In Situ Detection of Endogenous HIV Activation by Dynamic Nuclear Polarization NMR and Flow Cytometry

We demonstrate for the first time in-cell dynamic nuclear polarization (DNP) in conjunction with flow cytometry sorting to address the cellular heterogeneity of in-cell samples. Utilizing a green fluorescent protein (GFP) reporter of HIV reactivation, we correlate increased 15N resonance intensity with cytokine-driven HIV reactivation in a human cell line model of HIV latency. As few as 10% GFP+ cells could be detected by DNP nuclear magnetic resonance (NMR). The inclusion of flow cytometric sorting of GFP+ cells prior to analysis by DNP-NMR further boosted signal detection through increased cellular homogeneity with respect to GFP expression. As few as 3.6 million 15N-labeled GFP+ cells could be readily detected with DNP-NMR. Importantly, cell sorting allowed for the comparison of cytokine-treated GFP+ and GFP− cells in a batch-consistent way. This provides an avenue for normalizing NMR spectral contributions from background cellular processes following treatment with cellular modulators. We also demonstrate the remarkable stability of AMUPol (a nitroxide biradical) in Jurkat T cells and achieved in-cell enhancements of 46 with 10 mM AMUPol, providing an excellent model system for further in-cell DNP-NMR studies. This represents an important contribution to improving in-cell methods for the study of endogenously expressed proteins by DNP-NMR.


Introduction
Nuclear magnetic resonance (NMR) is an exceptional method for studying molecular structure and dynamics within complex biological systems including lipid-bilayers and membrane proteins [1], fibrils [2][3][4] and cell walls [5,6]. Viral systems [7][8][9][10] and intact cells [11][12][13][14][15][16] are particularly accessible to solid-state NMR due to its independence from molecular correlation times, and therefore particle We further improved the sensitivity of detection by employing fluorescence-activated cell sorting (FACS) to greatly enhance the homogeneity of the cell population. Furthermore, we demonstrated the potential of Jurkat T cells as a model system for in-cell studies due to slow AMUPol reduction in this cell line.

AMUPol is Stable in JLat T Cells
To develop a model system for studying endogenously produced HIV by NMR, we utilized JLat 10.6 cells which are a variant of the Jurkat T cell line [42]. The HIV genome has been integrated into the host cell genome and provides a valuable model for studying the HIV reactivation from latency [42][43][44]. These cells produce near complete virions (but which lack the nef and env genes) in an NFκB-dependent manner [42,45]. We prepared 15 N-labeled JLat cells for DNP-NMR by resuspension in 10% DMSO/90% phosphate buffered saline (PBS) and AMUPol to a final concentration of 10 mM (Figure 1). Rotor-packed cell samples were immediately frozen in liquid nitrogen to prevent radical reduction. We consistently achieved packing times of less than 2 min between the addition of the radical and the freezing in liquid nitrogen. intensities, demonstrating the feasibility of using DNP-NMR to monitor endogenous HIV reactivation. We further improved the sensitivity of detection by employing fluorescence-activated cell sorting (FACS) to greatly enhance the homogeneity of the cell population. Furthermore, we demonstrated the potential of Jurkat T cells as a model system for in-cell studies due to slow AMUPol reduction in this cell line.

