Digestive vacuolar pH of intact intraerythrocytic P. falciparum either sensitive or resistant to chloroquine

Molecular and Biochemical Parasitology 110 (2000) 107 – 124 www.elsevier.com/locate/parasitology

Digestive vacuolar pH of intact intraerythrocytic P. falciparum either sensitive or resistant to chloroquine Sergey M. Dzekunov, Lyann May B. Ursos, Paul D. Roepe * Department of Chemistry and Program in Tumor Biology, Lombardi Cancer Center, Georgetown Uni6ersity, 37 th and O Streets, Washington, DC 20057, USA Received 17 January 2000; received in revised form 3 April 2000; accepted 4 May 2000

Abstract We present the first single cell-level analysis of digestive vacuolar pH for representative chloroquine resistant (strain Dd2) versus sensitive (strain HB3) malarial parasites. Human red blood cells harboring intact intraerythrocytic parasites were attached to glass substrate, continuously perfused with appropriate buffer, and pH was analyzed via single cell imaging and photometry techniques. We find that digestive vacuolar pH (pHvac) is near 5.6 for HB3 parasites. Surprisingly, we also find that pHvac of Dd2 is more acidic relative to HB3. Notably, in vitro pH titration of hematin confirms a very steep transition between soluble heme (capable of binding chloroquine) and insoluble heme (not capable of binding chloroquine, but still capable of polymerization to hemozoin) with a distinct midpoint at pH 5.6. We suggest the similarity between the hematin pH titration midpoint and the measured value of HB3 pHvac is not coincidental, and that decreased pHvac for Dd2 titrates limited initial drug target (i.e. soluble heme) to lower concentration. That is, changes in pHvac for drug resistant Dd2 relative to drug sensitive HB3 are consistent with lowering drug target levels, but not directly lowering vacuolar concentrations of drug via the predictions of weak base partitioning theory. Regardless, lowering either would of course decrease the efficiency of drug/target interaction and hence the net cellular accumulation of drug over time, as is typically observed for resistant parasites. These observations contrast sharply with the common expectation that decreased chloroquine accumulation in drug resistant malarial parasites is likely linked to ele6ated pHvac, but nonetheless illustrate important differences in vacuolar ion transport for drug resistant malarial parasites. In the accompanying paper (Ursos, L. et al., following paper this issue) we describe how pHvac is affected by exposure to chloroquine and verapamil for HB3 versus Dd2. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Acridine orange; Antimalarial drug resistance; Single-cell photometry; Vacuolar pH

Abbre6iations: AO, acridine orange; ATP, adenosine 5% triphosphate; BCECF-AM,2%,7%-bis (carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester; CQ, chloroquine (7-chloro-4-amino-quinoline); IC50, concentration that inhibits 50% growth; MDR, multidrug resistance; pHvac, digestive vacuolar pH of intact intraerythrocytic malarial parasites; RBC, red blood cell. * Corresponding author. Tel.: +1-202-6877300; fax: + 1-202-6876209. E-mail address: [email protected] (P.D. Roepe). 0166-6851/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 6 - 6 8 5 1 ( 0 0 ) 0 0 2 6 1 - 9

108

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

1. Introduction Drug resistance in malarial parasites (principally Plasmodium falciparum and Plasmodium 6i6ax) is arguably the greatest challenge currently facing infectious disease research. Worldwide, millions of people die of malaria each year and the emergence of newly drug resistant strains along with the spread of existing resistant strains are both increasing at an alarming rate. To achieve better balance in this dire situation requires several things, including additional detailed information on the molecular mechanisms of antimalarial drug resistance. Currently, the primary technique for distinguishing chloroquine (CQ1) resistant from sensitive parasites in the field is the response of patients over several weeks of treatment, which is fairly inefficient. Relatively few cells are infected in patients and it is difficult and expensive to grow large quantities of parasites in vitro. Obtaining relevant samples for study is far from trivial. Also, events critical to antimalarial pharmacology occur in the subcellular digestive vacuole, which is inaccessible to many experimental techniques. Thus, elucidating the molecular pharmacology of antimalarial drug resistance is a major challenge. The majority of easily available antimalarial drugs (i.e. quinoline derivatives such as CQ) work at the intraerythrocytic stage of parasite development, and it is believed their principle mode of action is inhibition of hemozoin polymerization within the digestive vacuole. That is, as the parasite grows rapidly within the red cell it meets the severe metabolic challenge (in part) by digesting hemoglobin. The toxic by-product, free heme, is polymerized into an optically dense crystal called hemozoin, which is left behind within the lysed detritus that was previously the infected erythrocyte as new merozoites are released. Although the chemistry of drug/heme interactions and hemozoin polymerization are not completely understood, it is believed CQ and other antimalarials form non covalent complexes with either heme monomers or m-oxo dimers that ‘cap’ the growing hemozoin chain [1] or alter hemozoin polymerization kinetics in other ways [2]. This presumably leads to build-up of toxic heme or heme-drug conjugates within the vacuole [1].

As is also the case with drug resistant tumor cells (see [3] and references within) drug resistance in malarial parasites has typically been associated with decreased cellular accumulation of CQ and other drugs [4,5]. This most likely reflects reduced net drug retention within the vacuole [1] more than any kinetic alteration in transmembraneous transport of drug, since CQ and other antimalarials passively diffuse very rapidly, yet cellular accumulation does not plateau for at least 30–40 min [4,5]. If altered cellular drug accumulation was understood molecularly, additional strategies for circumventing resistance would become apparent. Several hypotheses for decreased drug accumulation in resistant parasites have been suggested, including one analogous to the ‘drug efflux pump’ model for drug resistant tumor cells. At least two homologues of human MDR 1 protein (P-glycoprotein) are expressed in P. falciparum (Pf mdr 1, Pf mdr 2), and their mutation or altered expression may (in specific examples) be associated with modulating resistance phenotypes [6]. However, it has also been found that increased antimalarial drug resistance is not related to elevated Pf mdr expression [7,8], and it has been pointed out (e.g. [4]) that the passive diffusion of CQ and other antimalarials is likely too fast to make a ‘drug efflux pump’ scenario attractive for explaining lowered accumulation. Thus, similar to recent analysis of tumor drug resistance [9,10] ‘altered partitioning’ models for decreased drug accumulation in drug resistant parasites have proven increasingly popular [11]. It is a straightforward physico–chemical prediction that the diprotic weak base CQ will concentrate within the parasite’s acidic digestive vacuole to an extent proportional to the square of the net pH gradient (vacuole to blood). Quite subtle increases in pHvac (0.1–0.2 units) are predicted to significantly decrease the steady state vacuolar water phase [CQ]. Previous studies [12–15] suggested that pHvac was acidic and that the acid pH influenced net vacuolar accumulation of drug. Since CQ accumulation is lower in resistant parasites, it is therefore commonly accepted that resistant parasites might well have elevated pHvac. However, some CQ resistant parasites are actually more sensitive to mefloquine, which is a

