Characteristics of the study population

Baseline clinical and biochemical characteristics for the entire cohort, as well as for two groups categorized by MASH status (non-MASH vs MASH), are summarized in Table 1. The overall population consisted of middle-aged individuals (46 ± 11 years), with severe obesity (BMI: 43.4 ± 6.0 kg/m2), predominantly women (75%), and exhibited insulin resistance (HOMA-IR: 3.0 [1.6; 4.8]) [24]. Other glucose metabolism parameters, including FPG, insulin and HbA1c were mostly normal or near abnormal thresholds, while plasma lipids, such as triglycerides and LDL-C, were slightly elevated, likely reflecting partial treatment within the population. Accordingly, 72.3% had dyslipidemia, but only 29.1% received lipid-lowering drugs, while 45.9% had diabetes with 43.9% on anti-diabetic medications (Table 1). The prevalence of hypertension was high (72.3%), yet only 35.8% of affected individuals received antihypertensive treatment. Overall, solely 15.5% of the study population had neither a diagnosis nor treatment for comorbidities.

Table 1 Sociodemographic, clinical and biological characteristics of the study population according to MASH status

Significant differences were observed between the non-MASH and MASH groups. Specifically, the diagnosis of MASH was associated with a marked difference in sex distribution, with women being significantly less represented in the MASH group (64.8% versus 80.9% in the non-MASH group, p = 0.03). Individuals with MASH were more insulin resistant, as documented by higher HOMA-IR values compared to the non-MASH (p = 0.002), along with significantly higher level of FPG, insulin, and HbA1c. Additionally, triglycerides were significantly higher, and HDL-C levels were significantly lower in the MASH group compared to the non-MASH group. Consistent with these metabolic differences, individuals with MASH had a higher prevalence of diabetes and dyslipidemia, but showed no difference in the prevalence of hypertension. Furthermore, individuals with MASH had higher levels of transaminases (AST and ALT), accompanied by a lower AST/ALT ratio (p < 0.001). No significant differences were observed between groups in terms of anthropometric measures (BMI, waist circumference) and lifestyle risk factors such as smoking and alcohol consumption.

Plasma concentration of apolipoproteins by MASH status

Baseline concentrations of 14 plasma apolipoproteins, stratified by MASH status, are presented in Table 2. Plasma levels of apoC-III and apoL1 were nominally higher in the MASH group compared with the non-MASH group (p = 0.038 and p = 0.025, respectively). Levels of apoC-I, apoC-II, apoE, and apoJ were also higher in the MASH group (p < 0.10). After correction for multiple testing, all six apolipoproteins (apoC-I, apoC-II, apoE, apoJ, apoC-III, and apoL1) showed an FDR-adjusted p-value (q) of 0.217 (Supplementary Table 3). In contrast, apoA-I, apoA-II, apoA-IV, apoB100, apoD, apoF, apoH, and apoM did not differ nominally between groups (p > 0.10).

Table 2 Plasma concentration of apolipoproteins according to MASH statusAssociation between plasma apolipoprotein levels and biochemical parameters

For the study population, Spearman’s rank correlation coefficients (rs) were calculated to assess associations among plasma apolipoproteins (Fig. 2 and Supplementary Table 1), as well as between apolipoproteins and biochemical parameters at baseline (Fig. 3 and Supplementary Table 2).

Fig. 2

Spearman correlations between plasma apolipoprotein concentrations. Plasma apolipoproteins were measured in the whole study population (n = 148). Size and color, according to the color scale detailed on the right, represent the strength of the correlation intensity between variables, with Spearman’s correlation coefficient (rs) ranging from −1 to + 1. Only correlations significant at the 5% threshold are shown. Non-significant correlations are indicated by empty boxes. Blue dot, positive correlation; red dot, negative correlation

Fig. 3

Spearman correlations between plasma apolipoprotein concentrations and bioclinical variables. Plasma apolipoproteins were measured in the whole study population (n = 148). Size and color, according to the color scale detailed on the right, represent the strength of the correlation intensity between variables, with Spearman’s correlation coefficient (rs) ranging from −1 to + 1. Only correlations significant at the 5% threshold are shown. Non-significant correlations are indicated by empty boxes. Blue dot, positive correlation; red dot, negative correlation. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; DBP, diastolic blood pressure; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; SBP, systolic blood pressure

