Promotion of liver regeneration by natural killer cells in a murine model is dependent on extracellular adenosine triphosphate phosphohydrolysis


  • Potential conflict of interest: Nothing to report.

  • See Editorial on page 1688


Nucleotides, such as adenosine triphosphate (ATP), are released by cellular injury, bind to purinergic receptors expressed on hepatic parenchymal and nonparenchymal cells, and modulate cellular crosstalk. Liver resection and resulting cellular stress initiate such purinergic signaling responses between hepatocytes and innate immune cells, which regulate and ultimately drive liver regeneration. We studied a murine model of partial hepatectomy using immunodeficient mice to determine the effects of natural killer (NK) cell-mediated purinergic signaling on liver regeneration. We noted first that liver NK cells undergo phenotypic changes post-partial hepatectomy (PH) in vivo, including increased cytotoxicity and more immature phenotype manifested by alterations in the expression of CD107a, CD27, CD11b, and CD16. Hepatocellular proliferation is significantly decreased in Rag2/common gamma-null mice (lacking T, B, and NK cells) when compared to wildtype and Rag1-null mice (lacking T and B cells but retaining NK cells). Extracellular ATP levels are elevated post-PH and NK cell cytotoxicity is substantively increased in vivo in response to hydrolysis of extracellular ATP levels by apyrase (soluble NTPDase). Moreover, liver regeneration is significantly increased by the scavenging of extracellular ATP in wildtype mice and in Rag2/common gamma-null mice after adoptive transfer of NK cells. Blockade of NKG2D-dependent interactions significantly decreased hepatocellular proliferation. In vitro, NK cell cytotoxicity is inhibited by extracellular ATP in a manner dependent on P2Y1, P2Y2, and P2X3 receptor activation. Conclusion: We propose that hepatic NK cells are activated and cytotoxic post-PH and support hepatocellular proliferation. NK cell cytotoxicity is, however, attenuated by hepatic release of extracellular ATP by way of the activation of specific P2 receptors. Clearance of extracellular ATP elevates NK cell cytotoxicity and boosts liver regeneration. (HEPATOLOGY 2013)

Extracellular adenosine triphosphate (ATP), released actively by stressed cells or derived from necrotic cell death, serves as a potent “danger” signal.1 The release of endogenous nucleotides represents a critical signal modulating cellular crosstalk, injury, and proliferation during liver regeneration.2-5 The regenerative process is closely coordinated by way of paracrine and cell contact-dependent interactions of parenchymal and nonparenchymal cells.6, 7 Extracellular nucleotides modulate immune responses through the activation of specific P2Y (G-protein coupled) and P2X (ligand-gated ion channel) receptors that are expressed on many cell types, including natural killer (NK) cells.8

NK cells are the major sinusoidal lymphocyte population in the human liver. A diverse range of receptors expressed on the surface of NK cells allows them to recognize and rapidly respond to damaged or stressed cells. NK cells also coordinate early events in the innate immune response to injury by rapidly producing cytokines and by cytotoxic activity.

We have previously demonstrated that hepatic NK cells express a repertoire of purinergic receptors for extracellular nucleotides, most prominently P2Y1, P2Y2, P2Y14, P2X3, and P2X6 receptors.9 Interestingly, NK cells, unlike natural killer T (NKT) cells, lack P2X7 receptor, making these cells resistant to ATP-induced apoptosis.7, 10 High concentrations of ATP analogs inhibit NK cell functions in vitro and abolish the production of interferon gamma (IFNγ), indicating that ATP may suppress NK cell activation and regulates inflammatory insults, such as after ischemia reperfusion injury.9 Furthermore, NK cells have been shown to modulate injury indirectly by way of interactions with other innate immune cells such as macrophages.11, 12 In the liver, NK cells interact with Kupffer cells to influence cytokine and chemokine levels within the proliferating tissues.13 Recently, Di Santo and colleagues proposed a beneficial role of NK cells on outcome in the regenerating liver.14 However, mechanisms modulating NK cell activation and function post-partial hepatectomy (PH) remain largely unexplored and are the focus of this study.

Here we show that NK cells exhibit specific markers of activation and cytotoxicity post-PH and that the absence of NK cells in Rag2/common gamma-null mice is associated with significantly reduced liver regeneration, potentially by way of interactions with Kupffer cells. Release of extracellular ATP in response to PH inhibits NK cell cytotoxicity in vivo and in vitro, and this process can be reversed by clearance of extracellular ATP. Thus, we propose that NK cell responses after PH are salutary but are inhibited by extracellular ATP—an effect that can have a potential therapeutic impact on liver regeneration.


