Immuno-metabolic dendritic cell vaccine signatures associate with overall survival in vaccinated melanoma patients

The study addresses why dendritic cell (DC)-based cancer vaccines often lack durable clinical efficacy and whether DC metabolic states underlie variability in immunogenicity and clinical outcomes. DCs orchestrate innate and adaptive immunity, but in cancer their function may be compromised. Prior work suggests murine DC activation depends on a switch to glycolysis, whereas human DCs show more complex, context-dependent metabolic wiring. Emerging single-cell methods enable resolving heterogeneous DC states not captured by bulk assays. The authors hypothesize that immuno-metabolic phenotypes of ex vivo–generated DC vaccines and of their circulating myeloid precursors in melanoma patients are altered compared with healthy donors, and that these metabolic features associate with immune phenotype and patient survival after DC vaccination. The purpose is to integrate transcriptomics, bulk functional metabolism, and single-cell functional and proteomic metabolic profiling to identify metabolic biomarkers predictive of DC function and clinical outcome in vaccinated melanoma patients.

Background studies demonstrate: (1) murine bone marrow–derived DCs undergo a shift from OXPHOS to glycolysis upon TLR activation and antigen presentation, but this does not directly translate to human DCs, where metabolic programs differ across inflammatory, tolerogenic, and lineage-defined DC states; (2) diverse metabolic pathways, including OXPHOS, glycolysis, and fatty acid oxidation (FAO), regulate DC survival, maturation, and T-cell priming; (3) tolerogenic DCs can exhibit metabolic hyperactivity with coordinated upregulation of multiple pathways, linked to altered mTOR/AMPK signaling; (4) bulk transcriptomics and extracellular flux analyses (OCR/ECAR) provide insights but cannot resolve cellular heterogeneity; (5) single-cell tools such as SCENITH and scMEP allow functional and protein-level metabolic profiling at single-cell resolution. Prior work by the authors mapped DC differentiation/maturation states to metabolic programs, identifying elevated phospho-mTOR:AMPK ratio with upregulated OXPHOS, glycolysis, and FAO in tolerogenic DCs. This foundation supports investigating immuno-metabolic phenotypes in patient-derived DC vaccines and circulating myeloid/DC precursors in melanoma.

Design and participants: Specimens were obtained from a Phase I, single-site clinical study (NCT01622933) of autologous DC vaccines in 35 subjects with recurrent unresectable stage III/IV melanoma or resected stage IIIB-C/IV melanoma. Endpoints included toxicity, immunogenicity, and clinical response. Enrollment occurred 09/2012–11/2015 with IRB approval (UPCI #09-021) and informed consent.

DC generation: Leukapheresis products were elutriated into myeloid and lymphoid fractions. Immature DC (iDC) were generated by culturing myeloid cells for 5 days in GM-CSF (1000 U/mL) and IL-4 (1000 U/mL) in DC medium, then matured for 24 h with IFN-γ (1000 U/mL) and LPS (250 ng) to yield mature DC (mDC). Some assays used Dorsomorphin during maturation to inhibit p-AMPK.

Transcriptomics: Total RNA from iDC, mDC, and vaccine DC was profiled on Affymetrix HUGENE 2.0 ST arrays. Differential expression used limma/voom with FDR 0.05 and |log2FC|≥2. Pathway analysis used gProfiler and GSEA (MSigDB C2). GSVA assessed sample-level enrichment.

Metabolic flux assays: Seahorse XFe96 measured OCR (mitochondrial respiration) and ECAR (glycolysis) with sequential additions of oligomycin, FCCP, rotenone/antimycin A, and 2-deoxyglucose. FAO was assessed using palmitate-BSA (XF Palmitate Oxidation Stress Test). Derived parameters included basal respiration, maximal OCR, spare respiratory capacity, proton leak, ATP-linked respiration, basal glycolysis, and glycolytic capacity.

