Edinburgh Research Explorer

Resolving the fibrotic niche of human liver cirrhosis at single-celllevelCitation for published version:Ramachandran, P, Dobie, R, Wilson-Kanamori, JR, Dora, E, Henderson, B, Luu, N-T, Portman, J, Matchett,K, Brice, M, MARWICK, JOHNALEXANDER, Richard, T, Efremova, M, Vento-Tormo, R, Carragher, NO,Kendall, TJ, Fallowfield, JA, Harrison, EM, Mole, DJ, Wigmore, SJ, Newsome, PN, Weston, CJ, Iredale, JP,Tacke, F, Pollard, JW, Ponting, CP, Marioni, JC, Teichmann, SA & Henderson, NC 2019, 'Resolving thefibrotic niche of human liver cirrhosis at single-cell level', Nature. https://doi.org/10.1038/s41586-019-1631-3

Digital Object Identifier (DOI):10.1038/s41586-019-1631-3

Link:Link to publication record in Edinburgh Research Explorer

Document Version:Peer reviewed version

Published In:Nature

General rightsCopyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s)and / or other copyright owners and it is a condition of accessing these publications that users recognise andabide by the legal requirements associated with these rights.

Take down policyThe University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorercontent complies with UK legislation. If you believe that the public display of this file breaches copyright pleasecontact openaccess@ed.ac.uk providing details, and we will remove access to the work immediately andinvestigate your claim.

Download date: 21. Jun. 2020

1

Resolving the fibrotic niche of human liver cirrhosis using single-cell

transcriptomics

Ramachandran P1*, Dobie R1, Wilson-Kanamori JR1, Dora EF1, Henderson BEP1, Luu

NT2,3, Portman JR1, Matchett KP1, Brice M1, Marwick JA1,4, Taylor RS1, Efremova M5,

Vento-Tormo R5, Carragher NO4, Kendall TJ1,6, Fallowfield JA1, Harrison EM7, Mole

DJ1,7, Wigmore SJ1,7, Newsome PN2,3, Weston CJ2,3, Iredale JP8, Tacke F9, Pollard

JW10,11, Ponting CP12, Marioni JC5,13,14, Teichmann SA5,13,15, Henderson NC1*

1University of Edinburgh Centre for Inflammation Research, The Queen’s Medical Research Institute, Edinburgh BioQuarter, Edinburgh, UK. 2NIHR Birmingham Biomedical Research Centre, University Hospitals Birmingham NHS Foundation Trust and University of Birmingham, UK 3Institute of Immunology and Immunotherapy, University of Birmingham, UK 4Cancer Research UK Edinburgh Centre, MRC Institute of Genetics and Molecular Medicine at the University of Edinburgh, Edinburgh, UK. 5Wellcome Sanger Institute, Wellcome Genome Campus, Hinxton, Cambridge, UK. 6Division of Pathology, University of Edinburgh, Edinburgh, UK. 7Clinical Surgery, University of Edinburgh, Royal Infirmary of Edinburgh, Edinburgh, UK. 8Office of the Vice Chancellor, Beacon House and National Institute for Health Research, Biomedical Research Centre, Bristol, UK 9Department of Hepatology and Gastroenterology, Charité University Medical Center, Berlin, Germany 10MRC Centre for Reproductive Health, The Queen's Medical Research Institute, University of Edinburgh, Edinburgh, UK. 11Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, New York, USA. 12MRC Human Genetics Unit, MRC Institute of Genetics and Molecular Medicine at the University of Edinburgh, Edinburgh, UK. 13European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI), Hinxton, Cambridge, UK. 14Cancer Research UK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, UK. 15Theory of Condensed Matter Group, The Cavendish Laboratory, University of Cambridge, JJ Thomson Ave, Cambridge, CD3 0EH, UK. *Address correspondence to:

Prakash Ramachandran, University of Edinburgh Centre for Inflammation Research, The

Queen’s Medical Research Institute, Edinburgh BioQuarter, 47 Little France Crescent,

Edinburgh, UK, EH16 4TJ. Phone: 0131.242.6654; Email: Prakash.Ramachandran@ed.ac.uk

or Neil Henderson, University of Edinburgh Centre for Inflammation Research, The Queen’s

Medical Research Institute, Edinburgh BioQuarter, 47 Little France Crescent, Edinburgh, UK,

EH16 4TJ. Phone: 0131.242.6688; Email: Neil.Henderson@ed.ac.uk

1

Abstract 1

Currently there are no effective antifibrotic therapies for liver cirrhosis, a major killer 2

worldwide. To obtain a cellular resolution of directly-relevant pathogenesis and to 3

inform therapeutic design, we profile the transcriptomes of over 100,000 primary 4

human single cells, yielding molecular definitions for the major non-parenchymal cell 5

types present in healthy and cirrhotic human liver. We uncover a novel scar-associated 6

