Cytokine-Responsive Gene-2/IFN-Inducible Protein-10 Expression in Multiple Models of Liver and Bile Duct Injury Suggests a Role in Tissue Regeneration¹

The chemokine family consists of ∼40 distinct secreted proteins that share 20 to 70% sequence identity. Their known activities include regulation of immune cell trafficking and function. Within the chemokine family, four distinct subfamilies have been defined based upon the presence and positioning of conserved cysteine residues. The two largest and best-characterized subgroups are the CXC and CC chemokines. Members of the CXC subfamily, also termed α-chemokines, contain a variable residue between the first two invariant cysteines; these cysteine residues are adjacent in the CC or β-chemokine subfamily. The CXC chemokines may be further subdivided based upon the presence or absence of a glutamic acid, leucine, arginine (ELR) motif near the N terminus. CXC chemokines with the ELR motif have potent neutrophil chemotactic activity and bind to the CXCR1 and CXCR2 receptors, while non-ELR CXC chemokines do not bind to these receptors and have greater chemotactic activity for NK cells, lymphocytes, and monocytes (1, 2). Studies are beginning to define novel activities for various chemokines beyond those involved in inflammatory cell trafficking. For example a CXC chemokine, stroma-derived factor-1, and its receptor, CXCR4, have been demonstrated to be essential for normal heart, cerebellum, B cell, and vascular development (3, 4, 5), while other members of the CXC chemokine family have been demonstrated to possess angiogenic and angiostatic activity, with some having associated endothelial growth-regulatory activity (2, 5, 6, 7).

IFN-inducible protein-10 (IP-10)⁴ is a non-ELR CXC chemokine. It is a chemoattractant for NK cells and T cells (2, 8), and is believed to be an important regulator of the Th1, IL-12-driven, inflammatory response as an inducer of cellular infiltration, including perhaps induction of IFN-γ (9). A single physiological receptor for IP-10, termed CXCR3, has been identified (10, 11). CXCR3 is a shared receptor for monokine induced by IFN-γ (Mig) (10, 12) and IFN-inducible T cell α-chemoattractant (13), the two chemokines most highly related to IP-10. CXCR3 is found on T, NK, and B cells (10, 14). Recently, IP-10 has been demonstrated to bind a second receptor, the Kaposi’s sarcoma-associated herpesvirus G protein-coupled receptor, although on this receptor IP-10 functions as an inverse agonist (15).

In addition to T cell and NK cell chemotaxis and activation, IP-10 has antiproliferative effects on endothelial cells (6), as well as angiostatic and antitumor activity (6, 7, 16, 17). IP-10 also inhibits fibroblast motility (18). The murine ortholog of IP-10 is cytokine-responsive gene-2 (CRG-2), and the two proteins are 68% identical (19). Elevated Crg-2/IP-10 expression has been noted in numerous infectious and autoimmune diseases (2, 20, 21, 22). In the liver, IP-10 is expressed in autoimmune liver disease (23), chronic hepatitis (24), and biliary atresia (25). The degree of IP-10 expression has been shown to correlate with the severity of inflammation in hepatitis, being most highly expressed in chronic active hepatitis (24), and in those patients with biliary atresia and secondary hepatic injury (25). IP-10 has been detected in most of these models late in the disease process or during active inflammatory cell recruitment. In this study, we describe that Crg-2 was expressed as an immediate early gene following multiple types of acute hepatic injury and resection, including models that involve hepatocyte, oval cell, and cholangiocyte regeneration. Furthermore, Crg-2 was expressed just before each known wave of hepatocyte proliferation following hepatectomy. These data suggest that IP-10 is a general marker of hepatic inflammation and injury, and furthermore that IP-10 may play a fundamental role in hepatic repair and regeneration.

Unless otherwise specified, all reagents were purchased from Sigma (St. Louis, MO). Recombinant human IP-10, IL-6, and TNF were purchased from PeproTech (Rocky Hill, NJ).

Freeze-clamped normal human liver was obtained at time of donor hepatectomy, from fatty liver obtained after hepatectomy and perfusion for transplantation, or from diseased liver obtained at time of transplant. All liver samples were obtained in accordance with institutional guidelines (Medical University of South Carolina). Frozen tissue was homogenized into RNAzol B (Tel-Test, Friendswood, TX), according to the manufacturer’s instructions.

