Dr Mark A Knepper, National Institutes of Health, Building 10, Room 6N260, 10 Center Drive, MSC 1603, Bethesda, MD 20892-1603, USA. Email: email@example.com
SUMMARY: Recent progress in biotechnology offers the promise of better medical care at lower costs. Among the techniques that show the greatest promise is mass spectrometry of proteins, which can identify proteins present in body fluids and tissue specimens at a large scale. Because urine can be collected in large amounts in a non-invasive fashion, the potential exists to use mass spectrometry to discover urinary biomarkers that are early predictors of renal disease, or useful in making therapeutic choices. Recently, the authors demonstrated that both membrane proteins and cytosolic proteins from renal epithelia are highly enriched in low-density urinary structures identified as exosomes. Exosomes were found to contain many disease-associated proteins including aquaporin-2, polycystin-1, podocin, non-muscle myosin II, angiotensin-converting enzyme, Na+K+2Cl- cotransporter (NKCC2), thiazide-sensitive Na-Cl cotransporter (NCC), and epithelial sodium channel (ENaC). Potentially, other disease biomarkers could be discovered by mass spectrometry-based proteomic studies in well-defined patient populations. Herein is described the advantages of using urinary exosomes as a starting material for biomarker discovery. In addition, the purpose of this review is to present an overall strategy for biomarker discovery in urine using exosomes and for developing cost-effective clinical assays for these biomarkers, which can potentially be used for early detection of disease, as a means of differential diagnosis, or as a means of guiding therapy. Finally, potential barriers that need to be overcome before urinary proteomics can be applied clinically, are emphasized.
A mass spectrometer is an instrument that measures the mass-to-charge ratio (m/z) of charged chemical species. Key breakthroughs were made in the early 1990s to create methods for ionizing peptides and delivering the resulting peptide ions to the near vacuum existing in a mass spectrometer.1,2 These breakthroughs accounted for Nobel prizes in Chemistry in 2002 to Fenn and Tanaka.1,2 In general, a given protein can be identified by digesting it first with trypsin (or another protease) and then measuring the molecular masses of the resulting peptides with a mass spectrometer (a peptide-mass fingerprint; Fig. 1a). Each protein has a unique peptide-mass fingerprint, allowing its identification by comparison with the theoretical mass fingerprints of every known human or animal protein. Because of the success of the human genome project, human protein sequence databases are nearly complete. The most typical approach to peptide-mass fingerprinting uses 2-D electrophoresis to isolate individual proteins and matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry to identify them.3
The power of mass spectrometry has been made even greater by the addition of tandem mass spectrometers (MS) and by coupling tandem MS to liquid chromatography (LC) to effect an initial separation of tryptic peptides (Fig. 1b). Using such an LC-MS/MS approach, the specific proteins contained in a complex mixture can be identified without the need for initial purification of individual proteins. This LC-MS/MS approach and peptide-mass fingerprint approach have already been applied to identification of proteins in urine by a number of laboratories.4–6
The purpose of this review is to present an overall strategy for biomarker discovery in exosomes isolated from urine and for developing cost-effective clinical assays for these biomarkers. We present a workflow paradigm for discovery of biomarkers, for validation of sensitivity and specificity of biomarker excretion in discrimination of target disease processes, and for creating cost-effective assays that can be made available for clinical use. We emphasize the major barriers to success with urinary proteomics including some seemingly mundane, but crucial questions such as how to collect urinary samples, how to store them, how to ship them, how to process them to enrich for low abundant markers, and how to quantify excretion rates of individual proteins. Finally, we will define criteria we believe that the clinical syndrome under investigation should meet for urinary proteomics to be useful, and we will provide examples to illustrate our point.
SOURCES OF URINARY PROTEINS
A major challenge of urinary proteomics is to develop a rational means of reducing the complexity of the urinary proteome in order to enhance the detectability of relatively low-abundance proteins that may have special pathophysiological significance. Theoretically, urinary proteins may originate from (i) glomerular filtration of plasma proteins; (ii) renal tubular secretion of soluble proteins; (iii) whole cell shedding (sloughed cells); (iv) apical plasma membrane shedding (non-specific or apoptotic processes); (v) glycosylphosphatidyl inositol (GPI) anchored protein detachment (e.g. Tamm–Horsfall protein); or (vi) exosome secretion.