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

monoprotic weak base that also presumably binds to heme [7]. It is difficult to envision how higher pHvac would confer increased sensitivity to a monoprotic weak base. Also, total accumulation of CQ within the living parasite is usually measured to be to a higher degree than would be predicted based on weak-base partitioning within water phase alone (perhaps up to 10times higher, see [4]). Thus, binding to drug target(s) strongly contributes to net total accumulation of drug within the vacuole (see also [1]), so decreased drug binding efficiency is also a reasonable explanation for decreased net drug accumulation. Identifying the factors that could contribute to lowering drug binding efficiency is essential. To test these models requires detailed knowledge of pHvac. At the time measurements of parasite pHvac were first attempted [12,13], epifluorescence microscopy approaches for analyzing living cells under constant perfusion were still in their infancy. Previous cuvette-based measurements with mass populations of intra- [13] or extra-erythrocytic [12] parasites are not analogous to measurements that are now in theory possible with modern imaging technology. Also, no measurements of parasite biophysical parameters have yet been made under the O2 tension at which the parasites normally propagate (5% O2). For these reasons, we have expanded upon previous work by modifying several techniques we have used in analysis of individual drug resistant tumor cells [9,16,17] and present the first analysis of pHvac for living intraerythrocytic P. falciparum under constant perfusion. In these measurements, parasites are perfused with physiologically relevant buffers balanced with 5% CO2 and 5% O2. We use CCD camera detection, the well defined pH-dependent partitioning of acridine orange (AO), and the strong fluorescence of this probe to quantitatively compare pHvac for Dd2 (CQ resistant) versus HB3 (CQ sensitive) parasites. Also, by analyzing the pH dependency of heme solubility, as well as the binding of CQ to soluble versus insoluble heme, we test whether acidification of pHvac is a logical pathway to resistance. The results are surprising, but logical nonetheless. In the accompanying paper we test whether chloroquine or verapamil have significant effects on pHvac.

109

2. Materials and methods

2.1. Reagents Nigericin, AO, BCECF-AM were from molecular probes (Eugene, OR) and CQ was from Sigma. Gas mixtures 5%CO2/5% O2 balance N2 were from Roberts Oxygen Co., Inc. (Rockville, MD). All other chemicals were reagent grade or better and were used without further purification.

2.2. Cell culture Starter HB3 (CQ sensitive) and Dd2 (CQ resistant) P. falciparum cultures were kindly provided by Dr. Pradipsinh K. Rathod (Catholic University, Washington DC) and grown at 2% hematocrit in sealed flasks containing 20% human serum in RPMI 1640 medium. The flasks were gassed with a mixture of 5% O2/5% CO2/90% N2. Dd2 is approximately 8-fold resistant to CQ relative to HB3 [18]. For single cell photometry experiments, parasitized red cells were attached to polylysine-pretreated glass coverslips (Corning no. 1; 18 mm2 per 0.1 mm thick) and incubated for 10 min with AO in a small volume of perfusion buffer with additional 10 units of streptolysin-O (Sigma, St Louis MO, 2000 units ml − 1). Control experiments (not shown) verified that staining was similar for sequential versus simultaneous incubation with streptolysin and AO, and that streptolysin treatment did not affect the viability of parasites or the mean intensity or localization of AO vacuolar staining (see below). In addition, pulse cytotoxicity experiments performed essentially as described previously [19] showed that low levels of AO exposure (100 nM–2 mM) for short times (10–15 min) as used in our measurements of pHvac (see below) were not toxic to the parasites. Although pharmacophores based on the acridine chemical backbone can be toxic to malarial parasites, IC50 of various compounds ranges from 0.1 to well over 20 mM [20], in continuous growth measurements, depending on the time of exposure and the strain of P. falciparum. Under short term exposure conditions like those used in this study, IC50 is raised significantly. Also, we found no evidence to indicate (see Section 4) that short

110

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

exposure to low levels of AO used in this work perturbed pHvac. The coverslips were mounted in a custom-built perfusion chamber [16] on the stage of a Nikon Diaphot epifluorescence microscope (see below) and perfused with 3 ml min − 1 buffer at 37°C. Buffers were either pH balanced with NaHCO3/ 5% CO2 and further gassed with 5% O2 (HBSS with O2), or, in a few selected experiments pH balanced with HEPES, pH 7.4, and not gassed with elevated O2 (HBS). Before beginning a fluorescence imaging experiment, the transmittance light images of individual cells were studied carefully to verify stage of infection (via the observation of dense hemozoin crystals characteristic of late-stage trophozoites, see below) and the viability of intraerythrocytic parasites (via slight ‘wiggling’ of viable parasites within the red cells).

2.3. Fluorescence microscopy and imaging We have extensively modified the hardware and software of a commercially available single cell photometry apparatus (the ‘InCyt Im2’ dual wavelength imaging system from Intracellular Imaging Inc., Cincinnati, OH) for these studies. Excitation was with a 300 W Xenon lamp, and stability of the intensity was 92%. AO fluorescence was monitored at 530 nm (490 nm excitation) by positioning Omega Optical 10 nm bandpass filters on a computer controlled rocker switch positioned between the lamp and a fiber optic directed to the microscope. A 530 nm dichroic mirror beneath the microscope stage reflected excitation and passed emission. 640× 480 pixel images were captured by an integrating Cohu CCD camera attached to the side port of a Nikon Diaphot inverted microscope equipped with a Nikon Fluor40 Ph3DL objective. Images were processed immediately between subsequent measurements by means of custom software, and light intensities in selected regions of interest were recorded as a function of time (e.g. Fig. 4 in Section 4). Periods of on-chip image integration were controlled by the image-processing software1 and automatically adjusted during the course of an experiment to compensate for any changes in

fluorescence intensity, thus allowing more effective use of the camera’s dynamic range. This was helpful since mature parasites are about 0.5 the volume of intact red cells and the digestive vacuole is B 0.04 of that volume (B 2 mm in diameter), and, as described above, viable intraerythrocytic parasites ‘wiggle’ which contributes to unavoidable fluctuations in pixel intensities over the course of an experiment. We also corrected for scatter of light from the infected erythrocytes by recording data at similar settings for cells not stained with AO, as well as uninfected red cells that were exposed to AO (see Section 4). In initial experiments (Dzekunov et al., unpublished data) we measured AO fluorescence intensities as above but using microscopy immersion oil (Stephens Scientific, Riverdale, NJ) to establish direct contact between the microscope objective and the glass coverslip to which the intraerythro1 Images were processed immediately, between subsequent measurements, by our custom software. Average pixel intensities (9 S.D.) in selected regions of interest (ROI) were analyzed for a given number of pixels, and intensities from vacuolar regions were recorded versus time. Selection of the pixels included for averaging was via extended statistical analysis. This analysis produced refined fluorescence patterns corresponding to the brightest regions within the vacuole. Specifically, ROI were subdivided into a number of small (3 ×3 to 5×5) pixel squares, and the closest surrounding region of a given square (the hollow square surrounding) constituted the conjugate ‘shell region’. Then, for each of the small squares, a z-test was performed to determine whether the mean intensity within was significantly (P] 0.05) different from the mean intensity of the shell: I1 −I2 /(s 21/N1 +s 22/N2)1/ 2 =za, where I, s and N are average intensity, standard deviation and number of pixels in the region, respectively, and where subscripts refer to the population of pixels in the central square (1) or the shell (2). If za was lower than a critical value (for a given value of P), the pixels in the center square were considered significant. Repetition of this procedure for all the pixels in a given ROI produces a contiguous pattern of significant pixels with the boundaries of the pattern corresponding to the sharpest intensity transitions across the ROI. This is essentially the pattern that would be obtained after more convenient (but less statistically rigorous) ‘thresholding’ methods of image processing. Our extended statistical analysis procedure allows much more stable readouts relative to image processing techniques that use fixed (and to some extent randomly chosen) threshold values to isolate the brightest pixels in an ROI.

