Deconvolution of Human Urine across the Transcriptome and Metabolome

Abstract

Background

Early detection of the cell type changes underlying several genitourinary tract diseases largely remains an unmet clinical need, where existing assays, if available, lack the cellular resolution afforded by an invasive biopsy. While messenger RNA in urine could reflect the dynamic signal that facilitates early detection, current measurements primarily detect single genes and thus do not reflect the entire transcriptome and the underlying contributions of cell type-specific RNA.

Methods

We isolated and sequenced the cell-free RNA (cfRNA) and sediment RNA from human urine samples (n = 6 healthy controls and n = 12 kidney stone patients) and measured the urine metabolome. We analyzed the resulting urine transcriptomes by deconvolving the noninvasively measurable cell type contributions and comparing to plasma cfRNA and the measured urine metabolome.

Results

Urine transcriptome cell type deconvolution primarily yielded relative fractional contributions from genitourinary tract cell types in addition to cell types from high-turnover solid tissues beyond the genitourinary tract. Comparison to plasma cfRNA yielded enrichment of metabolic pathways and a distinct cell type spectrum. Integration of urine transcriptomic and metabolomic measurements yielded enrichment for metabolic pathways involved in amino acid metabolism and overlapped with metabolic subsystems associated with proximal tubule function.

Conclusions

Noninvasive whole transcriptome measurements of human urine cfRNA and sediment RNA reflects signal from hard-to-biopsy tissues exhibiting low representation in blood plasma cfRNA liquid biopsy at cell type resolution and are enriched in signal from metabolic pathways measurable in the urine metabolome.

Introduction

Distinct cell type changes underlie many genitourinary disease etiologies (1). While current clinical urine assays can help guide diagnostic differentials, they lack high-resolution insights and diagnosis oftentimes requires an invasive biopsy. For example, renal tubular epithelia and prostate epithelia are regarded as the respective cell of origin for renal carcinomas and prostate adenocarcinomas. While these cancers and others generally exhibit favorable prognoses with early detection (2, 3), an invasive biopsy is required for definitive diagnosis. These procedures are not risk-free for the patient (4, 5), and gold standard screening tools are limited in their capacity for early detection (6, 7). Cell-free RNA (cfRNA) liquid biopsy measuring messenger RNA in the blood can reflect dynamic tissue (8) or cell type-specific (9) gene expression changes, and offers promise for the early diagnosis or monitoring of a broad array of physiological changes and diseases (8). However, genitourinary tissues (8) and cell types (9) exhibit low abundance in blood-based cfRNA measurements, limiting the sensitivity to capture early-stage disease changes.

Urine possesses both cellular (10) and cell-free nucleic acids (11) and directly interfaces with genitourinary tissues, rendering it a promising biofluid to reflect cell type-specific signal (12). However, existing measurements measure small RNA (13) or primarily quantify gene panels (11, 12), which do not enable full determination of its cell type origins. Practical challenges associated with urine (e.g., solute variability (14, 15), bacterial growth, and leukocytes/hematuria (10, 15, 16)) have impacted the measurement of the urine cell-free transcriptome (cf-transcriptome) (11) and current studies remain limited to urine sediment (10, 16). Moreover, urine contains a variety of metabolic products, including by-products from cellular metabolism (17). As urine interfaces with distinct tissues from blood and comprises its chemical waste products, we hypothesized that measuring the transcriptome of human urine could provide a noninvasive window into its contributing cell types.

In this work, we measure the transcriptomes of urine cfRNA and sediment RNA alongside the urine metabolome. We first report cell type repertoire of human urine across its cell-free and sediment transcriptomes. Next, we compare our findings to plasma cfRNA and demonstrate striking enrichment for genitourinary cell types and metabolic signal in the urine transcriptome. Finally, we integrate transcriptomic and metabolomic measurements, observing metabolites and genes enriched in the same pathways and that overlap with metabolic functions of the renal proximal tubule.