The animal care and use committee of the Johns Hopkins University School of Medicine approved all animal studies. Both surgical and toxic injury models used 5- to 7-wk-old male mice. CD-1 mice were purchased from Charles River Breeding Laboratories (Wilmington, MA). B6;129S wild-type and B6;129S Tnfrsf1a^{tm11 mx} Tnfrsf1b^{tm11 mx} (TNFR superfamily member 1a null and 1b null) mice were purchased from The Jackson Laboratory (stock number 003243; Bar Harbor, ME).

All surgical procedures were performed between the hours of 10 a.m. and 2 p.m. to reduce diurnal variation. All surgical models were performed through a midline abdominal incision under methoxyflurane (Metaphane; Mallinckrodt Veterinary, Mundelein, IL) general anesthesia. Controls for all surgical experiments underwent general anesthesia, midline abdominal incision, gentle hepatic or renal manipulation, and closure. Partial hepatectomy was performed with ligature of the middle and left hepatic pedicles, as described (26). Bile duct ligation was performed with 4-0 silk ligation of the extrapancreatic common bile duct (27). Five-sixths nephrectomy was performed with hilar ligation and excision of the right kidney and resection of the upper and lower left kidney poles. Following the two-thirds left nephrectomy, 30-s finger pressure was applied to the remaining kidney parenchyma to tamponade bleeding.

For toxic injury models, mice were challenged by i.p. injection of the toxic agent in solubilizing carrier, as described below; control groups received injection of toxin-free carrier. CCl₄ was administered by i.p. injection of 20 μl in 0.1 ml vegetable oil (Wesson, Fullerton, CA). d-Galactosamine (GalN) was injected i.p. in 0.1 ml saline at 0.7 g GalN per kg body weight. Methylene dianiline (DAPM) was administered by i.p. injection of 50 mg in 0.1 ml of a 50% ethanol/saline solution. LPS serotype 055:B5 (Sigma) was injected i.p. at a dose of 100 μg in 0.1 ml saline.

Implantable osmotic pumps (Alzet, Palo Alto, CA) were placed using standard sterile surgical technique under methoxyflurane anesthesia. A midline abdominal incision was performed, and the preloaded pump containing either 200 μg human IP-10 in saline or saline alone was placed in the peritoneal cavity. The abdominal incision was closed with a two-layer running closure to prevent subsequent fluid leakage. Following pump placement, daily i.p. injections of 1 ml cell-labeling reagent (bromodeoxyuridine (BrdU) and fluorodeoxyuridine; Amersham Pharmacia, Piscataway, NJ) were given at 10 a.m. on postoperative days 1 through 6. Animals were killed on postoperative day 7. BrdU-labeled nuclei were detected on frozen sections using the FLUOS in situ cell proliferation kit (Roche Molecular Biochemicals, Indianapolis, IN), according to the manufacturer’s instructions, counterstained with 4′,6′-diamidino-2-phenylindole (Molecular Probes, Eugene, OR), and mounted in ProLong antifade mounting medium (Molecular Probes).

Cyclohexamide pretreatment of mice was performed as described (28). Mice were injected with 50 mg/kg cycloheximide in 1 ml saline i.p. 1 h before two-thirds hepatectomy or CCl₄ injection. Animals were sacrificed, and RNA was isolated using RNAzol B.

Following 18 h of GalN treatment, C57BL/6 mice were placed under methoxyflurane anesthesia, and hepatic cell compartments were separated, as previously described (29). Briefly, a midline abdominal incision was performed, then the portal vein was identified and cannulated with a 24-gauge angiocatheter. Following return of portal blood, the liver was perfused for 5 min with liver perfusion medium, followed by 10 min of liver digest medium (Life Technologies, Gaithersburg, MD). After excision, the cell suspension was passed twice through a 70-μm filter (Falcon Cell Strainer, Franklin Lakes, NJ) and placed on ice. Separation into parenchymal and nonparenchymal fractions was performed using multiple low-speed centrifugations.