AMUPol is Stable in JLat T Cells
To develop a model system for studying endogenously produced HIV by NMR, we utilized JLat 10.6 cells which are a variant of the Jurkat T cell line [42]. The HIV genome has been integrated into the host cell genome and provides a valuable model for studying the HIV reactivation from latency [42][43][44]. These cells produce near complete virions (but which lack the nef and env genes) in an NFκBdependent manner [42,45]. We prepared 15 N-labeled JLat cells for DNP-NMR by resuspension in 10% DMSO/90% phosphate buffered saline (PBS) and AMUPol to a final concentration of 10 mM ( Figure  1). Rotor-packed cell samples were immediately frozen in liquid nitrogen to prevent radical reduction. We consistently achieved packing times of less than 2 min between the addition of the radical and the freezing in liquid nitrogen. To ensure the maximum consistency between the rotors in terms of cell mass per rotor, specially designed Teflon funnels were used in which the neck of the funnel occupied 26 μL of the rotor during the packing, leaving a filling volume of 36 μL for the cell samples (Supplementary Materials SI-1), while allowing sufficient space for the insertion of the rotor drive tip.
The stability of AMUPol in the JLat cells was assessed by EPR spectroscopy at room temperature. After 30 min, we observed an 8% reduction in the integrated area of the AMUPol EPR spectrum, giving a decay rate of 0.20% min −1 ± 0.08, which is remarkably slow, indicating a relatively stable environment for DNP (Figure 2a,b). A DNP enhancement of 46 was measured at 300 MHz for the 15 N amide resonances of JLat cells prepared in the same way ( Figure 2c). We have previously reported a 30% reduction in the integrated area of the AMUPol EPR spectrum after 30 min in HEK293 cells [24]. In addition, recently reported decay rates in bacteria are even greater with a 75% reduction of a paramagnetic radical within 10 min [46]. Xenopus laevis oocytes, a common model system for in-cell structural studies, show a range of radical lifetimes from 3.8 min (TEMPOL radical) [47] to over 1 h To ensure the maximum consistency between the rotors in terms of cell mass per rotor, specially designed Teflon funnels were used in which the neck of the funnel occupied 26 µL of the rotor during the packing, leaving a filling volume of 36 µL for the cell samples (Supplementary Materials SI-1), while allowing sufficient space for the insertion of the rotor drive tip.
The stability of AMUPol in the JLat cells was assessed by EPR spectroscopy at room temperature. After 30 min, we observed an 8% reduction in the integrated area of the AMUPol EPR spectrum, giving a decay rate of 0.20% min −1 ± 0.08, which is remarkably slow, indicating a relatively stable environment for DNP (Figure 2a,b). A DNP enhancement of 46 was measured at 300 MHz for the 15 N amide resonances of JLat cells prepared in the same way ( Figure 2c). We have previously reported a 30% reduction in the integrated area of the AMUPol EPR spectrum after 30 min in HEK293 cells [24]. In addition, recently reported decay rates in bacteria are even greater with a 75% reduction of a paramagnetic radical within 10 min [46]. Xenopus laevis oocytes, a common model system for in-cell structural studies, show a range of radical lifetimes from 3.8 min (TEMPOL radical) [47] to over 1 h (MTSSL radical) [48]. It is unclear why AMUPol is so stable in JLat T cells. The net redox potential of JLat cells might be different from other cell types previously studied with AMUPol as redox processes are the greatest source of DNP radical instability [46]. Regardless of the mechanism, the data suggest that JLat T cells are a good model cell type for DNP studies. (MTSSL radical) [48]. It is unclear why AMUPol is so stable in JLat T cells. The net redox potential of JLat cells might be different from other cell types previously studied with AMUPol as redox processes are the greatest source of DNP radical instability [46]. Regardless of the mechanism, the data suggest that JLat T cells are a good model cell type for DNP studies.

