Capturing the Kidney Transcriptome by Urinary Extracellular Vesicles-From Pre-Analytical Obstacles to Biomarker Research

Abstract

Urinary extracellular vesicles (uEV) hold non-invasive RNA biomarkers for genitourinary tract diseases. However, missing knowledge about reference genes and effects of preanalytical choices hinder biomarker studies. We aimed to assess how preanalytical variables (urine storage temperature, isolation workflow) affect diabetic kidney disease (DKD)-linked miRNAs or kidney-linked miRNAs and mRNAs (kidney-RNAs) in uEV isolates and to discover stable reference mRNAs across diverse uEV datasets. We studied nine raw and normalized sequencing datasets including healthy controls and individuals with prostate cancer or type 1 diabetes with or without albuminuria. We focused on kidney-RNAs reviewing literature for DKD-linked miRNAs from kidney tissue, cell culture and uEV/urine experiments. RNAs were analyzed by expression heatmaps, hierarchical clustering and selecting stable mRNAs with normalized counts (>200) and minimal coefficient of variation. Kidney-RNAs were decreased after urine storage at -20 °C vs. -80 °C. Isolation workflows captured kidney-RNAs with different efficiencies. Ultracentrifugation captured DKD -linked miRNAs that separated healthy and diabetic macroalbuminuria groups. Eleven mRNAs were stably expressed across the datasets. Hence, pre-analytical choices had variable effects on kidney-RNAs-analyzing kidney-RNAs complemented global correlation, which could fade differences in some relevant RNAs. Replicating prior DKD-marker results and discovery of candidate reference mRNAs encourages further uEV biomarker studies.

Keywords: diabetic kidney disease; exosomes; mRNA; miRNA; reference genes; sequencing; urinary extracellular vesicles; urine.

Figure 1

Effect of storage temperature on kidney top expressed miRNAs and kidney enriched genes in uEV isolates. Urine EV were isolated by UC from urine stored at −20 °C vs. −80 °C. (A): Heatmap depicts the expression level of 29 out of 40 top miRNAs expressed in kidney (miRNATissueAtlas2) and found in the uEV isolates (miRNAs with ≥1 raw count in at least 50% of the samples). (B): Heatmap depicts the expression of 33 out of 53 kidney enriched genes (human protein atlas) found in the uEV isolates (genes with ≥5 raw counts in at least 50% of the samples). Micro RNA (miRNA), messenger RNA (mRNA), ultracentrifugation (UC), urinary extracellular vesicles (uEV).

Figure 2

Effect of EV isolation workflows on kidney top expressed miRNAs and kidney enriched genes in uEV isolates. Urine EV were isolated by HFD, NG and UC workflows from urine samples of healthy controls (n = 5) and T1D patients with macroalbuminuria (n = 5). (A). Heatmap depicts the expression level of 36 out of 40 top miRNAs expressed in kidney (miRNATissueAtlas2) and found in the uEV isolates (miRNAs with ≥5 raw counts in at least 50% of the samples). (B). Heatmap depicts the expression level of 31 out of 56 kidney enriched genes (Human protein atlas) found in the uEV isolates (genes with ≥5 raw counts in at least 50% of the samples).

Figure 3

Urinary EV capture miRNAs associated with DKD. (A,B,E). Expression Heatmaps depict the expression of miRNAs associated with DKD and expressed in our uEV datasets. (C,D). Depict principal component analysis. (A). MiRNAs with evidence of dysregulation in kidney tissue and or cell lines and (B,C). miRNAs dysregulated in uEV/urine/urinary sediments. (D). MiRNAs with the highest fold changes from figure B which are part of the first and fourth cluster. The uEV expression data used in (A–D) corresponds to the UC isolation workflow dataset comprising healthy control and T1D macroalbuminuria groups ([13], part of our UC miRNA dataset in Table 3 and Table 4). (E). The 31 miRNAs that could separate individuals with DKD and macroalbuminuria (as shown in B) were analyzed in the PCa uEV miRNA dataset [6]. Diabetic kidney disease (DKD), Prostate cancer (PCa), ultracentrifugation (UC), urinary extracellular vesicles (uEV).

Figure 4

Stable mRNA in common across diverse uEV datasets. Venn diagrams depict elements in common between the different datasets. (A–C). A 3-step search for the stable uEV mRNA using the top 100 genes with the lowest CV from each dataset. Diabetic kidney disease (DKD), prostate cancer (PCa), urinary extracellular vesicles (uEV).

