pp. 57
66. Alan R. Liss, New York.

FIGURE 6.26 A gallery of F0F1-ATPases from different sources. Top-view projections of electron micrographs of negatively stained preparations of seven different ATPases: (A) from spinach chloroplasts; (B) from beef heart mitochondria; (C)
(G) from various bacteria, including E. coli and an archaebacterium. Taken, with kind permission, from G. Schafer et al. (1988) in “The Ion Pumps” (W. D. Stein, ed.), pp. 57
66. Alan R. Liss, New York.

Primary Active Transport Systems Chapter 6

|

287

FIGURE 6.27 Packing of its component subunits into the envelope of the F0F1 ATPase from bovine mitochondria. (Left half based on Figure 6.5 of Muench SP, Trinick J, Harrison MA., Q Rev Biophys. 2011 44:311
56 and right half based on Figure 6.5 of Stock et al. A. G. Stewart, M. Sobti, R. P. Harvey et al. Bioarchitecture, 3; 2013, 2
12.). Legend for right hand half: 1, nucleotide binding, αβ subunits (2WPD); 2, central stalk, γ subunit (“crankshaft”—2WPD; 3, rotor ring, c subunits (“turbine”—4B2Q); 4, ion channel forming subunit, a (no structure available as yet—April 2014); 5, peripheral stalk, b and F subunits (“pushrod”—4B2Q); 6, small central stalk, δ subunit (“ratchet” in prokaryotes—2WPD); 7, eukaryotic additional central stalk, ε subunit (“lock”—2WPD); 8, IF1 (“brake”—1OHH).

6.5.2 Mechanism of Action of the F0F1-ATPases The F0 portion can be incorporated into phospholipid bilayers to which it imparts a proton conductivity, inhibitable by DCCD. For the chloroplast enzyme, about 5 3 105 protons are transported per F0 molecule per second, when the transmembrane potential is 100 mV. This is a good deal higher than the rate of a typical carrier or pump (see Table 4.2). Indeed, translating this flux into a conductance (Section 3.1) gives a value of about 1 pS, a value that is not inconsistent with a channel being the route for the protons (Section 3.4). This is again far higher than the 16 molecules per second measured for the maximum rate of ATP synthesis driven by proton flow in an F0F1 system reconstituted in artificial liposomes (see Soga et al., 2011). It would seem that subunits of the F1 portion drastically reduce the proton flux through the F0 portion of the molecule. We know a good deal about how the F0F1-ATPase acts as an ATPase. A crucial experiment was performed by Paul Boyer, who showed that, even in the absence of a proton gradient, ATP can be synthesized (remaining bound to the ATPase on the enzyme) from ADP and phosphate. Thus, the high energy bond in the ATP can be formed, but this ATP does not dissociate readily from the enzyme. Such considerations led Boyer to propose the

288 | Channels, Carriers, and Pumps

above-depicted reaction scheme for ATP synthesis by mitochondrial F0F1 (Figure 6.28). This is Boyer’s binding change model. In sequence, at each β-subunit in turn, ATP is bound to a site in a form that binds it tightly. Next, ATP is released to form a site free of nucleotide and then ADP and Pi bind to form the loose binding state. It is the torque arising from the proton-driven rotation of F0 that forces the ATP off from its binding at the high affinity state. In the full reaction scheme of Figure 6.29 below E, E , and E are three interconvertible forms of the enzyme that differ in affinities for ADP (symbolized as “D” in the figure) for Pi and for ATP (symbolized as “T”). The form E binds both ATP (the complex written as E T) and ADP (and Pi) strongly (written as PE D). E has a moderate affinity for ADP (and Pi) but little affinity

Loose binding

Binding change mechanism

ADP + Pi

ATP

ATP Tight binding

Open

ATP

ADP + Pi

Repeat

ADP + Pi

ATP ATP

FIGURE 6.28 Scheme for the “binding change” model of Paul Boyer.

(A)

(B) PE*DH k”

ADP Pi

E

PED

PE*D

E*DH

k’

k’

k”

E°TH”

PEDH’ ADP, Pi

E°

E°T ATP

E*T

EH’ H’

E°H” H”

k”

E

E° k’

FIGURE 6.29 Kinetic schemes for the coupling of ATP synthesis and proton transport in the F0F1-ATPases. (A) reaction scheme for the F0F1 ATPase. Binding to E, the free enzyme, by ADP (symbolized as “D”) and Pj gives the complex PED. This interconverts to E T (“T” symbolizes ATP), from which ATP can dissociate. The forms labeled with the superscript circle, asterisk, or no symbol are three conformations of the enzyme having, respectively, an affinity for ATP but not ADP, high affinities for both ATP and ADP, and an affinity for ADP but not ATP. (B) Reaction scheme for the integral F0F1-ATP synthase that synthesizes ATP and transports protons. As (A) except that, in addition, H0 and Hv represent protons binding to the pump from sides 0 and v of the membrane, respectively. The steps labeled k0 and kv are conformation changes that, in a linked fashion, interconvert the three conformations of the pump enzyme. Figures (A) and (B) taken from W. D. Stein and P. Lauger (1990). Biophys. J. 57, 255
267. ¨

Primary Active Transport Systems Chapter 6

|

289

for ATP, while E has a moderate affinity for ATP and little for the other two ligands. Harvey Penefsky obtained experimental data for the reactions depicted in Figure 6.29 at a single site on the F1-ATPase. The reaction was performed at a very low ratio of ATP to enzyme (1:3), i.e., at “subenzyme levels” of the substrate. The enzyme
substrate complex is formed on E and the catalytic step takes place on E . The rate-limiting steps are the release of the product, from PE D and then PED. The crucial observation is that the rate constants for the steps involving the E forms are nearly the same in the forward and the backward directions, i.e., ATP bound to the enzyme is nearly in equilibrium with ADP and Pi bound to the enzyme. This is, of course, in great contrast to the situation where these substances are free in solution, when the equilibrium is far over to the side of ATP breakdown. Compensating for the apparent “low-energy” nature of the ATP is the fact that it is tightly bound to the E form of the enzyme. The dissociation constant for its release from the enzyme is approximately 10212 M, so that it will seldom be released unless some other circumstance intervenes. In the fully functioning F0F1-ATPase, this circumstance is the transport of proton into the mitochondrion. The chemical reaction depicted in Figure 6.29A can, for the intact F0F1 complex, be written formally (in one possible scheme) as in Figure 6.29B, where E is the F0F1 complex, and H0 and Hv represent the protons inside and outside the mitochondrion, respectively. The flow of protons through F0 is coupled to the release of ATP from its high affinity binding to F1. In this way, ATP is delivered to the cytoplasm at the concentration prevailing there, in spite of the fact that ATP is so tightly bound to F1 when no flow of protons is occurring. The precise details of the coupling between proton flow and ATP release remain to be established, but some wonderful experiments by the group of Yoshida and Kinosita have gone a long way to solving this problem. Box 6.5 summarizes their 20 years of work on this question. We should still discuss some important features of the F0F1 system. First, the stoichiometry of transport seems to be three protons being transported for each ATP molecule synthesized. The mitochondrion has a transmembrane potential of about 180 mV, inside negative. This, as we saw in Section 2.2, is equivalent to a concentration gradient of about 103. In addition, the pH inside the mitochondrion is about one unit higher than in the cytoplasm, a concentration gradient of 10. The effective gradient across which the protons move is thus about 104. With a stoichiometry of 3, the overall transport of the protons involved in the synthesis of a single ATP molecule uses the equivalent of a gradient of 1012, just enough to synthesize an ATP molecule. Second, the mitochondrion uses this combination of transmembrane potential difference and the transmembrane concentration gradient. In the chloroplast (which we should note, in passing, has the gradient for protons set up in the opposite direction to that of the mitochondrion) the contribution

290 | Channels, Carriers, and Pumps

BOX 6.5 Single-Molecule Studies on the F0F1-ATPase The group of Yoshida, Kinosita, and their colleagues in Japan has given us a wonderful series of single-molecule studies that have gone far toward providing a detailed confirmation of Boyer’s “binding change” model for the action of the rotator ATPases. Their first breakthrough study was to show that the central stalk, the γ-subunit, indeed rotated about its own axis with respect to the α3β3 subunits of the F1 component (see the models of the whole F1F0 ATPase in Figure 6.24). They did this by binding the α3β3γ complex of F1 to a glass coverslip, when the β-subunits had been engineered to have a polyhistidine tail, and the coverslip had been coated with nickel-NTA (nitrilotriacetic acid) to which the histidines bound strongly. The γ-subunit was biotinylated enabling it to bind to streptavidin-bound, fluorescent-labeled actin molecules, thus forming a tail which could be directly visualized by fluorescence microscopy (see Figure 6.30 for the setup, and Figure 6.31 for the data). Figure 6.31 shows a sequential set of pictures showing how the actin filaments moved when 2 mM ATP was admitted to the system. In (A) the rotation axis was at the end of the filament, in (B) the less frequent cases where the middle of the filament happened to be the axis of rotation. In (C), the centroid of the filament in each picture of (A) is tracked as a set of linked lines, the first rotation seen as the heavy line, with the subsequent progress of the centroid as the continuing thin line. Clearly, the γ-filament is rotating with respect to the tethered α3β3 complex. Controls performed in the absence of ATP showed no continuous rotation, merely rocking Brownian movements, observed also when azide, a blocker of the ATPase was added together with the ATP. Thus the first part of the Boyer model had been verified...the F1 system is a rotator motor.

(A)

β (empty form)

(B) Actin filament

α

α

Streptavidin

γ α3β3γ complex β

α

β

α

His-tag Coverslip coated with Ni-NTA

β (AMP-PNP form)

β (ADP form)

α FIGURE 6.30 The molecular engine visualized: A, side-view. The coverslip coated with nickel that binds to the histidine-tagged F1 molecules, and with the central stalk γ-subunit being bound to actin. B, view from above, showing the α and β subunits and their rotation. Reproduced by permission from: Direct observation of the rotation of F1ATPase. Noji H, Yasuda R, Yoshida M, Kinosita K Jr. Nature. 1997;386:299
302.

(Continued )

Primary Active Transport Systems Chapter 6

|

291

BOX 6.5 (Continued)

FIGURE 6.31 The data from the experimental set-up shown as Figure 6.30. Successive views of the γ-subunit visualised as an actin “tail.” (A) The rotation axis at an edge. (B) the axis in the center. (C) Trace of the centroid of the successive pictures in (A). Reproduced by permission from: Direct observation of the rotation of F1-ATPase. Noji H, Yasuda R, Yoshida M, Kinosita K Jr. Nature. 1997;386:299
302.

