Redox Index and Capacity Analysis (RICA) reveals NAD biology changes in T cell responses in vitro and ex vivo

Abstract

The importance of NAD metabolism in T cell differentiation and function has gained attention in recent years. However, technical limitations impede the specific interrogation of NAD dynamics in living immune cells. In this report, we present the redox index and capacity analysis (RICA) assay, a novel technique for measuring mitochondrial NAD content and redox balance. The RICA assay is a flow cytometry-based technique that uses NADH autofluorescence and mitochondrial inhibitors to assess NAD within specific phenotypic subsets of immune cells. We validated this technique using metabolic modulators and used it to examine murine CD8 T cell subsets in vitro and ex vivo. Consistent with previous findings, we observed that metabolically active, effector-like cells had a higher mitochondrial NADH:NAD⁺ ratio than quiescent cells. We discovered that cells with greater differentiation potential often possessed a larger pool of mitochondrial NAD than terminally differentiated cells in vitro and in a vaccinia viral immunization model. Mitochondrial NAD content fluctuated considerably in response to fuel availability and metabolic modulators, even within short treatment timeframes. Finally, tumor localization and differentiation status dramatically affected the mitochondrial NAD pool, but not the NADH:NAD⁺ ratio, of adoptively transferred CD8 T cells in a B16 melanoma model. This study establishes a tool for evaluating mitochondrial NAD biology in living immune cells at a greater level of detail than previously possible. It also highlights dynamic changes in mitochondrial NAD pool size as an important and novel element of CD8 T cell biology.

Introduction

Metabolism is a critical aspect of immune cell function, particularly that of CD8 T cells. Factors such as fuel usage, oxygen availability, and mitochondrial dynamics underly CD8 T cell activation and differentiation, their protective roles in various body compartments, and their dysfunction in chronic infections and cancer.^1–3 Another crucial element of T cell metabolism is redox biochemistry. Redox reactions underly almost every part of cellular metabolism, the redox coenzyme NAD being particularly important. NAD interconverts between its two forms, the oxidized form NAD⁺ and the reduced form NADH, by gaining or donating two electrons.⁴ Interconversion between NAD⁺ and NADH is critical in glycolysis, lactate formation, the TCA cycle, and ATP production via the mitochondrial electron transport chain (ETC).⁴ NAD reactions are also crucial in the catabolism of other fuels, such as amino acids and fatty acids, prior to their entry into the TCA cycle.^5–7 Therefore, all aspects of cellular energy production rely on interconversion between NAD⁺ and NADH and the relative amounts of these two forms. Additionally, NAD⁺ itself is a substrate for many important enzymes, including DNA repair enzymes and epigenetic modifiers.⁷ This means that the total pool of available NAD, as well as the balance between NAD⁺ and NADH, is critical to T cell function.

Despite the importance of NAD in T cell biology, NAD dynamics in immune cells are challenging to study, especially within specific compartments such as the mitochondria. Some approaches, including enzymatic biochemical assays and isotope tracing/mass spectrometry, require homogenization of large numbers of cells. This means they cannot be used in living samples.^8,9 Genetically encoded biosensors for NAD⁺/NADH bypass these issues, but are not practical for all applications, including ex vivo analysis of immune cell subsets.¹⁰ More recently, techniques have been developed that take advantage of NADH autofluorescence to measure redox state in living cells.¹¹ Many of these are microscopy-based, meaning they are low-throughput and rely on imaging equipment and expertise that are not readily accessible to many immunology labs. Additionally, most of these autofluorescence-based redox measurements are not specific to NAD, limiting the biological inferences that can be made from them.^11–13

Here, we present the mitochondrial NAD redox index and capacity analysis (RICA) assay, a novel autofluorescence-based technique to measure the total NAD pool and ratio of NADH:NAD⁺ in the mitochondria of living immune cells. The RICA assay employs flow cytometry to measure the NADH autofluorescence signal of cells at baseline and after treatment with two mitochondrial inhibitors: rotenone, which blocks complex I of the ETC from oxidizing NADH, and FCCP, which decouples electron transport from the mitochondrial proton gradient and maximally oxidizes mitochondrial NADH. This inducible range provides a more complete, specific picture of mitochondrial NAD biology than other techniques and allows more accurate comparisons across cell types. Additionally, surface marker staining can be used to interrogate mitochondrial NAD within heterogeneous cell populations. Because the RICA assay is flow cytometry-based, it is high-throughput, accessible to labs with standard immunology equipment, and easy to integrate into existing sample preparation workflows. It can be used with samples from both in vitro and ex vivo contexts. Here, we validated the RICA assay and used it to analyze the mitochondrial NAD biology of murine CD8 T cells in culture, in a viral immunization model, and in the context of B16 melanoma.

We found that the mitochondrial NADH:NAD⁺ ratio (defined here as the mitochondrial NAD redox index) and the total pool of mitochondrial NAD (defined as the mitochondrial NAD capacity) differed substantially between T cell subsets. In vitro, effector-like cells exhibited a higher mitochondrial NAD redox index than other subsets, while less differentiated cells tended to have a larger capacity. In ex vivo contexts, we found that the differences in redox index were less pronounced, while the capacity varied dramatically between subsets. This was especially true among tumor-infiltrating lymphocytes: we found that activated, precursor exhausted cells had an expanded mitochondrial NAD capacity that diminished sharply as cells became terminally exhausted. These results provide a deeper, more nuanced understanding of mitochondrial NAD biology in CD8 T cells than has been previously described and highlight the critical nature of NAD dynamics in T cell function. These insights gained using the RICA assay can inform strategies for overcoming T cell dysfunction in disease contexts.

Materials & Methods

Mice

Female C57BL/6N mice were purchased from Charles River Laboratories (027). CD45.1 mice (002014) and OT-I mice (003831) were originally purchased from The Jackson Laboratory and were backcrossed onto the Charles River C57BL/6N background for multiple generations. CD45.1 OT-I mice were crossed and bred in-house. For the comparison of C57BL/6N with C57BL/6J mice, female C57BL/6J mice were purchased from The Jackson Laboratory (000664) and were co-housed with C57BL/6N mice for at least one week prior to use in experiments. All mice were housed in the Center for Comparative Medicine and Research at Dartmouth-Hitchcock Medical Center. All animal experiments were approved by the Dartmouth College Institutional Animal Care and Use Committee.

RICA assay

Unless otherwise noted, all staining, washing, and flow cytometry steps were performed in RICA buffer (1x PBS + 2% FBS + 1 mM EDTA + 25 mM HEPES). Cells were plated in a 96-well U-bottom plate and stained on ice. After two washes in cold RICA buffer, cells were resuspended in 200 μL warm RICA buffer and incubated at 37 °C in a humidified incubator at 5% CO₂ for 10 minutes. “Basal” measurements were obtained by recording 25 μL each well using a Bio-Rad ZE5 flow cytometer with the plate loader at 37 °C and the “Return Sample” function off. The ZE5 DAPI channel was used to measure NADH autofluorescence. Additional fluorescence channels were used for stained surface markers and viability and were chosen to minimize overlap with each other and with NADH. After “Basal” measurements, remaining cells in each well were split 2 × 50 μL into two new sets of wells. 50 μL of 1 μM rotenone (Cayman Chemical 13995) in warm RICA buffer were added to the first set of wells (“Maximum”), and cells were incubated at 37 °C for 10 minutes. 50 μL each “Maximum” well were recorded on the ZE5. 50 μL of 2 μM FCCP (Cayman Chemical 15218) in warm RICA buffer were added to the second set of wells (“Minimum”), and cells were incubated at 37 °C for 10 minutes. 50 μL each “Minimum” well were recorded on the ZE5. Data were analyzed in FlowJo as described in the “Results” section. For some experiments, flow cytometry was performed in modified T cell medium (TCM) consisting of phenol red-free RPMI 1640 + 10% FBS, 1x MEM non-essential amino acids, 1 mM sodium pyruvate, and 25 mM HEPES. When TCM was used, the FCCP concentration was doubled (to 4 μM) to compensate for the presence of extra FBS. For the media comparisons experiment, flow cytometry for some samples was performed in modified Seahorse Mito Stress Assay medium consisting of Seahorse XF Base Medium (Agilent 103335–100) + 2 mM glutamine, 10 mM glucose, 1.2 mM pyruvate, and 25 mM HEPES +/− 10% FBS. The higher FCCP concentration (4 μM) was used for all groups in the media comparisons experiment.

Cell culture

T cells and splenocytes were cultured in T cell medium (TCM) consisting of RPMI 1640 + 10% FBS, 1x MEM non-essential amino acids, 1 mM sodium pyruvate, 1 mM HEPES, and 50 μM β-mercaptoethanol. B16-ova cells were cultured in B16 medium consisting of RPMI 1640 + 10% FBS + 1x MEM non-essential amino acids. BHK cells were cultured in DMEM + 10% FBS + 100 U/mL penicillin streptomycin.

Treatment with metabolic modulators

Naïve CD8 T cells were isolated from pooled spleens and lymph nodes of female C57BL/6N mice age 6–16 weeks old using the STEMCELL EasySep Mouse Naïve CD8+ T Cell Isolation Kit (19858A). Cells were resuspended at a density of 0.5×10⁶ cells/mL in TCM + 25 IU/mL recombinant human IL-2 (rhIL-2) (NCI Biological Research Branch Preclinical Biologics Repository) + 5 μg/mL anti-CD28 (BioXCell BE0015–1). Cells were plated on 24-well plates that had been previously coated with 10 μg/mL anti-CD3 (BioXCell BE0001–1). Cells were incubated at 37 °C for 48 hours before being used for experiments.

For the oligomycin treatment experiment, 200,000 recently activated CD8s per well were plated in a 96-well U-bottom plate. Cells were resuspended in TCM +/− 1 μM oligomycin (Cayman Chemical 11342) and incubated at 37 °C for 1 hour. Cells were washed with 200 μL cold RICA buffer, then Fc blocked (anti-CD16/CD32, BioXCell BE0307) on ice for 10 minutes. Cells were stained on ice for 20 minutes with Live/Dead Fixable NearIR Dead Cell Stain (Invitrogen L34976, 1:1500), CD8β PE (BioLegend 126608, 1:400), CD62L FITC (BioLegend 104406, 1:200), and CD44 APC (BioLegend 103012, 1:400). Cells were washed twice with cold RICA buffer, then resuspended in 200 μL warm buffer, incubated at 37 °C for 10 minutes, and analyzed using the RICA assay.

For the pyruvate supplementation experiment, 150,000 recently activated CD8s per well were plated in a 96-well U-bottom plate. Cells were washed with cold RICA buffer, Fc blocked, and stained as described above. Cells were washed twice with cold RICA buffer, then resuspended in 200 μL warm buffer +/− 1 mM sodium pyruvate (Corning 25–000-CI) and +/− 2 μM UK5099 (Cayman Chemical 16980). Cells were incubated at 37 °C for 10 minutes, then analyzed using the RICA assay. Pyruvate and/or UK5099 concentrations were kept consistent during the rotenone and FCCP additions.

For the NMN and FK866 treatment experiment, 300,000 recently activated CD8s per well were plated in a 96-well U-bottom plate. Cells were resuspended in 200 μL TCM + 25 IU/mL rhIL-2 +/− 1 mM NMN (MedChem Express HY-F004) and +/− 10 nM FK866 (Cayman Chemical 13287) and incubated at 37 °C for 5 hours. Cells were washed with cold RICA buffer, Fc blocked, and stained as described above. Cells were washed twice with cold RICA buffer, then resuspended in 200 μL warm buffer, incubated at 37 °C for 10 minutes, and analyzed using the RICA assay.

