Abstract

   The thioredoxin-1 (Trx1) system is an important contributor to cellular
   redox balance and is a sensor of energy and glucose metabolism. Here we
   show critical c-Myc-dependent activation of the Trx1 system during
   thymocyte and peripheral T-cell proliferation, but repression during
   T-cell quiescence. Deletion of thioredoxin reductase-1 (Txnrd1)
   prevents expansion the CD4^−CD8^− thymocyte population, whereas Txnrd1
   deletion in CD4^+CD8^+ thymocytes does not affect further maturation
   and peripheral homeostasis of αβT cells. However, Txnrd1 is critical
   for expansion of the activated T-cell population during viral and
   parasite infection. Metabolomics show that TrxR1 is essential for the
   last step of nucleotide biosynthesis by donating reducing equivalents
   to ribonucleotide reductase. Impaired availability of
   2′-deoxyribonucleotides induces the DNA damage response and cell cycle
   arrest of Txnrd1-deficient T cells. These results uncover a pivotal
   function of the Trx1 system in metabolic reprogramming of thymic and
   peripheral T cells and provide a rationale for targeting Txnrd1 in
   T-cell leukemia.
     __________________________________________________________________

   Thioredoxin (Trx), Trx reductase, Txnip and NADPH together comprise the
   Trx system. Here the authors make a T cell-specific thioredoxin
   reductase-1 knockout mouse to show how this system reprograms cellular
   metabolism to enable T cell development, proliferation and responses.

Introduction

   Aerobic organisms have developed a complex system of antioxidant
   proteins that maintains a reducing environment in the cell. The
   thioredoxin (Trx) system is one of the major cellular disulfide
   reducing systems comprising NADPH, Trx reductase (TrxR, encoded by
   Txnrd), and Trx (encoded by Txn). TrxR possesses the unique capacity to
   utilize electrons from NADPH to recycle oxidized Trx to its reduced
   form^[42]1,[43]2. The Trx system regulates multiple cellular processes,
   such as gene expression, antioxidant response, apoptosis, and
   proliferation. In particular, it acts as electron donor for the
   antioxidant enzymes peroxiredoxins and methionine sulfoxide
   reductases^[44]3,[45]4, modulates the activity of transcription factors
   such as NF-κB, Ref1, and HIF1α^[46]5–[47]7, and is involved in the
   regulation of apoptosis by suppressing ASK1 function^[48]8. In mammals,
   the glutathione (GSH)–glutaredoxin (Grx) system has many overlapping
   activities with the Trx system due to extensive crosstalk^[49]9. For
   instance, Trx and Grx redundantly deliver reducing equivalents for the
   enzymatic activity of ribonucleotide reductase (RNR)^[50]10. However,
   how antioxidant systems regulate immune cell function and to which
   extent the Trx and Grx systems can compensate for each other functions
   in vivo are poorly understood at present.

   Mammals express three TrxR isozymes that differ in their intracellular
   localization and tissue-specific expression pattern: TrxR1 localized in
   the cytoplasm, TrxR2 localized in mitochondria, and TrxR3 specifically
   expressed in the testis. Moreover, Trx1 and Trx2 are isoforms localized
   in the cytoplasm and in mitochondria respectively^[51]9. This
   implicates distinct roles at different cellular compartments.
   Thioredoxin-interacting protein (Txnip) binds to reduced Trx and acts
   as a negative regulator of Trx functions^[52]11. Txnip knockout in mice
   results in hypoglycemia, hyperinsulinemia, and liver steatosis^[53]12.
   Moreover, disruption of Txnip in obese mice strikingly improves
   hyperglycemia and glucose intolerance^[54]13, demonstrating a crucial
   role of Txnip in metabolic disorders. Independently of its Trx-binding
   function, Txnip represses cellular glucose uptake by both reducing
   Glut1 expression and inducing Glut1 internalization^[55]14–[56]17.
   Txnip expression is mainly mediated by the glucose-sensing
   transcription complexes chREBP–Mlx and MondoA–Mlx, which bind to
   carbohydrate response element on the Txnip promoter^[57]18,[58]19.

   Due to the critical functions of Txnip in regulating glucose
   metabolism, we hypothesized the Txnip–Trx system might play a role in
   the metabolic changes occurring upon T-cell stimulation. Notably, in
   contrast to naive T cells, activated T cells consume large amount of
   glucose and amino acids, thereby adjusting their metabolism toward
   increased glycolysis and glutaminolysis^[59]20,[60]21. Previously, the
   mitochondrial Trx system was described to be dispensable for
   development, maintenance, and proliferation of lymphocytes^[61]22. To
   establish a potential function the cytosolic Trx system in
   T-cell-mediated immunity and metabolism, we generated
   Txnrd1^fl/fl;Cd4-Cre αβ T-cell-specific and tamoxifen (TAM)-inducible
   (Txnrd1^fl/fl;Cre-ERT2) mice. We found that T-cell activation is linked
   to a drastic downregulation of Txnip and an increase in Txn1/Txnrd1
   expression, which is absolutely required for synthesis of
   2′-deoxyribonucleotides during T-cell metabolic reprogramming. These
   results consequently characterized a previously unknown function of the
   cytosolic Trx system in T-cell development and responses.

Results

Txnrd1 is essential for thymic iNKT cell development

   To investigate the function of the Trx system in T cells, we generated
   Txnrd1^fl/fl;Cd4-Cre mice by crossing mice with loxP-flanked Txnrd1
   alleles to mice expressing Cre recombinase from the Cd4 promoter. In
   these mice, deletion of Txnrd1 mainly occurs in CD4^+CD8^+ double
   positive (DP) thymocytes, and consequently both CD4^+ and CD8^+ αβ T
   cells and CD1d-resticted, invariant natural killer T (iNKT) cells lack
   Txnrd1. Cre-mediated deletion in DP and mature thymocytes, and in
   peripheral CD4^+ and CD8^+ T cells from Txnrd1^fl/fl;Cd4-Cre mice was
   complete at the genomic DNA and mRNA levels (Supplementary
   Fig. [62]1a,b). Wild-type (WT) and Txnrd1^fl/fl;Cd4-Cre mice showed
   comparable frequencies and numbers of thymic populations of CD4^−CD8^−
   double-negative (DN), CD4^+CD8^+ DP and CD4^+ and CD8^+ single-positive
   (SP) T cells (Fig. [63]1a). Furthermore, Txnrd1 deficiency had no
   effects on peripheral αβ T cell numbers in spleen, lymph nodes (LNs),
   and liver (Fig. [64]1b and Supplementary Fig. [65]1c). Expectedly, a
   proportion of peripheral CD4^+ and CD8^+ T cell in naive WT mice
   displayed an activated/memory phenotype (i.e., CD62L^hiCD44^hi and
   CD62L^loCD44^hi). However, Txnrd1^fl/fl;Cd4-Cre mice had a considerably
   lower percentage of such cells in the spleen, LNs, and the liver
   (Fig. [66]1c and Supplementary Fig. [67]1d).

Fig. 1.

   [68]Fig. 1
   [69]Open in a new tab

   Txnrd1 is required for thymic iNKT cell development. a–c T-cell
   populations in naive Txnrd1^fl/fl;Cd4-Cre and Txnrd1^fl/fl littermate
   control mice were analyzed by flow cytometry. Representative FACS plots
   (left) and quantification (right) are shown. a Thymic T-cell
   development was assessed by gating on CD4^−CD8^− DN, CD4^+CD8^+ DP,
   CD4^+TCRβ^+ (CD4^+T), and CD8^+ TCRβ^+ (CD8^+T) thymocytes (n = 3–4). b
   Splenic CD4^+ and CD8^+ T cells (n = 3–4). c Expression of CD62L and
   CD44 on splenic CD4^+ (top) and CD8^+ (bottom) T cells (n = 4–5). d, e
   Lethally irradiated WT mice were reconstituted with a 1:1 mixture of WT
   and Txnrd1^fl/fl;Cd4-Cre bone marrow expressing the congenic markers
   CD45.1 and CD45.2, respectively. After reconstitution, the contribution
   of Txnrd1^fl/fl;Cd4-Cre cells to the indicated thymic and splenic T
   cell populations was assessed. Values were normalized to non−Cre
   expressing CD45.2^+CD19^+ B cells. Values below 1 indicate reduced
   contribution of Txnrd1-deleted T cells to the indicated population
   compared to WT cells (n = 4). f iNKT cells gated as
   TCRβ^intCD1d-PBS57-tetramer^+ in the thymi (top), livers (middle), and
   spleens (bottom) of naive Txnrd1^fl/fl;Cd4-Cre (or control) mice
   (n = 4–5). g Analysis of the indicated stages of MACS-enriched,
   CD1d-PBS57-tetramer^+PLZF^+ iNKT cells. Stage 0, 1, 2, and 3 iNKTs were
   gated as CD24^hiCD44^loNK1.1^−, CD24^loCD44^loNK1.1^−,
   CD24^loCD44^hiNK1.1^−, and CD24^loCD44^hiNK1.1^+, respectively
   (n = 5–6). h Depicted are the expression levels of Txn1 and Txnip
   determined by RT-PCR for FACS-sorted ETP (lin^−CD44^hic-Kit^hiCD25^−),
   DN1-2 (lin^−CD44^hic-Kit^hiCD25^int), DN2
   (lin^−CD44^hic-Kit^int/hiCD25^hi), DN2–3 (lin^−CD44^intCD25^hi), DN3A
   (lin^−CD44^−CD28^−CD25^hi), DN3B (lin^−CD44^−CD28^+CD25^hi), DN3–4
   (lin^−CD44^−CD28^+CD25^int), DN4 (lin^−CD44^−CD28^+CD25^−), ISP
   (CD8^+CD24^+TCRβ^−), DP blast (CD4^+CD8^+FSC^hi), DP rest
   (CD4^+CD8^+FSC^lo), CD8^+ and CD4^+ thymocyte populations from WT mice.
   Circular arrows indicate proliferating populations (n = 4). Bar graphs
   show mean + standard deviation. Data are representative of five (a, b),
   four (c, f), and two (d, e, g, h) independent experiments. Student’s t
   test (two-tailed, unpaired) was used to compare Txnrd1^fl/fl and
   Txnrd1^fl/fl;Cd4-Cre groups (a–c, f, g): *P ≤ 0.05; **P ≤ 0.01;
   ***P ≤ 0.001; ****P ≤ 0.0001; ns not significant. One-sample t test
   with a hypothetical value of 1 was used in d: ****P ≤ 0.0001. One-way
   ANOVA adjusted by Tukey’s multiple comparison test was used in e:
   *P ≤ 0.0332; ***P ≤ 0.0002; ****P ≤ 0.0001

