Search
Close this search box.

Nondestructive, longitudinal, 3D oxygen imaging of cells in a multi-well plate using pulse electron paramagnetic resonance imaging – npj Imaging

Our first goal was to design hardware for environmental control during oxygen imaging of cells in 96-well strip-well plates by controlling temperature, CO2, and humidity. Recently, we introduced a 25 mT (720 MHz resonance frequency) EPROI instrument JIVA-25™, and we designed and tested a multi-well plate incubator resonator (MWIR) for the instrument. Figure 2 shows the schematics and components of MWIR. JIVA-25™ has a 10 cm air gap between the magnet poles where the MWIR was to be placed. The MWIR has three components: (a) a rectangular resonator for oxygen imaging (Fig. 2a) (b) a multi-well plate enclosure to keep the cells at desired high humidity and gas (95% air + 5% CO2) conditions (Fig. 2b, c), and (c) heating control to maintain the temperature of 37 °C (not shown).

Fig. 2: Schematics of the MWIR apparatus.
figure 2

a A rectangular resonator without air directors and covers. b a cross-section of multi-well plate enclosure and the resonator, air flow is shown using yellow direction arrows. c a resonator with air director (left) and cover (right) installed; a two-channel air intake is visible. d Three 8-well stripwells from a 96-well plate with columns C to F containing 150 μL of culture medium with 1 mM OX071. The 12 wells outlined in green were used for pO2 imaging. e Thermal image of the multi-well plate enclosure immediately after removal from MWIR showing that the temperature was maintained at 37 °C. f Relative humidity plot shows that a gas flow rate of 30 sccm sets the humidity at 100% while 3.75 sccm sets the humidity at 70%. g Representative pO2 maps (slice #4 at ~ 3.2 mm from the bottom of the well) of the middle 12 wells (columns C to F) filled with PBS during reoxygenation with 95% air + 5% CO2 after N2 bubbling. SCCM (standard cubic centimeter per min).

Resonators are special volumetric units in magnetic resonance instruments that confine radiofrequency energy and allow it to interact with the studied object52,53,54. Typical EPROI resonators are cylindrical. However, to accommodate a multi-well plate, we designed and tested a rectangular 2-gap loop-gap resonator of 42 mm L × 38 mm W × 26 mm H (Fig. 2a). This is one of the largest volume resonators used in the field of EPROI. The cross-section of this resonator can accommodate three strip wells of a 96-well strip-well plate (Fig. 1D). Only the middle 12 wells (3 × 4) fit inside the volume of the resonator are imaged. The rest of the wells, while being maintained at the constant temperature, humidity, and gas conditions, are outside the imaging volume.

The second component (Fig. 2b, c) is the sealed multi-well plate enclosure, with precisely controlled delivery of the necessary gas mixtures. For all experiments, the environment was maintained at a humidified state with 95% air plus 5% CO2. The system can accommodate any other gas conditions, such as 5% O2. An integrated two-channel (hot-cold) control system was implemented to maintain the temperature at 37 °C. Ducts and air directors were built into the MWIR to enable air circulation from all sides of the plate enclosure (Fig. 2b, c).

The MWIR can image the 12 middle wells of 3 single strips from a 96-well stripwell plate (Fig. 2d), while keeping the temperature constant at 37 °C ± 1 °C (Fig. 2e) and relative humidity (RH) between 70% and 100% (Fig. 2f) by controlling the flow rate of the gas mixture between 3.75 sccm (standard cubic centimeter per min) to 30 sccm. A high gas flow rate maintained high humidity, but the system became susceptible to condensation in some of the wells, which affected cell viability beyond 8 h. Reducing the gas flow rate eliminated condensation without compromising cell viability. Therefore, a gas flow rate of 3.75 sccm was used for all experiments with cells. The MWIR was tested for its ability to provide a controlled environment for oxygen imaging using PBS in the middle 12 stripwells (Fig. 2g).

