Serum Proteomics of Insulin Resistance Disorders Distinguish MASLD From Lipodystrophy and Insulin Receptor Defects

Abstract

Background and Aims

Metabolic dysfunction-associated steatotic liver disease (MASLD) is linked to obesity, metabolic syndrome, and insulin resistance (IR). In IR associated with obesity or lipodystrophy (LD), caused by missing adipose tissue, hyperinsulinemia stimulates hepatic lipogenesis, causing steatotic liver disease (SLD). In contrast, insulin receptor pathogenic variants (INSR), despite hyperinsulinemia and IR, do not promote hepatic lipogenesis and steatosis. We aimed to understand SLD pathophysiology by comparing serum proteomic signatures of MASLD, LD, and INSR.

Methods

Single-center study of fasting serum proteome using SomaScan assay in 30 LD, 29 INSR, and 16 people with MASLD. Key targets were assessed in the hepatic transcriptome of patients with MASLD (n = 19) and in HepG2 cells.

Results

Of 6412 proteins, 567 differed between at least 2 groups. The proteome profiles clearly separated INSR from the other groups, whereas MASLD and LD displayed similarity, primarily in proteins involved in metabolism, liver injury, and fibrosis. Several proteins were uniquely elevated in INSR and LD compared to MASLD, particularly Factor IX and liver-expressed antimicrobial peptide 2. Serum levels of Factor IX and liver-expressed antimicrobial peptide 2 correlated with their hepatic expression, with in vitro expression unaffected by leptin, insulin, or glucose treatment.

Conclusion

This first comparative proteomics study identified shared and unique pathways in IR disorders. MASLD and LD share features, suggesting similar drivers of liver disease progression. Key liver-derived proteins differ between MASLD, INSR, and LD because of unique disorder features. We demonstrate the utility of proteomics and rare disorder studies in interrogating pathogenesis of common disorders like MASLD.

Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most prevalent chronic liver disease in the United States and worldwide (1). It is a component of the obesity-associated metabolic syndrome and pathophysiologically linked to insulin resistance (IR). Energy surplus in MASLD, usually from overnutrition, accompanied by hyperinsulinemia, leads to high rates of de novo lipogenesis (DNL) and triglycerides (TG) deposition in liver tissue. Accumulation of lipotoxic metabolites in turn leads to steatohepatitis and liver fibrosis (2, 3). Studies to dissect pathogenic mechanisms of individual pathways, such as insulin signaling, are typically conducted using animal models where genetic manipulation is feasible but are difficult to perform in humans. In this context, rare diseases offer a unique opportunity for mechanistic studies.

Lipodystrophy (LD) syndromes are marked by a deficiency of adipose tissue. Despite the differences in adipose tissue mass, MASLD and LD share common adverse metabolic phenotypes, such as IR, diabetes, hypertriglyceridemia, atherosclerosis, and increased cardiovascular mortality (4). As in obesity-associated MASLD, hyperinsulinemia in LD promotes DNL, and as a result, steatohepatitis and fibrosis (5).

The rare condition of insulin receptor pathogenic variants (INSR) is characterized by hyperinsulinemia and marked insulin resistance because of a lack of insulin signaling through its receptor. Previous studies demonstrated that, in contrast to MASLD and LD, hyperinsulinemia in INSR does not promote hepatic lipogenesis, steatosis, or steatohepatitis, likely due to low DNL (5) and a shift in hepatic substrate use toward gluconeogenesis (6).

Prior research on proteomics in MASLD helped to gain an in-depth understanding of the pathophysiology of hepatic involvement (7) and identify biomarkers for fibrosis and risk for adverse outcomes (8, 9). However, no previous studies have explored the proteome in the context of liver disease in LD or INSR. The aim of this study was to use serum proteomics to explore the pathophysiology of hepatic involvement in 3 different disorders with IR. Our objective was to identify the commonalities and unique differences in serum proteome in common MASLD, LD, and INSR.

