What was used for the data normalization and scaling?

‘SCTransform’ was used for data normalization and scaling (based on top 2,000 features), followed by principal component analysis.

What is the role of fatty acid oxidation in tubulointerstitial ?

Decrease in fatty acid oxidation, resulting in a loss of ATP generation, has been shown to be a significant contributor to tubulointerstitial fibrosis 19.

How many libraries were needed to reidentify podocytes?

On average 12 and 15 libraries (~3,100 and 3,835 nuclei) allowed reidentification of seven of the top 10 predicted podocyte and proximal tubule MBCO SCPs, respectively, while 21 libraries (~5,462 nuclei) were sufficient to reidentify five of the topwas not certified by peer review) is the author/funder.

How many jensenlab confidences were used to identify each gene?

Subcellular localization of each gene was identified using the jensenlab human compartment database based on a jensenlab confidence of at least four (i.e. 80% of maximum confidence in the database) 28.

How many libraries are needed for a consistent detection of podocytes?

Their results indicate that for a consistent detection of podocytes (i.e. in more than 95% of all down sampled datasets with the same library counts), at least 16 (~11,727 cells) or 7 libraries (1,837 nuclei) are needed if subjected to single-cell RNASeq (Figure 4A) or single-nucleus RNASeq (Figure 4B), respectively.

How many differentially expressed genes and proteins were predicted by each assay?

Top 300 differentially expressed genes (DEGs) and proteins (DEPs) predicted by each assay for each analyzed cell type/tissue subsegment.

How many SCPs can be included in the top seven predictions?

Notice that the top seven predictions based on dynamic enrichment analysis can contain more than seven SCPs, since each prediction is either a single SCP or a unique combination of two or three SCPs.

How many samples were sufficient to reproduce the results for the full dataset?

For the LMD proteomics dataset, six to eight samples were sufficient to reproduce the results obtained for the full datasets with only minor variations in the correlation of identified DEGs (Figure 4C) and SCPs (Supplementary Figure 2E) or SCP rankings (Figure 4C).

What is the way to integrate the three different assays?

An idealized integration scenario would combine these assays synergistically such that they could complement the shortcomings of each other, improve quality control metrics across technologies, and increase rigor and reproducibility of the overall study.

How many predictions are needed to re-identify the top 10 or seven predictions?

the authors determined how many SCPs have to be considered in a down-sampled analysis to re-identify at least 70% (or 50%) of the top 10 or seven predictions obtained from standard or dynamic enrichment analysis with the full dataset, respectively.

What is the correlation between the gene expression profiles of cells and LCM segments?

To compute the Pearson correlation between the gene expression profiles of cells and LCM segments, the gene profiles were restricted to genes shared between the two datasets and showing variable expression in the single-cell dataset and correlations were computed between the logarithm of the mean ratio vector for each LCM segment and the scaled expression profile of each cell in the single cell dataset.

(Open Access) A reference tissue atlas for the human kidney (2021) | Jens Hansen

Q: What are the future works mentioned in the paper "Towards building a smart kidney atlas: network-based integration of multimodal transcriptomic, proteomic, metabolomic and imaging data in the kidney precision medicine project" ?

Their approach is amendable to future computational modeling studies that can further improve the proposed tissue atlas. In addition to the integrated analytics presented here, the KPMP is also building a community-based Kidney Tissue Atlas Ontology ( KTAO ), which will systematically integrate different types information ( such as clinical, pathological, cell and molecular ) into a logically defined tissue atlas, which can then be further utilized to support various applications 34.

Towards Building a Smart Kidney Atlas: Network-based integration of multimodal

transcriptomic, proteomic, metabolomic and imaging data in the Kidney Precision

Medicine Project

Jens Hansen

1,*

, Rachel Sealfon

2,*

, Rajasree Menon

3,*

, Michael T. Eadon

, Blue B. Lake

Becky Steck

, Dejan Dobi

, Samir Parikh

, Tara K. Sidgel

, Theodore Alexandrov

, Andrew

Schroeder

, Edgar A. Otto

, Christopher R. Anderton

9,10

, Daria Barwinska

, Guanshi Zheng

Michael P. Rose

, John P. Shapiro

, Dusan Velickovic

, Annapurna Pamreddy

, Seth

Winfree

, Yongqun He

, Ian H. de Boer

, Jeffrey B. Hodgin

, Abhijit Nair

, Kumar Sharma

Minnie Sarwal

, Kun Zhang

, Jonathan Himmelfarb

, Zoltan Laszik

, Brad Rovin

, Pierre C.

