Shortcomings in methodology complicate measurements of serum retinol binding protein (RBP4) in insulin-resistant human subjects
- First Online:
- Cite this article as:
- Graham, T.E., Wason, C.J., Blüher, M. et al. Diabetologia (2007) 50: 814. doi:10.1007/s00125-006-0557-0
- 919 Downloads
Levels of retinol binding protein (RBP4) are increased in the serum of insulin-resistant human subjects even before overt diabetes develops. RBP4 levels correlate with insulin resistance, BMI, WHR, dyslipidaemia and hypertension. Improvement of insulin sensitivity with exercise training is associated with reduction in serum RBP4 levels. Therefore serum RBP4 may be useful for early diagnosis of insulin resistance and for monitoring improvements in insulin sensitivity. We sought to determine the performance of assays for this application.
Subjects and methods
We compared quantitative western blotting and three commercially available multiwell immunoassays in parallel measurements of RBP4 concentrations in serum from insulin-sensitive subjects and from insulin-resistant subjects with impaired glucose tolerance or type 2 diabetes.
The assays yielded different absolute values and magnitudes of elevation of serum RBP4. Western blotting and a sandwich ELISA reported RBP4 concentrations that highly inversely correlated with insulin sensitivity measured by euglycaemic–hyperinsulinaemic clamp. However, western blotting yielded concentrations with a greater dynamic range and less overlap between control and insulin-resistant subjects. Two competitive enzyme-linked immunoassays undervalued serum RBP4 concentrations in insulin-resistant subjects, possibly due to assay saturation. Poor linearity of dilution also limited assay utility. All assays tested exhibited greater immunoreactivity with urinary (C-terminal proteolysed) RBP4 than with full-length RBP4, the predominant form in serum.
These findings support the use of quantitative western blotting standardised to full-length RBP4 protein as a ‘gold standard’ method for measuring serum RBP4 in insulin-resistant states. Other assays should use full-length RBP4 and be extensively cross-validated using other methods.
KeywordsAdipokine Insulin-resistance Metabolic syndrome Obesity Type 2 diabetes Vitamin A
glucose disposal rate
matrix-assisted laser desorption ionisation-time of flight
retinol binding protein-4
Insulin resistance plays a key role in the development of type 2 diabetes. Even without hyperglycaemia, insulin resistance confers increased risk of cardiovascular disease and early mortality [1, 2, 3]. The term metabolic syndrome denotes a constellation of insulin resistance and cardiovascular risk factors such as hypertension, hypertriglyceridaemia, and low HDL-cholesterol . While obesity is the most common cause of insulin resistance and the metabolic syndrome, lean individuals may be insulin-resistant, frequently due to genetic background .
We recently reported that serum retinol binding protein (RBP4) is increased in insulin-resistant mice  and in a variety of insulin-resistant human subjects with obesity, impaired glucose tolerance or type 2 diabetes, and even in lean, normoglycaemic subjects with strong family histories of type 2 diabetes . Furthermore, serum RBP4 levels correlate inversely with insulin sensitivity measured by the euglycaemic–hyperinsulinaemic clamp technique. Serum RBP4 levels fell in subjects in whom exercise training improved insulin sensitivity and not in those who showed no improvement in insulin sensitivity . Therefore, serum RBP4 may provide both a useful marker and a relative index of the severity of insulin resistance among individuals with a variety of clinical presentations.
Commercial assays for measuring serum or urine RBP4 were originally manufactured for assessing vitamin A (retinol) status. While upper limits for RBP4 levels in normal subjects have not been established, minimum RBP4 concentrations are reported to be ∼15 to 30 μg/ml among healthy subjects without vitamin A deficiency [8, 9]. In vitamin A deficiency, liver stores of retinol are depleted [10, 11]. Since RBP4 must be bound to retinol for efficient secretion from hepatocytes , vitamin A deficiency reduces serum RBP4. Consequently, serum RBP4 concentrations correlate well with other indicators of vitamin A status, including serum retinol concentrations measured by HPLC  or total body retinol status measured by retinol stable isotope dilution methods [13, 14]. Until recently, the potential value of measuring RBP4 levels in the setting of insulin resistance had not been recognised, although elevations in serum and urinary retinol or RBP4 levels have been reported in humans with type 2 diabetes [15, 16].