A succeeding project in the group’s tour de force was to tackle the reverse process. Could they show synthesis of ATP from ADP and Pi, brought about by applying a forced rotation to the γ-stalk with respect to the α3β3 complex? They already had, as we have just seen, the entire F1 complex tethered to a glass slide. Now they attached a streptavidin-coated magnetic bead to the biotinylated γ-subunit, as in Figure 6.32 (A). Being magnetic, the bead could be rotated by an external magnet. A set of six electromagnets, arranged in a circle (C and, in D, seen from above) were sequentially induced, forcing the rotation of the bead (and hence the biotinylated γ-subunit) in one or other direction, depending on the clockwise or anticlockwise direction of the switching on of the electromagnets. The F1 molecules were placed within liquid droplets, as in the figure. Any synthesis of ATP was measured by added firefly luciferase, which emitted photons of light in proportion to the accumulation (or disappearance) of ATP. These were counted by an imageprocessing system. Figure 6.33 shows four data sets taken from their paper. At S, the bead was rotated in a direction that raised the light signal (the quantity of light depicted as the y-axis on the chart). At H, the rotation direction was reversed leading to hydrolysis of ATP and a reduction in the emitted light. At N, there was no rotation. The blue dashed lines show the results of inverting the chamber from its position in the solid blue lines. A clockwise rotation in the “solid blue” experiment became an anticlockwise rotation in the dashed lines experiment and the peaks in photon emission in one data set exactly matched the troughs in the inverted data set. Thus, a forced rotation of the γ-subunit led to the production of ATP (or to its hydrolysis, if the rotation direction was reversed). (Continued )

292 | Channels, Carriers, and Pumps

BOX 6.5 (Continued)

FIGURE 6.32 The experiment to test whether synthesis of ATP can be produced by forced rotation of the γ-subunit stalk. Reproduced by permission from: Mechanically driven ATP synthesis by F1-ATPase. Itoh H, Takahashi A, Adachi K, Noji H, Yasuda R, Yoshida M, Kinosita K. Jr. Nature. 2004;427:465
8.

A third study reached an even higher level of sophistication. This time, the ATP or ADP nucleotides were present as their fluorescent analogs, the Cy3-derivatives (cyanine dyes http://en.wikipedia.org/wiki/Cyanine#Cy3_and_Cy5). Binding of the fluorescent nucleotide to F1 could be followed by total internal reflection fluorescence (http://en.wikipedia.org/wiki/Total_internal_reflection_fluorescence_microscope). The set-up is illustrated in Figure 6.34(A). The system of electromagnets again enabled the rotation of the γ-subunit with respect to the tethered F1 complexes, but this time the rotation was maintained at a controlled rate. Part (B) of the figure shows a series of paired records in which the upper part of each pair shows the orientation of the magnetic bead while the lower part shows the corresponding fluorescence signal at that angle of orientation of the bead. Each pair of records is separated from the next by 0.167 s. The bead rotation rate was 0.2 Hz, or 5 s for a complete rotation, so that each pair of pictures is separated from its neighbor by 12 of rotation. One can see that, as the beads rotate, ATP (in this case) binds to F1 and is later released. Parts (C) and (D) of the figure show records of the light intensity as a function of time during the experiment, while the beads are rotating in the direction (Continued )

Primary Active Transport Systems Chapter 6

|

293

BOX 6.5 (Continued)

30

200 H

N

N

N 150

20

15

100

10 50

Total number of photons (103)

Number of photons detected (103)

H 25

H

H

H

S

S

S

S

5

0

0 0

1

2

3

4

5

6

7

8

9

10

12

13

FIGURE 6.33 Four data sets obtained using the experimental arrangement of Figure 6.32. At positions marked by N the magnetic field was not rotated; at S, rotated in a direction favouring synthesis; at H, favouring hydrolysis. The blue dashed line shows the experiment in which the system depicted as the blue solid line was inverted. Modified by permission from: Mechanically driven ATP synthesis by F1-ATPase. Itoh H, Takahashi A, Adachi K, Noji H, Yasuda R, Yoshida M, Kinosita K. Jr. Nature. 2004;427:465
8.

of hydrolysis (C) or synthesis (D). The horizontal lines mark the intensity of light signal for 0, 1, or 2 Cy3 molecules per F1 complex. By counting the number of on and off events per degree of rotation, and knowing the speed of rotation, Kengo and colleagues could compute the rate constants for the binding/debinding of ATP or ADP. A part of their findings is depicted below (Figure 6.35). In (A), the computed association constants (calculated as the ratio of the measured on-constant to off-constant) are plotted on a log scale on the y-axis as a function of the angle of rotation of the γ-subunit (with respect to the tethered α3β3 complexes). The data are presented for ATP and ADP alone and for each in the presence of Pi. (B) shows the progress of the reaction and the associated binding/debinding of the nucleotides, while (C) interprets and summarizes the data of (A) as a cartoon of the cycling of the three component α/β subunits each through the states of low affinity to ATP, high affinity and ATP binding, ATP debinding, and ADP binding and its debinding to complete the cycle. (The thick (Continued )

294 | Channels, Carriers, and Pumps

BOX 6.5 (Continued)

FIGURE 6.34 Controlled rotation of the γ-subunit of F1 with respect to the α3β3 complex. (A) The experimental set-up (B) successive records of, above, the γ subunit visualized by the magnetic bead and, below, the fluorescence. (C) and (D) are records of the flurorescence signal as a function of time during the rotation. Several cycles are shown. Reproduced by permission, Nature Group, from Controlled rotation of the F1-ATPase reveals differential and continuous binding changes for ATP synthesis Kengo Adachi, Kazuhiro Oiwa, Masasuke Yoshida, Takayuki Nishizaka, Kazuhiko Kinosita Jr Nature Commun.; 3, 1
12, 2012:1022. doi: 10.1038.

arrow represents the orientation of the γ-subunit as it is forced to orient through a cycle by the torque exerted by the electromagnets). Note the finding from (A): the rates of debinding differ little between ATP and ADP, but the rates of binding do differ. Importantly, phosphate blocks ATP binding at angles where ADP binding is essential for ATP synthesis. In synthesis rotation, the affinity for ADP increases by .104, followed by a shift to high ATP affinity, and finally the affinity for ATP decreases by .104. All these angular changes are gradual, implying tight coupling between the rotor angle and site affinities. Thus, these experiments triumphantly confirmed an essential aspect of Boyer’s binding change model. Rotation of the central γ-subunit stalk with respect to the α3β3 complex forces the rise and fall of the affinity of the nucleotides to the α3β3 subunits, with the change in affinities of the ATP and ADP molecules being displaced with respect to each other. Interestingly and unexpectedly, binding of inorganic phosphate (Pi) plays an important role in modifying the affinity of ATP to the α3β3 complex. (Continued )

Primary Active Transport Systems Chapter 6

|

295

BOX 6.5 (Continued)

FIGURE 6.35 Summary of data obtained using the setup depicted in Figure 6.34. A, the computed association constants for ATP, ADP, ATP 1 Pi, and ADP 1 Pi, as indicated. B, the progress of the hydrolysis or synthesis of ATP. C, a cartoon depiction of the changes of association constants as a function of the rotation angle. Reproduced by permission, Nature Group from: Controlled rotation of the F1-ATPase reveals differential and continuous binding changes for ATP synthesis Kengo Adachi, Kazuhiro Oiwa, Masasuke Yoshida, Takayuki Nishizaka, Kazuhiko Kinosita Jr Nature Commun.; 3, 1
12, 2012:1022. doi: 10.1038.

of the potential is small, that of the concentration gradient large. In bacteria, one or other of these factors can dominate the overall driving force for protons, depending on the circumstances. It would appear that the F0F1-ATPase readily adds up the two components of the overall driving force, the electrical and the chemical gradient of protons. Since the binding sites for the protons transported seem to be deep within the body of F0 it is possible that a good deal of the transmembrane potential might fall between the membrane surface and these binding sites. This would ensure that the protons partition between the bulk medium and these binding sites in accordance with this potential difference. With such a system, it is of no significance whether the electrochemical gradient of protons across the entire membrane is composed mainly of its electrical or its chemical component. In both

296 | Channels, Carriers, and Pumps

circumstances, the proton-binding site within F0 will see a far higher concentration of protons that exists in the medium facing the F1 portion of the pump. Third, we have not discussed the implications of the fact that there are two binding sites for ATP on each a chain of F1 and that there are three such units per intact F0F1-ATPase. The importance of the threefold organization of the F1 molecule is not at all clear, but since it is preserved universally in living systems, it must be profound. The bacterial flagellum is a similarly constructed molecule of threefold symmetry that uses the transmembrane proton gradient to drive its rotation in the membrane, suggesting that there may be some real evolutionary link between these two membrane functions! Fourth, some members of the F-ATPase family (found so far only for certain rather specialized bacterial systems) can use sodium ions rather than protons for the coupling of ion transport to ATP synthesis. This observation suggests strongly that the proton and sodium ions are playing equivalent roles in the overall process. For both, the synthesis of ATP is brought about by coupling the ion flow to conformational changes in the protein. It seems that it is not any special property of protons (e.g., that they can bind covalently to acid groups on the enzyme) that is needed for the energy coupling.

6.6 THE VACUOLAR PROTON-ACTIVATED ATPASE Both animal and plant cells contain vacuoles whose internal pH is lower than that of the cytoplasm, often by as much as three pH units. The pH is under strict control, the gradient increasing steadily as these vacuoles pass through their developmental histories, from the time that they are newly pinched off from the plasma membrane, through the stages of primary and then secondary endosomes, until they finally become lysosomes. We saw in Section 5.3.6 how such vacuoles are used for the storage of neurotransmitters, these weak bases being trapped within the vacuole by the prevailing low pH. The low pH is maintained by a proton pump, one that is far closer in structure and properties to the F0F1-ATPases that we have just discussed than to the vanadate-inhibitable enzymes of Sections 6.1
6.4. There is a family of these vacuolar enzymes or “V-ATPases.” (For a model of their structure, see Figure 6.24.) Like the F0F1-ATPases, they are inhibited by DCCD, and have a complex subunit composition. There is no evidence for the presence of a phosphorylated intermediate during ATP hydrolysis by the V-ATPases.

6.7 BACTERIORHODOPSIN: A LIGHT-DRIVEN PROTON PUMP In lakes and ponds having high salt content, bacteria can be found (e.g., Halobacterium halobium) that have evolved the ability to harness light energy directly into osmotic energy. They do this using a simple yet

Primary Active Transport Systems Chapter 6

|

297

remarkable proton pump, powered by the absorption of light by a chromophore that is that same pigment, retinal, that the eye uses in the visual process. The retinal
protein complex that performs this light absorption in the eye is known as rhodopsin. The analogous protein in the bacterium, that absorbs light and pumps protons, is called bacteriorhodopsin. Bacteriorhodopsin is a small protein of molecular mass 26,000 Da, containing seven hydrophobic polypeptide sequences that straddle the membrane. A purple membraneous patch of bacteriorhodopsin, when isolated from the bacterium, is bleached when light shines on it. Walter Stoeckenius, working at the Rockefeller Institute in New York, found that, during this bleaching, protons were released from the patch toward the face that was originally at the exterior of the bacterium. If left to itself, the bacteriorhodopsin regenerates its purple color, in the process taking up protons from what was the cytoplasmic surface of the patch. Detailed analysis has shown that the absorption of a photon is followed by a sequence of changes in the spectral properties and hence the structure of the bacteriorhodposin (bR), the sequence closely paralleling that followed by the visual pigment, rhodopsin, when it absorbs light! Figure 6.36 shows the sequence and depicts also the change of

N–(lys)-opsin

13–cis K590 L550 H+ Medium 10 ps

M410

Cytoplasm

Light O640

H+

bR570

N–(lys)-opsin

Retinal trans

FIGURE 6.36 The light-induced cycle of events in bacteriorhodopsin. In the native state of bacteriorhodopsin, the form bR570, the retinal is in all-trans conformation as depicted on the lower part of the figure. The molecule absorbs a quantum of light to give the 13-cis-retinal (upper sketch in diagram). As the energy so absorbed is released, the conformation of the protein cycles through the forms labeled K590, etc. The subscripts refer to the wavelength (in nanometers) at which the protein maximally absorbs light. Taken, with kind permission, from F. M. Harold, “The Vital Force: A Study in Bioenergetics.” Copyright 1986 by W. H. Freeman and Company.

