Characterization of Methylphosphonate as a 31P NMR pH Indicator*

The 31P NMR pH indicator, methylphosphonate, has been extensively characterized, and the uncertainty in pH determination by its chemical shift has been ana-lyzed. The pKa decreases by 0.003 pH unit/"C and 0.06 pH unit/100 mM ionic strength. The pK, appears not to be sensitive to Ca2+ but is sensitive to Mg2+, resulting in an uncertainty of 20.05 pH unit. Substituting 300 mM Na+ for 300 mM K+ causes the pK,, to decrease by 0.1 pH unit. Taking the effects of temperature, ionic strength, and cation identity into account, the overall estimated uncertainty in pH determination can be as high as 20.1 pH unit. Methylphosphonate was tested as a pH indicator in Ehrlich ascites tumor cells. Our data indicate that both the unchanged and monoanion forms of methyl phosphonate are very permeable, rendering this compound unsuitable as a pH indicator in this system. However, the sensitivity of this compound's chemical shift to pH and the relative insensitivity to other parameters suggest that phosphonates, as a group, may be applicable as pH indicators by 31P NMR. or, appropriate, to either

* This work was supported by Grant DCB-8412577 from the National Science Foundation, Grant CB-246 from the American Cancer Society, and National Science Foundation Instrumentation Grant CHE-8208821 to the regional NMR center at Colorado State University. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed. The abbreviations used are: 2-dGlu-6-P, 2-deoxyglucose 6-phosphate; GPC, L-a-glycerophosphorylcholine; PIPES, piperazine-N,N" bis(2-ethanesulfonic acid); HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid; I , ionic strength; pK,, pKa' (Le. practical pK, values). indicator in Escherichia coli by Slonczewski et al. (4) who found that ionic strength had a negligible effect on its pK,. Studies by Gillies et al. (1) confirmed a negligible effect of ionic strength and also suggested that methylphosphonate was relatively impermeable to several mammalian cell types. More recently, methylphosphonate has been used by Lin et al. (5) as an external pH indicator in frog skin studies where it is impermeable and by Labotka and Kleps (6) and Stewart et al. (7) as an intracellular pH indicator in erythrocytes, where it apparently equilibrates rapidly across the membrane. Stewart et al. (7) have also suggested that the chemical shift of methylphosphonate is somewhat dependent on ionic strength.
In the present communication, we analyze the behavior of this indicator in vitro and in Ehrlich ascites tumor cells. We analyze the uncertainty in pH determined by the chemical shift of methylphosphonate as a function of temperature, ionic strength, Mg+, Ca2+, and Na' and K' . We show methylphosphonate to be nontoxic to EAT cells, yet it is not an ideal indicator of intracellular pH in this system due to its high permeability to cellular membranes.

MATERIALS AND METHODS
Cell Culture-Ehrlich lettr6 carcinoma cells, strain E, were obtained from American Type Culture Collection and were cultured in suspension using minimum essential medium-autoclavable (GIBCO) supplemented with 10% fetal bovine serum (HyClone partially characterized, Logan, UT). Cultures were initially inoculated at approximately 1 X lo4 cells/ml. For loading cells, methylphosphonate was added to a final concentration of 20 mM in this growth medium by dissolving the appropriate amount of methylphosphonic acid (Alpha Products, Danvers, MA) into serum-free medium, titrating to pH 7.2 with 10 N NaOH, and filtering through a Sterivex-GS 0.22-fim filter unit (Millipore, Bedford, MA) directly into the culture. Cells were grown at 37 "C in 5% CO,, 95% air to a density of 2-5 X lo5 cells/ml (5 or 6 days) prior to harvest.
