Research ArticleT CELL MEMORY

TRM maintenance is regulated by tissue damage via P2RX7

See allHide authors and affiliations

Science Immunology  14 Dec 2018:
Vol. 3, Issue 30, eaau1022
DOI: 10.1126/sciimmunol.aau1022

Tissue damage triggers a TRM trim down

Noncirculating tissue-resident memory T cells (TRM) provide rapid host defense after reexposure to a previously encountered pathogen. Stark et al. found that TRM in the liver and small intestine expressed high levels of the P2RX7 purinergic receptor compared with circulating T cells. Release of extracellular nucleotides that are P2RX7 ligands (ATP and NAD+) in a mouse model of sterile liver injury selectively enhanced the death of liver TRM compared with circulating T cells, whereas concurrent TCR engagement promoted survival of TRM. These findings help define the signaling pathways that shape the clonal composition of the TRM compartment on an ongoing basis.


Tissue-resident memory T cells (TRM) are noncirculating immune cells that contribute to the first line of local defense against reinfections. Their location at hotspots of pathogen encounter frequently exposes TRM to tissue damage. This history of danger-signal exposure is an important aspect of TRM-mediated immunity that has been overlooked so far. RNA profiling revealed that TRM from liver and small intestine express P2RX7, a damage/danger-associated molecular pattern (DAMP) receptor that is triggered by extracellular nucleotides (ATP, NAD+). We confirmed that P2RX7 protein was expressed in CD8+ TRM but not in circulating T cells (TCIRC) across different infection models. Tissue damage induced during routine isolation of liver lymphocytes led to P2RX7 activation and resulted in selective cell death of TRM. P2RX7 activation in vivo by exogenous NAD+ led to a specific depletion of TRM while retaining TCIRC. The effect was absent in P2RX7-deficient mice and after P2RX7 blockade. TCR triggering down-regulated P2RX7 expression and made TRM resistant to NAD-induced cell death. Physiological triggering of P2RX7 by sterile tissue damage during acetaminophen-induced liver injury led to a loss of previously acquired pathogen-specific local TRM in wild-type but not in P2RX7 KO T cells. Our results highlight P2RX7-mediated signaling as a critical pathway for the regulation of TRM maintenance. Extracellular nucleotides released during infection and tissue damage could deplete TRM locally and free niches for new and infection-relevant specificities. This suggests that the recognition of tissue damage promotes persistence of antigen-specific over bystander TRM in the tissue niche.


Tissue-resident memory T cells (TRM) form a crucial first line of defense against local infection and invading pathogens. They are specialized memory T cells that do not recirculate through blood and lymph but are permanently located in an organ. TRM have been detected in barrier tissues such as the intestine, skin, and lung; in internal organs such as brain and liver; and in secondary lymphoid organs. Because of their location at pathogen entry points and their ability to unleash immediate effector functions after activation, TRM confer superior protection against local infections compared with circulating T cells (TCIRC) (14). Upon recognition of their cognate antigen, TRM initiate a tissue-wide state of alert by recruiting other immune cells to the site of infection (46).

Their location in the front line of the immune system positions TRM at hotspots of pathogen encounter. Therefore, TRM are frequently exposed to tissue damage and inflammatory events. In many cases, TRM will not encounter their cognate pathogen but will receive nonspecific signals of inflammation and tissue damage. This history of local inflammation and tissue damage is an important aspect of TRM homeostasis that has not been studied so far.

Pathogen encounter, inflammation, and tissue damage lead to the release of danger signals. These so-called pathogen- or danger-associated molecular patterns (DAMPs) comprise exogenous pathogen-derived molecules (e.g., lipopolysaccharide, double-stranded RNA) as well as nonmicrobial endogenous signals derived from stressed and dying cells that are released during infection and inflammation and during tissue damage by toxins or ischemia (7). The tissue damage–associated DAMPs mainly comprise intracellular and intranuclear molecules such as HMGB1, S100 proteins, DNA, and RNA, as well as purine metabolites that are released by dying cells (8). DAMPs are recognized by a large array of pattern recognition receptors (PRRs), including Toll-like receptors (TLRs), RIG-I–like receptors (RLRs), AIM2-like receptors (ALRs), NOD (nucleotide-binding and oligomerization domain)–like receptors (NLRs), and purinergic receptors (P2XRs and P2YRs). Triggering of these receptors on innate immune cells initiates a tissue-wide alert and induces the production of proinflammatory cytokines and other signals that recruit and activate additional innate and adaptive immune cells. However, T cells can also sense DAMPs directly. Triggering of TLR2, TLR7, and NOD1 on T cells provide costimulatory signals for T cell activation (913). Similarly, activation of the purinergic receptor P2RX1, P2RX4, or P2RX7 by extracellular adenosine triphosphate (ATP) costimulates T cell activation (14, 15) and recently has been shown to play a role in the generation of memory CD8+ T cells (16). The RLR LGP2 improves survival of activated CD8+ T cells during viral infection (17). However, triggering of PRRs can also limit T cell responses. NLRX activation inhibits proliferation of T cells (18), and activation of the cytosolic DNA detector STING after T cell receptor (TCR) triggering as well as prolonged triggering of the purinergic receptor P2RX7 induces cell death in T cells (15, 19). Most of these studies focused on TCIRC. Despite the localization of TRM at hotspots of pathogen encounter, it is currently unknown how TRM respond to danger signals.

