Intrahepatic Lymphocyte Expression of Dipeptidyl Peptidase I-Processed Granzyme B and Perforin Induces Hepatocyte Expression of Serine Proteinase Inhibitor 6 (Serpinb9/SPI-6)1

Human proteinase inhibitor 9 (PI-9/serpinB9) and the murine ortholog, serine proteinase inhibitor 6 (SPI-6/serpinb9) are members of a family of intracellular serine proteinase inhibitors (serpins). PI-9 and SPI-6 expression in immune-privileged cells, APCs, and CTLs protects these cells against the actions of granzyme B, and when expressed in tumor cells or virally infected hepatocytes, confers resistance to killing by CTL and NK cells. The present studies were designed to assess the existence of any correlation between granzyme B activity in intrahepatic lymphocytes and induction of hepatic SPI-6 expression. To this end, SPI-6, PI-9, and serpinB9 homolog expression was examined in response to IFN-α treatment and during in vivo adenoviral infection of the liver. SPI-6 mRNA expression increased 10- to 100-fold in the liver after IFN-α stimulation and during the course of viral infection, whereas no significant up-regulation of SPI-8 and <5-fold increases in other PI-9/serpinB9 homolog mRNAs was observed. Increased SPI-6 gene expression during viral infection correlated with influxes of NK cells and CTL. Moreover, IFN-α-induced up-regulation of hepatocyte SPI-6 mRNA expression was not observed in NK cell-depleted mice. Additional experiments using genetically altered mice either deficient in perforin or unable to process or express granzyme B indicated that SPI-6 is selectively up-regulated in hepatocytes in response to infiltration of the liver by NK cells that express perforin and enzymatically active granzyme B.

I ncreased numbers of activated NK cells and CTL infiltrate the liver during systemic (1) and localized immune responses (2). These cytotoxic effector cells play an essential role in the clearance of noncytopathic viruses from infected tissues (3)(4)(5). NK and NKT cells also amplify hepatocellular injury initiated by noninfectious causes of liver injury (6). Virally infected hepatocytes are resistant to killing by the perforin and granzyme-dependent cytotoxic effector pathway (7), the predominant killing pathway utilized by CTL and NK cells in clearance of noncytopathic viruses from many nonhepatic tissues. Consequently, Fas ligand and TNFR-mediated killing mechanisms play a more prominent role in clearance of viral infection from the liver than is observed during clearance of viral infections from extrahepatic sites (7)(8)(9).
Granzyme B, a serine proteinase present in the specialized cytotoxic granules of NK cells and CTL, plays a key role in the induction of cell death mediated by granule exocytosis effector mechanisms (4). Granzyme B is synthesized as an inactive zymogen and requires the removal of its propeptide by cathepsin C/dipeptidyl peptidase I (DPPI) 3 for activation (10). Human proteinase inhibitor 9 (PI-9) and the murine ortholog serine proteinase inhibitor (SPI) 6 are members of a family of OVA serpin inhibitors present in immune-privileged cells, APCs, and CTLs that afford protection against the actions of granzyme B (11)(12)(13). When expressed in tumor cell lines or embryonic stem cells, PI-9 (14 -17) and SPI-6 (18 -20) confer resistance to perforin/granzyme B-mediated killing by CTL and NK cells. Recent studies have indicated that induction of SPI-6 expression in hepatocytes during hepatic viral infection (21) modulates sensitivity to cytotoxic lymphocyte effector mechanisms and the severity of liver injury during viral infection of the liver (22).
In contrast to the human genome, which appears to contain only a single serpinB9 gene, the mouse genome contains multiple serpinB9 homologs identified by sequence similarities (18). The target or cognate proteases associated with these additional serpinB9 homologs are not yet known. In contrast to stable patterns of SPI-6 expression in the spleen and various immune privileged sites, in liver, induction of SPI-6 expression from very low background levels has been observed only following viral infection (21) or following administration of IFN-␣ (22). The current studies were designed to further characterize the pattern of serpin expression in the liver during viral expression and initiate investigation of mechanisms responsible for hepatocyte expression of SPI-6 in vivo.