ADP, adenosine diphosphate; ADPβS, adenosine diphosphate beta S; ALT, alanine aminotransferase; AST, aspartate aminotransferase; ATP, adenosine triphosphate; ATPγS, adenosine triphosphate gamma S; BrdU, 5-bromo-2′-deoxyuridine; CD39/E-NTPDase1, ecto-nucleoside triphosphate diphosphohydrolase 1; HPF, high-power field; IFNγ, interferon gamma; MFI, mean fluorescent intensity; NK, natural killer cell; NKT, natural killer T cell; PH, partial hepatectomy; UDP, uridine diphosphate; UTP, uridine triphosphate.

Materials and Methods

Animal Model.

Animals were housed in accordance with the guidelines from the Swiss Veterinary Office, which approved all research protocols. C57BL/6 strains of wildtype mice (Harlan, Netherlands, and DKF animal facility, Bern, Switzerland) were studied. Rag1-null mice were purchased from Jackson Laboratories; Rag2/common gamma-null mice were purchased from Taconic (Germantown, NY); P2Y2-null mice were kindly provided by Wolfgang G. Junger (Beth Israel Deaconess Medical Center, HMS, Boston, MA). Mice had free access to a standard mouse chow. For adoptive transfer experiments, intravenous injections were performed in the saphenous vein in animals of 8 to 10 weeks of age under anesthesia using medetomidin 0.5 mg/kg, climazolamum 5 mg/kg, and fentanyl 0.05 mg/kg intraperitoneally. A total of 106 sorted splenic NK (NK1.1pos CD49bpos CD3neg) cells were injected intravenously in 100 μL of phosphate-buffered saline (PBS) with a 30G needle. At the time of sacrifice, mice were anesthetized, blood was taken from the inferior vena cava, and liver lobes were removed for further processing. Gadolinium chloride was injected intravenously at a concentration of 2 μg/g mouse into Rag2/common gamma-null mice 24 hours before PH. Anti-NK1.1 (300 μg per mouse) and anti-NKG2D (500 μg per mouse) were given intraperitoneally 24 and 6 hours before PH.

For PH experiments, ligature of the median and left lobes was performed in order to achieve standard hepatectomy; the right inferior lobe was additionally ligated in order to achieve extended hepatectomy. The ligated liver lobes were removed, weighed, and further processed. Liver regeneration was monitored by bromodeoxyuridine (BrdU) incorporation and Ki67 immunostaining. Fifty mg/kg of BrdU was injected intraperitoneally 2 hours prior to sacrifice. Forty-eight hours after PH, blood was harvested from the inferior vena cava and the remnant liver lobes were removed, weighed, and further processed.

Reagents and Antibodies.

The following antibodies were used for flow cytometry analysis: antimouse NK1.1 (APC, allophycocyanin), CD49b (PE-Cy7), CD3 (FITC, fluorescein isothiocyanate), CD4 (PE, phycoerythrin), CD8 (PE), CD19 (PE), CD45 (Pacific blue), CD11b (PE), NKG2AB6 (PE), NKG2D-clone CX5 (PE), CD16/32 (PE), CD107a (PE), and NKp46 (PE) were purchased from eBioscience; CD11b (APC-Cy7) and CD27 (PE-Cy7) were purchased from Biolegend. SORP LSRII (BD Bioscience) was used for flow cytometry and data were analyzed with FlowJo software. FACSAria (BD Bioscience) was used for cell sorting experiments. For in vivo experiments antimouse NK1.1 (clone PK-136) and antimouse NKG2D (clone HMG2D) were purchased from Bioxcell. The following antibodies were used for immunohistochemistry: goat antimouse NKp46/NCR1 from R&D Systems and rat antimouse F4/80 from Serotec, BrdU from BD Pharmingen, and Ki67 from DAKO. Apyrase, ATP, ATPγS, and ADPβS were purchased from Sigma, 2-thio-UTP (uridine triphosphate), MRS2500, and TNP-ATP were purchased from Tocris. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were measured using a standardized enzymatic assay (Roche Modular P800). Serum ATP levels were measured by ATP Bioluminescence Assay Kit (Roche). Phosphohydrolysis of ATP was stopped directly after blood sampling by adding a stop solution.15


