Neutrophil dysfunction in alcoholic hepatitis superimposed on cirrhosis is reversible and predicts the outcome

LIVER FAILURE/CIRRHOSIS/PORTAL HYPERTENSION

Neutrophil Dysfunction in Alcoholic HepatitisSuperimposed on Cirrhosis is Reversible and Predicts

the OutcomeRajeshwar P. Mookerjee,* Vanessa Stadlbauer,* Sukhwinderjit Lidder, Gavin A.K. Wright, Stephen J. Hodges,

Nathan A. Davies, and Rajiv Jalan

Mortality in patients with alcoholic hepatitis (AH) remains high, and although corticoste-roids are widely used for treatment, the results vary considerably. In AH, neutrophils areprimed and infiltrate the liver to produce injury, but paradoxically, the main cause of deathin such patients is infection. Our prospective study addressed this paradox of primed neu-trophils on the one hand and increased risk of infection on the other. We hypothesized thatthe full activation of neutrophils by a humoral factor such as endotoxin renders them unableto respond to further bacterial challenge. We analyzed neutrophil oxidative burst and phago-cytosis in whole blood by fluorescence-activated cell sorting analysis in 63 alcoholic patientswith cirrhosis and patients with cirrhosis with superimposed AH (cirrhosis�AH). In 16patients, ex vivo studies determined whether the removal of endotoxin restored neutrophilfunction. A resting burst greater than or equal to 5%, indicating neutrophil activation and areduced phagocytic capacity lower than 42%, was associated with significantly greater risk ofinfection, organ failure, and mortality. This defective neutrophil function was transmissiblethrough patients’ plasma to normal neutrophils, and patients’ neutrophil function could berestored by normal plasma. The ex vivo removal of endotoxin from patients’ plasma de-creased the resting burst and increased the phagocytic function. Conclusions: Our studyprovides the rationale for a goal-directed approach to the management of patients withcirrhosis and AH, in which the assessment of neutrophil function may be an importantbiomarker to select patients for immunosuppressive therapy. The neutrophil dysfunction incirrhosis and AH is reversible, with endotoxin-removal strategies providing new targets forintervention. (HEPATOLOGY 2007;46:831-840.)

In patients with severe alcoholic hepatitis (AH), infec-tion complicates the course of illness and is associatedwith significant morbidity and mortality.1-4 Neutro-

phils are an essential component of the innate immune

response and key players in the pathogenesis of AH.5 Dataon neutrophil function in AH are paradoxical, with somestudies suggesting neutrophil priming, indicating a readi-ness to respond to bacterial challenge (for a review, seeCondliffe et al.6), as measured by hydrogen peroxide over-production, decreased L-selectin expression,7 and highlevels of neutrophil elastase.8 In contrast, other studiesshow decreased neutrophil phagocytic capacity correlat-ing with disease severity.9 Endotoxin is known to be ele-vated in patients with AH10 and has the ability to bothprime6,11 and activate neutrophils.12,13

The clarification of this controversy has importantclinical implications. Current strategies for treating AHinclude the administration of immunosuppressive agentssuch as corticosteroids and, more recently, anti–tumornecrosis factor (TNF)� strategies, which may further po-tentiate susceptibility to infection. The results of studiesusing these drugs in patients with AH are conflicting,14,15

and although there are some data showing benefits from

Abbreviations: AH, alcoholic hepatitis; AUROC, area under the receiver opera-tor curve; CRP, c-reactive protein; CV, coefficient of variation; FACS, fluorescence-activated cell sorting; fMLP, N-formylmethionyl-leucyl-phenylalanine; GMFI,geometric mean of fluorescence intensity; IL, interleukin; LPS, lipopolysaccharide;MDA, malondialdehyde; MELD, model for end-stage liver disease; TNF, tumornecrosis factor.

From the Liver Failure Group, Institute of Hepatology, Division of Medicine,University College London, London, UK.

Received January 15, 2007; accepted March 20, 2007.Supported by the Austrian Science Foundation through an Erwin-Schrodinger

fellowship (J2547) to V.S.*These authors contributed equally to this work.Address reprint requests to: Rajiv Jalan, Institute of Hepatology, 69-75 Chenies

Mews, London WC1E 6HX, UK. E-mail: [email protected] © 2007 by the American Association for the Study of Liver Diseases.Published online in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/hep.21737Potential conflict of interest: Nothing to report.

831

anti-TNF� strategies,16,17 one trial showed increased riskof infection and mortality, which resulted in early trialtermination.18

To reconcile the apparent paradox, we assessedwhether neutrophils exist in a primed or fully activatedstate in patients with different alcoholic liver diseases [al-coholic cirrhosis and AH superimposed on cirrhosis(cirrhosis�AH)] and studied the association of this withthe development of infection and clinical outcome. Wehypothesized that full activation is associated with im-paired neutrophil responses to ongoing bacterial chal-lenge rendering patients more susceptible to infection.Alternatively, if neutrophils are primed but not activated,immunosuppressive therapy may be advantageous. Thus,our prospective study systematically examined neutrophiloxidative burst and phagocytosis in patients withcirrhosis�AH to evaluate whether altered neutrophilfunction was associated with infection, organ failure, andsurvival. In ex vivo studies, we investigated whether thedefect in neutrophil function was due to a humoral factorand whether endotoxin removal from plasma would re-store a patient’s neutrophil function.