Materials and Methods
Mice C57BL/6J (B6) and perforin knockout (B6.pfp) mice were obtained from The Jackson Laboratory. Granzyme B knockout mice backcrossed on the B6 background for Ͼ20 generations (B6.gzmb Ϫ/Ϫ ) were generously provided by Dr. T. Ley (Washington University, St. Louis, MO) (23). DPPI (B6.dpp1 Ϫ/Ϫ ) mice were generously provided by Dr. C. Pham (Washington University) (10). Mice used in individual experiments were age and sex matched and used before 12 wk of age. No gender differences in levels or regulation of SPI-6 expression were noted in these experiments. All animal studies were conducted in compliance with accepted standards of humane animal care and were approved by the University of Texas Southwestern Institutional Animal Care and Use Committee.

Real-time PCR
Real-time RT-PCR primer sets were designed using PRIMER EXPRESS software (Applied Biosystems). Real-time PCR were performed in a final volume of 10 l containing cDNA from 20 ng of reverse-transcribed total RNA, 150 nM forward and reverse primers, and SYBR Green Universal PCR Master Mix (Applied Biosystems). PCR was conducted in 384 well plates using the Applied Biosystems Prism 7900HT Sequence Detection System (Applied Biosystems). All reactions were performed in triplicate. Melting curve analyses and serial dilution were performed to identify primer sets and conditions yielding specific products quantitatively over a wide range of input cDNA. Primers validated by this technique and used in the present studies were serpinb9, serpinb9b, serpinb9c, serpinb9e, serpinb9f, serpinb9g, cyclophilin, 18S RNA, KLRK1, and CD3 (Table I). Relative levels of mRNA were calculated by the comparative cycle threshold method (User Bulletin No. 2; Applied Biosystems). Cyclophilin mRNA and/or 18S RNA levels were used as the invariant control for all studies.

Western blotting
Cells were washed, suspended in lysis buffer (20 mM HEPES (pH 7.2), 10 mM KCl, 5 mM MgCl 2 ,1 mM EDTA, 250 mM sucrose, and proteinase inhibitors (Sigma-Aldrich)), lysed by repeated freeze-thawing, and centrifuged for 10 min at 10,000 ϫ g to remove debris, as previously described (21). Protein concentrations in tissue homogenates were assayed by the bicinchoninic acid method with reagents purchased from Pierce and using BSA as a standard. Equal amounts of total cytosolic protein (20 or 40 g/lane) were separated on 15% bis-acrylamide gels by SDS-PAGE and electrophoretically transferred to nitrocellulose in pH 9.9 carbonate buffer. Immunodetection was performed using primary anti-SPI-6 and secondary HRP-conjugated anti-rabbit Ig and the ECL Western Blotting Analysis System (Amersham Pharmacia Biotech).

IFN-␣ administration
Recombinant human IFN-␣2b (INTRON A) was obtained from Schering-Plough. Mice were injected i.p. IFN-␣ and selected tissues were harvested 24 -48 h after treatment as previously described (9). Previous work has indicated that in vivo administration of human IFN-␣ A/D (25) or IFN-␣2b (26) in mice leads to hepatic accumulation of NK cells while in vitro exposure to murine IFN-␣ leads to induction of SPI-6 expression in hepatocytes (21).

Isolation of intrahepatic lymphocytes and hepatocytes
The abdomen of anesthetized mice was entered using a sterile technique, the portal vein was cut, and the abdominal portion of the vena cava was perfused with 10-ml of Ca 2ϩ -and Mg 2ϩ -free phosphate buffer solution preheated to 37°C. The liver was removed and passed through a 40-mesh pore size screen, then a 300-mesh pore size screen (VWR Scientific) (27). The cell suspension was centrifuged at 400 ϫ g for 15 min and the cell pellet was resuspended in 40% Percoll and underlayered with an equal volume of 70% Percoll. The gradient was centrifuged at 900 ϫ g for 20 min and lymphocytes were isolated from the 40 -70% interface. Cells were washed with RPMI 1640 medium (Invitrogen Life Technologies). Primary murine hepatocytes were isolated as previously described (7,28).

