Introduction

Regulatory T-cells (Treg), characterized by the expression of the alpha chain of the interleukin-2 receptor CD25 and the transcription factor Foxp3 1, are known to be key players in immuno-regulation in both humans and mice; too few may lead to autoimmune diseases whereas too many may prevent an efficient immune response to cancer2,3. Over the last few years, several new functions of Treg beyond immunoregulation have been identified in tissue regeneration or local regulation of metabolism46 for instance. In addition, there are accumulating evidences of a cross-talk between Treg and the nervous system. These includes promotion of oligodendrocyte differentiation or inhibition of neuroinflammation facilitating CNS repair process after brain injuries and preventing cognitive decline710. Furthermore, Treg have been involved in the regulation of pain in various model of nerve injury in rats and mice, such as in autoimmune neuritis or chronic constriction of the sciatic nerve1113. Depletion of Treg have been associated with enhanced pain sensitivity whereas increased Treg number or activity limit pain hypersensitivity14. Although the current view is that Treg control pain through their immuno-suppressive functions 15, whether Treg might regulate pain at steady state is currently unknown. Starting from our previous observation that the expression of the proenkephalin gene (Penk), encoding for analgesic enkephalins, is highly enriched in Treg (Aubert et al., 2019 biorXiv, https://doi.org/10.1101/638072), we characterize the expression of Penk in the immune system and investigate the role of Treg on pain in vivo at steady state using advanced genetically-engineered mouse models. Our results uncover a previously unknown mechanism by which Treg modulate basal somatic sensitivity.

Results and Discussion

A meta-signature of murine Treg and Penk expression in lymphoid and non-lymphoid organs

Using 11 available transcriptomes retrieved from the GEO website, we generated a Treg molecular signature at steady state in lymphoid tissues, improving the robustness of the analysis by several orders of magnitude relative to individual studies. The 25 most differentially expressed genes are depicted in Figure 1A and the complete list is shown in Table S1. As expected from the sorting strategies used to isolate Treg (Table 1 in the Material and Methods section), we observe that Il2ra and Foxp3 are the most differentially expressed genes in Treg compared to Tconv, followed by the well-known Treg markers CTLA-4, Itgae (CD103), Ikzf2 (Helios) and Klrg1. However, several unexpected genes are present in that signature, including the proenkephalin gene Penk, coding for opioid peptides endowed with analgesic properties. Using the Immuno-Navigator database16, a batch-corrected collection of RNA quantification across numerous studies, samples and cell types, we confirm that Penk is highly correlated to Foxp3 in mice (R=0.871, Figure 1B). Enriched expression of Penk in Treg was previously reported in several specific contexts including TCR-transgenic mice17, UVB-exposure6 or in the brain of mice recovering from stroke8. Using publicly available datasets, we also observe that Penk is enriched in Treg in the thymus, the visceral adipose tissue and in muscles (Figure 1C). Thus, Penk mRNA expression in Treg is consubstantial of their generation and is independent of their tissue localization at steady state.

A meta signature of murine Treg and Penk mRNA expression in lymphoid and non lymphoid organs.

A) Top 25 genes enriched in Treg compared to Tconv from lymphoid tissue from at least 10 of the 11 datasets analyzed ranked by fold change (LogFC) of mean expression relative to Tconv. The Penk gene is highlighted in red. B) Correlation of Penk and Foxp3 mRNA expression in all the cell types listed in the legend. The Pearson correlation coefficient is indicated. C) Expression of Penk mRNA in Treg (blue) and Tconv (red) isolated from thymus, visceral adipose tissue (VAT) and muscle. The source of the data is indicated below each graph.

Characteristics of the datasets used for the Treg meta-signature

(NA = not available; LN = Lymph Nodes)

Penk mRNA expression is regulated by TNFR signaling and the BATF transcription factor in Treg

To explore possible mechanisms that could explain the enrichment of Penk mRNA in Treg cells, we examined the Immuno-Navigator dataset to extract the genes that are most correlated with Penk and with each other (Figure 2A). Among these genes, we observe several members of the TNFR superfamily (Tnfrsf1b (TNFR2), Tnfrsf4 (OX40), Tnfrsf9 (4-1BB), Tnfrsf18 (GITR)) as well as the transcription factor Batf. Furthermore, Batf expression is highly correlated with Penk in Treg (R=0.599, Figure 2B). Thus, TNFR and BATF could be involved in the regulation of Penk in Treg. To examine this possibility, we reanalyzed our previously published dataset establishing the transcriptome of Treg stimulated with anti-CD3, anti-CD28 antibodies and TNFR agonists in vitro 18. We observe that addition of TNFR2, OX40 or 4-1BB agonists increases Penk expression at 36h post-stimulation relative to controls (Figure 2C). Interestingly, Batf is also increased by TNFR agonists at an earlier time point (18h), suggesting that TNFR signaling may regulate Penk expression through the BATF transcription factor. Accordingly, a dramatic decrease in Penk expression is observed in Treg lacking BATF 19(Figure 2D). In line with our hypothesis, BATF is a crucial partner of AP-1 transcription factor that has been shown to regulate Penk in the hippocampus20.

