Clinical trial registration number ISRCTN 86571875.
Potential conflict of interest: Nothing to report.
Liver failure is the major cause of death in alcoholic steatohepatitis (ASH). In experimental hepatitis, granulocyte-colony stimulating factor (G-CSF) mobilizes hematopoietic stem cells, induces liver regeneration, and improves survival. We studied the short-term effects of G-CSF on CD34+ stem cell mobilization, liver cell proliferation, and liver function in patients with ASH. Twenty-four patients (mean age 54 years) with alcoholic cirrhosis [Child-Turcotte-Pugh score 10 (7–12)] and concomitant biopsy-proven ASH [Maddrey score 36 (21-60)] were randomized to standard care associated with 5 days of G-CSF (10 μg/kg/day, group A, n = 13) or standard care alone (group B, n = 11). Serial measurement of CD34+ cells, liver tests, cytokines [hepatocyte growth factor (HGF); tumor necrosis factor α; tumor necrosis factor-R1; interleukin-6; alfa-fetoprotein], and 13C-aminopyrine breath tests were performed. Proliferating hepatic progenitor cells [HPC; double immunostaining (Ki67/cytokeratin 7)], histology, and neutrophils were assessed on baseline and day 7 biopsies. Abstinent alcoholic patients with cirrhosis served as controls for immunohistochemistry. G-CSF was well tolerated. At day 7, both CD34+ cells (+747% versus −6%, P < 0.003), and HGF (+212% versus −7%, P < 0.03) increased in group A but not in group B. Cytokines and aminopyrine breath test changes were similar between groups. On repeat biopsy, a >50% increase in proliferating HPC was more frequent in group A than in group B (11 versus 2, P < 0.003). Changes in Ki67+/cytokeratin 7+ cells correlated with changes in CD34+ cells (r = 0.65, P < 0.03). Neutrophils and histological changes were similar in both groups. Conclusion: G-CSF mobilizes CD34+ cells, increases HGF, and induces HPC to proliferate within 7 days of administration. Larger trials would be required to determine whether these changes translate into improved liver function. (HEPATOLOGY 2008.)
Alcoholic steatohepatitis (ASH) carries a poor short-term prognosis, with liver failure resulting from both acute liver injury and from chronic liver disease.1
Elevated proinflammatory cytokines during ASH have prognostic significance.2 Both anti-inflammatory and liver regeneration mechanisms participate in the repair of ASH-dependent liver damage. Although corticosteroids decrease proinflammatory cytokines and bring a clear survival benefit in severe, biopsy-proven ASH,3 regeneration is also crucial to restore liver function. Hepatocellular injury is believed to trigger proliferation of the remaining mature hepatocytes. Ductular proliferation is thought to be a source of progenitor cells (at the interface between portal tracts and the liver lobule) that can give rise to hepatocyte and biliary cells.4 However, regeneration processes are impaired in cirrhosis,5 and the influence on liver cell proliferation of alcohol itself6 or the presence of steatohepatitis7 remains unclear.
In animal models, after liver injury, bone marrow–derived circulating pluripotent cells have been reported to participate in liver repopulation with both nonparenchymal cells and hepatocytes. This repopulation process, however, seems to be highly dependent on the type of liver injury and experimental conditions (see Thorgeirsson and Grisham8 for review). Fusion with hematopoietic cells has been substantiated as a mechanism by which hepatocytes can regenerate, and studies have demonstrated improved histology and survival after liver injury following mobilization of bone marrow cells by granulocyte-colony stimulating factor (G-CSF).9, 10
In humans with chronic liver diseases, no therapy has proven useful to promote liver regeneration. However, G-CSF showed promising results in ischemic heart diseases11 and in a small uncontrolled pilot study of patients with cirrhosis.12 G-CSF may induce musculoskeletal pain. Allergic-type reactions (less than 1 in 4000 patients) have been reported, as well as spontaneous spleen rupture due to extramedullary hematopoiesis,13 a clinically relevant element in portal hypertension and splenomegaly.
