Deep metabolic profiling of immune cells by spectral flow cytometry—A comprehensive validation approach

Summary

The advancing field of immunometabolism requires tools that link single-cell metabolism with immune function. Metabolic flow cytometry provides this capability, but its broad adoption has been limited by costly custom reagents and a lack of standardized methods for validating metabolic targets. Here, we present a standardized and user-friendly spectral flow cytometry panel that profiles eight key metabolic pathways at single-cell resolution using only commercially available antibodies, enabling simultaneous analysis of immune phenotype and metabolic activity . Applying this approach to lung myeloid and T cells following intranasal adenoviral CD40L vaccination revealed distinct metabolic phenotypes between resident and infiltrating myeloid cells, as well as functionally divergent metabolic programs in naive, effector, and tissue-resident memory T cells. Additionally, leveraging NAD(P)H autofluorescence allowed label-free detection of glycolysis and expanded the panel’s utility. This standardized approach reduces cost and experimental complexity, enabling researchers to elucidate how metabolism drives immune function across broader immunological and clinical contexts.

Graphical abstract

Highlights

• •Spectral panel with commercial antibodies tracks eight metabolic pathways
• •Validated with inhibitors to confirm marker specificity and metabolic pathways
• •Reveals distinct metabolic programs in macrophages and T cells in lung vaccination
• •Exploits NAD(P)H autofluorescence for label-free detection of glycolytic activity

Immunological methods; Immune response; Biocomputational method; Metabolomics

Introduction

Cellular metabolism is intimately tied to function.¹ Intermediates of metabolic processes activate various signaling pathways, act as substrates for enzymatic reactions and also regulate gene expression, thereby dictating effector functions of cells. They can also act directly as signaling molecules to induce both pro- and anti-inflammatory functions. This is of particular interest to immunologists, as metabolic reprogramming is required for immune cell differentiation, function and fate. For example, metabolic reprogramming of macrophages and T cell subsets toward anti-inflammatory states has been explored in various disease contexts, e.g., multiple sclerosis.^2,3,4,5 These strategies highlight the practical importance of studying cellular metabolism, which has significantly advanced our understanding of immune cell biology and opened new avenues for treatment.

Conventional methods of probing cellular metabolism rely on extracellular flux assays (e.g., Seahorse) or metabolomics. However, these techniques have several limitations. Firstly, these bulk-based assays require homogenous samples, such as sorted cells or cell lines, making it difficult to acquire meaningful data from heterogeneous cell populations. Secondly, they require substantial numbers of cells which limits the analysis of rare cell populations. Thirdly, the Seahorse assay interrogates only a select few metabolic pathways; several other pathways important to cellular metabolism are not captured. Finally, the Seahorse assay is an in vitro culture system that does not replicate the complex, heterogeneous in vivo environment of naturally functioning immune cells.

Metabolic status can also be estimated by the autofluorescence of certain biological molecules, such as reduced nicotinamide adenine dinucleotide (NADH). NADH is an intracellular co-enzyme that is a key electron carrier in cellular energy metabolism. In its reduced form as NADH, this molecule exhibits autofluorescence, whereas the oxidized NAD⁺ does not.^6,7 During glycolysis, NAD⁺ is reduced to NADH, whereas in the electron transport chain (ETC), NAD⁺ is produced. Thus, a shift in cellular metabolism toward glycolysis and/or lower mitochondrial respiration corresponds to higher NADH autofluorescence intensity⁸ and can be used to estimate glycolytic status. However, this method has primarily been used in imaging approaches and has limited application to the study of cellular metabolism by single-cell techniques like flow cytometry.⁸ The recent availability of full spectrum cytometry to detect cellular autofluorescence should enable more precise identification of specific autofluorescent signatures linked to unique metabolic states using flow cytometry.

Advances in single-cell technologies have led to the development of several pioneering approaches that overcome the shortcomings of bulk analysis, enabling the analysis of metabolic profile and phenotypic identity simultaneously. The first of these methods leverages single-cell RNA sequencing of a wide array of expressed genes, laying out a comprehensive map of potential metabolic changes at the mRNA level. The second technique, embodied by cytometric methods, such as single-cell metabolic regulome profiling (scMEP)⁹ and Met-Flow,¹⁰ quantifies rate-limiting enzymes, nutrient transporters, and transcription factors as proxies for different metabolic pathways. These approaches allow for the indirect assessment of metabolic pathway activity at the single-cell level. An additional cytometric approach, single cell energetic metabolism by profiling translation inhibition (SCENITH), combines metabolic inhibitors and protein synthesis quantification to assess the metabolic dependencies on glucose catabolism, mitochondria, and fatty acid/amino acid oxidation in immune cells at the single-cell level.¹¹ These approaches have yielded insights into the relationship between metabolism and function across several cell types,^{9,10,12,13,14,15} thus demonstrating the versatility and reliability of examining cellular metabolism by flow cytometry.

Although pioneering metabolic panels such as scMEP and Met-Flow have yielded valuable insights into cellular metabolism, their widespread adoption has been hindered by the need for expensive custom conjugation of reagents.^{9,10,12,13,14,15} In addition, standardized methods for optimization and validation of metabolic targets as reliable representatives of their respective pathways are limited and have generally focused only on the upregulation of glycolysis or oxidative phosphorylation with immune cell activation, often combined with readouts from extracellular flux assays.^9,10 This approach does not adequately cover other metabolic pathways (e.g., fatty acid metabolism and amino acid metabolism) and may not correlate with single-cell measurements of metabolic proteins, as these assays are based on bulk measurements. These challenges were recently highlighted by Cosgrove et al.,¹⁶ who called for the development of standardized, validated, and conjugated metabolic panels targeting key metabolic regulators to reduce barriers to entry into immunometabolic studies and facilitate widespread adoption of cytometric approaches in the field.

Here, we present a spectral cytometric metabolic panel that combines commercially available antibodies with cellular autofluorescence to profile immunometabolism. Our carefully optimized panel uses species cross-reactive reagents and incorporates validation through metabolic inhibitors, providing an accessible research tool for investigators to examine cellular metabolism at the single-cell level in humans and mice. This approach directly addresses the current need for standardized methods while also enabling simultaneous analysis of metabolic profiles and phenotypic identity across heterogeneous cell populations.

Results

Validation of a novel metabolic panel through established immune cell activation models

Given the limitations of existing tools to study metabolism at a single-cell level, we developed a comprehensive spectral cytometry panel combining immune cell identification with metabolic profiling at single-cell resolution. With the potential to include various additional immune and functional markers, this proof-of-principle panel integrates 12 immune cell markers with 9 key metabolic targets spanning major metabolic pathways: glycolysis (GAPDH), the TCA cycle (IDH2), the ETC (cytochrome c), hypoxia response (HIF-1α), amino acid transport (CD98), fatty acid oxidation (CPT1A) and synthesis (ACAC), and arginine/NO metabolism (Arg1/iNOS) (Table 1, Methods S1). The majority of antibodies are available commercially as fluorescent conjugates (Table 2), with the exception of ACAC and CPT1A which require a secondary fluorescent antibody step during the staining protocol (detailed in Methods S1).

Table 1.

Metabolic targets used in this study compared to previously developed metabolic panels

Metabolic target	Full name	Metabolic pathway	Function in metabolic pathway	Hartmann et al.9	Heieis et al.15	Levine et al.13	Ahl et al.10	Core panela 116	Core panela 217
GAPDH	Glyceraldehyde 3-phosphate dehydrogenase	Glycolysis & fermentation	Glycolytic enzyme catalyzing the conversion of glyceraldehyde 3-phosphate to 1,3-bisphosphoglycerate	x		x		x	x
IDH2	Isocitrate dehydrogenase 2	TCA cycle	Conversion of isocitrate to oxoglutarate				x	x	x
CytoC	Cytochrome c	Electron transport chain	Essential electron carrier	x		x		x	x
CPT1A	Carnitine Palmitoyltransferase 1A	Fatty acid oxidation	Fatty acid shuttling into mitochondria	x	x		x	x	x
ACAC/ACC1	Acetyl-CoA carboxylase	Fatty acid synthesis	Malonyl-CoA production, rate-limiting enzyme in fatty acid synthesis	x	x		x	x	x
CD98	CD98	Amino acid metabolism	Essential amino acid transporter	x	x	x		x	x
HIF-1α	Hypoxia-inducible factor 1-alpha	Metabolic regulation/signaling	Hypoxia and inflammation-induced transcription factor	x		x		x	x
iNOS	Inducible nitric oxide synthase	Amino acid metabolism	NO production, initial rate-limiting enzyme involved in arginine degradation
Arg1	Arginase 1	Amino acid metabolism	Conversion of arginine to ornithine

^aCore metabolic panels of key metabolic proteins suggested by Cell Metabolism review (2020) by Artyomov et al.¹⁷ or Nature Metabolism commentary by Cosgrove et al.¹⁶ (2024).

Table 2.

