1. Introduction

Water is the most basic requirement of life, comprising at least 70% of the mass of most living organisms. Liquid water serves as the suspension medium for all normal physiologic processes, acting as a diffusive environment for the exchange of nutrients, ions, and waste, and allowing for the proper folding, structure, and function of complex biomolecules [1,2]. The ability of a cell to maintain a proper osmotic balance in response to outside stimuli is critical to the preservation of life [3].

The measurement of intracellular water content is a notoriously difficult endeavor, and early efforts have proved elusive. Somewhat primitive methods involving the mass weighing and drying of organs have been attempted to obtain bulk properties of tissues, though these cannot be applied to a single cell [4]. X-ray microanalysis (using oxygen concentration as a proxy for water) has been used in single cell layers to compute the weight percentage of water, though this requires frozen-hydrated tissue preparations, precluding analysis of dynamic events in living cells [5]. Suspended microchannel resonators (SMR) have been constructed which enable calculation of a single cell’s water mass after consecutive measurements of the cell’s buoyant mass in two fluids of different densities [6], however, the need to exchange immersion buffers rapidly also precludes the study of rapid water exchange.

One particularly promising technique for rapidly detecting the intracellular water content is quantitative phase imaging (QPI), an interferometric imaging modality which captures the quantitative phase delay $\varphi ({x,y} )$ of light passing through a thin sample according to the well-known relation

In this regard, measuring the cellular volume is key to unlocking intracellular water content from QPI images. Unfortunately, QPI is a projection technique which does not easily translate to the measurement of three-dimensional volumes. Several methods have been proposed to solve this problem. Recently, Park et al. utilized the narrow distribution of volumes among homogenously sized red blood cells to extract their water content [9], though this will not be useful for other cell types, which exhibit size heterogeneity. Several authors have combined QPI with a “transmission-through-dye” (TTD) approach [10,11], which surrounds the cell with a high concentration of cell-impermeable absorbing dye, and uses brightfield imaging to detect the spatially-varying reduction in absorption caused by the cell to compute the cell’s height at each point, and thus, the cell’s volume. While ingenious, this technique exhibits practical downsides, including the need for high concentrations of exogenous contrast agent in the imaging solution, custom imaging geometries requiring a narrow gap of roughly the cell’s height between two coverslips, the need for careful calibration of the absorption coefficient of the dye, and instrumentational complexities associated with multimodal imaging.

In this work, we utilize a priori knowledge of a spherical cellular geometry to extract the intracellular water content, by mass and by volume, from a single quantitative phase image. This is done by imaging freshly plated cells, which assume a highly spherical geometry [12–15]. Image processing techniques are applied to obtain the cellular volume from the projected phase image, while spatial integration of the image is used to derive the dry mass. Both measurements are combined with the externally measured refractive index of the imaging medium to compute the intracellular water content under the assumption of a two-component mixture model.

We utilize our method to study the influx and efflux of water across cellular membranes in response to a single, microsecond-duration pulsed electric field (PEF). PEF exposure is a well-studied cellular stressor, with applications in cancer therapeutics [16,17], electroporation [18,19], and fundamental biophysics [20,21]. While QPI has been used in previous works to report on the evolution of phase-based parameters in electropermeabilized cells [22,23], our work extends this approach by quantifying the absolute intracellular water content of electropermeabilized cells with high temporal precision.

One of the hallmark bioeffects of PEF exposure in cells is swelling, which occurs via membrane permeabilization and water uptake following the colloid osmotic gradient. While water uptake has been inferred using surrogate indicators such as cell area [24], diameter [25], volume [26], and exogenous fluorescent dye intensity [27], the precise water content as a function of time within cells exposed to PEF has yet to be quantified. In this study, we examine a population of CHO-K1 cells exposed to a PEF and measure an increase in their intracellular water content from approximately 1400 to 1580 femtoliters on average, moving from 87.7 ± 0.7% to 89.1 ± 0.4% of the cell’s content by mass. We further examine the same cells in a hyperosmotic environment and observe a loss in intracellular water from approximately 970 fL to 880 fL, moving from 80.0 ± 0.9% to 78.3 ± 1.0% of the cell’s mass following PEF-induced permeabilization. Finally, we demonstrate the utility of our method by examining intracellular water uptake in human T lymphocyte (Jurkat) cells following a 600 ns electric pulse in the presence of mercury and gadolinium, two ions which are known to affect water-transporting aquaporin channels, and to modify cell swelling in response to PEF, respectively. We believe our method represents a convenient approach for the study of water influx and efflux, PEF bioeffects, and osmotic balancing across cellular membranes.