   We next assessed the ability of Txnrd1-deficient T cells to refill the
   hematopoietic compartment of lethally irradiated hosts in a competitive
   situation with WT cells. Therefore, we reconstituted irradiated C57BL/6
   mice with an equal ratio of congenically marked donor bone marrow (BM)
   cells from Txnrd1^fl/fl;Cd4-Cre (CD45.2^+) and WT (CD45.1^+) mice. In
   this setting, Txnrd1-deficient CD4^+ and CD8^+ T cells had a clear
   disadvantage compared to WT cells in peripheral organs but not in the
   thymus (Fig. [70]1d and Supplementary Fig. [71]1e). This suggests that
   Txnrd1 is required to refill the peripheral hematopoietic compartment
   but not for thymic selection and maturation. In line with the low
   number of activated/memory T cells in Txnrd1-deficient naive mice
   (Fig. [72]1c and Supplementary Fig. [73]1d), the defect in contribution
   of Txnrd1-deficient T cells to this compartment was even more
   pronounced in a competitive setting (Fig. [74]1e and Supplementary
   Fig. [75]1f). Taken together, these data suggest that Txnrd1 is
   dispensable for selection of conventional αβ DP T cells in the thymus
   and their homeostasis in the periphery. Moreover, Txnrd1 is required
   intrinsically for expansion of T cells in a lymphopenic environment and
   steady state generation of activated/memory T cells.

   In contrast to conventional αβ T cells, we found that Txnrd1-deficient
   iNKT cells were massively reduced in the thymus, liver, and spleen
   (Fig. [76]1f), suggesting a crucial function of Txnrd1 in iNKT cell
   development. iNKT cells are known to arise from DP T cells and undergo
   massive thymic expansion thereafter^[77]23. In the absence of Txnrd1,
   thymic CD1d-PBS57-tetramer^+PLZF^+ iNKT cell numbers were unaffected at
   stage 0 (CD24^hiCD44^loNK1.1^−) but were reduced at stages 1
   (CD24^loCD44^loNK1.1^−), 2 (CD24^loCD44^hiNK1.1^−), and 3
   (CD24^loCD44^hiNK1.1^+), indicating a defect in expansion rather than
   failed thymic selection (Fig. [78]1g). In contrast to thymic iNKT
   cells, expansion of T cell precursors occurs mainly at the DN stage,
   and therefore prior Txnrd1 deletion in Txnrd1^fl/fl;Cd4-Cre mice.
   Interestingly, by examining expression of the three main components of
   the Trx1 system including TrxR1, Trx1, and Txnip, the inhibitor of
   Trx1, we found that both Txn1 and Txnrd1 expression was increased in DN
   compared to DP and SP T cells. In contrast, Txnip was predominantly
   expressed in DP and SP compared to DN T cells (Supplementary
   Fig. [79]1g). DN precursors undergo two separate phases of
   proliferation. Pre-β proliferation occurs prior to TCR β-chain
   rearrangement at the DN2 stage followed by pre-α proliferation before
   TCRα chain rearrangement at the DN4 stage^[80]24. Precursor
   proliferation ceases at a stage characterized as
   CD4^+CD8^+TCR^−^/^loFCS^hi (DP blasts) and it comes to a rest at the
   CD4^+CD8^+TCR^−^/^loFCS^lo stage (DP rest). To better resolve the
   expression profiles of the Trx system components during thymic
   development, we FACS-sorted distinct thymocyte populations. We
   consistently observed that phases of precursor proliferation (i.e., DN2
   and DN4) were associated with low levels of Txnip but high levels of
   both Txn1 and Txnrd1. By contrast, cells in resting phases (i.e., DN1,
   DN3, DP, and SP) displayed high Txnip, but low Txn1 and Txnrd1
   expression (Fig. [81]1h and Supplementary Fig. [82]1h). These data
   indicate that cell proliferation may be licensed by a high
   Txnrd1–Txn1:Txnip ratio and that TrxR1 may play a role for expansion of
   DN precursors, which would be missed in Txnrd1^fl/fl;Cd4-Cre mice.

Txnrd1 is essential for expansion of DN thymocytes

   To directly assess the effects of Txnrd1 deletion in hematopoietic
   precursors, we generated Txnrd1^fl/fl;Cre-ERT2 mice that allow Txnrd1
   gene deletion at will by TAM injection. We evaluated the function of
   Txnrd1 in early thymocyte development 2 weeks after TAM treatment.
   Txnrd1 was completely deleted in T cells from Cre-expressing mice at
   the genomic DNA and mRNA levels upon TAM administration (Supplementary
   Fig. [83]2a,b). These mice had visibly smaller thymi (Fig. [84]2a) due
   to a massively reduced cellularity (Fig. [85]2b), which affected all
   thymocyte subsets, including DN and DP precursors, mature SP αβ T
   cells, γδ T cells, as well as iNKTs (Fig. [86]2c, d). Importantly, we
   excluded the possibility that reduction of thymocytes occurred due to
   Cre-mediated toxicity rather than deletion of Txnrd1 in DN cells
   (Supplementary Fig. [87]2c). The massive reduction of DP precursors is
   consistent with the proliferation occurring during CD44^−CD25^− (DN4)
   to DP phase (pre-α proliferation). In line with this, DN4 thymocytes
   lacking Txnrd1 showed defective proliferation, measured as
   incorporation of the thymidine analog 5-Ethynyl-2′-deoxyuridine (EdU)
   (Fig. [88]2e). Analysis of known DN precursor subsets (i.e.,
   CD44^+CD25^− [DN1], CD44^+ CD25^+ [DN2], CD44^−CD25^+ [DN3] and DN4)
   showed that each was present in reduced numbers in TAM-treated
   Txnrd1^fl/fl;Cre-ERT2 mice, although DN2–DN3 stage where pre-β
   proliferation occurs was affected the most, highlighting the
   significance of Txnrd1 during proliferation (Fig. [89]2f).

Fig. 2.

   [90]Fig. 2
   [91]Open in a new tab

   Txnrd1 is critically required for thymic T-cell development.
   Txnrd1^fl/fl;Cre-ERT2 mice (and Txnrd1^fl/fl littermates as controls)
   were analyzed by flow cytometry 2 weeks after TAM-mediated Txnrd1
   deletion (n = 4–5). a Representative image of thymi (scale bar = 1 cm).
   b Total number of thymocytes. c Absolute cell counts of
   TCRβ^intCD1d-PBS57-tetramer^+ iNKT and TCRβ^-TCRγδ^+ (γδ T) cells in
   the thymus. d Representative FACS plots (left) and absolute numbers of
   CD4^−CD8^− DN, CD4^+CD8^+ DP, CD4^+TCRβ^+ (CD4^+ T), and CD8^+TCRβ^+
   (CD8^+ T) thymocytes (right). e Proliferation of DN4 thymocytes
   (CD25^−CD44^−) assessed by thymic EdU incorporation. f Representative
   FACS plots (left) and absolute cell numbers of CD25^−CD44^+ (DN1),
   CD25^+CD44^+ (DN2), CD25^+CD44^− (DN3), and CD25^−CD44^− (DN4; right).
   Bar graphs show mean + standard deviation. Numbers in the bar graphs
   next to the arrows indicate fold-changes between Txnrd1^fl/fl and
   Txnrd1^fl/fl;Cre-ERT2 groups (d, f). Data are representative of three
   (a–d, f) and two (e) independent experiments. Student’s t test
   (two-tailed, unpaired) was used to compare Txnrd1^fl/fl and
   Txnrd1^fl/fl;Cre-ERT2 groups (b–f): *P ≤ 0.05; **P ≤ 0.01;
   ***P ≤ 0.001; ****P ≤ 0.0001; ns not significant

   To assess whether Txnrd1 is cell-intrinsically required in T cells
   during development, we reconstituted irradiated WT mice (CD45.1^+) with
   an equal ratio of congenically marked BM cells from WT
   (CD45.1^+CD45.2^+) and TAM-treated Txnrd1^fl/fl;Cre-ERT2 (CD45.2^+) or
   control Txnrd1^fl/fl (CD45.2^+) mice. In line with the previous
   observation, we found reduced contribution of Txnrd1-deficient cells
   during all stages of T cell development with the most prominent
   decrease from the DN3 stage (Supplementary Fig. [92]2d). As expected,
   abolished thymocyte development resulted in an almost complete absence
   of T cells in the periphery (Supplementary Fig. [93]2e). Taken all
   together, these results indicate that the Trx1 system has a critical
   function for proliferation during thymic development, in line with the
   expression levels of Txn1, Txnrd1, and Txnip in the different thymic
   populations (Fig. [94]1h and Supplementary Fig. [95]1h).

