Functional Analysis of Immune Signature Genes in Th1* Memory Cells Links ISOC1 and Pyrimidine Metabolism to IFN-γ and IL-17 Production

This article is featured in Top Reads, p.1115

The IFN-γ production by CD4⁺ T cells plays an essential role in the persistent control of infections, including latent Mycobacterium tuberculosis (1–4). In humans, genetic mutations resulting in IFN-γ deficiency predispose to mycobacterial diseases (5), and animal studies show that mice lacking IFN-γ or IFN-γ–regulating factors such as T-bet, IL-12, and STAT1 are highly susceptible to mycobacteria infection (6–9). Nonconventional IFN-γ–producing CD4⁺ T cells termed Th1* cells account for >80% of M. tuberculosis–reactive memory T cells in the periphery (10–13). Th1* cells exhibit a CCR6⁺CXCR3⁺CCR4⁻ phenotype and share lineage-specific characteristics with conventional Th1 (CCR6⁻CXCR3⁺CCR4⁻) and Th17 (CCR6⁺CXCR3⁻CCR4⁺) memory cells (10, 14–17), including expression of T-bet and RORC, the transcriptional regulator of IL-17 production. Th1* cells are significantly expanded in latent tuberculosis (TB) infection (LTBI) (10), and IFN-γ–producing Th1* cells comprise more than 50% of M. tuberculosis–specific cells during active TB (17), supporting a protective role in M. tuberculosis control. Th1* cells are also linked to proinflammatory signaling in rheumatoid arthritis, multiple sclerosis, and Crohn disease (14, 16, 18). Thus, this unique subset is important in the context of protective responses to infection and in controlling autoimmunity and inflammation.

Previous studies provided transcriptomic analysis of immune signature genes in human Th1* cells (10, 19). More than 400 genes demonstrate differential expression in Th1* compared with Th1, Th2, and Th17 populations (10). Functional network analysis of genes selectively upregulated in Th1* shows enrichment of factors linked to T cell proliferation, effector functions, and M. tuberculosis immunity such as CCR2 (20), RORC (21), and components of IL-23 signaling (22–26). We reasoned that these genes are involved in Th1* effector function. In this study, we report a loss-of-function screen of immune signature genes in endogenous human Th1* cells. New regulators of IFN-γ production included ISOC1, a metabolic enzyme of unknown function in mammalian cells. We find that the loss of ISOC1 reduced IFN-γ and IL-17 production by perturbing cellular pyrimidine metabolism. Results demonstrate the utility of functional screening for validating immune signature genes and highlight the importance of metabolic fitness for the adaptive immunity.

For the RNA interference (RNAi) screen, 209 genes were selected based upon enriched expression in Th1* cells compared with Th1, Th2, and Th17 (10). To prioritize functionally relevant genes in this subset, gene regulatory network analysis using the C3NET approach (27) was performed using the published gene expression dataset for Th1* cells (10) and additional Th1* and activated CD4⁺ T cell datasets from the Database of Immune Cell Expression, Expression Quantitative Trait Loci (eQTLs) and Epigenomics (28). For each gene, C3NET identifies paired genes showing the highest coexpression as “neighbors” in the regulatory network, and genes with multiple neighbors are central network “hub” genes. We used a number of neighbors for each gene in the network analysis to rank the Th1*-enriched genes. Combining this information with gene ontology, a set of 100 gene-specific short hairpin RNAs (shRNAs) (including a nontargeting shRNA) was used for screening, with T-bet and LCK included as established positive regulators of IFN-γ.

PBMC were isolated from healthy donor whole blood at the La Jolla Institute for Immunology Clinical Core and normal blood donor program (VD-057). PBMC were purified by density gradient centrifugation using Lymphoprep (Cosmo Bio). Isolated cells were resuspended in heat-inactivated FBS (Sigma-Aldrich) containing 10% DMSO and cryopreserved in liquid N₂. CD4⁺ cells were positively selected and isolated using CD4 Dynabeads and magnetic separation (Dynal; Life Technologies), according to the manufacturer’s protocol. After purification, ∼98% of cells were CD4⁺ as assessed by flow cytometry. For isolation of Th1* and other memory cell subsets, PBMC were labeled with positive and negative selection Abs (10). Briefly, viable lymphocytes were identified by forward/side scatter and LIVE/DEAD Aqua (eBioscience) staining. Dump gating excluded CD8, CD14, CD19-positive populations, and dead cells. CD4⁺CD3⁺CD25⁻ memory cells were identified based on CD45RA and CCR7 expression (i.e., excluding CD45RA⁺CCR7⁺ naive cells). Cells were sorted in CCR6⁺CXCR3⁺CCR4⁻ (Th1*), CCR6⁺CXCR3⁻CCR4⁺ (Th17), CCR6⁻CXCR3⁺CCR4⁻ (Th1), and CCR6⁻CXCR3⁻CCR4⁺ (Th2) subsets. Th1* cells were further expanded in culture for screening experiments. Cell sorting was performed on a BD FACSAria-3 or FACSAria-4 Fusion instruments. The Abs used were as follows: CD4 allophycocyanin eFluor 780 (RPA-T4) and CD45RA eF450 (HI100) (from eBioscience), CD3 Alexa Fluor 700 (UCHT1), CXCR3 allophycocyanin (IC6), CCR6 biotin (11A9), CD19 V500 (HIB19), CD14 V500 (145E2), CD8 V500 (RPA-T8), CD25 FITC, CCR4 PECy7 (1G1) (all from Becton Dickinson), and CCR7 PerCPCy5.5 (G043H7) (BioLegend).