2. Materials and methods

2.1 Quantitative phase microscope

A QPI microscopy platform, previously utilized by our laboratory to describe bleb mass dynamics [28] and to quantify bioeffects associated with nanosecond-duration electric pulses [29], was utilized to acquire quantitative phase images of cells. Briefly, a supercontinuum laser (NKT Fianium) was utilized as a light source for a custom-built transmission microscope with illumination wavelength $\lambda $ = 633 nm and a bandwidth of $\Delta \lambda $ ∼ 1 nm. An objective lens (Olympus PlanApo N, 60x/1.42 NA, oil immersion) and tube lens (f = 180 mm) relayed the sample plane to a CMOS camera (FLIR Blackfly S, Sony IMX250 sensor) with 3.45 µm pixels, for an effective pixel size of 57.5 nanometers at the sample plane. A reference beam, matched in optical pathlength to ensure temporal interference, was collimated onto the camera at a slight angle, creating a spatial interference pattern encoding the complex optical field at the sample. Standard methods in off-axis digital holography were employed [30] to extract the phase image, $\varphi ({x,y,t} )$. The spatial phase noise of the system, computed as the standard deviation of background regions, was typically 10-30 mrad.

2.2 Phase retrieval

The phase image was recovered from the raw interferogram using a custom MATLAB (MathWorks, 2019b) script. A two-dimensional fast Fourier transform (FFT) was used to separate the sample contribution in the interferogram, which was re-centered in frequency space. An inverse FFT operation was utilized to recover the complex field. From here, the argument of the complex field was taken to retrieve the phase image. A two-dimensional phase unwrapping algorithm based on the transport of intensity equation was applied to remove 2$\pi $ ambiguities [31], and a two-dimensional polynomial was fit to the background to de-trend and produce a flat field. Further details on these methods may be found in previous works [15,32].

2.3 Cell culture

Chinese hamster ovarian cells (CHO-K1, ATCC CCL-61) were cultured according to standard protocols. We selected CHO-K1 cells after examining CHO-K1, U937 (ATCC CRL-1593.2), and HEK 293TN (System Biosciences, LV900A-1) cell lines, and finding that U937 and HEK 293TN cells exhibited occasional to moderately frequent blebbing and asymmetric swelling in response to an applied PEF, precluding the analysis method, which assumes a spherical cellular geometry. In contrast with U937 and HEK 293TN, CHO-K1 cells rarely blebbed, and swelled isotopically in response to a PEF. All lines were cultured according to manufacturer protocols.

CHO-K1 cells were maintained at 37°C and 95% humidity with 5% CO₂ in air. A complete growth medium consisted of Kaighn’s modification of Ham’s F-12 Medium, 10% fetal bovine serum, 2 mM L-glutamine, and 1% by volume 100 U/mL penicillin/streptomycin (ATCC 30-2300). Cells were passaged prior to imaging, diluted using fresh media, and allowed to settle only briefly on a glass-bottom dish (35 mm, No. 1.5 coverslip, Poly-D-Lysine coated, Mattek Corporation) to allow adhesion without losing the spherical geometry of the cell. The dish was then washed with either physiological buffer (Live Cell Imaging Solution, Invitrogen, HEPES buffered, pH = 7.4, mOsm = 300), or hyperosmotic buffer (custom, see Imaging Buffers below) and cells were immersed in their respective solutions for imaging. All imaging was completed within 45 minutes to ensure cell sphericity, and only cells with a high degree of circularity were included for analysis. Cells with an obvious lack of spherical symmetry (before or after the pulse) or which exhibited blebbing were excluded from the dataset, as these would represent sources of error in our cell volume computation. Blebs were identified as any portion of the cell which extended visibly beyond the boundary of the circle assigned by the Hough transform.

For the experiments involving mercury and gadolinium, Jurkat clone E6-1 cells (ATCC, TIB-152) were maintained at 37°C and 95% humidity with 5% CO₂ in air. A complete growth medium consisted of Gibco RPMI 1640 Medium, 10% fetal bovine serum, 2 mM L-glutamine, and 1% by volume 100 U/mL penicillin/streptomycin (ATCC 30-2300). In preparation for the experiment, 500 µL of Jurkat cells suspended in their medium were removed from their flask and centrifuged for 2 min. Cells were then resuspended in medium without fetal bovine serum to increase their ability to adhere to surfaces. The cells in serum-free solution were then pipetted onto a glass-bottom dish and allowed to settle. As suspension cells, Jurkats maintain a naturally spherical shape and do not flatten over time. Cells were incubated in serum-free media for 1 hour prior to QPI experiments. After incubation, the dish was carefully washed using physiological buffer solution, and cells were immersed in this solution containing either 5 µM HgCl₂ (mercuric chloride) or 100 µM GdCl₃ (gadolinium chloride). Control dishes containing physiological buffer solution only were used in both experiments. As with CHO-K1, cells exhibiting blebbing or anisotropic swelling were excluded from the analysis.

