Current Methodological Challenges of Single-Cell and Single-Nucleus RNA-Sequencing in Glomerular Diseases

Abstract

Single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA-seq (snRNA-seq) allow transcriptomic profiling of thousands of cells from a renal biopsy specimen at a single-cell resolution. Both methods are promising tools to unravel the underlying pathophysiology of glomerular diseases. This review provides an overview of the technical challenges that should be addressed when designing single-cell transcriptomics experiments that focus on glomerulopathies. The isolation of glomerular cells from core needle biopsy specimens for single-cell transcriptomics remains difficult and depends upon five major factors. First, core needle biopsies generate little tissue material, and several samples are required to identify glomerular cells. Second, both fresh and frozen tissue samples may yield glomerular cells, although every experimental pipeline has different (dis)advantages. Third, enrichment for glomerular cells in human tissue before single-cell analysis is challenging because no effective standardized pipelines are available. Fourth, the current warm cell-dissociation protocols may damage glomerular cells and induce transcriptional artifacts, which can be minimized by using cold dissociation techniques at the cost of less efficient cell dissociation. Finally, snRNA-seq methods may be superior to scRNA-seq in isolating glomerular cells; however, the efficacy of snRNA-seq on core needle biopsy specimens remains to be proven. The field of single-cell omics is rapidly evolving, and the integration of these techniques in multiomics assays will undoubtedly create new insights in the complex pathophysiology of glomerular diseases.

Single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA-seq (snRNA-seq) are powerful new methods that allow transcriptomic profiling of thousands of cells from a renal biopsy specimen at single-cell resolution.^1,2 When performing scRNA-seq and snRNA-seq, four different steps can be recognized in the experimental workflow.³ First, a high-quality, single-cell or single-nucleus suspension is created from the tissue sample (Figure 1). In scRNA-seq, viable single cells are dissociated from their extracellular matrix; in snRNA-seq, single nuclei are isolated using stronger dissociation methods that lyse the cell membranes and free the nuclei.³ In the second step, the mRNA within these cells or nuclei is captured, reverse transcribed into cDNA, amplified, and sequenced (Figure 2).³ The third step consists of raw data processing, which is subsequently analyzed in the final step, by using advanced bioinformatics pipelines that translate sequencing data into expression profiles of individual cells, leading to the identification of distinct cell populations (Figure 2).³

Figure 1.

General workflow of scRNA-seq and snRNA-seq experiments. Renal tissue is mechanically and enzymatically dissociated into a single-cell or single-nucleus suspension. In scRNA-seq, viable single cells are dissociated from their extracellular matrix; in snRNA-seq, stronger dissociation techniques are used to dissociate nuclei from the cells.³ Next, cells or nuclei are sorted into a microtiter plate or loaded into a microfluidics device containing integrated fluidic circuits (IFCs), nanowells, or droplets, ready for subsequent processing steps.

Figure 2.

General workflow of scRNA-seq and snRNA-seq experiments, continued. Individual cells or nuclei are lysed and transcripts from single cells or single nuclei are captured using poly[T] oligonucleotide primers that hybridize with the polyadenylated (poly[A]) mRNA.^3,82 These primers may also contain other oligonucleotide sequences, including a cell barcode (BC), unique molecular identifier (UMI), and adaptor sequence for PCR or RNA polymerase promotor (T7 promotor) for in vitro transcription (IVT).³ Next, a reverse transcription (RT) reaction synthesizes the first and second strand of cDNA, which is subsequently amplified by either PCR or IVT. Amplified cDNA molecules are fragmented (using enzymatic tagmentation or chemical fragmentation) and ligated with adaptors for sequencing.³ A sequence library is created and, finally, using advanced bioinformatics analysis, sequencing data are translated into cellular expression patterns, leading to the identification of distinct cell clusters.⁸³ Adapted from refs. ^3,4,81,82. UMAP, Uniform Manifold Approximation and Projection.

The first two steps of the workflow are crucial, because they are prone to several methodologic pitfalls. First, the total number of cells that needs to be profiled increases significantly when investigating less abundant cell populations, such as podocytes.⁴ Second, the choice of the dissociation procedure comes with a certain degree of “dissociation bias,” especially in scRNA-seq protocols.^5,6 Indeed, in mild dissociation protocols, cells strongly embedded in the extracellular matrix remain attached and are, therefore, under-represented in the scRNA-seq data.^7–9 In contrast, with stronger dissociation protocols, fragile cells may become damaged, resulting in a reduction in cell viability.^7–9 In snRNA-seq, difficult-to-dissociate cell types are more easily isolated, because stronger dissociation protocols are used and only nuclei are isolated, as opposed to viable cells in scRNA-seq.³ Finally, the dissociation process itself could also result in transcriptional stress responses and subsequent RNA degradation, which could mask true biologic signals within the data.^7,10

Theoretically, all cells from the glomerular microenvironment—such as podocytes, glomerular endothelial cells, mesangial cells, and leukocytes—can be studied with single-cell transcriptomics. In reality, however, isolation and analysis of glomerular cells is very challenging.¹¹ On the basis of the current literature on scRNA-seq and snRNA-seq experiments on renal tissue, this review aims to answer five key questions to optimize the design of single-cell transcriptomics experiments that focus on glomerular pathology (Figures 3 and 4; search strategy is outlined in Supplemental Figures 1 and 2).

Figure 3.

Technical flowchart of scRNA-seq and snRNA-seq experiments on renal tissue of humans or mice. The first step involves choosing the renal tissue of interest. Human renal tissue can be derived from nephrectomy tissue or core needle biopsies.

Figure 4.