Immunohistochemistry for BrdU incorporation, Ki67 immunostaining, and NKp46 and F4/80 double-staining was performed on paraffin-embedded and snap-frozen liver sections. The tissue was fixed with 4% paraformaldehyde (PFA) for 24 hours. Heat treatment with citrate buffer (pH 6) was performed, sections were incubated with 3% H2O2 for 10 minutes, blocked with normal serum for 15 minutes, and incubated with primary antibody for BrdU and Ki67 or NKp46 and F4/80 O/N at room temperature (RT). Sections were then rinsed with PBS and incubated with secondary antibody for Ki67 for 1 hour at RT, rinsed with PBS, and incubated with streptavidin-HRP (horseradish peroxidase) for 30 minutes at RT. Staining and detection with DAB were performed with the kit from BD Pharmingen. Total numbers of BrdU and Ki67-labeled hepatocytes were determined by counting positively stained hepatocyte nuclei in 40× high-power microscope fields per liver. For BrdU and Ki67 positive percentages of total hepatocytes unstained nuclei were additionally counted in 40× high-power microscope fields per liver. Four fields for each liver section were counted.

Purification of Liver Mononuclear cells and Splenocytes.

Livers were excised and passed through a 200 G stainless steel mesh. The filtrate was centrifuged at 50g for 3 minutes and the supernatant was collected. The nonparenchymal cell supernatant fraction was washed once. Cells were resuspended in a 35% Percoll (GE Healthcare) solution and overlaid on a 70% Percoll solution. After centrifugation at 800g for 20 minutes, the interphase was collected, stained with specific antibodies, and analyzed by BD SORP LSR II. For adoptive transfer experiments and 51Cr release assay, NK cells were purified from the spleen. To enrich NK cells, CD4, CD8, and CD19 (all PE-labeled) positive cells were depleted using PE electromagnetic beads and LS columns according to the manufacturer's protocol (Miltenyi Biotec, Auburn, CA). The flow-through was labeled with NK1.1-APC, CD49b-PECy7, and CD3-FITC for sorting by BD FACSAria.

51Cr-release Assay.

A standard 4-hour 51Cr-release assay was performed using the NK cell-sensitive murine lymphoma cell line YAC-1 as target cells. 106 target cells were radiolabeled with 500 mCi of 51Cr (PerkinElmer, NEZ03000) per 500 μL cell suspension for 1 hour at 37°C. Radiolabeled cells were washed twice and incubated for 1 hour at 37°C and then washed again. Primary murine NK cells (effectors) were sorted by flow cytometry. Targets and effectors were mixed at a ratio of 1:20 in a final volume of 200 μL in a U-bottomed 96-well plate for cell suspension. The mixtures were incubated for 4 hours at 37°C and 5% CO2/air atmosphere. After incubation, 20 μL 10% Triton X-100 was added to each positive control. The supernatants were measured for 51Cr release on a gamma-counter (COBRA II, Packard S/N 404470) and specific lysis was calculated.

Statistical Analysis.

Results are expressed as mean, where data is log-normal distributed and mean ± standard deviation or standard error of the mean for linear datasets. For statistical analysis, Student's t test was used. Statistical significance was defined as P < 0.05.


NK Cells Are Crucial for Optimal Liver Regeneration.

The impact of NK cells on liver regeneration was assessed in a mouse model of PH. We contrasted the regenerative capacity of the liver in C57BL/6 wildtype, Rag2/common gamma-null mice (lacking T, B, NKT, and NK cells), and Rag1-null mice (lacking T, B, and NKT cells but retaining NK cells). Rag2/common gamma-null mice displayed a significant impairment in hepatic regeneration as assessed by incorporation of BrdU (Fig. 1A,B) and immunostaining of Ki67 (Fig. 1C) for proliferating hepatocytes. Plasma levels of AST and ALT were used as markers of injury in response to impaired liver regeneration. Correspondingly, both AST (Fig. 1D) and ALT (data not shown) were significantly increased after PH in the absence of T, B, and NK cells as assessed in Rag2/common gamma-null mice compared to C57BL/6 wildtype mice. There was no impairment of liver regeneration (Fig. 1E), and there was no difference of liver injury (data not shown) after PH in the absence of only T and B cells, as assessed in Rag1-null mice when compared to C57BL/6 wildtype mice. Taken together, these data demonstrate a critical role for liver lymphocytes in driving regeneration of the liver and indicate an important role for common gamma-chain-dependent cell types, most likely NK cells over T, B, and NKT cells, in promoting liver regeneration after PH.