We undertook this investigator-initiated, randomized, phase II controlled trial to study the mobilizing effects of G-CSF on hematopoietic stem cells and the short-term changes on hepatic regeneration in patients with biopsy-proven ASH associated with cirrhosis and a variable degree of liver failure. The primary endpoint was the ability of G-CSF to increase circulating CD34+ cells, a surrogate marker for hematopoietic stem cell mobilization.14 Secondary endpoints included filgrastim safety in liver failure, its effects on liver cell regeneration using biological markers and immunohistochemistry, and its possible influence on liver function.
From September 2005 to August 2006, all patients with ASH were considered eligible. The inclusion criteria were as follows: recent heavy alcohol intake (>80 g/day), biopsy-proven ASH, Maddrey score ≥20 and ≤70, leukocytes ≤15 G/L, age 18 to 70 years, and ability to give informed written consent. Exclusion criteria were as follows: platelets <20 G/L; international normalized ratio >1.9; known hypersensitivity to filgrastim; creatinine >150 μmol/L; infection or hemorrhage within the last 10 days; documented hepatocellular carcinoma; hepatitis B, C, or human immunodeficiency virus seropositivity; and pregnancy.
Liver biopsies of a group of 10 patients with alcoholic cirrhosis abstinent from alcohol and awaiting liver transplantation served as controls for immunohistochemistry. Data from 10 healthy peripheral blood stem cell (PBSC) donor candidates undergoing G-CSF mobilization were used to compare values obtained in our patients with cirrhosis and ASH.
The sample size was calculated as follows: after standard stem cell mobilization with G-CSF in normal controls, CD34+ cells are expected to range between 0.5% and 1%. Our initial assumption was that circulating CD34+ cells would be 0.20% ± 0.10% in non-G-CSF–treated patients with ASH (secondary to stress) versus 0.80% ± 0.40% in G-CSF treated patients. In patients with ASH, the sample size required to verify the null hypothesis, with a power of 95% and an alpha error of 5%, would be 10 patients in each group. To allow for an expected 20% loss at follow-up in this population, we decided to enroll a minimum of 24 patients.
A randomization code was generated in blocks of four. Randomization was done with sequentially numbered envelopes to either standard medical therapy with or without G-CSF. The design of the study is illustrated in Fig. 1.
In our hospital, when ASH is suspected on clinical grounds, both liver biopsy and the histological diagnosis are available within days of admission to inform the decision for steroid treatment.
As soon as ASH was histologically confirmed and informed consent was obtained, blood sampling was performed in all patients. Standard medical therapy included a 28-day course of prednisone, 40 mg daily in case of severe ASH.3 In addition, patients randomized to group A (G-CSF+) received filgrastim (for dose determination, see below).
At day 7 after randomization, liver biopsy and blood sampling were repeated. The day-28 follow-up visit included blood sampling. Group A (G-CSF+) patients had a daily physical examination and were asked to report any side effects during the treatment. Every clinically significant event was recorded.
Hematopoietic stem cell mobilization was performed using G-CSF in the same manner as stem cell mobilization in bone marrow transplantation.14 Group A (G-CSF+) patients received filgrastim, a human G-CSF (Neupogen; Amgen SA, Zug, Switzerland), at a dose of 500,000 UI/kg subcutaneously two times daily (=10 μg/kg/day) for 5 days starting within 24 hours of diagnosis of ASH. This dose of G-CSF was similar to that given in a recent uncontrolled pilot study.12
Circulating Neutrophils and Hematopoietic Stem Cells.
We measured neutrophils (days 0, 7, and 28) and CD34+ cells (days 0 and 7) in peripheral venous blood (5 mL in ethylene diamine tetraacetic acid). The leukocyte count was determined on a Sysmex XE-2100 automated counter (Sysmex-Digitana, Horgen, Switzerland). Samples were diluted in phosphate-buffered saline (PBS) and 1–2 × 105 cells were used for labeling with anti-CD34-phycoerythrin (PE; BD Biosciences, Allschwil, Switzerland) and anti-CD45-fluorescein isothiocyanate (FITC; Beckman Coulter, Roissy, France). A control tube was also prepared with mouse isotype antibodies to immunoglobulin G1-FITC and immunoglobulin G1-PE (Beckman Coulter). After red-cell lysis and washes, cells were diluted in 100 μL PBS before evaluation by flow cytometry (FACScalibur; BD Biosciences, San Jose, CA). Based on side-scatter versus forward-scatter analysis, the leukocytes (neutrophils, monocytes, and lymphocytes) were gated on 50,000 events. The PBSC compartment (CD45dimCD34+) was identified by sequential analysis of side scatter versus CD45-FITC and CD34-PE dot-plot distributions. The PBSCs constituted the group of cells with low side scatter, low CD45 expression, and CD34 positivity. Results were given as number of CD34+ cells/μL.
