Conundrums of choice of “normal” kidney tissue for single cell studies

Abstract

Purpose:

Defining molecular changes in key kidney cell types across lifespan and in disease states is essential to understand the pathogenetic basis of disease progression and targeted therapies. Various single cell approaches are being applied to define disease associated molecular signatures. Key consideration include the choice of reference tissue or “normal” for comparison to diseased human specimens and a benchmark reference atlas. We provide an overview of select single cell technologies, key considerations for experimental design, quality control, choices and challenges associated with assay type and source for reference tissue.

Recent Findings:

Several initiatives including Kidney Precision Medicine Project, Human Biomolecular Molecular Atlas Project, Genitourinary Disease Molecular Anatomy Project, ReBuilding a Kidney consortium, Human Cell Atlas and Chan Zuckerburg Initiative are generating single cell atlases of “normal” or disease kidney. Different sources of kidney tissue are used as reference. Signatures of injury, resident pathology and procurement associated biological and technical artifacts have been identified in human kidney reference tissue.

Summary:

Committing to a particular reference or “normal” tissue has significant implications in interpretation of data from disease samples or in ageing. Voluntarily donated kidney tissue from healthy individuals is generally unfeasible. Having reference datasets for different types of “normal” tissue can aid in mitigating the confounds of choice of reference tissue and sampling biases.

Introduction

Single cell technologies are increasingly being adopted by the kidney community to understand the biology in healthy and disease states [1–9]**. The high content and sensitive nature of these technologies also creates a chance of introducing artifacts and sources of variation that can confound the interpretation of results [7*,10*]. Unlike studies in well-controlled preclinical model systems, a healthy or “normal” reference tissue is not feasible to obtain form humans without posing significant risks including death. Therefore, other sources for obtaining kidney tissue that can be used as reference for disease samples is generally used. There are potential sources of variation from procurement, preservation, processing and background physiological changes present in reference tissue processed for single cell data that need to be considered if these tissues are to be used as reference to compare with disease and have not been formally addressed in the field. Follows is an overview of different single cell technologies, key areas in single cell technology pipeline to consider minimizing sources of variations, advantages and disadvantages of various types of reference tissue being used and quality control criteria to minimize technical and random variation. Despite the limitations associated with different types of reference tissue, potential solutions are suggested to leverage these resources and derive biological insights. We will primarily focus on single cell gene expression methods and application to human kidney tissue.

Overview of single cell experimental pipeline

For this review we sue single cell to represent data generated from either single cell or single nucleus sequencing. Since running a single cell experiment involves highly complex process, we provide an overview of many of the key points to consider in this process first.

Steps in generating single cell gene expression data.

Conducting single cell experiments require thorough planning and defining the goals of the experiment. There are several steps in the process of generating single cell gene expression data (Figure 1a). In this “follow the tissue” pipeline, the steps include tissue type selection and size, derived products, assay to be performed, measurements that are made during the assay, analytical methods and dissemination of the results to the community.

Figure 1.

Illustration shows the various steps involved in the process of conducting single cell experiment. A) Process shows various steps from tissue procurement to sharing data. (B) Illustration shows examples of choices that are available at each of the steps depicted in part A. (C) Illustration shows that for each choice in part B there are challenges or limitations thus underscoring the need for carefully planning the objectives of the single cell experiment.

Conundrum with choices.

In each of the steps in the pipeline depicted in Figure 1a, there are choices that impact the outcome. Therefore, it is critical to first identify the goals of the study that will use human reference tissue. For example, the objectives may range from surveying cell diversity, identifying gene networks, profiling as many cells possible (cell throughput), identifying rare cells or pathobiology and this will determine what reference tissue is used (see later). In Figure 1b, we illustrate a few examples of choices that may be available or selected for each of the steps in the single cell experimental pipeline. For example, if one needs to generate a comprehensive cell diversity atlas then wider sampling and tissue sources where large amount of tissue is available than a typical biopsy may be preferred. On the other hand, if determining cellular composition in a disease sample that can be correlated with clinical and pathological information then it may be preferable to use biopsy tissue. However, smaller number of datasets will be generated and differences in sampling may need to be considered in interpretation. Further, the technology chosen here needs to be compatible in working on small biopsy tissue.

Challenges in choices made.