Detection of Endogenously Expressed HIV Proteins by DNP-MAS NMR
We used two distinct JLat T cell lines in which the HIV genome was latently integrated and commercially available [42]. JLat 10.6 and JLat 9.2 cells differ only in the chromosomal region in which the HIV genome is integrated [43]. This results in varying HIV activation between the two cell lines upon stimulation, where the JLat 10.6 cell line is easier to activate. These cells are infected with a recombinant HIV in which the nef gene has been replaced with the reporter green fluorescent protein (GFP), in addition to a mutation in env that prevents the formation of infectious virions [42]. Therefore, JLat cell lines can safely be assayed and analyzed in a standard biosafety 2 laboratory. Both cell lines exhibit low basal HIV activation, as monitored by GFP fluorescence (Figure 3a), consistent with published results [43]. The expression of GFP and HIV genes are both driven by the same promoter and therefore results in the coexpression of GFP with Gag-pol/vpu/vif/vpr/rev and tat polypeptides. GFP fluorescence intensity has been shown to directly correlate with HIV expression levels in this cell line by Western blotting for the Gag protein [49,50]. The addition of T cell activators induces HIV and GFP production in JLat cells [38,51].
To detect endogenously expressed HIV proteins by NMR, JLat 10.6 cells were cultured for the exponential growth phase in unlabeled media. The cells were then transferred to selectively 15 Nenriched media and stimulated with the T cell activator, TNF-α for 12 h. HIV activation was assessed by flow cytometry. The stimulation with TNF-α resulted in 89% of JLat 10.6 cells upregulating GFP compared to 0.4% of the cells treated with 1% DMSO (vehicle control) (Figure 3a). The analysis of these same cells by DNP-NMR showed that the TNF-α treatment increased 15 N amide resonance intensities by 51% and amine resonances by 57%, compared to the DMSO controls ( Figure 3b). Thus, the increased signal intensity is likely attributed to endogenous HIV and GFP expression. The comparison of the amino acid sequences of all the HIV proteins produced with that of GFP indicates that under our labeling conditions, 89% of the 15 N signal intensity can be attributed to HIV proteins and 11% to GFP, assuming all other things are equal (Supplementary Materials SI-2). Due to the well documented and published overexpression of HIV-related proteins in this cell line (in excess of 85,000 pg/mL) [42] we expect a large proportion of the 15 N signal to come from HIV proteins through overexpression, although all other proteins produced during the labeling period will also be labeled.
TNF-α is known to increase protein expression in stimulated T cells, predominantly in the form of cytokine production, and might partially account for the observed increase in 15 N resonance