For experiments with cells, two identical sets of wells were prepared by seeding HEK-293 cells into strip wells. One set was used for oxygen imaging in the MWIR, while the second set was placed inside the BOD incubator as a control. Three cell densities were used in quadruplicate: control medium with no cells (CM), 5000 cells per well (5 K), and 50,000 cells per well (50 K). The cells were arranged in the middle 12 wells as shown in Fig. 3a. Figure 3b shows the representative pO2 maps from these middle 12 wells at the beginning of the experiment. Interestingly, the wells with 50 K cells had lower oxygen tension than CM or 5 K cells, even though all wells were individually supplied with 19.95% O2 gas flow. The lower oxygen tension for wells with 50 K cells represents a higher oxygen demand. Figure 3c provides the box plots of pO2 values vs cell density at t = 0. The difference in pO2 values between CM and 5 K cells (p-value = 0.0002) and between 5 K cells and 50 K cells (p-value < 0.0001) is statistically significant, showing that pO2 maps can be used to assess cell viability.

Fig. 3: Oxygen imaging of HEK-293 cells in MWIR.
figure 3

a The pattern of cell seeding for control (CM), 5 K, and 50 K cells per well (n = 4). Each well had 150 μL of medium. b Representative pO2 maps of wells (slice #4, ~3.2 mm from the bottom of the well) in the transverse plane at t = 0 h (first imaging experiment). c Box plot of pO2 for each cell density; differences are statistically significant (p ≤ 0.001). d Cell viability measured by MTT assay after 24 h of imaging shows no significant difference between cells in the MWIR and BOD incubator at 5 K and 50 K densities (p > 0.05). e pH of the medium from wells that were in MWIR and in the incubator after 24 h. f Cell morphology after 24 h imaging from MWIR and BOD.

We performed the oxygen imaging experiments for 2 h, 4 h, 8 h, and 24 h and compared cell viability for the MWIR versus BOD using the MTT assay (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) for cellular dehydrogenase activity. The pO2 images and cell viability using MTT assay for MWIR versus BOD are similar for all time points and no significant differences were observed (Fig. 3d). The pH values (Fig. 3e) and cell morphology (Fig. 3f) of cells in the MWIR system were also consistent with the BOD incubator. These results demonstrate that EPROI imaging in the MWIR does not affect cell viability or metabolic activity during a 24 h experiment.

Next, control experiments using dead HEK-293 cells were performed to verify oxygen imaging results. The same three seeding densities and seeding patterns were used as in Fig. 3a. The cells were killed either by exposing them to 4% paraformaldehyde (PFA) for 30 min or to 70 °C for 15 min, as confirmed by morphological assessment. The pO2 images of dead cells were taken immediately after killing the cells. As expected, the difference in pO2 maps of dead cells with 0 (CM), 5000 (5 K), and 50,000 (50 K) cells per well is less pronounced but statistically significant compared to the live cells in both sets, as shown in Fig. 4a, c (live cells), b & d (dead with PFA) and c & g (dead with heat-shock). Note that the medium was also treated with heat or PFA for dead cell experiments. The MTT assay (Fig. 4g–i) shows no significant difference between the MWIR and BOD. Figure 4j–l shows trypan blue staining for assessing cell morphology and viability. Some cells are still alive after PFA treatment (Fig. 4k), which could explain the lower pO2 in the wells for this treatment. It is possible that dead cells and the treatment of the medium with PFA or heat-shock change the medium viscosity and oxygen permeability in this medium, which is reflected by the statistically significant difference in pO2 values for the dead cells. These results confirm that pO2 imaging is a sensitive tool that can differentiate between live and dead cells.

