Can You Trust Your Gut? Implicating a Disrupted Intestinal Microbiome in the Progression of NAFLD/NASH

NAFLD covers a wide range of liver morbidities, with accumulation of lipid droplets being its mildest manifestation, and liver failure or cirrhosis being the worst. When the accumulation of lipid droplets exceeds 5% of the total liver weight, an individual may be characterized as having fatty liver, hepatic steatosis, or non-alcoholic fatty liver (NAFL) (1). About 30% of individuals with NAFL progress to NASH which is characterized by inflammation in addition to lipid accumulation (2). About 20% of NASH patients with advanced fibrosis will progress to cirrhosis, which marks an irreversible decline in liver function, in addition to being a risk factor for hepatocellular carcinoma (2).

NAFLD/NASH progression is hypothesized to be due to the combination of insults, termed the two-hit hypothesis (Figure 1) (3). This theory postulates that the “first hit” is the development of fatty liver. The “second hit” is characterized by a multitude of factors including inflammatory cytokines, oxidative stress and/or insulin resistance (IR), although the sequence of these events is unclear (3). Since a two-hit model did not sufficiently explain the complex pathophysiology of NAFLD, a more inclusive theory was proposed, the “multiple hit theory” (4). Fatty liver still remains the first hit, but the complex secondary insults reflect broader metabolic dysfunction that involves crosstalk with other organs central to metabolism such as adipose tissue, pancreas, and gut microbiota (4). However, the multiple hit model is still an oversimplification, and additional factors have yet to be fully explored, with age, obesity, and genetic pre-disposition being just a few of them.

On top of a complex etiology, tracking progression of NAFLD is an additional challenge. Serum ALT and fibrosis score are surrogate markers to determine liver damage; however, liver biopsy remains the gold standard for diagnosing and characterizing the different stages of NASH (5). Less invasive alternatives, such as ultrasonography and MRI, allow for visualization of fatty liver, but do not evaluate inflammation or accurately assess fibrosis (6). Limited functionality makes these techniques less viable alternatives as disease development/progression indicators. To complement imaging, biomarker research is an active area of the NAFLD/NASH field. The gut microbiome composition or associated metabolites could be one such biomarker, although additional research is needed to confirm the utility of these approaches.

Fecal microbiota transplant (FMT) studies have implicated the microbiome and NAFLD development in mice (7, 8). However, no particular microbial signature has emerged in human NAFLD, making it difficult to trace disease development back to any particular cluster of bacterial taxa. Sharpton et al. reviewed the reasons behind discordant results obtained from studies trying to draw correlations between the microbiome and human NAFLD (9). A number of confounders could underlie why no one signature has emerged across multiple studies, including differences in patient age, presence of other metabolic co-morbidities, and geographic location. Differences in handling of stool samples, sequencing and statistical analyses performed also could skew the results of individual studies (9). Additionally, compared to 16S ribosomal RNA sequencing, metagenomic analysis allows for a better understanding of the functional and metabolic potential of the gut microbiome (9).

Despite heterogeneity in specific taxa associated with disease, several cross-sectional studies have shown associations between an unhealthy change in the normal bacterial ecology, also known as dysbiosis, and all stages of NAFLD, including fatty liver, NASH, advanced fibrosis and also cirrhosis and hepatocellular carcinoma (10). Aron-Wisnewsky et al. divided human studies into steatosis to NASH, and NAFLD fibrosis to NASH cirrhosis signatures. In doing so, the authors found significant overlap in microbial signatures in both simple steatosis and NASH (10). In brief, steatosis and NASH patients have increased abundance of Proteobacteria (13.5%) at the phylum level, increased Enterobacteriaceae (12.02%) and decreased Rikenellaceae (0.41% in NASH versus 1.97% in healthy patients) and Ruminococcaceae (7.01% in NASH versus 18.82% in healthy patients) at the family levels, and increased Escherichia (2.36% versus 0.3% in healthy patients), Peptoniphilus (4.1% versus 0.36% in healthy patients) and decreased Anaerosporobacter (1.08% versus 2.02% in healthy patients), Coprococcus (1.03% versus 3.69% in healthy patients), Eubacterium (0.29% versus 1.18% in healthy patients), Faecalibacterium (4.27% versus 8.15% in healthy patients) and more discordant changes in Prevotella at the genera level (6, 11–24). The authors did acknowledge that despite these differences there is widespread divergence in the literature across all levels of taxonomy, with some studies even reporting trends opposite to the ones discussed above (10).