Using transcriptome profiling of lymphocytic choriomeningitis virus (LCMV)–specific CD8+ T cells from liver, spleen, and small intestine, we analyzed which sensors for danger signals are specifically expressed by TRM and identified the tissue damage receptor P2RX7 to be highly up-regulated in TRM. P2RX7 is an ion channel that responds to extracellular ATP and NAD+ (nicotinamide adenine dinucleotide). The concentration of NAD+ and ATP is high intracellularly (100 μM to 2 mM) but in the low nanomolar range in the interstitium of healthy tissues (20, 21). The local concentration of extracellular ATP and NAD+ is strongly increased under pathological conditions and during tissue damage, tumor development, and inflammation, as both nucleotides are released into the extracellular space (2224). Necrotic cells can release their intracellular contents, and apoptotic or stressed cells in inflammatory settings can secrete NAD+ and ATP actively via pannexin and connexin channels (25, 26). The function of P2RX7 has been previously studied in detail in natural killer T (NKT) cells and follicular and regulatory T cells, which also express P2RX7. Triggering of P2RX7 by high levels of ATP or NAD+ induces massive efflux of K+ ions and the formation of a P2RX7 pore, resulting in cell death of the triggered T cell (15, 27, 28). Using experimental infection models as well as sterile induction of tissue damage by acetaminophen, we demonstrate that P2RX7 activation in vivo leads to a specific depletion of TRM, whereas TCIRC are maintained. Activation of TRM by TCR ligation induced a marked down-regulation of P2RX7 and resistance to P2RX7-mediated cell death, whereas interleukin-12 (IL-12) up-regulated P2RX7, suggesting that in local infections, the presence of antigen and inflammatory cytokines modulate TRM persistence.

These results describe P2RX7-mediated signaling as a critical pathway for the regulation of TRM maintenance. Extracellular nucleotides released during inflammation and tissue injury could deplete TRM locally and free niches for new specificities. Down-regulation of P2RX7 by TCR stimulation suggests that the pathway promotes persistence of antigen-specific cells over activated bystander cells in the tissue niche.


The tissue damage receptor P2RX7 is selectively expressed on CD8+TRM

To analyze which DAMPs TRM could respond to, we used previously published RNA-sequencing (RNA-seq) data of tissue-resident as well as circulating LCMV-specific memory CD8+ T cells from liver, small intestine, and spleen (29). TRM were defined by CD69 expression and absence of CD62L and expressed TRM-associated molecules such as CXCR6 and CD103 or the transcription factor Hobit (figs. S1 and S2). Of the 47 selected PRRs, 7 were differentially expressed between the memory T cell subsets. AIM2, P2RX7, and P2RY10 were highly expressed in TRM from liver and small intestine. Whereas AIM2 and P2RY10 were expressed at intermediate or high levels in TCIRC [naïve T cells (TN), central memory T cells (TCM), and effector memory T cells (TEM)], P2RX7 was lowly expressed (Fig. 1A). The high expression of P2RX7 on TRM compared with TCIRC (TCM and TEM) was confirmed on pathogen-specific T cells 30+ days after infection (Fig. 1, B and C, and fig. S3, A and B). P2RX7 was highly expressed not only on TRM in the small intestine and liver but also on TRM in the lung, bone marrow (BM), kidney, and salivary gland, suggesting that P2RX7 expression is a universal feature of TRM. The danger receptor P2RX7 can be activated via two different routes. Extracellular ATP triggers transient activation of P2RX7, resulting in the opening of a cation channel. Extracellular NAD+ induces constitutive activation of P2RX7 through the ectoenzyme ARTC2.2, which catalyzes adenosine diphosphate (ADP) ribosylation of P2RX7 and induces constitutive P2RX7 activation. Compared with TCIRC, TRM expressed higher levels not only of P2RX7 but also of ARTC2.2 (Fig. 1D and figs. S3C and S4), which suggests that these memory cells are more susceptible to NAD+-mediated activation of P2RX7 compared with TCIRC. Local concentrations of extracellular NAD+ and ATP are also influenced by enzymatic activity of cell membrane proteins. The two ectoenzymes CD38 and CD39 are known to drive the degradation of extracellular NAD+ and ATP. CD39 hydrolyzes ATP to ADP, and CD38 catalyzes the formation of adenosine diphosphate ribose and nicotinamide from NAD+ (26, 30). The high expression of both ectoenzymes on TRM compared with TCIRC (Fig. 1E and fig. S4) might provide a regulatory mechanism that limits the duration of the activating signal of extracellular nucleotides on TRM. Previously published transcriptional profiles (29) show similar expression patterns of P2RX7, ARTC2.2, and CD38 on innate lymphocytes. Tissue resident NKT cells and type 1 innate lymphoid cells (ILC1) display higher expression of these molecules than conventional natural killer (cNK) cells (fig. S4), suggesting that P2RX7-associated purinergic signaling could be a universal feature of tissue-resident lymphocytes.