Dagher

, John Cijiang He

, Tarek M. El-Achkar

, Sanjay Jain

, Olga G. Troyanskaya

2,#

Matthias Kretzler

3,#

, Ravi Iyengar

1,#

, Evren U. Azeloglu

1,#

for the Kidney Precision Medicine

Project Consortium

* Contributed equally, joint first authors

Affiliations:

1. Icahn School of Medicine at Mount Sinai, New York, New York

2. Princeton University, Princeton, New Jersey and Flatiron Institute, New York, New York

3. University of Michigan School of Medicine, Ann Arbor, Michigan

4. Indiana University School of Medicine, Indianapolis, Indiana

5. University of California San Diego, Jacobs School of Engineering, San Diego, California

6. University of California San Francisco School of Medicine, San Francisco, California

7. Ohio State University College of Medicine, Columbus, Ohio

8. European Molecular Biology Laboratory, Heidelberg, Germany

9. Pacific Northwest National Laboratory, Richland, Washington

10. UT-Health San Antonio School of Medicine, San Antonio, Texas

11. University of Washington, Schools of Medicine and Public Health, Seattle, Washington

12. Washington University in Saint Louis School of Medicine, St. Louis, Missouri

Corresponding Authors, joint senior authors:

Evren U. Azeloglu, Ph.D.

Assistant Professor of Medicine, Nephrology

Icahn School of Medicine at Mount Sinai, New York, NY

Email: evren.azeloglu@mssm.edu

Twitter: @azeloglu

Ravi Iyengar, Ph.D.

Dorothy H and Lewis H Rosenstiel Professor of Pharmacological Sciences

Icahn School of Medicine at Mount Sinai, New York, NY

Email: ravi.iyengar@mssm.edu

Matthias Kretzler, M.D.

Professor of Medicine, Nephrology

University of Michigan School of Medicine, Ann Arbor, MI

Email: kretzler@med.umich.edu

Olga Troyanskaya, Ph.D.

Professor of Computer Science

Princeton University, Princeton, NJ

Email: ogt@genomics.princeton.edu

The copyright holder for this preprint (whichthis version posted July 24, 2020. ; https://doi.org/10.1101/2020.07.23.216507doi: bioRxiv preprint

ABSTRACT

The Kidney Precision Medicine Project (KPMP) plans to construct a spatially specified

tissue atlas of the human kidney at a cellular resolution with near comprehensive molecular

details. The atlas will have maps of healthy, acute kidney injury and chronic kidney disease

tissues. To construct such maps, we integrate different data sets that profile mRNAs, proteins

and metabolites collected by five KPMP Tissue Interrogation Sites. Here, we describe a set of

hierarchical analytical methods to process, combine, and harmonize single-cell, single-nucleus

and subsegmental laser microdissection (LMD) transcriptomics with LMD and near single-cell

proteomics, 3-D nondestructive and immunofluorescence-based Codex imaging and spatial

metabolomics datasets. We use nephrectomy, healthy living donor and surveillance transplant

biopsy tissues to create a harmonized reference tissue map. Our results demonstrate that

different assays produce reliable and coherent identification of cell types and tissue

subsegments. They further show that the molecular profiles and pathways are partially

overlapping yet complementary for cell type-specific and subsegmental physiological

processes. Focusing on the proximal tubules, we find that our integrated systems biology-

based analyses identify different subtypes of tubular cells with potential for different levels of

lipid oxidation and energy generation. Integration of our omics data with pathways from the

literature, enables us to construct predictive computational models to develop a smart kidney

atlas. These integrated models can describe physiological capabilities of the tissues based on

the underlying cell types and pathways in health and disease.

