Human embryonic kidney (HEK)-293 T cells (ATCC) were cultured in DMEM (BioWest). Peripheral blood mononuclear cells (PBMC) were obtained from buffy coats from healthy donors (Centro de Transfusiones of the Comunidad de Madrid, Spain), using Ficoll density gradients. CD8⁺ T cells were isolated by negative selection (EasySep human CD8⁺ T cell, Stem Cell Technologies; 86–95.5% purity), and cultured in RPMI-1640 medium (BioWest).

For activation, CD8⁺ T cells were incubated (1:3.5 ratio) with tosyl-activated magnetic beads (Dynabeads M-450; Thermo Scientific) coated with 8% anti-CD3 (HIT3a, BD Biosciences), 10% anti-CD28 (CD28.2, BioLegend), and 82% control IgG₁ (T_ACT), or with anti-CD3, anti-CD28, and 82% PD-L1-Fc chimeric protein (R&D Systems) (T_{ACT + PD1}); IgG₁-coated beads were used as control (T_CTRL). In some experiments, PD-L1-Fc was used at 16.4, 3.3% or 0.66%. In some experiments, CD8⁺ T cells were incubated with T_{ACT + PD1} beads (48 h, 37 °C), which were mechanically released, removed with a magnet, and the cells restimulated with T_ACT or T_CTRL beads (48 h, 37 °C). As positive control, naïve CD8⁺ T cells were incubated (48 h, 37 °C) with plate-bound anti-CD3 (5 μg/ml; UCHT1, BD Biosciences) and soluble anti-CD28 (2 μg/ml) antibodies.

T cell activation was confirmed by FACS (Cytomics FC500 or Gallios cytometers; Beckman Coulter) using anti-CD25-PE (B1.49.9, Beckman-Coulter), −CD279-APC (MIH4, eBioscience), −CD69-PCy5 (TP1.553, Inmunotech), and -CD8-FITC (B9.11, Beckman-Coulter) antibodies. IFNγ was detected by intracellular staining using anti-IFNγ-PE (B27, Pharmingen) antibody in permeabilized cells (Beckman-Coulter) pretreated with brefeldin A (10 μg/ml, 4 h, 37 °C; eBioscience). Dead cells were detected with propidium iodide (2.5 μg/test, 1 min), or the LIVE/DEAD stain kit (Invitrogen). Appropriate isotypes were used as negative controls. Data were analyzed using Kaluza and FlowJo software.

hCD8⁺ T cell proliferation was determined by [methyl-³H] thymidine (1 μCi/well; Perkin Elmer) incorporation into DNA, in a 1450 Microbeta liquid scintillation counter (Perkin Elmer).

The RNA-seq libraries were prepared using an Illumina TruSeq Stranded Total RNA Sample Preparation kit (Illumina). Library size and quality was assessed in an Agilent DNA 7500 Bioanalyzer assay (Agilent). Each library was sequenced using TruSeq SBS Kit v3-HS, in paired end mode with read length 2 × 76 bp. On average, we generated 36 million paired-end reads for each sample in a fraction of a sequencing lane on HiSeq2000 (Illumina). Image analysis, base calling, and quality scoring of the run were processed by Real Time Analysis (RTA 1.13.48) software, followed by generation of FASTQ sequence files by CASAVA 1.8.

RNA-seq reads were aligned with the human reference genome (gencode v19) using the GEMtools RNA-seq pipeline v1.7 (http://gemtools.github.io), which is based on the GEM mapper [21]. Expression quantification at the gene level was calculated with Flux (http://sammeth.net/confluence/display/FLUX/Home). RNA-seq data were analyzed using the DESeq2 R Bioconductor package [22]. Raw counts of sequencing reads were normalized to the effective library size. Real-time quantitative PCR (qPCR) was performed in an ABI PRISM7900HT system (Applied Biosystems) with indicated primers (Additional file 2: Table S1).

A likelihood ratio test (LRT) was used to test for differences over multiple time points. This test compares a full model, including an interaction term class:time, with a reduced model without the interaction term; this permits to determine whether PD-1 treatment induces change of a specific gene at any point after time 0. This class-specific effect is measured as a p value for interaction (p_inter) and FC values for T_{ACT + PD1} vs. T_ACT cells at each time point. Genes with a significant p_inter were analyzed by STEM (Short Time-series Expression Miner) software [23] for cluster analysis and integration with the Gene Ontology (GO) database (http://geneontology.org/). These genes were analyzed for enrichment in KEGG signaling pathways using the online tool Webgestalt (http://www.webgestalt.org). Genes involved in metabolic pathways (KEGG hsa011000) were further explored for known interactions using Cytoescape (http://www.cytoscape.org/). GO enrichment analysis was performed using BINGO. GO categories were summarized and visualized using ClueGO or REVIGO.