TREM2+CD9+ macrophage subpopulation, which expands in human and mouse liver 7

fibrosis, has a distinct differentiation trajectory from circulating monocytes and 8

displays a pro-fibrogenic phenotype. In the endothelial compartment, we show that 9

newly-defined ACKR1+ and PLVAP+ endothelial cells expand in cirrhosis, are 10

topographically located in the fibrotic septae and enhance leucocyte transmigration. 11

Multi-lineage ligand-receptor modelling of specific interactions between the novel 12

scar-associated macrophages, endothelial cells and PDGFRα+ collagen-producing 13

mesenchymal cells in the fibrotic niche, reveals intra-scar activity of several major pro-14

fibrogenic pathways including TNFRSF12A, PDGFR and NOTCH signalling. Our 15

work dissects unanticipated aspects of the cellular and molecular basis of human organ 16

fibrosis at a single-cell level, and provides the conceptual framework required to 17

discover rational therapeutic targets in liver cirrhosis. 18

2

Main 19

Liver cirrhosis is a major global healthcare burden. Recent estimates suggest that 844 20

million people worldwide have chronic liver disease, with a mortality rate of two 21

million deaths per year and a rising incidence1. In health, the liver serves a myriad of 22

functions including detoxification, metabolism, bile production and immune 23

surveillance. Chronic liver disease, the result of iterative liver injury secondary to any 24

cause, results in progressive fibrosis, disrupted hepatic architecture, vascular changes 25

and aberrant regeneration, defining characteristics of liver cirrhosis2. Importantly, the 26

degree of liver fibrosis predicts adverse patient outcomes, including the development 27

of cirrhosis-related complications, hepatocellular carcinoma and death3. Hence, there is 28

a clear therapeutic imperative to develop effective anti-fibrotic approaches for patients 29

with chronic liver disease4–7. 30

Liver fibrosis involves a complex, orchestrated interplay between multiple non-31

parenchymal cell (NPC) lineages including immune, endothelial and mesenchymal 32

cells spatially located within areas of scarring, termed the fibrotic niche. Despite rapid 33

progress in our understanding of the cellular interactions underlying liver fibrogenesis 34

accrued using rodent models, there remains a significant 'translational gap' between 35

putative targets and effective patient therapies4,5. This is in part due to the very limited 36

definition of the functional heterogeneity and interactome of cell lineages that 37

contribute to the fibrotic niche of human liver cirrhosis, which is imperfectly 38

recapitulated by rodent models4,6. 39

Single-cell RNA sequencing (scRNA-seq) has the potential to deliver a step change in 40

both our understanding of healthy tissue homeostasis as well as disease pathogenesis, 41

allowing the interrogation of individual cell populations at unprecedented resolution8–42 11. Here, we have studied the mechanisms regulating human liver cirrhosis, using 43

scRNA-seq to analyse the transcriptomes of 106,616 single cells obtained from ten 44

healthy and cirrhotic human livers and peripheral blood. 45

Our data define: (1) a single-cell atlas of non-parenchymal cells in healthy and cirrhotic 46

human liver; (2) a new subpopulation of scar-associated TREM2+CD9+ pro-fibrogenic 47

macrophages; (3) new subpopulations of scar-associated ACKR1+ and PLVAP+ 48

3

endothelial cells; and (4) key ligand-receptor interactions between novel scar-49

associated macrophages, endothelial subpopulations and collagen-producing 50

myofibroblasts in the fibrotic niche. Thus, we have simultaneously identified a series 51

of intra-scar pro-fibrogenic pathways which represent hitherto unsuspected therapeutic 52

targets for the treatment of liver fibrosis, whilst demonstrating the applicability of 53

scRNA-seq to define pathogenic mechanisms for other human fibrotic disorders. 54

Results 55

Single-cell atlas of human liver non-parenchymal cells 56

Hepatic NPC were isolated from fresh healthy and cirrhotic human liver tissue spanning 57

a range of aetiologies of cirrhosis (Fig. 1a, Extended Data Fig. 1a). Importantly, to 58

minimise artefacts12, we developed a rapid tissue processing pipeline, obtaining fresh 59

non-ischaemic liver tissue taken by wedge biopsy prior to the interruption of the hepatic 60

vascular inflow during liver surgery or transplantation, and immediately processing this 61

for FACS. This enabled a workflow time of under three hours from patient to single-62

cell droplet encapsulation (Methods). 63

We used an unbiased approach, FACS sorting viable single cells from liver tissue into 64