Purified T cells were maintained in IMDM containing 5% heat-inactivated FBS (Sigma-Aldrich), 2% human serum (Cellgro), and 25 μg/ml gentamicin (Life Technologies), denoted as T cell medium. Freshly sorted cells were stimulated with anti-CD3/CD28–coated Dynabeads (Invitrogen) at 1:1 cell:bead ratio and expanded in T cell medium supplemented with 60 U/ml rIL-2 (eBioscience). Cell numbers were determined by trypan blue staining using Countess Cell Counter (Thermo Fisher Scientific) or MOXI Z Cell Counter (Orflo Technologies).

Genome-wide human lentiviral MISSION shRNAs (Sigma, developed by The RNAi Consortium [TRC]) were obtained at the La Jolla Institute for Immunology Functional Genomics Core (29, 30). shRNA clones were collected from master glycerol stock plates using Beckman Biomek FXp liquid handler and amplified in Luria–Bertani medium in deep-well 96-well plates. Plasmid DNA was isolated using Omega E.Z.N.A. Plasmid DNA Mini Kit and quantified on a NanoDrop One (Thermo Fisher Scientific). In the screen, four distinct shRNAs were combined into gene-specific pools during bacterial culture after normalization by OD. The culture was then used for plasmid DNA purification using GenElute HP 96-Well Plasmid Miniprep Kit (Sigma-Aldrich) and lentivirus production. For validation of selected genes, three to five individual shRNAs were tested for each gene. Lentivirus for pooled or individual shRNAs were produced in HEK293T cells plated in six-well plates in antibiotic-free DMEM (Thermo Fisher Scientific) supplemented with glutamine and FBS. Cells cultured in 10% FBS-containing DMEM were transfected with 375 ng shRNA plasmid and 375 ng of 3:1 (psPAX2:pMD2.G) lentiviral packaging mix using jetPRIME transfection reagent (Polyplus). Packaging plasmids were obtained from Addgene. After 24 h, the media was changed to the culture media containing 30% FBS. Viral supernatants were collected at posttransfection days 2 and 3, filtered through 0.45 μm low protein-binding filters, aliquoted, and stored at −80°C. Titers of pooled shRNA viruses used in the primary screen were measured using p24 ELISA Kit (ZeptoMetrix). Nontargeting control corresponds to SHC002 from the TRC library.

Cells were stimulated for 1–2 d with anti-CD3/CD28–coated Dynabeads (Life Technologies) at a 1:1 bead to cell ratio prior to transduction with shRNA lentiviruses. After removal of Dynabeads, cells were infected with high-titer shRNA lentiviruses (multiplicity of infection [MOI] ∼8–10) in 96-well plates (100 μl viral supernatants per ∼9 × 10⁵ cells per well). Cells were transduced in triplicate wells (assigned for each shRNA target). Polybrene (Sigma-Aldrich) was added to the mixture at a final concentration of 8 μg/ml. Cells were centrifuged at 2000 Rpm for ∼2 h at 30°C; viral media was removed after 4 h and replaced with T cell medium containing 60 U/ml IL-2. One day after transduction, puromycin (InvivoGen) was added at a final concentration of 2 μg/ml. Cells were cultured for 3–4 d in IL-2 containing T cell medium in the presence of puromycin, followed by restimulation with anti-CD3/CD28 Dynabeads (in the absence of added IL-2). Viable cells were analyzed for anti-CD3/28–induced IFN-γ production and used in other downstream analyses. For several experiments, functional viral titers were determined by titrating viral preparations in the presence and absence of puromycin (Fig. 1A). Titration curves were fit according to the Poisson distribution. Briefly, a fraction of infected cells P follows equation P = (1 – e^−m), where m is MOI. For a given dose of viral stock V, P is a variable that can be experimentally quantified by dividing the number of viable puromycin-resistant cells by the control cell number in the absence of puromycin. The corresponding MOI is a function of an unknown variable, viral titer T; MOI can be expressed as T*V/N, where N is the number of cells in the sample. Therefore, P can be expressed as (1-e^−T*V/N). However, experimental titration curves show that quantification can be confounded by the presence of transduction-resistant cells that are not infected at increasing amounts of virus (Fig. 1A). Thus, an additional parameter, a fraction of “transducible” cells (A_max), is introduced to the analysis. The resulting equation P = A_max *(1 – e ^{−T*V/Amax *N}) was used for nonlinear regression analysis of titration curves to determine the viral titer.

