Human Tumor Cell Proliferation Evaluated Using Manganese-Enhanced MRI

Background Tumor cell proliferation can depend on calcium entry across the cell membrane. As a first step toward the development of a non-invasive test of the extent of tumor cell proliferation in vivo, we tested the hypothesis that tumor cell uptake of a calcium surrogate, Mn2+ [measured with manganese-enhanced MRI (MEMRI)], is linked to proliferation rate in vitro. Methodology/Principal Findings Proliferation rates were determined in vitro in three different human tumor cell lines: C918 and OCM-1 human uveal melanomas and PC-3 prostate carcinoma. Cells growing at different average proliferation rates were exposed to 1 mM MnCl2 for one hour and then thoroughly washed. MEMRI R1 values (longitudinal relaxation rates), which have a positive linear relationship with Mn2+ concentration, were then determined from cell pellets. Cell cycle distributions were determined using propidium iodide staining and flow cytometry. All three lines showed Mn2+-induced increases in R1 compared to cells not exposed to Mn2+. C918 and PC-3 cells each showed a significant, positive correlation between MEMRI R1 values and proliferation rate (p≤0.005), while OCM-1 cells showed no significant correlation. Preliminary, general modeling of these positive relationships suggested that pellet R1 for the PC-3 cells, but not for the C918 cells, could be adequately described by simply accounting for changes in the distribution of the cell cycle-dependent subpopulations in the pellet. Conclusions/Significance These data clearly demonstrate the tumor-cell dependent nature of the relationship between proliferation and calcium influx, and underscore the usefulness of MEMRI as a non-invasive method for investigating this link. MEMRI is applicable to study tumors in vivo, and the present results raise the possibility of evaluating proliferation parameters of some tumor types in vivo using MEMRI.


Introduction
Uncontrolled cellular proliferation is the hallmark of cancer, and proliferation rate, i.e., the rate of tumor cell division, is linked to prognosis for several types of cancer [1,2,3,4]. Currently the only method to spatially monitor local tumor cell proliferation in vivo is positron emission tomography (PET), which uses the accumulation of 18 F-labeled 39-deoxy-39-fluorothymidine ( 18 F-FLT), fluorodeoxyglucose ( 18 F-FDG), or 2-11 C thymidine (11 C TdR) as a proliferation marker [5]. While application of PET as a method of detecting proliferation in vivo remains promising, its spatial resolution is limited compared to other imaging modalities, such as MRI. Bading and Shields acknowledge that ''an effective and clinically practical means for the imaging of cell proliferation is still an unrealized objective.'' [5].
Cell proliferation is usually associated with an increase in cytoplasmic calcium ion, either from the extracellular space or from intracellular calcium stores [6,7,8]. Much of the extracellular Ca 2+ enters the cell via calcium-permeable channels [6,7,8].
Indeed, tumor cell proliferation has been specifically linked to calcium ion channel activity in some, but not all, tumors [6,7,9,10,11], suggesting that calcium ion channel activity could be a useful surrogate marker of tumor cell proliferation. A powerful method for investigating calcium ion channel activity in vivo is monitoring the extent of tissue uptake of manganese ion, Mn 2+ , a Ca 2+ analog [12,13]. Manganese can enter cells via calcium ion channels, particularly through voltage-gated channels [12,13], although other routes, including transferrin receptormediated or DMT1-dependent routes, may also contribute [14,15]. Importantly, Mn 2+ accumulates intracellularly due to a slow rate of efflux and acts as an MRI contrast agent by increasing the tissue longitudinal relaxation rate (R 1 = 1/T 1 ) in proportion to manganese concentration [16,17]. Manganese-enhanced MRI (MEMRI) has been successfully used to functionally image brain [16,18,19,20,21] and retinal [22,23,24] activity, as well as the activity of other tissues [25]. These considerations suggest that MEMRI might be usefully applied to monitor tumor cell proliferation.
Free Mn 2+ ion is known to accumulate in tumors in vivo [25,26,27,28,29]. Previously, using MEMRI in a nude rat model, we demonstrated significant Mn 2+ uptake by C918 uveal melanoma xenografts relative to surrounding tissues and speculated that the tumor-specific Mn 2+ uptake may have been a result of high proliferation rates within the tumor [30].
In this study, we test the hypothesis that tumor cell uptake of a calcium surrogate, Mn 2+ (measured with MEMRI), is linked to tumor cell proliferation rate and will be a biomarker of proliferation in at least some tumors. Specifically, the growth of three human tumor cell lines was characterized in vitro and their proliferation rates were correlated to MEMRI R 1 (1/T 1 ) values.