For primary culture, hepatocytes were isolated from normal CD-1 male mice by the two-step collagenase perfusion method (30). Hepatocytes were plated on collagen-coated dishes in William’s E medium with 10% FBS, HEPES, penicillin/streptomycin, gentamicin, glutamine, 10 mM glucose, 10 nM insulin, and 1 nM dexamethasone, and maintained in the same medium without serum. Human IP-10 was added to the maintenance medium at concentrations ranging from 1 ng/ml to 1μg/ml, in the presence or absence of 20 ng/ml epidermal growth factor (Collaborative Biomedical Products, Bedford, MA). Medium was changed every 24 h. For measurement of DNA synthesis, hepatocytes were pulse labeled for 24 h with [³H]thymidine at 0-, 24-, 48-, 72-, and 96-h postseeding. Cells were lysed, then fixed with 10% TCA, and the amount of [³H]thymidine incorporation into glass filter-bound DNA was measured by liquid scintillation counting.

AML-12 cells (American Type Culture Collection (ATCC) CRL-2254, Manassas, VA) were grown to near confluence. CCl₄ was added directly to each plate to a final concentration of 20 or 40 mM. Control flasks were treated identically, but received no addition of CCl₄. Subsequently, all flasks were sealed and incubated at 37°C and 30 rpm on an orbital platform. RNA was harvested using RNAzol B.

MRC-5 cell lines (ATCC CCL-171) were grown in 10% FBS in DMEM until nearly confluent. Cells were washed with PBS and stimulated with human IP-10 for 48 h. Cells were harvested and RNA was isolated as described for the AML-12 cells. In a separate experiment, cells were treated with IP-10 in 0.1% FBS in Eagle’s medium for 24 h, and supernatants were collected and frozen. Hepatocyte growth factor (HGF) content was determined using the Quantikine human HGF immunoassay (R&D Systems, Minneapolis, MN).

CRG-2-secreting lines were generated as previously described (17). Control cell lines were transfected with the empty vector and similarly selected. Cell lines were grown to near confluence in MEM (Life Technologies) with 10% dialyzed FBS (Life Technologies), proline, glutamine, penicillin, streptomycin, and 1.6 μM methotrexate (amithopterin). Before injection, cells were washed in PBS and trypsinized. Following two PBS washes, cells were resuspended at 5 × 10⁶ cells in 0.2 ml PBS and injected i.m. into the thighs of female athymic nude mice (Harlan Sprague Dawley, Indianapolis, IN).

RNA was isolated by homogenization in guanidinium thiocyanate or RNAzol B, as described (31). Where specified, once selected poly(A)⁺ RNA was generated using oligo(dT) cellulose, as described (31), or by using an Oligotex mRNA isolation kit (Qiagen, Valencia, CA). Total RNA (20 μg/lane) or poly(A)⁺ RNA (2 μg/lane) was subjected to Northern blotting analysis on GeneScreen Plus (NEN Life Science Products, Boston, MA), as described (31).

The right liver lobes were collected 3 h following treatment and snap frozen in liquid nitrogen. Liver samples were weighed, and equal amounts were homogenized in PBS. Protein concentration was determined by Coomassie Plus-200 Protein Assay Kit (Pierce, Rockford, IL). Equal amounts of protein from each sample were prepared in sample buffer and run on reducing SDS-PAGE. Samples were transferred to nitrocellulose (Schleicher & Schuell, Relliehausen, Germany) and blotted using a rabbit polyclonal anti-mouse CRG-2 antiserum (5171) diluted 1/20,000 in TBS with 0.03% Tween 20 and 0.5% nonfat dried milk (Bio-Rad, Richmond, CA). Bands were detected using HRP-conjugated goat anti-rabbit IgG (Pierce) and the SuperSignal West Pico detection kit (Pierce), according to instructions provided by the manufacturer.

Northern blots were prepared with RNA isolated from freeze-clamped normal human liver obtained at time of donor hepatectomy, from fatty liver obtained after hepatectomy and perfusion for transplantation, or from diseased liver removed from the recipient at time of transplant. As shown in Fig. 1, markedly increased expression of both IP-10 and the related chemokine gene, MIG, was observed in several patients with end-stage liver disease.

Given the observed induction of IP-10 in livers of patients with end-stage liver disease, we set out to examine whether murine parenchymal cells could express Crg-2 directly following injury. The differentiated, nontransformed hepatocyte cell line, AML-12 (32), was treated with the potent hepatotoxin CCl₄ in vitro. A dose-responsive induction of Crg-2 was seen following administration of 20 and 40 μM CCl₄, with peak expression at 3 h (Fig. 2). Of note, no such induction was seen for the Mig gene (data not shown). These data show that Crg-2 may be induced as a direct consequence of cellular injury, occurring in the absence of inflammatory or other mediating cells.