Fig. 1 P2RX7 is selectively expressed by CD8

+ TRM. (A) RNA-seq of naïve and LCMV-specific memory CD8+ T cell subsets. Heatmap of PRR mRNA that is differentially expressed between the memory T cell subsets (TN: CD44 CD62L+; TCM: CD44+ CD62L+ CD69; TEM: CD44+ CD62L CD69; TRM: CD44+ CD62L CD69+). (B) Mice were infected with LCMV or (C) LM-OVA and day 30+ postinfection pathogen-specific CD8+ T cells were analyzed [GP33 (KAVYNFATC) tetramer+ for LCMV or OVA (SIINFEKL) tetramer+ for LM-OVA]. Thirty minutes before sacrifice, mice were injected with ARTC2.2-blocking nanobody s+16a to prevent P2RX7 ADP ribosylation (60), and P2RX7 expression on pathogen-specific TCM, TEM, and TRM in the different organs was analyzed (representative of more than three experiments). (D) WT-P2RX7 KO mixed BM chimeras were infected with LM-OVA. More than 30 days after infection, mice were injected with anti-ARTC2.2 nanobody, and the expression of P2RX7 and ARTC2.2 was analyzed by flow cytometry. LM-OVA–specific (SIINFEKL-tetramer) WT CD8+ T cells are displayed. P2RX7 KO CD8+ T cells from the same mouse were used for gating (representative of n = 5). (E) At day 30+ after LCMV infection, LCMV-specific (GP33-tetramer) CD8+ T cells were analyzed (n = 4/5).

P2RX7 deficiency does not impair effector or memory differentiation of CD8+T cells

Because of the high expression of P2RX7 on TRM, we hypothesized that P2RX7 might influence TRM generation or maintenance. Hence, we investigated the role of P2RX7 during the CD8+ T cell response. To analyze how P2RX7 affects the generation of CD8+ T cells under homeostatic conditions, we generated mixed BM chimeras with wild-type (WT) and P2RX7 knockout (KO) BM (ratio, 1:1). The WT and P2RX7 KO cells contributed equally to CD8+ T cells in all analyzed tissues (fig. S5A), and more than 80% of the T cells were of donor origin 8 to 12 weeks after the generation of the mixed BM chimeras. The distribution of TCIRC and TRM was equal in the WT and P2RX7 KO compartment (fig.S5A), suggesting that P2RX7 does not influence the development of CD8+ T cells. To investigate the CD8+ T cell response after antigen challenge, we infected WT and P2RX7 KO mice as well as WT-P2RX7 KO mixed BM chimeras with LCMV and analyzed the effector response at day 8 after infection. In both experimental setups, the effector response of LCMV-specific WT and P2RX7 KO CD8+ T cells was of similar magnitude and displayed the same short-lived effector cell/memory precursor effector cell phenotype (fig. S5, B to D). These findings were expected as P2RX7 expression was low or absent on WT effector CD8+ T cells in the blood (fig.S5E). Similarly, analysis of LCMV-infected mice at day 30+ after infection showed that the memory response of WT and P2RX7 KO CD8+ T cells was of similar size and composition (Fig. 2, A and B). The frequency of TRM and TCIRC subsets (TCM and TEM) was comparable between WT and P2RX7 KO mice in all analyzed organs. The WT and P2RX7 KO compartments of mixed BM chimeras generated similar frequencies of pathogen-specific TRM at day 30+ after infection with LCMV or Listeria monocytogenes–ovalbumin (LM-OVA) (Fig. 2C and fig. S6), and the expression of ARTC2.2 and CD38 was not influenced by P2RX7 deficiency (fig. S7). These data demonstrate that P2RX7 does not play a role in the generation of CD8+ T cells and is dispensable for the generation of TRM and other CD8+ T cell subsets after infection.

Fig. 2 P2RX7 deficiency does not impair memory differentiation of CD8

+ T cells. (A and B) WT or P2RX7 KO mice were infected with LCMV. More than 30 days after infection (A), the frequency of LCMV-specific CD8+ T cells and (B) the phenotype of the LCMV-specific CD8+ T cells were analyzed [n = 4 to 10, one (BM) or two experiments]. (C) WT:P2RX7 KO mixed BM chimeras were infected with LCMV. More than 30 days after infection, the phenotype of LCMV-specific (GP33 tetramer) WT or P2RX7 KO donor CD8+ T cells was analyzed (n = 5, one experiment). Sal. gl., salivary glands.

P2RX7 activation by NAD+induces cell death of TRMbut not TCIRC

P2RX7 is activated via the extracellular nucleotides ATP and NAD+. To investigate how P2RX7 triggering influences TRM maintenance, we treated TRM isolated from the small intestine or TCIRC from the spleen in vitro with increasing concentrations of NAD+. CD8αβ+ TRM but not TCIRC died after adding NAD+ in a dose-dependent manner (Fig. 3, A and B). In contrast, TRM isolated from P2RX7 KO mice were insensitive to NAD+ treatment (Fig. 3, C and D), suggesting that NAD+-induced cell death of TRM is P2RX7 dependent. Treatment with inhibitors of P2RX7 or the P2RX7-activating ectoenzyme ARTC2.2 confirmed these findings (Fig. 3E). Direct inhibition of P2RX7 activation by the P2RX7 inhibitors A438079 or KN-62 as well as inhibition of ARTC2.2 with nanobody rescued TRM survival after NAD+ treatment, whereas no effect on TCIRC was observed (Fig. 3E). Elevated concentrations of 1 to 10 μM extracellular NAD+ have been detected in inflamed tissues in vivo (22), suggesting that during inflammation and tissue damage, TRM can succumb to NAD+-induced cell death.

