Impacts of the Osmolality and the Lumenal Ionic Strength on Osmosensory Transporter ProP in Proteoliposomes*

Background: Transporter ProP serves as a paradigm for the study of osmosensing. Results: ProP activity does not correlate better with the lumenal ionic strength than the osmolality in proteoliposomes. Anion effects follow the Hofmeister series. Conclusion: Effects of electrolytes and large solutes on ProP are non-coulombic in nature. Significance: An understanding of osmoregulatory mechanisms is fundamental to cell physiology and protein structure-function relations. H+ symporter ProP serves as a paradigm for the study of osmosensing. ProP attains the same activity at the same osmolality when the medium outside cells or proteoliposomes is supplemented with diverse, membrane-impermeant solutes. The osmosensory mechanism of ProP has been probed by varying the solvent within membrane vesicles and proteoliposomes. ProP activation was not ion specific, did not require K+, and could be elicited by large, uncharged solutes polyethylene glycols (PEGS). We hypothesized that ProP is an ionic strength sensor and lumenal macromolecules activate ProP by altering ion activities. The attainable range of lumenal ionic strength was expanded by lowering the phosphate concentration within proteoliposomes. ProP activity at high osmolality, but not the osmolality, yielding half-maximal activity (Π½/RT), decreased with the lumenal phosphate concentration. This was attributed to acidification of the proteoliposome lumen due to H+-proline symport. The ionic strength yielding half-maximal ProP activity was more anion-dependent than Π½/RT for proteoliposomes loaded with citrate, sulfate, phosphate, chloride, or iodide. The anion effects followed the Hofmeister series. Lumenal bovine serum albumin (BSA) lowered the lumenal ionic strength at which ProP became active. Osmolality measurements documented the non-idealities of solutions including potassium phosphate and other solutes. The impacts of PEGS and BSA on ion activities did not account for their impacts on ProP activity. The effects of the tested solutes on ProP appear to be non-coulombic in nature. They may arise from effects of preferential interactions and macromolecular crowding on the membrane or on ProP.

lality for each preparation, and no ion specificity was evident. However, the osmolality required to activate ProP varied with the lumenal composition, implying that ProP does not detect the osmolality (or water activity) per se. The osmolality (⌸/RT), which is the osmotic pressure (⌸) at a particular temperature (T) in molal units, is related to the water activity (a w ) by ⌸/RT ϭ Ϫln a w /V w (Eq. 1) where V w is the partial molar volume of water (0.01801 liter mol Ϫ1 ). In this report we extend that work by testing the hypothesis that ProP activity in PRLs is determined by the lumenal ionic strength.
In the previous study a higher osmolality was required to activate ProP in PRLs loaded with glucose or a membrane-impermeant PEG than those containing only buffer (10). ProP activity correlated with the lumenal potassium phosphate concentration when the lumenal solvent included glucose or a small PEG (degree of polymerization up to 13). However, PEGs with a degree of polymerization greater than 13 enhanced ProP activity over that attained with phosphate buffer alone. The degree of enhancement correlated with the molecular size of the PEG. Here we show that lumenal bovine serum albumin (BSA) can also elevate ProP activity and test the hypothesis that macromolecules alter the impact of potassium phosphate on ProP activity by altering ion activities.
The following modifications were employed for analyses of the impacts of electrolytes on ProP activity. In preparing PRLs, mixtures of detergent-solubilized protein and Triton X-100destabilized liposomes were encased in a dialysis membrane to protect them from contact with BioBeads during the detergent removal step of the reconstitution procedure (SpectraPor2, molecular weight cutoff 12,000 -14,000; Spectrum Laboratories, Inc., Rancho Dominguez, CA). The potassium phosphate concentration of PRL preparations was decreased by diluting them with water, harvesting them by centrifugation at 300,000 ϫ g for 22 min at room temperature, resuspending them in buffer of the desired composition, and extruding them through Nucleopore Track-Etch Membranes with 0.4 m (diameter) pore size. Unless otherwise indicated, the assay and wash buffers were 0.1 M sodium phosphate, 0.5 mM EDTA, pH 6.4. The osmolalities of assay and wash media were adjusted with sorbitol.
