Sex and Age but Not Body Mass Index (BMI) Predict Serum Retinol Binding Protein 4 (RBP4) and Transthyretin (TTR) Concentrations

ABSTRACT

Increased serum concentrations of retinol (vitamin A) and its binding protein retinol binding protein 4 (RBP4) have been linked to progressive obesity, type 2 diabetes, and reduced liver or kidney function. It has been suggested that body mass index (BMI) and serum RBP4 concentrations correlate. Whether this correlation is due to obesity or associated disorders such as insulin resistance or steatotic liver disease is not known. In circulation, RBP4 forms a complex with transthyretin (TTR). Whether the formation of the RBP4‐TTR complex is altered with progressive obesity remains to be determined. The goal of this study was to establish whether serum RBP4, retinol, and TTR concentrations change with BMI in a population without diabetes or organ impairment. Fasting serum concentrations of RBP4, TTR, and retinol were measured by LC–MS/MS in individuals with a broad range of BMI (21–56 kg/m²), and RBP4‐TTR‐retinol complex formation was modeled. Sex and age explained a significant portion of the inter‐individual variability in RBP4 (p < 0.01) and TTR (p = 0.02) concentrations, while BMI did not. Kinetic modeling suggested that the impact of sex on total RBP4 concentrations is driven by the sex difference in TTR concentrations. A negative correlation was observed between retinol concentrations and BMI but not between RBP4 concentrations and BMI. Collectively, our study shows that sex differences and possibly obesity‐associated comorbidities rather than BMI explain prior findings of correlations of RBP4 and retinol concentrations with BMI. Our results emphasize the importance of measuring TTR, retinol, and RBP4 to assess disease associations.

Serum retinol (ROL), retinol binding protein 4 (RBP4), and transthyretin (TTR) concentrations were measured by LC–MS/MS. Backward stepwise multiple linear regression analysis revealed age and sex as significant correlates for RBP4 and TTR and BMI for ROL serum concentrations. Our developed kinetic binding model predicted that most of total serum RBP4 circulates complexed with TTR. An increase in total TTR concentrations was predicted to increase RBP4:TTR complex concentrations, likely resulting in higher total RBP4 concentrations. Figure created using Biorender.com.

1. Introduction

Progressive obesity and associated comorbidities affect nearly half of the United States population and are a major public health burden globally. Vitamin A (retinol, ROL) is an essential dietary micronutrient required for immunity, vision, reproduction, and cell cycle regulation [1]. The active metabolite of retinol, all‐trans retinoic acid, contributes to the regulation of adipogenesis and lipid metabolism in vitro and in preclinical studies [2], suggesting a key role of retinoid signaling in adiposity and energy homeostasis. Circulating retinol concentrations reflect nutritional vitamin A status [3, 4] and pathological conditions, including liver disease and chronic kidney disease [5, 6]. In mouse models of obesity, serum retinol concentrations were elevated and liver retinol concentrations were depleted, suggesting that hepatic vitamin A stores are mobilized in obesity [7]. However, whether these findings translate to humans remains unclear. Moreover, it is not known if obesity alone, in the absence of comorbidities, alters vitamin A homeostasis in humans, and the data on vitamin A homeostasis in obesity in the presence of comorbidities are inconsistent.

In blood, lipophilic retinol binds tightly to retinol binding protein 4 (RBP4) (K _d = 70–190 nM) [8, 9, 10]. Serum RBP4 concentrations are increased with diabetes and hypothesized to contribute to the pathology of insulin resistance [11, 12, 13, 14]. Detection of RBP4 mRNA in human and mouse adipose tissue led to a hypothesis that RBP4 may be an adipokine synthesized and secreted by adipose tissue, resulting in increased serum RBP4 concentrations with adipose tissue expansion [14, 15]. Indeed, circulating RBP4 concentrations correlated with increased body mass index (BMI) in men with and without obesity and type 2 diabetes [16]. Subsequent studies in men and women with insulin resistance also showed a positive association between RBP4 and BMI and between RBP4 and insulin resistance [17, 18], but no relationship between retinol and BMI [17]. When diabetes was an exclusion criterion, one study observed a positive correlation between RBP4 and BMI in men and women [19], whereas a large cross‐sectional study found an association only in men [20]. In contrast, studies exclusively in women did not show a relationship between RBP4 and BMI [21, 22, 23]. Collectively, these clinical findings suggest that diabetes and/or insulin resistance rather than BMI is the main correlate with increased RBP4 concentrations observed in individuals with obesity, and that the relationship between BMI and RBP4 may be sexually dimorphic. However, the discrepant findings across studies may also be due to the variety of analytical methods used [11, 24]. Prior quantification of RBP4 has been achieved by western blots or enzyme linked immunoassays (ELISAs), where reported concentrations can vary based on method, disease state, assay, and sample processing methods [25]. As such, absolute quantification of RBP4 and TTR with rigorous quantitative methods in the context of increasing BMI is necessary to advance the field.

Most of the body's vitamin A stores reside in the liver, with 10%–20% stored in the adipose tissue [26]. Stored liver retinyl esters are hydrolyzed and secreted from the liver as a retinol‐RBP4 (ROL:RBP4) complex [2, 27]. While monomeric RBP4 is a 21 kDa protein that is filtered by the kidneys [28], in serum, ROL:RBP4 binds tightly to transthyretin (TTR) tetramer (K _d = 215–294 nM) [29, 30, 31] resulting in a complex that is too large (~75 kDa) to be filtered by the kidney [28, 32]. The critical role of TTR in conserving circulating RBP4 suggests that changes in TTR tetramer concentrations or in the formation of the RBP4:TTR complex (one RBP4 and TTR tetramer) will result in changes in vitamin A homeostasis and ROL:RBP4 complex concentrations. Yet, whether TTR concentrations and RBP4:TTR complex formation are altered in progressive obesity is unknown.