Patients and MethodsPatient Selection

All patients gave written informed consent, and thestudy was approved by the local ethics committee. Pa-tients admitted with evidence of alcoholic cirrhosis19 wereenrolled into the study at the time of a liver biopsy, whichwas performed to evaluate the presence or absence of AH.The included patients had been admitted with acute de-compensation of alcoholic cirrhosis manifested by in-creasing jaundice, with no clinical or microbiologicalevidence (chest radiographs or routine cultures of urine,blood, sputum, and ascites) of infection. Prophylactic an-tibiotics (cefotaxime) were prescribed following initialcultures if there was a suspicion of infection, and theywere stopped if subsequent cultures proved negative. Pa-tients were excluded if they were less than 18 or greaterthan 75 years and had evidence of organ failure (inotroperequirement, creatinine � 150 �mol/L, hepatic enceph-alopathy � grade 2, requirement for mechanical ventila-tion), hyponatremia, and hepatic/extrahepatic malignancy,less than 3 days post gastrointestinal bleeding or if theyreceived any immunomodulatory therapy prior to entryin the study. Blood from 20 age-matched and sex-matched healthy volunteers with no history of liver dis-ease was studied to serve as controls for the comparison ofneutrophil function.

Patients were classified histologically into those with cir-rhosis and superimposed inflammatory AH (cirrhosis�AH)with a histological grading system similar to that of nonalco-

holic steatohepatitis,20 taking into account steatosis, neutro-phil infiltration, the ballooning of hepatocytes, and theformation of Mallory’s hyaline, and compared with cirrhosisalone.

Study DesignFollowing the correction of any electrolyte distur-

bances or hypovolemia, peripheral venous blood wasaseptically collected into pyrogen-free tubes (BD Vacu-tainer lithium-heparin, 60 U per tube, BD, Plymouth,UK) and used for routine biochemistry, neutrophil func-tion, cytokine profile, and detection of markers of oxida-tive stress. For experiments with cells, blood was kept atroom temperature (maximum 1 hour); for harvestingplasma, blood was placed immediately on ice. After cen-trifugation, plasma was aliquoted under pyrogen-freeconditions into nonpyrogenic cryotubes (Corning Inc.,Corning, NY) and stored at �80°C until further analysis.Whole blood or isolated neutrophils mixed with plasmawere used to perform the Phagoburst or Phagotest (Orpe-gen Pharma GmbH, Heidelberg, Germany). For all ex-periments, we took strict precautions to avoid endotoxincontamination by working aseptically and using endotox-in-free equipment. Bilirubin, albumin, liver functiontests, coagulation parameters, full blood count, and c-reactive protein (CRP) were routinely performed. Mad-drey’s discriminant function,21 the model for end-stageliver disease (MELD),22 and the Pugh score23 were calcu-lated. The patients were followed prospectively over aperiod of 90 days. The occurrence of organ dysfunctionand mortality was recorded.

NeutrophilsNeutrophils were investigated either in a whole blood

assay (as described below) or after isolation by a 1-stepgradient centrifugation.

Neutrophil Isolation. Whole blood (4 mL) was lay-ered over 5 mL of Polymorphprep (Axis-Shield, Oslo,Norway) and spun for 30 minutes at 400g at room tem-perature. Neutrophils were harvested from the second in-terface and washed with phosphate-buffered saline(Sigma Aldrich, St. Louis, MO). Neutrophils werecounted in a Thoma hemocytometer and resuspended inphosphate-buffered saline at a density of 5 � 105 in 50�L. Fifty microliters of the cell suspension and 50 �L ofplasma were used per assay. The viability was tested byTrypan Blue exclusion and was over 98%.

Neutrophil Activation (Oxidative Burst). ThePhagoburst kit (Orpegen Pharma, Heidelberg, Germany)was used to determine the percentage of neutrophils thatproduced reactive oxidants with or without stimulation ac-cording to the manufacturer’s instructions. In brief, 100 �L

832 MOOKERJEE, STADLBAUER, ET AL. HEPATOLOGY, September 2007

of heparinized whole blood or 50 �L of isolated neutrophilsand 50 �L of plasma (as indicated) were incubated for 20minutes with 20 �L of opsonized Escherichia coli (2 � 107

bacteria), N-formylmethionyl-leucyl-phenylalanine (fMLP;5�M), and phorbol-12-myristate-13-acetate (8.1 �M) orwithout a stimulus at 37°C. The formation of reactive oxi-dants was monitored by the oxidation of dihydrorhodamine123 to rhodamine, which produced green fluorescence. Toidentify neutrophils, cells were stained with anti–CD16-PEantibody (Immuntools, Friesoythe, Germany) and analyzedby fluorescence-activated cell sorting (FACS; Becton Dick-inson FACScan, San Jose, CA) with Cellquest software.Neutrophils were gated on forward and side scatter charac-teristics (Fig. 1A), and subsequently, the percentage ofCD16-positive cells producing reactive oxygen metabolites(green fluorescence) was calculated (Fig. 1C,E,G). The sam-ples were analyzed in duplicate or triplicate. The interassaycoefficient of variation (CV) for the resting burst was 5.4%,and for the stimulated burst, it was 4.2%. The intra-assayCV for the resting burst was 4.7%, and it was 2.4% for thestimulated burst.

Neutrophil Phagocytic Capacity. The Phagotest(Orpegen Pharma) was used to measure phagocytosiswith fluorescein isothiocyanate–labeled opsonized E. colibacteria. Whole blood (100 �L) or 50 �L of isolatedneutrophils and 50 �L of plasma (as indicated) were in-cubated with 20 �L of bacteria (2 � 107) at 37°C for 20minutes, whereas a negative control sample remained onice. Neutrophils were identified and gated as describedpreviously, and subsequently, the geometric mean of flu-orescence intensity (GMFI), corresponding to the num-ber of bacteria engulfed by a single cell, was analyzed (Fig.1B,D,F,H). To avoid misinterpretation of the results dueto batch-to-batch variability of the bacteria, the resultswere normalized to the mean of at least 3 healthy controlsamples for each new batch of bacteria used. The sampleswere analyzed in duplicate or triplicate. The interassayCV was 10.1%, and the intra-assay CV was 1.6%.

