Refining the Composition and Significance of Human Kidney Intratubular Casts Using Spatial Protein Imaging

Visual Abstract

Abstract

Key Points

• Used multiplex protein imaging to elucidate the intratubular cast components and the associated tubular alterations.

• Identified (Prominin-1/CD133), a dedifferentiation marker, as a major constituent of intratubular casts.

• Protein components within casts were associated with the injury of the surrounding tubular epithelium.

Background

Kidney intratubular casts are frequently observed in the distal nephron segments of the kidney and have long been regarded as a sign of kidney disease. However, the composition and pathologic significance of intratubular casts have remained understudied.

Methods

We leveraged Hematoxylin and Eosin (H&E) staining to identify intratubular casts along with concurrent codetection by indexing multiplexed spatial protein imaging on human kidney biopsy sections from the Kidney Precision Medicine Project. We also conducted immunoblotting of Prominin-1 (PROM1/CD133) in urine and assessed its levels from publicly available urinary proteomics datasets of the Kidney Precision Medicine Project consortium.

Results

We analyzed 493 intratubular casts across 42 individuals with kidney disease or healthy controls. We identified PROM1 and insulin-like growth factor binding protein 7 as major constituents of casts (positive staining in 89.0% and 39.1%, respectively). Staining for uromodulin, an established cast component, was present in 86.6%. These components showed variable patterns across disease states. Intratubular casts were predominantly detected in the distal nephron segments, and their presence was associated with a marked loss of sodium-chloride cotransporter and aquaporin-2 expression in the cast-containing tubular epithelium, suggesting underlying injury. The loss of these transporters correlated with protein components within casts, and the presence of intracast PROM1 showed the strongest association, with an odds ratio of 26.7 (95% confidence interval, 13.1 to 54.7). Urinary PROM1 secretion was confirmed by immunoblotting and was greater in patients with AKI compared with healthy controls (P = 0.01).

Conclusions

We identified PROM1, a dedifferentiation and injury marker expressed in epithelial cells, as a novel major constituent of intratubular casts. Our studies suggest that protein composition signature within casts varies with disease state and is associated with tubular injury in distal nephron segments. Our study also suggests that urinary PROM1 may have potential as a biomarker for AKI.

Introduction

Casts are cylindrical structures formed in the lumen of distal nephron segments of the kidney, mainly in distal convoluted tubules (DCT) and collecting ducts (CD).^1,2 Casts can comprise protein aggregates and cellular debris, including epithelial cells.¹ Microscopic analysis of excreted urinary casts is a common diagnostic method in clinical nephrology. Urinary casts can be observed in healthy individuals, and an increased abundance of urinary casts is generally regarded as evidence of intrakidney pathology.^3,4 Casts that are retained within kidney tubules, termed intratubular casts, may form obstructive plugs.⁵ While tubular obstruction with intratubular casts has been proposed to cause epithelial damage,⁴ the underlying composition and mechanisms are poorly defined.

Uromodulin (UMOD) is a protein synthesized in the cells of the thick ascending limb (TAL) and is the most abundant secretory protein in normal urine.⁶ UMOD is a major constituent of hyaline casts,^2,7 the most common type of tubular casts. However, the role of UMOD in tubular cast formation and pathology remains enigmatic.⁸ In addition, the presence and role of other proteins in intratubular casts remain unclear. In situ multiplex protein expression profiling in kidney biopsy specimens is a key first step to understand intratubular cast composition and associated tubular damage.

Codetection by indexing (CODEX) is an emerging technology for highly multiplexed spatial protein imaging using DNA-conjugated antibodies. CODEX allows simultaneous in situ visualization of over 60 markers, enabling protein expression profiling at a single-cell spatial resolution.⁹ We have established CODEX imaging and analytical pipeline on human kidney specimens.¹⁰ In this study, we used CODEX imaging to elucidate the spectrum of tubular cast components and the associated tubular alterations in human kidney biopsy specimens. We identified Prominin-1 (PROM1, also known as CD133), a marker associated with injury and altered tubular cell state, as a major constituent of casts. Intratubular casts were detected mainly in distal nephron segments and were associated with a remarkable loss of sodium-chloride cotransporter (NCC) and aquaporin-2 (AQP2) expression in the surrounding tubular epithelium, suggesting that intratubular casts contribute to the distal nephron injury.