Txnrd1 is critical for T-cell immune responses in vivo

   Given the requirement for expansion of T-cell precursors but not the
   homeostatic maintenance of naive αβ T cells, we next investigated the
   requirement of the gene during an in vivo immune response. To address
   this, we infected mice with lymphocytic choriomeningitis virus
   (LCMV)-WE strain. We found a drastic reduction in activated
   CD62L^−CD4^+ and CD62L^−CD8^+ T cells in spleen and blood of
   Txnrd1^fl/fl;Cd4-Cre compared to control mice (Fig. [96]3a and
   Supplementary Fig. [97]3a). Moreover, virus-specific CD4^+ and CD8^+ T
   cells were undetectable by gp[61–80] and gp[33–41] tetramer staining,
   suggesting a crucial role of Txnrd1 in T-cell expansion (Fig. [98]3b
   and Supplementary Fig. [99]3b). To assess whether this impaired
   expansion in Txnrd1-deficient T cells arose from defective in vivo
   cycling or increased cell death, we next measured cell proliferation
   during the course of a viral infection by incorporation of EdU. A
   drastic reduction in EdU incorporation was observed in
   Txnrd1^fl/fl;Cd4-Cre T cells at all time points analyzed with the
   largest difference at the peak of expansion (7 days post infection;
   Fig. [100]3c and Supplementary Fig. [101]3c). Moreover, staining of
   7-AAD and Annexin-V during infection showed comparable frequencies of
   dead/dying CD4^+ and CD8^+ T cells (Supplementary Fig. [102]3d,e). In
   line with the lack of virus-specific T cells, Txnrd1^fl/fl;Cd4-Cre mice
   failed to clear LCMV unlike control mice (Fig. [103]3d). Similarly,
   Txnrd1 was also essential for Th1-mediated protection against the
   protozoan parasite L. major, as judged by increased footpad swelling
   (Fig. [104]3e), elevated parasite load (Fig. [105]3f), and impaired
   numbers of activated CD4^+ and CD8^+ T cells (Fig. [106]3g) in foodtpad
   and draining LNs of infected Txnrd1^fl/fl;Cd4-Cre mice. Taken all
   together, these data suggest a pivotal role of the Trx1 system in
   expanding T cells during in vivo immune responses.

Fig. 3.

   [107]Fig. 3
   [108]Open in a new tab

   Txnrd1-deficient T cells fail to expand following viral and parasite
   infection. a–c Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre mice were infected
   i.v. with 200 focus-forming units (FFU) LCMV WE. a, b On day 7 of
   infection, activation of CD4^+ and CD8^+ T cells was assessed by
   downregulation of CD62L (a), and virus-specific T cells were stained
   with gp[33–41] and gp[61–80] tetramers (b). c Proliferation of CD8^+ T
   cells was assessed by EdU incorporation in spleens of LCMV-infected
   mice at the indicated days of infection. d LCMV titers over time were
   determined by focus assay of blood from mice infected with 500 FFU LCMV
   WE i.v. at indicated times after infection. Dotted line represents the
   detection limit (DtL; n = 5). e–g Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre
   mice were inoculated with 2 millions L. major promastigotes into the
   right and left hind footpad. e Footpad swelling. f Parasite load in
   infected footpads and dLNs 25 days after infection. g Total number of
   activated (CD62L^−CD44^+) CD4^+ and CD8^+ T cells in footpads (left)
   and dLNs (right) 25 days after infection. Means + standard deviations
   and +standard error of the mean are shown for a–c, f, g, and for e,
   respectively. Data are representative of two independent experiments
   (a–g). Student’s t test (two-tailed, unpaired) was used to compare
   Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre groups (a–c, f, g): *P ≤ 0.05;
   **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Two-way ANOVA adjusted by
   Bonferroni’s multiple comparison test was used to compare Txnrd1^fl/fl
   and Txnrd1^fl/fl;Cd4-Cre groups in e: **P ≤ 0.0021; ****P ≤ 0.0001

Induction of the Trx system upon T-cell activation

   To further elucidate mechanisms of defective T-cell expansion, we
   studied responses of naive T cells to αCD3/CD28 stimulation in vitro.
   TCR and IL-2R triggering are known to induce c-Myc which is essential
   for proliferation of T cells (as well as cancer and embryonic stem
   cells)^[109]25,[110]26 and metabolic reprogramming^[111]21. Expectedly,
   c-Myc expression was upregulated upon T-cell activation (Fig. [112]4a
   and Supplementary Fig. [113]4a). Concomitantly, both Txn1 and Txnrd1
   but not Grx1 expression were upregulated (Fig. [114]4b, c), while the
   prominent expression of Txnip in naive T cells was abrogated within 4 h
   of T-cell activation (Fig. [115]4d), suggesting that the Trx system is
   the critical pathway controlling T-cell activation and expansion.
   Txnrd1 deficiency did not alter Txnip, Txn1, and c-Myc expression
   indicating that TrxR1 acts downstream and is not involved in T-cell
   activation (Fig. [116]4a–d). To assess a potential regulation of Txnip
   by c-Myc^[117]27, we used JQ1, a transcriptional inhibitor of c-Myc.
   Indeed, JQ1 treatment of T-cell stimulation cultures dose-dependently
   inhibited c-Myc protein production and cell surface expression of CD71,
   a well-established c-Myc target gene (Fig. [118]4e). Notably, c-Myc
   blockade prevented Txnip downregulation induced by T-cell activation
   (Fig. [119]4f), indicating that TCR-triggering induction of c-Myc is
   responsible for Txnip repression.

Fig. 4.

   [120]Fig. 4
   [121]Open in a new tab

   Txnrd1-deficient T cells have a proliferation defect independent of
   ROS. Magnetically sorted T cells from spleen and peripheral LNs of
   Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre mice were stimulated in
   αCD3/CD28-coated plates. a, b Shown are the expression levels of c-Myc
   (a), Txn1 and Txnrd1 (b) determined by RT-PCR in CD4^+ T cells
   (n = 3–4). c Normalized copy numbers of Txn1 and Grx1 determined by
   RT-PCR. The obtained Ct values were converted to DNA concentration
   using a standard curve (n = 3–4). d Expression level of Txnip
   determined by RT-PCR in CD4^+ T cells (n = 3–4). e c-Myc (left) and
   CD71 (right) protein expression levels measured by flow cytometry after
   incubation of stimulating CD4^+ T cells with the indicated
   concentrations of the c-Myc inhibitor (+)-JQ1 for 10 h. f Txnip
   expression levels in CD4^+ T cells stimulated in the presence of the
   indicated concentrations of (+)-JQ1 (n = 3). g T cell counts at
   indicated times (n = 2). h CFSE dilution of CD8^+ T cells measured by
   flow cytometry at the indicated times of stimulation. i CFSE-dilution
   profiles were used to calculate mean division time (t[division];
   reciprocal of the slope) and time to first division (t[1°division];
   intersection with mean division number y = 1) with the precursor cohort
   method. j CD8^+ T cells leaving G[0] phase and entering cell cycle were
   detected by staining for Ki-67 at the indicated times of αCD3/CD28
   stimulation. Numbers in the plots indicate average percentage of
   Ki-67^+ cells from triplicate samples. k CD8^+ T cells were treated
   with the antioxidants ascorbic acid (0.8, 4, 20 μM),
   Catalase-polyethylene glycol (catalase; 278, 1136, 4545 U/ml),
   Diphenyleneiodonium chloride (DPI; 62.5, 250, 1000 nM),
   dl-Dithiothreitol (DTT; 111, 333, 1000 μM) and N-acetyl-l-cysteine
   (NAC; 2, 8, 32 mM) during stimulation and counted after 3 days (n = 4).
   l Expression of indicated activation markers in CD8^+ T cells measured
   by flow cytometry during the first 8 h of stimulation (n = 3). Dots and
   bar graphs are mean ± standard deviation. Data are representative of
   two (a–g, k, l), five (h) and three (i, j) independent experiments.
   One-way ANOVA followed by Dunnett’s correction was used in a, b, d:
   **P ≤ 0.0021; ****P ≤ 0.0001. Student’s t test (two-tailed, unpaired)
   was used in c, g: *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.
   One-way ANOVA adjusted by Tukey’s multiple comparison test was used in
   f: ****P ≤ 0.0001; ns not significant