Technical flowchart of scRNA-seq and snRNA-seq experiments on renal tissue of humans or mice, continued. ^IsnRNA-seq has not yet been performed on human core needle biopsies. ^IIMethanol fixation of single cells has not yet been performed on human renal tissue. ^IIIMagnetic bead isolation of glomeruli is not possible in human patients/tissue. ^IVManual selection of glomeruli before tissue dissociation has only been performed in one scRNA-seq study and still requires validation.^{18 V}FACS/MACS have successfully been used to enrich for glomerular endothelial cells in scRNA-seq experiments on mice,^44,50 but not yet for podocytes or mesangial cells. ^VICold dissociation protocols have been successfully used on renal tissue of mice,^33,34,45,52 but studies on human renal tissue have yet to be published. (+), advantage; (-), disadvantage.

QUESTION 1: WHICH RENAL BIOPSY SPECIMEN CAN BE USED FOR THE ISOLATION OF GLOMERULAR CELLS?

Single-cell transcriptomics experiments have been performed on both mouse and human kidneys (Supplemental Tables 1 and 2), and many different scRNA-seq protocols have been used (Table 1). Studies on mice allow resection of one or both kidneys, which yields ample starting tissue material. In humans, however, renal tissue is only available after core needle biopsies or nephrectomy procedures (Figure 3).

Table 1.

Overview of different scRNA-seq protocols used in studies on renal tissue

Method	SMART-seq SMART-seq2	Fluidigm C1 System (SMART-seq)	Fluidigm C1 System (mRNA Seq HT)	CEL-seq2	Drop-seq	inDrops	10x Genomics Chromium Single Cell Gene Expression Solution
Platform	Plate (FACS)	Microfluidics (IFC)	Microfluidics (IFC)	Plate (FACS)	Microfluidics (droplet)	Microfluidics (droplet)	Microfluidics (droplet)
Reaction volume	Microliter	Nanoliter	Nanoliter	Microliter	Nanoliter	Nanoliter	Nanoliter
Cell-size limitations	Independent of cell size	Three types of IFCs available: 5–10 µm, 10–17 µm, and 17–25 µm	Two types of IFCs available: 5–10 µm and 10–17 µm	Independent of cell size	Adjustable with microfluidic and droplet size	Adjustable with microfluidic and droplet size	Maximum 50–60 µm
Transcript coverage	Full length	Full length	3′ end	3′ end	3′ end	3′ end	3′ end and 5′ end
Cell barcode incorporation	Tagmentation	Tagmentation	Oligo(T) primer 3′	Oligo(T) primer 3′	Oligo(T) primer 3′	Oligo(T) primer 3′	Oligo(T) primer 3′
Use of UMI possible	No	No	No	Yes	Yes	Yes	Yes
cDNA amplification	PCR	PCR	PCR	IVT	PCR	IVT	PCR
Fragmentation method	Tagmentation	Tagmentation	Tagmentation	Random priming	Tagmentation	Chemically (KOAc, MgOAc)	Enzymatic fragmentation
Cellular capture rate of sample	Depending on sorting settings	Approximately 10% of cells	Approximately 10% of cells	Depending on sorting settings	<5% of cells	>75% of cells	>50% of cells
Typical cell output per run (microfluidics)	N/A	Maximum 96 cells per run	Maximum 800 cells per run	N/A	Thousands of cells	Thousands of cells	Thousands of cells
Landmark study	V184 and V285	86	see Fluidigm website	87	88	89	90
Application on renal tissue	V148 and V229,91,92	13,26,40,41,46	11,15,43	14	9,32,34,49,52,56	2	6,14,16,17,19,21–23,25–27,33,34,39,42,44,45,47,50,51,55,58,69,91–98

The data included in this table was derived from previously published reviews^{3,4,7,82,99,100} and additional literature search. Two additional, infrequently used, plate-based protocols (Microwell-seq²⁸ and STRT-seq¹⁸) were not included in this table. IFC, integrated fluidics circuits; UMI, unique molecular identifier; IVT, in vitro transcription; KOAc, potassium acetate; MgOAc, magnesium acetate; N/A, not available.

Human Core Needle Biopsy Specimens

Renal tissue in glomerular diseases is almost exclusively available after core needle biopsy. A major disadvantage of this is the small amount of cells contained in one core: approximately 40,000–75,000 cells in total.² Strategies to obtain more tissue include the use of larger biopsy needles, taking one extra biopsy core per patient,¹² or pooling the material of several patients.

Since 2017, eight studies have performed scRNA-seq experiments on human renal tissue obtained from core needle biopsies (Table 2).^2,6,13–18 Six studies used unsorted tissue,^{2,6,13,15–17} one study analyzed FACS-isolated leukocytes and epithelial cells,¹⁴ whereas the most recent study used a stepwise dissociation protocol in which glomeruli were selected before tissue dissociation.¹⁸ Three of these studies were able to isolate podocytes from core needle biopsy specimens.^6,17,18 The first study identified all glomerular cell types from tissue obtained from both core biopsies (preperfusion biopsy specimens from kidney donors, n=3; surveillance biopsy specimens from kidney allografts, n=5) and nephrectomy tissue (n=16).⁶ However, only 11 podocytes were isolated from the tissue from the eight core biopsies (11 of 7524 cells, 0.15%), whereas 159 podocytes originated from nephrectomy tissue (159 of 14,744 cells, 1.08%). A second study, by the same research group, also identified all glomerular cells from 18 core biopsy specimens from healthy kidney donors (25,163 cells) and 44 core biopsy specimens from patients with diabetic kidney disease (85,872 cells), but this study did not mention podocyte numbers or percentages.¹⁷ The last study analyzed 13 core biopsy specimens from patients with IgA nephropathy, and six nephrectomy specimens, and used a stepwise dissociation approach to enrich for glomeruli.¹⁸ The renal tissue was mildly digested, glomeruli were manually picked under a microscope with a needle and pipette, and the selected glomeruli and remaining tissue were separately dissociated further into a single-cell suspension and analyzed. Although glomerular cells were successfully isolated, the total numbers of podocytes (26 of 2875 cells, 0.9%) and mesangial cells (71 of 2875 cells, 2.5%) were still low and comparable with some scRNA-seq studies that did not use any sorting techniques.^6,19 Manual picking of glomeruli is labor intensive, low throughput, and might induce transcriptional artifacts or reduce the viability of fragile cells (such as podocytes), which might explain the low cell numbers.