Antibodies used in this paper

Antibodies	Source	Identifier
Fc block
Anti-CD16/32 (clone 93) – purified (for Fc receptor blockade)	Biolegend	Cat # 101301
Extracellular labeling
Anti-CD11b (clone M1/70) – BUV395	BD	Cat # 563553
Anti-CD11b (clone M1/70) – PE	Biolegend	Cat # 101208
Anti-CD11b (clone M1/70) – APC-Cy7	BD	Cat # 557657
Anti-CD45 (clone 30-F11) – BUV661	BD	Cat # 612975
Anti-CD45.2 (clone 30-F11) – biotin	Biolegend	Cat # 103104
Anti-CD11c (clone F10/21A3) – BUV737	BD	Cat # 748723
Anti-CD11c (clone N418) – BV785	Biolegend	Cat # 117336
Anti-CD11c (clone N418) – PE	Biolegend	Cat # 117308
Anti-CD98 (clone H202-141) – BUV615	BD	Cat # 752360
Anti-CD8a (clone 53–6.7) – BUV805	BD	Cat # 612898
Anti-MHC-II (clone MP6-XT22) – BV510	Biolegend	Cat # 506339
Anti-MHC-II (M5/114.15.2) – BV480	BD	Cat # 566088
Anti-Ly6C (clone HK1.4) – BV605	Biolegend	Cat # 128036
Anti-Ly6G (clone 1A8) – BV650	Biolegend	Cat # 127641
Anti-Ly6G (clone 1A8) – PE	Biolegend	Cat # 127608
Anti-CD4 (clone GK1.5) – BV750	Biolegend	Cat # 100467
Anti-B220 (clone RA3-6B2) – BV785	Biolegend	Cat # 103245
Anti-B220 (clone RA3-6B2) – BV570	Biolegend	Cat # 103237
Anti-B220 (clone RA3-6B2) – PE	Biolegend	Cat # 103207
Anti-NK1.1 (clone PK136) – PE-Cy5	Biolegend	Cat # 108716
Anti-NK1.1 (clone PK136) – PE	Biolegend	Cat # 108707
Anti-F4/80 (clone BM8) – BV711	Biolegend	Cat # 123147
Anti-F4/80 (clone BM8) – Pacific Blue	Biolegend	Cat # 123123
Anti-CD3e (clone 500A2) – AF700	BD	Cat # 557984
Anti-CD3e (clone 17A2) – PE	Biolegend	Cat # 100206
Anti-Siglec-F (clone E50-2440) – BV480	BD	Cat # 746668
Anti-Siglec-F (clone RP/14) – PE	BD	Cat # 552128
Anti-CD103 (clone 2E7) – BB700	BD	Cat # 748240
Anti-CD62L (clone MEL-14) – BV650	BD	Cat # 564108
Anti-CD44 (clone) – BV605	Biolegend	Cat # 103047
Streptavidin PE-Cy5.5 (to detect biotin 5.2)	BD	Cat # 544062
Anti-CD3e BUV395 (clone SK7)	Biolegend	Cat # 564001
Anti-CD16 BV421 (clone 3G8)	Biolegend	Cat # 302037
Anti-CD14 BV510 (clone 63D3)	Biolegend	Cat # 367124
Anti-CD45R BV570 (clone RA3-62B)	Biolegend	Cat # 103237
Anti-HLA-DR BV650 (clone G46-6)	BD	Cat # 564231
Anti-CD8a BV711 (clone RPA-T8)	Biolegend	Cat # 301044
Anti-CD11c BV785 (clone Bu15)	Biolegend	Cat # 337251
Anti-CD45 PerCP (clone 2D1)	Biolegend	Cat # 368505
Anti-CD4 PerCP-Cy5.5 (clone RPA-T4)	Biolegend	Cat # 300530
Anti-CD19 APC-Cy7 (clone HIB19)	Biolegend	Cat # 302218
Anti-CD56 AF700 (clone 5.1H11)	Biolegend	Cat # 362521
Intracellular labeling
Anti-iNOS (clone CXNFT) – BV421	Invitrogen	Cat # 404-5920-82
Anti-Arginase 1 (clone A1exF5) – BV480	Invitrogen	Cat # 414-3697-82
Anti-GAPDH (clone W17079A) – AF488	Biolegend	Cat # 607905
Anti-Cytochrome C (clone 6H2.B4) – AF647	Biolegend	Cat # 612310
Anti-IDH2 – PE	Abcam	Cat # ab212122
Anti-HIF-1α (clone 241812) – APC	R&D systems	Cat # IC1935A
Anti-CPT1α (clone 8F6AE9) – Purified	Abcam	Cat # ab128568
Anti-ACAC – Purified	Abcam	Cat # ab72046
Anti-GAPDH (clone W17079A) – Purified	Biolegend	Cat # 607901
Anti-rabbit IgG – Dylight594 (Secondary Ab for ACAC)	Invitrogen	Cat # 35560
Anti-mouse IgG2b (clone RMG2b-1) – PE-Cy7 (Secondary Ab for CPT1A)	Biolegend	Cat # 406713
Anti-Rat IgG2a (clone MRG2a-83) – AF647	Biolegend	Cat # 407511
Intracellular labeling – isotype controls
Rat IgG2a, k Isotype – BUV615	BD	Cat # 613005
Rat IgG2a, k Isotype – AF488	BioLegend	Cat # 400525
Rabbit IgG Isotype – PE	Abcam	Cat # Ab209478
Mouse IgG1, k Isotype – APC	BioLegend	Cat # 400121
Mouse IgG1, k Isotype – AF647	BioLegend	Cat # 400130
Mouse IgG2b, k isotype purified	BioLegend	Cat # 401201
Rabbit IgG isotype purified	Invitrogen	Cat # 02-6102

Our selection of metabolic targets aligned with core recommendations in the field,^16,17 while also incorporating markers, iNOS and Arg1, to enable deeper analysis of macrophage metabolism (Table 1). We validated each antibody for staining using both fluorescence-minus-one (FMO) and isotype controls (Figure S1, Methods S1) and determined optimal antibody concentrations (Methods S1). All selected antibodies, except CD98, are cross-reactive with human samples (Figure S2), and for CD98, a human-specific equivalent is available on the identical fluorophore.

To compare and validate our panel against established approaches,^9,10,15 we examined two well-characterized immune cell activation models: murine bone marrow leukocytes treated for 16 h to stimulate classical (M1; LPS + IFN-γ) or alternative (M2; IL-4) macrophage activation, and splenic T cells activated with anti-CD3/CD28 beads for 4, 24, and 48 h (Figure 1).

Figure 1.

Metabolic profiling of stimulated macrophages and CD8⁺ T cells

(A) Fold change in metabolic marker expression (median fluorescence intensity) in bone marrow-derived macrophages (BMDM) following 16 h stimulation, normalized to unstimulated controls (M0).

(B) Heatmap analysis showing expression patterns of metabolic and functional markers in stimulated BMDM.

(C) Representative flow cytometry histograms of metabolic marker expression, with fluorescence minus one (FMO) controls or 2° antibody controls (CPT1A and ACAC) shown in gray.

(D) Temporal analysis of metabolic marker expression in splenic CD8⁺ T cells, comparing untreated (UT) cells to those stimulated with anti-CD3/CD28 beads at 4, 24, and 48 h.

(E) Heatmap visualization of metabolic marker expression dynamics in CD8⁺ T cells.

(F) Representative flow cytometry histograms showing metabolic marker expression, with FMO controls or 2° antibody controls (CPT1A and ACAC) shown in gray. MFI calculations are based on the gated positive population as determined by the isotype/FMO controls.

Data represent one independent experiment with 5–6 biological replicates per group, analyzed in technical duplicate. Each data point represents cells from an individual animal, with matched samples across experimental groups. Statistical analysis: repeated measures RM one-way ANOVA with Tukey’s correction for multiple comparisons (A and D). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Error bars are representative of mean ± SD.

Classically (M1) and alternatively (M2) activated macrophages exhibited distinct metabolic signatures that aligned with their established bioenergetic profiles.¹⁵ Within the gated macrophage population, approximately 50% of M1-stimulated cells expressed iNOS above background, whereas 10%–15% of M2-stimulated cells expressed Arg1, consistent with partial polarization potentially influenced by additional factors present in the mixed leukocyte bone marrow culture. M1 macrophages displayed significant upregulation of glycolytic markers (GAPDH) and HIF-1α signaling, coupled with enhanced amino acid transport (CD98) and nitric oxide production (iNOS) (Figures 1A and 1B). This metabolic shift toward glycolysis and inflammatory mediator production supports the rapid energy demands and pro-inflammatory functions of M1 macrophages. In contrast, M2 macrophages showed increased expression of oxidative phosphorylation markers (IDH2, cytochrome c), fatty acid oxidation enzymes (CPT1A), and enhanced arginine metabolism via Arg1 (Figures 1A and 1B), reflecting their role in tissue repair and homeostasis. Notably, M2 macrophages exhibited elevated expression of the fatty acid synthesis enzyme ACAC (Figure 1A). While ACAC is traditionally associated with M1 polarization, this finding aligns with previous metabolic profiling data showing enhanced fatty acid synthesis in M2 macrophages¹⁵ and supports the emerging understanding of fatty acid metabolism in M2 function.¹⁸ All markers demonstrated significant expression above FMO controls under these conditions (Figure 1C).