DPPI and N-␣-benzyloxy-carbonyl-L-lysine thiobenzyl (BLT) esterase assays
DPPI activity was assayed by the hydrolysis of glycylphenylalanyl-␤napthylamide (100 M) as previously described (7,29,30). Substrate hydrolysis was monitored by the fluorescence of ␤-naphthylamine released per minute per milligram of protein. BLT esterase activity was assayed as previously described (7). The mixture was incubated at 37°C and the change in absorbance at 410 nm was measured and compared with that in samples containing no cell extract.

Serpinb9 (SPI-6) is the only homolog of serpinB9 (PI-9) that is up-regulated Ͼ10-fold during hepatic viral infection
SerpinB9 (PI-9) and the murine homolog serpinb9 (SPI-6) were found in previous studies to be up-regulated in hepatocytes in vitro by the antiviral cytokine IFN-␣ (21). To investigate the in vivo effect of IFN-␣ on hepatic expression of SPI-6 and other serpinB9 homologs, B6 mice were treated with varying doses of human rIFN-␣. At 24 h after administration, liver tissue was harvested and murine serpinB9 homolog mRNA levels were analyzed using real-time PCR. As illustrated in Fig. 1, top panel, SPI-6 mRNA expression was increased ϳ7-fold above background in recipients of 0.4 U/g IFN-␣ with higher levels of SPI-6 mRNA induction (30-to 40-fold) observed after administration of larger IFN-␣ doses. Of note, up-regulation of SPI-6 mRNA was observed following i.p. administration of lower doses of IFN-␣ than required for induction of hepatic mRNA levels of the well characterized IFN response gene PKR. Following administration of lower doses of IFN-␣, SPI-6 was the only serpinB9 homolog gene significantly up-regulated. Although modest induction of serpinb9b, serpin9c, and serpin9f mRNA expression was observed after administration of very high doses of IFN-␣, expression of these serpin genes was not increased to levels comparable to those of SPI-6. mRNA levels of two additional serpinB9 homologs, serpin9g and serpin9e, remained unaltered even in recipients of very high doses of IFN-␣ (data not shown). mRNA expression of constitutively expressed genes, such as cyclophilin and SPI-8, a serpin previously observed to be expressed in normal liver (21,32), also remained unaltered over a broad range of IFN-␣ dosing.
To assess serpinB9 homolog expression during viral infection, B6 mice were infected with 4 ϫ 10 10 OPU of AdCMV-LacZ via a single tail vein injection. As illustrated in Fig. 1, bottom panel, a nearly 10-fold increase in SPI-6 mRNA expression was detected at day 1, with further increases over time. Peak gene expression, nearly 100-fold baseline, was detected at day 8. In contrast, serpinb9b, serpin9c, and serpin9f mRNA levels in the liver were not increased Ͼ3-fold at any point during adenoviral infection while serpin9g and serpin9e mRNA levels remained unaltered during the course of viral infection (data not shown).

