Functional Education of Invariant NKT Cells by Dendritic Cell Tuning of SHP-1

Invariant NKT (iNKT) cells are a conserved subset of T lymphocytes that express an invariant TCR (iTCR) α-chain (Vα14-Jα18 in mice) paired with a restricted set of Vβ genes (Vβ8.2, Vβ7, or Vβ2 in mice) (1). This TCR is restricted for the MHC-like molecule CD1d. iNKT cells are prototypical “innate lymphocytes” that acquire effector-memory characteristics in the thymus and display effector functions without prior Ag sensitization (1). iNKT cells undergo a unique developmental program, in which CD4⁺CD8⁺ “double-positive” (DP) precursors are positively selected by CD1d-expressing DP thymocytes (1). Postselection iNKT cells (stage 0 CD24^hiCD44^loNK1.1⁻) undergo subsequent intrathymic differentiation into naive (stage 1, CD44^loNK1.1⁻) and Ag-experienced (stage 2, CD44^hiNK1.1⁻) cells (1). Stage 2 iNKT cells expand dramatically and undergo the final NK maturation (stage 3 CD44^hiNK1.1⁺) in thymus and in periphery (1).

iNKT cells recognize lipid Ags derived from several pathogenic bacteria (2, 3), but are also potently autoreactive against CD1d-expressing APCs that present stress-related endogenous lipid Ags (4–6). Activated iNKT cells rapidly produce different cytokines and license APC functions, which, in turn, stimulate both innate and adaptive immune responses, promoting the control of infectious pathogens, tumors, and autoimmunity (1).

Because of their strong autoreactivity, iNKT cell responsiveness requires regulation to maintain effective host defense while preventing hyperstimulation and, possibly, autoimmunity. Upregulation of agonist self-lipids by APCs under stress/inflammation is one mechanism to limit iNKT cell activation in normal conditions (4–6). It is unclear, however, how the iNKT cell activation threshold is physiologically tuned to lipid Ags generated by microbial infections and stress. iNKT cells developing in transgenic (tg) mice expressing human or murine CD1d only on thymocytes display increased Ag reactivity (7, 8), suggesting that iNKT cells undergo “functional education” to establish TCR-responsiveness by interacting with CD1d⁺ cells other than thymocytes. Multiple CD1d⁺ cells of hematopoietic, epithelial, or endothelial origin might influence this process (9).

In this study, we sought to gain mechanistic insight into the functional education of iNKT cells and identify the CD1d-mediated cellular interactions that regulate their responsiveness.

C57BL/6N mice were from Charles River. pLck-hCD1d, mCD1d^−/−, pLck-mCD1d, μT, and ΔDC mice were previously described (8, 10–12). TCRα^−/− mice (13) were obtained from Centre de Distribution, Typage, Archivage Animal–Centre National de la Recherche Scientifique (Orleans, France). Nonviable motheaten (me/+) (C57BL/6J-Hcph^me/J) mice were purchased from The Jackson Laboratories. Mice were kept in a specific pathogen–free environment. All animal studies were reviewed and approved by the Institutional Animal Care and Use Committee, San Raffaele Scientific Institute (protocol N.509).

iNKT cells were stained with αGalCer-loaded mouse or human CD1d-IgG1 dimers (BD Pharmingen) (8) or with PBS57-loaded mCD1d-PE tetramers (NIH Tetramer Core Facility), and with mAbs specific for TCRβ, CD44, NK1.1, hCD1d, CD3ε, anti-CD45.1, CD24/HSA, CD19, and H-2-A^b (BioLegend). Nonspecific binding was blocked by preincubation with anti-CD32/CD16 mAb. Intracellular SHP-1 or cytokines were detected by staining fixed/permeabilized cells with anti–SHP-1 mAb Y476 (Abcam) and secondary Ab (Southern Biotech), or with anti–IFN-γ or anti–IL-4 mAbs (BioLegend). The specificity of intracellular SHP-1 detection is shown in Supplemental Fig. 2. Samples were acquired on Canto-II flow cytometer (BD Biosciences), excluding aggregates and dead cells, and analyzed using FlowJo (Tree Star). CD69 and SHP-1 expression level was calculated as relative fluorescence intensity (RFI) (sample mean fluorescence intensity/isotype control mean fluorescence intensity).