Analytical Procedures-Liposomes were prepared by detergent dialysis as described (4). The lumenal pH of liposomes and PRLs was recorded by monitoring the fluorescence emission of pyranine (16). Liposomes and PRLs were loaded with 0.4 M pyranine by extrusion (described above). External pyranine was removed from 1-ml aliquots of liposomes by gel exclusion chromatography as described (17) and from 75-l aliquots of PRLs by dialysis against the appropriate buffer lacking pyranine using Slide-a-Lyzer Mini Dialysis Units (Thermo Fisher Scientific) with 1 liter of a dialysis buffer that matched the PRL loading buffer. Fluorescence was monitored with a PTI Quantamaster C-61 steady-state fluorimeter (Photon Technology International, London, ON, Canada) with excitation wavelength 460 nm, emission wavelength 510 nm, and 4-nm band widths. The membrane potential (⌬⌿) was monitored using 3,3Ј-dipropylthiadicarbocyanine iodide as previously described (4,15).
For dynamic light scattering experiments, liposomes were loaded with D-glucose, PEG1000, or BSA by extrusion 21 times through Nucleopore Track-Etch Membranes with a 0.2-m (diameter) pore size. They were diluted in isosmolal, NaCl-supplemented buffer to a lipid concentration of 0.1 mg/ml. The intensity and intensity fluctuations of light scattered at 90°to the incident beam were recorded and analyzed as described (18).
Calculations-The following procedures were used to obtain data reported in Figs. 2, 4, and 5 and supplemental Fig. S2. The lumenal composition of PRLs was determined experimentally. The corresponding osmolality (⌸ 0 /RT) and the osmolalities of transport assay media (⌸/RT) were measured. The osmolality yielding half-maximal ProP-His 6 activity (⌸1 ⁄ 2 /RT) was determined by fitting the initial rate of proline uptake via ProP-His 6 (a 0 ) at the corresponding assay medium osmolality (⌸/RT) to Equation 2, where a 0 is the initial rate of substrate (radiolabeled proline) uptake, X (or ⌸/RT) is the osmolality, A max is the rate that would be attained at infinite osmolality, B is a constant, X1 ⁄ 2 (or ⌸1 ⁄ 2 /RT) is the value of X at which a 0 ϭ 1 ⁄ 2A max , R is the gas constant (8.314 joules⅐K Ϫ1 ⅐mol Ϫ1 ), and T is the temperature (298 K).  (15). Ionic strengths (I) were calculated with the assumption that the proportions of monobasic and dibasic phosphate did not vary significantly over the relevant ionic strength range (15). If PRLs behave as ideal osmometers, the concentrations of solutes in the lumen of PRLs subjected to osmotic upshifts can be calculated by multiplying their initial values by the ratio of the assay medium osmolality (⌸/RT) and the initial osmolality (⌸ 0 /RT). The ionic strength of the PRL lumen can then be determined with Equation 3, where c i is the concentration of the ith solute and z i is the charge on that solute. However, the responses of PRLs to osmotic shifts will be influenced by thermodynamic non-ideality of their lumenal contents. In that case a calibration curve must be used to determine the lumenal composition at which osmotic equilibrium will be reached (e.g. supplemental Fig. S1). This approach was used to correct the ionic strengths of PEG-containing PRLs.

RESULTS
The Analysis of Osmosensing-The composition of the solvent outside PRLs is easily altered, but special measures must be taken to vary the composition of the lumenal solvent (10). PRLs are routinely loaded with the solvent in which they are prepared. They can then be diluted with water, harvested by centrifugation, resuspended in a solvent of the desired composition, and extruded to equilibrate the external and internal solutions (see "Experimental Procedures"). Components of the initial solvent are not completely removed, but variations in the reconstitution conditions that may affect subsequent functional comparisons are avoided.
Data illustrating the osmotic activation of ProP are fit to empirical Equation 2 (see Experimental Procedures and Fig.  2A). The phospholipid membrane is highly water-permeable so all lumenal solutes are concentrated within seconds as the external osmolality is increased with membrane impermeant solutes (9). This means that the osmolalities inside and outside PRLs are the same on the time scale of transport assays, which take tens of seconds to complete. PRLs are usually assumed to behave as ideal osmometers (e.g. a 2-fold increase in external osmolality will result in a 2-fold decrease in lumenal volume). The lumenal concentrations of individual solutes and the lumenal ionic strength can be calculated on that basis, and the ionic strength at half-maximal ProP activity can be determined using Equation 2, with X ϭ I and X1 ⁄ 2 ϭ I 1/2 . ProP Activity Is Determined by the Lumenal Buffer Capacity of PRLs-We sought to test the hypothesis that ProP activity correlates better with the ionic strength of the PRL lumen than with the osmolality. Ion concentrations, the ionic strength, and the osmolality vary independently only for solutions that include polyvalent ions. In previous experiments the initial potassium phosphate concentration was fixed at 0.1 M as other solutes were added to the PRL lumen. The high background concentrations of monovalent (K ϩ , H 2 PO 4 Ϫ ) and divalent (HPO 4 2Ϫ ) ions limited the scope for variation of the lumenal ionic strength. For this study the lumenal potassium phosphate concentration was varied in the range 10 to 100 mM.