Materials and Methods

Human Subjects

This was a single-center retrospective study of subjects with MASLD, LD, and INSR. Subjects with MASLD participated in a study exploring postprandial metabolism (NCT02520609) (10). MASLD was confirmed either by liver biopsy or by presence at least 2 of the following: hepatic steatosis on imaging, elevated aminotransferases levels, and metabolic syndrome. Subjects with INSR or LD were enrolled in a long-term natural history study of IR (NCT00001987). Subjects with INSR were matched to those with LD based on age, sex, pubertal stage, and diabetes control as measured by hemoglobin A1c (HbA1c), whereas MASLD subjects were recruited separately and later from an adult population. This approach was chosen due to extreme rarity of INSR and LD, which necessitated prioritizing comparisons between these conditions over age-matched subjects. The studies were approved by the institutional review board of the National Institutes of Health (NIH), approval ID 15-DK-0174 and 76-DK-0006, and conducted in accordance with both the Declarations of Helsinki and Istanbul. All subjects or their guardians provided written informed consent; minors provided assent when age appropriate. Pooled serum from deidentified healthy subjects was used as an internal control for cross-plate calibration.

Studies of liver gene expression were done using liver biopsy samples from a separate cohort of people with MASLD (n = 19) participating in a clinical trial of vitamin E (clinicaltrials.gov NCT01792115). All samples were obtained before initiating treatment. Inclusion and exclusion criteria were previously described (11). The study was approved by the National Institute of Diabetes and Digestive and Kidney Diseases/National Institute of Arthritis and Musculoskeletal and Skin Disease institutional review board, and all subjects provided written informed consent.

Clinical Testing

Anthropometric data including weight, height, and body mass index (BMI) were measured in the fasting state. Blood was collected after an 8- to 12-hour fast for measurement of glucose, insulin, HbA1c, liver-associated enzymes, lipids, and blood count using standard methods at the Department of Laboratory Medicine, NIH Clinical Center.

Proteomic Analysis

Serum samples were collected in fasting state and frozen at −8 °C. Samples were analyzed using the proteomic aptamer based SomaScan assay v4.1 (SomaLogic). SomaScan assay performance and characteristics were previously described (12, 13). Protein concentrations are expressed in relative fluorescence units and log-transformed before analysis. All data were normalized, calibrated, and underwent quality control check.

Liver Gene Expression

Hepatic gene expression was obtained by RNA sequencing from human liver biopsies as previously described in detail (11). Cohort characteristics are presented in Table S1 (14). Briefly, percutaneous liver biopsy samples were placed in RNALater or flash frozen at the bedside and stored in −8 °C for batch processing. After RNA extraction, PolyA-enriched RNA libraries were used for 75-bp paired-end sequencing and gene expression counted with htseq-count.

In Vitro Studies

HepG2 cells were cultured in DMEM (Corning, 10-013-CV) with 10% fetal bovine serum (Sigma F2442-500ML) and 1% penicillin/streptomycin (Corning 30-002-Cl). In a separate experiment, the medium was supplemented with oleate and palmitate (1:1 ratio, 250 μM in BSA) or BSA control. Cells were maintained at 37 °C at 5% CO₂. Cells were treated with insulin (Millipore Sigma, I9278-5ML) or leptin (Millipore Sigma, L4146-1MG) at indicated concentrations. For glucose treatments, media comprised a low-glucose DMEM (Corning 10-014-CV, 5.5 mM glucose), 10% fetal bovine serum, and 1% penicillin/streptomycin was prepared. Glucose was supplemented up to indicated concentrations, and cells were treated for 24 hours. RNA was harvested using the Qiagen RNeasy Mini kit (74106) and DNase (Qiagen, 79256) according to manufacturer's instructions, and cDNA was synthesized (ThermoFisher, 4368814). Gene expression was quantified using Sybr Green (ThermoFisher, 100029284) using primers purchased from Integrated DNA Technologies with sequences listed in Table S2 (14).

Statistical Methods

Statistical analysis was performed using RStudio version 4.2.1 (Posit Software PBC, Boston, Massachusetts) and GraphPad Prism version 9.4.0 (GraphPad Software, Boston, Massachusetts). Baseline characteristics of the 3 groups are presented as median and interquartile range or percentage. Continuous data with normal distribution were compared by ANOVA with post hoc pairwise comparisons using the Tukey method, and data with nonnormal distribution were compared using the Kruskal-Wallis test with post hoc pairwise comparisons using the Dunn method. Chi-square test or Fisher exact test was used for categorical variables.