Many apolipoproteins were intercorrelated, consistent with their shared transport by lipoproteins [25] (Fig. 2). Specifically, apoA-I, the primary HDL apolipoprotein, showed positive correlations with apoA-II, apoE, apoL1, and apoM. Similarly, significant positive intercorrelations were observed among apoB100—a major protein constituent of VLDL and LDL—apoE, apoC-I, apoC-II, and apoC-III. Among other apolipoproteins exhibiting multiple correlations, apoL1 was positively correlated with apoA-I, apoC-I, apoC-III, apoE and apoM. ApoM displayed a similar correlation profile to apoL1, with additional correlations to apoA-II, apoC-II, and apoD, consistent with its transport by both HDL and LDL [26]. In contrast, apoA-IV was positively correlated only with apoA-I and negatively correlated with apoB100, while apoF was negatively correlated only with apoD and apoH. Additionally, apoD was positively correlated with both apoC-I and apoC-II, while apoH showed positive correlations with apoB100 and apoC-III. Notably, apoJ was the only apolipoprotein that did not show any correlation with the other apolipoproteins (Fig. 2).

As expected, most plasma apolipoproteins were associated with lipid levels (Fig. 3). Half of the measured apolipoproteins, including apoA-I, apoA-II, apoB100, apoC-I, apoC-II, apoE, and apoM, showed a positive correlation with total cholesterol. ApoB100, apoC-I and apoE were positively associated with LDL-C, while apoC-III exhibited a negative correlation with HDL-C. In contrast, apoA-I, apoA-II and apoM were positively correlated with HDL-C. Additionally, apoC-I, apoC-II, apoC-III and apoE displayed a positive correlation with triglycerides.

Regarding glycemic parameters (Fig. 3), apoA-IV, apoC-I, apoC-III, and apoF levels were positively correlated with FPG. Additionally, apoC-I was positively correlated with insulin, and both apoC-I and apoC-III showed significant positive correlations with HOMA-IR. ApoA-IV and apoC-III were also positively correlated with HbA1c.

ApoC-III, apoE and apoL1 exhibited positive correlations with plasma transaminase levels (AST and ALT). ApoA-IV and apoC-III were positively correlated with age, whereas apoA-I was negatively correlated with BMI and waist circumference. Furthermore, apoA-IV was positively associated with SBP, while apoC-I and apoL1 were positively associated with DBP. Noteworthily, no correlation was observed between apoH or apoJ and any of the measured biochemical parameters (Fig. 3).

Plasma apolipoprotein levels and MASH diagnosis

The results of multivariable logistic regression analyses examining plasma apolipoproteins, HDL-C, and LDL-C levels in relation to MASH status are presented in Fig. 4.

Fig. 4

Multivariable logistic regression models assessing MASH status with plasma apolipoprotein levels as independent variables. Model 1: Adjusted for age and sex. Model 2: Model 1 + HOMA-IR, plasma triglyceride levels, waist circumference, and the AST/ALT ratio. Model 3: Model 1 + diabetes, dyslipidemia, and hypertension. The 95% confidence interval (95% CI), and p-value associated with each odds ratio (OR) are reported. Dyslipidemia was defined as LDL-C ≥ 160 mg/dL and/or triglyceride levels ≥ 150 mg/dL and/or use of lipid-lowering treatment. Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mmHg and/or diastolic blood pressure (DBP) ≥ 90 mmHg at rest, and/or antihypertensive treatment. Diabetes was defined as fasting plasma glucose (FPG) ≥ 126 mg/dL and/or hypoglycaemic treatment. ALT, alanine aminotransferase; AST, aspartate aminotransferase; HOMA-IR, Homeostatic Model Assessment of Insulin Resistance

In a model adjusted for age and sex (Model 1), plasma apoE, apoJ, and apoL1 were positively associated with a diagnosis of MASH. The respective odds ratios (OR) per 1-SD increase were as follows: apoE (OR = 1.23, 95% CI [1.00–1.51], p = 0.046), apoJ (OR = 1.07, 95% CI [1.00–1.14], p = 0.040), and apoL1 (OR = 1.53, 95% CI [1.09–2.16], p = 0.015).