In CD8⁺ T cells, our results aligned well with established metabolic profiling approaches.^9,10 Upon T cell receptor engagement, we observed the expected simultaneous increase in glycolysis and oxidative phosphorylation, demonstrated by enhanced expression of the glycolytic enzyme GAPDH and TCA/ETC proteins, IDH2 and cytochrome c (Figures 1D and 1E). The temporal dynamics of metabolic remodeling in naive CD8⁺ T cells matched previous findings following antigen exposure. Interestingly, we detected an early downregulation of metabolic proteins including GAPDH, IDH2, and ACAC at 4 h post-stimulation (Figures 1D and 1E). CPT1A expression was significantly increased at 4 h post-stimulation, suggesting that CD8⁺ T cells may initially rely on fatty acid oxidation for early activation. Subsequently, there was a sharp increase at 24 h in GAPDH, IDH2, cytochrome c, CD98, ACAC, and HIF-1α, indicating enhanced glycolysis, OXPHOS, amino acid uptake, and lipid synthesis (Figures 1D and 1E). The timing of peak metabolic activation at 24 h coincided with maximal CD69 expression, confirming optimal T cell activation (Figure 1D). Our observations of lipid metabolism markers also replicated previous findings,⁹ with CPT1A showing an early spike at 4 h followed by progressive downregulation, while ACAC increased following activation from 24 h and remained stable (Figure 1D). The progressive increase in CD98 expression over time is consistent with the increasing amino acid demands of proliferating T cells (Figure 1D), matching the previously established pattern of peak proliferation at 48 h post-stimulation as measured by BrdU incorporation.⁹ All markers showed clear separation from FMO controls (Figure 1F). Of note, similar trends were observed for CD4⁺ T cells (Figure S3).

Validation of metabolic targets involved in central carbon metabolism

To validate our metabolic panel, we employed targeted pathway inhibition—an approach based on the principle that metabolic blockade triggers cells to rely on alternative pathways and results in compensatory changes in protein expression in the inhibited pathway. This strategy enabled validation through acute, reversible perturbations with well-characterized metabolic consequences. We focused on stimulated monocytes/macrophages as a model system for our validation approach. While inhibition was pathway specific, we could not exclude the possibility that other metabolic pathways were also impacted, as this reflects the interconnected nature of cellular metabolism.

We first examined markers of central carbon metabolism, including glycolysis, HIF1α signaling, TCA cycle, and ETC (Figures 2A and 2B). GAPDH was expressed in >97% of cells relative to FMO controls (Figure 2C). To validate GAPDH as a glycolytic marker, we treated mouse bone marrow cells with LPS + IFN-γ in the presence or absence of 2-DG, a hexokinase 1 inhibitor (Figure 2D). GAPDH expression was significantly decreased, both with 2-DG treatment in monocytes (Figure 2E) and in glucose-deprived media (Figure S4).

Figure 2.

Pathway-specific validation of metabolic targets involved in central carbon metabolism

(A) Schematic of central carbon metabolic targets for flow-based analysis.

(B) Summary of central carbon metabolic targets.

(C, F, I, and L) Dot plots showing positive fluorescent staining of the indicated metabolic targets on cultured CD45⁺ bone marrow cells relative to the fluorescence-minus-one (FMO) control.

(D, G, J, and M) Mechanistic schematics showing metabolic pathway stimulation (green) or inhibition (red) by the indicated drugs of glycolysis (D), HIF-1α signaling (G), tricarboxylic acid (TCA) cycle (J), and the electron transport chain (ETC) (M).

(E, H, K, N) Change in median fluorescence intensity (MFI) of positively stained monocytes under various culture conditions for GAPDH (E), HIF-1α (H), IDH2 (K), and cytochrome c (N). Statistics were calculated using RM one-way ANOVA with Tukey’s correction for multiple comparisons (E, H, and N) or paired t test (K).

Each data point represents cells isolated from a different animal, with data points matched across groups within each experiment. Data show one independent experiment with 3–4 biological samples per group performed in technical duplicate. ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Error bars are representative of mean ± SD.

We detected clear staining for HIF-1α in CD45⁺ cells, distinct from the FMO control (Figure 2F). When bone marrow cells were cultured under normoxia (20% O₂) or hypoxia (3% O₂) with HIF-1α modulators (Figure 2G), HIF-1α increased significantly under hypoxic conditions with dimethyloxalylglycine (DMOG) compared to normoxia with chrysin in monocytes (Figure 2H), confirming that this marker is a reliable readout of hypoxia-induced metabolic signaling.

Oxidative phosphorylation markers included IDH2 (TCA cycle) and cytochrome c (ETC), both showing clear separation from FMO controls (Figures 2I and 2L). IDH2 expression decreased with oligomycin treatment in alternatively (IL-4) activated macrophages (Figures 2J and 2K), confirming its regulation with OXPHOS inhibition. Similarly, cytochrome c expression decreased in alternatively activated macrophages treated with oligomycin, with or without FCCP (Figures 2M and 2N).

Validation of metabolic targets involved in fatty acid and amino acid metabolism

We next investigated fatty acid metabolism, encompassing both anabolic (fatty acid synthesis) and catabolic (fatty acid oxidation) pathways, along with amino acid metabolism markers CD98, which facilitates the uptake of large neutral amino acids, and iNOS, which metabolizes the amino acid L-arginine into NO (Figures 3A and 3B).

Figure 3.

Pathway-specific validation of metabolic targets involved in fatty acid and amino acid metabolism

(A) Schematic summary of fatty acid (FA) and amino acid (AA) metabolic targets for flow-based analysis.

(B) Summary of metabolic targets for fatty acid and amino acid metabolism.

(C, F, I, and L) Dot plots showing positive fluorescent staining of the indicated metabolic targets on cultured CD45⁺ bone marrow cells relative to the 2° antibody control (C and F) or fluorescence-minus-one (FMO) control (I, L).

(D, G, J, and M) Mechanistic schematics showing metabolic pathway stimulation (green) or inhibition (red) of fatty acid oxidation (D), fatty acid synthesis (G), large neutral (LN) AA uptake (J), and L-arginine metabolism by iNOS (M).

(E, H, K, and N) Change in median fluorescence intensity (MFI) of positively stained monocytes under various culture conditions for CPT1A (E) and ACAC (H) or percentage of cells expressing CD98 (K) and iNOS (N). Statistics were calculated using RM one-way ANOVA with Tukey’s correction for multiple comparisons (E, H, N) or paired t test (K).

Each data point represents cells isolated from a different animal, with data points matched across groups within each experiment. Data show one independent experiment with 3–4 biological samples per group performed in technical duplicate. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. Error bars are representative of mean ± SD.

CPT1A, our fatty acid oxidation marker, showed reliable staining compared to the FMO control (Figure 3C). We cultured cells with palmitate, with or without a low dose of etomoxir (3 μM) to specifically inhibit CPT1A¹⁹ (Figure 3D), and added 2-DG to promote fatty acid utilization over glucose.¹² Monocytes showed downregulation of CPT1A with etomoxir treatment, but importantly, this effect only occurred in palmitate-supplemented conditions (Figure 3E), confirming its specificity for active fatty acid oxidation pathway usage.

For fatty acid synthesis, ACAC showed reliable staining relative to the FMO control (Figure 3F). When classically activated monocytes/macrophages were treated with the fatty acid synthase inhibitor C75 (Figure 3G), ACAC expression decreased in a dose-dependent manner (Figure 3H), validating it as a marker of fatty acid synthesis. In contrast, the direct ACAC inhibitor 5-(tetradecyloxy)-2-furoic acid (TOFA) only affected ACAC expression at high concentrations (40 μM) (Figure S4), highlighting the different impacts these two inhibitors have on detectable ACAC protein levels despite both targeting the fatty acid synthesis pathway.

Investigation of large-neutral amino acid uptake capacity by CD98 staining demonstrated that this molecule was lowly expressed under homeostatic conditions relative to the FMO control (Figure 3I). To evaluate whether this marker is an accurate readout of amino acid uptake capacity, we stimulated bone marrow cells with lipopolysaccharide (LPS) and interferon gamma (IFN-γ) in culture media depleted of amino acids or supplemented with amino acids at levels in standard culture media (Figure 3J). This demonstrated that the proportion of CD98-expressing monocytes was modestly, but significantly, increased with amino acid supplementation (Figure 3K), indicating this marker is a reliable readout of amino acid uptake.

Finally, iNOS activity, which is linked to inflammation and glycolysis (Figure 3M), was clearly detectable relative to the FMO control (Figure 3L). While few cells expressed iNOS under homeostatic conditions, inflammatory stimulation with LPS and IFN-γ induced expression in approximately 60% of monocytes, which decreased to approximately 15% with 2-DG treatment (Figure 3N), demonstrating its strong connection to both inflammation and glycolysis.