Proteins that differed between groups were identified by ANOVA, followed by pairwise comparison using the Mann-Whitney U test. Statistical significance was determined after adjustment for multiple comparisons using the Benjamini-Hochberg procedure with q-value ≤.01 and fold change cutoff ≥1.5. The proteome of each group was visualized using a 3-dimensional (3D) volcano plot, a method developed and previously published by Lewis et al for the analysis of 3D data from the human synovial transcriptome (15). For unsupervised clustering, z-scores of protein levels in each of the groups, MASLD, LD, and INSR were calculated, and visualization was performed using R package pheatmap. Spearman rank correlation and linear regression were used to determine relationships between key proteins and clinical parameters with P-value <.05 considered significant. Log transformation was performed for fasting insulin, Homeostatic Model Assessment for Insulin Resistance (HOMA-IR), and triglycerides.

Pathway Analysis

Gene Ontology (GO) enrichment analysis was used to determine the affected processes within distinct protein clusters. Visualization was performed using R package goplot.

Results

Cohort Description

The study included 30 subjects with LD, 29 with INSR, and 16 with MASLD, who were seen between 2001 and 2020 and met eligibility criteria. Sixteen of 30 (53.3%) LD subjects had generalized lipodystrophy (GLD), and 14/30 (46.7%) had partial lipodystrophy. Subjects with GLD had pathogenic variants in AGPAT2 (7/16), BSCL2 (5/16), and LMNA (2/16), whereas subjects with partial lipodystrophy had pathogenic variants in LMNA (9/14) and PPARG (4/14). Two subjects with LD had unknown genetics, and 1 had nonhereditary acquired GLD. Among subjects with INSR, 18/29 (62.1%) had biallelic, recessive pathogenic variants and 11/29 (37.9%) had monoallelic, dominant negative pathogenic variants. Genetic assessment in MASLD group was not available.

Subjects with LD and INSR were younger and had lower BMI compared to subjects with MASLD. As expected, the groups differed in their baseline characteristics, especially in those related to metabolic processes (Table 1). MASLD and LD had higher liver-associated enzymes compared to INSR. Subjects with LD had the highest levels of triglycerides and fasting glucose, while subjects with INSR had the highest levels of fasting insulin, HOMA-IR, and HbA1c. The medications taken by the study subjects are listed in Table S3 (14).

Table 1.

Baseline characteristics of the enrolled participants

	MASLD (n =1 6)	LD (n = 30)	INSR (n = 29)	P value
Age, years	52 (37-56)	17 (13-28)	17 (12-24)	<.0001*,†
Female (%)	75	67	65	.8
Race (%)				.002
Asian	6	10	14
Black	6	27	28
White	38	57	55
Multiple/other	50	6	3
Ethnicity (%)				.06
Hispanic	50	27	17
Non-Hispanic	50	73	83
BMI, kg/m2	32.9 (29.6-37.7)	23.6 (18.6-26.1)	22.9 (18.9-27.7)	<.0001*,†
Platelet count, K/mcl	247 (207-307)	263 (217-292)	280 (221-340)	.47
Alanine aminotransferase, U/L	48 (29-90)	44 (26-100)	18 (11-24)	<.0001†,‡
Aspartate aminotransferase, U/L	34 (20-67)	29 (22-49)	17 (13-20)	<.0001†,‡
Total bilirubin, mg/dLa	0.4 (0.3-0.8)	0.4 (0.3-0.6)	0.4 (0.3-0.6)	.9
Albumin, g/dLa	4.4 (4.2-4.6)	4.2 (3.9-4.6)	4.1 (3.7-4.3)	.0008†,‡
Fib-4 scoreb	0.96 (0.71-1.63)	0.30 (0.20-0.58)	0.26 (0.19-0.36)	<.0001*,†
Hemoglobin A1C, %	5.7 (5.3-6.3)	7.6 (5.7-9.6)	8.5 (5.6-9.6)	<.0001*,†
Fasting glucose, mg/dL	107 (89-117)	140 (97-234)	93 (79-158)	.025‡
Fasting insulin, mcU/mLc	27.3 (21.5-2.9)	31.9 (18.5-52)	231.0 (93.3-310.0)	<.0001†,‡
HOMA-IRc	6.5 (5.6-10.5)	11.6 (4.7-28.4)	54 (20.6-149.6)	<.0001†,‡
Total cholesterol, mg/dL	183 (155-194)	164 (131-192)	145 (126-176)	.051
High-density lipoprotein cholesterol, mg/dL	43 (36-47)	29 (23-35)	58 (44-74)	<.0001*,†,‡
Low-density lipoprotein cholesterol, mg/dLd	103 (93-133)	81 (60-96)	75 (55-95)	.002*,†
Triglycerides, mg/dL	178 (98-264)	278 (169-457)	69 (47-87)	<.0001†,‡
TSH, mcIU/mLe	2.27 (1.65-3.57)	1.62 (1.03-2.73)	1.56 (0.99-2.46)	.16
Comorbidities (%)
Diabetes mellitus	31	73	83	.0013
Chronic kidney disease	0	3	0	>.99
Hypertension	56	33	7	.00014
Cardiovascular disease	0	17	0	.022