For the G6PD inhibition experiment, 200,000 recently activated CD8s or naïve splenocytes per well were plated in a 96-well U-bottom plate. Naïve splenocytes were obtained by harvesting a sex- and age-matched C57BL/6N spleen and mashing it through sterile nylon mesh (Bally Ribbon Mills 2672) in TCM using a syringe plunger. Splenocytes were spun down, resuspended in 3 mL RBC lysis buffer (155 mM NH₄Cl+ 10 mM KHCO₃ + 0.0005% phenol red), and incubated at 37 °C for 3 minutes, then washed with TCM. Cells were resuspended in 200 μL TCM +/− 25 μM or 40 μM G6PDi-1 (MedChem Express HY-W107464) and incubated at 37 °C for 2 hours. For the RICA assay, cells were washed with cold RICA buffer, Fc blocked, and stained as described above. Cells were washed twice with cold RICA buffer, then resuspended in 200 μL warm buffer, incubated at 37 °C for 10 minutes, and analyzed using the RICA assay. G6PDi-1 concentrations were kept consistent during all washing steps and RICA assay measurements. For MitoSOX Red staining, cells were washed with warm RICA buffer, then Fc blocked and stained with Live/Dead Fixable NearIR Dead Cell Stain (1:1500) and CD8α BV421 (BioLegend 100737, 1:400) in warm RICA buffer at 37 °C for 20 minutes. Cells were then stained with 5 μM MitoSOX Red (Invitrogen M36008) in warm HBSS containing calcium and magnesium at 37 °C for 10 minutes. Cells were washed twice with warm RICA buffer and analyzed using the ZE5 flow cytometer.

In vitro OT-I experiments

Spleens from male or female OT-I mice age 6–16 weeks old were harvested and mashed through sterile nylon mesh in TCM using a syringe plunger. Splenocytes were spun down, resuspended in 3 mL RBC lysis buffer, and incubated at 37 °C for 3 minutes. Splenocytes were washed with TCM and plated in 24-well plates at a density of 0.5×10⁶ cells/mL + 1 μg/mL SIINFEKL peptide (New England Peptide P621046). Cells were incubated at 37 °C for 48 hours.

For IL-2 and IL-15 OT-I cultures, cells were collected after the 48-hour activation period, resuspended in TCM, and underlaid with Ficoll-Paque Premium (Sigma-Aldrich GE17–5442-02). Cells were centrifuged at 400 × g for 25 minutes at 18 °C with low acceleration and low brake to remove debris. Live cells were collected from the TCM-Ficoll interface, washed with TCM, and resuspended in TCM at a density of 0.5×10⁶ cells/mL. For IL-2 cultures, 25 IU/mL rhIL-2 was added to the medium. For IL-15 cultures, 20 μg/mL recombinant mouse IL-15 (rmIL-15) (BioLegend 566302) was added to the medium. Cells were collected, washed, and re-seeded at 0.5×10⁶ cells/mL with the appropriate cytokine every 2 days for the following 5 days.

For T_scm cultures, OT-I splenocytes were prepared and plated with SIINFEKL as above, but 50 IU/mL rhIL-2 was added to the medium at the start of the activation period. Three hours later, CAL-101 (Cayman Chemical 15279) was added to the medium at a final concentration of 30 μM. T_scm cells were activated for 3 days and were not subjected to a Ficoll gradient. After the 3-day activation period, cells were collected, washed, and re-seeded at 0.5×10⁶ cells/mL with 50 IU/mL rhIL-2 and 30 μM CAL-101. Three days later, cells were re-seeded again under the same conditions. T_scm OT-Is were analyzed alongside IL-2, IL-15, and naïve OT-Is on day 7.

On the day of the RICA assay, naïve OT-Is were obtained by harvesting a sex-matched OT-I spleen, mashing it, and performing RBC lysis as described above. Splenocytes were resuspended in TCM and kept on ice while the other cells were prepared.

For the RICA assay, 500,000 cells per well were plated in a 96-well U-bottom plate. Cells were washed with cold RICA buffer, Fc blocked, and stained as described above. Cells were washed twice with cold RICA buffer, then resuspended in 200 μL warm buffer, incubated at 37 °C for 10 minutes, and analyzed using the RICA assay.

For TCF-1 staining, T_scm cells were Fc blocked on ice for 10 minutes. Cells were stained with Live/Dead Fixable NearIR Dead Cell Stain (1:1500), CD8α BV421 (1:400), CD62L FITC (1:200), and CD44 PE (BioLegend 103008, 1:200) on ice for 20 minutes. Cells were washed with PBS, then fixed on ice for 30 minutes with Fixation/Permeabilization Buffer from the eBioscience Foxp3 / Transcription Factor Staining Buffer Set (71–5775-40). Cells were washed with eBioscience Permeabilization Buffer and stained with TCF-1 AF647 (BD Pharmingen 566693, 1:50) in Perm Buffer on ice for 30 minutes. Cells were washed twice with Perm Buffer, then resuspended in stain wash buffer (SWB) (1x PBS + 2% normal calf serum + 0.1% sodium azide) and analyzed using a Beckman Coulter CytoFLEX S flow cytometer.

Mitochondrial enrichment and NADH/NAD⁺ quantification

For mitochondrial isolations and NADH/NAD⁺ quantification, IL-2 OT-I cultures were prepared as described above, but cells were not subjected to a Ficoll gradient. Cells were collected on day 6 of culture. For the UK5099 and pyruvate experiment, 1×10⁸ cells each treatment group were resuspended in 80 mL warm RICA buffer +/− 1 mM sodium pyruvate or 2 μM UK5099. Cells were incubated at 37 °C for 10 minutes, then spun down and washed with ice-cold PBS. Cells were resuspended in 500 μL mitochondrial extraction buffer (MEB) (10 mM Tris-Cl, pH 7.4 + 220 mM mannitol + 68 mM sucrose + 50 mM KCl + 2 mM MgCl₂ + 1 mM DTT + 1x Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Scientific PI78441)) +/− sodium pyruvate or UK5099 and incubated on ice for 30 minutes. Cells were transferred to frosted glass Dounce homogenizers on ice and homogenized with a tight-fitting pestle for 50 strokes. Homogenates were collected in microcentrifuge tubes on ice. Homogenizers were washed with MEB +/− sodium pyruvate or UK5099, and washes were collected and pooled with homogenates to a final volume of 1.5 mL each. Homogenates were centrifuged at 1,300 × g for 5 minutes at 4 °C to remove unlysed cells, nuclei, and heavy membranes (cell membrane and rough ER). Supernatants were transferred to new tubes and centrifuged at 14,000 × g for 15 minutes at 4 °C to pellet mitochondria. Supernatants were discarded, and mitochondria were washed by adding 500 μL MEB +/− sodium pyruvate or UK5099 without resuspending the pellet, inverting the position of the pellet in the microcentrifuge, and centrifuging at 14,000 × g for 5 minutes at 4 °C. Isolated mitochondria were lysed in NAD/NADH Lysis Buffer from the Abcam NAD/NADH Fluorometric Assay Kit (ab176723) at 37 °C for 15 minutes, then centrifuged at 2,000 × g for 5 minutes at 4 °C. NADH and total NAD⁺/NADH in mitochondrial extracts were quantified following the kit protocol; fluorescence was measured using a SpectraMax i3x Microplate Reader. The purity of isolated mitochondria was confirmed by western blot. Separate samples of the same cell cultures were analyzed using the RICA assay as described above. For the FK866 experiment, the same protocol was used, except that cells were incubated in TCM + 25 IU/mL rhIL-2 +/− 10 nM FK866 at 37 °C for 5 hours prior to analysis, and the same concentration of drug was maintained during mitochondrial isolations.

In vitro CD4 experiments

Naïve CD4 T cells were isolated from pooled spleens and lymph nodes of female C57BL/6N mice age 6–10 weeks old using the STEMCELL EasySep Mouse Naïve CD4+ T Cell Isolation Kit (19765A). For Th0 skewing, cells were resuspended at a density of 1×10⁶/mL in TCM + 30 IU/mL rhIL-2. For Th1 skewing, cells were resuspended at a density of 1×10⁶/mL in TCM + 30 IU/mL rhIL-2 + 15 ng/mL recombinant mouse IL-12 (rmIL-12) (BioLegend 577002) + 5 μg/mL anti-IL-4 (BioXCell BE0045). For Th2 skewing, cells were resuspended at a density of 0.5×10⁶/mL in TCM + 30 IU/mL rhIL-2 + 10 ng/mL recombinant mouse IL-4 (rmIL-4) (BioLegend 574302) + 5 μg/mL anti-IFNγ (BioXCell BE0055). For iTreg skewing, cells were resuspended at a density of 1×10⁶/mL in TCM + 100 IU/mL rhIL-2 + 15 ng/mL recombinant human TGFβ (rhTGFβ) (BioLegend 781802) + 5 μg/mL anti-IL-4 + 5 μg/mL anti-IFNγ. Cells were plated on a 96-well U-bottom plate that had been previously coated with 1 μg/mL anti-CD3 and 1 μg/mL anti-CD28. Cells were incubated at 37 °C for 48 hours, then collected and counted. TCM was added to achieve the same density at which the cells were originally seeded, and cells were re-seeded in a 24-well plate. rhIL-2 was added to all cells at the same concentration it was originally added, and rmIL-4 was similarly added to Th2s. Cells were re-seeded in the same way the following day, then analyzed on day 4.

On the day of the RICA assay, naïve CD4 T cells were obtained by harvesting a female C57BL/6N spleen, mashing it, and performing RBC lysis as described above. Splenocytes were resuspended in TCM and kept on ice while the other cells were prepared.

For the RICA assay, 250,000 cells per well were plated in a 96-well U-bottom plate. Cells were Fc blocked, stained, washed, and analyzed in phenol red-free TCM supplemented with 100 IU/mL rhIL-2. Cells were stained on ice for 20 minutes with Live/Dead Fixable NearIR Dead Cell Stain (1:1500), CD4 APC (eBioscience 14–0041-83, 1:400), CD62L FITC (1:200), and CD44 PE (1:800). CD25 PE (BioLegend 102007, 1:200) was added to the staining cocktail for naïve CD4s to exclude endogenous Tregs. The FCCP concentration was doubled (50 μL of 4 μM FCCP added per well) to compensate for the presence of extra FBS.

For Th1 and Th2 restimulation, 250,000 cells were plated per well in a 96-well U-bottom plate. Cells were resuspended in 200 μL TCM + 10 ug/mL brefeldin A (AdipoGen AG-CN2–0018-M025) +/− eBioscience Cell Stimulation Cocktail (00–4970-93) and incubated at 37 °C for 3.5 hours. Cells were washed with PBS + 10 ug/mL brefeldin A, then Fc blocked and stained with Live/Dead Fixable NearIR Dead Cell Stain (1:1500) on ice for 15 minutes. Cells were washed again with PBS + brefeldin A and fixed with 4% paraformaldehyde at room temperature for 10 minutes. Cells were washed with PBS + brefeldin A, then permeabilized with CD4 Permeabilization Buffer (50 mM NaCl + 5 mM EDTA + 0.02% sodium azide + 0.5% Triton X-100) at room temperature for 10 minutes. Cells were stained for intracellular cytokines in CD4 Permeabilization Buffer at room temperature for 20 minutes. Th1 cells were stained with IFNγ FITC (BioLegend 505806, 1:400) and IL-17 APC (BioLegend 506916, 1:400). Th2 cells were stained with IFNγ FITC (1:400) and IL-4 PE (BioLegend 504104, 1:400). Cells were washed with PBS, then resuspended in SWB and analyzed using a Beckman Coulter CytoFLEX S flow cytometer.