Fig. 3 P2RX7 activation by NAD

+ induces cell death of TRM but not of TCIRC. (A to E) IEL or splenocytes were cultured in medium or in the presence of different concentrations of NAD+. After 20 hours, the number of viable CD8αβ+ TRM (IEL) or TCIRC (spleen) was assessed by flow cytometry and normalized to the medium control. (A and B) TRM or TCIRC of WT mice (n = 10/11, four experiments). (C) TRM of WT or P2RX7 KO mice. (D) TRM or TCIRC of WT and P2RX7 KO mice cultured in medium with or without 10 μM NAD+ (n = 7 to 11 per group, five experiments). ns, not significant. (E) TRM or TCIRC of WT mice were treated with P2RX7 inhibitors (A438079, KN-62) or anti-ARTC2.2 nanobody s+16a or dimethyl sulfoxide (DMSO) as control for 30 min before addition of 10 μM NAD+ (n = 4 to 5, three experiments). Unpaired t test (B and D) and paired t test (E). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

P2RX7 expression is modulated by TCR activation and cytokine signals

During infection, inflammation, and tissue damage, the local microenvironment of TRM changes quickly, with alterations in the cellular composition and the presence of antigen and inflammatory cytokines. These signals could modulate the expression of P2RX7 and the ability of CD8+ T cells to respond to extracellular ATP and NAD+. Whereas type I interferons (IFNs) did not change P2RX7 expression, IL-12 up-regulated P2RX7 on TRM (Fig. 4A). ARTC2.2 expression was unchanged (fig. S8A). In addition, during priming of naïve CD8+ T cells, IL-12 induced P2RX7 expression at similar levels to retinoic acid (RA) (fig. S8B), which is known as a P2RX7-inducing agent (31). This suggests that during inflammation IL-12 may sensitize TRM to cell death induced by extracellular nucleotides through up-regulation of P2RX7. The cytokines IL-33, tumor necrosis factor–α (TNF-α), and transforming growth factor–β (TGF-β) associated with TRM differentiation and maintenance did not induce P2RX7 and did not substantially influence P2RX7 maintenance (fig. S8, B and C). Last, we analyzed how recognition of antigen and TCR-mediated activation influenced TRM homeostasis. TCR triggering reduced P2RX7 expression on TRM (Fig. 4, B and C, and fig. S8D) and made them less sensitive to NAD-induced cell death (Fig. 4D and fig. S8, E and F). These data suggest that TCR triggering may selectively rescue antigen-specific over nonspecific TRM from P2RX7-driven cell death induced by tissue damage.

Fig. 4 P2RX7 expression is modulated by inflammatory cytokines and TCR activation.

(A) SI IEL were cultured for 3 days with homeostatic cytokines (recombinant murine IL2, IL7, and IL15) alone or additional IFN-α or IL-12 (10 ng/ml each), and the expression of P2RX7 on CD8αβ+ TRM was assessed by flow cytometry (n = 6, two experiments). (B) IEL or splenic CD8+ T cells were stimulated with plate-bound anti-CD3 or left untreated and cultured for 3 days with homeostatic cytokines. P2RX7 expression on CD8ab+ TRM (IEL) or TCIRC (spleen) was assessed (n = 5, three experiments). (C and D) CD8+ TRM (IEL) or TCIRC (spleen) of OT1 mice were stimulated with plate-bound anti-CD3 or SIINFEKL peptide or left untreated and cultured for 1 day with soluble anti-CD28 and homeostatic cytokines. (C) P2RX7 expression on CD8αβ+ OT1 cells was assessed, or (D) cells were treated with 10 μM NAD+ for 20 hours and the number of viable CD8αβ+ T cells relative to medium control was quantified (n = 4, two experiments). Unpaired t test (A to D). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

P2RX7 activation depletes TRMbut not TCIRCin vivo

To study the impact of P2RX7 activation in vivo, we injected mice with NAD+ and quantified CD8+ T cell subsets in liver, BM, and small intestine intraepithelial lymphocytes (SI IEL) 24 hours after injection. In both liver and BM, we observed a decreased frequency of CD8αβ+ TRM in NAD+-injected mice compared with the phosphate-buffered saline (PBS)–injected control mice (Fig. 5, A and B). Absolute quantification of CD8αβ+ T cell numbers in the organs demonstrated that TCIRC were undisturbed by NAD+ injection and TRM numbers were selectively decreased by 5-fold (liver and BM) or 19-fold (SI IEL) (Fig. 5C). Similar results were obtained for LCMV-specific memory T cells. The frequencies of LCMV-specific TRM in the SI IEL and liver were reduced by 18-fold (SI IEL) or 6-fold (liver) (Fig. 5, D and E) at 24 hours after NAD+ injection. In contrast, the frequency of LCMV-specific TEM cells in the liver remained unchanged in NAD+-injected compared to the PBS-injected control mice (Fig. 5E). To investigate whether the NAD+-induced depletion of TRM was dependent on TRM-expressed P2RX7, we generated mixed BM chimeras. As described in Fig. 2, no defects in the generation of TRM were observed in the absence of P2RX7. WT and P2RX7 KO compartments contributed equally to TEM and TRM in liver, BM, kidney, spleen, and SI IEL in the PBS-injected control mice (Fig. 5F and fig. S9). NAD+ injection led to the loss of WT TRM. In contrast, P2RX7 KO TRM as well as WT and P2RX7 KO TEM were maintained after NAD+ injection (Fig. 5F and fig. S9). We conclude that the elevated concentration of extracellular NAD+ during tissue injury could induce NAD+-driven activation of P2RX7 and regulate TRM maintenance during tissue damage. These data also demonstrate the potential of NAD+ injections as a tool to selectively deplete TRM while leaving TCIRC responses intact.