Several commercial RBP4 assays use standard curves with ranges that could accommodate measurements of increased concentrations of serum RBP4 as observed in insulin resistance. However, these assays have not been validated for this application. Therefore, we sought to compare the performance of several methods for measuring RBP4. We analysed serum RBP4 from insulin-resistant subjects with impaired glucose tolerance or type 2 diabetes and from insulin-sensitive control subjects with normal glucose tolerance using three commercial assays and a quantitative western blotting assay previously validated for measuring serum RBP4 in insulin-resistant subjects . We found considerable discrepancy among the assays, particularly for serum RBP4 measurements in insulin-resistant subjects. We identify several factors that may influence assay performance, including assay saturation due to high concentrations of RBP4, poor linearity of dilution, and preferential reactivity of antibodies with the urinary form more than the serum form of RBP4. Urinary RBP4 is frequently used as a standard in commercial assays.
Subjects and methods
Comparison of assay performance
Quantitative western blotting
Competitive EIA no. 1
Competitive EIA no. 2
RBP4 (μg/ml, range)
23 ± 8(15–35)
64 ± 39 (30–151)
72 ± 33 (31–132)
34 ± 9 (25–49)
66 ± 26 (37–114)
65 ± 24 (31–104)
25 ± 4 (20–30)
37 ± 9 (23–58)
42 ± 13 (25–67)
34 ± 5 (25–39)
36 ± 6 (28–43)
28 ± 12 (6–40)
Intra-assay CV (%)
10.4 ± 4
12.5 ± 5
10.9 ± 4
4.6 ± 2
Inter-assay CV (%)
15.6 ± 6
15.2 ± 6
17.3 ± 8
Pearson correlation with GDR (r)
−0.80 (−0.91 to −0.58)dp < 0.001
−0.70 (−0.86 to −0.43)dp < 0.001
−0.58 (−0.80 to −0.25)dp = 0.002
0.11 (−0.2 to 0.09)dp = 0.69
Pearson correlation with western blot RBP4 (r)
0.80 (0.62 to 0.90)dp < 0.001
0.73 (0.51 to 0.87)dp < 0.001
−0.47 (−0.71 to −0.25)dp = 0.009
Assay dynamic range (% GDR dynamic range)
Serum collection and storage
Serum samples were collected after overnight fasting using standard clinical venipuncture methods and glass vacuum collection tubes without clot-activating beads, protease inhibitors, or anticoagulants. Samples were centrifuged and serum supernatant fractions collected according to standard clinical laboratory procedures. Samples were divided into aliquots in polypropylene microtubes for long-term storage at −80°C; two freeze–thaw cycles were made in the process of collection and dividing into aliquots for storage. Measurements were performed on samples that were thawed and analysed simultaneously using each assay method. Standard vacuum collection tubes containing heparin, EDTA or sodium citrate were used for plasma preparation.