The copyright holder for this preprint (whichthis version posted July 24, 2020. ; https://doi.org/10.1101/2020.07.23.216507doi: bioRxiv preprint

INTRODUCTION

The kidney is one of the most diverse organs in the human body in terms of its cellular

heterogeneity, and possibly second only to the brain in its spatial complexity. Accordingly,

decoding the functional and pathogenic mechanisms of kidney disease has been challenging;

as such, nephrology has consistently ranked behind all other subspecialties of medicine in

terms of the drug discovery pipeline

. Delineating the cell types and subtypes in different

regions of the kidney during health and disease will help identify the tissue-level, cellular and

subcellular pathways and processes involved in disease initiation and progression, and aid in

drug discovery.

The Kidney Precision Medicine Project (KPMP) is a consortium funded by the National

Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) that aims to ethically and

safely obtain kidney biopsies from participants with chronic kidney disease (CKD) or acute

kidney injury (AKI); create a reference kidney atlas; characterize disease subgroups to stratify

patients based on molecular features of disease; and identify critical cells, pathways, and

targets for novel therapies and preventive strategies. The KPMP features an expanding set of

complementary set of high throughput assays for molecular entities that span transcriptomic,

proteomic, metabolomic profiles and spatial/structural properties of kidney tissue. These

assays, described here for the five initially funded Tissue Interrogation Sites (TISes), will be

integrated to create a comprehensive knowledge environment for the human kidney. This

knowledge environment will be compiled by the KPMP Central Hub to serve as a foundation

for a spatially specified interactive smart tissue atlas that will include molecular and

physiological information on healthy and diseased states of all individual cell types within the

adult human kidney.

The KPMP envisions that harmonization and integration of different types of molecular data

from omics assays, combined with state-of-the-art pathological and clinical descriptors, will

allow us to classify different disease subtypes and states for diagnostic and therapeutic

purposes. Numerous groups have proposed the use of integrated multiomics analysis to

characterize disease phenotypes using tools that include Bayesian, correlative, network-based

and machine learning-based clustering algorithms

2-4

. The goals of these approaches include

prediction of clinical outcomes, identification of underlying disease mechanisms and

stratification of patients

. KPMP further envisions that the final integrated analytical

environment will serve as a knowledge base for the entire field that will empower a molecular

anchored outcome prediction and development of targeted treatments.

Here, we present an overview of KPMP’s strategies to harmonize and integrate multiple

data types through identification of subcellular pathways and functions that delineate cell-level

biochemical and physiological functions. Using reference kidney pilot tissue samples, we have

performed data harmonization and integration to investigate the complementarity of different

data types and develop a pipeline for the generation of tissue maps.

RESULTS

Outline of KPMP Data Types

In these analyses, there were four transcriptomic, two proteomic, one imaging-based, and

one spatial metabolomics tissue interrogation assays that consisted of 3 to 48 different

datasets obtained from 3 to 22 participants (Supplementary Table 1). These assays and their

detailed tissue pre-analytical, tissue processing, data acquisition and analytical data

processing pipelines are outlined in Figure 1. We also summarize the steps whereby the data

sets were integrated and harmonized in the upper right side of this descriptive map view of the

KPMP data integration paradigm.

The copyright holder for this preprint (whichthis version posted July 24, 2020. ; https://doi.org/10.1101/2020.07.23.216507doi: bioRxiv preprint

Pathway- and network-level integration of multiple molecular interrogation techniques

reveals cell- and tissue-specific biological processes that are critical for renal

physiology

To overcome the inherent challenges of multiomics integration and assay dependent

divergence, we employed dynamic enrichment analysis

and network mapping

. We

evaluated the convergence of subcellular processes (SCPs) and pathways that are over-

represented in different cell types or subsegments within the kidney (in comparison to the other

cell types or subsegments), using single cell RNASeq data from PREMIERE TIS (Michigan,

Princeton, Broad)

, single nucleus RNASeq data from UCSD/WU TIS

, Laser microdissected

(LMD) bulk RNASeq (Supplementary Table 2) and LMD proteomics (Supplementary Table 3)

from the OSU/IU TIS, Near Single Cell (NSC) proteomics from the UCSF TIS (Supplementary

Table 4) and spatial metabolomics from the UTHSA-PNNL-EMBL TIS (Supplementary Table

5A/B/C from 3 different participants).