In contrast to simple steatosis, identification of microbial signatures in NASH with fibrosis is less well established, in part due to differences in the threshold for “fibrosis” between studies. For example, some human fibrosis studies have made comparisons between mild to moderate (F0-F2), versus severe fibrosis (F3-F4), while some others have compared no to little fibrosis (F0-F1) to moderate and severe fibrosis (F2-F4), which has created discrepancies in the literature (15, 17). Even then, microbial signatures associated with advanced fibrosis have emerged. In general, advanced fibrosis correlated with increased Gram-negative bacteria, increased Fusobacteria phylum, and decreased Enterobacteriaceae family and Gram-positive bacteria, Firmicutes phylum, Prevotellaceae family, and Prevotella genus (15, 17, 20). One of these studies utilized metagenomic sequencing along with serum metabolomics which allowed the authors to overlay bacterial abundance with pathway and metabolite enrichment data. This approach provided a more holistic microbial profile of patients with mild/moderate fibrosis, and severe fibrosis with NASH (17). While the gut microbiome signature was consistent with previous studies, serum metabolite analysis revealed increased nucleoside metabolism in severe fibrosis and increased amino acid and carbon metabolism related metabolites in mild/moderate fibrosis (16). In terms of pathway enrichment, mild/moderate fibrosis stool samples were enriched in nucleotide and steroid degradation pathways, while severe fibrosis stool samples were enriched in carbon metabolism and detoxification pathways (19). These data suggest the possibility of harnessing the microbiome to differentiate mild/moderate fibrosis from severe fibrosis with NASH. More studies with the same study design and larger cohort sizes are needed to confirm whether these microbiome-derived signatures can truly be used as a diagnostic tool.

The gut barrier is the first line of defense between intestinal luminal contents and circulation, and mostly consists of the epithelial barrier and the over-laying mucus layer. The epithelial barrier consists of a monolayer of adjacently aligned epithelial cells, a vast majority of which are enterocytes/colonocytes. This layer is also interspersed with four other epithelial cell types—goblet cells, enteroendocrine cells, Paneth cells, and microfold cells (69). Underneath the epithelial cell monolayer is the lamina propria, which houses innate and adaptive immune cells such as T cells, B cells, macrophages, and dendritic cells (70). Finally, under the lamina propria lies a vascular network that eventually converges into the portal vein which, in turn, empties into the liver.

Goblets cells are specialized mucus secreting cells embedded within the epithelial monolayer (71). Secreted mucus is composed of glycosylated mucin proteins that form a gel-like layer and sit above the epithelial monolayer (71). The small intestine (SI) and colon have very distinct physiologies (72). The SI has Immunoglobulin As (IgA) and anti-microbial peptides (AMPs) which are secreted into the mucus layer by plasma cells within the lamina propria, and Paneth cells respectively, making the SI relatively less hospitable for bacterial growth (73, 74). Compared to the SI, the colon has a significantly greater number of goblet cells, and hence more mucus. Unlike the SI, the colon has two layers of mucus, with the bottom layer sitting right above the epithelial monolayer, and is more “tight” in consistency (72). A “loose” mucus layer overlays the bottom layer. This outer mucus layer serves as a habitat for colonic gut microbes (72). Since the colon has fewer Paneth cells, there is less IgA and AMP secretion, which in combination with more mucus production and thickness, makes it a more fertile ground for gut microbes (72). Owing to these differences between the SI and colon, gut microbiome composition varies along the gastrointestinal tract (GI) as well, with more aerobic and facultative anaerobes in the duodenum and jejunum, and more fiber-fermenting, bile acid-metabolizing anaerobes in the colon.

Under the mucus layer lies the intestinal epithelial monolayer. Transport of molecules between the intestinal lumen and the underlying vascular layer is regulated by junctional complexes between epithelial cells within the monolayer (71). The three most important junctional complexes are tight junctions (TJs), adherens junctions (AJs), and gap junctions (75). TJs include proteins like zona occludin-1 (ZO-1), occludin, and members of the claudin family which seal intercellular space. AJs are found below TJs, and along with gap junctions, they help maintain the integrity of the epithelial monolayer and facilitate cell-cell communication. Intracellularly, TJs and AJs are attached to actin and myosin, thereby playing important roles in cytoskeletal dynamics. It should be noted that the gut barrier is not a static organ, but is actually rather dynamic and sensitive to changes occurring in the gut (75).

