Microbial Consortia-Immunoglobulin G Glycosylation Combination Scores Identify Severe Hepatic Steatosis and Significant Fibrosis in Patients with Metabolic Dysfunction-Associated Steatotic Liver Disease

Abstract

INTRODUCTION:

Severe hepatic steatosis (HS) and significant hepatic fibrosis indicate severity of metabolic dysfunction-associated steatotic liver disease (MASLD), and the models for early detection are important. Recent studies suggest immunoglobulin G (IgG) glycosylation and microbial consortia may involve in the development of MASLD. This study investigates the utility of IgG glycosylation and microbial consortia in early detection models.

METHODS:

A total of 111 patients with MASLD categorized into high vs low controlled attenuation parameter (for HS) or high vs low FIB-4 (for fibrosis). We analyzed differences in microbial consortia and IgG1/IgG2 glycosylation features, creating scores for identifying severe HS and significant hepatic fibrosis. The combination scores for identifying severe HS and significant hepatic fibrosis in MASLD patients were derived based on the microbial consortia and IgG1/IgG2 glycosylation features accordingly.

RESULTS:

Megamonas was linked to severe HS, and Christensenellaceae R-7, UCG-005, CAG-56, and Lachnospira were associated with significant hepatic fibrosis in patients with MASLD. The areas under curve of severe HS- and significant hepatic fibrosis-microbial scores were 0.731 and 0.773, respectively. The ratios of IgG2-N4H5F1/IgG1-N4H5F1 and bisected-IgG2/bisected-IgG1 predicted severe HS, while IgG2-N4H3F1/IgG1-N4H3F1, bisected-IgG1/bisected-IgG2, and IgG1-galactosylation/IgG2-galactosylation index identified significant hepatic fibrosis in MASLD. The combination scores including microbial consortia and glycosylation features achieved areas under curve of 0.776 and 0.810, showing comparable performance with established indices like HS index and FibroAST score.

DISCUSSION:

The study highlights the effectiveness of combining fecal microbiota and serum IgG glycosylation features in identifying MASLD severity, which are noninferior to the current regularly used models, and suggests the intervention of glycosylation as a promising therapeutic approach for MASLD.

INTRODUCTION

Metabolic dysfunction-associated steatotic liver disease (MASLD) is currently the fastest-growing cause of hepatocellular carcinoma (HCC) among liver transplant recipients and candidates on the waiting list in the United States (1,2). Liver biopsy is the gold standard for assessing hepatic fibrosis; however, biopsy has some limitations such as invasiveness, intraobserver/interobserver variability, severe procedure-related complications, sampling error and difficult to repeat during follow-up (3). Several noninvasive models have been established to evaluate severe hepatic steatosis (HS) and significant hepatic fibrosis, including the widely used hepatic steatosis index (HSI) and FibroScan-AST (FAST) score. However, both models incorporate aspartate transaminase (AST), which can fluctuate because of various treatments and may originate from extrahepatic tissues in patients with MASLD with multiple comorbidities. In addition, the application of FibroScan is operator-dependent and requires specific equipment (4,5). Thus, there is a pressing need for novel assessing models that do not rely on easily changeable and operator-dependent tools.

Recent research has highlighted the relationship between intestinal microbiota and chronic liver diseases (CLDs). Although several diversity indices have been reported to identify the likelihood of MASLD with cirrhosis, assessing models incorporating specific gut microbiota species for MASLD progression remain limited (6). One model including 25 variables, such as serum biochemistry data, gut metabolites, immunologic proteins, and microbiota, has shown promise in predicting HS. However, the complexity and high costs associated with these labile substances hinders clinical application (7). Conversely, intestinal microbiota consortia have emerged as effective models for the early detection of HCC, outperforming traditional alpha-fetoprotein (8). Thus, evaluating whether intestinal microbial consortia can serve as models for identifying MASLD progression is warranted, particularly in comparison with widely used models.

The post-translational modification of immunoglobulin G (IgG) through N-glycosylation has garnered attention in CLDs because the liver is the primary organ for glycoprotein synthesis, leading to altered glycosylation patterns (9). Previous studies have identified specific glycosylation features of total IgG as predictive models for MASLD (10). Furthermore, glycan analysis of IgG subclasses has shown potential for predicting liver fibrosis in CLDs (11). Combining IgG glycosylation features with serum alpha-fetoprotein has improved HCC detection (12). However, investigations into the IgG subclass glycosylation features in relation to severe HS or significant fibrosis in MASLD remain limited.

Both gut dysbiosis and alterations in IgG glycosylation play crucial roles in the development and progression of chronic inflammatory diseases (13). Previous studies have reported associations between microbiota and specific IgG glycosylation features in the general population (14). Changes in intestinal microbial consortia can modulate systemic IgG responses, influencing the proportions of IgG subclasses toward either proinflammatory or anti-inflammatory reactions (15,16). However, the direct links between microbiota and specific IgG glycosylation features in CLDs are scarce. This study aims to develop combined models using IgG glycosylation features and microbial consortia to identify severe HS and significant hepatic fibrosis in MASLD.

