Application of Liquid Chromatography–Mass Spectrometry Technology for Early Detection of Microalbuminuria in Patients with Kidney Disease

Proteinuria is an early and sensitive marker of renal damage in many forms of chronic kidney disease. In diabetes mellitus, mild albuminuria occurs early in the course of renal damage, and this microalbuminuria, defined as urinary albumin excretion of 30–300 mg/day (∼20–200 μg/min), is the best noninvasive predictor of subsequent clinical diabetic nephropathy. Without proper treatment, early diabetic kidney damage progresses to end-stage kidney failure in 5–7 years. With early detection, however, the onset of kidney disease can be slowed, halted, and in some cases reversed through treatment with drugs such as inhibitors of angiotensin-converting enzyme and blockers of angiotensin-2 receptor, particularly if drug treatments is combined with tight blood sugar control and, possibly, decreased protein intake (1)(2). Therefore, strong interest is directed toward detecting patients at risk of diabetic nephropathy as early as possible (3). Current microalbuminuria assays use various immunochemical methods (4). These assays may lead to underestimation of albuminuria in some patients, possibly contributing to delayed diagnosis of early diabetic nephropathy, as indicated by data generated with a recently developed, FDA-cleared size-exclusion HPLC with ultraviolet detection (SEC-HPLC-UV) method (Accumin™, AusAm Biotechnologies). Urine sample analysis with this assay has yielded consistently higher albumin concentrations than immunoassays (5). In one study, microalbuminuria was detected by SEC-HPLC-UV in 53% of urine samples from diabetic patients that were normalbuminic according to an immunochemical method (6). In another study, the SEC-HPLC-UV method was reported to detect microalbuminuria 1–12 years earlier than widely used commercial tests (7). It has been hypothesized that the higher values obtained with the SEC-HPLC-UV assay are attributable to detection of modified forms of albumin. These redox and glycation modifications of the albumin do not change the size of the albumin markedly, and thus with the SEC-HPLC-UV method they are detected under the same peak as albumin. In immunochemical methods, the specificity of the antibody used in the immunoturbidometric method precludes detection of these modified forms (6)(8)(9). Most recently, several other proteins were found to be coeluted along with albumin during SEC-HPLC-UV analysis (10). This finding could explain disparity between immunoassays and SEC-HPLC-UV in urinary albumin measurements. According to this study, the size-exclusion limit of glomerular filtration is not specific for albumin, and therefore urine also contains other plasma proteins with a relative molecular mass (Mr) similar to albumin. We developed an assay based on liquid chromatography–mass spectrometry (LC-MS) for measurement of urinary albumin, which could serve as a candidate reference method for detection and quantification of intact albumin in urine. Further objectives were: (a) to develop such an assay suitable for clinically laboratories with standard LC-MS equipment, (b) to validate the method’s analytical and functional sensitivity, precision, accuracy, and linearity, and (c) to compare the new method with immunoassay and SEC-HPLC-UV methods. We used purified preparations of human serum albumin (HSA) (Sigma-Aldrich) and bovine serum albumin (BSA) (Sigma-Aldrich) for preparation of calibrators and internal standard, respectively. We confirmed >98% purity of calibrator and internal standard preparations with a Micromass® Q-Tof™ MS (Waters Corporation), and then verified the concentrations by absorbance spectroscopy at 280 nm. In our on-line extraction method, 0.1 mL of patient urine was mixed with 20 μL of 1 g/L BSA (internal standard) dissolved in water purified by reverse osmosis. The urine sample was injected into the LC-MS with an HPLC autosampler onto a 2.0 cm × 2.1 mm, 5 μm Supelco Discovery® BIO wide-pore C-8 column. Mobile phases consisted of (a) water + 0.1% formic acid, and (b) acetonitrile + 0.1% formic acid; total flow rate was 0.3 mL/min. During the first 5 min, the organic mobile phase content was increased from 5% to 30% with a linear gradient, and the flow was diverted to waste with the help of a multiplexing HPLC system (Cohesive Technologies Inc.). During the next 4 min, albumin was eluted by use of an increase in the gradient content of the organic phase from 30% to 95%, and sample flow was redirected to the mass spectrometer (ABI Sciex API 5000 LC-MS/MS system). We performed ion suppression studies according to a method described by Annesley (11), and no signal suppression was observed (Fig. 1A1 ). It has been previously shown that it is possible to induce fragmentation of the first 30 residues from the N terminus of albumin. These authors also showed that any substantial, abundant chemical modifications occur in a region beyond the first 30 residues from the N terminus of the albumin. Therefore, the first 30 residues should be representative of both modified and unmodified albumin. We achieved the N-terminal fragmentation of albumin during ionization in the MS electrospray ion source by raising the declustering potential to 350 V. The multiple ion mode was used to detect the resulting b244+ and b243+ fragment ions at m/z 685.1 (HSA b244+), 913.2 (HSA b243+), 698.5 (BSA b244+), and 930.9 (BSA b243+) (bn; n = number of amino acids). The MS spectra of N-terminal fragments of HSA and BSA and a chromatogram of a typical patient urine sample are shown in Fig. 1A1 . The lowest analyte concentration indistinguishable from blank, determined by assaying 20 replicate blanks, was 