Table 2.

scRNA-seq studies on human renal core needle biopsy specimens

Author	Protocol	Age of Patients	Txa	Diseaseb	Biopsyc	T°d	Cellse	Depthf	Genesg	Glomh	Remarks
Der et al.13	Fluidigm C1 SMART-seq	A	N	SLE (n=10)	Partial core or full core (n=16)	(+), W	899*	500,000	700	—	*Pooling of renal and skin cells and PBMCs
Wu et al.2	inDrops	A	T	Acute TCMR (n=1)	Full core (n=1)	(+), W	4487	50,000	827	—	Study also performed snRNA-seq on healthy tissue (Supplemental Table 2)
Der et al.15	Fluidigm C1 mRNA Seq HT	A	N	SLE (n=21); healthy (Pre) (n=3)	Partial core (n=24)	(-),W, cryo	4019*	200,000	?	M	*Pooling of renal and skin cells
Arazi et al.14	CEL-seq2* 10x Chromium**	A	N	SLE (n=24); healthy (Pre) (n=10)	Partial core or full core (n=34)	(-), W, cryo	2838*; 122**	Approximately 106*; ?**	1000–5000*; 250–3500**	—	scRNA-seq on FACS-isolated leukocytes (CD45+) versus epithelial cells (CD45-, CD10+). *CEL-seq2; **10x on two healthy donor biopsy specimens
Menon et al.6	10x Chromium	A	N; T	Healthy (Tu) (n=16); healthy (Pre) (n=3); healthy (S) (n=5)	Partial core (n=8); neph (n=16)	(-), W, cryo	7524*; 14,744**	3971*,***; 3089**,***	1339*; 1134**	P, E, M	*CNB; **nephrectomies; ***mUMIs per cell; P°, *11 cells (0.14%), **159 cells (1.08%)
Malone et al.16	10x Chromium	A	T	ABMR (n=3); nonrejection AKI (n=2)	Full core (n=5)	(+), W	81,139	2497*	1124	—	*Mean transcripts per cell, not raw reads
Menon et al.17	10x Chromium	A	N	Healthy (Pre) (n=18); DKD (n=44)	Partial core (n=62)	(-), W, cryo	25,163*; 85,872**	?	500–5000	P, M	*Healthy; **DKD; P°, not mentioned
Zheng et al.18	STRT-seq	A	N	IgAN (n=13)*; healthy (Tu)** (n=6)	Partial core (n=13); neph (n=6)	(+), W ***	2022*; 763**	37,875*; 38,391**	3348*; 3105**	P, M	*IgAN;**healthy;
											***stepwise dissociation:
											(1) glomerulus isolation (pipetting),
											(2) MACS CD14+ cells,
											(3) MACS CD326+ cells,
											(4) unselected cells
											P°, *22 cells (1%), **four cells (0.52%)

Tx, tissue; T°, temperature; glom, glomerular cells; A, adult; N, native kidney; (+), fresh tissue; W, warm dissociation; −, no glomerular cells; T, transplant kidney; TCMR, T cell–mediated rejection; Pre, preperfusion or pretransplant core needle biopsy of living donor kidney; (-), frozen tissue; cryo, cryopreservation; ?, data could not be found in published paper; M, mesangial cells; Tu, tumor-free tissue at maximal distance of mass or tumor; S, surveillance kidney transplant biopsy; neph, biopsy from partial/total nephrectomy; P, podocytes; E, glomerular endothelial cells; CNB, core needle biopsy; mUMIs, mean unique molecular identifiers; P°, absolute and relative number of podocytes isolated in scRNA-seq studies; ABMR, antibody-mediated rejection; DKD, diabetic kidney disease; IgAN, IgA nephropathy; MACS, magnetic-activated cell sorting.

^aTissue from native kidneys or transplant kidneys.

^bHealthy or pathologic kidney tissue. n=number of patients.

^cBiopsy technique. n=biopsy samples taken.

^dFresh versus frozen tissue and warm versus cold dissociation.

^eTotal number of cells isolated and analyzed after quality control.

^fSequencing depth defined as “mean (raw) reads per cell,” “mean transcripts per cell,” or “mean UMIs per cell.”

^gMean number of genes per cell.

^hIsolation of glomerular cells.

Interestingly, another scRNA-seq study analyzed a large number of cells (total of 81,139 cells) from five core needle biopsy specimens from renal allografts (antibody-mediated rejection, n=3; nonrejection AKI, n=2), but was unable to identify glomerular cells.¹⁶ A low number of glomeruli in the core biopsy samples might have played a major role (sample bias); however, differences in dissociation methods, tissue preservation, and the presence of renal inflammation might have also influenced the researchers’ ability to identify glomerular cells.

Human Nephrectomy Tissue

In the context of common renal pathologies (e.g., diabetic nephropathy, hypertensive nephrosclerosis), residual (tumor-free) tissue from partial or total nephrectomy specimens might be available.^20,21 In contrast, nephrectomy tissue is generally unavailable in patients with rare glomerular diseases. Indeed the majority of single-cell transcriptomics studies (both scRNA-seq and snRNA-seq) on human nephrectomy tissue have involved healthy renal tissue.^{2,6,18,19,22–28} In all three snRNA-seq experiments, glomerular cells were present, including podocytes.^2,20,24 Similarly, most of the scRNA-seq studies detected glomerular cells, including podocytes,^{6,18,19,21,23,25} glomerular capillary endothelial cells,^{6,21,23,25,28} and mesangial cells.^6,18,19,28

Concluding Answer

Renal tissue in glomerular diseases is almost exclusively available after core needle biopsy. scRNA-seq on specimens from core needle biopsies is able to characterize podocytes and other glomerular cells, although cell counts remain very low.^6,18 Manually enriching for glomeruli before tissue dissociation might modestly increase the chances of identifying podocytes; however, this technique is low throughput, labor intensive, and should be validated in future studies.¹⁸ The analysis of more cells per sample might increase the chance of identifying podocytes, and the inclusion of more samples might circumvent the risk of sample bias when using tissue from core needle biopsies. Whether or not snRNA-seq applied to core needle biopsy specimens is a valid alternative for studying glomerular cells remains to be investigated.