The choices made will impact feasibility and pose challenges at each step (Figure 1c). For example, well-preserved biopsy tissue should be available that is amenable for single cell sequencing if biopsy is the source tissue. Due to limited tissue available from biopsies, rare cells may be missed and this may not be desirable if these are important in contributing to pathobiology or the focus of the study.

Spectrum of technologies.

Several single cell technologies have emerged in the last ten years. They range from targeted interrogation of expression of select genes in a few cells to genome wide transcriptome analysis of thousands of cells or nuclei. The assays vary in depth of sequencing, extent of coding region covered, method of isolation of single cell or nuclei and costs per cell (Table 1) and have advantages and disadvantages [11*–13*]. Therefore, identifying the desired outcome will aid in selecting the assay best suited for purpose. For example, if the goal is to identify RNA isoforms, then a method covering as much of the coding region should be used compared to droplet-based approaches. On the other hand, if the purpose is to determine cellular diversity, then high content droplet-based methods such as 10X Chromium may be preferable. The cost may also determine the use of a commercial source or an in-house assay and source of tissue used.

Table 1.

Single cell technologies and characteristics. This table is partly derived from [11,13,27].

Technology/ parameter	C1 (SMARTer)	Smart-seq2	MATQ-seq	MARS-seq	CEL-seq	Drop-seq	InDrop	10X Genomics	SEQ-well	SPLIT-seq
Transcript coverage	Full length	Full length	Full length	3’-end counting	3’-end counting	3’-end counting	3’-end counting	3’ or 5’-end counting	3’-end counting	3’-end counting
Platform	Microfluidics	Plate-based	Plate-based	Plate-based	Plate-based	Droplet	Droplet	Droplet	Nanowell array	Plate-based
Throughput (number of cells)	102–103	102–103	102–103	102–103	102–103	103–104	103–104	103–104	103–104	103–105
Typical read depth/cell	106	106	106	104–105	104–105	104–105	104–105	104–105	104–105	104
Genes/cell	~10,000	~5000	~800	~2000	~5000	~800	~800	~2000	~5000	~800
Reaction volume	Nanoliter	Microliter	Microliter	Microliter	Nanoliter	Nanoliter	Nanoliter	Nanoliter	Nanoliter	Microliter
Cost/cell	$$	$$$$	$$$	$$$	$$$	$	$	$	$	$
Reference	[28]	[29]	[30]	[31]	[32]	[33]	[34]	[35]	[36]	[27]

Choices of single cell versus single nucleus RNA-seq.

Single cell RNA-seq data for creating cell diversity maps can be generated from isolated cells or nuclei. There are advantages and disadvantages of each approach [2**,4**,14,15*]. Both technologies work on fresh and stored frozen tissue. Cells harbor more than 10-fold amount of mRNA than the nucleus. Whole cell RNA has more mature transcripts, closer to being translated into a protein product. However, nuclear RNA consists of pre-mRNA and mature mRNA thus earlier in the process of transcription to translation. One advantage of using cells is that specific populations can be isolated using immunoselection procedures and then used for single cell experiments. However, in solid tissues such as the human kidney, harsher dissociation steps, often using enzymes, are needed to isolate cells that can activate cell stress pathways and affect viability. Epithelial cells are particularly sensitive to ischemia and dissociation and their isolation can result in increase in ambient RNA. In this regard, nuclear membrane is more resilient to isolation procedures and helps maintain nuclear RNA intact. In the Kidney Precision Medicine Project (KPMP) [16], we developed procedures that allow snRNA-seq on biopsy-equivalent tissue using frozen OCT-embedded tissue blocks that completely eliminate enzymatic dissociation step and produces minimum stress artifacts compared to other preservation methods [3]. We have found that compared to scRNA-seq, snRNA-seq achieves broad representation of epithelial and stromal cell populations while single cell isolation procedures lead to more versatile immune cell representation [2**,4**,17]. Thus, both technologies provide overlapping and complementary information and the choice should be made judiciously.

Quality assurance and control in single cell technologies.

Single cell technologies generate a vast amount of gene expression data that is highly sensitive to technical and procedural variations as small changes can cause large differences in results. This affects reproducibility and interpretation of results wasting precious resources and time. It is imperative that rigorous procedures are in place to assure high quality tissue and data are obtained consistently. Below are key criteria that we have established in the Human Biomolecular Molecular Atlas Project (HuBMAP) [18] and KPMP [16] for single cell atlas [19**].