The goal of this study was to test whether concentrations of serum retinol, RBP4, and TTR associate with BMI in metabolically healthy men and women and to determine the relative distribution of retinol, RBP4, and TTR tetramer in different complexes in human serum in this population. Serum retinol, RBP4, and TTR concentrations were measured using quantitative LC–MS/MS‐based assays. A kinetic model was developed to establish the binding equilibria of RBP4, TTR, and retinol binding in serum in progressive obesity.

2. Materials and Methods

2.1. Clinical Study Design

Participants were 31 healthy adult men and women (aged 18 to 65 years) undergoing elective abdominal surgery at the University of Washington (UW) Medical Center [33]. The study was conducted in accordance with the Declaration of Helsinki principles and approved by the UW Institutional Review Board (ID: CR00006134, study ID: STUDY00005135). All participants signed written informed consent and none dropped out of the study or withdrew consent. Exclusion criteria included a history of diabetes; significant cardiac, pulmonary, kidney, or liver disease; endocrine or secondary causes of obesity; use of weight loss or glucose‐lowering medication; major systemic illness; infection; malabsorptive gastroenterological disease; anemia; bleeding disorders; or vitamin A supplementation ≥ 10 000 international units within the last 3 months.

Blood was collected in the morning prior to surgery following an overnight fast. For retinol, RBP4, and TTR analysis, blood was collected into serum separator blood collection tubes, light‐protected, allowed to coagulate on ice, and centrifuged at 3000 g for 20 min at 4°C before serum was aliquoted into amber vials and stored at −80°C until analysis. Serum retinol, RBP4, and TTR concentrations were quantified using previously validated LC–MS/MS methods [34, 35] as described below. Surrogate peptides for RBP4 (FSGTYAMAK) and TTR (GSPAINVAVHVFR) were quantified using corresponding stable isotope labeled peptides as internal standards. As each molecule of TTR results in 1 M equivalent of the surrogate peptide, TTR concentrations are reported for single molecules of TTR.

Blood was also collected into sodium citrate tubes, centrifuged at 3000 g for 20 min at 4°C, and plasma was aliquoted. Insulin, glucose, adiponectin, and leptin concentrations were quantified from the plasma samples according to manufacturer recommendations using the following ELISA kits: Ultrasensitive Insulin Kit (ALPCO, Salem, NH), Glucose Assay Kit (Abcam, Cambridge, UK), Adiponectin Quantikine Kit (R&D Systems, Minneapolis, MN), and Leptin Quantikine Kit (R&D Systems, Minneapolis, MN). Plasma creatinine concentrations were quantified using LC–MS/MS as described below. Other relevant clinical chemistries were extracted from patient records when screening for study enrollment.

NAFLD (non‐alcoholic fatty liver disease) fibrosis score was calculated from 1.675 + 0.037 × Age (year) + 0.094 × BMI (kg/m²) + 1.13 × hyperglycemia or diabetes (yes = 1, no = 0) + 0.99 × AST/ALT ratio—0.013 × platelet count (×10⁹/L) − 0.66 × albumin (g/dL) [36]. This scoring system is also appropriate for MAFLD (metabolic dysfunction‐associated fatty liver disease) [37]. BMI was calculated as weight (kg) divided by height (m) squared. Homeostatic model assessment of insulin resistance (HOMA‐IR) was calculated as fasting glucose (mmol/L) × fasting insulin (mU/L)/22.5 [38]. Estimated glomerular filtration rate (eGFR) was calculated as 142 × min (Scr/κ, 1) ^α  × max (Scr/κ, 1)^−1.200 × 0.9938^Age(year) where α is −0.241 if female and −0.302 if male, Scr/κ is the serum creatinine divided by κ (0.7 if female and 0.9 if male), at min (Scr/κ, 1) and max (Scr/κ, 1) the minimum or maximum of Scr/κ or 1 is input into the equation, and the equation is multiplied by 1.012 if the individual is female [39].

DNA was extracted from whole cell containing samples, and the participants were genotyped for RBP4 non‐coding genetic variant rs10882272 (C_111756878_10) and TTR non‐coding genetic variant rs1667255 (C_1278223_10). The genotyping was done via rt‐PCR using a StepOne plus instrument and Taqman reagents as previously described [40].

2.2. Analyte Quantitation

For all LC‐MS/MS analyses standard curves and quality controls (QCs) for all analytes passed the criteria set by the FDA Bioanalytical Guidance [41].

2.2.1. Retinol Analysis

Retinol‐d₆ was purchased from Toronto Research Chemicals (Toronto, Ontario) and all‐trans‐retinol was purchased from Sigma‐Aldrich (St. Louis, MO). Charcoal stripped human DC Mass Spect Gold serum (PN MSG4000, Lot F04005) was purchased from Golden West Diagnostics (Temecula, CA). Optima LC–MS grade acetonitrile, water, acetic acid, and formic acid were purchased from Fisher Scientific (Pittsburg, PA). Retinol purity was confirmed by LC‐ultraviolet detection.