Figure 1.

NK cells are required for optimal liver regeneration. (A) Representative liver sections of BrdU incorporation in C57BL/6 wildtype and Rag2/common gamma-null mice 48 hours after PH. (B) Liver regeneration was assessed by BrdU incorporation and (C) Ki67 immunostaining in C57BL/6 wildtype (n = 5) and Rag2/common gamma-null mice (n = 5) 48 hours after PH (counts of BrdU-positive or Ki67-positive cells per high-power field [HPF] with a magnification of 40×). (D) Liver injury was assessed by AST levels in C57BL/6 wildtype (n = 5) and Rag2/common gamma-null mice (n = 5) 48 hours after PH. (E) BrdU incorporation in C57BL/6 wildtype (n = 6) and Rag1-null (n = 8) mice are shown 48 hours after PH. Each data point represents an individual mouse and the line represents the mean.

PH Stimulates NK Cell Cytotoxicity.

Previous reports have shown that NK cell infiltration to the regenerating liver occurs after PH7, 16, 17 and that hepatic NK cells are a highly heterogeneous cell population.18 We examined the composition of NK cell subpopulations, in particular markers of NK cell activation and cytotoxicity, during liver regeneration after PH (Figs. 2, 3). The percentage of NK cells displaying the surface degranulation marker CD107a (LAMP-1) was increased 24 hours after PH (Fig. 2A,B), indicating that surgical resection provides a signal for NK cell cytotoxicity in vivo.

Figure 2.

NK cells exhibit elevated markers of in vivo cytotoxicity. (A,B) Isolated hepatic NK (NK1.1pos CD3neg) cells from C57BL/6 wildtype mice were characterized in vivo pre-PH and 4 and 24 hours post-PH. NK cell cytotoxicity was assessed by using the marker CD107a, an endosomal membrane protein, which appears on the cellular surface during granule exocytosis, revealing a specific temporal pattern of NK cell activation post-PH. Each data point represents an individual mouse and the line represents the mean. (C) Localization was performed by immunohistochemistry with alkaline phosphatase (blue) for NK cells (NKp46pos) and with peroxidase (brown) for macrophages (F4/80pos) on liver tissue pre- and 4 hours post-PH. Arrows indicate representative colocalizations of NKp46pos with F4/80pos cells. (D) Anti-NK1.1 and anti-NKG2D monoclonal antibodies were administered to C57BL/6 wildtype mice (n = 8) revealing a significant decrease of hepatocellular proliferation in response to anti-NKG2D. (E) Rag2/common gamma-null mice treated with gadolinium chloride (n = 6) 24 hours before PH in order to inactivate Kupffer cells were compared to Rag2/common gamma-null mice treated with PBS (n = 8). Liver regeneration post-PH was significantly elevated after pretreatment with gadolinium chloride compared to PBS. Data are expressed as mean ± SEM.

Figure 3.

PH associates with differential expression of NK cell-specific surface markers. (A,B) Isolated hepatic NK (NK1.1pos CD3neg) cells from C57BL/6 wildtype mice were characterized in vivo pre- and post-PH at 4 and 24 hours by flow cytometry and compared to controls before PH. NK cell maturation was assessed by using CD27 and CD11b as differentiation markers. (C) Immature (CD27high CD11blow) and mature (CD27low CD11bhigh) NK (NK1.1pos CD3neg) cell subsets are represented as fractions of total CD45pos cells. NK cell phenotype was assessed using (D) activating NKG2D receptors on NK (NK1.1pos CD49bpos CD3neg) cells, (E) activating CD16/32 receptors on NK (NK1.1pos CD3neg) cells, and (F) inhibitory NKG2A receptors on NK (NK1.1pos CD49bpos CD3neg) cells. Data in (A) are representative of two independent experiments. Each data point represents an individual mouse and the line represents the mean.