Since the gut barrier serves to keep intestinal luminal contents from entering the underlying vascular network, any disruption in its integrity leads to a condition called the “leaky gut”. Under certain conditions, expression of TJPs is reduced leading to increased permeability between adjacent epithelial cells. Increased paracellular permeability gives luminal contents access to the underlying lamina propria and vascular network. Leakage of bacterial antigens into the vascular network, portal vein and liver leads to increased hepatic inflammation due to activation of immune signaling (76–78) (Figure 3). Indeed, metabolic diseases are often associated with a loss of intestinal barrier function and an increase in passive transport of microbial pathogen associated molecular patterns (PAMPs) into the body (79) (Figure 3). Emerging evidence suggests a link between a dysfunctional gut barrier and human NAFLD (80–82). A meta-analysis based on five clinical studies demonstrates a linear relationship between increased gut permeability and NAFLD progression, with stronger correlations as disease severity increases (82). Specifically, 39.1% of NAFLD patients had displayed increased intestinal permeability versus only 6.8% of HCs. In addition, it was found that patients with NASH were more likely to have this phenotype with the incidence of gut permeability in this subgroup being 49.2% higher compared to patients with NAFLD as a whole. These data suggest that inflammatory events occurring in the pathophysiology of NASH might be a function of increased gut permeability.

To demonstrate that increased intestinal permeability precedes NASH, Mouries et al. showed that intestinal epithelial barrier (IEB) disruption in mice occurs within 48 h of HFD-feeding (83). This was evidenced by reduced expression of ZO-1 and increased bacterial translocation into the ileum and cecum lamina propria. Plasmalemma vesicle-associated protein 1 (PV1), a marker for gut vascular barrier (GVB) damage was unchanged at 48 h. Following 1 week of HFD-feeding, PV1 expression and intestinal permeability were significantly upregulated along all sections of the gut, and stayed that way until the end of the 24-week study. In both 1-week and 6-week HFD-fed mice, disruptions in the IEB and GVB preceded signs of hepatic steatosis and IR, indicating that these are early events in the development of NASH. Impaired GVB is then maintained through development of IR and inflammatory NASH. To further support the hypothesis that a dysbiotic gut disrupts epithelial barrier integrity, when SPF mice were transplanted with fecal matter from control diet and HFD-fed mice, mice receiving FMT from HFD-fed mice had increased adipose mass and expression of PV1, suggesting that HFD-feeding induces dysbiosis, which disrupts the GVB, which, in turn, correlates with increased intestinal blood vessel permeability. Collectively, these data suggest a linear sequence of events- IEB disruption, GVB disruption, IR and hepatic steatosis, and finally NASH (83).

In contrast to the above study, Thaiss et al. absolved the gut microbiome of any culpability in IR-driven gut permeability (84). In addition to showing increased gut permeability in db/db and ob/ob mice, the authors were able to show similar gut barrier dysfunction in STZ-treated mice, therefore demonstrating that IR-driven gut barrier perturbations are associated with, but do not require obesity. To determine the consequence of barrier dysfunction, the authors used a bioluminescent variant of Citrobacter Rodentium to track infection in vivo, which mimics human enteropathogenic E. coli infections. In addition to being hyperglycemic, STZ-treated mice exhibited reduced expression of ZO-1 along with increased gut permeability. Upon receiving C. rodentium, these mice showed increased susceptibility to infection and systemic translocation, enhanced bacterial growth, epithelial adherence and systemic spread. To determine whether these gut dysfunction signals were arising from microbiome alterations upon STZ treatment, FMTs were performed with feces from STZ-untreated and treated mice. No gut barrier dysfunction and associated increase in bacterial infection was seen in mice receiving FMTs from STZ-treated mice, demonstrating that the gut barrier dysfunction observed in these mice is independent of the microbiome. RNA sequencing of IECs revealed global reprogramming of the epithelial transcriptome of STZ-treated mice. In particular, it was found that the transcription of the GLUT2 gene which is responsible for glucose uptake in IECs was significantly upregulated in STZ-treated mice. IEC specific GLUT2 knockouts did not show increased permeability, reduced TJPs or increased susceptibility to infection upon STZ treatment. Taken together, these data suggested that IR-driven gut barrier dysfunction is independent of changes in the gut microbiome, and instead is dependent on GLUT2 dependent signaling in IECs (84).

A key feature that sets the above study apart from the work of Mouries et al. is the animal model used. STZ injection is typically representative of Type 1 (T1D), while ob/ob, db/db, and HFD-feeding models are more representative of Type 2 diabetes (T2D. In humans however, T1D, much like T2D is often accompanied by the presence of metabolic syndrome, thereby making it challenging to investigate microbiome-independent mechanisms behind gut barrier dysfunction in human T1D (85). In conclusion, in mouse models of diabetes, gut barrier dysfunction in T1D is driven by GLUT2 signaling in IECs and in T2D is driven by disruptions in the gut microbiome, and precedes the development of NASH.