Given the induction of IP-10 in end-stage cirrhosis, and Crg-2 in AML-12 cells following CCl₄ exposure, we examined expression of Crg-2 in mouse models of liver and bile duct injury that preferentially injure and stimulate distinct liver cell populations. Examination of Crg-2 expression following these models of liver injury also demonstrated dramatic induction. Crg-2 was induced following bile duct ligation and administration of DAPM (Fig. 3, A and B), two interventions that preferentially injure bile ducts and induce subsequent biliary epithelial cell proliferation (27, 33). Administration of the RNA synthesis inhibitor, GalN, which is hepatotoxic and requires the liver to recruit oval cells for subsequent liver regeneration (34, 35), also potently induced Crg-2 (Fig. 3 C).

Induction of Crg-2 in the AML-12 cell line suggested that Crg-2 could be induced in hepatocytes in response to injury. We next evaluated the expression of Crg-2 in hepatic cell compartments following exposure to GalN. Parenchymal and nonparenchymal cell fractions were obtained from mouse livers isolated 18 h after GalN injection when Crg-2 expression peaked. By Northern analysis, Crg-2 expression was noted in both parenchymal and nonparenchymal fractions. Excellent separation of the parenchymal and nonparenchymal populations was achieved, as evaluated by Northern analysis of the cell type-specific markers albumin (parenchymal), and Hgf (nonparenchymal), respectively (Fig. 4).

Given the dramatic induction of Crg-2 following various forms of hepatic injury, we expanded our studies to evaluate Crg-2 in liver regeneration using two-thirds hepatectomy. In contrast to the models examined above, partial hepatectomy produces only minimal injury to the remnant liver around the time of surgery, but results in a 70% loss of total hepatic mass, thereby eliciting a vigorous regenerative response (26). This compensatory hyperplasia results in a return to normal liver mass after two coordinated cycles of hepatic cell division within days of surgery (26). We performed Northern blot analysis of liver mRNA isolated from CD-1 mice subjected to two-thirds hepatectomy. Crg-2 was dramatically induced in a bimodal fashion following two-thirds hepatectomy in the mouse (Fig. 5). The observed increased level of Crg-2 expression began within 1 h of hepatic resection, peaked at 3 h, and returned to baseline by 12 h. A second peak of expression was noted at 4 days following hepatectomy. Therefore, expression of Crg-2 is associated with both the early injury and the later repair and precedes each of the two peaks of DNA synthesis known to occur in the mouse during liver regeneration (36). In contrast, Mig was induced late in the regenerative response after hepatectomy, coinciding with only the second observed peak of Crg-2 expression (Fig. 5).

Given the dramatic induction of Crg-2 mRNA following various forms of hepatic injury, we sought to determine whether CRG-2 protein levels were also elevated. Indeed, Western blot analysis using a polyclonal antiserum generated against rCRG-2 revealed increased CRG-2 protein in extracts from injured livers taken 3 h after injury as compared with normal livers (Fig. 6). Injection of LPS, a known stimulant of Crg-2 production, was used as a positive control.

We considered whether the expression of Crg-2 immediately after partial hepatectomy might be part of a systemic response to injury. Examination of nonhepatic tissues following two-thirds hepatectomy demonstrated increased expression of Crg-2 in kidney, small bowel, thymus, and spleen 1 h after hepatectomy (Fig. 7), but not at 4 days during the second peak of Crg-2 expression in the liver. Therefore, Crg-2 appears to be expressed in the liver and other organs within 1 h of two-thirds hepatectomy in a time course suggestive of a growth-regulatory or tissue-repair factor.

Given the observation that two-thirds hepatectomy induced renal Crg-2 production (Fig. 7), Crg-2 expression following five-sixths nephrectomy was also examined. As shown in Fig. 8, dramatic up-regulation of Crg-2 was observed in this model as well.