Vaccination induces distinct metabolic signatures in lung myeloid subsets

To investigate whether our developed metabolic panel could detect metabolic differences between immune cell subsets in vivo, we focused on the lung microenvironment, which harbors both T cells and myeloid cells with distinct metabolic demands due to its unique conditions, including high oxygen and low glucose concentrations. We specifically aimed to evaluate metabolic differences between resident and peripherally derived myeloid cells, as well as between resident and circulating T cell subsets.

To examine these differences, we administered MemVax, a replication-deficient adenovirus serotype 5 vector expressing a membrane-bound chimeric CD40L intranasally. This approach induces a robust memory T cell response while simultaneously sensitizing lung myeloid cells. CD40L signaling plays a crucial role in T cell priming and myeloid cell activation, enhancing antigen presentation and cytokine production, and we used MemVax to assess how this targeted immune modulation influences metabolic adaptation in both myeloid and T cell compartments. Mice were culled 8-week post-vaccination (Figure 4A). To distinguish intravascular from tissue-resident cells, we injected biotin-conjugated anti-CD45 antibody (biotin 5.2) intravenously shortly prior to euthanasia. Biotin-labeled cells were detected flow cytometrically using PE-Cy5.5-conjugated streptavidin to distinguish circulating immune cells from tissue-resident populations.

Figure 4.

Metabolic profiling following vaccination reveals distinct metabolic profiles across myeloid populations

(A) Experimental set up. C57BL/6 mice (n = 4/group) were vaccinated intranasally with MemVax or PBS alone at days 0 and 14 (week 0 and 2), then allowed to rest for 8 weeks prior to organ collection. Isolated lung cells (week 10) were re-stimulated with LPS (100 ng/mL) or PBS alone overnight.

(B) MHC-II median fluorescence intensity (MFI) on alveolar macrophages (AM; Lin⁻, CD11b^low, CD11c⁺, Siglec-F⁺), interstitial macrophages (IM; Lin⁻, CD11b⁺, Ly6C^low, F4/80^hi), and intravascular Ly6C^hi monocytes (Lin⁻, CD11b⁺, Ly6C^hi, biotin 5.2⁺).

(C) Representative dot plots depicting iNOS expression in AM, IM and monocytes from MemVax or PBS-treated groups following LPS stimulation.

(D) Representative histograms showing metabolic marker expression in myeloid subsets. Note that although AM histograms for GAPDH and HIF-1α appear to have higher fluorescence than monocytes, this arises from a bimodal GAPDH distribution and higher background for HIF-1α; MFI calculations and gating against isotype/FMO controls confirm that AM exhibit lower true expression of these glycolytic markers than do monocytes.

(E and F) Heatmap visualization of metabolic marker MFI across myeloid populations from MemVax and PBS-treated group, treated with LPS (E) or PBS (F).

(G) Metabolic marker MFI across myeloid subsets.

Statistics were calculated by unpaired t test (B) or ordinary one-way ANOVA with Tukey’s correction for multiple comparisons (G). ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Error bars are representative of mean ± SD.

In the lung, we focused on several key myeloid cell subsets, including alveolar macrophages (AM; Lin⁻, CD11b^−, CD11c^hi, Siglec-F^hi), interstitial macrophages (IM; Lin⁻, CD11b⁺, F4/80⁺, Ly6C^low), and monocytes (Lin⁻, CD11b⁺, F4/80^low-int, Ly6C^hi, biotin 5.2⁺). Previous research has established that lung myeloid cells can develop a memory-like phenotype that enhances host defense through generally non-specific mechanisms.^20,21 To investigate this phenomenon, we re-stimulated myeloid cells isolated from both MemVax, and control-treated (PBS) groups with LPS or left them unstimulated (Figure 4A). LPS was selected as a potent toll-like receptor 4 ligand that reliably activates myeloid cells, providing a standardized secondary stimulus to assess trained immunity effects induced by MemVax.

We observed that myeloid cells from MemVax-treated mice exhibited a significantly heightened response to LPS stimulation, as evidenced by a marked increase in major histocompatibility complex class II (MHC-II) expression compared to cells from the PBS-treated group (Figure 4B). This finding suggests the establishment of a trained immune response in myeloid subsets following vaccination.²⁰ Notably, monocytes were found to be the most responsive to LPS stimulation, as demonstrated by elevated iNOS expression, a marker indicative of NO production (Figure 4C), with an approximately 24-fold increase in the percentage of monocytes expressing iNOS upon LPS exposure (Figure 4C), compared to the unvaccinated group, highlighting their enhanced responsiveness to non-specific stimuli.

Metabolic characterization of these myeloid populations revealed highly distinct metabolic signatures specific to each cell type (Figures 4D–4F). Hierarchical clustering based on metabolic markers demonstrated that cells clustered primarily by cell type rather than by vaccination status, in both LPS-stimulated and unstimulated conditions (Figures 4E and 4F). This indicates that the inherent metabolic phenotype of each myeloid cell type was the predominant determinant of their metabolic signature. Upon LPS stimulation, monocytes adopted a more glycolytic and classically activated (M1) phenotype, characterized by upregulation of GAPDH, CD98, HIF-1α, and iNOS expression (Figure 4G)—a profile consistent with the M1 polarization observed in BMDM in vitro (Figure 1). In contrast, alveolar macrophages exhibited a metabolic phenotype more aligned with the alternatively activated (M2) profile, displaying high expression of oxidative phosphorylation metabolic targets IDH2 and cytochrome c and upregulation of the pathways for fatty acid oxidation, CPT1A, and synthesis, ACAC, mirroring patterns observed with M2 stimulation (Figure 4G). Interestingly, interstitial macrophages exhibited an intermediate metabolic profile between monocytes and AMs (Figure 4G), potentially reflecting their heterogeneous origins developing from both peripheral and lung-resident myeloid cells.^22,23

CD8⁺ T cell subsets display characteristic metabolic phenotypes in response to vaccination

Given the established metabolic differences between naive, effector, and memory CD8⁺ T cells, we next examined metabolic variation among lung CD8⁺ T cell subsets. These included tissue-resident memory (TRM; Lin⁻, CD3ε⁺, CD4⁻, CD8α⁺, biotin 5.2^-, and CD103⁺), effector/effector memory (T_Eff/T_EM; Lin⁻, CD3ε⁺, CD4⁻, CD8α⁺, biotin 5.2^-, CD103^-, CD62L⁻, and CD44⁺), and intravascular (IV; Lin⁻, CD3ε⁺, CD4⁻, CD8α⁺, CD103^-, and biotin 5.2⁺) CD8⁺ T cells (Figure 5A). MemVax significantly increased TRM and T_Eff/T_EM populations while reducing IV CD8⁺ T cells (Figure 5B), which were predominantly naive (CD62L⁺ and CD44⁻) (Figure 5A), confirming a robust memory response.

Figure 5.

Metabolic profiling captures differential metabolic regulation of CD8⁺ T cell subsets following vaccination

(A) Characterization of lung CD8⁺ T cell populations, including tissue-resident memory (TRM; Lin⁻, CD3ε⁺, CD4⁻, CD8α⁺, biotin 5.2^-, CD103⁺), effector/effector memory (T_Eff/T_EM; Lin⁻, CD3ε⁺, CD4⁻, CD8α⁺, biotin 5.2^-, CD103^-, CD62L⁻, CD44⁺), and intravascular (IV; Lin^−, CD3ε⁺, CD4⁻, CD8α⁺, CD103^-, biotin 5.2⁺).

(B) Frequency of CD8⁺ T cell subsets as a proportion of total CD8⁺ T cells from the lung of MemVax and PBS-treated groups.

(C) UMAP visualization of T cell populations clustered by metabolic marker MFI, shown by experimental group.

(D) Flow cytometry plots showing T cell subset distribution across groups.

(E and F) UMAP plot (E) or dot plot (F) pseudocolored by CD8⁺ T cell subset.

(G) Metabolic marker MFI across CD8⁺ T cell subsets.

(H) Representative histograms of metabolic marker expression in CD8⁺ T cell subsets.

Data represent one independent experiment (n = 4 mice/group). Statistical analysis calculated by unpaired t test (B) or ordinary one-way ANOVA with Tukey’s correction for multiple comparisons (G). ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Error bars are representative of mean ± SD.

Unsupervised clustering of metabolic markers showed distinct metabolic profiles between circulating IV T cells and tissue-resident subsets (Figures 5C–5G). TRM exhibited increased expression of cytochrome c and CPT1A, indicative of enhanced oxidative phosphorylation and fatty acid oxidation, along with high CD98 expression, suggesting increased amino acid metabolism (Figures 5G and 5H). T_Eff/T_EM cells predominantly upregulated GAPDH, reflecting their reliance on glycolysis, whereas naive IV CD8⁺ T cells expressed high levels of IDH2, supporting their dependence on oxidative phosphorylation via the TCA cycle (Figure 5G and 5H). These metabolic profiles align with established paradigms of naive CD8⁺ T cells favoring oxidative metabolism, effector T cells relying on glycolysis for rapid energy production, and memory T cells utilizing fatty acid oxidation for longevity and survival.