These associations persisted for apoJ and apoL1 after further adjustment for traditional MASH risk factors, including HOMA-IR, plasma triglyceride levels, waist circumference, and the AST/ALT ratio (Model 2: apoJ, OR = 1.08, 95% CI [1.00–1.16], p = 0.047; apoL1, OR = 1.54, 95% CI [1.00–2.37], p = 0.052), as well as for comorbidities such as diabetes, dyslipidemia, and hypertension (Model 3: apoJ, OR = 1.07, 95% CI [1.00–1.15], p = 0.055; apoL1, OR = 1.60, 95% CI [1.09–2.35], p = 0.017).

Notably, among plasma apolipoproteins, those with differential concentration based on MASH status (apoC-III and apoL1, Table 2) and/or those associated with MASH in multivariable logistic regression (apoE, apoJ, and apoL1, Fig. 4) were further examined at the transcriptional level by measuring their corresponding mRNA expression in the liver samples from the current cohort (Fig. 5), as well as the plasma levels in an independent external cohort of obese individuals, similarly classified as non-MASH (no liver alterations or simple steatosis) or MASH [22] (Fig. 6). As shown in Fig. 5, hepatic mRNA levels of APOC3, APOJ, and APOL1—but not APOE—were higher in individuals with obesity who were histologically diagnosed with MASH compared to those without MASH (p = 0.0348, 0.0025, and 0.007, respectively). In the external cohort, plasma levels of apoJ were also higher in the MASH group compared to the non-MASH group (p = 0.004, Fig. 6), whereas no differences were observed for apoL1, apoC-III, or apoE.

Fig. 5

APOC3, APOE, APOJ, and APOL1 mRNA expression levels in liver based on the MASH status of individuals with obesity. Subjects who underwent liver biopsy (n = 141) during bariatric surgery were stratified into two groups: those without liver disease or with simple steatosis (non-MASH, n = 87) and those with MASH (n = 54). Data are presented as mean ± SEM. Two-tailed unpaired Student’s t-test was used for MASH status comparison, and p values < 0.05 are reported

Fig. 6

Signal intensity of apoC-III, apoE, apoJ, and apoL1 peptides in plasma from an external cohort of individuals with obesity undergoing bariatric surgery, stratified into two groups: those without liver disease or with simple steatosis (non-MASH, n = 127) and those with MASH (n = 33). Data are presented as median with interquartile range (IQR); whiskers indicate minimum and maximum values. Comparisons between groups were performed using a two-tailed unpaired Student’s t-test, and p values < 0.05 are reported

Incremental predictive values of apoJ and apoL1 for MASH diagnosis beyond traditional risk factors

We evaluated whether apoJ and apoL1 improved prediction beyond a core risk factor model (age, sex, HOMA-IR, plasma triglycerides, waist circumference, and AST/ALT ratio) using likelihood ratio tests (LRT), net reclassification improvement (NRI), and integrated discrimination improvement (IDI) (Table 3).

Table 3 Incremental predictive values for apoJ and apoL1 for MASH diagnosis beyond clinical and biological risk factors

LRT analyses showed that apoJ (ΔDeviance = 4.08, p = 0.043) and apoL1 (ΔDeviance = 3.95, p = 0.047) each significantly improved model fit, with their combination yielding additional improvement (ΔDeviance = 7.53, p = 0.023). NRI confirmed the strongest effect for the combined model (NRI total = 0.39, p = 0.026), driven mainly by improved classification of non-MASH individuals (NRI non-event = 0.31, p = 0.0023), corresponding to 31% of true negatives correctly reclassified into lower-risk categories compared with the core model. When assessed separately, apoJ showed a greater effect than apoL1, also predominantly in non-MASH individuals (NRI non-event = 0.29, p = 0.005 versus 0.22, p = 0.037, respectively). Consistently, IDI was highest for the combined model (IDI = 0.04, p = 0.034), indicating a 4% increase in discrimination between MASH and non-MASH individuals compared with the core model.

Overall, adding apoJ and apoL1 to the core model significantly improved predictive performance, with the greatest benefit observed in the correct classification of non-MASH individuals.