Much like NAFLD/NASH, diabetes also has been linked with gut dysbiosis and barrier dysfunction. For example, examination of 345 patients microbiome samples demonstrated a reduction in butyrate producers and increase in opportunistic pathogens in the diabetic microbiome (86). Another study confirmed a significant reduction in the population of Bifidobacteria and Verrucomicrobia (87). Specifically, Akkermansia muciniphila of the Verrucomicrobia phylum has been shown to be significantly reduced in diabetes in both mouse and human studies (87–89). Indeed, oral administration of A. muciniphila resulted in marked improvements in metabolic parameters in genetic and diet-induced models of diabetes, positioning it as a beneficial microbe (89). It is most abundantly found in the loose outer mucus layer of the colon, and uses polysaccharides in the mucus as substrates to generate SCFAs like acetate and propionate (90). A. muciniphila supplementation was reported to restore the colonic mucus layer to its normal thickness in HFD-fed mice, the mechanisms behind which remain unclear (89). In another report, pasteurized A. muciniphila and Amuc_1100, the pili protein in A. muciniphila, were shown to improve gut barrier integrity by upregulating the expression of some TJ proteins (91). Perhaps, most strikingly, A. muciniphila supplementation was shown to significantly reduce circulating LPS levels, which suggests that it lead to improvements in gut barrier integrity (89, 91). In the strongest case yet for using A. muciniphila supplementation as therapy for gut-related and hepatic pathologies, a small exploratory proof-of-concept study was conducted on 40 obese male and female individuals (92). Participants were divided into three groups- placebo, pasteurized A. muciniphila and live A. muciniphila treated groups. At the 3-week end-point, in addition to demonstrating no adverse responses associated with A. muciniphila supplementation, the group receiving pasteurized A. muciniphila had modest, yet statistically significant, reductions in circulating LPS, AST and ALT levels (92). Although more studies with larger patient cohorts are required to confirm these findings, the use of A. muciniphila as a therapeutic agent still holds promise.

Another important function of A. muciniphila is its ability to support the growth of butyrate producing bacteria by a method known as cross-feeding (93). More specifically, in using mucus as a substrate, A. muciniphila produces the SCFA’s acetate and propionate, which are, in turn, utilized by bacteria such as Faecalibacterium prausnitzii, which produce butyrate (93). Much like A. muciniphila, loss of F. prausnitzii also correlates with development of T2D (94) (Figure 2). By producing butyrate, F. prausnitzii enhances mitochondrial function in colonocytes, thereby stabilizing HIF-1α in the gut (95). HIF-1α although considered “bad” in other contexts, has been shown to improve gut barrier integrity through yet unclear mechanisms. In addition to maintaining hypoxic conditions in the gut, butyrate supplementation to colonocyte and epithelial cell lines has led to increased transepithelial resistance (TEER), marking improved barrier function (96). Most importantly, one study found that F. prausnitzii supplementation in HFD-fed mice led to a significant reduction in diet induced steatosis, ALT and AST levels, thereby suggesting that increased butyrate production improves barrier integrity and consequently improves NASH (97).

While the above data paints A. muciniphila as a good player, another study showed that consuming diets depleted of fiber led to significant proliferation of A. muciniphila, correlating with a significant reduction in colonic mucus thickness and a compromised gut barrier (28, 98). While it is unclear why A. muciniphila appears to be a negative component of the microbiome in this study, it is possible that consuming a diet low in fiber deprives A. muciniphila and associated cross-feeders of classic substrates. A known mucus degrader, it is possible that A. muciniphila instead shifts its metabolism to use mucus as a substrate, thereby feeding into a cycle of mucus consumption, reduction in mucus thickness and increased proliferation. Eventually, when mucus consumption exceeds production, a scarcity of substrate availability results, and the population of A. muciniphila declines. It is possible then, that reduced A. muciniphila population size in diabetic patients is a consequence, rather than cause of compromised gut integrity. This might also be the reason why pasteurized forms of A. muciniphila show an improvement in metabolic endpoints, because this form does not have mucus degrading activity. A prospective study where A. muciniphila populations are measured from the onset to full- fledged development of diabetes might be able to answer some of these questions, but until then, the jury is out on the role of A. muciniphila in gut barrier integrity.

