A novel model for predicting the risk of liver cirrhosis in people living with HIV/HBV coinfection: a study based on proteomic analysis

Abstract

Background

Easily accessible clinical warning signals of liver cirrhosis for optimal management in people living with HIV/HBV co-infection remains limited. Using proteomic techniques, a risk prediction model for liver cirrhosis through the screening of differentially expressed proteins (DEPs) in people living with HIV/HBV coinfection was developed.

Methods

Quantitative liquid chromatography-mass spectrometry (LC–MS) was used to identify DEPs in plasma collected from HIV/HBV co-infected patients with or without liver cirrhosis. The annotation information of DEPs were obtained by mapping identified proteins to the Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes(KEGG), and Disease Ontology(DO) databases. Differential expression levels of plasma L-selectin(CD62L) were validated in HIV/HBV co-infected patients.

Results

Proteomic profiling of 13 matched patient pairs with HIV/HBV coinfection revealed 111 differentially expressed proteins (DEPs) associated with liver cirrhosis. Functional analysis showed significant enrichment in immune processes and the Complement and Coagulation Cascades pathway, underpinning the chronic inflammatory and fibrotic aspects of the disease. A diagnostic signature derived from these DEPs was developed, with a model containing IGHV5-37 and L-selectin showing high predictive value. Critically, the elevated plasma level of L-selectin in cirrhosis was confirmed in a separate validation cohort of 90 patients, underscoring its clinical relevance.

Conclusions

Our findings demonstrate that cirrhosis in HIV/HBV coinfected patients is characterized by a specific plasma proteomic profile. From this profile, we derived and validated a diagnostic model, highlighting L-selectin as a key validated biomarker. This model holds significant promise for development into a clinical assay to improve the detection and management of liver disease in this high-risk population.

Clinical trial

Not applicable

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-025-12298-z.

Introduction

Chronic hepatitis B virus (HBV) co-infection represents a major clinical challenge in the management of people living with human immunodeficiency virus (HIV). This specific population faces a significantly accelerated trajectory of liver disease progression, driven by the combined virological impact of both chronic infections [1]. Consequently, HIV/HBV co-infected individuals experience a high burden of liver-related morbidity, with rapid progression to cirrhosis being a particularly pressing outcome [1–3]. The early and often subclinical nature of fibrogenesis creates a critical window where intervention could be most effective. Therefore, the development of robust, non-invasive tools for the early identification of cirrhosis is imperative. Such strategies are essential to enable proactive management, mitigate serious liver-related complications, and improve overall outcomes in this high-risk cohort.

Thrombocytopenia, AST-to-platelet ratio index (APRI) and AST/ALT ratio (AAR), cirrhosis discriminant score (CDS), age-platelet index (API), Pohl score, FIB-4 index, and Lok’s model are well-known surrogates for liver cirrhosis that are based on routine laboratory data and thus readily available in clinical practice. In some patients, these criteria have been shown to indicate the presence of significant and extensive fibrosis. Some non-invasive or minimally invasive fibrosis biomarkers, such as miR-214 [4] and M2BPGi [5], have shown remarkable diagnostic potential for early diagnosis of liver cirrhosis without biopsy in recent years. These tests necessitate a high level of technology and technical personnel, and the standard equipment of general medical facilities is incapable of determining these signs. The development of non-invasive and simple serological indicators for early cirrhosis diagnosis is both urgent and clinically significant.

A glycoprotein known as L-selectin (CD62L) has been linked to numerous disorders and has a wide range of biological functions [6–9]. Its function has been investigated in NASH in humans and two steatohepatitis mouse models, and it was found that L-selectin knockout mice exhibited enhanced liver infiltration of Treg cells and strong activation of antioxidant stress response, resulting in reduced liver fibrosis in L-selectin-/- mice [10]. These results indicate that L-selectin is a promising target for NASH intervention. In addition, Previous studies have shown that L-selectin dependent lymphocyte infiltration can induce inflammatory bowel disease (IBD) and selective inhibition of L-selectin has reached clinical application in patients with IBD [11], but its function in the liver, especially in liver cirrhosis of patients with HIV/HBV co-infection, remains unclear.

Therefore, in this study, plasma samples were taken from HIV/HBV coinfected people with or without liver cirrhosis for quantitative proteomic analysis, and an L-Selectin-included model for predicting the likelihood of hepatic cirrhosis in this population was established. The purpose of this study was to provide easily accessible clinical warning signals of liver cirrhosis for optimal management in people living with HIV/HBV co-infection.

Methods

Clinical characteristics of the study subjects

The patient selection criteria for this study have been described in detail elsewhere [12]. In brief, we enrolled HIV/HBV co-infected adults (30–70 years old) at Zhongnan Hospital during the timeframe of January 1, 2021, to December 31, 2022. Individuals with incomplete clinical data, other liver diseases, or key comorbidities (e.g., pregnancy, organ dysfunction) were excluded.

This study included 58 age- and sex-matched pairs of HIV/HBV co-infected patients, comprising a cirrhosis group and a non-cirrhosis control group.

The diagnosis of liver cirrhosis was definitively established by meeting at least one of the following criteria: (1) Liver Histology: Evidence of cirrhosis on liver biopsy (Metavir stage F4). (2) Imaging Findings: Presence of radiological features suggestive of cirrhosis and portal hypertension on abdominal ultrasound, computed tomography, or magnetic resonance imaging (e.g., nodular liver surface, coarse parenchymal echo-texture, splenomegaly, or portosystemic collateral vessels). (3) Clinical Signs of Portal Hypertension: Documentation of esophageal or gastric varices on endoscopy, or a history of variceal bleeding or ascites.