298 | Channels, Carriers, and Pumps

structure of the pigment retinal when it absorbs a photon, altering in structure from an all-trans to a 13-cis configuration. The form labeled bR570 is the stable form, purple in color and with protons bound to the protein. The transient intermediates K590 and L550 are also protonated, with the proton now accessible at the outside of the bacterium, while M410 is deprotonated. (The subscripts refer to the wavelength (in nanometers) at which the protein maximally absorbs light.) Somewhere between L550 and M410 the proton is expelled into the medium. It is clear from low-temperature infrared studies that certain carboxyl groups belonging to aspartyl residues of the bacteriorhodopsin are alternately charged and discharged during a cycle of proton pumping. These data are consistent with the following model of this proton pump (the reader should refer to Figures 6.36 and 6.37). In the form bR570 in Figure 6.36, a proton is bound to the aldehyde-amino link (a so-called Schiff’s base) that holds the retinal to a lysine residue of the bacteriorhodopsin (Figure 6.37 shows an early, schematic view of the process). The charge on this proton suffices to form a salt bridge with the negatively charged carboxyl group of the aspartyl residue labeled Asp-2 in Figure 6.37. The retinal is in the all-trans configuration. When it absorbs a photon, its conformation alters to 13-cis, the salt bridge with Asp-2 is broken and the proton is carried through the protein (and hence effectively across the membrane) to the region around Asp-1, where a salt bridge is formed once again. This is now the form K of Figure 6.36. As the protein relaxes through forms L, M, and O the protons are released from Asp-1 to the exterior and further protons are taken up by Asp-2 from the interior. The retinal returns to its all-trans configuration, with the establishment of the salt bridge again in conformation bR. With the crystallization and structure determination of bacteriorhodopsin (see 1MOL in the Protein Data Bank), this early scheme can be put onto a firmer foundation. Figure 6.38 shows a view of the bacteriorhodopsin molecule in the trimeric form in which it crystallizes, viewed from the side (top, left) or from the surface (top, right). In the surface view one can see the bound retinal molecule in each subunit of the trimer. The retinal is perhaps more easily seen in the view from above of a single monomer (bottom picture). A current description of the mechanism of the coupling of assimilation of a photon by retinal with the pumping of a proton across the membrane sees the proton as moving through a chain of linked water molecules, where the proton resides within an H-bonded network consisting of water molecules, stabilized by aspartate residues E204, E194, and arginine R82. Such a proton would behave similarly to excess protons in water, which can be delocalized by being shared between two H2O molecules in an H5O2 complex, a Zundel cation (see the article by Garczarek and his colleagues, cited in Figure 6.39). The H bonds in the Zundel cation are characterized by low-to-nonexistent

Primary Active Transport Systems Chapter 6

|

299

FIGURE 6.37 Path of the proton through bacteriorhodopsin. A tentative scheme for the movement of protons through the protein during the photon-linked isomerization of the protein. The seven cylinders represent the seven transmembrane helices of bacteriorhodopsin. Asp1 and Asp2 are two aspartyl residues that are thought to be successfully protonated and deprotonated during the photocycle and indicate the route of proton transfer through the molecule. The retinal is seen as the horizontal double-bonded structure, forming a Schiff’s base with a lysine residue of the protein. Taken, with kind permission, from Gerwert and Hess (1988) Prog Clin Biol Res. 1988;273:321
6.

barriers for proton transfer and short equilibrium distances. Figure 6.39 (top left) depicts a molecule of bacteriorhodopsin with internally located water molecules (the red spheres), holding an additional proton in such a Zundel cation, with two of the water molecules seen in schematic form in Figure 6.39 (top right). The bacteriorhodopsin molecule is conceived of as being composed of two halves. The depiction in part (A) of Figure 6.39

300 | Channels, Carriers, and Pumps

FIGURE 6.38 Three views of bacteriorhodopsin, 1MOL in the Protein Data Bank.

shows the water molecules held in the lower half. Part (C) of the figure shows the results of a molecular dynamics computation that located water molecules in the upper half of the bacteriorhodopsin molecule after retinal had assimilated a photon, and undergone the conformation change that opened a “gate” between the upper and lower halves. A cycle of light absorption and coupled proton pumping has occurred. The electrochemical gradient of protons that bacteriorhodopsin establishes with the absorption of light is itself harnessed by the cell in a secondary transport system, an antiport of sodium and protons, whereby sodium is pumped out of the bacterium, maintaining its osmotic balance. The resulting electrochemical gradient of sodium is used to concentrate metabolites within the cell. In addition, this cell possesses a light-driven chloride pump, known as halorhodopsin, which pumps chloride into the cell. The interrelationship of these many metabolic systems in the bacterium is only now beginning to be understood. Finally, consider the figure in part (D) of Figure 6.39. This depicts the structure of the bacteriorhodopsin molecule and three other bacterial rhodopsins with various functions, superimposed. These are all members of the enormous family of 7-helix proteins that includes bovine rhodopsin (1U19), squid rhodopsin (2Z73), the β2-adrenergic receptor (2RH1), the β1-adrenergic receptor (2VT4), and also the A2A adenosine receptor (3EML). Therefore,

Primary Active Transport Systems Chapter 6

|

301

FIGURE 6.39 (A) View of the extracellular side of bR, showing water density clusters IV, V, and VI. (These water density clusters characterize the space occupied by the fluctuating water molecules in their positions as suggested by molecular dynamics computations.) The red balls demonstrate the discrete positions of water molecules taken from x-ray structure. Water cluster V is located in the region suggested for the protonated water cluster, schematized in (B). Taken, with permission, from Florian Garczarek, Leonid S. Brown, Janos K. Lanyi, Klaus Gerwert, Proton binding within a membrane protein by a protonated water cluster, Proc Natl Acad Sci USA. 2005; 102: 3633
3638. (C) Channel opening in a simulation with deprotonated D96 and protonated D85 and D115. The snapshot depicts the last of an 87 ns simulation of the photon assimilation and coupled proton transport event. The model is viewed from cytoplasmic side. Six water molecules are displayed in the D96-K216 (upper) cavity (red ball). The pink surface depicts a contiguous water channel, connecting the cytoplasm with K216 and passing by D96. Four water molecules were in the channel, forming a hydrogen-bonded water chain. Taken, with permission, from Wang T, Sessions AO, Lunde CS, Rouhani S, Glaeser RM, Duan Y, Facciotti MT. Deprotonation of D96 in bacteriorhodopsin opens the proton uptake pathway. Structure. 2013;21:290
7. (D) Structural similarity among various bacterial rhodopsins in the seven-helix family. Bacteriorhodopsin (PDB code; 1M0L), halorhodopsin (1E12), sensory rhodopsin II (1JGJ), and xanthorhodopsin (3DDL) are aligned using residues 11
29, 43
59, 82
98, 110
124, 137
153, 170
189, and 205
223 of bacteriorhodopsin and equivalent residues of others and the proteins shown colored in blue, light green, yellow, and coral respectively. Taken, with permission, from Teruhisa Hirai, Sriram Subramaniam, and Janos K. Lanyi, Structural snapshots of conformational changes in a seven-helix membrane protein—lessons from bacteriorhodopsin. Curr Opin Struct Biol. 2009; 19: 433
439.

this 7-helical structure has been adapted to a wide variety of signaling processes in bacteria and eukaryotes, ranging far beyond the photon assimilating systems of the rhodopsins—see the article by Teruhisa Hirai and his colleagues cited in the legend to Figure 6.39.

302 | Channels, Carriers, and Pumps

6.8 MDR—DRUG PUMPS MDR (multidrug resistance) is a phenomenon that is observed both in microorganisms, such as bacteria, fungi, and parasites that have become insensitive to various antibiotics, and in tumor cells, where MDR is a major impediment to the effective chemotherapeutic treatment of cancer. The mechanism of MDR has been investigated for more than 40 years, and now we know that it is mainly—though not merely—a matter of membrane transport!

6.8.1 The Discovery of MDR The first breakthrough in MDR research came in 1973, when the Danish physician, Keld Danø showed that daunomycin (also known as daunorubicin)—a widely used anticancer agent—is extruded from multidrug resistant cells via an energy-dependent mechanism (Danø K. BBA 1973; 323: 466
483). Danø studied the transport kinetics of daunomycin and observed the following: 1. Daunomycin accumulation was higher in isolated nuclei than in whole cells, and similar in nuclei from drug-sensitive and drug-resistant cells (suggesting that the mechanism of resistance was not associated with the nucleus). 2. The intracellular steady-state accumulation of drug was lower in drugresistant cells than in the corresponding wild-type cells (implying lower permeability in drug-resistant cells, see Figure 6.40A). 3. Metabolic inhibitors, such as 2-deoxyglucose or azide, were able to increase the intracellular concentration of daunomycin in drug-resistant cells to the level observed in drug-sensitive cells (indicating an energydependent transport mechanism). 4. Structural analogs, such as N-acetyl daunomycin, enhanced the accumulation of daunomycin in MDR cells (indicating competition between closely related substrates for a common efflux pathway). 5. Incubation with drugs to which the cells were cross-resistant, such as vincristine and vinblastine, also increased the steady-state accumulation of daunomycin (suggesting that competition for transport could be extended also to structurally unrelated compounds). To explain the above observations, Danø proposed the simple leak and pump model (Figure 6.40B), in which drug enters the cell by passive diffusion, but before it can reach the nucleus it is extruded by an active efflux mechanism.

Primary Active Transport Systems Chapter 6

|

303

(A)

(B)

FIGURE 6.40 (A) Visualization by confocal microscopy of accumulation of daunorubicin in drug sensitive MCF-7 breast cancer cells (left) and the corresponding MDR MCF-7 AdVp cell line (right). The sensitive cells quickly accumulate the (red) fluorescent drug, which binds to both the plasma membrane and to the nucleus. In contrast, the MDR cells are virtually empty. Original micrographs from the author’s experiments, performed at NCI in collaboration with Dr. Susan Bates. The MCF-7 AdVp cells were selected in the presence of adriamycin and verapamil in the medium, and express ABCG2 (MXR), a recently identified multidrug transporter. (Litman T, Druley TE, Stein WD, Bates SE. CMLS 2001; 58: 931
59). For a “live” time-lapse view of fluorescent drug accumulation in sensitive and drug resistant colon cancer cells, please see Video Clip 1 on the companion website listed on page ii in the frontmatter of this book. (B) The leak and pump model as its “father,” Keld Danø is depicted imagining it: The anticancer drug diffuses into the cell, but before it can enter the nucleus, it is pumped out. The figure is from the Journal of NIH Research (December 1994), which nominated Danø’s discovery of the multidrug resistance mechanism as a milestone in the history of cancer research.