METHODS

Study design

This cross-sectional study enrolled individuals older than 20 who underwent abdominal ultrasound for suspected fatty liver between May 2023 and October 2024 at an outpatient department. Participants included those with or without MASLD, defined as histologically or radiologically confirmed HS with at least 1 cardiometabolic risk factor: (i) body mass index ≥23 kg/m² or waist circumference >90 cm in men and >80 cm in women; (ii) fasting serum glucose ≥100 mg/dL, hemoglobin A1c ≥5.7%, or specific drug treatment; (iii) blood pressure ≥130/85 mm Hg or specific drug treatment; (iv) triglycerides ≥150 mg/dL or specific drug treatment; (v) high-density lipoprotein cholesterol <40 mg/dL for men and <50 mg/dL for women (17). Exclusion criteria included recent treatment with immunomodulators, cytotoxic agents, chemotherapy, steroids, or antibiotics within 28 days and unstable conditions that could hinder accurate measurements. All analyses were conducted in a secure environment, adhering to the Declaration of Helsinki under protocol code 2022-12-010BCF#2, approved by the Research Ethics Committee of VGHTPE.

Data collection

Participant medical histories were retrieved from medical records. Body mass index was calculated based on height and weight measurements. HS was assessed using transient elastography (VCTE, FibroScan 630 Expert), performed by experienced gastroenterologists, with controlled attenuation parameter (CAP) values of ≥292 dB/m indicating severe HS (18,19). After these measurements, blood samples were collected for biochemical analysis, including IgG glycosylation feature identification. FIB-4 index was calculated as (age [year] × AST [U/L])/(platelet [10⁹/L]) × (alanine aminotransferase [U/L] 1/2), of which ≥1.3 indicating high risk of significant hepatic fibrosis (20).

Method of the purification and measurement of glycosylation features of IgG

Total IgG was purified from plasma samples using recombinant Protein A beads (21). The filter-aided sample preparation method facilitated in-solution protease digestion for glycopeptide analysis (22). Peptide samples were subsequently cleaned using C18 ziptips, dried in a SpeedVac evaporator, and stored at −20°C until analysis with liquid chromatography/mass spectrometry (MS). NanoLC−nanoESI-MS/MS analysis was performed on a Thermo UltiMate 3000 RSLCnano system connected to a Thermo Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Glycan structures associated with glycopeptides were identified using Byonic software version 4.2.10 (Protein Metric), with results manually verified. For protein identification, raw files were analyzed with MaxQuant (version 2.5.2.0), using the integrated Andromeda search engine. MS/MS spectra were compared against the UniProt Human database (UP000005640), with specific search parameters set for enzymatic digestion and modifications (23). The detailed protocols were mentioned in described in Supplementary Material and Methods (Supplementary Digital Content 1, http://links.lww.com/CTG/B431).

Measurement of gut microbiota

Fecal samples were processed by The PureLink Microbiome DNA Purification kit (Invitrogen, Carlsbad, MA) according to manufacturer's instructions. The V3-V4 region of the 16S rRNA gene was amplified, and raw data were analyzed using the QIIME 2 platform (24). Taxonomic alignment was performed on the amplicon sequence variants using the SILVA database version 138.1 (25,26). Taxonomic abundances were calculated through the phyloseq and vegan packages, and differential taxa analysis was conducted with the ANCOM-BC2 test using the microbiomeMarker package (27,28). The detailed protocols were mentioned in described in Supplementary Material and Methods (Supplementary Digital Content 1, http://links.lww.com/CTG/B431).

Intercorrelations between IgG1/IgG2 glycosylation features and their correlations with microbiota consortia

The association between microbial consortia and IgG1/IgG2 glycosylation features was examined, differentiating between low (non-severe HS)/high CAP (severe HS) groups, as well as low/high FIB-4 groups. Intercorrelations among various IgG1 and IgG2 glycosylation features were analyzed. Subsequently, the consistently positively correlated pairs of IgG1/IgG2 glycosylation features across the defined groups were compared and evaluated to identify severe HS or significant hepatic fibrosis, using receiver operating characteristic (ROC) analyses. Correlation heatmaps were generated using Prism 8.3.0 (GraphPad Software, San Diego, CA).

Establishment of models for identifying severe hs and significant hepatic fibrosis based on microbiota consortia and IgG glycosylation features

To establish models for identifying severe HS, we developed a scoring system and evaluated its classification ability using ROC curves. We first identified significant differences in the abundance of fecal microbial consortia at the genus level between low/high CAP and low/high FIB-4 groups. Cutoff values for each dominant fecal bacterium, associated with either nonsevere or severe HS, were determined through ROC analysis using the Youden index. For each bacterium linked to severe HS (dominant in the high CAP group), a score of 1 was assigned if its abundance exceeded the cutoff. Conversely, a score of 1 was assigned for nonsevere HS-related bacteria (dominant in the low CAP group) if their abundance fell below the cutoff. The severe HS-microbial consortia score was calculated by summing the scores of severe and nonsevere HS-related bacteria for each individual with MASLD, followed by ROC curve analysis. Similarly, a significant hepatic fibrosis-microbial consortia score was established through the summation of points from significant and nonsignificant hepatic fibrosis-related bacteria and evaluated for identifying MASLD with significant hepatic fibrosis. We further integrated microbiota scores and IgG subclass glycosylation ratios into combination scores to assess their ability for identifying severe HS or significant hepatic fibrosis in MASLD.