Materials and Methods

Sample Collection and Preprocessing

Clean catch urine specimens were collected from kidney stone patients (n = 12) and healthy controls without known kidney disease (n = 6) with Stanford Institutional Review Board approval. Voided specimens were stored at +4 °C until processing; samples were processed within 6 h of collection. Whole urine was aliquoted for metabolomic analysis, and a standard urine dipstick (Fisherbrand) was performed. The remaining sample was spun at 4°C and 3000g for 30 min. Urine sediment RNA was prepared as previously described (18): 0.1% v/v betamercaptoethanol (Millipore) and 1 mL Trizol (Ambion) were added to the pellet following centrifugation and frozen at −80°C. Spot creatinine was measured using an aliquot of frozen urine (Biotechne).

RNA Isolation, Library Preparation, and Sequencing

Urine sediment RNA isolation was performed as previously described (18). Frozen Trizol pellets thawed on ice were isolated with the RNeasy Mini Kit (Qiagen). Urine supernatant was thawed at room temperature, and cfRNA was isolated from 2-8 mL urine with the QIAamp Circulating Nucleic Acid Kit (Qiagen). Isolated RNA was treated with Baseline-ZERO DNase (Lucigen), cleaned and concentrated (Zymo), and quantified (Agilent RNA 6000 Pico Kit). All libraries were prepared using the SMARTer Stranded Total RNaseq Kit v2 and barcoded (Takara). Urine samples were sequenced with either the 2 × 150 bp or 2 × 75 bp configuration (Supplemental Fig. 1) to a mean depth of 30 M reads per sample on an Illumina NextSeq.

Metabolomics Instrumentation

Sample preparation is described in the Supplemental Materials. Metabolites were measured using a hydrophilic interaction liquid chromatography column coupled to a Thermo Q Exactive HF Hybrid Quadrupole-Orbitrap mass spectrometer. High resolution full scan mass spectrometry and data-dependent tandem mass spectrometry (MS²) were collected in both positive and negative mode ionization (separate injections). MS² selection was prioritized for metabolites with known mass to charge ratios and LC-MS/MS method specific retention times using an inclusion list from an in-house library generated from authentic standards. Additional MS² data were collected in a data-dependent manner when scan bandwidth was available. Mass range was 60 to 900 m/z. High resolution full scan mass spectrometry resolution was 60k; MS², 15k. Data was collected in centroid mode, loop count was 4, and quadrupole isolation window was 1.0 Da.

Data Preprocessing

RNA: Reads were trimmed (trimmomatic v.0.36) and mapped (STAR v.2.7.3a) to the human reference genome (hg38). Reads were deduplicated (MarkDuplicates v.4.1.1) and counted (htseq-count v.0.11.1). Read statistics were estimated (FastQC v.0.11.8). Quality parameters for 3’ bias, intron to exon ratio, and ribosomal read fraction were computed as previously described (19) and rounded to the nearest tenth. Associated thresholds for these parameters (19) were then applied; the ribosomal read fraction threshold was 0.5 given that samples were sequenced to sufficient depth and were not principal component analysis outliers. Passing samples were used in subsequent analyses (Supplemental Table 1).

Metabolomics: data were processed using MS-Dial (v.4.60) (20) for peak picking, alignment, and annotation. Metabolites were annotated by accurate mass and MS² spectral matching. When a metabolite’s method-specific retention time was known, a retention time matching ±0.4 min was required. MS² reference spectra were obtained from MassBank of North America (https://mona.fiehnlab.ucdavis.edu/) or in-house analysis of authentic standards (21). MS1 accurate mass tolerance was 0.01 Da; MS² tolerance was 0.015 Da. All annotations were manually reviewed (Supplemental Materials).

Cell Type Deconvolution

Deconvolution of cell type-specific RNA was performed as previously described (9). Sediment RNA sample N4 was excluded from subsequent comparisons given the outlier deconvolved fraction of keratinocyte cell type-specific RNA. Deconvolved relative fractions of cell type-specific RNA in urine were compared to deconvolved plasma cfRNA [BioIVT cohort (22)] with a 2-sided Mann–Whitney U-test; fractions less than 0.1% were set to 0. Associated deconvolved fractional confidence intervals were determined (Supplemental Materials). As with the differential expression analyses, a given cell type comparison was made only if the proportion of nonzero coefficients was greater than or equal to the smallest relative proportion of samples of a given type. Multiple hypothesis (MH) correction was performed using the Benjamini–Hochberg procedure with a minimum significance cutoff level of 0.05.