broad leucocyte (CD45+) or other NPC (CD45-) fractions (Extended Data Fig. 1b), prior 65

to scRNA-seq. To facilitate discrimination between liver-resident and circulating 66

leucocytes, we also performed scRNA-seq on CD45+CD66b- peripheral blood 67

mononuclear cells (PBMCs) (Extended Data Fig. 1c, f). In total, we analysed 67,494 68

human cells from healthy (n=5) and cirrhotic (n=5) livers, 30,741 PBMCs from 69

cirrhotic patients (n=4) and compared our data with a publicly-available reference 70

dataset of 8,381 PBMCs from a healthy donor. 71

Tissue cells and PBMCs could be partitioned into 21 distinct clusters, which we 72

visualized using t-distributed stochastic neighbourhood embedding (t-SNE) (Extended 73

Data Fig. 1d). Clusters were annotated using signatures and integrating with known 74

lineage markers (Extended Data Fig. 1e; signature gene lists available in Supplementary 75

Table 1). All PBMC datasets contained the major blood lineages, with excellent 76

reproducibility between samples (Extended Data Fig. 1g, h). To generate an atlas of 77

4

liver-resident cells, contaminating circulating cells were removed from the liver tissue 78

datasets, by excluding individual cells from the tissue samples which mapped 79

transcriptionally to blood-derived clusters 1 and 13 (Extended Data Fig. 1d). 80

Re-clustering the 66,135 liver-resident cells revealed 21 clusters (Fig. 1b), each 81

containing cells from both healthy and cirrhotic livers (Fig. 1c). Gene signature analysis 82

enabled annotation of each cluster by major cell lineage (Fig. 1d, Extended Data Fig. 83

2a, b). We noted heterogeneity in the post-normalised detected number of genes and 84

unique molecular identifiers (UMIs) per cell, dependent on cell lineage (Extended Data 85

Fig. 2c, d). All samples contained the expected cell lineages (Extended Data Fig. 2e, g) 86

and reproducibility between livers was excellent for the main NPC populations 87

(Extended Data Fig. 2f). 88

We used an area-under-curve classifier to identify cell subpopulation markers across 89

all 21 clusters and 11 lineages (Fig. 1e; Supplementary Tables 2 and 3). Expression of 90

collagens type I and type III, the main extracellular matrix components of the fibrotic 91

niche, was restricted to cells of the mesenchymal lineage (Fig. 1e). To gain further 92

resolution on NPC heterogeneity, we then iterated clustering and marker gene 93

identification on each lineage in turn, for example defining 11 clusters of T cells and 94

innate lymphoid cells (ILCs) (Extended Data Fig. 3a) and four clusters of B cells and 95

plasma cells (Extended Data Fig. 3f, g, Supplementary Table 6). No major differences 96

in B cell or plasma cell composition between healthy and cirrhotic livers were observed 97

(Extended Data Fig. 3h), and plasmacytoid dendritic cells (pDC) showed no additional 98

heterogeneity. 99

To further annotate the 11 T cell and ILC clusters (36,900 cells from 10 livers) we 100

assessed expression of known markers (Extended Data Fig. 3c) and computationally 101

identified differential marker genes (Extended Data Fig. 3d, Supplementary Table 4). 102

We also performed imputation of gene dropouts, which enhanced detection of 103

discriminatory marker genes for each cluster but did not yield additional T cell or ILC 104

subpopulations (Extended Data Fig. 3e, Supplementary Table 5). All T cell and ILC 105

clusters expressed tissue residency markers CD69 and CXCR4. Clusters 1 and 2 were 106

CD4+ T cells, with CD4+ T cell(2) expanding significantly in cirrhotic livers (Extended 107

Data Fig. 3a, b, e) and expressing SELL and CCR7, indicating an expansion of memory 108

5

CD4+ T cells in liver cirrhosis. Sparse expression of FOXP3, RORC, IL17A and IFNG 109

in both CD4+ T cell subpopulations suggested the presence of Tregs, Th17 and Th1 110

cells in these clusters. Clusters 3, 4 and 5 were CD8+ T cells, with features of effector 111

T cells expressing GZMA, GZMH and IFNG. Two resident CD56bright IL7R- NK cell 112

clusters were defined (NK cell(1) and NK cell(2)), as well as a distinct cytotoxic 113

CD56dim NK cell population (cNK), with specific expression of FCGR3A and GZMB. 114