Liver Tissue Studies.
The baseline liver biopsy was performed early after admission (median time 3 days, range 1–7) and prior to steroid therapy. Both baseline and day-7 repeat liver biopsies were performed using the transjugular route. Histological and immunohistochemical studies were undertaken in patients at baseline and at day 7. Ten patients with alcoholic cirrhosis durably abstinent from alcohol, without ASH, and awaiting liver transplantation, served as controls for baseline studies.
Biopsy material was fixed in 10% buffered formaldehyde, paraffin-embedded, and processed for light microscopy (5-μm sections) with hematoxylin and eosin, Masson's trichrome, and reticulin stainings. Histopathological examinations were performed by a single pathologist (L.R.-B.) expert in liver diseases, while blinded to the patient characteristics and treatment allocation. We evaluated the liver's architecture and fibrosis, and the features for the histological diagnosis of ASH, as accepted.15, 16 We also performed a semiquantitative analysis of histological features of alcoholic liver disease, including steatosis, inflammation, cell death, ballooning degeneration, and fibrosis, using an appropriate score.17 Using this system, values may vary between no lesion (score of 0) to severe damage (score of 17). Assessment of lobular and portal inflammation was performed by a manual count of leukocytes on hematoxylin and eosin in the whole tissue specimen, using ×400 magnification, and expressed as the number of cells per high-power field (HPF). Polynuclear neutrophils were identified using chloroacetate esterase staining, as described.16
Using double immunostaining methods on serial biopsy sections, we determined both the total number of hepatic cells entered into proliferation and the proportion of these cells specifically involved in regeneration processes. Thus, we identified cells with the morphology and typical immunoreactivity of mature hepatocytes, of hepatic progenitor cells (HPCs) appearing either as isolated cells (smaller than hepatocytes, with an ovoid nucleus and a scant basophilic cytoplasm) in the lobule or within the ductular reaction, and of cells intermediate between hepatocyte and progenitors, termed intermediate hepatocyte-like cells.4 Both intermediate hepatocyte-like cells and HPCs constitute the ductular reaction, the human counterpart to oval cell activation in rodents, and demonstrate cytokeratin 7 (CK7) immunoreactivity,4 whereas mature hepatocytes stain positively with CK18.18
Immunohistochemical staining was performed in serial 3-μm-thick to 5-μm-thick sections prepared from each tissue block, and mounted as described.19 The following antibodies were used (DakoCytomation, Zug, Switzerland): mouse monoclonal human antibody MiB1 against the proliferation marker Ki67 (1:20 dilution), CK7 (1:50 dilution), and CK18 (1:20 dilution). Microwave or pressure cooker pretreatment was performed. Sections were incubated for 1 hour at room temperature with the diluted primary antibodies, which were then revealed by ENVISION (DakoCytomation). Peroxidase activity was revealed with 30% 3,3′ diaminobenzidine as chromogen in PBS containing 0.015% H2O2. Sections were weakly counterstained with Mayer's hematoxylin and mounted in Eukitt (Kindler GmbH, Freiburg, Germany). After a first incubation with MiB1 diluted antibody, sections were secondly incubated with the CK7 or CK18 antibody, and then revealed by streptavidin alkaline phosphatase and fast red chromogen.
Proliferative activity was scored by manual count of cells staining positively with the Ki67 antigen in the whole biopsy specimen under high magnification (×400). Then, in the double staining method, we determined the number of Ki67+/CK7+ cells with the morphology of HPCs or intermediate hepatocyte-like cells. Finally, we determined the number of Ki67+/CK18+ cells with the characteristics of mature hepatocytes. Values are expressed as the median number of cells per HPF.