2.4 Imaging buffers

To demonstrate the influx and efflux of water in response to PEFs, two imaging solutions were used. To investigate CHO-K1 swelling and water uptake in response to PEF, a physiological saline buffer was purchased commercially (Invitrogen Live Cell Imaging Solution, pH = 7.4, mOsm = 300). To demonstrate water efflux, a hyperosmotic solution was created by modifying an isosmotic solution developed by Nesin et. al. [26] consisting of 35 mM NaCl, 5 mM KCl, 4 mM MgCl, 3 mM HEPES, 10 mM glucose, 2 mM Na-EGTA, and 26 mM PEG-4000. This isosmotic solution was made to be hyperosmotic by adding an additional 5 mM PEG-4000, for a final concentration of 31 mM. The solution was confirmed to have an osmolarity of ∼309 mOsM using a freezing point osmometer (Advanced Instruments Osmo1). Prior to imaging, the cells were washed with the buffer of interest (either physiological or hyperosmotic), followed by immersion in the buffer. Test solutions for Jurkat experiments were modified by adding either 5 µM HgCl₂ (mercuric chloride) or 100 µM GdCl₃ (gadolinium chloride) to the physiological buffer solution.

The variable RI of the outside solutions can somewhat alter the signal obtained using QPI. This is accounted for by utilizing an additional term in the dry mass computation, as shown in Eq. (6). Using a commercial refractometer (Fisherbrand, HDR-P6, accuracy ±0.0003), the physiological buffer was found to have a mean RI of 1.3350, while the hyperosmotic buffer had a mean RI of 1.3559. These figures are displayed in the text as 1.335 and 1.356 to reflect the accuracy of the instrument. Each of these was used in the custom MATLAB code as the refractive index of the medium (${n_{medium}}$) while the refractive index of water (${n_{water}}$) was obtained using distilled water and measured to be 1.3331, in good agreement with literature values. All refractive index measurements were taken at a room temperature of 22 °C.

2.5 Electric pulse delivery and timing

A scheme for delivery of PEF to the cell has been extensively described in previous works [28,29]. Briefly, a tungsten electrode pair with inter-electrode spacing of 225 microns was placed such that with the electrode tips just contacted the glass surface at a 45° angle. The cell of interest was then centered between the electrode tips. Prior to imaging, the electrode tips were retracted 50 microns laterally in the image plane towards the proximal end of the electrodes to remove diffractive imaging artifacts which degraded the phase image. Once in place, a square electric pulse (∼610 V, 4 µs duration) was delivered using a custom pulse generator. This produced a PEF at the sample location with an amplitude of 25 kV/cm and duration of 4 µs, as verified using COMSOL Multiphysics modeling, also previously described [29]. A duration of 4 µs was chosen as it sits between regimes of microsecond to millisecond PEF (used for irreversible electroporation and traditional gene delivery [33]) and nanosecond PEF (which are known to exhibit interesting and unique bioeffects [34,35]).

For experiments involving Jurkats, a slightly wider electrode pair with an inter-electrode spacing of ∼353 microns was used to reduce the field strength to ∼17.2 kV/cm, while the pulse duration was reduced from 4 µs to 600 ns. This pulse regime produced reliable water influx in Jurkat cells, which are more sensitive to PEF [36]. Future studies are being planned to track water influx/efflux resulting from the wide range of pulse durations and amplitudes currently available.

2.6 QPI imaging

The cell of interest was first centered between the electrode pair, followed by a slight retraction of the electrodes, as described in prior sections. The imaging session was initiated using a delay generator (DG535, Stanford Research Systems), which triggered camera acquisitions via a data acquisition (DAQ) digital I/O PCIE card from National Instruments (PCIe-6612) at 1 frame/second for two minutes. The integration time of the camera was set to 20-40 ms to fill the well depth of the camera without saturation. After a baseline recoding period of five seconds, the delay generator triggered the custom pulsing electronics to deliver a 4 µs, 25 kV/cm PEF to the cell. QPI raw interferograms were saved to the disk at a rate of 1 frame/second, and post-processed for phase retrieval and determination of the intracellular water content. Each imaging window lasted two minutes, for a total of 120 frames/exposure.

2.7 Extraction of water content from QPI images

The water content (by mass and by volume) of each cell was computed using a three-step process. First, the dry mass for the cell was computed from the phase image using previously described methods [29,37]. The dry mass of intracellular constituents is typically given by

This formulation has been well-validated in the literature, however, it assumes that the phase object is immersed in a medium with a RI very close to water. While this approximation is fairly accurate for cells in physiological buffer, it will underestimate the dry mass of cells in a medium with high RI (such as the hyperosmotic buffer). To account for the contribution of the refractive index of the imaging medium, a corrective factor was added to the dry mass, consisting of the dry mass encompassed by a sphere of equivalent size to the cell, with a RI equivalent to the imaging medium as measured using a commercial refractometer.