Cell Type Gene Profiles and Signature Scoring

Following cell type gene profile derivation (SI), the cell type signature score for the j^th sample is given by:

Where N is the number of measured genes in the cell type gene profile and $g_{ij}$ is the log-transformed, counts per million and trimmed mean of M values normalized gene expression values. Counts tables used for signature scoring were filtered (Supplemental Materials). Bladder urothelial cell signature scoring in bladder cancer was performed with data from Sin et al. (Supplemental Materials).

RNA Sequencing Differential Expression

Differential expression was performed using limma (v.3.52.2) and edgeR (v.3.38.4) to compare (a) urine sediment RNA vs urine cfRNA and (b) either urine transcriptome to the plasma cf-transcriptome. MH correction was performed using the Benjamini–Hochberg procedure with a minimum log2 fold change of 0 and a minimum significance cutoff level of 0.05. Of the genes passing MH correction, 95% CIs were bootstrapped to estimate the log fold change. Further information on gene filtering, differential expression procedure, and CI bootstrapping are described in the Supplemental Materials.

Pathway and Cell Type Enrichment Analysis

Pathway and cell type enrichment analyses were performed using a hypergeometric test. Pathway enrichment analyses were restricted to the Kyoto Encyclopedia of Genes and Genomes to compare the same pathway between the transcriptome and metabolome. For all tests within a given enrichment analysis, MH correction was performed across all raw P values using a Benjamini–Hochberg test with alpha = 0.05 (statsmodels v.0.10.1) (Supplemental Materials).

Metabolomics Analysis

Chemical taxonomy on identified metabolite InChiKeys were determined using ClassyFire (v.1.0). The Human Metabolic Atlas (23) (v.3.1) was used to map the metabolite reactants and products or a given gene to a given reaction and subsequent pathway (SI).

Results

Cell Type Deconvolution of the Urine Transcriptome

We measured the urine transcriptome (n = 18 patients) (Fig. 1A) by isolating and sequencing RNA from the sediment and cell-free (supernatant) fractions of urine specimens (Supplemental Fig. 1) from a cohort of patients with kidney stones and healthy controls (n = 12, n = 6 respectively, Supplemental Table 1). We then deconvolved these bulk measurements into their relative fractional contributions of cell type-specific RNA (9) using Tabula Sapiens version 1.0 (TSP) (24), a multidonor whole-body cell atlas spanning 24 tissues and organs (Fig. 1B, Supplemental Figs. 2A and 3). In urine sediment RNA from healthy male controls, we observed large fractions of cell type-specific RNA from kidney epithelia and prostate epithelia, in addition to keratinocytes and immune cell types (Fig. 1B). Urine directly interfaces with the kidney and prostate (in males), consistent with the observed large relative fractional contributions. Contributions from bladder urothelial cells and Schwann cells were observed, highlighting the ability to measure the whole spectrum of genitourinary tract cell types. In urine cfRNA from healthy male controls, we observed elevated contributions of cell type-specific RNA from prostate and secretory cell types (Supplemental Fig. 3A). We additionally observed small contributions of nongenitourinary tract cell type-specific RNA throughout the body, including intrahepatic cholangiocytes, pericytes, respiratory cell types, and intestinal cell types.

We were struck by the relative fractional contributions of intestinal cell type-specific RNA. We therefore investigated the expression of intestinal cell type marker genes ANPEP and MUC2 (25). ANPEP encodes alanyl aminopeptidase, which facilitates peptide breakdown in the small intestine (26) and exhibits the highest expression in the intestine across all tissues in the human body despite being somewhat expressed in renal villi (27). MUC2 encodes a secretory mucin protein that is a primary constituent of gastrointestinal mucus and exhibits expression unique to the intestine (28). We observed elevated expression of both marker genes in urine sediment RNA relative to urine cfRNA (Fig. 1C).

We then independently assessed the relative abundance of genitourinary cell type-specific signal across both urine transcriptomes. Using signature scoring, we observed elevated luminal prostate epithelia signature in urine cfRNA relative to urine sediment RNA and no significant difference in the bladder urothelial and the proximal tubule signatures, suggesting that either urine fraction is conducive to measuring these cell types (Fig. 1D). Intersecting the differentially upregulated genes in urine cfRNA relative to urine sediment RNA with the cell type-specific gene profiles yielded significant enrichment for luminal prostate epithelial cells (P = 4.92 * 10⁻⁷, hypergeometric test), consistent with the observed elevated deconvolved fractions of cell-type specific RNA from this cell type.