Fig. 5 P2RX7 activation depletes T

RM but not TCIRC in vivo. (A to C) Mice were injected with PBS or NAD+, and organs were analyzed 24 hours later. (A) Frequencies of CD8αβ+ T cell subsets in the liver and (B) in the BM as well as (C) absolute numbers were analyzed by flow cytometry (n = 4 to 6 per group, three experiments). (D and E) More than 30 days after LCMV infection, mice were injected with PBS or NAD+. Frequencies of LCMV GP33-specific CD8+ T cells with TEM or TRM phenotype were quantified in the (D) IEL and (E) liver 24 hours after NAD+ treatment (n = 4 per group, one experiment). (F) WT:P2RX7 KO mixed BM chimeras were infected with LCMV. More than 30 days after infection, mice were treated with NAD+ or PBS as control. Twenty-four hours after NAD+ treatment, GP33-specific donor CD8+ T cells were quantified (n = 7 per group, two experiments). IEL, small intestine intraepithelial lymphocytes. Unpaired t test (C to F). *P ≤ 0.05, ***P ≤ 0.001.

Tissue damage depletes TRMbut not TCIRCin a P2RX7-dependent manner

To investigate the effect of tissue damage on T cell subsets, we used the circumstance that the disaggregation of organs is often a prerequisite for the isolation of T cells from these locations. TRM or TCIRC subsets were isolated from the liver, which involves mashing of the tissue and therefore exposes T cells within the liver to tissue damage. The isolated subsets were cultured for 6 hours at 37°C. Whereas TCIRC (TCM and TEM) were more than 90% viable after 6 hours of culture, less than 40% of the liver TRM remained viable during this culture period (Fig. 6A). To inhibit activation of P2RX7 by extracellular NAD+ during tissue preparation, mice were injected with ARTC2.2-blocking nanobody before sacrifice, which has previously been shown to enhance viability of P2RX7-expressing lymphocytes (3234). The injection of anti-ARTC2.2 nanobody s+16a improved liver TRM viability during culture to the same level of TCM and TEM (Fig. 6A). CD8αβ+ TRM isolated from SI IEL with minimal mechanical destruction did not demonstrate increased cell death (fig. S10A). These data suggest that the short-term tissue damage induced during the mashing of the liver is sufficient to induce NAD+-induced cell death of TRM. To test the survival of TRM after induction of tissue damage in vivo, we treated LCMV-immune mice (30+ days after LCMV infection) with acetaminophen to induce acute sterile liver injury. In acetaminophen-treated mice, TRM were underrepresented in LCMV-specific CD8+ T cells in the liver compared to PBS-treated controls 24 hours after liver injury induction (Fig. 6B). The tissue damage–induced loss of TRM was P2RX7 specific, because no reduction in the TRM frequency was observed in P2RX7 KO mice after liver injury induction (Fig. 6B). Similar effects were observed in a competitive setting of acetaminophen-induced liver injury in LCMV-immune mixed (WT-P2RX7 KO) BM chimeras (Fig. 6C) as well as in mice that received cotransfer of TCR transgenic WT and P2RX7 KO CD8+ T cells (Fig. 6D). TRM depletion was restricted to the damaged organ, as reduction of WT TRM was observed in the liver (Fig. 6D) but not in the kidney (Fig. 6E) or spleen (fig. S10, B and C) of acetaminophen-treated mice. The depletion of LCMV-specific TRM was long lasting, as 30+ days after liver injury induction, acetaminophen-treated mice still showed a reduction in liver TRM in the WT but not in the P2RX7 KO compartment (Fig. 6, C and D). These data demonstrate that tissue damage during injury leads to a selective P2RX7-dependent depletion of TRM.

Fig. 6 Tissue damage depletes TRM but not TCIRC.

(A) Liver CD8+ T cells from untreated mice or mice that were injected with ARTC2.2-blocking nanobody s+16a were isolated and CD8+ T cell subsets were sorted. Subsets were then cultured for 6 hours, and viability was assessed by flow cytometry (n = 9/12 per group, four experiments). (B) WT or P2RX7 KO mice were infected with LCMV, and 30+ days after infection, mice were treated with acetaminophen to induce liver injury or PBS as control. Twenty-four hours after acetaminophen injection, GP33-specific CD8+ T cells in the liver were analyzed, and the frequency of TRM was quantified (n = 3/4 per group). (C) WT-P2RX7 KO mixed BM chimeras were infected with LCMV. More than 30 days after infection, mice were treated with acetaminophen or PBS as control. More than 30 days after acetaminophen treatment, LCMV-specific donor CD8+ T cells (GP33 tetramer) in the liver were analyzed, and the frequency of TRM within the WT and P2RX7 KO compartment was quantified (n = 4 per group, one experiment). (D and E) Mice were injected with a 1:1 mixture of naïve WT and P2RX7 KO OT1 CD8+ T cells and infected with rLCMV-OVA. More than 30 days after infection, mice were treated with acetaminophen, and 30+ days later, the frequencies of TRM in transferred WT and P2RX7 KO OT1 CD8+ T cells in the (D) liver and (E) kidney were quantified (n = 8 per group, one experiment). Mean ± SEM. Unpaired t test. *P ≤ 0.05, ***P ≤ 0.001.


Our results introduce tissue damage as an important signal that regulates TRM maintenance. Inflammation and tissue damage during priming are important cofactors for the generation of TRM in tissues (35, 36). Activated CD8+ T cells get drawn into inflamed tissues and form TRM at the site of inflammation (prime-pull strategy) (3, 37), and lung CD8+ TRM are preferentially formed at foci that show signs of previous tissue injury (36). However, despite the location of TRM at hotspots of pathogen encounter, it was unknown how inflammation, tissue damage, and unrelated secondary infections affect preexisting TRM. We describe that TRM undergo cell death upon recognition of elevated levels of extracellular nucleotides via the DAMP receptor P2RX7 and that the recognition of tissue damage during sterile tissue injury deletes TRM in a P2RX7-dependent manner.