The dry mass of this theoretical sphere is given by

Because $\mathrm{\smallint\!\!\!\smallint }d({x,y} )dxdy$ is simply the volume of the object ${V_{cell}}$, our corrective factor simplifies to

Here, the volume ${V_{cell}}$ of the sphere is easily computed from its diameter D, obtained using a circular Hough transform on the phase image, such that ${V_{cell}} = \; \frac{4}{3}\pi {\left( {\frac{D}{2}} \right)^3}$. This is a well-validated method for determining the morphology of spherical cells from QPI images [12,13,15].

The medium-adjusted dry mass (accounting for the RI of the imaging solution, and thus, capturing all non-aqueous material) is

Once the dry mass is obtained, we compute the intracellular water content through the assumption of a two-component mixture (dry and fluid) within a known volume. The medium-adjusted dry mass $m_{dry}^\ast $ can be used to compute a dry volume ${V_{dry}} = \; \frac{{m_{dry}^\ast }}{{{\rho _{dry}}}}$, using the average mass density of proteins ${\rho _{dry}}$ = 1.37 g/mL obtained from the literature [40,41]. The fluid volume is simply the remaining fraction of the total volume ${V_{fluid}} = {V_{cell}} - \; {V_{dry}}$, and the fluid mass may be calculated using the fluid density,

Finally, the water content by volume (${\theta _{{H_2}O}}$) of the cell is found by dividing the fluid volume by the cell volume, ${\theta _{{H_2}O}} = \,{\raise0.7ex\hbox{${{V_f}}$} \!\mathord{\left/ {\vphantom {{{V_f}} {{V_{cell}}}}} \right.}\!\lower0.7ex\hbox{${{V_{cell}}}$}}$. Alternatively, the water content by mass is measured by dividing the fluid mass by the sum of the fluid mass and medium-adjusted dry mass, ${u_{{H_2}O}} = \frac{{{m_{fluid}}}}{{{m_{fluid}} + m_{dry}^\ast }}$. Both quantities are expressed as percentages for convenience.

While our methods are aimed primarily at describing the intracellular water content, other biophysical parameters become accessible using our processing scheme. The cellular refractive index may be computed from the phase image of a spherical body, as described in previous works [12,13,15]. Furthermore, the mass density of the cell can be calculated using the sum of the masses of dry and fluid constituents, divided by the volume of the cell

Lastly, the cellular volume is easily computed from the sphere diameter as described earlier. We report these three biophysical parameters in addition to the water content to show their evolution in response to a PEF.

2.8 Experimental design

Two extracellular environments were initially studied: 1) CHO-K1 cells in standard physiological imaging buffer, and 2) CHO-K1 cells in hyperosmotic imaging buffer. These were chosen to model the influx and efflux of water across the cell membrane, respectively, after permeabilization by the PEF. For each condition, between 16 and 20 image sequences of unique cells were acquired to capture the sample’s response to PEF. An equivalent number of sham exposures were taken for each case, obtained by turning the high voltage power supply off and initiating exposures in the same manner. Any cell which experienced noticeable anisotropic deformation or blebbing was removed from consideration, as this violates the assumption of a spherical sample. All cell images were acquired within ∼45 minutes of plating to ensure sphericity. To demonstrate the ability of the system to report intracellular water content by both mass and volume, the intracellular water volume (in fL, where 1 fL =10⁻¹⁵L = 1 µm³) and the intracellular water fraction (by mass) are reported for all three exposure conditions as a function of time. A demonstration of these methods is shown in Fig. 1.

 

Fig. 1. Demonstration of the experimental method. From the raw phase images (left) a Hough transform is used to fit the profile of the spherical cell to extract a volume. The gray dashed line indicates the circular profile identified by the Hough transform. Intracellular water content (right) can be computed as a function of time from the dry mass and cell volume using a two-component mixture model. The dashed red line indicates delivery of the PEF. Note that the absolute phase intensity will be affected by the refractive index of the medium. This is corrected using Eq. (6). Scale bars are 5 µm.