Single-cell

and -nucleus

RNASeq analysis resulted in the grouping of multiple cells or

nuclei into clusters that were assigned to a particular cell type based on the expression of

essential genes. The top 300 most significantly differentially expressed genes (DEGs) and

proteins (DEPs) of each cluster or subsegment compared to all other clusters or subsegments

as well as the metabolites assigned to glomerular and non-glomerular kidney regions

(Supplementary Table 6) were subjected to enrichment analysis to create pathway maps

(Supplementary Table 7) for the three representative cell types contributing diverse function to

kidney physiology: proximal tubular epithelial cells (Figure 2A, Supplementary Figure 1A for

nonspecific pathways), podocytes (Supplementary Figure 1B) and principal cells of the

collecting ducts (Supplementary Figure 1C). The final maps revealed highly interrelated SCPs

that are intimately linked to the physiological function of the respective cell types. Furthermore,

these SCPs are highly overlapping between assays and datasets with up to 74% of them being

repeatedly enriched in two or more assays, confirming the inherent agreement among these

different assays. While the individual significant genes or gene products coming from multiple

assays were not necessarily the same, placement of these gene products into an

interconnected pathway map showed innate congruence between the assays. The key

subcellular processes (SCPs) for the different cell types differed significantly.

Cell-type specific SCP networks predict overlapping and complementary pathways that

accurately support each cell type’s whole cell function. Proximal tubule networks predict a high

metabolic activity and describe ion reabsorption and ion-triggered glucose reabsorption

pathways as well as ammonia metabolism and detoxification pathways (Figure 2A). The

predictions are in agreement with the energy intensive ion, glucose and other small molecule

reabsorption by the proximal tubule cells

and their predominant function in ammonium

excretion and renal drug clearance

. The identification of cellular iron homeostasis pathways

documents the iron storage capacity of proximal tubule cells

that among other functions,

also mitigates kidney damage during acute kidney injury

. Podocyte/glomerular networks

focus on cell-cell/cell-matrix adhesion, glomerular basement membrane/extracellular matrix

(ECM) and actin dynamics (Supplementary Figure 1B), all pathways fundamental for barrier

generation and consequently for glomerular filtration. Principal cell/collecting duct networks

concentrate on ion reabsorption (Supplementary Figure 1C), emphasizing the important role of

the collecting duct in fine-tuning these mechanisms, thereby regulating systemic electrolyte

and water balance.

These networks document that 13% (principal cells/collecting duct), 27% (proximal tubule

cells/tubulointerstitium) and 74% (podocytes/glomerulus) of all predicted SCPs were

The copyright holder for this preprint (whichthis version posted July 24, 2020. ; https://doi.org/10.1101/2020.07.23.216507doi: bioRxiv preprint

discovered by at least two different technologies. A closer investigation of the SCPs further

highlights that the overlap is even higher, if only the SCPs that describe cell type specific

functions are considered. Furthermore, the different datasets describe complementary

subfunctions of the same physiological processes. For example, both proteomic datasets of

the proximal tubule subsegments describe fatty acid transport via carnitine shuttling into the

mitochondrial matrix, where the enzymes for mitochondrial beta oxidation are localized (Figure

2A). The PREMIERE SC RNASeq dataset predicts carnitine biosynthesis, i.e. synthesis of the

central molecule of the carnitine shuttle.

Integration of pathways that were predicted based on the tubulointerstitial metabolites, such

as ‘Glycolysis and Gluconeogenesis’ and ‘D-Arginine and D-ornithine metabolism’

(Supplementary Figure 1D), into the Molecular Biology of the Cell Ontology (MBCO) SCP-

networks (Figure 1A) further underline the predicted high metabolic activity of the proximal

tubule cells. Glomerular metabolites enrich for pathways (Supplementary Figure 1C), such as

sphingolipid and arachidonic acid metabolism, that support cell-matrix/cell-cell adhesion and

gap junctions, respectively

. Dynamic enrichment analysis of both single-cell RNA-seq

datasets predicts the involvement of another metabolic pathway, i.e. retinol metabolism, in

podocyte function, in particular as a regulator of tight junctions (Supplementary Figure 1B).