Incubation with EndotoxinEndotoxin (E. coli 0111:B4 lot 085K4068, Sigma,

Poole, United Kingdom) was prepared as a stock solutionof 1 mg/mL and diluted with phosphate-buffered saline(Sigma) to required concentrations at the time of experi-mentation. Whole blood was incubated for 1 hour withthe respective endotoxin concentration in a water bath at37°C before the Phagotest or Bursttest was performed.

Endotoxin Removal from Patient Plasma

Using Polymixin B. Detoxi-Gel affinity-pack pre-packed columns (Pierce Biotechnology, Rockford, IL)containing immobilized polymixin B, which binds the

lipid A portion of bacterial lipopolysaccharide (LPS),were used to remove endotoxin from plasma samples (di-luted 1:1 with phosphate-buffered saline). The endotox-in-free, diluted plasma (100 �L) was incubated with 50�L of a cell suspension, and the Bursttest or Phagotest wasapplied.

Fig. 1. Representative FACS analysis plots for Phagotest and Bursttestare shown. (A) Neutrophils are gated according to their forward and sidescatter characteristics. (B) Analysis of phagocytosis: on a sample withoutbacteria, markers are set so that more than 99% of the gated neutrophilsare within the first maker. (C) Representative FACS plot from a healthycontrol subject (resting oxidative burst � 8.3%). The percentage ofdouble positive cells (shown in the right upper quadrant) is measured.(D) Corresponding FACS plot of a healthy control with a normal phago-cytic capacity (100%). (E) Representative FACS plot from a patient witha low resting oxidative burst (39.2%). (F) Corresponding FACS plot of apatient with a relative GMFI of 50%. (G) Representative FACS plot froma patient with a high resting oxidative burst (99.9%). (H) CorrespondingFACS plot of the same patient in part C with a relative GMFI of 17%.

HEPATOLOGY, Vol. 46, No. 3, 2007 MOOKERJEE, STADLBAUER, ET AL. 833

Using Anti-CD14 Antibody. Plasma (50 �L) and 50�L of isolated neutrophils were incubated for 60 minuteswith 5 �L of an anti–human anti-CD14 antibody (Clone11D18, Immuntools, Friesoythe, Germany) known toneutralize LPS before the Phagotest or Bursttest was per-formed.

CytokinesTNF�, interleukin-6 (IL-6), and interleukin-8 (IL-8)

were determined by an enzyme-linked immunosorbentassay from ethylenediaminetetraacetate anticoagulatedplasma samples with commercially available antibody sets(BioSource International, Nivelles, Belgium) in accor-dance with the manufacturer’s instructions as described.17

The lower limit for the detection of the cytokines was 3pg/mL. The intra-assay CV was 5.4%-6.4%. IL-6 andIL-8 were undetectable in the controls.

Malondialdehyde (MDA) and Prostaglandin F2�MDA was determined with a modified thiobarbituric aci-

d–reactive substance assay described by Lapenna et al.24 inwhich the major interfering/oxidizable component in theplasma is inhibited by the addition of sodium sulfate.

Free 8-isoprostane F2� was assayed with a commercialenzyme immunoassay kit (Cayman Chemical, Ann Ar-bor, MI) according to the manufacturer’s instructions.

StatisticsFor a comparison of 2 groups, a chi-square test, t test, or

Mann-Whitney test was used (whatever was appropriate),whereas an analysis of variance test with Tukey’s (equal vari-ances) or Dunnett C (nonequal variances) post hoc analysiswas used for the comparison of more than 2 data sets. Toassess the diagnostic accuracy, the areas under the receiveroperator curves (AUROCs) were calculated. Differences insurvival were analyzed by the log-rank test. Pearson’s corre-lation coefficient was used to assess relationships betweenvariables. The results are presented as means � SEM, withP � 0.05 considered significant.

ResultsPatient Characteristics

Sixty-three patients, all of whom fulfilled the studycriteria, were enrolled. Patients classified histologically ashaving cirrhosis�AH (n � 23, all with moderate to severesteatohepatitis) were more severely ill, as evidenced byhigher MELD and Pugh scores (P � 0.001) in compari-son with patients with only cirrhosis (n � 40). Patientswith cirrhosis�AH also had significantly higher CRP(P � 0.005) and white cell counts (P � 0.001). All pa-tients had higher levels of TNF�, IL-6, IL-8, MDA, andisoprostane F2� than healthy controls. As anticipated,

patients with cirrhosis�AH had significantly higher levelsof IL-6 and IL-8 than those with cirrhosis alone, but themeasures of oxidative stress did not differ significantlybetween these patient subsets, nor did they correlate withliver disease severity. Furthermore, there was no correla-tion between levels of cytokines and markers of oxidativestress and neutrophil function. Table 1 outlines baselinecharacteristics for all patients and distinguishes betweenthe groups of cirrhosis�AH and cirrhosis alone. For exvivo experiments, whole blood, isolated neutrophils, orplasma from 16 of the 63 patients was used (6 hadcirrhosis�AH and 10 had cirrhosis alone), and the base-line clinical characteristics for these 16 patients were notsignificantly different from those of the whole cohort.