Figure 5

The mRNA sequencing read counts of candidate reference genes in pre-analytical and DKD uEV datasets from men. The uEV datasets included healthy controls and individuals with type 1 diabetes and different stages of albuminuria as well as comparisons of preanalytical variables (all male). (A–D). Line graphs depicts CPM of HSPD1, SRSF3, VAPA, RAB1A and MORF4L1 across samples and (E,F). Boxplots depict CPM per candidate reference genes. A reference gene used commonly for normalization (GAPDH) and a gene with high CV in all datasets (UPK1A) were included. (A,E). EV isolation workflows, (B,F). In column DNAse treatment during uEV RNA extraction, (C,G). A technical dataset (type of urine collection, pre-clearing the urine before freezing, and technical replicates) and (D,H). DKD cohort 1. Sample pairs or triplicates are named similarly apart from the abbreviation of the tested variable. Centrifuged (C), Coefficient of variation (CV), counts per million (CPM), hydrostatic filtration dialysis (HFD), macroalbuminuria (Macro), microalbuminuria (Micro), normoalbuminuria (Normo), ultracentrifugation (UC).

Figure 5

The mRNA sequencing read counts of candidate reference genes in pre-analytical and DKD uEV datasets from men. The uEV datasets included healthy controls and individuals with type 1 diabetes and different stages of albuminuria as well as comparisons of preanalytical variables (all male). (A–D). Line graphs depicts CPM of HSPD1, SRSF3, VAPA, RAB1A and MORF4L1 across samples and (E,F). Boxplots depict CPM per candidate reference genes. A reference gene used commonly for normalization (GAPDH) and a gene with high CV in all datasets (UPK1A) were included. (A,E). EV isolation workflows, (B,F). In column DNAse treatment during uEV RNA extraction, (C,G). A technical dataset (type of urine collection, pre-clearing the urine before freezing, and technical replicates) and (D,H). DKD cohort 1. Sample pairs or triplicates are named similarly apart from the abbreviation of the tested variable. Centrifuged (C), Coefficient of variation (CV), counts per million (CPM), hydrostatic filtration dialysis (HFD), macroalbuminuria (Macro), microalbuminuria (Micro), normoalbuminuria (Normo), ultracentrifugation (UC).

Figure 6

The mRNA sequencing read counts of the candidate reference genes in uEV datasets from DKD study of women and from prostate cancer patients. (A,B). Line graphs depicts CPM of HSPD1, SRSF3, VAPA, RAB1A and MORF4L1 across samples and (C,D). Boxplots depict CPM per candidate reference genes. A reference gene used commonly for normalization (GAPDH) and a gene with high CV in all datasets (UPK1A) were included. The uEV datasets included A. DKD cohort 2 (women with type 1 diabetes and different stages of albuminuria) and B. PCa patients and healthy controls (technical replicates, R1-3). Samples PCa1, 3 and 4 were obtained before prostatectomy. Sample PCa2 was obtained after prostactectomy from the same donor as PCa1. Coefficient of variation (CV), counts per million (CPM), macroalbuminuria (Macro), microalbuminuria (Micro), normoalbuminuria (Normo), prostate cancer (PCa).

Figure 6

The mRNA sequencing read counts of the candidate reference genes in uEV datasets from DKD study of women and from prostate cancer patients. (A,B). Line graphs depicts CPM of HSPD1, SRSF3, VAPA, RAB1A and MORF4L1 across samples and (C,D). Boxplots depict CPM per candidate reference genes. A reference gene used commonly for normalization (GAPDH) and a gene with high CV in all datasets (UPK1A) were included. The uEV datasets included A. DKD cohort 2 (women with type 1 diabetes and different stages of albuminuria) and B. PCa patients and healthy controls (technical replicates, R1-3). Samples PCa1, 3 and 4 were obtained before prostatectomy. Sample PCa2 was obtained after prostactectomy from the same donor as PCa1. Coefficient of variation (CV), counts per million (CPM), macroalbuminuria (Macro), microalbuminuria (Micro), normoalbuminuria (Normo), prostate cancer (PCa).

Figure 7

Protein-protein interaction network analysis for the reference gene candidates. Network was generated using STRING (https://string-db.org/, accessed on 28 April 2023) and reproduced under Creative Commons BY 4.0 license (https://creativecommons.org/licenses/by/4.0/, accessed on 28 April 2023). Only interactions with experimental validation evidence in the literature are shown.