No expansion of these populations was observed in cirrhotic livers. 115

We provide an interactive gene browser freely-available online 116

(http://www.livercellatlas.mvm.ed.ac.uk), to allow assessment of individual gene 117

expression both in all human liver NPC and in specific lineages, comparing healthy 118

versus cirrhotic livers. 119

(Note to referees: http://www.livercellatlas.mvm.ed.ac.uk will be made freely 120

available. For review purposes, log-in details are username: Edinburghlivercellatlas; 121

password: Cirrhosis; please refresh your browser if the webpage does not load at first 122

attempt.) 123

Distinct macrophage subpopulations inhabit the fibrotic niche 124

Macrophages are critical to tissue homeostasis and wound-healing13. Previous studies 125

have highlighted phenotypically-distinct macrophage populations orchestrating both 126

liver fibrosis progression and regression in rodent models14,15, with preliminary 127

evidence of heterogeneity in fibrotic human livers16. Here, we define unique 128

subpopulations of macrophages which populate the fibrotic niche of cirrhotic human 129

livers. Unsupervised clustering of all 10,737 mononuclear phagocytes (1,074±153 cells 130

from each liver), isolated from the combined liver-resident cell dataset, identified nine 131

MP clusters and one cluster of proliferating MP cells (Fig. 2a). We annotated these nine 132

clusters as subpopulations of scar-associated macrophages (SAM), Kupffer cells (KC), 133

tissue monocytes (TMo), and conventional dendritic cells (cDC) (Fig. 2a; see below). 134

Strikingly, clusters MP(4) and MP(5), named SAM(1) and SAM(2) respectively, were 135

expanded in cirrhotic livers (Fig. 2b), a finding that was confirmed by quantification of 136

the MP cell composition of each liver individually (Fig. 2c), and reproduced in all 137

cirrhotic livers irrespective of liver disease aetiology. 138

6

To enable MP cell annotation, we initially assessed expression of known MP marker 139

genes (Extended Data Fig. 4a), classifying clusters MP(8) and MP(9) as conventional 140

dendritic cells, cDC2 and cDC1 respectively, based on CD1C and CLEC9A specificity. 141

However, the remaining markers did not demarcate the other monocyte and 142

macrophage subpopulations. Instead, these were identified using differential expression 143

analysis across all MP clusters (Fig. 2d, Supplementary Table 7). Clusters MP(1), 144

MP(2) and MP(3) were distinguished by expression of S100 genes, FCN1, MNDA and 145

LYZ, in keeping with a tissue monocyte (TMo) phenotype and informing annotation as 146

TMo(1), TMo(2) and TMo(3) respectively (Fig. 2d, e, Extended Data Fig. 4a, 147

Supplementary Table 7). 148

Clusters MP(6) and MP(7) were enriched in CD163, MARCO, TIMD4 and CD5L 149

(Extended Data Fig. 4b); multiplex immunofluorescence staining confirmed these as 150

Kupffer cells (KC; resident liver macrophages), facilitating annotation of these clusters 151

as KC(1) and KC(2) respectively (Extended Data Fig. 4c). Application of these markers 152

enabled the definitive distinction between KC and other MP cells for the first time in 153

human liver tissue. KC displayed characteristic morphology and sinusoidal topography 154

in healthy livers but were absent from areas of scarring in cirrhotic livers (Extended 155

Data Fig. 4c). A lack of TIMD4 expression distinguished KC(2) from KC(1) (Extended 156

Data Fig. 4b); CD163+MARCO+TIMD4- cells were identifiable in healthy livers but 157

rare in cirrhotic livers (Extended Data Fig. 4c), concordant with a significant reduction 158

of KC(2) cells in cirrhosis (Fig. 2c). Automated histological cell counting demonstrated 159

TIMD4+ cell numbers to be equivalent between healthy and cirrhotic livers, but showed 160

a loss of MARCO+ cells, consistent with selective reduction in MARCO+TIMD4- KC 161

in liver fibrosis (Extended Data Fig. 4d, e). 162

Scar-associated clusters SAM(1) and SAM(2), expanded in diseased livers and 163

expressed the unique markers TREM2 and CD9 (Fig. 2c-e). These newly-defined 164

macrophages displayed a hybrid phenotype, with features of both tissue monocytes and 165

KC (Fig. 2d, e), analogous to monocyte-derived macrophages in murine liver injury 166

models15,17. Multi-colour flow cytometry confirmed expansion of these TREM2+CD9+ 167

macrophages in human fibrotic livers (Fig. 2f, Extended Data Fig. 4f). Tissue 168

immunofluorescence staining and single-molecule fluorescent in situ hybridization 169