Methods

Human Kidney and Urine Specimens

Human kidney specimens collected by the Kidney Precision Medicine Project (KPMP) consortium were acquired with informed consent and approved under a protocol by the KPMP single Institutional Review Board (IRB) of the Washington University in St. Louis (Approval No. 20190213). Human kidney tissue biopsies were preserved in Optimal Cutting Temperature (OCT) medium. The healthy reference specimens were obtained from living kidney donors. For kidney disease patients, the primary adjudicated category was used to classify a subject as AKI versus CKD. In detail, the adjudication team at KPMP, consisting of clinicians and pathologists, reviewed the whole clinical and pathologic characteristics of each case and reached a consensus on one of the following categories: (1) Acute Interstitial Nephritis, (2) Acute Tubular Injury, (3) Diabetic Kidney Disease, (4) Hypertensive Kidney Disease, (5) other (including IgA nephropathy, FSGS, membranous nephropathy, fibrillary GN, and other kidney conditions), and (6) cannot be determined. In this study, categories (1) and (2) were classified as AKI, and categories (3) to (5) were classified as CKD. Further details of KPMP manuals of procedures regarding tissue acquisition, handling, recruitment sites, case adjudication, and processes for tissue access are referred to in the publicly available link: https://www.kpmp.org/for-researchers. Our initial cohort included consecutively received and therefore likely unbiased samples (n=40) obtained by Indiana University from the KPMP Central Biorepository, where all biospecimens are sent. Among these, individuals with an adjudicated category of “cannot be determined (n=4)” and unavailable (n=3) were excluded, and the remaining n=33 specimens were used for the analysis. Considering that there was a pronounced sex imbalance among the groups in the initial cohort, we added “additional cohort (n=9)” from the subsequent consecutively received samples, based on sex as an additional selection criterion: male samples for the reference and CKD groups, and female samples for the AKI group. Human urine was obtained from Biopsy Biobank Cohort of Indiana under the approval of IRB (Approval no. 1906572234).¹¹

Codetection by Indexing (CODEX/Phenocycler) Multiplexed Tissue Imaging

CODEX imaging using its most recent Phenocycler-Fusion 2.0 (PCF) platform was performed, as described previously,¹⁰ with modifications. In brief, 10-μm thick kidney tissue sections were cut from OCT blocks onto SuperFrost Gold+charged slides, followed by OCT removal, fixation with 1.6% paraformaldehyde, and incubation with an antibody solution. Oligonucleotide probe staining and automated tile-scan imaging of the tissue between probe staining rounds were performed on the PCF fluidics handler and microscope. Images were processed as part of the PCF imaging workflow and visualized using FIJI/ImageJ. The panel used for this study includes 42 markers as recently described based on organ mapping antibody panels.^10,12 Imaging data are publicly available in the KPMP atlas (www.KPMP.org).

Cast Identification and Characterization

Identification of casts was conducted based on Hematoxylin and Eosin (H&E) staining images under the supervision of two pathologists (J. Hata and A.Z. Rosenberg). Intratubular casts were defined as intratubular cylindrical materials, and we excluded detached cell clusters. Large casts were defined by casts with an area >5000 µm²_. Extrusion of cast materials into the interstitium¹³ was defined by the extratubular signal of UMOD adjacent to cast-containing tubules in CODEX images.

Immunoblotting of Human Urine

Human spot urine was collected, centrifuged at 2000× g for 10 minutes to remove cell debris, and stored in a −80°C freezer. For immunoblotting, the frozen urine was thawed at room temperature and was diluted twofold (2×) with distilled water and then denatured for 20 minutes at 60°C using twofold (2×) concentrated Sample Buffer (S3401, Sigma). Immunoblotting was conducted, as described previously.¹⁴ Rabbit anti-PROM1/CD133 antibody (1:2000, 18470-1-AP, Proteintech) was used as a primary antibody. Horseradish peroxidase-conjugated anti-IgG antibody (Proteintech) was used as a secondary antibody.

Statistics

Statistical significance was evaluated using an unpaired t test, Welch t test (volcano plots), or Fisher exact test between two conditions. One-way ANOVA with embedded comparisons between two individual groups was used among three conditions. The relationship between the two variables was evaluated using simple linear regression analysis. The coefficient of determination (R²) and the corresponding P value was calculated. We used a significance level of P < 0.05.