IFN-γ in cell supernatants was measured using human IFN-γ ELISA Kit (Invitrogen), according to the manufacturer’s protocol. Cells plated at 5 × 10⁵ cells/100 μl per well of a 96-well plate were stimulated with anti-CD3/28 Dynabeads for 24 h, unless otherwise indicated. Background signal was assessed using unstimulated cells. After stimulation, cells were transferred to U-bottom 96-well plates and centrifuged for 5 min (800 × g). Supernatants (50 μl from each well) were collected and stored at −80°C prior to analysis. Cell viability was measured using ATP-based CellTiter-Glo Viability Assay (Promega). After supernatant collection, cells pellets were resuspended in 100 μl T cell media, and 20 μl aliquots from each well were transferred to black-wall 96-well plates (Costar) for ATP measurements. IFN-γ (OD₄₅₀) and ATP (luminescence) were measured using a PE EnVision 1800 high throughput microplate reader. ATP values were used to normalize IFN-γ levels. To correct for confounding inhibitory effects on ATP, dose-dependent effects of azide on ATP production and IFN-γ levels were measured (Fig. 3B). The nonlinear relationship was fit into a third-order polynomial function (31). This model was used to calculate normalized IFN-γ levels. Screening data are expressed as percentages of corresponding values obtained with control (nontargeting) shRNAs. IFN-γ concentrations in the supernatants were determined based on the calibration curve (using human IFN-γ standards) generated on each ELISA plate. A nonlinear (quadratic) regression was applied to fit the data.

IL-17 in cell supernatants was measured using Human IL-17 ELISA Kit (Sigma), according to the manufacturer’s protocol and as described above for IFN-γ, except that 60–80 μl supernatant was analyzed. IL-17 concentrations were determined from a calibration curve using IL-17 standards and a nonlinear (quadratic) regression analysis.

Intracellular IFN-γ was measured in anti-CD3/28–stimulated cells. After 24-h stimulation, Fixation & Permeabilization Buffer (Affymetrix eBioscience) was added prior to staining for intracellular proteins. Brefeldin A (Sigma-Aldrich) was added at 10 μg/ml during the last 4 h of stimulation. Prior to fixation, cells were stained with LIVE/DEAD Aqua or Violet Fixable Viability Dyes (Invitrogen). Fixed and permeabilized cells were stained with IFN-γ–FITC (eBioscience) or IFN-γ–allophycocyanin Ab (BD Pharmingen) for 40 min at room temperature. Other Abs used for surface and intracellular staining were as follows: CD160–allophycocyanin, CD4–allophycocyanin eFluor 780 or CD4–FITC (all from eBioscience), and T-bet–PE (Miltenyi Biotec). For T-bet staining, samples of anti-CD3/28–stimulated cells were prepared using Fixation & Permeabilization Buffer Set for transcription factors and nuclear proteins (Affymetrix eBioscience). Samples were acquired in a BD FACSCanto II instrument and analyzed using FlowJo 10.2 software.

Proliferation of anti-CD3/28–stimulated CD4 T cells was measured by carboxyfluorescein dilution. Cells were labeled using CellTrace CFSE Cell Proliferation Kit (Thermo Fisher Scientific), following the manufacturer’s protocol. After 2- or 3-d stimulation with anti-CD3/28 Dynabeads (at a 1:1 bead to cell ratio), carboxyfluorescein-loaded cells were analyzed by flow cytometry. CFSE staining at day 0 (in unstimulated cells) was performed in select experiments to verify equal CFSE loading in control and test samples. Cell survival during cell culture was assessed using MOXI Z Cell Counter (Orflo Technologies). Alternatively, cells were counted after trypan blue staining using Countess Automated Cell Counter (Thermo Fisher Scientific). Apoptosis was measured using FITC-labeled Annexin V/propidium iodide (PI) staining kit (BD Pharmingen) and flow cytometry. Initial forward and side scatter gating excluded most of the dead (PI-positive) cells from further analysis. Early apoptosis was quantified as a percentage of Annexin V–positive and PI-negative cells. Samples were analyzed in a BD FACSCanto II instrument within 1 h after staining.