QUESTION 2: SHOULD FRESH OR FROZEN TISSUE BE USED TO ISOLATE GLOMERULAR CELLS?

When using scRNA-seq on fresh tissue, a biopsy sample should be dissociated immediately into a single-cell suspension and processed further. This does not allow for histopathologic evaluation of the biopsy sample (which takes several days) before processing and, therefore, hampers the selection of well-selected disease phenotypes.

Using a hypothermic preservation solution might be an alternative for immediate processing.²⁹ HypoThermosol FRS (HTS-FRS) can be used to store fresh tissue at 2–8°C for 24–72 hours before starting the dissociation process.²⁹ Mouse kidney cells that were treated with HTS-FRS and preserved for up to 72 hours showed no reduced viability, and scRNA-seq of FACS-isolated immune cells showed a similar cell type heterogeneity with an equal average number of detected genes per cell, when compared with immediately dissociated cells.²⁹ However, to date, studies on human renal tissue have only used HTS-FRS to store samples for a few hours before downstream processing¹³ or cryopreservation.^6,14,15

scRNA-Seq on Frozen Tissue

Snap freezing of tissue results in low-quality single-cell suspensions once the samples are thawed and subjected to standard dissociation protocols.³⁰ Therefore, cryoprotectants are necessary to keep the cells viable and intact upon thawing. There are two preservation protocols for renal tissue: (1) methanol fixation of a single-cell suspension, and (2) cryopreservation of undissociated tissue or a single-cell suspension in DMSO.

Methanol is a coagulating fixative that does not alter the structure of nucleic acids, and dissociated, methanol-fixed cells can be frozen at −80°C and stored for several weeks.³¹ To date, no studies have used methanol fixation on human kidney cells, this has only been performed in mice (Supplemental Table 3).^9,32–34 One study directly compared scRNA-seq on fresh tissue versus methanol-fixed frozen cells of mouse kidneys.³³ The methanol-fixed cell composition was comparable with fresh tissue and included podocytes, although some cell types, especially macrophages, were slightly under-represented.³³ In line with previous studies, methanol fixation appeared to damage cells, resulting in ambient RNA contamination.^33,35

In the cryopreservation technique, whole tissue specimens or single-cell suspensions are immersed in a cryopreservation medium containing DMSO, and subsequently frozen and stored at −80°C.³⁵ In one study on renal tissue from healthy mice, scRNA-seq of cryopreserved, single-cell suspensions performed poorly in isolating epithelial cells, such as proximal tubule cells, and also upregulated a cellular stress response.³³ However, this result contrasts with five studies that successfully used scRNA-seq methods on cryopreserved tissue from human core needle biopsies^6,14,15,17 or nephrectomy tissue,^6,19 in which tissue was frozen in cryopreservation medium and only dissociated after thawing (Supplemental Table 4). These studies did not report problems with isolating tubular cells,^6,15,17,19 and were also able to isolate glomerular cells, including podocytes.^6,17,19

snRNA-Seq on Frozen Tissue

In the third technique, undissociated renal tissue is snap frozen without fixation or cryopreservation and stored at −80°C. This tissue appears to only be compatible with snRNA-seq pipelines. Several studies, both in mice and humans, have successfully used snRNA-seq on snap-frozen renal tissue.^{9,20,24,33,34,36–39} All of these studies were able to isolate podocytes and many also identified other glomerular cells, although no studies used tissue from core needle biopsies.

Concluding Answer

scRNA-seq of frozen tissue might be an attractive alternative to fresh biopsy material when studying glomerular diseases. To date, frozen, methanol-fixed, single-cell suspensions of human kidneys have not been studied. Cryopreservation with DMSO before single-cell dissociation has yielded good and consistent results in human biopsy specimens, including biopsy cores. snRNA-seq on snap-frozen tissue is efficient in capturing glomerular cells, but its efficacy in biopsy cores still needs to be proven.

QUESTION 3: CAN GLOMERULAR CELLS BE ENRICHED IN SINGLE-CELL TRANSCRIPTOMICS EXPERIMENTS?

Using selected tissue sections or cell-sorting techniques in combination with scRNA-seq or snRNA-seq to enrich for rare kidney cells is an appealing, but challenging, option. In scRNA-seq, every manipulation of tissue or cells before mRNA capture and sequencing poses the risk of inducing transcriptional artifacts or decreased cell viability of fragile cells. In snRNA-seq, cell-surface staining is not possible and alternative, complex, nuclear-staining methods are required to sort nuclei.

Selection of Glomeruli before Tissue Dissociation

In mice, glomeruli can be selectively isolated using magnetic bead perfusion. In this technique, the kidneys of a euthanized mouse are perfused with a solution containing magnetic beads that accumulate in the glomerular vessels and are used for magnetic isolation of glomeruli after partial enzymatic tissue digestion.^{11,32,40–43} This method effectively isolates viable glomerular cells, but is not possible in humans. Alternatively, cortical renal tissue can be dissected from the medulla before dissociation to enrich for glomeruli,^34,44,45 which might be useful in nephrectomy tissue, but not in human core needle biopsy specimens. In humans, only one recent scRNA-seq study performed isolation of glomeruli before single-cell dissociation, using a labor-intensive and low-throughput method in which glomeruli were manually selected under a microscope after mild digestion of the renal tissue.¹⁸ The yield of glomerular cells remained low, and no direct comparison of this technique with scRNA-seq of unsorted tissue has been published. Taken together, to date, no validated techniques are available that efficiently enrich for glomeruli in human renal tissue before dissociation.

