Paneth cell dysfunction mediates alcoholic steatohepatitis through promoting bacterial translocation in mice: role of zinc deficiency

Abstract

Microbial dysbiosis is associated with alcoholic hepatitis (AH) with the mechanisms yet to be elucidated. The present study aimed to determine the effects of alcohol and zinc deficiency on Paneth cell (PC) antimicrobial peptides - α-defensins, and to define the link between PC dysfunction and AH. Translocation of pathogen-associated molecular patterns (PAMPs) was determined in patients with severe AH and in a mouse model of alcoholic liver disease. Microbial composition and PC function were examined in mice. The link between α-defensin dysfunction and AH was investigated in α-defensin deficient mice. Synthetic human α-defensin 5 (HD5) was orally given to the alcohol-fed mice to test the therapeutic potential. The role of zinc deficiency in α-defensin was evaluated in acute and chronic mouse models of zinc deprivation. Hepatic inflammation was associated with PAMP translocation, and lipocalin-2 (LCN2) and CXCL1 elevation in AH patients. Antibiotic treatment, lipopolysaccharide injection to mice, and in vitro experiments showed that PAMPs, but not alcohol, directly induced LCN2 and CXCL1. Chronic alcohol feeding caused systemic dysbiosis and PC α-defensin reduction in mice. Knockout of functional α-defensins synergistically affected alcohol-perturbed bacterial composition and gut barrier, and exaggerated PAMP translocation and liver damage. Administration of HD5 effectively altered cecal microbial composition, especially increased Akkermansia muciniphila, and reversed alcohol-induced deleterious effects. Zinc regulated PC homeostasis and α-defensins function at multiple levels, and dietary zinc deficiency exaggerated the deleterious effect of alcohol on PC bactericidal activity. Conclusions: Taken together, the study suggests that alcohol-induced PC α-defensin dysfunction is mediated by zinc deficiency and involved in the pathogenesis of AH. HD5 administration may represent a novel and promising therapeutic approach for treating AH.

Introduction

Alcoholic liver disease (ALD) is one of the major causes of liver-related mortality worldwide, which encompasses steatosis, hepatitis and cirrhosis.¹ Alcoholic hepatitis (AH), characterized by jaundice and progressive inflammation, has high mortality rates, and few effective medical therapies have been established. Translocation of pathogen associated molecular patterns (PAMPs) (e.g. bacterial endotoxin) from the intestinal tract to the circulation represents a crucial inflammatory signaling that triggers hepatic proinflammatory response and subsequent injury.² Accumulating evidence demonstrate that alcohol-perturbed gut microbiota composition (dysbiosis) promotes ALD, and treatment with commensal bacterial species, such as Lactobacillus spp. or Akkermansia muciniphila (A. muciniphila), or antibiotics protects against alcohol-induced liver damage.^{2, 3}

Early studies reveal that alcohol disrupts the intestinal barrier, increases intestinal permeability, which facilitates the leakage of PAMPs. Emerging data suggest that alcohol-impaired intestinal immune function, such as antimicrobial peptides (AMPs), also influences the development of ALD. Indeed, antimicrobial C-type regenerating islet-derived-3 lectins (Reg3β and Reg3γ) have been reported to be reduced by alcohol, and overexpressing Reg3γ in mice ameliorated experimental ALD.⁴ Paneth cells (PCs) are highly specialized AMP-secreting epithelial cells residing at the base of the crypts of Lieberkühn in the small intestine.⁵ These cells are innate immune cells that synthesize and secrete substantial quantities of AMPs, including α-defensins, lysozyme, Reg3γ, and phospholipase A2, with α-defensins being the dominant type. AMPs are enwrapped in cytoplasmic secretory granules, and rapidly released into the crypt lumen upon stimulation. They regulate the symbiosis of commensal microbiota and prevent pathogenic microbial invasion. However, the roles of PC α-defensins in the pathogenesis of ALD still remain obscure.

Alterations in zinc metabolism and zinc deficiency consistently occur in alcoholic patients and experimental models of ALD,⁶ which further complicated the disease progression. In particular, clinical studies showed that the reduction of serum zinc levels correlates with the severity of ALD.⁶ Zinc, an essential trace element, is required for multiple cellular functions, including proliferation, redox balance, and immune regulation. The intestine is the site of zinc absorption and excretion. Interestingly, PC contains abundant cellular zinc, which is extruded along with AMPs during degranulation.⁷ However, why PC has adopted such mechanism and the potential interaction between zinc and AMPs still remain a mystery, needless to say with the complication of alcohol.

In this study, we aimed to determine the impact of alcohol on PC α-defensins, the pathological effect of α-defensin dysfunction on alcohol-induced PAMP translocation and hepatic inflammation, and identify potential therapeutic targets modulating host-microbiota homeostasis for ALD treatment. We also investigated the regulatory role of zinc in α-defensins and how alcohol-induced zinc deficiency mediates with α-defensin dysfunction.