Cellular oxygen consumption (OCR) and extracellular acidification rates (ECAR) were determined in Seahorse XF Base Medium supplemented with 25 mM glucose (Sigma-Aldrich), 2 mM L-glutamine and 1 mM sodium pyruvate (both from BioWest) using the XF cell Mito Stress Kit (SeaHorse Bioscience), in an XF24 Extracellular Flux Analyzer (SeaHorse Bioscience; Agilent Technologies). Fatty acid oxidation (FAO) was determined in Krebs-Henseleit buffer (KHB) supplemented with 0.5 mM carnitine (Sigma-Aldrich) and 2.5 mM glucose, using palmitate as substrate, in the Agilent Seahorse XF96 Extracellular Flux Analyzer.

Lactate levels were determined enzymatically in extracts from T_CTRL, T_ACT and T_{ACT + PD1} cells after 48 h stimulation, using a fluorometric lactate assay kit (Cell Biolabs) according to supplier’s protocol; fluorescence was quantified in a Filter Max F5 microplate reader (Molecular Devices) at 530/590 nm excitation/emission. A lactate standard curve was generated in all assays and used to extrapolate relative fluorescent units (RFU) measured in the samples.

Equal amounts of Triton X-100-based cells lysed were analyzed by SDS-PAGE analysis and immunoblotted with specific antibodies (see Additional file 1) [24]. For blue native analyses, we obtained a mitochondria-enriched fraction by cell lysis with hypotonic buffer and homogenization with a polypropylene pestle homogenizer. Nuclei and unbroken cells were removed, and mitochondria obtained by centrifugation (12,000×g) from the cytosolic fraction. The enriched mitochondrial fraction was suspended in 50 mM Tris-HCl pH 7.0 containing 1 M 6-aminohexanoic acid, lysed in 10% digitonin at 4 g/g mitochondrial proteins, and mitochondrial proteins fractionated in blue native gels.

Total mitochondrial mass, mitochondrial membrane potential (ΔΨm) and reactive oxygen species (ROS) were determined by FACS using MitotrackerGreen FM, tetramethylrhodamine, methyl ester (TMRM) and MitoSOX probes (Thermo Fisher), respectively. DNP (2,4-dinitrophenol) was used as ΔΨm negative control. Dead cells were excluded by diamino-2-phenylindol (DAPI) staining. Mitochondrial DNA (mtDNA) was extracted from hCD8⁺ cells with the DNeasy Blood and Tissue kit (Qiagen) and quantified by RT-qPCR using primers for MT-TL1 tRNA (Leu)(UUR) [25]; the α2-microglobulin gene was used for normalization.

Immunofluorescence analyses were performed in paraformaldehyde-fixed CD8⁺ T cells, permeabilized with Triton X-100 (0.1%). After blocking, cells were stained sequentially with anti-human aconitase-2 (6F12BD9, Abcam) and goat anti-mouse Alexa 488 (Molecular Probes). Samples were mounted in Prolong Gold Antifade Reagent with DAPI (Cell Signaling) and images captured in a Leica Microsystems microscope (LAS X v2.01; 60x objective). Mitochondrial morphology was determined with ImageJ [26].

For transmission electron microscopy, fixed cells were treated sequentially with 1% osmium tetroxide (TAAB Laboratories) and 2% aqueous uranyl acetate, dehydrated with acetone, embedded in EPON 812 resin, and polymerized. Ultrathin sections (70 nm-thick; Ultracut EM UC6, Leica Microsystems) in 200 mesh nickel EM grids (Gilder) were stained with 3% aqueous uranyl acetate and lead citrate, and analyzed on a JEOL JEM 1011 electron microscope. The number of mitochondria per cell and cristae length were quantified by two independent observers blind to the experiment.

Lentiviruses encoding CHCHD3 or control short hairpin RNA (shRNA; Genecopoeia) were produced in HEK-293 T cells. Prior to transduction, hCD8⁺ cells were stimulated with anti-CD3- and -CD28 antibody-coated beads, then transduced with viral supernatants at 10–20 m.o.i in the presence of polybrene. CHCHD3 silencing was determined by qPCR and by immunoblot.