Thymocytes, splenic lymphocytes, and hepatic lymphocytes were obtained as described previously (8). Thymic iNKT cells (TCRβ^int CD1d-dimers⁺) and peripheral iNKT cells (CD19⁻MHC-II⁻ TCRβ^intCD1d-dimer⁺ or TCRβ^intCD1d-tetramer⁺) were sorted (>95% purity) using a MoFlo cell sorter (Beckman Coulter). Dendritic cells (DCs) were obtained from bone marrow (BM) cells cultured 5 d with GM-CSF (25 ng/ml) and IL-4 (5 ng/ml; R&D Systems) (14).

For iTCR cross-linking, sorted iNKT cells were placed in 96-well plates precoated with 10 μg/well goat anti-rat IgG and 2 μg/ml anti-CD28 mAb (BD Pharmingen). Goat anti-rat IgGs cross-links the iTCR by binding rat anti-mouse IgG1-PE mAbs used to detect the CD1d-dimer⁺ iNKT cells. For activation by DCs, sorted iNKT cells and DCs were cocultured with increasing αGalCer concentrations. PMA/ionomycin stimulation was performed with 25 ng/ml PMA and 1 μg/ml ionomycin (Sigma) for 1 h at 37°C plus brefeldin A (10 μg/ml) for 1 h. iNKT cells were activated in vivo by injecting 1.25 μg anti-CD3 mAb 3C11 i.v. into mice.

Total RNA from hepatic iNKT cells was amplified twice, biotin-labeled, and purified using the RNeasy Mini Kit (Qiagen). The cRNA (10 μg) was hybridized to M430 2.0 chips (Affymetrix). All samples were generated in triplicate. Raw data were normalized using the RMA algorithm of GenePattern (15). Nonvariable low expression values were excluded using “Preprocess Dataset” of GenePattern. Data were visualized using the “Multiplot Preprocess,” “Multiplot,” and “Hierarchical Cluster” of GenePattern. Gene ontology analysis and visualization were performed using Gorilla software (16). Gene expression data are deposited in National Center for Biotechnology Information Gene Expression Omnibus and are accessible through Gene Expression Omnibus Series accession number GSE42835 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE42835).

RNA was amplified using MessageAmp Premier aRNA Amplification (Ambion) and reverse transcribed. TaqMan was performed in triplicate with 100 ng cDNA/reaction using specific primers and Gadph or Hprt1 endogenous controls (Applied Biosystems) on ABI Prism 7900 and analyzed with SDS 2.2.1 software. Relative quantitation of gene expression was determined using the comparative cycle threshold method (17). For absolute quantification, Ptpn6 exons 8–12 cDNA was cloned into ZeroBlunt TOPO Cloning plasmid (Invitrogen). Plasmid DNA was diluted from 10⁹ to 0 copy number into irrelevant DNA. Gapdh and Hprt1 genes were amplified concurrently to normalize the cDNA of samples. Standard curves for both housekeeping genes were prepared using serial dilution of a cDNA pool (from 150 to 12.5 ng) derived from wild-type (WT) iNKT cells. Ptpn6 mRNA copy number present in each sample after normalization was calculated using the standard curves.

BM chimeras were generated as described previously (11) and analyzed 6–8 wk after transplantation.

The two-tailed Student t test was used for unpaired samples, applying correction for unequal variances where required. The p values <0.05 were considered significant.

Functional maturation of iNKT cells requires interaction with CD1d⁺ APCs, so we investigated reactivity of iNKT cells that developed in pLck-hCD1d^+/− mCD1d^−/− tg mice (pLck-hCD1d tg) in the absence of “educational” CD1d⁺ APCs (8). iNKT cells were sorted from the liver of WT and pLck-hCD1d tg mice using αGalCer-loaded human CD1d-IgG dimers (hCD1d-dim) and were activated by either iTCR cross-linking or αGalCer presentation by DCs. hCD1d-dim bind selectively to mouse Vβ8.2⁺ iNKT cells (18), and the hCD1d transgene selects essentially the same population (8), thus enabling robust identification of the same iNKT cell repertoire in both hCD1d tg and WT mice. pLck-hCD1d tg-derived hepatic iNKT cells produced significantly more IFN-γ and IL-4 than WT iNKT cells after either iTCR cross-linking (Fig. 1A) or αGalCer-loaded DCs (Fig. 1B). The cells from both mice expressed similar TCR levels (data not shown). To assess whether iNKT cell hyperreactivity in tg mice depended on enhanced early TCR signaling or instead reflected increased capacity for cytokine expression, we activated hepatic lymphocytes from tg and WT mice with PMA/ionomycin, which induce cytokine production bypassing the TCR–proximal complex. Similar fractions of iNKT cells from tg and WT mice expressed CD69 (Fig. 1C, left panels) and produced IFN-γ (Fig. 1C, right panels), suggesting that the hyperreactivity of tg iNKT cells depended on enhanced early TCR signaling.