For harvesting, the cells were first placed at 4 'C for 1 h/liter of cells. After this time, they were collected by centrifuging for 7 min at 1000 X g in 250-ml polypropylene tubes (Corning Glass Works, Corning, NY). The cell pellets were resuspended (typically 1 ml of pelleted cells/liter of suspension was obtained), pooled, and washed with 250 ml of cold (4 "C) buffer A (in mM) 5 KCl, 140 NaCl, 1.2 CaCl,, 2 MgSO,, 4.2 or 20 NaHC03, 1 KH,PO,, and 50 or 25 PIPES at pH 6.7. The cells were centrifuged again and resuspended in buffer A to cell densities of 20-30% in either a 10-or 20-mm outer diameter NMR tube (Wilmad Glass Co., Inc., Buena, NJ) containing 25 ~1 of Antifoam B emulsion (Sigma). The suspensions were kept in an ice bath for transfer to the spectrometer (about 5 min). At 37 "C, buffer A pH becomes 7.0.
NMR Spectroscopy-Depending on accessibility, either a Nicolet NT-360 with a 20-mm probe operating at 145.78 MHz or a Bruker WP-200 with a 10-mm broadband probe operating at 81.02 MHz was used for the in vivo experiments. Those performed on the Nicolet were accomplished using a hollow fiber bioreactor (2). This bioreactor contains a fiber basket for perfusion and aeration and an air-driven impeller to keep the cells suspended. Experiments have shown that 1 0 ' cells/ml (30% cell density) can be maintained in this reactor for up to 10 h (2). The in vivo experiments performed on the Bruker were oxygenated by bubbling carbogen through the suspension at a Methylphosphonate as a 31P N M R p H Indicator 11033 rate of about 2 bubbles/s. In vivo spectra were initially referenced to the a-phosphate of ATP at -10.05 ppm or, when appropriate, to either diethylmethyl phosphonate at +36.07 ppm or GPC at +0.49 ppm relative to 85% phosphoric acid. All in vitro experiments were performed on the Bruker instrument. The titration spectra were referenced to GPC and/or diethylmethyl phosphonate. Control studies have shown that the chemical shift of diethylmethyl phosphonate is not affected by changes in Ca2+, M e , Z, temperature, Na+, or K+ and that it is nontoxic to EAT cells at concentrations of 10 mM. Spectra were obtained with acquisition parameters as indicated in the figure legends. Titrations-Intracellular environment titrations were accomplished using a buffer containing 10 mg/ml BSA, 1 /AM Ca2+, 140 mM K+, 5 mM Na+, 5 mM Pi, 10 mM HCO;, 20 mM HEPES, 10 mM methylphosphonate, 10 mM GPC, and 10 mM diethylmethyl phosphonate. The BSA was treated with chelating resin (Dow) and extensively dialyzed against ultrapure water (Millipore) to remove contaminating metal ions. The bench titrations were accomplished using a Beckman pH1 71 meter, a Beckman 39522 glass electrode, and a Microscribe 4500 strip chart recorder. The samples were maintained at the appropriate temperatures using a Precision 254 circulating water bath. The pK,, values were determined graphically by the base equivalent midpoint method and compared to the inflection point of the titration curve located by assuming equal deviation on either side of the pK.. NMR titrations were accomplished using the same pH meter and electrode and a Cole-Parmer 1095-00 water bath. The spectrometer was equipped with a temperature control. The pK, values were determined graphically as being the midpoint of the two chemical shift end points.

RESULTS
Effect of Temperature-pK, data obtained from titrations at various temperatures at two ionic strengths are illustrated in Table I. These data indicate that the pK, decreases consistently with increasing temperature. At Z = 306 mM the ApK,/"C is -0.005 with a linear correlation coefficient of 0.96, whereas for I = 93 mM it is -0.003 (r = 0.79). Similar results were obtained for KC1 with 0.2 molar ratio Mg2+/0.08 molar ratio Ca2+ or for 300 mM NaCl, in which the ApK,/"C were -0.005 and -0.003, respectively. Titration of KC1 at Z = 93 mM on the NMR yielded pK, values of 7.65 and 7.62 at 25 (not shown) and 37 "C (Table 11), respectively. These data are consistent with a change of about -0.004 pH unit/"C increase.