Penk mRNA expression is regulated by TNFR signaling and the BATF transcription factor in Treg.

A) A network of the genes most correlated to Penk in Treg is shown. The Pearson correlation values were extracted from the Immuno-Navigator database (taking only Treg for the analysis) and integrated into Cytoscape v3.730. Each node is a gene linked by edges with width proportional to the Pearson correlation. (edge range: 0.538-0.758). B) Illustration of the correlation between expression of Penk and Batf in Treg. Each dot is a sample from the Immuno-Navigator database. The Pearson correlation coefficient is indicated on the figure. C) Penk and Batf mRNA expression after in vitro stimulation of purified Treg with the indicated TNFR agonists prior (0), and at 18 and 36hrs after stimulation. Each dot is a biological replicate from a single experiment. D) GEO2R analysis of the GSE89656 dataset for Penk mRNA variations between wild type control Treg (WT) and BATF-KO Treg.

Penk is prominently expressed by Treg at steady state

Based on these findings, we further investigated the expression of Penk mRNA in various immune cell types across multiple tissues, as defined by a combination of lineage markers. (Figure S1A). To do so, we crossed a transgenic mouse model expressing Tamoxifen (TMX)-inducible Cre recombinase under the promoter of Penk (PenkCre-ERT2) with the ROSA26TdTomato reporter mice. In these mice, any cell that would expressed Penk at the time of TMX administration would become permanently tagged with the tdTomato reporter a few days later. In order to improve detection of those cells by flow cytometry, we also used an anti-mCherry that cross-react with TdTomato (Figure S1B). In lymphoid tissue, such as in lymph node (LN), Penk expression is mainly restricted to Treg and to a small subset of activated CD4+ T-cells (Figure 3A). Similar results are obtained in non-lymphoid tissues, such as in the colon, although Penk expression in activated T cells is further increased relative to LN (Figure 3B). As summarized in Figure 3C, Treg is the immune subset expressing the highest level of Penk in all tissues analyzed. Expression of Penk is also detected in Tconv, but at lower frequencies than in Treg, in agreement with the molecular meta-signature established above (Figure 1). All other cell types show low or undetectable Penk expression. Interestingly, in CD4+ T-cells (Tconv and Treg), Penk expression is higher in the activated CD62Llow CD44high fraction compared to the naive CD62Lhigh CD44low phenotype (Figure S2). Thus, we observed that Penk mRNA is predominantly found in Treg and activated CD4+ T cells in the immune system at steady state. Among these, the highest expression of Penk was detected in Treg cells present in the skin and the colon. However, we cannot exclude the possibility that the distribution of Penk may be broader in an inflammatory context. In line with this, a recent study demonstrated that Penk expression in M2 macrophages is dependent on IL-4 and occurs in response to nerve injury21.

Penk is prominently expressed by Treg at steady state

A-B) (right) UMAP representing all major cell types indicated in the figures determined by flow cytometry from lymph nodes (A) and colon (B) (left). Projection of Penk (blue) on the UMAP of lymph node and colon (right). Subsets were manually gated as depicted in Figure S1A. C) Scatter plot displaying the average population size and percentage of Penk+ cells for the listed cell populations and organs. Population size was calculated as the percentage out of total CD45+ single cells and represented on a log10 scale. (n = 3 mice for the VAT, spleen and bone marrow, n = 6 mice for the other groups; results cumulative of two independent experiments).