Results

CODEX Multiplex Imaging Analysis of Human Kidney Intratubular Casts

To understand the protein expression profile of tubular casts and associated tubules in human kidneys, we analyzed tissue sections imaged with CODEX followed by H&E staining from 42 individuals in total: healthy (n=11) and with kidney disease (AKI, n=13; CKD, n=18) from the KPMP consortium¹⁵ (Figure 1A). Participant demographics of the combined, initial, and additional cohorts are presented in Figure 1B and Supplemental Tables 1 and 2, respectively. We used H&E staining to detect intratubular casts and identified a total of 493 intratubular casts. We identified the protein expression profile within casts, surrounding tubules and interstitium using CODEX data. Since CODEX imaging and H&E staining were conducted on the same section, we were able to perform in situ analysis of protein expression within casts and the surrounding tubules. The cast number per specimen exhibited a variability, especially within CKD patients (Figure 1, C and D). The cast number, normalized per area, was trending higher in AKI patients and significantly higher in patients with CKD compared with the reference samples (Figure 1E).

Figure 1.

Study design. (A) We performed CODEX multiplex imaging and H&E staining on the same section of human kidney biopsy samples from the KPMP cohort. H&E staining images were used for cast identification and histologic analysis. CODEX images were used to evaluate the protein expression profile of tubular casts, surrounding tubules and interstitium. Scale bars=1 mm (left) and 50 μm (right). (B) Demographics of reference and disease kidney tissue specimens. Age was presented as mean±SD. (C) Cast number and area (mm²) of the specimens in each group. Data were presented as median (IQR). (D) Number of casts per specimen in each disease group. (E) Cast number for each specimen normalized by area. AQP2, aquaporin-2; CODEX, Codetection by indexing; DAPI, 4′,6-diamidino-2-phenylindole; H&E, Hematoxylin and Eosin; IQR, interquartile range; KPMP, Kidney Precision Medicine Project; LRP2, LDL receptor–related protein 2; NCC, sodium-chloride cotransporter; UMOD, uromodulin; αSMA, α-smooth muscle actin.

UMOD, PROM1, and Insulin-Like Growth Factor Binding Protein 7 Are Major Components of Tubular Casts

We first assessed the protein expression within intratubular casts (Figure 2A). UMOD, an already established major component of tubular casts,² was positive in 87% (427 of 493) of casts. Notably, PROM1, a stem cell marker recently shown to be associated with dedifferentiated states in kidney epithelial cells,^10,15,16 was detected in 89% (439 of 493) of casts. Insulin-like growth factor binding protein 7 (IGFBP7), a urinary secretory protein and a biomarker for AKI,¹⁷ was found in 39% (193 of 493) of intratubular casts. Importantly, kidney injury molecule-1 (KIM-1) and vascular cell adhesion protein-1, markers associated with injury of proximal tubule cells,^18,19 were rarely detected within intratubular casts (0.2% [1 of 493] and 2% [8 of 493], respectively). Both the initial and additional cohorts exhibited similar results (Supplemental Table 3). These results suggest that PROM1, UMOD, and IGFBP7 specifically adhere to or become incorporated into casts. Since the number of casts varied markedly among patients (Figure 1, C and D), patients with a large number of casts may have an effect on the overall positivity rate. Therefore, we also calculated the positivity rate of these components on a per-specimen basis (Supplemental Table 4) for samples that contained casts (total, n=35; reference, n=8; AKI, n=11; CKD, n=16 specimens). The results were consistent with those obtained from the per-cast analysis, suggesting that the observed trend was not driven by a subset of specimens but rather reflected a common pattern across all samples.

Figure 2.

PROM1, UMOD and IGFBP7 are major components of tubular casts. (A) Representative images of casts and each protein (top). Bars=50 μm. Percentage of casts positive for each protein among all casts (bottom). (B) Ratio of PROM1-, UMOD- and IGFBP7-positive casts in each specimen in each condition. The graphs with error bars were generated using the median and IQR. (C) Percent of cast subtype based on the combination of the presence or absence of PROM1, UMOD and IGFBP7 within casts in each condition. (D) Odds ratio of cast subtype (defined in [C]) for AKI versus CKD. (E) Odds ratio of cast subtype based on the presence or absence of UMOD and PROM1 for AKI versus CKD. IGFBP7, insulin-like growth factor binding protein 7; PROM1, Prominin-1; UMOD, uromodulin.

We next examined the positivity rate of these major components per specimen for each disease state (Figure 2B). PROM1 was present in all casts from reference, but its positivity rate was significantly lower in AKI and CKD specimens. UMOD was detected in nearly 100% casts in reference, but its positivity rate was reduced in CKD kidneys. The positivity of IGFBP7 in casts did not differ among conditions. These data suggest that the three major cast components behave independently in a condition-specific manner.