Quantitative western blotting
Full-length recombinant RBP4 was expressed in E. coli, purified, and retinol-binding and transthyretin (TTR)-binding were verified as described . Proper disulfide bond formation was verified by matrix-assisted laser desorption ionisation-time of flight (MALDI-TOF) mass spectrometry. Protein concentration was measured by the Lowry method (DC kit; BioRad, Hercules, CA, USA) which reported concentrations comparable to the Bradford method (BioRad); however, the bicinchoninic acid/CuSO4 method underestimated concentrations of purified RBP4 (not shown). Standard solutions of 15, 30, 60 or 120 μg/ml RBP4 were prepared in Tris-buffered saline containing 1% NP-40. Standards and sera were diluted 1:100 in SDS-PAGE buffer and boiled for 5 min. Next, 15 μl of diluted standards or sera and molecular weight markers were electrophoresed on 18% Tris-glycine SDS-PAGE gels (Criterion brand; BioRad) and transferred to nitrocellulose. Western blotting was performed with non-fat milk (BioRad) as a blocking agent. Blots were incubated overnight at 4°C with primary antibody (A0040, DAKOCytomation, DAKO USA, Carpinteria, CA, USA) diluted 1:500 or 1:1,000 and for 1 h at room temperature with horseradish-peroxidase-conjugated secondary antibody (GE Healthcare Bio-Sciences, Piscataway, NJ, USA) diluted 1:2,000. Bands were detected by enzymatic chemiluminescence (Pierce Biotechnology, Rockford, IL, USA) and quantified with GeneSnap software (Synoptics/Syngene, Frederick, MD, USA). A single band for RBP4 migrated at about 21 kDa. Second-order polynomial curves were used to fit the purified RBP4 standards (r2 > 0. 99) for each individual gel and calculate concentrations in serum samples. This method provides absolute measurements of RBP4 concentrations with minimal gel to gel variability (∼10% CV for the same sample from duplicate gels). When RBP4 concentrations exceeded the upper limits of the standard curve (120 μg/ml), samples were diluted 1:2 and reanalysed. To avoid irregularities from gel quality or protein transfer, RBP4 standards were placed in the same position at the centre of each gel, and the four outermost lanes were not used for samples. To monitor assay quality, serum reference samples with low and high concentrations of RBP4 (28 and 110 μg/ml, respectively) were included on every gel. Blots on which reference sample values varied by more than 15% from expected values were considered defective and not used for measurements.
Commercial multiwell immunoassay kits
Quantitative western blotting
Untagged RBP4b (full length)
Competitive EIA no. 1
FLAG-RBP4c (full length)
FLAG-RBP4c (full length)
FLAG-RBP4c (full length)
Competitive EIA no. 2
Analysis of full-length and urinary RBP4 proteins
To characterise full-length recombinant and purified urinary RBP4 (Sigma-Aldrich, St Louis, MO, USA), 1 mg/ml stock solutions were made in phosphate buffered saline, desalted with a Ziptip (Millipore, Billerica, MA, USA), diluted in matrix buffer, spotted on a MALDI target, and analysed with an ABI4700 TOF/TOF mass spectrometer (Applied Biosystems, Foster City, CA, USA) in linear TOF mode. For comparison of western blot immunoreactivity, 20 to 160 μg/ml stock solutions of full-length recombinant or urinary RBP4 were measured on the same blot as described above for human serum. For comparison of ELISA immunoreactivity, a serum sample known to contain 15 μg/ml of serum RBP4 was supplemented (‘spiked’) with increasing concentrations of urinary or full-length RBP4 protein. Therefore non-RBP4 serum ‘matrix’ components were the same for every sample. Control samples containing 15 μg/ml serum not supplemented with purified RBP4 protein were analysed in parallel. RBP4 concentrations were measured using the ELISA manufacturer’s protocol and urinary RBP4 standards. Recovery of ‘spiked’ RBP4 was calculated as the concentration measured in each sample minus the quantity of RBP4 measured in the control sample.
Data are expressed as means±SD unless otherwise noted. Simple regression analyses with Pearson correlations were determined using Analyse-It software version 1.73 (Analyse-It, Leeds, UK). Dynamic range for each assay was calculated as the average variance of each individual RBP4 concentration from the mean concentration of all the samples in that assay. Dynamic range for the GDR values measured by euglycaemic–hyperinsulinaemic clamp was calculated using the same method. Since RBP4 and GDR are measured with different units, variance was normalised as a fraction of the mean for each parameter. Since GDR values displayed a greater dynamic range than the RBP4 concentrations we expressed RBP4 dynamic range as a percentage of the GDR dynamic range (arbitrarily defined as 100%) for each assay.