Effect of Ionic Strength- Table I also illustrates the effect of ionic strength on the pK, of methylphosphonate. The two extremes of ionic strength were chosen to encompass ionic     (Table II), a difference in pK, values of 0.15 over the same range.
The first three lines of Table I11 show the direct effect of ionic strength on the chemical shift of methylphosphonate at pH 7.5 and 37 "C. A plot of these data gives an effect of -0.135 ppm/100 mM change in ionic strength. Since the sensitivity of methylphosphonate is 2.2 ppm/pH unit at pH unit at pH 7.5 (see below), these data correspond to an effect of 0.061 pH unit/100 mM ionic strength.
Ca2+ and Mg2+ Effect-Since cells can contain varying amounts of either Ca2+ or Mg2+, several experiments were designed to investigate the effects of these divalent cations on the titration curves and chemical shift of methylphosphonate. Table I lists pK, values at 22 and 37 "C with or without both 0.08 molar ratio (2 mM) Ca2+ and 0.2 molar ratio (5 mM) Mg2+. Correcting for the change in ionic strength from 306 to 325 mM with both cations present (which would account for a pK, change of approximately -0.011) the pK, , values are observed to change by -0.02 at 22 "C and -0.04 pH unit at 37 "C.
For Ca2+ alone, Table I11 shows an apparent effect of -0.08 ppm change for either 0.4 or 1.0 molar ratio Ca2+. However, in another experiment, NMR titrations with Ca2+ of a 300 mM KC1 solution at 37 "C showed no change in the chemical shift of methylphosphonate up to a molar ratio of 0.4. When the effect of ionic strength is taken into account, the change in chemical shift becomes -0.04 ppm for a molar ratio of 0.4 and +0.02 ppm for a molar ratio of 1.0 resulting in -0.1 ppm/ molar ratio and +0.02 ppm/molar ratio, respectively. Thus, it appears that the effects for Ca2+ alone are inconsistent and small.
For Mg2+ alone, Table I11 indicates chemical shift changes of -0.1 ppm for a molar ratio of 0.4 and -0.38 ppm for a molar ratio of 1.0. Taking into consideration the effect of ionic strength on chemical shift, these values become -0.06 and -0.28 ppm, respectively, which correspond to a M$+ effect -0.15 to -0.28 ppm/molar ratio. Na+ versus K+ Effect- Table I shows a change in the pK, of -0.1 pH unit when 300 mM Na+ is substituted for 300 mM K+ at either 22 or 37 "C. NMR titration data (Table  11) provide similar results indicating that a difference in the pK, values of 0.05 pH unit occurs between successive samples. NMR data in Table I11 indicate a -0.1 ppm change at pH 7.5. Similar results were obtained for Pi, where the pK, appeared to decrease about 0.05 pH unit between the two tested samples (Table 11).
Intracellular Environment Titrations-Buffers approximating the intracellular environment of EAT cells were titrated at 37 "C to yield calibration curves for intracellular pH determinations. Data for 0.04 molar ratio, 1 molar ratio, and 2 molar ratio (0.4,10, and 20 mM, respectively) M e are shown in Table 11. A titration at 0.5 molar ratio M$+ produced a curve coincident with the 0.04 molar ratio curve (not shown). Comparing the 1 and 2 molar ratio data, in the worst case the pK, decreased by 0.1 pH unit with the increase in M$+. Table   I1 also shows the response of Pi in the same titrations. The Mg2+ effect is similar in that the maximum effect occurs near the pK,. The effect on Pi however, is not observed until Mg2+ is raised to a molar ratio of 4, which decreases the pK, by 0.15 pH unit.