When casts were classified based on the presence or absence of these three major components, two major subtypes (UMOD-positive/PROM1-positive/IGFBP7-positive casts and UMOD-positive/PROM1-positive/IGFBP7-negative casts) were predominant. However, in AKI and CKD samples, additional subtypes became more prevalent, suggesting greater heterogeneity of cast composition (Figure 2C). We next reasoned that cast subtypes defined by the cast components could be associated with the stage of kidney injury (AKI or CKD). Two UMOD-positive/PROM-1 negative cast subtypes (UMOD-positive/PROM1-negative/IGFBP7-positive casts and UMOD-positive/PROM1-negative/IGFBP7-negative casts) were associated with higher odds of AKI rather than CKD (Figure 2D). We then examined whether a more simplified cast classification using UMOD and PROM1 could be used, since IGFBP7 does not appear to be a major discriminant. UMOD-positive/PROM1-negative casts were associated with higher odds of AKI rather than CKD, with an odds ratio of 4.61 (95% confidence interval [CI], 2.29 to 8.99; Figure 2E). Conversely, PROM1-positive/UMOD-negative casts were associate with lower odds of AKI, with an odds ratio of 0.25 (95% CI, 0.06 to 0.94; Figure 2E). We also analyzed at the specimen level, where specimens were counted as positive if the subtype constituted ≥25% of the total casts (Supplemental Figure 1). UMOD-positive/PROM1-negative casts and PROM1-positive/UMOD-negative casts showed similar trends to those observed in the per-cast analysis (Figure 2E), although they did not reach statistical significance, likely due to the limited sample size.

Tubular Casts Are Associated with Loss of Transporter Expression in Distal Nephron Segments

We next evaluated protein expression profile in cast-containing tubular epithelium to understand the localization and potential pathologic role of intratubular casts. Our CODEX panel includes markers for each nephron segment (Figure 3A). We first assessed LDL receptor–related protein 2 (proximal tubule marker) and E-cadherin (distal nephron segment marker) expression in cast-containing tubules. Most cast-containing tubules were E-cadherin positive (Figure 3B and Supplemental Figure 2A), consistent with previous reports where intratubular casts were predominantly detected in the distal nephron segment.^2,4 We then focused on E-cadherin–positive (distal nephron segment) tubules and assessed the detailed segmental markers: UMOD for TAL, NCC for DCT, and AQP2 for CD (Figure 3A). In the absence of casts, 92% (982 of 1062) of E-cadherin–positive tubules expressed at least one of the segment-specific markers UMOD, NCC, or AQP2. Strikingly, only 13% (62 of 473) of E-cadherin–positive tubules expressed any of these segment-specific markers in the presence of intratubular casts (Figure 3C; P < 0.001, Fisher exact test). The results were consistent even when the initial cohort and the additional cohort were analyzed separately (Supplemental Figure 2B). In addition, when the positive rates of each marker were calculated per patient, the results were similar; UMOD/NCC/AQP2 triple-negative tubules were significantly more frequent among tubules containing casts (P < 0.001, unpaired t test; Supplemental Table 5).

Figure 3.

Tubular casts are associated with loss of transporter expression in distal nephron segments. (A) Markers for each nephron segment used in this study. (B) Distribution of LRP2 and E-cadherin expression in tubular epithelium surrounding casts. (C) Qualitative analysis of UMOD, NCC, and AQP2 expression in E-cadherin-positive tubules with or without casts. Bars=50 μm. (D) Mean fluorescence intensity of NCC in tubules with and without casts at the specimen-level (based on the mean value per specimen). Bars=50 μm. n=10 specimens. (E) Mean fluorescence intensity of AQP2 in tubules with and without casts at the specimen-level (based on the mean value per specimen). Bars=50 μm. n=9 specimens. (F) Mean fluorescence intensity of UMOD in tubules with and without casts at the specimen-level (based on the mean value per specimen). Bars=50 μm. n=3 specimens. (G) Factors associated with the loss of NCC and AQP2 expression in tubules surrounding casts. CD, collecting duct; DCT, distal convoluted tubule; NCC, sodium-chloride cotransporter; PT, proximal tubule; TAL, thick ascending limb.