Single-cell functional metabolism (SCENITH): Cultured mDC, elutriated fraction cells, or PBMCs were treated with inhibitors (2-DG, oligomycin, etomoxir, CB-839, DGO), pulsed with puromycin, and stained for puromycin and phenotypic markers by spectral flow cytometry. Percentual parameters: glucose dependence, mitochondrial dependence, FAO dependence, glutaminolysis dependence, and glycolytic capacity were computed from puromycin MFI under inhibitor conditions.

Single-cell metabolic regulome (scMEP): Mass cytometry (CyTOF) quantified metabolic enzymes, transporters, and signaling proteins across OXPHOS, glycolysis, FAO, glutaminolysis, and biosynthesis pathways, along with immune phenotypes, in iDC and mDC. Data were arcsinh transformed and analyzed by clustering, PCA/tSNE/UMAP, and differential expression.

Secretome and metabolite assays: DC culture supernatants were assayed by Thermo-Fisher ProcartaPlex 65-plex cytokine/chemokine panel and 14-plex checkpoint/costimulatory panel. Glucose and lactate in supernatants were measured via handheld meters.

T-cell immunogenicity: IFN-γ ELISPOT measured CD8 and CD4 responses to vaccine antigens (tyrosinase, MART-1, MAGE-A6); positivity defined as >10 spots and ≥2× over baseline after AdVLacZ subtraction.

Circulating myeloid/DC profiling: SCENITH combined with a high-dimensional immunophenotyping panel to define monocyte subsets (classical cMo, intermediate iMo, non-classical ncMo) and DC subsets (pDC, pre-DC, cDC1, CD5+ and CD5− cDC2, CD14− and CD14+ DC3). Frequencies, metabolic parameters, and marker expression were compared between healthy donors (HD) and melanoma patients and associated with outcomes.

Statistics: Group comparisons used ANOVA with Tukey’s post-hoc or t/Wilcoxon tests as appropriate. Survival analyses used Kaplan–Meier and Cox proportional hazards. Maximally selected rank statistics determined cutoffs for SCENITH parameters. Data processing used established R packages for cytometry and transcriptomics. Public data: HD microarray GSE111581; melanoma GSE157738; code/data on GitHub.

• Transcriptomics: Melanoma patient mDC vs HD mDC showed extensive differential expression (HD: 507 up, 221 down mDC vs iDC; melanoma: 1008 up, 1069 down; patient vs HD comparison: 725 up and 818 down specific to melanoma mDC). Pathways in HD mDC upregulated included MHC class I antigen processing/presentation and CCR5 binding; melanoma mDC showed upregulated VEGFA, TGFβ receptor, NLRP3 inflammasome and Oncostatin M signaling, with downregulated antigen processing and PRR activity. GSEA indicated selective downregulation of TCA/ETC/OXPHOS in HD mDC and FA/phospholipid metabolism and PPAR pathway in melanoma mDC. Between clinical outcome groups, favorable outcomes were associated with LPS/inflammatory response, NF-κB targets, DC maturation, VEGF/Hypoxia, APC/MHC/Interleukin, Matrisome/Integrins, and FAO/sphingolipid metabolism; worse outcomes had upregulated DNA repair, TCA/ETC, mRNA processing, interferon signaling, and Golgi-ER transport/glycosylation. A 57-gene signature was elevated in good-outcome mDC (e.g., IL1A, CCL24, CXCL6, IFNG, MMP1/9/10/12).

• Seahorse metabolic profiling: Melanoma mDC exhibited higher basal glycolysis (ECAR) in both outcome groups and increased glycolytic capacity significantly in the bad-outcome mDC. FAO capacity (palmitate-stimulated OCR) was significantly reduced in melanoma mDC vs HD. Proton leak increased stepwise from HD to good to bad outcome groups; ATP-linked respiration was significantly decreased in bad outcome mDC, consistent with mitochondrial dysfunction.