Normal or parametric distribution of the data was analyzed. For comparison between two conditions, data were analyzed with the Mann-Whitney U test. For multiple non-parametric comparisons, Kruskal-Wallis followed by Dunn’s post-test was used. For multiple parametric comparisons, data were analyzed by one- or two-way ANOVA with the Bonferroni post-hoc test. For the same samples with different treatments, paired Student’s t-test was performed for two comparisons or paired repeated-measures one-way ANOVA for more than two conditions. Differences were considered significant when p < 0.05. All statistical analyses were performed using Prism 7.0 software (GraphPad).

To determine how PD-1 signals change gene expression during activation of human (h)CD8⁺ T cells, we used an in vitro system that mimics the simultaneous engagement of PD-1 and the TCR/CD3 complex. Purified hCD8⁺ T cells were stimulated with magnetic beads conjugated with stimulating anti-CD3 and -CD28 antibodies (T_ACT cells), or with anti-CD3, anti-CD28, and PD-L1-Ig fusion protein (T_{ACT + PD1} cells); hCD8⁺ T cells incubated 6 h with polyclonal IgG-conjugated beads were used as control (T_CTRL cells). In these conditions, PD-1 consistently inhibited hCD8⁺ T cell activation and effector functions, determined by a reduction in CD25, CD69 and IFNγ expression (Fig. 1a-d), as well as decreased proliferation (Fig. 1e). The PD-1-induced reduction was dose-dependent (Additional file 3: Figure S1).Fig1Fig. 1

Characterization of gene expression profiles in CD8⁺ T cells after PD-1 ligation. a Representative dot plots showing CD25 and CD69 staining of primary human CD8⁺ T cells after 48 h stimulation with T_CTRL, T_ACT, and T_{ACT + PD1} beads. b Quantification of CD25- and CD69-expressing cells from dot plots as in a. Each dot represents a donor (n = 18). c Representative histograms showing IFNγ production by cells stimulated as in a for 24 and 48 h. d Quantification of data from c (n = 4). e Thymidine ([³H]-TdR) incorporation by cells stimulated as in a (n = 5). f PCA plot using the rlog-transformed values from the RNA-seq analysis. Each unique combination of cell stimulation and time is assigned a distinct color. g Hierarchical clustering dendrogram of the top 200 most-variable genes across samples. The heat map color code (left) uses the combination of cell stimulation/time as in f. h Venn diagrams showing the number of differentially expressed genes between indicated conditions at different times. For d and e, data show mean ± SEM. ** p < 0.01; *** p < 0.001, one-way (b) or two-way ANOVA (d, e) with Bonferroni post-test

We used LRT to identify genes expressed differentially over time. This type of analysis identifies genetic patterns impaired by PD-1 engagement more reliably than direct comparison between T_CTRL, T_ACT and T_{ACT + PD1} RNAseq data at each time point. LRT analysis identified 1651 genes with divergent expression between T_ACT and T_{ACT + PD1} (p_inter < 0.05), but only 578 passed FDR correction (Adj-p_inter < 0.05); Additional file 5: Table S2 shows the top 20 genes in this analysis. KEGG pathway analysis using these 578 genes indicated that, in addition to pathways related to cell cycle and immune function, there was significant enrichment in metabolic genes, with 43 genes in this category (Fig. 2a; Additional file 6: Table S3). The primary metabolic processes with the most differentially regulated genes were amino acid, nucleotide and carbohydrate (glycolysis and the pentose phosphate) metabolism, the citrate cycle and OXPHOS (Additional file 7: Figure S3).Fig2Fig. 2