We next assessed the responsiveness of tg iNKT cells to rapid activation with anti-CD3 mAb in vivo, because the absence of CD1d⁺ APCs in these animals precludes the use of αGalCer. Within 90 min of mAb injection, both splenic and hepatic iNKT cells in tg mice upregulated CD69 and produced significantly more IFN-γ than iNKT cells from WT mice (Fig. 1D). Similarly, IL-4 production was upregulated by tg iNKT cells in liver. In these same assays, T cells began to upregulate CD69 similarly in both animals, whereas they did not produce cytokines (data not shown). Consistent with the intracellular staining data, tg mice administered with anti-CD3 mAb exhibited substantially increased IFN-γ, and slightly higher IL-4, levels in serum compared with WT mice, upon normalization for the total iNKT cell number present in the spleens of each type of mice (Fig. 1E).

Mature T cells of pLck-hCD1d tg mice express low hCD1d levels (8). However, this leaky hCD1d expression on peripheral T cell played a role in the acquisition of NK1.1, but not in the altered responsiveness, of iNKT cells from pLck-hCD1d tg mice, as shown by the phenotype and TCR hyperreactivity exhibited by peripheral iNKT cells developing in mixed BM chimeras in which hCD1d expression was confined to DP thymocytes (Supplemental Fig. 1). Furthermore, the iNKT cell hyperresponsiveness was not due to xenogenic CD1d expression, because iNKT cells isolated from an alternative model in which the mouse CD1d cDNA is expressed under the proximal Lck promoter (10) (pLck-mCD1d tg) exhibited comparable Ag hyperreactivity (Fig. 1F). Together, these data suggested that iNKT cells that develop in pLck-hCD1d tg mice display increased responsiveness to TCR-dependent stimulation.

To define the molecular basis of iNKT cell hyperreactivity in pLck-hCD1d tg mice, we compared the global gene expression profile of WT and tg iNKT cells using DNA microarrays. Hepatic iNKT cells were sorted from tg and WT mice using hCD1d-dimers before RNA was extracted for hybridization to genome-wide microarrays (M430v2). iNKT cells from WT and tg mice exhibited very similar transcription profiles, with only 38 known genes upregulated and 14 downregulated >2-fold in tg cells compared with WT iNKT cells (Fig. 2A, 2B, Supplemental Table I). Gene ontology analysis of the biological processes associated with these differentially expressed genes did not reveal any specific function or pathway (data not shown). The differential expression of a set of these genes with potential roles in the functional regulation of iNKT cells was validated by RT-quantitative PCR (Fig. 2C). Among the candidate genes downregulated ∼2-fold in tg iNKT cells, Ptpn6 encodes the SHP-1 tyrosine phosphatase that is involved in negative regulation of TCR signaling (19, 20). Among the confirmed upregulated genes, Ceacam1 encodes Carcinoembryonic-Ag–related adhesion molecule 1, which is expressed by activated T cells and inhibits TCR signaling through interaction with SHP-1 (21). Also upregulated were genes Tyrobp, encoding for the NK adaptor protein DAP12, which can exert both activating and inhibitory effects (22), and gene FcgRIIb, encoding the low-affinity IgG receptor expressed by B cells and activated T cells that mediates inhibitory functions by recruiting SHP-1, SHP-2, and SH2-containing complex inositol 5-phosphatase SHIP1 (23). Finally, upregulated genes Klra5 and Klra7 encode the inhibitory NK receptors Ly49E and Ly49G2 that recruit SHP-1 via their cytoplasmic ITIM regions (24).

These data indicated that downregulation of SHP-1 may underpin the increased responsiveness of iNKT cells that develops in the absence of peripheral CD1d. The upregulation of receptor molecules able to interact with SHP-1 in tg iNKT cells suggested some level of functional compensation, as described in NK cells expressing inactive SHP-1 (25).