Abbreviations: BMI, body mass index; HOMA-IR, Homeostatic Model Assessment for Insulin Resistance; INSR, pathogenic variants of insulin receptor gene; LD, lipodystrophy; MASLD, metabolic dysfunction-associated steatotic liver disease.

Post hoc pairwise comparison for continuous data:

^*MASLD vs. LD P < .05.

^†MASLD vs. INSR P < .05.

^‡LD vs. INSR P < .05.

^a Data available for 73 participants.

^b Data available for 72 participants.

^c Data available for 74 participants.

^d Data available for 67 participants.

^e Data available for 68 participants.

Within the MASLD group, 10 subjects had liver biopsies available for review, of whom 8 had steatohepatitis; 3 of the 8 had advanced fibrosis (stage F3). In the LD group, 11 subjects had a liver biopsy for review, 9 of whom had steatohepatitis; 4 of the 9 had advanced fibrosis or cirrhosis (stages F3/F4). No subjects with INSR had available liver biopsy. One subject with INSR with had active hepatitis C infection at the time of sample collection. All MASLD subjects but 1 had measurements performed with vibration-controlled transient elastography: the median liver stiffness was 5.7 kPa (4.8-7.0) and median controlled attenuation parameter was 320 dB/m (298-356).

Proteomic Analysis Comparison of Proteome Between Groups

A total of 6412 unique proteins were measured in each serum sample. There were 567 differentially expressed proteins between groups (Fig. 1A, Fig. S1 (14)). The polar plot (Fig. 1B) provides a 2-dimensional view of the 3D volcano plot and illustrates that the INSR group had the highest number of differentially expressed proteins (n = 290), with 196 of them (represented in blue) higher compared to MASLD and LD groups, and 94 (orange) with levels lower than in MASLD and LD. In contrast, MASLD had 172 uniquely differentially expressed proteins (80 increased, 92 decreased), whereas LD had 105 (59 increased and 46 decreased). This suggests a unique proteome that distinguishes the INSR group from the other groups. Among proteins increased in 2 groups, LD and MASLD (n = 94), share the majority, indicating similarities between these conditions. Conversely, INSR and MASLD (n = 46) have the fewest upregulated proteins in common, highlighting the distinct differences between them. A sensitivity analysis was performed after excluding the INSR patient with hepatitis C virus: a minor change was observed in a few proteins; none of these were target proteins included in pathway analysis (Fig. S2 (14)).

Figure 1.

(A) 3-Dimensional cylindrical volcano plot of differentially regulated proteins by subject group with x, y, and r calculated using Z-scores; the polar angle conveys the degree to which a protein is associated with each of the groups. Z-axis denoting statistical significance by ANOVA. Proteins significantly differentially expressed at q ≤ .01 are highlighted in color. (B) 3-Axes polar plot of the data from 1A. M+ proteins higher in MASLD (compared to LD and INSR), L+ proteins upregulated in LD only, I+ proteins upregulated in INSR only, L+M+ proteins upregulated in LD and MASLD compared to INSR, I+M+ proteins upregulated in INSR and MASLD compared to LD, I+L+ proteins upregulated in INSR and LD compared to MASLD. (C) Heatmap and unsupervised clustering of 130 differentially expressed proteins with q ≤ .001, fold change ≥1.5 on pairwise comparisons.

To better understand similarities and differences between groups as reflected in the proteome, we performed an unsupervised clustering analysis using a strict subset of 130 differentially expressed proteins, with q-value ≤.001 from ANOVA analysis and a ≥1.5-fold change from pairwise comparisons. A clear difference was observed between the 3 subject groups, resulting in near-perfect separation (Fig. 1C). All MASLD subjects clustered together. The LD group clustered closer to MASLD, suggesting many similarities between the 2 conditions, whereas the INSR group was distinct. At the protein level, the differentially expressed proteins formed 5 distinct clusters (Fig. 1C). To better understand the underlying biology, we performed functional enrichment analysis.