Depending on the clinical syndrome, it might be prudent to select one or more of these protein types as a starting material for biomarker discovery. Because we are interested in structural and/or functional defects in renal epithelia, in the present paper we emphasize urinary exosomes as a starting material for biomarker discovery. Exosomes are small vesicles derived indirectly from the apical endosomal system and which contain many proteins already known to be associated with renal diseases.6 The water channel aquaporin-2 (AQP2), an apical plasma membrane protein, is one biomarker present in exosomes that can be readily measured in urine,7 and which has been exploited in studies of various water balance disorders.8
Secretion of urinary exosomes
After the discovery of AQP2 in the urine, various studies have been performed to determine the mechanism of its excretion. Immunoblot studies of different urinary fractions obtained from sequential centrifugation demonstrated that urinary AQP2 and apical Na+ transporter proteins are present chiefly in small, low-density membrane vesicles obtained by high-speed ultracentrifugation.7,9,10 In addition, immunoelectron microscopy has confirmed the presence of AQP2 in small, low-density membrane vesicles.7,9 This evidence suggested that whole cell or apical plasma membrane shedding was unlikely to be the main mechanism of apical plasma membrane protein excretion. However, further evidence was needed to explain the origin of small, low-density membrane vesicles. Wen et al. revealed that urinary AQP2 is predominantly excreted via an apical pathway in the collecting duct because AQP2, but not AQP3 (a basolateral membrane marker), was excreted in urine at a significant level.9
Recent studies have demonstrated that these small vesicles are so-called ‘exosomes’.6 Urinary exosomes are the internal vesicles of multivesicular bodies (MVB) that are delivered to the urinary space by fusion of the outer membrane of MVB with the apical plasma membrane of renal tubule epithelial cells. This mechanism is shown and described in more detail in Fig. 2. Exosomes have been shown to derive from every epithelial cell type facing the urinary space, from the podocyte to the transitional epithelium of the bladder. Exosomes have been previously demonstrated to be secreted by other cell types, including B cells, T cells, dendritic cells, reticulocytes, mastocytes, enterocytes, platelets, erythroleukaemia cells and other tumour cells.
The proteomic analysis of urinary exosomes using LC-MS/MS coupled with upstream 1-D sodium dodecylsulfate–polyacrylamide gel electrophoresis (SDS-PAGE) fractionation identified a roster of proteins directly involved in exosome biogenesis such as class E vacuolar-sorting proteins, which have been found to mediate MVB formation in yeast.6 Overall, the proteomic analysis of urinary exosomes identified 295 proteins, including at least 19 proteins already known to be associated with various renal diseases (Table 1).
Table 1. Proteins identified in urinary vesicles that are associated with kidney diseases or hypertension
Advantages of urinary exosomes as a starting material for biomarker discovery
The use of urinary exosomes for biomarker discovery has a number of important advantages, three of which will be addressed here.
Exosome isolation reduces the complexity of the urine proteome
The urine proteome contains a number of highly abundant proteins filtered from plasma as well as produced by renal tubular cells, especially Tamm–Horsfall protein. The presence of these abundant proteins complicates the detection of relatively low-abundance proteins that may have special pathophysiological significance. This common dilemma is analogous to the proteomic analysis of blood plasma, which has been limited by the presence of highly abundant proteins such as albumin and immunoglobulins. Exosome isolation helps to minimize highly abundant proteins in urine and also enriches a subproteome that includes both membrane proteins and cytosolic proteins trapped inside the exosomes. In the Barriers section we will describe the urinary exosome isolation method in more detail, including strategies to remove high abundant proteins.
Exosomes contain important information from broad epithelial origins
Urinary exosomes contain cytosolic proteins entrained when the exosomes are formed by invagination of the outer membranes of multivesicular bodies (Fig. 2), plus the proteome of the apical plasma membrane endocytic pathway, which plays a crucial role in the regulation of renal epithelial cell function. Our proteomic analysis of urinary exosomes identified proteins from all renal epithelia including the glomerular podocytes, the proximal tubule, the thick ascending limb of Henle, the distal convoluted tubule, the collecting duct, and the transitional epithelium of urinary bladder. Thus, based on our findings, urinary exosomes can be used to investigate the physiological or pathophysiological processes in virtually every epithelial cell type facing the urinary space.