For iTreg restimulation, cells were plated and restimulated for 3.5 hours as described above for Th1s/Th2s. Cells were washed with PBS + brefeldin A, then Fc blocked and stained for viability as described above. Cells were washed with PBS + brefeldin A, then fixed on ice for 30 minutes with eBioscience Fixation/Permeabilization Buffer. Cells were washed with eBioscience Permeabilization Buffer and stained with FoxP3 PE (BioLegend 126404, 1:100) and IL-17 APC (1:400) in Perm Buffer on ice for 1 hour. Cells were washed with Perm Buffer, then resuspended in SWB and analyzed using a Beckman Coulter CytoFLEX S flow cytometer.

MVA-ova viral immunization

Attenuated vaccinia virus, strain Ankara expressing ovalbumin (MVA-ova) was a gift from Dr. Pamela Rosato (Geisel School of Medicine at Dartmouth College). MVA-ova was grown and titered using BHK cells (originally from Dr. William Green, Dartmouth affiliation as above) as previously described.^14,15 For viral immunizations, naïve CD8 T cells were isolated from pooled spleens and lymph nodes of female CD45.1+ OT-I mice age 6–16 weeks old as described above. 50,000 naïve OT-Is were administered retro-orbitally (r.o.) to female C57BL/6N mice ages 8–12 weeks under isoflurane anesthesia. The following day, mice were immunized intraperitoneally (i.p.) with 1×10⁷ PFU MVA-ova. Separate age-matched cohorts of mice were immunized 30 days or 7 days prior to euthanasia for the day 30 and day 7 timepoints, respectively. Unmanipulated, age-matched female OT-I mice were used as naïve controls.

For the RICA assay, spleens of immunized and naïve mice were collected, mashed, and RBC lysed as described above. ~2×10⁶ cells per well were plated in a 96-well U-bottom plate. Cells were washed with cold RICA buffer, then Fc blocked for 10 minutes and stained for 20 minutes on ice. All cells were stained with Live/Dead Fixable NearIR Dead Cell Stain (1:1500). For CD44 × CD62L comparisons, cells were stained with CD62L FITC (1:200) and CD44 APC (1:200). Effector and memory cells were stained with CD45.1 PE (BioLegend 110708, 1:200), while naïve cells were stained with CD8β PE (1:400). For the KLRG1 × CD127 comparisons, cells were stained with CD45.1 PE (1:200), KLRG1 FITC (BioLegend 138410, 1:200), and CD127 APC (BioLegend 135012, 1:200). Cells were washed twice with cold RICA buffer, then resuspended in 200 μL warm buffer, incubated at 37 °C for 10 minutes, and analyzed using the RICA assay.

B16-ova tumor experiments

B16 mouse melanoma cells expressing ovalbumin (B16-ova) were a gift from Dr. Mary Jo Turk (Dartmouth affiliation as above). 250,000 B16-ova cells were administered intradermally (i.d.) to female C57BL/6N mice age 10–12 weeks old. On the same day, a spleen from one female CD45.1+ OT-I mouse age 6–16 weeks old was harvested, mashed, and RBC lysed as described above. Splenocytes were washed with TCM and plated in 24-well plates at a density of 0.5×10⁶ cells/mL + 1 μg/mL SIINFEKL + 25 IU/mL rhIL-2. Cells were incubated at 37 °C for 48 hours, then collected, washed with TCM, and re-seeded at 0.5×10⁶ cells/mL in TCM + 25 IU/mL rhIL-2. Cells were collected, washed, and re-seeded with rhIL-2 every 2 days for the following 5 days.

Six days after tumor inoculation, tumor-bearing mice received 500 centigray whole-body irradiation. One day later (on day 7 of OT-I culture), 1×10⁶ OT-I cells were administered r.o. to each mouse. Tumor size and mouse condition were monitored every two days. Mice were euthanized 7 days after adoptive cell transfer.

Tumors, tumor-draining lymph nodes (right inguinal), and spleens were collected into B16 medium on ice. Tumors were mashed through sterile nylon mesh in B16 medium using a syringe plunger. Tumors were spun down and resuspended in 4 mL 44% Percoll (Cytiva 17–0891-02) in RPMI underlaid with 3 mL 67% Percoll in PBS. Tumors were centrifuged at 805 × g for 25 minutes at 18 °C with low brake. Debris was aspirated off the top, and live leukocytes were collected from the density interface and washed with B16 medium. Cells were RBC lysed in 1 mL RBC lysis buffer at 37 °C for 1 minute, then washed with B16 medium. Lymph nodes were mashed, spun down, resuspended in B16 medium, and filtered through 35 μm cell strainer tube caps (Falcon 352235). Spleens were mashed, spun down, and RBC lysed in 2 mL RBC lysis buffer at 37 °C for 3 minutes. Spleens were washed with B16 medium and filtered through 35 μm cell strainer tube caps.

For the RICA assay, cells were plated in a 96-well U-bottom plate and washed once with cold phenol red-free TCM. Cells were Fc blocked in RICA buffer on ice for 10 minutes, then stained in RICA buffer on ice for 20 minutes. All cells were stained with Live/Dead Fixable NearIR Dead Cell Stain (1:1500). Tumors and lymph nodes were stained with CD45.1 FITC (BioLegend 110706, 1:200), PD-1 APC (BioLegend 109112, 1:200), and TIM3 PE (BioLegend 119704, 1:400). Spleens were stained with CD8α PE (eBioscience 2–0081-83, 1:400), CD45.1 APC (BioLegend 110714, 1:400), CD44 APC (1:400), and CD62L FITC (1:200). Cells were washed twice with cold phenol red-free TCM, then resuspended in 150 μL warm TCM. Each well was split 3 × 50 μL into new wells. 50 μL of 1 μM rotenone in TCM were added to the first set of wells. Cells were incubated at 37 °C for 10 minutes, and then 50 μL each “Maximum” well were recorded on the ZE5. 50 μL warm TCM were added to the next set of wells, and then 50 μL each “Basal” well were recorded on the ZE5. 50 μL of 4 μM FCCP in warm TCM were added to the final set of wells. Cells were incubated at 37 °C for 10 minutes, and then 50 μL each “Minimum” well were recorded on the ZE5.

Statistical analysis

All statistical analysis was performed using GraphPad Prism. Comparisons involving two groups were performed using two-tailed unpaired Student’s t-test, and those involving more than two groups were performed using Welch’s ANOVA with Dunnett’s T3 multiple comparisons test. For MVA-ova samples originating from the same mouse, differences between samples were validated using two-tailed paired Student’s t-test as described in the “Results” section. Data from the tumor experiments were analyzed using repeated measures one-way ANOVA with Tukey’s multiple comparisons test.

Results

RICA Assay Design

To interrogate NAD biology in living cells, the RICA assay involves three parts: (i) Basal NADH measurement of untreated cells, (ii) Maximum mitochondrial NADH measurement by inhibition of complex I of the ETC, and (iii) Minimum mitochondrial NADH measurement by treatment with the proton ionophore FCCP. We designed this assay to integrate easily with existing flow cytometry workflows. First, live cells from in vitro or ex vivo samples are collected, stained for viability and surface markers, and washed. All staining and washing steps are done on ice or at 4 °C to preserve cell viability (Fig. 1A). In most experiments, we used RICA buffer (1x PBS + 2% FBS + 1 mM EDTA + 25 mM HEPES) for staining, washing, and flow cytometry. RICA buffer is cell purification buffer routinely used for FACS, with the addition of a HEPES buffering agent, which we and others have found protects against pH changes that can disrupt NADH autofluorescence measurements (data not shown).^16,17 Importantly, RICA buffer lacks preservative agents such as sodium azide, which inhibits the mitochondrial electron transport chain (Table 1). This convenient, cost-effective buffer uses reagents that are readily available in most labs that regularly do flow cytometry. While our data demonstrate that standard RICA buffer works well for a variety of metabolic experiments, it can be further supplemented or replaced with other media if required (see below).

Figure 1.

The RICA assay measures mitochondrial NAD redox index and capacity within subsets of living immune cells. (A) Workflow of the RICA assay. (B) Example data demonstrating gating on the NADH-hi population within a phenotypic subset of interest (T_eff) in the Basal, Maximum, and Minimum measurements. Cells were gated on single, live, CD8+, CD44+ CD62L- lymphocytes as shown in Supplemental Fig. 1A. (C) Example Basal, Maximum, and Minimum NADH autofluorescence histograms, and a diagram showing multiple replicates used for calculating mitochondrial NAD redox index and capacity. (D) Example mitochondrial NAD redox index and capacity values for the T_eff subset shown in (B). (E) Example data showing the size-normalized capacity comparison between the T_eff subset and a control subset (T_n). (F-H) RICA assay measurements on recently activated CD8 T cells treated with metabolic modulators. Naïve CD8s were activated in vitro with anti-CD3/anti-CD28 + IL-2 for 48 hours prior to treatment. Cells were gated as in Supplemental Fig. 1A, but with the final gate encompassing the entire CD44+ population. (F) Effects of 1-hour oligomycin treatment on the mitochondrial NAD redox index. (G) Effects of pyruvate and/or UK5099 addition to RICA buffer on the mitochondrial NAD redox index. (H) Effects of 5-hour NMN and/or FK866 treatment on the mitochondrial NAD capacity. Each group consisted of 4–6 technical replicates, and each experiment was performed 2–3 times. Data were analyzed using two-tailed unpaired Student’s t-test or Welch’s ANOVA with Dunnett’s T3 multiple comparisons test. Bars that lack a shared letter in the same graph are significantly different from one another.

Table 1.

Technical considerations and recommendations for the RICA assay. Buffer fuel source(s) and staining panel design may be tailored to each experiment. Acquisition parameters may be adjusted for different flow cytometers.

Buffer
Component	Rationale	Recommended
Fuel source	Enables mitochondrial respiration	2% FBS; others as needed
Buffering agent	Minimizes pH disruption	25 mM HEPES added day of assay
Note: Preservative agents such as sodium azide should not be used
Panel Design
Fluorophores/ Dyes	Must minimize emission overlap with each other and spillover into NADH autofluorescence detector (see below)	FITC, PE, APC NearIR viability dye
Flow Cytometer
UV laser	Must excite ~350 nm	Bio-Rad ZE5 355 nm UV laser
Detector	Must detect ~400–500 nm	Bio-Rad ZE5 UV447 filter (“DAPI”)
Sample collection	Must be able to stop acquisition after “Basal” measurements, leaving enough volume to split remaining cells in half for “Maximum” and “Minimum” measurements	On Bio-Rad ZE5, record 25 μL “Basal” each well, then split remaining cells 2 × 50 μL and record 50 μL after drug additions. Toggle “Return Sample” OFF.
Temperature control	Ideally, maintain samples at 37 °C*	Bio-Rad ZE5 stage temperature control setting

^*Note: If cytometer temperature control is not available, samples and buffers should be equilibrated to room temperature prior to running the assay.