Glomerular Cell Sorting after Tissue Dissociation (FACS and Magnetic-Activated Cell Sorting)

FACS and magnetic-activated cell sorting (MACS) are cell-sorting techniques that have been used in dissociated renal cells to selectively enrich for a target cell population. In mouse kidneys, scRNA-seq studies enriched for collecting duct cells,⁴⁶ immune cells,^26,29,47 myofibroblasts,⁴⁸ and endothelial cells^44,49,50 (including glomerular endothelial cells^44,50). One snRNA-seq study involved the creation of a reporter mouse that expressed green fluorescent protein in nephron progenitor cells, which enabled the selective enrichment for nuclei of mature nephron cells by sorting for green fluorescent protein–positive nuclei.⁵¹ In human kidneys, scRNA-seq studies enriched for immune cells,^14,18,26 epithelial cells,¹⁸ proximal tubule cells,²¹ and myofibroblasts.²¹

However, to date, no studies have used cell-sorting techniques to enrich for podocytes or mesangial cells before scRNA-seq. Indeed, selective enrichment of podocytes might be very challenging. It is unclear whether cell-enrichment methods are efficient enough to isolate the few podocytes that are present in tissue from human core biopsies. FACS and MACS both require cell staining, washing, and sorting steps that might damage the podocytes, alter the transcriptional profile, and reduce viability, with low cell yield in downstream scRNA-seq analysis. However, this hypothesis should be tested in future experiments.

Concluding Answer

In mice, glomeruli can be isolated using magnetic beads, which successfully enriches for glomerular cells. In humans, current techniques to selectively enrich for glomeruli before dissociation are still low throughput and require validation and optimization. Sorting of podocytes and mesangial cells with FACS or MACS before single-cell capture and scRNA-seq has not been performed and should be the focus of future research.

QUESTION 4: WHICH DISSOCIATION PROTOCOL IS ABLE TO ISOLATE GLOMERULAR CELLS?

A single-cell suspension should meet the following ideal characteristics: the detected cell types should be representative of the sampled tissue, cell viability should be high, and the cellular transcriptomes should have as little dissociation-induced artifacts as possible.⁵² Therefore, an important question is how to adapt dissociation protocols when studying difficult-to-dissociate cell types, such as podocytes, in scRNA-seq experiments.

Dissociation into Single-Cell Suspension

Standard dissociation is performed at 37°C, which can trigger a cellular stress response and dissociation-induced artifacts.⁵² The use of cold active protease in mouse kidneys and dissociating at 6°C or 12–16°C results in less dissociation-induced artifacts.^45,52 A recent study compared scRNA-seq on warm versus cold dissociated mouse renal tissue, and used this cold active protease in a dissociation protocol on ice.³³ This study confirmed that warm dissociation induces a cellular stress response and that cells prone to these transcriptional changes, such as podocytes, are under-represented after warm dissociation. However, cold dissociation protocols were less efficient in isolating other difficult-to-dissociate cell types, such as cells of the loop of Henle.³³ In total, four studies reported an efficient tissue dissociation of mouse kidney using cold active protease, and all studies identified podocytes.^33,34,45,52 However, another study that analyzed glomeruli of mice explicitly mentioned they were unable to effectively dissociate renal cells with cold dissociation protocols, whereas an alternative warm dissociation protocol was successful.⁴² Furthermore, to date, no published study has successfully used cold active protease to dissociate human renal tissue.

Concluding Answer

The use of warm dissociation protocols in scRNA-seq experiments on human renal tissue is still recommended, because cold dissociation protocols using cold active protease have produced conflicting results in mice and have not yet been successfully used on human renal tissue. Therefore, future studies should test the performance of cold dissociation protocols on human renal tissue.

QUESTION 5: WHICH METHOD IS PREFERRED, SCRNA-SEQ OR SNRNA-SEQ?

snRNA-seq might be an appealing alternative to scRNA-seq.^{2,9,20,24,33,34,36–39,51} First, some difficult-to-dissociate cell types can be detected with snRNA-seq, because dissociation methods that isolate nuclei are stronger.³³ Second, nuclei can be dissociated on ice, thereby minimizing dissociation-induced artifacts.⁹ Third, because snap-frozen tissue can be used, snRNA-seq is compatible with historically collected biobanks of frozen samples.⁵³ Finally, although nuclei contain lower amounts of mRNA, snRNA-seq can detect an equal average number of genes per cell when compared with scRNA-seq.^9,33,54