   In line with the in vivo proliferation defect, we observed lower
   numbers of T cells lacking Txnrd1 compared to control T cells upon
   stimulation (Fig. [122]4g). This defect was confirmed by CFSE dilution
   analysis (Fig. [123]4h and Supplementary Fig. [124]4b). The application
   of the precursor cohort method^[125]28 on this data allowed calculation
   of the mean division time (t[division]) and the time to perform the
   first cell division (t[1°division]) upon T-cell stimulation. We found
   that the mean division time in both CD4^+ and CD8^+ T cells lacking
   Txnrd1 was ~30% longer. Moreover, completion of the first cell division
   was also delayed in the absence of Txnrd1 (Fig. [126]4i and
   Supplementary Fig. [127]4c), suggesting that additional defects in cell
   cycle entry might play a role. Consistently, stimulated
   Txnrd1-deficient T cells had a delayed upregulation of Ki-67, an
   indicator of cell cycle entry (Fig. [128]4j and Supplementary
   Fig. [129]4d). Finally, we also evaluated whether increased cell death
   in T cells lacking Txnrd1 might contribute to impaired expansion. We
   found that viability of Txnrd1-deficient T cells was marginally
   affected (Supplementary Fig. [130]4e). Yet, the use of classical
   apoptosis, necroptosis, ferroptosis inhibitors, or anti-oxidants was
   insufficient to rescue this death (Supplementary Fig. [131]4f) and
   proliferation (not shown) indicating that absence of Txnrd1 did not
   activate a pathway of controlled cell death.

   The Trx system is a major player in cellular antioxidant
   responses^[132]2. To test whether the defective proliferation in
   Txnrd1-deficient T cells is a consequence of increased cellular
   reactive oxygen species (ROS) levels, we stimulated T cells from
   Txnrd1^fl/fl;Cd4-Cre and Txnrd1^fl/fl mice in the presence of classical
   antioxidants. However, supplementation failed to restore proliferation
   of Txnrd1-deficient T cells (Fig. [133]4k). Moreover, generation of
   cellular ROS detected by CM-H[2]DCFDA staining in stimulated T cells
   was comparable (Supplementary Fig. [134]4g).

   Since secreted Trx has been reported to induce IL-2R^[135]29, we
   stimulated mixtures of T cells from congenically marked WT (CD45.1^+)
   and Txnrd1^fl/fl;Cd4-Cre (CD45.2^+) mice at different ratios. Presence
   of Txnrd1-competent T cells did not rescue the defective expansion of
   Txnrd1-deficient T cells (Supplementary Fig. [136]4h), indicating a
   cell-intrinsic requirement of TrxR1. Furthermore, CD25, CD44, and CD69
   upregulation was comparable between Txnrd1-deficient and control T
   cells upon stimulation (Fig. [137]4l and Supplementary Fig. [138]4i),
   excluding impaired activation as a cause for defective expansion.

   Overall, these results indicate that T-cell proliferation requires
   c-Myc-dependent unlocking of Trx1 from Txnip-mediated inhibition and
   that Txnrd1 deficiency delays cell cycle entry and slows down cell
   cycling independent of T-cell activation, cell death, and cellular ROS.

Txnrd1 is required for nucleotide biosynthesis in T cells

   Growing evidence demonstrates that T-cell activation is coupled to
   drastic changes in cellular metabolism that regulate the
   differentiation and function of activated T cells^[139]20,[140]30. To
   explore whether the defective proliferation of Txnrd1-deficient T cells
   was the result of impaired metabolic reprogramming, we stimulated naive
   T cells in vitro and measured oxygen consumption rate (OCR), an
   indicator of oxidative phosphorylation (OXPHOS), and extracellular
   acidification rate (ECAR), a measure of aerobic glycolysis. We observed
   that Txnrd1-deficient T cells displayed substantially higher basal and
   maximal OCR compared to control cells. Moreover, they also had higher
   baseline ECAR as well as maximal glycolytic capacity (Fig. [141]5a),
   indicating that Txnrd1-deficient T cells were metabolically more active
   compared to control cells. Phosphorylation of mTOR was comparable
   between Txnrd1-deficient and control T cells upon stimulation
   (Supplementary Fig. [142]5a). However, we observed increased
   phosphorylation of the metabolic stress sensor 5′AMP-activated protein
   kinase (AMPK) in the absence of the Trx system (Fig. [143]5b). We
   additionally measured decreased ATP-to-AMP ratio in Txnrd1-deficient T
   cells upon stimulation, indicating that AMPK activation is due to
   energy crisis (Fig. [144]5c). AMPK is well known to respond to nutrient
   availability and to reprogram cellular metabolism toward increased
   glycolysis, β-oxidation and mitochondrial function^[145]31. These data
   are therefore consistent with and explain the higher ECAR and OCR
   observed in Txnrd1-deficient T cells (Fig. [146]5a).

Fig. 5.

   [147]Fig. 5
   [148]Open in a new tab

   Txnrd1-deficient T cells have defective nucleotide biosynthetic
   pathway. Magnetically sorted T cells from spleen and peripheral LNs of
   Txnrd1^fl/fl (WT) and Txnrd1^fl/fl;Cd4-Cre mice (KO) were stimulated in
   αCD3/CD28-coated plates for 2 days in vitro. a Oxygen consumption rate
   (OCR; left) and extracellular acidification rate (ECAR; right) were
   measured with a Seahorse Extracellular Flux XF-96 analyzer under basal
   conditions and following addition of the indicated metabolic inhibitors
   (n = 6; data are mean + standard deviation). b Western blot of p-AMPK
   expression with β-actin as a loading control in T cells. c–e
   Metabolites were analyzed by non-targeted metabolomics (four replicates
   were examined, each analyzed twice). c Depicted is the ATP-to-AMP
   ratio. d Metabolic pathway enrichment analysis comparing WT and KO T
   cells both before and after stimulation. KO > WT and WT > KO indicate
   higher metabolite levels in KO and WT T cells, respectively. e Box
   plots displaying cellular ribonucleotide levels in WT and KO T cells.
   Numbers in brackets represent the mass-to-charge ratio of the detected
   ions. f Ribonucleotide and 2′-deoxyribonucleotide metabolites were
   analyzed by targeted metabolomics at 0 and 48 h after stimulation
   (n = 4). Bar graphs are mean + standard deviation (c, f). The boxes
   represent the 25th to 75th percentiles, the horizontal lines within the
   boxes indicate the medians, and the whiskers are drawn to indicate the
   1.5 × Inter Quartile Range (3rd quartile–1st quartile; e). Data are
   representative of two independent experiments (a–f). Student’s t test
   (two-tailed, unpaired) was used in c, f: *P ≤ 0.05; **P ≤ 0.01;
   ****P ≤ 0.0001. P-values were adjusted using Benjamini–Hochberg
   procedure (d). Significance of the box plot data was assessed using a
   Student’s t test (two-tailed, unpaired) corrected by the
   Benjamini–Hochberg multiple testing (e)

   To perform a more comprehensive analysis of the metabolism in
   proliferating Txnrd1-deficient T cells and to understand the cause of
   the observed metabolic stress, we next utilized high-throughput,
   non-targeted, flow injection time-of-light metabolomics^[149]32. This
   method allows rapid and deep profiling of metabolites involved in
   central carbon metabolism and in peripheral pathways, such as urea
   cycle, amino acid metabolism, and nucleotide biosynthesis. We
   stimulated WT and Txnrd1-deficient T cells in vitro and extracted polar
   metabolites for metabolomics analysis. We identified a total of 989
   distinct ion species and putatively annotated them to metabolites
   listed in the human metabolome database (HMDB) version 3.0^[150]33. Out
   of 989 metabolites, 101 were significantly changed between
   Txnrd1-deficient and control T cells upon stimulation
   (|Log2(fold-change)| > 0.5; adjusted p[Benjamini–Hochberg] < 0.01), but
   only eight in unstimulated conditions, indicating that the absence of
   the Trx system influences the metabolism of T cells mainly after
   stimulation (Supplementary Fig. [151]5b and Supplementary Data [152]1).
   To obtain a general overview of the metabolic differences, we performed
   a pathway enrichment analysis of the metabolites changing in
   Txnrd1-deficient versus control T cells both before and after
   stimulation. This revealed significantly increased levels of
   metabolites involved in purine and pyrimidine metabolisms in
   Txnrd1-deficient T cells (Fig. [153]5d). In addition, Txnrd1-deficient
   T cells also showed significant accumulation of metabolites involved in
   the aspartate and glutamate metabolism (Fig. [154]5d and Supplementary
   Fig. [155]5c). However, no significant enrichment for metabolic
   pathways that have been described to regulate immune cell function was
   observed, such as glycolysis, tricarboxylic acid (TCA) cycle, and fatty
   acid synthesis and oxidation (Fig. [156]5d and Supplementary
   Fig. [157]5d,e). Among the components of purine and pyrimidine
   biosynthetic pathways, we observed accumulation of nucleotides
   containing ribose sugar, suggesting that the final reduction of RNA
   into DNA building blocks during DNA biosynthesis was defective
   (Fig. [158]5e). To confirm this, we performed a targeted metabolomics
   study measuring the cellular levels of the abundant purine
   ribonucleotides and 2′-deoxyribonucleotides. As expected, T-cell
   stimulation in the absence of the Trx system led to accumulation of
   ribonucleotides but reduced abundance of their 2′-deoxyribonucleotide
   forms (Fig. [159]5f). Taken together, these data imply that the Trx
   system in T cells regulates the reduction of RNA into DNA building
   blocks at the last step of DNA biosynthesis.