Cluster 1 included proteins that were increased in both MASLD and LD compared to INSR. Many of these proteins are highly expressed in the liver. Thus, this clustering may be explained by similar hepatic manifestations between MASLD and LD. GO enrichment analysis showed that this cluster includes proteins related to metabolic processes, including those involved in lipid transport and alcohol, fatty acid, and hormone metabolism (Fig. 2A). These proteins included aldo-keto reductase family 1 members C2, C4, and D1, aldolase B, glutathione S transferases α, fructose-bisphosphatase 1, alcohol dehydrogenases 1A, 1C, and 4, growth arrest-specific protein 2, GH receptor (GHR), and prostaglandin reductase 1. Upregulation of apolipoprotein (APO) E, necessary for catabolism of triglycerides, may be explained by high levels of triglycerides in both MASLD and LD. Subjects with LD and MASLD also demonstrated high levels of leukocyte cell-derived chemotaxin 2 and afamin, both highly expressed in liver. Non-liver-expressed proteins in Cluster 1 included Wnt family member 5A and triggering receptor expressed on myeloid cells 2, both involved into extracellular matrix organization.

Figure 2.

GO enrichment analysis for (A) Cluster 1. (B) Cluster 2. (C) Cluster 3. (D) Cluster 4. (E) Cluster 5.

Cluster 2 contained proteins predominantly increased in MASLD; however, some of the proteins were also increased in INSR. Among proteins upregulated in MASLD were S100 calcium-binding protein A9, fibronectin 1, matrix metallopeptidase 9 (MMP9), and microfibril associated protein 4 (MFAP4). Fibronectin 1, MMP9, and MFAP4, all are related to extracellular matrix and may play roles in liver fibrosis. Several adipose tissue-enriched proteins showed higher levels in MASLD and lower levels in LD and INSR, including adipsin/complement factor D, fatty acid-binding protein-4, and TIMP metallopeptidase inhibitor 4. This may be explained by the excess of adipose tissue in MASLD and its deficiency in LD and, to a lesser extent, in INSR. Both MASLD and INSR demonstrated higher levels of the adipose tissue-derived hormone leptin compared to LD (Fig. S1 (14)). GO enrichment analysis grouped proteins in Cluster 2 related to thermogenesis and temperature homeostasis (Fig. 2B).

Cluster 3 included proteins mostly increased in INSR and LD compared to MASLD. GO enrichment analysis revealed enrichment of pathways related to ossification, extracellular matrix, and development of the nervous system (Fig. 2C). Proteins in this cluster include collagen type X and XI, hyaluronan and proteoglycan link protein 1, fibromodulin, matrilin 3, MMP13, tenascin, and osteomodulin. This cluster may be attributable to the younger age of subjects with LD and INSR compared to those with MASLD. Many subjects with LD and INSR were children and still undergoing age-appropriate growth and bone development.

LD and INSR also shared upregulation of kininogen 1 and ficolin 2, both predominantly expressed in the liver. LD alone demonstrated upregulation of APOC3, a component of triglyceride-rich proteins, likely related to high levels of TG in this group.

Proteins predominantly increased in INSR, INSR and LD, or INSR and MASLD comprised Cluster 4. As expected, insulin was highest in INSR, correlating with the severe degree of IR in this condition (Fig. S3 (14)). Proteins increased in both INSR and MASLD included inter-alpha-trypsin inhibitor heavy chain 3 and IGF binding protein 2, both of which are liver-enriched. Adiponectin was higher in both INSR and MASLD compared to LD. Functional enrichment analysis grouped proteins related to steroid and lipid metabolism and reproductive processes (Fig. 2D).

Cluster 5 included proteins upregulated in INSR, or both INSR and LD. Liver-expressed antimicrobial peptide 2 (LEAP2) and procoagulant factor IX (F9) were higher in both LD and INSR compared to MASLD, both of which are exclusively expressed in liver. Compared to MASLD subjects with INSR and LD demonstrated higher levels of glucagon, and gut-derived hormone peptide YY, usually suppressed in fasting. GO enrichment analysis demonstrated that proteins in this cluster relate to protein transport and localization (Fig. 2E).