304 | Channels, Carriers, and Pumps

Such an active mechanism was indeed identified in 1979 by Jack Riordan and Victor Ling at the University of Toronto.1 It was named P-glycoprotein (P for permeability, often abbreviated P-gp), and is a 170 kDa membrane protein highly expressed in MDR cells, but not in sensitive cells. The gene encoding P-gp, MDR1 (human) and mdr1 (mouse) (or, following the HUGO nomenclature: ABCB1, see Table 6.1 for details), was cloned independently in 1986 by Igor Roninson and colleagues at Michael Gottesman’s lab at NCI,2 and by Philippe Gros et al. at McGill University.3 The same two groups soon demonstrated that transfection with the mdr1 gene was sufficient to confer the full phenotype of MDR to hitherto drug-sensitive cells, thus providing the ultimate proof that P-gp can cause MDR.4

6.8.2 The ABC Superfamily Based on the mRNA sequence data, the primary and secondary structure of P-gp could be deduced and it appeared to be a member of the ATP-binding cassette (ABC) superfamily of transporters, which includes numerous conserved proteins, most of which are involved in transfering molecules (such as amino acids, peptides, sugars, metals, vitamins, lipids, and drugs) across cell membranes. Homology analyses have identified several hundred different ABC transporters: from E. coli to man. The genome sequence of E. coli codes for more than 80, while the human genome contains 48 ABC proteins. Some are specific for a single substrate or even act as an ion channel, like the cystic fibrosis transmembrane regulator (CFTR). Others—like P-gp—display very broad substrate specificity. Table 6.1 lists the 48 members of the human ABC transporter family, including their function (if known), phenotype and disease association.

6.8.3 Topology Common to all ABC transporters is a conserved cytosolic nucleotide-binding domain (NBD) and a less-conserved transmembrane domain (TMD, also called the membrane-spanning domain, MSD). The NBD domain contains the catalytically active Walker A (GXXGXGKS/T) and Walker B (φφφφD, where φ is a hydrophobic residue) motifs, which are common for all proteins that bind ATP. Between these motifs, ABC transporters also have a unique C signature (LSGGQ) as well as a flexible Q-loop (containing a glutamine, Q), which is presumably involved in the interaction between the NBD and TMD. The TMD is embedded in the membrane and typically consists of six alpha 1. 2. 3. 4.

Riordan JR & Ling V. J Biol Chem 1979; 254: 12701
5. Roninson IB et al. PNAS USA 1986; 83: 4538
52. Chen C-J et al. Cell 1986; 47: 381
9. Gros P et al. Nature 1986; 323: 728
31. Ueda K et al. PNASA USA 1987; 84: 3004
8.

Primary Active Transport Systems Chapter 6

|

305

TABLE 6.1 Human ABC Transporters Transporter (Synonym)

Protein Size (Amino Acids)

Phenotype

Function/ Substrates

ABCA1 (ABC1)

2261

Mutations associated with Tangier disease and familial HDL deficiency

Effluxes PS and cholesterol from macrophages

ABCA2

2436

Colocalizes with betaamyloid; perhaps protective role in AD

Transports cholesterol into ER for esterification

ABCA3

1704

Mutations associated with surfactant deficiency and abnormal lamellar bodies

Expression induced by glucocorticoids (GRE promoter motif). Transports lipids and cholesterol

ABCA4

2273

Mutations found in Stargardt disease-1 Also associated with retinitis pigmentosum

Retina-specific. A N-retinylidene-PE flippase

ABCA5

1642

Knockout mice develop lysosomal disease symptoms

Lysosome-specific. Autolysosome role?

ABCA6

1617

Possible role in macrophage lipid homeostasis

ABCA7

2146

Expressed in myelolymphatic tissue Transports cholesterol

ABCA8

1581

Xenobiotic transporter

ABCA9

1624

Possible role in monocyte differentiation and lipid homeostasis

ABCA10

1543

May play a role in macrophage lipid homeostasis

ABCA11P

0

Pseudogene

ABCA12

2595

Mutations cause harlequin ichthyosis. Loss of function results in impaired lamellar membrane formation in stratum corneum

Found in lamellar granules in keratinocytes Transports lipid glucosylceramides to the extracellular space (Continued )

306 | Channels, Carriers, and Pumps

TABLE 6.1 (Continued) Transporter (Synonym)

Protein Size (Amino Acids)

Phenotype

Function/ Substrates

ABCA13

5058

Mutations are linked to schizophrenia, bipolar disorder, and depression

Largest ABC protein known. Unknown substrate

ABCB1 (MDR1, Pgp)

1280

P-glycoprotein (P-gp) is the main cause of multidrug resistance (MDR) in cancer

Xenobiotic transporter. Integral part of the blood
brain barrier

ABCB2 (TAP1)

808

ABCB3 (TAP2)

653

Mutations are associated with ankylosing spondylitis, diabetes, and celiac disease

TAP1 and TAP2 are half-transporters forming a heterodimer that pumps peptides across the ER for presentation to MHC Class I molecules

ABCB4 (MDR2, MDR3, PGY3)

1279

Mutations associated with progressive familial intrahepatic cholestasis (PFIC)

PC (phospatidylcholine) flippase. Involved in biliary PC secretion

ABCB5

1257

May cause resistance to doxorubicin in melanoma

Effluxes rhodamine and doxorubicin

ABCB6

842

The Langereis (Lan) blood group antigen at the plasma membrane of erythrocytes

Half-transporter involved in heme uptake in mitochondria

ABCB7

753

Mutations associated with X-linked sideroblastic anemia and with ataxia

Half-transporter, heme transport from mitochondria to the cytosol

ABCB8

735

Half-transporter, mitochondrial

ABCB9

766

Half-transporter, translocates peptides from cytosol to the lysosomal lumen

ABCB10

738

Half-transporter, mitochondrial (Continued )

Primary Active Transport Systems Chapter 6

|

307

TABLE 6.1 (Continued) Transporter (Synonym)

Protein Size (Amino Acids)

Phenotype

Function/ Substrates

ABCB11 (BSEP)

1321

Mutations cause PFIC-2

The major canalicular bile salt export pump

ABCC1 (MRP1)

1531

Involved in MDR in cancer

Organic anion transporter, e.g., glutathione conjugates (LTC4)

ABCC2 (MRP2, CMOAT)

1545

Mutations associated with Dubin
Johnson syndrome

Expressed in canalicular part of the hepatocyte Transports anionic conjugates (bilirubin)

ABCC3 (MRP3)

1527

May be involved in drug resistance in cancer

Transports glucuronides and bile salts

ABCC4 (MRP4)

1325

May be involved in drug resistance

Organic anion transporter; cyclic nucleotides, e.g., PMEA

ABCC5 (MRP5)

1437

May be involved in drug resistance

Organic anion transporter; cyclic nucleotides, e.g., PMEA

ABCC6 (MRP6)

1503

Mutations cause pseudoxanthoma elasticum (PXE)

Transports anionic peptides and GSH conjugates (BQ123)

ABCC7 (CFTR)

1480

Mutations cause cystic fibrosis (CF)

chloride channel

ABCC8 (SUR1)

1581

Mutations associated with neonatal diabetes mellitus

Subunit of beta-cell ATP-sensitive K channel

ABCC9 (SUR2)

1549

Mutations associated with cardiomyopathy

Subunit of muscle ATP-sensitive K channels

ABCC10 (MRP7)

1464

May be involved in drug resistance in cancer

Transports lipophilic anions, e.g., glucuronide conjugates; paclitaxel and taxanes (Continued )

308 | Channels, Carriers, and Pumps

TABLE 6.1 (Continued) Transporter (Synonym)

Protein Size (Amino Acids)

Phenotype

Function/ Substrates

ABCC11 (MRP8)

1382

Lipophilic anion transporter. Probably secretes of earwax!

ABCC12 (MRP9)

1359

Candidates for paroxysmal kinesigenic choreoathetosis

ABCC13 (MRP10)

325

pseudogene, truncated

No ATP-binding site, no transport activity

ABCD1 (ALD)

745

Mutations cause adrenoleukodystrophy

Peroxisomal halftransporter. Imports fatty acids (e.g., VLCFA) and acyl-CoAs

ABCD2 (ALDL1)

740

Associated with adrenoleukodystrophy and Zellweger syndrome

Peroxisomal halftransporter. Possible dimerization partner of ABCD1 and other ABCs

ABCD3 (PXMP1, PMP70)

659

Mutations associated with Zellweger syndrome

Peroxisomal halftransporter. Possible dimerization partner of ABCD1 and ABCD2

ABCD4 (PXMP1L)

606

May be associated with adrenoleukodystrophy

Peroxisomal halftransporter

ABCE1 (RNASELI)

599

Antagonizes antiviral effect of the interferon-regulated 2
5A/RNase L pathway

Probably heterodimerizes with RNASEL, blocking its activity

ABCF1

807

Regulated by TNF-α, perhaps involved in inflammation

No TM domain, probably not involved in transport

ABCF2

623

No TM domain, probably not involved in transport

ABCF3

709

No TM domain, probably not involved in transport

Derived of duplication with ABCC11

(Continued )

Primary Active Transport Systems Chapter 6

|

309

TABLE 6.1 (Continued) Transporter (Synonym)

Protein Size (Amino Acids)

Phenotype

Function/ Substrates

ABCG1 (ABC8)

638

Deletion (ko mice) causes accelerated atherosclerosis

Half-transporter. Macrophage phospholipid and cholesterol efflux pump

ABCG2 (MXR, BCRP, ABCP)

655

Involved in MDR in cancer. Mutations associated with hyperuricemia and gout

Half-transporter, homodimer, xenobiotic efflux pump. Also transports uric acid. Important part of the placenta barrier

ABCG4

627

ABCG5

651

ABCG8

673

May be involved in macrophage lipid homeostasis Mutations associated with sitosterolemia. Required for cholesterol secretion into bile

ABCG5 and ABCG8 are sterol halftransporters

An overview of the 48 known human ABC transporters grouped according to their seven (A-G) subfamilies (the table is based on the original compilation by Michael Mu¨ller, available at http://nutrigene.4t.com/humanabc.htm) and named according to the official HUGO (Human Genome Organization) nomenclature. The table lists 50 genes, but ABCA11P and ABCC13 are pseudogenes and should therefore not count as true ABC transporters. Note that the size of the protein may vary due to alternative splicing resulting in multiple transcript variants.

helices. A fully functional ABC transporter, such as P-gp, consists of 2 NBDs and 2 TMDs. Other ABC proteins, such as ABCG2 (which also can cause MDR), are “half-transporters.” That is, they need to homo- or heterodimerize in order to form a functional unit. Figure 6.41 shows the topology of three typical ABC transporters based on their mRNA sequence and hydrophobicity plot (such as those shown in Figure 1.6) analysis.