Fig. 4: Comparison of pO2 of live and dead HEK-293 cells.
figure 4

The cells were seeded in the same pattern shown in Fig. 3a. Representative pO2 maps and box plots of pO2 values as a function of cell density (a, d) live cells (same as Fig. 3a), (b, e) dead cells (killed with 4% PFA), and (c, f) dead cells (killed with heat-shock). A significant difference in the pO2 of 5 K and 50 K cells per well is observed for all three cases. d (live cells) CM vs 5 K: p = 0.0002; 5 K vs 50 K: p < 0.0001. e (PFA dead cells) CM vs 5 K: p = 0.0008; 5 K vs 50 K: p < 0.0001. f (Heat-shock) CM vs 5 K: p = 0.2598; 5 K vs 50 K: p < 0.0001. Note that the medium was also treated with PFA or heat for the dead cell experiments. gi MTT assay of live and dead cells after imaging shows no significant difference between MWIR and BOD (p > 0.05). jl Trypan blue exclusion assay for the confirmation of cell death before imaging. Some cells were still alive when killed using 4% PFA (live cells are indicated by red arrows).

Next, the sensitivity of oxygen imaging to cell density was assessed. HEK-293 cells were seeded at six different densities: 0 (CM), 10,000 (10 K), 25,000 (25 K), 50,000 (50 K), 75,000 (75 K), and 100,000 (100 K) cells per well. Figure 5a shows the arrangement of cells in the wells with two replicates of each cell density. Figure 5b shows the representative pO2 maps of these wells and confirms that pO2 mapping can differentiate between these cell densities. The box plot at t = 0 shown in Fig. 5c shows that the wells with higher cell densities had lower pO2, and the difference between different cell densities is statistically significant. Figure 5d shows how the average pO2 evolved over the course of 4 h. As cells differentiate, proliferate, become dormant, or die, their oxygen consumption rate changes, which can be reflected in pO2 values. Most interesting are the time trends of 50 K, 75 K, and 100 K cells per well. We can see that within 4 h, the mean pO2 values of wells with 50 K cells matched the pO2 of wells with 100 K cells, where cell dormancy or cell death was probably happening because of overcrowding. The pO2 for 25 K cells dropped over time, indicating increased metabolic activity in these wells. A comparison of the MTT assay for cells in the MWIR with cells in the BOD incubator shows no significant difference in cell viability except for 100 K cells per well (Fig. 5e). These experiments demonstrate the ability of EPROI to observe longitudinal cellular oxygen consumption and to differentiate between six different cell densities. One limitation of pO2 imaging technology could be when cells change their metabolism due to the environment; they may show similar pO2 as is the case for 50 K, 75 K, and 100 K cells in Fig. 5d. A similar cell morphology was observed for cells at all seeding concentrations after 4 h in the MWIR or BOD (Fig. 5f).

Fig. 5: The sensitivity of pO2 imaging.
figure 5

a Six different cell densities of HEK-293 cells (CM (0 K), 10 K, 25 K, 50 K, 75 K, and 100 K cells per well) were seeded in the given pattern (n = 2). b A representative pO2 map at t = 0 h showing visual assessment of cell metabolic activity. The image is taken ~3.2 mm above the bottom of the well. c Box plot of pO2 as a function of cell density. The mean and standard error obtained from the pO2 map for CM (144.06 ± 4.82 torr), 10 K (116.61 ± 1.45 torr), 25 K (49.58 ± 0.74 torr), 50 K (27.97 ± 0.70 torr), 75 K (30.15 ± 0.94 torr), and 100 K (32.15 ± 1.18 torr), (n = ~ 200–300 voxels per well, n = 2 per cell density). There is a statistically significant difference (p < 0.0001) in pO2 between CM vs 10 K, 10 K vs 25 K, 25 K vs 50 K, 50 K vs 75 K, and 75 K vs 100 K. d The change in pO2 as a function of time. e MTT assay after 4 h revealed no significant differences between MWIR and BOD (p > 0.05), except for 100 K cells (p ≤ 0.05). f Morphological assessment of cells in MWIR and BOD after 4 h of measurements.