PD-1 ligation impairs mainly CD8⁺ T cell metabolism. a KEGG signaling pathways with the highest scores significantly enriched in the 578 transcripts selected by LRT. b Lactate production in hCD8⁺ T cells stimulated 48 h with T_CTRL, T_ACT and T_{ACT + PD1} beads. c Lactate production in hCD8+ T cells stimulated 48 h with T_{ACT + PD1} beads containing the indicated amounts of PD-L1-Fc. d-f hCD8⁺ T cells were stimulated with beads as in b and analyzed with SeaHorse. Basal extracellular acidification rate (ECAR; D), basal O₂ consumption rate (OCR; e), and basal OCR/ECAR ratio (f). g OCR obtained during mitochondrial stress test in cells stimulated as in b, performed by injection of oligomycin, the mitochondrial uncoupler FCCP, and the electron transport chain inhibitors antimycin A/rotenone. h-j Maximal OCR obtained after FCCP injection (h), spare respiratory capacity (SRC; i) calculated as the difference between maximal and basal OCR, and relative proton leak (j) determined as OCR after oligomycin and subsequent injection of rotenone plus antimycin A. k-n hCD8⁺ T cells were stimulated with beads as in b, treated with etomoxir or vehicle and analyzed with SeaHorse, using palmitate as substrate. Basal OCR with vehicle (solid) or with etomoxir (hatched) (k), FAO-specific OCR from data in k (l), maximal OCR after FCCP injection in vehicle or etomoxir-treated cells (m), FAO-specific maximal OCR calculated from m (n). o Representative immunoblots for CPT1A and β-actin (loading control) in CD8⁺ T cells stimulated as indicated. p Densitometric analysis of immunoblots as in o. The CPT1A/β-actin ratio is shown, with the value for T_CTRL cells as reference (n = 3 donors). Data are mean ± SEM from six (b, d-j), four (k-n) or three (p) donors; for c, data are mean ± SD representative of one donor from two. ** p < 0.01, * p < 0.05, Kruskal-Wallis with Dunn’s post-hoc test for multiple comparisons (b, d-f, h-k, m), two-tailed Student’s t-test (l, n), or two-way ANOVA with Newman-Keuls post-hoc test for multiple comparisons (p)

To validate the transcriptional changes with metabolic alterations, we focused on glycolysis and OXPHOS, key metabolic pathways for T cell differentiation and function [13, 27]. We found that lactate production, a glycolysis indicator, was reduced in T_{ACT + PD1} compared with T_ACT cells, in a dose-dependent manner (Fig. 2b, c). T_{ACT + PD1} cells similarly showed a significant ECAR reduction (Fig. 2d), which suggested that PD-1 ligation effectively inhibited the glycolytic pathway in CD8⁺ T cells. When we used high glucose levels as an energy source, basal OCR, an OXPHOS indicator, was significantly higher in T_ACT than in T_CTRL and T_{ACT + PD1} cells (Fig. 2e); the OCR/ECAR ratio was nonetheless higher in T_{ACT + PD1} than in T_ACT cells (Fig. 2f), which suggested that T_{ACT + PD1} cells preferentially use OXPHOS rather than glycolysis to generate ATP.

To analyze additional parameters of mitochondrial metabolism, we measured OCR in real time in basal conditions and after addition of several mitochondrial inhibitors (Fig. 2g). Addition of FCCP, which uncouples ATP synthesis from the electron transport chain, showed that maximal respiration capacity was higher in T_ACT than in T_CTRL and T_{ACT + PD1} cells (Fig. 2h). T_CTRL and T_{ACT + PD1} cells nonetheless had a substantial mitochondrial SRC, as indicated by the difference between maximal and basal OCR (Fig. 2i). The elevated SRC, a parameter associated with long-term survival [14], and the higher OCR/ECAR ratio suggest more efficient OXPHOS in T_{ACT + PD1} than in T_ACT cells. Confirming this idea, proton leak (determined as OCR after oligomycin relative to OCR after rotenone and antimycin A) was significantly lower in T_{ACT + PD1} than in T_ACT cells (Fig. 2j); there was also a tendency to lower proton leak in T_{ACT + PD1} than in T_CTRL cells (Fig. 2j).

To further study metabolic differences in the mitochondria of PD-1-stimulated cells, we measured OCR using palmitate as a substrate, alone or in the presence of etomoxir, which inhibits carnitine palmitoyltransferase 1A (CPT1A), a central enzyme for long-chain fatty acid oxidation in mitochondria. Etomoxir led to greater inhibition of basal and maximal (after oligomycin and FCCP treatment) OCR in T_{ACT + PD1} than in T_CTRL and T_ACT cells (Fig. 2k-n), which indicated greater OXPHOS dependence on FAO in T_{ACT + PD1} cells than in the other conditions. We also found time-dependent induction of CPT1A in T_{ACT + PD1} compared to T_CTRL and T_ACT cells (Fig. 2o, p), which might explain the mechanism underlying the higher FAO capacity of PD-1-stimulated cells. These results indicate that PD-1 signals reprogram CD8⁺ T cell metabolism for efficient use of FAO-dependent mitochondrial OXPHOS, which resembles some aspects of long-lived memory T cells [14]. Moreover, the distinct FAO-dependent OXPHOS between T_CTRL and T_{ACT + PD1} cells (Fig. 2l, n) suggests that PD-1-induced metabolic changes are not simply blockade of T cell activation, but involve unique, time-dependent programs induced by PD-1 engagement.