6.8.4 Function When the molecular biology of MDR had been solved, the next burning question was: How does the transport actually take place? That is: what is the protein’s structure
function relationship? And how is energy coupled to the translocation process? The short answer is: we do not know all the

310 | Channels, Carriers, and Pumps TMD0

TMD1

TMD2

N

ABCC1

ATP

ATP

NBD1

NBD2

ABCB1

ATP

N

c

TMD2

TMD1

ATP

NBD2

NBD1

c

TMD

c

ABCG2 ATP

NBD

N

FIGURE 6.41 Cartoon showing the typical domain organization of the three most important ABC transporters involved in drug resistance. ABCC1 (MRP1) contains, in addition to the TMD1-NBD1-TMD2-TMD2 topology, an extra N-terminal 5-helix TMD0, which apparently is not critical for its transport function. ABCB1 (P-gp) has the TMD1-NBD1-TMD2-TMD2 topology in a single polypeptide chain, with two homologous halves (suggesting its evolution from a single half-structure by gene duplication). ABCG2 (MXR) is a half-transporter that needs to homodimerize to form a functional protein, and with reverse topology: NBD-TMD-NBD-TMD when compared to P-gp. Each TMD typically consists of 6 hydrophobic alpha helices that associate to form a hydrophobic core (and pore) that binds to, and moves, compounds across the membrane. The two black bars in each NBD illustrate the cytoplasmic, conserved Walker A and B motifs that bind and hydrolyze ATP, thereby powering the substrate translocation process. In bacteria, the functional unit is often a tetramer, where the individual TMD and NBD domains are encoded as separate proteins.

molecular events yet! But based on experimental evidence we can provide some educated guesses. First, a few words on the enigmatic transport function of P-gp. A function, which has not yet been resolved, and this in spite of numerous mutational studies that have identified residues involved in determining the substrate specificity and transport activity of P-gp.5 This lack of knowledge is mainly due to absence of high-resolution structural information on human P-gp (the protein is extremely difficult to crystallize, as is the case for many other membrane proteins). What is so peculiar about P-gp is its extremely wide substrate specificity. It seems to be able to transport hundreds—if not thousands—of structurally

5. In particular, the comprehensive cysteine-scanning mutagenesis of all transmembrane segments of P-gp, performed by Loo and Clarke, should be noted. For a review, see Loo TW & Clarke DM. J Membr Biol 2005; 206; 173
85.

Primary Active Transport Systems Chapter 6

|

311

unrelated compounds and to be affected by an even wider range of chemosensitizers, i.e., molecules that are able to inhibit its transport. Why is this protein so promiscuous? It does not easily fit (!) the simplistic “lock and key” interpretation of enzyme kinetics (one substrate for one enzyme). Rather, it seems to accommodate a wide range of very different keys, so its lock mechanism is probably better explained as a large, pliable domain, with multiple overlapping substrate- and inhibitor-binding sites. Importantly, these binding sites appear to be localized in the lipid bilayer of the plasma membrane, where the substrates, which are all hydrophobic, partition at high concentrations, allowing for a high Km for P-gp interaction. Thus, P-gp differs from conventional pumps, such as P-type ATPases that efflux water-soluble substrates (such as Na1, K1, Ca21, or protons) from one side of the membrane to the other (Figure 6.42A). Rather, it seems to act as a hydrophobic vacuum cleaner (Figure 6.42C), a term coined by Yossi Raviv and colleagues in 1990, based on a study using the lipophilic, photosensitive probe 5-[125I]Iodonaphtalene-1-azide (INA).6 However, INA is a hydrophobic probe, and—when photoactivated—it selectively binds to protein domains embedded in the plasma membrane. Raviv et al. used doxorubicin as a photosensitizer to photoactivate INA, and demonstrated that in MDR cells only P-gp was labeled specifically by INA, while in drug-sensitive cells many different proteins were labeled (Figure 6.42D). This remarkable difference shows that doxorubicin distributes in a diffuse manner in plasma membranes prepared from sensitive cells (which do not contain P-gp), whereas in MDR membranes, the drug is concentrated in the hydrophobic domains by P-gp. Therefore, the interaction between doxorubicin and P-gp must take place directly within the plasma membrane. Strong supporting evidence for P-gp mediated efflux directly from the cell membrane was provided in 1993 in a fluorescence study by Balazs Sarkadi and colleagues at The National Medical Center in Hungary.7 They showed that several fluorescent probes, such as fura-2, BCECF, and calcein, are readily taken up by sensitive cells, but not by P-gp-expressing MDR cells. These probes enter the cell in their membrane permeant ester form (as the nonfluorescent acetoxymethyl ester, AM), and once inside, they are immediately hydrolyzed by endogenous cytoplasmic esterases to the fluorescent, negatively charged acid form, which is not a substrate for P-gp. Therefore, the AM compounds must have been extruded before they could reach, and be cleaved by, cytoplasmic esterases. Thus, the transport must have taken place from within the membrane, in support of the hydrophobic vacuum cleaner model. Today, calcein-AM is widely used as a simple, sensitive and robust (it is neither sensitive to intracellular pH such as BCECF) nor to calcium (as fura-2)) screening tool for 6. Raviv Y, Pollard HB, Bruggemann EP, Pastan I, Gottesman MM. 1990; 265: 3975
80. 7. Homolya L, Hollo Z, Germann UA, Pastan I, Gottesman MM, Sarkadi B. JBC 1993; 268: 21493
6.

312 | Channels, Carriers, and Pumps

FIGURE 6.42 Three models for the mechanism of action of P-gp. For simplicity, the ATPbinding domains are not shown in this drawing. o is the outer, and i is the inner leaflet of the membrane bilayer. (A) Pump. A conventional pump, such as a proton pump, effluxes its watersoluble substrate across the lipid bilayer without the requirement for any substrate-lipid interaction. (B) Flippase. First, the hydrophobic drug intercalates between the lipids in the inner leaflet of the membrane. Next, P-gp flips the compound to the outer leaflet, where it is in equilibrium with extracellular drug. Strictly speaking, this mode of action is properly referred to as a floppase (floppases move phospholipids, such as phosphatidylcholine and sphingolipids from the inner to the outer leaflet of the plasma membrane, while flippases flip phosphatidylserine and phospatidylethanolamine from the outer to the inner leaflet. (For details on flippases and floppases, see Clark MR. Flippin’ lipids. Nature Immunology 2011; 12: 373
5.) (C) Hydrophobic vacuum cleaner. As for the flippase, the drug first partitions into the membrane bilayer, from which it is directly extruded into the extracellular phase. Thus, P-gp here acts as a phase separator, removing the drug from the hydrophobic membrane, and putting it into the hydrophilic (aqueous) medium. (D) Experimental evidence in favor of both the flippase and hydrophobic vacuum cleaner model. Selective INA-labeling of P-gp in living cells by energy transfer from doxorubicin. Multidrug-resistant (R) or sensitive (S) cells were incubated with [125I]INA and doxorubicin followed by irradiation with visible light for doxorubicin-induced photosensitized labeling with INA. The figure shows an autoradiogram of a 10% SDS-PAGE gel loaded with resistant (R) or sensitive (S) cells after labeling. Sensitive cell membranes display a diffuse labeling pattern, while MDR cells only label one membrane protein at 170 kDa: P-gp! Figure redrawn from Raviv Y, Pollard HB, Bruggemann EP, Pastan I, Gottesman MM.JBC 1990; 265: 3975
80.

P-gp
drug interactions; the higher the intracellular fluorescence observed, the stronger the inhibition of P-gp transport activity (see Figure 6.43 for an outline of the assay principle). An alternative model for the function of P-gp, which is compatible with drug efflux directly from the plasma membrane, is the flippase model (Figure 6.42B), which Higgins and Gottesman proposed already in 1992.8 According to this model, P-gp works like a phospholipid flippase—or rather: floppase—that moves phospholipids from the inner to the outer leaflet of the plasma membrane. Indeed, the closest relative to P-gp, namely ABCB4 (formerly known as MDR2, with 78% sequence identity to ABCB1) has been identified as a liver-specific PC flippase. Because the vast majority of P-gp substrates are hydrophobic molecules with high 8. Higgins CF & Gottesman MM. Is the multidrug transporter a flippase? Trends Biochem Sci 1992; 17: 18
21.

Primary Active Transport Systems Chapter 6

|

313

FIGURE 6.43 Calcein-AM efflux assay to study P-gp transport. (A) When nonfluorescent calcein-AM (C-AM) is added to the medium, it will diffuse through the plasma membrane and into the cytosol. Here it is quickly hydrolyzed by intracellular esterases into free, fluorescent calcein, (symbolised as star-surrounded spheres) which cannot cross the membrane. Cells that express functional P-gp will extrude C-AM from the plasma membrane, before it is cleaved to free calcein, which is not a P-gp substrate. (B) If a reverser, R, (i.e., a chemosensitizer, a compound that interferes with P-gp’s transport function) such as verapamil is added to the incubation medium (here, at time 10 min), it will block P-gp’s outward transport of C-AM. C-AM will be converted to free calcein, whose intracellular concentration will rapidly increase (here, measured as an increase in AFU [arbitrary fluorescence units]). This can readily be observed as an increase in fluorescence, and monitored in e.g. a microplate reader.The graph illustrates an experiment, where we incubated P-gp expressing Ehrlich ascites tumor cells together with test compounds, such as cyclosporin A, verapamil, and vinblastine, to study these drugs’ effect on calcein accumulation. The experiments are compatible with pumping (see Box 6.6) of calcein-AM from the plasma membrane (Litman T, Skovsgaard T, Stein WD. Pumping of drugs by P-glycoprotein: A two-step process? JPET 2003; 307: 846
53.)

affinity for phospholipids, they will partition directly into the lipid bilayer. Therefore, if the spontaneous flip rate is slow compared to the P-gpmediated “flop” rate, and the drugs in the two bilayers are in equilibrium with the aqueous phase on each side of the membrane, then an outward flipping of drug substrate would result in a reduced intracellular drug accumulation. Such a model could also—at least partially—account for the polyspecificity of P-gp: Assuming a two-step recognition mechanism, the primary determinant of specificity would then be the ability of the drugs to intercalate into the lipid bilayer; the second determinant would be interaction with a relatively non-selective, substrate-binding domain in P-gp. As we can see from Figure 6.42, the flippase and the vacuum cleaner models are fairly alike: The flippase moves drug to the outer leaflet

314 | Channels, Carriers, and Pumps

from which it rapidly equilibrates with the extracellular medium, whereas the vacuum cleaner extracts drug from the membrane directly into the extracellular aqueous phase, where it repartitions into the outer half of the membrane. Therefore, in each case a similar and rapid equilibrium state is reached, which makes it rather difficult to distinguish between the two models. From an energetic point of view, however, flipping of a lipophilic substrate into the outer leaflet appears more favorable than its direct extraction into a hydrophilic water phase. Finally, evidence for P-gp mediated flippase activity has been provided by Frances Sharom and coworkers at University of Guelph, Ontario, Canada. They showed that purified, reconstituted P-gp can translocate a number of fluorescent, NBD-labeled phospholipids from the inner to the outer leaflet of P-gp enriched proteoliposomes, and that the transport is inhibitable, both by vanadate (inhibiting ATP hydrolysis) and by P-gp substrates in a dose-dependent manner.9 Among the molecules that have been shown to be transported by P-gp are platelet-activating factors (PAFs) and steroid hormones, which are speculated to be endogenous, physiological substrates for P-gp. This may also account for the high basal activity of P-gp, in the absence of any added exogenous substrate.

6.8.5 ATPase Activity Now that we have suggested how the TMDs of P-gp might work let us turn to the cytoplasmic NBDs. These constitute an important feature of P-gp, as they bind and cleave ATP—a process that directly triggers the transient conformational changes resulting in drug transport. That drug transport and ATPase activity are tightly coupled is evident from experiments showing that the basal ATPase activity of P-gp is stimulated in the presence of transport substrates. Some typical P-gp ATPase stimulation readouts are illustrated in Figure 6.44, which shows that both transport substrates, such as progesterone (Figure 6.44B), and known chemosensitizers, such as verapamil (Figure 6.44A), stimulate the ATPase activity at low concentrations but inhibit it at high concentrations. Such a biphasic behavior (the characteristic bell-shaped ATPase activity profile) can be explained by the presence of two drug-binding sites: one of high affinity, activating, and the second, low-affinity site, inhibiting the ATPase. Other agents, such as cyclosporin A and gramicidin S, seem to act as pure inhibitors of the P-gp ATPase (Figure 6.44C). Therefore, one cannot directly infer from a compound’s effect on P-gp’s ATPase activity, whether it is a transport substrate or a transport inhibitor, only that it has affinity for and interacts with P-gp as assessed at the level of the catalytic site. 9. See Sharom FJ. Frontiers in Oncology 2014; 4: 1
19.