To assess a functional link between SHP-1 and iNKT cell reactivity, we first determined that the similarly hyperreactive iNKT cells from pLck-hCD1d and pLck-mCD1d tg mice exhibited also a comparable downregulation of SHP-1 transcript (Fig. 2D). SHP-1 transcript was also similarly downregulated in the hyperresponsive iNKT cells from pLck-hCD1d mice and from BM chimera expressing hCD1d only on DP thymocytes, described earlier (Supplemental Fig. 1). Next, we investigated iNKT cell responses in motheaten mice (me/−), which bear a defective SHP-1 protein because of a single nucleotide deletion in the Ptpn6 gene (26). Mice heterozygous for the motheaten mutation (me/+) exhibit a 50% reduction in SHP-1 transcript and protein in hematopoietic cells, resulting in T cell hyperreactivity upon TCR stimulation (26, 27). Unlike the homozygous me/− mice, the heterozygous me/+ mice are healthy and do not exhibit systemic inflammation (28). Frequency and NK1.1 expression of thymic and hepatic iNKT cells in me/+ and WT mice were comparable (data not shown). We confirmed that both SHP-1 transcript and protein levels were similarly reduced in hepatic iNKT cells from both me/+ and pLck-hCD1d tg mice compared with WT animals (Fig. 3A, 3B, Supplemental Fig. 2). As expected, a lower SHP-1 protein level was observed in T cells from me/+ mice, but not in T cells from pLckhCD1d tg mice (Fig. 3B), thereby confirming the selective downregulation of SHP-1 in the tg iNKT cells. Hepatic iNKT cells sorted from me/+ mice produced more cytokines than cells from controls (me^+/+), upon activation either by αGalCer-pulsed DCs (Fig. 3C) or by iTCR cross-linking (Fig. 3D), supporting a key role for SHP-1 in regulating iNKT cell responsiveness.

We next determined whether reintroducing physiological contact with endogenous mCD1d molecules normalized iNKT cell responsiveness in pLck-hCD1d tg mice. We hence crossed tg and C57BL6 mice to generate pLck-hCD1d tg mCD1d^+/+ mice and investigated the reactivity of the peripheral iNKT cells. Hepatic iNKT cells from pLck-hCD1d tg, pLck-hCD1d tg mCD1d^+/+, and WT mice were sorted using hCD1d-dimers and activated in vitro either by αGalCer-pulsed DCs (Fig. 4A, upper panels) or by iTCR cross-linking (Fig. 4A, lower panels). Both pLck-hCD1d tg mCD1d^+/+ and WT iNKT cells produced levels of cytokines significantly lower than those observed in iNKT cells from pLck-hCD1d tg mice (Fig. 4A). Accordingly, SHP-1 expression was significantly increased in both pLck-hCD1d tg mCD1d^+/+ and WT iNKT cells compared with pLck-hCD1d tg iNKT cells (Fig. 4B). The hyperreactivity of iNKT cells that develop in pLck-hCD1d tg mice was thus normalized by changes in SHP-1 expression induced by contact with CD1d⁺ cells other than thymocytes.

To determine whether the iNKT cell hyperreactivity and SHP-1 downregulation in pLck-hCD1d tg mice is imprinted in thymus, iNKT cells from the thymus, spleen, and liver of tg and WT mice were sorted with hCD1d-dimers and then activated by TCR cross-linking to directly compare their effector functions. Thymic iNKT cells from pLck-hCD1d tg and WT mice secreted comparable amounts of IFN-γ and IL-4 (Fig. 5A), and displayed equivalent SHP-1 protein expression (Fig. 5B, Supplemental Fig. 3). In WT animals, peripheral iNKT cells secreted significantly less cytokine than thymic iNKT cells. In contrast, cytokine secretion by peripheral iNKT cells from pLck-hCD1d tg was sustained at levels comparable with those detected in thymic iNKT cells, and cytokine production was significantly higher than that displayed by WT peripheral iNKT cells. The expression of SHP-1 protein increased markedly from thymus to periphery in iNKT cells from WT mice, but not in iNKT cells from pLck-hCD1d tg mice (Fig. 5B). Thus, under physiological conditions, iNKT cells upregulate SHP-1 expression in the peripheral compartment. Both the NK1.1⁻ and NK1.1⁺ subsets of peripheral iNKT cells from pLck-hCD1d tg mice displayed similar reductions in SHP-1 protein expression compared with the corresponding populations in WT animals (Fig. 5C), indicating that reduced SHP-1 expression by tg iNKT cells is independent of their maturation.