After centrifugation, we kept samples frozen at −70°C until measurement of circulating levels of tumor necrosis factor (TNF)α, the soluble form of TNFα receptor-1 (sTNF-R1), interleukin-6 (IL-6), hepatocyte growth factor (HGF), and alfa-fetoprotein (AFP) using commercially available specific immunoassays (R&D Systems Europe, Abingdon, UK). Serum levels of TNFα, sTNF-R1, and IL-6 are typically elevated in ASH,2 whereas AFP and HGF concentrations increase following alcoholic liver injury.20 The detection limits for TNFα, sTNF-R1, IL-6, AFP, and HGF were 4 pg/mL, 3 pg/mL, 1.5 pg/mL, 0.75 ng/mL, and 670 pg/mL, respectively. For all assays, the intraassay and interassay variabilities were below 6% and 5%, respectively.
At baseline and at the day 7 and day 28 visits, we collected clinical variables and laboratory values to calculate the Child-Turcotte-Pugh and the Model for End-Stage Liver Disease (MELD) scores as indicators of liver failure and prognosis. We also performed an aminopyrine breath test to evaluate the hepatic microsomal capacity (supplementary material).
The study protocol was approved by our institution's Ethics Committee (accepted on August 30, 2005) and was in accordance with International Committee for Harmonization/Good Clinical Practice (ICH/GCP). All patients gave informed written consent to participate.
Tabulated data are expressed as median and range unless otherwise noted. Values in the figures are represented as box-plots (median and interquartile range, twenty-fifth and seventy-fifth percentiles). Statistical comparisons were performed using the intention-to-treat principles. We used the Mann-Whitney, two-sided Fisher exact test, Wilcoxon signed rank test, and analysis of variance with a Bonferroni correction for multiple comparisons, as appropriate. Correlations between variables were calculated using the nonparametric Spearman rank test. The statistical level of significance was set at P < 0.05. All calculations were made using the StatView 5.0 program (SAS Institute Inc., Cary, NC).
Figure 1 illustrates the study design and flow diagram of patient selection21 over 12 months (September 2005 to August 2006). During this time, 54 alcoholic patients presented with ASH on liver biopsy. Thirty patients were not included for the following reasons: Maddrey score <20 or >70 (n = 7); associated viral hepatitis (n = 4); recent gastrointestinal bleeding or infection (n = 7); renal failure (n = 2); or refusal to participate (n = 10). Thirteen patients were randomly assigned to standard of care and G-CSF treatment (group A, G-CSF+) and 11 to standard of care alone (group B, G-CSF−; Table 1). Standard therapy included steroids in five and seven patients of the G-CSF+ and G-CSF− groups, respectively, based on the Maddrey score indicating disease severity and expected survival benefit.3
Table 1. Patient Characteristics at Baseline
Group A (G-CSF+)
Group B (G-CSF−)
Age (year, range)
Received steroids (n)
Hepatic venous pressure gradient measured during transjugular liver biopsy demonstrated clinically significant portal hypertension in all patients.
G-CSF Tolerance and Clinical Follow-Up
All except one patient (who presented a variceal bleed after 3 days of treatment) received the full 5-day course of G-CSF. Drug administration was well tolerated. We detected no spleen tenderness at clinical examination. Three patients complained of transient mild lower back pain reversible upon cessation of G-CSF. One G-CSF–treated patient died at day 27 of variceal hemorrhage. All clinical events are listed in Table 2.
Table 2. Clinical Events
Group A (G-CSF+)
Group B (G-CSF−)
GI indicates gastrointestinal.
Neutrophils and CD34-Positive Cells
At baseline, circulating neutrophils [6.1 (1.97 to 8.6) versus 7.1 (2.5 to 8.7) G/L, P = 0.6], and CD34+ cells [1604 (0 to 3606) versus 2199 (913 to 5131) cells/mL, P = 0.5] were comparable in the G-CSF+ and in the G-CSF− groups, respectively. The evolution in the number of cells at day 7 is depicted in Figs. 2 and 3. Percent changes in CD34+ cells were more important in the G-CSF+ group than in the G-CSF− group (+379 versus −40%, P < 0.003, respectively; Fig. 4). More patients in the G-CSF+ group than in the G-CSF− group achieved a >50% increase in CD34+ cells (11 versus 1, P < 0.0001). However, mobilization of CD34+ cells was lower than in the 10 healthy stem cell donors (Fig. 3).