Next, we looked at the localization of NK cells in the liver pre- and post-PH. Immunohistochemistry revealed that NKp46pos (a highly specific NK cell marker, coexpressed with NK1.1 marker on the large majority of liver NK cells) cells are localized throughout the sinusoids (Fig. 2C). Costaining of NKp46 and F4/80 (a specific macrophage marker) suggested that NK cells colocalize with Kupffer cells in the liver after PH (Fig. 2C). Recent literature revealed interactions of NK cells with macrophages.11, 12 In particular, NK cells recognize and kill macrophages in an NKG2D-dependent manner and thereby exhibit regulatory functions in terms of macrophage activity. After administration of monoclonal anti-NKG2D antibody hepatocellular proliferation was significantly reduced compared to untreated wildtype controls (Fig. 2D). Inactivation of Kupffer cells using gadolinium chloride increased liver regeneration only in Rag2/common gamma-null mice but not in parallel with anti-NKG2D administration in wildtype mice, suggesting a pleiotropic effect of NKG2D (Fig. 2D,E).

NK Cell Differentiation In Vivo Alters in Response to PH.

NK cell maturation stage and repertoire of activating and inhibitory receptors were assessed before and after PH to describe phenotypic changes. In the regenerating liver, there was an accumulation of immature CD27high CD11blow NK cells (Fig. 3 A,B). Correspondingly, there was a loss of the mature CD27low CD11bhigh NK cell subset, suggestive of activation and subsequent depletion of the mature NK cell fraction (Fig. 3A,B). The percentage of immature CD27high CD11blow NK cells also increased as a proportion of total liver leukocytes (CD45pos cells), which indicates a recruitment of immature NK cells from the periphery (Fig. 3C). There was a corresponding decrease of the mature CD27low CD11bhigh NK cell fraction 4 hours after PH (Fig. 3C). The expression of activating (NKG2D, CD16/32) and inhibitory (NKG2A/C) receptors was determined on NK cells post-PH. There was a significant enrichment of NK cells expressing NKG2D receptors at 4 hours, which returned to baseline by 24 hours post-PH (Fig. 3D). Moreover, NK cells expressing CD16/32 receptors were enriched at both 4 and 24 hours post-PH (Fig. 3E). However, the expression of NKG2A/C receptors was not significantly altered post-PH (Fig. 3F).

ATP Is Released Post-PH.

It has recently been suggested that extracellular ATP is released after PH and can modulate liver regeneration.19 We sampled blood at short time intervals after surgery up to 20 minutes post-PH and confirmed that ATP is released in our murine model of PH. Concentrations of extracellular ATP reached a peak at 5 minutes after PH and then returned to baseline levels (Fig. 4A). Concentrations of extracellular ATP were significantly decreased 20 minutes post-PH in the portal vein compared to the systemic circulation, indicating that the liver rather than the intestine is the main source of extracellular ATP in the serum (Fig. 4B).

Figure 4.

ATP is released post-PH. (A) Extracellular ATP concentrations were measured by luminescence at different timepoints during surgical liver resection in the inferior vena cava and (B) 20 minutes after partial hepatectomy in the portal vein. (A) C57BL/6 wildtype mice (n = 4) per timepoint; data are expressed as mean ± SEM. (B) C57BL/6 wildtype mice (n = 3), data are expressed as mean ± standard deviation.

NK Cells Improve Liver Regeneration in Response to ATP Clearance.

We examined the effect of extracellular ATP on liver regeneration and in particular its effect on NK cell activity in vivo using C57BL/6 wildtype mice. The enzyme apyrase (soluble CD39/NTPDase1), a highly efficient ATP phosphohydrolase, was administered to decrease extracellular ATP levels.20 Standard (60%) and extended (80%) PH were performed with and without the administration of exogenous apyrase and liver regeneration was assessed by immunohistochemistry 48 hours after surgery. BrdU incorporation was significantly elevated in response to apyrase compared to controls after standard (Fig. 5A,B) and extended (Fig. 5C) hepatectomy. Ki67 staining showed analogous results (data not shown). Therefore, a reduction of extracellular ATP levels is associated with an improvement of liver regeneration in C57BL/6 wildtype mice. To demonstrate the importance of ATP on NK cell function in liver regeneration, adoptive transfer experiments were performed in Rag2/common gamma-null mice. Sorted NK cells from C57BL/6 wildtype mice were adoptively transferred in Rag2/common gamma-null mice pretreated with apyrase 30 minutes before liver resection and compared to controls without adoptive transfer. BrdU incorporation (Fig. 5D) and Ki67 immunostaining (Fig. 5E) revealed a significant increase in hepatocyte proliferation after NK cells were adoptively transferred in the immunodeficient mice. These data suggest that ATP depletion by apyrase improves liver regeneration and that this beneficial effect is NK cell-dependent.