Primary Active Transport Systems Chapter 6

315

|

350 Basal activity, V0 Maximum activity, V1 Increase in activity, V1–V0

6

300 250

40 °C

200 37 °C

150 100

30 °C

50

loge (ATPase activity)

ATPase activity (nmol Pi /min/ mg)

(A)

5

4

3

22 °C

1

10

100

3.15

3.20

3.30

3.35

3.40

1/T (10 )

(B)

(C) 2.0

1.5

1.0

0.5

1

10

100

1000

Drug concentration (μm)

ATPase activity (relative scale)

ATPase activity (relative scale)

3.25

–3

Verapamil (μm)

2.0

1.5

1.0

0.5 0.1

1

10

100

Drug concentration (μm)

FIGURE 6.44 P-glycoprotein ATPase activity. (A) P-glycoprotein ATPase activity as a function of temperature. Left panel, verapamil-activated, vanadate-sensitive ATPase activity of P-gp enriched microsomes at 22, 30, 37, and 40 C. The lines drawn are fitted to a modified form of the Michaelis
Menten equation taking into account both the ascent and the decline of the ATPase activity curve. Right panel, the corresponding Arrhenius plot: ln k 5 ln A 2 (E )/(RT). The basal activity, V0, the maximum stimulated activity, V1, and the increase in activity, V1 2 V0 as a function of the reciprocal of the absolute temperature is depicted. The lines are first order regressions to the data, the slopes of which give the activation energies: E V1 5 77.2 kJ/mol, and E V12V0 5 72.9 kJ/mol. (B) Effect of steroids and a hormonal analogue on the ATPase activity of P-glycoprotein: progesterone (x), spironolactone (K), fucidin (Δ), and tamoxifen (¢). (C) Effect of hydrophobic peptides on the ATPase activity of P-glycoprotein: cyclosporin A (x), PSC-833 (K), valinomycin (Δ), and gramicidin S (¢). The above experiments were all performed with membranes derived from a highly P-gp overexpressing CHO (Chinese Hamster Ovary) cell line, where P-gp constitutes up to 32% (w/w) of the total plasma membrane protein, and thus gives a very high specific signal. For details of this two-site model for drug-P-gp interaction, please see Litman T, Zeuthen T, Skovsgaard T, Stein WD. Structure-activity relationships of P-glycoprotein interacting drugs: kinetic characterization of their effects on ATPase activity. BBA 1997; 1361: 159
68.

316 | Channels, Carriers, and Pumps

Of note also is P-gp’s rather high basal ATPase activity: it seems to run idle even in the absence of any added exogenous substrate. This could suggest the presence of endogenous substrates, such as phospholipids or other molecules associated with the plasma membrane. Furthermore, point mutations in the residues lining the drug-binding pocket result in decreased basal ATPase activity, supporting the notion that interaction of lipids or endogenous substrates at the TMD drug-binding pocket contribute to the basal activity.10 Finally, P-gp is extremely sensitive to inactivation by membrane active detergents, such as Triton X-100, digitonin, and SDS, and its ATPase activity depends very much on the lipid composition of the microsomal membrane preparation. Altogether, these observations support the flippase mode of action of P-gp. It is straightforward to monitor the ATPase activity of P-gp by measuring the amount of liberated inorganic phosphate by a colorimetric assay. Therefore, the ATPase assay is commonly used as a screening tool for P-gp
drug interactions. However, due to the bell shape of the ATPase activity profile, it is important to assay a wide concentration range of the test compounds (e.g., from 1 nM to 100 μM) to obtain meaningful kinetic information.

6.8.6 Substrates and Inhibitors of P-gp—Clarification of Concepts Before we go further it seems necessary to clarify what exactly we mean by substrate and inhibitor. Strictly speaking, and from an enzymologist’s point of view, drugs that are transported by P-gp are not substrates, because they do not undergo a molecular transformation. In this view, the true substrate for the P-gp ATPase is ATP, because it is being hydrolyzed to ADP and free phosphate. This, however, is not what we mean, when we talk about P-gp transport. The substrates we refer to here are indeed transport substrates, those that are being pumped by P-gp. Then, when we focus on the ATPase activity of P-gp, we can define catalytic substrates as those that stimulate the ATPase activity of the transporter, just as we have seen in Figure 6.44. In the same manner, we can define transport inhibitors as compounds that inhibit the efflux of other drugs (by competitive or noncompetitive mechanisms), and at the level of the catalytic site we can define ATPase inhibitors as molecules that decrease the endogenous ATPase activity of P-gp. Because the catalytic activity of P-gp is directly coupled to its transport activity, it would be natural to conclude that an ATPase inhibitor is also a transport inhibitor, and vice versa: that a

10. See Chufan EE et al. PLoS One 1993; 8: e82463.

Primary Active Transport Systems Chapter 6

|

317

compound that stimulates the P-gp ATPase is also a transport substrate. Or is it? What about verapamil? It was the first drug to be recognized as an efficient inhibitor of P-gp transport. Further, it is one of the best activators of P-gp-associated-ATPase activity; typically, we observe two- to five-fold stimulation with an EC50 (concentration required for 50% stimulation) of 0.5
2 μM, depending on the membrane preparation. Therefore, how can verapamil at the same time be both a strong transport inhibitor and an ATPase activator? By acting as a competitive substrate! That is, if verapamil is itself pumped by P-gp, it is a transport (and ATPase) substrate competing with the efflux of other drugs, and thus behaving as a classical competitive inhibitor, which leads us to the next question: is verapamil a true transport substrate of P-gp? Here, the experimental data are conflicting, as some studies favor verapamil as a P-gp transport substrate, while others do not. We have shown that the fluorescent verapamil analogue BODIPY-verapamil is indeed pumped out by P-gp overexpressing cells, but one could argue that the attached (bulky) BODIPY moiety changes the transport properties of the native compound, and studies with radioactively labeled verapamil have given equivocal results. What may explain why verapamil behaves more like an inhibitor of P-gp than a transport substrate is its membrane permeability. Because verapamil is a very lipophilic compound, its diffusion rate across the plasma membrane is fast. Therefore, even if P-gp indeed pumps out verapamil, its fast passive diffusion back into the cell balances the pump rate, and the observed net flux is zero. In other words, verapamil is a “fast diffuser,” and therefore it will enter the cell as fast as it is pumped out by P-gp, rendering its ATPase activity futile, and apparently uncoupling it from transport.11 The relation between the leak rate constant p (passive permeability) and the pump rate constant k, is illustrated in Box 6.6, where different leakpump scenarios are modeled. Progesterone is another strong activator of the P-gp ATPase (Figure 6.44B), which is not classified as a P-gp transport substrate. And the reason is, as for verapamil, that progesterone is a fast diffusing compound, so the passive diffusion balances the active efflux, resulting in no net efflux. Extending the above argument, we can now categorize different compounds with respect to their P-gp interaction in the following scheme, which takes into account both transport and ATPase activity:

11. The concept of fast and slow diffusing agents has been elaborated on by Gera Eytan at Technion, Haifa. See, the comprehensive text: Eytan GD & Kuchel PW. Mechanism of action of P-glycoprotein in relation to passive membrane permeation. Int Rev Cytol. 1999; 190: 175
250.

318 | Channels, Carriers, and Pumps

Category

Diffusion

Transport

ATPase Activity

Examples

Agonist

Slow

1

1

Anthracyclines, vinca alkaloids

Partial agonist

Fast

(2)

1

Verapamil, progesterone

Antagonist

Fast or slow







Vanadate (noncompetitive)

Partial antagonist

Slow

(2)




Cyclosporin A, PSC-833a

Nonsubstrates

Fast or slow

0

0

Methotrexate

Silent substrates

Slow

1

(0)

Colchicineb

a These bulky compounds are very slow transport substrates of P-gp, slowing down its turnover rate (reflected in the inhibition of its ATPase activity), and obstructing its transport machinery. b A “silent” substrate is effluxed by P-gp, but it apparently neither activates nor inhibits its ATPase activity. This could be because its activation profile resembles that of an endogenous substrate. To characterize such compounds’ effect on the P-gp ATPase, it is necessary to either pre-stimulate it (e.g., with verapamil) or to pre-inhibit it (e.g., with PSC-833) before establishing a dose
response relationship. Good inhibitors of P-gp, however, do not fall in this group; they are either strong activators, or strong inhibitors of its ATPase activity.

When assessing drug interaction with P-gp, the choice of assay is not trivial. As mentioned, the calcein-AM assay is very popular due to its ease of use and scalability (Figure 6.43). But it may not pick up all drug
P-gp interactions, only those that are competitive (common binding site) or non-competitive (different/overlapping binding sites) with calcein-AM transport. Because P-gp has multiple binding sites for multiple substrates, and perhaps even more than one transport-active site for each substrate—as recently suggested by site-directed mutagenesis and docking studies from Suresh Ambudkar’s lab at NIH12—this necessitates the use of multiple probes to fully characterize and understand a drug’s interaction with P-gp. Already in 1998, Shapiro and Ling showed that the two fluorescent probes, rhodamine-123 and Hoechst 33342 have different binding sites within P-gp13: The “R” site with high affinity for rhodamine-123 (and anthracyclines), and the “H” site that preferentially recognizes Hoechst 33342 (as well as colchicine). Thus, a multiprobe approach, including calcein-AM, rhodamine-123, and Hoechst 33342, in

12. Chufan EE, Kapoor K, Sim H-M, Singh S, Talele TT, Durell SR, Ambudkar SV. PLoS One 2013; 8: e82463. 13. Shapiro AB & Ling V. Acta Physiol Scand Suppl 2013; 643: 227
34.