Expression and processing of granzyme B modulates SPI-6 expression
To assess whether SPI-6 expression correlates with an increase in hepatic lymphocyte granzyme B expression following IFN-␣ administration, a therapy previous noted to induce hepatic infiltration of NK cells (26), IFN-␣ was administered to wild-type and genetically modified B6 mice with selected defects in the perforin and granzyme B CTL effector pathway. In wild-type B6 ( Fig. 2A) and B6.dpp1 Ϫ/Ϫ mice (Fig. 2B), granzyme B mRNA increased and reached peak expression at 8 h after IFN-␣ administration. As illustrated in Fig. 2A, SPI-6 mRNA levels in livers from B6 mice increased after granzyme B mRNA levels had peaked and were increased 140-fold at 16 h after IFN-␣ administration. However, a Ͻ5-fold increase in SPI-6 expression was detected in livers of B6.dpp1 Ϫ/Ϫ mice at 16 or 24 h after IFN-␣ administration (Fig. 2B). Since B6.dpp1 Ϫ/Ϫ mice lack the DPPI activity required for posttranslational processing of granzyme B to an enzymatically active, mature enzyme, these results suggested a role for active granzyme B in the induction of SPI-6 expression in IFN-␣-treated mice.
To determine whether SPI-6 expression was indeed dependent on normal expression and function of granzyme B, IFN-␣ was administered to other strains of mice with different defects in perforin/granzyme B-mediated cytotoxicity. In addition to B6.dpp1 Ϫ/Ϫ mice that are defective in granzyme B processing, B6.pfp Ϫ/Ϫ mice that are defective in granzyme-and perforin-induced cytotoxicity, and B6.gzmb Ϫ/Ϫ mice that lack any granzyme B expression were examined for hepatic SPI-6 induction after IFN-␣ administration. As illustrated in Fig. 2C, a 40-fold increase in SPI-6 mRNA expression was observed in wild-type B6 mice at 24 h after IFN-␣ treatment while a Ͻ3-fold increase in SPI-6 mRNA expression in liver isolated from B6.dpp1 Ϫ/Ϫ , B6.gzmb Ϫ/Ϫ , or B6.pfp Ϫ/Ϫ mice was observed.
To determine whether hepatic SPI-6 protein also was up-regulated during adenoviral infection in concert with changes in mRNA levels, AdCMV-LacZ was administered via a single tail vein injection to B6 and B6.dpp1 Ϫ/Ϫ mice and SPI-6 expression in liver and spleen was assessed by immunoblotting. As illustrated in Fig. 3A, SPI-6 protein was not observed in uninfected livers of B6 or B6.dpp1 Ϫ/Ϫ mice. However, constitutive expression was readily detected in spleens of both control (Fig. 3A) and AdCMV-LacZ-infected (data not shown) wild-type and B6.dpp1 Ϫ/Ϫ mice. During viral infection, SPI-6 protein expression increased in the liver of B6 mice, becoming readily detectable at day 3 and remaining elevated for Ͼ20 days following viral infection (Fig. 3B). In contrast, no readily apparent SPI-6 protein expression was detectable by immunoblot analysis in livers of B6.dpp1 Ϫ/Ϫ mice at days 3, 10, 15, or 25 after infection (Fig. 3C). Of note, similar ␤-galactosidase transgene expression levels were detected in the livers of each strain of mice in the experiments detailed in Figs. 2 and 3, confirming equivalent levels of viral infection (data not shown). Moreover, similar levels of DPPI and BLT esterase activity (Fig. 4) in day 8 intrahepatic lymphocyte populations indicated equivalent levels of intrahepatic CTL/NK cell activation in B6.gzmb Ϫ/Ϫ and B6 mice during viral infection in these studies. Of note, the very low levels of BLT esterase activity in B6.dpp1 Ϫ/Ϫ intrahepatic lymphocytes results from a lack of FIGURE 2. Analysis of SPI-6 mRNA expression in wild-type and granzyme B-deficient transgenic mice. RNA was isolated from the livers of experimental mice at specific time points after IFN-␣ administration and granzyme B and SPI-6 mRNA expression in B6 (A) or B6.dppi Ϫ/Ϫ mice (B) was assessed by real-time PCR and normalized to endogenous 18S RNA. C, SPI-6 mRNA expression in liver of B6, B6.dpp1 Ϫ/Ϫ , B6.gzmb Ϫ/Ϫ , and/or B6.pfp Ϫ/Ϫ mice was analyzed at specific time points after IFN-␣ administration or during the viral infection in the liver of wild-type and granzyme B-transgenic mice in which RNA was isolated from the livers of experimental mice at specific time points during AdCMV-LacZ infection and SPI-6 mRNA expression in liver of B6, B6.dpp1 Ϫ/Ϫ , or B6.gzmb Ϫ/Ϫ mice (n ϭ 6 -9 mice/time point). Each bar represents the mean Ϯ SD of triplicate determinations of mRNA expression relative to control and normalized to 18S RNA. DPPI-mediated processing of pro-granzyme A to an enzymatically active form (29,30).