Human Tumor Cell Lines
Three different human tumor cell lines were used in this study. The human uveal melanoma cell lines C918 and OCM-1 were used, because we had previously shown that C918 cells took up Mn 2+ in vitro and in vivo [30], and we wished to further investigate this class of tumor. The OCM-1 cells were originally cultured from a human choroidal melanoma specimen in the 1980's [31], while the C918 cells were derived from a patient tumor in 1996 at the University of Iowa [32]. To extend the analysis of Mn 2+ uptake to a type of cancer with a higher clinical incidence, we also investigated the human prostate carcinoma line PC-3, which was originally generated from a bone metastasis of a grade IV prostatic adenocarcinoma [33]. All cells were maintained in RPMI media +10% fetal bovine serum (FBS) + antibiotic under standard incubating conditions. Cells were seeded into 6-well plates or standard tissue culture flasks at a density of 20.8 cells/mm 2 .

Measurement of Proliferation Rate of Human Tumor Cell Lines
To describe the growth of the tumor cells in vitro, cells were trypsinized, harvested, and counted on different days after seeding. The cell density was determined by dividing the cell number by the area of the well or flask. The number of cell divisions over that period of days, D, was calculated as: where D = the number of cell divisions, C = tumor cell density (cells/mm 2 ), and C 0 = initial cell density, i.e., the seeding concentration (cells/mm 2 ). These data were fit to a Weibull growth model [34] using nonlinear least-squares regression, while setting C 0 = 20.8 cells/mm 2 (GraphPad Prism, GraphPad Software, Inc., La Jolla, CA). The Weibull model was chosen after fit comparisons with other common growth models, including the Gompertz and sigmoid logistic models [35]. Model comparisons using Akaike's Information Criterion (AIC) method [36] showed that the Weibull model was the better model (AIC probabilities.99.5%) compared to either the Gompertz model or the sigmoid logistic model. Since the Weibull model best described the growth of the cells, we used the corresponding equation to describe the number of cell divisions as a function of time: where C max = maximum cell density (cells/mm 2 ), k = the inverse of the time constant (1/days n ), and n = dimensionless constant.
The proliferation rate, dD/dt (divisions/day), is given by:

MEMRI of Human Tumor Cell Pellets
Based on the tumor growth curves, cells were used for these experiments on different days after seeding. First 19.8 ml of stock MnCl 2 solution (10 mg MnCl 2 N4H 2 O/ml of 0.9% saline; 50.5 mM MnCl 2 ) was added to each ml of RPMI medium in the flask to reach a concentration of 1 mM MnCl 2 . The solution was left on the cells for one hour in the incubator. The cells were then rinsed with Hanks' balanced salt solution and trypsinized. After cell detachment was evident, the trypsin was quenched with RPMI media +10% FBS. The cells were centrifuged, and the resultant pellet was washed twice more in RPMI +10% FBS. The cells were counted and resuspended to a concentration of 1.5610 6 cells/100 ml. Two hundred ml of the suspension were placed in a 0.65 ml microcentrifuge tube, yielding a total of 3610 6 cells in the tube. The cells were pelleted by gravity for at least 20 minutes. At early time points, e.g., Days 2 or 3, it was sometimes necessary to use more than one flask to obtain 3610 6 cells. The microcentrifuge tubes were placed in a 7 T magnet (Bruker ClinScan , Billerica, MA), and the average longitudinal relaxation time (T 1 ) for each pellet was determined using a partial saturation T 1 approach as follows. Several spin-echo images (two 2.0 mm thick slices -one through pellets and one through supernatants -spaced 4 mm apart; matrix size 4646576, field of view 46657 mm, TE 13 ms) were acquired at different repetition times in the following order (with number of acquisitions per TR in parentheses): TR 0.15 s (6), 3.50 s (1), 1.00 s (2), 1.90 s (1), 0.35 s (4), 2.70 s (1), 0.25 s (5), and 0.50 s (3). Images acquired with the same TR were averaged offline using ImageJ (http://rsbweb.nih.gov/ij/). Average signal intensity (SI, based on a circular region-of-interest encompassing most of the cells in each tube) from each sample varies as a function of TR according to a monoexponential function: where a, b, and T 1 are the fitted parameters. This function was fit to data using the Levenberg-Marquardt nonlinear least-squares algorithm in the minpack.lm library (v.1.1.1, by Timur V. Elzhov and Katharine M. Mullen) for R [37]. The R 1 (1/T 1 ) values (longitudinal relaxation rates) directly reflect manganese levels [16,17].
Each pellet likely contained a heterogeneous cell population, e.g., cells in different phases of the cell cycle, and, potentially, correspondingly distinct values for R 1 , and this could confound our estimates of the pellet R 1 . As a first approximation, we assumed a fast-exchange limit wherein the exchange of water between heterogeneous compartments (including cell populations) is fast relative to the acquisition of the T 1 data set. In this case, if the individual populations had different T 1 values, that information would be blurred and would show up as a single exponential.
To check this assumption, in a preliminary study using images from a typical run of nine pellets, we compared fits of monoexponential and biexponential functions to the intensity vs. TR data. When the data were fit to the monoexponential model, the fitted parameters had tight confidence intervals. In contrast, the biexponential fits resulted in parameter values with very large confidence intervals, indicative of non-unique parameter values (data not shown). Since these models are nested, i.e., one model is an extension of the other, an extra sum-of-squares F test was used to statistically compare the fits [36]. The F-test showed that the monoexponential fit was adequate to describe the data in every case and that there was no reason to invoke the more complex biexponential model (p.0.13, n = 9). These results are consistent with results from phantom experiments, which showed that T 1 values have to differ by a factor of 2 to 3 to be reliably distinguished by a biexponential fit [38]. Therefore, all of the signal intensity data in this study were fit to the monoexponential equation noted above.