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

cytic parasites are attached. In subsequent experiments (analyzed in this paper) designed to thoroughly quantify the difference in pHvac for HB3 versus Dd2 parasites, we omitted the oil. We note that non-specific scatter (background intensity) via the two procedures differs, and that use of this oil can lead to significant ‘dampening’ of the difference we calculate for HB3 versus Dd2 via the method presented within. Thus, we recommend oil be used with caution in single-cell photometry investigations of malarial parasites, at the recommendation of the objective manufacturer.

2.4. Localization of AO in parasites We used parallel inspection of bright field and fluorescence images of the same individual cell to verify vacuolar localization of probe before photometry experiments under constant perfusion (see Section 4, Fig. 2). Vacuolar localization of AO within parasites was also confirmed by acquiring simultaneous bright field (transmittance with Normarski optics) and fluorescence images using laser confocal microscopy (see Section 4, Fig. 3). In this case an Olympus Fluoview system was used to image successive 0.2 mm slices of interest in the Z-plane, using a 488 nm Argon line and living cells attached to polylysine-coated coverslips as described above. We first identified ‘wiggling’ parasites by eye in the absence of laser illumination and focussed manually by eye, followed by fine adjustment via inspection of several transmittance images acquired quickly (2 – 3 s). Once the desired plane of view was verified (e.g. panel 1 in Fig. 3, see Section 4), the z– axis series was obtained within 1 – 2 min under 488 nm illumination. Thus, the total time the parasite was subjected to laser illumination was the shortest possible, and at the lowest laser power level possible (20% output).

3. Theory Because it is a hydrophobic weak base, AO accumulates in acidic compartments (see [3] and references within for discussion of weak base

111

partitioning theory), and the fold accumulation is directly proportional to the magnitude of the pH gradient across the compartmental membrane. When more than one membrane separates the acidic compartment from the exterior phase (in this case, vacuolar, parasite plasma, and red cell plasma membranes separate the vacuole from blood) the net pH gradient across all membranes is relevant. The amount of AO in a parasite digestive vacuole thus equals the external concentration multiplied by 10(pHex – pHvac), with pHex being extraerythrocytic pH. When this DpH is constant, changes in the external concentration of AO will result in proportional changes in vacuolar AO concentration. In other words, the vacuolar concentration of AO depends linearly on its external concentration, with the slope of this dependence being equal to 10DpH (see Fig. 1 in Section 4). Because of this exponential relationship between DpH and vacuolar concentration of AO, relatively small changes in the magnitude of DpH result in approximately tripled changes of AO vacuolar concentration, as is illustrated in Section 4 (see Fig. 5). The linearity of this dependence is a good control for validity of the experimental epifluorescence data. This technique has obvious advantages as opposed to similar measurements done in a cuvette, because in the latter case the dependence of fluorescence on the probe concentration is typically nonlinear and more difficult to model. In our epifluorescence experiments (at constant working distance and consistent focal plane) we are essentially performing ‘thin layer’ experiments where photo-filtering effects, etc., are much less problematic. In addition, use of a CCD camera and proper data analysis software allows for the processing of a select number of pixels corresponding solely to fluorescence from vacuolar regions (see Section 4). The point is, most of the traditional difficulties for using AO fluorescence as a semi-quantitative probe of compartmental pH are reduced dramatically. In theory, one could obtain the precise value of the net pH gradient via measurements with a carefully calibrated apparatus by monitoring fluorescence intensity and thereby quantifying the

112

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

Fig. 1. Calculated AO partitioning between the digestive vacuole and the extraerythrocytic space at three different values of pHvac: 5.0 (circles); 5.5 (squares); and 6.0 (triangles). The vacuolar concentration of AO was calculated via [AO]vac =[AO]ext ×10DpH, where DpH is pHext − pHvac and pHext = 7.30. The INSET shows a thin layer calibration of AO fluorescence obtained with our single-cell photometry apparatus (see Section 2) and standard AO solutions prepared as described in the text. Fluorescence emission intensity versus concentration is linear out to several hundred mM, thus, it is safe to assume that a strict linear relationship exists for measured changes in vacuolar AO fluorescence intensity versus changes in vacuolar AO concentration.

concentration. In practice, physico-chemical properties of AO can lead to complications that could make exact calibration difficult. With these considerations in mind, we have paid considerable attention to the major factors that can affect fluorescence of AO entrapped in the digestive vacuole. In Section 4 we show that the present epifluorescence data agree quite well with theoretical predictions; thus, in our hands thin layer epifluorescence of AO is considerably less complicated in interpretation than bulk AO fluorescence in mixed cuvette experiments. Also, in calibrating differences in pHvac, we do not use fluorescence data acquired at only one external concentration of AO, rather, we plot vacuolar fluorescence intensities versus external AO concentration, obtain the slope of these straight line relationships and use any difference in slope to calculate pHvac differences (see Section 4).

4. Results

4.1. Vacuolar-specific AO staining as a measure of pH6ac Fig. 1 illustrates the theory behind our AO partitioning-based method for precise calibration of differences in pHvac for drug sensitive (HB3) versus resistant (Dd2) malarial parasites that avoids problems due to the non-ratiometric nature of the probe. The method relies on the ability to rapidly monitor AO fluorescence at the singlecell level. As described in Section 2, because initial fast partitioning of AO is highly DpH dependent, perfusion of intraerythrocytic malarial parasites with buffer containing a fixed concentration of AO leads to rapid concentration of the probe within the acidic vacuolar compartment and intense vacuolar fluorescence (see below, Fig. 2). In

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

addition, at constant perfusate pH, different pHvac for Dd2 versus HB3 would lead to distinctly different slope in plots of vacuolar AO fluores cence intensity versus external AO concentration ([AO]ex), as shown in Fig. 1. For example, pHvac of 5.0 (circles, Fig. 1) would generate a plot with slope 3.16 times greater than that generated for pHvac of 5.5 (squares, Fig. 1). Due to the logarithmic relationship between the slope and DpH (vacuole to perfusate), the method is an extremely sensitive method for computing pHvac differences. Moreover, non-specific effects (not due to fast DpH dependent partitioning) on AO vacuolar fluorescence at some AO concentration, if they occur, would appear as deviations from linearity in the plot. Note also that this method assumes that vacuolar AO fluorescence intensity is directly proportional to vacuolar AO concentration. This is indeed expected for these ‘thin layer’ experiments, in spite of non linearities in concentration versus fluorescence that are sometimes noted for bulk solution-based AO experiments. To test this directly, linearity was assessed using the same photometry apparatus and AO fluorescence under thin layer conditions.

Fig. 2.

113

4.2. Thin layer calibration of AO epifluorescence We mimicked the ‘thin layer’ aspects of the epifluorescence signal originating from the parasite vacuole in a calibration of AO fluorescence versus concentration (Fig. 1 INSET). In bulk solutions, [AO] versus fluorescence relationships can be complex (depending on concentration, pH, etc.) due to filtering effects, aggregation phenomena, etc. However, the fluorescence data presented herein were obtained in a very different ‘thin layer’ fashion. The calibration that is obtained via careful epifluorescence microscopy for thin AO solutions sandwiched between glass layers (essentially the thickness of perfused intraerythrocytic parasite layers, see below) is linear and goes through the origin (Fig. 1 INSET). In theory this