Figure 5.

Increased liver regeneration post-PH in response to decreased levels of extracellular ATP. (A,B) Parameters of hepatic proliferation were assessed by BrdU incorporation in C57BL/6 wildtype mice (n = 6) 48 hours after standard (60%) and (C) extended (80%) PH. Data are expressed as mean ± standard deviation. (D,E) Adoptive transfer of sorted splenic NK (NK1.1pos CD49bpos CD3neg) cells from C57BL/6 wildtype mice was performed in Rag2/common gamma-null mice (n = 8) and compared to controls without adoptive transfer (n = 6). Six days after adoptive transfer, mice were treated with apyrase (50 U/mouse) 30 minutes before PH was performed. (D) Parameters of liver regeneration were assessed by BrdU incorporation and (E) Ki67 immunostaining. Data are expressed as mean ± SEM.

NK Cell Cytotoxicity Is Increased in Response to ATP Clearance Post-PH.

Next, we investigated the phenotype of NK cells post-PH in response to decreased levels of extracellular nucleotides. In mice treated with apyrase there was an increase in the fraction of NK cells expressing the degranulation marker CD107a compared to nontreated controls 4 hours after PH (Fig. 6A,B). Likewise, apyrase increased the fraction of CD107pos NK cells 4 hours after PH compared to apyrase-treated controls before PH. Thus, the reduction of extracellular ATP levels appears to boost NK cell cytotoxicity in vivo post-PH.

Figure 6.

Clearance of ATP elevates NK cell cytotoxicity post-PH. (A,B) Isolated hepatic NK (NK1.1pos CD3neg) cells from C57BL/6 wildtype mice were characterized in vivo pre- and 4 hours post-PH in response to administration of apyrase (50 U/mouse) as compared to controls injected with PBS. NK cell cytotoxicity was assessed using CD107a as marker for degranulation before and 4 hours after PH and administration of apyrase (blue) or PBS (red). Each data point represents an individual mouse and the line represents the mean.

Extracellular ATP Inhibits NK Cell-Mediated Cytotoxicity.

To investigate further the effect of ATP on NK cell activation, we examined the cytotoxic function of NK cells in vitro. 51Chromium release assays were used to determine the effect of extracellular nucleotides on cytotoxicity of purified NK cells toward YAC-1 target cells. NK cell-mediated cytotoxicity was decreased in response to extracellular ATP in a dose-dependent manner with concentrations of ATP up to 100 μM (Fig. 7A). ATPγS, a nonhydrolyzable ATP analog, exerted an almost identical suppressive effect on NK cell cytotoxicity (Fig. 7A). This indicates that the inhibitory effect on NK cell-mediated cytotoxicity can be attributed to recognition of ATP rather than natural breakdown products of ATP such as adenosine.

Figure 7.

Extracellular nucleotides decrease NK cell cytotoxicity. (A) Cytotoxicity of sorted splenic C57BL/6 wildtype NK (NK1.1pos CD49bpos CD3neg) cells to radioactively labeled YAC-1 cells (NK cell-sensitive tumor cell line) was assessed by measuring 51Cr release. NK cells were incubated at a target to effector ratio of 1:20 for 4 hours with ATP and ATPγS (nonhydrolyzable ATP analog). (B) Cytotoxicity was assessed in vitro by measuring CD107apos NK cells excluding direct toxic effects of extracellular ATP on target cells. Splenic NK cells isolated by MACS (depletion of CD4, CD8, and CD19-positive cells) were incubated with NK cell-sensitive YAC-1 tumor cell line at a target to effector ratio of 1:10 for 4 hours with ATPγS 10 μM (blue, mean fluorescent intensity [MFI] 399) and 100 μM (green, MFI 366) compared to control (red, MFI 609). (C) Only CD27low but not CD27high sorted NK (NK1.1pos CD49bpos CD3neg) cells were responsive to ATP-mediated inhibition of cytotoxicity. (D) Sorted splenic NK (NK1.1pos CD49bpos CD3neg) cells were incubated with ADPβS (nonhydrolyzable ADP analog) at different concentrations 1, 10, 100 μM. (E) Sorted NK (NK1.1pos CD49bpos CD3neg) cells from C57BL/6 wildtype or P2Y2-null mice and YAC-1 target cells were incubated with 10 and 100 μM 2-thio-UTP (specific P2Y2 agonist). (F) Sorted NK (NK1.1pos CD49bpos CD3neg) cells from C57BL/6 wildtype or P2Y2-null mice and YAC-1 target cells were incubated with 1 μM ATPγS and 10 nM MRS 2500 (highly selective P2Y1 receptor antagonist). (G) Sorted NK (NK1.1pos CD49bpos CD3neg) cells from C57BL/6 wildtype mice and YAC-1 target cells were incubated with 1 and 10 μM TNP-ATP (P2X1, P2X3, P2X4 receptor antagonist). Representative data of two or three independent experiments are shown as mean ± standard deviation from triplicate values.