Download Full Size | PDF

2.9 Confocal microscopy

Experiments were performed using a confocal microscope to confirm both cell sphericity and membrane permeabilization. A Zeiss LSM 710 inverted microscope (Carl Zeiss Microscopy, Berlin, Germany), connected to ZEN black software v. 2.3SP1 (ZEN Digital Imaging for Light Microscopy RRID:SCR_013672) was utilized to acquire fluorescent and differential interference contrast (DIC) images. To confirm sphericity, approximately 6 × 10³ CHO-K1 cells were plated on a 35 mm Poly-d-lysine coated glass bottom dish in physiological buffer solution containing 2 µL of FluoVolt dye (1000x in DMSO, ThermoFisher Scientific, Waltham, MA) to label cell membranes. A 63x/1.4 NA oil immersion objective (Zeiss EC Plan Neofluar) was utilized to view cells. ZEN Black software automatically optimized Z-stack acquisition intervals to satisfy Nyquist sampling criteria, and a 0.88 RI corrective factor (RI of the buffer solution divided by the RI of Zeiss Immersol immersion oil) was applied. Fiji [42] was utilized to measure x, y and z axes of spherical CHO-K1 cells (N = 25), with measurements taken from the inner edge of the plasma membrane. Ratios of axes lengths were used as a surrogate measure of sphericity, with ideal axis ratios approaching 1. All fluorescence images were captured using a 488 nm argon laser to excite the dye, along with standard FITC filter settings and a 512 × 512 frame size. All images were acquired within 1 hour to replicate cell adhesion conditions used during QPI experiments.

Finally, membrane permeabilization was confirmed using YO-PRO-1 Iodide, a normally cell-impermeant dye which exhibits increased fluorescence upon penetration of a permeabilized plasma membrane [43]. Cells were plated in 3 mL of physiological buffer solution containing 15 µL YO-PRO-1 Iodide (491/509) dye (ThermoFisher Scientific, Waltham, MA). ZEN Black software was utilized to initiate a time series fluorescence experiment, in which the frame rate, imaging window duration, pulse delivery, timing, and electrode geometry were identical to the QPI experiments described above. A 40x/1.3 NA oil immersion objective (Zeiss Plan-Apochromat) was used to collect fluorescence images, and Fiji was utilized to measure the mean pixel intensity within the cell boundary. Sham exposures were obtained by initiating the acquisition with the high-voltage power supply turned off. N = 15 exposed cells and N = 15 sham cells were used to confirm membrane permeabilization.

2.10 Mercury and gadolinium-induced water uptake modulation in Jurkats

To further demonstrate our method, we measured water uptake following a nanosecond-duration PEF in immortalized human T-lymphocyte (Jurkat) cells immersed in either 5 µM mercuric chloride (HgCl₂) or 100 µM gadolinium chloride (GdCl₃). Jurkat cells were chosen for their physiological relevance as immortalized human lymphocytes, as well as their natural sphericity, which is synergistic with our approach. Since Jurkats are more susceptible to pulsed electric fields [36], we reduced the PEF duration from 4 µs to 600 nanoseconds, a common pulse duration in the literature for nanosecond electric pulse bioeffects [26]. We further reduced the field by utilizing an electrode pair with a larger inter-electrode spacing of ∼353 microns, resulting in a field delivered to the cell of roughly 17.2 kV/cm. This pulse regime produced reliable water influx in Jurkat cells.

The two ions (mercury and gadolinium) were chosen due to their known effects on biological membranes. Mercury is a known broad-spectrum blocker of aquaporins [44], a class of membrane channels known to transport water between intracellular and extracellular compartments under imbalanced osmotic conditions. Gadolinium, by contrast, has been shown to inhibit the effects of electropermeabilization by modifying the plasma membrane [45]. A comparison of water uptake following pulsing in the presence of these ions helps to elucidate the underlying biophysical mechanisms of water uptake following PEF exposure. The concentrations of mercury and gadolinium were selected from literature values which produced inhibition of cellular response under water efflux [46] or electropermeabilization [47] conditions, respectively.

3. Results

3.1 Confirmation of cell sphericity and permeabilization

Figure 2 shows the results of our confocal microscopy experiments confirming both membrane permeabilization and cell sphericity. Clear YO-PRO-1 uptake was observed following pulsing ($\Delta F/F$ ∼ 548 ± 50%, versus no change in the sham population), indicating permeabilization of the plasma membrane. Furthermore, confocal imaging indicated that cells were highly spherical, with X/Z, Y/Z, and X/Y aspect ratios of 1.01 ± 0.01, 1.02 ± 0.01, and 0.98 ± 0.01, respectively (errors are standard error of the mean). These results are useful for vindicating our method of deducing cell volume using the assumption of sphericity for freshly plated cells.

 

Fig. 2. Confirmation of membrane permeabilization and cell sphericity. (Left) YO-PRO-1 uptake is clearly observed, indicating permeabilization of the plasma membrane. Error bars represent standard error of the mean. (Middle) Axis ratios confirm that cells are highly spherical, validating our method. The center line indicates the median of the data, while the box indicates 25^th and 75^th percentiles, and whiskers indicate the data range. (Right) Representative 3D reconstruction and cross-section of confocal fluorescence image stack used to measure axis ratios. Scale bar = 10 µm.