Methylphosphonate as an Intracellular pH Indicator in EAT Cells-Adding methylphosphonate directly to a cell suspension while acquiring 31P spectra fails to give rise to a methylphosphonate peak corresponding to intracellular pH within 5 h (data not shown). Therefore, we have attempted to load methylphosphonate into cells by growing cultures in its presence. Methylphosphonate can be loaded into EAT cells by incubation in the growth media as described under "Materials and Methods." Growth assays are illustrated in Fig. 1 and indicate that concentrations as high as 20 mM methylphosphonate do not inhibit growth. When cells are grown in the presence of 20 mM methylphosphonate for several days, exhaustively washed, and then acid-extracted, 31P NMR spectra such as those shown in Fig. 2 are obtained. Similar results have been obtained in as little as 24 h of incubation. Taking differential saturation into account, the concentrations of intracellular methylphosphonate are only slightly higher than those of ATP in the same spectrum. Data from a number of experiments indicate that intracellular methylphosphonate concentrations are about 5 mM when cells are incubated in 20 mM methylphosphonate. I n vivo 31P spectra of methylphosphonate-loaded cells in a buffer containing 1 mM methylphosphonate at pH 6.8 show a methylphosphonate peak that corresponds to pH = 7.14 ( Fig. 3). However, as illustrated in the time course of Fig. 4, after 15 min a second peak appears, corresponding to pH = 6.78. After 25 min, this downfield peak has increased while the upfield peak has virtually disappeared. At times after 25 min there appears to be only the downfield peak.
The above results indicate that the methylphosphonate is leaking out of these cells. This hypothesis was tested by placing washed methylphosphonate-loaded cells for 1 h in a methylphosphonate-free buffer in an NMR tube through which was bubbled carbogen. 31P spectra of the resulting supernatant showed a methylphosphonate peak of intensity equivalent to about 0.08 mM as determined by standard addition. This amount corresponded to the expected total amount in the cells, further indicating that it had leaked out during incubations.

DISCUSSION
Uncertainty of pH Measurement-In NMR measurements, pH is inferred from the chemical shift of a suitable indicator. Hence, a relevant question is: "What is the uncertainty in pH from the measured chemical shift?" In the present study, uncertainty arises from two sources, i.e. in assigning chemical shift and in relating chemical shift to pH.
Uncertainty in assigning chemical shift is a function of both signal-to-noise ratio and line widths. The absolute error in assigning chemical shift can be estimated to be one-half of the peak width at one noise level below the peak height. For methylphosphonate in vivo, this value is about +25 Hz, corresponding to errors of up to 0.16 pH unit in assigning pH at the pK. On the other hand, the standard error of the mean, Exponential MP multiplication of the free induction decay was accomplished with 30-Hz line broadening. Chemical shifts of peaks were assigned as the average between the shifts at half-height.
AT P 2 0 arrived at by repetitive sampling, is +7 Hz or 0.04 pH unit.
Uncertainty in relating pH to chemical shift is illustrated in Fig. 5 , which shows two titration curves, A and B, representing an indicator under two different conditions and thus two different pK, values. At a measured chemical shift (do) there are two possible values of pH (pHol or pHo2) that can be estimated. Using the median pH value between the two curves as the determined pH, an uncertainty of 5% (pHo2 -pHol) results. At some point, one has to decide what level of uncertainty is tolerable. Generally, an uncertainty in assigning intracellular pH of k0.05 pH unit is tolerable, since most other measurements of intracellular pH are not more accurate (3, 9).
Temperature Effect-Titration data have shown that a -0.004 pH unit/"(= change in the pK, of methylphosphonate occurs with temperature (Table I). This effect can be considered negligible as long as the temperature is maintained within k10 "C, which is easily attainable. This is borne out by NMR experiments which have measured no significant differences IO 0 -10 -2 0 pprn between the relative chemical shifts of methylphosphonate at 20 and 37 "C.