Heat hyperalgesia in mice deficient for Penk in Treg

To determine if enkephalins produced by Treg had an effect on pain at steady state, we generated mice deficient for enkephalins in Treg by crossing TMX-inducible Cre recombinase under the control of the Foxp3 promoter (Foxp3Cre-ERT2) with mice transgenic for LoxP sequences flanking exon 2 of Penk (Penkflox). In these mice (hereafter referred to as Lox), any cell expressing Foxp3 at the time of TMX administration would become deficient for exon 2 of Penk a few days later. Since exon 2 is the precursor of Met-Enkephalin, an endogenous opioid that act on thermal pain sensation22, we evaluated the sensibility of these mice to pain induced by heat (Figure 4). As controls, we used mice expressing the Cre recombinase and wild-type (WT) at the locus of LoxP sequences insertion. Mice were treated with TMX and evaluated for heat sensitivity without further additional treatment at ten different time points (4 before and 6 after administration of TMX, 2 to 3 days apart). Under these conditions, a significant trend towards lower latency periods (signing heat hyperalgesia) is observed in the Lox group compared to WT mice. Interestingly, the effect is not apparent until 7 days after TMX administration (Figure S4). Indeed, before TMX administration, WT and Lox female and male mice do not differ in their response to heat (Figure 4A-B). However, after TMX administration, Lox mice develop hyperalgesia compared to control mice, with a 20 percent reduction in their median latency period from D7 onwards (8.1 vs. 6.5 sec, Figure 4C). Interestingly, the effect of Penk deletion in Treg on heat hyperalgesia is sex independent, since it is observed in both female and male (Figure 4D), although the reduction was slightly higher in female (22% reduction in latency period) than in male (16% reduction). This can be explained by the fact that males tend to respond faster (7.9s vs 6.1s) than females (8.4s vs 7.0s), irrespective of the genotype (Figure 4D). Penk expression by CD4+ T-cells has been associated with analgesia in mice models of visceral pain2325, but the specific contribution of Treg was not investigated. In neuroinflammatory settings, such as constriction of the sciatic nerve, pain control by Treg could be dependent on their immuno-suppressive function 12,2628. On the contrary, our tests were performed at steady state, ruling out the possibility that hyperalgesia could be due to a defect in the immuno-suppressive function of Treg. They strongly suggest a direct role of enkephalins produced by Treg on nociception, uncovering a new non-immune fundamental function of Tregs on endogenous regulation of basal somatic sensitivity.

Heat hyperalgesia in mice deficient for Penk in Treg.

A) Withdrawal latency of WT and Lox mice before the administration of TMX (baseline). The mean of all 4 measures before administration of TMX for each mouse is shown. Statistical modeling was performed using a non parametric unpaired Mann-Whitney t-test and was deemed as non-significant (ns). B) Withdrawal latency of WT and Lox mice before administration of TMX (baseline) shown separately for female and male mice, as in A. Statistical modeling was performed using two-way ANOVA and the effect of the genotype on the results was deemed non-significant (ns). C) Withdrawal latency of WT and Lox mice post-TMX treatment. The mean of all 4 measures taken from D7 onward is represented for each mouse. Statistical modeling was performed using a non parametric unpaired Mann-Whitney t-test and was deemed significant (p=0,0038). D) Withdrawal latency of WT and Lox mice post-TMX treatment shown separately for female and male mice, as in C. Statistical modeling was performed using two-way ANOVA and the effect of the genotype on the results was deemed significant (p=0,0037). A multiple comparisons test using the Benjamini-Hochberg False Discovery Rate model was used. The adjusted p values were deemed significant for both female and male mice (q=0,043). Results shown in this figure are cumulative of two independent experiments with n = 44 mice (26 WT and 18 Lox).

Materials & Methods

Extraction of Treg meta-signatures

The datasets used were selected based on a “Treg* AND (Tconv* OR Teff*) AND Mus Musculus” search in the GEO dataset web site (https://www.ncbi.nlm.nih.gov/gds). GEO datasets were manually inspected for inclusion of studies comparing fresh Treg with fresh Tconv from lymphoid organs. Characteristics of selected datasets are summarized in Table 1. For each dataset, we generated a list of differentially expressed genes with a cutoff based on a false discovery rate (FDR) inferior to 0.05 and a log2 fold change superior to 1 with the GEO2R embedded algorithm.

Mice

All male and female mice were on a C57Bl/6J background. Foxp3tm9(EGFP/cre/ERT2)Ayr/J (Foxp3Cre-ERT2) (catalog #016961), B6.Cg-Gt(ROSA)26Sortm9(CAG–tdTomato)Hze/J (ROSA tdTomato) (catalog #007909), B6.Cg-Penktm1.1(cre/ERT2)Hze/J, (PenkCre-ERT2) (catalog #022862) were purchased from The Jackson Laboratory. C57BL/6JSmoc-Penkem1(flox)Smoc (Penkflox) were purchased from Shanghai Model Organisms (catalog NM-CKO-210032). For the Penk mapping experiments, PenkCre-ERT2 were bred with ROSAtdtomato. All mice were confirmed to be homozygous for the inducible Cre and at least heterozygous for tdTomato by touchdown PCR (primers sequences available on request). To specifically knock-out the Penk gene in Tregs, Foxp3Cre-ERT2 were crossed with Penkflox leading to double heterozygous mice (F1) that were crossed together resulting in double homozygous F2 littermates’ mice. All mice used in this study were of F3 generation. Mice were genotyped by touchdown PCR (primers sequences available on request). Mice were housed under specific pathogen-free conditions and were used for experiments at 8 weeks or older. All protocols were approved by the Ethics Committee for Animal Experimentation Charles Darwin (APAFIS #32284-2021070513305185 v5).

Tamoxifen administration

Tamoxifen (ThermoFisher, Les Ulis, France) was dissolved in peanut oil at the concentration of 8 mg/ml under 37°C agitation and delivered by 200µL oral gavage. All mice were administered tamoxifen only once, 7 days before the start of each experiment.