The time course of Crg-2 induction after tissue or cell injury suggested that Crg-2 is part of the acute response to tissue damage or insufficiency. To investigate this possibility further, CD-1 mice were pretreated with the protein synthesis inhibitor cycloheximide and challenged with either two-thirds hepatectomy or CCl₄ injection. In both cases, expression of Crg-2 was noted to be dramatically up-regulated in animals pretreated with cycloheximide, characteristic of an immediate early gene (Fig. 9, A and B). As has been previously described (28), c-myc superinduction was also observed following cycloheximide pretreatment (Fig. 9 A).

TNF signaling is essential for normal liver regeneration following partial hepatectomy or CCL₄ treatment (37, 38, 39). To examine the potential induction of Crg-2 by TNF, mice were injected with TNF and liver mRNA was isolated. Dramatic up-regulation of liver Crg-2 was observed following i.p. injection of murine rTNF (Fig. 10,A). IL-6, another known regulator of liver regeneration, did not induce Crg-2 (Fig. 10,A). To determine whether TNF signaling is required for Crg-2 expression following liver resection or injury, TNFR superfamily members Ia and Ib doubly null C57BL/6;129SVJ hybrid mice were subjected to two-thirds hepatectomy, and total liver mRNA was evaluated by Northern blot analysis. Crg-2 induction in double knockouts and wild-type controls following CCl₄ toxic hepatic injury was similar (Fig. 10 B), suggesting that although exogenous TNF induced an elevation in Crg-2, the pathway for Crg-2 induction following liver injury could be TNF independent.

The expression of Crg-2 before the peaks of liver DNA synthesis following hepatectomy led us to examine whether IP-10 may be functioning as a direct regulator of hepatic growth. We examined the ability of IP-10 to stimulate hepatocyte DNA synthesis and proliferation in primary murine hepatocyte cell cultures and in vivo. No increase in [³H]thymidine incorporation was observed in murine hepatocytes treated for 24–96 h with 1 ng/ml to 1 μg/ml IP-10 (data not shown). As well, no increase in liver mass or hepatocyte nuclear BrdU labeling was observed in mice treated for 1 wk with an i.p. pump releasing IP-10 at a rate of 1 μg/h (data not shown). Furthermore, nude mice bearing Chinese hamster ovary cell tumors secreting CRG-2 experienced only a slight increase in liver mass when compared with control mice bearing Chinese hamster ovary cell tumors not expressing CRG-2 (data not shown).

We next evaluated the possible role of IP-10 in regulating the expression of the potent hepatic mitogen, HGF. Interestingly, expression of Hgf, like that of Crg-2, occurs in a bimodal fashion following partial hepatectomy, with each peak of expression preceding a known peak of hepatocyte DNA synthesis (40, 41, 42), with the described expression of Hgf following that observed for Crg-2. Moreover, an Hgf-inducing activity has been isolated from serum fractions obtained from rats with injured and regenerating liver (42, 43). The human lung fibroblast cell line, MRC-5, can be used to assay for this activity. Therefore, we used this cell line to evaluate the ability of IP-10 to induce HGF protein expression by ELISA and Hgf mRNA by Northern blotting. MRC-5 cells demonstrated secretion of HGF in response to IP-10 in a dose-dependent fashion. The EC₅₀ of IP-10 for induction of HGF secretion was 300 ng/ml (Fig. 11,B), but an increase of Hgf mRNA was noted in response to 10 and 100 ng/ml IP-10 (Fig. 11 A). These data suggest a potential role for IP-10 in HGF production before the regenerative response.

IP-10 and its murine analogue, Crg-2, are expressed in a wide variety of cell types in response to administration of LPS (19, 44), IFN-γ (45), IL-1β (2), and TNF (45). Systemic injection of LPS, IFN-γ, or TNF has been noted to induce Crg-2 expression potently in the liver (45, 46). In human liver disease, increased expression of IP-10 has been demonstrated in chronic hepatitis (24) and biliary atresia (25). In both disorders, the extent of IP-10 expression correlates with the degree of inflammation. Rat studies have demonstrated increased expression of rat IP-10 in severe, chronic alcohol-induced hepatitis and have correlated degree of expression with necrosis (47). Elevated Crg-2 expression has also been noted in the lung during the inflammatory response following hepatic ischemia-reperfusion injury (48). Our results demonstrating increased IP-10 expression in diseased human livers are consistent with these observations and further demonstrate concomitantly elevated expression of Mig, a related chemokine.