• SCENITH functional metabolism: Mitochondrial dependence was the dominant pathway in mDC but trended lower in bad outcomes (84.4% to 76.4%), with corresponding increase in glycolytic capacity (15.6% to 23.6%) and a 7% decrease in glutaminolysis dependence. HD DC trended toward higher mitochondrial and glutaminolysis dependence and lower glycolytic capacity vs melanoma mDC. Dichotomized SCENITH parameters associated with survival: higher mitochondrial dependence significantly associated with longer OS (Cox p=0.0296) and PFS (p=0.0032); FAO and glutaminolysis dependence showed trends for benefit (p≈0.08–0.13). KM analyses confirmed benefits for high mitochondrial dependence (OS log-rank p=0.023) and trending benefits for FAO and glutaminolysis (p≈0.07). Trends suggested higher mitochondrial/FAO dependence linked to stronger antigen-specific T-cell responses, though not significant.

• Immune phenotype and metabolic states: Worse-outcome mDC exhibited higher expression of HLA-DR, CD86, CD206, CD40, and ILT3 within highly glycolytic states. Oligomycin-based SCENITH binning showed a higher proportion of highly glycolytic cells in worse responders. p-AMPK and p-mTOR increased in glycolytic cells, with a higher p-AMPK:p-mTOR ratio in mitochondrial-dependent cells. Pharmacologic p-AMPK inhibition (Dorsomorphin) reduced HLA-DR, CD86, PD-L1, and CD206 expression, decreased mitochondrial mass, reduced glucose uptake, and decreased lactate levels without affecting viability.

• scMEP metabolic regulome: HD mDC expressed higher OXPHOS/TCA markers (CytC, ATP5A) and glutaminase (GLS) vs melanoma mDC, consistent with functional assays. PGC1α (mitochondrial biogenesis) trended down in melanoma mDC, aligning with increased proton leak. GSS (glutathione synthesis) was lower in worst-outcome mDC versus HD, potentially contributing to ROS and proton leak. MCT1 (lactate transporter) was increased in melanoma mDC; HADHA (β-oxidation) trended down. PD-L1 was overall downregulated in melanoma mDC vs HD in scMEP. Melanoma DC consumed more glucose and secreted more lactate; increased lactate secretion by iDC significantly correlated with inferior OS (KM).

• Secretome associations: Patients with progressive disease secreted the lowest levels of many analytes. Higher secretion of multiple cytokines/chemokines associated with positive outcomes. IL-12p70 was generally low and did not correlate with outcomes. Several analytes were highest in HD then good then bad outcomes (e.g., CXCL13, eotaxin, IL-23, IL-31, IL-5, MCP-1, MIG, sCD40L, TIM3, TRAIL). IFN-α, IL-18, IL-1α, IL-21 were strong in HD and good but reduced in bad outcomes. SCENITH glucose dependence inversely correlated with secretion of IDO, BTLA, and GITR; LAG3 inversely correlated with glycolytic capacity; IP-10 positively correlated with maximal OCR and inversely with APRIL, TNFβ, IL-23, IL-27.

• Circulating myeloid/DC compartments: Melanoma patients had increased pDC and decreased CD5+ cDC2 and CD14− DC3 frequencies; cDC1 frequency unchanged; classical and intermediate monocytes showed trends (cMo decreased, iMo increased). PCA separated HD vs melanoma, notably in monocyte and pDC clusters. PD-L1 and CD36 were higher in some HD populations; ILT3 was upregulated in iMo and cDC1s in non-responders, while PD-L1 and CD206 were down in iMo. Cox models: higher ILT3 expression on selected monocyte/DC subsets associated with decreased OS; CD11c on monocytes was protective; PD-L1 expression on conventional DC subsets associated with improved PFS. SCENITH of circulating cells showed decreased mitochondrial dependence (and trends for lower FAO/glutaminolysis) in cMo and ncMo in melanoma; cDC1 and CD14+ DC3 had reduced glucose dependence; cDC2 subsets had decreased mitochondrial dependence. Glutaminolysis dependence was broadly reduced across DC subsets in melanoma, especially in worse prognosis. iMo metabolic profiles correlated best with cultured mDC metabolic parameters (FAO dependence significantly correlated).