Increased Cdkn1a expression in Txnrd1-deficient T cells

   To further investigate the cellular function of TrxR1 in T cells, we
   performed RNA sequencing (RNA-seq) of in vivo-activated T cells. Since
   virus-specific T cells failed to expand in the absence of Txnrd1 during
   LCMV infection (Fig. [160]3b and Supplementary Fig. [161]3b), this
   model would not have allowed us to isolate enough cells for the
   experiment. To circumvent this issue, we injected SEB superantigen to
   obtain enough activated T cells for transcriptomics. After 2 days post
   SEB injection, Vβ8.1/8.2^+CD8^+ T cells from Txnrd1^fl/fl;Cd4-Cre and
   Txnrd1^fl/fl controls were FACS-sorted for analysis. Total CD8^+ cells
   from PBS-treated animals served as a control. Similar to αCD3/CD28 T
   cell stimulation in vitro, superantigen-mediated T cell activation in
   vivo resulted in downregulation of Txnip and upregulation of Txnrd1 and
   Txn1 expression (Fig. [162]6a). Moreover, the frequency of Vβ8.1/8.2^+
   among CD8^+ T cells was significantly reduced in Txnrd1^fl/fl;Cd4-Cre
   upon SEB administration but not in the PBS-treated group, indicating
   defective expansion in the absence of Txnrd1 (Fig. [163]6b and
   Supplementary Fig. [164]6).

Fig. 6.

   [165]Fig. 6
   [166]Open in a new tab

   Txnrd1 deficiency increases expression of the cell cycle inhibitor
   Cdkn1a. a–d Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre mice were treated
   i.p. with 100 μg of Staphylococcal enterotoxin B (SEB) or PBS.
   Vβ8.1/Vβ8.2^+CD8^+ T cells were FACS sorted at day 2, and gene
   expression profiles determined by RNA-seq (two biological replicates
   per group). a Expression of Txn1, Txnrd1, and Txnip in Vβ8.1/V8.2^+
   CD8^+ T cells from WT mice after PBS (naive) and SEB treatment. b
   Percentages of Vβ8.1/Vβ8.2^+ cells among CD8^+ T cell population are
   shown. c Heat map showing expression changes between SEB-stimulated
   Txnrd1-deficient and control T cell groups (|GFOLD| > 1.5; FDR < 0.2).
   d Heat map indicating the top differentially expressed genes involved
   in the cell cycle. Only genes with a |GFOLD| larger than 1.5 between
   naive and stimulated groups are displayed. e Magnetically sorted T
   cells from spleen and peripheral LNs of naive Txnrd1^fl/fl and
   Txnrd1^fl/fl;Cd4-Cre mice were stimulated in αCD3/CD28-coated plates.
   Cdkn1a expression was assessed using RT-PCR at the indicated times
   (n = 2–3). Bar graphs show mean + standard deviation (a, b, e). Data
   are representative of one (a, c, d) and two (b, e) independent
   experiments. Student’s t test (two-tailed, unpaired) was used to
   compare Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre groups (a, b, e):
   *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns not significant

   Among the top differentially expressed gene sets (|GFOLD| > 1.5; false
   discovery rate [FDR] < 0.2) between SEB-stimulated Txnrd1-deficient and
   control T cell groups, the cell cycle inhibitor Cdkn1a (the gene
   encoding p21; FDR = 0.05) was significantly enriched in
   Txnrd1-deficient T cells (Fig. [167]6c). Analysis of the differentially
   expressed gene sets involved in cell-cycle progression and regulation
   pathways revealed that Cdkn1a was the only gene significantly enriched
   in Txnrd1-deficient T cells upon SEB stimulation (Fig. [168]6d).
   Conversely, levels of Cdkn1a expression in unstimulated T cells were
   comparable between Txnrd1-deficient and -sufficient T cells
   (Fig. [169]6d). Furthermore, Cdkn1a also appeared as the top single
   best predictor, when we used a linear regression model on all detected
   genes in SEB-treated groups as an additional statistical tool
   (Supplementary Table [170]1). Consistent with in vivo RNA-seq results,
   in vitro T cell stimulation with αCD3/CD28 and real-time quantitative
   PCR (RT-PCR) confirmed higher expression levels of Cdkn1a in the
   absence of Txnrd1 (Fig. [171]6e). Since p21 is known to play a critical
   role in the DNA damage response by inducing cell cycle arrest^[172]34,
   these data are consistent with the slower proliferative capacity and
   indicate increased DNA damage response in Txnrd1-deficient T cells.

   In addition to Cdkn1a, the expression levels of the Nrf2 target genes
   Nqo1 and Srxn1 were also increased in Txnrd1-deficient T cells, as
   suggested by RNAseq (Fig. [173]6c) and confirmed by RT-PCR
   (Supplementary Fig. [174]7a,b), indicating oxidative stress-induced
   Nrf2 activation as the response to maintain redox balance.

Increased replication stress in Txnrd1-deficient T cells

   It is well established that low levels of DNA building blocks
   contribute to replication fork stalling and consequently activation of
   the DNA damage response. This process involves phosphorylation of the
   histone family member H[2]AX, leading to accumulation of its
   phosphorylated form, γH[2]AX^[175]35,[176]36. Consistently, stimulation
   of normal T cells in the presence of hydroxyurea (HU) dose-dependently
   increased H[2]AX phosphorylation (Fig. [177]7a). HU specifically blocks
   the catalytic activity of RNR, the enzyme that reduces RNA into DNA
   building blocks in the last step of nucleotide
   biosynthesis^[178]37,[179]38. Indeed, T cells with higher replication
   stress accumulated in the S-phase of the cell cycle (i.e., intermediate
   levels of DNA in the FACS plot) during which DNA building blocks are
   most required (Fig. [180]7a).

Fig. 7.

   [181]Fig. 7
   [182]Open in a new tab

   Activated Txnrd1-deficient T cells have an increased replication
   stress. a–c Magnetically sorted T cells from spleen and peripheral LNs
   of Txnrd1^fl/fl and Txnrd1^fl/fl;Cd4-Cre mice were stimulated in
   αCD3/CD28-coated plates in vitro. Levels of the DNA damage marker
   γH[2]AX and DNA content were assessed by flow cytometry. a CD8^+ WT T
   cells were αCD3/28-stimulated in the presence of hydroxyurea (HU; 0,
   10, 100 μM) for 30 h before analysis. b Cell cycle analysis of CD8^+ T
   cells with representative FACS plot with gating strategy (top;
   pre-gated on CD8^+) and quantification (bottom) showing percentage of
   cells in the indicated cell cycle stages (n = 3). c Representative FACS
   plot of CD8^+ T cells (top) and MFI of γH[2]AX in S phase cells after
   30 h αCD3/28 stimulation (bottom; n = 3). d Txnrd1^fl/fl and
   Txnrd1^fl/fl;Cd4-Cre mice were injected i.v. with αCD3 antibody
   (10 μg), and γH[2]AX were stained in splenic T cells 24 h after
   injection. Representative FACS plots (left) show γH[2]AX expression in
   CD4^+Ki-67^+ or CD8^+Ki-67^+ T cells (+αCD3) compared to total CD4^+ or
   CD8^+ T cells from PBS-treated mice (Control). Bar graphs on the right
   quantify γH[2]AX in CD4^+Ki-67^+ or CD8^+Ki-67^+ T cells from
   αCD3-treated mice (n = 3). e CD8^+ T cells were stimulated in the
   presence of increasing concentrations of BSO. After 3 days of αCD3/28
   stimulation, live cells were counted (n = 4). Bar graphs show
   mean + standard deviation (b–e). Data are representative of two (a, d,
   e) and four (b, c) independent experiments. Student’s t test
   (two-tailed, unpaired) was used to compare Txnrd1^fl/fl and
   Txnrd1^fl/fl;Cd4-Cre groups (b–d): **P ≤ 0.01; ****P ≤ 0.0001