A previously established method was used to quantify serum retinol concentrations [42]. Standard curves ranged from 0.25–3 μM and QCs were at 0.6, 1.25, and 2 μM. In brief, 50 μL of serum, calibration curve samples, and QCs were aliquoted into PCR plates and precipitated by the addition of 100 μL of ice‐cold acetonitrile containing 750 nM retinol‐d₆. The plate was centrifuged at 4°C at 3000 g for 40 min, followed by transfer to a new plate and centrifugation at 3000 g for 30 min before the supernatant was transferred to a new 96‐well plate for LC–MS/MS analysis on a SCIEX 6500 QTRAP mass spectrometer (Sciex, Framingham, MA) coupled to an Agilent 1290 Infinity LC (Agilent, Santa Clara, CA) with an Ascentis Express RP Amide column, 2.7 μm, 2.1 × 150 mm, and guard column, 2.7 μm, 2.1 × 5 mm (Sigma‐Aldrich, St. Louis, MO), as previously described [34, 42]. The following transitions were monitored for quantification: m/z 269 > 93 retinol and m/z 275 > 96 retinol‐d₆. Peaks were integrated and quantified using MultiQuant 3.0.

2.2.2. RBP4 and TTR Analysis

Sodium deoxycholate, iodoacetamide (IAA), ammonium bicarbonate, bovine serum albumin (Part Number A6003), trifluoroacetic acid, and yeast enolase were purchased from Sigma‐Aldrich (St. Louis, MO). Dithiothreitol (DTT) and Ringer's solution were purchased from Thermo Fisher Scientific (Waltham, MA). Trypsin Platinum (PN VA9000) was purchased from Promega (Madison, WI). Optima LC–MS grade acetonitrile, water, acetic acid, and formic acid were purchased from Fisher Scientific (Pittsburg, PA). Phosphate buffered saline (PBS) was purchased from Corning (Corning, NY). TTR purified from human serum and RBP4 purified from human urine were purchased from Bio‐Rad Laboratories (Hercules, CA). Lyophilized RBP4 and TTR were reconstituted according to supplier guidelines to 1 mg/mL using 1× Phosphate Buffered Saline (PBS). Protein concentration was confirmed by bicinchoninic acid (BCA) assay (Thermo Scientific, Waltham, MA) at three concentrations of protein (0.3, 0.5, and 0.7 mg/mL) in triplicate.

Serum RBP4 and TTR concentrations were determined as previously described [35]. In brief, 40 μL of samples diluted 100‐fold in 100 mM ammonium bicarbonate were aliquoted along with standard curves and QCs in 96‐well PCR plates. Standard curves ranged from 0.5–6 μM for RBP4 and 5.77–69.3 μM for TTR, and QC levels were 0.6, 0.9, 1.8, and 3.6 μM for RBP4 and 6.93, 10.4, 20.8, and 41.6 μM for TTR. Samples were reduced with 8 μL of 100 mM DTT for 20 min at room temperature; 10 μL of 10% sodium deoxycholate was added, and the plate was heated to 95°C for 10 min on an Eppendorf ThermoMixer to denature proteins. After the plate was cooled to room temperature, 16 μL of 200 mM IAA was added under reduced light for alkylation of cysteines, and the samples were incubated for 20 min at room temperature before initiation of digestion with trypsin at 37°C for 5 h. Digestions were quenched with 40 μL of acetonitrile with 8% TFA containing 50 nM FSGTWYAMAK[¹³C₆ ¹⁵N₂] and YWGVASF[¹³C₉ ¹⁵N]LQK and 250 nM GSPAINVAVHVFR[¹³C₆ ¹⁵N₄] internal standard peptides. The plate was centrifuged at 3000 g for 30 min at 4°C, the orientation of the plate flipped 180°, and centrifuged for another 30 min, and supernatants transferred to a new 96‐well plate for LC–MS/MS analysis.

Peptides were separated using an Aeris Peptide column (50 × 2.1 mm, 1.7 μm) with a SecurityGuard Ultra C18‐peptide cartridge (Phenomenex, Torrance, CA) on an Agilent 1290 LC (Agilent, Santa Clara, CA) coupled to a SCIEX 5500 QTRAP mass spectrometer (Sciex, Framingham, MA), as previously described [35]. For RBP4 peptide, FSGTWYAMAK, the transitions were as follows: precursor(⁺²) 581.3 > 1014.5 (y9), 927.4 (y8), 769.4 (y6), and for the labeled peptide: precursor(⁺²) 585.3 > 1022.5 (y9), 935.5 (y8), 777.4 (y6). For TTR peptide, GSPAINVAVHVFR, the transitions were precursor(⁺²) 683.8 > 941.5 (y8), 728.4 (y6), 611.8 (y11)⁺² and for the labeled peptide: precursors(⁺²) 688.9 > 951.5 (y8), 738.4 (y6), 616.9 (y11)⁺². The sum of three transitions was used for quantitation. Peaks were integrated and quantified in Skyline version 22 [43].

2.2.3. Creatinine Analysis

Creatinine was purchased from Sigma‐Aldrich (St. Louis, MO) and creatinine‐d₃ was purchased from Cayman Chemical (Ann Arbor, MI). Optima LC–MS grade acetonitrile, water, acetic acid, and formic acid were purchased from Fisher Scientific (Pittsburg, PA). Calibration curves and QCs were prepared by spiking 10‐fold water stocks of creatinine into representative serum matrix with 40 mg/mL human serum albumin in Ringer's solution. The calibration curve was prepared at a concentration range of 40–150 μM. Ten microliters of serum samples, calibration curves, and QCs were mixed with 75 μL of ice‐cold acetonitrile containing creatinine‐d₃ to precipitate serum proteins in a 96‐well plate. The plate was centrifuged at 3000 g at 4°C for 45 min. The supernatants were transferred to clean wells, and centrifugation was repeated at 3000 g at 4°C for 30 min. Supernatant was sequentially diluted 20‐fold and then twofold with water.