Detection of Endogenously Expressed HIV Proteins by DNP-MAS NMR
We used two distinct JLat T cell lines in which the HIV genome was latently integrated and commercially available [42]. JLat 10.6 and JLat 9.2 cells differ only in the chromosomal region in which the HIV genome is integrated [43]. This results in varying HIV activation between the two cell lines upon stimulation, where the JLat 10.6 cell line is easier to activate. These cells are infected with a recombinant HIV in which the nef gene has been replaced with the reporter green fluorescent protein (GFP), in addition to a mutation in env that prevents the formation of infectious virions [42]. Therefore, JLat cell lines can safely be assayed and analyzed in a standard biosafety 2 laboratory. Both cell lines exhibit low basal HIV activation, as monitored by GFP fluorescence (Figure 3a), consistent with published results [43]. The expression of GFP and HIV genes are both driven by the same promoter and therefore results in the coexpression of GFP with Gag-pol/vpu/vif/vpr/rev and tat polypeptides. GFP fluorescence intensity has been shown to directly correlate with HIV expression levels in this cell line by Western blotting for the Gag protein [49,50]. The addition of T cell activators induces HIV and GFP production in JLat cells [38,51].
To detect endogenously expressed HIV proteins by NMR, JLat 10.6 cells were cultured for the exponential growth phase in unlabeled media. The cells were then transferred to selectively 15 N-enriched media and stimulated with the T cell activator, TNF-α for 12 h. HIV activation was assessed by flow cytometry. The stimulation with TNF-α resulted in 89% of JLat 10.6 cells upregulating GFP compared to 0.4% of the cells treated with 1% DMSO (vehicle control) (Figure 3a). The analysis of these same cells by DNP-NMR showed that the TNF-α treatment increased 15 N amide resonance intensities by 51% and amine resonances by 57%, compared to the DMSO controls ( Figure 3b). Thus, the increased signal intensity is likely attributed to endogenous HIV and GFP expression. The comparison of the amino acid sequences of all the HIV proteins produced with that of GFP indicates that under our labeling conditions, 89% of the 15 N signal intensity can be attributed to HIV proteins and 11% to GFP, assuming all other things are equal (Supplementary Materials SI-2). Due to the well documented and published overexpression of HIV-related proteins in this cell line (in excess of 85,000 pg/mL) [42] we expect a large proportion of the 15 N signal to come from HIV proteins through overexpression, although all other proteins produced during the labeling period will also be labeled.
intensity following the stimulation of JLat 10.6 cells compared to the stimulated Jurkat cells ( Figure  3c). Thus, the increase in 15 N resonance intensity cannot be accounted for by cytokine production, which was expected as cytokines are secreted proteins and they are washed away in the sample preparation process. This is a clear demonstration of endogenous HIV and GFP detection by DNP solid-state NMR. Furthermore, the proportion of GFP+ cells also correlates reasonably well with increases in 15 N signal intensity. Increases in the 15 N resonance intensity due to the bulk activation of JLat 10.6 cells were readily detectable. However, the proportion of latently infected T cells in vivo is typically very low [52]. In order to assess the feasibility of using MAS-DNP with flow cytometry to study the clinically relevant HIV latency, we used the JLat 9.2 cell line. Latent HIV activation is much more difficult to achieve in JLat 9.2 cells compared to JLat 10.6. Therefore, the characterization of HIV activation in JLat 9.2 cells is important as it more closely mimics the behavior of HIV reservoirs in vivo, which are difficult to activate [32].
Due to the inherently low levels of HIV activation in JLat 9.2 cells, we determined the optimal stimulation period for GFP expression in JLat 9.2 cells over 24 h. We observed only 2% GFP+ JLat 9.2 cells at 12 h post stimulation, which increased to 12% after 24 h (Supplementary Materials SI-3). In contrast, 46% of the stimulated JLat 10.6 cells were GFP+ at 12 h, increasing to 68% at 24 h (Supplementary Materials SI-3). Thus, not only does a lower proportion of the JLat 9.2 population become activated, but the time course of activation is also slower. As a result, we chose to stimulate the JLat 9.2 cells for 24 h prior to the DNP-NMR analysis. We did not extend the activation period beyond 24 h in order to minimize the incorporation of the 15 N-enriched amino acids into proteins TNF-α is known to increase protein expression in stimulated T cells, predominantly in the form of cytokine production, and might partially account for the observed increase in 15 N resonance intensity [51]. We addressed this possibility by comparing the TNF-α-stimulated JLat 10.6 cells with stimulated Jurkat T cells, which do not contain any HIV DNA but are equally responsive to TNFα stimulation. The comparison of 15 N amide resonance intensities showed a 54% increase in signal intensity following the stimulation of JLat 10.6 cells compared to the stimulated Jurkat cells (Figure 3c). Thus, the increase in 15 N resonance intensity cannot be accounted for by cytokine production, which was expected as cytokines are secreted proteins and they are washed away in the sample preparation process. This is a clear demonstration of endogenous HIV and GFP detection by DNP solid-state NMR. Furthermore, the proportion of GFP+ cells also correlates reasonably well with increases in 15 N signal intensity.
Increases in the 15 N resonance intensity due to the bulk activation of JLat 10.6 cells were readily detectable. However, the proportion of latently infected T cells in vivo is typically very low [52]. In order to assess the feasibility of using MAS-DNP with flow cytometry to study the clinically relevant HIV latency, we used the JLat 9.2 cell line. Latent HIV activation is much more difficult to achieve in JLat 9.2 cells compared to JLat 10.6. Therefore, the characterization of HIV activation in JLat 9.2 cells is important as it more closely mimics the behavior of HIV reservoirs in vivo, which are difficult to activate [32].
Due to the inherently low levels of HIV activation in JLat 9.2 cells, we determined the optimal stimulation period for GFP expression in JLat 9.2 cells over 24 h. We observed only 2% GFP+ JLat 9.2 cells at 12 h post stimulation, which increased to 12% after 24 h (Supplementary Materials SI-3). In contrast, 46% of the stimulated JLat 10.6 cells were GFP+ at 12 h, increasing to 68% at 24 h (Supplementary Materials SI-3). Thus, not only does a lower proportion of the JLat 9.2 population become activated, but the time course of activation is also slower. As a result, we chose to stimulate the JLat 9.2 cells for 24 h prior to the DNP-NMR analysis. We did not extend the activation period beyond 24 h in order to minimize the incorporation of the 15 N-enriched amino acids into proteins other than GFP or HIV. We also note the variability in the percentage of the JLat 10.6 GFP+ cells between experiments. This highlights the importance of combining flow cytometry and other reporting techniques with in-cell NMR techniques for the accurate assessment of NMR spectra.
We recorded the 15 N cross polarization magic angle spinning (CPMAS)-DNP spectra of the JLat 9.2 cells stimulated with TNF-α for 24 h and observed a 15% and 17% increase in the intensity of the amide and amine resonances, respectively (Figure 4b). This correlates reasonably well with the flow cytometry data, where 10% of the stimulated JLat 9.2 cells were GFP+ (Figure 4a). Furthermore, the observation of this increase demonstrates the excellent sensitivity of the DNP-NMR to detect small populations of latently activated JLat cells within a heterogenous sample. However, the correlation between the percent GFP+ cells and the 15 N signal intensity is therefore not accurate enough to allow the use of 15 N amide resonance intensity as a proxy for the GFP/HIV expression alone. Additional factors must therefore influence the 15 N signal intensity in the activated JLat cells. other than GFP or HIV. We also note the variability in the percentage of the JLat 10.6 GFP+ cells between experiments. This highlights the importance of combining flow cytometry and other reporting techniques with in-cell NMR techniques for the accurate assessment of NMR spectra. We recorded the 15 N cross polarization magic angle spinning (CPMAS)-DNP spectra of the JLat 9.2 cells stimulated with TNF-α for 24 h and observed a 15% and 17% increase in the intensity of the amide and amine resonances, respectively (Figure 4b). This correlates reasonably well with the flow cytometry data, where 10% of the stimulated JLat 9.2 cells were GFP+ (Figure 4a). Furthermore, the observation of this increase demonstrates the excellent sensitivity of the DNP-NMR to detect small populations of latently activated JLat cells within a heterogenous sample. However, the correlation between the percent GFP+ cells and the 15 N signal intensity is therefore not accurate enough to allow the use of 15 N amide resonance intensity as a proxy for the GFP/HIV expression alone. Additional factors must therefore influence the 15 N signal intensity in the activated JLat cells.