We analyzed mitochondria bioenergetics in live cells by combining the ΔΨm-sensitive TMRM and the ΔΨm-independent MitotrackerGreen probes; the depolarizing agent DNP was used as a TMRM staining control (Fig. 3a). Compared with T_CTRL cells, CD8⁺ T cell activation caused a significant increase in both the number of cells with polarized mitochondria (Fig. 3b) and the TMRM fluorescence bound to these mitochondria (Fig. 3c-d). PD-1 ligation abrogated the ΔΨm increase caused by activation stimuli (Fig. 3c-d). Reactive oxygen species (ROS) production nonetheless did not differ statistically between T_{ACT + PD1} and T_ACT cells (Fig. 3e). It appears that although PD-1 affects mitochondrial function, these organelles retain some respiratory capacity compared to that of resting T_CTRL cells.Fig3Fig. 3

PD-1 inhibits mitochondrial function in activated CD8⁺ T cells. a Representative dot plots of CD8⁺ T cells stained with MitoTracker Green and TMRM to determine the effect of the indicated stimuli on mitochondrial polarization. Incubation with the depolarizing agent DNP was used as negative control. b Time-dependent expansion of TMRM⁺ cells after indicated stimuli (n = 5). c Representative histograms of T_CTRL, T_ACT and T_{ACT + PD1} cells after 48 h stimulation. TMRM fluorescence is shown of DNP-treated T_ACT cells (negative control; dotted line). d Mean fluorescence intensity of TMRM⁺ CD8⁺ T cells at different times post-stimulation, assessed from data as in c (n = 5). e Percentage of ROS⁺ cells as detected with the MitoSOX Red probe. f Scheme for analyzing the reversibility of PD-1 effects on mitochondrial potential. g, h Percentage of TMRM⁺ cells and TMRM mean fluorescence intensity in T_ACT and pre-treated T_{ACT + PD1} cells re-stimulated with T_CTRLand T_ACT beads (n = 3). Data shown as mean ± SEM. *** p < 0.001, two-way ANOVA with Bonferroni post-hoc test; * p < 0.05, two-tailed paired Student’s t-test

Of the 578 genes selected by LRT, 84 coded for transcripts enriched in mitochondrial-related GO categories (Additional file 11: Figure S7). These 84 genes were not only related to metabolic pathways, but also included those involved in mitochondrial DNA replication and repair (FEN1, TOP2A, XRCC3), translation (POP7, MRPL39, MRPS12), protein import machinery (TIMM22, TIMM23, TOMM34), fusion/fission (MIEF1, MTCH1), cristae structure and organization (CHCHD3, CHCHD10, HSPA9), and assembly of protein complexes of the respiratory chain (ATP5G1, COX8A, NDUFB3, SELRC1, UQCRC2) (Additional file 12: Table S4).

We used STEM software [23] to analyze and cluster our gene expression dataset more stringently. STEM clustering of logFC values generated eight model expression profiles significantly enriched (FDR < 0.05) for transcripts expressed longitudinally in T_ACT vs T_{ACT + PD1} cells (Fig. 4a). Profile A, which clustered transcripts whose expression increased over time in T_ACT compared to T_{ACT + PD1} cells, was specifically enriched for genes in Mitochondrial Protein Complex (including ATP5G1, CHCHD3, COX8A, DNA2, NDUFAB1, NDUFB3, NDUFB7, PPIF, TIMM22, TIMM23, TOMM40, TOMM40L and UQCRC2), as well as in 27 transcripts of other mitochondrial-related profiles (Fig. 4b). This finding suggests that genes involved in mitochondrial structure and function tend to be upregulated in T_ACT rather than T_{ACT + PD1} cells.Fig4Fig. 4