7

(smFISH) demonstrated the presence of TREM2+MNDA+ and CD9+MNDA+ 170

macrophages in fibrotic livers (Extended Data Fig. 4g-i). Multiplex 171

immunofluorescence further confirmed the presence of TREM2+CD9+ cells localised 172

in collagen-positive scar regions in cirrhotic livers (Fig. 2g), and automated cell 173

counting of stained sections confirmed expansion of TREM2+ and CD9+ cells in 174

cirrhotic livers (Fig. 2h, i). 175

Strikingly, TREM2+ and CD9+ cells were rarely identified in the parenchyma of healthy 176

livers, but were consistently located within areas of scar in cirrhotic livers. To confirm 177

this, automated cell counting was applied to immunohistochemically-stained cirrhotic 178

livers morphologically segmented into regions of fibrotic septae and parenchymal 179

nodules (Fig. 2j). This demonstrated a significant accumulation of TREM2+ and CD9+ 180

cells in fibrotic regions, whilst negligible numbers of KC populated the fibrotic septae 181

(Fig. 2j). Hence, we annotated TREM2+CD9+ MP cells as scar-associated macrophages 182

(SAM). 183

Local proliferation has been shown to play a significant role in the expansion of 184

macrophage subpopulations at sites of inflammation and fibrosis in experimental rodent 185

models15,18,19, but has not been extensively characterised in human inflammatory 186

disorders. To investigate MP proliferation in human liver fibrosis, we isolated the 187

cycling MP cluster (Fig. 2a; cluster 10), which was enriched for multiple cell cycle-188

related genes (Supplementary Table 7). Cycling MP cells subclustered into four, 189

yielding cDC1, cDC2, KC and SAM subpopulations (Extended Data Fig. 4j). We 190

observed a significant expansion of cycling SAM in cirrhosis, representing 1.70±0.52% 191

of total TREM2+ MP cells in cirrhotic livers (Extended Data Fig. 4k). In contrast 192

0.99±0.63% of KC were proliferating in healthy livers, with none detected in cirrhotic 193

livers (Extended Data Fig. 4k). These data highlight the potential role of local 194

macrophage proliferation in driving the accumulation of SAM in the fibrotic niche of 195

human chronic liver disease. 196

Pro-fibrogenic phenotype of scar-associated macrophages 197

To delineate the functional profile of SAM we generated self-organising maps using 198

the SCRAT package, visualising co-ordinately expressed gene groups across the MP 199

8

subpopulations. This created a landscape of 3600 metagenes on a 60x60 grid and 200

highlighted 44 metagene signatures overexpressed in the MP lineage (Fig. 3a). 201

Mapping the nine MP cell clusters to this landscape (Extended Data Fig. 5a) identified 202

six optimally-differentiating metagene signatures, denoted as A-F (Fig. 3a, 203

Supplementary Table 8). Signatures A and B defined the scar-associated macrophages 204

and were enriched for ontology terms relevant to tissue fibrosis and associated 205

processes such as angiogenesis, in addition to known macrophage functions such as 206

phagocytosis and antigen processing (Fig. 3b). These SAM-defining signatures 207

included genes such as TREM2, IL1B, SPP1, LGALS3, CXCR4, CCR2, and TNFSF12; 208

a number of which are known to regulate the function of scar-producing myofibroblasts 209

in fibrotic liver diseases20–25. The remaining MP subpopulations were defined by 210

signature C (KC), signatures D, E (TMo) and signature F (cDC1); ontology terms 211

matched known functions for the associated cell type (Fig. 3b, Extended Data Fig. 5b, 212

Supplementary Table 8). In particular, the KC clusters showed significant enrichment 213

for ontology terms involving endocytosis, lipid and iron homeostasis, known functions 214

of KC in mice26. Importantly, macrophage populations did not conform to either an M1 215

or M2 phenotype, again highlighting the limitation of this classification. 216

In mice, there are two main origins of hepatic macrophages, either embryologically-217

derived or monocyte-derived27. Under homeostatic conditions, tissue-resident KC 218

predominate and are embryologically-derived self-renewing cells28–32. However, 219

following liver injury, macrophages derived from the recruitment and differentiation of 220

circulating monocytes accumulate in the liver and regulate hepatic fibrosis15,33. The 221

ontogeny of human hepatic macrophage subpopulations has never previously been 222

investigated. Scar-associated TREM2+CD9+ macrophages demonstrated a monocyte-223

like morphology (Fig. 2g, Extended Data Fig. 4g-i) and a distinct topographical 224

distribution from KC (Fig. 2j), suggesting they may represent monocyte-derived cells. 225

To computationally assess the origin of these SAM, we performed in silico trajectory 226

analysis on a combined dataset of peripheral blood monocytes and liver-resident MPs. 227

We visualised the transcriptional profile of these cells using a diffusion map, mapped 228

them along a pseudotemporal trajectory (using the monocle R package) and interrogated 229

their directionality via spliced and unspliced mRNA ratios (RNA velocity34,35) (Fig. 230