By analyzing NK cell cytotoxicity at the single-cell level, the percentage of NK cells displaying CD107a on their surface was greatly reduced by the presence of 10 and 100 μM ATPγS (Fig. 7B). This demonstrates that ATP exerts its effect directly on NK cell cytotoxicity, and not on the target cells.

We further demonstrated that ATP-mediated inhibition acts only on mature (CD27low) NK cells, whereas cytotoxicity of immature (CD27high) NK cells was not reduced by the presence of 10 and 100 μM ATP (Fig. 7C). The same effect was observed using ATPγS (data not shown).

To identify which P2 receptor is responsible for the effect of ATP on NK cell-mediated cytotoxicity, we tested specific P2 receptor agonists and antagonists on NK cell cytotoxicity. P2 receptors are well characterized; P2Y are G-protein coupled receptors, which are activated by several naturally occurring nucleotides (e.g., ATP, ADP, UTP, UDP, UTP-glucose), whereas P2X are ligand-gated ion channel receptors and activated principally by ATP.21 To determine if P2Y receptors are also potentially involved in NK cell-mediated cytotoxicity ADPβS, a nonhydrolyzable ADP analog, was tested. Cytotoxicity toward YAC-1 cells was significantly decreased in response to ADPβS (Fig. 7D), which indicates also a P2Y receptor-mediated effect.

A limited repertoire of P2 receptors is expressed on NK cells, e.g., P2Y1, P2Y2, P2Y14, P2X3, and P2X6,9 and screening this list for receptors responsive to ATP and ADP suggested P2Y1, P2Y2, and P2X3 receptors as strong candidates. The receptor P2Y14 was excluded because it is a specific UTP receptor and the receptor P2X6 was excluded because it does not function as a homomultimer but requires P2X2 or P2X4 receptors. We next employed isolated NK cells from genetic knockout mice and specific agonists and antagonists to examine the role of P2Y2 receptors. Two-thio-UTP, a specific P2Y2 receptor agonist, did not inhibit the cytotoxic activity of C57BL/6 wildtype and P2Y2-null NK cells (Fig. 7E). Intriguingly, MRS2500, a specific P2Y1 receptor antagonist, blocked the effect of ATPγS in P2Y2-null NK cells but not in C57BL/6 wildtype NK cells, implicating a complementary effect of these two receptors (Fig. 7F). A further candidate receptor expressed on NK cells and responsive to ATP is P2X3. We observed a significant effect with TNP-ATP, a specific inhibitor of P2X1, P2X3, and P2X4 receptors (Fig. 7G). Because P2X1 and P2X4 receptors are not expressed on NK cells, these data suggest that P2X3 receptors also contribute to ATP-mediated suppression of cytotoxicity.


The liver has the remarkable and unique ability to regenerate and restore full functions after tissue damage. Here we report an important part of the complex regulatory network governing this process. We note that NK cells degranulate after PH and play a crucial role during liver regeneration by increasing hepatocyte proliferation and decreasing liver injury potentially by way of crosstalk with Kupffer cells. Extracellular ATP that is released after hepatectomy modulates this process by way of the activation of specific P2 type receptors on NK cells.

We have three lines of evidence that demonstrate a relevant role of NK cells and extracellular ATP in the promotion of liver regeneration: (1) decreased levels of liver regeneration in Rag2/common gamma-null mice when compared to Rag1-null and wildtype mice; (2) detection of cytotoxic NK cells post-PH, with associated recruitment of immature NK cells and decreases in mature NK cell subsets; and (3) improvement in liver regeneration when NK cells are reconstituted in Rag2/common gamma-null mice by adoptive transfer in the presence of extracellular ATP-phosphohydrolyzing enzymes.