After washing with cold RICA buffer, cells are plated in 200 μL warm RICA buffer per well of a 96-well U-bottom plate. Multiple wells of each sample are plated to provide technical replicates. Cells are incubated at 37 °C in a humidified incubator at 5% CO₂ for 10 minutes, then transferred to the flow cytometer (ideally programmed to maintain samples at 37 °C). Using staining controls, appropriate voltages are determined for each fluorophore of interest and for NADH autofluorescence detection. After setting acquisition parameters, 25 μL of each well are recorded for “Basal” measurements. After this, 50 μL each of the remaining cells in each well are split into two new wells for the “Maximum” and “Minimum” measurements. 50 μL 2x rotenone in warm RICA buffer are added to each “Maximum” well and mixed. Cells are incubated at 37 °C for 10 minutes, after which 50 μL each “Maximum” well are recorded on the cytometer. This process is then repeated for the “Minimum” wells - 50 μL 2x FCCP in warm RICA are added to each “Minimum” well, cells are incubated at 37 °C for 10 minutes, and then 50 μL each well are recorded (Fig. 1A). Recording twice the volume after drug additions ensures that roughly an equal number of cells are collected for the Basal, Maximum, and Minimum measurements of each sample. These acquisition volumes were optimized in our hands for using the Bio-Rad ZE5 cytometer without the “Return Sample” function, but they may be adjusted for different cytometers (Table 1).

For analysis, cell populations of interest are gated based on surface marker phenotype in the Basal, Maximum, and Minimum conditions (Fig. 1B, Supplemental Fig. 1A). Notably, compensation cannot be applied to these data, as it is not possible to obtain NADH-free cells to act as a negative control for NADH autofluorescence. Therefore, fluorophores and dyes used for viability and surface marker staining must be chosen carefully to minimize emission overlap with one another, and to prevent spillover into the NADH detection filter (Table 1). The same gate should be used for the Basal, Maximum, and Minimum conditions for each phenotypic population of interest. Then, within each population of interest, cells are gated for the NADH-high population (Fig. 1B). This omits dead or dying cells that were not excluded by viability gating, as they likely died after the viability staining step. Importantly, the optimal location of the NADH-hi gate often differs between the Basal, Maximum, and Minimum measurements (as expected after perturbing NAD metabolism) (Fig. 1B). Within each measurement, the same gate should be used for all technical replicates of each sample condition, and across conditions if possible. However, the most important consideration is to faithfully capture the NADH-hi population without erroneously including dying cells or excluding viable cells.

Once the NADH-hi population is obtained in each measurement, the geometric mean fluorescence intensity (gMFI) of NADH in this population is calculated (Fig. 1C). Each technical replicate will then have three measurements associated with it – Basal, Maximum, and Minimum NADH gMFI. These values can be plotted together to visualize their relationship (Fig. 1C). Additionally, the mitochondrial NAD capacity (a readout of the total pool of mitochondrial NAD in cells) and the mitochondrial NAD redox index (the percentage of mitochondrial NAD that is reduced, or in the form of NADH) can be calculated as follows (Fig. 1C, 1D):

$\mathrm{M}\mathrm{i}\mathrm{t}\mathrm{o}\mathrm{c}\mathrm{h}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{r}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{N}\mathrm{A}\mathrm{D}\mathrm{C}\mathrm{a}\mathrm{p}\mathrm{a}\mathrm{c}\mathrm{i}\mathrm{t}\mathrm{y}=\mathrm{M}\mathrm{a}\mathrm{x}\mathrm{i}\mathrm{m}\mathrm{u}\mathrm{m}\mathrm{N}\mathrm{A}\mathrm{D}\mathrm{H}\mathrm{g}\mathrm{M}\mathrm{F}\mathrm{I}‐\mathrm{M}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{m}\mathrm{u}\mathrm{m}\mathrm{N}\mathrm{A}\mathrm{D}\mathrm{H}\mathrm{g}\mathrm{M}\mathrm{F}\mathrm{I}$

$$\text{Mitochondrial NAD Redox Index}=\left(\frac{\text{Basal NADH gMFI}-\text{Minimum NADH gMFI}}{\text{Mitochondrial NAD Capacity}}\right)\times 100\text{\%}$$

The mitochondrial NAD capacity is most informative when normalized for cell size and compared across cell subsets (Fig. 1E). To normalize for cell size, the gMFI of forward scatter area (FSC-A) of the Basal NADH-hi population is calculated for each replicate. The size-normalized mitochondrial NAD capacity is calculated as follows:

$$\text{Size-Normalized Capacity}=\frac{\text{Mitochondrial NAD Capacity}}{\text{Basal FSC-A gMFI}}$$

The size-normalized mitochondrial NAD capacity can then be compared across different cell subsets by dividing each value by the mean size-normalized capacity of one subset (Fig. 1E).

The use of mitochondrial inhibitors that specifically disrupt NAD metabolism provides several advantages over other autofluorescence-based methods for measuring cell redox state. Inducing the maximum and minimum mitochondrial NADH signal omits the effects of background autofluorescence from NADPH and other molecules.^11,18 It also puts the basal mitochondrial NADH signal into a more meaningful biological context – for example, two cell types may have similar basal NADH autofluorescence values, but differences in the size of their mitochondrial NAD pools may mean that one is operating at a much higher mitochondrial NAD redox index than the other. This approach also allows the size of the mitochondrial NAD pool to be more accurately compared across cell types. Baseline measurements of NADH autofluorescence may differ dramatically between cell types due to differences in cell size.¹⁹ By inducing the minimum and maximum NADH signal, and normalizing this capacity to cell size, we can compare the mitochondrial NAD pool per unit cell area across cells of varying sizes. Finally, because the RICA assay is based on flow cytometry, it is more high-throughput than microscopy-based approaches. By including phenotypic markers, it can be used to compare mitochondrial NAD dynamics across immune cell subsets and in samples containing multiple different cell types. This powerful technique can be used to interrogate redox metabolism in T cells both in vitro and ex vivo, as described below.

Validating the RICA assay

We validated our measurements of mitochondrial NAD redox index and capacity by pharmacologically perturbing NAD redox balance and cellular NAD content. For these experiments, we used murine CD8 T cells that had been activated in vitro with anti-CD3/anti-CD28 for 48 hours, as these cells are highly metabolically active.^20,21 To perturb the mitochondrial NAD redox index, we treated cells for 1 hour with oligomycin, an inhibitor of mitochondrial ATP synthase that disrupts ETC activity and should cause a buildup of NADH relative to NAD⁺ inside the mitochondria. Indeed, the mitochondrial NAD redox index was significantly increased by oligomycin treatment (Fig. 1F, Supplemental Fig. 1B). As an orthogonal approach, we supplemented the RICA buffer with 1 mM pyruvate during the assay. Pyruvate is a fuel source catabolized in the mitochondrial TCA cycle, and pyruvate supplementation has been shown to increase the ratio of NADH to NAD⁺ inside the mitochondria.¹⁰ We observed an increase in mitochondrial NAD redox index upon pyruvate supplementation, which was reversed by treatment with UK5099, an inhibitor of the mitochondrial pyruvate carrier (Fig. 1G, Supplemental Fig. 1C).

To validate our assessment of mitochondrial NAD capacity, we pre-treated cells for 5 hours with nicotinamide mononucleotide (NMN), a major biosynthetic precursor of NAD⁺ via the salvage pathway.⁴ NMN treatment slightly increased the mitochondrial NAD capacity in recently activated CD8 T cells (Fig. 1H, Supplemental Fig. 1D). This small increase may be attributed to the fact that these cells are highly metabolically active, and therefore may be close to their maximum rate of NAD⁺ synthesis already. More notably, treatment with FK866, an inhibitor of the NAD⁺ biosynthetic enzyme NAMPT, dramatically lowered the mitochondrial NAD capacity (Fig. 1H, Supplemental Fig. 1D). Since NMN is downstream of NAMPT in the salvage pathway,⁴ NMN supplementation partially, but not completely, restored the mitochondrial NAD capacity in the context of FK866 treatment (Fig. 1H, Supplemental Fig. 1D).

In addition to altering the mitochondrial NAD redox index, we observed some interesting effects of oligomycin and pyruvate treatment on the mitochondrial NAD capacity. Oligomycin increased the mitochondrial NAD capacity (Supplemental Fig. 1B), whereas pyruvate decreased the capacity, even in the presence of UK5099 (Supplemental Fig. 1C). Similarly, our interventions to manipulate mitochondrial NAD capacity also affected the redox index. NMN supplementation decreased the mitochondrial NAD redox index slightly, while FK866 increased it. Combining the two treatments returned the redox index to its value in untreated cells (Supplemental Fig. 1D). These effects on the redox index were not unexpected, as promoting NAD⁺ synthesis with NMN could decrease the ratio of NADH to NAD⁺ in the mitochondria, while inhibiting NAD⁺ synthesis would have the reverse effect. This could occur through direct import of newly-synthesized NAD⁺ into the mitochondria via the mitochondrial NAD⁺ importer SLC25A51,^22–24 which is supported by the observed changes in mitochondrial NAD capacity. Additionally, the effects of NMN and FK866 on cytosolic NADH/NAD⁺ balance could be transmitted to mitochondrial NADH/NAD⁺ through the activity of mitochondrial electron shuttles, such as the malate-aspartate shuttle, which are known to be important for T cell function.^25–27 Collectively, our results point to mitochondrial NAD redox index and capacity being closely linked, rather than two independent parameters.

Given that several of our metabolic modulations affect whole-cell NAD metabolism, we wanted to confirm that our observations truly reflected changes in the mitochondrial NAD compartment. The use of rotenone and FCCP should make our measurements specific to mitochondrial NAD, as rotenone inhibits mitochondrial ETC complex I (which converts mitochondrial NADH to NAD⁺), and FCCP decouples ETC activity from the proton gradient to induce maximal oxidation of mitochondrial NADH to NAD⁺.¹⁸ However, we wished to confirm this by quantifying NAD⁺ and NADH in isolated mitochondria following our metabolic perturbations. To generate sufficient CD8 T cell numbers for these experiments, we activated OT-I T cells in vitro with SIINFEKL peptide and expanded them in the presence of IL-2 for six days. We then treated cells with pyruvate, UK5099, or FK866 at the same concentrations used above, followed by mitochondrial isolations and quantification of NAD⁺ and NADH in the isolated mitochondria using a fluorometric cycling assay. Enrichment of mitochondria in the isolated fractions was confirmed by western blot (Supplemental Fig. 1E). We observed similar results in the RICA assay using day 6 OT-Is as we did with recently-activated CD8 T cells; for example, pyruvate increased the mitochondrial NAD redox index and decreased capacity, while UK5099 by itself had a very modest effect on redox index and no significant impact on capacity (Supplemental Fig. 1F, 1G). FK866 very modestly boosted the redox index and more dramatically decreased the capacity (Supplemental Fig. 1H, 1I). The variance in the data from the mitochondrial isolations was larger than that from the RICA assay, owing to the technical challenges of isolating mitochondria from primary T cells and quantifying NAD⁺/NADH in the relatively small resulting samples. However, the results follow similar trends as those from the RICA assay: pyruvate increased the mean proportion of NADH in the mitochondria and decreased mean overall NAD (NAD⁺ plus NADH) content; UK5099 had more modest effects; and FK866 did not affect the mean proportion of NADH but decreased the mean size of the total NAD pool (Supplemental Fig. 1F-I). These results confirm that our metabolic modulations perturbed mitochondrial NAD redox state and content specifically, validating that the RICA assay measures changes in mitochondrial NAD. Additionally, they highlight how the RICA assay is more precise and feasible for detecting changes in mitochondrial NAD than traditional NAD⁺/NADH biochemical assays, especially in primary immune cells.