We then focused on TAL, DCT, and CD cells, which still express UMOD, NCC, or AQP2, respectively, even in the presence of intratubular casts. Expression levels of these proteins in the tubular epithelium were compared with adjacent TAL, DCT, or CD epithelium without casts across the entire biopsies. Since each specimen contained multiple kidney tubules, we evaluated the data using two complementary approaches¹: a specimen-level analysis based on the mean value per specimen (Figures 3, D–F) and² a tubule-level analysis comparing all individual tubules (Supplemental Figure 2C). In both analyses, we found that the NCC and AQP2 expression in DCT and CD cells were lower in the presence of casts, whereas UMOD expression in TAL epithelium was not affected by casts. These data suggest that intratubular casts are associated with suppression of NCC and AQP2 expression in DCT and CD cells. Therefore, the UMOD/NCC/AQP2-negative tubules surrounding casts (Figure 3C, pie chart on the right) are likely to be explained by DCT or CD cells which have lost NCC or AQP2 expression. This is consistent with the previous findings that intratubular casts are mainly observed in the lumen of DCT and CD cells.^1,2

We next aimed to understand the factors that lead to the loss of transporters in the cast-surrounding tubules in DCT and CD. To this end, we compared the characteristics of NCC/AQP2-negative and NCC/AQP2-positive tubules surrounding casts. UMOD-positive tubules were excluded from the analysis to focus on the DCT and CD epithelium. Interestingly, the odds ratio for observing NCC/AQP2-negative tubules was significantly altered by the protein components of casts but not by disease states, cast size, flattened epithelia, extrusion of casts, or adjacent inflammation (Figure 3G). Among the protein components, PROM1 had the strongest association with NCC/AQP2 loss in the surrounding tubules, with an odds ratio of 26.74 (95% CI, 13.08 to 54.68). The effects of intracast presence of PROM1 on the odds ratios were consistent between the initial and additional cohorts (Supplemental Figure 3, A and B).

PROM1 Is a Urinary Secretory Protein and a Potential Biomarker in Kidney Disease

PROM1 is a transmembrane glycoprotein expressed in kidney epithelial cells, but its extracellular secretion has not been well characterized in the literature. Immunoblotting of human urine confirmed its presence in the urine (Figure 4A). The observed mol wt of PROM1 was approximately 115 kDa, suggesting the excretion of full-length or near-full-length protein, rather than fragments. We assessed PROM1 expression in publicly available urinary proteomics datasets of the KPMP consortium (demographics are shown in Figure 4B). Note that the values from proteomics analysis indicate relative quantification based on a fixed amount of total urinary protein. In this dataset, PROM1 was detected in the urine, and was significantly higher in AKI patients (P = 0.01), but not in CKD patients (P = 0.12), compared with healthy reference (Figure 4C). Urinary KIM-1 and neutrophil gelatinase-associated lipocalin (NGAL), established biomarkers for AKI,^20,21 were also higher in AKI patients, supporting the fidelity of the proteomics analysis (Figure 4C). Urinary UMOD was lower in both AKI and CKD patients, consistent with previous observations.²² Urinary IGFBP7 was significantly lower in CKD patients. To understand how these proteins fit within the proteomic landscape in kidney disease, we compared the whole urinary protein expression in the setting of (1) AKI versus healthy reference (Figure 4D and Supplemental Table 6) and (2) CKD versus healthy reference (Figure 4E and Supplemental Table 7). In the comparison between AKI and reference groups, PROM1 showed a smaller mean log₂ fold change but a higher −log₁₀ P value than KIM-1 and NGAL. In the comparison between CKD and reference groups, PROM1, KIM-1, and NGAL all showed smaller changes.

Figure 4.

PROM1 is a urinary secretory protein and a potential biomarker in kidney disease. (A) Immunoblotting of PROM1 in human urine. (B) Demographics of reference and kidney disease individuals used for proteomics analysis. (C) Normalized urinary PROM1, KIM-1, NGAL, UMOD and IGFBP7 levels, measured by proteomics analysis. Proteomics (SomaScan) analysis is from publicly available data generated by the KPMP. Accessed Sep 18th, 2024. https://www.kpmp.org. n=18–73 per group. (D and E) Volcano plots, generated from the proteomics dataset, illustrate differential urinary protein abundance between healthy reference and AKI patients (D), and between healthy reference and CKD patients (E). The top three significant proteins with greatest and lowest expression, as well as the protein presented in Figure 4C, are highlighted. KIM-1, kidney injury molecule-1; NGAL, neutrophil gelatinase-associated lipocalin.