Materials & Methods

Human liver samples with severe alcoholic hepatitis

Liver explant specimens from patients with severe alcoholic hepatitis (SAH) and wedge biopsies from the donor livers (normal control) were collected at Johns Hopkins University under the support of NIAAA-funded Clinical Resource for Alcoholic Hepatitis Investigations (R24AA025017). Hematology results of these patients (age range 32–49 yr old) indicated severe liver damage, including ALT 52.4±13.9 U/L, AST 125.6±19.6 U/L, and bilirubin 32.4±12.6 mg/dL. Collecting tissues from explanted livers or biopsies from donor livers had been approved by Institutional Review Boards at Johns Hopkins Medical Institutions (IRB00107893, IRB00021325).

Mice

C57BL/6J wild type (WT) mice and MMP7 knockout mice (Mmp7^−/−, deficient in active α-defensins; stock no. 005111) were purchased from the Jackson Laboratory (Bar Harbor, ME). C57BL/6J WT and Mmp7^−/− mice were crossed to generate Mmp7^+/− heterozygous mice, which were then bred to produce WT and Mmp7^−/− littermates for animal studies. Mice were handled and all experiments were performed in accordance with the protocol approved by the North Carolina Research Campus Institutional Animal Care and Use Committee (project number 16019).

Chronic alcohol feeding and treatments

Male mice at 12-wk old were fed an ethanol-containing Lieber-DeCarli liquid diet (alcohol-fed; AF) or an isocaloric control liquid diet (pair-fed; PF) for 8 wk as described previously.⁸ For the dietary zinc deficiency model, the liquid diets were modified based on the Lieber-DeCarli formula as reported previously.⁹ Zinc adequate diets (ZnA) contained 33 mg elemental zinc/L in the form of zinc sulfate, whereas marginal zinc deficient diets (ZnD) contained 6.6 mg elemental zinc/L. All ingredients used in the liquid diets were obtained from Dyets (Bethlehem, PA) except for ethanol and zinc sulfate (Sigma-Aldrich, St. Louis, MO). Polymyxin B (150 mg/kg/day) and neomycin (200 mg/kg/day; Sigma-Aldrich) were given to the liquid diets-fed WT mice every other day for the last 2 wk of the 8-wk feeding. LPS (Sigma-Aldrich) was administrated intraperitoneally at 4 mg/kg 4 hours before finishing the 8-wk feeding. Synthetic human α-defensin 5 (HD5; Peptide International, Louisville, KY) was added to the diet at 10 μg/mouse every other day for the last 2 wk of the 8-wk feeding.

Acute zinc deprivations

In the acute zinc deprivation experiments, male WT mice were injected with 75 mg/kg dithizone (Sigma-Aldrich) intraperitoneally, and intestines were collected 6 h later; or isolated crypts from male WT mice were treated with 10 μmol/L carbamyl choline to induce granule release, and then treated with 2 μmol/L N,N,N′,N′-Tetrakis(2-pyridylmethyl)ethylenediamine (TPEN; Sigma-Aldrich) for 30 min.

Cell culture and treatments

Hepa1c1c7 mouse hepatoma cells (American Type Culture Collection, Rockville, MD) were pre-conditioned with alcohol toxicants, including 100 mmol/L ethanol, 100 μmol/L acetaldehyde, 100 μmol/L hydrogen peroxide (Sigma-Aldrich) or 10 μmol/L 4-hydroxynonenal (Cayman Chemical, Ann Arbor, MI) for 3 days, respectively, followed by 100 ng/mL LPS for 6 h.

Microbial 16S ribosomal RNA sequencing

Cecal content and liver specimens were collected and stored at −80°C until processed. DNA was extracted from these samples, and the V4 hyper variable region of the 16S rRNA gene was amplified and sequenced on an Illumina MiSeq platform using the MiSeq Reagent Kit v2 (Illumina, Inc., San Diego, CA). The output FASTQ files were analyzed using mothur v1.39.5¹⁰ according to the MiSeq SOP (https://www.mothur.org/wiki/MiSeq_SOP).¹¹ After quality-filtering and alignment against SILVA v132 database, sequences were clustered into operational taxonomic units (OTU) with 97% similarity and classified using the RDP classifier.¹² The α-diversity was measured using Observed OTU. Bray-Curtis distance matrices were calculated and visualized using principal coordinates analysis (PCoA). ANOSIM (analysis of similarity) was performed by using mothur v1.39.5 to determine the similarity in community structure between treatments. LEfSe (http://huttenhower.sph.harvard.edu/galaxy/)13 was performed to detect differentially represented bacterial taxa between groups using the default parameters.