The previous section elucidates how gut microbiome dysbiosis occurring during metabolic syndromes can alter intestinal biology to make the gut more permeable. This allows passive transport of microbial PAMPs from the intestinal lumen into the portal vein, and eventually the liver (Figure 2). Before diving into how PAMPs contribute to NASH, it is first important to appreciate its pathophysiology. As mentioned, the first step of NAFLD is almost always the development of fatty liver. The second step involves multiple hits like IR, increased gut permeability, inflammation, and reactive oxygen species (ROS) production which leads to the progression of a more inflammatory, fibrotic NASH phenotype. Progression of fatty liver to fibrosis affects all liver cell types (99). While hepatocytes appear injured and undergo a form of cell death termed apoptosis, the resident liver macrophages, Kupffer cells (KCs), start secreting pro-inflammatory chemokines and cytokines. Finally, quiescent stellate cells (SCs), which are the major storage site for retinoids, are activated (99). Activation of stellate cells leads to loss of retinoids and increased expression of signaling receptors including the transforming growth factor β (TGF-β) receptor. Activated SCs proliferate and secrete extracellular matrix proteins to form a fibrous scar, which imparts a “fibrotic” phenotype to NASH (99). PAMP receptors such as the toll-like and nod-like receptors (TLRs and NLRs) are expressed on the cell surface of hepatocytes, Kupffer cells (KCs), and stellate cells (SCs), and are known to contribute to the inflammatory and fibrotic phenotype of NASH (Figure 2). While suppressed in healthy liver, TLR signaling is activated in the presence of pathogenic microorganisms and bacteria-derived molecules. Since other reports have already reviewed the role of TLRs in NAFLD in great detail, this section will briefly highlight some of the major findings (100, 101).

Of all the TLRs, TLR2, -4, -5, and -9 have been shown to contribute to the inflammatory and fibrotic signaling that characterizes NASH. In hepatocytes, LPS binding to TLR4 recruits MyD88, an adapter protein, which, in turn, leads to the activation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and mitogen-activated protein kinase (MAPK) signaling pathways (102). NF-κB is a transcription factor which upregulates the transcription of pro-inflammatory cytokines including interleukin (IL)-1, 2, 6, and 8 (103). In addition to its role in increasing hepatocyte inflammation, TLR4 plays a role in KC and SC crosstalk. LPS binding to TLR4 in SCs leads to increased production of adhesion molecules and chemokines like vascular cell adhesion protein (VCAM) and methyl-accepting chemotaxis protein (MCP) (104, 105). Adhesion molecules attract KCs and these recruited KCs secrete the pro-fibrogenic TGF-β which binds to TGF-β receptors on SCs (104, 105). This stimulates the secretion of collagen from SCs into hepatocytes, marking the beginning of liver fibrosis. Indeed, KC-specific knockdown of TLR4 in mice on a methionine choline-deficient (MCD) diet led to a significant reduction in hepatic TGs, reduced expression of inflammatory and fibrosis markers, and a resultant reduction in histological markers of NASH (106). Similar findings, demonstrating increased TLR4 mediated signaling contributing to NASH development, have been reported by other groups (106, 107).

In addition to TLR4, other TLRs mentioned in the paragraph above also have been implicated in NAFLD, but there are only a handful of reports elucidating their roles. TLR2 is expressed by HSCs and KCs and is a receptor for bacterial peptidoglycan. To investigate the role of TLR2 in hepatic inflammation, Miura et al. treated KCs with a synthetic TLR2 ligand Pam₃CK_4, and an endogenous ligand palmitic acid (PA) (108). While priming with Pam₃CK₄ alone was enough to increase the expression of NLRP3, IL-1β and IL-1α, caspase-1 activity was only induced when KCs were treated with PA after priming with Pam₃CK₃ first, indicating that both signals were required for activation of the inflammasome complex (108). This further led to the cleavage of the pro-inflammatory cytokines IL-1β and IL-1α to their active form, thereby upregulating hepatic inflammation (108). On the other hand, knock-out of TLR2 has yielded conflicting results in different mouse models, with the more conventional metabolic syndrome models suggesting that loss of TLR2 is protective against NASH (108–110). TLR5 is expressed in hepatocytes, and is a receptor for bacterial flagellin. While the exact role of TLR5 in NAFLD remains unknown, two separate studies have shown that knock-down of TLR5 accelerates hepatic steatosis, susceptibility to liver injury and NASH, thereby assigning it a more protective rather than harmful role (111, 112). Finally, TLR9 is expressed in Kupffer cells, and is a receptor for bacterial DNA. TLR9 activation signals through NF-κB to increase the expression of the cytokine IL-1β from KCs, and induces chemotaxis of macrophages and neutrophils, thereby leading to hepatic steatosis, inflammation and fibrosis (113, 114).

In conclusion, with the exception of TLR5, hepatic TLRs, upon binding by gut bacteria-derived products, set into motion a cascade of inflammatory and fibrotic signals, thereby abetting the progression of fatty liver to NASH.