ProP activity was measured as the osmolality was varied by adding sorbitol as an external, membrane-impermeant osmolyte ( Fig. 2A). Surprisingly, the maximum ProP activity (indicated by A max ) decreased monotonically with the initial lumenal potassium phosphate concentration, whereas the osmolality at which activity was half-maximal (⌸1 ⁄ 2 /RT) did not (Fig. 2B). To determine whether the decrease in A max reflected the decrease in K ϩ and/or phosphate concentration, the pro-portions of lumenal K ϩ and Na ϩ were varied reciprocally at constant initial lumenal phosphate concentration (0.1 M) (Fig.  2C). There was a small decrease in A max , restricted to the lowest K concentrations, as the K ϩ concentration decreased at constant phosphate (Fig. 2D). The data in Fig. 2A were replotted to directly illustrate the relationship between the initial rate of proline uptake and the lumenal phosphate concentration (Fig.  2E). To do this the lumenal phosphate concentration for each measurement was calculated on the basis of the initial phosphate concentration and the degree of osmotic shrinkage (see "Experimental Procedures").
In our system, H ϩ -proline symport is powered by an imposed protonmotive force (⌬ H ϩ) comprised of a pH gradient (⌬pH, 60 mV) due to a pH downshift from 7.4 to 6.4 and a membrane potential (⌬⌿, Ϫ136 mV) imposed by diluting K ϩ -loaded PRLs 200-fold into K ϩ -deficient buffer in the presence of K ϩ ionophore valinomycin (7,15). In transport assays involving osmotic upshifts, K ϩ is also added externally to clamp the membrane potential at Ϫ136 mV despite the elevation of lumenal K ϩ concentration by osmotic shrinkage (15). Previous analyses based on bacterial membrane vesicles showed that passive proline efflux via ProP accelerated as the external pH increased, whereas the rate of proline exchange did not (7). This suggested that proton binding precedes proline binding to ProP, proton release follows proline release, and proton binding and/or release is rate-limiting (7). In PRLs, lumenal phosphate buffers proton influx due to ProP-mediated proline-proton symport. We postulated that acidification of the PRL lumen may limit the rate of proline uptake when the concentration of the phosphate buffer (and hence the buffer capacity) is low.
Pyranine fluorescence is a sensitive function of pH, with a pK a of 7.2. Thus lumenal pyranine can be used to monitor pH changes associated with proton translocation into or out of PRLs (16). Pyranine fluorescence was used to monitor the lumenal pH of PRLs under our transport assay conditions. First, pyranine-loaded liposomes were diluted into sodium phosphate buffers with pH 7.4 or 6.4, and pyranine fluorescence was monitored as K ϩ ionophore valinomycin and H ϩ ionophore FCCP were added sequentially (Fig. 3, A and B). In accord with previous measurements, decreasing pyranine fluorescence indicated acidification of the liposome lumen upon imposition of a pH gradient (a drop in external pH from 7.4 to 6.4). A further decrease in fluorescence, indicating further acidification, followed the addition of valinomycin. This proton influx would be favored, as the valinomycin-mediated increase in K ϩ permeability dissipated ⌬⌿, positive in (19), and imposed ⌬⌿, positive out. As expected, valinomycin triggered greater lumenal acidification for liposomes containing 10 mM potassium phosphate than for those containing 100 mM potassium phosphate. Subsequent FCCP addition collapsed the ion gradients, adjusting the lumenal pH to 6.4 and further decreasing the fluorescence. These data indicated that before the addition of valinomycin, the lumenal pH of PRLs diluted into pH 6.4 buffer would be no higher than 7.3.