Statistics analysis

Continuous variables were expressed as means ± SDs, while categorical variables were reported as counts (percentages). Statistical significance was assessed using a P value threshold of <0.05, applying Student paired t test for continuous variables and χ² tests for categorical variables. Age and sex were controlled using multivariate linear regression. Analyses were conducted with SPSS 25.0 (IBM, Armonk, NY).

RESULTS

Demographic, clinical characteristics, and fecal microbial consortia in individuals with MASLD

Among the 164 individuals, 111 patients with MASLD and 29 non-MASLD individuals were identified for further analysis (Supplementary Figure 1, Supplementary Digital Content 2, http://links.lww.com/CTG/B425). Demographic information and metabolic parameters are summarized in Supplementary Table S1 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432). Among patients with MASLD, stratification by CAP levels indicated that the high CAP group displayed significantly elevated triglycerides and decreased albumin levels (Supplementary Table S2, Supplementary Digital Content 3, http://links.lww.com/CTG/B432). Conversely, classification based on the FIB-4 index revealed that the high FIB-4 group had increased levels of AST, hemoglobin A1c, and a higher prevalence of diabetes mellitus (Supplementary Table S2, Supplementary Digital Content 3, http://links.lww.com/CTG/B432). No significant differences were noted in age, sex, or other metabolic parameters between low and high CAP or FIB-4 groups.

Differences of microbial consortia in low/high CAP or low/high FIB-4 groups and microbial consortia scores as prediction models

The study found that the abundances of 12 bacterial species were significantly higher in patients with MASLD compared with non-MASLD individuals, as detailed in Supplementary Figure 2 (Supplementary Digital Content 4, http://links.lww.com/CTG/B426). Notably, among patients with MASLD, species such as Eubacterium eligens, Clostridia UCG-014, and others were significantly more abundant in patients with low CAP (non-severe HS), with a log fold change greater than 0.5. These species were categorized as non-severe HS-related bacteria. By contrast, Megamonas exhibited a dominant abundance in patients with MASLD with high CAP (severe HS) and was classified as a severe HS-related bacterium. The cutoff values for each bacterium are presented in Supplementary Table S3 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432). A microbial consortia score, predictive of severe HS, was established by summing points from both severe and nonsevere HS-related bacteria. ROC analysis indicated an area under curve (AUC) of 0.731 for predicting severe HS, with a sensitivity of 0.855 and specificity of 0.571 (P < 0.001, Figure 1b).

Figure 1.

Differences in the fecal microbial abundance at the genus level and derivation of prediction models for severe HS and significant fibrosis based on the microbial consortia scores. The differential abundances, represented as log fold changes, of statistically different genera in fecal microbial consortia were examined between (a) low and high CAP groups and (c) low and high FIB-4 groups using the ANCOM-BC test. Prediction models based on microbial consortia scores for (b) severe HS (high CAP) and (d) significant hepatic fibrosis (high FIB-4) were evaluated using ROC analyses. For each patient with MASLD, a severe HS-microbial consortia score was established by the summation of points from severe and nonsevere HS-related bacteria. CAP, controlled attenuation parameter; Gal index, galactosylation index; HS, hepatic steatosis; MASLD, metabolic dysfunction-associated steatotic liver disease; ROC, receiver operating characteristic.

In hepatic fibrosis, Weisella was predominantly abundant in patients with MASLD with low FIB-4 (indicating nonsignificant fibrosis), while Christensenellaceae R-7, UCG-005, and others were more prevalent in those with high FIB-4 (indicating significant fibrosis) (Figure 1c). The cutoff values for individual bacteria are shown in Supplementary Table S3 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432). A significant hepatic fibrosis-microbial consortia score was created by summing points from both nonsignificant and significant fibrosis-related bacteria. ROC analysis for predicting significant hepatic fibrosis revealed an AUC of 0.773, with a sensitivity of 0.538 and specificity of 0.864 (P < 0.001, Figure 1d).

Differences of IgG subclass glycosylation features and correlations to microbial consortia in low/high CAP or low/high FIB-4 groups

Supplementary Table S4 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432) summarizes IgG1/2 glycosylation features among non-MASLD and individuals with MASLD. Notably in Supplementary Table S4 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432), there are no differences in the glycosylation features between MASLD and non-MASLD groups. These findings suggested that there is no role of changes of IgG glycosylation features in the poor glycemia/lipid profiles of MASLD. After adjusting for age and sex, there are differences of IgG1/2 glycosylation features among low and high CAP groups in patients with MASLD (Table 1). In addition, patients with low and high FIB-4 levels showed differences in certain IgG1/2 glycosylation features (Table 1).

Table 1.