Oxidative Burst and Phagocytosis in Patients withAlcoholic Liver Disease

Patients with cirrhosis�AH had a significantly higherresting oxidative burst than patients with cirrhosis alone(P � 0.001) or controls (P � 0.001, Fig. 2). Patients withcirrhosis alone also had a higher resting oxidative burstthan controls (P � 0.01, Fig. 2). Stimulation with fMLP,a synthetic peptide that triggers minimal oxidative burstin unstimulated neutrophils but a significantly higher re-sponse in primed neutrophils, resulted in a much greateroxidative burst in patients with cirrhosis (P � 0.01) andcirrhosis�AH (P � 0.001) in comparison with controls,suggesting prior priming of these patients’ neutrophils.There was no difference in response to E. coli or phorbol-12-myristate-13-acetate as a high stimulus between thepatient subgroups. There was no difference between theresting burst and response to fMLP in patients withcirrhosis�AH anymore, whereas in patients with cirrho-sis alone, the fMLP response was 22% higher than theresting burst (P � 0.05). Furthermore, upon stimulationwith E. coli, the relative increase in the oxidative burstfrom resting levels was significantly reduced in patientswith cirrhosis�AH in comparison with patients with cir-rhosis alone (P � 0.001) or controls (P � 0.001, Fig. 2).Furthermore, patients with cirrhosis�AH engulfed sig-nificantly less bacteria than controls (P � 0.05, Fig. 2).

Association of Resting Oxidative Burst andPhagocytosis with Organ Failure and Survival

Seventeen patients (27%) developed organ failure, and14 (22%) died within 90 days (30-day mortality 13%).Renal failure occurred in 15 (88%) and was associatedwith ventilatory and circulatory failure in 4. The restingoxidative burst was found to be predictive of 90-day mor-tality (AUROC 0.77, P � 0.005, Fig. 3A,B) and organfailure (AUROC 0.76, P � 0.001). A resting burst ofgreater than 55% had a sensitivity of 77% and a specificity


of 69% for predicting death. The phagocytic function wasalso predictive of death (AUROC 0.80, P � 0.05, Fig.3C,D) and organ failure (AUROC 0.91, P � 0.0001). Aphagocytic capacity less than 42% of normal among thestudied patients had a sensitivity of 86% and a specificityof 70% to predict mortality.

Association of the Resting Oxidative Burst andPhagocytosis with the Development of Infection

Although none of the included patients had a proveninfection at the time of assay for neutrophil function, in

38 (60%) patients, infection was clinically suspected dur-ing the course of the hospital admission. In 19 of 38patients (50%), a culture-positive infection was verifiedwithin 30 days of admission, albeit our management pro-tocol necessitates the use of broad-spectrum antibiotics assoon as an infection is suspected. In 8, more than 1 or-ganism was found. The following species were cultured:Enterococcus (n � 9), coagulase-negative Staphylococcus(n � 8), Candida sp. (n � 3), E. coli (n � 2), methicillin-resistant Staphylococcus aureus (n � 2), S. aureus (n � 2),and Propionibacterium sp. (n � 1). Patients withcirrhosis�AH were more likely to develop culture-posi-tive infections (57% versus 15%, P � 0.001). Further-more, when patients were stratified for resting burst, thosewith a high resting burst (�55%) were more likely todevelop culture-positive infections than patients with alow resting burst (48% versus 15%, P � 0.005, Fig. 4A).Moreover, only patients with a high burst developedinfections with more than 1 organism, and they alsodeveloped these infections earlier during their hospitalstay in comparison with patients with a low burst (7versus 16 days, P � 0.005). Additionally, 53% of thepatients with a phagocytic capacity lower than 42%developed a culture-positive infection versus 13% ofthe patients with a phagocytic capacity greater than orequal to 42% (P � 0.05). Those patients who devel-oped culture-positive infections were more likely todevelop organ failure (P � 0.001) and to die (P �0.0001, Fig. 4B).

Table 1. Patient Characteristics

NormalRange

All Patients(n � 63)

Cirrhosis(n � 40)

Cirrhosis�AH(n � 23)

Patients Selected forEx Vivo Studies

(n � 16)

Age (years) 50.3 � 1.3 52.2 � 1.7 47.1 � 1.9 51.8 � 2.7Biochemistry/liver functionBilirubin (�mol/L) 3–17 151.2 � 20.9 94.0 � 17.4 249.2 � 39.9* 145.8 � 47.7Prothrombin time (seconds) 10–12 15.3 � 0.6 13.6 � 0.5 17.2 � 1.0* 13.3 � 0.6Albumin (g/L) 35–53 29.8 � 1.1 32.7 � 1.2 27.6 � 1.4* 33.9 � 1.3White blood cell count (109/L) 3–10 10.1 � 0.9 7.9 � 0.9 14.1 � 1.5* 7.9 � 1.2CRP (mg/L) 0–5 30.9 � 4.5 21.4 � 5.2 45.7 � 7.0* 22.6 � 7.1ALT (U/L) 8–63 62.8 � 10.8 73.9 � 15.8 40.8 � 5.7 82.7 � 34.7Maddrey’s discriminant function — — 43.3 � 6.2 —Pugh score 9.3 � 0.4 8.0 � 0.3 11.1 � 0.4* 8.7 � 0.6MELD 15.6 � 1.8 10.7 � 1.1 23.5 � 4.1* 15.1 � 2.8Cytokine/inflammationTNF� (pg/mL) 0–5 19.6 � 6.5† 18.3 � 6.9† 22.3 � 14.6† 39.8 � 14.6IL-6 (pg/mL) 0–5 49.4 � 14.9† 21.9 � 7.9† 106.1 � 39.6*,† 43.2 � 17IL-8 (pg/mL) 0–5 180.5 � 56.9† 101.8 � 55.8† 337.9 � 122.8*,† 39.7 � 10.1Oxidative stressMDA (�mol/L) 1.2–2.5 3.2 � 0.5† 2.7 � 0.5† 4.3 � 0.8† 2.3 � 0.5Isoprostane F2� (pg/mL) 40–100 346.8 � 49.6† 296.9 � 43.8† 394.7 � 81.2† 343.9 � 46.3OutcomeDeath n (%) 14 (22) 3 (8) 11 (48) 3 (19)Organ failure n (%) 17 (27) 3 (8) 14 (61) 4 (25)

*P � 0.001 versus cirrhosis.†P � 0.05 versus control.