Collectively, our metabolic panel successfully discriminated distinct signatures across diverse immune cell populations in the complex lung microenvironment, validating its utility for investigating immune cell metabolism in vivo and providing insights into how metabolic adaptations support specialized immune functions in tissue- and disease-specific contexts.

Single-cell autofluorescence detected by spectral flow cytometry is linked to glycolytic state

The Cytek® Aurora’s 5-laser spectral flow cytometry system captures cellular autofluorescence across 64 detectors (365–829 nm), enabling detection of metabolically relevant molecules such as NADH. This endogenous fluorophore serves as a natural indicator of metabolic state.^24,25,26 During glycolysis, NAD⁺ is converted to NADH, which is subsequently consumed by the ETC during oxidative phosphorylation; thus, increased NADH autofluorescence predominantly reflects enhanced glycolytic activity. We used this capability to develop a simple, cost-effective alternative to traditional metabolic assays by analyzing glycolytic activity in unstained cells. Although NADH and NADPH participate in distinct metabolic reactions—NADPH primarily supporting biosynthesis and antioxidant defense—they emit identical autofluorescence signals and are therefore henceforth collectively referred to as NAD(P)H. In our experiments, NAD(P)H was excited using the UV laser at 335 nm and its emission detected in the UV5 channel at 450 nm.

Analysis of unstained myeloid cells demonstrated that classically activated (M1) cells displayed the highest NAD(P)H autofluorescence compared to alternatively activated (M2) and naive (M0) cells (Figures 6A and 6B), consistent with the known glycolytic preference of classically activated macrophages.

Figure 6.

NAD(P)H autofluorescence reveals metabolic state and glycolytic activity in immune cell populations

(A) Quantification of NAD(P)H autofluorescence (UV5) in myeloid cells under different activation states: naive (M0), classically activated (M1), and alternatively activated (M2).

(B) Representative histogram overlay depicting NAD(P)H autofluorescence distribution across M0-, M1-, and M2-stimulated conditions.

(C and D) Lysed bone marrow cells stimulated with M0 (PBS), M1 (LPS, 100 ng/mL), or M2 (IL-4, 40 ng/mL) showing total NAD (C) and NADH (D) levels quantified enzymatically.

(E) Overlay histogram demonstrating the spectral independence of NAD(P)H autofluorescence (UV5) and GAPDH-AF647 staining.

(F) Correlation analysis between NAD(P)H autofluorescence intensity (UV5 channel) and GAPDH expression (R2 channel) in GAPDH-AF647 stained myeloid cells.

(G) Impact of glycolysis inhibition on NAD(P)H autofluorescence in M1-stimulated myeloid cells following treatment with 2-DG (10 mM). Statistical analysis performed using repeated measures RM one-way ANOVA with Tukey’s post-hoc test (A) or paired t test (C, D, and G). Data shown are representative of one independent experiment with 4–6 biological replicates per condition in technical duplicate. ∗p < 0.05 and ∗∗_∗p < 0.001. Error bars are representative of mean ± SD.

To validate these flow cytometry findings, we measured total NAD and NADH enzymatically with commercial assays. The results confirmed an increase in total NAD levels in M1-stimulated myeloid cells (Figure 6C), with a corresponding trend toward increased NADH (Figure 6D). Further confirmation of the relationship between glycolytic activity and NAD(P)H levels was obtained by examining GAPDH expression using AF647, a fluorophore selected specifically for its lack of spectral overlap with the UV5 NAD(P)H channel (Figure 6E). This analysis revealed a strong positive correlation between GAPDH expression and NAD(P)H autofluorescence (Figure 6F).

As a mechanistic validation, we subjected M1-polarized cells to glycolytic inhibition with 2-DG, which resulted in a significant reduction in NAD(P)H autofluorescence (Figure 6G). This direct manipulation of glycolytic flux indicates that the detected NAD(P)H signal predominantly represents glycolytic activity rather than contributions from alternative metabolic pathways.

Together, these findings demonstrate that spectral flow cytometry-based detection of NAD(P)H autofluorescence provides a reliable, label-free approach for assessing cellular glycolytic activity, offering a valuable complement to traditional metabolic assays.

Discussion

The ability to measure the metabolic state of immune cells is essential for a fundamental understanding of cellular function. Here, we present a metabolic spectral flow cytometry panel to simultaneously measure multiple metabolic pathways across diverse immune cell subsets at a protein level in single cells. The metabolic flow cytometry panel presented here employs commercially available reagents, eliminating the need for custom conjugation. We have meticulously validated each metabolic target through targeted inhibition or stimulation, demonstrating that their expression directly correlates with the corresponding pathway activity. This includes glycolysis, the TCA cycle, the ETC, HIF-1α signaling, fatty acid oxidation and synthesis, amino acid uptake, and L-arginine metabolism by iNOS and Arg1. This comprehensive validation approach has not been previously performed in other metabolic panels.

Our metabolic panel encompasses the core pathways essential for immune cell function and phenotypic definition. Through systematic evaluation using established activation models, i.e., classical/alternative macrophage polarization and T cell receptor stimulation, we demonstrate that our panel captures the fundamental metabolic shifts observed with pioneering platforms such as scMEP and MetFlow.^9,10,15 While these earlier approaches revealed important metabolic diversity across immune cell states, they rely on custom-conjugated antibodies, limiting user accessibility and introducing potential inter-laboratory variability. Moreover, previous validation strategies have typically been restricted to correlating protein expression with bulk Seahorse measurements during immune activation, without directly confirming pathway representation through metabolic manipulation.^9,10 We address this limitation through comprehensive validation using pathway-specific inhibitors and substrate depletion, establishing that our selected metabolic targets accurately represent their corresponding metabolic pathways. This validation approach provides strong evidence for metabolic target specificity, and by combining this thorough validation with commercially available reagents, our panel offers analytical depth comparable to existing platforms while improving accessibility and reproducibility for the broader immunology community. To further facilitate adoption, we provide a detailed technical guide that includes staining protocols, recommendations for essential controls, and a comprehensive troubleshooting section (Methods S1). This integrated approach enables researchers to investigate metabolic reprogramming in heterogeneous immune populations without requiring specialized conjugation expertise or custom reagents, thus lowering barriers to entry in the rapidly evolving field of immunometabolism.

A key strength of our approach lies in the thorough validation methodology we employed, using targeted metabolic inhibition, stimulation, and nutrient depletion to confirm that each selected marker accurately represents its nominated pathway. We found that metabolic target labeling correlated well with pathway activity in our study, although this may be dependent on cell type and experimental conditions. A recent study comparing co-expression patterns of scMEP markers with SCENITH-derived metabolic capacity in dendritic cells showed that out of 5 measured ETC/TCA regulators, cytochrome c correlated least with mitochondrial dependence.¹² On the other hand, CD98 correlated highly with fatty acid/amino acid oxidation capacity. Of the glycolytic enzymes, MCT1 and PFKB4 correlated highest with glycolytic capacity in dendritic cells, although GAPDH, the glycolysis target used in our study, was not evaluated. However, it is worth noting that GAPDH can also play a role in gluconeogenesis in certain cell types, such as hepatocytes and should be considered when evaluating glycolysis in non-immune cells. Overall, these findings highlight the importance of validation and optimization by cell type, metabolic label and experimental conditions to accurately reflect the activity of the pathway of interest. Finally, it should be noted that a lack of correlation may point to other active elements that could uncover useful information about metabolic pathway responses.

Our application of the validated panel in the lung microenvironment demonstrates its capacity to detect significant metabolic differences between immune cell subtypes in complex tissue contexts. Following vaccination with MemVax, we successfully distinguished the metabolic properties of diverse immune populations, revealing how metabolic programming supports specialized functional states. In CD8⁺ T cell subsets, we observed clear metabolic distinctions between naive, effector, and memory populations. This discrimination is crucial for understanding how metabolic reprogramming influences effector function, longevity, and recall responses. Tissue-resident memory T cells exhibited enhanced expression of the fatty acid oxidation enzyme CPT1A, aligning with previous studies highlighting the importance of this pathway in memory T cell generation and maintenance,^27,28 while effector T cells showed preferential engagement of glycolysis to support their rapid need for energy in effector functions.²⁹

This metabolic distinction was also evident in myeloid populations, where our panel’s ability to simultaneously resolve multiple metabolic pathways at single-cell resolution uncovered distinct profiles between AM, IM, and monocytes, revealing how tissue-specific environmental factors shape myeloid cell metabolism and innate immune memory. Within this myeloid compartment, AM displayed a distinctive metabolic signature characterized by upregulated fatty acid oxidation enzyme CPT1A and fatty acid synthesis enzyme ACAC—a pattern that underpins their capacity for innate immune memory formation.²¹ Notably, these AM maintained relatively low expression of glycolytic enzyme GAPDH even after LPS exposure, aligning with previous studies demonstrating that AM, unlike other macrophage populations, do not substantially increase glycolysis during inflammatory stimulation.^30,31 This glycolytic restraint appears functionally important, as glycolysis has been shown to be dispensable for establishing innate immune memory in AM,²¹ which instead preferentially engages fatty acid metabolism pathways for their specialized functions. In contrast to AM, monocytes in MemVax vaccinated groups exhibited a more inflammatory profile upon LPS stimulation, characterized by heightened iNOS expression. This enhanced inflammatory responsiveness mirrors previous observations that monocyte-derived AM produce elevated inflammatory cytokines and chemokines after exposure to inflammatory environments such as bacterial infection.³² The metabolic basis for this functional difference becomes apparent in our observation that lung monocytes maintain significantly higher glycolytic capacity than resident AM. This metabolic flexibility has previously been shown to provide monocyte-derived cells with a competitive advantage during inflammation, enabling them to outcompete embryonically derived resident AM for niche occupation³³—a process with substantial implications for disease outcomes during recurrent infections.³³ Collectively, our spectral flow cytometry panel not only identifies these metabolic signatures but also reveals the key metabolic drivers that underpin immune cell functions such as innate immune memory, offering deeper insights into the metabolic programs that regulate myeloid cell behavior.