Together, these data suggested that iNKT cell responsiveness and SHP-1 expression in pLck-hCD1d tg mice are not imprinted in thymus. Instead, tg iNKT cells fail to upregulate SHP-1 expression in the absence of CD1d contacts in the periphery.

To identify the CD1d⁺ cells that might regulate iNKT cell responsiveness in peripheral compartments, we considered B cells and DCs as primary candidates, because both cell types express CD1d and interact with iNKT lymphocytes (29). Furthermore, hyperreactivity of peripheral iNKT cells from pLck-mCD1d tg mice has been reported to normalize upon crossing these animals with tg mice that express mCD1d on MHC II⁺ APCs (7).

We first investigated the reactivity of iNKT cells in μT mice that lack B lymphocytes (30). The percentage of iNKT cells in these animals was normal in thymus and in periphery, but the numbers were decreased in periphery (data not shown), consistent with the observed T and NK cell reduction in these animals (31, 32). Splenic and hepatic iNKT cells of μT and WT mice that were sorted and activated in vitro by TCR cross-linking were comparably responsive, although IL-4 secretion by splenic iNKT cells was somewhat reduced in μT mice (Fig. 6A). The normal responsiveness of peripheral iNKT cells from μT mice was associated with normal levels of SHP-1 protein expression (Fig. 6B).

The role of DCs in the regulation of iNKT cell responsiveness was investigated using mice in which DCs are constitutively deleted by the expression of an inducible tg diphtheria toxin (ΔDC mice) (12). ΔDC mice display >90% reduction in myeloid DCs, lymphoid DCs, plasmacytoid DCs, and Langerhans cells in thymus, spleen, lymph nodes (LNs), and skin, leading to lethal systemic autoimmunity by 7–8 wk of age (12). We therefore sorted liver iNKT cells from 5-wk-old ΔDC mice and WT mice using mCD1d dimers, and activated these cells by iTCR cross-linking in vitro. iNKT cells from ΔDC mice were hyperresponsive (Fig. 6C), suggesting that DCs are key regulators of the iNKT cell activation threshold. To confirm these data in mice not affected by autoimmune pathology, we reconstituted lethally irradiated CD1d^−/− mice with 20% BM cells from either WT mice or CD1d^−/− mice in combination with 80% BM cells from ΔDC mice (Fig. 6D). The frequencies of iNKT cells were similar in the two types of BM chimera (Fig. 6E), but the proportion of mature NK1.1⁺ iNKT cells in the liver and spleen was reduced in animals that lacked CD1d⁺ DCs (Fig. 6E). iNKT cells from mice that lacked CD1d⁺ DCs were also hyperresponsive to activation in vitro (Fig. 6F). Consistent with this hyperreactivity, iNKT cells from chimera deficient in CD1d⁺ DCs expressed a lower SHP-1 mRNA copy number compared with cells from control chimera (Fig. 6G). In the thymus, iNKT cells from BM chimera deficient in CD1d⁺ DCs displayed normal reactivity, and their SHP-1 transcript copy number was comparable with that expressed by WT thymic iNKT cells (data not shown).

Collectively, these data indicated that stimulation by CD1d expressed on peripheral DCs functionally educates iNKT cells by tuning SHP-1 expression.

Our study reveals a novel mechanism of iNKT cell functional education that occurs in the periphery and is exerted by tuning the TCR signaling threshold. This process is mediated by the upregulation of SHP-1 in iNKT cells, induced upon engaging CD1d⁺ DCs. In the absence of this functional checkpoint, expression of SHP-1 remains low, resulting in altered reactivity of iNKT cells.

Tyrosine phosphatase SHP-1 exerts a global negative regulatory function on TCR-mediated activation of T lymphocytes by dephosphorylating several proximal signal transduction mediators, thereby setting the threshold for thymocyte selection and activation of mature T cells (20). SHP-1 levels are physiologically modulated during T cell development or upon activation, increasing the avidity of immature thymocytes and recently activated T cells for Ag signals (20). Our finding that iNKT cells upregulate SHP-1 and reduce TCR responsiveness in the periphery is consistent with recent data from tg mice expressing a GFP reporter that directly responds to TCR signaling (33). Thymic iNKT cells from these animals are brightly fluorescent, whereas peripheral iNKT cells are not, suggesting that these cells may become desensitized to CD1d stimulation in the periphery.