LD and INSR Share Some Common Liver-expressed Proteins Despite Opposite Hepatic Phenotypes

Although clustering and physiology suggest that LD is pathophysiologically closer to MASLD, several proteins were differentially expressed similarly in LD and INSR compared to MASLD. Of those, LEAP2 and F9 stood out as significantly higher in both LD and INSR compared to MASLD (Fig. 3A) and their serum levels were highly correlated with each other (Fig. 3B). We further assessed the mRNA expression of LEAP2 and F9 in the hepatic transcriptome of a separate cohort of people with MASLD, where they were also highly correlated (Fig. 3C), suggesting that serum levels are reflective of their hepatic source. Therefore, we explored the physiology of LEAP2 and F9 in these populations by testing their associations with clinical parameters and their regulation in vitro.

Figure 3.

(A) Serum levels of liver-expressed antimicrobial peptide 2 (LEAP2) and factor IX (F9) by subject group (RFU, relative fluorescence units). (B) Correlation between LEAP2 and F9 serum levels. (C) Correlation between hepatic LEAP2 and F9 mRNA expression. (D) Effect of insulin or leptin on F9 and LEAP2 expression in HepG2 cells. (E) Effect of glucose concentration on F9 and LEAP2 expression in HepG2 cells.

Correlation of LEAP2 and F9 with clinical parameters

Correlation (Table S4 (14)) and regression analyses (Fig. S4 (14)) were performed to identify associations between LEAP2 and F9 and clinical parameters in each of the groups.

There was no significant association between LEAP2 or F9 and clinical parameters in MASLD. In LD there were significant positive correlations between LEAP2 and TG, fasting glucose, insulin, HbA1C, and HOMA-IR, and between F9 and BMI and TG with the associations between LEAP2 or F9 and TG being linear. In the INSR group both LEAP2 and F9 positively correlated with age and BMI, and negatively correlated with fasting insulin, HbA1c, and HOMA-IR. The relationship between LEAP2 and BMI, fasting insulin, HbA1c and HOMA-IR and F9 and BMI, fasting insulin and HOMA-IR were also linear.

LEAP2 and F9 are expressed in a non-insulin and non-leptin-dependent manner and not affected by glucose

To determine whether the higher levels of LEAP2 and F9 in LD and INSR are driven by hormonal regulation, we treated HepG2 hepatoma cells with various concentrations of leptin and insulin for 30 or 120 minutes, and LEAP2 and F9 expression was measured. There was no difference in expression observed after insulin or leptin treatment (Fig. 3D). We further placed HepG2 cells in low-glucose medium to mimic a fasting state. Direct exposure of HepG2 cells to various concentrations of glucose did not lead to increase in expression of LEAP2 or F9 (Fig. 3E). To mimic the lipid environment associated with SLD, we treated HepG2 cells with a mixture of oleic and palmitic acids for 24 hours, together with varying concentrations of insulin, but no change in LEAP2 or F9 expression was induced by insulin treatment (Fig. S5 (14)).

Discussion

Overnutrition leads to obesity and ectopic accumulation of TG, resulting in hepatic steatosis and IR in common MASLD. In LD, partial or complete absence of subcutaneous adipose tissue is associated with similar, often more severe, hepatic steatosis than in MASLD, along with IR. In INSR, insulin resistance occurs at the receptor level and is not associated with hepatic steatosis and typically not with overweight or obesity. This first comparative proteomic study of different disorders with IR provided an opportunity to evaluate the roles of adipose tissue and IR in hepatic impairment. Because of the predominance of liver-derived proteins in the serum proteome (16), we used serum proteomics to gain insight into mechanisms of liver disease. We demonstrated that, despite differences in adipose tissue mass, MASLD and LD share many similarities in the serum proteome, likely because of their common pathophysiology of fatty acid overload to the liver leading to ectopic fat storage. Despite the commonality between MASLD and LD, our analysis also shows the ability of the serum proteome to distinguish the 2, highlighting known differences, as we have previously described (17).