Validation of changes in expression of mitochondrial-related genes after PD-1 ligation. a STEM clusters of expression profiles in T_ACT and T_{ACT + PD1} cells. Only significant profiles are shown, ordered by p value (bottom left). The line in each STEM cluster represents the average temporal expression profile for the genes assigned to the cluster. The number of genes in each profile is indicated (top right). b Scatter plot showing GO terms of STEM profiles A, E and F, represented as circles and clustered according to semantic similarities as determined by REViGO. Circle area is proportional to the significance of GO term overrepresentation; color indicates the log₁₀ of the corrected p value for enrichment. c Time-course variation of the relative quantity (Rq) of indicated transcripts in T_CTRL, T_ACT, and T_{ACT + PD1} cells isolated from independent donors (n ≥ 3). d Representative immunoblots for proteins in CD8⁺ T cells stimulated as indicated (n ≥ 3 donors). e Densitometric analysis of immunoblots as in d. The Rq was calculated as the ratio between each protein and β-actin, taking the value for T_CTRL cells as reference. For C and E, data are mean ± SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, using two-way ANOVA with Bonferroni post-test (c) or Kruskal-Wallis with Dunn’s post-hoc test for multiple comparisons (e); only significant differences are indicated

We analyzed whether different treatments influenced cell mitochondrial mass. HSP60 is a marker of mitochondrial biogenesis [26]. Consistent with the tendency to HSP60 downmodulation in T_{ACT + PD1} cells, mitochondrial number was significantly reduced in T_{ACT + PD1} compared to T_ACT cells, as determined by direct counting (Fig. 5a; Additional file 15: Figure S9A-C), relative mtDNA quantity (Fig. 5b), or MitotrackerGreen staining (Fig. 5c, d). In contrast, mitochondrial mass was statistically unchanged between T_ACT and T_CTRL cells (Fig. 5a-d). As for ΔΨm, cell pre-incubation with PD-L1 reduced mitochondrial mass, which was not reversed after their re-stimulation with T_ACT beads (Fig. 5e).Fig5Fig. 5

PD-1 stimulation reduces the number of mitochondria but does not affect mitochondrial dynamics. a Number of mitochondria per cell as determined by direct counting from transmission electron microscopy images (n ≥ 83 cells/condition). Results are the average of counting by two independent observers, one of them blind to the experiment. b Relative mitochondrial DNA quantity determined by qPCR (n = 3). c Representative histogram of T_CTRL, T_ACT, and T_{ACT + PD1}-stimulated cells (48 h) stained with the MitoTracker Green probe. d Quantification of mean fluorescence intensity from cells as in c (n = 7 donors). e Quantification of MitoTrackerGreen mean fluorescence intensity in T_ACT and pre-treated T_{ACT + PD1} cells restimulated with T_CTRLand T_ACT beads (n = 3). f Representative confocal images of T_CTRL, T_ACT, and T_{ACT + PD1} cells stained with aconitase-2. g Quantification of mitochondrial circularity, determined from confocal images as in e using ImageJ software (n ≥ 31 cells/condition). h Quantification of OPA-1 and DRP-1 mRNA levels in T_ACT and T_{ACT + PD1}-stimulated cells. Values were normalized to those from T_CTRL cells. i Representative OPA-1 and DRP-1 immunoblots in cells treated as indicated. The line indicates removal of an empty lane. j Densitometric analysis of immunoblots as in h. The Rq was calculated as the ratio between each protein and β-actin, taking the value for T_CTRL cells as reference (n = 3 donors). In all cases, data were compared using one-way (a, b, g), two-way ANOVA (d, h, j) with Bonferroni’s post-test, or paired two-tailed Student’s t-test (e); * p < 0.05, ** p < 0.01, n.s., not significant

Although several reports linked PD-1 to functional mitochondrial impairment [15, 17–19], the structural changes in mitochondria from PD-1-stimulated CD8⁺ T cells have not been described in detail. PD-1 downregulated two genes, CHCHD3 (also termed Mic19) and CHCHD10 (Mic14; Fig. 4d, e), which form part of the mitochondrial contact site and MICOS [30]. In mammalian cells, the MICOS is a multimeric complex composed of nine known subunits and putative interactors, which links the inner boundary to the outer mitochondrial membranes and stabilizes cristae junctions [30].