To further confirm the specificity of RICA assay for mitochondrial NAD, we wanted to exclude potential contributions of NADPH to our measurements, as NADPH has identical autofluorescent properties to NADH.¹¹ NADPH autofluorescence should theoretically be subtracted out when the FCCP-induced “Minimum” signal is subtracted from both “Basal” and “Maximum” measurements.¹⁸ We attempted to experimentally validate this by inhibiting glucose-6-phosphate dehydrogenase (G6PD), a cytosolic enzyme that generates NADPH from NADP⁺ in the pentose phosphate pathway. We treated naïve or recently activated murine CD8 T cells with the commercially available G6PD inhibitor G6PDi-1 at 25 or 40 μM for 2 hours. These conditions have been published to deplete NADPH in murine CD8 T cells.²⁸ We confirmed this in our hands by measuring mitochondrial superoxide production, which should increase upon loss of NADPH due to the importance of NADPH in generating reduced glutathione to scavenge mitochondrial ROS. Indeed, there was a dose-dependent increase in MitoSOX Red staining in activated CD8 T cells treated with G6PDi-1 (Supplemental Fig. 1J). No significant effect was seen in naïve cells, likely due to their low levels of G6PD expression and activity.²⁸

Given the depletion of NADPH, we expected to see a decrease in background autofluorescence with G6PDi-1 treatment, which would be reflected by an overall drop in the Basal, Maximum, and Minimum measurements. Surprisingly, we instead observed an increase in background autofluorescence, shifting all three of these measurements higher in both activated and naïve cells (Supplemental Fig. 1K). This increased background autofluorescence could reflect an increase in cytosolic NADH levels upon G6PD inhibition via several mechanisms. Firstly, G6PD inhibition could cause a buildup in glucose-6-phosphate that is then routed into glycolysis instead of the pentose phosphate pathway, promoting the production of cytosolic NADH. Secondly, buildup of NADP⁺ due to G6PD inhibition could cause feedback inhibition of NAD kinase (NADK), an enzyme which converts NAD⁺ to NADP⁺. This may lead to accumulation of NAD⁺ in the cytosol and its subsequent reduction to NADH to maintain redox balance.²⁹ These and other mechanisms have been attributed to the increase in NADH levels observed upon G6PD knockout or inhibition in several studies.^30,31 Despite these substantial changes in background autofluorescence, G6PD inhibition had only modest effects on RICA parameters, slightly decreasing the mitochondrial NAD redox index or capacity at different G6PDi-1 concentrations in activated cells and increasing the redox index while decreasing the capacity in naïve cells (Supplemental Fig. 1L). These differences may reflect changes in mitochondrial NAD redox state and content concomitant with overall cellular changes in redox state upon G6PD inhibition. Overall, the large changes in background autofluorescence but small effects on mitochondrial NAD redox index and capacity in response to NADPH depletion confirm the specificity of the RICA assay for mitochondrial NADH.

Finally, we compared CD8 T cells from Charles River/NCI C57BL/6N (CR/NCI 6N) and Jackson Laboratories C57BL/6J (JAX 6J) mice using the RICA assay. Among other genetic differences between the two strains, 6J mice have a deletion in the gene for nicotinamide nucleotide transhydrogenase (NNT), a mitochondrial enzyme that reversibly transfers a hydride ion between NADH and NADP⁺ to generate NAD⁺ and NADPH.³² The loss of NNT function in 6J mice might alter NAD⁺/NADH and NADP⁺/NADPH ratios in the mitochondria, the effects of which could be detected using the RICA assay. Indeed, when we compared naïve and recently activated CD8 T cells isolated from CR/NCI 6N and JAX 6J spleens, we observed very small but statistically significant and reproducible differences between the two strains. (Supplemental Fig. 1M). These variations in mitochondrial NAD might be influenced by differential NNT activity as well as broader differences in metabolism, especially glucose metabolism, that exist between 6N and 6J mice.³² Notably, all other experiments with B6 mice in this study utilized the C57BL/6N strain with intact NNT.

Overall, our metabolic manipulations validate the RICA assay and its calculations of mitochondrial NAD redox index and capacity as a method for measuring and comparing mitochondrial NAD redox balance and content across cell subsets.

Differentiation state markedly affects NAD redox parameters

We next evaluated mitochondrial NAD redox index and capacity in murine CD8 T cell subsets in vitro. We first compared naïve OT-I CD8 T cells from a freshly isolated, unmanipulated spleen (T_n) with OT-Is that had been activated with SIINFEKL peptide in vitro and cultured in the presence of IL-2 or IL-15 to skew them towards effector (T_eff) or effector memory/central memory (T_em/T_cm) phenotypes, respectively.^33,34 Distinct CD8 T cell subsets were analyzed by gating based on CD44 and CD62L expression (Supplemental Fig. 2A). We found that T_eff cells had the highest mitochondrial NAD redox index, as well as the largest capacity. Conversely, T_n cells had the lowest redox index and the smallest capacity (Fig. 2A). Among the memory-skewed subsets, T_cm cells were equivalent in their redox index to T_n cells, while T_em cells had an intermediate redox index between these and T_eff cells. T_em and T_cm had the same mitochondrial NAD capacity, which was between that of T_n and T_eff cells (Fig. 2A). The elevated redox index and capacity of T_eff cells reflects the high metabolic activity of these cells,^20,35 while the higher capacity in T_em and T_cm compared to T_n corresponds to the heightened mitochondrial respiratory reserve of memory CD8 T cells compared to naïve cells.^34,36 Our results concur with previously understood aspects of metabolism in CD8 T cell differentiation and deepen our understanding of the role of mitochondrial NAD dynamics in this process.

Figure 2.

Mitochondrial NAD redox index and capacity differ between T cell subsets and media conditions. (A) RICA assay measurements in subsets of OT-I CD8 T cells. OT-I cells were grown with either IL-2 or IL-15 to generate T_eff or T_em/T_cm, respectively. These were compared to naïve OT-Is (T_n) from a freshly isolated, unmanipulated spleen. Cells were gated on single, live, CD8+ lymphocytes and further gated based on CD44 and CD62L expression as shown in Supplemental Fig. 2A. (B-C) Mitochondrial NAD redox index (B) and capacity measurements (C) of OT-I T_n and T_eff cells in RICA buffer, T cell medium (TCM), and Seahorse medium. (D-E) RICA assay measurements in OT-I T stem cell memory (T_scm) cells compared to other OT-I subsets. OT-I cells were grown with IL-2 and CAL-101 to generate T_scm cells. (D) TCF-1 expression in CD44^hi and CD44^lo T_scm cells. FMO = “fluorescence minus one” staining control. (E) Mitochondrial NAD redox index and capacity measurements for T_scm CD44^hi and CD44^lo cells compared to other OT-I subsets. (F) Mitochondrial NAD redox index and capacity measurements for CD4 T cell subsets. Naïve CD4s were activated in vitro and cultured with appropriate cytokines and blocking antibodies to generate Th1, Th2, iTreg, or Th0 cells. These were compared to naïve, CD25- CD4s (CD4 T_n) from a freshly isolated, unmanipulated spleen. Cells were gated on single, live, CD4+ lymphocytes and were further gated based on CD44 (Th1, Th2, iTreg, Th0) and CD62L/CD25 expression (CD4 T_n) as shown in Supplemental Fig. 2D. Each group consisted of 4–6 technical replicates, and each experiment was performed 2–3 times. Data were analyzed using Welch’s ANOVA with Dunnett’s T3 multiple comparisons test. Bars that lack a shared letter in the same graph are significantly different from one another.

These initial subset experiments, as well as our earlier validation experiments (Fig. 1), were performed in our standard RICA buffer. In our hands, this buffer was suitable for numerous metabolic interrogations of different subsets of primary CD8 T cells. However, later experiments with other cell types required more nutrient-rich media to sustain cell viability during the assay, as described below. Therefore, we decided to explore the effects of different media on RICA assay parameters in CD8 T cell subsets that we knew could withstand varying conditions. We first assessed our RICA buffer versus Seahorse Mito Stress Assay medium (Agilent Seahorse XF Base Medium + 2 mM glutamine, 10 mM glucose, 1.2 mM pyruvate, and 25 mM HEPES) and our lab’s standard T cell medium (TCM) (RPMI 1640 + 10% FBS, 1x MEM non-essential amino acids, 1 mM sodium pyruvate, and 25 mM HEPES). Phenol red was omitted from all media, and the HEPES concentration was increased to 25 mM in Seahorse medium and TCM to match that of the RICA buffer. We compared these media using CD8 T_n and T_eff cells, as these showed the largest differences in our initial experiments (Fig. 2A). Cells were stained and washed in cold RICA buffer, and transferred to their respective media for the ten-minute pre-Basal incubation and all steps afterward. Surprisingly, the media composition had a dramatic effect on the mitochondrial NAD redox index and capacity, even in such a short time. Within each medium, the T_eff cells maintained a higher redox index than the T_n cells (Fig. 2B). However, the redox index was highest in Seahorse medium, and lowest in TCM, for both T_n and T_eff cells (Fig. 2B). Even more strikingly, the mitochondrial NAD capacity was much lower in TCM than in the other two media in both T_n and T_eff cells (Fig. 2C). The T_eff cell capacity was also slightly lower in Seahorse medium than in RICA buffer (Fig. 2C). Notably, the differences between TCM and Seahorse medium were not solely driven by the presence of serum, as the redox index and capacity in T_eff cells differed between the two media even when serum was added to or omitted from both (Supplemental Fig. 2B, 2C). The differences in redox index are likely explained by the varying fuel sources available in each medium, which can rapidly alter the mitochondrial NADH:NAD⁺ ratio through both direct contributions to the TCA cycle and, in the case of carbohydrates, through contributions to cytosolic NADH via glycolysis that are then transmitted to mitochondrial NADH via the malate-aspartate shuttle.^10,37 The differences in the capacity are more surprising, as they indicate rapid changes in the size of the mitochondrial NAD pool in response to different nutrient availability, which is a potentially novel aspect of CD8 T cell biology.

We next expanded our analysis to other T cell subsets of interest. We generated CD8 T stem cell memory (T_scm) cells from OT-I splenocytes in vitro using the PI3Kδ inhibitor idelalisib (CAL-101).³⁸ We analyzed both the CD44^hi and CD44^lo subsets of these cells and confirmed that both subsets expressed the canonical stemness transcription factor TCF-1 (Fig. 2D). We next compared these T_scm subsets to our other in vitro generated CD8 T cell subsets in standard RICA buffer (Fig. 2E, Supplemental Fig. 2E). Similar to T_em and T_cm cells, both CD44^hi and CD44^lo T_scm cells had a mitochondrial NAD redox index intermediate between that of T_n and T_eff cells, with the CD44^lo T_scm cells being slightly higher than the CD44^hi cells (Fig. 2E). However, both subsets of T_scm cells had a much larger mitochondrial NAD capacity than any other CD8 T cell subset, more than twice that of T_n cells (Fig. 2E). This large pool of mitochondrial NAD in T_scm cells could help explain their elevated spare respiratory capacity (SRC) compared to exhausted progenitor-like cells, which promotes their enhanced antitumor efficacy in the context of adoptive cell transfer therapy.^38,39

Finally, we analyzed in vitro generated CD4 T cell subsets. We activated murine naïve CD4 T cells with anti-CD3/anti-CD28 in the presence of appropriate cytokines and blocking antibodies to skew them towards a Th1, Th2, iTreg, or Th0 phenotype.⁴⁰ We analyzed the CD44^hi population of each of these subsets in the RICA assay, while naïve CD4 T cells (CD4 T_n) were obtained by gating on CD4+ CD44^lo CD62L^hi CD25- cells from an unmanipulated, freshly-isolated C57BL/6N spleen (Supplemental Fig. 2D). We confirmed skewing to the appropriate subsets by restimulating the cells and measuring cytokine production and/or FoxP3 expression (Supplemental Fig. 2F). We found that RICA buffer was not sufficient to maintain the survival and mitochondrial function of all CD4 subsets in the RICA assay. Specifically, the Th2 and iTreg cells required more nutrients to survive the assay, and iTreg cells required a high concentration of IL-2 (100 IU/mL rhIL-2) (data not shown). Therefore, for all CD4 T cell subsets, we performed the RICA assay in TCM supplemented with 100 IU/mL rhIL-2. We also used a lower cell number per well than in our in vitro OT-I experiments (see “Materials and Methods”).