Flow Cytometric Sorting Improves Cell Homogeneity
The contribution of background T cell activation to 15 N resonance intensity is difficult to ascertain when comparing non-stimulated with stimulated JLat cells. To do this requires the separation of GFP+ TNF-α stimulated JLat cells from their GFP− counterparts. We achieved this by using flow cytometry to sort GFP+ cells and GFP− cells from a heterogenous population of TNF-αstimulated JLat 9.2 cells. Due to the limited sorting capacity of flow cytometers, only 3.6 × 10 6 GFP+ and GFP− cells could be collected. However, we could still resolve a 34% increase in 15 N-amide resonance intensity in the GFP-enriched samples compared to GFP− samples ( Figure 5). This is a further demonstration that the increased signal intensity is due to the expression of HIV and associated GFP as the incorporation of flow cytometric sorting normalizes for any additional protein production due to stimulation. Flow cytometric sorting, in general, yields good cellular homogeneity but at the expense of total cellular mass. Alternative methods for bulk cellular purification such as antibody-based sorting could be employed and will be an important step for the analysis of primary cells and clinical samples where the occurrence of virally infected cells is very low.
The difference in the signal intensity between GFP+ and GFP− was expected to be much larger than what was observed. Given the 15% increase in the 15 N signal intensity of the stimulated JLat 9.2 cells compared to the unstimulated cells, which represents nearly a 1% increase in the 15 N signal intensity per 1% GFP expression, we expected close to a 100% increase in the 15 N signal intensity of the sorted GFP+ JLat 9.2 cells compared to the sorted GFP− JLat 9.2 cells. This may be due to the upregulated basal protein production by TNF-α stimulation which partly cancels out the enhancements we would expect to see if the increase in protein content could be solely attributed to