Download Full Size | PDF

3.2 CHO-K1 in physiological buffer

Figure 3 shows the dynamics of water influx for CHO-K1 cells in physiological buffer in response to a PEF. Water content is shown both as an absolute volume (in fL) and as a percentage of the cell’s mass. Due to the variance in initial water content, we additionally plot both metrics normalized to their respective initial conditions, which helps to better visualize the effects of pulse exposure. In response to the PEF event, we observed substantial water uptake across the cell membrane. Within the pulsed population (N = 18) the cells exhibited an initial average water content of 87.7 ± 0.7% by mass, which increased to 89.1 ± 0.4% at the end of the two-minute imaging window. Measured using the volume of water within the cell, this represents an increase from 1406 ± 108 fL to 1578 ± 114 fL, a mean increase of approximately 12%. Note that because the cell is mostly water, a substantial increase in the cell’s intracellular water volume corresponds to a more muted, but observable increase in the intracellular water fraction, ${u_{{H_2}O}}$. The effects of the pulse are confirmed in the sham population, where the average intracellular water content did not substantially change during the exposure window, moving from 87.2 ± 0.6% to 87.3 ± 0.6%. The fluid volume was also constant during this period, moving from 1392 ± 132 fL to 1405 ± 131 fL.

 

Fig. 3. Population averages of changes in intracellular water volume and percent water by mass, both absolute values (top) and normalized to their initial condition (bottom) for cells in a physiological buffer. Substantial water uptake is observed for the pulsed, but not the sham populations. The dashed red line indicates delivery of the PEF. Shaded regions represent standard error of the mean. N = 18 (pulsed) and 20 (sham).

Download Full Size | PDF

3.3 CHO-K1 in hyperosmotic buffer

Figure 4 displays the intracellular water content for CHO-K1 cells in hyperosmotic buffer. As might be expected, the initial water content (prior to pulsing) is reduced for this population in comparison with the cells in standard imaging solution, with an average initial water content of approximately 80%, substantially lower than the ∼87% figure for cells in physiological buffer. The cellular diameter was also smaller for this population, with a mean diameter of 12.9 µm versus 14.1 µm for the cells in physiological buffer. This observation helps to validate our method, as the cells are confirmed to follow the quintessential pattern of shrinking and expelling water in a hyperosmotic environment.

 

Fig. 4. Population averages of changes in intracellular water volume and percent water by mass, both in absolute units (top) and normalized to the starting value (bottom), for cells in a hyperosmotic buffer. Significant water loss is observed for the pulsed, but not the sham populations. The dashed red line indicates delivery of the PEF. Shaded regions represent standard error of the mean. N = 17 for both pulsed and sham populations.

Download Full Size | PDF

For cells in this environment, the PEF induced a net water efflux, with the mean intracellular water content reducing from 971 ± 57 fL to 880 ± 54 fL, and the intracellular water content (by mass) falling from 80.0 ± 0.9% to 78.3 ± 1.0% in response to the PEF. The effects of the pulse were again confirmed by the sham population, which maintained its water volume across the two-minute window, only changing from 981 ± 111 fL to 965 ± 107 fL. These results are summarized in Table 1.

Table 1. Intracellular water volume, ${\boldsymbol{V}_{{\boldsymbol{H}_2}\boldsymbol{O}}}$, mass fraction of water, ${\boldsymbol{u}_{{\boldsymbol{H}_2}\boldsymbol{O}}}$, and net changes to these values, for each exposure condition.^a

View Table

3.4 Refractive index, density, and volume

As noted in our Materials and Methods section, our processing scheme involves the computation of the cellular refractive index, mass density (in g/cm³), and volume (in fL). These parameters and their time evolutions for all experiments are shown in Figs. 5 (in absolute units) and 6 (normalized to the initial value). Collectively, these results mirror those expected from our observations of the intracellular water content – as more water enters the cell, it reduces the RI and density, and increases the volume. In physiological buffer, the cellular RI drops following a pulse exposure, from approximately 1.359 to 1.356 on average. The opposite effect is observed in a hyperosmotic buffer, with the RI increasing from 1.375 to 1.379. Similarly, the average cell density decreased from 1.035 g/cm³ to 1.030 g/cm³ in the physiological buffer, and increased from 1.057 g/cm³ to 1.063 g/cm³ in hyperosmotic buffer. Finally, the cell volume followed the opposite trend from the density and refractive index. In physiological buffer, cell volume swelled substantially, from ∼1540 fL to ∼1710 fL during the 2-minute window. The opposite effect was observed in hyperosmotic buffer, with a reduction in volume from ∼ 1150 fL to 1060 fL.

 

Fig. 5. Other biophysical parameters measured using our QPI detection scheme. The intracellular RI, density, and volume are shown as a function of time. For cells in physiological buffer, the RI and density are reduced as water enters the cell, while the cell volume expands. For cells in hyperosmotic buffer, the RI and density increases as water exits the cell, while the cell volume is reduced. The dashed red line indicates delivery of the PEF at t = 5 seconds. Shaded regions represent standard error of the mean.