Ultrastructural analyses showed clear differences in the organization of the inner mitochondrial membrane and cristae (Fig. 6a). Mitochondria from T_ACT cells had a large number of tight cristae, with a parallel-oriented lamellar profile (Fig. 6a). This contrasted with the loose vesicular profile of cristae in T_CTRL cells. T_{ACT + PD1} cell mitochondria also had some swollen cristae, although they did not show the clear vesicular profile observed in T_CTRL cells (Fig. 6a); this is consistent with the loss of respiratory capacity and transcriptomic downregulation of structural proteins. Moreover, T_{ACT + PD1} cell mitochondria often lacked visible cristae (Fig. 6a). The percentage of mitochondria without cristae was significantly larger in T_{ACT + PD1} than in T_ACT cells (Fig. 6b). Although T_CTRL cells also had a larger number of mitochondria without cristae than T_ACT cells (Fig. 6b), the differences were not significant (p = 0.14; Fisher’s exact test). The number of cristae per mitochondrion and the length of these cristae was significantly reduced in T_{ACT + PD1} compared to T_ACT cells (Fig. 6c, d). The results suggest that PD-1-induced downmodulation of these MICOS-associated proteins affect cristae organization.Fig6Fig. 6

PD-1 reduces the number and length of mitochondrial cristae. a Representative micrographs showing magnified mitochondria from T_CTRL, T_ACT, and T_{ACT + PD1}-stimulated cells (48 h). b-d Percentage of mitochondria without cristae (b), average number of cristae per mitochondrion in each cell (c), and length of cristae in each mitochondrion (d) in CD8⁺ T cells stimulated for 48 h, as indicated. e Relative CHCHD3 mRNA levels in shRNA_CTRL- or shRNA_CHCHD3-transduced CD8⁺ T cells. Data are mean ± SEM (n = 3). f Representative immunoblot showing CHCHD3 protein levels in shRNA_CTRL- or shRNA_CHCHD3-transduced cells. The densitometric CHCHD3/β-actin ratio was calculated, using the value for shRNA_CTRL cells as reference (n = 2). g, h Percentage of shRNA_CTRL- or shRNA_CHCHD3-transduced CD8⁺ T cells showing polarized mitochondria, as determined by TMRM staining (g), and producing IFNγ (h). Each pair of points represents an independent donor. For a, b and d, n = 127 (T_CTRL), 170 (T_ACT) and 222 (T_{ACT + PD1}) mitochondria analyzed; for c, n = 17 (T_CTRL), 23 (T_ACT) and 33 (T_{ACT + PD1}) cells. * p < 0.05, ** p < 0.01, *** p < 0.001, one-way ANOVA with Bonferroni’s post-test (b-d) or two-tailed paired Student’s t-test (e, g-h)

Individual respiratory chain complexes can be organized in quaternary supramolecular structures termed supercomplexes (RCS) [31, 32]. These RCS reside in the inner mitochondrial membrane, and establish an efficient proton gradient for complex V to synthesize ATP [33]. Although the precise RCS arrangement is largely unknown, high-resolution structural models of the mammalian respirasome have been described [34–37]. Since RCS are highly enriched in the cristae membrane [31, 32] and their formation/stability is linked to cristae shape [38], we tested whether the morphological changes in the T_{ACT + PD1} cell cristae affected RCS formation. To our surprise, we found greater enrichment of RCS containing complexes I and III in mitochondrial membranes of T_{ACT + PD1} and T_CTRL than of T_ACT cells (Fig. 7a-d); in contrast, complex III dimers were represented equally in all cell types (Fig. 7a-d).Fig7Fig. 7

PD-1 increases the formation of supercomplexes. a Representative blue native PAGE showing RCS formation in T_ACT, and T_{ACT + PD1}-stimulated cells (48 h). Blots were hybridized sequentially with anti-NDUFS3 (complex I), −Core2 (complex III) and -βF1-ATPase (complex V) antibodies. b Densitometric quantification of the blots shown in A (n = 4; *, p < 0.05, paired two-tailed Student’s t-test). c Blue native PAGE showing RCS formation in T_ACT and T_CTRL cells (48 h); hybridizations were as in a. d Densitometric quantification of the blots shown in c. Data shown are from a pool of three donors. e Relative MCJ/DnaJC15 mRNA levels in T_CTRL, T_ACT and T_{ACT + PD1} cells at different times post-stimulation with indicated beads. Values were normalized to unstimulated cells (time 0). Data are mean ± SEM (n = 3 independent donors). *** p < 0.001, two-way ANOVA with Bonferroni’s post-hoc test