As summarized in the Table 1, the two groups were well-matched in terms of baseline characteristics. However, as expected, the cirrhosis group exhibited significant alterations in markers of synthetic liver function and systemic inflammation, including higher International Normalized Ratio (INR), prolonged prothrombin time, and elevated levels of C-reactive protein (CRP) and interleukin-6 (IL-6). Among the 58 patients in the cirrhosis group, 38 were classified as having compensated cirrhosis and 20 had decompensated cirrhosis. The disease severity within this group was quantified by a Child-Pugh score of 6 ± 2 and a Model for End-Stage Liver Disease (MELD) score of 9 ± 3. Figure 1 depicts the flow chart of study object selection.

Table 1.

Comparative analysis of clinical and laboratory parameters in HIV/HBV co-infected patients with and without liver cirrhosis

	People living with HIV/HBV co-infection		t-value	p
	With liver cirrhosis	Without liver cirrhosis
Cases	58	58	/	/
Male (n,%)	41(70.7)	41(70.7)	/	/
Age, mean ± SD	46.8 ± 2.4	48.8 ± 3.5	0.449	0.654
BMI (kg/m²)	20.8 ± 0.58	22.0 ± 0.45	3.506	0.012
cART treatment (n,%)	29(50.0)	29(50.0)	/	/
Anti-HBV therapy (n,%)	29(50.0)	29(50.0)	/	/
CD4+ T cell count (cells/µl)	145 ± 28	150 ± 32	0.202	0.840
Log10HIV-RNA (mean ± SD)	4.61 ± 0.20	4.58 ± 0.33	0.502	0.617
HBeAg (n,%)	17	16
Log10HBV-DNA (mean ± SD)	3.04 ± 0.42	3.41 ± 0.32	0.705	0.482
Blood routine test (mean ± SD)
White blood cells (109/L)	4.21 ± 0.24	5.92 ± 0.56	2.320	0.022
Hemoglobin (g/L)	118.2 ± 13.2	120.3 ± 16.4	0.716	0.475
Platelets (109/L)	112.2 ± 32.1	227.4 ± 40.2	1.874	0.063
Liver function test (mean ± SD)
ALT (U/L)	58.6 ± 6.5	39.7 ± 6.5	2.894	0.039
AST (U/L)	56.5 ± 4.2	39.2 ± 6.7	2.814	0.040
Tbil (µmol/L)	15.2 ± 3.7	8.9 ± 2.9	1.536	0.420
Alb(g/L)	40.2 ± 5.5	45.6 ± 6.3	2.036	0.044
Diabetes (n, %)	10 (17.2)	3 (58)	4.245	0.039
Antidiabetic Therapy	10 (17.2)	3 (58)	4.245	0.039
FBG(mmol/L)	5.62 ± 1.71	5.44 ± 1.85	0.668	0.505
Alcohol Use (n, %)	5 (8.6)	7 (12.1)	0.372	0.542
Child-Pugh Score	6 ± 2	N/A	/	/
MELD Score	9 ± 3	N/A	/	/
INR	1.3 ± 0.5	0.9 ± 0.4	3.253	0.032
Prothrombin Time (s)	15.2 ± 0.7	11.9 ± 1.0	3.860	0.029
CRP (mg/L)	3.56 ± 0.18	2.35 ± 0.22	2.536	0.040
IL-6 (pg/mL)	1.42 ± 1.21	0.85 ± 0.43	3.027	0.036

Abbreviations: SD, standard deviation; BMI, Body Mass Index; cART, Combination antiretroviral therapy; ALT, alanine aminotransferase; AST, aspartate aminotransferase; Tbil, total bilirubin; Alb, albumin; FBG, fasting blood glucose; MELD, Model for End-Stage Liver Disease; INR, International Normalized Ratio; CRP, C-Reactive Protein; IL-6, Interleukin-6

Fig. 1.

The flow chart of study object selection

This study was conducted in accordance with the Helsinki Declaration and its recent modifications. Human Ethics and Consent to Participate declarations were acquired in this study.

Proteomic profiling by liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Sample preparation and protein digestion

Proteomic analysis of the current cohort was performed as previously described [12]. Protein extraction was performed using a lysis buffer containing 8 M urea, 100 mM Tris-HCl (pH 8.5), and protease inhibitors. The protein concentration was determined using the Bradford assay. Subsequently, 50 µg of protein from each sample was subjected to reduction with 5 mM dithiothreitol (DTT) at 37 °C for 45 min, followed by alkylation with 15 mM iodoacetamide (IAA) at room temperature in the dark for 30 min. The urea concentration was then diluted to below 2 M with 100 mM Tris-HCl, and proteins were digested with sequencing-grade modified trypsin (Promega) at an enzyme-to-substrate ratio of 1:50 (w/w). The digestion was carried out at 37 °C for 16 h and quenched by acidification with formic acid (FA) to a final concentration of 1%. The resulting peptides were desalted using C18 solid-phase extraction (SPE) cartridges, vacuum-dried, and reconstituted in 0.1% FA for LC-MS/MS analysis.

Liquid chromatography and tandem mass spectrometry (LC-MS/MS)

The peptide mixtures were analyzed using a nano-flow liquid chromatography system (EASY-nLC 1200, Thermo Fisher Scientific) coupled online to a high-resolution tandem mass spectrometer (Orbitrap Exploris™ 480, Thermo Fisher Scientific).