Flow Cytometric Sorting Improves Cell Homogeneity
The contribution of background T cell activation to 15 N resonance intensity is difficult to ascertain when comparing non-stimulated with stimulated JLat cells. To do this requires the separation of GFP+ TNF-α stimulated JLat cells from their GFP− counterparts. We achieved this by using flow cytometry to sort GFP+ cells and GFP− cells from a heterogenous population of TNF-α-stimulated JLat 9.2 cells. Due to the limited sorting capacity of flow cytometers, only 3.6 × 10 6 GFP+ and GFP− cells could be collected. However, we could still resolve a 34% increase in 15 N-amide resonance intensity in the GFP-enriched samples compared to GFP− samples ( Figure 5). This is a further demonstration that the increased signal intensity is due to the expression of HIV and associated GFP as the incorporation of flow cytometric sorting normalizes for any additional protein production due to stimulation. Flow cytometric sorting, in general, yields good cellular homogeneity but at the expense of total cellular mass. Alternative methods for bulk cellular purification such as antibody-based sorting could be employed and will be an important step for the analysis of primary cells and clinical samples where the occurrence of virally infected cells is very low.
is frozen out and can manifest as line broadening in DNP spectra, greatly reducing sensitivity and resolution. In addition, in-cell experiments further challenge NMR sensitivity due to the low concentration of spins of interest. Further gains in DNP and NMR sensitivity through technological developments will significantly improve the implementation of multi-dimensional NMR analysis to proteins at physiological concentrations (nano and micromolar concentrations) and in the heterogeneous environment of intact cells. The combination of cell sorting and purification strategies provides a means for addressing the challenge of cellular heterogeneity for in-cell NMR studies to improve sensitivity and sample normalization. We hope to foster the integration of new developments in NMR methods and instrumentation with rigorous and well established biological techniques. Further development of our integrated approach could open up more biologically relevant systems to the in situ study at the atomic level.

Cell Lines and Cell Culture
JLat 10.6 and JLat 9.2 cells were obtained from the NIH AIDS Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases (catalog #1340), NIH.: J-Lat Full Length Cells from Dr. Eric Verdin [43]. This cell line has been previously described [42,43]. Briefly, the Jurkat T cells were infected with HIV-R7/E-/GFP retroviral vector and displayed no virus production under The difference in the signal intensity between GFP+ and GFP− was expected to be much larger than what was observed. Given the 15% increase in the 15 N signal intensity of the stimulated JLat 9.2 cells compared to the unstimulated cells, which represents nearly a 1% increase in the 15 N signal intensity per 1% GFP expression, we expected close to a 100% increase in the 15 N signal intensity of the sorted GFP+ JLat 9.2 cells compared to the sorted GFP− JLat 9.2 cells. This may be due to the upregulated basal protein production by TNF-α stimulation which partly cancels out the enhancements we would expect to see if the increase in protein content could be solely attributed to HIV and GFP. Furthermore, it is not clear how much protein needs to be produced to result in a 1% increase in signal intensity in this endogenous system.
The JLat MAS-DNP system is a promising new tool for the study of HIV virion production and latency reversal at the molecular level. The work presented here is the first demonstration of the DNP-enhanced solid-state NMR detection of endogenously produced HIV virions. The detection of increased amide resonance intensities by the MAS-DNP of HIV reactivated JLat cells provides an excellent model for future multi-dimensional studies of HIV virions within intact human cells. The benefits of multi-dimensional spectroscopy are the significant gains in data resolution through reduced spectral overlap with every additional dimension. This is a significant caveat for DNP-solid-state NMR as at cryogenic temperatures, molecular motion that is present at ambient temperatures is frozen out and can manifest as line broadening in DNP spectra, greatly reducing sensitivity and resolution. In addition, in-cell experiments further challenge NMR sensitivity due to the low concentration of spins of interest. Further gains in DNP and NMR sensitivity through technological developments will significantly improve the implementation of multi-dimensional NMR analysis to proteins at physiological concentrations (nano and micromolar concentrations) and in the heterogeneous environment of intact cells.
The combination of cell sorting and purification strategies provides a means for addressing the challenge of cellular heterogeneity for in-cell NMR studies to improve sensitivity and sample normalization. We hope to foster the integration of new developments in NMR methods and instrumentation with rigorous and well established biological techniques. Further development of our integrated approach could open up more biologically relevant systems to the in situ study at the atomic level.