The study demonstrates that DC metabolic state is tightly linked to vaccine DC immune phenotype and clinical outcomes in melanoma. Patient-derived mDC display elevated glycolysis, reduced FAO, increased proton leak, and decreased ATP-linked respiration, indicating mitochondrial bioenergetic dysfunction compared to HD. Single-cell functional profiling reveals that higher mitochondrial dependence in vaccine DC associates with longer OS and PFS, while higher glycolytic capacity is enriched in worse outcomes and correlates with aberrant overexpression of activation and inhibitory markers in more heterogeneous cell states. These findings address the central question by identifying mitochondrial dependence as a marker of effective maturation and functional competency of human DCs, challenging oversimplified extrapolations from murine models where glycolysis dominates activation. The scMEP data connect functional metabolism with protein-level regulome alterations (reduced CytC/ATP5A/GLS/PGC1α; increased MCT1) and with increased lactate secretion, which predicts poor survival. Baseline metabolic skewing in circulating myeloid/DC subsets in patients suggests that the in vivo disease state imprints precursor cells used for vaccine manufacture, potentially limiting vaccine efficacy. Transcriptomic signatures only partially reflected functional metabolism, highlighting the necessity of functional single-cell approaches. Overall, the results support metabolism—particularly mitochondrial dependence and lactate handling—as actionable biomarkers for DC vaccine quality and as potential targets to optimize vaccine manufacturing and patient selection.

This work integrates transcriptomics, extracellular flux, single-cell functional metabolism, and metabolic proteomics to show that immuno-metabolic features of DC vaccines and their precursors strongly associate with clinical outcomes in melanoma. Key contributions include: identification of high mitochondrial dependence (and FAO/glutaminolysis reliance) as favorable DC vaccine features linked to longer OS/PFS; demonstration that elevated glycolysis, increased MCT1 expression, and lactate secretion mark dysfunctional DC states and predict poorer survival; and evidence that circulating myeloid/DC compartments in melanoma are metabolically skewed prior to vaccine generation. These findings advocate for incorporating metabolic profiling (e.g., SCENITH, scMEP, lactate assays) into DC vaccine quality control and for developing metabolically informed culture conditions to enhance mitochondrial fitness and metabolic flexibility. Future research should: (1) validate metabolic biomarkers prospectively and across larger, independent cohorts; (2) mechanistically dissect how manipulating AMPK/mTOR, amino acid metabolism, and lactate transport impacts DC function and vaccine efficacy; (3) optimize ex vivo culture conditions to restore balanced mitochondrial metabolism; (4) evaluate lactate secretion as a practical potency or release biomarker; and (5) interrogate the causal links between circulating myeloid metabolic states, DC vaccine performance, and clinical outcomes.

• Sample size and heterogeneity: Healthy donor numbers were small in several assays (e.g., HD n=3–4), and patient mDC cultures were heterogeneous, potentially limiting power and generalizability. Single-site, Phase I cohort may not reflect broader populations.
• Cross-sectional/steady-state measurements: Metabolic profiling was performed at steady state ex vivo and may not capture dynamic metabolic shifts during antigen presentation and T-cell priming; metabolic parameters were not strongly predictive of antigen-specific T-cell responses.
• Transcriptome-function discordance: Bulk transcriptional changes did not consistently align with functional metabolic readouts, reflecting post-transcriptional regulation and cellular heterogeneity.
• Culture differences: Some differences in iDC culture timing between HD (3-day) and melanoma (5-day) are acknowledged, though considered unlikely to affect maturation outcomes.
• Causality: Associations between metabolic markers (e.g., mitochondrial dependence, MCT1, lactate) and outcomes are correlative; interventional validation is needed. Some analyses used trends near significance and maximally selected cutoffs that require external validation.