Fig. 2. Bright field (A, top) and fluorescence (B, bottom) images of the same field of view at 100 X magnification, obtained within 15 s of each other, showing both infected (arrow marks) and non-infected erythrocytes. The cells were under constant perfusion with HCO− 3 /CO2 balanced buffer (see Section 2) which also contained 1 mM AO. The dark hemozoin crystal that defines the interior of the digestive vacuole is easily visible for the infected red cell near the center of panel A, as are the outlines of red cell and parasite plasma membranes (arrows labelled 1 and 2, respectively, the contours of these membranes are also outlined in B with dashed white lines). Note the parasite also contains a parasitophorous membrane that should be essentially coincident with the plasma membrane in these images. The periphery of AO fluorescence (B, bottom) lies just outside the boundary defined by the vacuolar membrane (arrow labelled ‘3’ top panel, and innermost dashed line, bottom panel) because of scattering effects within the focal plane. However, note image processing software allows precise definition of spatial regions of interest (e.g. the center of the AO emission well within the vacuole). It is clear from dozens of side-by-side comparisons such as these, as well as additional confocal data (Fig. 3) that the most prominent AO staining is by far for the digestive vacuole of mature intraerythrocytic parasites within the infected cells in these preparations. In contrast, essentially no significant AO staining is seen for uninfected erythrocytes at the AO concentrations and exposure times used in this work, nor for the cytosolic or intra parasitophorus space in parasites. The common assumption that AO staining distinguishes between infected versus uninfected erythrocytes based on intercalation of AO into parasite DNA is (at least in part) incorrect; in fact, intense AO staining of parasites is due to the predicted (see Fig. 1) concentration of the weak base probe within the acidic digestive vacuolar compartment.

114

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

in vitro calibration provides a reliable and simple way to calculate the steady state pHvac from measurements of absolute vacuolar AO fluorescence intensity (see below). However, we note absolute steady state pHvac values computed using this calibration curve (see below) are quantitative estimates that could conceivably be influenced by additional features of the vacuolar compartment that we cannot mimic in thin layer AO solutions. In contrast, Fig. 1 demonstrates differences in pHvac found in different experiments or for different strains under the same conditions are extremely reliable due to the steep pHvac dependency for the slope from plots of [AO]ex versus vacuolar fluorescence intensity. Additionally, for this approach to be viable the measured fluorescence intensity must correspond spatially to the vacuolar compartment. Fig. 2A, B shows conjugate bright field and fluorescence images of an intraerythrocytic HB3 late stage trophozoite, stained for 10 min with 1 mM AO, that were acquired on our photometry apparatus within 10 s of each other. In Fig. 2A the arrows point to different membranes for the infected erythrocyte (1, red blood cell plasma membrane; 2, parasite plasma membrane; 3, parasite vacuolar membrane). The optically dense hemozoin crystal located within the digestive vacuole (indicative of the late-trophozoite stage of parasite development) is easily visualized under bright field illumination (dark black spot in Fig. 2A) and helps to orient the viewer in defining the vacuolar compartment (note faint outline of the vacuolar mem-

brane surrounding the hemozoin). Note these images are taken for living parasite samples (which ‘wiggle’, see Section 2) under constant perfusion using an epifluroescence microscope, not for fixed cells using a confocal apparatus. Thus, subcellular resolution in Fig. 2A is not expected to correspond to the best possibly available (see below, Fig. 3). Nonetheless, upon switching to fluorescence mode, intense AO fluorescence (Fig. 2B) is easily found to be coincident with the outline of the vacuolar compartment, indicating the pH probe localizes primarily to the digestive vacuole. This is entirely expected based on the predicted acidic pHvac and the weak base nature of AO (12). To test this more thoroughly, living intraerythrocytic parasites (late stage trophozoites) were also examined by laser confocal microscopy. Nomarski transmittance (left side, ‘A’, Fig. 3) and AO fluorescence (right side, ‘B’, Fig. 3) were obtained simultaneously as the z axis was stepped in 0.2 mM increments (e.g. panels labelled ‘2’ in Fig. 3 are images acquired approximately 0.5–1.0 s after panels labelled ‘1’ and after the objective was moved 0.2 mM). In panel A1 of Fig. 3, the view of the infected red cell is just at the periphery of the digestive vacuole, since the hemozoin crystal is barely visible. The parallel fluorescence image (B1, Fig. 3) is correspondingly faint. Within 2–3 steps in the z axis direction (panels A3, A4; 0.4–0.6 mM further into the parasite) the hemozoin crystal comes quite sharply into view, and the diameter of the hemozoin is at its maximum.

Fig. 3. Verification of digestive vacuolar localization of AO fluorescence via laser confocal microscopy. Again, we examine living cells in buffer, not fixed or imbedded cells, thus subcellular resolution is not as good as seen for confocal images of fixed cells, due to slight ‘wiggling’ of the viable parasites. Nomarski transmittance (left side) and AO fluorescence (right side) images were obtained simultaneously as the z-axis was stepped in 0.2 mm increments (panels labelled 1, 2, 3, etc.). As described in the text, the top panels illustrate the periphery of the digestive vacuole, and within 2– 3 steps in the z-axis direction (panels 3,4) the hemozoin crystal within the center of the vacuole comes quite sharply into view. Concomitantly, AO fluorescence intensity (right side) reaches its maximum, verifying co-localization of AO with hemozoin (i.e. within the interior of the vacuole as expected based on weak-base partitioning theory (see Section 3). Not coincidentally, the outline of the vacuolar membrane reaches its largest diameter at this point as well (panel 4, left side). In panel 4, the outline of the vacuolar (interior concentric circle) and plasma membranes of the parasite are reproduced as white lines on the parallel fluorescence image. Note these lines are hand drawn by eye and do not necessarily represent the precise diameter of these compartments. An additional step in the z-direction (panel 5) brings the hemozoin back out of plane and reduces the vacuolar perimeter diameter; thus the apparent center of the vacuole defined by this method is 0.4 – 0.6 mm from the periphery (i.e. panel 4 vs. panel 1), which corresponds to the expected vacuolar dimesions of a late-stage trophozoite. As a frame of reference, note also the uninfected red cell to the right of the infected red cell, both of which are laying flat on the glass surface; the characteristic flattened-toroidal shape of the red cell also comes in and out view as successive z-axis slices are obtained.

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

Thus, at this position we are well within the interior of the vacuole. Concomitantly, AO fluorescence intensity (B3, B4, Fig. 3) reaches its maximum, and exhibits a characteristic ‘half

115

moon’ shape. It is likely the dark shadow within the crest of the ‘moon’ is due to optical interference from the dense hemozoin crystal. In panel A4, the outline of the vacuolar and plasma mem-

Fig. 3.

116

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

Table 1 Vacuolar AO fluorescence intensity versus [AO]

a ex

HB3

Slope Calculated pH

Dd2

Mean

SEM

N

Mean

SEM

N

284.5 5.64

6.4 0.03

38 38

770.1 5.21

17.4 0.04

54 54

a

Average slope computed from many plots of AO vacuolar fluorescence intensity versus external [AO] generated in different experiments as shown in Fig. 4b. There is a consistent trend for higher slope (corresponding to lower predicted pHvac, cf. Fig. 1) in the resistant Dd2 parasites relative to drug sensitive HB3 that corresponds to 0.43 pH units. Based on our estimated pHvac of 5.64 for HB3 (see text), this predicts a pHvac of 5.21 for Dd2. We stress the relative difference between the strains is precise and accurate, whereas the steady state values are based on calculations using the thin layer calibration curve described in the text.