• Chromatography: Peptides were loaded onto a C18 analytical column (75 μm × 25 cm, 2 μm particle size) and separated with a 120-minute linear gradient from 5% to 28% mobile phase B (0.1% FA in 80% acetonitrile) at a flow rate of 300 nL/min.

• Mass Spectrometry: The MS data were acquired in Data-Independent Acquisition (DIA) mode. Full MS1 spectra were collected at a resolution of 120,000 (at m/z 200) with a scan range of 350–1200 m/z. For MS2, DIA windows were set to cover the same m/z range with variable window widths, and fragments were detected in the Orbitrap at a resolution of 30,000. The key ion source settings were as follows: electrospray voltage, 2.0 kV; capillary temperature, 320 °C; normalized higher-energy collisional dissociation (HCD) energy, 28%.

Data processing and protein identification/quantification

The protein identification was based on widely accepted criteria, including a minimum of two unique peptides per protein and a false discovery rate (FDR) of < 1% at the protein level. The raw DIA files were processed using DIA-NN software (version 1.8). A spectral library was generated in-silico from the Swiss-Prot human protein database (reviewed, containing ~ 20,000 sequences) using the library-free mode. The search parameters included: trypsin/P as the enzyme with up to one missed cleavage; fixed modification of carbamidomethylation on cysteine; and variable modification of methionine oxidation. The precursor and fragment mass accuracies were set to 10 ppm.

• Quantification and Filtering: Protein quantification in DIA-NN was performed using the MaxLFQ algorithm integrated into the software. The reported protein intensities were based on the median of the top three precursor ions (Top3 approach). Protein intensity values were log2-transformed and normalized across all samples using quantile normalization. Low-abundance proteins were filtered out, retaining only those detected in at least 70% of samples in either the liver cirrhosis (LC) or non-cirrhosis (NC) group. To ensure high-quality data, proteins were filtered to include only those with at least two unique peptides and detected in 100% of the samples within at least one experimental group.

• False Discovery Rate (FDR): Both peptide-spectrum matches (PSMs) and protein identifications were controlled at a global FDR of 1%, as estimated by DIA-NN through a reverse decoy database strategy. Significance was defined as a false discovery rate (FDR) < 0.05 with an absolute log2(fold change) > 0.58.

Data normalization and quality control

To correct for systematic technical variance, peptide abundance data were normalized using the median normalization method within DIA-NN. Several quality control (QC) measures were implemented to ensure data reliability:

• Technical Replicates: A pooled sample from all individuals was created as a quality control and run repeatedly throughout the acquisition sequence to monitor instrument performance and reproducibility.

• Blank Samples: Solvent blanks were injected periodically to monitor and account for potential background contamination.

• Peptide Intensity and Retention Time Stability: The coefficient of variation (CV) of peptide intensities in the pooled QC samples was calculated, and only analyses with a median CV of less than 15% were considered for further statistical analysis.

Differential expression and statistical analysis

Differentially expressed proteins (DEPs) were identified by applying a dual-threshold criterion to the normalized protein intensities. Proteins with an absolute fold change (FC) ≥ 1.5 and a p-value < 0.05 (calculated using a two-sided Student’s t-test, adjusted for multiple comparisons using the Benjamini-Hochberg procedure) were considered statistically significant.

Protein validation

To confirm the reliability of the proteomic findings, the expression levels of two selected candidate proteins were independently validated using Western blot analysis on a subset of the original samples. This orthogonal method provided additional confidence in the mass spectrometry-based quantification.

Biological information analysis

Functional annotation of the differentially expressed proteins (DEPs) was carried out using several public databases: Gene Ontology (GO), the Kyoto Encyclopedia of Genes and Genomes (KEGG), and Disease Ontology (DO). These resources facilitated the determination of the DEPs’ functional properties. Subsequently, protein-protein interaction (PPI) networks were constructed by integrating data from the STRING database with CytoScape software (version 3.10.1), visualizing interactions between DEPs and known human proteins.

The construction of clinical plasma modeling

To rigorously evaluate the model’s performance and avoid potential overfitting, we partitioned the dataset into a training set (20 cases) and an independent validation set (6 cases). The Least Absolute Shrinkage and Selection Operator (LASSO) regression, implemented using the “glmnet” package in R, was applied solely to the training set to identify the optimal combination of differentially expressed proteins (DEPs). This optimal protein combination was then used to construct a multivariate logistic regression model. The model’s performance, including the Area Under the Curve (AUC), was evaluated on the held-out independent validation set. Decision curve analysis (DCA) was employed to ascertain the clinical utility of the proposed liver cirrhosis prediction model. This analysis quantifies the net benefit across a spectrum of decision thresholds, thereby integrating the clinical consequences of false-positive and false-negative predictions. Furthermore, a nomogram was constructed to provide a visual representation of the model, enabling clinicians to readily compute an individualized risk estimate for each patient based on the contribution of each predictor variable.

Plasma L-selectin validation detection

In this study, apart from the plasma samples used for proteomic analysis, the remaining 90 plasma samples were utilized for L-selectin validation. Plasma L-selectin levels were quantified using a human CD62L ELISA kit (Abcam, UK) according to the manufacturer’s protocol. Briefly, samples and standards were added to antibody-coated microtiter wells. After incubation and washing, a streptavidin-HRP conjugate and TMB substrate were sequentially added. The reaction was stopped with an acid solution, and the absorbance was measured at 450 nm. The concentration of L-selectin was determined by interpolation from the standard curve.