Serum creatinine was quantified using an AB Sciex 5500 QTRAP mass spectrometer (Sciex, Framingham, MA, USA) coupled to an Agilent 1290 Infinity II liquid chromatograph (Santa Clara, CA, USA) operated in electrospray ionization mode. An Agilent Zorbax XDB‐C18 2.1 × 50 mm, 5 μm column was used for analyte separation. The mobile phase consisted of (A) water with 0.1% formic acid and (B) methanol, with a 5 μL injection volume and the column at 45°C. A gradient elution was carried out at 0.4 mL/min starting from 10% B for 1.5 min, increasing to 90% B for 1 min, held for 2.5 min, before returning to initial conditions and held for 2.5 min. The following transitions in positive ion mode were monitored: m/z 114 > 44 for creatinine and m/z 117 > 47 for creatinine‐d₃. Peaks were integrated and quantified using MultiQuant 3.0. Standard curves were weighted 1/x.

2.3. Kinetic Modeling

A kinetic model was developed in MATLAB to simulate the concentrations of unbound TTR tetramer, RBP4, and retinol (ROL) along with retinol bound RBP4 (ROL:RBP4), RBP4:TTR, and the ternary complex, ROL:RBP4:TTR, in each participant. TTR is assumed to be entirely tetrameric in human serum and in each of these complexes based on prior studies regarding the tetramer dissociation kinetics and analysis of TTR tetramers in plasma [44, 45]. The model was set up akin to an enzyme inhibition reaction coordinate with parallel reactions (Figure 1). Alpha (α) was solved from experimentally determined dissociation constants (K _d). The K _d associated with retinol binding to RBP4 (ROL:RBP4), K _d1, was set as 146 nM, which is the mean of dissociation constants from three previous experimental findings: 178 nM [9], 190 nM [8] and 70 nM [10]. The K _d associated with RBP4:TTR complex formation, K _d2, is 1.2 μM [29]. From literature, mean αK _d2, associated with TTR binding to ROL:RBP4, was set as 215 nM, calculated as a mean from published values of 150 nM [30], 294 nM [31], and 200 nM [29] and αK _d1 is 35 nM [10]. Based on these values, α was set as 0.235 in the model. As a result, αK _d1 solved to 34 nM and αK _d2 to 282 nM. The complex of two ROL:RBP4 binding to one TTR tetramer, (ROL:RBP4)₂:TTR, was not included in the model, as experimental (ex vivo) formation of (ROL:RBP4)₂:TTR requires RBP4 in twofold molar excess to TTR—a supraphysiological ratio [10, 31, 46]. To determine the potential impact of (ROL:RBP4)₂:TTR on model results, a model was built with αK _d2 as the dissociation constant for ROL:RBP4, ROL:RBP4:TTR, and (ROL:RBP4)₂:TTR. Incorporation of the (ROL:RBP4)₂:TTR complex resulted in a minimal decrease in free ROL and ROL:RBP4. This decrease is unlikely to be detectable under physiological conditions, but the simulated values here should be considered as the estimated upper limit of circulating free ROL and ROL:RBP4.

FIGURE 1.

Schematic of the kinetic model used to simulate concentrations of free RBP4, unbound retinol (ROL), retinol‐RBP4 complex (ROL:RBP4), RBP4 complexed to TTR tetramer (RBP4:TTR), and the ternary complex ROL:RBP4:TTR. The kinetic constants estimated from literature values are shown. Alpha was solved from the mean of experimentally determined published dissociation constants (K _d) as described under kinetic modeling in the methods. TTR refers to the tetrameric form of TTR.

For data analysis, initial conditions in the model were set as the total measured retinol, RBP4, and TTR concentrations in individual study participants, and the model was run to steady state. The concentrations of each binding state and the total at the end of the simulation were collated for all participants. The total retinol, RBP4, and TTR at the end of the simulation were compared to input concentrations to confirm mass balance and equivalence.

A sensitivity analysis was also performed using the model and the same dissociation constants, holding total RBP4 concentration constant at 2 μM and increasing retinol and TTR concentrations. The concentration range was determined from the observed total concentrations in the study participants. The MATLAB (R2024A, MathWorks, Natick, MA; RRID:SCR_001622) code and system of ordinary differential equations can be found at https://github.com/Isoherranen‐Lab/RBP4_TTR_Retinol_Binding_Model.

2.4. Statistical Analysis

Regressions and statistical analyses were performed in R statistical computing software (version 4.2) [47]. The following packages were used: tidyverse, car, forcats, ggpubr, ggally, rigr, broom, officer, kableExtra, grid, flextable, and knitr. Normality was tested by Shapiro–Wilk test, and variables with p < 0.05 were log‐transformed.

A multiple linear regression model including BMI, HOMA‐IR, sex, and age as independent variables was used to test whether these variables explain the observed variability in retinol, RBP4, and TTR concentrations in the study population. Regression was considered significant if p < 0.05. Variance inflation factors (VIF) for BMI, HOMA‐IR, sex, and age were 1.59, 1.29, 1.23, and 1.26, respectively, indicating acceptable inclusion of the parameters in the multiple linear regression model [48]. Results of the correlation analysis between the independent variables are shown in Figure S1. Stepwise backward elimination prioritizing a lower Akaike information criterion (AIC) was used to determine which independent variables and interaction terms were statistically significant. When a lower AIC was observed in the model with additional parameters, analysis of variance (ANOVA) was performed to compare the calculated F‐statistic to the critical value, using p < 0.05 as the threshold to determine whether the inclusion of additional parameters statistically improved model fit. Genotype effects on circulating retinol, RBP4, and TTR in the entire study participant population were tested by ANOVA. R markdown statistical analysis and outputs can be found at https://github.com/Isoherranen‐Lab/RBP4_TTR_Retinol_Binding_Model.