Statistical analysis

All human data are expressed as mean±standard error of the mean (SEM), and mouse data are expressed as mean±standard deviation (SD). The data were analyzed using GraphPad Prism software (La Jolla, CA). Statistical significance was carried out using unpaired two-tailed Student’s t-test, Wilcoxon rank sum test, or one-way ANOVA followed by post hoc Newman-Keuls test where appropriate, and considered significant at P<0.05.

Results

Alcoholic hepatitis is associated with hepatic PAMP translocation

In the livers of patients with SAH, we detected increased infiltrated immune cells (CD11b⁺ cells) and activated Kupffer cells (enlarged CD68⁺ cells) compared with the healthy controls (Fig. 1A,B). Meanwhile, SAH patients had significantly elevated levels of lipocalin-2 (LCN2), mainly in hepatocytes (Fig. 1C). Hepatic IL-8 levels were 14 times higher in SAH patients than that in healthy subjects (Fig. 1D). Accompanied with inflammation, hepatic endotoxin levels (Fig. 1E), total bacterial DNAs, and bacterial DNAs of certain pathogenic species (Escherichia coli/E. coli and Helicobacter hepaticus/H. hepaticus; Fig. 1F), were all significantly higher in SAH patients compared with the controls. Similarly, in the AF mice, systemic PAMP translocation was also profound as indicated by elevated plasma endotoxin levels and increased translocation of E. coli to the intestinal lamina propria and to the livers (Fig. S1).

Figure 1.

AH is associated with PAMP translocation. (A) Immunofluorescence (IF) staining of CD11b⁺ and (B) CD68⁺ cells, (C) quantification and immunohistochemistry (IHC) staining of LCN2, (D) quantification and IHC staining of IL-8, (E) endotoxin levels, (F) total 16S bacterial rRNAs, E. coli, and H. hepaticus DNAs relative to β-globin in the livers of patients with SAH (n=5) and healthy controls (n=5). Scale bar, 20 μm (A,B) and 50 μm (C,D). *P<0.05, ^**P<0.01, ^***P<0.001. SAH, severe alcoholic hepatitis.

PAMPs directly contribute to alcohol-induced hepatic LCN2 and CXCL1

To determine the cause-effect relationship between PAMP translocation and the observed LCN2 and CXCL1 induction, mice were given antibiotics (Abx; Fig. 2A) to minimize the amount of translocated PAMPs. Abx administration effectively reversed alcohol-induced hepatic Lcn2 and Cxcl1 expressions (Fig. 2B), and reduced the number of CD11b⁺ cells (Fig. S2). In the subsequent experiment, mice were injected with a single dose of LPS 4-hour prior to finishing the chronic alcohol feeding (Fig. 2C) to test the hepatic response to bacterial pathogens. Surprisingly, unlike the PF mice which were capable of producing dramatically upregulated Lcn2 and Cxcl1, the AF mice only mildly upregulated Lcn2 and did not further alter Cxcl1 expression in response to LPS injection (Fig. 2D), indicating an immunosuppression to PAMPs caused by alcohol. We then did in vitro experiment to dissect the roles of alcohol and LPS in inducing LCN2 and CXCL1. LPS directly stimulated Lcn2 and Cxcl1 expressions in Hepa1c1c7 hepatocytes (Fig. 2E). Pre-treatment of ethanol or other related metabolites, including acetaldehyde, H₂O₂, and 4-hydroxynonenal, did not potentiate LPS-induced Lcn2 and Cxcl1 induction, but rather, pre-treatment of acetaldehyde diminished the response (Fig. 2F). Taken together, the observed LCN2 and CXCL1 induction in SAH patients is a direct consequence of PAMP translocation rather than a bona fide alcoholic toxicity.

Figure 2.

PAMPs are responsible for hepatic LCN2 and CXCL1 induction. (A) Experimental design. Black arrows represent antibiotic (Abx) treatments. (B) Relative hepatic Lcn2 and Cxcl1 levels in mice of the Abx experiment (n=6). (C) Experimental design. Black arrow indicates LPS injection that was given 4 hours before sample collection. (D) Relative hepatic Lcn2 and Cxcl1 levels in mice of the LPS experiment (n= 6). (E) Expressions of Lcn2 and Cxcl1 in Hepa1c1c7 cells treated with LPS for 6 h (n=3). (F) Expressions of Lcn2 and Cxcl1 in Hepa1c1c7 cells pre-treated with alcohol toxicants for 3 d followed by LPS for 6 h (n=3). *P<0.05, ^**P<0.01, ^***P<0.001 except for (M) and (N), where * represents the difference between control and LPS treatments (white bar vs. gray bar), and # represents difference between AcH and other treatments (EtOH, H₂O₂, and 4-HNE; gray bars). PF, pair-fed; AF, alcohol-fed; Ctrl, control; EtOH, ethanol; AcH, acetaldehyde.