3c). These analyses suggested a branching differentiation trajectory from peripheral 231

9

blood monocytes into either SAM or cDC (Fig. 3c). Additionally, applying RNA 232

velocity indicated a lack of differentiation from KC to SAM, and no progression from 233

SAM to KC (Fig. 3c). 234

To further investigate the pseudotemporal relationship between SAM and KC, we 235

visualised the combined blood monocyte and liver-resident MP dataset using a UMAP, 236

and performed additional RNA velocity analyses34,35 (Fig. 3d). Evaluation of spliced 237

and unspliced mRNAs showed expected downregulation (negative velocity) of 238

monocyte gene MNDA in SAM, expected upregulation (positive velocity) of SAM 239

marker gene CD9 in tissue monocytes, and a lack of KC gene TIMD4 velocity in SAM 240

(Fig. 3e). This infers an absence of pseudotemporal dynamics between KC and SAM. 241

Furthermore, assessment of the probabilities of cells in this dataset transitioning into 242

SAM, indicated a higher likelihood of tissue monocytes than KC differentiating into 243

SAM (Fig. 3f). Overall, these data suggest that scar-associated macrophages in human 244

fibrotic liver are monocyte-derived, and imply that SAM represent a terminally-245

differentiated cell state within the fibrotic niche. 246

To further characterise the phenotype of scar-associated macrophages, we identified 247

differentially expressed genes along the branching monocyte differentiation trajectories 248

(Fig. 3g). We defined three gene co-expression modules by hierarchical clustering, with 249

module 1 representing genes that are upregulated during blood monocyte-to-SAM 250

differentiation (Fig. 3g). Module 1 was over-expressed in SAM, and contained multiple 251

pro-fibrogenic genes including SPP1, LGALS3, CCL2, CXCL8, PDGFB and VEGFA20–252 23,36–38(Fig. 3h). Analogous to signatures A and B (Fig. 3b), module 1 displayed 253

ontology terms consistent with promoting tissue fibrosis and angiogenesis, including 254

the regulation of other relevant cell types such as fibroblasts and endothelial cells (Fig. 255

3h). Co-expression module 2 contained genes that were downregulated during 256

monocyte-to-SAM differentiation, confirming a loss of characteristic monocyte genes 257

(Extended Data Fig. 5e). Module 3 encompassed a distinct group of genes that were 258

upregulated during monocyte-to-cDC differentiation (Extended Data Fig. 5f). Full lists 259

of genes and ontology terms for all three modules are available (Supplementary Table 260

9). These data highlight that SAM acquire a specific pro-fibrogenic phenotype during 261

differentiation from circulating monocytes. 262

10

To further assess the function of human SAM, macrophage subpopulations were 263

isolated from cirrhotic human livers by FACS (Fig. 2f, Extended Data Fig. 4f). SAM 264

demonstrated enhanced protein secretion of several of the mediators identified by 265

transcriptional analysis (Extended Data Fig. 5c). Additionally, conditioned media from 266

SAM promoted fibrillar collagen expression by primary human hepatic stellate cells 267

(HSC) (Extended Data Fig. 5d), confirming that SAM have a pro-fibrogenic phenotype. 268

To enable cross-species comparison, we performed scRNA-seq on liver MP cells 269

isolated from control (uninjured) mice or mice treated with chronic carbon tetrachloride 270

(CCl4), a well-established mouse model of liver fibrosis15. MP cells from fibrotic livers 271

were isolated 24 hours following the final CCl4 injection, a time of active fibrogenesis15. 272

Five clusters of MP cells were defined (Extended Data Fig. 6a), with differentially 273

expressed marker genes (Extended Data Fig. 6c, d, Supplementary Table 10) which 274

facilitated cell type annotation (Extended Data Fig. 6a). An injury-specific macrophage 275

population, cluster mMP(2), was identified (Extended Data Fig. 6a, b) and was 276

differentiated by high expression of Cd9, Trem2, Spp1 and Lgals3 (Extended Data Fig. 277

6c, d). We confirmed expansion of this murine CD9+ SAM (mSAM) population in liver 278

fibrosis (Extended Data Fig. 6e, f). Co-culture of mSAM with quiescent primary murine 279

HSC promoted fibrillar collagen expression in HSC (Extended Data Fig. 6g), indicating 280

a pro-fibrogenic phenotype of mSAM. To confirm that mSAM represent the corollary 281

population to human SAM (hSAM), we performed an unbiased canonical correlation 282

analysis (CCA) between human and mouse MP datasets39. Both hSAM and mSAM 283

clustered together (h&mMP(2); Extended Data Fig. 6h,i) and this cluster was enriched 284

for the SAM markers CD9, TREM2 and SPP1 (Extended Data Fig. 6j). Similar cross-285

species conservation was also observed for KC (h&mMP(3); Extended Data Fig. 6h-j). 286