Retinoic acid has a regulatory effect on tight junctions

15, 16

and plays a significant role in

mitigating podocyte apoptosis and dedifferentiation during podocyte injury

The enrichment results suggest that proximal tubular cells have the capacity to meet the

high energy demand by not only fueling the citric acid cycle via beta oxidation, but also via

glucose and glutamine catabolism. Nevertheless, beta oxidation is most consistently predicted,

in agreement with previous studies documenting lipid metabolism as the preferential energy

source in proximal tubule cells

18, 19

. Investigation of the pathway components of these SCPs

documents that the different omics technologies identify different components of these

pathways that integrate into a comprehensive description of the relevant biochemical pathways

(Figure 2B). Each technology contributes genes, proteins and metabolites for a fuller

description of the pathways than would be obtained by a single technology. Tubulointerstitial

metabolites, for example, contain glucose, cofactors of the pyruvate dehydrogenase complex

and multiple adenosine nucleotides/nucleosides (i.e. metabolites of the energy carrier ATP). In

agreement with the results of the pathway predictions, network mapping

revealed that cell-

type specific DEGs and DEPs lie within the same area of the human interactome

(Supplementary Figure 1E), indicative of close functional relationships.

In parallel, we identified modules in a kidney-specific functional network using the top

ranked 300 marker genes and proteins across all datatypes in order to detect sets of cell-type

specific, functionally related genes

20, 21

. The module detection algorithm finds groups of genes

that form tightly connected communities within a kidney-specific functional network, which is

constructed using a data-driven approach from gene-gene relationships across thousands of

experimental assays. After module detection, gene enrichment analysis is performed within

each module to understand the key functions of the genes in each module. As with dynamic

enrichment analysis, the modules display clear cell-type specific functional enrichments

(Supplementary Table 8). For example, the network of proximal tubule marker genes includes

modules enriched in anion transport and cellular response to metal ions (Figure 2C), the

network of podocyte marker genes includes modules enriched in glomerulus development and

cell-cell adhesion (Supplementary Figure 1F), and the network of principal cell marker genes

includes modules enriched in sodium ion transport (Supplementary Figure 1G)

The copyright holder for this preprint (whichthis version posted July 24, 2020. ; https://doi.org/10.1101/2020.07.23.216507doi: bioRxiv preprint

A reference tissue atlas for the human kidney

Citations

How to Get Started with Single Cell RNA Sequencing Data Analysis.

Perspectives in systems nephrology.

An optimized approach and inflation media for obtaining complimentary mass spectrometry-based omics data from human lung tissue

References

Kidney physiology and susceptibility to acute kidney injury: implications for renoprotection.

Novel renal amino acid transporters

Kidney Single-Cell Atlas Reveals Myeloid Heterogeneity in Progression and Regression of Kidney Disease.

Apical Transporters for Neutral Amino Acids: Physiology and Pathophysiology

Single cell regulatory landscape of the mouse kidney highlights cellular differentiation programs and disease targets.

Related Papers (5)

A genome-wide transcriptomic analysis of protein-coding genes in human blood cells

The Human Lung Cell Atlas: A High-Resolution Reference Map of the Human Lung in Health and Disease

A molecular cell atlas of the human lung from single cell RNA sequencing

Single-cell transcriptomics reveals gene expression dynamics of human fetal kidney development.

The Tabula Sapiens: a single cell transcriptomic atlas of multiple organs from individual human donors

Frequently Asked Questions (15)

Q1. What have the authors contributed in "Towards building a smart kidney atlas: network-based integration of multimodal transcriptomic, proteomic, metabolomic and imaging data in the kidney precision medicine project" ?

Q2. What are the future works mentioned in the paper "Towards building a smart kidney atlas: network-based integration of multimodal transcriptomic, proteomic, metabolomic and imaging data in the kidney precision medicine project" ?

Q3. What was used for the data normalization and scaling?

Q4. What is the role of fatty acid oxidation in tubulointerstitial ?

Q5. How many libraries were needed to reidentify podocytes?

Q6. How many jensenlab confidences were used to identify each gene?

Q7. What are the metabolites of the energy carrier ATP?

Q8. How many libraries are needed for a consistent detection of podocytes?

Q9. How many differentially expressed genes and proteins were predicted by each assay?

Q10. How many SCPs can be included in the top seven predictions?

Q11. How many samples were sufficient to reproduce the results for the full dataset?

Q12. What is the way to integrate the three different assays?

Q13. How many predictions are needed to re-identify the top 10 or seven predictions?

Q14. What is the role of the collecting duct in regulating systemic electrolyte and?

Q15. What is the correlation between the gene expression profiles of cells and LCM segments?