Alcohol exposure causes Paneth cell α-Defensin reduction and systemic dysbiosis

As altered gut microbiota composition is evidenced in ALD, and affects susceptibility to the disease, we analyzed bacterial species in both the cecal lumen and the liver. LEfSe analysis (cladograms in Fig. 3A) showed distinguished bacterial populations in both the cecal lumen and liver between the PF and AF groups, suggesting systemically changed bacterial taxa by alcohol (green: alcohol-decreased; red: alcohol-increased). At the family level, Pseudomonadaceae, which belongs to Proteobacteria phylum, constituted the most abundant family in the liver (PF 47.0±17.5%), and was significantly increased in the AF group (86.3±4.7%, P=0.005) (Fig. 3B). Of note, many members in this family are recognized as opportunistic pathogens of clinical relevance.¹⁴ In the cecal contents, members of Bacteroidetes and Firmicutes were dominant, together representing 91.6–97.2% of the total community. Compared with the PF group, the AF group was characterized by a relatively lower abundance of Bacteroidetes (AF 50.5±15.5% vs. PF 70.7±4.2%, P=0.08) and a higher abundance of Firmicutes (AF 42.3±14.3% vs. PF 22.0±4.3%, P=0.03). Furthermore, differences in community structure between the PF and AF groups were observed in both the liver and cecal contents based on Bray-Curtis distances (Fig. S3; ANOSIM, R=0.57, P=0.027 for liver; R=0.63, P=0.027 for cecal contents).

Figure 3.

Chronic alcohol feeding causes systemic bacterial dysbiosis and Paneth cell dysfunction in mice. (A) Cladogram and (B) family-classified taxonomic abundances of bacteria in the livers and cecal contents of mice (n=4). P, phylum. (C) Transmission electron microscopy (TEM) of the ileal crypts of mice (n=3). Arrowheads indicate PCs, and stars indicate secretary granules. Scale bar, 10 μm (upper), 2 μm (lower). (D) Relative ileal mRNA levels of α-defensins in mice (n=6). *P<0.05, ^**P<0.01. (E) IF staining of DEFA5 (arrowheads). Scale bar, 20 μm. PF, pair-fed; AF, alcohol-fed.

Based on the fact that AMPs play an important role in modulating gut microbial homeostasis and combating dysbiosis under disease conditions, we next analyzed the effect of alcohol on PC and α-defensins. Transmission electron microscopy revealed that alcohol exposure decreased the densities and sizes of PC granules (Fig. 3C). Concomitantly reduced mRNA levels of α-defensins and protein levels of α-defensin 5 (DEFA5) were found in the AF mice compared with the control (Fig. 3D,E).

Genetic α-Defensin dysfunction promotes alcohol-induced PAMP translocation and liver injury

To assess the impact of PC α-defensin reduction on ALD, we utilized the Mmp7^−/−mouse model that is deficient in active α-defensins (MMP7 hydrolyzes pro-α-defensins to active α-defensins in mice; Fig. S4), and fed the mice with Lieber-DeCarli liquid diets for 8 wk (Fig. 4A). Compared with WT mice, Mmp7^−/− mice had fewer numbers of PC granules, which was further reduced by alcohol (Fig. 4B). Knockout of MMP7 did not affect alcohol-reduced α-defensin expression (Fig. 4C). Bacterial killing activities of isolated ileal crypts were lower in the AF groups than that in the PF groups of both WT and Mmp7^−/− mice, and the Mmp7^−/−_AF mice showed the worst ability (Fig. 4D). Accordingly, cecal microbiota was analyzed by α-diversity, β-diversity, and assigned at the family level. Statistical differences in microbial community diversity were observed by the number of observed OUTs between the AF groups and its genetic controls, and also between Mmp7^−/− mice and WT mice. The Mmp7^−/−_AF group showed the highest microbial richness (Fig. 4E). PCoA revealed 4 clearly separated clusters between Mmp7^−/− mice and WT mice fed either control or alcohol diet (ANOSIM, P ≤ 0.001, detailed in Table S1), suggesting significantly different bacterial community structures among these groups. Indeed, annotation of sequence data to the family level revealed distinct patterns of bacterial composition in each group (Fig. 4E). Deficient in active α-defensins also led to impaired intestinal barrier. Examination of the ultrastructure of the intestinal epithelium showed more severely disrupted apical junctional complex and reduced/disarranged microvilli in Mmp7^−/−_AF mice than in WT_AF mice (Fig. S5). In WT mice, alcohol feeding reduced the distribution of tight junction proteins, ZO-1 and claudin-1, whereas in Mmp7^−/− mice the impairment was even worse (Fig. 4F).

Figure 4.