Download Full Size | PDF

While the responses to PEF events are interesting, the initial values of the RI, density, and volume reflect the condition of the cell in its environment. For example, prior to pulsing, the average density and RI of cells in physiological buffer was substantially smaller than those of cells in hyperosmotic buffer, while the average cellular volume was significantly greater. These reflect the homeostatic response of cells to their external environment prior to PEF exposure. To account for the differences, we also plot the RI, density, and cell volume normalized to their initial conditions in Fig. 6. This allows for improved visualization and understanding of the bioeffects associated with PEF exposure in different environments.

 

Fig. 6. Other biophysical parameters measured using our QPI detection scheme normalized to their starting value to better compare effects from different exposure conditions. The dashed red line indicates delivery of the PEF at t = 5 seconds. Shaded regions represent standard error of the mean.

Download Full Size | PDF

3.5 Effects of mercury and gadolinium on water uptake in Jurkats

Finally, the effects of mercury and gadolinium on water uptake in Jurkat cells are displayed in Fig. 7. In physiological buffer, the intracellular water content increased following application of a 600 ns, ∼17.2 kV/cm pulse by approximately 1.6 ± 0.2%, while in the presence of 5 µM mercury, this number was 1.2 ± 0.2%. This change was not significant, p > 0.2, measured using a two-tailed t-test. In contrast, 100 µM GdCl₃ significantly reduced the change in intracellular water uptake, from 1.4 ± 0.2% in the sham condition to 0.8 ± 0.1%, p < 0.005. These results suggest the limited involvement of aquaporins in the water uptake process in Jurkats following nanosecond PEF exposure, and adds precision to results from the literature, which showed reduced swelling of pulsed Jurkat cells exposed to gadolinium [47].

 

Fig. 7. (Left) Water uptake for a population of Jurkat cells in physiological buffer which included either 0 µM (N = 13) or 5 µM (N = 11) HgCl₂, a known broad-spectrum blocker of aquaporins. No significant effect was observed, p > 0.2. (Right) An identical experiment, with cells exposed to either 0 µM (N = 13) or 100 µM (N = 14) GdCl₃, which has been shown to inhibit cell swelling following electric pulse exposure. Here, water uptake is reduced, p < 0.005. Error bars represent standard error of the mean.

Download Full Size | PDF

4. Discussion

In this work, we have developed a system for tracking intracellular water content, label-free, in a microscopy platform utilizing QPI. Our method was validated by tracking the influx and efflux of water across the membrane of CHO-K1 cells in response to a PEF event. Underlying assumptions of membrane permeabilization and cell sphericity were confirmed using confocal microscopy. Dynamic changes in biophysical parameters of CHO-K1 cells were also reported in response to a PEF. Finally, a case study of water uptake in naturally spherical Jurkat cells was presented, where water uptake was modulated in the presence of gadolinium, but not mercury, helping to elucidate the mechanism of PEF-induced water transport.

While it is generally understood that the optical phase signal is heavily influenced by the intracellular water content, many previous studies have used QPI to track osmotic movement by reporting changes to the phase as a surrogate marker [48]. While somewhat informative, there is an ambiguity between changes to the cell RI and thickness, preventing conclusive analysis of the water content even when considering the dry mass. Our method attempts to greatly simplify QPI-based acquisition of the intracellular water content using a simple geometric assumption which has been independently verified for freshly plated cells [12] and which we confirmed experimentally here.

As some authors have shown, it is possible to decouple the RI from the cell thickness [49], and even to compute the intracellular water content, by acquiring multiple QPI images after washing with immersion media with different refractive indices. Indeed, this technique has been known for some time [50]. Medium-exchange methods have obvious drawbacks similar to the SMR methods mentioned earlier. The need for rapid exchange of the immersion medium will limit the acquisition rate of the technique, preventing the study of fast water dynamics such as those observed here. This method would also likely induce sample movement, which QPI will detect even at the nanometer scale. Furthermore, swapping of the extracellular medium is likely to induce rapid changes in the intracellular water content without very precise osmotic balancing. Osmolarity is not the only concern here, as large molecules (which cannot cross the electropermeabilized membrane) nevertheless contribute to the colloid osmotic gradient. This is particularly apparent in our dataset, as our CHO-K1 population in hyperosmotic buffer exhibited significantly lower initial intracellular water (80.0 ± 0.9%) compared to the cells in physiological buffer (87.7 ± 0.7%), likely due to the cell-impermeant PEG-4000 acting to draw water from within the cellular compartment. The difference here is notable and indicates the effects of homeostatic water balance for cells under distinct osmotic conditions.