Among CD4 T cell subsets, we found that Th2 and iTreg cells had the highest mitochondrial NAD redox index, while Th1 and Th0 cells were not significantly different from T_n (Fig. 2F, Supplemental Fig. 2G). iTreg and Th1 cells had the highest mitochondrial NAD capacity, followed by CD4 T_n and Th0 cells, with Th2 cells being the lowest (Fig. 2F, Supplemental Fig. 2G). These findings highlight metabolic differences among CD4 T cell types and show the versatility of the RICA assay in evaluating various lymphocyte subsets. Overall, our results reveal the diversity of mitochondrial NAD redox index and capacity among CD8 and CD4 T cell subsets in vitro. They also uncover potentially novel T cell biology regarding the rapid changes in the size of the mitochondrial NAD pool in response to different nutrient environments.

NAD redox state differs in effector and memory subsets in an in vivo antiviral response

Having compared mitochondrial NAD state between CD8 T cell subsets in vitro, we next turned our attention to ex vivo applications. Thanks to its high-throughput nature and relatively low cell number requirements, the RICA assay is ideally suited for ex vivo analyses. We began with a viral immunization model using attenuated vaccinia Ankara virus expressing ovalbumin (MVA-ova). Naïve CD45.1+ OT-I T cells were adoptively transferred into congenically distinct recipients, which were immunized with MVA-ova one day later. Cohorts of mice were analyzed together on day 30 post-immunization (memory time point) and day 7 post-immunization (effector time point) (Fig. 3A, Supplemental Fig. 3A). CD45.1+ cells from the spleen of each immunized mouse were analyzed based on CD44 × CD62L (T_eff, T_em, T_cm) and KLRG1 × CD127 (short-lived effector cell/memory precursor effector cell (SLEC/MPEC) and KLRG1^lo CD127^hi memory). For the CD44 × CD62L comparisons, CD8+ CD44^lo CD62L^hi cells from the spleens of uninfected OT-I mice were used for the T_n subset (Fig. 3A, Supplemental Fig. 3A). We chose to use naïve cells from uninfected mice for the T_n subset, rather than endogenous T_n cells from the immunized mice, because these cells were unaffected by systemic viral exposure. Furthermore, having one external naïve group facilitated comparisons across time points.

Figure 3.

The RICA assay captures differences in mitochondrial NAD parameters ex vivo. (A) Experimental design for MVA-ova immunization experiments. Separate age-matched cohorts of mice were immunized 30 days or 7 days prior to euthanasia. Unmanipulated, age-matched female OT-I mice were used as naïve controls. Splenocytes of immunized and naïve mice were gated as shown in Supplemental Fig. 3A. (B) Mitochondrial NAD redox index and capacity measurements for T cell subsets based on CD44 × CD62L expression. (C) Mitochondrial NAD redox index and capacity measurements for T cell subsets based on KLRG1 × CD127 expression. Each group consisted of 5 mice, and the experiment was performed 3 times. Data were analyzed using Welch’s ANOVA with Dunnett’s T3 multiple comparisons test. Bars that lack a shared letter in the same graph are significantly different from one another.

Based on our in vitro observations, we hypothesized that the more effector-like cells (T_eff and SLEC) would have the highest mitochondrial NAD redox index and capacity compared to other subsets, with T_n being the lowest and memory cells/precursors (T_em, T_cm, KLRG1^lo CD127^hi memory, MPEC) in the middle. Indeed, CD44^hi CD62L^lo T_eff cells had a higher redox index than T_n, T_em, and T_cm cells (Fig. 3B, Supplemental Fig. 3B), and SLECs had a higher redox index than MPECs and KLRG1^lo CD127^hi memory cells (Fig. 3C, Supplemental Fig. 3C). However, among the CD44 × CD62L subsets, the capacity of T_eff and T_em cells was lower than that of T_n and T_cm cells (Fig. 3B, Supplemental Fig. 3B). Similarly, SLECs had a lower capacity than MPECs and KLRG1^lo CD127^hi memory cells (Fig. 3C, Supplemental Fig. 3C). These results show that, ex vivo, effector cells maintain a higher mitochondrial NAD redox index than other subsets due to their high metabolic activity, and that cells with higher differentiation potential (naïve, memory precursor, and memory) have an elevated mitochondrial NAD capacity compared to more terminally differentiated subsets.

Of note, these data were analyzed using Welsh’s ANOVA with Dunnett’s T3 multiple comparisons test, which treats all sample groups as unpaired. However, each memory timepoint mouse provided both a T_em and a T_cm sample, and each effector timepoint mouse provided both a SLEC and an MPEC sample, meaning that these groups (T_em – T_cm and SLEC – MPEC) were technically paired. To address this, we performed paired t-tests for the mitochondrial NAD redox index and capacity for these groups. The statistically significant differences (or lack thereof) identified by the t-tests matched those identified between the relevant groups in the ANOVA (Supplemental Fig. 3D, 3E; Fig. 3B, 3C). Overall, our results here provide insight into how mitochondrial NAD utilization changes during CD8 T cell differentiation in response to an in vivo viral immunization. They also broaden our understanding of the metabolic differences between in vitro- and in vivo-activated CD8 T cells.

Mitochondrial NAD capacity is elevated in TILs but declines with exhaustion

Finally, we expanded our analysis to ex vivo interrogation of tumor-infiltrating lymphocytes (TILs). C57BL/6N mice were inoculated intradermally with B16 melanoma tumor cells expressing ovalbumin (B16-ova). At the same time, CD45.1+ OT-I splenocytes were activated in vitro with SIINFEKL and IL-2 and expanded for adoptive cell transfer (ACT). One day before ACT, tumor-bearing mice were sub-lethally irradiated to partially deplete their endogenous lymphocytes (Fig. 4A). This freed “space” and allowed for better engraftment and expansion of transferred cells. Activated OT-I T cells were adoptively transferred into irradiated mice, which were sacrificed and analyzed 7 days later. We found that the TILs were extremely sensitive to our standard rotenone treatment; therefore, we ran these samples using phenol-red free TCM to maximize viability and modified the protocol to run the Maximum measurements before the Basal measurements (see “Materials and Methods”).

Figure 4.

The RICA assay reveals changes in the mitochondrial NAD capacity of TILs. (A) Experimental design for the B16-ova tumor model. TILs were gated on single, live, CD45.1+ lymphocytes, and further gated based on TIM3 and PD-1 expression as shown in Supplemental Fig. 4A. tdLN cells were gated on single, live, CD45.1+ lymphocytes as shown in Supplemental Fig. 4A. Splenocytes were gated on single, live, CD8+ lymphocytes as shown in Supplemental Fig. 3A, and further gated based on CD44/CD45.1 and CD62L expression. (B) Mitochondrial NAD redox index and capacity measurements for cells from tumor-bearing mice. Points connected by lines were derived from the same mouse. Each experiment used 6 mice, and the experiment was performed 3 times. Data were analyzed using repeated measures one-way ANOVA with Tukey’s multiple comparisons test. Bars that lack a shared letter in the same graph are significantly different from one another.

CD45.1+ cells in the tumor and tumor-draining lymph node (tdLN) were analyzed in each mouse (Fig. 4A, Supplemental Fig. 4A). In the tumor, TILs were divided between PD-1^hi TIM3^hi terminally exhausted cells (T_exh) and PD-1^hi TIM3^lo precursor exhausted (T_pre) cells, the latter being activated but not yet fully differentiated to a terminally exhausted state (Fig. 4A). In the tdLN, the number of transferred cells was much lower, so we analyzed the entire CD45.1+ population. Endogenous naïve (CD45.1- CD44^lo CD62L^hi) CD8 T cells from the spleens of tumor-bearing mice were used as T_n controls (Fig. 4A). The mitochondrial NAD redox index and capacity of these endogenous naïve CD8s from tumor-bearing spleens were similar to those of naïve CD8s from unmanipulated C57BL/6N or OT-I spleens (Supplemental Fig. 4B).

The redox index was remarkably similar across subsets in the tumor, tdLN, and spleen (Fig. 4B, Supplemental Fig. 4C). However, we saw dramatic differences in the mitochondrial NAD capacity between subsets. T_pre TILs consistently had the largest mitochondrial NAD capacity, which was sharply reduced among the T_exh TILs (Fig. 4B, Supplemental Fig. 4C). Surprisingly, CD45.1+ cells in the tdLN had the lowest capacity of any subset, even lower than that of naïve CD8s from the spleen (Fig. 4B, Supplemental Fig. 4C). Since the tdLN samples included the entire CD45.1+ population, they were more heterogeneous than the samples from other compartments, which could help explain why their capacity differed from others. Alternatively, the unique metabolic microenvironment of the tdLN may enforce a smaller capacity on these cells. The enlarged mitochondrial NAD capacity of activated PD-1^hi TIM3^lo T_pre cells is consistent with the observation that NAMPT expression and cellular NAD content greatly increase following TCR stimulation.^41–44 NAMPT expression subsequently declines in TILs.⁴³ This, along with defects in mitochondrial structure and function associated with T cell exhaustion,^45–47 could underly the loss of mitochondrial NAD capacity as cells progress from the T_pre to the T_exh state. These results highlight the profound effects of tumor localization and differentiation state on the size of the mitochondrial NAD pool in CD8 T cells. They provide greater detail to our understanding of the metabolic changes that underly T cell exhaustion, which can be used to inform strategies for counteracting TIL dysfunction and improving therapeutic efficacy.

Discussion

Here, we have presented the RICA assay, a novel flow-cytometry based technique for measuring mitochondrial NAD redox balance and content in living cells. We have defined the mitochondrial NAD redox index, a measure of the proportion of total mitochondrial NAD that is in the reduced form, and the mitochondrial NAD capacity, a measure of the total pool of mitochondrial NAD in cells. These parameters can be calculated and compared across immune cell subsets to provide a rigorous understanding of how NAD dynamics differ between cell types. More specifically, the mitochondrial NAD redox index reflects the balance between NADH and NAD⁺ in the mitochondria at a given time. This balance is dynamically influenced by many metabolic processes within the cell, including glycolytic flux, mitochondrial TCA cycle and ETC activity, the function of mitochondrial electron shuttles, and the consumption of NAD⁺ as a substrate by several mitochondrially localized enzymes. Therefore, the redox index can be used as a gauge for the overall metabolic activity of cells, with a higher redox index generally indicating cells with a higher rate of flux through pathways that generate NADH and ultimately ATP. The mitochondrial NAD capacity reflects the size of the total pool of mitochondrial NAD in cells. A larger NAD capacity might indicate cells with more mitochondria, or a higher concentration of NAD within each mitochondrion, among other possibilities. This in turn may indicate a cell’s capacity to engage mitochondrial ATP production when stimulated to do so. Together, these parameters provide a more detailed understanding of NAD biology than is possible with other techniques.