Deficiency of active α-defensins exaggerates alcohol-induced intestinal antibacterial dysfunction, dysbiosis, and barrier disruption. (A) Experimental design. (B) TEM of ileal crypts of mice. Scale bar, 10 μm. (C) Relative ileal mRNA levels of α-defensins in mice (n=6). (D) Bactericidal activity of isolated small intestinal crypts against E. coli ML35. Data are presented as percentage of remaining live bacteria (n=6). (E) Alpha-diversity measurement of observed number of OTUs, PCoA plot showing dissimilarity in bacterial community structures based on Bray-Curtis distances, and barplot showing the bacterial composition at family level of the intestinal microbiome from mice cecal contents (n=8). Legend of prominent families is shown at right. (F) IF staining of ileal ZO-1 and Western blot of ileal ZO-1 and claudin-1. White arrowheads indicate disarranged ZO-1. Scale bar, 20 μm. *P<0.05, ^**P<0.01, ^***P<0.001 except 4F where letters indicate difference at P<0.05. PF, pair-fed; AF, alcohol-fed.

In accordance with the changes in the intestine, alcohol-elevated plasma and hepatic endotoxin levels were both higher in Mmp7^−/− mice than those in WT mice (Fig. 5A). In addition, we found that the levels of bacterial DNAs, including E. coli and H. hepaticus, were increased in the livers by alcohol, and further elevated by MMP7 deletion (Fig. 5A). Plasma ethanol levels and protein levels of hepatic alcohol metabolizing enzymes, CYP2E1 and ALDH2, were comparable in the two AF groups (Fig. S6). Hepatic LCN2 and CXCL1 were both induced by alcohol and synergistically elevated by deletion of MMP7 (Fig. 5B and Fig. S7A,B). After alcohol intoxication, plasma ALT levels were significantly increased in both WT and Mmp7^−/− mice compared with their genetic controls; with MMP7 deletion, the mice showed even higher values (Fig. 5C). H&E staining showed that alcohol feeding led to lipid accumulation and inflammatory cell infiltration in the liver, whereas MMP7 deletion exacerbated alcohol-induced hepatic inflammation (Fig. 5D). That observation was further confirmed by immunofluorescence staining of F4/80⁺ (Fig. 5E) and CD11b⁺ cells (Fig. S7C), and by flow cytometry analysis of F4/80⁺CD80⁺ cells (activated macrophages) and CD11b⁺Ly6G⁺ cells (neutrophils; Fig. 5F. Gating strategy in Fig. S8). The numbers of analyzed immune cells were all significantly increased by alcohol, slightly increased by MMP7 deletion per se, and further elevated when the two factors were combined. In contrast to inflammation, the two AF groups had comparable levels of hepatic lipids (Fig. S9), suggesting that α-defensin dysfunction does not affect alcoholic steatosis.

Figure 5.

Deficiency of active α-defensins promotes alcohol-induced PAMP translocation and hepatic inflammation. (A) Plasma and hepatic endotoxin levels, and relative hepatic 16S bacterial rRNAs in WT and Mmp7^−/− mice fed control or alcohol diet for 8 wk (n=8). (B) mRNA and protein levels of LCN2 and CXCL1 (n=6). (C) Plasma ALT levels (n=8). (D) H&E staining of the liver sections (n=3). Arrowheads indicate lipid accumulation, and arrows indicate inflammatory cells. Scale bar, 50 μm. (E) IF staining of hepatic F4/80⁺ cells (n=3). Scale bar, 20 μm. (F) Flow cytometry analysis of activated macrophages and neutrophils (n=3). *P<0.05, ^**P<0.01, ^***P<0.001. PF, pair-fed; AF, alcohol-fed.

Synthetic HD5 effectively reverses the detrimental effects of alcohol

Although antibiotic treatment reduced alcohol-induced hepatic inflammation, it is not an ideal option for treating ALD due to the risk of antibiotic resistance. In contrast, AMPs have been reported to effectively correct dysbiosis and improve symptoms of diseases. Therefore, we next evaluated the therapeutic potential of synthetic human α-defensin 5 (HD5) in mice ALD. WT mice were fed alcohol diet for 6 wk, and then given HD5 every other day for another 2 wk with concomitant alcohol feeding (Fig. 6A). The bioavailability of orally administrated HD5 was estimated by mass spectrometry analysis of HD5 after in vitro pepsin digestion in gastric fluid, showing 75% of intact HD5 bypassed gastric digestion (Fig. S10). HD5 treatment did not significantly change the total OTUs of cecal bacteria (Fig. 6B). However, analysis of dissimilarities in the bacterial composition revealed 2 clearly clustered groups on the PCoA plot, according to the presence or absence of HD5 (Bray-Curtis, PC1 accounting for 47.63% of the total variation, ANOSIM, R=0.93, P<0.001). HD5 effectively shifted the intestinal microbial composition as indicated at the family level (Fig. 6B). Of note, Verrucomicrobiaceae was dramatically enriched in the cecal contents of the AF+HD5 mice than in the AF mice (20.39±5.29% vs 2.03±0.69%, P<0.001). Subsequent PCR analysis confirmed that A. muciniphila, a species of Verrucomicrobiaceae family, was significantly enriched by HD5 (Fig. 6C). A. muciniphila has recently been reported to be reduced in alcoholic patients, and supplementation with it ameliorated ALD in mice.³ H&E staining of the intestine showed minimal structural changes, such as slight villous blunting, in the AF group compared with the PF group, which were improved in the AF+HD5 group (Fig. S11). Moreover, HD5 unexpectedly upregulated the expression of α-defensins, including Defa4, Defa20 and Defa21 (Fig. 6D). Alcohol-impaired distribution of ZO-1 and claudin-1 was also reversed by HD5 (Fig. 6E). Intestinal immune regulators were analyzed to decipher the observed protection. Interestingly, HD5 notably augmented the expression of intestinal Il22, and lowered the levels of inflammatory cytokines/chemokines, including Ip10, Tnfa, and Mip2 (Fig. 6F).