Serum IgG glycosylation features between either low (<292)/high CAP (≥292) or low (<1.3)/high FIB-4 (≥1.3) groups of patients with MASLD

	MASLD, CAP<292 (n = 41)	MASLD, CAP ≥ 292 (n = 70)	P valuea	MASLD, FIB-4 < 1.3 (n = 59)	MASLD, FIB-4 ≥ 1.3 (n = 52)	P valuea
[IgG1]/[IgG] (%)	40.56 ± 10.02	39.90 ± 8.11	0.715	41.59 ± 9.20	39.07 ± 8.69	0.294
IgG1 level	7.29 ± 3.79	7.78 ± 4.15	0.324	7.33 ± 3.60	7.91 ± 4.44	0.637
IgG1 glycosylation features (%)
IgG1-N3H3F1	0.35 ± 0.23	0.40 ± 0.23	0.382	0.44 ± 0.23	0.34 ± 0.22	0.036
IgG1-N4H3	0.90 ± 0.76	1.03 ± 0.87	0.466	1.09 ± 0.74	0.87 ± 0.85	0.25
IgG1-N4H3F1	18.73 ± 4.82	21.43 ± 5.77	0.037b	21.25 ± 4.59	19.89 ± 6.08	0.650
IgG1-N4H4	2.62 ± 1.93	2.36 ± 1.46	0.343	2.45 ± 1.17	2.41 ± 1.90	0.771
IgG1-N4H4F1	35.55 ± 7.07	36.36 ± 3.29	0.273	34.65 ± 5.79	37.46 ± 3.70	0.040b
IgG1-N4H5	3.06 ± 1.33	2.54 ± 0.98	0.133	2.59 ± 0.83	2.81 ± 1.32	0.587
IgG1-N4H5F1	22.91 ± 3.59	21.11 ± 4.45	0.036b	20.88 ± 4.01	22.32 ± 4.20	0.268
IgG1-N5H4	0.98 ± 0.70	0.92 ± 0.56	0.348	1.02 ± 0.48	0.86 ± 0.68	0.343
IgG1-N5H4F1	11.83 ± 2.55	11.88 ± 2.45	0.304	12.67 ± 2.35	11.18 ± 2.41	0.041b
IgG1-N5H5F1	1.91 ± 0.77	1.89 ± 0.78	0.770	2.08 ± 0.68	1.77 ± 0.82	0.164
IgG1-N5H5F1S1	0.11 ± 0.08	0.09 ± 0.05	0.467	0.11 ± 0.06	0.09 ± 0.06	0.810
Bi-antennary IgG1	84.12 ± 7.07	85.23 ± 3.33	0.244	83.36 ± 6.03	86.10 ± 3.24	0.043b
Bisected IgG1	14.84 ± 3.38	14.77 ± 3.33	0.219	15.88 ± 3.13	13.90 ± 3.24	0.038b
IgG1-Gal− index	1.07 ± 0.10	1.03 ± 0.11	0.515	1.02 ± 0.10	1.06 ± 0.11	0.770
IgG1-Fuc− index	0.91 ± 0.08	0.93 ± 0.04	0.361	0.92 ± 0.06	0.93 ± 0.04	0.582
IgG1-Sia− index	0.001 ± 0.0008	0.0009 ± 0.0005	0.867	0.001 ± 0.0006	0.001 ± 0.0003	0.881
[IgG2]/[IgG] (%)	52.48 ± 11.22	52.85 ± 8.10	0.846	53.59 ± 9.54	51.32 ± 9.07	0.202
IgG2 level	9.70 ± 5.74	10.20 ± 5.38	0.470	10.43 ± 5.77	9.55 ± 5.19	0.391
IgG2 glycosylation features (%)
IgG2-N1H1	0.30 ± 0.68	0.16 ± 0.66	0.330	0.20 ± 0.85	0.09 ± 0.18	0.061
IgG2-N3H3F1	0.02 ± 0.02	0.04 ± 0.03	0.015b	0.03 ± 0.02	0.04 ± 0.02	0.046b
IgG2-N3H4F1	0.02 ± 0.02	0.02 ± 0.01	0.369	0.02 ± 0.01	0.02 ± 0.01	0.738
IgG2-N3H5F1S1	0.04 ± 0.06	0.04 ± 0.03	0.743	0.04 ± 0.06	0.03 ± 0.03	0.385
IgG2-N3H6F1S1	0.07 ± 0.09	0.08 ± 0.08	0.579	0.08 ± 0.09	0.06 ± 0.06	0.373
IgG2-N4H3	0.16 ± 0.23	0.14 ± 0.22	0.640	0.12 ± 0.16	0.18 ± 0.27	0.372
IgG2-N4H3F1	28.33 ± 6.17	32.16 ± 6.22	0.049b	29.35 ± 6.63	32.46 ± 6.22	0.042b
IgG2-N4H4F1	41.83 ± 5.89	40.05 ± 2.56	0.047b	41.58 ± 4.96	39.46 ± 2.56	0.048b
IgG2-N4H5F1	20.57 ± 3.64	18.51 ± 4.05	0.031b	20.19 ± 3.88	18.19 ± 4.34	0.022b
IgG2-N4H5F2S1	0.013 ± 0.015	0.008 ± 0.010	0.283	0.011 ± 0.014	0.009 ± 0.012	0.049b
IgG2-N5H4F1	6.96 ± 3.08	7.24 ± 2.17	0.101	6.70 ± 2.84	7.77 ± 1.89	0.035b
IgG2-N5H5F1	1.57 ± 0.84	1.48 ± 0.60	0.635	1.50 ± 0.78	1.58 ± 0.58	0.538
IgG2-N5H5F1S1	0.11 ± 0.12	0.07 ± 0.05	0.306	0.08 ± 0.07	0.09 ± 0.09	0.535
Bi-antennary IgG2	91.05 ± 3.97	91.03 ± 2.86	0.490	91.42 ± 3.77	90.45 ± 2.40	0.311
Bisected IgG2	8.63 ± 3.90	8.79 ± 2.73	0.494	8.28 ± 3.61	9.44 ± 2.39	0.047b
IgG2-Gal− index	0.92 ± 0.10	0.87 ± 0.11	0.042b	0.91 ± 0.11	0.87 ± 0.11	0.027b
IgG2-Fuc− index	1.00 ± 0.003	1.00 ± 0.002	0.582	1.00 ± 0.002	1.00 ± 0.003	0.433
IgG2-Sia− index	0.002 ± 0.002	0.002 ± 0.001	0.359	0.002 ± 0.003	0.002 ± 0.002	0.731