We then asked whether cellular pathophysiology in bladder cancer was resolvable in urine, where bladder urothelial cells are the primary cell type of origin for bladder cancer and can abnormally proliferate, resulting in the development of tumors that left unchecked can become invasive (2). We defined an associated cell type-specific gene profile (9) and applied it to published bladder cancer sediment RNA data (10). We observed an increase in the bladder urothelial cell signature score across bladder cancer patients (Fig. 1E), highlighting the potential for urine transcriptomic measurements to reflect cell type-specific changes in genitourinary tissues.

Urine Sediment RNA And cfRNA Are Distinct From Plasma cfRNA

Urine and blood interface with distinct tissues and could potentially reflect distinct cell type-specific signal. We therefore compared the cell type spectra of the urine cell-free and sediment transcriptomes to the plasma cf-transcriptome(Fig. 2, Supplemental Fig. 2B). In plasma cfRNA, we observed elevated relative fractions of cell type-specific RNA originating from hematopoietic cell types, including erythrocytes and platelets. In either urine transcriptome, we observed enrichment for genitourinary cell type-specific RNA, including prostate epithelial cell types and kidney epithelia, and secretory cell types and ciliated cells in urine cfRNA. We observed nongenitourinary contributions of cell type-specific RNA including intestinal enterocytes enriched in both urine fractions and endothelial cells enriched in plasma cfRNA, highlighting the distinct cell type landscapes of the urine transcriptome relative to the plasma cf-transcriptome.

A few nongenitourinary cell types significantly enriched in the urine transcriptome relative to plasma cfRNA could reflect practical challenges with measuring urine. In urine sediment RNA, keratinocytes were enriched and variable in relative fractional contribution (Fig. 3A), consistent with the clinical observation that these cells are a source of contamination in conventional urine cytology (29). Several genitourinary diseases are often associated with inflammation, which is oftentimes reflected by a positive dipstick urinalysis for leukocyte esterase, which is released by neutrophils (30). A subset of patients with kidney stones were dipstick positive for leukocyte esterase. Across these urine specimens, we observed large deconvolved fractions of neutrophil-specific RNA in the sediment transcriptome and increased recovery of nonimmune cell type-specific signal in the urine cfRNA (Fig. 3A). Comparison of bulk-level differentially expressed genes upheld these observations, where sediment RNA from leukocyte dipstick-positive urine specimens was enriched for Gene Ontology terms reflecting stimulus responses and immune system processes compared to the corresponding cf-transcriptome (Supplemental Fig. 4A). These observations suggest that urine cfRNA is more robust than urine sediment RNA to noninvasively detect genitourinary cell types and tissues even in patients with inflammation, and that different urine fractions may be more informative to measure depending on the disease and tissue of interest.

We observed that the deconvolved cell type spectra in urine exhibited higher sample-to-sample heterogeneity relative to plasma cfRNA (Fig. 3B). This variability is consistent with challenges of solute concentration in urine due to factors such as patient hydration and voided volume (15). We controlled for solute variability and determined the differentially expressed genes in either urine transcriptome relative to plasma cfRNA (see Materials and Methods). We intersected the differentially upregulated genes in either urine fraction relative to plasma, again observing significant enrichment for bladder urothelial cells and proximal tubules as well as prostate luminal epithelia in urine cfRNA (Supplemental Table 2). Taken together, our observation of consistently elevated signal from these cell types in the urine transcriptome relative to plasma cfRNA highlights the enriched signal from hard-to-biopsy genitourinary tissues exhibiting low abundance in plasma cfRNA.

Comparison of the bulk urinary cell-free and sediment and transcriptomes with the plasma cf-transcriptome yielded striking differences in gene expression patterns. We observed a bimodal distribution of the median log2-fold change; both peaks centered around an absolute value of 2 (Supplemental Fig. 4B–D). Pathway enrichment on differentially upregulated genes in either urine fraction relative to plasma exhibited striking enrichment for metabolic activity (Fig. 3C), further highlighting the distinct transcriptional landscape of urine from plasma.