Sample Preparation
After stimulation, the JLat cells were prepared for NMR by pelleting 36 × 10 6 cells at 170× g for 5 min and then washing with 10 mL 1 × PBS to remove the cell culture media, unused isotopes and secreted proteins. The cell pellet (30 µL volume) was then resuspended in 30 µL of DNP solution containing 1 × PBS, 10% DMSO (v/v) and 20 mM AMUPol (Cgiving a final AMUPol concentration of 10 mM. The cell suspension was funneled directly into a Y 2 O 3 -stabilized ZrO 2 NMR rotor without further washing at 800× g for 20 s using custom-made Teflon funnels and frozen immediately in liquid nitrogen to halt the cellular reduction of AMUPol. The total packing time, defined as the time from the addition of AMUPol (to the cell pellet) to the freezing of the sample in liquid nitrogen, was less than 2 min.
Samples for the flow cytometry were prepared by taking 100 µL aliquots of isotopically-labeled JLat cells (~100,000 cells) and then fixing in 2% formaldehyde overnight at 4 • C prior to flow cytometric analysis.

NMR Data Collection
All experiments were performed on a 7.05 T magnet at 95 K with Larmor frequencies of 1 [54] to determine intensities.

EPR Analysis
JLat cells (12 × 10 6 cells) were prepared as described previously. Cell pellets (10 µL) were resuspended in 10 µL of 1 × PBS with 10% DMSO and 20 mM AMUPol and centrifuged into EPR tubes at 800× g for 20 s. Excess supernatant was removed and the sample was frozen immediately in liquid nitrogen with a total packing time less than 3 min, to halt any reduction of AMUPol prior to EPR analysis. For analysis, the capillary tube was thawed to room temperature then immediately inserted into a 5 mm NMR tube to fit the cavity of a 9.4 GHz EPR spectrometer (JEOL-JES-FA 100). Continuous wave EPR spectra was acquired at 9.4 GHz with 2 mW microwave power and a sweep time of 30 s (0.1 s time constant) using a modulation frequency of 100 kHz over 0.1 mT and an amplitude of 50. All spectra were acquired at room temperature for 30 min.

Flow Cytometry Analysis
GFP fluorescence of fixed, isotopically labeled JLat 10.6 and JLat 9.2 cells was analyzed using a FACScan flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). Live cells were gated based on forward (FSC) and side scatter (SSC) profiles. The data were analyzed using Flowing Software v2.5.1 (Perttu Terho). For the GFP time course analyses, 1 × 10 6 cells were plated at 3 × 10 6 /mL and treated with 10 ng/mL TNF-α. The samples were collected at 0, 12, and 24 h and analyzed by flow cytometry.
For the cell-sorting experiments, 100 × 10 6 JLat 9.2 cells were plated at 3 × 10 6 /mL in 15 N-(G,L,K,Q,E,T,V,P)-cRPMI media and treated with 10 ng/mL TNF-α for 24 h. Cells were then sorted on a FACSAria flow cytometer (BD Biosciences). Live cells were gated based on forward and side scatter profiles as depicted in Figure 5. After sorting, the cells were counted, and 3.6 × 10 6 GFP+ and GFP− cells were packed into 3.2 mm Y 2 O 3 -stabilized ZrO 2 rotors as described above.