The use of spectral flow cytometry demonstrated that autofluorescence can uncover metabolic properties of immune cells. Stimulating glycolysis in vitro with LPS induced a glycolytic-specific autofluorescence signature in bone marrow cells, with autofluorescence intensity associated with increased expression of the glycolytic enzyme GAPDH. This association between autofluorescence and glycolysis is consistent with the emission profile of NAD(P)H. Typically, the autofluorescence signal originating in the cell cytoplasm is almost entirely attributable to NAD(P)H and flavins.^34,35 These findings build on established methods of evaluating cell metabolism by NAD(P)H, as both glycolysis and oxidative phosphorylation can be approximated by fluorescence-lifetime methods in imaging microscopy.⁸ This technique has identified metabolic changes in macrophage polarization^24,25,26 and T cell activation.³⁶ In addition, glioma stem cells can be sorted and functionally defined by their NAD(P)H autofluorescence profiles,³⁷ where cells with higher NAD(P)H autofluorescence had a higher capacity for multilineage differentiation, tumorigenesis, and invasive ability,³⁷ consistent with more aggressive tumor cells being highly glycolytic.³⁸ Here, we provide a simple method that harnesses the sensitivity of spectral cytometers to detect NAD(P)H as a glycolysis indicator by autofluorescence. This simple and convenient method obviates the need to stain cells with antibodies to measure this pathway. However, the correlation between GAPDH and autofluorescence should be rigorously examined in each experimental setting as a baseline, with the consideration that different organs and experimental conditions may induce molecules with different autofluorescence properties. This in itself may open the way for discovery and future use of other autofluorescent cellular components. By combining different detected autofluorescent signatures enabled by the increasing sensitivity of spectral cytometry, this method can provide a wealth of information about the metabolic status of a cell without the need for additional labeling and manipulation. This approach not only preserves the natural state of the cell, but also offers insights into its functional identity that can potentially be further tested by sorting, cell culture and adoptive transfer experiments.

Overall, we present a systematically validated metabolic panel that can be used to investigate the relationship between phenotypic identity and metabolic profile at the single-cell level. This versatile research tool will likely garner significant insights into immune cell biology but also provides a powerful, standardized framework for dissecting the metabolic basis of immune cell function in diverse contexts, paving the way for broader applications in immunometabolism research and therapeutic development.

Limitations of the study

A key limitation of our validation strategy is the reliance on pharmacological inhibitors rather than genetic knockout or knockdown approaches. Our rationale for using metabolic inhibitors was to acutely and reversibly modulate metabolic pathways to observe immediate, detectable changes in target protein expression by spectral flow cytometry. This approach mirrors those used in Seahorse assays, a gold standard for metabolic analysis, and allows for temporal precision without the compensatory adaptations associated with long-term genetic manipulation. Nonetheless, we acknowledge that pharmacological inhibitors may have off-target effects and that their use provides indirect evidence of pathway activity.

While genetic knockout models are commonly used for pathway validation, the majority of our metabolic targets are essential for embryonic development or cellular viability, precluding their use in our study. For instance, knockout of HIF-1α,³⁹ CPT1A,⁴⁰ ACAC (ACC1),⁴¹ CD98,⁴² or cytochrome c⁴³ each results in embryonic lethality, and GAPDH knockdown induces apoptosis in metabolically active cells,⁴⁴ complicating interpretation. Although we opted not to use siRNA in this study, it offers certain advantages that merit consideration in future work. siRNA-mediated knockdown enables gene-specific suppression and can be used to confirm the functional relevance of individual enzymes in metabolic pathways. In contexts where the target protein is non-essential or where long-term modulation is desirable, siRNA may provide greater specificity and mechanistic insight than small-molecule inhibitors. Nevertheless, for the purpose of validating flow cytometry-based detection of metabolic proteins, siRNA knockdown presents challenges, including reduced protein expression below antibody detection thresholds and variable silencing efficiency, as well as attendant unknown potential impacts on other metabolic pathways in cells undergoing siRNA treatment that limit its utility for confirming cytometry-based measurements.

In addition, our metabolic panel provides an indirect measurement of metabolic pathway activity rather than direct pathway flux. Complementary approaches, such as metabolic probes and fluorescent analogs (e.g., BODIPY for lipid storage, MitoTracker for mitochondrial mass, and MitoSOX for mitochondrial reactive oxygen species), could enhance metabolic assessments in flow cytometry-based studies. Furthermore, techniques such as extracellular flux assays or SCENITH could be used to directly measure metabolic flux, providing real-time insights into pathway activity.

The metabolic panel developed in this study has several limitations compared to previously established approaches. Using CyTOF, scMEP offers a similar advantage to spectral flow cytometry in that it enables highly parametric use of markers and thus can assess a greater number of metabolic proteins, although at a substantially greater cost, both in reagents and extended run times. Additionally, MetFlow, using 10 markers, investigates some metabolic pathways not covered in this study, including the pentose phosphate pathway, and incorporates the assessment of post-translational modifications, such as phosphorylation status, which enables monitoring of signaling pathways that play vital roles in regulating immunometabolism like mTOR and AMPK. However, the use of spectral cytometry in this work permits expansion to include additional metabolic markers that would allow more detailed investigation of major pathways not covered by our original panel.

Beyond marker selection, our approach has additional technical constraints. Two key metabolic targets, CPT1A and ACAC, necessitate secondary antibody labeling steps that extend staining time and require additional controls, introducing constraints on panel design (detailed in Methods S1). Furthermore, while we prioritized antibodies with mouse-human cross-reactivity, not all antibodies in our panel are cross-reactive with human samples. Species-specific metabolic differences must also be considered for translational applications; for example, iNOS expression is minimal in human macrophages,^45,46,47 limiting its utility for assessing L-arginine metabolism in human studies. In addition, human monocyte-derived macrophages preferentially engage oxidative phosphorylation rather than glycolysis following LPS stimulation, unlike their murine counterparts.⁴⁸ These species-specific metabolic divergences, while presenting challenges for translational research, nevertheless provide unique opportunities to investigate how metabolic programs have evolved to support distinct immune functions across species.

Finally, a limitation of our approach is the inability of spectral flow cytometry to distinguish between NADH and NADPH, as both cofactors emit nearly identical autofluorescence signals. While NADH is primarily associated with glycolysis and oxidative phosphorylation, NADPH plays a central role in biosynthesis and redox balance. Consequently, the detected NAD(P)H signal represents a combined pool of both metabolites. Despite this, the consistent correlation between NAD(P)H autofluorescence, GAPDH expression, and glycolytic manipulation in our study supports the interpretation that changes in signal intensity primarily reflect alterations in glycolytic activity. Nonetheless, future refinement, such as fluorescence lifetime-based approaches, may help resolve specific contributions of each cofactor.

Resource availability

Lead contact

Further information and request for resources and reagents should be directed to and will be fulfilled by the lead contact, Nicholas J.C. King (nicholas.king@sydney.edu.au).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data: All study data are included in the article and/or supplemental information.
• •Code: This paper does not report previously unpublished custom code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Acknowledgments

This work was supported by the Australian National Health and Medical Research Council (1088242), and a Scholarship to C.L.W. from the Merridew Foundation. We would like to thank Dr Mark Cantwell from Memgen, Inc. (TX, USA) for generously providing MemVax for this work, and Kate Pilkington for her expertise on autofluorescence extraction and analysis of spectral flow cytometry data. We also wish to acknowledge the support of the University of Sydney’s Laboratory Animal Services and the Sydney Cytometry facilities, in addition to the animal facility at the Centenary Institute. Illustrations were created with BioRender.