   We hypothesized that defective nucleotide biosynthesis in
   Txnrd1-deficient T cells (Fig. [183]5) would cause increased
   replication stress and phosphorylation of H[2]AX, leading to enhanced
   expression of p21 (as shown in Fig. [184]6), similar to HU-treatment of
   WT T cells. To explore this possibility, we first assessed the role of
   Txnrd1 in the progression of in vitro activated T cells through the
   different cell cycle stages. Notably, T cells lacking Txnrd1
   accumulated in the S-phase upon stimulation and were conversely reduced
   in the G[0]/G[1]- and G[2]/M-phases (Fig. [185]7b). Furthermore,
   Txnrd1-deficient T cells accumulating in the S-phase had higher levels
   of γH[2]AX (Fig. [186]7c) and increased expression of the DNA damage
   sensors Ddb2 and Xpc compared to control cells (Supplementary
   Fig. [187]8a). Comparable γH[2]AX levels in T cells from Txnrd1^fl/fl
   and Cd4-Cre mice excluded the possibility that Cre-mediated
   toxicity/DNA damage rather than Txnrd1 deletion was responsible for
   increased H[2]AX phosphorylation (Supplementary Fig. [188]8b). In line
   with these in vitro results, in vivo activation using α-CD3
   administration similarly led to increased phosphorylation of H[2]AX
   among proliferating cells (Ki-67^+). As a control, unstimulated T cells
   did not accumulate γH[2]AX irrespectively of the presence or absence of
   Txnrd1 (Fig. [189]7d). To address whether the increased γH[2]AX level
   in the absence of the Trx system was due to stalled replication forks,
   we measured phosphorylation of ATR, the sensor that responds to this
   type of DNA damage^[190]39. We observed higher ATR phosphorylation in
   Txnrd1-deficient T cells (Supplementary Fig. [191]8c), in line with the
   lower amount of 2′-deoxyribonucleotides and the increased
   phosphorylation of H[2]AX. Taken together, these data show increased
   replication stress in the absence of the Trx system, thereby mimicking
   the HU-mediated phenotype caused by reduction of DNA building block
   pools.

   Our results are consistent with previous reports indicating the Trx and
   the GSH systems as electron donors for RNR in bacteria^[192]40,[193]41.
   Therefore, we next investigated to which extent the two cellular
   antioxidant systems are involved in nucleotide biosynthesis in T cells.
   We stimulated Txnrd1-deficient and control T cells in the presence of
   L-buthionine-sulfoximine (BSO), which blocks GSH synthesis by
   inhibiting glutamate-cysteine ligase, thereby reducing the cellular GSH
   levels^[194]42. We observed that BSO treatment further reduced the
   expansion capacity of Txnrd1-deficient T cells in a
   concentration-dependent manner. Conversely, control T cells were
   resistant to BSO treatment and were able to expand at BSO
   concentrations that completely blocked proliferation of cells lacking
   Txnrd1 (Fig. [195]7e). BSO-treatment further increased the already
   higher γH[2]AX levels in Txnrd1-deficient T cells in a
   concentration-dependent manner, leading to cell-cycle arrest in the
   S-phase (Supplementary Fig. [196]8d). This demonstrated that blocking
   GSH in addition to the Trx system led to even higher replication
   stress. Notably, BSO treatment on control T cells did not affect levels
   of γH[2]AX (Supplementary Fig. [197]8d). Taken all together, these data
   demonstrate that the Trx system is the major electron donor for RNR
   during nucleotide biosynthesis in stimulated T cells. In contrast, the
   GSH system is dispensable and only inefficiently compensates for the
   lack of the Trx system in T cells.

Discussion

   T-cell activation and proliferation are coupled to drastic changes in
   cellular metabolism to support their increased bioenergetic and
   biosynthetic demands. Although less efficient than OXPHOS at producing
   ATP, activated T cells switch to aerobic glycolysis to generate
   metabolic intermediates that are crucial for cell growth and
   proliferation^[198]20,[199]30,[200]43,[201]44. In particular,
   glucose-6-phosphate generated in the first step of glycolysis can be
   metabolized in the pentose phosphate pathway to generate ribose,
   thereby promoting nucleotide biosynthesis^[202]20,[203]21. RNR
   catalyzes the rate-limiting step of nucleotide biosynthesis by reducing
   ribonucleotides into the corresponding 2′-deoxyribonucleotides^[204]10.
   Both the Trx and Grx systems are known electron donors for this
   reaction^[205]40,[206]41. Previous studies showed that proliferation
   and development of embryonic fibroblasts, cardiomyocytes, and
   hepatocytes was unaffected in the absence of Txnrd1^[207]45–[208]47 and
   redundantly covered by the Grx system^[209]48. Through a
   mass-spectrum-based metabolomics approach, we found impaired pyrimidine
   and purine biosynthesis after T-cell stimulation in the absence of
   Txnrd1 (Fig. [210]5), consistent with the Trx1 system at providing
   reducing equivalents to RNR. Glutamine and aspartate are crucial
   nitrogen donors for synthesis of purines and pyrimidines and can
   potentially accumulate in case RNR-catalyzed reaction is
   impaired^[211]49,[212]50. Interestingly, metabolomics revealed also the
   accumulation of metabolites in the glutamate and aspartate metabolism,
   further indicating defective nucleotide biosynthesis in the absence of
   the Trx1 system. Seahorse-based measurements showed that
   Txnrd1-deficient T cells were metabolically more active, in line with a
   recent publication indicating that mouse embryonic fibroblasts lacking
   Txnrd1 displayed increased metabolic flux, glycogen storage,
   lipogenesis, and adipogenesis^[213]51. Moreover, the higher metabolic
   activity in Txnrd1-deficient T cells is consistent with the observed
   increased phosphorylation of the metabolic stress sensor AMPK, thereby
   boosting glycolytic flux, β-oxidation, and mitochondrial
   function^[214]31. Increased glutamine levels in the absence of Txnrd1
   may also fuel the TCA cycle and potentially contribute to the increased
   OCR. Consistent with impaired nucleotide biosynthesis, we detected
   increased phosphorylation of H[2]AX in the S-phase of the cell cycle in
   the absence of Txnrd1 after stimulation (Fig. [215]7). During normal
   cell-cycle progression, γH[2]AX decorates single-stranded DNA at the
   replication forks. The higher levels of γH[2]AX in the absence of
   Txnrd1, which was mimicked by direct inhibition of RNR by HU, is
   consistent with replication fork stalling events due to reduced
   availability of DNA building blocks^[216]35,[217]36. Moreover,
   expression of the cyclin-dependent kinase inhibitor p21, known to
   participate in DNA damage response downstream of γH[2]AX^[218]34, was
   upregulated (Fig. [219]6). Thus, in the absence of Txnrd1, cellular
   pools of DNA building blocks are depleted, thereby increasing
   replication stress, which in turn promotes p21-dependent inhibition of
   cell cycle progression. Taken together, our results demonstrate that
   the Trx system is the major electron donor for RNR during nucleotide
   biosynthesis in rapidly proliferating T cells during development and
   immune responses. Indeed, blockade of Grx system by targeting
   gamma-glutamylcysteine synthetase with BSO did not affect proliferation
   of WT T cells, while it abrogated the remaining proliferative capacity
   observed in activated Txnrd1-deficient T cells in vitro. Notably, this
   Grx-dependent partial capacity of Txnrd1-deficient T cells to
   proliferate in vitro was not observed upon LCMV infection in vivo,
   where expansion of anti-viral T cells was completely abrogated in the
   absence of Txnrd1. By contrast, Grx1 was poorly expressed in peripheral
   T cells and not upregulated upon T-cell activation, thereby providing
   an explanation why the Grx system cannot fuel DNA synthesis required
   for T-cell expansion.

   Txnip plays a central role in the regulation of glucose metabolism and
   glycolysis by repressing cellular glucose uptake^[220]14–[221]17. Naive
   T cells utilizing low glucose and OXPHOS for energy generation^[222]20
   showed high Txnip expression, while T-cell stimulation resulted in
   rapid Txnip downregulation, in line with activated T cells requiring
   glucose uptake during metabolic reprogramming^[223]30. We found that
   the transcription factor c-Myc, which controls the metabolic
   reprogramming occurring in activated T cells^[224]21, links T-cell
   activation with downregulation of Txnip, in line with other reports
   demonstrating the potential of c-Myc to directly repress Txnip in tumor
   cells^[225]27. Increased glycolysis due to Txnip downregulation may
   dynamically deplete glucose metabolites that are sensed by chREBP–Mlx
   or MondoA–Mlx complexes^[226]16,[227]17 preventing Txnip gene
   expression and thereby generating a positive feedback loop further
   promoting glucose uptake and augmenting the flux of glycolysis^[228]52.
   In addition to facilitate glucose import, diminishment of Txnip
   releases Trx1 from inhibition. Concomitant with Txnip downregulation,
   T-cell activation resulted in Txn1 upregulation. Trx1 is essential for
   the activation of the low energy sensor AMPK during metabolic stress by
   suppressing its oxidation^[229]53. Activated AMPK phosphorylates Txnip
   resulting in its degradation, which, in turn, increases Glut1 function
   and glucose import^[230]15. This mechanism may sustain proliferation in
   activated T cells. Consistently, T cells lacking AMPKα1 show strikingly
   reduced expansion upon viral and bacterial infection in vivo^[231]54,
   similar to Txnrd1-deficient T cells.