To identify potential transcriptional regulators of human SAM we used the SCENIC 287

package to define sets of genes co-expressed with known transcription factors, termed 288

regulons. We assessed the cell activity score for differentially-expressed regulons along 289

the tissue monocyte-macrophage pseudotemporal trajectory and in KC, allowing 290

visualisation of regulon activity across liver-resident macrophage subpopulations 291

(Extended Data Fig. 5g, h, Supplementary Table 11). This identified regulons and 292

corresponding transcription factors associated with distinct macrophage phenotypes, 293

11

highlighting NR1H3 and SPIC activity in human KC (Extended Data Fig. 5g, h), which 294

are known to regulate KC function in mice40,41. Scar-associated macrophages are 295

enriched for regulons containing the transcription factors HES1 and EGR2 (Extended 296

Data Fig. 5g, h), both of which have been associated with modulation of macrophage 297

phenotype and tissue fibrosis42–45. 298

To determine whether SAM also expand in earlier stage human liver disease, we 299

analysed cohorts of patients with non-alcoholic fatty liver disease (NAFLD). We 300

applied differential gene expression signatures of human SAM, KC and TMo to a 301

deconvolution algorithm46, which enabled the assessment of the hepatic monocyte-302

macrophage composition in whole liver microarray data across the spectrum of early-303

stage NAFLD severity47 (Extended Data Fig. 7a). This demonstrated an expansion of 304

SAM in patients with non-alcoholic steatohepatitis (NASH) (Extended Data Fig. 7a, b), 305

with an increased frequency of SAM with worsening histological NASH activity (NAS) 306

and fibrosis scores (Extended Data Fig. 7c). No association was observed between 307

SAM frequency and patient demographics such as gender, age or body mass index 308

(Extended Data Fig. 7d). We were also able to histologically identify SAM in a locally 309

generated NASH biopsy cohort (Extended Data Fig. 7e). SAM expansion increased 310

with NASH activity (Extended Data Fig. 7e) and there was a positive correlation 311

between SAM number and degree of fibrosis across the full severity spectrum of 312

NAFLD-induced liver fibrosis (Extended Data Fig. 7f). 313

In summary, multimodal computational, functional and histological analysis 314

demonstrates that TREM2+CD9+ scar-associated macrophages derive from the 315

recruitment and differentiation of circulating monocytes, are conserved across species, 316

display a pro-fibrogenic phenotype and expand early in the course of liver disease 317

progression. 318

Distinct endothelial subpopulations inhabit the fibrotic niche 319

In rodent models, hepatic endothelial cells are known to regulate both fibrogenesis48,49 320

and macrophage recruitment to the fibrotic niche37. Unsupervised clustering of human 321

liver endothelial cells identified seven subpopulations (Fig. 4a). Clusters Endo(6) and 322

Endo(7) significantly expanded in cirrhotic compared to healthy livers, whilst Endo(1) 323

12

contracted (Fig. 4a, b). Classical endothelial cell markers did not discriminate between 324

the seven clusters, although Endo(1) was distinct in lacking CD34 expression 325

(Extended Data Fig. 8a). In order to fully annotate endothelial subpopulations 326

(Extended Data Fig. 8h), we identified differentially expressed markers (Fig. 4c, 327

Supplementary Table 12), determined functional expression profiles (Extended Data 328

Fig. 8g, Supplementary Table 13), performed transcription factor regulon analysis (Fig. 329

4g, Supplementary Table 14) and assessed spatial distribution via multiplex 330

immunofluorescence staining (Fig. 4d, e). 331

Disease-specific endothelial cells Endo(6) and Endo(7), CD34+PLVAP+VWA1+ and 332

CD34+PLVAP+ACKR1+ respectively (Fig. 4c, Extended Data Fig. 8b), expanded in 333

cirrhotic liver tissue (Fig. 4f) and were spatially restricted to the fibrotic niche (Fig. 4d, 334

e, Extended Data Fig. 8c), allowing their annotation as scar-associated endothelia 335

SAEndo(1) and SAEndo(2) respectively. Scar-associated endothelial cells displayed 336

enhanced expression of the ELK3 regulon (Fig. 4g), a transcription factor known to 337

modulate angiogenesis50. Metagene signature analysis found that Endo(6) (SAEndo(1)) 338

cells expressed pro-fibrogenic genes including PDGFD, PDGFB, LOX, LOXL2 and 339

several basement membrane components51–53; associated significant ontology terms 340

included extracellular matrix organization and wound healing (signature A; Extended 341