Correlations between RBP4, TTR, and retinol concentrations were performed as Spearman correlations with p < 0.01 as the cut‐off for significance, and graphs were prepared using GraphPad Prism 10.2 (GraphPad Software, LaJolla, CA; RRID:SCR_002798).

3. Results

3.1. Study Participants

Individuals in the study were overall healthy, without a history of kidney or liver disease. However, one participant unexpectedly had liver fibrosis on gross morphology observed by the surgical team during surgery. The participant was retained in the analysis, and sensitivity analyses were performed when this participant appeared to be a high‐leverage data point, to determine their influence on regression results. The geometric mean age of participants was 42 (25–65) years, and the geometric mean BMI was 36.9 (20.7–55.7) kg/m² (Table 1). The median age and range were similar between participants segmented by BMI and HOMA‐IR. Elective surgical procedures included cholecystectomy (n = 4), hernia repair (n = 7), and metabolic surgery (n = 20). Use of glucose‐lowering medication was an exclusion criterion. Participant demographics and clinical characteristics are listed in Table 1. The matrix of clinical characteristics is shown in Figure S1 and association with sex in Table S1.

TABLE 1.

Demographics and clinical lab characteristics of study participants. The data are expressed as geometric means with range except for sex, race, and ethnicity (n = 31).

Parameter	Geometric mean (range)
Age (years)	42 (25–65)
Weight (kg)	103 (57–164)
Height (cm)	168 (157–184)
Sex, (female, n (%))	20 (65%)
BMI (kg/m2)	36.9 (20.7–55.7)
eGFR (mL/min/1.73 m2)	108 (79–131)
Serum ALT (U/L)	21 (8–73) a
Serum AST (U/L)	19 (12–36) a
NAFLD fibrosis score	−1.8 (−4.0–0.7) a , b
Fasting glucose (mM)	4.6 (2.2–5.9)
Fasting insulin (mU/L)	10 (0.2–63)
HOMA‐IR	2.1 (0.02–16)
Leptin (ng/mL)	24 (3.4–76)
Adiponectin (ng/mL)	6.1 (1.7–19)
Race (n)
White	25
Black	5
Asian	1
Ethnicity (n)
Hispanic or Latino	6
Not Hispanic or Latino	25

Abbreviations: ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; eGFR, estimated glomerular filtration rate; HOMA‐IR, homeostatic model assessment of insulin resistance; NAFLD, non‐alcoholic fatty liver disease.

^a Dataset incomplete for these parameters (data available for n = 28).

^b NAFLD fibrosis score shown as arithmetic mean due to negative values.

3.2. Quantification of RBP4, TTR, and Retinol and Association With BMI, HOMA‐IR, Sex, and Age

The absolute total concentrations of RBP4 and TTR in serum were quantified using LC–MS/MS. The molar concentrations of RBP4 and TTR (Table 2) were comparable to previously reported concentrations measured via immunoassays in different populations [16, 20, 21, 49]. The concentrations of TTR molecules exceeded RBP4 concentrations in all participants by about 10‐fold (Table 2). Assuming TTR is entirely in a tetrameric form, this corresponds to a 2.5‐fold (1.9‐ to 3.5‐fold) molar excess of tetrameric TTR relative to RBP4. TTR concentrations correlated significantly with RBP4 concentrations (Figure S2). Since complexation with the TTR tetramer prolongs the half‐life of RBP4 in circulation by preventing renal filtration [28, 32], the correlation of RBP4 and TTR is expected, yet stronger here than previously observed [50].

TABLE 2.

Measured total serum concentrations of molar units of RBP4, TTR, and retinol. Study participants are grouped by body mass index (BMI) and Homeostatic Model Assessment of Insulin Resistance (HOMA‐IR). Data, including BMI, are shown as geometric means with range for all participants.

Variable	Without obesity (n = 8; 2 F, 6 M)	With obesity HOMA‐IR < 2.5 (n = 10; 9 F, 1 M)	With obesity HOMA‐IR > 2.5 (n = 13; 9 F, 4 M)	All (n = 31, 20 F, 11 M)
BMI (kg/m2)	24.1 (20.7, 27.5)	40.1 (33.6, 46.6)	44.8 (39.1, 55.7)	36.9 (20.7, 55.7)
RBP4 (μM)	2.2 (1.2, 3.2)	2.2 (1.2, 3.2)	1.9 (1.0, 3.2)	2.1 (1.0, 3.2)
TTR (μM)	22 (13, 30)	21 (17, 27)	19 (10, 30)	20 (10, 30)
Retinol (μM)	1.5 (1.0, 1.9)	1.5 (1.0, 2.0)	1.4 (0.6, 2.9)	1.4 (0.6, 2.9)
TTR to RBP4 a	9.9 (8.7, 11)	9.5 (7.7, 14)	10 (8.2, 13)	9.9 (7.7, 14)
Retinol to RBP4	0.7 (0.6, 0.9)	0.7 (0.6, 0.8)	0.7 (0.6, 0.9)	0.7 (0.6, 0.9)

Note: Without obesity: BMI ≤ 28 kg/m², with obesity; BMI > 30 kg/m².

Abbreviations: F, female; M, male.