Statistical analysis

The screening pipeline for biomarkers integrated magnitude of change (fold change, FC) and statistical testing (p ≤ 0.05). All statistical analyses were performed using R software (version 4.3.1). Feature selection was conducted with the glmnet package (version 4.1-8) via LASSO regression. The optimal features identified were then used to construct a multivariate logistic regression model without regularization for clinical interpretation and nomogram development.

Results

Patient characteristics

As shown in Table 1, a total of 116 individuals living with HIV/HBV co-infection were included in this study, comprising 58 patients with liver cirrhosis and 58 matched controls without cirrhosis. The two groups were well-balanced in terms of sex distribution (70.7% male in both) and mean age (46.8 ± 2.4 vs. 48.8 ± 3.5 years, p = 0.654).

Notable differences were observed in several parameters. Patients with cirrhosis had a significantly lower body mass index (BMI) compared to the non-cirrhotic group (20.8 ± 0.58 vs. 22.0 ± 0.45 kg/m², p = 0.012). As expected, markers of liver synthetic function and injury were significantly altered in the cirrhosis group. This was evidenced by elevated levels of alanine aminotransferase (ALT, 58.6 ± 6.5 vs. 39.7 ± 6.5 U/L, p = 0.039), aspartate aminotransferase (AST, 56.5 ± 4.2 vs. 39.2 ± 6.7 U/L, p = 0.040), total bilirubin (15.2 ± 3.7 vs. 8.9 ± 2.9 µmol/L), and a reduced albumin level (40.2 ± 5.5 vs. 45.6 ± 6.3 g/L, p = 0.044). Coagulation profiles were also impaired in cirrhotic patients, demonstrated by a higher international normalized ratio (INR, 1.3 ± 0.5 vs. 0.7 ± 0.4, p = 0.032) and prolonged prothrombin time (14.5 ± 0.7 vs. 11.9 ± 1.0 s, p = 0.029). Furthermore, the cirrhosis group exhibited significant thrombocytopenia (platelets: 182.2 ± 3.23 vs. 227.4 ± 4.02 × 10⁹/L, p = 0.053).

The prevalence of diabetes were significantly higher in the cirrhotic group (17.2% vs. 5.8%, p = 0.039), and all enrolled patients with diabetes were receiving antidiabetic therapy. Systemic inflammation was more pronounced in cirrhotic patients, as indicated by higher levels of C-reactive protein (CRP, 3.56 ± 0.18 vs. 2.35 ± 0.22 mg/L, p = 0.040) and interleukin-6 (IL-6, 1.42 ± 1.21 vs. 0.85 ± 0.43 pg/mL, p = 0.036). Among the cirrhotic patients, the mean Child-Pugh score was 6 ± 2, and the mean Model for End-Stage Liver Disease (MELD) score was 9 ± 3.

No significant differences were found between the groups in CD4⁺ T-cell count, HIV viral load, HBV viral load, HBeAg status, hemoglobin, fasting blood glucose, or reported alcohol use.

Differential protein analysis

Our comparative proteomic analysis successfully distinguished individuals with liver cirrhosis (LC) from non-cirrhosis (NC) controls. Data-Independent Acquisition (DIA) was used to profile 26 proteomes, which were then analysed in DIA-NN (v1.8). 7,978 precursors corresponding to 643 proteins were quantified (Fig. 2A and B). Unsupervised principal component analysis (PCA) based on the intensities of 111 differentially expressed proteins (DEPs) revealed clear segregation between the LC and NC groups (Fig. 2D), confirming the discriminatory power of this protein set. These DEPs, identified using a fold-change cutoff (|log2FC| ≥ 0.58) and a significance threshold (p ≤ 0.05), included 65 upregulated and 46 downregulated proteins (Fig. 2C) in the LC group compared to the NC controls. Finally, a heatmap visualization further underscored the profound proteomic alterations associated with liver cirrhosis by displaying the distinct expression patterns of these DEPs (Fig. 2E).

Fig. 2.

Differently expressed proteins in patients with liver cirrhosis. (A, B) The distribution of numbers of quantified (A) proteins and (B) precursors. (C) the volcano plot of DEPs. (D) PCA plots. (E) Heat map visualization of DEPs. G0 represents HIV/HBV co-infected patients without liver cirrhosis, and G1 represents HIV/HBV co-infected patients with liver cirrhosis

Functional enrichment analysis of differentially abundant proteins

Functional enrichment analysis of the differentially expressed proteins (DEPs) revealed significant alterations in two core biological processes: immune response and lipid metabolism. These findings are consistent with the known pathophysiological features of liver cirrhosis in the context of HIV/HBV co-infection.

Within the Biological Process domain, the DEPs were markedly enriched in immune-related processes including complement activation, immunoglobulin-mediated response, and B cell immunity (Fig. 3A) in the LC group compared to the NC controls. These processes are closely associated with chronic viral infection and liver inflammation.

Fig. 3.

GO, KEGG and DO analysis for 111 DEPs. (A) GO-based enrichment analysis of DEPs(two-sided hypergeometric test, p ≤ 0.05), GO terms were sorted by p-value, and top 5 terms of each category were displayed. (B) KEGG-based enrichment analysis of DEPs(two-sided hypergeometric test, p ≤ 0.05), KEGG terms were sorted by p-value, and top 15 terms were displayed. (C) DO-based enrichment analysis of DEPs(two-sided hypergeometric test, p ≤ 0.05), DO terms were sorted by p-value, and top 15 terms were displayed

KEGG pathway analysis identified the Complement and Coagulation Cascades pathway as the most significantly enriched in the LC group compared to the NC controls (Fig. 3B). This pathway plays a crucial role in chronic liver injury and fibrosis progression.