Figure 6.

HD5 treatment alters intestinal microbiome, and reverses alcohol-impaired intestinal bactericidal function and barrier disruption. (A) Experimental design. Green arrows indicate HD5 treatment. (B) Alpha-diversity measurement of observed number of OTUs, PCoA plot showing dissimilarities in bacterial community structures based on Bray-Curtis distances, and barplot showing the bacterial composition at family level of the intestinal microbiome from cecal contents of the AF mice with or without HD5 treatment (n=8). Legend of prominent families is shown at right. (C) Relative abundance of A. muciniphila determined by qPCR (n=4). (D) Relative mRNA levels of ileal α-defensins (n=6). (E) IF staining of ileal ZO-1 and Western blot of ZO-1 and claudin-1 (n=3). Arrowheads indicate disarranged ZO-1. Scale bar, 20 μm. Letters indicate difference at P<0.05. (F) Relative mRNA levels of Il22 and inflammatory cytokines/chemokines in mice ileum. *P<0.05, ^**P<0.01, ^***P<0.001. PF, pair-fed; AF, alcohol-fed.

PAMP translocation, including endotoxemia and increased hepatic bacterial DNAs, was effectively reversed by HD5 treatment (Fig. 7A). In line with that, alcohol-induced Lcn2 and Cxcl1 elevation (Fig. 7B,C), macrophage activation, and immune cell infiltration (Fig. 7D) were all reduced by HD5. Concomitantly, HD5 administration alleviated alcohol-induced liver injury as indicated by decreased plasma ALT levels and improved histology (Fig. 7E,F) though plasma ethanol levels and protein levels of CYP2E1 and ALDH2 remained the same (Fig. S12).

Figure 7.

HD5 treatment alleviates alcohol-induced liver damage in mice. (A) Plasma endotoxin levels and hepatic 16S bacterial rRNAs relative to β-actin (n=8). (B) mRNA levels of Lcn2 and Cxcl1 (n=6). (C) IHC staining of LCN2 and CXCL1 (n=3). (D) IF staining of hepatic F4/80⁺ and CD11b⁺ cells (n=3). (E) Plasma ALT levels (n=8). (F) H&E staining of liver sections. Scale bar, 50 μm (C,F), 20 μm (D). *P<0.05, ^**P<0.01. PF, pair-fed; AF, alcohol-fed.

Zinc deficiency critically mediates alcohol-induced intestinal α-defensin dysfunction

To determine the role of zinc in PCs, we examined the acute and chronic effects of zinc deprivation on α-defensin function. We measured the antibacterial activity of the intestinal crypts after acute zinc deprivation by dithizone.¹⁵ Dithizone treatment caused rapid PC degranulation, impaired crypt bacterial killing, and reduced expression of Defa5 and Lyz1 (Fig. 8A). We also directly chelated zinc from released crypt components (mainly AMPs) by TPEN, and determined the antibacterial activity. A similar trend of impaired bacterial killing in the zinc deprivation group was observed (Fig. 8A), suggesting direct participation of zinc in PC antibacterial activity. To validate this, the bactericidal activity of HD5 with or without zinc was determined. Both oxidized and reduced forms of HD5 effectively killed bacteria, with the reduced form being more potent (Fig. 8B). Zinc alone did not affect bacterial viability, but it significantly improved the antibacterial function of both forms of HD5. Moreover, zinc helped in stabilizing HD5_red. HD5_red was incubated overnight with or without zinc. In the absence of zinc, HD5_red all became oxidized, suggesting high vulnerability to proteolytic breakdown in vivo, whereas with zinc incubation, a majority of HD5 remained in reduced forms (Fig. 8C and Fig. S13). In terms of alcohol, we previous reported that chronic alcohol feeding significantly decreased ileal zinc levels in mice.¹⁶ Fluorescent staining of zinc showed that compared to the intestinal epithelia, the crypts had more zinc reduction after alcohol exposure (Fig. S14). To determine the effect of chronic ZnD in PC α-defensin function, WT mice were subjected to alcohol feeding along with adequate or deficient zinc (Fig. 8D). Dietary ZnD reduced PC granules, especially after alcohol intoxication (Fig. 8E). Mice fed alcohol with dietary ZnD showed the most significant reduction in Defa5 and Mmp7 mRNAs, and the worst antibacterial activity among the 4 groups (Fig. 8F). Histological examination of the small intestine revealed blunted villi, crypt hyperplasia, and reduced proliferative epithelia in the ZnD+AF mice compared with those with the ZnA+AF mice (Fig. S15A). Comparing to ZnA, ZnD also exaggerated alcohol-induced endotoxemia (Fig. S15B), increased plasma ethanol levels (without affecting protein levels of hepatic CYP2E1, ADH, and ALDH2; Fig. S16), and liver damage (Fig. S17).