Some of our findings confirmed existing knowledge about these 3 disorders, providing an internal “positive control.” For example, as expected, we observed higher levels of leptin in MASLD and INSR compared to LD, reflecting differences in adipose tissue mass. Adiponectin has been previously shown to be negatively correlated with insulin resistance regardless of adiposity; surprisingly, people with INSR have been previously shown to have high adiponectin levels through an unclear mechanism (18, 19). In our study, we found higher adiponectin levels in MASLD compared to LD (consistent with different IR) and confirmed that adiponectin was disproportionately high relative to severity of IR in INSR. Both leptin and adiponectin play roles in liver fibrogenesis, acting as profibrogenic and antifibrogenic adipokines, respectively (20); a decrease in both can potentially lead to a disruption of homeostasis, and consequently to more severe liver disease. Furthermore, other adipose-derived proteins such as complement factor D, TIMP metallopeptidase inhibitor 4, and fatty acid-binding protein-4 were increased in MASLD, likely related to increased fat mass. Similarly, subjects with INSR demonstrated the highest levels of insulin among all 3 groups, as evidence of severity of IR in this condition and the lowest levels of GHR compared to LD and MASLD, consistent with prior data from our group (21). INSR is considered a GH-resistant state; low GHR expression is expected to increase bioavailable GH. Thus, the confirmation of expected results for insulin, leptin, and adiponectin confirms the validity of the serum SomaLogic proteomic assay and validates the additional novel insights we describe.

Prior investigation of the proteome in MASLD provided a spectrum of target proteins that can be used as markers of histological features and liver disease severity. Our data demonstrated that many of those proteins were increased in both MASLD and LD, supporting common mechanisms of steatotic liver disease in these 2 distinct populations. Sanyal et al identified proteomic signatures that can detect histological features of steatohepatitis in MASLD (22). Of those proteins, both MASLD and LD shared upregulation of PTRG1 and Wnt family member 5A, included in published models to predict steatosis, ballooning, and fibrosis in liver biopsies. Similarly, triggering receptor expressed on myeloid cells 2 and alcohol dehydrogenase 4, both previously validated as markers of fibrosis in MASLD (23, 24), were increased in both MASLD and LD. Several proteins increased in both MASLD and LD have been described as potential markers of drug-induced liver injury, including aldolase B, glutathione S transferases α, and fructose-bisphosphatase 1 (25, 26). These findings suggest that MASLD and LD share similar mechanisms of liver injury and progression to fibrosis. Supporting our observations, individuals with INSR, who do not have steatotic liver disease, had higher levels of IGF binding protein 2, which has been shown to be negatively correlated with hepatic lipid content (27).

Compared to MASLD, subjects with LD had significant increases in proteins that have been associated with adverse metabolic phenotypes in the general population. LD demonstrated higher levels of leukocyte cell-derived chemotaxin 2 and afamin, both of which are implicated in the pathogenesis of pediatric and adult steatotic liver disease associated with obesity (28, 29), as well as APOC3, associated with atherosclerotic plaque calcification in patients with cardiovascular disease (30). These findings are consistent with more severe metabolic syndrome in LD compared to MASLD, although the underlying mechanisms appear to be the same.

Our attention was drawn to LEAP2 and F9, proteins studied in obesity and MASLD but not previously described in association with LD and INSR. LEAP2 has been recently discovered as a physiological antagonist of ghrelin, reversing activation of GH secretagogue receptor. LEAP2 was shown to be an independent predictor of MASLD; it is associated with IR, obesity, and hepatic lipogenesis in general population and is considered as a novel target for treatment of obesity and metabolic disorders (31, 32). In our study, the LEAP2 levels in MASLD were lower compared to LD and INSR, as was F9. F9 is associated with thrombosis and impaired fibrinolysis (33, 34), and is increased in obesity and MASLD, interpreted as altered coagulability. We found F9 and LEAP2 to be highly correlated with each other in the serum proteome and also in the hepatic transcriptome, suggesting both are regulated similarly in the liver at the transcriptional level. Mani et al (32) suggested that blood LEAP2 levels are regulated by obesity, glucose levels, and feeding status. In contrast, in our data set neither LEAP2 nor F9 correlated with any clinical parameter in MASLD, including BMI and glucose. However, in LD, both LEAP2 and F9 positively correlated with triglycerides, whereas LEAP2 also positively correlated with insulin and HOMA-IR similar to previously published data, suggesting that LEAP2 may have the same effects in those who have both excess and deficient adipose tissue. Interestingly, in INSR, LEAP2 negatively correlated with fasting insulin and HOMA-IR. Using an in vitro model, we showed that the transcriptional regulation of LEAP2 and F9 is not mediated through insulin or leptin pathways, nor is it affected directly by glucose, suggesting that the findings of Mani et al (32) are due to a more complex control. Although ghrelin was similar across the 3 groups in the proteomics assay, we were not able to measure acylated ghrelin, the physiologically relevant form of this protein. Thus, we could not determine if differences in ghrelin activity relate to differences in LEAP2 in our cohorts.