Preparation of cell suspensions

Organs were harvested on ice in PBS 3% fetal calf serum (FCS). Inguinal, brachial and mesenteric LNs and spleens were directly mashed through a 70µm filter and resuspended in

PBS 3% FCS. Lungs, colons, livers, VAT and skin were dissected, minced then incubated in appropriate digestion buffers (Miltenyi Biotech, Paris, France) at 37°C for various duration according to manufacturer protocols. Cell suspensions were then passed through a 70µm cell strainer and suspended in PBS 3% FCS. To eliminate dead cells and debris, liver cell suspensions was isolated on a 70:30 Percoll gradient. Rings were collected, washed, and cell pellets were resuspended in PBS 3% FCS. ACK Lysing Buffer was used to lyse red blood cells in the lungs, livers and spleens (lysis performed for 1 min at RT, followed by two washes with PBS), prior to staining for flow cytometry.

Antibodies and flow cytometry analysis

The monoclonal antibodies (mAb) and fluorescent reagents used in this study are listed below in the Table. Up to 4.106 cells were incubated for 30 min at 4°C with fixable live/dead dye and with the anti-CD16/CD32 (clone 2.4G2) to block FcgRII and FcgRII receptors. Cells were then stained with a combination of antibodies. Cell-surface staining was performed in PBS 3% FCS for 20 min at 4°C. Permeabilization and intracellular staining was performed using the Foxp3/Transcription Factor Staining Buffer Set kit and protocol (ThermoFisher). Stained cells were washed with PBS 1X before acquisition on a Cytek Aurora flow cytometer (Cytek Bioscience, Fremont, CA, USA). Data were analyzed and UMAPs were generated using FlowJo software, version 10.8.1 (TreeStar, Ashland, OR, USA).

Hot plate test

The method used was described by Baker et al. (2002)29. Briefly, mice were placed on a metal hot plate maintained at 55°C. The plate was enclosed with Plexiglas walls so that mice could not escape. The response latency, which is the time taken to observe a nocifensive behavior such as jumping, licking or flicking of the hind paws, was recorded. The mice were then immediately removed from the plate upon the recording of a reaction, or within 25s if no response was observed to prevent tissue damage. The test was repeated every two days for a total of four measures before the administration of tamoxifen and four measures starting from D3 post-tamoxifen. Male and female mice were tested separately. Mice remained in their home cage except when being tested on the hot plate. The experimenter was blind to the genotype of the mice until the end of the experiment

Statistical analysis

All statistical tests are reported in the figure legends and were performed with Prism v8 (Graph Pad Inc, La Jolla, CA, USA). The statistical power of the analyses (alpha) was set at 0.05.

Funding

This study was funded by a research grant from Sorbonne University (SU) (Emergence 2021) to GM, and by the National Institute for Health and Medical Research (INSERM). NA was funded by a doctoral fellowship from Fondation ARC and by a post-doctoral fellowhip from EGLE TX, MF by a doctoral fellowship from the French Ministère de l’Enseignement Supérieur et de la Recherche, MP and MN by pregraduates fellowships from SU and EGLE TX respectively. The funders had no role on the supervision of the research, nor in the analysis of the results.

Acknowledgements

The authors would like to thank Olivier Brégerie, Flora Issert, and Doriane Foret for taking care of our mice; Dr Martina Lubrano di Ricco and Morgane Hilaire for technical help, Dr Benoit Salomon for providing Foxp3Cre-ERT2 and ROSAtdTomato mice; Francois-Xavier Lejeune (ICM, Paris) for help with biostatistics analysis and Dr Jean-Luc Teillaud for critical review of the manuscript.

Supplementary figures and table

Representative flow cytometry profiles in PenkCre-ERT2x ROSAtdtomato

A) Gating strategy used to define the population for the mapping of Penk expression by immune cells. B) Representative staining of TdTomato and mCherry

Penk expression in Tconv and Treg according to activation status in lymph node

Representative staining of TdTomato and mCherry in naive (CD62Lhigh CD44low), central memory (CD62Lhigh CD44high) and effector memory (CD62Llow CD44high) Tconv (top) and Treg (bottom).

A kinetics of latency periods in response to heat in WT and Lox mice.

Each dot is the mean value (± SD) of latency period in response to heat stimulation for all mice of the WT or Lox genotype. The results are represented as a function of days relative to TMX gavage (D0). Statistical modeling of the results was performed with linear regression curve fitting with a comparison of intercept and slopes. The low p value allow to reject the hypothesis that one curve fits all the datasets.

Table S1: List of the differentially expressed genes between Tconv and Treg. Only genes upregulated in 10 out of 11 datasets are presented with the associated log2 fold change as determined by GEO2R. (NA = not available)