One interesting phenomenon was the volumetric behavior of the sham, or unexposed CHO-K1 populations. In physiological buffer, the sham population swelled slightly during the exposure period, with an increase in water volume from 1392 ± 132 fL to 1405 ± 131 fL. The sham population in hyperosmotic buffer, by contrast, shrunk slightly, and lost a small amount of water, moving from 981 ± 111 fL to 965 ± 107 fL of water within the cell. Due to the large variation in cell-to-cell water content, these effects are better visualized when normalized to their initial conditions, where it becomes apparent that during the imaging window, sham cells gained ∼1% additional water in physiological buffer, while similar sham cells lost ∼1% of their water in hyperosmotic buffer. While these are small changes, they are still interesting, as they correspond to the natural tendencies expected of CHO-K1 cells under these osmotic conditions, the effects of which are dramatically increased by membrane permeabilization following PEF.

Our observation that gadolinium (but not mercury) reduced PEF-driven water uptake in Jurkat cells suggest that the cell membrane, rather than aquaporins, is the primary mediator of PEF-induced water transport. This aligns with the prevailing theory of electropermeabilization by sub-microsecond pulses [51], which considers the downstream bioeffects of PEF to be primarily membrane-driven. Still, it is possible that other cell lines which express a higher number of aquaporins (such as glial cells) will respond differently, or alternatively, that aquaporins mediate recovery from PEF, as cells return to homeostatic equilibrium through osmotic diffusion. Our lab is conducting ongoing efforts to quantify the expression of aquaporins in various cell lines. Future work will link water uptake under mercury exposure, and recovery from PEF, with aquaporin expression to conclusively determine whether aquaporins play a role in the biophysics of PEF.

One potential source of error in our measurements is the ability to accurately fit the circular profile of CHO-K1 cells using the Hough transform. This is particularly a concern when there is lesser contrast in QPI imaging caused by a medium with a higher RI, such as the cells in hyperosmotic buffer, which can artificially reduce the phase signal. Even in this case, we can still very reliably identify the cell border. In the sham population of cells in hyperosmotic buffer, for example, where no swelling is expected, the average change in the measured cell diameter between t = 0 and t = 2 minutes was ∼0.1 µm, below the spatial resolution of our system, and indicating that our measurements are highly repeatable across images. Furthermore, all circular fits were verified by visual inspection, and any cell whose diameter was not accurately described by the circular profile was removed from consideration.

In this work, we measured the RI of our solutions at $\lambda $ = 589 nm, however, our QPI microscope utilizes light at a slightly longer wavelength of $\lambda $ = 633 nm, due to the availability of suitable spectral filters. Even though chromatic dispersion is non-zero, this will be a small source of error. For example, the RI of pure water is expected to differ by only 0.0007 between these wavelengths [52] which approaches the instrumental accuracy of our refractometer (±0.0003). Furthermore, the additional corrective term in our dry mass equation (Eqs. (5) and (6)) accounts for the difference in RI between the medium and distilled water, both of which are measured by our refractometer. Since chromatic error will be directionally the same in the indices which are subtracted from one another, we expect much of this error to cancel in the corrective term.

While our system is useful for most cell types which assume spherical geometries after plating, a primary limitation of our method is the need for this spherical geometry, which imposes a limited time window between cell plating and cell adhesion, precluding long-term studies and hindering the use of more geometrically complex cells like neurons, especially in their fully adhered and differentiated stages. In the future, it may be possible to modify tomographic QPI modalities [53] to extract the cellular water content, possibly by using a thresholding operation in the RI tomogram to compute the cell’s volume as a function of time. Of course, the need for several QPI holograms generally results in significantly longer acquisition times for RI tomography systems, which may preclude monitoring of rapid water dynamics such as those studied in this work. Imaging of quantitative water fractions within living, adhered cells throughout their life cycle should be a focus of future efforts.

This study represents a proof-of-concept for our method of tracking water content in cells exposed to PEF. While our results are interesting, substantial work remains to quantify water content across different cell types, PEF dose responses, pulse durations and energies, and to quantify the effects of multipolar pulses and long-term exposures. Ongoing work is underway to measure water dynamics across a plethora of experimental variables.

5. Conclusion

In this work, we utilized quantitative phase imaging and a priori knowledge of a spherical cellular geometry to compute the intracellular water content, by mass and volume, of single cells. We used this system to study rapid water flux across biological membranes caused by PEF-induced membrane permeabilization. We demonstrated our ability to quantify water influx for CHO-K1 cells in physiological buffer, and water efflux for cells in hyperosmotic buffer, all noninvasively. Underlying assumptions of cell sphericity and membrane permeabilization were confirmed using confocal microscopy. Finally, we presented a case study demonstrating that gadolinium, but not mercury, reduced water uptake following a 600 nanosecond PEF in Jurkat cells, pointing towards the cell membrane (rather than aquaporins) as a primary mediator of PEF-induced water influx. We hope our technique will be useful for researchers interested in measuring the precise water content and kinetics of water movement in living cells.