Cell Cycle Analysis
For C918 cells, separate flasks were seeded at a concentration of 20.8 cells/mm 2 , and cells were harvested on different days after seeding. The washed cells were fixed in 70% ethanol by adding absolute ethanol dropwise to the cell suspension. The cells were stored at 220uC until flow cytometry could be performed on multiple samples. Three hours before flow cytometric analysis, the cells were washed three times in Hanks' balanced salt solution and were stained with a 50 mg/ml propidium iodide solution in the presence of 0.5 mg/ml RNase A. DNA content of each sample was determined by flow cytometry on a BD LSR II flow cytometer (BD Biosciences, San Jose, CA). The percentages of cells in G 0 /G 1 phase, S phase, and G 2 /M phase were determined using ModFit LT software (Verity Software House, Topsham, ME). Flow cytometry was performed by the Microscopy, Imaging, and Cytometry Resources Core at The Karmanos Cancer Institute, Wayne State University.
For the PC-3 cells, a portion of the cells from the same flask that was used to perform the MEMRI experiments was processed for flow cytometry and analyzed in the same manner as described above.
To functionally describe the relationship between the percentage of cells in S phase and the proliferation rate, dD/dt, three different models were compared using Akaike's Information Criterion (AIC) method [36]: the sigmoid logistic model, the Weibull model, and a simple monoexponential model. The 3parameter logistic model resulted in fits with lower sum-ofsquares errors than the 3-parameter Weibull model for both cell lines. The AIC test revealed that the symmetric sigmoid logistic model was superior to the monoexponential model (AIC probability.70%). Therefore, the relationship is best described by: where S = percentage of cells in S phase (%), dD/dt = proliferation rate (divisions/day), S max = maximum percentage of cells in S phase (%), d S = constant (days/division), and t S = value of dD/ dt at which S is half its maximum value (divisions/day). Forms of the same three functions were similarly compared to test their ability to adequately describe the relationship between the percentage of cells in G 0 /G 1 phase and dD/dt. The 4parameter logistic model resulted in fits with lower sum-ofsquares errors than the 4-parameter Weibull model for both cell lines. Since the AIC test revealed that the symmetric sigmoid logistic model was the better model (probability.60%) compared to the monoexponential model, the sigmoid logistic model was used: where G 01 = fraction of cells in G 0 /G 1 phase (%), G A = constant (%), G B = constant (%), d G = constant (days/division), and t G = constant (divisions/day). Note that G A or G B taken individually has no biological significance and could be .100, but G A -G B is the minimum percentage of cells in G 0 /G 1 phase (%). The sum of all three cell fractions must be 100: where G 2M = fraction of cells in G 2 /M phase (%). Solving this equation for G 2M yields: Substituting Equations 4 and 5 into Equation 6 yields the following expression for G 2M : Statistical Analysis The fits of the cellular growth data to the logistic, Gompertz, and Weibull models were compared using Akaike's Information Criterion (AIC) method [36]. This test was used rather than an extra sum-of-squares F test, because the models are not interrelated, i.e., they are non-nested [36]. The AIC method was also used to compare the fits of the percentage of cells in S phase or G 0 /G 1 phase versus proliferation rate data to the sigmoid logistic, Weibull, and exponential models.
Two-way ANOVA analysis was used to compare intracellular R 1 values among the three cell lines with or without Mn 2+ exposure. Tukey's HSD post-hoc comparison tests were used to check for differences between any two groups. Two-way ANOVA and Tukey's HSD were performed using R [37]. A p-value,0.05 was considered statistically significant.
Correlations between R 1 values and proliferation rate were determined using linear regression analysis (GraphPad Software, Inc., La Jolla, CA). A regression p-value,0.05 was considered statistically significant.
The lack of fit of the weighted-average model to the experimental data was determined using the lack-of-fit ANOVAbased F-test [39,40]. Using the sum of squares of the measurement and modeling errors, it tests the hypothesis that the lack of fit of the model curve is much greater than the measurement error [40]. A p-value,0.05 was considered statistically significant, indicating the model provided an incomplete description of the data.

Tumor Cell Growth and Proliferation Rate
As shown in Figure 1, the growth data for all three cell lines were well described by the Weibull model (Equation 2), with all coefficients of determination equal to 0.97. Proliferation rates were calculated from the fitted parameters using Equation 3 ( Figure 1B). C918 cells had the highest maximum proliferation rate of 2.1 divisions/day, which occurred 1.7 days after seeding. The OCM-1 cell line showed a maximum proliferation rate of 1.5 divisions/day at 2.4 days after seeding. The PC-3 cells grew the slowest with a maximum proliferation rate of 1.3 divisions/day at 2.5 days after seeding.

Tumor Cell Mn 2+ Uptake
A 263 two-way ANOVA analysis revealed a significant impact of Mn 2+ exposure (p,0.0001) and of cell type (p,0.0001) on the R 1 value of the cell pellets, regardless of their proliferation rate ( Figure 2). Post-hoc analysis demonstrated that, as expected, for each cell line, the R 1 of cells exposed to Mn 2+ for one hour and then rinsed was significantly greater than the R 1 of corresponding cells not exposed to Mn 2+ (p,0.0001, Tukey's HSD test, Figure 2). R 1 , and thus Mn 2+ uptake, in the PC-3 and OCM-1 cell lines was greater than R 1 in the C918 cells (p,0.0001). In addition, the R 1 value of the PC-3 cells was significantly less than that of the OCM-1 cells (p = 0.001). When no Mn 2+ was added, there were no differences among any of the pellet R 1 values (Tukey's HSD test, p = 1.000). Supernatant R 1 values, regardless of whether the cells were exposed to Mn 2+ or not, ranged between 0.63 and 0.87 sec 21 .