Continuous variables were expressed as mean ± SD.

CAP, controlled attenuation parameter; Fuc− index, fucosylation index; Gal− index, galactosylation index; IgG, immunoglobulin G; MASLD, metabolic dysfunction–associated steatotic liver disease; Sia− index, sialylation index.

^aP value was adjusted by age and sex.

^bP < 0.05 indicated statistically significant.

In with patients MASLD with low CAP, 5 positive correlations were identified between nonsevere HS-related glycosylation features and microbial consortia, specifically Lachnospiraceae ND3007 and Monoglobus (Figure 2a; Supplementary Figure 3, Supplementary Digital Content 5, http://links.lww.com/CTG/B427). Conversely, 3 negative correlations were noted between Lachnospiraceae ND3007 and Monoglobus and severe HS-related glycosylation features. Notably, Megamonas lacked correlation with severe HS-related glycosylation features in the low CAP group, though it positively correlated with several features in the high CAP group (Figure 2b). Besides, in patients with high FIB-4, a positive correlation was noted between the nonsignificant fibrosis-related bacterium Weisella and the nonsignificant fibrosis-related glycosylation feature IgG2-N4H4F1. Conversely, no correlations were observed in the low FIB-4 group (Figure 3a and b). These results suggest that complicated interactions between microbial consortia and glycosylation may influence the development of HS and significant hepatic fibrosis.

Figure 2.

Heatmaps of correlation coefficients of microbial consortia to IgG1 or IgG2 glycosylation features in nonsevere and severe HS, according to low (CAP <292) and high CAP (CAP ≥292) value, in patients with MASLD. Among the patients with MASLD, the heatmaps depict correlation coefficients (r) for nonsevere HS-related and severe HS-related microbial consortia in relation to IgG1 or IgG2 glycosylation features. Among the patients in (a) low and (b) high CAP group. The value within each colored square of the heatmap represents the r for the respective paired data. The “*” indicates the correlation pairs with significant P value of correlation. The squares with bright colors such as red and pink indicated positive correlations, whereas cooler colors like light blue and deep blue signify negative correlations between the paired data in the correlation heatmaps. CAP, controlled attenuation parameter; IgG, immunoglobulin G; MASLD, metabolic dysfunction-associated steatotic liver disease.

Figure 3.

Heatmaps of correlation coefficients of microbiota to IgG1 or IgG2 glycosylation features between nonsignificant and significant hepatic fibrosis, according to low (FIB-4 <1.3) and high FIB-4 (FIB-4 ≥1.3) value, in patients with MASLD. Among the patients with MASLD, the heatmaps depict correlation coefficients (r) for nonsignificant hepatic fibrosis-related and significant hepatic fibrosis-related microbial consortia in relation to IgG1 or IgG2 glycosylation features among the patients in the (a) low and (b) high FIB-4 group. The glycosylation features selected for analysis were those that differed between the low FIB-4 groups. The value within each colored square of the heatmap represents the r for the respective paired data. The “*” indicates the correlation pairs with significant P value of correlation. The squares with bright colors such as red and pink indicated positive correlations, whereas cooler colors like light blue and deep blue signify negative correlations between the paired data in the correlation heatmaps. IgG, immunoglobulin G; MASLD, metabolic dysfunction-associated steatotic liver disease.

Intercorrelations between IgG1 and IgG2 glycosylation features in low/high CAP or low/high FIB-4 groups among patients with MASLD

The analysis of intercorrelations among glycosylation features in IgG1 and IgG2 across low and high CAP groups revealed several moderate to high positive correlations, particularly in nonsevere HS patients (Figure 4a; Supplementary Figure 4, Supplementary Digital Content 6, http://links.lww.com/CTG/B428). Three positively correlated glycosylation features maintained consistent correlations across both low and high CAP groups: IgG1-N4H3F1 with IgG2-N4H3F1, IgG1-N4H5F1 with IgG2-N4H5F1, and bisected-IgG1 with bisected-IgG2 (Figure 4b and c; Supplementary Figure 4, Supplementary Digital Content 6, http://links.lww.com/CTG/B428).

Figure 4.