Figure 8.

Zinc regulates α-defensin bactericidal activity and stability, and dietary zinc deficiency exaggerates alcohol-induced PC α-defensin dysfunction. (A) Representative image of isolated crypts, bactericidal activity, relative ileal mRNA levels of Defa5 and Lyz1 of mice treated with or without dithizone (n=6); and bactericidal activity of released AMPs from isolated small intestinal crypts after direct zinc chelation by TPEN (n=6). (B) E. coli ML35 killing of HD5 with or without zinc determined by flow cytometry (n=6). (C) Analytical nUHPLC-MS traces of HD5 with or without zinc. (D) Experimental design of chronic alcohol feeding with zinc adequate (ZnA) or zinc deficient (ZnD) diets. (E) TEM of ileal crypts of mice (n=3). Scale bar, 10 μm. (F) Relative ileal mRNA levels of Defa5 and Mmp7, and bactericidal activity of isolated small intestinal crypts of mice (n=6). *P<0.05, ^**P<0.01. PF, pair-fed; AF, alcohol-fed.

Discussion

Endotoxemia and gut barrier dysfunction in patients with ALD have been well-documented.¹⁷ A recent study showed that changes in circulating endotoxin and bacterial DNAs are associated with the severity of AH.¹⁸ Here, for the first time, we reported significantly higher levels of endotoxin and bacterial DNAs in the livers of SAH patients compared with non-alcohol-drinking healthy subjects. Moreover, we were able to define a causative link between translocated PAMPs and hepatic inflammation. We found that PAMPs, rather than alcohol, is the driving factor of hepatic LCN2 and CXCL1 elevation that observed in SAH patients and experimental models of ALD. We also found that alcohol actually mediates an immunosuppressive effect on hepatic response to bacterial pathogens, which is being increasingly recognized.^{19, 20} These findings suggest that the role of PAMPs in AH is more profound than thought before. Furthermore, we found that the major source of hepatic LCN2 is hepatocytes. Consistently, it has been reported that hepatocytes are the major cell type responsible for LCN2 production after bacterial infection or partial hepatectomy.²¹ CXCL1, on the other hand, is expressed by many cell types the liver, including hepatocytes,²² macrophages,²³ hepatic stellate cells,²⁴ and sinusoidal endothelial cells.²⁵ Studies are necessitated to further explore the molecular mechanisms of alcohol-induced LCN2 and CXCL1 induction and the outcomes. Dysbiotic changes in the intestinal microbiota are associated with various liver diseases.²⁶ In patients with ALD, colonic mucosa-associated bacteria exhibited a lower abundance of Bacteroidetes and a higher abundance of Proteobacteria.²⁷ A significant reduction of Bacteroidetes was also detected in the blood of subjects with a history of alcohol abuse compared with non-alcohol-drinking individuals.¹⁸ In our model, we found striking differences in cecal and hepatic microbiomes of the AF mice, suggesting a systemic dysbiosis status. Proteobacteria was the most abundant phylum in the liver, and was substantially increased by alcohol, whereas Firmicutes and Bacteroidetes constitute the majority in the cecal contents, and an increased ratio of the two was detected in the AF group. The distinct compositions and alcohol-mediated alterations between cecal and hepatic microbiota indicate that factors targeting microbial symbiosis are involved in PAMP translocation at the gut-liver axis during the onset and progression of ALD apart from the disrupted gut barrier.