TRM are fast-reacting memory cells at the front line of the immune system. After recognition of cognate antigen, TRM initiate a tissue-wide alert state and recruit other innate and adaptive immune cells that contribute to the protective function of TRM (4, 5, 38). TRM express elevated levels of mRNA for effector molecules such as IFN-γ and granzyme B (39, 40) and are able to initiate these effector mechanisms ad hoc after activation. However, misguided or uncontrolled TRM responses lead to immunopathology in acute immune responses such as contact dermatitis as well as chronic autoimmune diseases (41). Accordingly, TRM must be tightly controlled. The P2RX7-dependent deletion of TRM during sterile tissue damage could provide a negative feedback mechanism to prevent the unleashing of the TRM-dependent deployment of a proinflammatory immune response. Thereby, TRM-associated immunopathology is limited, and prompt regeneration fostered. The observed further increase of P2RX7 expression on TRM by the proinflammatory cytokine IL-12 suggests that, once inflammation is established, TRM are deleted after recognition of extracellular nucleotides in a P2RX7-dependent manner to quench further immune activation.

On the basis of our results on TRM susceptibility to sterile tissue damage in acetaminophen-induced liver injury, we hypothesize that local tissue damage and inflammation could also deplete preexisting TRM in the course of secondary infections. Because TRM are crucial for the protection of peripheral tissues, depletion of all TRM during infection-induced tissue damage would curtail TRM function. We demonstrate that TCR activation of TRM reduced P2RX7 expression and rescued TRM from cell death induced by extracellular nucleotides. Similar observations have been made for P2RX7 expression on follicular helper T cells (TFH) in Peyer’s patches (42). This suggests that modulation of P2RX7 expression during infections could ensure that infection-relevant TRM are maintained. Extracellular nucleotides released during infection and tissue damage could deplete non–antigen-specific TRM locally and free niches for new and infection-activated specificities. It has been shown that in secondary infections, new TRM can be generated from naïve, effector, and circulating memory T cells (43). As we have demonstrated, TCIRC have both lower expression of P2RX7 and are insensitive to tissue damage–induced cell death. This suggests that the recognition of tissue damage promotes persistence of antigen-specific over bystander TRM in the tissue niche. Because extracellular nucleotides are quickly metabolized, P2RX7-dependent deletion of TRM might be a very localized effect during infection of tissues. Susceptibility of TRM in this context might depend on their exact location as well as the tissue damage, lytic potential, and immunopathology inflicted by the pathogen; e.g., infection with LCMV WE causes substantial liver damage compared with LCMV Armstrong (44). Analysis of the lungs after influenza infection shows distinct patches of regenerating tissue (36), suggesting that within the parenchyma, tissue damage might influence TRM very locally. Accordingly, infections would not completely wipe out all of the TRM but just locally punch holes in the TRM repertoire at sites of tissue damage.

TRM are dependent on survival signals such as IL-15 in the tissues (45). These signals are provided by accessory cells and might be a limiting factor for the dissemination of TRM in the tissue. In line with this, TRM in the skin are generated at the expense of dendritic epidermal γδ T cells (46). With a series of consecutive unrelated infections and immunizations, however, recent data demonstrate that preexisting TRM in the skin are only minimally dislodged by newly generated TRM of different specificities (43). In these experiments, no direct tissue damage or inflammation was induced at the site of primary TRM generation because “new” TRM are generated by immunizations/infections at different unrelated locations. Additionally, the lower P2RX7 mRNA levels in skin TRM compared with TRM in liver and small intestine (29) suggest that TRM maintenance during tissue damage in the skin might be differently regulated. Although we detect P2RX7 expression on most of the TRM in all tissues that we analyzed, TRM in other tissues or induced in other contexts might lack P2RX7 expression, as shown for a subset of influenza-specific TRM in the lung (47).

TRM are underrepresented in lymphocyte preparations from nonlymphoid tissues (48). It is possible that P2RX7-driven recognition of tissue damage during the isolation of lymphocytes contributes to poor TRM retrieval from peripheral tissues. Moreover, other P2RX7-expressing lymphocytes succumb to cell death after triggering of P2RX7 by extracellular NAD+, ATP, or tissue damage induced during the preparation of lymphocytes from the organs (31, 32, 42, 4952). Inhibition of ARTC2.2 or P2RX7 activity is crucial for isolation of functional regulatory T cells (Tregs) as well as liver NKT cells and liver TRM and allows for in vivo survival of these P2RX7- and ARTC2.2-expressing cell populations after adoptive transfer (3234, 53). We found that the ectoenzymes CD38 and CD39, which limit the extracellular concentration of ATP and NAD+ (54), are highly expressed on TRM. Although TRM have counterregulatory mechanisms to reduce extracellular nucleotides, they are responsive to exogenous NAD+ and tissue damage in vivo. CD38 and CD39 are also highly expressed by other P2RX7-expressing lymphocytes such as NKT cells, ILC1, and Tregs (55, 56), as well as TRM-like tumor-infiltrating CD8+ T cells (57). This suggests that CD38 and CD39 function as a fail-safe to prevent inadvertent P2RX7 activation under homeostatic conditions. However, these nucleotide-degrading ectoenzymes do not appear to substantially reduce local concentrations of extracellular nucleotides in the event of tissue damage.