Total RNA was isolated from ∼1 to 2 × 10⁶ cells using Quick-RNA Miniprep Kit (Zymo Research). RNA concentrations were obtained on a NanoDrop Spectrophotometer; cDNA was reverse transcribed from 5 μg RNA with qScript cDNA SuperMix (Quanta Biosciences). Primers and probes for real-time quantitative PCR analysis were predesigned TaqMan Gene Expression Assays (Applied Biosystems), including TBX21 (Hs00894392_m1), CD160 (Hs00199894_m1), HPRT1 (Hs02800695_m1), ISOC1 (Hs0011372_m1), and IL-23R (Hs00332759_m1). Measurements were done in Bio-Rad CFX96 quantitative PCR instrument. All values were normalized to mRNA levels of the housekeeping gene HPRT1 (hypoxanthine phosphoribosyltransferase).

Untargeted analysis of endogenous metabolites in CD4⁺ T cells was performed using “deep [¹³C] labeling” of amino acids and glucose (32). CD4⁺ cells transduced with nontargeting and ISOC1 shRNAs were cultured in RPMI-based [¹³C]-labeled (Cambridge Isotope Laboratories) media supplemented with 25 μg/ml gentamicin and 5% FBS and IL-2 (60 U/ml). Transduction and selection of shRNA-expressing cells was performed as described above in [¹³C]-labeled medium and 1 μg/ml puromycin 24 h after transduction. Cells were cultured for 5 d in the labeling media, and media was freshly added every day. In parallel, control and ISOC1 shRNA cells were cultured in unlabeled media under the same conditions. On day 4, cells were transferred to puromycin-free labeled or unlabeled media and stimulated with anti-CD3/CD28. On day 5, cells (∼1.5 × 10⁶ for triplicate samples per condition) were centrifuged, and supernatant aliquots were collected and frozen at −80°C. Cell pellets were washed with cold PBS and stored at −80°C. Cell pellets were lysed in cold 4:1 methanol:water and cycled three times between 40 and −80°C for 30-s intervals. Extracts were centrifuged at 14,000 Rpm for 10 min at 4°C to sediment insoluble cell debris. Supernatants were removed and dried via vacuum concentration at 40°C until dry before resuspension in cold 4:1 methanol:water. Extracts were incubated at −20°C for 30 min to sediment any remaining insoluble material, which was removed through an additional centrifugation at 14,000 Rpm for 10 min at 4°C. Liquid chromatography–tandem mass spectrometry (LC-MS/MS) was performed as described (32). Briefly, chromatography was performed using a Thermo Vanquish UHPLC system (Thermo Fisher Scientific) with a ZIC-pHILIC polymeric column (150 mm × 2.1 mm × 5 μm) (SeQuant) at 25°C. All cell samples used an injection volume of 5 μl, constituting 80,000 cell equivalents. Mobile phases were 20 mM ammonium bicarbonate in water (pH 9.60) and acetonitrile. Gradient mobile phase followed a 19-min course, starting at 90% acetonitrile for 2 min, followed by a linear gradient to 55% acetonitrile at 16 min, sustained for three additional minutes. The column was then re-equilibrated for 11 min at 90% acetonitrile. Positive and negative ions were generated from a heated electrospray ionization source coupled to a Thermo Q Exactive Orbitrap Mass Spectrometer. MS1 spectra were acquired at 35,000 resolution with 100-ms ion trap time; MS2 spectra were acquired at 17,500 resolution and 50-ms ion trap time from a 1.0 m/z range around parent MS1 and with a normalized collision energy of 35 V. Mass range was 67–1000 m/z. Sheath and auxiliary gas flow rates were 40 and 20 U, respectively, with sweep gas flow rate at 2 U. Spray voltage was 3.5 kV for positive mode and 2.5 kV for negative mode. Capillary inlet and auxiliary gas heater temperatures were 275 and 350°C, respectively. LC-MS/MS feature extraction, putative Human Metabolome Database identification, and mass isotopomer enrichment analysis were performed as previously described (32).