Quality assurance and control in enrolment, tissue procurement and preservation.

It is important to ensure that rigorous procedures are in place to procure and preserve tissue in a timely manner for single cell assays or other complementary technologies. These high content datasets can best serve the community if both raw and processed data are available, therefore. Best efforts should be made to consent patients for broad sharing and open access with mechanisms in place to protect patient confidentiality. Tissue procurement and relevant clinical parameters should be recorded. For example, ischemia times, time from harvest to preservation and type of preservation and functional state of the organ should be available. These parameters can significantly affect cell types isolated, genes detected and injury profiles that can confound biological interpretation. Ideally, preservation method should be compatible with several omics or spatially resolved technologies to yield maximum biological information and orthogonal validation. In KPMP we have developed methods to use fresh frozen, OCT-embedded biopsy for dual snRNAseq-snATACseq, histology, CODEX, spatial transcriptomics, regional proteomics and imaging cytometry

Quality assurance and control in tissue morphology and cell preparations.

An adjacent section of the material used for cell or nuclei isolation should be evaluated for tissue integrity, morphology, composition and rule out non-kidney tissue contaminants to aid in interpretation of cell types and avoid wasting precious resources. For example, if tissue is overly fragmented, hemorrhagic, necrotic or has non-kidney tissue (skeletal muscle, fat or blood clot) it should be discarded. The isolated cells or nuclei should be evaluated for viability or morphology and yield needed to run the assay successfully. For example, the nuclei should be round, with minimal debris in the background and ideally less than10% forming clumps.

Quality assurance and control in assay pre-sequencing.

The assay procedures should clearly define the threshold used for the cDNA and libraries generated. If possible, methods should include kits used, controls for sequencing and a reference control that can be periodically used in each batch to ensure the assay generally worked. This will also help gauge assay drift and performance over time. The instruments should be routinely calibrated with controls (most core facilities perform this).

Quality assurance and control in post-sequencing data.

The experimenter should establish quality control criteria that can be applied to all samples. These could range from quality of sequencing reads (for example Q30 > 85%), number of cells, mapped reads per cell, sequencing depth, percent genome mapped and genes per cell [19**].

Quality assurance and control in data analaytics.

Finally, uniform QC cutoffs should be applied to all data that is desired to be analyzed together ensuring that quality of data from each cell is high. For example, for single nuclear RNA sequencing we assess genes, mitochondrial reads, ER stress, injury markers for each nuclei. It is also advisable to use the same human reference genome version for mapping. An additional step we include is to monitor assay drift over time by evaluating key parameters for all samples using Levey-Jennings plot (with 2 SD around the mean as a guide to detect drift). For example, one could use genes per nucleus as a parameter and monitor how it varies across samples over time. Deviations more than 2 SD should prompt a thorough evaluation of the metadata in the pipeline at each step to help troubleshoot if a decline is a general trend or related to only one sample due to low quality or biology.

Types of kidney tissue used as a reference source for healthy atlas

Choice of kidney tissue used for reference should be considered keeping in mind the goals of the experiment as summarized in Figure 1. I will discuss four different types that have been used in omics studies by various groups and highlight advantages and limitations of each (Table 2).

Table 2.

Overview of pros and cons of different sources of adult healthy reference kidney tissue.

Category	Deceased Donor	Surgical tumor free	Living Donor	Biopsy (stone)
Region sampling	Broad, can span entire kidney	Broad, but limited to tumor free areas	Limited	Limited
Cortico-medullary depth	Yes	Yes	No	No
multiple omics assays on same tissue	Yes	Yes	Limited	Limited
Volumetric imaging	Yes	Yes	Limited	Limited
Cell Diversity	High	High	Medium	Medium
Ease of tissue availability	Medium	High	Low	Low
Risk to patient for research tissue	None	None	High	Minimal
Stress Injury (procurement ischemia)	Yes	Yes	Limited	Limited
Medical intervention confounds	Yes	Yes	Limited	Limited
Representation of physiological state	Medium	Medium	High	High

Deceased donor kidney.