2.5 mg/L. The limit of detection (analytical sensitivity), defined as the lowest analyte concentration that can be distinguished from blank with >95% certainty was determined to be 4.84 mg/L by assaying 20 replicates of a 5 mg/L sample. The limit of quantification (functional sensitivity interassay CVs <20%) was 20 mg/L. Precision studies of patient pooled samples yielded intraassay CVs (n = 20) of 10% (20 mg/L), 9.7% (70 mg/L), and 5.3% (270 mg/L); the corresponding interassay CVs (n = 14) were 14.6%, 8.2%, and 10.6%. We assessed the linearity and interassay variability for the 6-point calibration curve (12–420 mg/L) with a set of calibrators assayed with each analysis and monitored for 6 days. The calibration curves were linear and reproducible with the following linear regression equation: y = 0.004x + 0.10; r2 = 0.998. Recovery of purified human albumin (40, 80, and 170 mg/L) into patient samples (n = 5) was 80%–124% (mean, 99%) of predicted; dilution–linearity (1:2, 1:4, and 1:8) was 92%–112% (mean, 102%). For 68 patient samples, we compared the results of urine albumin measurements obtained with our LC-MS method with corresponding measurements obtained by an automated immunoturbidometric assay (Roche Hitachi 912, Roche Diagnostics) and by an Accumin SEC-HPLC-UV method. The comparison shows that the LC-MS method is relatively in agreement with the immunoturbidometric method (y = 1.06x − 15.49; R = 0.82), whereas albumin concentration measurements obtained with the Accumin SEC-HPLC-UV method were ∼38% higher (y = 1.38x − 1.89; R = 0.86) (Fig. 11 , B and C). Because we verified purity and concentrations of our standards, we are confident that calibration of our LC-MS assay is accurate, as is, by implication, the calibration of the Hitachi assay. As previously suggested by Sviridov et al. (10), because the Accumin method is intrinsically unable to clearly separate transferrin, α1 acid glycoprotein, and α1 antitrypsin from albumin. The overestimation of albumin could be caused by coelution of these proteins (10). Several other proteins, such as monomeric IgG, α1-microglobulin, and free immunoglobulin light chains, may also overlap with the SEC-HPLC-UV albumin peak (10). All of these variables would tend to contribute to the overestimation of albumin measurement by SEC-HPLC-UV compared with immunoassays. In contrast, MS would appear to be a superior technique in this regard because it can distinguish albumin unequivocally from coeluting interferences. It has been suggested that urinary albumin is biochemically modified by lysosomal enzymes in healthy individuals, leading to the excretion of <1% intact albumin and >99% albumin-derived fragments of Mr <10 000 (13)(14). Although this hypothesis is not universally accepted (15), changes in excretion patterns or concentrations of these fragments might indicate progression toward nephropathy. Clinically, this hypothesis is supported by the finding that some normoalbuminuric patients with type 1 diabetes have advanced glomerular lesions (16). For these patients, biomarkers other than albuminuria are needed to assess the risk of nephropathy (17). Because correlation with LC-MS was not ideal with currently used methods, neither SEC-HPLC-UV nor immunochemical methods appear suitable for this task. The former is not specific for albumin and its fragments, and the latter can detect only fragments with Mr >12 000 and albumin aggregates (18)(19). Similarly, although our current LC-MS method may detect albumin fragments with an intact N terminus, it will not detect fragments that lack the N-terminal portion. More experiments are needed to modify this method to target N-truncated albumin fragments, with or without concomitant C truncation. In conclusion, we have developed a sensitive and specific LC-MS assay for detection and measurement of urinary albumin. The method can be performed with standard clinical LC-MS instrumentation and has high throughput and good precision because of online extraction. Preliminary data suggest that this method yields measurements that are in general agreement with results obtained by a traditional immunoassay method, suggesting that underdetection of albumin by immunoassay methods might not be a major problem. Further efforts to improve early diagnosis of nephropathy should therefore concentrate on albumin fragments and alternative biomarkers, rather than different detection methods for intact albumin. LC-MS appears to be an ideal method for such future studies. Total ion chromatogram and comparison of urinary albumin concentrations. (A), total ion chromatogram of a microalbuminuric patient urine sample. The abscissa shows elution time, the ordinate signal intensity in counts per second. The internal standard (BSA) 24 amino acid N-terminal fragments in their 3+ charge and 4+ charge states are depicted in green and gray, respectively. The corresponding 3+ and 4+ fragments of the albumin in the patient sample are shown in blue and red, respectively. The purple trace shows the result of ion suppression test where the analyte was injected directly into the mobile phase. The inset shows the corresponding ion scan mass spectra of the N-terminal fragments obtained for HSA (top) and BSA (bottom). The b24 fragments observed are unique to human albumin and BSA, justifying our choice of BSA as an internal standard. (B) and (C), comparison of urinary albumin concentrations obtained by LC-MS (abscissa on both panels), with corresponding results obtained by automated immunoturbidometry (B, ordinate; Roche-Hitachi 912), and SEC-HPLC-UV (C, ordinate; Accumin) with 68 patient samples. Passing-Bablock linear fit (solid lines) and 95% confidence intervals of fit (dashed lines) are superimposed.