Relationship between Tumor Cell Mn 2+ Uptake and Proliferation Rate
The R 1 values for pellets of cells following MnCl 2 exposure are shown as a function of proliferation rate in Figure 3

Phases of the Cell Cycle and Proliferation Rate
Since there was a positive correlation between MEMRI R 1 values and proliferation rate in the PC-3 and C918 cell lines, cell cycle changes in those cell lines were also investigated. As expected, the fraction of cells in each of the phases of the cell cycle changed with proliferation rate (Figure 4). In both cell lines, the fraction of cells in S phase increased with proliferation rate, while the fraction of cells in G 0 /G 1 decreased. The percentage of cells in S phase as a function of proliferation rate, dD/dt, was well described by Equation 4 for both the PC-3 cells (r 2 = 0.921, n = 29, Figure 4A) and the C918 cells (r 2 = 0.920, n = 20, Figure 4B). Similarly, Equation 5 adequately described the percentage of cells in the G 0 /G 1 phase as a function of dD/dt for both the PC-3 cells (r 2 = 0.967, n = 29, Figure 4A) and the C918 cells (r 2 = 0.848, n = 20, Figure 4B). Equation 10 reasonably described the changes in the G 2 /M fraction of cells as well (Figure 4).

Modeling the Relationship between Tumor Cell Mn 2+ Uptake and Proliferation Rate
It is possible to derive a simple, general model to describe the positive correlation between Mn 2+ -induced changes in R 1 and proliferation rate in the PC-3 and C918 cell lines (Figure 3). The cell pellets at each proliferation rate have at least three distinct subpopulations of cells in different phases of the cell cycle: G 0 /G 1 , S, and G 2 /M ( Figure 4). If each subpopulation takes up a different amount of Mn 2+ , each subpopulation would have its own R 1 value, and the average R 1 of any mixed cell population would be the weighted average of these individual cell cycle-specific values. Thus, the overall R 1 of the pellet is described by the following weighted average: where (R 1 ) pellet = R 1 value of cell pellet (sec 21 ), R S = R 1 value of a cell in S phase (sec 21 ), R 01 = R 1 value of a cell in G 0 /G 1 phase (sec 21 ), and R M = R 1 value of a cell in G 2 /M phase (sec 21 ). From Figure 4, it is known that the size of each subpopulation in the pellet changes with proliferation rate, and the parameters S, G 01 , and G 2M can be expressed as functions of proliferation rate. Substituting Equation 6 into Equation 8 and simplifying yields: Substituting Equations 4 and 5 into Equation 9 yields: The only unknowns in this equation are R S , R 01 , and R M . It is not known whether the amount of Mn 2+ taken up by each subpopulation remains constant or changes with proliferation rate, i.e., R S , R 01 , and R M are either constants or functions of dD/ With the available data, it is possible to directly test the first possibility, i.e., that the three subpopulations demonstrate unique amounts of ion flux (R 01 , R S , and R M ), but these amounts are not affected by proliferation rate. In other words, the correlation between R 1 and proliferation is solely due to the relative size of each subpopulation. Application of this simple model will help guide future studies by indicating whether the second more complicated model needs to be further investigated.
The PC-3 and C918 cellular R 1 values following MnCl 2 exposure as a function of proliferation rate shown in Figure 3 were fit to Equation 10 (with R S , R 01 , and R M as constants) using nonlinear least-squares regression (GraphPad Prism,GraphPad Software, Inc., La Jolla, CA). Cell line-specific constant parameters were obtained from the fits shown in Figure 4.
For the PC-3 cells, the model fit of all 26 data points ( Figure 5A, solid line, r 2 = 0.304, n = 26) passes through the mean values (open symbols). A lack-of-fit ANOVA-based F-test [39,40] revealed that the model adequately described the data (p = 0.978). In other words, given the variance in the PC-3 data and the quality of the present model fit, it is not possible for an alternative model to produce a significantly better fit. The best-fit values of R 01 , R S , and R M with the 95% confidence intervals (95% CI) were 1.16 (95% CI: 0.82 to 1.51) sec 21 , 1.86 (95% CI: 0.66 to 3.07) sec 21 , and 2.02 (95% CI: 21.87 to 5.91) sec 21 , respectively. Despite the wide confidence intervals (a consequence of uncertainty in R M ), if the value for R 01 is fixed within its 95% CI (between 0.82 and 1.51 sec 21 ), all solutions yield fits in which R 01 ,R S . As an additional check that R 01 ,R S , the four mean values were fit to the model ( Figure 5A, dashed line). In this case, the fit was indistinguishable from the curve obtained when all of the individual points were fit to the model (solid line). The resulting fit yielded an r 2 of 1.000 (n = 4) and the following parameter values: R 01 = 1.16 (95% CI: 1.07-1.25), R S = 1.86 (95% CI: 1.54-2.18), and R M = 2.02 (95% CI: 0.98-3.05). From the confidence intervals, it is evident that R 01 is always less than R S , but that R M cannot be as reliably predicted. Based on these results, the correlation between PC-3 pellet R 1 and proliferation rate is consistent with a change in the sizes of the cellular subpopulations during proliferation, without a change in their individual R 1 values.
For the C918 cells, a fit of the 40 cell pellet R 1 values following MnCl 2 exposure to Equation 10 yielded the curve shown in Figure 5B (solid line, r 2 = 0.436, n = 40). The best-fit values of R 01 , R S , and R M with the 95% confidence intervals (95% CI) were 0.67 (95% CI: 0.48 to 0.86) sec 21 , 1.57 (95% CI: 1.17 to 1.97) sec 21 , and 1.07 (95% CI: 20.20 to 2.33) sec 21 , respectively. Although, on visual inspection, the model does follow the general trend of the data, there are notable deviations. At low proliferation rates, the model predicts a more rapid change in pellet R 1 than seen in the data, resulting in a large overestimate of R 1 at a proliferation rate of 0.9 divisions/day ( Figure 5B). Similarly, at high proliferation rates, the model predicts a plateau in the pellet R 1 value, which is not evident in the experimental data. Consistent with these observations, a lack-of-fit ANOVA-based F-test [39,40] revealed that the model does not adequately describe the data (p = 0.017).
The assumption of constant R 1 values for C918 cells in the different phases of the cell cycle appears questionable.