Liver Tissue Studies
All but one patient had biopsy repeated at day 7. The median size of the biopsies was 13 mm (range, 9–20 mm).
Liver architecture showed cirrhosis in all patients. Groups were comparable at baseline. Table 3 summarizes the values of the day 0 and day 7 liver biopsies.
Table 3. Liver Tissue Studies
Day 7 liver tissue studies were performed in 12 of 13 G-CSF-treated patients.
P < 0.01 versus patients with ASH at baseline.
P < 0.01 versus Ki67+/CK7+ cells from abstinent cirrhotic controls.
In abstinent control patients with cirrhosis, the number of CK7+/Ki67− cells, not in the proliferation cycle, was higher than CK7+/Ki67+ cells (Table 3), as reported.19
At baseline, the number of Ki67+ cells was higher in patients with ASH as compared to abstinent patients with alcoholic cirrhosis [6.3 (2.5 to 17) versus 2.1 (2 to 2.4) cells per HPF, P < 0.02]. Groups were similar at baseline, with a median value of 4.7 (1.6 to 13.5) in the G-CSF+ group and 4 (2.4 to 18.7) cells per HPF in the G-CSF− group (P = 0.7). On the repeat liver biopsy at day 7, there was a mean 80% increase in the Ki67+ cells in the G-CSF+ group as compared to a −35% decrease in the G-CSF− group (P < 0.007).
Ki67/CK7 and Ki67/CK18 Double Immunostaining.
Figure 5 illustrates typical findings in both groups. At baseline, both Ki67+/CK7+ and Ki67+/CK18+ cells were more numerous in patients with ASH compared to abstinent alcoholic patients with cirrhosis (P < 0.01; Table 3). Groups were similar regarding baseline values of Ki67+/CK7+ (P = 0.14) and Ki67+/CK18+ (P = 0.32) cells.
Table 3 provides immunohistochemical data of the day 0 and day 7 liver biopsies. The evolution of the numbers of double-stained hepatocyte lineage cells, expressed as percent changes, is illustrated in Fig. 6. We defined a >50% increase in the number of Ki67+/CK7+ cells as indicative of proliferation of HPCs. This cutoff was met in 11 out of 13 patients in the G-CSF+ group compared to two out of 11 patients in the G-CSF− group (P < 0.003). Regarding mature hepatocytes, however, the threshold of a >50% increase in Ki67+/CK18+ cells was reached in six out of 13 group A patient (G-CSF+) compared to one out of 11 in group B patients (G-CSF−; P = 0.077).
At baseline, as expected, circulating TNFα, sTNF-R1, and IL-6 were elevated.1, 2 Figure 7 illustrates the evolution of these cytokines at day 7 and day 28. As reported,20 baseline HGF concentrations were elevated as compared to values reported in healthy subjects. The evolution of HGF, expressed as absolute values, was statistically similar between groups (Fig. 8), but percent changes at day 7 were more important in the G-CSF+ group as compared to the G-CSF− group (+60% versus −21%, P < 0.03).
The increase in serum AFP at day 7 and at day 28 was significant in both groups (Fig. 8).
Evolution of Liver Function
The evolution of liver function, depicted in Fig. 9, showed a parallel improvement in the Child-Turcotte-Pugh score and a decrease in the MELD score that reached statistical significance only in the G-CSF− group. Values of 13C-aminopyrine breath tests measured at day 7 and at day 28 remained stable in both groups.
We observed a positive correlation at baseline between Ki67+/CK7+ cells with the MELD score (r = 0.41, P < 0.05), but not with the Maddrey score (r = 0.26).
Changes in circulating CD34+ and Ki67+/CK7+ cells in liver tissue between baseline and day-7 values were positively correlated (r = 0.65, P < 0.03).