When Podocytes Are the Focus of Research

snRNA-seq methods perform well in isolating glomerular cells, including podocytes,^{2,9,20,24,33,34,36–39,51} mesangial cells and/or glomerular endothelial cells.^{9,20,24,33,37} Three studies directly compared snRNA-seq with scRNA-seq methods.^2,9,33 In the first study on human tissue, podocytes were present in the snRNA-seq experiment, but not in the scRNA-seq experiment.² However, differences in the biopsy specimen (core needle biopsy versus nephrectomy) and renal pathophysiology (cellular rejection in renal allograft versus healthy native renal tissue) could confound this observation. A second study compared three snRNA-seq methods (sNuc-Drop-Seq, DroNC-seq, sNuc-10x) to a methanol-fixed (frozen) scRNA-seq method (Drop-seq) on healthy mouse tissue; again, all snRNA-seq methods identified podocytes, whereas the scRNA-seq method did not.⁹ However, Drop-seq is a suboptimal scRNA-seq method with a low cellular capture rate (<5%), and methanol fixation might have damaged cells, including podocytes. The third study systematically compared snRNA-seq with scRNA-seq on tissue from healthy adult mice using 10x Genomics.³³ Both techniques identified podocytes, but, surprisingly, the percentage of podocytes was lower with the snRNA-seq technique in this study (0.7% versus 3.28%).³³ So, although snRNA-seq is able to identify glomerular cells in general, scRNA-seq methods are not necessarily inferior to snRNA-seq when their experimental design is optimized. Indeed, the heterogeneity of scRNA-seq studies explains their varying performance in isolation of glomerular cells. Some studies exclusively performed scRNA-seq on magnetically isolated glomeruli from mouse kidneys and all were successful in isolating podocytes.^{11,32,40–43} In unsorted renal tissue from mice, all seven snRNA-seq studies^{9,33,34,36–39} versus half (eight of 16) of the scRNA-seq studies^{33,34,45,52,55–58} were able to isolate podocytes (Supplemental Table 5). Almost all successful scRNA-seq studies analyzed >20,000 cells in total. Four scRNA-seq studies used a cold dissociation protocol and all identified podocytes,^33,34,45,52 whereas the unsuccessful studies all used a warm dissociation protocol. In unsorted human nephrectomy tissue, all three snRNA-seq studies^2,20,24 versus half (four of eight) of the scRNA-seq studies^6,19,23,25 were able to identify podocytes (Table 3). The successful human scRNA-seq studies analyzed more nephrectomy samples (n=3–20 samples) when compared with the unsuccessful studies (n=1–3 samples). Overall, in both the snRNA-seq studies and the successful scRNA-seq studies on unsorted renal tissue from humans and mice, the percentage of podocytes to total isolated cells was extremely low (0.03%–2.78% for scRNA-seq; 0.25%–4.86% for snRNA-seq).

Table 3.

scRNA-seq or snRNA-seq studies on unsorted human nephrectomy tissue

Author	Protocol	Age of Patients	Txa	Diseaseb	Biopsyc	T°d	Cellse	Depthf	Genesg	Glomh	Remarks
Gillies et al.19,i	10× Chromium	A	N	Healthy (Tu) (n=3)	Neph (n=3)	(-), W, cryo	4734	?	?	P, M	Study on eQTL and integration with scRNA-seq; P°, 49 cells (1%)
Wu et al.2,j	inDrops	A	N	Healthy (D) (n=1)	Neph (n=1)	(+), C	4259	50,000	>400	P	P°, not mentioned
Young et al.23,i	10x Chromium	F, Ch, A	N	Fetal (n=2); healthy (D) (n=1); healthy (Tu)* (n=8); tumor* (n=8)	Neph (n=20); whole (n=2); other** (n=16)	(+), W	72,501	?	?	P, E	*Same patients; **tissue from renal pelvis, ureter, and tumor; P°, 259 cells (0.3%)
Wilson et al.20,j	10x Chromium	A	N	Healthy (Tu) (n=3); DKD (Tu) (n=3)	Neph (n=6)	(-), C, snap	23,980	6894*	2541	P, E, M	*mUMIs per cell; P°, 663 cells (2.76%)
Lake et al.24,j	snDrop-seq	A	N	Healthy (Tu) (n=14); healthy (D) (n=2)	Neph (n=19)	(+), C; (-), C, snap	17,659	1082*	589	P, E, M	*Mean transcripts per cell, not raw reads; P°, 859 cells (4.86%)
Stewart et al.25,i	10x Chromium	F, Ch, A	N	Fetal (n=6); healthy (Tu) (n=10); healthy (D) (n=3)	Neph (n=14); whole (n=6)	(+), W	40,268*; 27,203**	?	?	P, E	*Mature kidneys; **fetal kidneys; P°, 126 cells (0.3%) for F and Ch
Menon et al.6,i	10x Chromium	A	N, T	Healthy (Tu) (n=16); healthy (Pre) (n=3); healthy (S) (n=5)	CNB (n=8); neph (n=16)	(-), W, cryo	7524*; 14,744**	3971*,***; 3089**,***	1339*; 1134**	P, E, M	*CNB; ** nephrectomies; *** mUMIs per cell; P°, *11 cells (0.14%), **159 cells (1.08%)
Liao et al.27,i	10x Chromium	A	N	Healthy (Tu) (n=3)	Neph (n=3)	(+), W	23,366	Approximately 30,000	Approximately 800	—
Liu et al.95,i	10x Chromium	A	T	Chronic rejection (n=2)	Neph* (n=2)	(+), W	27,197	Approximately 2500**	200–2500	—	*Reason for nephrectomy unclear; **mUMIs per cell
Han et al.28,i	Microwell-seq	F*, A**	N	Fetal (n=4)*; healthy (D) (n=1)**; healthy (Tu) (n=2)**	Whole* (n=4); Neph** (n=3)	(+), W	22,439*; 22,692**	Approximately 858*,***; approximately 1251**,***	?	P*, M* E**	Study constructing the “human cell landscape.” *Fetal kidney; **adult kidney; ***mUMIs per cell; P°, not mentioned
Deng et al.22,i	10x Chromium	A	N	Healthy (Tu) (n=1)	Neph (n=1)	(+), W	6138	?	?	?	Study also performed Sanger sequencing on FACS-isolated PTs of two additional kidneys

Tx, tissue; T°, temperature; glom, glomerular cells; A, adult; N, native kidney; Tu, tumor-free tissue at maximal distance of mass or tumor; neph, biopsy from partial/total nephrectomy; (-), frozen tissue; W, warm dissociation; cryo, cryopreservation; ?, data could not be found in published paper; P, podocytes; M, mesangial cells; eQTL, expression quantitative trait loci; P°, absolute and relative number of podocytes isolated in scRNA-seq and snRNA-seq studies; D, healthy tissue from declined/discarded kidneys that were donated for potential transplantation; (+), fresh tissue; C, cold dissociation; F, fetal; Ch, child; tumor, tissue invaded by tumor; whole, dissection of whole fetal kidney; E, glomerular endothelial cells; DKD, diabetic kidney disease; snap, snap-frozen tissue; mUMIs, mean unique molecular identifiers; T, transplant kidney; Pre, preperfusion or pretransplant core needle biopsy of living donor kidney; S, surveillance kidney transplant biopsy; CNB, core needle biopsy; PTs, proximal tubular cells; −, no glomerular cells.