The observed P2RX7-dependent depletion of TRM during tissue damage appears to be mainly mediated by NAD+-dependent constitutive P2RX7 activation via ARTC2.2, as ARTC2.2 inhibition in vivo nearly completely rescued cell death of TRM induced during tissue preparation. Additional studies will be needed to determine whether this or similar pathways also limit TRM maintenance in humans, in which ARTC2.2 is not functional. Recently, a role of P2RX7 in extracellular ATP–driven metabolic programming of CD8+ T cell memory has been demonstrated (16). In this study, P2RX7 was described to affect the maintenance of long-lived memory CD8+ T cells, in contrast to our findings. Differences in genetic background or microbiota might have resulted in discrepancies. Additionally, we build on the results of Borges da Silva et al. (16) and demonstrate here that P2RX7 is highly up-regulated specifically on TRM CD8+ T cells in various tissues. P2RX7 expression has been demonstrated on other T cells that share a tissue-resident phenotype. Liver NKT cells, Tregs, IEL, and TFH in the Peyer’s patches of the small intestine express P2RX7 at high levels (31, 32, 42, 4952). This implies that P2RX7-mediated recognition of tissue damage might be a feature that different types of tissue-resident lymphocytes have in common. The high expression of P2RX7 and ARTC2.2 on liver ILCs suggests that at least ILC1 can respond to tissue damage similarly to NKT cells and TRM.

TRM are retained in the tissue in an alerted mode with elevated levels of effector molecules. Tight control of transcription, translation, and activation of TRM is needed to maintain tissue homeostasis. On the basis of our data, we propose an additional level of TRM regulation in the tissue. Recognition of tissue damage could equip TRM with a suicide mechanism to prevent excessive immunopathology.


Study design

The main aim of this study was to assess the impact of danger signals and tissue damage on TRM CD8+ T cells in peripheral tissues. For the identification of danger signal receptors expressed on TRM, an existing RNA-seq dataset was used. The impact of the identified receptor P2RX7 on TRM generation and maintenance was analyzed with mouse models of LM-OVA and LCMV infection. Local tissue damage in the liver was studied using the acetaminophen-induced liver injury model. Detailed descriptions of the experimental parameters can be found below and in the figure captions. Sample sizes, number of experimental replicates, and statistical tests are indicated in the figure legends.


C57BL/6JRj mice were purchased from Janvier or bred in the animal facility of the NKI (Netherlands Cancer Institute). P2RX7 KO (B6.129P2-P2rx7tm1Gab/J), CD45.1 (B6.SJL-PtprcaPepcb/BoyJ), and OT1 mice [C57BL/6-Tg(TcraTcrb)1100Mjb/J] were purchased from the Jackson Laboratory and maintained in the animal facility of the NKI. For the generation of mixed BM chimeras, B6.SJL-PtprcaPepcb/BoyJ × C57BL6/JRj (Ly5.1 × Ly5.2) recipients were irradiated (2 × 5 gray) and reconstituted with intravenous transfer of 2 × 107 BM cells. Mixed BM chimeras were used 8 to 16 weeks after reconstitution. Chimerism of lymphocytes was confirmed before usage. All mice were maintained under specific pathogen–free conditions. Animal experiments were conducted according to institutional and national guidelines.

Infections and other in vivo treatments

Mice were infected intraperitoneally with 1 × 105 plaque-forming units (PFU) of LCMV Armstrong and 1 × 105 PFU of rLCMV OVA (58) (provided by D. Merkler, University of Geneva) or by feeding of 2 × 109 colony-forming units of LM-OVA InlAM (provided by B. Sheridan, Stony Brook University) as described previously (59). For some experiments, naïve CD8+ T cells (CD44low and CD62Lhigh) of OT1 and P2RX7 KO OT1 mice were sorted and mixed in a ratio of 1:1, and 1 × 104 cells were transferred intravenously 1 day before infection. To trigger P2RX7 in vivo, we injected mice intravenously with 60 mg of NAD+ (Sigma) in PBS or NaCl (pH 7) (200 μl) or PBS as control. To induce liver injury, male mice were fasted overnight, mice were weighed, and acetaminophen (400 mg/kg) (Sigma) in PBS at a concentration of 15 mg/ml was injected intraperitoneally. Equivalent volumes of PBS were injected as a control. For detection of P2RX7, mice were injected intraperitoneally or intravenously with 50 μg of ARTC2.2-blocking nanobody s+16a (BioLegend) 30 min before sacrifice (32).

Tissue preparation

Single-cell suspension of spleen and liver was generated by passing organs over a 70-μm cell strainer. IEL were prepared from the small intestine. After removal of residual fat tissue, Peyer’s patches, and feces, pieces of small intestine were incubated for 30 min at 37°C in Hanks’ balanced salt solution (Gibco) with 10% fetal calf serum (FCS), 5 mM EDTA, and 1 mM dithiothreitol and vortexed extensively. The IEL fraction was isolated by filtering over 70-μm strainers. IEL-depleted intestine pieces, kidney, salivary glands, or lungs were chopped into small pieces and digested for 30 min at 37°C with collagenase type I (750 U/ml) (Invitrogen) and deoxyribonuclease I (0.31 mg/ml) (Roche, from bovine pancreas, grade II) in RPMI with 10% FCS, and single-cell suspensions were generated by filtering over 70-μm strainers. The isolated lymphocytes were purified by density centrifugation on a 60%/40% Percoll gradient (GE Healthcare). CD8ab+ T cells from IEL preparations contained >95% CD69+/CD62L cells. BM was isolated from tibia and femur by crushing the bones in PBS or flushing them with PBS and passing through a 70-μm cell strainer. Contaminating erythrocytes were removed using red blood cell lysis buffer (155 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA).