Heatmaps of correlation coefficients of IgG1 to IgG2 glycosylation features between nonsevere and severe HS, according to low (CAP <292) and high CAP (CAP ≥292), or low (FIB-4 <1.3) and high FIB-4 (FIB-4 ≥1.3), in patients with MASLD. The heatmaps of correlation coefficients (r) of IgG1 to IgG2 glycosylation features in the (a) low CAP and (b) high CAP group. The “*” indicates the glycosylation features pairs with significant P value of correlation. (c) In the red and blue circles, the nonoverlap area in the red circle indicated the positively correlated glycosylation features dominant in only low CAP group. The nonoverlap area in the blue circle indicated the positively correlated glycosylation features dominant in only high CAP group. The heatmaps of correlation coefficients of IgG1 to IgG2 glycosylation features in the (d) low FIB-4 and (e) high FIB-4 group. (f) In the red and blue circles, the nonoverlap area in the red circle indicated the positively correlated glycosylation features dominant in the only low FIB-4 group. The squares with bright colors such as red and pink indicated positive correlations, whereas cooler colors like light blue and deep blue signify negative correlations between the paired data in the correlation heatmaps. CAP, controlled attenuation parameter; Gal index, galactosylation index; IgG, immunoglobulin G; MASLD, metabolic dysfunction-associated steatotic liver disease.

The intercorrelations among glycosylation features in IgG1 and IgG2 across low and high FIB-4 groups, and several glycosylation features exhibited significant correlations (Figure 4d and e). Notably, 4 positively correlated glycosylation features maintained consistent correlations across low and high FIB-4 groups: IgG1-N4H3F1 with IgG2-N4H3F1, IgG1-N4H5F1 with IgG2-N4H5F1, bisected-IgG1 with bisected-IgG2, and IgG1-galactosylation index with IgG2-galactosylation index (Figure 4e and f). Supplementary Table S5 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432) showed that their ratio consistent positively correlated glycosylation features were different between low/high CAP or low/high FIB-4 Groups, which means the ratio is associated with different degree of HS and fibrosis.

Models for identifying severe HS and significant hepatic fibrosis by the ratio of the consistent positively correlated glycosylation features

Focusing on consistent positively correlated glycosylation features in both low and high CAP groups, the ratios were evaluated as candidates for identifying severe HS, among IgG2-N4H5F1/IgG1-N4H5F1 and bisected-IgG2/bisected-IgG1 can identify severe HS (Figure 5a and c). The other glycosylation ratio, IgG2-N4H3F1/IgG1-N4H3F1, cannot identify severe HS (Figure 5b). Similarly, the ratios of consistent positively correlated features, including IgG2-N4H3F1/IgG1-N4H3F1, bisected-IgG1/bisected-IgG2, and IgG1-galactosylation index/IgG2-galactosylation index, could identify significant hepatic fibrosis (Figure 5d–f). However, IgG1-N4H5F1/IgG2-N4H5F1 cannot identify significant hepatic fibrosis (Figure 5g).

Figure 5.

ROC analyses of the ratio of the positively correlated common glycosylation features in both nonsevere (low CAP)/severe HS (high CAP) or nonsignificant (low FIB-4)/significant hepatic fibrosis (high FIB-4) for identification of severe HS or significant hepatic fibrosis in patients with MASLD. (a) The ROC curve of IgG2-N4H5F1/IgG1-N4H5F1 ratio for severe HS. (b) The ROC curve of IgG1-N4H5F1/IgG2-N4H5F1 ratio for severe HS. (c) The ROC curve of bisected-IgG2/bisected-IgG1 ratio for severe HS. (d) The ROC curve of IgG2-N4H3F1/IgG1-N4H3F1 ratio for significant hepatic fibrosis. (e) The ROC curve of bisected-IgG1/bisected-IgG2 ratio for significant hepatic fibrosis. (f) The ROC curve of IgG1-galactosylation index/IgG2-galactosylation index ratio for significant hepatic fibrosis. (g) The ROC curve of IgG1-N4H5F1/IgG2-N4H5F1 ratio for significant hepatic fibrosis. CAP, controlled attenuation parameter; Gal index, galactosylation index; IgG, immunoglobulin G; ROC, receiver operating characteristic.

Combination scores incorporating microbial consortia and glycosylation features ratios as identifying models for severe HS and significant hepatic fibrosis

To evaluate the overall effects of microbial consortia and glycosylation features on identifying severe HS and significant hepatic fibrosis among patients with MASLD, we developed combination scores that included both the severe HS-microbial consortia score and the ratios of glycosylation features. For identifying severe HS, a score of 1 was assigned if either IgG2-N4H5F1/IgG1-N4H5F1 or bisected-IgG2/bisected-IgG1 exceeded their respective cutoff values. The severe HS combination score was calculated as the sum of the glycosylation feature ratios score and the severe HS-microbial consortia score (Figure 6a). The AUC for this combination score was 0.776 (P < 0.001), which was higher than the HSI (AUC = 0.645, P = 0.01). The calibration curves of Figure 6 showed that P values of Hosmer–Lemeshow test are 0.109 and 0.797 for model of detecting of severe HS and significant hepatic fibrosis indicating that the appropriateness of the model (Supplementary Figure 5A and B, Supplementary Digital Content 7, http://links.lww.com/CTG/B429).