^aTissue from native kidneys or transplant kidneys.

^bHealthy or pathologic kidney tissue. n=number of patients.

^cBiopsy technique. n=biopsy samples taken.

^dFresh versus frozen tissue and warm versus cold dissociation.

^eTotal number of cells isolated and analyzed after quality control.

^fSequencing depth defined as “mean (raw) reads per cell,” “mean transcripts per cell,” or “mean UMIs per cell.”

^gMean number of genes per cell.

^hIsolation of glomerular cells.

ⁱscRNA-seq experiment.

^jsnRNA-seq experiment.

In summary, snRNA-seq methods are efficient in identifying glomerular cells, including podocytes, but their efficacy on human core needle biopsy specimens remain to be proven. scRNA-seq pipelines are less robust and are more reliant on the number of included samples, the total number of isolated cells, the dissociation protocol, and—possibly—renal pathophysiology. Because absolute podocyte counts in unsorted renal tissue with both scRNA-seq and snRNA-seq remain very low, future research should focus more on developing sorting techniques for glomeruli or individual glomerular cells to enrich for this rare cell population.

When Leukocytes Are the Focus of Research

Although efficient in identifying glomerular cells, snRNA-seq does not perform as well in isolating leukocytes, especially lymphocytes, when directly compared with scRNA-seq.^2,33 Indeed, some snRNA-seq studies could not identify leukocytes,^2,36,39 whereas others only identified macrophages^9,24,33,34; however, nearly all of these studies used healthy renal tissue or tissue from noninflammatory kidney disease. In contrast, two large snRNA-seq studies that analyzed inflammatory renal pathologies (ischemia-reperfusion injury in mice,³⁷ diabetic kidney disease²⁰) successfully isolated clusters of T lymphocytes, B lymphocytes, dendritic cells, and plasma cells.^20,37 It is possible that snRNA-seq underestimates the presence of leukocytes in the sample, but scRNA-seq might also artificially inflate leukocyte rates due to under-representation of other cell types due to incomplete cell dissociation.³³

Concluding Answer

When human glomerular cells are the focus of research and nephrectomy tissue is available, an snRNA-seq method is the first choice. Whether or not snRNA-seq can be used on renal core needle biopsy samples remains to be investigated. When using scRNA-seq to study glomerular cells, it is advised to include many samples; analyze a large number of cells (>20,000); and use successful, previously published dissociation techniques. When leukocytes are the focus of research, scRNA-seq might be preferred over snRNA-seq.

Validation Experiments in Single-Cell Transcriptomics

Single-cell transcriptomics experiments are prone to different forms of transcriptional bias. Tissue processing, (cryo)preservation, tissue dissociation, and cell-sorting techniques may all introduce transcriptional artifacts that can obscure true, biologically relevant, gene expression signals. Therefore, it is necessary to perform validation experiments on proposed marker genes and transcriptional signatures from single-cell transcriptomics experiments to confirm that they do indeed result from a true biologic signal and originate from the proposed cell type. Single-cell gene expression profiles can be validated using other well-established techniques that detect gene expression (e.g., bulk RNA-seq, RT-PCR, and RNA in situ hybridization [ISH]), as long as they are not subject to the same transcriptional bias. Alternatively, validation could focus on protein expression levels (e.g., immunohistochemistry [IHC], immunofluorescence, or emerging single-cell proteome technologies). A first in silico validation can be performed by comparing gene expression with previously published bulk RNA-seq datasets¹⁴ or protein atlases (e.g., The Human Protein Atlas,⁵⁹ http://www.proteinatlas.org).^6,24 Validation experiments using bulk RNA-seq on unsorted or undissociated tissue will help detect cell sorting– and dissociation-induced transcriptional artifacts and will also indicate whether the dissociation process introduced bias toward cell types that are dissociated more easily.^{32,33,46,49,58} RNA staining of renal tissue using ISH techniques is an alternative, lower-throughput method to validate the expression of a small set of marker genes.^6,17,21 Additionally, this method visually identifies gene expression in cells, and can, therefore, validate whether the proposed marker gene of a cell cluster is indeed visualized in the same cell type on the tissue slide. Finally, routine protein-staining techniques (IHC, immunofluorescence) can be used to visually validate whether gene expression indeed correlates with protein synthesis in the proposed cell type.^{2,18,20,21,23,24} An emerging field aims to obtain high-multiplex IHC or immunofluorescence readouts by using more innovative approaches. For instance, the CODEX platform (Akoya Biosciences) is already able to detect >40 cell markers in one staining step.⁶⁰ This method uses validated antibodies that are conjugated to a unique oligonucleotide, which hybridizes to “reporting fluorophores” in a cyclical process to finally image all of the cell markers in the tissue.⁶⁰

Future Perspectives

The combination of several multiomics measurements (e.g., genomics, epigenomics, proteomics, spatial information) on the same single cells creates an integrated view on cellular identity, cell state, and behavior.⁶¹ The different single-cell modalities and their integration in multimodal assays have recently been reviewed.^61,62 We briefly highlight three promising, multimodal, single-cell assays that combine gene expression data with (1) spatial information, (2) detection of cell surface proteins, or (3) analysis of chromatin accessibility.