Quantitative western blotting and three commercial multiwell immunoassay kits (sandwich ELISA, EIA no. 1 and EIA no. 2) were used to measure serum RBP4 from subjects with normal glucose tolerance, impaired glucose tolerance, and type 2 diabetes (Table 1). Each assay used different combinations of antibody types and purified RBP4 proteins (Table 2). Quantitative western blotting and EIA no. 1 reported RBP4 concentrations in subjects with normal glucose tolerance of 23 ± 8 and 25 ± 4 μg/ml, respectively, values consistent with some reports for normal subjects without vitamin A deficiency (∼15–30 μg/ml) [8, 9]. In contrast, the sandwich ELISA and competitive EIA no. 2 reported higher RBP4 concentrations in subjects with normal glucose tolerance (34 ± 9 and 34 ± 5, respectively).
To determine why EIA no. 2 reported consistently lower RBP4 values among insulin-resistant subjects, we examined the raw optical absorbance unit data from that assay. Absorbance values for samples fell within the standard curve and expected values reported by the manufacturer (data not shown). However, EIA no. 2 employs a standard curve with a maximum concentration (40 μg/ml) that is lower than RBP4 concentrations found in many insulin-resistant subjects by western blotting or ELISA . We hypothesised that this discrepancy might be accounted for by assay saturation and loss of competitive binding kinetics due to increased quantities of RBP4 present in sera of subjects with impaired glucose tolerance or type 2 diabetes . Consistent with the notion of assay saturation, we found that RBP4 values measured by EIA no. 2 correlated negatively with RBP4 values measured by western blotting (Table 1). This observation suggests that higher actual concentrations of RBP4 produce greater saturation of the assay and lower apparent concentrations as measured by EIA no. 2.
To test for assay saturation, we lowered RBP4 concentrations in serum samples from subjects with impaired glucose tolerance or type 2 diabetes by making additional dilutions (1:2, 1:4, or 1:8) beyond that recommended by the manufacturer and re-measured the samples using competitive EIA no. 2. Dilution factors were selected individually for each sample based on expected RBP4 concentrations determined by quantitative western blotting. RBP4 concentrations obtained after these dilutions with EIA no. 2 (Fig. 1, adjusted) positively correlated with concentrations measured by western blotting (r = 0.76, CI 0.64–0.88, p < 0.001), and inversely correlated with GDR (r = −0.70, CI −0.86 to –0.42, p < 0.001; data not shown), confirming that assay saturation in competitive EIA no. 2 masks elevations of serum RBP4 in insulin-resistant subjects. The intra-assay %CV after adjustment (12.1%) was greater than before adjustment (4.6%) and similar to values for the other assays, also indicating assay saturation was overcome by the additional dilutions. We speculate that some degree of assay saturation could also contribute to lower RBP4 concentrations for subjects with impaired glucose tolerance or type 2 diabetes and the overall decreased dynamic range observed in competitive EIA no. 1 (Table 1).
Lack of linearity of dilution in RBP4 measurements by EIA no. 2
RBP4 standards (μg/ml)
Raw absorbance (AU)
Measured concentration (μg/ml)
Average measured concentration (μg/ml)
Final calculated concentration based on dilution (μg/ml)
Concentration measured by quantitative western blotting (μg/ml)
Serum sample A
Serum sample B
Specific antibodies and proteins used for standard curves are likely to influence measurements of serum RBP4. Except for competitive EIA no. 1, each assay used either urinary RBP4 or antibodies raised against urinary RBP4 (Table 2). Urinary RBP4 is modified by proteolysis at the carboxyl terminus [19, 20] and does not bind TTR. Thus, it is filtered by the glomerulus and excreted in the urine . Only a small fraction of serum RBP4 (<10%) is proteolysed in normal subjects [19, 22, 23]. Since carboxyl-terminus amino acid differences may influence the binding of antibodies to the protein, we hypothesised that assays might exhibit different affinities for full-length serum RBP4 and proteolysed urinary RBP4.