Up-regulation of SPI-6 mRNA correlates with increases in NK cell and CTL-specific gene expression in liver during viral infection
To determine whether increased SPI-6 expression correlated with increased infiltration of granzyme B-expressing NK cells and/or CTL, liver tissue was harvested from B6 mice at select time points during viral infection and levels of NK and CTL receptor mRNA expression were analyzed. As shown in Fig. 5, mRNA encoding KLRK1, an activation receptor present on the surface of NK cells, was increased ϳ65-fold by day 3. A Ͼ100-fold increase in mRNA levels of CD3 at day 8 during viral infection was detected (Fig. 5).
These results suggest that initial up-regulation of SPI-6 expression in liver during AdCMV-LacZ infection immediately follows infiltration by NK cells and persists in parallel with the more prolonged presence of infiltrating T cells. To investigate whether the induction of hepatic SPI-6 expression following IFN-␣ treatment (Fig. 1) also correlates with the NK cell influx observed after i.p. IFN-␣ administration (26), mice were treated with an NK celldepleting Ab, anti-asialo GM1, or an isotype control Ab 48 h before administration of IFN-␣. In intrahepatic lymphocytes from control Ig-treated B6 mice, IFN-␣ administration resulted in an   8-fold increase in SPI-6 mRNA levels (Fig. 6). This increase in SPI-6 expression occurred in association with a 4-to 5-fold increase in the numbers of NK1.1 ϩ CD3 Ϫ NK cells in livers of IFN-␣-treated mice (data not shown). SPI-6 mRNA expression in hepatocytes isolated from control Ig-treated B6 mice was increased Ͼ50-fold at 24 h after IFN-␣ administration (Fig. 6B). In contrast, hepatocytes and intrahepatic lymphocytes isolated from NK celldepleted mice did not show a significant increase in SPI-6 mRNA expression after IFN-␣ administration. These results suggest that the influx of granzyme B-expressing activated NK cells after IFN-␣ administration or during viral infection leads to an increase in cytoprotective SPI-6 expression by hepatocytes to protect against perforin/granzyme B-mediated killing.

Reconstitution of DPPI Ϫ/Ϫ mice with wild-type bone marrow results in SPI-6 expression after IFN-␣ stimulation
To determine whether the lack of SPI-6 expression in B6.dpp1 Ϫ/Ϫ mice was due to the lack of enzymatically active granzyme B expression in intrahepatic lymphocytes or a defect in SPI-6 expression by DPPI-deficient hepatic parenchymal cells, B6.dpp1 Ϫ/Ϫ mice were irradiated and reconstituted with bone marrow from wild-type mice and maintained for 6 wk to allow reconstitution of the immune system. DPPI assays of spleen samples confirmed DPPI expression in B6.dpp1 Ϫ/Ϫ mice transplanted with wild-type bone marrow, indicating establishment of chimerism (data not shown). As shown in Fig. 7A, B6 or B6.dpp1 Ϫ/Ϫ mice reconstituted with wild-type bone marrow exhibit SPI-6 protein expression in the liver after treatment with IFN-␣. Similar results were obtained with B6.gzmb Ϫ/Ϫ mice reconstituted with wild-type bone marrow (data not shown). In contrast, wild-type B6 mice reconstituted with bone marrow derived from B6.dpp1 Ϫ/Ϫ mice did not exhibit an increase in SPI-6 expression after IFN-␣ stimulation (Fig. 7B, lanes 4 and 5). These results indicate that the lack of hepatic SPI-6 expression in B6.dpp1 Ϫ/Ϫ after IFN-␣ treatment or adenoviral infection is not related to any intrinsic defect in hepatic SPI-6 expression, but rather relates to lack of granzyme B activity in bone marrow-derived lymphocytes infiltrating the liver in response to these immune stimuli.