Exosomes contain many disease-associated proteins
The proteomic analysis of urinary exosomes identified at least 19 proteins that are known to be involved in specific kidney diseases or in blood pressure regulation (Table 1). Additional studies are likely to reveal many other disease-related proteins in urinary exosomes. In the Barriers section we will discuss how these potential biomarkers may lead to clinical studies of urinary proteomics.
WORKFLOW PARADIGM FOR DISCOVERY AND VALIDATION OF URINARY BIOMARKERS
Exosome isolation may provide an efficient first step in biomarker discovery in urine. Figure 3 shows the major steps that will be required for the development of urinary biomarker assays for routine clinical application.
The first element is discovery of potential biomarkers. For this, the power of LC-MS/MS is well established and application to the analysis of urinary exosomes may be predicted to identify one or more proteins excreted at a greater or lesser rate in the affected population than in an appropriate control group. Extremely well-defined patient populations are needed for these studies to avoid heterogeneity. An alternative to LC-MS/MS is to use 2-D electrophoresis,3 although this technique may be somewhat less efficient in the detection of membrane proteins and low abundance proteins. Certain biomarkers may be identified that are completely absent in the control subjects but strongly positive in the patients. More likely, however, quantitative changes in biomarker excretion rate will have to rely on discrimination of patients from controls or on discrimination between different subpopulations of patients. In principle, identification of differentially expressed proteins and quantification of proteins can be done by use of the isotope-coded affinity tag (ICAT) method11 or by related methods.
Once a set of potential biomarkers is discovered, it is necessary to carry out validation studies in larger groups of patients to test the hypothesis that measurements of biomarker excretion can discriminate patients from controls, or can discriminate among different subgroups of patients (Fig. 3). In general, the strength of mass spectrometry is in protein identification and not in quantification, so it is likely that the validation phase can be completed more efficiently using antibody-based assays of protein abundance, rather than by quantitative mass spectrometry. The objective of these studies will be to obtain estimates of the sensitivity and specificity of the biomarker measurements in discriminating clinical groupings, prior to development of a product that can be applied to everyday clinical medicine. It is possible that a single biomarker will allow adequate sensitivity and specificity for some renal disease processes, and a single enzyme-linked immunosorbent assay (ELISA) might prove cost-effective. However, it is more likely that a combination of several biomarker measurements will prove necessary for disease discrimination. In this case, antibody arrays may provide an appropriate means of completing measurements for all relevant biomarkers at the lowest cost.12 Antibody arrays are ensembles of antibodies specific for the relevant biomarkers, laid down as individual spots on a solid substrate such that urinary proteins will bind to them in proportion to their abundance in the original samples. Labelling of these antibodies with one fluorescent dye and a reference sample with another fluorescent dye allows quantification relative to the reference by two-channel fluorescence quantification.13 Biomarkers relevant to a number of disease markers could be included in a single antibody array to allow a single array to be used for broad screening. For a more indepth review on the workflow paradigm, we recommend a recent review by Hewitt et al.14
Before urinary proteomics can proceed from bench to bedside, there are a number of important barriers that need to be overcome. These barriers are discussed here.
Collection, processing and storage of urinary samples
For quantitative comparisons it is necessary to measure the rate of excretion of putative biomarkers. Thus, a 24-h urine collection would be desirable. However, this is an inconvenient procedure that is characterized by low patient compliance and thus unreliable results.15 An alternative is to collect a spot urine sample and normalize the biomarker concentration by the creatinine concentration. For this, the first morning urine may be used, although this has the potential drawback of bladder and bacterial contamination due to a long residence-time in the bladder.16 Thus, from the perspective of practicality and sterility, ideal initial conditions for urine collection may be the second morning urine. Studies on proteinuria have shown that the method of collecting random urine samples is valid and that the rate of protein excretion correlates well with that found in 24-h urine.15
Once the urine is collected, another potential problem is proteolysis,17 which can be dealt with through the use of protease inhibitors in the collection vessel.17 However, some agents that are commonly used in the laboratory such as leupeptin and phenyl methyl sulphonyl fluoride (PMSF) are expensive. Hence, it will be important to determine the lowest cost set of protease inhibitors that are compatible with the protection of the collected human samples from degradation. Alternatively, it is possible that protease inhibitors may not be needed for studies involving spot urine collections that are processed or frozen immediately.