^a Molar ratio of TTR to RBP4 concentrations. TTR is expected to circulate entirely as a tetramer. Considering the tetramer stoichiometry, the molar ratio of TTR tetramer to RBP4 monomer is overall 2.5 (1.9, 3.5). For each subgroup, the TTR tetramer to RBP4 monomer ratios are 2.5 (2.2, 2.8) without obesity, 2.4 (1.9, 3.5) with obesity, HOMA‐IR < 2.5, and 2.5 (2.1, 3.3) with obesity, HOMA‐IR > 2.5.

Serum retinol concentrations were within vitamin A sufficiency for all study participants except for two (Table 2). Serum retinol was insufficient (0.72 μM) in one participant (BMI = 52.7 kg/m²), and one participant (BMI = 55.7 kg/m²) appeared vitamin A deficient (retinol < 0.7 μM). The participant with apparent vitamin A deficiency had liver fibrosis on gross morphology. The total retinol concentrations correlated with total RBP4 (Spearman r = 0.91, p < 0.0001) (Figure S2). The retinol concentrations were lower than RBP4 concentrations in all participants (Table 2). This finding confirms using modern methods and analytical techniques the prior data [51, 52] that RBP4 is in molar excess in human serum. The high degree of correlation of RBP4 and retinol is expected, as RBP4 is critical for the mobilization of retinol from the liver [53].

RBP4 genetic variant rs10882272 has been associated with decreased plasma RBP4 and retinol [54, 55], while TTR variant rs1667255 has been linked to lower serum retinol [54]. Study participants were genotyped for these polymorphisms, but no significant differences in retinol, RBP4, or TTR serum concentrations were observed across genotypes (Tables S2 and S3), possibly due to the small number of participants in the different genotype groups.

BMI, HOMA‐IR, age, and sex were assessed as potential variables contributing to inter‐individual variability in RBP4, TTR, and retinol serum concentrations. Starting with all variables in the model, backward stepwise multiple linear regression analysis identified age and sex as significant correlates and interaction terms for total RBP4 (p = 0.009, adjusted R ² = 0.38) and TTR (p = 0.022, adjusted R ² = 0.36) (Figure 2A,B). Forward addition of BMI or HOMA‐IR did not improve the model fit for either RBP4 or TTR (Tables S4 and S5). BMI and HOMA‐IR were not significant predictors of variability in total RBP4 or TTR concentrations. As age and eGFR are correlated, a post hoc linear regression analysis was performed with total serum RBP4 concentration and eGFR. The association of RBP4 with eGFR was not significant (p = 0.12) alone or with eGFR and sex interaction terms (p = 0.16).

FIGURE 2.

Final regression models after multiple linear regression analyses of total RBP4, TTR and retinol. Total RBP4 (A) and TTR (B) concentrations versus age were stratified by sex, as sex and age were significant interaction terms. BMI was found to negatively correlate with retinol (C). One study participant was found to have liver fibrosis on gross morphology and is shown as a red triangle. Each symbol represents a study participant, the shaded areas represent the 95% confidence interval of the linear regression and the p‐value for each model is shown above the plot.

For total serum retinol, BMI was a significant negative correlate (p = 0.016, R ² = 0.18, β‐coefficient = −0.013 on linear scale) (Figure 2C). Adding parameters of sex and age did not significantly improve the model fit (F = 2.21, p = 0.15) for retinol (Table S6). Notably, the participant with gross morphology of liver fibrosis and overt retinol deficiency was also the participant with the highest BMI. Omission of that participant from the regression model increased the p‐value of correlation between BMI and retinol to p = 0.09, resulting in a lack of significant correlation between retinol and any participant variables.

3.3. Modeling of free RBP4, ROL:RBP4, RBP4:TTR, and ROL:RBP4:TTR Concentrations

To assess the distribution and binding equilibria of retinol, RBP4, and TTR in different complexes in serum, a model was developed for the overall binding kinetics using experimentally determined binding constants (Figure 1). The concentrations of free ROL, TTR, RBP4, and ROL:RBP4, RBP4:TTR, ROL:RBP4:TTR complexes were simulated for each participant based on their measured total retinol, RBP4, and TTR concentrations and reported binding kinetics (Table 3). In all of the simulations, TTR was assumed to be tetrameric, and the total measured TTR concentration was divided by four to reflect the tetramer concentrations.

TABLE 3.

Simulated concentrations of unbound ROL, RBP4, and TTR tetramer and ROL:RBP4, RBP4:TTR, and ROL:RBP4:TTR complex in the study participants. The simulations were conducted for each participant based on their measured total retinol, RBP4, and TTR concentrations, assuming TTR in a complex is tetrameric and binding is at steady state. Data are shown as geometric means with range for all participants.

Simulated species	All (n = 31)
ROL (μM)	0.03 (0.01, 0.08)
RBP4 (μM)	0.2 (0.1, 0.3)
TTR (μM)	3.3 (1.8, 4.8)
ROL:RBP4 (μM)	0.1 (0.1, 0.2)
RBP4:TTR (μM)	0.5 (0.1, 1.1)
ROL:RBP4:TTR (μM)	1.3 (0.5, 2.6)

Based on the simulations and the published kinetic constants, in the study participants, 2.1% of the total retinol was unbound in circulation, with about 90% bound in the ROL:RBP4:TTR complex and 7.6% as ROL:RBP4 (Figure 3). Based on this analysis, circulating RBP4 is not a single entity. On average, the majority (63%) of the total RBP4 in circulation was predicted to be in the ROL:RBP4:TTR complex. Overall, 68% of RBP4 was bound by retinol either with or without TTR. However, RBP4 is also present in serum in its unliganded form. The model predicted that 8.4% of the serum RBP4 is free, unbound RBP4, and 23% of RBP4 circulates as RBP4:TTR (Table 3, Figure 3).