Figure 6.

ROC analyses of combination score of various microbial consortia-glycosylation feature for identifying severe hepatic steatosis and comparison with hepatic steatosis index. (a) The overall severe HS-glycosylation ratios and the severe HS-microbial consortia score were applied to establish the severe HS-combination score, which was used as an identifying marker for severe HS, and the comparison with the HSI. (b) Similarly, the significant hepatic fibrosis combination score was used as an identifying marker, and the comparison with FAST score for significant hepatic fibrosis. FAST, FibroScan-AST; HS, hepatic steatosis; HSI, hepatic steatosis index; ROC, receiver operating characteristic.

Similarly, the significant hepatic fibrosis-combination score was calculated by summing the significant hepatic fibrosis-microbial consortia score and the ratios of glycosylation features. A score of 1 was assigned if any of the following ratios exceeded their respective cutoff values: IgG2-N4H3F1/IgG1-N4H3F1, bisected-IgG1/bisected-IgG2, or IgG1-galactosylation index/IgG2-galactosylation index. As shown in Figure 6b, the identifying ability of this combination score was compared with FAST score for significant hepatic fibrosis. The AUC for the combination score was 0.810 (P < 0.001), which was higher than that of the FAST score (AUC = 0.690, P = 0.02).

In addition to cross-validation with preexisting model HSI and FAST score, the internal validation was performed with 10-fold with bootstrap validation showed good AUC curves (Supplementary Figure 6, Supplementary Digital Content 8, http://links.lww.com/CTG/B430).

DISCUSSION

Recent studies have highlighted a connection between the relative agalactosylation of IgG-GlcNAc and visceral fat deposition, which is linked to HS and fibrosis in patients with MASLD (29). Hypogalactosylation of IgG has also been associated with hepatic inflammation in pediatric patients with MASLD (30,31). Our research found that a low galactosylation index of IgG2 in patients with MASLD significantly correlates with severe HS and substantial hepatic fibrosis. Notably, specific ratios of glycosylation features between IgG1 and IgG2 emerged as potential predictors for severe HS and significant hepatic fibrosis in these patients.

Alterations in microbial consortia have been implicated in the pathogenesis of MASLD (32). Increased abundance of the intestinal bacterium Megamonas is associated with the development of MASLD, and our study further established a link between Megamonas and severe HS (33). In this study, similar analysis had been done between MASLD vs non-MASLD or low vs high CAP groups to explore the impacts of gut microbiota on clinical data and HS of MASLD or severe HS. Notably in Figure 1, microbe Megamonas is related to severe HS (high CAP) than nonsevere HS group (low CAP), but also more abundant in non-MASLD group than MASLD (Supplementary Figure 2, Supplementary Digital Content 4, http://links.lww.com/CTG/B426). Differently, the abundance of microbe Megamonas is a protective factor in the comparison between non-MASLD vs MASLD groups, whereas the abundance of microbe Megamonas is a harmful factor in the comparison between severe HS vs non-severe HS groups. Notably, the negative fold change of relative abundance of microbe Megamonas between MASLD vs non-MASLD group is smaller than the positive fold change of relative abundance of microbe Megamonas between low-CAP vs high-CAP groups. As shown in Supplementary Table S1 (Supplementary Digital Content 3, http://links.lww.com/CTG/B432), the significant better glycemia and lipid profiles were observed in non-MASLD group than those in MASLD group. In fact, previous study had reported that high abundance of microbe Megamonas is associated with better glycemia and lipid profiles (34). Accordingly, it is probably that the microbe Megamonas had different roles the glycemia/lipid profiles and HS (33,34).

Previous research has indicated decreased levels of Eubacterium and Lachnospiraceae in relation to HS; our findings support that these species diminish as HS severity progresses (35). In addition, changes in intestinal microbial consortia correlate with the progression of hepatic fibrosis in MASLD (36). We identified specific microbial consortia alongside glycosylation feature ratios as potential predictors of MASLD severity, particularly severe HS and significant hepatic fibrosis.

The glycosylation features of IgG subclasses were significantly associated with severe HS and significant hepatic fibrosis in our study. Interventions targeting glycosylation have been explored in anticancer therapies, with ongoing clinical trials assessing their efficacy (37). Consistent with our findings, microbial consortia influence IgG glycosylation in various inflammatory conditions (38). Furthermore, modifications to microbial consortia through fecal microbiota transplantation have shown potential to reduce chronic inflammation by altering serum N-glycosylation patterns (39). Thus, modifying IgG glycosylation may alleviate HS and steatohepatitis through probiotic-based alterations in intestinal microbial consortia (40).

Compared with established predictive models like the HSI and FAST score, our combination models demonstrated comparable efficacy in predicting severe HS and significant hepatic fibrosis within our cohort. Notably, serum AST levels in the HSI and FAST scores lack specificity for liver conditions and may be influenced by multiple factors. The HSI is limited in distinguishing severe steatosis and may be confounded by concurrent fibrosis (41). Moreover, the requirement for specialized equipment, such as FibroScan, complicates FAST score measurements. The accuracy of fibroscan-based liver stiffness (kPa) diagnostics for advanced fibrosis in patients with MASLD remains controversial. In fact, hepatic fibrosis (collagen fiber deposition) develops gradually, not linearly (42). At the earlier stages, fibers are deposited with a slower increment. Alternatively, during advanced CLD, deposition of fibers happens with higher speed (exponentially). Therefore, the accuracy of liver stiffness (kPa) measurements and reference standards can be negatively affected by these nonlinear differences and random accumulation of collagen (43).