Fig. 2. The column bars represent the percentage of cells undergoinga resting oxidative burst in untreated neutrophils (white) and afterstimulation with fMLP (light gray) or E. coli (dark gray) in samples fromcontrols (n � 13), patients with cirrhosis (n � 40), and patients withcirrhosis�AH (n � 23). The last column bar (black) depicts the phago-cytic capacity (relative GMFI) in the study groups.


Effect of Patients’ Plasma and Normal Plasma onthe Neutrophil Oxidative Burst

Plasma from patients with a high resting burst (�55%;n � 6) induced a high resting burst in normal neutrophils(P � 0.005), whereas plasma from patients with a lowresting burst (�55%; n � 6) failed to do so (Fig. 5A).

This result suggests the presence of a transmissible factorin patients’ plasma promoting neutrophil activation.

When isolated neutrophils from patients with a knownhigh resting burst were incubated with normal plasma,the resting burst decreased significantly in comparisonwith isolated neutrophils incubated with the patients’own plasma (P � 0.02; Fig. 5B). This may suggest theremoval of a factor present in patients’ plasma by thetransfer of their neutrophils into a more healthy environ-ment.

Effect of Patients’ Plasma and Normal Plasma onPhagocytosis

Neutrophils from healthy controls incubated withplasma from patients with a low resting burst did notdiffer from those of controls, whereas neutrophils fromhealthy controls incubated with plasma from patientswith a high resting burst showed a 22% decrease in thephagocytic capacity (P � 0.05, n � 6). Conversely, incu-bating patients’ neutrophils for 60 minutes with plasmafrom healthy controls showed a 22% increase (P � 0.05,n � 6) in phagocytosis in comparison with patients’ neu-trophils incubated with their own plasma (Fig. 5C).These results concur with derangements in oxidativeburst and indicate that impaired phagocytic function maybe due to a transmissible and reversible serum factor.

Effect of Endotoxin on Oxidative Burst andPhagocytosis

Incubating neutrophils from healthy volunteers (n �5) with increasing endotoxin concentrations resulted in aconcentration-dependent increase in the resting burst

Fig. 3. (A) AUROC shows the predictive utility of the measurement ofthe oxidative burst in determining survival. (B) A resting burst of greaterthan 55% had the best sensitivity and specificity on AUROC analysis togroup the patients and was used to construct a Kaplan-Meier survivalcurve; the log-rank analysis shows a significantly higher mortality (P �0.01) in the group of patients with a high burst (�55%). (C) AUROCanalysis also reveals that the phagocytic capacity is predictive in deter-mining survival. (D) A phagocytic capacity of less than 42% had the bestsensitivity and specificity on AUROC analysis and was used to constructa Kaplan-Meier survival curve; the log-rank analysis shows a significantlyhigher mortality (P � 0.001) in the group of patients with decreasedphagocytic capacity (�42%).

Fig. 4. (A) An analysis of the time to infection (Kaplan-Meier andlog-rank) shows that patients with a high resting burst (�55%) have asignificantly higher risk (P � 0.005) of developing culture-positiveinfections within 30 days. (B) According to Kaplan-Meier and log-rankanalyses used to determine differences in survival between patients withand without culture-positive infections, those with infections have asignificantly higher 90-day mortality (P � 0.001).


(P � 0.0001, Fig. 6A). Furthermore, incubating patients’neutrophils with endotoxin reduced the phagocytic ca-pacity by 20% (n � 8, P � 0.05, Fig. 6B). These resultssuggest that endotoxin modulates the neutrophil function

in a manner comparable to that seen after exposure topatient plasma.

Effect of Removing Endotoxin from a Patient’sPlasma

Using Polymixin B Columns. When patient plasmaevoking a high resting burst in the whole blood assay waspassed over polymixin B columns, the resting burst wasreduced by 32% (P � 0.001, n � 9). Plasma from pa-tients with a low resting burst (n � 4) and normal plasma(n � 3) did not change the resting burst of isolatedhealthy donor neutrophils (Fig. 7A). Endotoxin removalfrom the plasma of patients with a high resting burst (n �11) also increased the phagocytic capacity by 31% (P �0.05) in comparison with cells incubated with untreatedplasma but had no significant effect on the plasma ofpatients with a low burst (n � 8) or normal plasma (n �5; Fig. 7B). This set of experiments suggests that endo-toxin removal by polymixin B reverses the burst-inducingand phagocytosis-decreasing effect of patients’ plasma.

Using LPS-Neutralizing Antibodies. Incubationwith an LPS-neutralizing anti–human anti-CD14 anti-body prevented the induction of increased burst in nor-mal neutrophils by the plasma of patients with a highburst (P � 0.001, n � 7) but had no effect on neutrophilsincubated with the plasma of patients with a low burst(n � 8) or healthy control plasma (n � 3, Fig. 7A). Theincubation of plasma of patients with a high burst withthe anti-CD14 antibody increased the phagocytic capac-ity by 20% (P � 0.04, n � 11), whereas this antibody hadno effect on the phagocytic capacity when plasma frompatients with a low burst (n � 8) or healthy controls (n �

Fig. 6. (A) Dose-dependent increase in the neutrophil resting burstfollowing incubation with endotoxin. (B) Incubation with endotoxin doesnot change phagocytosis in normal neutrophils but decreases phagocy-tosis further in patients’ neutrophils.