FIGURE 3.

Percentage of total RBP4 (A, B) or total retinol (C, D) in simulated binding states in study participants segmented by BMI ≤ 28 kg/m² (A, C) and BMI ≥ 30 kg/m² (B, D). Arithmetic means and ranges from individually calculated percentages for each participant are presented for each binding species.

To determine how alterations in total TTR and retinol concentrations affect the distribution of RBP4 between different binding states, sensitivity analyses were performed keeping total RBP4 concentration constant and varying TTR tetramer and retinol concentrations. The ROL:RBP4:TTR complex concentrations were highly sensitive to total retinol concentrations (Figure 4A). This analysis translated to the simulated concentrations of ROL:RBP4:TTR complex in the study participants. An increase in simulated ROL:RBP4:TTR complex concentrations was observed with increased total retinol concentrations across participants (Table S7). Although total TTR tetramer concentrations are in molar excess of RBP4, increasing TTR concentrations shifted the distribution of total RBP4 toward RBP4:TTR compared to free RBP4 (Figure 4B). This is consistent with simulated concentrations of RBP4:TTR and free RBP4 in study participants as well. In participants with higher measured total TTR concentrations, higher concentrations of the RBP4:TTR complex were predicted, along with an overall decrease in percentage of RBP4 that was free (Table S8). This is important as RBP4 is subject to glomerular filtration while RBP4:TTR is not and RBP4 is proposed to have retinoid‐independent pharmacological activity [13]. When RBP4 concentration is constant, predicted free RBP4 concentrations are inversely associated with total TTR and retinol, not just retinol. This serves to emphasize the importance of integrated quantification of TTR and retinol when considering circulating RBP4 concentrations, the presence of different RBP4 complexes, and retinoid biological effects.

FIGURE 4.

Sensitivity analysis of the parameters in developed kinetic model. Simulations were run to determine the sensitivity of predicted concentrations of (A) RBP4, RBP4:TTR, ROL:RBP4 and ROL:RBP4:TTR, to increasing concentrations of TTR tetramer and retinol. RBP4 concentration was held constant at 2 μM. Panel (B) shows a magnification of the sensitivity analysis.

4. Discussion

Our study was designed to test if circulating RBP4, TTR, or retinol concentrations are correlated with BMI in the absence of major obesity‐associated comorbidities. We also developed a kinetic binding model based on experimentally determined dissociation constants to assess the binding kinetics of RBP4, TTR tetramer, and retinol in serum with regard to the formation of the different complexes of RBP4, TTR, and retinol. Our data show that sex and age but not BMI are significant correlates for serum RBP4 and TTR. A negative correlation was observed between BMI and total retinol concentrations. Further, the model‐predicted binding states of RBP4, TTR, and retinol provide an overall paradigm of binding equilibria with relative proportions of each binding species in circulation (Figure 5).

FIGURE 5.

Overview of retinol, RBP4, and TTR disposition with the contribution of the adipose tissue, liver and kidney highlighted. Unbound retinol (ROL) can be secreted from the adipose and bound by circulating free RBP4 to form ROL:RBP4. The liver, which stores retinoids and synthesizes RBP4 and TTR, can secrete ROL:RBP4, along with TTR which can form ROL:RBP4:TTR complex in circulation. After delivery of retinol to target tissues, RBP4:TTR also circulates. In the kidney only RBP4, ROL:RBP4, and ROL will be filtered by the glomerulus due to their low molecular weight. Figure created in https://BioRender.com.

Our data show that in the absence of diabetes and liver or kidney disease, serum RBP4 concentrations do not correlate with increasing BMI. This is consistent with previous findings. Reviews on the relationship between BMI and RBP4 have noted a lack of consistent associations [11, 12, 56] and a lack of evidence that increasing BMI causes increased RBP4. Clinical studies have observed that higher serum RBP4 concentrations correlate with insulin resistance [13, 14, 15, 18], a relationship supported by molecular mechanisms demonstrating that increased RBP4 impairs insulin sensitivity [13, 14, 57]. Altered kidney function also correlates with altered circulating RBP4 concentrations. Renal impairment, even in the absence of diabetes, has been independently associated with increased serum RBP4 and retinol [5]. Thus, the lack of exclusion criteria for liver or kidney disease in three studies reporting a positive correlation between RBP4 and BMI in non‐diabetic populations limits interpretation [16, 19, 20]. In our cohort, age and eGFR were strongly and inversely correlated. Yet, RBP4 concentrations did not correlate with eGFR, indicating a direct association between RBP4 concentrations and age.

In the current study, sex and age were significant correlates for both serum RBP4 and TTR, consistent with previous findings of higher RBP4 and TTR concentrations in men [12, 49]. The underlying mechanisms of these sex differences remain unclear, but our modeling data suggest that sexual dimorphism in TTR concentrations may play a role in modulating RBP4 concentrations. RBP4 complexation with TTR tetramer to form RBP4:TTR and ROL:RBP4:TTR prevents glomerular filtration of RBP4 [28, 32], while free RBP4 and ROL:RBP4 are subject to filtration. Decreased filtration clearance prolongs the half‐life of RBP4 and results in higher overall circulating RBP4 concentrations [32]. The model suggests that higher TTR concentrations in men result in increased RBP4:TTR complex concentrations (Figure 4A) potentially resulting in lower clearance of RBP4 in men than in women and higher total RBP4 concentrations.