We next examined the association between PROM1 positivity rate within casts and urinary PROM1 concentration in 23 individuals for whom both CODEX imaging data and urinary proteomics data were available (Supplemental Figure 4). No significant correlation was observed between the two factors. UMOD showed a similar result as PROM1. IGFBP7 showed a very weak correlation but did not reach statistical significance.

Discussion

Kidney tubular casts are frequently observed in the kidneys and have long been recognized as a sign of kidney disease, yet their protein composition and pathologic roles have been largely unknown. In this study, we took advantage of CODEX multiplex protein imaging to investigate the intratubular cast components and associated tubular alterations in human kidneys. CODEX multiplex imaging and H&E staining were performed on the same section, enabling precise in situ analysis of protein expression in the casts and surrounding tubules.

We identified PROM1 as a novel major component of intratubular casts. PROM1 is a transmembrane glycoprotein known as a stem cell marker.²³ PROM1 is also highly expressed in kidney epithelial cells.²⁴ In AKI, proximal tubular cells undergo dedifferentiation, characterized by the upregulation of injury markers and the acquisition of progenitor-like properties that facilitate proliferation and repair.¹⁶ PROM1 is one of the representative markers of these dedifferentiated tubular cells,^10,15,16 and has been shown to promote proliferation of kidney epithelial cells after injury,²⁵ suggesting a role in regeneration. In chronic settings, however, persistent dedifferentiation without redifferentiation results in maladaptive repair. A recent study has shown that elevated PROM1 expression is associated with disease progression in diabetic kidney disease, supporting that sustained PROM1 upregulation may contribute to the pathogenesis of CKD.²⁶ KIM-1 and vascular cell adhesion protein-1 are other tubular proteins associated with injury, but they were rarely detected within casts. Therefore, the presence of PROM1 within casts is likely specific. The differential incorporation of proteins into casts may be explained by the intrinsic biochemical properties of each protein, such as their tendency to aggregate or bind to other urinary components. Intra-cast PROM1 presence showed the strongest correlation with the loss of transporters in the epithelium of distal nephron segments. The role of PROM1 within casts needs further investigation. We also showed that PROM1 is secreted extracellularly into the urine as a full-length or near-full-length protein. Urinary PROM1 levels showed a significant statistical association with AKI in comparison between three groups and even higher −log₁₀ P values compared with KIM-1 and NGAL in the comparison between the AKI and reference groups, suggesting its potential as a biomarker. To evaluate the potential of PROM1 as an AKI biomarker, it will be necessary to analyze a larger number of clinical specimens, incorporating AKI severity and time-course data, as well as the validation using complementary modalities such as ELISA. Note that the difference in PROM1 level between AKI and CKD patients was relatively small as a discriminative measurement. Therefore, the clinical utility of urinary PROM1 to distinguish between AKI and CKD may be limited to use in combination with other biomarkers or clinical parameters. The mechanism governing the extracellular secretion of PROM1 in healthy and diseased conditions also needs to be investigated. Interestingly, urinary PROM1 concentration was higher in AKI patients, while PROM1-negative casts were observed exclusively in patients with AKI or CKD patients, but not in the reference individuals. Moreover, there was no association between urinary PROM1 levels and PROM1 positivity rate within casts. These data suggest that a higher urinary concentration of a protein does not necessarily correspond to greater cast positivity, possibly because incorporation into casts can reduce its free urinary fraction, thereby obscuring the apparent relationship between the two parameters. Therefore, evaluating PROM1 expression within kidney casts, in addition to measuring urinary PROM1 levels, will provide complementary insights into kidney disease.

UMOD is a secretory protein in the urine and has been established as a primary constituent of hyaline casts. While UMOD plays protective roles in the urinary space against stone formation and infection,^6,27 its role in intratubular/urinary casts remains unknown. We have reported that UMOD knockout mice still generate intratubular casts after injury, indicating that UMOD is not necessary for cast formation in mice.²⁸ The presence of UMOD-negative intratubular casts was also reported in a mouse model with distal tubule injury.²⁹ Here, we described the existence of UMOD-negative casts (13%) in human kidneys. This further supports that UMOD is not indispensable for cast formation. This study also showed that UMOD-negative casts are significantly enriched in CKD kidneys. It has been reported that all urinary casts were UMOD-positive in healthy and diseased individuals.³⁰ Therefore, UMOD-negative casts might be more likely to be retained within the tubules rather than being excreted into the urine. This raises the hypothesis that UMOD may have a role in preventing tubular retention of casts, which could be explained by its negative charge³¹ or flexibility of its polymeric form.² A side-by-side comparison of the protein expression profiles between intratubular casts and urinary excreted casts will help to understand the protein component that may facilitate casts to pass through tubules into the urine. This study suggests an exploratory hypothesis that the protein composition of casts may differ depending on the underlying disease states. Further study is needed to determine if the cast subtypes defined by the presence or absence of PROM1 and UMOD serve to distinguish AKI and CKD.