Next, pyranine-loaded PRLs were diluted into pH 6.4 buffer at low and high osmolality to simulate our transport assay conditions. Pyranine fluorescence was monitored as valinomycin without or with ProP substrate proline, and then FCCP were FIGURE 2. Impacts of K ؉ , Na ؉ , and phosphate on ProP activity. PRLs were prepared in 0. 1 M potassium phosphate, 0.5 mM EDTA, 2 mM 2-mercaptoethanol, pH 7.4, and the lumenal buffer was modified as explained under "Experimental Procedures." ⌬ H ϩ (protonmotive force) was imposed by diluting PRLs prepared in potassium phosphate, pH 7.4, 200-fold into sodium phosphate, pH 6.4, in the presence of valinomycin, and proline uptake activities were determined as a function of the osmolality using assay and wash buffers of equivalent sodium phosphate concentration (e.g. 10 mM sodium phosphate, pH 6.4, for PRLs loaded with 10 mM potassium phosphate, pH 7.4). The assay medium osmolality was adjusted with sorbitol. The data were fit to Equation 2 to obtain A max and ⌸1 ⁄2 /RT as previously described (10). A, PRLs were loaded with potassium phosphate at a series of concentrations (10 -100 mM) so that K ϩ and phosphate concentrations varied in parallel. B, A max and ⌸1 ⁄2 /RT values derived by fitting data in panel A to Equation 2 are plotted versus the initial lumenal phosphate concentration of each preparation. C, PRLS were loaded with potassium phosphate/sodium phosphate mixtures so that the K ϩ and Na ϩ concentrations varied reciprocally, but the phosphate concentration remained constant at 100 mM. D, A max and ⌸1 ⁄2 /RT values derived by fitting data in panel C to Equation 2 are plotted versus the initial lumenal phosphate concentration contributed by potassium phosphate. E, data from panel A are replotted to show the relationship between the initial rate of proline uptake (a 0 ) and the lumenal phosphate concentration (calculated assuming that the PRLs behave as ideal osmometers). F, PRLs were loaded with 10 mM KPi alone (white circles) or 10 mM KPi plus KCl (160 mM, gray circles) or NaPi (90 mM, black circles).
added sequentially (Fig. 3, C-F). Again, decreasing pyranine fluorescence indicated lumenal acidification upon valinomycin addition. As expected, this acidification was proline-independent at low osmolality (conditions of very low ProP activity; Fig.  3, C and D), and it was exacerbated by proline at high osmolality (conditions of high ProP activity, Fig. 3, E and F). Thus H ϩ -proline symport via ProP acidified the PRL lumen. The lumenal pH of PRLs loaded with 10 mM potassium phosphate reached that of the external medium within seconds after the addition of valinomycin plus proline. It is, therefore, likely that, as the lumenal phosphate concentration decreased, ProP activity became limited by the lumenal pH.
The experiments reported in Fig. 3 did not explain why A max was lower when PRLs were loaded with high phosphate (0.1 M) and low K ϩ (10 mM potassium phosphate/90 mM sodium phosphate or 20 mM potassium phosphate/80 mM sodium phos-phate) than when they contained 100 mM potassium phosphate (Fig. 2, C and D). We reasoned that ⌬⌿ may be compromised when the K ϩ concentration was very low even though the phosphate concentration remained high. Fluorescent indicator 3,3Јdipropylthiadicarbocyanine iodide (20) was used to verify that ⌬⌿ was generated in each of the experiments reported above and that it decreased as expected when K ϩ was added to the external medium (data not shown). However, this technique did not indicate whether a ⌬⌿ of the same magnitude was generated with each preparation.
ProP activity was higher when the PRL lumen contained 160 mM KCl in addition to 10 mM potassium phosphate than when it contained 10 mM potassium phosphate alone (Fig. 2F). This KCl addition rendered the lumenal K ϩ concentration equal to that of 100 mM potassium phosphate, pH 7.4 (178 mM). A lumenal K ϩ concentration of 178 mM was, therefore, used to maximize ProP activity for subsequent experiments. These adjustments to the PRL system facilitated attainment of our central goal, which was to test the hypothesis that lumenal ionic strength determines ProP activity in PRLs.