The RICA assay overcomes limitations in existing NAD measurement techniques in several ways. It requires relatively few cells, making it useful for ex vivo analysis and other situations in which cell number is limiting. It avoids cell lysis, preserving the metabolic state of living cells and limiting perturbations to NADH/NAD⁺ levels that can occur during sample processing.^8,9

The RICA assay also improves upon several aspects of other autofluorescence-based NAD measurement techniques. Many approaches utilize fluorescence lifetime imaging microscopy (FLIM) or microscopy-based intensity measurements of NADH autofluorescence to estimate cellular metabolic state. For example, FLIM has been used to discern the activation state of T cells, B cells, and NK cells;^12,48 to distinguish between CD4 and CD8 T cells within bulk human CD3+ populations;¹² and to observe alterations in redox metabolism of T cells infiltrating murine tumors of different sizes⁴⁹ and in response to anti-CTLA4 therapy.⁵⁰ Such “touch-free” techniques allow rapid assessment of the metabolic state of living cells, making them an attractive option for streamlining the production of adoptive cell therapy products.^51,52 However, FLIM and other microscopy-based techniques have important limitations. Firstly, the redox cofactor NADPH, which plays important roles in metabolic processes distinct from those of NADH, is almost identical to NADH in its autofluorescence properties.¹¹ FLIM cannot reliably distinguish between NADH and NADPH, and therefore, FLIM-based measurements of NADH autofluorescence entail a combination of NADH and NADPH.^11,12 The RICA assay circumvents this issue using rotenone and FCCP, which specifically disrupt mitochondrial NAD metabolism. Any autofluorescence signal from NADPH (or other autofluorescent molecules in the cell) is removed by subtracting out the “Minimum” NADH measurement from both the “Basal” and “Maximum” measurements during calculations of the mitochondrial NAD redox index and capacity.^11,18

Furthermore, certain microscopy-based methods for estimating metabolic state using NAD(P)H autofluorescence involve an additional cofactor, FAD.^12,13,49 Like NADH, FAD is autofluorescent, and its reduced form FADH₂ contributes electrons to the mitochondrial ETC to power OXPHOS. Because of this, the ratio of NAD(P)H to FAD autofluorescence signal has been used to estimate the balance of glycolysis versus mitochondrial respiration in immune cells.^12,13,53 However, this approach provides an indirect representation of metabolic state, limiting the depth to which biology can be inferred from it. Additionally, these composite measurements are not specific to NAD, which has numerous biochemical roles distinct from those of NADP(H) and FAD(H₂). The RICA assay avoids these issues through the use of inhibitors specific to mitochondrial NAD metabolism. This provides a more focused picture of redox state, allowing specific interrogation of NAD metabolism, which itself is a critical aspect of immune cell function. Finally, since multiple parameters can be used to estimate redox state with microscopy-based methods, there are numerous different ways of defining “redox ratio” in the literature.^12,48,49,53 This makes comparing results across studies difficult, and different approaches can lead to different conclusions.⁵³ The RICA assay provides a standardized, simplified approach for calculating mitochondrial NAD redox index and capacity, facilitating comparisons across cell types and conditions.

In addition to FLIM and other microscopy-based approaches, several techniques using flow cytometry to measure NADH autofluorescence have been reported, including one from Abir and colleagues that uses mitochondrial inhibitors in a manner similar to the RICA assay.^19,44,54 However, Abir and colleagues approach their measurements differently, calculating the optical redox ratio (ORR) using both NADH and FAD autofluorescence after adding each mitochondrial inhibitor.⁵⁴ To our knowledge, ours is the first report of a flow cytometry-based assay that specifically focuses on mitochondrial NAD and condenses the basal, maximum, and minimum NADH autofluorescence measurements into two parameters (mitochondrial NAD redox index and capacity) that can be readily compared between cell types. Besides the benefits mentioned earlier, focusing on NAD is particularly useful in flow cytometry, as omitting FAD opens more fluorescence channels for surface marker staining.¹¹ Additionally, our approach normalizes for cell size, which can vary dramatically as immune cells differentiate and will affect baseline NADH levels.¹⁹ Finally, ours is the first report to systematically characterize CD8 T cell NAD metabolism using flow cytometry within a variety of phenotypic subsets and under multiple in vitro and ex vivo conditions. Our optimizations of these experiments and in other contexts (for example, in CD4 T cells) provide a useful starting point for others who might use this technique to study additional, potentially challenging cell types.

Our study using the RICA assay uncovered aspects of NAD biology in T cells that agree with previous findings while also contributing novel observations. In our preliminary validation experiments, we found that either oligomycin or pyruvate treatment increased the mitochondrial NAD redox index in recently activated CD8 T cells, in line with NADH:NAD⁺ measurements from previous studies.^10,54 The effects of pyruvate likely stem from its direct contribution to the TCA cycle to generate mitochondrial NADH; this is reinforced by the observation that blocking mitochondrial pyruvate import with UK5099 removed its effect on the redox index. In addition to producing mitochondrial NADH through the TCA cycle, activated CD8 T cells generate substantial amounts of NADH in the cytosol due to their high rate of glycolysis.¹ Electrons from this cytosolic NADH can be passed into the mitochondria via the malate-aspartate shuttle and glycerol phosphate shuttle to generate mitochondrial NADH and FADH₂, respectively, which go on to power ATP production via the ETC.^37,55 When ATP synthase activity is blocked by oligomycin, ETC activity slows down, preventing the oxidation of NADH to NAD⁺. This, alongside continued activity of mitochondrial electron shuttles (particularly that of the malate-aspartate shuttle), leads to NADH buildup within the mitochondrial matrix, driving the observed increase in mitochondrial NAD redox index.

The NAMPT inhibitor FK866 dramatically lowered the mitochondrial NAD capacity, in accordance with published observations of the importance of NAMPT for maintaining NAD levels in activated T cells.^41–44 Conversely, supplementation with the NAD⁺ precursor NMN partially rescued the effect of FK866 and boosted the capacity in untreated cells. Our results concur with previous findings that precursor supplementation increases intracellular NAD levels and can restore NAD under conditions that deplete it.^41,43,44,56

Our unexpected findings that oligomycin increased the mitochondrial capacity while pyruvate decreased it may provide novel insight into mitochondrial NAD balance. These changes occurred in a short time frame (1 hour pre-treatment for oligomycin experiments; pyruvate supplementation 10 minutes prior to Basal measurements), meaning that whatever mechanism contributed to the change in mitochondrial NAD content must act relatively rapidly. The mammalian mitochondrial NAD⁺ importer SLC25A51 was recently discovered and shown to be critical for maintaining mitochondrial NAD⁺ levels.^22–24 The increase in mitochondrial NAD capacity after oligomycin treatment could be attributed to increased NAD⁺ import via this transporter, potentially as an attempt to counteract decreased energy production upon ATP synthase inhibition.

To our knowledge, no transporter for exporting NAD from mitochondria to the cytosol has been reported. However, mitochondrial NAD⁺ can be consumed by sirtuins and poly(ADP-ribose) polymerase 1 (PARP1) and can leak out via nonspecific mitochondrial permeability transition (MPT) pores under conditions of stress.⁵⁷ It is possible that introducing a bolus of pyruvate may have caused metabolic stress that induced mitochondrial NAD depletion through one or more of these mechanisms, perhaps concomitant with reduced NAD⁺ import via SLC25A51. Alternatively, there may be an undiscovered mechanism for exporting NAD⁺ from mitochondria under conditions where rapid NAD rebalancing is required between the mitochondria and the cytosol. Introducing a large amount of pyruvate to the cytosol could deplete cytosolic NAD⁺ as it is used to convert this pyruvate to lactate, meaning that cells might theoretically draw more NAD⁺ out of the mitochondria to compensate. This possibility merits further investigation. Interestingly, SLC25A51 knockout Jurkat T cells maintained the same whole-cell level of NAD despite diminished mitochondrial NAD content,²² potentially indicating a controlled balance of NAD between mitochondrial and extramitochondrial compartments in T cells.

As part of our validation experiments, we interrogated the potential influence of NADPH metabolism on our NAD measurements. As described above, treating cells with rotenone and FCCP creates an inducible range specifically of mitochondrial NADH; any contribution from NADPH to the signal is subtracted out as background. Indeed, depleting NADPH via G6PD inhibition dramatically altered the background signal in activated and naïve CD8 T cells, but had more mild effects on the mitochondrial NAD redox index and capacity that could reflect changes in NAD itself upon NADPH depletion. We also saw very minor differences between mitochondrial NAD redox index and capacity in CD8 T cells from Charles River/NCI C57BL/6N and Jackson Laboratories C57BL/6J mice, despite the fact that 6J mice lack a functional NNT enzyme that catalyzes hydride transfer between NADH and NADP⁺. These small differences indicate that this reaction has a very minor contribution, if any, to our measured RICA assay parameters, at least in the absence of additional manipulations. Furthermore, all of the data presented here were from mice on the 6N background with intact NNT activity. However, the metabolic effects of NNT absence (as well as additional metabolic differences between 6N and 6J mice) may significantly impact NAD biology in other situations. Therefore, we advise other groups to be mindful of the C57BL/6 strain used when performing or interpreting the results of metabolic experiments.

After validating our measurements from the RICA assay, we turned our attention to mitochondrial NAD redox state in various T cell subsets in vitro. It is well established that CD8 T cells undergo profound metabolic changes over the course of their differentiation. These include changes to their relative reliance on mitochondrial metabolism and glycolysis for producing ATP and biosynthetic precursors, as well as alterations to mitochondrial structure and organization.^{20,21,34–36,58,59} Our results provide greater insight into the role of NAD in the metabolic adaptations underlying CD8 T cell differentiation. Previous studies using biochemical assays and metabolomics showed that IL-2 cultured T_eff CD8s have a higher total NAD content and NADH:NAD⁺ ratio than IL-15 cultured memory cells, and that T_n CD8s have a lower NADH:NAD⁺ ratio than any activated subset.^34,60 Studies using FLIM and flow-cytometry based autofluorescence readouts have also shown that stimulated T cells, B cells, and NK cells have higher NAD(P)H/(NAD(P)H + FAD) redox ratios compared to their unstimulated counterparts.^{12,13,48,53,54} In line with these results, we consistently observed higher mitochondrial NAD redox indices in activated T cells compared to naïve T cells, the highest being in IL-2 cultured T_eff cells. This elevated redox index reflects the high metabolic activity of effector cells, which rapidly catabolize fuels through both glycolysis and the TCA cycle to fulfill their energetic and biosynthetic needs.^20,35,59

Using the RICA assay, we were also able to compare the relative size of the mitochondrial NAD pool between CD8 T cell subsets, which has proven challenging using other techniques. We were initially surprised by the observation that IL-2 T_eff cells had a higher mitochondrial NAD capacity than IL-15 T_em and T_cm cells, even when normalized to cell size, as we expected the more OXPHOS-efficient structure and organization of memory cells to equate to a larger mitochondrial NAD pool.^34,36 However, this result is in line with whole-cell NAD measurements in IL-2 and IL-15 cultured CD8s.³⁴ It is possible that a higher number of mitochondria and/or concentration of NAD within each mitochondrion might account for the high mitochondrial NAD content of IL-2 T_eff cells. Additionally, recently activated CD8 T cells have been shown to be extremely reliant on NAD⁺ synthesis to support their metabolism.^43,44 It is therefore reasonable to conclude that effector CD8 T cells require a large pool of mitochondrial NAD to maintain their high level of metabolic activity.