The experiments in Fig. 5 show that cell viability was similar for the MWIR and BOD, even though low pO2 was observed for high-cell density wells (50 K, 75 K, 100 K). This raised a question of whether the high cell density wells in the BOD incubator were also experiencing low pO2. This was tested by the following experiment. Two identical sets of plates with four cell densities (0 K (CM), 10,000 (10 K), 50,000 (50 K), and 100,000 (100 K) cells per well, triplicate for each cell density) were made. One was placed in the MWIR and the other in the BOD incubator and, after 4 h, pO2 images were acquired for both sets. Figure 6a shows the arrangement of cells in the middle 12 wells in the plate, and Fig. 6b, c shows the representative pO2 maps of the cells in MWIR and in the BOD incubator for 4 h, respectively. Interestingly, similar to MWIR, the cells in the BOD incubator also experienced low pO2 for high cell densities. Figure 6d shows the pO2 values for MWIR and BOD. In all cases, there are no significant differences between MWIR and BOD (except 10 K, *p ≤ 0.05).

Fig. 6: Assessment of pO2 for HEK- 293 cells cultured 4 h in the BOD incubator.
figure 6

a The layout of the wells for 0 K (CM), 10 K, 50 K, and 100 K cells per well (n = 3). Representative pO2 maps at 4 h of cells in (b) MWIR and (c) BOD incubator taken ~ 3.2 mm from the bottom of the wells. d pO2 quantification between the MWIR and BOD incubator. The mean and standard error of pO2 are: MWIR- CM (148.27 ± 2.97 torr), 10 K (91.77 ± 1.16 torr), 50 K (34.9 ± 0.95 torr), 100 K (35.98 ± 0.99 torr) or BOD -CM (106.42 ± 1.00 torr), 10 K (110.69 ± 1.85 torr), 50 K (37.10 ± 1.57 torr), 100 K (25.95 ± 1.05 torr), (n = 300–750 voxels per well, n = 3 per cell density). There is a significant difference (p < 0.0001) in pO2 between different cell densities. The MWIR vs BOD p-values among wells are as follows: CM, p = 0.1938; 10 K, p = 0.0179; 50 K, p = 0.5495; 100 K, p = 0.01863; MWIR: CM vs 10 K, p = 0.0015; 10 K vs 50 K, p = 0.00014; and 50 K vs 100 K p = 0.9415.

The above experiments were performed using HEK-293 adherent cells that were attached to the bottom of the well. Next, we performed oxygen imaging of a non-adherent cell line, Jurkat cells, using MWIR. Cell densities of 0 K (CM) 5 K, and 50 K cells per well were used in the arrangement shown in Fig. 7a. Representative pO2 maps (Fig. 7b) and box plots (Fig. 7c) at t = 0 show higher pO2 values for these cells compared to HEK-293 (Fig. 3b, c), indicating different metabolic activity for Jurkat cells. The pO2 heterogeneity may be due to the Jurkat cells growing in aggregates. The MTT viability assay (Fig. 7d), pH test (Fig. 7e), and cell morphology (Fig. 7f) after 4 h of oxygen imaging show that there is no significant difference between the Jurkat cells in MWIR versus the BOD incubator, and EPROI using the MWIR is a noninvasive and robust method.

Fig. 7: Oxygen imaging of Jurkat cells using MWIR.
figure 7

a The cell seeding pattern of Jurkat cells CM, 5 K, and 50 K cells per well (n = 4). b A representative pO2 map taken at t = 0 h (~ 3.2 mm from the bottom of the well). c The box plot of pO2 values from t = 0 h. Differences in pO2 are statistically significant: CM vs 5 K, p = 0.0002 and 5 K vs 50 K, p < 0.0001. d The MTT assay after 4 h of imaging showed no significant difference in cell viability between the MWIR and BOD incubator (p > 0.05). e, f pH and cell morphology confirm that the MWIR emulates incubator-like conditions to enable in situ oxygen mapping of live cells.