branes of the parasite are visible; the outer limit of these outlines are reproduced as white lines on the parallel fluorescence image (B4; note the shape of the vacuole is not necessarily a perfect sphere). Clearly, the AO fluorescence corresponds spatially to the digestive vacuole. An additional step in the z direction (panel 5) now brings the hemozoin crystal (A5) and the AO intensity (B5) back out of plane. Knowing the average diameter of the vacuole for a late stage trophozoite (1 – 1.5 mM), along with the ability to quickly acquire simultaneous Nomarski and fluorescence images in precise z – axis directional steps, allows us to firmly define vacuolar AO fluorescence for li6ing intraerythrocytic parasites. In the following photometry experiments, we only processed pixels that corresponded to AO fluorescence defined within the vacuolar compartment of each individual parasite, following the above method of co-localization with hemozoin, and using our image processing procedure. Since these cells were under constant perfusion (see Section 2), it was straightforward to rapidly change [AO]ex during the course of an imaging experiment. Fig. 4a shows raw data for single infected erythrocytes as [AO]ex in the perfusate is changed between 2.0, 1.0, 0.50, and 0 mM. As [AO]ex is lowered, intensity of vacuolar AO fluorescence decreases concomitantly because AO rapidly diffuses down the newly established chemical gradient. We did not notice any appreciable difference in the kinetics of vacuolar AO release for HB3 versus Dd2 (not shown). The INSET to Fig. 4a shows a plot of vacuolar fluorescence

intensity versus [AO]ex generated from these data that illustrates how linear the relationship truly is (at relatively low [AO]ex). Thus, the method outlined in Fig. 1 for calculating differences in pHvac between strains (via comparison of slopes of these lines obtained in different experiments) is applicable. Note also that at [AO]ex = 0, there is residual, very slight AO fluorescence in the vacuole, even several minutes after perfusion with AO free solution. We suspect this is due to some minor association of AO to proteins, vacuolar membrane, or perhaps free (unpolymerized) heme in the vacuole (see below, section Section 4.4). Regardless, upon careful inspection, and after monitoring vacuolar fluorescence upon small changes in external pH at constant [AO]ex (see below, Fig. 5A) or upon perfusion with solutions harboring the protonophore nigericin (see below, Fig. 5B) it is clear that the vast majority of vacuolar AO fluorescence intensity is a result of DpH (external-to-vacuole) dependent concentration of the probe via weakbase partitioning effects (see [3] and references within). When multiple experiments such as that shown in Fig. 4a are performed for many drug sensitive HB3 versus resistant Dd2 parasites and the data averaged, a highly reproducible difference in slopes of the resultant vacuolar AO fluorescence versus [AO]ex plots (Fig. 4b) is found for HB3 (circles Fig. 4b) versus Dd2 (squares Fig. 4b) parasites (Fig. 4b and Table 1). As outlined in Fig. 1, this clearly indicates a difference in pHvac of 0.43 pH units between the two strains. However, surprisingly, pHvac is found to be more acidic for Dd2, relative to HB3 (reflected by a

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

117

Fig. 4. (a) Vacuolar AO fluorescence intensity for infected erythrocytes at different external concentrations of AO. Initially, the cells were perfused at 37°C with buffer containing 2 mM AO for 8 min, at which point accumulation of the probe plateaus (not shown) and at which point the X axis is labelled ‘0’ s. At the first arrow, perfusate was switched to a similar solution with the same pH, but with [AO] = 1.0 mM; second arrow, to 0.5 mM AO; third arrow, to no AO in the perfusate. The solid line is the average of six individual traces obtained from six different infected cells, and the dashed lines above and below the solid trace delimit the 95% confidence interval for the average. Insert: linear fit of intensities vs. external [AO] in the perfusate. Intensities were obtained at the plateau of each relaxation phase by curve fitting with a single exponential (not shown) and averaged between individual cells. (b) Average of multiple plots of [AO]ex vs. vacuolar AO intensity obtained for many experiments as shown in Fig. 4a. Data points are shown9 S.E.M. Top line is for Dd2 (drug resistant), bottom line is for HB3 (drug sensitive) parasites. The Y axis is absolute counts from the CCD camera, thus the Y intercept is not predicted to be 0 (i.e. raw data not corrected for background). As described in Fig. 1 and Section 2, the different slopes indicate different pHvac (Table 1).

118

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

Fig. 5. Verification that parasite digestive vacuolar AO fluorescence intensity is directly related to the relative magnitude of the pH gradient vacuole-to-extraerythrocytic space. (a) Measured changes in vacuolar fluorescence intensity for an infected erythrocyte in response to small changes in perfusate pH at constant AO concentration (0.5 mM). Initially, perfusate pH was 7.13, then perfusate was switched to similar perfusate but at pH 7.18 (1st arrow, 200 s); 7.25 (2nd arrow, 450 s); and back to 7.13 (3rd arrow, 650 s). The horizontal thin lines illustrate the predicted fluorescence changes following the assumption (see Fig. 1 INSET) that vacuolar AO fluorescence intensity vs. vacuolar AO concentration is linear, and using the following equation: [AO]vac2/[AO]vac1 =([AO]ext2/ [AO]ext1) (10DpH2/10DpH1)= 10pH2 − pH1. (b) Changes in vacuolar fluorescence for an infected erythrocyte in response to perfusion with buffer containing nigericin but at constant AO concentration (0.5 mM) and pH (7.40). Initially, the cells were perfused with a HEPES buffered perfusate (37°C) that contained 5 mM K+/140 mM Na+. At the 1st arrow (275 s), perfusate was switched to the same buffer but with 7 mM of nigericin; at the 2nd arrow: 7 mM nigericin and 100 mM K+/45 mM Na+; 3rd arrow; back to 7 mM nigericin and 5 mM K+/140 mM Na+.

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

steeper slope in these [AO]ex versus vacuolar AO fluorescence intensity plots, compare Fig. 4b and Fig. 1). This is entirely surprising based on expectations from standard models for altered CQ vacuolar diffusion in resistant parasites (see, for example, [12]). Again, these slope data allow for accurate and precise determination of pHvac differences between HB3 and Dd2; with regard to a quantitative estimate of the value of steady state pHvac for control HB3 parasites under physiological perfusion conditions, if we analyze vacuolar AO fluorescence intensity at fixed [AO]ex and compare this to the thin layer calibration curve (Fig. 1, INSET), a value of 5.64 is computed for HB3 (Table 1). Thus, the 0.43 pH unit difference we measure when comparing HB3 and Dd2 (Fig. 4b) suggests a pHvac near 5.21 for Dd2 (Table 1). We note that ionic composition of perfusate in these experiments reflects that of human blood, however, parasites may be bathed (for at least part of their life cycle) in an ionic environment that is different, namely, that of the RBC cytosol. The chief difference is that blood is high Na+/low K+, whereas RBC cytosol is low Na+/high K+. Thus, we analyzed AO fluorescence for streptolysin permeabilized infected RBC while perfusate was switched from 145 mM Na+/5 mM K+ to 145 mM K+/5 mM Na+. A very small (B 5%) change in vacuolar fluorescence of similar magnitude was observed for both HB3 and Dd2 (not shown). Thus, elevated K+ only mildly affects pHvac, and not any differently for Dd2 versus HB3 parasites.