Fig. 5. (A) The bars show the resting oxidativeburst in the whole blood of controls, in the wholeblood of patients, and in normal neutrophils incu-bated with patients’ plasma. Plasma from patientswith a high resting burst induced a high burst innormal neutrophils, whereas plasma from patientswith a low resting burst failed to do so. (B) Theelevated resting oxidative burst in patients’ neutro-phils incubated with their own plasma is reversedafter incubation with normal plasma. (C) The incu-bation of normal neutrophils with plasma frompatients with a low resting burst does not changephagocytosis; plasma from patients with a highburst decreases phagocytosis. The incubation ofpatients’ neutrophils with normal plasma restoresphagocytic function.


3) was used on normal neutrophils (Fig. 7B). This furthersupports the concept that endotoxin may be responsiblefor the observed induction of high resting burst in neu-trophils from patients with AH.

DiscussionThe results of this study suggest a severe functional

failure of neutrophils in a proportion of patients withcirrhosis�AH and that these defects are associated withincreased risk of infection, organ failure, and mortality.The ex vivo studies support the notion that this neutrophildysfunction is contributed by endotoxin and is reversibleif the plasma is treated with endotoxin-removal strategiessuch as passing the plasma over a polymixin B column orusing LPS-neutralizing anti-CD14 antibodies. Our re-sults provide important clues to explain the apparent par-adox in neutrophil function in the context ofinflammation on the one hand and increased susceptibil-ity to infection on the other.

AH is a proinflammatory condition in which TNF� isthought to promote a cascade of proinflammatory signalsthat perpetuate liver injury.25-27 Evidence for an enhancedproinflammatory cytokine environment was demon-strated in the cirrhosis�AH group in our study. More-over, circulating neutrophils have been shown to beprimed in AH,7,8 resulting in enhanced responses to amicrobial challenge.6 Our data confirm that neutrophilsin patients with alcoholic cirrhosis and cirrhosis�AH areprimed as shown by increased oxidative burst upon stim-ulation with fMLP in comparison with controls. Thepresence of primed neutrophils would suggest good bac-tericidal function but with the potential for promotinglocal damage if further activated.

Paradoxically, despite data suggesting neutrophilpriming, it is also established that patients with alcoholiccirrhosis, especially those with AH, are prone to infec-tions,28 with a higher prevalence of sepsis,29 and are morelikely to die from sepsis-related complications.14 Thisclinical finding is supported by several studies showingdefects in phagocytosis, bactericidal capacity, chemotaxis,and neutrophil locomotion.9 It is therefore important todistinguish between primed (but not activated) neutro-phils and those that are fully activated. Activated neutro-phils were defined in this study as neutrophils that have ahigh resting burst above and beyond the primed state(increased oxidative burst following fMLP stimulation).Additionally, the presentation of a bacterial challenge toactivated neutrophils was associated with an inability ofthese cells to generate a further oxidative burst in compar-ison with primed cells. The neutrophils in our study pop-ulation of cirrhosis�AH showed both neutrophilactivation and impairment of phagocytosis. We can onlyspeculate on the mechanism by which an increased restingoxidative burst might be related to a decreased phagocyticcapacity. Because both oxidative burst and phagocytosisare energy-dependent processes, the induction of oxida-tive burst may lead to energy depletion and reduction inphagocytosis. This is supported by data showing that theactivation of adenosine monophosphate kinases in a stateof low cellular energy contributes to a reduction in hostdefense mechanisms.30

The clinical importance of these neutrophil abnormal-ities, which are identified by a high resting oxidative burstgreater than or equal to 55% and a reduced phagocyticcapacity (relative GMFI � 42%), are highlighted by theobservation of increased risk of infection and the associa-tion with organ failure and mortality in these patients. Itis notable that this group of patients in our study hadinfections with multiple organisms despite antibiotictherapy within a relatively short time frame. Although weattempted to exclude all patients with preexisting infec-

Fig. 7. (A) A high resting oxidative burst is reversible by the passageof plasma over an endotoxin-removal column or following incubation withanti-CD14 antibodies. Plasma from patients with a low burst or controlplasma passed over the column or incubated with anti-CD14 antibodiesdoes not influence the resting burst. (B) Impaired phagocytosis isreversible by the passage of plasma from patients with a high burst overan endotoxin-removal column or incubation with anti-CD14 antibodies.Plasma from patients with a low burst or from controls passed over thecolumn or incubated with anti-CD14 antibodies does not influencephagocytosis.


tions at the baseline, we cannot fully exclude the possibil-ity of underlying subclinical infections that were notdetected by current routine microbiology techniques,which might precipitate a high resting oxidative burst insome patients. It has been shown that 32% of patientswith cirrhosis with culture-negative ascites have detect-able bacterial DNA in the blood, which may indicatesystemic bacterial seeding.31 An important observationmade through ex vivo experiments in this study was thedemonstration that the neutrophil functional defectcould be transmitted to normal neutrophils incubatedwith patients’ serum. We also showed that the neutrophildysfunction in the patients was reversible following incu-bation with normal serum. As endotoxin is known to beelevated in patients with alcoholic liver disease, we ex-plored whether this transmissible factor in the patients’plasma imparting the functional disturbance in neutro-phils was endotoxin. Endotoxin not only is a primingagent6 but has been reported to fully activate neutro-phils12,13 by up-regulating the NADPH oxidase assem-bly.32 The highest levels of endotoxin reported in patientswith liver disease are found in portal venous blood,33,34

highlighting the importance of the bowel and altered gutpermeability as a possible source of endotoxin.35 How-ever, the difficulties with the measurement of endotoxinare reflected in the wide range of values in the literatureand make this an unreliable indicator on which to basefurther management.33,34 Our data do support the notionthat endotoxin may be the putative humoral factor inpatients’ plasma responsible for the activation of neutro-phils by showing that endotoxin increased the restingburst in healthy neutrophils while reducing phagocytosisin neutrophils from patients with alcoholic cirrhosis.