Our binding model supports the prevailing hypothesis that circulating retinol exists predominantly in the ternary ROL:RBP4:TTR complex, formation of which may be limited by total retinol concentrations. Model simulations indicate that one‐third of RBP4 is not bound to retinol and ~85% of total RBP4 is complexed with TTR tetramer (Figure 3). Although some studies have measured both serum RBP4 and retinol and used ratios as a proxy of retinol free RBP4 [17, 58], many have quantified RBP4 in isolation [11, 56]. This is a critical gap in knowledge, as RBP4 may have a retinoid‐independent role in insulin sensitivity [13]. Only the ROL:RBP4 complex binding to STRA6 has been shown to activate the Janus kinase 2 (JAK2) and the signal transducer and activator of transcription 5 (STAT5) pathway, which may contribute to insulin resistance [57]. By quantifying absolute RBP4 and TTR concentrations using LC–MS/MS, we were able to apply our kinetic model to investigate the binding kinetics and complex formation between retinol, RBP4, and TTR in serum. The developed model and measurement of absolute concentrations provide a framework for investigators to explore the potential roles of RBP4 in different binding states in vivo.

Our finding that RBP4 and TTR regression lines for male and female participants intersect at about age 50 to 55 is consistent with previous reports of higher RBP4 concentrations in men and postmenopausal women compared to premenopausal women [59, 60]. This suggests hormonal changes during menopausal transition may impact RBP4 and TTR concentrations. While the sex difference in TTR concentrations is well established [49], an interaction between sex and age has not been previously described. Although sex steroid regulation of RBP4 has been proposed, in vitro and observational clinical data are equivocal [12] and the underlying mechanisms require further study. Our model and measurements suggest that the sex difference in RBP4 may be driven by TTR concentrations rather than direct regulation of RBP4.

Independent of age or sex, measured total serum retinol inversely correlated with BMI in our study population. This is in contrast to two prior studies that found no association between retinol and BMI alone [17, 61], but in agreement with a study reporting an inverse correlation in individuals with obesity and fatty liver disease [62]. Perturbations in liver health can result in altered hepatic vitamin A homeostasis [63, 64] and decreased circulating retinol [6]. We cannot exclude the possibility that liver function was an important determinant of retinol concentrations in our study as well. Notably, excluding the participant with the highest BMI and liver fibrosis rendered the correlation between BMI and retinol non‐significant, suggesting liver disease may have played a role.

Circulating RBP4, TTR, and retinol concentrations depend on both clearance and secretion processes. Here, based on our kinetic modeling and measured total concentrations, we can explore the role of adipose tissue in contributing to circulating RBP4. It has been postulated that RBP4 is also secreted by adipose tissue [12, 14]. Yet, data in mice suggest that serum RBP4 is entirely derived from the liver. When RBP4 is knocked out in the liver, RBP4 is not detectable in circulation [53, 65]. Further, under fasting conditions, mouse data suggest unbound retinol is secreted from adipose tissue and captured by plasma free RBP4 [66]. Our model and absolute quantification support this concept as we found that there is an excess of RBP4 in circulation in the fasting condition. Though not the main objective of our study, our findings appear consistent with the mouse data and that, in human serum, RBP4:TTR and RBP4 concentrations are sufficient to capture retinol released from the adipose tissue (Figure 5).

This work does have some important limitations. This study was cross‐sectional with a limited number of participants and employed a static measurement of RBP4, TTR, and retinol in the fasting state. Thus, it is difficult to comment on the longitudinal progression of obesity and associated comorbidities and the mechanistic role of retinoids. Additionally, a strong correlation between sex and BMI was observed in our study population. Larger clinical studies that enroll men and women across different age groups are needed to further evaluate the sex differences in RBP4 and TTR concentrations and the potential interactions of sex and BMI in the context of retinol, RBP4, and TTR concentrations. The potential impact of menopausal transition on retinol, RBP4, and TTR concentrations in women should also be further evaluated, as our study did not collect data on menopausal status. Quantification of RBP4 and its binding partners, retinol and TTR, in future prospective and longitudinal studies will be important to define whether retinol and/or RBP4 signaling alters the development of diabetes and cardiometabolic diseases. Further, as alluded to earlier, vitamin A homeostasis is a dynamic process and may be different after feeding. Thus, future studies should quantify RBP4, TTR, and retinol concentrations in a fed state.

In conclusion, our data suggest that in the absence of diabetes or liver disease, sex and age, but not BMI, can explain some of the interindividual variability in RBP4 and TTR concentrations. With the kinetic modeling results, it can be postulated that the sex difference in TTR is driving the observed dimorphism in RBP4 and that kidney function may also influence circulating RBP4 and TTR concentrations. The observed negative correlation between BMI and retinol may be influenced by hepatic health.

Author Contributions

K.B.R. and N.I. conceived and designed the research. A.S.Y., L.C.C., J.Z., J.L., A.K.A., L.W., J.Y.C., E.W., Z.P., S.K., D.K., K.B.R., and N.I. performed the research. A.S.Y., K.B.R., and N.I. analyzed data. A.S.Y., L.C.C., K.B.R., and N.I. wrote or contributed to the writing of the manuscript. All authors reviewed the final manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Appendix S1.

Yadav A. S., Czuba L. C., Zhu J., et al., “Sex and Age but Not Body Mass Index (BMI) Predict Serum Retinol Binding Protein 4 (RBP4) and Transthyretin (TTR) Concentrations,” The FASEB Journal 39, no. 14 (2025): e70842, 10.1096/fj.202501772R.

Data Availability Statement

The authors declare that all the data supporting the findings of this study are contained within the paper. The raw data are available upon request from the corresponding author.