4.3. Tests for DpH dependence of 6acuolar AO signal We wished to firmly test that AO partitioning was indeed an accurate measure of pHvac. Fig. 5A presents the results of a simple experiment, wherein vacuolar AO fluorescence is monitored at constant [AO]ex while pH of the perfusate flowing over the cells is mildly and rapidly altered (from 7.13 to 7.18, 1st arrow; 7.25, 2nd arrow; then back to 7.13, 3rd arrow). Again, net DpH outsideto-vacuole determines the vacuolar concentration of AO (see Section 2), and assuming these mild changes in pHex do not promote significant

119

changes in pHvac (not expected because pHex changes are quite small) the experiment tests how closely experimentally determined vacuolar AO fluorescence intensity corresponds to prediction based on the magnitude of computed DpH. The actual recorded vacuolar AO fluorescence intensity data for the single parasite corresponds nearly exactly to predicted changes in AO concentration (recall in these ‘thin layer’ experiments fluorescence intensity is linearly related to concentration, see Fig. 1 INSET) deduced from simple application of the Henderson–Hasselbalch equation (see caption to Fig. 5a; theoretical predictions are denoted by thin solid horizontal lines). The close agreement between experiment and theory, the reversibility of the change in fluorescence intensity upon return to initial external pH (i.e. compare beginning and end of the recorded trace), and the fact that very mild and fast changes in external pH are used in this experiment (which should have negligible effects on metabolism that might then indirectly alter pHvac) supports our conclusion that the measured intensity of the vacuolar AO signal is primarily reflecting relative DpH outside – to – vacuole (and hence pHvac at known pHex). Fig. 5b tests this conclusion another way. Fluorescence is recorded for an individual vacuole while perfusate is changed to the same perfusate with 7 mM nigericin (first arrow), nigericin with 100 mM K+ instead of the normal 5 mM (second arrow), and then back to nigericin and 5 mM K+ (third arrow). Introduction of the K+-dependent protonophore quickly reduces the AO signal by nearly 50% in low K+ medium (first arrow), indicating loss of vacuolar AO due to partial collapse of the vacuolar membrane pH gradient. Because nigericin is basically a K+/H+ exchanger, simplistically, at this plateau [K+]vac/[K+]cyt = [H+]cyt/[H+]vac, with the equilibrium driven by endogenous K+ and H+ transport along with the K+-dependent protonophore. Upon switch to perfusate plus nigericin and 100 mM K+ (second arrow) an additional decrease in fluorescence (reflecting additional collapse of DpH) is observed, which is reversed by then switching back to nigericin/5 mM K+ (third arrow). Taken together, the results firmly support the notion that

120

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

vacuolar AO fluorescence intensity directly reflects the net pH gradient vacuole to extracellular. Vanadate also collapsed vacuolar membrane DpH (not shown), but high concentrations (10 mM) were required for efficient collapse, presumably because the vanadate needs to cross at least two membranes (RBC and parasite plasma membranes) before it reaches the vacuolar H+ ATPase acting to acidify the compartment.2

4.4. Possible complexities in interpretation Since AO is a hydrophobic compound we worried about possible complexities in interpretation due to any hypothetical association of AO to heme within the digestive vacuole. Indeed, some other acridine derivatives (e.g. quinacrine and pyronaridine, see [2]) are actually antimalarial compounds that could conceivably inhibit heme polymerization and perhaps even bind to heme similar to CQ. We titrated AO fluorescence versus hematin concentration (not shown) and note that in bulk solution some quenching of AO fluorescence by heme compounds does occur. By titrating possible AO:heme interactions at low pH where hematin aggregates (see below) we also confirmed that AO only binds to freely soluble hematin, not to insoluble heme, and that binding is responsible for the quenching we measure in solution. Quenching by freely soluble hematin likely reflects some electrostatic interaction between AO and the porphyrin ring structure. Some mild binding of AO to heme is also consistent with the observation that quenching of AO by freely soluble hematin is reversed by excess CQ (not shown) which likely has a much higher affinity for heme. However, importantly, the ratio of hematin: AO must be 10:1 or greater for a significant effect on AO fluorescence. Thus, it is highly unlikely that freely soluble vacuolar heme 2 We assume the P. falciparum vacuolar ATPase is ‘V-type’ and hence vanadate inhibitable, however, the high concentrations of vanadate necessary for efficient collapse of vacuolar DpH could also indicate a lower sensitivity to vanadate for this ATPase, the pharmacology of which has not yet been studied in detail.

is directly affecting the efficiency of vacuolar AO fluorescence to a significant extent. Section 4 in Fig. 5a and b in Section 4 confirm this expectation, as does the linearity of [AO]ex versus vacuolar intensity shown in the inset to Fig. 4a. Regardless, if we neglect these arguments, to affect our determination of pHvac differences between Dd2 and HB3, an unrealistic difference in freely soluble heme would need to exist Dd2 versus HB3 (we estimate\ 10-fold lower for Dd2).3

4.5. Effects of pH on [soluble heme] Observation of more acid pHvac for resistant parasites was initially quite surprising to us, until we recalled that the essential aspects of CQ pharamacology involve disrupting heme/hemozoin polymer dynamics and that net parasite accumulation of CQ and other antimalarials is greater than can be explained by weak base partitioning alone. As suggested by others (e.g. [1,22,23]) this then makes the investigator interested in drug resistance mechanisms consider drug/target interactions that contribute to net drug accumulation over relatively long time, not just initial diffusion of drug or its initial binding to target. We wondered how acid pH might affect the dynamics of heme polymerization and hence possibly titrate relevant target instead of drug. Fig. 6 presents a simple titration of soluble hematin versus pH. We note a particularly sharp drop in solubility with a calculated midpoint at pH 5.55. This, and our estimate of vacuolar pH (about 5.6) suggests that pH of the digestive vacuole is held near a value that may be designed to precisely control the availability of soluble free heme. Importantly, polymerized hemozoin forms from free heme even at low pH (e.g.B 5.0; [1,2]) where the vast major3

The fold decrease in freely soluble heme that would need to exist for Dd2 relative to HB3 in order to cause the percent increase in AO fluorescence we measure would depend on the normal concentration of free heme found in sensitive parasites, for which there are only estimates. However, it is unlikely that the ratio of soluble heme: AO in HB3 is higher than about 0.5. In this case, the fold decrease would need to be about five. If the freely soluble heme is closer to most estimates (near 20 mM or so), then a 10-fold decrease would be necessary.

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

121

ity of free heme would be expected to be in some type of aggregate form (Fig. 6). However, as we also confirm (Fig. 7) CQ does not bind efficiently to heme aggregates, only to soluble heme. Thus, although the precise value of pHvac is likely important for optimizing activity of proteases that

Fig. 6. Solubility of hematin vs. pH. Two hundred micromolar hematin solutions were incubated for 1 h at different pH (100 mM MES – Tris buffer) in 1.6 ml eppendorf tubes. Hematin from a single stock solution was added to the tubes at different pH and incubated at room temperature. Control experiments indicated the addition of stock hematin did not alter the initial pH of the tubes (not shown). After incubation, tubes were centrifuged at 12 000 rpm for 20 min in an eppendorf microfuge to pellet aggregated material and absorbance for the supernatants (diamonds) and re-dissolved pellets (squares) were collected. The solid line plots the absorbance at 343 nm for the supernatants (343 nm is near the 380 nm maximum for hematin) vs. pH of the initial solutions. Note pelleted material is solubilized quickly upon resuspension at pH ] 6.0, which indicates it is an aggregate form of heme, not polymerized hemozoin. The dashed line plots the 343 nm absorbance for the pellets obtained for each tube, after solubilization for 2–5 min in MES – Tris buffer at pH 7.0. The plots show the average of data from two experiments; when these data are fit to equations of the form y = {a/[1 + exp(b× (x −c))]}+ d, where: a, range of hematin absorbance values; b, slope coefficient; c, pH at the maximum rate of change in absorbance vs. pH; and d, minimum hematin absorbance (near 0.003), we obtain a value for c near 5.55 for each curve. Thus, the drop in hematin solubility is exceedingly steep over a quite narrow range of pH, and the midpoint of this drop is conspicuously close to the value of vacuolar pH we compute for HB3 malarial parasites. Similar behavior is observed over a wide range of hematin concentrations (data not shown).