In contrast, our combination score relies solely on accessible serum and fecal samples, and allows for simultaneous detection of severe HS and significant hepatic fibrosis, facilitating easier collection and comprehensive assessment of MASLD severity.

This study has several limitations, primarily focusing on the glycosylation of dominant IgG subclasses, specifically IgG1 and IgG2. Future research should investigate IgG3 and IgG4 glycosylation to provide a more comprehensive understanding. In addition, the diet composition between MASLD and non-MASLD and those with high and low CAP may be different in calories, fiber intake, processed foods, etc, which might affect the differences of gut microbiota between groups (44). So, future study needs to explore the effects of diet composition on the gut microbiota between groups. Although Fibroscan have a cost and limited availability, fecal microbiota and serum IgG glycosylation analysis will not be easy to obtain clinically due to variety abilities in different labs and insurance applicability. But the initial results of this study still can be the basis for further pharmacologic and pathogenic exploration of hepatic fibrosis.

In our study, cross validation with preexisting model, internal validation with 10 time, and calibration curves confirmed the model's accuracy for identifying severe HS and significant hepatic fibrosis in MASLD case. Nonetheless, the limited sample size precluded the inclusion of a validation cohort and with estimate medium power and effect size of 0.65, necessitating further external validation of these models in subsequent large-scale studies. Despite these limitations, our findings establish a significant association between IgG glycosylation features and MASLD severity that proved by calibration curves, underscoring the potential of glycosylation modifications as a promising therapeutic avenue for further exploration.

In conclusion, this study demonstrated the potential of combination models utilizing fecal microbiota and serum IgG subclass glycosylation features, which were found to be noninferior to conventional models for identifying severe HS and significant hepatic fibrosis. Furthermore, the research highlights glycosylation modification as a potential therapeutic approach for MASLD.

CONFLICTS OF INTEREST

Guarantor of the article: Ying-Ying Yang, PhD.

Specific author contributions: T.-H.L. and Y.-Y.Y.: conceptualization. Y.-Y.Y. and Y.-H.L.: data curation. J.-W.R., C.-C.M., and S.-Y.T.: formal analysis. Y.-Y.Y.: funding acquisition. H.-Y.Y., H.-C.T., and H.-C.S.: investigation. P.-C.C., S.-Y.H., T.-L.H., and Y.-Y.Y.: methodology. Y.-Y.Y.: project administration. Y.-Y.Y.: resources. T.-H.L., J.-W.R., C.-C.M., and S.-Y.T.: software. M.-C.H. and H.-C.L.: supervision. C.-W.S.: validation. C.-C.M., S.-Y.T., and P.-C.C.: visualization. T.-H.L. and S.-Y.T.: roles/writing—original draft. Y.-Y.Y. and J.-W.R.: writing—review and editing.

Financial support: This research was funded by grants from Taipei Veterans General Hospital (grant numbers: V114C-012, V114EA-004, V113D701-001-MY2-1, V113D71-001-MY2-2, VTA113-A-4-2, VN114-01, VTA114-V1-5-1, 114Q159502), Ministry of Science and Technology (Taiwan; grant numbers: NSTC112-2314-B-A049-043-MY3,NSTC114-2410-H-A49-029-MY2, MOST 111-2314-B-001-009).

Potential competing interests: None to report.

Study Highlights.

WHAT IS KNOWN

• ✓ Metabolic dysfunction-associated steatotic liver disease (MASLD) is the fastest-growing cause of hepatocellular carcinoma.

• ✓ Total immunoglobulin G (IgG) glycosylation features are associated with development of MASLD, but unknown in progression.

• ✓ Currently existed non-invasive models for severe hepatic steatosis and significant fibrosis in MASLD have several limitations.

WHAT IS NEW HERE

• ✓ IgG1/IgG2 subclass glycosylation features are associated with severe hepatic steatosis and significant fibrosis in MASLD.

• ✓ Models combining microbiota and IgG1/IgG2 glycosylation well identify severity of hepatic steatosis and fibrosis in MASLD.

• ✓ The combination models provide noninferior and easily accessible methods comparing with currently used models.

Supplementary Material

ct9-17-e00957-s007.docx ^{(30.7KB, docx)}

ct9-17-e00957-s008.docx ^{(51.2KB, docx)}

ABBREVIATIONS:

AST

aspartate transaminase

AUC

area under curve

CAP

controlled attenuation parameter

CLDs

chronic liver diseases

FAST

FibroScan-AST

HCC

hepatocellular carcinoma

HS

hepatic steatosis

HSI

hepatic steatosis index

IgG

immunoglobulin G

MASLD

metabolic dysfunction-associated steatotic liver disease

MS

mass spectrometry

ROC

receiver operating characteristic