The Impact of Lumenal Ionic Strength on ProP Activity-We previously showed that the relationship between ProP activity and the osmolality was the same for PRLs loaded with 100 mM potassium phosphate plus the Cl Ϫ salt of Na ϩ , K ϩ , Rb ϩ , or Cs ϩ (10). Experiments performed with membrane vesicles, in which the protonmotive force was generated by respiration, revealed osmotic activation of ProP in preparations loaded with either sodium phosphate or potassium phosphate (7). This lack of specificity suggested that ions may affect ProP by contributing to the ionic strength of the lumen in PRLs or membrane vesicles.
This hypothesis was tested by loading PRLs with the Na ϩ or K ϩ salts of citrate ((C 6 H 5 O 7 ) 2 3Ϫ ), SO 4 2Ϫ , supplementary phosphate, Cl Ϫ , or I Ϫ in a background of 100, 40 or 10 mM potassium phosphate. The anions were varied because polyvalent cations at the concentrations required for these experiments, profoundly affect membrane structure and permeability (21,22). The anion selection was constrained by the fact that the selected anions must not exist in a membrane-permeant form in the relevant pH range (6.4 -7.4). The Na ϩ or K ϩ salts were selected, and the concentrations of those salts were adjusted to fix the initial, lumenal K ϩ concentration at 178 mM. ProP activity was measured as a function of the osmolality, and A max , ⌸1 ⁄ 2 /RT, and I 1/2 were determined by fitting the resulting data to Equation 2. The A max values reflected the phosphate concentration dependence documented above, but A max did not vary systematically with other aspects of the composition of the PRL lumen (supplemental Fig. S2). The osmolality and ionic strength at which ProP activity was half-maximal were both anion-dependent (Fig. 4).
The left panels of Fig. 4 show the dependence of ⌸1 ⁄ 2 /RT and I 1/2 on ion composition for proteoliposomes loaded with 100, 40, or 10 mM potassium phosphate plus the K ϩ or Na ϩ salt of an additional anion. The right panels show how the ionic strength of the PRL lumen varied with the osmolality for these preparations. As expected, the ionic strength was more clearly differentiated from the osmolality as the phosphate content of the PRLs decreased. If ProP activity was determined by the ionic strength, half-maximal ProP activity would be attained at the same ionic strength (I 1/2 ) regardless of the ion composition of the PRL lumen. In fact, the ionic strength yielding half-maximal ProP activity (X1 ⁄ 2 ϭ I 1/2 ) was more dependent on the lumenal ion composition than was the osmolality yielding half-maximal ProP activity (X1 ⁄ 2 ϭ ⌸1 ⁄ 2 /RT) (Fig. 4). The anion dependence of these parameters became more marked as the lumenal phosphate concentration decreased (left panels), and the distinction between the ionic strength and the osmolality increased (right panels). ProP activity clearly did not correlate better with the ionic strength than with the osmolality of the PRL lumen, contradicting our hypothesis that lumenal ionic strength determines ProP activity in PRLs.
Neither PEG1000 nor BSA Activates ProP by Altering Ion Activities-Previous work showed that high molecular weight PEGs (degree of polymerization greater than 13) affected the osmotic activation of ProP (10) (e.g. Fig. 5A). BSA-loaded PRLs were prepared to assess the impact of a more physiological, high molecular weight solute on ProP activity.
First, dynamic light-scattering spectroscopy was used to show that liposomes (and hence PRLs) could be loaded with BSA by extrusion (Fig. 6). In this experiment the scattering intensity reflects the refractive index of the liposome lumen, PRLs were loaded with potassium phosphate at the indicated concentration plus the K ϩ or Na ϩ salt of a listed anion to adjust the initial lumenal K ϩ concentration to 178 mM, and proline uptake rates were measured as outlined in the legend to Fig. 2. The assay and wash buffers were identical to the PRL loading buffers except that Na ϩ replaced K ϩ , and sorbitol was used in place of NaI because I Ϫ interfered with scintillation counting. These variations in ion composition did not affect the membrane potential as indicated by 3,3Ј-dipropylthiadicarbocyanine iodide fluorescence (data not shown). The assay medium osmolality was adjusted with sorbitol. The data were fit to Equation 2, with X as the osmolality or the lumenal ionic strength, the latter calculated with Equation 3. Estimates of the osmolality (⌸1 ⁄2 /RT) and the ionic strength (I 1/2 ) at which ProP activity was half-maximal are shown. The primary data are provided in supplemental Fig.  