Furthermore, Disease Ontology analysis confirmed the expected association with human immunodeficiency virus infection and additionally identified disorders of glucose metabolism as significantly enriched terms in the LC group compared to the NC controls (Fig. 3C). Glucose metabolism disorders represent a common complication in chronic liver disease, including cirrhosis.

We constructed a protein-protein interaction (PPI) network for the DEPs using the STRING database and visualized it with Cytoscape software (Fig. 4A). To identify pivotal nodes, we applied the CytoHubba plugin with four distinct algorithms: DEGREE, DMNC, MCC, and MNC. The top 10 proteins from each algorithm were compared using a Venn diagram. This cross-analysis culminated in the identification of six hub proteins: GC, ORM1, C3, APOA4, HPR, and HP (Fig. 4B).

Fig. 4.

PPI network analysis of DEPs. (A) Network graph depicting the correlation of proteins derived from DEPs. (B) The hub proteins calculated under DEGREE. (C) The hub proteins calculated under DMNC. (D) The hub proteins calculated under MCC. (E) The hub proteins calculated under MNC. (F) Venn diagram based on cross-analysis under four algorithms (DEGREE, DMNC, MCC and MNC). Circles indicated the gene symbol of protein, in which red and green indicated up-regulated and down-regulated proteins, respectively. The thickness of the lines represented the strength of the interaction. The darkness of the colors represented the magnitude of the discrepancy

Notably, the identified hub proteins have established roles in biological processes central to liver pathology. C3 is a key component of the complement system, which is critically involved in chronic liver inflammation and fibrosis. HP, ORM1, and HPR are acute-phase proteins whose expression is altered in response to inflammation and infection. APOA4 is a component of lipid metabolism, a process frequently dysregulated in liver cirrhosis. GC (Vitamin D-binding protein) has been implicated in immune modulation and chronic liver disease. The central role of these proteins in the PPI network underscores the interplay between immune response, inflammation, and metabolic dysregulation in the context of HIV/HBV co-infection and liver cirrhosis.

Screening protein by LASSO

To construct a diagnostic model, we employed Least Absolute Shrinkage and Selection Operator (LASSO) regression for feature selection. The LASSO model was fitted across a range of penalty parameter (λ) values, which controls the strength of regularization. As λ increases, the coefficients of less contributory proteins are shrunk towards zero, and only those with the strongest associations to the outcome are retained. Through this process, a solution path was generated, illustrating how each protein’s coefficient varied with λ (Fig. 5A).

Fig. 5.

DEPs selection using LASSO regression model. (A) LASSO coefficient profiles of the DEPs. (B) The relational graph between fitting error and log(λ). Dotted vertical lines were drawn at the optimal values by using the minimum criteria(min criteria) and the 1 standard error of the minimum criteria (1se criteria). λ value was chosen when the fitting error was minimum

The optimal λ value (lambda.min) was determined via five-fold cross-validation as the value that yielded the minimal mean cross-validated error (Fig. 5B). At this specific λ value (lambda.min), the model identified a panel of 13 proteins with non-zero coefficients, which constituted the final biomarker signature. The final set of proteins included IGLV5-37, Pulmonary surfactant-associated protein B, Zinc finger protein basonuclin-2, and others listed in Fig. 5A.

Development of a liver cirrhosis risk prediction model

To develop a clinical prediction model for liver cirrhosis, we evaluated random combinations of the 13 DEPs (limited to 5 proteins per combination) using five-fold cross-validation based on AUC. The optimal combination, which included IGLV5-37, L-selectin, and three other proteins, was selected based on its superior and consistent discriminatory performance across both training and test cohorts.

The model demonstrated strong discriminatory power, with AUCs of 0.93 (95% CI: 0.8243-0.93) in the training cohort and 0.8889 (95% CI: 0.5809–0.8889) in the test cohort (Fig. 6A). Further validation confirmed its high classification accuracy via confusion matrices (Fig. 6B) and excellent calibration with the ideal curve (Fig. 6C). Within a decision-analytic framework, the decision curve analysis (Fig. 6D) confirmed that our model provides a superior net benefit across a clinically relevant threshold probability range (10–70%), underscoring its value in reducing the net rate of clinical decision errors relative to default strategies. This finding implies that the model optimizes patient outcomes by maximizing true positives and minimizing false positives. The nomogram (Fig. 6E) operationalizes this benefit by facilitating point-of-care risk stratification. Clinicians can efficiently compute an individualized risk probability by summing the points corresponding to a patient’s predictor values and projecting the total points to the final risk axis, thereby generating a quantitative evidence base for shared decision-making.

Fig. 6.

The risk of liver cirrhosis model evaluation. (A) The receiver operating characteristic curve(ROC). (B) Confusion matrix. (C) Plot depicting the calibration. (D) Decision curve analysis curve. (E) Nomogram for predicting the risk of liver cirrhosis

The selection of these proteins is supported by their established biological relevance to liver disease processes. L-selectin (encoded by the SELL gene) is a key adhesion molecule involved in lymphocyte homing to the liver, a process central to the inflammatory response in chronic viral hepatitis. IGLV5-37 represents an immunoglobulin lambda variable region, reflecting the significant involvement of B-cell and humoral immune responses in HIV/HBV co-infection. The complete panel of selected proteins encompasses functions related to immune regulation, inflammatory response, and metabolic pathways known to be dysregulated in the progression of liver cirrhosis.