Our study provides an insight into the potential pathologic role of intratubular casts. The presence of intratubular casts was associated with a substantial loss of NCC and AQP2 expression in the surrounding tubules. These data suggest that intratubular casts may contribute to injury in the distal nephron segments. Tubules occluded by casts, which have reduced luminal flow, may undergo dedifferentiation, resulting in lower transporter expression. We cannot exclude the possibility that loss of NCC and AQP2 precedes the tubular cast formation. That is, injured tubules could be the cause of cast retention rather than the result of it. Interestingly, protein composition within intratubular casts, but not disease state, cast size, tubular atrophy or adjacent inflammation, showed the strongest correlation with loss of transporters in the distal nephron segments. How the protein composition within casts is associated with the potential distal nephron injury needs to be determined. It is possible that each protein within casts may play a role by affecting the flexibility or charge of casts.

Urinary IGFBP7 did not increase in AKI samples compared with healthy references, although IGFBP7 has been reported as a biomarker of early AKI. This discrepancy may be explained by three factors. First, urine samples in the KPMP cohort were collected at the time of kidney biopsy and therefore, the timing of sampling relative to the onset of AKI may slightly differ from other cohorts. Second, the SomaScan presents relative protein abundance on the fixed amount of urinary protein. Therefore, the signal of IGFBP7 may relatively decrease when other proteins are more strongly upregulated. Third, potential structural modifications of IGFBP7 during disease state may lead to discrepancies between aptamer-based (SomaScan) and antibody-based assays (e.g., NephroCheck and ELISA). Importantly, while IGFBP7 showed an unexpected trend, the direction of urinary levels of KIM-1, NGAL, and UMOD were consistent with previous findings, supporting the overall robustness of the SomaScan measurements.

Our study had several notable limitations. First, in our workflow, histologic evaluation is limited to H&E staining images on fresh-frozen sections, which does not allow for further classification of intratubular casts (hyaline, granular, and waxy casts) and specific casts with established pathologic significance, including light chain casts.⁷ Applying CODEX to formalin-fixed paraffin-embedded specimens along with H&E or periodic acid–Schiff staining, and integrating additional histologic modalities, will enable detailed histologic cast classification, which can be associated with distinct protein composition within the cast. In addition, a machine learning-based method to annotate intratubular casts using Masson's trichrome-stained images has been recently reported.¹ Combining these automated methods will enable more high-throughput analysis. Second, urinary casts in the urine samples were not assessed in this study. A side-by-side comparison of the counts and protein components in urinary and intratubular casts will provide a better understanding of the molecular mechanism underlying cast plugging within tubules. Third, the expression profiles of molecules other than UMOD, NCC, and AQP2 in cells surrounding the casts remain unexplored. Incorporating additional markers (e.g., apoptosis) to the CODEX panel, as well as spatial transcriptomic analysis will provide the gene expression landscape to define the cell state in cast-containing tubular epithelium. Forth, the sample size for cast analysis was relatively small, and the demographics, including age, sex, and diabetes status, were not fully matched across disease states, even after including the additional cohort. Therefore, cast number and composition analysis comparing disease groups is exploratory rather than conclusive. To determine the disease-related differences, a larger sample size with marched demographics, will be crucial.

In conclusion, this study suggests that protein composition signature within casts varies with disease state and is associated with loss of transporter expression in distal nephron segments. Our work provides a foundation for interpreting kidney pathology at the molecular level through advanced spatial analysis performed on the same histologic sections.

Supplementary Material

cjasn-21-822-s001.pdf ^{(2.1MB, pdf)}

cjasn-21-822-s002.pdf ^{(894KB, pdf)}

cjasn-21-822-s003.xlsx ^{(257.8KB, xlsx)}

cjasn-21-822-s004.xlsx ^{(260KB, xlsx)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C611.

Author Contributions

Conceptualization: Tarek M. El-Achkar, Azuma Nanamatsu, Avi Z. Rosenberg.

Data curation: Daria Barwinska, William S. Bowen, Michael Ferkowicz, Azuma Nanamatsu, Angela R. Sabo.

Formal analysis: Tarek M. El-Achkar, Azuma Nanamatsu, Avi Z. Rosenberg.