Ionic Strength Effects-The pK, of methylphosphonate is dependent on ionic strength both in theory (8) and in practice (our data and Ref. 7). The effect of ionic strength on the pK, of methylphosphonate is similar to that observed for Pi and 2-dGlu-6-P (3) in that the end points of the titrations are not affected by ionic strength, yet the pK, is affected. This effect on pKa is translated into an effect on the chemical shift. Our data indicate a change in chemical shift of -0.994 ppm/100 mM ionic strength, while Stewart et al. (7) shows -0.829 ppm/ "ionic strength unit." However, the dependence of chemical shift on ionic strength at a particular pH is not the relevant parameter. The parameter of interest is the uncertainty in the pH determined by a particular chemical shift, as discussed above. Titration curves must be performed at ionic strengths which encompass those estimated to occur physiologically. In the present study, such data indicate a maximum deviation in estimated pH of 0.12 pH unit. This maximum occurs at pH 30 20 PPm  FIG. 4. Sequential 6-min regional 31P spectra of EAT cells preincubated in 20 m M methylphosphonate. Cells were cultured for 4 days in the presence of methylphosphonate, harvested, and resuspended as detailed in the text. At t = 0, cold cell suspension was placed in a Bruker 200-MHz magnet that was prewarmed to 37 "C. Spectra were collected in 5-min blocks thereafter. Data indicate loss of upfield peak and subsequent appearance of downfield peak. pH values near the pK,. Therefore, the maximum uncertainty in absolute pH determination is k0.06 pH unit. The uncertainties for Pi and 2-dGlu-6-P are k0.08 and kO.1 pH unit, respectively (1). The uncertainties of determining changes in p H are much less, since cells presumably do not change their ionic strength by large amounts. Therefore, although the chemical shift of methylphosphonate is dependent upon ionic strength, the effect of ionic strength is negligible in determining the pH. Ca2+ Effect-The effect of Ca2+ on the chemical shift of methylphosphonate is inconsistent and small. The worst case, at 0.4 molar ratio, indicated a -0.1 ppm/molar ratio change. Increasing the molar ratio to 1.0 causes a change in the chemical shift in the opposite direction. We interpret this inconsistency, combined with the small changes observed, as Ca2+ having a negligible effect on the pH determined by the chemical shift of methylphosphonate.
Mg2+ Effect-Our data suggest that Mg2+ has some effect on the pK, and, hence, chemical shift of methylphosphonate. At 1.0 molar ratio, neglecting effects of ionic strength, Mg2' has a -0.3 ppm/molar ratio effect on the chemical shift of methylphosphonate at its pK,. However, in the presence of other compounds which stimulate the intracellular environment, no effect of Mg2+ was observed up to 1.0 molar ratio.
The reason for this difference is currently unexplainable. It is possible that the BSA or HEPES in the intracellular environment buffer may be chelating some of the M%+. BSA is an unlikely candidate, however, because titrations with unchelated and chelated BSA are identical. Also, Roberts et al. ( 3 ) showed that BSA at 50 mg/ml had no effect on the titration curves of Pi or 2-dGlu-6-P in the presence or absence of 5 mM MgC12. Nonetheless, the relevant analysis involves the uncertainty of pH as discussed above. From this type of analysis, Mg2+ at a molar ratio of 2.0 in the intracellular environment causes an uncertainty of k0.05 pH unit in the pH determined by the chemical shift of methylphosphonate. The data suggest that a 2:l complexation of Mg2+ to methylphosphonate is occurring. However, this is not conclusive since it appears from data in Table I11 that molar ratios less than 2 also have an effect.
It appears from the above data that the major effect of divalent cations on the chemical shift of methylphosphonate is due mostly to Mg2+ and that Ca2+ has little or no effect. This is also suggested by the following observations. (a) The change in chemical shift by Mg2+ alone accounts for the combined Mg2+/Ca2+ effect (Table 111); and (b) NMR titration curves containing 0.2 molar ratio, Mg2+ with and without 0.08 molar ratio Ca2+ are coincident (not shown). These data imply that either there is no effect due to Ca2+ or that the Ca2+ effect is masked by the M P effect.
The Effect of Monovalent Cations-The differences in pK, between methylphosphonate in solutions of 300 mM K+ and 300 mM Na' is as large as 0.1 pH unit. This would represent an uncertainty due to monovalent cations of k0.05 pH unit. Under physiological conditions, this uncertainty would be much less, since the extra-and intracellular concentrations of both of these ions are well known and relatively constant.