We determined how the different purified proteins behave in the sandwich ELISA assay, which utilises urinary RBP4 as the standard and an antibody raised against urinary RBP4. Identical serum samples were supplemented (‘spiked’) with different quantities of full-length or urinary RBP4 and the ELISA kit was used to measure RBP4 (Fig. 3c). Recoveries of the added proteins were calculated as described in Subjects and methods. As shown in Fig. 3c, the ELISA kit exhibited greater reactivity for urinary RBP4 than for full-length RBP4 at all concentrations. Whereas the assay failed to detect or underestimated concentrations of full-length RBP4 at lower concentrations, it exaggerated concentrations of urinary RBP4 at higher concentrations. These data indicate that differences in the reactivity of antibodies for urinary and full-length RBP4 may influence absolute values of RBP4 measured by immunoassays. Furthermore, these data predict that levels of serum RBP4 will be underestimated by assays that use urinary RBP4 instead of full-length RBP4 as a standard.
Serum RBP4 is elevated in insulin resistance and may indicate risk of type 2 diabetes and cardiovascular disease [6, 7]. Moreover, studies in mice indicate that elevated serum RBP4 can cause insulin resistance . In humans, serum RBP4 concentrations correlate highly with the severity of insulin resistance, and in subjects in whom exercise training improves insulin sensitivity, serum RBP4 concentrations are lowered . Therefore, accurate serum RBP4 measurements could provide a quantitative index of disease severity and a benchmark by which to follow interventions aimed at improving insulin sensitivity. Furthermore, since RBP4 levels are elevated in serum of lean, non-diabetic but insulin-resistant people with a strong genetic risk of type 2 diabetes , RBP4 may be useful as an early marker of diabetes risk.
In prior studies, western blotting and an ELISA sandwich assay produced RBP4 measurements that distinguished normal individuals from insulin-resistant individuals and correlated with magnitude of insulin resistance . That ELISA assay is no longer available. In the present study, we found that some commercial assays report higher RBP4 levels among insulin-sensitive subjects than other methods, and the levels are higher than expected for a normal population [8, 9]. The cause of this is not clear, but could reflect the background signal of the assay or differences in reactivity between the full-length RBP4 protein, which is the dominant form in serum, and the proteolysed urinary RBP4 used in these assays (discussed below). Therefore it seems advisable to routinely compare RBP4 levels of insulin-sensitive control subjects with normative population data. We find this practice also identifies samples in which RBP4 is degraded resulting in serum levels considerably lower than the reported range of normal for a given population.
We found that competitive EIA assays may selectively underestimate serum RBP4 levels in the setting of insulin resistance due to assay saturation. Additional dilution of serum samples rectified this problem in the case of one assay (competitive EIA no. 2). However, this may not be a practical strategy for correcting these assays, since universally increasing the dilution of all the samples could lower serum RBP4 concentrations below assay range for subjects with normal glucose tolerance. Furthermore, since EIA no. 2 did not exhibit linearity of dilution (Table 3), a prior knowledge of RBP4 concentrations measured by western blotting was needed to guide the selection of the dilution factors that enabled accurate measurements (Fig. 1, adjusted values). Therefore, adaptation of competitive assays for the measurement of RBP4 in insulin-resistant subjects will require protocol modifications more extensive than increased sample dilutions and may necessitate adjustments of antibody and/or competitor protein concentrations.
An unexpected finding was that RBP4 immunoassays (quantitative western blotting and sandwich ELISA) exhibited different reactivities for full-length RBP4 and urinary RBP4 modified by carboxyl terminus-proteolysis. Since urinary RBP4 is used as a protein standard and/or competitive binding reagent in some commercial assays, these differences could influence serum RBP4 measurements. Since full-length RBP4 is the dominant form in serum [19, 22, 23], use of full-length RBP4 as a protein standard or use of monoclonal antibodies that react equally with full-length and carboxyl terminus-proteolysed RBP4 would be preferable.