As we observe changes in the frequency of NK cells expressing specific activating and inhibitory receptors, it remains possible that specialized subsets of NK cells are responsible for these divergent functions and are activated by different stimuli. In other organs, such as the intestine and the uterus, it has been shown that subsets of NK cells are required for tissue repair and regeneration.22, 23 After clonal expansion, mature NK cells do not change their repertoire of receptors; therefore, these shifts must represent differential recruitment and/or cell death within the liver NK cell population during the early stages of liver regeneration. Recent studies that have characterized distinct subsets of hepatic NK cells highlight the complexity and heterogeneity of NK cells in the liver. In particular, in vivo fate-mapping experiments revealed the existence of two distinct subsets of NK cells based on the expression of NKp46 and CD11b.18

Previous studies have investigated the effect of “NK cell activation” induced by systemic administration of the Toll-like receptor 3 agonist, poly I:C, or by virus infection. However, these treatments are also associated with systemic type 1 interferon production,24, 25 which is known to promote tissue apoptosis and shutdown of proliferative signaling. By comparison with these previous studies, our data suggest that the nature of NK cell activation observed in response to cellular damage during PH produces cells with distinct functions (promotion of organ repair) to those induced by “infectious”-type stimuli, which rather promote hepatocyte lysis. Further, in our experiments standard hepatectomy in the mouse strains used was not associated with decreased survival seen in other studies.26 Thus, in order to address clinical applicability, we performed extended hepatectomy and observed an increase of liver proliferation in response to decreased levels of extracellular ATP (Fig. 5C).

NK cells function in the lytic clearance of cells during viral infection or tumorigenesis.27, 28 It may seem counterintuitive for a cell type typically associated with proinflammatory cytotoxic clearance of infected or transformed cells to improve tissue regeneration. However, recent literature revealed crosstalk between NK cells and macrophages that is associated with altered phagocytic and/or inflammatory response of targeted macrophages.12, 29, 30 We have shown a spatial correlation of NK and Kupffer cells and a significant change of phenotype after inactivation of NKG2D. Furthermore, inactivation of Kupffer cells in the absence of NK cells in Rag2/common gamma-null mice significantly improved outcome. Thus, NKG2D-dependent interactions between NK and Kupffer cells are likely to modulate liver regeneration and impact outcome post-PH.

In agreement with previous studies, we observed a release of extracellular ATP post-PH.5 However, in this study performed in rats, a single injection of an unspecific P2 (mainly P2X) receptor antagonist was performed directly into the portal vein post-PH. This boost injection was associated with impaired liver regeneration, potentially by way of stunning of hepatocytes. Extracellular ATP has many pleiotropic effects and its function depends on the concentration of the nucleotide, on the cell type, and on the distribution of P2-type receptors on the target cell. In our current and previous studies, we showed that both continuous and single reduction of extracellular ATP levels are associated with elevated hepatocellular proliferation in response to signaling responses by NK cells or sinusoidal endothelial cells. Interestingly, the impact on outcome of elevated levels of extracellular ATP critically depends on the context and the effector function of NK cells. We previously observed that ATP reduces the secretion of interferon gamma and thereby is protective in a model of hepatic ischemia/reperfusion injury.9 Now we demonstrate that extracellular ATP inhibits NK cell-derived cytotoxicity and thereby impairs liver regeneration.

The interplay between “danger signals,” innate lymphocytes and tissue regeneration and repair appears highly orchestrated. This process may serve to balance the clearance of potentially transformed or infected cells while minimizing further injury and also accelerating restoration of tissue function. The physiological role of ATP-dependent inhibition of NK cell activation is likely to represent an evolutionarily ancient pathway to limit inflammation in badly damaged tissues. Such inhibition of inflammation by ATP, however, limits liver regeneration in a deleterious manner. Conversely, reduction of extracellular ATP levels may enhance components of tissue repair. Thus, pharmacological strategies boosting rates of ATP phosphohydrolysis could improve and accelerate liver tissue regeneration.


The authors thank Andrew J. Macpherson and Kathleen D. McCoy for critical review of the article and support throughout the project, and Wolfgang G. Junger for generously providing the P2Y2-null mice.