The major strength of our work is the well curated, unique cohort of individuals with insulin resistant conditions that vary by fat mass and by signaling through the insulin receptor, allowing us to dissect out the contributions of these variables to the serum proteome. There are, however, several important limitations to our study. First, our power to detect differences with a low effect size is limited by the small sample size; however, LD and INSR are very rare diseases, limiting the ability to recruit a larger cohort. Second, there are inherent differences between the groups in age and BMI. However, as shown in the example of LEAP2 and F9, the lack of association with BMI or age within the MASLD group suggests that our findings are not merely a reflection of adiposity and age. Third, for practical purposes and because of the scarcity of samples, we relied on the SomaLogic assay without corroboration by direct measurement of target proteins. Fourth, we did not have a control cohort of healthy individuals for comparison, which limits our ability to determine which findings can be solely attributed to the presence of IR. Fifth, we did not have a sufficient number of subjects who underwent liver biopsies to allow us to corroborate past findings related to histological features such as fibrosis and ballooning, nor to imply that our data can be used by itself to derive clinical prediction models. The relatively small MASLD cohort limited our ability to adequately assess associations between the target proteins with markers of hepatic steatosis, steatohepatitis and fibrosis. Finally, some participants were receiving medications that could potentially influence both proteomic profiles and the clinical correlations reported, representing a potential confounding factor in our analyses.

In conclusion, we demonstrated that, despite differences in adipose tissue mass and severity of IR, MASLD and LD have largely convergent serum proteome profiles. That MASLD and LD share common expression of proteins related to liver fibrosis and liver injury suggests that predictive models of liver disease severity based on proteomic studies in MASLD may be potentially applied to the rarer condition of lipodystrophy. We further show how the use of rare disease populations can shed light on the regulation of the secretory proteome in more common disorders such as MASLD. In addition, we demonstrate that the regulatory mechanisms for LEAP2 and F9 secretion may be more complex than previously thought and require further investigation. Our work could pave the way for novel therapeutic targets and a move toward more personalized medicine in the treatment of insulin resistance and hepatic complications.

Abbreviations

3D

3-dimensional

APO

apolipoprotein

BMI

body mass index

DNL

de novo lipogenesis

F9

factor IX

GHR

GH receptor

GLD

generalized lipodystrophy

GO

Gene Ontology

HbA1c

hemoglobin A1c

HOMA-IR

Homeostatic Model Assessment for Insulin Resistance

INSR

insulin receptor pathogenic variants

IR

insulin resistance

LD

lipodystrophy

LEAP2

liver-expressed antimicrobial peptide 2

MASLD

metabolic dysfunction-associated steatotic liver disease

MMP

matrix metallopeptidase

NIH

National Institutes of Health

TG

triglycerides

Funding

This work was supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. M.C.P. received funding from the NIH Office of Dietary Supplements (ODS) Research Scholars Program.

Author Contributions

All authors made a substantial contribution to this work. All authors provided approval for the final submitted version of the manuscript. M.M.: data curation, formal analysis, investigation, methodology, visualization, writing original draft. A.W.: data curation, formal analysis, methodology, writing, review and editing. B.S.A.: data curation. M.C.P.: data curation, methodology. R.J.B. and Y.R.: supervision; data curation; investigation; funding acquisition; conceptualization; resources; methodology; writing, review and editing.

Disclosure

The authors have nothing to disclose.

Data Availability

Aggregate proteomic data, cell culture experimental data, 3D volcano widget, and R code will be deposited in Mendeley Data, and the link will be made publicly accessible upon publication.

Ethics Approval

Research was approved by Institutional Review Board of the National Institutes of Health.

Patient Consent

Written informed consent was provided by all the enrolled participants.

Clinical Trial Information

Clinical trial registration number NCT02520609, NCT00001987.