Discussion
In this study, in two of three cancer cell lines, we found that MEMRI R 1 values, which reflect cellular Mn 2+ uptake, changed with tumor cell proliferation rate. These data underscore the usefulness and sensitivity of MEMRI to the known heterogeneous nature of different tumor cell proliferation rates. Although we and others have shown that tumor cells accumulate Mn 2+ when exposed to MnCl 2 and that Mn 2+ changes the cellular MEMRI R 1 values [25,27,28,29,30], this is the first time that the Mn 2+ -induced R 1 changes have been correlated with changes in proliferation rate. In other studies, increased tumor Mn 2+ uptake has been correlated with increased tumor neuroendocrine activity [27] or tumor cell Mn-superoxide dismutase (Mn-SOD) levels [29].

Tumor Cell Mn 2+ Uptake
Tumor cell Mn 2+ uptake could occur by several different mechanisms, including expression and/or activity of Ca 2+ channels, transferrin receptors, and divalent metal-ion transporter-1 (DMT1) channels [14,15]. Both voltage-gated Ca 2+ channels (VGCC) [13,41,42] and some TRP channels [43,44,45] are permeable to Mn 2+ , and the expression of these channels has been linked to tumor cell proliferation [6,7,10,11,46,47,48]. Although we cannot rule out a role for transferrin receptor-mediated or DMT1-dependent Mn 2+ uptake, preliminary studies in our laboratory suggest that VGCCs, most likely T-type, are important regulators of Mn 2+ uptake in the PC-3 and C918 cell lines. Therefore, we speculate that the difference in MEMRI R 1 values, i.e., Mn 2+ uptake, among the three cell lines, is related primarily to differences in the expression or activity of VGCCs.
The differences in Mn 2+ uptake and cellular R 1 values among the three cell lines (Figure 2) are most likely the result of different expression or activity of various calcium ion channels or other transport mechanisms responsible for Mn 2+ entry into the cells. Interestingly, the C918 cells took up the least Mn 2+ , i.e., had the lowest pellet R 1 values, even though they had the highest proliferation rates ( Figure 1B). The OCM-1 cells, in which Mn 2+ uptake was not correlated with proliferation rate, had the highest R 1 values and an intermediate proliferation rate. These results demonstrate that the absolute level of Mn 2+ uptake (R 1 value) is not correlated with proliferation rate across cell types. In other words, a high R 1 value does not mean that a particular cell line proliferates more rapidly than another cell line with a lower average R 1 . Rather, it is the change in the pellet R 1 value for a specific cell line that is indicative of a change in cellular proliferation rate.
As expected, substantial leakage of Mn 2+ out of the cells into the overlying supernatant did not occur, suggesting stable intracellular accumulation of manganese and adequate wash procedures. The lack of any significant Mn 2+ leakage out of the cells after removal of the MnCl 2 and washing of the cells implies a relatively slow rate of Mn 2+ efflux out of the cells, similar to what has been reported in neuronal tissues [23,49].