Fig. 7. Results of experiments that test whether CQ binds to soluble or insoluble hematin, or to both. The CQ absorbance maximum is at 343 nm, hematin is at 380 nm, but there is considerable overlap between the two absorbance spectra at room temperature. For the first bar (labelled ‘CQ’) 20 mM CQ was added to a tube of 24 mM MES – Tris buffer at pH 4.5, equilibrated for 1 h, and then centrifuged as in Fig. 6 and the absorbance of the supernatant measured. Second bar (labelled ‘HT’), the same experiment was performed for 200 mM hematin originally dissolved at pH 7.0 but titrated to pH 4.5; note low absorbance of the supernatant is due to pelleting insoluble hematin at pH 4.5 (see Fig. 6). For the third bar, CQ was added to the tube first, then HT, and possible complexes allowed to form for 1h at pH 4.5 before centrifugation. For the fourth, HT was added first, the tube equilibrated for 1 h, then CQ was added, the tube vortexed vigorously, and then centrifuged. Comparing the 3rd and 4th bars shows that when HT aggregates are allowed to form completely prior to introduction of CQ (4th bar) much more CQ remains in solution. Finally, the 5th bar shows the results obtained when HT and CQ are given 1 h to interact in soluble form; in this case, CQ and HT were mixed at pH 7.0 where HT is soluble (Fig. 6), incubated for 1 h at RT, and then the mixture was titrated quickly to pH 4.5, centrifuged, etc. Less CQ absorbance is found relative to the 3rd bar. Data shown are the average of four experiments, 9S.E.

degrade hemoglobin, perhaps the parasite also uses this parameter to influence the physical state of heme destined for polymerization to hemozoin.

5. Discussion We stress that these data are obtained with only one set of CQ resistant versus sensitive laboratory

122

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

strains of malarial parasites. Nonetheless, when considered along with the data in the accompanying manuscript (following paper, this issue), we feel these data highlight an important way in which altered ion transport likely contributes to antimalarial drug resistance [11]. In addition, techniques perfected in this work should be useful in further investigations of malarial parasite physiology. Recently, several laboratories have applied advanced cellular imaging technology to the study of intraerythrocytic malarial parasites. For example, a recent elegant study [21] pre-loaded erythrocytes with BCECF-dextran, infected these with parasites, and then used confocal microscopy to image different macromolecular transport pathways for the intraerythrocytic parasite. However, to our knowledge no previous photometry studies have been with living intraerythrocytic parasites under constant perfusion with buffers balanced with appropriate CO2 and O2. Application of single-cell photometry or confocal methods to analysis of living intraerythrocytic malarial parasites under constant perfusion, along with the use of trace levels of vital dyes, is an extremely promising approach for analyzing the physiology of malarial parasites. To our knowledge, this is the first study that applies these methods to the analysis of parasite vacuolar biophysical parameters, which have been predicted for nearly two decades to have profound effects on antimalarial drug pharmacology, yet have only been examined in a handful of studies. Determining these parameters is essential for modeling levels of soluble heme and antimalarial drugs within the vacuole, as well as for predicting the bioavailability of new drugs and resistance-reversal agents. Our results illustrate that bulk optical measurements made with mass populations of parasites in cuvettes without constant perfusion and/or in the absence of physiologically relevant CO2 and O2 tension are not analogous to single-cell epifluorescence studies of living parasites under perfusion. The importance of vacuolar (and other compartmental) biophysical parameters for understanding antimalarial pharmacology and drug resistance, as well as for perhaps defining new ‘signatures’ for various resistance phenotypes is enormous. Whether pHvac

is alkaline, acid, or unchanged for resistant malarial parasites relative to sensitive is one particularly fundamental issue. Slightly more acid pHvac for Dd2 relative to HB3 contrasts with most investigators’ initial expectations regarding pHvac differences between drug resistant and sensitive parasites (including ours), since a decrease in pHvac would be predicted to increase vacuolar concentrations of weak base antimalarial drugs (see Section 3). Since the vacuole is the site of action of these drugs, this would provide a higher effective dose near the drug target and should (simplistically) make the parasite more drug-sensitive, not resistant. However, Bray, Ward and colleagues [23] recently suggested that the accessibility of free heme to CQ, and not necessarily the total amount of heme or of drug, was the relevant parameter for determining reduced net CQ accumulation in resistant strains. We believe our data support this idea and provide an attractive explanation for CQ accumulation data that originates with the classic studies of Fitch and others [24]. We suggest that a precise relationship between ‘soluble’ (monomeric and/or m-oxo dimer) heme, ‘insoluble’ heme, and ‘polymerized’ (hemozoin) heme exists in the vacuole (Fig. 8), and that quinoline drugs such as CQ bind to soluble heme, but not insoluble heme or hemozoin (see also Fig. 7). We further speculate that soluble monomeric heme proceeds through an ‘insoluble’ state similar to that promoted in vitro by acidic pH (Fig. 6) before polymerization to hemozoin. We propose that pHvac tightly controls the soluble/insoluble heme ratio and that decreasing this ratio is a principle mechanism of CQ resistance. Thus, total heme concentration does not change in resistant parasites, but the amount of CQ sequestered by free heme within the vacuole is lowered. An additional important observation that also supports this general idea is that Dorn and co-workers [2] have recently shown that polymerization to hemozoin actually becomes slightly more efficient as pH is lowered from 6.0 to 4.5. Thus, by at least two mechanisms (titrating to a non polymerized form of heme with lower affinity for CQ, and increasing the rate at which heme is polymerized) more acid pHvac is predicted to lower the efficiency of drug/target interaction.

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

That is, we propose that the organism uses the value of pHvac to manipulate interaction with drugs by altering the physical arrangement of target, without necessarily altering the total net amount of that target or appreciably compromising relevant biochemistry essential to its survival (i.e. hemozoin polymerization). We are reminded of an oddly similar (but converse) scenario in drug resistant tumor cells; elevation of cytosolic pH in MDR tumor cells likely titrates soluble tubulin (target of vinca chemotherapeutic drugs and colchicine) to tubulin polymers (to which vincas and colchicine do not bind) thereby decreasing net accumulation of the drugs over time without altering the total net amount of target

123

[25] or compromising relevant biochemistry (i.e. formation of tubulin fibers). Finally, we note that another recent study [26] used photometry methods to analyze cytosolic pH for resistant versus sensitive parasites. Interestingly, these investigators found slightly alkaline cytosolic pH for the resistant parasites, but the mechanistic explanation for this is currently disputed [22]. Nonetheless, observations of altered pH in both the cytosol and vacuole of resistant malarial parasites lend support to the idea that emergence of CQ resistance is linked to alterations in ion transport that then perturb pH, membrane potential, and volume regulation in such a way as to decrease the net efficiency of drug/target interaction [11]. These alterations in ion transport could occur through mutation or altered expression of ion pumps, channels, exchangers, or co-transporters. Thus, it will be important in future work to identify the ion transport reactions at both plasma and vacuolar membranes that are dysregulated in resistant parasites.