DCs play a key role in the activation of iNKT cells (29) and mediate their thymic negative selection (8, 34). B cells are abundant CD1d⁺ APCs in secondary lymphoid organs and are also able to interact with iNKT cells (35–37). Yet, our data suggest that DCs are key in tuning the TCR responsiveness of iNKT cells in peripheral tissues in the steady-state. DCs, unlike B cells, might provide costimulatory signals that, in addition to the CD1d-dependent one, may cooperate to upregulate SHP-1 expression by peripheral iNKT cells. iNKT cells might also preferentially contact DCs in unchallenged situations, as they colocalize with DCs in the spleen marginal zone and periarteriolar lymphoid sheaths, LNs subcapsular area, and liver sinusoids (38–40). We cannot exclude that also CD1d⁺ macrophages, which locate proximally to iNKT cells in the splenic marginal zone and in LNs (38, 40), and may be depleted in ΔDC mice, may contribute to tuning iNKT cell responsiveness. Conditional deletion of CD1d expression by DCs was recently reported to exert little effect on iNKT responsiveness (41). This model is different from ours, and a differential persistence of residual CD1d⁺ DCs in these mice and in the ΔDC mice used in this study might account for the variable responsiveness observed.

We show that B lymphocytes do not play a major role in regulating SHP-1 expression in iNKT cells. However, upon TCR cross-linking in vitro, we find a reduced IL-4 secretion by splenic iNKT cells from μT mice. Upon αGalCer injection in vivo, a normal iNKT cell response was found either in μT mice (42) or in mice that exhibit only CD1d-deficient B cells (41), whereas other studies suggested a possible CD1d-dependent suppressive role for B cells in regulating iNKT cell function (29). It is difficult to compare the different iNKT cell activation regimens used in these studies; nevertheless, these data imply that B cells might also regulate iNKT cell effector response, possibly in specific anatomical sites and likely by mechanisms distinct from DCs.

Peripheral contacts with CD1d are required for the final phenotypic maturation of iNKT cells (7, 43, 44). We find that contacts with peripheral CD1d⁺ DCs results in the upregulation of both NK1.1 and SHP-1 expression by iNKT cells. iNKT cells may complete their maturation upon engaging self-lipid Ags (45), as well as exogenous lipid(s) derived from the intestinal microbiota of mice (46), following uptake and presentation by CD1d-expressing DCs. We also show that hCD1d expression by peripheral T cells in the pLck-hCD1d tg mice can upregulate NK1.1, but not SHP-1, in iNKT cells. Human CD1d may deliver a stronger “basic” stimulation to peripheral iNKT cells than mouse CD1d (8), sufficient for NK1.1 upregulation. SHP-1 upregulation in peripheral iNKT cells may require additional costimulatory signals, selectively provided by DCs and not by B or T lymphocytes.

Unlike their peripheral counterparts, thymic iNKT cells from pLck-hCD1d tg mice exhibited normal reactivity and SHP-1 levels, in line with published data showing that thymic and peripheral NKT cells represent different functional subsets (7, 47), controlled by distinct molecular and/or environmental cues (7, 48).

The mechanisms by which CD1d⁺ DC-dependent signaling upregulates SHP-1 expression in peripheral iNKT cells remain to be defined. In a different context, it was recently shown that weak stimulation of human iNKT cells by CD1d-expressing APCs in vitro opens the IFNG locus and makes it accessible to transcription triggered by innate signaling (49). The TCR triggering of mouse iNKT cells by CD1d⁺ DCs may poise the PTPN6 locus to transcription, induced by concomitant costimulatory signals directly delivered by DCs or present in the microenvironment.

In conclusion, our study defines a functional education step of iNKT cells occurring in the periphery, which may be critical to control the innate effector state of iNKT cells by tuning their reactivity toward stronger stimulatory signals, generated in stress-related conditions and bacterial infections.

We thank L. Guidotti and C. Benoist for critical manuscript reading.

The authors have no financial conflicts of interest.