Blood Proteomics for Biomarkers of Kidney Pathology

Biomarkers have particular utility when they can direct clinical care to avoid invasive diagnostic or therapeutic procedures, such as a kidney biopsy. Discovery of novel biomarkers has recently been accelerated by the use of high-throughput proteomics. In this issue of JASN, Kim et al.¹ employ aptamer-based proteomics on human plasma samples paired with biopsies from the Boston Kidney Biopsy Cohort to discover biomarkers of specific lesions on kidney biopsy. These efforts aim at enabling noninvasive evaluation of histopathology that may help defer kidney biopsy in certain clinical contexts. Kim et al. report 35 profiled proteins that significantly associate with specific forms of histopathology, including testican-2 with glomerulosclerosis and NELL1 with interstitial fibrosis and tubular atrophy (IFTA).

For many years, liquid chromatography tandem mass spectrometry (LC-MS/MS) has been the method of choice for proteomics. Proteins in a sample are digested with trypsin, and resulting peptides are purified before being analyzed with mass spectrometry (MS). The peptides' mass-to-charge ratios are determined, and the peptides are ionized and further fragmented to enable sequencing. The identity and quantitative abundance of proteins in a sample is determined from the MS/MS spectra produced by the peptides and peptide fragments, which also enable a determination of the sequence coverage of proteins. Among the strengths of LC-MS/MS, it has the ability to identify both proteins and isolated protein domains or fragments, as well as their post-translational modifications, whether those modifications result from normal physiology or disease states.

Significant challenges remain in applying LC-MS/MS to biomarker identification in bodily fluids, such as plasma. The range of concentrations of plasma proteins spans more than ten orders of magnitude, and peptides from abundant proteins, such as albumin or immunoglobulins, can mask the signal from scarce proteins. Accordingly, LC-MS/MS sensitivity is often attenuated in samples with a large range of protein concentrations, entailing complex workflows to fractionate biological specimens before the LC-MS/MS analysis.

In part to address this problem, novel proteomics platforms are being developed, including ones based on aptamers (SOMAScan) and antibodies (Olink). Kim et al. employed the SOMAScan platform that utilizes single-stranded DNA oligonucleotides, or SOMAmers, that recognize nearly 7000 target proteins.² Each SOMAmer is engineered to be specific to a single protein target, and fluorescent tags on the SOMAmers facilitate quantitation of protein abundance. Multiple groups have now employed aptamer-based proteomics to make associations between circulating plasma proteins and progression and risk of CKD.^3–6 By contrast, Olink targets >5400 proteins with pairs of antibodies tagged with complementary single-stranded DNA oligonucleotides. The oligonucleotides hybridize when the antibody pair binds the target protein, and this double-stranded DNA enables quantification of protein abundance through next-generation sequencing. Importantly, SOMAScan and Olink can yield divergent biomarker–phenotype associations when analyzing the same samples,⁷ warranting careful validation of results from these new platforms.

In this issue of JASN, Kim et al. now extend the use of SOMAScan to the proteomics of specific histopathologic lesions, aiming to identify biomarkers of specific biopsy findings. The authors identify 35 of 6592 profiled proteins with significant associations to distinct phenotypes on biopsy in a cohort of 434 patients from the Boston Kidney Biopsy Cohort. Testing for associations of log₂-fold protein abundance with biopsy lesions was performed using regression models after adjustment for age, sex, ethnicity, eGFR, and proteinuria. The biopsy lesions associated with significant changes in protein levels included IFTA, arteriolar sclerosis, and glomerulosclerosis. To interrogate whether changes in abundance of plasma proteins associated with changes in kidney expression, protein gradients were measured across the renal artery and vein for each protein identified in the SOMAScan assay.

Of special interest in the dataset are NELL1 and testican-2, two proteins that had negative correlations with IFTA and glomerulosclerosis, respectively. Both these proteins had statistically significant changes in measured concentrations in renal arteriovenous gradients, suggesting that they are secreted by kidney cells into the passing blood. Moreover, single-cell RNA expression data from the Kidney Precision Medicine Project (KPMP) demonstrated kidney cell type–specific gene expression of NELL1 in tubules and testican-2 in podocytes. Consistent with the authors' results that testican-2 negatively correlates with glomerulosclerosis severity, testican-2 has been previously shown to positively correlate with eGFR in the Jackson Heart Study and Framingham Heart Study.⁶ Taken together, these results suggest that circulating levels of NELL1 and testican-2 may represent new markers of tubular and podocyte health, respectively.