Energy metabolism fluxes was assessed using Seahorse XF^e96 analyzer (Agilent Technologies). Optimized assay medium contained DMEM base (D5030-1L; Sigma), 1.85 g/L NaCl, 5 mM HEPES (pH 7.6) (at room temperature), and 3 mg/L Phenol Red supplemented with 10 mM glucose, 10 mM pyruvate, and 4 mM glutamine. Cells were acutely attached to Seahorse-manufactured 96-well cell culture plates precoated with Cell-Tak (no. 354240; Corning). Cell-Tak was used essentially as described in the manufacturer’s protocol, except that the coating density of 12 μg/cm² (1.3 μg/well in 10 μl) was applied. For seeding, cells were pelleted (5 min at 400 g), washed with the assay medium, plated at a density of 3 ×10⁵ cells per well, and spun down (5 min at 500 g) for attachment 30 min before the assay. Prior to seeding, cell concentrations were determined from triplicate live cell counts using the MOXI Z Cell Counter. Equal cell numbers were verified in each well by microscopy. For stimulation, anti-CD3/CD28–coated Dynabeads were washed with PBS + 0.1% BSA twice followed by wash with the assay buffer and added to the cells in the ratio 2:1 immediately prior to plating. Oxygen consumption rate (OCR) was measured as the indicator of mitochondrial activity. Extracellular acidification rate (ECAR) reflects the rate of lactate production from glucose as well as carbon dioxide production from glycolytically produced pyruvate. Thus, ECAR serves as a semiquantitative measure of glycolytic flux. The contribution of nonglycolytic acidification in ECAR was derived from the responses of cells incubated in the absence of added glucose and subtracted from the responses in complete, glucose-containing medium to determine glycolytic fluxes. The basal state of stimulated or unstimulated cells was followed for 30 min by which time TCR-mediated bioenergetics responses manifested. Monitoring was followed by the addition of mitochondrial inhibitors (Sigma-Aldrich): 2 μg/ml oligomycin, two pulses of 150 μM uncoupler DNP, and 2 μM respiratory inhibitor myxothiazol to assess resting respiration (an indicator of mitochondrial membrane intactness), maximal respiratory capacity, and nonmitochondrial/baseline oxygen consumption, respectively. ECAR measurements corrected for the nonglycolytic component were used to determine basal and maximally stimulated glycolytic activity; the latter was induced by oligomycin but increased over time and reached the maximum in the last reading in the presence of myxothiazol. A conversion factor of ECAR units (millipH/min) to acid production rates (pmol H⁺/min) was determined in separate experiments by injecting a known amount of acid to the medium and was found to be 6.1.

Cells were centrifuged at a low speed and pellets resuspended in 50 μl ice-cold lysis buffer containing 0.5% Nonidet P-40, 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 1× complete protease inhibitor mixture (Roche). After 20 min on ice, whole cell lysates were centrifuged at 15,000 × g for 15 min. Protein concentrations were determined by Pierce BCA Protein Assay (Thermo Fisher Scientific). Whole cell extracts were loaded on NuPage 4–12% Bis-Tris gels (Life Technologies) at 20–40 μg per lane. Gels were run at 200 V for 42 min, and proteins were transferred to nitrocellulose membrane (Bio-Rad Laboratories) at 30 V for 75 min. Membranes were blocked in 2.5% milk overnight at 4°C and probed with primary Abs for RORC, ISOC1 (Aviva Systems Biology), and STIM1 (Cell Signaling Technology) and secondary HRP-conjugated anti-mouse and anti-rabbit Abs (Sigma-Aldrich). Proteins were detected using ECL (GE Healthcare) or SuperSignal West Femto reagents (Thermo Fisher Scientific).

Student t test and two-way ANOVA with post hoc Bonferroni tests were performed using GraphPad Prism software: ***p < 0.001, **p < 0.01, and *p < 0.05.

To characterize the Th1* immune signature, we designed an RNAi screen examining IFN-γ production. Target genes were selected based upon enrichment in Th1* cells and gene network analysis prioritizing regulatory or signaling proteins (Supplemental Table I). Lentiviral transduction in CD4⁺ T cells yielded ∼50% shRNA-expressing, puromycin-resistant cells (Fig. 1A). Higher viral doses did not significantly increase transduction efficiency, and cell viability decreased at MOI >10. Pooled shRNA preparations used in the screen typically elicited 60–80% target mRNA and protein depletion after puromycin selection (Fig. 1B–D).