These findings further suggest that the abundance of proteolysed RBP4 endogenously present in serum could influence measurement of RBP4 by immunoassays. It has been proposed that RBP4 carboxyl terminus proteolysis occurs in circulation but results in rapid clearance of RBP4 by the kidney . It is not known whether proteolysed forms of RBP4 are increased in serum in insulin resistance or type 2 diabetes without renal failure. In 60 subjects with normal glucose tolerance, impaired glucose tolerance, or type 2 diabetes without renal impairment, we found no correlation between serum RBP4 concentrations and creatinine or urinary albumin concentrations (; Graham, Blüher and Kahn, unpublished data). However, since proteolysed RBP4 can accumulate in renal failure [19, 22, 23], we advise caution in interpreting serum RBP4 measurements in patients with advanced nephropathy.
Another potential cause for assay inaccuracy is interference from serum TTR. TTR circulates in a three- to fivefold molar excess over RBP4. Circulating RBP4 is highly bound to TTR in a 1:1 stoichiometric ratio, and there is little or no ‘free RBP4’ in circulation . Using gel filtration chromatography to analyse the RBP4-TTR complex, we found that increased serum RBP4 remains bound to TTR in insulin-resistant states (not shown). Nevertheless, because the affinity of RBP4-TTR binding is very strong , the relative stoichiometry and affinity of the two proteins in serum could conceivably influence kinetics of RBP4-antibody binding under the non-denaturing conditions employed by the commercial EIAs and ELISAs. In contrast, western blotting, which involves a denaturing SDS-PAGE step that separates RBP4 and TTR, is not affected by the RBP4-TTR interaction. The potential for TTR to interfere with RBP4 measurements by immunoassays deserves study.
Collection methods can alter RBP4 concentrations independently of measurement techniques (Fig. 4). In particular, plasma anticoagulants (e.g. heparin, EDTA, citrate or lithium chloride) may cause spurious results. Similar observations have been made by other investigators with regard to clot activating beads. Therefore, we advise use of glass vacuum tubes with no additives until optimal methods for collecting blood for RBP4 measurements are better defined. Subject characteristics that could also influence serum RBP4 include fasting/fed state and physical activity. Assays for rodents also require rigorous validation.
In conclusion, quantitative western blotting is the most reliable method for assaying serum RBP4 elevations associated with insulin resistance, because it has the greatest dynamic range and inverse correlation with insulin sensitivity (clamp GDR). Furthermore, quantitative western blotting reports RBP4 concentrations with the least overlap between insulin-sensitive and insulin-resistant subjects. A commercially available sandwich ELISA reports RBP4 concentrations that inversely correlate with insulin resistance, but values in normal subjects are higher than expected. Therefore, this assay underestimates the differences between insulin-sensitive and insulin-resistant people. Other assay methods, especially nephelometry, deserve testing. At present, we advocate the use of quantitative western blotting employing full-length RBP4 protein standards as the ‘gold standard’ method for measuring serum RBP4 in insulin resistance and for validating the accuracy of other assays.
The authors wish to acknowledge A. Oberbach for euglycaemic–hyperinsulinaemic clamp measurements. This study was supported by grants from the National Institutes of Health: R01 DK-43051 (to B. B. Kahn), K08 DK-69624 (to T. E. Graham); and from the Deutsche Forschungsgemeinschaft (German Research Foundation) BL 580/3-1 (to M. Blüher) and the Takeda Pharmaceutical Company (to B. B. Kahn).
Duality of interest
T. E. Graham and B. B. Kahn are listed as inventors on patent applications related to RBP4, and the latter also receives grant support from Takeda Pharmaceutical Company.