S2. Right panels, the ionic strength of the PRL lumen attained at each (measured) osmolality during the titrations outlined above was calculated with the assumption that the PRLs behaved as ideal osmometers. FIGURE 5. The impacts of PEG1000 and BSA on ProP activity in PRLs. A and B, PRLs were prepared in 0.1 M potassium phosphate and loaded with nothing (circles) or with PEG1000 (triangles, 83 g/liter). Initial rates of proline uptake were measured as described under "Experimental Procedures." A, Relative proline uptake rates, determined as a function of the osmolality, were derived from Fig. 5 of Culham et al. (10). They are replotted here as a basis for further analyses. The A max values were 1.6 and 2.1 mol/min/mg ProP protein for unloaded and PEG1000-loaded PRLs, respectively. B, relative proline uptake rates were plotted versus the lumenal ionic strength, calculated with the assumption that the PRLs acted as ideal osmometers (open symbols) or corrected to account for the non-ideality of potassium phosphate-PEG1000 mixtures illustrated in panel E (solid triangles, see "Experimental Procedures"). C and D, PRLs were prepared in 0.1 M potassium phosphate and loaded with nothing (circles) or with BSA (triangles, 181 g/liter). Initial rates of proline uptake were measured as described under "Experimental Procedures." C, relative proline uptake rates are plotted as a function of the osmolality, which was adjusted with NaCl (open symbols) or glucose (closed symbols). The A max values were 0.9 and 1.1 mol/min/mg ProP protein for unloaded PRLs and BSA-loaded PRLs, respectively. D, relative proline uptake rates are plotted versus the lumenal ionic strength, calculated with the assumption that the PRLs acted as ideal osmometers. E and F, solutions of potassium phosphate (KPi), KP i plus PEG1000, and KP i plus BSA were prepared to simulate the conditions in our osmotically shrunken PRLs then diluted to simulate the range of concentrations that would occur during an osmolality titration (see "Experimental Procedures" and the legend to supplemental Fig. S1). Corresponding solutions containing only KP i , PEG1000, or BSA were also prepared. The osmolalities of the solutions were measured and plotted versus the corresponding KP i or equivalent PEG1000 (E) or BSA (F) concentration. The data in panel E were used as explained under "Experimental Procedures" to correct the relationship between relative proline uptake rate and lumenal ionic strength illustrated in panel B.
whereas fluctuations in scattered light intensity reflect liposome size, which determines the rate of liposome diffusion. The sizes and size ranges of unloaded and BSA-loaded liposomes were similar (Fig. 6B). However, the intensity of the light scattered by the BSA-loaded liposomes was much higher than the intensity of the light scattered by the unloaded, glucose-loaded, or PEG-loaded liposomes (Fig. 6A). This result was consistent with the higher refractive index of the BSA solution and equilibration of the BSA across the liposome membrane during extrusion.
PRLs were loaded with BSA to an initial lumenal concentration of 181 mg/ml, and ProP activity was measured as the external osmolality varied. This was the highest BSA concentration that could be attained in our experimental system. BSA loading did not alter the relationship between ProP activity and the osmolality (Fig. 5C). However, plotting the data as a function of the lumenal ionic strength revealed that like PEG1000, BSA increased the activity of ProP in solute-loaded PRLs relative to that attained by concentrating potassium phosphate alone (Fig.  5, B and D).
We postulated that high molecular weight PEGs and BSA may activate ProP indirectly by altering K ϩ and phosphate activities in the PRL lumen. To detect interactions among the solutes used for these studies, solutions containing potassium phosphate alone, solute alone, or potassium phosphate plus solute (e.g. inorganic salts, PEGs, or BSA) were prepared to simulate the concentrations that would have occurred after maximal osmotic shrinkage of the PRLs in our experiments. These solutions were then diluted stepwise to simulate the lumenal concentrations that would have occurred during an osmolality titration, and the osmolalities were measured. Interactions among K ϩ , phosphate, added solutes, and water were evident if the sums of the osmolalities of solutions containing potassium phosphate alone and solute alone were not equal to the osmolalities of solute-containing potassium phosphate solutions (supplemental Fig. S1; Fig. 5, E and F). This effect was pronounced for PEG-and BSA-containing solutions, and it increased with both PEG molecular weight and solute concentration.