Discussion
These studies illustrate that increases in hepatocyte SPI-6 expression during viral infection correlate chronologically with the presence of activated, bone marrow-derived, granzyme B-expressing NK and CTL in the liver. Furthermore, mice lacking intact perforin or granzyme B do not exhibit similar levels of increased hepatic SPI-6 mRNA or protein expression after IFN-␣ administration or during AdCMV-LacZ infection. Thus, these findings indicate that the presence and likely release of DPPI-processed, enzymatically active granzyme B and perforin by bone marrow-derived cells that infiltrate the liver after IFN-␣ administration or during AdCMV-LacZ infection leads to the induction of SPI-6 expression in hepatocytes.
The mouse genome contains multiple serpinB9 homologs (18,33). Putative mouse PI-9 homologs share Ͼ90% sequence homology with human PI-9. However, these homologs differ greatly from PI-9 in that they do not possess negatively charged residues in the P1 region of their reactive center loop (13,18,33,34). Other than sequence comparisons, the mouse homologs of PI-9 have not been well characterized previously. The current studies evaluated hepatic expression of SPI-6 and of additional PI-9 homologs, such as serpinb9b/R86, serpin9c/NK9, serpin9d/NK21, serpin9e/NK21b, and serpin9g/NK26, posttreatment with antiviral cytokines and during viral infection. The current findings indicate that SPI-6 is the only PI-9 homolog whose expression is up-regulated 10-to 100-fold in the liver after IFN-␣ administration or during the course of AdCMV-LacZ infection. Furthermore, the results of the present studies illustrate that SPI-6 expression correlates with granzyme B expression in that up-regulation in SPI-6 expression directly follows infiltration of the liver by granzyme B-expressing cytotoxic lymphocytes, but not by cytotoxic lymphocytes deficient in granzyme B expression, granzyme B processing, or perforin, an effector molecule that facilitates granzyme B entry into target cell cytosol.
Previous work has illustrated that the liver does not constitutively express PI-9 or SPI-6, which may indicate that these genes are tightly regulated and bound to a transcriptional repressor during normal conditions (13,18). The need for both enzymatically active granzyme B and perforin expression by intrahepatic lymphocytes during induction of high levels of SPI-6 expression in virally infected hepatocytes suggests a unique mechanism of SPI-6 gene regulation not previously noted in other cell lineages expressing SPI-6. Once inside the cytosol, granzyme B cleaves substrates that can induce apoptosis. Up-regulation of SPI-6 may occur as a direct result of granzyme B entry into hepatocytes or may be due to enzymatic cleavage and/or up-regulation of additional factors present in the cytosol that bind and induce SPI-6 gene transcription. Similar proteinase-mediated regulation of hepatocyte gene transcription has been noted to be mediated by caspase 3 (35), a known substrate of granzyme B (4). However, it is of interest that in the present studies, up-regulation of the granzyme B inhibitor SPI-6 occurred both during adenoviral infection, when caspasedependent NK and CTL killing of virally infected hepatocytes is anticipated, and after IFN-␣ was administered in doses known to induce hepatic infiltration of highly activated NK cells but not known to induce perforin and granzyme B-mediated hepatocellular injury (25). It remains to be determined whether granzyme B-and perforin-dependent, IFN-␣-induced SPI-6 expression also is related to mild levels of caspase-dependent hepatocellular injury induced by IFN-␣-activated NK cells. Previous studies from our laboratory indicate that an important role of serpinb9/SPI-6 expression in hepatocytes is to protect these cells from killing by granzyme B-expressing NK cells and CTL (22). Moreover, regulated expression of this specific inhibitor of granzyme B (13,18,33) in the liver accounts, at least in part, for the previously observed resistance of virally infected hepatocytes to perforin-and granzyme-dependent effector mechanisms (7) and, in turn, the greater importance of Fas and related death receptordependent cytotoxic effector mechanisms in immune-mediated injury and clearance of viral infections from the liver (8,9). Indeed, inhibition of SPI-6 expression during hepatic viral infection leads to vigorous cytopathic immune responses resulting in severalfold higher serum alanine aminotransferase levels than observed in control mice and to the development of rapid onset, fatal acute liver failure in mice with deficient SPI-6 expression (22). The present findings indicate that unique feedback mechanisms allow the liver to rapidly up-regulate SPI-6 expression after exposure to granzyme B and perforin. This counterregulatory mechanism is likely, at least in part, responsible for the observation that vigorous cytopathic immune responses to hepatic viral infection only rarely lead to acute liver failure (36).