Having established a potential role for endotoxin, wedetermined whether different strategies to remove endo-toxin from a patient’s plasma ex vivo would result in im-proved neutrophil function. We chose to use 2approaches, a gel containing bound polymixin B and ananti-CD14 antibody, both of which have been validatedin their ability to bind endotoxin. The removal of endo-toxin by both these methods resulted in a effective reduc-tion in the high resting oxidative burst and also improvedthe phagocytic capacity. This novel observation, showingthe potential of endotoxin removal to affect the functionalstatus of circulating neutrophils, may have importanttherapeutic implications. Endotoxin removal has beentested in patients with sepsis and multiorgan failure, withimprovements in clinical and biochemical parametersnoted.36 However, a definitive survival benefit has notbeen shown to date, most likely because of the heteroge-neity of the conditions studied.

Our results may also have important implications inthe selection of patients for immunosuppressive therapyfor AH. Although controversy exists around the use ofcorticosteroids and anti-TNF strategies for the routinetreatment of AH, there is little doubt of their effectivenessin selected patients. Our data provide insight into theconflicting results of clinical trials and may suggest a ra-tionale to select an appropriate therapy tailored to thepatient. The use of oxidative burst measurements to cat-egorize patients in whom the neutrophils are primed butnot activated (indicated by an increased fMLP responsebut a resting burst lower than 55%) would identify thosepatients with AH that are likely to respond to anti-inflam-matory strategies but have a low risk of precipitated infec-tion. However, patients in whom the neutrophils areactivated (indicated by a high resting burst greater than orequal to 55%) are likely to have a high risk of infection,and strategies to remove endotoxin may be more benefi-cial. This hypothesis will need to be tested in appropri-ately powered clinical trials.

In conclusion, our study data provide an explanationfor the apparent paradox of increased inflammation andsimultaneous heightened risk of infection and suggest animportant role for endotoxin as the humoral factor thattriggers neutrophil activation. We also show the revers-ibility of neutrophil dysfunction following endotoxin re-moval and suggest that neutrophil activation rendersthem hyporesponsive to further bacterial challenge, ac-counting for the increased rate of infection and mortality.The mechanism for phagocytic dysfunction in fully acti-vated neutrophils requires further study, but our data pro-vide a rationale for improved selection of patients forcurrent therapies and suggest new therapeutic approachesfor the management of AH.

References1. Mathurin P, Duchatelle V, Ramond MJ, Degott C, Bedossa P, Erlinger S,

et al. Survival and prognostic factors in patients with severe alcoholichepatitis treated with prednisolone. Gastroenterology 1996;110:1847-1853.

2. Aggarwal A, Ong JP, Younossi ZM, Nelson DR, Hoffman-Hogg L, Arro-liga AC. Predictors of mortality and resource utilization in cirrhotic pa-tients admitted to the medical ICU. Chest 2001;119:1489-1497.

3. Navasa M, Fernandez J, Rodes J. Bacterial infections in liver cirrhosis. ItalJ Gastroenterol Hepatol 1999;31:616-625.

4. Linderoth G, Jepsen P, Schonheyder HC, Johnsen SP, Sorensen HT.Short-term prognosis of community-acquired bacteremia in patients withliver cirrhosis or alcoholism: a population-based cohort study. AlcoholClin Exp Res 2006;30:636-641.

5. Jaeschke H, Hasegawa T. Role of neutrophils in acute inflammatory liverinjury. Liver Int 2006;26:912-919.

6. Condliffe AM, Kitchen E, Chilvers ER. Neutrophil priming: pathophysi-ological consequences and underlying mechanisms. Clin Sci (Lond) 1998;94:461-471.


7. Taieb J, Mathurin P, Elbim C, Cluzel P, Arce-Vicioso M, Bernard B, et al.Blood neutrophil functions and cytokine release in severe alcoholic hepa-titis: effect of corticosteroids. J Hepatol 2000;32:579-586.

8. Stanley AJ, MacGregor IR, Dillon JF, Bouchier IA, Hayes PC. Neutrophilactivation in chronic liver disease. Eur J Gastroenterol Hepatol 1996;8:135-138.

9. Rajkovic IA, Williams R. Mechanisms of abnormalities in host defences againstbacterial infection in liver disease. Clin Sci (Lond) 1985;68:247-253.

10. Thalheimer U, Triantos CK, Samonakis DN, Patch D, Burroughs AK.Infection, coagulation, and variceal bleeding in cirrhosis. Gut 2005;54:556-563.

11. Urbaschek R, McCuskey RS, Rudi V, Becker KP, Stickel F, Urbaschek B,et al. Endotoxin, endotoxin-neutralizing-capacity, sCD14, sICAM-1, andcytokines in patients with various degrees of alcoholic liver disease. AlcoholClin Exp Res 2001;25:261-268.