Funding acquisition: Pierre C. Dagher, Michael T. Eadon, Tarek M. El-Achkar, Azuma Nanamatsu.

Investigation: Tarek M. El-Achkar, Jessica Hata, Azuma Nanamatsu.

Methodology: Daria Barwinska, William S. Bowen, Tarek M. El-Achkar, Jessica Hata, Azuma Nanamatsu, Avi Z. Rosenberg, Angela R. Sabo.

Project administration: Pierre C. Dagher, Michael T. Eadon, Tarek M. El-Achkar, Azuma Nanamatsu.

Resources: Michael T. Eadon, Tarek M. El-Achkar.

Software: Tarek M. El-Achkar.

Supervision: Tarek M. El-Achkar.

Validation: Tarek M. El-Achkar, Azuma Nanamatsu, Angela R. Sabo.

Visualization: Daria Barwinska, Tarek M. El-Achkar, Azuma Nanamatsu, Angela R. Sabo.

Writing – original draft: Azuma Nanamatsu.

Writing – review & editing: Daria Barwinska, Pierre C. Dagher, Michael T. Eadon, Tarek M. El-Achkar, Takashi Hato, Avi Z. Rosenberg, Angela R. Sabo.

Funding

T.M. El-Achkar: National Institute of Diabetes and Digestive and Kidney Diseases (R01DK111651), Office of Academic Affiliations, Department of Veterans Affairs (5I01BX003935), Dialysis Clinic Inc. A. Nanamatsu: Takeda Science Foundation. This study was supported by National Institute of Diabetes and Digestive and Kidney Diseases (U01DK133081, U01DK133091, U01DK133092, U01DK133093, U01DK133095, U01DK133097, U01DK114866, U01DK114908, U01DK133090, U01DK133113, U01DK133766, U01DK133768, U01DK114907, U01DK114920, U01DK114923, U01DK114933, U24DK114886, UH3DK114926, UH3DK114861, UH3DK114915, and UH3DK114937).

Declarative Statements

This study includes clinical experimentation and received Institutional Review Board or Ethics Committee approval. All patients provided written informed consent. This study includes clinical experimentation and complies with the Declaration of Helsinki.

Data Availability Statements

Original data generated for the study are or will be made available in a public access repository upon publication. Data Type: Image Data; Raw Data/Source Data. Repository Name: KPMP. Linkable Citation: www.KPMP.org. De-identified participant-level data, including CODEX multiplex imaging data and urinary proteomics data, are publicly available in the KPMP website (www.KPMP.org).

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/C612, http://links.lww.com/CJN/C613 and http://links.lww.com/CJN/C614.

Supplemental Table 1. Demographics of reference and disease kidney tissue specimens in the initial cohort. Age was presented as mean±SD.

Supplemental Table 2. Demographics of reference and disease kidney tissue specimens in the additional cohort. Age was presented as mean±SD.

Supplemental Table 3. Protein component in casts by cohort. Percentage of casts positive for each protein among all casts. Data were presented as median n (%).

Supplemental Table 4. Protein component in casts per patient. Data were presented as median [interquartile range [IQR]].

Supplemental Table 5. Qualitative analysis of UMOD, NCC, and AQP2 expression in E-cadherin-positive tubules with or without casts per specimen. Data were presented as mean±SD.

Supplemental Table 6 (Excel sheet). Differential urinary protein abundance between healthy reference and AKI patients. Group comparisons were performed using Welch's t test, followed by adjustment with the Benjamini–Hochberg method.

Supplemental Table 7 (Excel sheet). Differential urinary protein abundance between healthy reference and CKD patients. Group comparisons were performed using the Welch t-test, followed by adjustment with the Benjamini-Hochberg method.

Supplemental Figure 1. Odds ratio of two cast subtypes (UMOD-positive/PROM1-negative casts and UMOD-negative/PROM1-positive casts) for AKI versus CKD at the specimen level.

Supplemental Figure 2. (A) Distribution of LDL receptor–related protein 2 and E-cadherin expression in tubular epithelium surrounding casts in the initial and additional cohorts.

Supplemental Figure 3. (A) Factors associated with the loss of NCC and AQP2 expression in tubules surrounding casts in the initial cohort.

Supplemental Figure 4. Association between urinary PROM1/UMOD/IGFBP7 concentration and PROM1/UMOD/IGFBP7-positive ratio of tubular casts.

Supplemental Methods.

List of the KPMP collaborators.