Subsequently, before the urinary exosomes are isolated, procedures can be carried out to remove tubular casts, cells and abundant proteins that may obscure lower abundant proteins. Examples of the latter include the Tamm–Horsfall protein6,18 and albumin when there is proteinuria.19 Casts and intact cells can be removed by low-speed centrifugation, albumin can be removed by albumin removal devices that use affinity ligands and, finally, Tamm–Horsfall protein can be removed with the use of reducing agents such as dithiothreitol.6 Although the removal of Tamm–Horsfall protein is necessary for the analysis of exosomes, it was recently shown that this protein itself may have pathophysiological significance.20 Hence, it could be possible that a simultaneous analysis of exosomes and other non-exosome proteins such as Tamm–Horsfall may provide the most complete pathophysiological information.
At least four different isolation protocols for urinary proteomics have been described.21 Due to the variability in protein chemical and physical properties, different methods will isolate different subpopulations of urinary proteins,21 each of which may be appropriate for different clinical questions. We have used a protocol that employs ultracentrifugation to precipitate exosomes from urine.6 This approach is potentially most valuable for the detection of defects in renal tubular function. It has been noted that this isolation protocol will enhance the detection of hydrophobic proteins in the apical endosomal pathway, whereas other isolation protocols, such as the acetone precipitation technique, will increase the identification of hydrophilic proteins that are derived from sources other than exosomes.21
Another point that merits emphasis is that ultracentrifugation will be an impractical method to use for large-scale studies and for the ultimate clinical application of urinary proteomics because of the time and cost involved. Possible alternative methods for isolation of exosomes include filtration, adsorption, evaporation, dialysis, and/or immuno-isolation.
Storage of samples may be required either directly after the initial urine collection or after the actual exosome isolation. In addition to storage of samples, shipping of samples may also be required, especially if different laboratories collaborate in urinary proteomics projects. Prior to the undertaking of biomarker discovery studies in urine it would be important to develop standardized methods for storing and shipping of samples.
Freezing of samples can potentially cause loss of exosomes due either to aggregation of vesicles or adsorption of exosomes to the surface of the collection vessel. Both problems could hypothetically be dealt with by adjusting sample pH or by adding a surfactant prior to freezing to minimize attraction between the vesicles. With regard to the degradation of frozen proteins over time, evidence from previous studies suggests that there is considerable interprotein variation in the degree of degradation.22,23 Furthermore, these studies concluded that if urinary proteins can be measured within 4 weeks after sampling, storage at 4°C appears to give the best results. It may be desirable to add biocides such as sodium azide or thymol. For longer time-periods, a storage temperature of −70°C or −80°C may be optimal. In both studies, storage at −20°C appeared least favourable both due to a higher degradation rate and the formation of a calcium-phosphate precipitate that was observed at this specific temperature.22,23 The adjustment of pH prior to freezing did not appear to affect protein stability.23
Quantification of urinary protein excretion and determination of its normal variability
A crucial step in urinary proteomics is the translation from biomarker discovery in a research setting to the diagnostic application in a clinical setting. This translation requires that a differentially expressed protein or group of proteins provides sufficient sensitivity, specificity and positive predictive value to be applied as a diagnostic test. As was discussed in the working paradigm, this translation step should occur in the validation phase (Fig. 3), where antibody-based assays of protein abundance rather than quantitative mass spectrometry is likely to be the most cost-effective technique. However, before differences in protein abundance can be interpreted, two questions should be addressed. The first question is how to quantify urinary protein excretion rates and the second question is how to establish the normal variability in excretion of a given urinary protein.