The non-ideality illustrated in supplemental Fig. S1 and Fig.  5, E and F, affects the relationship between ProP activity and lumenal ionic strength for PRLs loaded with PEG or BSA and potassium phosphate (Fig. 5, A and B). It is usually assumed that PRLs shrink in direct proportion to a -fold increase in extracellular osmolality. For example, a 2-fold increase in osmolality (imposed by adding membrane-impermeant solutes to the external medium) would result in a 2-fold decrease in lumenal volume and a 2-fold increase in lumenal concentration of each chemical species. Because the lumenal osmolality of PRLs loaded with BSA or PEG and phosphate is a non-linear function of the concentration of either species, a 2-fold increase in external osmolality will result in less than a 2-fold increase in the concentrations of the internal solutes. The data for potassium phosphate-PEG1000 mixtures in Fig. 5E were used as a standard curve to estimate the concentrations of K ϩ and phosphate that would occur after osmotic equilibration of PEG-PRLs. Those estimates were used to recalculate the lumenal ionic strength. This correction was significant, but it did not fully account for the effect of PEG1000 on ProP activity (Fig. 5B, solid  triangles). Comparison of Fig. 5, E and F, shows that the nonideality of potassium phosphate-BSA mixtures would not be sufficient to account for the effect of BSA on ProP activity illustrated in Fig. 5D.

DISCUSSION
There is abundant evidence that prokaryotic and eukaryotic cells sense and respond to changes in the osmotic pressure of their environment (23). Diagnostically, these cells respond in the same way to environments with the same osmotic pressure that have diverse chemical compositions. Bacterial osmosensory proteins, including mechanosensitive channels and osmosensory transporters, were identified more than a decade ago (4,24). Mechanosensitive channels clearly respond to increases in turgor pressure-induced membrane strain that arise when cells imbibe water in response to osmotic down-shocks (25). In contrast, we do not fully understand how osmosensory transporters detect and respond to osmotic pressure changes.
Most protein receptors function via stereospecific ligandprotein interactions with physiologically relevant K d values in the femtomolar to low millimolar range. It is difficult to imagine how a protein would directly detect a change in the osmotic pressure or water activity. Water is the solvent in biological systems (not a solute), and biological systems detect very small changes in water activity. For example, osmosensory transporter ProP can make the transition from an inactive to an active state in vivo with an osmolality change from 0.1 to 0.3 mol/kg (a water activity change from 0.998 to 0.995) (10).
Changes to the osmotic pressure of the external medium may alter many cellular properties simultaneously. Those properties include turgor pressure, membrane strain, and intrinsic membrane curvature as well as the activities of individual solutes or groups of solutes or the crowding of macromolecules in the cytoplasm (3). In principle an osmosensory transporter could detect and respond to any of these changes. To determine which are relevant, researchers seek to correlate transporter activity with systematic changes to each candidate property in turn. Early studies performed with cells and PRLs failed to correlate the activities of representative transporters BetP, OpuA, and ProP with turgor pressure or membrane strain (2). Thus, attention shifted toward osmolality-sensitive cytoplasmic properties.
PRLs were used so that the activities of transporters BetP, OpuA, and ProP could be monitored as the composition of the lumenal solvent (representing the cytoplasm) was varied systematically (Refs. 4 -6 and others reviewed by Wood (1)). Such work is based on the premise that solute effects will be additive; effects of individual solutes applied in vitro will simulate effects of the complex solutions that confront these proteins in vivo. It led to the concept that BetP responds to the lumenal (and hence the cytoplasmic) K ϩ concentration with an apparent K d of 0.2 to 0.4 M (26). BetP is, therefore, expected to possess a K ϩ -specific regulatory site (27). Analogous work led to the view that OpuA responds to the lumenal and cytoplasmic ionic strength with an I 1/2 of 0.2 M (22). On this basis, osmoregulation of OpuA would involve modulation of electrostatic interactions (28).
Previous studies detected no specificity for lumenal ions, as ProP was osmotically activated in PRLs or cytoplasmic membrane vesicles (7,10). In this work ProP activity correlated with the phosphate concentration of the PRL lumen, but that effect could be attributed to the role of phosphate as lumenal buffer rather than to a particular role of phosphate in osmosensing (Fig. 2E, 3). ProP activity did not correlate better with the ionic strength of the PRL lumen than with the osmolality (Fig. 4), and the effects of PEGs and BSA on ProP activity could not be rationalized in terms of their effects on ion activities (Fig. 5). We, therefore, conclude that ProP does not respond to the ionic strength of the PRL lumen or cytoplasm. This suggests that, like those of PEGs and BSA, the effects of lumenal ions on ProP are non-coulombic in origin (29,30).