Spatially Resolved Transcriptomics

Many glomerular diseases have a focal distribution and the integration of spatial information with single-cell transcriptomics might, therefore, help to identify cell (sub)populations and cell states in diseased microenvironments. Currently, there are four major spatially resolved transcriptomics techniques⁶³: (1) computational reconstruction of the spatial information of dissociated tissue^64–66; (2) single-cell analysis of cells or tissue of interest isolated through laser-capture microdissection; (3) high-throughput in situ transcriptomics (e.g., fluorescence-ISH techniques and in situ sequencing); and (4) spatial barcoding techniques (e.g., Slide-seq⁶⁷ and spatial transcriptomics,⁶⁸ which is now commercialized as the Visium Spatial Gene Expression Solution by 10x Genomics). To date, two studies have used a computational method (de novo Spatial Reconstruction of Single-cell Gene Expression, NovoSpaRc)⁶⁶ on single-cell or single-nucleus datasets of dissociated mouse renal tissue to reconstruct the spatial position of single cells on the basis of transcriptome similarity.^34,66 The landmark study that introduced the Slide-seq technique demonstrated the spatial distribution of tubular cells and podocytes in a mouse kidney,⁶⁷ and another recent study used the Visium platform (10x Genomics) in a murine endotoxemia model to map the location of a proximal tubular cell subtype (S3 type 2) in the mouse kidney.⁶⁹ More studies using in situ sequencing or spatial barcoding techniques on renal tissue are anticipated in the near future.

Integrated Detection of Cell Surface Proteins

The simultaneous measurement of single-cell gene expression and cell surface proteins is possible with the Cellular Indexing of Transcriptomes and Epitopes by sequencing (CITE-seq) and RNA Expression And Protein sequencing (REAP-seq) assays.^70,71 Using DNA-barcoded antibodies against cell surface proteins, both the intracellular mRNA and antibody barcodes are converted into cDNA and sequenced.⁶¹ To date, no studies using these techniques on non-tumor renal tissue have been published.

Integrated Analysis of Chromatin Accessibility

The study of the epigenome and its correlation with the transcriptome on a single-cell level creates a deeper understanding of gene expression regulation.^61,62 A popular technique to study chromatin accessibility in bulk cells is the assay for transposase-accessible chromatin using sequencing (ATAC-seq), in which DNA from open chromatin is tagged and excised by barcoded transposase Tn5, after which the excised DNA fragments are sequenced.⁷² Analogous to transcriptomics, single-cell methods for chromatin accessibility emerged with single-cell ATAC-seq, which is compatible with existing single-cell microfluidic devices, including plate- and droplet-based methods.^73,74 Alternatively, a combination of barcodes can be used to identify single cells, as used in the sci-ATAC-seq protocol, which uses a single-cell combinatorial indexing (“sci”) strategy.⁷⁵ One study provided a single-cell atlas of chromatin accessibility from 13 different tissues in adult mice, including renal tissue, using this sci-ATAC-seq technique.⁷⁶ Integrated multiomics assays that combine single-cell transcriptomics with chromatin accessibility exist in different protocols, including sci-CAR,³⁶ scCAT,⁷⁷ Paired-seq,⁷⁸ SHARE-seq,⁷⁹ and SNARE-seq,⁸⁰ and have been commercialized in the Chromium Single Cell Multiome ATAC + Gene Expression assay. To date, only one study has used a combined chromatin accessibility and gene expression assay (sci-CAR) on adult mouse renal tissue.³⁶

Conclusions

Single-cell research has gained traction in the field of nephrology and has already proven to be a powerful and useful tool to dissect the (patho)physiology of healthy kidneys and glomerular diseases. This review addressed five key methodologic questions that arise when designing single-cell transcriptomics experiments that focus on glomerular cells. Currently, several challenges remain when core needle biopsy samples are used for scRNA-seq experiments, and no data are yet available on the performance of snRNA-seq protocols. Single-cell transcriptomics experiments on frozen renal tissue are feasible using three possible approaches. To date, no efficient techniques are available that can enrich for glomeruli or podocytes in human renal tissue before dissociation, and the creation of a high-quality single-cell suspension remains the Achilles heel of scRNA-seq, with cold dissociation protocols resulting in less artifacts but also less efficient cell dissociation. Finally, the choice between scRNA-seq or snRNA-seq also depends upon the cells of interest: when glomerular cells are the focus of research, an snRNA-seq method might be preferable, whereas scRNA-seq methods may be superior when profiling immune cells. The delay between biopsy and histopathologic diagnosis currently hampers the widespread use of fresh core needle biopsy specimens and future studies should, therefore, further optimize snRNA-seq and frozen-tissue pipelines and glomerular cell enrichment techniques applicable to human renal core needle biopsy specimens. The field of single-cell transcriptomics is rapidly evolving, and combining these techniques with several other single-cell measurements and multiomics assays will undoubtedly create new insights in the complex pathophysiology of glomerular diseases.

Disclosures

M. Naesens reports serving as an advisor for the European Medicines Agency (EMA), and on the editorial boards of several journals. B. Sprangers reports serving as an ad hoc expert for the EMA. All remaining authors have nothing to disclose.

Funding

D. Deleersnijder holds a KU Leuven Special Research Fund PhD scholarship (DB/19/010/BM. J. Callemeyn is supported by a Fonds Wetenschappelijk Onderzoek (Research Foundation Flanders) fellowship grant, number 1196119N. B. Sprangers is a Fonds Wetenschappelijk Onderzoek senior clinical investigator, via grant 1842919N. M. Naesens is a Fonds Wetenschappelijk Onderzoek senior clinical investigator, via grant 1844019N.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2021020157/-/DCSupplemental.

Supplemental Table 1. Overview of all studies reporting on scRNA-seq and/or snRNA-seq on mouse tissue

Supplemental Table 2. Overview of all studies reporting on scRNA-seq and/or snRNA-seq on human tissue

Supplemental Table 3. scRNA-seq studies on methanol-fixed or cryopreserved tissue of mouse kidneys

Supplemental Table 4. scRNA-seq studies on cryopreserved human renal tissue

Supplemental Table 5. scRNA-seq or snRNA-seq studies on unsorted mouse renal tissue

Supplemental Figure 1. Literature search strategy

Supplemental Figure 2. PRISMA 2009 flow diagram of the literature search strategy