Kidneys from deceased donors that could not be used for transplant are a common source of kidney tissue. Donors selected for transplant of various organs are generally healthy but sometimes the kidneys are not suitable for transplant due to underlying pathology in the kidney such as percent glomerulosclerosis and acute kidney injury or unusable due to other reasons including unfavorable aortic cuff, history of certain viruses, age and procedural errors (degloving or contamination of the field) and are discarded. If carefully selected (see later) these kidneys can be very useful as a reference for research. One of the key advantages is that the whole kidney and sometimes both kidneys are available for sampling of different anatomic regions (outer and inner cortex, outer and inner medulla including papilla, different poles). Multiple samples can be acquired for different technologies that span micro to macroscopic scales. This allows for optimization of protocols, determining differences due to location in the kidney so variations due to sampling can be assessed and preservation in several ways for current and future technologies that may not be apparent at present. Both discarded donor kidneys and those that are selected for transplant may sometimes be sampled through a cortical wedge biopsy for pathological evaluation. These are typically fresh frozen and left over tissue can be used for research. However, there is potential time lag due to transport from organ procurement to the pathology lab that may affect molecular profiles. Deceased donor kidney tissue is generally well-preserved due to perfusion with preservation media routinely used in organ transplant thus minimizing warm ischemia time. A disadvantage of this source of tissue is that circumstances around death may not be entirely known and the patients may have gone through medical interventions during resuscitation and support; availability of detailed metadata here is critical to interpret biology.

Kidney tissue from surgical resection.

One of the common sources for healthy reference kidney tissue is non-tumor tissue from kidneys resected (full or partial) from patients undergoing nephrectomy. The tumor size varies from a few centimeters to more than 10 cm. However, there is typically plenty of tumor-free kidney tissue that is not needed for pathological assessment. Nephrectomy tissue is an important source of reference tissue in patients with no clinical kidney disease or evidence of histopathological abnormalities. The tissue availability for research is usually not limited thus sampling can be done from various regions as long as they are distant from the tumor to minimize compression artifacts and tumor microenvironment. Gross examination should include ruling out obstruction of the pelvis-calyces by the tumor from the regions being sampled. Since renal tumors occur across life-span, affects both sexes and all races, tissue obtained can represent much broader biological diversity than other sources. Additional advantage is nephrectomy tissue can also serve as an important internal reference for potential future prediction or risk of kidney disease in these patients. Compared to deceased donor nephrectomy tissue, tumor-free reference tissue has fewer confounding factors related to medical interventions including life-support that occur in the former. However, a concern is underlying cancer and ischemia associated with surgical procedure (usually less than 30 min at our center). Detailed microscopic assessment of the reference tissue from nephrectomies should be done to rule out significant pathology.

Living donor biopsy tissue.

Healthy reference kidney tissue can also be obtained from living donors donating kidneys to patients with end stage kidney disease. These patients have undergone extensive screening for their suitability as healthy donors and normal kidney function. This source provides the closest to healthy reference tissue one can obtain due to minimal confounds of other existing medical conditions or procedures, have minimal pathology. Special consent may be needed as there is a defect created due to biopsy with risks of bleeding in the recipient. A needle core biopsy (one or two cores) is usually taken once the donor kidney is out of the body and in some cases at time zero after the kidney is transplanted in the recipient [5]. The KPMP has a mechanism established to procure healthy tissue from living donor kidneys that is being used as a source of reference for AKI and CKD. The limitations of amount of tissue sampled, exact coordinates and ischemia/anoxia before a biopsy is procured/preserved still remain.

Reference kidney tissue from patients with kidney stones.

Sometimes patients with stones in the papilla or obstructing one of the calyces undergo a percutaneous endoscopic procedure that provides access to renal cortical tissue. In this procedure access to the pelvis is through renal parenchyma via a port that is distant from the region impacted by the stone. In these cases due to a surgical defect created by the procedure there is access to uninvolved kidney tissue when the port is removed. If patient consents, a small piece of cortical tissue (less than 3 mm) can be sampled for research. Because often this procedure is guided by imaging, the location of the region sampled can be ascertained. With normal kidney function and histology this can be another important source of healthy tissue as there is minimal anoxia/ischemia and was recently used to assess papillary stone disease [20]. There is a potential confound of mechanical stress created on the tissue sampled due to access port; these patients could also have cortical disease not known before. Healthy reference tissue from these patients is also being used in KPMP to construct a healthy kidney tissue atlas. Note that this procedure also provides limited sampling of kidney regions.