Th1*, Th1, Th2, and Th17 subsets were cosorted from human PBMC (Fig. 2A). Endogenous Th1* cells typically constitute 2–21% of total CD4⁺ T cells in the periphery (10 and Fig. 2B). Anti-CD3/28–stimulated Th1* coproduced IFN-γ (Fig. 2C) and IL-17 (Fig. 2D), which was preserved after repeated polyclonal stimulations ex vivo. The screen was conducted using ex vivo expanded Th1* cells by measuring secreted IFN-γ in puromycin-selected cells after anti-CD3/28 stimulation (Fig. 2E). To normalize IFN-γ levels, cell viability was measured poststimulation by quantifying ATP. These measurements typically demonstrate linear dependence upon metabolically active cell numbers over a large dynamic range. However, ATP readouts varied significantly in the shRNA-expressing Th1* populations (Fig. 3A), likely because of effects upon cellular bioenergetics, puromycin sensitivity, or other factors affecting cellular ATP levels. To reduce false positives, IFN-γ measurements were corrected for decreased ATP. Dose-dependent effects of the mitochondrial inhibitor azide on IFN-γ production were examined, confirming a reciprocal relationship between acute inhibition of ATP synthesis and IFN-γ levels. Interestingly, dependence of IFN-γ on ATP was nonlinear, in that a 2-fold decrease in ATP did not significantly affect IFN-γ that was sharply decreased at >50% depletion of ATP (Fig. 3B). This implies that IFN-γ production in CD4 T cells can be sustained under conditions of impaired mitochondrial function and energy production. In line with this, a previous study demonstrated that preactivated CD4 T cells retained proliferative capacity in the presence of mitochondrial inhibitors for 2 d postactivation (33). Azide titration results justify utility of the ATP assay for measuring viable cells to normalize IFN-γ in microplate format. The relationship between the two parameters was fit into a cubic polynomial function to generate a calibration curve for IFN-γ values (Fig. 3B, 3C). In contrast to the relative independence of IFN-γ production on acute ATP depletion using azide, inhibition of protein synthesis by puromycin profoundly affected IFN-γ levels (Supplemental Fig. 1). Accordingly, shRNA pools targeting ribosomal proteins RPL39 and RPL17 were among the top inhibitory candidates (Table I). IFN-γ and ATP data (Fig. 3A) and ATP-normalized IFN-γ data (Fig. 3C) are presented in Supplemental Tables I, II, respectively.

Candidate emerging from the shRNA screen fell into diverse functional groups (Table I). Among the positive regulators of IFN-γ were proteins involved in M. tuberculosis susceptibility (IL-23R and IL-12RB2) or enhanced T cell immunoreactivity (TNFSF13B/BAFF). Knockdown (KD) of the Th17 lineage–specific factor RORC decreased Th1* viability and IFN-γ production, in agreement with a previous report (21). Candidates included proteins previously implicated in innate and adaptive immunity (Table I), including the HSPE1-encoded chaperonin (HSP10), a potential therapeutic target in autoimmune diseases (34).

CD160 was a high-ranking positive regulator emerging from the screen (Table I). CD160 binds MHC class I molecules and herpes virus entry mediator (HVEM) and is critical for IFN-γ production in NK cells (35). The function of CD160 in T cells is less well understood and is linked to stimulatory and inhibitory signals (36–40). Validation experiments confirmed that CD160 shRNA significantly decreased IFN-γ production in Th1* cells isolated from different donors (Fig. 3D). We identified multiple individual CD160 shRNAs that reduced target gene and protein expression without affecting cell viability (Fig. 4A, 4B, Supplemental Table III). Reminiscent of CD4⁺ and CD8⁺ T cells (36), expression of CD160 in Th1* was detectable in a subpopulation of cells (Fig. 4B). Measurements of secreted or intracellular IFN-γ by ELISA and immunostaining confirmed that CD160 KD in Th1* cells decreased IFN-γ production (Fig. 4C–E). Together, data support a stimulatory role for CD160 in memory CD4 T cells.

ISOC1 is an isochorismatase domain-containing protein of unknown function. In prokaryotes, isochorismatase (Enzyme Commission identifier 3.3.2.1) catalyzes the conversion of isochorismate to 2,3-dihydroxybenzoate and pyruvate; however, the metabolic function of its mammalian orthologs remains obscure. Transcriptome profiles demonstrated a 2-fold increase in ISOC1 expression in Th1* cells (10). The effect of ISOC1 depletion in reducing anti-CD3/28–stimulated IFN-γ and IL-17 production was confirmed in bulk CD4 and Th1* cells transduced with individual ISOC-specific shRNAs (Fig. 5).

To explore ISOC1 function, we performed metabolic analysis of ISOC1-deficient CD4⁺ T cells using validated ISOC1 shRNAs that produce efficient KD and minimal toxicity (Fig. 5A, 5B). To examine glycolysis and oxidative phosphorylation, respectively, the ECAR and OCR were measured (Fig. 6A–E). T cell activation is accompanied by rapid upregulation of glycolysis and dynamic changes in ECAR. Accordingly, anti-CD3/28–stimulated CD4 T cells demonstrated markedly increased ECAR versus nonstimulated cells (Fig. 6A). Consistent with the higher energy demands of activated T cells, mitochondrial respiration (OCR) was also increased upon anti-CD3/28 stimulation, as evidenced by a higher basal OCR within 1 h after stimulation (Fig. 6C, 6E). The addition of oligomycin, an inhibitor of mitochondrial ATP synthase, suppressed mitochondrial respiration and enhanced ECAR, a reflection of increased glycolysis as a compensatory mechanism for ATP production. ISOC1 KD cells showed significant reduction in both ECAR and OCR (Fig. 6A–E) compared with controls. The basal and maximal ECAR, the latter measured in the presence of the respiratory inhibitor myxothiazol, were significantly decreased upon depletion of ISOC1. Likewise, basal (i.e., ATP-producing) and maximal (i.e., DNP-stimulated) mitochondrial respiration were partially inhibited. Overall, ISOC1 depletion produced ∼40% inhibition of mitochondrial respiration and glycolytic activity (Fig. 6D, 6E). Because both basal and maximal OCR and ECAR were equally suppressed in ISOC1 KD cells, our results suggest that ISOC1 deficiency broadly perturbs T cell metabolism. Impaired bioenergetics is consistent with reduced proliferative capacity of ISOC1 KD cells (Fig. 6F, 6G). Effects on cell numbers manifested after 6 d in culture (Fig. 6F), which likely reflects the relatively low energy requirements of resting T cells. Delayed proliferation because of ISOC1 depletion also manifested after TCR restimulation, which imposes high energy demands (Fig. 6G).