Emerging evidence show how the intestinal microbiota composition can be altered by a deficiency in host antimicrobial ability provided by AMPs.²⁸ It has been reported that an AMP specifically targeting Gram-positive bacteria, Reg3γ, was suppressed by alcohol, and mice overexpressing Reg3γ were protected from alcohol-induced liver damage.⁴ We hypothesized that PC α-defensins may represent another key contributor to dysbiosis and barrier impairment during ALD. PC α-defensins have a broad spectrum of antimicrobial activity against Gram-positive and Gram-negative bacteria, fungi, and viruses. They contribute about 70% of PC-related antimicrobial activity.⁵ We found that alcohol reduced PC α-defensins, and determined a causative role of α-defensin dysfunction in ALD development. Deficiency of functional α-defensins led to impaired bacterial killing, dysbiosis, and a compromised epithelial barrier. It is noteworthy that Mmp7^−/− mice have an increased abundance of segmented filamentous bacteria (SFB) and more IL-17A producing CD4⁺ T cells in the intestine.²⁹ Additionally, the knockout model we used (Mmp7^−/− mouse) is not Paneth cell-specific, so non-Paneth cell and non-intestinal effects of MMP7 may interfere with the findings though MMP7 is undetectable in normal liver. Therefore, we cannot rule out the possibility that other veiled effects may also be involved in the pathogenesis of α-defensin dysfunction-mediated ALD.

We next focused on mechanisms by which zinc regulates PC α-defensin function, and synergistically induces dysbiosis as well as hepatic inflammation with alcohol when deficient. Interestingly, zinc is known to be concentrated in PCs³⁰ even though the exact mechanisms are not well-defined. Oxidized HD5 has been shown to have high zinc chelating affinity, and the complex was resistant to protease hydrolysis.³¹ Here, we found that zinc directly binds to HD5, protects it from proteolytic degradation, and more importantly, enhances its antibacterial activity. This is in agreement with another study showing that zinc has a significant effect on microbial population and diversity,³² and our findings may represent one of the underlying mechanisms. Further analysis of zinc coordination to other α-defensin family members would be valuable. In addition, functional MMP7 is bound by four metal ions, including a catalytic zinc ion and a structural zinc ion.³³ Therefore, it is reasonable to speculate that PC zinc status also affects the activation of α-defensins in mice, and that the regulation by zinc on PC α-defensins is essential and comprehensive. Furthermore, zinc deficiency has been frequently documented in clinical and experimental ALD.⁶ Although direct liver injury and intestinal barrier disruption have been analyzed in the context of zinc deficiency, our study further demonstrated that zinc deficiency also mediates alcohol-induced PC α-defensin reduction and dysbiosis. We found that plasma ethanol levels were higher in zinc deficient mice than in zinc adequate mice after alcohol feeding, which may be due to impaired ADH function as ADH is a zinc-dependent enzyme.³⁴

Alcohol-induced pathological changes in intestinal microbial compositions and host antibacterial defense lay the foundation for novel therapies targeting bacteria and bacterial-related inflammatory mediators in treating ALD. Probiotics and antibiotics have been reported to be protective against ALD.² A small pilot study involving 8 patients received fecal microbiota transplants (FMT) from healthy donors and 18 matched patients with SAH showed promising efficacy of FMT in improving intestinal dysbiosis and clinical outcomes.³⁵ The present study showed that HD5 treatment effectively altered intestinal microbiota (i.e. strikingly enriched Verrucomicrobiaceae family), and counteracted alcohol-induced liver lesions in mice. A recent study reported similar findings that HD5 increased Akkermensia sp. without decreasing microbial diversity in mice.³⁶ Another study reported that HD5 treatment attenuated intestinal injury caused by alcohol and colitis.³⁷ Beyond a direct role in microbial sensing, HD5 also exhibits immunomodulatory functions. Indeed, in the present study, we observed elevated IL-22 levels after HD5 treatment. It has been reported that HD5 can act as a chemoattractant that recruits macrophages, T lymphocytes, and mast cells.³⁸ Moreover, HD5 overexpressing mice had a dramatic loss in SFB and lamina propria IL-17-producing T cells.²⁹ Thus, it has to be taken into consideration that the mechanism of HD5-mediated protection against experimental ALD involves microbiota rehabilitation, altered immune response, and a secondary effect of the improved intestinal microbial ecology.

Taken together, the present study revealed a causative role of alcohol-impaired PC α-defensins in systemic dysbiosis, PAMP translocation, and hepatic inflammation, which could be reversed by HD5 treatment (Fig. S18). Zinc deficiency critically modulates PC α-defensin dysfunction at multiple levels, including translational regulation and post-translational modifications. These observations might usher in a new paradigm of developing novel therapies targeting intestinal AMPs for patients with AH.

Abbreviations:

AF

alcohol-fed

ALD

alcoholic liver disease

AMP

antimicrobial peptides

HD5

human α-defensin 5

LCN2

lipocalin-2

SAH

severe alcoholic hepatitis

PF

pair-fed

PAMPs

pathogen-associated molecular patterns

PC

Paneth cell