We demonstrate that a 5-day course of G-CSF induces proliferation of hepatocyte lineage cells in the biopsy performed after 7 days in patients with ASH, cirrhosis, and liver failure. This effect was associated with an increased level of circulating CD34+ stem cells and more important changes in serum HGF compared with standard of care–treated patients. In contrast, inflammatory cytokine concentration remained stable over a 28-day period. Despite these effects on liver cell proliferation, we failed to demonstrate any improvement in liver function.
Regeneration of the diseased organ is of paramount importance in decompensated liver disease.22 In the presence of cirrhosis, hepatocellular injury with hepatocyte loss stimulates the remaining mature hepatocytes to proliferate. However, in cirrhosis, regeneration of the resident hepatocyte population is often limited.5 This situation stimulates the proliferation of intrahepatic progenitor cells, mostly present within the ductular reaction, to generate cells of the hepatocyte lineage and improve residual liver function of the cirrhotic liver. Accordingly, we observed a correlation between the number of proliferating HPCs and the severity of liver disease as expressed by the MELD score. Both groups were comparable at inclusion with respect to the severity of liver injury (for example, the stimulus for regeneration), as assessed by the Maddrey score, and other characteristics such as age and gender, both known to influence liver regeneration.23 Thus, the marked increase in the number of proliferating HPCs found on repeat biopsy 7 days after initiation of treatment represents a proof of concept that G-CSF causes histologically-visible changes in the liver. Although we observed a concomitant increase in circulating CD34+ stem cells in G-CSF–treated patients, confirming previous results,12 interpreting the precise role of these bone marrow–derived cells in the processes of endogenous organ repair remains delicate. Published experimental data report conflicting results, with either a significant,9 marginal,10 or no effect24 in liver regeneration. In view of the early changes in serum HGF, more prominent in G-CSF–treated patients, a facilitating and supportive role of circulating growth factors on HPC proliferation could be proposed.8 In a recent study, the role of G-CSF in acute liver injury was evaluated in a rat partial hepatectomy model using a hepatocyte proliferation inhibitor.25 In this model, both intrahepatic HPC and bone marrow–derived engrafted cells were observed, supporting the hypothesis of a double effect of G-CSF.
Despite an increased proliferation of HPC, we failed to observe an improved liver function. Moreover, one G-CSF–treated patient died of massive variceal bleeding. However, our study was not powered to detect clinical improvements. Other clinical adverse events during follow-up, although slightly more frequent in the G-CSF group, are not rare in patients with cirrhosis and poor hepatic reserve,1 and do not suggest an a priori a role for G-CSF.
Preliminary results of bone marrow–derived cells in improving liver function in patients with cirrhosis are encouraging,26 but result from an uncontrolled study. In our trial, we speculate that factors such as advanced age and cirrhosis may have blunted the regenerative effects of G-CSF on liver function. Moreover, CD34+ cell mobilization was blunted as compared to healthy stem cell donors, a possible consequence of impaired bone marrow function together with cellular sequestration in an enlarged spleen.12 Although steroids, given to a subgroup of our patients with ASH, do not seem to impair hepatocyte proliferation,27 the effects of alcohol on liver regeneration is still a matter of debate.6 As animals exposed to long-term alcohol become less responsive to cytokines and growth factors that regulate hepatic regeneration,28 the toxic effect of alcohol on mature hepatocytes provides a strong stimulus for the replication of progenitor cells to repopulate the damaged organ.6
In conclusion, we demonstrated that a 5-day course of G-CSF safely stimulates the HPCs to proliferate in patients with cirrhosis, superimposed ASH, and moderate to severe liver failure. These signs of liver regeneration, evident as early as 7 days after initiation of treatment, together with the increased number of circulating stem cells and the more important changes in serum HGF as compared to standard of care–treated patients, should encourage larger clinical trials to determine how these changes may translate into improved liver function over a longer observation period.
We thank Corinne Charrin and Colette Grand for their expert work in flow cytometry measurements, Carole Verdan for immunohistochemistry studies, Rafael Quadri and Martha Jordan for performing the breath tests, and Tom McKee for critical reading of the manuscript.