Validation of L-selectin as a plasma marker in HIV/HBV co-infected patients with liver cirrhosis

A cohort of 90 plasma samples from HIV/HBV co-infected individuals, with or without liver cirrhosis, was assessed for L-selectin expression. As presented in Fig. 7, the plasma L-selectin concentration was significantly elevated in cirrhotic patients (25.78 ± 1.79 ng/ml) compared to their non-cirrhotic counterparts (13.45 ± 1.03 ng/ml) (t = 5.981, p < 0.05). The diagnostic performance evaluation revealed that L-selectin exhibited a sensitivity of 75.6% (34/45) and a specificity of 84.4% (38/45) for detecting liver cirrhosis. The positive and negative predictive values were calculated to be 82.9% (34/41) and 77.6% (38/49), respectively.

Fig. 7.

Plasma L-selectin levels in HIV/HBV co-infected patients with and without liver cirrhosis

Discussion

HIV co-infection hastens the progression of liver cirrhosis, a stage of HBV liver disease that is frequently irreversible. It is still challenging to predict liver cirrhosis and to take early action using noninvasive serological techniques. Here, for the first time in people living with HIV/HBV coinfection, we performed a proteome study to profile plasma protein abnormalities in patients with and without liver cirrhosis, as well as develop a model to predict the risk of liver cirrhosis. Ultimately, 13 core proteins, including L-selectin, were successfully selected out of 111 differentially expressed proteins (DEPs) by the LASSO model, and a clinical risk prediction model of liver cirrhosis based on L-selectin was built. Furthermore, the accuracy of L-selectin as a novel biomarker for liver cirrhosis prediction was validated in a larger cohort of HIV/HBV co-infection. This study will aid in the investigation of the mechanism by which HIV influences the progression of HBV liver disease, as well as the identification of possible treatment targets for liver cirrhosis.

In people with HIV/HBV coinfection with and without liver cirrhosis, 111 proteins were discovered to be differentially expressed, with 65 upregulated and 46 downregulated in those with liver cirrhosis. This finding suggested that plasma protein differences between patients with and without liver cirrhosis were relatively extensive. The GO terms (BP) for these DEPs were discovered to be highly involved in immune system process regulation, complement activation, immunoglobulin mediated immune response, B cell mediated immunity, and humoral immune response mediated by circulating immunoglobulin. Our study consistently observed significantly lower counts of white blood cells (WBC) and platelets (PLT) in the cirrhosis cohort compared to non-cirrhosis control. This clinical phenomenon, often attributed to hypersplenism, appears to be intricately linked to the altered immune responses identified through our bioinformatic analyses. The leukopenia and thrombocytopenia in cirrhosis are not merely hematological side effects but are deeply embedded in the disease’s immunopathology. Our bioinformatic analysis provides a molecular rationale, demonstrating that these cytopenias are closely associated with a state of chronic immune activation and inflammation.

It also found that these DEPs were highly implicated in complement and coagulation cascades, as well as cell adhesion molecules, using KEGG pathway enrichment analysis. Although HIV infection primarily causes an decrease in the number and function of T lymphocytes, particularly CD4 + T lymphocytes, the human body has a complex immune network. The enrichment of differential proteins in plasma suggests that humoral immunity changes significantly. The complement system is a proteolytic cascade in blood plasma that serves as a conduit for non-specific natural defenses against pathogens. The main effects of complement activation include pathogen psonization, recruitment of inflammatory and munocompetent cells, and direct pathogen death. Another study [13] discovered complement and coagulation signaling pathways in HIV/HBV coinfected patients. As a result, we can conclude that during HBV coinfections, HBV may block complement system activation via a mechanism involving all of these proteins. The DEPs that were primarily enriched in DO items were associated with AIDS and glucose metabolism disorders in the LC group compared to the NC controls. This finding is exactly compatible with the AIDS population in this study, and the glucose metabolism illness caused by cirrhosis will be discussed later in this study. The disease ontology (DO) enrichment analysis confirmed the accuracy of the results.

The significant enrichment of the Complement and Coagulation Cascades pathway in our cirrhosis cohort underscores the critical role of systemic inflammation and dysregulated immunity in disease progression. This finding aligns with the contemporary understanding of cirrhosis as a state of persistent immune activation. Recent research elucidates that chronic liver injury, regardless of etiology, leads to a vicious cycle of hepatocyte death, inflammatory cell infiltration, and aberrant tissue repair. As highlighted by [14], damaged hepatocytes and hepatic stellate cells (HSCs) release damage-associated molecular patterns that perpetuate innate immune signaling and contribute to a pro-fibrotic microenvironment. Concurrently, there is a well-documented dysbiosis of the gut-liver axis. The study by [15] further demonstrates that gut-derived microbial products, such as lipopolysaccharide, continuously fuel this process via Toll-like receptor signaling, leading to the activation of HSCs and their transformation into collagen-producing myofibroblasts. This sustained inflammatory drive not only promotes fibrogenesis but also directly activates the complement and coagulation systems within the liver sinusoids, creating a pro-thrombotic milieu that exacerbates portal hypertension and compromises hepatic perfusion. Therefore, our pathway analysis corroborates the model of cirrhosis as a chronic wound-healing response gone awry, driven by intertwined mechanisms of innate immune hyperactivation and coagulation dysfunction.