Given the wide expression pattern of IP-10 and Crg-2 in chronic liver diseases, we investigated the expression of Crg-2 following hepatic injury. The hepatocyte cell line, AML-12, was injured with the known hepatotoxin, CCl₄. A dose-responsive induction of Crg-2 was observed, suggesting that initial Crg-2 expression precedes the inflammatory response following liver injury, and further that Crg-2 expression may occur as a direct parenchymal response to cellular injury in the absence of inflammatory cells. To define Crg-2 expression following liver injury in vivo, we next examined the time course of Crg-2 expression in multiple models of liver injury. Three distinct models of liver injury were examined: bile duct ligation, DAPM injury, and GalN toxic injury. In contrast to CCl₄, which induces hepatocellular injury and a hepatocyte proliferative response in vivo, bile duct ligation and DAPM injury induce extensive biliary tree injury and a resultant bile duct proliferative response (27, 33). Expression of Crg-2 was increased markedly and rapidly following the injury in both models, again in a fashion suggesting that Crg-2 induction preceded the inflammatory response. The third model examined exposure to the hepatotoxic RNA synthesis inhibitor GalN, which induces irreversible injury and widespread hepatocellular death and requires liver progenitor oval cells to undergo a subsequent proliferative response to reestablish hepatocellular mass (35). In this model also, increased expression of Crg-2 was observed. Interestingly, expression of Crg-2 is delayed in this model, consistent with the known onset of hepatotoxicity after GalN injury (34).

In situ hybridization studies have reported conflicting results for hepatic Crg-2 expression following injection of LPS, TNF, or IFN-γ, with both parenchymal and nonparenchymal sources having been suggested (24, 46). Given the novel observation that the transformed hepatocyte line, AML-12, expressed Crg-2 in vitro following exposure to CCl₄, we examined Crg-2 expression in an in vivo injury model as well. The GalN injury model was selected because it causes parenchymal injury as well as a nonparenchymal proliferative response along with a prolonged elevation of Crg-2 expression. Northern blot analysis of mRNA isolated from these fractions revealed a signal for Crg-2 in both parenchymal and nonparenchymal cells. Taken together, these data suggest that both parenchymal and nonparenchymal liver cells can induce Crg-2 expression as a direct response to liver injury. This is reminiscent of Crg-2 expression occurring as a direct response to viral infection through an NF-κB pathway (20). Of note, NF-κB activation, presumably through TNF signaling, is a well-described phenomenon occurring within 30 min of liver injury, and thus may be involved in Crg-2/IP-10 expression (36).

In the models of liver injury in which Crg-2 was induced, death of hepatocytes or biliary epithelial cells is followed by a potent proliferative response. Therefore, we wished to investigate the possibility that Crg-2 might be functioning not only in response to acute and chronic liver and biliary injury, but as a signal of regeneration as well. Partial (two-thirds) hepatectomy is a model that involves surgical removal of the left lateral and medial liver lobes, with minimal, early hepatocellular injury to the remaining right liver lobes that subsequently undergo a regenerative response (26) (for review, see Refs. 36 and 49, 50, 51). In this model, rapid, highly reproducible induction of hepatocyte cell division occurs with an initial wave of DNA synthesis peaking at 40 h. In the mouse, a second peak of DNA expression occurs at ∼4 days. Liver mass, through compensatory hyperplasia, is reestablished within 1 wk of resection. Interestingly, we found that Crg-2 RNA and protein, as in the previously examined models of liver injury, were induced within 1 h of two-thirds hepatectomy. Examination of other organs following two-thirds hepatectomy also demonstrated increased expression of Crg-2 in small bowel, kidney, thymus, and spleen, demonstrating that the Crg-2 expression is initially a systemic response to hepatic injury. A second peak of Crg-2 expression occurs at 4 days following two-thirds hepatectomy, a time that corresponds with the second wave of hepatic mitosis during regeneration. At this time, Mig was also expressed. Incidentally, these results suggest that no IFN-γ is released immediately post liver resection because no Mig expression was observed at that time. However, it does suggest IFN-γ is present at 4 days following two-thirds hepatectomy because Mig induction is known to be highly IFN-γ specific.