The RICA assay allowed us to measure mitochondrial NAD redox state within specific phenotypic subsets of CD8 T cells without purification. For example, we were able to distinguish subtle differences in the NAD redox index of IL-15 cultured T_em and T_cm cells which could not be measured in bulk IL-15 culture.³⁴ This also allowed us to compare parameters between CD44^hi and CD44^lo subsets of T_scm cells, which are of great interest in the field of immunotherapy for their high differentiation potential and enhanced tumor control capacity.³⁸ Both T_scm subsets had a high mitochondrial NAD redox index relative to most other CD8 T cell types, consistent with previous metabolomics findings.⁶⁰ Furthermore, the mitochondrial NAD capacity of both T_scm subsets far exceeded that of T_eff, T_em, T_cm, and T_n cells. This may again reflect elevated metabolic activity in these cells, as well as their high mitochondrial content and spare respiratory capacity that promote their antitumor function.³⁹ Indeed, a recent study in CAR-T cells showed that aging-related declines in T cell NAD content drive mitochondrial dysfunction, the loss of stemness, and impaired antitumor efficacy in CAR-T products from older individuals, highlighting the importance of the mitochondrial NAD pool in T cell stemness and therapeutic potential.⁶¹

Notably, these Tscm cells were generated in vitro by inhibition of PI3Kδ, which orchestrates numerous pathways for metabolic reprogramming during T cell activation.⁶² Therefore, the observed impact on mitochondrial NAD redox index and capacity might arise from both direct effects of PI3Kδ inhibition and indirect effects of memory cell differentiation on overall cellular metabolism. It has been shown that the effects of PI3Kδ inhibition on T_scm formation are independent of mTOR, a master regulator of metabolic reprogramming during T cell activation.⁶³ Furthermore, T_scm cells retain their metabolic characteristics when PI3Kδ inhibition is removed upon tumor antigen exposure in vitro, and maintain their antitumor efficacy in the absence of inhibitor in vivo.^38,39 These data demonstrate that at least some of the metabolic characteristics of T_scm cells are intrinsic to their differentiation state, rather than a direct consequence of PI3Kδ inhibition. Further investigation is merited to determine the extent to which this includes their mitochondrial NAD biology, and whether the high mitochondrial NAD capacity of in vitro-generated T_scm cells is maintained in vivo and contributes to their antitumor efficacy. Overall, our results offer greater insight into the NAD dynamics of these metabolically flexible, therapeutically important cells, which could guide further research into metabolic interventions that enhance CD8 T cell antitumor efficacy.

While not our primary focus, we also analyzed mitochondrial NAD redox index and capacity in in vitro generated CD4 T cell subsets. It is known that CD4 T cell subsets vary in their rates of glycolysis and OXPHOS and in their functional sensitivity to inhibition of mitochondrial ETC complexes.^25,42,64,65 For example, Th1 cells generated in vitro are highly reliant on OXPHOS for their proliferation and function.²⁵ Th2 cells, while also possessing high mitochondrial activity, are especially dependent on glycolysis.^64,65 We found that Th1 cells possessed a low mitochondrial NAD redox index and a relatively high capacity, similar to mitochondrially-reliant CD8 T_n cells assayed in TCM. In contrast, Th2 cells had a high redox index and a lower capacity, more closely resembling highly glycolytic CD8 T_eff cells in TCM. iTregs, which are overall less metabolically active than effector CD4s and less reliant on NAMPT activity for their survival and function,^42,64,65 had characteristics of both groups: their redox index mirrored that of Th2 cells, while their capacity resembled that of Th1s. These observations further highlight the metabolic diversity among CD4 T cell subsets. Additionally, our CD4 experiments demonstrate the adaptability of the RICA assay. We showed how media composition and cytokine conditions can be tailored to support analysis of various cell types, even those that are sensitive to standard RICA conditions. This shows that our assay can be optimized to explore metabolism in lymphocytes, and also potentially other cell types.

Given that changing media composition alters cellular metabolism, we also wanted to address how the mitochondrial NAD redox index and capacity were affected by the medium used in the RICA assay. Even after a short (10 minute) pre-incubation in RICA buffer, TCM, or Seahorse medium, these parameters profoundly changed in CD8 T_n and T_eff cells. The fact that the redox index varied between media is not surprising, as the differing availabilities of glucose, amino acids, pyruvate, and FBS metabolites would reasonably influence cells’ reliance on glycolysis and OXPHOS, and therefore the ratio of NADH:NAD⁺ in the mitochondria. What are more striking are the changes in mitochondrial NAD capacity, even in such a short timeframe. It is especially interesting that cells in RICA buffer, which is the least nutrient replete of the three media (containing only 2% FBS as a fuel source), had the highest capacity, while those in TCM, which is the most nutrient dense, had the lowest. Nutrient stress and NAD rebalancing mechanisms could affect NAD capacity, as discussed above; however, less cellular stress would be expected in the richer TCM. Possibly another NAD rebalancing mechanism is at play, which would represent a novel finding in T cell biology, particularly relevant for adoptive T cell therapy. It is known that culture conditions impact the effectiveness of adoptive T cell therapy,⁶⁶ and it is likely that mitochondrial NAD pool size is a contributing factor.

Besides our extensive in vitro analyses, the RICA assay allowed us to interrogate NAD metabolism ex vivo, enabling the analysis of T cell metabolic status directly from the animal. In our MVA-ova immunization model, the most effector-like cells (T_eff and SLEC) had the highest redox indices in their respective groups, reflecting their high metabolic activity in a manner consistent with our in vitro results. In terms of mitochondrial NAD capacity, our results differed from our in vitro observations, but aligned more closely with our expectations based on the varying reliance of effector vs naïve and memory cells on mitochondrial ATP production. In the CD44 × CD62L group, T_n and T_cm cells had a higher capacity than T_eff and T_em cells, while in the KLRG1 × CD127 group, MPECs and memory cells had a higher capacity than SLECs. Additionally, the differences between subsets in both redox index and capacity tended to be less dramatic ex vivo than they were in vitro. It is well documented that the metabolism of CD8 T cells differs when activated in vitro vs ex vivo, or in normal culture media vs physiologic media.^35,67 Most studies have focused on the differential utilization of glucose and other nutrient sources between these contexts. However, since NAD reactions are crucial for enabling glycolysis, lactate production, and mitochondrial metabolism, mitochondrial NAD redox balance and content would also be reasonably expected to differ between CD8 T cells activated in vitro vs ex vivo.

To determine the effects of in vivo metabolic stress within the TME on NAD redox parameters, we turned to a B16-ova melanoma model. The RICA assay enabled us to interrogate mitochondrial NAD metabolism within specific subsets of CD8 T cells in the tumor and in secondary lymphoid organs, something which has proven challenging with other techniques. For example, we were able to distinguish NAD redox state between TIM3^hi and TIM3^lo PD1^hi antigen-specific TILs, and to directly compare these to antigen-specific CD8s in the tdLN and naïve CD8s in the spleen. We did not observe significant differences in the redox index between subsets. Chronic antigen stimulation in vitro has been shown to increase the NADH:NAD⁺ ratio in CD8 T cells.⁶⁸ However, similar to our MVA-ova results, our observations in TILs may reflect the disparity between mitochondrial NADH:NAD⁺ balance in vitro vs ex vivo.

Our most striking findings were again in mitochondrial NAD capacity. TILs had a higher capacity than both antigen-specific CD8s in the tdLN and naïve CD8s in the spleen. This was surprising, as studies on T cell exhaustion have shown that T_exh TILs have altered mitochondrial morphology, abnormal cristae structure, and OXPHOS deficits,^45–47 which we expected to coincide with decreased mitochondrial NAD capacity. Indeed, work comparing mouse spleen CD8s and human patient PBMCs with matched TILs has shown that TILs tend to have a lower NAD⁺ content than T cells outside the tumor.^43,69 However, several factors may contribute to this disparity. Firstly, prior studies analyzed the entire CD8+⁶⁹ or even the entire CD3+⁴³ pool within mouse and human samples, whereas we focused on specific subsets within each tissue (such as naïve CD8s within the spleen). A sizable pool of activated, circulating T cells in tumor-bearing mice and patients may have disproportionately contributed to the measured NAD⁺ pool in spleen and PBMCs in previous studies. Conversely, the low metabolic activity of T_n cells in the spleen, and the heterogenous nature of the CD45.1+ population in the tdLN, may help explain why our capacity measurements were lower in these cells relative to TILs. Furthermore, prior work quantified NAD⁺ alone in total cell lysates, whereas the RICA assay measures total NAD (NAD⁺ + NADH) in the mitochondria specifically. Failing to account for NADH may omit a substantial contribution to the total NAD pool.

Finally, NAD⁺ quantification in total cell lysates cannot distinguish between mitochondrial and extramitochondrial (cytosolic and nuclear) contributions. Under stressful conditions in the TME, extramitochondrial NAD⁺ might be consumed by epigenetic modifiers and DNA repair enzymes such as PARP1, lowering the overall cellular NAD content.^2,4,70 Additionally, TILs often face glucose restriction and lactate accumulation in the TME,^46,70 and PD-1 signaling itself has been shown to repress glycolysis in T cells.^71,72 TILs might theoretically respond to this glycolytic repression by shunting NAD⁺ from the cytosol to the mitochondria in an attempt to maintain ATP production. They might also do this to rebalance the mitochondrial NADH:NAD⁺ ratio and alleviate reductive stress in the mitochondria, which can occur under chronic antigen conditions.^26,68 Altogether, these factors could cause a lower measurement for overall cellular NAD content in TILs vs extratumoral CD8s, while the mitochondrial NAD content itself is higher in TILs. Regardless, our data support the idea that terminally exhausted CD8s have reduced mitochondrial function, as within TILs, the PD-1^hi TIM3^hi T_exh compartment had a lower mitochondrial NAD capacity than the PD-1^hi TIM3^lo T_pre compartment. Further work is required to better understand the changes in mitochondrial NAD dynamics that occur along the timeline of T cell exhaustion, and how these changes might be prevented or reversed to reinvigorate TIL function. Related to this, recent work has shown that T cell NAD content and mitochondrial function decline with aging, and that the efficacy of CAR-T cell products derived from aged hosts can be improved by restoring NAD levels.⁶¹ Cancer and the use of CAR-T cell therapies are more prevalent among older individuals. In order to improve therapeutic outcomes for these patients, strategies for boosting the mitochondrial NAD capacity of tumor-specific T cells to counteract the effects of aging and T cell exhaustion merit further investigation.

As with any technique, there are limitations to the RICA assay. Since the NADH measurements are based on arbitrary fluorescence intensity units, the assay is not strictly quantitative for NAD. It is better used for comparing the relative amounts of mitochondrial NAD⁺, NADH, and total NAD between cell types. Additionally, while the RICA assay can be used for ex vivo metabolic analysis, it cannot fully replicate in vivo metabolic conditions. Finally, the use of mitochondrial inhibitors and the relatively long duration of the assay (several hours) may prevent the analysis of sensitive cell types.

Overall, the RICA assay is a novel method for measuring mitochondrial NAD redox state that complements and builds upon existing techniques. It provides deeper insights into the dynamics of mitochondrial NAD specifically. Through the use of surface staining, it allows metabolic data to be interpreted in light of the wealth of phenotypic and functional information that exists about specific immune cell subsets. While we have presented numerous insights into CD8 T cell biology gained using this technique, future studies using this and other methods will deepen our understanding of the dynamics of this crucial metabolite in immune cells.

Data availability

Data will be made freely available in compliance with journal guidelines