   The Trx system has also been proposed to inhibit apoptosis by binding
   and inactivating ASK1^[232]9,[233]55, and to maintain survival of DP
   thymocytes^[234]56. However, our data revealed no relevant increase of
   cell death in Txnrd1-deficient T cells upon activation in vitro and in
   vivo, and the numbers and viability of DP cells were unaffected in
   thymi of Txnrd1^fl/fl;Cd4-Cre mice. Furthermore, as the Trx system is
   one of the main antioxidant pathways in mammalian cells, we originally
   speculated that increased ROS levels and consequent aberrant cellular
   functions were the major cause of the impaired proliferation of
   Txnrd1-deficient T cells. Yet, supplementation of antioxidants was
   found ineffective and ROS levels were comparable in WT and
   Txnrd1-deficient T cells, indicating that defective proliferation is
   independent of cellular ROS. It is feasible that the GSH system acts as
   the main cellular antioxidant^[235]9 and protects T cells from
   ROS-mediated cell death. Indeed, crucial roles of the GSH system in the
   regulation of the antioxidant response and in promoting survival of
   activated T cells have been shown recently. GSH buffers ROS in order to
   promote c-myc-dependent metabolic reprogramming^[236]57. Moreover,
   deficiency of Glutathione peroxidase 4 results in fatal accumulation of
   membrane lipid peroxides and death by ferroptosis upon T-cell
   activation^[237]58. Thus, our results point to the essential
   requirement of the Trx system for donating electrons to RNR during
   rapid T-cell proliferation, while the GSH system is indispensable as
   antioxidant and for survival of activated T cells.

   Trx and TrxR upregulation has been reported in several types of
   cancers, including lymphomas and aggressive T-cell acute lymphoblastic
   leukemia (T-ALL), and it was associated with a poor clinical
   outcome^[238]59,[239]60. The Trx system has been suggested to promote
   development and propagation of cancer by various mechanisms including
   inhibition of apoptosis, promotion of cell growth, and sustainment of
   angiogenesis^[240]59–[241]61. Several chemotherapeutic drugs that
   target TrxR are in the clinics and are considered to mediate antitumor
   activity primarily through induction of oxidative stress and
   apoptosis^[242]62,[243]63. However, abrogation of T-cell expansion in
   the absence of Txnrd1 was not associated with a significant increase in
   ROS production and cell death possibly indicating off-target effects of
   the inhibitors. Indeed, currently, there is no clinical drug that
   targets TrxR1 specifically. Our data suggest that such inhibitors may
   be efficacious and sufficient in treatment of lymphoma and leukemia by
   direct inhibition of DNA synthesis.

   In conclusion, we propose a model where c-Myc-dependent Txnip
   downregulation by T-cell activation is critical for release of active
   Trx1 that in turn provides electrons for the synthesis of
   2′-deoxyribonucleotides independently of the GSH–Grx system.

Methods

Mice

   Txnrd1^fl/fl mice^[244]46 were provided by M. Conrad (Helmholtz
   Zentrum, Munich, Germany) and were backcrossed for more than eight
   generations to C57BL/6. To obtain Txnrd1^fl/fl;Cd4-Cre and
   Txnrd1^fl/fl;Cre-ERT2 mice, Txnrd1^fl/fl mice were crossed with
   Cd4-Cre^[245]64 and Cre-ERT2 mice^[246]65, respectively. B6 Ptprc^a
   (CD45.1) animals were purchased from The Jackson Laboratory (Bar
   Harbor, Maine, USA). About 6–12-week-old age- and sex-matched mice
   (either male or female) were used for the experiments. For deletion of
   Txnrd1 in Txnrd1^fl/fl;Cre-ERT2 mice, mice were treated with 2 mg TAM
   (Sigma-Aldrich) intraperitoneally (i.p.) on two consecutive days. For
   EdU labeling, mice were injected i.p. with 0.5 mg EdU (Life
   Technologies) and analyzed 4 or 16 h later for thymic expansion or
   T-cell proliferation during LCMV infection, respectively. Mice were
   housed in individually ventilated cages under specific pathogen free
   conditions at ETH Phenomics Center (EPIC) (Zurich, Switzerland). All
   animal experiments were approved by the local animal ethics committee
   (Kantonales Veterinärsamt Zürich, licenses 167/2011, ZH270/2014,
   113/2012, ZH135/15, and ZH161/15), and performed according to local
   guidelines (TschV, Zurich) and the Swiss animal protection law (TschG).

Bone marrow chimeras

   CD45.1^+ or CD45.1^+ CD45.2^+ recipients were irradiated twice with
   4.75 Gy with a 4-h break in a RS 2000 (Rad Source Technologies Inc.,
   Alpharetta, USA). The following day, mice were reconstituted by
   intravenous (i.v.) injection of bone marrow cells from hind legs of
   donor mice, treated with antibiotics (0.024% Borgal, MSD Animal Health,
   in the drinking water) for 6 weeks, and used for experiments 8 or more
   weeks after reconstitution. Txnrd1^fl/fl;Cre-ERT2 (and Txnrd1^fl/fl as
   control) donor mice were pre-treated twice with 2 mg TAM i.p. to delete
   Txnrd1 gene before BM transfer.

LCMV and Leishmania major infection

   LCMV WE was originally provided by Rolf Zinkernagel (University of
   Zurich, Zurich, Switzerland) and propagated in L929 cells with virus
   aliquots stored at −80 °C. Mice were i.v. infected with 200 or 500 FFU
   LCMV WE. Blood was diluted 1:5 in MEM medium (Life Technologies) +  2%
   FCS (Gibco) was stored at −80 °C until LCMV titers were determined in
   MC57 cells as previously described^[247]66.

   L. major parasite (MHOM/IL/81/FEBNI) were grown in Schneider’s
   Drosophila medium (Invitrogen) with 20% FBS and 100 U/ml penicillin
   6-potassium and 100 μg/ml streptomycin sulfate. Mice were injected with
   2 millions promastigote parasites into the left and right hind footpad,
   and they were analyzed 25 days after infection. Footpad swelling was
   monitored every 3–4 days. To quantify parasite load in footpads and
   draining LNs, single-cell suspensions were plated on 96-well plates in
   duplicates of threefold serial dilutions and were cultured at 30 °C for
   7 days.

In vivo stimulation

   In total, 100 μg Staphylococcal enterotoxin B (Sigma-Aldrich) was
   injected i.p. per mouse and blood/spleen were analyzed at indicated
   days. For DNA damage analysis in vivo, 10 μg of anti-CD3 (145-2C11;
   home-made) was injected i.v., and splenocytes were isolated 24 h later
   for staining.

Flow cytometry

   The LCMV glycoprotein peptides gp[33–41] (KAVYNFATM), gp[61–80]
   (GLNGPDIYKGVYQFKSVEFD), and CD1d-PBS57 tetramer were kindly provided by
   the National Institutes of Health (Bethesda, MD). Fixable viability dye
   eFluor® 780 (eBioscience) and LIVE/DEAD® fixable yellow dead cell stain
   (Life Technologies) were stained in pure PBS before antibody staining.
   For EdU proliferation experiments, cells from EdU-injected mice were
   labeled before staining with antibodies with the Click-iT Plus EdU
   Alexa Fluor® 488 Flow Cytometry Assay Kit (Life Technologies) according
   to manufacture’s instructions. Antibodies for extracellular stains were
   incubated with cells for 15 min in FACS buffer (PBS + 2% FCS). For
   c-Myc, γH[2]AX, PLZF, pATR, and Ki-67 intracellular stainings, cells
   were fixed and permeabilized with Foxp3 Staining Buffer Set
   (eBioscience) according to manufacture’s instructions and subsequently
   stained for 30 min. The anti-pATR and the anti-c-Myc antibodies were
   detected with an anti-rabbit secondary antibody labeled with the FITC
   fluorophore. To analyze cell death, cells were stained with
   Annexin-V-APC (BD Bioscience) and 7-AAD (eBioscience) in Annexin-V
   binding buffer. For cell cycle analysis, DNA was stained with Fxcycle
   violet stain (Thermo Scientific) according to manufacture’s
   instructions. Total cellular ROS was quantified by adding 1 μM
   CM-H[2]DCFDA (Life Technologies) to cells and measuring fluorescein
   signal after 45 min recovery in supplemented medium. Cells were
   acquired on FACSCanto II or LSRFortessa (BD Bioscience), or sorted on
   FACSAria III (BD Bioscience). Data were analyzed in FlowJo software
   (Tree Star). A complete list of all antibodies and staining reagents
   used in this study can be found in Supplementary Table [248]2. All the
   gating strategies used for flow cytometry plots are shown in
   Supplementary Figs [249]9–[250]13.

Magnetic cell sorting

   For analysis of the different iNKT developmental stages, single thymi
   were MACS-enriched with APC-labeled CD1d-PBS57-tetramer and anti-APC
   microbeads (MACS, Miltenyi Biotec). For T-cell stimulation experiments
   in vitro, T cells were enriched from spleen and lymph node (inguinal,
   axillary, brachial) by positive selection using a MACS system with
   microbeads conjugated to monoclonal anti-mouse CD4 (L3T4), CD8a (Ly-2),
   or CD90.2 (Miltenyi Biotec) following the manufacturer’s instructions.
   Briefly, 100 millions cells/ml were stained with microbeads (1:10
   dilution) for 15 min at 4 °C. After washing, 100 millions cells were
   resuspended in 500 μl of buffer, and the cell suspension was applied to
   LS columns and positively sorted in the magnetic field of a MACS
   separator.