12. Etheredge EE, Spitzer JA. Chronic endotoxemia reversibly alters respira-tory burst activity of circulating neutrophils. J Surg Res 1993;55:261-268.

13. Bohmer RH, Trinkle LS, Staneck JL. Dose effects of LPS on neutrophils ina whole blood flow cytometric assay of phagocytosis and oxidative burst.Cytometry 1992;13:525-531.

14. Mathurin P, Mendenhall CL, Carithers RL Jr, Ramond MJ, MaddreyWC, Garstide P, et al. Corticosteroids improve short-term survival inpatients with severe alcoholic hepatitis (AH): individual data analysis of thelast three randomized placebo controlled double blind trials of corticoste-roids in severe AH. J Hepatol 2002;36:480-487.

15. Christensen E, Gluud C. Glucocorticoids are ineffective in alcoholic hep-atitis: a meta-analysis adjusting for confounding variables. Gut 1995;37:113-118.

16. Menon KV, Stadheim L, Kamath PS, Wiesner RH, Gores GJ, Peine CJ, etal. A pilot study of the safety and tolerability of etanercept in patients withalcoholic hepatitis. Am J Gastroenterol 2004;99:255-260.

17. Tilg H, Jalan R, Kaser A, Davies NA, Offner FA, Hodges SJ, et al. Anti-tumor necrosis factor-alpha monoclonal antibody therapy in severe alco-holic hepatitis. J Hepatol 2003;38:419-425.

18. Naveau S, Chollet-Martin S, Dharancy S, Mathurin P, Jouet P, PiquetMA, et al. A double-blind randomized controlled trial of infliximab asso-ciated with prednisolone in acute alcoholic hepatitis. HEPATOLOGY 2004;39:1390-1397.

19. Runyon BA. Management of adult patients with ascites due to cirrhosis.HEPATOLOGY 2004;39:841-856.

20. Brunt EM, Janney CG, Di Bisceglie AM, Neuschwander-Tetri BA, BaconBR. Nonalcoholic steatohepatitis: a proposal for grading and staging thehistological lesions. Am J Gastroenterol 1999;94:2467-2474.

21. Maddrey WC, Boitnott JK, Bedine MS, Weber FL Jr, Mezey E, White RIJr. Corticosteroid therapy of alcoholic hepatitis. Gastroenterology 1978;75:193-199.

22. Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM,Kosberg CL, et al. A model to predict survival in patients with end-stageliver disease. HEPATOLOGY 2001;33:464-470.

23. Pugh RN, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Tran-section of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646-649.

24. Lapenna D, Ciofani G, Pierdomenico SD, Giamberardino MA, Cuccu-rullo F. Reaction conditions affecting the relationship between thiobarbi-turic acid reactivity and lipid peroxides in human plasma. Free Radic BiolMed 2001;31:331-335.

25. McClain CJ, Cohen DA. Increased tumor necrosis factor production bymonocytes in alcoholic hepatitis. HEPATOLOGY 1989;9:349-351.

26. Felver ME, Mezey E, McGuire M, Mitchell MC, Herlong HF, Veech GA,et al. Plasma tumor necrosis factor alpha predicts decreased long-termsurvival in severe alcoholic hepatitis. Alcohol Clin Exp Res 1990;14:255-259.

27. Tilg H, Diehl AM. Cytokines in alcoholic and nonalcoholic steatohepati-tis. N Engl J Med 2000;343:1467-1476.

28. Fernandez J, Navasa M, Gomez J, Colmenero J, Vila J, Arroyo V, et al.Bacterial infections in cirrhosis: epidemiological changes with invasive pro-cedures and norfloxacin prophylaxis. HEPATOLOGY 2002;35:140-148.

29. Verma S, Ajudia K, Mendler M, Redeker A. Prevalence of septic events,type 1 hepatorenal syndrome, and mortality in severe alcoholic hepatitisand utility of discriminant function and MELD score in predicting theseadverse events. Dig Dis Sci 2006;51:1637-1643.

30. Alba G, El Bekay R, Alvarez-Maqueda M, Chacon P, Vega A, MonteseirinJ, et al. Stimulators of AMP-activated protein kinase inhibit the respiratoryburst in human neutrophils. FEBS Lett 2004;573:219-225.

31. Such J, Frances R, Munoz C, Zapater P, Casellas JA, Cifuentes A, et al.Detection and identification of bacterial DNA in patients with cirrhosisand culture-negative, nonneutrocytic ascites. HEPATOLOGY 2002;36:135-141.

32. DeLeo FR, Renee J, McCormick S, Nakamura M, Apicella M, Weiss JP, etal. Neutrophils exposed to bacterial lipopolysaccharide upregulateNADPH oxidase assembly. J Clin Invest 1998;101:455-463.

33. Prytz H, Holst-Christensen J, Korner B, Liehr H. Portal venous and sys-temic endotoxaemia in patients without liver disease and systemic endo-toxaemia in patients with cirrhosis. Scand J Gastroenterol 1976;11:857-863.

34. Lumsden AB, Henderson JM, Kutner MH. Endotoxin levels measured bya chromogenic assay in portal, hepatic and peripheral venous blood inpatients with cirrhosis. HEPATOLOGY 1988;8:232-236.

35. Garcia-Tsao G, Wiest R. Gut microflora in the pathogenesis of the com-plications of cirrhosis. Best Pract Res Clin Gastroenterol 2004;18:353-372.

36. Hanasawa K. Extracorporeal treatment for septic patients: new adsorptiontechnologies and their clinical application. Ther Apher 2002;6:290-295.


Neutrophil dysfunction in alcoholic hepatitis superimposed on cirrhosis is reversible and predicts the outcome

Documents