A significant issue for assessing tissue-engineered medical products is that once the cells are seeded into a hydrogel scaffold, there is no way to track their metabolic activity without destroying the construct. To address this issue, we performed oxygen imaging of Jurkat cells seeded in a VitroGel, a commercially available glucose-based hydrogel. The cell seeding scheme with three densities, 0 K (CM), 5 K, and 50 K cells per well is shown in Fig. 8a. Figure 8b shows the representative pO2 maps. The box plot in Fig. 8c shows that the pO2 values were significantly different for the wells with 5 K and 50 K cells. Interestingly, the pO2 values are higher for Jurkat cells in VitroGel (Fig. 8c) compared to Jurkat cells in culture medium (Fig. 3c), suggesting that Jurkat cells have reduced metabolic activity upon cell-seeding or that cells are lost during the VitroGel encapsulation process.

Fig. 8: Oxygen imaging of Jurkat cells seeded in VitroGel.
figure 8

a The cell seeding pattern for 0 K (CM), 5 K, and 50 K cells per well (n = 4). b pO2 maps of the wells at time t = 0 (~3.2 mm from the bottom of the wells). c Box plot of pO2 vs cell density at t = 0. There is a statistically significant difference between 5 K and 50 K cells (p < 0.0001), but no significant difference between CM and 5 K cells (p = 0.6581). d Plot of pO2 versus cell density (0 K, 5 K, and 50 K cells per well, n = 4) for HEK-293, Jurkat, and Jurkat in VitroGel. Error bars are standard errors.

Figure 8d shows the normalized pO2 values for HEK-293, Jurkat, and Jurkat in VitroGel as a function of cell seeding density. The normalized pO2 values were calculated by dividing the mean pO2 of wells with cells by the mean pO2 of wells with control medium. The normalized pO2 of Jurkat cells in VitroGel is higher than Jurkat cells in medium or HEK-293 cells. HEK-293 cells have the lowest pO2 or the highest oxygen consumption rate (OCR). For Jurkat cells in VitroGel, the metabolic activity of cells is reduced by 17.3% for 50 K cells and by 4% for 5 K cells compared to Jurkat cells in medium, providing a quantitative assessment of the impact of cell seeding on viable cells.

Finally, we demonstrate nondestructive 3D imaging of cells seeded in a scaffold using EPROI. Figure 9a shows the schematic of the three cell densities (0 K (CM), 5 K and 50 K cells per well). Figure 9b shows three slices of HEK-293 cells at 3.21 mm (Bottom), 3.57 mm (Middle), and 3.92 mm (Top) from the bottom. The three slices represent a complete 3D volume of the 150 μL of medium in each well. A gradient is visible from the bottom to the top of each well, especially for the wells with 50 K HEK-293 cells, which are adhered to the bottom of the well, as shown in Fig. 9c for three selected wells denoted as A, B, and C in Fig. 9b. Figure 9d shows the change in the pO2 values for control, 5 K and 50 K cells, going from the bottom of the well to the top. Figure 9e shows the change in pO2 at t = 0 and t = 4 h for HEK-293, Jurkat, and Jurkat in VitroGel for the 50 K cells. There is a statistically significant difference between pO2 at the bottom and top of the well (p ≤ 0.05) for all conditions except Jurkat at 4 h.

Fig. 9: 3D oxygen imaging of HEK-293 cells, Jurkat cells and Jurkat cells in VitroGel.
figure 9

a The cell seeding pattern of 0 K (CM), 5 K, and 50 K cells per well (n = 4). b Three slices from the bottom to top showing the change in pO2 over depth for HEK-293 cells. c pO2 maps of three wells with 50 K cells denoted as A, B, and C in b for better visualization of vertical gradient. For each well, the central slice in the yz plane was cut and reported. The central slice was identified by choosing the slice with the highest number of pixels in the “y” direction. d The pO2 at the bottom, middle, and top of the wells at t = 0 for CM, 5 K and 50 K HEK-293 cells. e The pO2 for 50 K cells at the bottom, middle and top of the wells for HEK-293, Jurkat and Jurkat seeded in VitroGel at t = 0 and 4 h. There is a significant difference in pO2 between the bottom of the well and the top of the well for all cases except for Jurkat cells at 4 h.