Our KEGG pathway analysis also revealed other significantly enriched entries included pathways related to COVID-19 in the liver cirrhosis (LC) cohort, except for the Complement and Coagulation Cascades pathway. This concurrent enrichment may not be coincidental but could reflect a shared underlying pathophysiological mechanism. A growing body of evidence has established that SARS-CoV-2 infection can lead to profound dysregulation of the complement and coagulation systems, contributing to the immunothrombosis seen in severe COVID-19 [16, 17]. Furthermore, similar to the chronic liver injury in cirrhosis, COVID-19 can induce acute hepatic damage through systemic inflammatory responses and endothelial dysfunction [18]. It is important to note that while all study participants had lived through the 2019–2020 pandemic period, they were asymptomatic for COVID-19 at the time of recruitment and did not undergo specific SARS-CoV-2 testing as part of this study. Therefore, the enrichment of the COVID-19 pathway is unlikely to be driven by an active infection. Instead, we postulate that the activated complement and coagulation cascades in cirrhosis create a systemic pro-inflammatory and hypercoagulable state that mirrors, in part, the host response to SARS-CoV-2. This shared immunological milieu, characterized by innate immune activation and a predisposition to thrombosis, may explain the observed pathway overlap. This finding underscores the profound inflammatory burden in advanced liver disease, which can parallel pathways activated during major viral infections.

L-selectin, a member of the selectin family, is constitutively expressed on lymphocytes, monocytes, granulocytes and some hematopoietic malignancies, and can isolate from the cell surface after cell activation [19]. Through selectin-mediated platelet and endothelial cell interactions, L-selectin binding activates leukocytes and initiates signaling cascades that govern immune response in the tumor microenvironment [20]. This study developed a model incorporating L-Selectin to predict the risk of liver cirrhosis in patients with HIV/HBV coinfection.

The correlation between L-selectin and liver disease has been demonstrated in a few studies. In chronic liver disease, monocyte L-selectin expression is specifically elevated in decompensated cirrhosis and shows a positive correlation with the Model for End-Stage Liver Disease (MELD) score, as reported in one study [21]. Cirrhosis-associated immune dysfunction is an immunological imbalance that frequently exacerbates liver cirrhosis. As previously observed [21], elevated percentages of L-selectin⁺ monocytes was observed in cirrhotic patients [22]. These results may suggest increased hepatic recruitment of inflammatory monocytes exacerbates tissue injury in patients with chronic liver disease. A small proportion of chronic liver illnesses that are not appropriately managed may develop to hepatocellular carcinoma. A study using CIBERSORT assay to determine the composition of immune cells in tumors and adjacent tissues found that the novel LINC00261/MiR105-5p/SELL pathway was associated with overall survival of patients with HCC and influenced the efficacy of immunotherapy in HCC [23].

L-selectin not only influences inflammatory monocyte recruitment in the liver, but it also plays an important role in HIV entry and infection. L-selectin was identified as a viral adhesion receptor on CD4 + T cells and infected CD4⁺ T cells are stimulated to gradually shed L-selectin after HIV entry [9, 24]. The significant expression of L-selectin in cirrhotic patients’ peripheral blood may imply that HIV entry and infection are easier, and active HIV replication is more likely once ART protection is eliminated. Furthermore, a higher level of plasma L-selectin may imply that there are more infected CD4 + T cells, which could be one of the causes of liver cirrhosis in this population.

L-selectin has multiple roles mediating inflammatory diseases and metastatic adhesion receptors. Increased L-selectin expression on the surface of leukocytes is a marker of leukocyte activation. Recent studies have confirmed that L-selectin can cause endothelial injury and intravascular microthrombus formation by promoting the adhesion of white blood cells, platelets and endothelial cells, which plays an important role in the occurrence and development of diabetic macroangiopathopathy and microangiopathopathy. The circulating polymorphonuclear leukocytes(PMN) of patients with acute diabetic ketoacidosis (DKA) can display a primed phenotype characterized by L-selectin shedding. Thus, the novel neutrophil phenotypic signature is proposed as a new biomarker for DKA risk stratification in patients with diabetes [25]. In a follow-up study, multiple logistic regression analysis reveals that level of L-selectin is associated with diabetic nephropathy and retinopathy [26]. A liver cirrhosis risk prediction model that included L-selectin was developed in this work, and differential protein enrichment analysis revealed that glucose metabolism was strongly enriched in DO entries. The aforementioned research bolsters the validity of the findings of this investigation and raises the possibility that liver cirrhosis and blood glucose metabolism are strongly associated.

The liver has insulin receptors, which plays a critical role in glucose homeostasis. A certain degree of increased insulin resistance, pancreatic β cell dysfunction and abnormal glucose metabolism were found in subjects of chronic hepatitis B (CHB) with liver cirrhosis [27–29]. Glucose metabolism disorder is also common in patients with chronic hepatitis B cirrhosis. The prevalence of diabetes was significantly higher in subjects of CHB with liver cirrhosis than that in those without liver cirrhosis (47.1% vs. 16.8%) [30]. In a study, dysfunctional energy metabolism was found in B cells from subjects with decompensated liver cirrhosis related to HBV Chronic infection [31]. In addition, the study revealed that AKT/mTOR pathway plays a critical role in the regulations of B cell metabolic process [31]. The inclusion of glucose metabolism in the DO enrichment analysis of differential proteins was discovered in this study, which has significant clinical guiding significance. Clinicians should be encouraged to pay particular attention to early warning indications of glucose metabolism anomalies in order to standardize the management of patients with HBV-induced cirrhosis from the standpoint of proteomics.