Relationship between Tumor Cell Pellet MEMRI R 1 and Proliferation Rate
Our initial hypothesis was based on reports that intracellular Ca 2+ levels change during cellular proliferation and that these changes may, at least in part, be caused by an increase in the uptake of extracellular Ca 2+ [6,7,8]. Since Mn 2+ is a calcium surrogate, this uptake of Ca 2+ might be detectable as a change in intracellular Mn 2+ measured using MEMRI. Thus, we hypothesized a positive correlation between average MEMRI tumor cell pellet R 1 value and cellular proliferation rate.
Two of the three tumor cell lines tested, PC-3 and C918, revealed such a positive correlation between average cell pellet MEMRI R 1 and proliferation rate. No correlation was found for the OCM-1 uveal melanoma cell line. There are several possible, but not mutually exclusive, explanations for this result. First, it is possible that the proposed link between increased Ca 2+ influx and proliferation is not universal to all tumor cell lines. For example, in the OCM-1 cells, the cytosolic Ca 2+ change that helps drive proliferation may be dominated by release of intracellular Ca 2+ stores, rather than uptake of extracellular Ca 2+ [6,7,8]. Alternatively, extracellular Ca 2+ may enter the OCM-1 cells through Mn 2+ -impermeable Ca 2+ channels, so that the Ca 2+ changes are not detected by MEMRI. Third, while a proliferation-related increase in Mn 2+ uptake may occur through certain routes, the change in MEMRI R 1 could be too small to be detected over the background of high baseline Mn 2+ permeability through alternative proliferation-independent routes. This last possibility is consistent with the fact that the OCM-1 cells had the highest Mn 2+ uptake (R 1 ) of all three cell lines, regardless of the proliferation status (Figure 2). At this point, additional work is needed to identify the underlying reasons for a lack of a significant correlation in the OCM-1 cells. In any event, these data highlight the sensitivity of MEMRI to differential cell calcium handling in various tumor cell types.
As presented in the Results section, the simple weighted-average model (Equation 10) assumes that the changes in pellet R 1 can be fully explained by changes in the relative distribution of cells in different phases of the cell cycle (each with a cell-cycle specific constant R 1 ) as proliferation rate changes. For the PC-3 cells, the relationship between the average pellet R 1 and proliferation rate was adequately described by this model ( Figure 5A). The results of the modeling suggest that PC-3 cells in the G 0 /G 1 phase of the cell cycle took up less Mn 2+ , i.e., had a lower R 1 value, than cells in S phase. The dramatic increase in the fraction of cells in S phase during proliferation and concomitant decrease in the fraction of G 0 /G 1 cells ( Figure 4A) seems most likely responsible for the positive correlation between PC-3 pellet R 1 and proliferation rate (Figure 2A).
For the C918 cells, the poorer fit of the model to the data ( Figure 5B) suggests that the simple weighted-average model with constant cell cycle-specific R 1 values is insufficient to completely describe the relationship between average cell pellet R 1 and proliferation rate. This result suggests that the Mn 2+ uptake and the R 1 value of one or more of the cell subpopulations changed with proliferation rate, but more work is needed to investigate this possibility.
In summary, MEMRI is a useful non-invasive method for accurately measuring the link between tumor calcium channel activity and tumor proliferation in vitro. Future studies will investigate whether these proliferation-related changes in MEMRI R 1 can be confirmed in vivo.