Flow cytometry

Cells were incubated with antibodies at 4°C for 30 min in the dark. Antibodies were purchased from BioLegend, eBioscience, or Novus Biologicals. H-2 Db KAVYNFATC (GP33), H-2 Db FQPQNGQFI (NP396), H-2 Db SGVENPGGYCL (GP276), and H-2 Kb SGYNFSLGAAV (NP238) tetramers were provided by R. Arens (Leiden University Medical Center, Leiden) and H-2 Kb SIINFEKL tetramer (OVA) was a gift of A. ten Brinke (Sanquin). Exclusion of dead cells was performed with a LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (Molecular Probes). After staining, cells were fixed with the eBioscience Transcription Factor Staining Buffer Set according to the manufacturer’s specifications. Samples were acquired on an LSRFortessa flow cytometer (BD Biosciences), and data were analyzed using FlowJo (TreeStar). Cell sorting was performed using Aria III (BD Biosciences). Exemplary gating strategies are shown in fig. S1.

In vitro treatments

Cells were cultured in vitro in RPMI (Gibco) supplemented with 10% FCS (Bodingo BV), penicillin (100 U/ml) (Sigma), streptomycin (100 μg/ml) (Sigma), 2 mM l-glutamine (Sigma), and 55 μM β-mercaptoethanol (Gibco). To activate or inhibit P2RX7 NAD+ (Sigma), we used A438079 (Abcam, 20 μM), KN-62 (Abcam, 20 μM), or anti-ARTC2 nanobody s+16a (BioLegend, 20 μg/ml). Cells were stimulated with plate-bound anti-CD3 (clone 17A2, 5 μg/ml) or SIINFEKL peptide (GenScript, 1 μg/ml) and soluble anti-CD28 (clone PV-1, 1 μg/ml, both a gift of L. Boon, Bioceros) with or without antigen-presenting cell-containing splenocytes and cultured in the presence of the following recombinant murine cytokines: IL-2, IL-7, and IL-15 (all from R&D Systems, 10 ng/ml), recombinant murine IFN-α or IFN-β (both from eBioscience, 10 ng/ml), IL-12, IL-33, TNF-α (PeproTech, 10 ng/ml), recombinant human TGF-β (PeproTech, 5 ng/ml), or all-trans RA (10 nM).

RNA-seq analysis

Previously published and normalized RNA-seq data (29) of sorted murine CD8+ T cell subsets after LCMV infection were analyzed for PRR-associated genes. Expression was detected for 47 manually selected PRR genes (table S1). Seven of the selected transcripts were differentially expressed (adjusted P < 0.05, fold change > 2, 8 reads per kilobase per million mapped reads) between TRM and TCIRC (TCM or TEM) from liver, small intestine, and spleen.

Statistical analysis

Statistical analysis was carried out using GraphPad Prism. The test used is indicated in the figure legends. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.


Fig. S1. Gating strategies.

Fig. S2. CD69+/CD62L CD8+ T cells have features of TRM cells.

Fig. S3. P2RX7 and ART2.2 are highly expressed by TRM cells.

Fig. S4. Expression of P2RX7, ARTC2.2, CD38, and CD39 on lymphocyte subsets.

Fig. S5. P2RX7 deficiency does not impair effector differentiation of CD8+ T cells.

Fig. S6. P2RX7 deficiency does not impair memory differentiation of CD8+ T cells.

Fig. S7. P2RX7 deficiency does not impair ARTC2.2 and CD38 expression of memory CD8+ T cells.

Fig. S8. Modulation of P2RX7 expression cytokines and TCR activation.

Fig. S9. P2RX7 activation depletes TRM but not TCIRC in vivo.

Fig. S10. Effect of tissue damage on TRM in IEL and spleen.

Table S1. Selected PRR genes expressed in RNA-seq data set.

Table S2. Raw data.


Acknowledgments: We thank A. ten Brinke for providing SIINFEKL tetramer and B. Nota for bioinformatical analysis of the RNA-seq data. We thank B. Sheridan (Stony Brook University) for providing LM-OVA InlAM and D. Merkler (University of Geneva) for providing rLCMV OVA. Funding: R.S. was supported by a Fellowship from the Alexander von Humboldt Foundation and by Veni grant 016.186.116 from the Netherlands Organization for Scientific Research (NWO). F.M.B. and K.P.J.M.v.G. were supported by Vidi grant 917.13.338 from NWO and a fellowship of the Landsteiner Foundation of Blood Transfusion Research. F.K.-N. was supported by grant no. 310/11 from the DFG. Author contributions: R.S., K.P.J.M.v.G., and R.A.W.v.L. designed the experiments. R.S., T.H.W., F.M.B., and N.A.M.K. performed the experiments and analyzed data. R.S. did the statistical analysis. R.A. and F.K.-N. contributed reagents and helpful discussions. R.S., K.P.J.M.v.G., and R.A.W.v.L. prepared the manuscript. Competing interests: F.K.-N. is co-inventor on a patent application on P2RX7-specific nanobodies. F.K.-N. receives a share of antibody and nanobody sales via MediGate GmbH, a wholly owned subsidiary of the University Medical Center Hamburg-Eppendorf. All other authors declare that they have no competing interests. Data and materials availability: The RNA-seq data are available under GSE70813 at GEO.

Stay Connected to Science Immunology

Navigate This Article