We acknowledge several limitations in the present study that warrant careful consideration. First, the proteomic analysis was conducted on a relatively small patient cohort, which may restrict the generalizability of the identified differentially expressed proteins (DEPs) and introduce potential bias in the biomarker discovery phase. Nevertheless, the subsequent validation of L-selectin in a larger HIV/HBV co-infected population strengthens its relevance as a candidate biomarker for liver cirrhosis in this specific clinical context. Second, the cirrhosis prediction model developed in this study was derived from and validated in a population with HBV-related liver disease. Its applicability to cirrhosis resulting from other etiologies, such as alcohol-induced liver disease or HCV infection, remains unclear and necessitates external validation in broader and more diverse patient cohorts. Third, due to constraints in the availability and performance of current commercial assay kits, this study focused solely on the experimental validation of L-selectin. Other promising DEPs identified through proteomic screening, such as IGLV5-37, could not be verified in the same manner. Future studies should prioritize the development of reliable detection methods to experimentally confirm the roles of these proteins in cirrhosis pathogenesis. Fourth, another limitation was that no specific techniques for the depletion of high-abundance proteins was employed prior to proteomic analysis. However, other strategies to “penetrate” the interference from albumin were adopted, which primarily including the use of High-Resolution Fractionation and High-Accuracy Mass Spectrometry. Our findings are representative of the “detectable” plasma proteome under these conditions. Fifth, one limitation of this study is the lack of healthy controls or mono-infected control groups. Therefore, our findings primarily reveal differences between cirrhotic and non-cirrhotic individuals within the HIV/HBV-coinfected population, and comparisons with broader populations should be interpreted with caution. Future studies incorporating such control groups will help to further clarify the distinct characteristics of HIV/HBV coinfection in the context of liver cirrhosis progression. Finally, the cross-sectional design of this study limits the ability to establish a causal relationship between L-selectin levels and the progression of liver cirrhosis. Longitudinal studies tracking dynamic changes in L-selectin throughout the course of liver disease will be essential to clarify its temporal role and potential utility in monitoring disease evolution. Despite these limitations, the bioinformatics and partial experimental findings presented here offer valuable insights and a foundation for further investigation into the molecular mechanisms underlying cirrhosis in HIV/HBV co-infected individuals.

Conclusions

In this cohort of 13 matched patient pairs with HIV/HBV coinfection (with vs. without liver cirrhosis), we identified 111 differentially expressed proteins (DEPs), including 65 that were upregulated and 46 that were downregulated in the cirrhosis group. Functional enrichment analysis revealed that these DEPs were significantly involved in immune-related biological processes critical to chronic viral infection and hepatic inflammation. This finding was further supported by KEGG pathway analysis, which highlighted the Complement and Coagulation Cascades pathway - a key driver of chronic liver injury and fibrosis - as the most significantly enriched in the LC group compared to the NC controls. Disease Ontology analysis confirmed the expected link to human immunodeficiency virus infection and also identified disorders of glucose metabolism, a common complication of chronic liver disease. From the 13 DEPs, a clinical prediction model for cirrhosis was developed using five-fold cross-validation based on AUC. The optimal model included immunoglobulin heavy variable 5–37 (IGHV5-37) and L-selectin. The clinical relevance of L-selectin was validated in an independent cohort of 90 HIV/HBV coinfection patients, where its plasma levels were significantly elevated in those with cirrhosis.

Supplementary information

Below is the link to the electronic supplementary material.

Abbreviations

DEPs

Differentially expressed proteins

LC–MS

Liquid chromatography-mass spectrometry

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

DO

Disease Ontology

CD62L

Plasma L-selectin

HBV

Chronic hepatitis B virus

HIV

Human immunodeficiency virus

APRI

AST-to-platelet ratio index

AAR

AST/ALT ratio

CDS

Cirrhosis discriminant score

API

Age-platelet index

IBD

Inflammatory bowel disease

SD

Standard deviation

BMI

Body Mass Index

cART

Combination antiretroviral therapy

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

Tbil

Total bilirubin

Alb

Albumin

FBG

Fasting blood glucose

MELD

Model for End-Stage Liver Disease

INR

International Normalized Ratio

CRP

C-Reactive Protein

IL-6

Interleukin-6

Author contributions

YX and YXZ conceived of and designed the study. RRY performed the experiments, collected and analyzed the data. PX collected the samples. XEG provided suggestions. All authors contributed to the study and approved the submitted version.

Funding

This study was supported by National Natural Science Foundation of China (Grant No. 82003511), Hubei Province Key Research and Development Project (Grant No. 2020BCB025), Discipline and Platform Construction Project of Center for AIDS Research, Wuhan University (PTPP2023002).

Data availability

The data presented in the study are deposited in the ProteomeXchange Consortium via the iProX repository, accession number PXD046369. IPX0007140001 in PXD046369 is the original sequence information of this study.

Declarations

Ethical approval

No patient identifying data was used. A signed informed consent form was obtained the participating ID specialists. This study was approved by the Institutional Ethics Committee of Zhongnan Hospital of Wuhan University (Approval No: 2021022), which was conducted in accordance with the principles of the Declaration of Helsinki and complied with the Good Clinical Practice Guidelines

Consent for publication

Not applicable.

Artificial intelligence software use

No artificial intelligence software was used to write this report.

Competing interests

The authors declare no competing interests.