In vitro T-cell stimulation

   MACS-sorted T cells (10^5/well) were seeded in IMDM + GlutaMAX, 10 %
   FCS, 100 U/mL penicillin, 100 μg/mL streptomycin, 50 μM
   β-mercaptoethanol (all Gibco) into 96 well plates pre-coated with
   anti-CD3 (4 μg/ml; 145-2C11; home-made) and anti-CD28 (2 μg/ml; 37.51;
   home-made) for stimulation. For time-course experiments, cells were
   fixed with 4% Formaldehyde in PBS (Sigma-Aldrich). For CFSE-labeling,
   10^7/ml PBS-washed cells were incubated for 10 min at 37 °C with 5 μM
   CFSE (5-(and-6)-Carboxyfluorescein Diacetate, Succinimidyl Ester; life
   technologies) in PBS + 0.1% BSA (bovine serum albumin; Sigma-Aldrich).
   Analysis of proliferation kinetics of CFSE-labeled cells was performed
   as described elsewhere (precursor cohort method^[251]28). Inhibitors
   and compounds were added at the beginning of culture. The following
   substances were used (purchased from Sigma-Aldrich unless indicated
   otherwise): l-ascorbic acid, l-Buthionine-sulfoximine,
   Catalase-polyethylene glycol, Ciclopirox olamine, Deferoxamine,
   diphenyleneiodonium chloride, dl-Dithiothreitol, Ferrostatin-1
   (ChemBridge Corporation), Hydroxyurea, N-Acetyl-l-Cysteine,
   Necrostatin-1 (inactive) (Merck), (±)-α-Tocopherol,
   Z-Val-Ala-Asp(OMe)-CH2F (zVAD-FMK; Peptitde Institute), (+)-JQ1.

RNA analysis by real-time quantitative PCR

   Total RNA was extracted using TRIzol (Life Technologies), followed by
   reverse transcription using GoScript Reverse Transcriptase (Promega)
   according to the manufacturer’s instructions. Real-time quantitative
   PCR (RT-PCR) was performed using Brilliant SYBR Green (Stratagene) on
   an i-Cycler (Bio-Rad Laboratories) according to manufacturer’s
   protocol. Expression was normalized to the housekeeping gene Tbp for
   mRNA expression, or to genomic Txnrd1 for addressing DNA recombination
   efficiency in Txnrd1-deficient cells. The sequences of all used primers
   are listed in Supplementary Table [252]3.

Immunoblotting

   Purified splenic T cells were lysed on ice with RIPA buffer (20 mM
   Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 1 mM Na[3]VO[4], 1% Triton
   X-100, supplemented with protease inhibitor [Sigma-Aldrich] and
   phosphatase inhibitor [Sigma-Aldrich]). Samples were then spun for
   10 min at 4 °C to remove all cell debris. Protein concentrations were
   determined using the Pierce^TM BCA Protein Assay Kit (Thermo
   Scientific). Whole cell extracts (30 μg of proteins) were fractionated
   by SDS-PAGE and transferred to a Polyvinylidene difluoride (PVDF)
   membrane using a transfer apparatus according to manufacturer’s
   instructions (Bio-Rad). After blocking with 4% nonfat milk in TBST
   (50 mM Tris, pH 8.0, 150 mM NaCl, 0.1% Tween 20) for 45 min, the
   membrane was washed once with TBST and incubated with primary
   antibodies (1:1000 in TBST with 4% BSA) at 4 °C for 12 h. Membranes
   were washed three times for 10 min and incubated with a 1:2000 dilution
   (in TBST with 4% nonfat milk) of horseradish peroxidase-conjugated
   anti-rabbit antibody for 1 h. Blots were washed with TBST three times
   and developed with the ECL system (Thermo Scientific) according to
   manufacturer’s instructions. Antibodies for phospho-AMPKα (Thr172;
   40H9; Cell Signalling Technology) and β-Actin (AC-15; Sigma-Aldrich)
   were used in this study. The uncropped immunoblots are shown in
   Supplementary Fig. [253]14.

RNA preparation and sequencing

   Mice were injected with 100 μg Staphylococcal enterotoxin B i.p.
   (Sigma-Aldrich; or PBS as control). Vβ8.1/8.2^+ CD8^+ and total CD8^+ T
   cells were FACS-sorted from SEB- and PBS-injected mice, respectively.
   Approximately 300,000 sorted cells were resuspended in TRIzol (life
   technologies), phase separation was achieved by addition of chloroform
   (Sigma-Aldrich) and RNA was precipitated from the aqueous layer with
   isopropanol (Sigma-Aldrich) using glycogen (Roche) as a carrier. The
   TruSeq RNA Stranded Sample Prep kit (Illumina) was used to construct
   the sequencing libraries. In brief, total RNA samples (100 ng) were
   poly A enriched and reverse transcribed into double-stranded cDNA.
   TruSeq adapters were then ligated to double-stranded cDNA. Fragments
   containing TruSeq adapters on both ends were selectively enriched with
   PCR and subsequently sequenced on the Illumina NextSeq 500 in
   single-end mode, 150 cycles at the Functional Genomics Center Zurich
   (FGCZ). The fragments were mapped to the ensemble mouse reference
   genome GRCm38 (Version25.06.2015) using the STAR aligner^[254]67. For
   normalization, the read counts were scaled with the use of the trimmed
   mean of M-values (TMM) method proposed by Robinson and Oshlack^[255]68.
   For differential expression analysis, the Bioconductor package edgeR
   was used. For the generation of the heat maps, genes with a
   |generalized fold change (GFOLD)| larger than 1.5 and a false-discovery
   rate of <0.2 were chosen for visualization. For the analysis of genes
   involved in the cell cycle, the KEGG cell cycle pathway (mmu04110) was
   used.

Extracellular flux analysis

   Extracellular flux assay was performed with a Seahorse XF-96
   Extracellular Flux Analyzer (Seahorse Bioscience) according to
   protocols described elsewhere^[256]69. Oxygen consumption rates (OCR)
   and extracellular acidification rates (ECAR) were measured in
   unbuffered RPMI 1640 with 2 mM l-glutamine containing 25 mM or without
   glucose. Baseline levels and responses to the indicated compounds were
   determined. To address mitochondrial stress, 1 μM oligomycin, 1.5 μM
   FCCP, and 500 nM rotenone + 1 μM antimycin A were provided (all from
   Seahorse Bioscience). For glycolysis stress analysis, 25 mM glucose
   (Sigma-Aldrich), 1 μM oligomycin, and 45 mM 2′-deoxy-d-glucose
   (Sigma-Aldrich) were added.

Metabolomics

   Freshly-isolated, MACS-sorted (with CD90.2 microbeads) T cells were
   either analyzed directly after isolation or at different time points
   after αCD3/CD28 stimulation. Approximately 500,000 cells were washed
   twice with PBS and snap frozen in liquid nitrogen. Intracellular
   metabolites were extracted twice with cold (−20 °C) 40:40:20
   acetonitrile:methanol:water. Non-targeted metabolite profiling was
   performed by flow injection analysis on an Agilent 6550 QTOF instrument
   in negative mode as previously reported^[257]32. Ions were putatively
   annotated to metabolites based on the measured mass using the HMDM v3.0
   database^[258]32,[259]33. For targeted metabolomics, intracellular
   nucleotides were extracted with the same procedure, dried in a
   speed-vac and resuspended in 60 μl of ddH[2]O. Relative levels of XMP
   and dXTP were quantified by ion-pairing LC–MS/MS in negative mode with
   multiple reaction monitoring using the MRM settings listed in
   Supplementary Table [260]4. The data were analyzed as previously
   described^[261]70.

Statistical analysis

   Two-group comparisons were assessed with a Student’s t test
   (two-tailed, unpaired). One-sample t test was used to compare the mean
   of a group with specified value. Multi-group comparisons were assessed
   by one-way ANOVA followed by either Tukey’s or Dunnett’s corrections.
   For metabolomics data, the p-values were adjusted using the
   Benjamini–Hochberg multiple testing correction. The data are
   represented as mean + standard deviation, and the method of statistical
   evaluation and the significance levels are indicated in each figure
   legend.

Data availability

   Sequence data that support the findings of this study have been
   deposited in GEO with the primary accession code [262]GSE107090. The
   authors declare that all the other data supporting the findings of this
   study are available within the article and its supplementary
   information files and from the corresponding authors upon reasonable
   request.

Electronic supplementary material

   [263]Supplementary Information^ (6.8MB, pdf)
   [264]Peer Review File^ (186.6KB, pdf)
   [265]41467_2018_4274_MOESM3_ESM.pdf^ (80.3KB, pdf)

   Description of Additional Supplementary Files
   [266]Supplementary Data 1^ (1.7MB, xlsx)

Acknowledgements