The rapid induction of Crg-2 in the liver following systemic TNF administration is similar to that observed following two-thirds hepatectomy, suggesting that TNF, which is detectable in liver venous blood within 1.5 h of hepatic injury or resection (52), may be inducing its production. To examine whether Crg-2 may be induced as a TNF-dependent signal, TNFR superfamily member Ia and Ib doubly null mice were compared with colony controls following CCl₄-induced toxic liver injury. Expression of Crg-2 appeared uninhibited in the livers of doubly null mice, demonstrating that Crg-2 expression, although TNF inducible, can occur independently of TNF. One may speculate that the known LPS release from the gut after hepatectomy (53) could be a distinct, redundant inducer of Crg-2 after liver resection.

Because Crg-2/IP-10 was induced in a pattern suggesting a role in the regenerative response to hepatectomy, we investigated a possible direct effect of CRG-2/IP-10 on liver growth. No direct mitogenic effects upon hepatocytes could be demonstrated with either IP-10 or CRG-2 in vivo or in vitro. However, CRG-2/IP-10 could be having an indirect effect on regeneration through induction of liver growth factors. In this regard, Matsumoto and coworkers (42) have reported a partially purified activity that they termed injurin, a serum protein fraction with potent HGF-inducing activity produced following two-thirds hepatectomy or CCl₄ injury in the rat. Interestingly, CRG-2 has many similarities to injurin. The time course for Crg-2 expression is similar to that reported for injurin and for HGF, even demonstrating a bimodal expression pattern (40, 41). Elevated injurin activity has been observed at ∼2 days after hepatectomy in the rat, a time point that corresponds to the second peak of DNA synthesis that occurs at 4 days in the mouse. Increased renal Crg-2 expression was also observed acutely following five-sixths nephrectomy, another injurin-inducing procedure. Moreover, the partially purified activity noted by Matsumoto (42) was a low molecular mass, stable protein.

To examine the possibility that CRG-2 may possess injurin activity, the HGF-secreting cell line, MRC-5, was treated with IP-10. Elevations of Hgf mRNA and increased HGF secretion by MRC-5 cells were noted following stimulation with IP-10. Unfortunately, neither injection of IP-10 into normal animals, nor growth of CRG-2-secreting tumors in nude mice elicited liver growth. Nor could we reproducibly demonstrate increased Hgf mRNA in vivo following injection of IP-10. Our inability to correlate the increase in HGF secretion in vitro with induction of liver growth in vivo may be due either to insufficient HGF stimulation by exogenously supplied CRG-2/IP-10 or to the requirement for additional, priming signals that are present following liver injury (see Ref. 50).

Given its known activities, another indirect role for CRG-2/IP-10 in liver regeneration might be in recruitment of T cells and NK cells. Injection of IP-10 is known to induce a local, predominantly lymphocytic cellular infiltrate (54). T cells, particularly after activation, as well as NK cells, are known to respond to IP-10 (55, 56, 57). Thus, the early expression of Crg-2 may function as a chemotactic signal to bring T cells and NK cells into the liver. Interestingly, Tamaru and coworkers (58) have recently reported that Crg-2 is required for liver infiltration of T cells in the Propionibacterium acnes/LPS-induced model of hepatitis. Moreover, it is well known that a T cell infiltrate occurs during normal hepatic regeneration. The importance of this T cell role has been underscored by the observation that athymic, T cell-deficient nude mice have a markedly delayed hepatic regenerative response and an increased mortality rate following hepatic resection or injury (59, 60). Similarly, C/EBP-β-deficient mice, which possess an abnormal Thl response, also demonstrate abnormal liver regeneration, although evidence for abnormal hepatocyte function in these animals also exists (61). Finally, NK cells, another CRG-2/IP-10 target, have recently been implicated in immune surveillance during liver regeneration (62). If CRG-2/IP-10 supports liver regeneration through the recruitment of T or NK cells, then local production of CRG-2/IP-10 may be critical, explaining our failure to demonstrate an effect with protein administered systemically.

Taken together, these results imply that CRG-2/IP-10 may be present in a wider spectrum of conditions than have previously been recognized. As well, these data suggest a potential role for CRG-2/IP-10 in mediating not only the hepatic inflammatory response, but also liver regeneration following hepatocyte loss. CRG-2/IP-10 may well be a cytokine that links tissue injury and inflammation to repair and regeneration.

We thank Dr. Se-Jin Lee for permitting this work to be done in his laboratory and for his advice on the project.