Comparison Across the Urine Transcriptomes and the Untargeted Metabolome

We next examined the metabolite landscape of human urine in conjunction with our RNA measurements. Untargeted LC-MS/MS-based metabolomics yielded a broad chemical spectrum from several distinct small molecule classes (Fig. 4A). We observed predominant relative contributions of amino acids and derivatives, as well as nucleic acids, carbohydrates, fatty acids and conjugates, carnitines, and exogenous compounds, consistent with prior characterization of the human urinary metabolomic landscape (17).

Pathway enrichment analysis on the measured polar urine metabolites included various amino acid metabolic pathways, purine metabolism, and vitamin B6 metabolism (Supplemental Fig. 4E). Comparison of the Kyoto Encyclopedia of Genes and Genomes-enriched pathways across both measurements yielded joint enrichment for different amino acid metabolic pathways in addition to glyoxylate and dicarboxylate metabolism and beta-alanine metabolism, demonstrating the ability to noninvasively measure genes and metabolites enriched in the same pathways (Fig. 4B).

Upon jointly measuring metabolites and transcripts enriched in the same pathways in urine, we sought to determine associated metabolic subsystems that overlap the proximal tubule, a highly metabolic renal cell type (31). Intersecting measured genes from the proximal tubule cell type gene profile and the metabolome in the Human Metabolic Atlas (23) reflected its high metabolic activity across both the urine transcriptome and the metabolome (Fig 4C). For example, the proximal tubule transports molecules for bloodstream reabsorption, and we observed expression of proximal tubule-specific genes in urine encoding proteins participating in metabolic transport reactions (including SLC13A3 and SLC5A12) and several corresponding metabolites in urine that were jointly annotated as belonging to the Transport metabolic subsystem in the Human Metabolic Atlas. The proximal tubule also metabolizes several amino acids (32). As one example, we measured genes encoding proteins that participate in glycine, serine, and threonine metabolism (GLYAT, HAO2, AGXT2, and BHMT) alongside pathway metabolites tiglyl glycine, guanidinoacetate, isobutyrylglycine, isovalerylglycine, methionine, and ornithine (23).

Discussion

Noninvasively measuring genitourinary cell types in urine could improve clinical management spanning early detection, patient prognosis, and therapy response for several genitourinary diseases presently requiring invasive biopsy. In this study, we demonstrated that RNA originating from genitourinary tract tissues is resolvable at cell-type resolution in human urine by reporting the cell-free and sediment transcriptomes. Our observations highlight the potential of transcriptome-wide urine measurements in developing much-needed surrogates for invasive biopsy of genitourinary tissues at cell type resolution in subsequent validation studies.

By measuring the metabolome alongside the transcriptome, we demonstrate that various metabolic pathways are noninvasively enriched across both measurements in urine. To the best of our knowledge, this study is among the first measurements of cfRNA and the metabolome from the same biofluid. We anticipate that coupling transcriptional and metabolic measurements in future cell-free nucleic acid studies will enable increased potential for early disease detection, subtyping, and prognosis in addition to advancing fundamental molecular insights underlying basic disease biology.

In addition to observing RNA from bladder, kidney, and prostate cell types, we observed contributions from secretory cell types in urine. The prostate gland comprises ductal and secretory cell types whose secretions form the basis for seminal fluid and pass through the urethra. The Cowper’s glands in men comprise tubuloalveolar glands whose secretions enter the male urethra (33). While these and other highly specialized genitourinary cell types are absent from TSP (18), deconvolution analyses likely assigned other secretory and glandular cells given the transcriptional overlap with nongenitourinary tract cell types represented in TSP. We hypothesize that our observation of cell type contributions from solid organs beyond the genitourinary tract, which are among highest cellular turnovers in the body (34), in urine cfRNA could be derived from RNA in the blood stream and that was filtered by the glomerulus into the urine. Our ability to measure genes from nonurological cell types is consistent with prior reports observing nonurinary tract tumor-derived DNA (35, 36) and fetal DNA (37) in urine.

The deconvolved fractional estimates of bladder urothelial-specific RNA were reduced relative to urothelium estimates in methylated urine cell-free DNA (38). We hypothesize that this may be due to correlations in the basis matrix of this cell type with prostate epithelia and secretory cells (Supplemental Note 1). This highlights an inherent challenge in deconvolving a complex bulk mixture such as urine with a reference matrix of cell types across the human body given the resulting correlations of diverse cell types owing to overlapping transcriptional functions. Independent signature scoring identified the transcriptional contributions of bladder urothelial cells in urine and its changes in bladder cancer, confirming the presence of this cell type across the urine transcriptome and offering a generalizable means to assess cell type contributions in any cfRNA measurement.