Acknowledgements Fig. 8. Proposed model for the interactions between CQ and heme within the digestive vacuole that is consistent with the observations reported herein. In this model, we propose that CQ only binds to soluble heme, but not insoluble (aggregated) heme nor hemozoin, and that conversion of soluble (presumably monomeric) heme to an insoluble form (aggregated, but not yet polymerized to hemozoin) does not eliminate the possibility of eventual conversion to hemozoin polymer. Thus, CQ inhibits steps 1 and 4 in this branched scheme, but not steps 3 or 2. That is, to explain more acidic pH for the digestive vacuole of drug resistant malarial parasites, we propose that a balance exists between two forms of unpolymerized heme within the digestive vacuole (what we call soluble and insoluble), that conversion to polymerized heme (hemozoin) can proceed through either form, and that acid pH promotes formation of the insoluble form (Fig. 6). Thus, since CQ does not bind well to the insoluble form (Fig. 8) but in vitro formation of hemozoin crystals occurs readily at acidic pH [1,2], titration of vacuolar pH to lower values should inhibit association of CQ to soluble heme and thereby confer drug resistance. The relative degree of resistance confered by a given reduction of pHvac would depend upon the relative rates of steps 1 – 4, and the concentrations of soluble heme and CQ in the vacuole.

We thank Dr J. Martiney (Robert Warren Laboratories) for many illuminating and helpful conversations, Dr P. Rathod and colleagues (Catholic University) for providing HB3 and Dd2 parasites and advice on their culture, Dr S. Mueller (Lombardi Cancer Center) for advice and help with our laser confocal studies, and Ms S. Hamilton for help with the graphics in Figs. 2 and 3. We also thank Drs D. Taylor, E. Davidson and D.C. Gowda (Georgetown), Kevin Baird (US Navy Malaria Program), and T. Wellems and D. Fiddock (NIH) for additional advice and encouragement. We thank the Burroughs Wellcome Fund and the Chemistry Department of Georgetown University for providing funds essential to the completion of this work, and our laboratory colleagues Ellen Howard and Catherine Santai for additional help and discussion. L.M.B.U. was a Georgetown Eppenscheind fellow during the course of this work.

124

S.M. Dzekuno6 et al. / Molecular and Biochemical Parasitology 110 (2000) 107–124

References [1] Sullivan DJ, Gluzman IY, Russell DG, Goldberg DE. On the molecular mechanism of chloroquine’s antimalarial action. Proc Natl Acad Sci 1996;93:11865–70. [2] Dorn A, Vippagunta SR, Matile H, Jaquet C, Vennerstrom JL, Ridley RG. An assessment of drug-haematin binding as a mechanism for inhibition of haematin polymerisation by quinoline antimalarials. Biochem Pharm 1998;55:727 – 36. [3] Wadkins RM, Roepe PD. The biophysics of multidrug resistance. Int Rev Cyt 1997;171:121–65. [4] Martiney JA, Cerami A, Slater AFG. Verapamil reversal of chloroquine resistance in the malaria parasite Plasmodium falciparum is specific for resistant parasites and independent of the weak base effect. J Biol Chem 1995;270:22393– 8. [5] O’Neill PM, Bray PG, Hawley SR, Ward SA, Park BK. 4-Aminoquinolines-past, present and future: a chemical perspective. Pharmacol Ther 1998;77:29–58. [6] Reed MB, Saliba KJ, Caruana SR, Kirk K, Cowman AF. Pgh 1 modulates sensitivity and resistance to multiple antimalarials in Plasmodium falciparum. Nature 2000;403:906 – 9. [7] Barnes DA, Foote SJ, Galatis D, Kemp DJ, Cowman AF. Selection for high-level chloroquine resistance results in deamplification of the pfmdr1 gene and increased sensitivity to mefloquine in Plasmodium falciparum. EMBO J 1992;11:3067–76. [8] Ritchie GY, Mungthin M, Green JE, Bray PG, Hawley SR, Ward SA. In vitro selection of halofantrine resistance in Plasmodium falciparum is not associated with increased expression of Pgh 1. Mol Biochem Parasitol 1996;83:35– 46. [9] Hoffman MM, Roepe PD. Analysis of ion transport perturbations caused by hu MDR 1 protein overexpression. Biochemistry 1997;36:11153–68. [10] Roepe PD, Wei LY, Cruz J, Carlson D. Lower electrical membrane potential and altered pHi homeostasis in multidrug-resistant (MDR) cells: further characterization of a series of MDR cell lines expressing different levels of P-glycoprotein. Biochemistry 1993;32:11042–56. [11] Martiney JA, Roepe PD. Are ion exchange processes central to understanding drug-resistance phenomena? Trends Pharmacol Sci 1999;20:62–5. [12] Yayon A, Cabantchick ZI, Ginsburg H. Identification of the acidic compartment of Plasmodium falciparum-infected human erythrocytes as the target of the antimalarial drug chloroquine. EMBO J 1984;3:2695–700. [13] Krogstad DJ, Schlesinger PH, Gluzman IY. Antimalarials increase vesicle pH in Plasmodium falciparum. J Cell Biol 1985;101:2302– 9.

.

[14] Krogstad DJ, Schlesinger PH. The basis of antimalarial action: non-weak base effects of chloroquine on acid vesicle pH. Am J Trop Med Hyg 1987;36:213 – 30. [15] Ginsburg H, Nissani E, Krugliak M. Alkalinization of the food vacuole of malaria parasites by quinoline drugs and alkylamines is not correlated with their antimalarial activity. Biochem Pharmacol 1989;38:2645 – 54. [16] Luz JG, Wei LY, Basu S, Roepe PD. Transfection of mu MDR 1 inhibits Na + -independent Cl−/HCO− 3 exchange in Chinese hamster ovary cells. Biochemistry 1994;33:7239 – 49. [17] Wei LY, Hoffman MM, Roepe PD. Altered pHi regulation in 3T3/CFTR clones and their chemotherapeutic drug selected derivatives. Am J Physiol 1997;272:C1642– 53. [18] Su X-z, Kirkman LA, Fujioka H, Wellems TE. Complex polymorphisms on an :330 kDa protein are linked to chloroquine-resistant P. falciparum in Southeast Asia and Africa. Cell 1997;91:593 – 603. [19] Hoffman MM, Wei LY, Roepe PD. Are altered pHi and membrane potential in hu MDR 1 transfectants sufficient to cause MDR protein-mediated multidrug resistance? J Gen Physiol 1996;108:295 – 313. [20] Gamage SA, Tepsiri N, Wilairat P, Wojcik SJ, Figgit DP, Ralph RK, et al. Synthesis and in vitro evaluation of 9-anilino-3,6-diaminoacridines active against a multidrugresistant strain of the malaria parasite Plasmodium falciparum. J Med Chem 1994;37:1486 – 94. [21] Goodyear I, Pouvelle B, Schneider EG, Trelka DP, Taraschi TF. Characterization of macromolecular transport pathways in malaria infected erythrocytes. Mol Biochem Parasitol 1997;87:13 – 28. [22] Wellems TE, Wootton JC, Fujioka H, Su X-z, Cooper R, Baruch D, et al. P. falciparum CG2, linked to chloroquine resistance, does not resemble Na+/H+ exchangers. Cell 1998;94:285 – 6. [23] Bray PG, Mungthin M, Ridley RG, Ward SA. Access to hematin: the basis of chloroquine resistance. Mol Pharmacol 1998;54:170 – 9. [24] Fitch CD. Plasmodium falciparum in owl monkeys: drug resistance and chloroquine binding capacity. Science 1970;169:289 – 90. [25] Bornmann WG, Roepe PD. Analysis of drug transport kinetics in multidrug-resistant cells using a novel coumarin-vinblastine compound. Biochemistry 1994;33: 12655 – 75. [26] Wu¨nsch S, Sanchez CP, Gekle M, Große-Wortmann L, Wiesner J, et al. Differential stimulation of the Na+/H+ exchanger determines chloroquine uptake in Plasmodium falciparum. J Cell Biol 1998;140:335 – 45.