Although deferring biopsy in the context of nephrosis, nephritis, or CKD of unknown cause would seem to require biomarkers of exceptional specificity, already serologic approaches to diagnosing and treating kidney disease are taking hold in certain clinical contexts, among the most famous being phospholipase A2 receptor antibody in membranous nephropathy.⁸ In patients with positive phospholipase A2 receptor serologic tests and no evidence of kidney insufficiency or secondary causes of nephrosis, some clinicians will now diagnose membranous nephropathy in the absence of kidney biopsy. Although Kim et al.'s findings do not have this level of specificity, their results raise the possibility that blood biomarkers may provide useful risk information that could help in guiding clinical considerations for rebiopsy. If a histopathologic lesion is confirmed on an initial biopsy for which there is an established biomarker with strong correlation, longitudinal measurement of the biomarker may potentially inform clinical decision making.

Several important limitations of this study need to be mentioned. First, the authors uncover new correlations between circulating proteins and histology, but these correlations do not necessarily reflect causal relationships. For example, the observed inverse correlation of blood testican-2 levels with glomerulosclerosis may be reflective of a protective role of testican-2 in podocytes or may be entirely secondary to podocyte loss in the setting of glomerulosclerosis. Mechanistic studies would be needed to distinguish these possibilities. Second, some of the positive correlations between circulating proteins and histology markers of disease severity, such as IFTA, may simply reflect reduced renal clearance that is not adequately accounted for by adjustments for eGFR, an imperfect proxy of kidney function. This may explain why most putative biomarkers of IFTA exhibited positive coefficients of association while most of the corresponding genes had no significant enrichment of expression in specific kidney cells. Third, the reported associations have not been independently validated yet. KPMP may be one of the best cohorts to follow-up on these observations, since both plasma and urine samples are being collected at the time of kidney biopsy. To avoid the “winner's curse” bias, detailed biomarker performance characteristics (including area under the receiver operator characteristic curve, variance explained, etc.) should also be derived based on an independent validation set. Ultimately, combining plasma and urine proteomics with intrarenal molecular profiles will provide the most powerful approach to define biomarkers of specific intrarenal pathology and will soon become possible with the expanding sample size of the KPMP cohort.

Given that high-throughput proteomics studies with SOMAScan and Olink are increasingly common, it also bears noting some of the limitations intrinsic to these new technologies. Since these platforms utilize aptamers or antibodies that recognize proteins through affinity for a specific binding site, failure to recognize a protein may occur through truncation that cleaves the binding site or through chemical modification of the binding site. There are numerous mechanisms in CKD whereby plasma proteins can undergo modification. The uremic toxins present in plasma in CKD can lead to O-sulfation of tyrosine residues, reactive oxygen species can produce dityrosine or novel carbonyl groups, and isocyanic acid deriving from urea metabolism can cause carbamylation, such as in the conversion of lysine to homocitrulline after binding of isocyanate to the amino group of the lysine side chain. Type 2 diabetes mellitus, a frequent comorbidity in the CKD population, is associated with glycation of diverse plasma proteins, including hemoglobin. These modifications have the potential to reduce or even eliminate the affinity of aptamers or antibodies for their targets, and these modifications themselves are not identifiable by the new proteomic platforms.

Alongside protein modifications, CKD is also frequently accompanied by metabolic acidosis and hypocalcemia, conditions that can lead to the dissociation of protein complexes that are dependent either on charge or calcium coordination for their association, thus releasing isolated proteins into plasma. While an aptamer- or antibody-binding site may be occluded in the protein complex, it can be revealed after dissociation due to acidosis or hypocalcemia, leading to an increase in signal even if the absolute protein abundance has not changed.

Although the increasing throughput and resolution of the new proteomic platforms in complex biological samples such as plasma ensures their continued use, there is still an important role for LC-MS/MS in proteomic profiling, especially in states such as CKD where some biomarkers may harbor multiple chemical modifications induced by the disease. LC-MS/MS can also accurately measure protein abundance, regardless of dynamic changes in protein conformation and protein assemblies due to disease. Adaptations of LC-MS/MS to facilitate its use on plasma and urine are ongoing, including the use of high-resolution isoelectric focusing and nanoparticles.^9,10 Thanks to the ever increasing sophistication of proteomic approaches, we are now on the verge of major advances in biomarker discovery for CKD and related traits.

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/JSN/E686.

Funding

K. Kiryluk: NHGRI (U01 HG013201) and NIDDK (2R01 DK105124, 1R01 DK136765, and RC2 DK116690). A. Beenken: NIDDK (K08 DK132511).

Author Contributions

Conceptualization: Krzysztof Kiryluk.

Writing – original draft: Andrew Beenken.

Writing – review & editing: Andrew Beenken, Krzysztof Kiryluk.