Primary Active Transport Systems Chapter 6

|

319

BOX 6.6 Pump
Leak Theory and the Meaning of r Pumping from inside only

Simple diffusion Steady-state drug accumulation, no pumping, no reverser

Steady-state drug accumulation, no reverser

p = leak rate constant (~ passive permeability)

p = leak rate constant

p

p

Si / So = p / p = 1

p

p

k

si = intracellular steady-state concentration s = Extracellular steady-state concentration o

Pre-emptive pumping

Pumping from inside only—with reverser /

Steady-state drug accumulation, no reverser

Steady-state drug accumulation, reverser I present

p = leak rate constant k = pump rate constant

k0 = uninhibited pump rate

Si / So = p / (p+k0Z) Z = fraction of pumping activity in presence of I

p

p p

p

Si / So = (p – k) / p

k

k

Z = Ki / (Ki+I) Ki = Intrinsic pump/reverser dissociation constant

Pumping from inside only—with reverser /I At the apparent Ki = I½ half of the pumping activity is blocked:

Pre-emptive pumping—with reverser At the apparent Ki = I½ half of the pumping activity is blocked:

Half-way between p/(p+k0) and 1

Fraction of blocked activity 1

Si / So = (p +k0/2) / (p +k0)

Si / So = (p - k0/2) / p

or p p + k0

Z½ = value of Z at apparent ki

or

Si / So = p /(p +k0Z½) = (p +k0/2)/(p +k0) I½

I

I½ = Ki X (p + K0) / p

Half-way between (p-k0)/p and 1

Fraction of blocked activity

... which gives: or

I½ = Ki X r

Si / So = (p - k0Z½) / p = (p - k0 / 2) / p

p – k0 p I½

I

Z½ = ½

I½ = Ki

r = ratio of blocked of unblocked uptake

r5

p1k p

If pck; then r-1:

Thus, the pumping rate k will not affect the distribution of a fast (high p) partitioning drug.

addition to the ATPase assay, is recommended for a complete characterization of a compound’s binding to and transport by P-gp. Table 6.2 lists some commonly used functional assays to screen for interactions with P-gp. Among these, the transepithelial transport assay, applying a confluent monolayer of polarized epithelial cells, appears physiologically relevant, as it mimics the biology of P-gp the best. Because P-gp is expressed

320 | Channels, Carriers, and Pumps

TABLE 6.2 P-gp Functional Assays—Pros and Cons Assay

Pros

Cons

Calcein
AM efflux

Sensitive, simple, rapid, robust strong signal, real-time signal suitable for HTS and FACS

Cell growth slow, only competition with calcein transport (other drug-binding sites not assayed) serum causes hydrolysis of calcein-AM

Epithelial transport (Caco-2 or MDCKII monolayers)

Close to “real-life” polarized transport (apical
basolateral)

Transport substrates may be missed if they are fast diffusers Other transport mechanisms (and cellular metabolism) may blur the signal

Inside-out vesicle transport

Sensitive (fluorescent or radioactive compound measured)

Vesicle preparation laborious, separation by filtration, some hydrophobic compounds may stick to the filters and vesicles

Confocal microscopy

Spatial information, colocalization information, real-time signal

Time consuming, not all compounds are fluorescent, fluorescent analogs may not be representative of native compound

ATPase activity

Easy, fast, inexpensive, enzyme kinetics , both activation and inhibition

Membrane preparation laborious, no direct information on drug transport

in the apical membrane, the basolateral-to-apical (b-to-a) flux for a P-gp transport substrate would be expected to be substantially higher than the corresponding apical-to-basolateral (a-to-b) movement. Thus, measurement of the polarized transport of a compound (provided that it is available in a radioactive or fluorescent form. Alternatively, HPLC or mass spectrometry may be used to detect it) across an epithelial monolayer will provide information on the compound’s physiologically relevant interaction with P-gp. But again: if the compound in question is a fast diffuser, the observed flux ratio (a
b/b
a) may be 1, and we will still not know if the drug actually interacts with P-gp at the molecular level. To answer this question, the ATPase assay appears more appropriate. Another caveat in using epithelial cell cultures is their expression of multiple transporter systems—both in the apical and in the basolateral membrane—that can blur the P-gp specific signal. Therefore, careful control experiments, applying specific inhibitors of the transport system(s) studied, are necessary to aid the interpretation of these data.

Primary Active Transport Systems Chapter 6

|

321

6.8.7 Catalytic Cycle of P-gp How is ATP hydrolysis coupled to transport by P-gp? From data on mutational studies (rendering one or both of the Walker sites inactive, or changing amino acids in the drug-binding pocket), on nonhydrolyzable ATP analogs (such as ATP-γ-S), on vanadate trapping experiments (Box 6.7), on biochemical data, and (lately on) improved structural information by X-ray diffraction, we now have a fairly good idea of the order of events in the catalytic cycle of P-gp. A scheme outlining these events is shown in Figure 6.45. First, drug and ATP bind to P-gp (Figure 6.45(1)). Drug enters the drugbinding site, which is in its high-affinity form, directly from the lipid bilayer. Then, ATP enters the NBD from the cytoplasm. The two events—drug and ATP binding—occur independently of each other. Note that in the figure, the ATP-binding site to the right is occupied throughout the catalytic cycle (and does not hydrolyze ATP), while the one to the left is catalytically active (binding of the red ATP in Figure 6.45(2)). This is in accord with Alan BOX 6.7 Vanadate as a Tool to Trap and Study ATPase Activity Vanadate—specifically the tetrahedral orthovanadate VO3
4 anion (abbreviated Vi)— is a transition state analog in phosphoryl transfer reactions as it easily forms a structure similar to that of a phosphate ester during hydrolysis. Therefore, P-type ATPases like the Na1/K1- and Ca21-pump, which employ a covalent phosphorylated intermediate, are most sensitive to vanadate inhibition with Ki values in the nanomolar range. The P-gp ATPase has a relatively low sensitivity toward vanadate inhibition (Ki 2
10 μM), suggesting that the reaction intermediate is not of the covalent phosphorylated type. Rather, the catalytic cycle of the P-gp ATPase seems similar to that of the myosin ATPase, where vanadate-induced inhibition is associated with noncovalent trapping of a single ADP molecule per catalytic site. That is, vanadate forms a Mg  ADP  Vi complex, mimicking the Mg  ADP  Pi transition intermediate. This is strong evidence for release of Pi before dissociation of ADP in the translocation cycle of P-gp (as illustrated in Figure 6.45). Alan Senior and colleagues at University of Rochester, NY, observed that for P-gp, the vanadate-trapped nucleotide was always ADP, independent of whether they preincubated P-gp with Mg-ATP or Mg-ADP. This showed that (for Mg-ATP) at least one turnover round must have occurred before inhibition. And importantly, vanadate-trapped ADP at one catalytic site was enough to block all ATPase activity, while release of the trapped ADP caused complete regeneration of ATPase activity, suggesting that the two nucleotide-binding sites interact. Based on these experiments, Senior et al. proposed the alternating catalytic sites cycle model for P-gp.14 According to this model, the two NBDs work like a twocylinder engine; they are both able to hydrolyze ATP, but take turns in doing so. (Continued )

14. Senior AE, Al-Shawi MK, Urbatsch IL. The catalytic cycle of P-glycoprotein. FEBS Letters 1995; 377:285
9.

322 | Channels, Carriers, and Pumps

BOX 6.7 (Continued)

FIGURE 6.45 Scheme for the alternating catalytic sites cycle of P-gp. The red hexagon represents a transport substrate, such as a hydrophobic drug. (1) P-gp is in its resting, open “apo” state, with disengaged NBDs, and no drug bound. One NBD is empty, the other has bound ATP. (2) Drug and ATP bind independently of each other to P-gp: drug enters directly from to plasma membrane; ATP binds to the vacant cytoplasmic NBD. (3) The two NBDs dimerize and form an occluded “ATP sandwich.” Addition of ATP-γ-S, which is a nonhydrolyzable ATP analog, can also generate the occluded pre-hydrolysis ATP conformation, which presumably provides the power stroke for the conformation change shown in (4). The TMD forms an outward facing, low affinity configuration of the drug-binding site allowing for release of the drug to the extracellular side of the membrane. Addition of vanadate traps P-gp in the ADP-bound post-hydrolysis conformation (5), indicating the sequential release of first Pi, and then ADP (6). The latter seems to be the rate-limiting step in the catalytic cycle, resetting P-gp from its low-affinity outward-facing to its high-affinity inward-facing “inverted V” conformation (1).15 The figure is made following personal communication with Dr. Suresh V. Ambudkar, and is a modified version of the scheme described in Ambudkar SV, Kim I-W, Sauna ZE. Eur J Pharm Sci 2006; 27: 392
400. The ATP-γ-S experiments are described in Sauna ZE, KIM IW, Nandigama K, Kopp S, Chiba P, Ambudkar SV. Biochemistry 2007; 46: 13787
99.

15. Convincing evidence for ADP release being the rate-limiting step in P-gp’s catalytic cycle is provided in Kerr KM, Sauna ZE, Ambudkar SV. JBC 2001; 276: 8657
64.

Primary Active Transport Systems Chapter 6

|

323

Senior’s original model of alternating catalysis (see Box 6.7) by P-gp, where the two sites take turns in ATP hydrolysis. Before hydrolysis can take place, the two NBDs have to dimerize. This is indicated in Figure 6.45(3), where the NBDs go from a disengaged open configuration to form a closed “ATP sandwich” dimer. This occluded prehydrolysis intermediate state can also be obtained with the nonhydrolyzable ATP analog, ATP-γ-S. Only one of the two ATPs is occluded at a time, and this is the ATP that subsequently undergoes hydrolysis (Figure 6.45(4)). Importantly, it is probably this pre-hydrolysis transition-like state that induces the TMD conformational change, which turns the drug-binding site from a high- to a low-affinity site, and allows for debinding and release of the drug. The post-hydrolysis P-gp  ADP transition state can be captured by vanadate to form a P-gp  ADP  Vi complex (Box 6.7), implying that phosphate is released (Figure 6.45(5)) before ADP (Figure 6.45(6)). The latter ADP release appears to be the rate-limiting step in the transport cycle of P-gp, and the one which resets the protein to its basal (empty, apo, highaffinity) inward-facing conformation (Figure 6.45(1)). The precise stoichiometry of P-gp—how many ATP molecules are split per drug molecule transported—is not known with certainty, although careful kinetic analysis of the transport of vinblastine compared to the vinblastinestimulated ATPase activity, suggests that this rate is 2 to 1 (Ambudkar SV, Cardarelli CO, Pashinsky I, Stein WD. J Biol Chem 1997; 272: 21160
6.). How the energy from ATP hydrolysis is actually transmitted to the TMDs to fuel drug transport, we still do not know. To clarify this, high resolution structural information is needed—which brings us to the next section on P-gp structure.

6.8.8 Structure Currently, the most consistent model structure for human P-gp in its “open,” “inverted V” apo conformation (in the absence of ATP or transport substrate) is a homology model based on the structure of mouse P-gp, as shown in Figure 6.46A. Further, the best available model for the corresponding “closed” conformation (the post-hydrolysis ADP-bound state) of P-gp is based on the structure of the bacterial ABC transporter Sav1866 (Figure 6.46B). There are no high-resolution structures of human P-gp available yet. This is because human P-gp—in spite of having 88% sequence identity to its mouse ortholog—is more flexible, and therefore, more difficult to crystallize (Suresh V. Ambudkar, personal communication). In fact, the X-ray crystal structure data ˚ on ABC transporters are at best medium-to-low resolution (in the 3
4 A ˚ range) while no true high-resolution (1 A) data exist. That interpretation of X-ray diffraction data can be challenging—and even erroneous—follows from a recent report by Stephen Aller’s group at the University of Alabama at Birmingham. This year (2014), they published a

324 | Channels, Carriers, and Pumps

FIGURE 6.46 Model of the structure of P-gp before and after ATP hydrolysis. (A) Homology model of human P-gp based on the improved murine P-gp structure, 4M1M (Li J, Jaimes KF, Aller SG. Protein Science 2014; 23: 34
46). P-gp is seen in its open, inverted V form, and the drug is on its way to the high-affinity binding site in the central cavity of P-gp. (B) Homology model of human P-gp based on the multidrug transporter Sav1866 from Staphylococcus aureus, 2HYD (Dawson RJP, Locher KP. Nature 2006; 443: 180
5.). After ATP hydrolysis has taken place, P-gp is seen in its outward-facing, regular V shape (closed, ADP-bound form with two ADP molecules shown in pink). The conformation change has turned the high-affinity drug-binding site into a low-affinity site, allowing the substrate to dissociate from P-gp and leave into the extracellular space. Figure courtesy of Dr. Suresh V. Ambudkar, NCI/NIH.