Next, we investigated how ISOC1 deficiency globally alters CD4 T cell metabolism by applying high-resolution mass spectrometry and untargeted metabolomics using deep [¹³C] isotope labeling (32) (Fig. 7A). A total of 896 liquid chromatography–mass spectrometry isotopically labeled features corresponding to known cellular metabolites were measured in control and ISOC1-depleted T cells (Fig. 7). To identify metabolic perturbations in ISOC1-depleted T cells, the differentially produced [¹³C]-labeled metabolites were subjected to pathway enrichment analysis using MetaboAnalyst (41, 42), identifying the most prominently affected pathways as pyrimidine and purine biosynthesis (Fig. 7D). As both the building blocks for nucleic acid synthesis and bioactive signaling molecules, purine and pyrimidine nucleotides play a critical role in T cell survival, differentiation, and expansion (43). Thus, we reasoned that lower nucleotide levels in ISOC1-deficient T cells impairs IFN-γ production by interfering with multiple biosynthetic pathways and accounting for the decline in bioenergetic parameters (Fig. 6).

To further investigate the relationship between pyrimidine deficiency and diminished effector functions of ISOC1-deficient T cells, rescue experiments were performed by nucleoside supplementation. Nucleosides and pyrimidine bases such as uridine and cytidine are transported across the plasma membrane by facilitated diffusion via equilibrative nucleoside transporters expressed in T cells. Previous studies demonstrate that uridine or cytidine supplementation (at 100–200 μM) interferes with chemical inhibitors of pyrimidine biosynthesis in lymphocytes (44). Exogenous cytidine or uridine was added to the growth media, resulting in potentiation of IFN-γ production in ISOC1 KD cells (Fig. 8A) and negligible effects on control cells. Independent replicate experiments demonstrated significantly augmented IFN-γ response in cytidine-supplemented ISOC1-deficient cells, with the less significant effect of uridine (p < 0.05). ISOC1 KD moderately increased stimulation-dependent early apoptosis, which was reversed by uridine but not cytidine (Fig. 8D, 8E). Supporting the hypothesis that nucleotide deficiency compromises the metabolic demands of ISOC1-deficient T cells, supplementation of either uridine or cytidine stimulated ECAR and OCR (Fig. 8B, 8C). These data suggest that replenishment of nucleotide pools in ISOC1-deficient T cells enables increased energy demands and restores effector functions. Thus, our data highlight a novel regulatory role for ISOC1 and pyrimidine metabolism in TCR-mediated IFN-γ production.

The function of most candidates emerging from unbiased molecular profiling or genome-based network predictions is unknown. To understand how gene expression signatures can influence specific disease outcomes, it is imperative to understand the function of signature genes in a given cell type. In this study, we report a functional analysis of immune signature genes enriched in Th1* cells, also known as Th1/Th17 cells because of coexpression of IFN-γ and IL-17 (10). Genes from the Th1* signature group were prioritized for RNAi screening based upon a predictive gene network analysis pinpointing regulatory and signaling proteins. This “targeted” genomics approach can explain the relatively high number of hits with functional significance in the screen. In addition to T-bet and RORC, our results supported substantial contribution of genes previously implicated in M. tuberculosis and/or regulation of effector functions (Table I).

