We initially observed that HBV transgenic mice injected with HBV-specific CTLs and C57BL/6J mice infected with a replication-deficient adenovirus expressing lacZ (RAd35) showed platelet aggregates in the liver alongside apoptotic hepatocytes and inflammatory cells (Supplementary Fig. 1 online). To verify whether platelets contribute to the development of these lesions, we injected mice with antibodies (α-PLT) against mouse glycoprotein (GP) Ibα (Supplementary Methods online) that caused a >97.5% reduction in platelet count within 30 min and for up to 6 d (Supplementary Table 1 online). Platelet counts remained normal (8–10 × 10⁵/μl of blood) in mice that received saline (NaCl) solution or an irrelevant antibody (α-Irr). We then evaluated liver disease severity in Thy-1.2⁺ HBV transgenic mice injected with α-PLT or α-Irr before transfer of Thy-1.1⁺ HBV-specific CTLs. In agreement with previous results^5–8, the liver of α-Irr–treated mice killed at the peak of serum alanine aminotransferase (sALT) elevation showed scattered necroinflammatory foci, in which loss of hepatocytes (hepatocellular dropout) was readily detectable (Fig. 1a). The degree of hepatocellular dropout as well as the size of necro-inflammatory foci (mean size, 1,245 μm² versus 6,724 μm²; Fig. 1b) was smaller in thrombocytopenic mice, which also showed 80% lower sALT levels (Fig. 1c) and ~40% (P < 0.01) and 30% (P < 0.01) fewer intrahepatic Thy-1.1⁺ HBV-specific CTLs at days 1 and 2 after transfer, respectively (Fig. 1d). CTL function was unaffected by the low platelet count, as Thy-1.1⁺ HBV-specific CTLs recovered from the liver of normal or thrombocytopenic mice had identical cytotoxic activity against target cells that express hepatitis B surface antigen (Fig. 1e) and identical interferon (IFN)-γ (encoded by Ifng) production ex vivo (data not shown). In thrombocytopenic mice, intrahepatic Ifng mRNA (a marker of antigen recognition by CTL^6,9) was reduced (Supplementary Fig. 2 online) in proportion to the number of CTLs in the liver (Fig. 1d), but probably yielded sufficient levels of this antiviral cytokine^6,9 to abolish viral replication (Supplementary Fig. 2 online).

To investigate further the role of platelets in CTL-induced liver disease, we transfused thrombocytopenic mice with washed mouse platelets lacking endogenous but expressing human GP-Ibα (Supplementary Methods online). These platelets (TKK-PLT) are not recognized by α-PLT (Supplementary Fig. 3 online) and function normally in mice¹⁰. Transfusion of 6 × 10⁸ TKK-PLTs into C57BL/6J mice treated with α-PLT 1 d earlier caused platelet counts of ~40% (at 0.5, 3 and 24 h) and ~22% (at 48 h) of NaCl-injected mice, whereas the same number of C57BL/6J-derived platelets yielded counts <3% of control as early as 0.5 h after transfusion (Supplementary Fig. 3 online). TKK-PLT transfusion restored severity of liver disease, intra-hepatic accumulation of CTL and levels of Ifng mRNA (Supplementary Fig. 4 online), unless TKK-PLTs were treated with the activation inhibitor prostaglandin (PG) E₁ before transfusion (Fig. 1f and data not shown). Unlike untreated TKK-PLTs, circulating PGE₁-treated TKK-PLTs isolated at autopsy did not aggregate in response to collagen stimulation in vitro (data not shown). Thus, activated platelets contribute to CTL-mediated immunopathological responses. Accordingly, we found that CTLs adhere to platelets activated on a model-reactive surface ex vivo (Supplementary Video 1 online).

Livers from HBV transgenic mice that received HBV-specific CTLs after injection of α-Irr showed abundant fibrin deposition as compared to NaCl-injected mice; fibrin deposits were reduced in mice rendered thrombocytopenic before CTL injection, but restored in thrombocytopenic mice that received TKK-PLT transfusion before CTL injection (Supplementary Fig. 5 online). This indicates that platelets are required for fibrin deposition. Treatment with the anticoagulant warfarin at a dose that caused a tenfold increase in whole-blood clotting time before CTL transfer also prevented liver fibrin deposition but had no impact on liver damage (Fig. 1g), CTL accumulation or intrahepatic Ifng mRNA levels (data not shown). Fibrin deposition, therefore, is a consequence and not a cause of CTL-induced liver injury.

To confirm our findings in a different model of acute viral hepatitis, we immunized C57BL/6J mice with a plasmid expressing lacZ before infecting them with RAd35. This allowed us to measure the pathogenic lacZ-specific CTL response and define the role of platelets independently of CTL priming. Mice with normal platelet counts killed at the peak of sALT elevation showed an overwhelming disease resembling fulminant hepatitis in humans (Fig. 2a). Thrombocyto-penia markedly improved this picture, with most hepatocytes retaining normal cytological appearance (Fig. 2b), paralleled by a ~90% reduction in sALT activity (Fig. 2c), and an 88% (P < 0.01) and 66% (P < 0.01) reduction in the intrahepatic accumulation of lacZ-specific CTLs at days 3 and 4 after infection, respectively (Fig. 2d). The ability of CTLs recovered from the liver of thrombocytopenic mice to express granzyme B or to produce IFN-γ ex vivo was not impaired (data not shown). Accordingly, intrahepatic Ifng mRNA levels in thrombocytopenic and control mice (Fig. 2e) paralleled the number of intrahepatic CTLs at the same time points (Fig. 2d). By day 5, thrombocytopenic mice showed a more abundant expression of lacZ RNA (Fig. 2f) and higher percentage of β-galactosidase–positive hepatocytes (Fig. 2g), an indication that RAd35 was not readily cleared. This suggests that, in the absence of platelets, the number of functionally normal lacZ-specific CTLs that had accumulated in the liver was not sufficient for viral clearance within the time frame of our experiments. As occurred in HBV transgenic mice, TKK-PLT transfusion in RAd35-infected thrombocytopenic mice restored severity of liver disease, intrahepatic accumulation of CTLs, levels of Ifng mRNA (Supplementary Fig. 4 online) and deposition of fibrin (Supplementary Fig. 5 online).

In conclusion, our findings show that platelet activation is necessary to accumulate virus-specific CTLs at the site of inflammation, thus mediating immunopathogenic responses. Events of this kind may occur during liver infections caused by noncytopathic viruses (such as hepatitis B and C viruses), in which CTLs have a crucial role in the progression of organ damage^1,11,12. The experiments described in this report were approved by the Animal Research Committee of The Scripps Research Institute.

Supplementary Video 1 (MOV 40M)