paper correcting the originally reported structure of mouse P-gp (Li J, Jaimes KF, Aller SG. Refined structures of mouse P-glycoprotein. Protein Science 2014; 23: 34
46.). While in the original 2009 Science paper (Aller SG et al. Science 2009; 323: 1718
22.) only 57% of the amino acids were modeled in the favorable Ramachandran region, this fraction has now increased to 95%. This underscores the importance of careful inspection of the structural information, and shows that models based on such data—such as molecular dynamics (MD) simulations—are only as reliable as the original structures on which they are based. Therefore, such models should always be validated by experimental data, such as mutational and biochemical experiments. What the X-ray structural data have clearly revealed is a large, central ˚ 3, which can explain the polyspecificity, which is a cavity of around 6000 A manifestation of both structural and chemical “flexibility” (a term coined by Suresh V. Ambudkar at NCI), of P-gp. This is its membrane-embedded, pliable drug-binding domain with multiple drug-binding sites that can

Primary Active Transport Systems Chapter 6

|

325

accommodate several substrates simultaneously, mainly via hydrophobic, aromatic, and hydrogen-bonding interactions. With respect to the large sepa˚ apart (depending on ration of the two NBD’s—they appear to be 30
50 A the model)—it is still a matter of controversy whether this represents a physiologically relevant conformation, or rather an experimental artifact of the crystallization conditions and the detergent used for the purification of the protein. We still need a crystal structure of ADP-bound P-gp and of human P-gp in any form. Future experiments, such as chimeric mouse
human fusion P-gp, combined with mutagenesis studies, fluorescence resonance energy transfer (FRET) (see Section 1.7) and docking approaches will hopefully fill our gap in understanding the complex structure
function relationship of this most intriguing transporter.

SUGGESTED READINGS This list contains both classic and recent papers. (We suggest that you consult also the papers cited in the legends to the figures, only some of which are listed below).

General Blostein R. Ion pumps. Curr Opin Cell Biol 1989;1:746
52. Harold FM. The vital force: a study of bioenergetics. New York: Freeman; 1986. Pedersen PL, Carafoli E. Ion motive ATPases. TIBS 1987;12:186
9. Stein WD. Transport and diffusion across cell membranes. Orlando, Florida: Academic Press; 1986 [chapter 6] Stein WD, editor. The ion pumps: structure, function, and regulation. New York: Alan R. Liss; 1988. Burroughs AM, Allen KN, Dunaway-Mariano D, Aravind L. Evolutionary genomics of the HAD superfamily: understanding the structural adaptations and catalytic diversity in a superfamily of phosphoesterases and allied enzymes. J Mol Biol 2006;361:1003
34. http://www.tcdb.org/ A comprehensive and up-to-date database containing functional and phylogenetic classification of membrane transport proteins.

Thermodynamics of Pumping Hill TH. Free energy transduction and biochemical cycle kinetics. New York: Springer-Verlag; 1989. Lauger P. Thermodynamic and kinetic properties of electrogenic ion pumps. Biochim Biophys Acta 1984;779:307
41.

Sodium Pump Apell H-J. Electrogenic properties of the Na,K pump. J Membr Biol 1989;110:103
14. Glynn IM, Karlish SJD. Occluded cations in active transport. Annu Rev Biochem 1990;59:171
205. Jorgensen PL, Andersen JP. Structural basis for E,-E2 conformational transitions in Na, K-pump and Ca-pump proteins. J Membr Biol 1988;103:95
120. Kawakami K, et al. Primary structure of the alpha-subunit of Torpedo californica (Na1 1 K1) ATPase deduced from cDNA sequence. Nature (London) 1985;316:733
6.

326 | Channels, Carriers, and Pumps Post RL, et al. Flexibility of an active center in sodium-plus-potassium adenosine triphosphatase. J Gen Physiol 1969;54:306
26. Shull GE, et al. Amino-acid sequence of the catalytic subunit of the (Na1 1 K1) ATPase deduced from a complementary DNA. Nature (London) 1985;316:691
5. Skou JC, editor. “The Na1, K1-Pump,” Parts A and B. New York: Alan R. Liss; 1988. Toyoshima C, Kanai R, Cornelius F. First crystal structures of Na1,K1-ATPase: new light on the oldest ion pump. Structure 2011;19:1732
38 [Review]. Toyoshima C, Cornelius F. New crystal structures of PII-type ATPases: excitement continues. Curr Opin Struct Biol 2013;23:507
14. Haviv H, Habeck M, Kanai R, Toyoshima C, Karlish SJ. Neutral phospholipids stimulate Na,KATPase activity: a specific lipid-protein interaction. J Biol Chem 2013;288:10073
81.

Calcium Pump of Sarcoplasmic Reticulum Hasselbach W, Oetliker H. Energetics and electrogenicity of the sarcoplasmic reticulum calcium pump. Annu Rev Physiol 1983;45:325
39. MacLennan DH, et al. Amino acid sequence of a Ca21, Mg21-dependent ATPase from rabbit muscle sarcoplasmic reticulum, deduced from its complementary DNA sequence. Nature (London) 1985;316:696
700. Martonosi AN, et al. The Ca21 pump of skeletal muscle sarcoplasmic reticulum. In: Stein WD, editor. The ion pumps: structure, function, and regulation. New York: Alan R. Liss; 1988. p. 7
14. Toyoshima C. Structural aspects of ion pumping by Ca21-ATPase of sarcoplasmic reticulum. Arch Biochem Biophys 2008;476, 3
11 [Review]. Toyoshima C, Iwasawa S, Ogawa H, Hirata A, Tsueda J, Inesi G. Crystal structures of the calcium pump and sarcolipin in the Mg21-bound E1 state. Nature 2013;495:260
64.

Calcium Pump of Plasma Membrane James PH, et al. Structure of the c-AMP-dependent phosphorylation site of the plasma membrane calcium pump. Biochemistry 1989;28:4253
8. Lew VL, et al. Physiological [Ca21]j level and pump-leak turnover in intact red cells measured using an incorporated Ca chelator. Nature (London) 1982;298:478
81. Schatzman HJ. The red cell calcium pump. Annu Rev Physiol 1983;45:303
12. Lopreiato R, Giacomello M, Carafoli E. The plasma membrane calcium pump: new ways to look at an old enzyme. J Biol Chem 2014;289:10261
68 [an insightful review].

Gastric H1K1-ATPase

Lorentzon P, et al. The gastric H , K1-ATPase. In: Stein WD, editor. The ion pumps: structure, function, and regulation. New York: Alan R. Liss; 1988. p. 247
54. Polvani C, Sachs G, Blostein R. Sodium ions as substitutes for protons in the gastric H,KATPase. J Biol Chem 1989;264:17854
9. Sachs G, et al. The ATP-dependent component of gastric acid secretion. In: Slayman CF, editor. Current topics in membranes transport “Electrogenic Ion Pumps”, vol. 16. New York: Academic Press; 1982. p. 135
59. Shin JM, Munson K, Sachs G. Gastric H1,K1-ATPase. Compr Physiol 2011;1:2141
53.

Multidrug Resistance Litman T, Skovsgaard T, Stein WD. Pumping of drugs by P-glycoprotein: a two-step process? JPET 2003;307:846
53.

Primary Active Transport Systems Chapter 6

|

327

Litman T, Druley TE, Stein WD, Bates SE. From MDR to MXR: new understanding of multidrug resistance systems, their properties and clinical significance. CMLS 2001;58: 931
59. Kapoor K, Sim HM, Ambudkar SV. Multidrug resistance in cancer: a tale of ABC drug transporters. In: Bonavida B, editor. Molecular mechanisms of tumor cell resistance to chemotherapy. New York: Springer; 2013. Li J, Jaimes KF, Aller SG. Refined structures of mouse P-glycoprotein. Protein Science 2014;23:34
46. http://nutrigene.4t.com/translink.htm A nice table compiled by Michael Mu¨ller, listing all human ABC transporters (including links).

F0F1 ATPases Boyer PD, Kohlbrenner WE. The present status of the binding-change mechanism and its relation to ATP formation by chloroplasts. In: Selman BR, Selman-Reimer S, editors. Energy coupling in photosynthesis. North-Holland, Amsterdam: Elsevier; 1981. p. 231
40. Graber P. Primary charge separation and energy transduction in photosynthesis. In: Milazzo G, Blank M, editors. Bioelectrochemistry II. New York: Plenum; 1987. p. 379
429. Penefsky HS. Energy-dependent dissociation of ATP from high affinity catalytic sites of beef heart mitochondrial adenosine triphosphatase. J Biol Chem 1985;260:13735
41. Senior AE. ATP synthesis by oxidative phosphorylation. Physiol Rev 1988;68:177
231. Bernal RA, Stock D. Three-dimensional structure of the intact Thermus thermophilus H1ATPase/synthase by electron microscopy. Structure 2004;12:1789
98. Soga N, Kinosita Jr K, Yoshida M, Suzuki T. Efficient ATP synthesis by thermophilic Bacillus FoF1-ATP synthase. FEBS J 2011;278:2647
54. Muench SP, Trinick J, Harrison MA. Structural divergence of the rotary ATPases. Q Rev Biophys 2011;44:311
56 [comprehensive review]. Stewart AG, Lee LK, Donohoe M, Chaston JJ, Stock D. Nat Commun 2012;21:687. Stewart AG, Sobti M, Harvey RP, Stock D. Rotary ATPases: models, machine elements and technical specifications. Bioarchitecture 2013;3:2
12.

Vacuolar and Anion Pumps

Moriyama Y, Nelson N. The vacuolar H1-ATPase, a proton pump controlled by a slip. In: Stein WD, editor. The ion pumps: structure, function, and regulation. New York: Alan R. Liss; 1988. p. 387
94. Rosen BP, et al. Molecular characterization of a unique anion pump: the ArsA protein is an arsenite (antimonate)-stimulated ATPase. In: Stein WD, editor. The ion pumps: structure, function, and regulation. New York: Alan R. Liss; 1988. p. 105
12.

Bacteriorhodopsin Butt HJ, et al. Aspartic acids 96 and 85 play a central role in the function of bacteriorhodopsin as a proton pump. EMBO J 1989;8:1657
63. Engelhard M, et al. Light-driven protonation changes of internal aspartic acids of bacteriorhodopsin: an investigation by static and time-resolved infrared difference spectroscopy using [4,3C] aspartic acid labeled purple membrane. Biochemistry 1985;24:400
7. Gerwert K, Hess B. Light induced intramolecular processes of bacteriorhodopsin monitored by time resolved FTIR (Fourier transform infrared) spectroscopy. In: Stein WD, editor. The ion pumps: structure, function, and regulation. New York: Alan R. Liss; 1988. p. 321
6.

328 | Channels, Carriers, and Pumps Stoeckenius W, Bogomolni RA. Bacteriorhodopsin and related pigments of Halobacteria. Annu Rev Biochem 1982;51:587
616. Lanyi JK. Proton transfers in the bacteriorhodopsin photocycle. Biochim Biophys Acta 2006;1757:1012
18. Lo´renz-Fonfrı´a VA, Heberle J. Channelrhodopsin unchained: Structure and mechanism of a light-gated cation channel. Biochim Biophys Acta 2014;1837:626
42 [a sister molecule to bacteriorhodopsin].