Among positive regulators of IFN-γ production, CD160 was unexpected as it is linked to an HVEM-dependent inhibitory effect on TCR-mediated CD4⁺ T cell activation (36). However, the role of CD160 in T cell effector function is complex because other studies support a costimulatory function in T cells (37, 38). Likewise, in other immune cells, CD160 acts as both a positive and negative regulator of Ag-induced signals, depending on the specific receptor/ligand involved. In NK cells, CD160 is critical for IFN-γ production (35), whereas it acts as a negative regulator of NKT cells in early innate immune response (45). A previously described unconventional mechanism of HVEM activation upon CD160 engagement (40) further highlights the versatile function of this protein. In the context of M. tuberculosis immunity, CD160 expression is downregulated in CD8 T cells isolated from actively infected TB patients versus LTBI subjects, although differences in cytokine profiles were not detected between these groups (46). Interestingly, detectable CD160 expression is limited to a minor subset (8–10%) of Th1* cells, whereas its downregulation produced an inhibitory effect on IFN-γ production in the entire population (data not shown). This result warrants further investigation. Other potentially interesting candidates emerging from the screen include cancer-relevant apoptotic factors IER5 (a p53 target gene) and mitochondrial chaperonin HSPE1/HSP10 (47, 48). IER5 promotes apoptosis in cancer cells (49) and was one of the markedly downregulated proapoptotic factors in transcriptional profiles of LTBI subjects versus uninfected control (19). Together with concomitant upregulation of antiapoptotic proteins, these data indicate an enhanced propensity of Th1* cells to survival (10, 19). HSP10 is a multifunctional protein that is involved in procaspase 3 activation (47) and has been also implicated in promoting tumorigenesis and alleviating proinflammatory immune responses (34). As a component of the mitochondrial protein quality control system, HSP10 may play an auxiliary role in IFN-γ production via maintaining mitochondrial functional integrity.

With respect to novel potential regulators required for optimal IFN-γ response, we were intrigued by the prominent effect of ISOC1, a recently identified isochorismatase domain-containing protein presumed to elicit metabolic activity. ISOC1 expression is upregulated in several cancers (50–52), and its depletion markedly reduces the proliferation of tumorigenic cells (50, 52). Additionally, ISOC1 was identified among the top 10 differentially expressed genes in a mouse strain highly susceptible to Staphylococcus aureus and other pathogens (53), suggesting a regulatory role in the immune response. However, literature data related to the functional significance of ISOC1 are rare, and its molecular function is unknown. Comprehensive metabolomic profiling of ISOC1 KD CD4 T cells linked ISOC1 deficiency to severely compromised cellular pyrimidine and purine metabolism. Furthermore, bioenergetic flux analysis demonstrated a rapid stimulation of glycolysis by anti-CD3/28 addition that was abrogated in ISOC1 KD cells. Of note, the signaling pathways controlling acute upregulation of glycolysis, known as “glycolytic switch,” in activated hematopoietic cells are incompletely defined, as is its mechanistic link to cytokine production (33, 54–57). It has been suggested that glycolysis drives TCR-mediated IFN-γ synthesis via a posttranscriptional mechanism rather than fueling ATP production (33, 54), providing an explanation as to why glycolysis, which is inefficient compared with oxidative phosphorylation and yields only two molecules of ATP per glucose, is indispensable for activated T cell proliferation and IFN-γ production (33). We observed that mitochondria in stimulated CD4 T cells also rapidly responded to increased energy demands, as indicated by the enhanced oxygen consumption in different metabolic states. Because both mitochondrial and glycolytic function were equally suppressed in ISOC1 KD cells, our results suggest that ISOC1 deficiency broadly perturbs cellular metabolism. This drives the decreased IFN-γ and IL-17 expression, slower proliferation, and accordingly, reduced cellular energy demands. Supporting ISOC1-dependent metabolic defects, supplementation of ISOC1-deficient cells with exogenous nucleosides restored IFN-γ production and potentiated mitochondrial and glycolytic energy transduction. Notably, our data indicate that ISOC1 regulates IFN-γ production in CD4⁺ T cells and the Th1* subset.

Our results are supported by reports of compromised lymphocyte function after treatment with immune-suppressive inhibitors of pyrimidine biosynthesis, which was rescued by supplementation of cytidine or uridine (43, 44, 58). Further highlighting the role of pyrimidine metabolism in T cell abundance, loss-of-function mutations in the gene encoding the de novo pyrimidine synthesis enzyme CTP synthase 1 suppressed lymphocyte proliferation and induced T cell apoptosis, which was restored by the addition of cytidine (59). Our functional analysis of CD4⁺ T cells indicates that compromised nucleotide metabolism has downstream repercussions on CD4 T effector function, glycolysis, and mitochondrial respiration. As noted above, inhibitors of pyrimidine biosynthesis have potential therapeutic applications in alleviating inflammation by restricting lymphocyte proliferation. Conversely, a fully functional memory T cell immune response is critically important to combat M. tuberculosis and other infections. In these scenarios, targeting the pyrimidine supply could be explored as a beneficial approach. In conclusion, this study provides a necessary functional dimension to Th1* immune signature genes and rationalizes further studies of their roles in M. tuberculosis–specific and other immune cells.

We thank David Freeman and Mehak Kaur for general technical assistance and Dr. Benjamin Schmiedel for helpful advice.

The authors have no financial conflicts of interest.