We observed elevated prostate cell type-specific signal in urine cfRNA, which could help address an unmet need of improved prostate cancer screening biomarkers (39). The prostate gland is lined by luminal epithelial cells that interface with seminal fluid, which passes through the urethra in addition to urine. Across urine cfRNA from male donors, we observed variability of this cell type, suggesting that sexual activity prior to specimen collection can impact measurement of this cell type. In future work, comparison of prostate cell type-specific signal in urine cfRNA to current collection strategies for increasing prostate-derived signal through a digital rectal exam or first morning void (40) could be performed. We additionally observed low, nonzero expression of genes nominally annotated as prostate specific and deconvolved fractions of prostate cell type RNA in female urine specimens (Supplemental Note 1). Moreover, inspecting the prostate cell type gene expression distribution with TSP (24) identified cell types from other tissues with overlapping functions; for example, SPDEF is a prostate transcription factor and also participates in mucosal tissue maintenance and exhibits expression in goblet cells (41). BMPR1B exhibits comparable median gene expression in the cervix in the Human Protein Atlas RNA consensus dataset (27) and is expressed in female reproductive tissues (42). Although these genes exhibit high expression in the prostate relative to other tissues in the body, they also exhibit nonzero expression in female tissues and select cell types in other tissues of both sexes.

A challenge with urine measurements is the variability in solute concentration. We therefore used a RNA-library normalization method that directly accounts for variability in library size that can skew differential comparisons (43). To further address this challenge, we measured the spot urine creatinine (15) and the urine volume used for the RNA isolation and included them as covariates during differential expression.

In conclusion, our findings demonstrate that transcriptome-wide measurements on human urine can facilitate noninvasive deconvolution of cell types with low abundance in plasma cfRNA that are challenging to access with conventional invasive biopsy. We anticipate that measurements across the urinary transcriptome and metabolome will advance our understanding of the underlying basic disease biology in genitourinary tissues and inform efforts to develop noninvasive tests across the fields of oncology, urology, nephrology, and transplantation for disease prognosis and early detection.

Data Deposition

Raw sequencing data has been deposited in the Sequence Read Archive under the accession PRJNA1141572. Metabolomics data have been made available in the National Institutes of Health Metabolomics Workbench under the accession ST003356.

Code Availability

Code for the work in this manuscript is available on GitHub at www.github.com/sevahn/urine and at https://doi.org/10.6084/m9.figshare.26828236.v1.

Author Declaration

A prior version of this manuscript has been posted to bioRxiv (https://doi.org/10.1101/2023.10.20.563226).

Supplemental Material

Supplemental materials are available at Clinical Chemistry online.

Nonstandard Abbreviations

cfRNA, cell-free RNA; cf-transcriptome, cell-free transcriptome; MH, multiple hypothesis; TSP, Tabula Sapiens version 1.0.

Human Genes

ANPEP, alanyl aminopeptidase, membrane; MUC2, mucin 2, oligomeric mucus/gel-forming; SLC13A3, solute carrier family 13 member 3; SLC5A12, solute carrier family 5 member 12; GLYAT, glycine-N-acyltransferase; HAO2, hydroxyacid oxidase 2; AGXT2, alanine–glyoxylate aminotransferase 2; BHMT, betaine–homocysteine S-methyltransferase; SPDEF, SAM pointed domain containing ETS transcription factor; BMPR1B, bone morphogenetic protein receptor type 1B.

Author Contributions

The corresponding author takes full responsibility that all authors on this publication have met the following required criteria of eligibility for authorship: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. Nobody who qualifies for authorship has been omitted from the list.