The effects of citrate, sulfate, phosphate, and chloride on the osmolality at which ProP activates (indicated by ⌸1 ⁄ 2 /RT) follow the Hofmeister series (31-33): citrate Ͼ sulfate Ͼ phosphate Ͼ chloride Ͼ iodide. In this anion group the osmolality at which ProP became active was lowest for citrate (a kosmotrope) and highest for chloride (which is more chaotropic). The effects of iodide at low phosphate contradict this trend. Iodide may exert additional effects with a different origin.
The Hofmeister series reflects interactions among macromolecules, other solutes, and water. Solutes may be preferentially accumulated in or excluded from water of hydration at macromolecular surfaces (29). Effects of solute partitioning between bulk and surface water on protein folding and proteinnucleic acid interactions have been extensively documented (29). In such systems excluded solutes (kosmotropes) favor processes that decrease the exposure of macromolecular surfaces to water. Osmotically induced increases to the activities of ionic and non-ionic kosmotropes in the cytoplasm of living cells may increase ProP activity by favoring a transporter conformation with minimal, cytoplasm-exposed surface area. In fact, the conformation of ProP is osmolality-dependent, as indicated by the chemical reactivities of introduced cysteine residues, and high osmolality favors the reactivities of periplasm-proximal residues in vivo (34,35). However, other factors must be considered before a structural mechanism for osmosensing can be confirmed.
It is not yet clear whether the osmolality determines ProP structure and activity directly via ProP-solvent interactions and/or indirectly via membrane-solvent interactions and alteration of the membrane environment of ProP. Membrane composition is certainly a determinant of osmosensing. Zwitterion phosphatidylethanolamine and anions phosphatidylglycerol and cardiolipin (diphosphatidylglycerol) are the predominant lipids in E. coli membranes. The proportion of cardiolipin varies from 2-3 to 6 -7 mol % in bacteria cultivated in low and high osmolality media, respectively (36,37). ProP concentrates with cardiolipin at the poles of E. coli cells, where the cardiolipin content may exceed 10 mol % (38). The osmolality at which ProP attains half-maximal activity (⌸1 ⁄ 2 /RT) varies directly with the anionic lipid content of the membrane both in vivo and in vitro (36 -38). Phospholipid headgroup composition is also a key determinant of osmosensing for transporters BetP (39) and OpuA (6) (reviewed by Wood (1)). Furthermore, Hofmeister effects on membrane lipid phase behavior are well documented (40,41). Thus solvent-membrane interactions could affect the structure and function of these osmosensory transporters by affecting key membrane properties (3,42). The membrane could serve as an osmosensory antenna, and observed outcomes could arise from distinct and even opposing effects of solvent changes on the membrane and the transporter.
Finally, how do large PEGs and BSA act on ProP? Each raised ProP activity to levels much higher than those attained by concentrating potassium phosphate alone in the PRL lumen (Ref. 10 and Fig. 5, B and D). Note that the potassium phosphate concentration scales with the ionic strength in these figures because the data are derived from PRLs that were not loaded with electrolytes other than potassium phosphate. Large molecules like PEGs and BSA can favor minimum surface area conformations via steric exclusion from macromolecular surfaces (43) and by occupying solution volume, the latter phenomenon known as macromolecular crowding (44). In this study BSA, a 66.5-kDa, globular protein, was incorporated in the PRL lumen at the maximum attainable concentration (0.18 g/ml, rising to a maximum of ϳ0.5 g/ml with maximal osmotic shrinkage). In previous studies the fraction of cytoplasmic volume occupied by macromolecules was estimated to vary from ϳ0.15 to 0.28 as E. coli was cultured in minimal media with osmolalities varying from 0.1 to 1 mol/kg (45), and it was estimated that the degree of macromolecular crowding in the cytoplasm of E. coli could be simulated by a globular protein with a molecular mass close to 75 kDa at a concentration of 0.34 g/ml (46). Thus the degree of crowding attained in our experimental system would mimic that occurring in vivo. Unless the shrinkage of PRLs not loaded with large solutes invokes the analogous phenomenon "macromolecular confinement," crowding cannot be essential for ProP activation (e.g. Figs. 2 and 4). However, crowding may exert physiologically relevant effects on ProP. For example, the activation of ProP at a much lower osmolality in cells than in PRLs (10) may reflect the occurrence of crowding in the former system.