Utilizing kidney reference tissue to inform single cell and spatial kidney maps in disease samples

The sources of reference tissue provided above have all been used or actively being procured for generating single cell molecular and spatial healthy reference kidney tissue atlas by individual labs or in large consortia efforts including KPMP [16], HuBMAP [18], HCA [21], RBK [22], GUDMAP [23] and SenNet [24]. The anatomical organization of the kidney into repetitive lobes of cortical-medullary pyramids and nephrons suggests that physiologically there may not be significant differences in key filtration, absorptive and secretory functions of the kidney. However, there are anatomical variations in both left and right kidney and the regional representation is quite heterogeneous due to complexity of cell types in the kidney. Further, it is currently not clear if there are molecular differences among similar regions (example, cortex) sampled from different kidney lobes; molecular profiles may be more sensitive to local environment and physiology. Since disease biopsy samples may be quite heterogeneous in sampling, a robust reference atlas should represent as many different kidney regions as possible for appropriate comparative analysis. Further, it is not uncommon to see a biopsy aimed to get cortical tissue sample significant portion of medulla. Since some cell types in cortex and medulla may share markers, it will be optimal to have a benchmark reference that includes both cortical and medullary cell types for appropriate interpretation of disease phenotypes that may manifest in the medulla. Having a wide representation of reference from different anatomical locations thus helps better representation of kidney cell types and comparisons to disease biopsies for gaining insights into biology. Ideally a reference atlas should be constructed with all these tissue sources so users can choose the most appropriate comparison. We expect that the atlas efforts of HuBMAP and KPMP will expand existing data [4] from several samples from above sources of reference tissue to differentiate procurement related molecular effects from biological effects and adequate spatial and regional annotation of disease samples from tissue biopsies.

Several sources of reference kidney tissue can help in identifying common sources of technical variation due to procurement and processing. In recent joint HuBMAP and KPMP kidney atlas efforts the authors had described reference and altered states. These altered states could represent many different possibilities including physiological, disease associated or technical artifacts. One example is the degenerative state where the cells have lost differentiation markers and are enriched in cell stress genes and injury marks. Some of these changes could be due to acquisition and processing methods. Once identified, validation with orthogonal means such as spatial imaging or transcriptomic technologies can help in distinguishing these possibilities [25]. Since certain amount of ischemia in procuring reference tissue from any of the sources discussed above is unavoidable, one can expect some degree of acute kidney injury in any sample. As such one important outcome of these altered acute injury state cell identities can inform on early time course of AKI in human samples. This is relevant since biopsies showing AKI changes are rarely done during the early state of injury due to a number of factors including timing of patient presentation and that biopsies for AKI are not typically indicated. To gain confidence in early ischemic injury associated genes in in kidney tissue correlation with AKI mouse models with time course data can be useful. An example of this was done recently using mouse IRI time course ranging from few hours after AKI to 6 weeks [4**,26**] and used in human kidney reference atlas to find early AKI associated cell states and genes in reference and disease samples.

Conclusion

Cataloguing cellular identities based on molecular signatures and standardized nomenclature to define mature healthy state from altered states due to injury including increased cell stress markers, low quality due to increased mitochondrial fraction, increased ER stress and low genes per cell in reference tissue that do not validate in at least two different omics methods is a plausible method to distinguish biological signals from procurement/processing associated artifacts. Overlapping cellular identities observed in different technologies for different reference tissues increase the confidence in cell annotations of healthy cells and is therefore encouraged. This way pitfalls in kidneys obtained from deceased donors and nephrectomies can be minimized and providing sufficient datasets with extensive sampling to deliver robust artificial intelligence and machine learning tools to the community and more robust benchmark atlases.

Key points.

• Several complex steps in performing single cell experiments

• Several sources of biases and technical variations exist in each of the steps in a single cell tissue analysis pipeline and these should be carefully considered before starting an experiment

• Tissue procurement and processing are one of the key factors that impact the output of single cell experiments

• Rigorous quality control procedures should be in place with meticulous preanalytical metadata to ensure the artifacts and variations introduced are minimized and interpretable

• All sources of reference kidney tissue have advantages and disadvantages that vary from ischemic injury to sampling biases.