Sevahn Vorperian (Conceptualization-Lead, Data curation-Lead, Formal analysis-Lead, Investigation-Lead, Methodology-Lead, Project administration-Lead, Validation-Equal, Visualization-Equal, Writing—original draft-Lead, Writing—review & editing-Lead), Brian DeFelice (Data curation-Equal, Formal analysis-Supporting, Investigation-Lead, Methodology-Supporting, Writing—review & editing-Equal), Joseph Buonomo (Investigation-Equal, Methodology-Equal, Writing—review & editing-Equal), Hagop Chinchinian (Formal analysis-Supporting, Methodology-Supporting), Ira Gray (Investigation-Supporting, Writing—review & editing-Supporting), Jia Yan (Investigation-Supporting), Kathleen Mach (Methodology-Supporting, Project administration-Supporting, Visualization-Supporting, Writing—review & editing-Equal), Vinh La (Investigation-Supporting), Timothy Lee (Investigation-Supporting), Joseph Liao (Investigation-Supporting, Project administration-Supporting, Resources-Equal, Supervision-Supporting, Writing—review & editing-Supporting), Richard Lafayette (Investigation-Supporting, Supervision-Supporting, Writing—review & editing-Supporting), Gabriel Loeb (Conceptualization-Supporting, Investigation-Supporting, Methodology-Supporting, Supervision-Supporting, Writing—review & editing-Equal), Carolyn Bertozzi (Project administration-Supporting, Resources-Supporting, Supervision-Supporting, Writing—review & editing-Supporting), and Stephen Quake (Conceptualization-Supporting, Formal analysis-Supporting, Funding acquisition-Lead, Investigation-Equal, Methodology-Supporting, Project administration-Equal, Resources-Lead, Supervision-Lead, Validation-Equal, Visualization-Equal, Writing—original draft-Supporting, Writing—review & editing-Equal)

Authors’ Disclosures or Potential Conflicts of Interest

Upon manuscript submission, all authors completed the author disclosure form.

Research Funding

This work is supported by the Chan Zuckerberg Biohub (S.R. Quake). S.K. Vorperian is supported by the National Science Foundation Graduate Research Fellowship (Grant # DGE1656518), the Benchmark Stanford Graduate Fellowship, and the Stanford ChEM-H Chemistry Biology Interface training program. J.A. Buonomo was supported by a National Institutes of Health Ruth L. Kirschstein NRSA F32 Postdoctoral Fellowship (5F32GM134689-03). J.C. Liao is supported by National Cancer Institute Grant R01 CA244526 and the Department of Veterans Affairs Merit Review I01 BX004962. C.R. Bertozzi is an investigator of the Howard Hughes Medical Institute and received support from National Institutes of Health Merit Award AI51622.

Disclosures

S.K. Vorperian, G.B. Loeb, and S.R. Quake are inventors on a patent application submitted by the Chan Zuckerberg Biohub and Stanford University for nucleic acid measurement in urine. S.K. Vorperian and S.R. Quake are inventors on a patent application submitted by Chan Zuckerberg Biohub and Stanford University for noninvasive profiling of cell types in biofluids. B.C. DeFelice is a member of the Metabolomics Association of North America Early Career Members Council and is a consultant for Surrozen Inc. G.B. Loeb is a member of the scientific advisory board for Judo Bio. J.A. Buonomo is a consultant for Boston Consulting Group. C.R. Bertozzi is a cofounder and scientific advisory board member of Grace Science LLC, Lycia Therapeutics, Palleon Pharmaceuticals, Enable Bioscience, Redwood Biosciences (a subsidiary of Catalent), OliLux Bio, InterVenn Bio, GanNA Bio, Firefly Bio, Neuravid Therapeutics, and Valora Therapeutics, and is a member of the board of directors of Alnylam Pharmaceuticals and OmniAb. C.R. Bertozzi received a Nobel Prize in Chemistry (2022). S.R. Quake is a founder and shareholder of Superfluid Diagnostics, Karius Dx, Mirvie, and Clear Note Health.

Role of Sponsor

The funding organizations played no role in the design of the study, choice of enrolled patients, review and interpretation of data, preparation of the manuscript, or final approval of the manuscript.

Acknowledgments

We thank C. Liou, N. Mehta, J. Long, E. Sattely, and Y. Dai for metabolomics discussions. We thank J. Zou for confidence interval discussions. We thank B. Yu for sequencing discussions. We thank T. Meyer for renal discussions and urine specimens. We thank Quake lab members past and present for helpful discussions. We thank the ChEM-H Knowledge Center for QTOF access. We thank R. LaMantia for making the rainbow cell spectrum in Fig 1A; the cells in Fig. 1A were generated using BioRender.

References