Author contributions

Conceptualization, C.L.W., A.G.S., J.T., N.J.C.K., and L.M.; data curation, C.L.W. and C.C.; investigation, C.L.W.; formal analysis, software, and visualization, C.L.W.; funding acquisition and supervision, N.J.C.K., L.M., and J.A.T.; writing – original draft, C.L.W.; and writing – review and editing, C.L.W., A.G.S., J.T., and N.J.C.K.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Antibodies for flow cytometry	This paper (Table 2)	N/A
Chemicals, peptides, and recombinant proteins
Pharm Lyse™ Lysing Buffer	Invitrogen	555899
Ack Lysing Buffer	Gibco	A1049201
HEPES	Sigma-Aldrich	SRE0065
L-Glutamine	Thermo Fisher Scientific	25030081
D-(+)-Glucose	Sigma-Aldrich	G8270-100G
2-deoxy-D-glucose	Sigma-Aldrich	D8375-5G
Lipopolysaccharides from Escherichia coli K12	InvivoGen	NC9355258
Purified interferon gamma	Biolegend	575308
Purified Iinterleukin-4	Miltenyi Biotec	130-097-760
Dimethyloxalylglycine (DMOG)	Sigma-Aldrich	D3695-10MG
Chrysin	Abcam	ab141230-1g
Oligomycin	Sigma-Aldrich	75351-5MG
Carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP)	Sigma-Aldrich	C2920-10MG
Palmitic acid	Sigma-Aldrich	P0500-10G
Bovine serum albumin (BSA)	Sigma-Aldrich	A7906-100G
5-(Tetradecyloxy)-2-furoic acid (TOFA)	Sigma-Aldrich	T6575-5MG
C75	Sigma-Aldrich	C5490-5MG
L-leucine	Sigma-Aldrich	L8912-25G
L-isoleucine	Sigma-Aldrich	I7403-25G
L-valine	Sigma-Aldrich	V0513-25G
L-histidine	Sigma-Aldrich	H5659-25G
L-methionine	Sigma-Aldrich	M5308-25G
L-phenylalanine	Sigma-Aldrich	P5482-25G
L-tyrosine	Sigma-Aldrich	T8566-25G
L-tryptophan	Sigma-Aldrich	T8941-25G
Lymphoprep™	StemCell Technologies	18061
MojoSort™ Mouse anti-PE Nanobeads	Biolegend	480079
Dynabeads™ Mouse T-Activator CD3/CD28 beads	Gibco	11456D
Critical commercial assays
True-Nuclear™ Transcription Factor Buffer Set	Biolegend	424401
NAD/NADH Assay Kit	Abcam	Ab65348
Experimental models: Organisms/strains
C57BL/6	Animal BioResources, NSW
Software and algorithms
SpectroFlo	Cytek Biosciences	https://cytekbio.com/pages/spectro-flo
FlowJo v10.10.0	TreeStar	https://www.flowjo.com
GraphPad Prism software version 10.1.1	GraphPad	https://graphpad.com
R Studio	Posit	https://posit.co/download/rstudio-desktop/
Spectre	Immune Dynamics49	https://immunedynamics.io/spectre/
Other
Ultra-low adherence 96-well plates	Corning	3474
SepMate™ PBMC Isolation Tubes	StemCell Technologies	85450
RPMI 1640 Medium with L-glutamine	Lonza	12-702Q
DMEM, no glucose	Sigma-Aldrich	D5030-10X1L

Experimental model and subject details

PBMC isolation

For the isolation of PBMC, blood samples were collected into K2-EDTA collection tubes from ten healthy adult male and female donors. Ethical approval was granted by the Ethics Review Committee (Royal Prince Alfred Hospital Zone) of the Sydney Local Health District (Protocol Number X17–0130 & HREC/17/RPAH/192) and conducted in accordance with the National Statement on Ethical Conduct in Human Research (NHMRC, 2007, updated 2018). PBMC was isolated using the SepMate PBMC Isolation Tubes (StemCell Technologies), according to the manufacturer’s instructions. Briefly, blood was diluted 1:1 with PBS (15 mL each) and carefully layered over Lymphoprep density gradient medium (StemCell Technologies) in a SepMate tube and centrifuged for 1200 x g for 10 min at room temperature with brakes. The PBMC-containing upper layer was collected and washed twice with PBS containing 2% FBS: first at 400 × g for 8 min, then at 200 × g for 10 min to remove platelets. Cell pellets were resuspended in PBS/2% FBS for counting, then centrifuged at 400 × g for 8 min at 4°C. Finally, cells were resuspended in FBS containing 10% DMSO at a concentration of 1×10⁷ cells/mL.

Mice

Female 9 to 10-week-old C57BL/6 mice were obtained from Animal BioResources (NSW, Australia). All experiments involving mice were approved by the University of Sydney Animal Ethics Committee (Protocol 1696) or Sydney Local Health District (SLHD) Animal Ethics and Welfare Committee (2020/007 SLHD) and conducted in accordance with the Australian Code for the Care and Use of Animals for Scientific Purposes (NHMRC, 2013, updated 2021). Mice maintained specific pathogen-free at the Charles Perkins Centre (Sydney, Australia) or Centenary Institute Bioresources animal facility (Sydney, Australia) under Biosafety Level (BSL) II or III conditions. Female mice were used to facilitate compatibility with cage mates. Mice were group-housed in individually ventilated cages at approximately 21°C and a relative humidity of 45–46% on a 12 h day/night light cycle with ad libitum access to standard rodent chow and water. Sterile cardboard and shelter were provided for environmental enrichment. Mice were randomly assigned to experimental groups, and order of treatments was randomised to minimise potential confounders. Procedures were not performed until after a week of acclimatisation.

Primary cells

Primary splenocytes and bone marrow cells were isolated from female 9-10-week-old C57BL/6 mice obtained from Animal BioResources (NSW, Australia), in accordance with the relevant ethics and maintenance/care guidelines as described above. Primary cells were isolated according to the procedures described below.

Method details

Mouse treatments

For all vaccinations, mice were anaesthetised with gaseous isoflurane (3–4%, O₂ 1 L/min) to reduce distress. Mice were treated with MemVax (IFSD35), a replication-deficient adenovirus serotype 5 (Ad5) vector expressing a membrane-bound chimeric CD40L. Animals received two intranasal vaccinations, administered 14 days apart, consisting of 20 μL of phosphate-buffered saline (PBS) containing 1x10¹⁰ plaque-forming units of MemVax. Control mice received equivalent volumes of PBS alone via the same route. Following vaccination, mice underwent an 8-week resting period before being humanely euthanized for tissue collection.

For intravascular staining of leukocytes, mice were administered 3 μg of biotin-conjugated anti-CD45 monoclonal antibody (catalogue #103104, BioLegend, San Diego, CA, USA) in 200 μL of PBS via lateral tail vein injection 3 min prior to euthanasia. The detection of biotin-labelled cells was performed using PE-Cy5.5-conjugated streptavidin (BioLegend).

Tissue isolation and processing

The bone marrow (femur and tibia) or spleen were collected from mice deeply anesthetized with an intraperitoneal injection of avertin, followed by thoracotomy and transcardial phosphate-buffered saline (PBS) perfusion. Cells from the bone marrow were isolated by flushing the femur and tibia with PBS using a 30-gauge needle. Spleens were gently mashed through a 70 μM nylon sieve using 1 mL of PBS. Red blood cells were then lysed using 1x Pharm Lyse Buffer (Invitrogen).

For vaccination experiments, mice were not perfused prior to sacrifice to allow for the analysis of circulating leukocytes in the lung. Murine lungs were collected and mechanically dissociated with the GentleMACS™ dissociator (Miltenyi Biotec) before enzymatic digestion with DNAse I (10 U/mL; Sigma-Aldrich) and Collagenase IV (10 U/mL; Sigma-Aldrich) for 45 min at 37°C. Single-cell suspensions were prepared by filtering through a 70 μM nylon cell strainer and red blood cells were lysed with ACK lysis buffer (Gibco) before washing and resuspension in RPMI 1640 (Life Technologies, ThermoFisher Scientific) supplemented with 10% FCS (Sigma-Aldrich), 2-ME (0.05 mM), and penicillin-streptomycin (100 U/mL; Sigma).

Primary cell culture

Primary splenocytes and bone marrow cells were derived from from female 9-10-week-old C57BL/6 mice. In all conditions, media was supplemented with 10 mM of 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES, Sigma-Aldrich), 2 mM glutamine (Thermo Fisher Scientific), 1% FCS, and 2 mg/mL glucose, and the cells were incubated at 37°C with 20% oxygen unless otherwise stated.

T cell isolation and activation

Murine T cells were enriched from splenocytes by negative selection using magnetic beads with MojoSort Mouse anti-PE Nanobeads (BioLegend). Briefly, 10⁸ splenocytes were stained in 300 μL of a PE-conjugated antibody cocktail containing anti-B220, anti-CD11b, anti-Ly6G, anti-CD11c, and anti-NK1.1 (Table 2) in 1X MojoSort Buffer for 15 min. After washing, samples were resuspended in 100 μL of 1X MojoSort buffer and 10 μL of anti-PE nanobeads were added to each sample and incubated for 20 min. After washing, T cells were isolated by negative selection in a magnetic separator (Miltenyi Biotec, Germany) through 30 μM columns (Miltenyi Biotec, Germany) and washed three times with 1X MojoSort Buffer. For T cell activation, 5 × 10⁵ T cells were plated in triplicate in a U-bottom, 96-well tissue culture-treated plate (Corning) and cultured with anti-CD3/anti-CD28 beads (Dynabeads, Gibco) at a 1:1 cell-to-bead ratio for 4, 24 and 48 h in complete RPMI (Lonza Biosciences).