Theoretically, the best choice for quantification of protein excretion is to carry out a timed urine collection and express the excretion rate in conventional rate units such as nmol/h. However, as discussed previously, patient compliance is a frequent problem with timed urine collections and most investigators have adopted approaches that can be applied to spot urines. A frequently used approach is to normalize by the excretion rate of creatinine, inulin or other filterable but non-reabsorbable markers, thereby eliminating the time term in the variable.7 If creatinine is used, it must be recognized that creatinine excretion rate is not invariant.24
In the discovery phase, validation phase and clinical application phase, sample measurements in patients must be related to normal reference samples to draw conclusions. Thus, it is important to define a method for obtaining normal reference samples. To approach this problem, it would be useful to determine the intrinsic variability in the urinary proteome of normal human subjects. Potentially, the possible effects of age, gender, circadian rhythm, high- and low-salt diets and water loading and restriction need to be investigated. The sensitivity, specificity and thus the clinical value of a candidate biomarker can be defined only relative to a carefully chosen control.
CRITERIA FOR AND EXAMPLES OF URINARY PROTEOMICS STUDIES
Regardless of the specific diseases in which urinary proteomics may be applied, a number of general criteria that a disorder should meet for urinary proteomics to be useful should be defined. We argue that in this time of rising health-care costs, biomarker discovery studies should be pursued only for patient populations for which measurements of the identified biomarkers have a realistic possibility of providing a significant improvement in patient care. We further argue that priority should be given to diseases for which interventions are likely to lead both to increased quality of life and reduced patient care costs. Finally, the diagnostic tools that are currently available for a specific disorder should be taken into consideration. Biomarker discovery may be pursued especially if the current tools are non-existent or inadequate due to cost, complexity, risk or accuracy.
In this paradigm, improved clinical decision making through the use of urinary proteomics may include the early detection of disease, the differential diagnosis of diseases with similar presentation but different therapeutic implications, and guidance in the selection, evaluation or adjustment of a therapeutic regimen. A good but ambitious example in nephrology would be a situation where urinary proteomics could direct the physician to better diagnostic and therapeutic strategies that would either delay or even prevent end-stage renal failure.
Candidate biomarkers for the near future include the disease-associated proteins that were identified in urinary exosomes (Table 1), many of which may have diagnostic potential. One candidate biomarker is the Na+/H+ exchanger isoform 3, which in patients with acute renal failure was shown to be a marker of renal tubule damage, and may therefore help to differentiate prerenal azotemia, acute tubular necrosis, and intrinsic acute renal failure other than acute tubular necrosis.25 Another important biomarker may be polycystin-1, the protein product of a gene responsible for autosomal dominant polycystic kidney disease, the most common genetic disease leading to renal failure.26 Polycystin-1 is of low abundance in kidney tissue, but is readily detectable in urinary exosomes.6 Lee et al. demonstrated that immunoblot analysis of urinary exosomes can differentiate two different types of mutations for the thiazide-sensitive Na-Cl cotransporter of the distal convoluted tubule, and thus may become a very simple and useful diagnostic tool to subclassify Gitelman's syndrome.27 Furthermore, urinary proteomics has recently been applied in predicting diabetic nephropathy,28 in the early detection of acute renal failure14,29 and acute allograft rejection,30 and in the assessment of the progression of renal disease in the renal Fanconi syndrome.5
Besides the application of urinary proteomics to patient populations, another potential of clinical proteomics is what has been coined ‘patient-tailored’ or ‘personalized molecular medicine’.31 In a model for personalized cancer care, clinical proteomics was proposed for the early detection of cancer via serum proteomics patterns and subsequently in therapeutic fine-tuning via proteomic pathway profiling in diagnostic, post-therapy and post-relapse biopsies.31 In analogy with this model, one could hypothesize a similar future role for urinary proteomics in hypertension, the treatment of which is currently largely empirical. With the kidney being the major regulator of blood pressure,32 one could argue that subtypes of essential hypertension may have unique fingerprints in the urinary proteome. If so, these fingerprints may be used to predict the optimal antihypertensive drug regimens for each subtype and may eventually help to personalize therapy and classify this heterogeneous disorder.
It should be borne in mind that the examples cited here represent first-level urinary proteomics studies and that the clinical value of these approaches remains to be determined in comparative studies. Furthermore, as was pointed out recently,33 the focus for the near future should be to further increase our understanding of the regulation of the urinary proteome and to design clinical urinary proteomics studies with small and extremely well-characterized patient cohorts to identify new biomarkers for clinically important renal disease processes.