Macrophage activation and inhibitor treatments

For the NADH assay, a total of four mice were pooled per sample (8x tibia and 8x femurs). Bone marrow cells were plated at a density of 5x10⁵ cells/well in quadruplicate in flat-bottom, ultra-low adherence 96-well plates (Corning) and stimulated with LPS (100 ng/mL) and IFN-γ (100 ng/mL), IL-4 (40 ng/mL), or PBS for 16 h.

For all other experiments, mouse bone marrow cells were plated at a density of 1.2 x 10⁶ cells per well in flat-bottom, ultra-low adherence 96-well plates (Corning). Mouse bone marrow cells were treated either with IL-4 (40 ng/mL), or LPS (100 ng/mL) and IFN-γ (100 ng/mL), to induce alternative or classical macrophage activation, respectively, or PBS for 16 h as an unstimulated control. For glycolysis and inducible nitric oxide synthase (iNOS) experiments, bone marrow cells were pre-treated with 10 mM 2-Deoxy-D-glucose (2-DG, Sigma-Aldrich) or PBS for 1 h. The cells were then cultured with 100 ng/mL IFN-γ and 100 ng/mL LPS (InvivoGen) (Biolegend, USA) with 10 mM 2-DG or PBS for 4 h (glycolysis) or 16 h (iNOS). Alternatively, cells were cultured in custom glucose-free media with or without 2 mg/mL glucose and 100 ng/mL LPS or PBS as a vehicle control. To validate hypoxia inducible factor-1α (HIF-1α) signalling, bone marrow cells were cultured in media supplemented with either 0.2 mM Dimethyloxalylglycine (DMOG, Sigma-Aldrich) or PBS in 3% oxygen, or 0.01 mM Chrysin (Abcam) or PBS in 20% oxygen. For TCA cycle experiments validating isocitrate dehydrogenase 2 (IDH2), bone marrow cells were stimulated with IL-4 (40 ng/mL) and cultured with 2 μM oligomycin (Sigma-Aldrich) or vehicle control overnight. To validate cytochrome C as a marker of the electron transport chain, bone marrow cells were stimulated with IL-4 (40 ng/mL) cultured with 1 μM carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, Sigma-Aldrich) with or without 2 μM oligomycin (Sigma-Aldrich) overnight. For experiments involving fatty acid metabolism, 2 mM palmitic acid (Sigma-Aldrich) was conjugated with 20% BSA (Scientifix Australia) before use. To validate carnitine palmitoyltransferase 1A (CPT1A) as a marker for fatty acid oxidation, bone marrow cells were cultured with 167 μM palmitate conjugated to bovine serum albumin (BSA) or BSA alone, ±10 mM 2-DG, and ±3 μM etomoxir or vehicle control for 16 h. In some experiments, CPT1A carnitine palmitoyltransferase 1A (CPT1A) expression was assessed in splenocytes cultured in glucose-free Dulbecco’s Modified Eagles Medium (DMEM, Sigma-Aldrich) supplemented with palmitate conjugated to BSA (0.175 mM) and/or 4.5 g/L glucose. To validate acetyl-coA carboxylase (ACAC) as a marker of fatty acid synthesis, bone marrow cells were cultured with the indicated concentrations of either C75 (Sigma-Aldrich) or 5-(Tetradecyloxy)-2-furoic acid (TOFA, Sigma-Aldrich) with 100 ng/mL LPS and 100 ng/mL interferon-gamma (IFN-γ) for 16 h. To validate CD98 as a marker of amino acid uptake, bone marrow cells were cultured with 100 ng/mL LPS in custom RPMI media free of large neutral amino acids or supplemented with large neutral amino acids at normal RPMI levels: L-leucine (0.3817 mM), L-isoleucine (0.3817 mM), L-valine (0.1709 mM), L-histidine (0.097 mM), L-methionine (0.1006 mM), L-phenylalanine (0.0909 mM), L-tyrosine (0.1111 mM), and L-tryptophan (0.0245 mM) (all from Sigma-Aldrich).

To detect metabolic changes in myeloid cells following vaccination, 5x10⁵ lung cells were plated in duplicate in flat-bottom, ultra-low adherence 96-well plates (Corning) with 100 ng/mL of LPS for 16 h.

Spectral flow cytometry

Single-cell suspensions were Fc receptor-blocked with anti-CD16/32 (Biolegend, USA) and stained with Zombie UV Fixable Viability kit (Biolegend, USA) in PBS for 30 min on ice. After washing, cells were stained with a cocktail of fluorescently-labelled surface-staining antibodies in FACS buffer on ice for 30 min (Table 2). Following this, cells were washed twice and permeabilized with True-Nuclear 1x Fix Concentrate (Biolegend, USA) for 45 min at room temperature prior to staining with transcription factor or intracellular antibodies for 30 min on ice (Table 2). For antibodies targeting ACAC and CPT1A, cells were stained with anti-ACAC and ant-CPT1A for 30 min on ice during the intracellular stain, washed three times, and then stained with secondary antibodies 30 min on ice.

Optimal antibody concentrations for metabolic targets were determined empirically by titrating each antibody and assessing signal-to-noise ratio using an isotype control (Figure S1). Since the selected antibodies often did not exhibit a clear bimodal separation between positive and negative populations, the isotype control was used to define background staining. The optimal concentration was chosen based on two criteria: (1) maximal separation between the isotype control and antibody-stained population, and (2) minimal background staining in the negative population, ensuring it remained close to baseline fluorescence levels.

The 5-Laser Aurora spectral cytometer (Aurora 5L, Cytek Biosciences, USA) was used to acquire stained cells. Unstained controls for each specific condition were used as reference controls for spectral unmixing in each experiment. Autofluorescence extraction was applied to each data set during spectral unmixing. The data acquired was analysed using FlowJo (v10.8, BD Biosciences, USA). Quality control measures such as time, single cells, non-debris, and Live/Dead staining were applied to exclude debris, doublets, and dead cells.

Analysis of spectral flow cytometry data

The FCS files were compensated in FlowJo and gated down to the indicated populations. Dimensionality reduction and clustering of lung CD8⁺ T cells was performed using Spectre.⁴⁹ Total CD8⁺ T cells were gated and the median fluorescence intensity of markers was exported from FlowJo. Clustering was performed on metabolic targets only.

To examine changes in autofluorescence, unstained, raw data was extracted from unstained samples from M1- (100 ng/mL LPS +100 ng/mL IFN-γ), or M2- (40 ng/mL IL-4, or PBS) stimulated bone marrow cells or M1-stimulated (100 ng/mL LPS +100 ng/mL IFN-γ) bone marrow cells treated with 10 mM of 2-DG (Sigma-Aldrich, USA), using the Aurora 5L (Cytek Biosciences, USA). The fluorescence of NADH was measured by exciting cells with the 355 nm UV laser and the emission was detected at the UV5 channel (centre wavelength of 458 nm).

NAD/NADH assay

Cellular NADH levels were quantified using a commercial NAD/NADH Assay kit (Abcam ab65348). After culture, technical replicates were combined and recounted using trypan blue exclusion and a hemocytometer. A total of 1x10⁶ cells were collected as described above and suspended in 400 μL of NADH/NAD Extraction Buffer. The cell extracts were filtered through a 10 kD Spin Column (ab93349, Abcam) by centrifugation at 10,000 x g for 40 min at 4°C. The filtered sample was split into two equal portions. To measure total NAD, one portion was retained unmodified. To measure NADH alone, the other portion was heated at 60°C for 30 min to degrade NAD⁺. From each portion, 20 μL was combined with 30 μL Extraction Buffer and 100 μL Reaction Mix, then incubated at room temperature for 5 min. NADH Developer (10 μL) was added to each sample and mixed. The reaction proceeded at room temperature for 20 min, after which absorbance was measured at 450 nm using a microplate reader in kinetic mode.

Quantification and statistical analysis

The FCS files were compensated and gated down to individual cell populations prior to exporting cell proportions and MFI in FlowJo. Cell proportions and live cell counts were used to quantify cell numbers. The positive population for each metabolic target, as determined by the isotype and/or fluorescence-minus-one control (FMO), were used to quantify the change in MFI of each of metabolic target. Statistical analyses were carried out in Prism 10.1.1 (GraphPad). For cell culture experiments, statistical significance was computed by paired t-tests (two groups) or RM one-way ANOVA with Geisser-Greenhouse correction and Tukey’s multiple comparisons tests (three or more groups). For mouse experiments, statistical significance was computed by unpaired t-tests (two groups) or ordinary one-way ANOVA (three or more groups). Significance is indicated as follows: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Error bars represent mean ± SD.

Published: June 13, 2025