Data Fig. 8g). Endo(7) (SAEndo(2)) cells displayed an immunomodulatory phenotype 342

(signature B; Extended Data Fig. 8g). Furthermore, the most discriminatory marker for 343

this cluster, ACKR1, is restricted to venules in mice54 and has a role in regulating 344

leucocyte recruitment55. We isolated endothelial cells from healthy and cirrhotic human 345

livers and confirmed increased expression of PLVAP, CD34 and ACKR1 on cells from 346

diseased livers (Extended Data Fig. 8d). Flow-based adhesion assays56 demonstrated 347

that cirrhotic endothelial cells showed enhanced leucocyte transmigration (Extended 348

Data Fig. 8e), an effect that was attenuated by ACKR1 knockdown (Extended Data Fig. 349

8f). These data demonstrate that SAEndo regulate inflammatory cell recruitment to the 350

fibrotic niche. 351

Using CLEC4M as a discriminatory marker of cluster Endo(1) (Fig. 4c, Extended Data 352

Fig. 8b), immunofluorescence confirmed these CLEC4M+CD34- cells as liver 353

sinusoidal endothelial cells (LSEC), restricted to perisinusoidal cells within the liver 354

13

parenchyma (Fig. 4d). Cluster Endo(1) demonstrated known features of LSEC, 355

including GATA4 transcription factor regulon expression57 (Fig. 4g), and a metagene 356

signature enriched for ontology terms including endocytosis and immune response58 357

(signature D, Extended Data Fig. 8g). There was a reduction in CLEC4M staining in 358

cirrhotic livers with an absence in fibrotic septae (Fig. 4d, f), indicating that LSEC do 359

not inhabit the fibrotic niche in chronic liver disease. This was further supported by 360

trajectory analysis, suggesting a lack of clear pseudotemporal dynamics between the 361

LSEC and clusters SAEndo(1) and SAEndo(2) (Extended Data Fig. 8i). 362

We annotated cluster Endo(2) (PDPN+CD34+PLVAP-) as lymphatic endothelial cells 363

based on marker gene expression, relevant ontology terms (signature E; Extended Data 364

Fig. 8g) and FOXC2 and HOXD8 regulon activity (Fig. 4g)59,60. Lymphatics populated 365

the portal region of healthy livers (Fig. 4e). Hierarchical clustering of the endothelial 366

subpopulations demonstrated that clusters Endo(3) and Endo(4) were closely related to 367

LSEC (dendrogram not shown), co-expressing markers including CLEC4G (Extended 368

Data Fig. 8b). Endo(4), defined as RSPO3+CD34+PLVAP+ (Fig. 4c, Extended Data Fig. 369

8b), expressed a metagene signature overlapping with LSEC (signature D, Extended 370

Data Fig. 8g), and were identified as central vein endothelial cells (Fig. 4e). This 371

mirrors murine liver zonation data indicating RSPO3 as a marker of pericentral 372

endothelial cells61. Similar to LSEC, central vein endothelial cells did not inhabit the 373

fibrotic niche in cirrhosis (Fig. 4e). 374

Cluster Endo(5), AIF1L+CD34+PLVAP+ cells, were mapped to periportal thick-walled 375

vessels, consistent with hepatic arterial endothelial cells (Fig. 4e). Of note, these cells 376

were also topographically associated with fibrotic septae in cirrhotic livers (Fig. 4e). 377

The arterial identity of this cluster was further indicated by SOX17 regulon expression62 378

(Fig. 4g), and it displayed a metagene signature enriched for Notch pathway ligands 379

JAG1, JAG2 and DLL4; ontology terms included animal organ development, 380

angiogenesis and Notch signalling (signature C; Extended Data Fig. 8g) in keeping with 381

the known requirement of the Notch pathway in the development and maintenance of 382

hepatic vasculature63. Endo(5) was annotated as HAEndo for subsequent analysis of 383

cellular interactions within the fibrotic niche. 384

385

14

PDGFRA expression defines scar-associated mesenchymal cells 386

Hierarchical clustering of human liver mesenchymal cells identified four populations, 387

including a population of mesothelial cells (Fig. 5a, b, Extended Data Fig. 9a, 388

Supplementary Table 15). MYH11 expression distinguished cluster Mes(1) (Fig. 5b, 389

Extended Data Fig. 9a), and labelled a population of vascular smooth muscle cells 390

(VSMC) (Fig. 5c). Cluster Mes(2) expressed high levels of RGS5 (Fig. 5b, Extended 391

Data Fig. 9a). RGS5 immunofluorescence staining demonstrated that these 392

mesenchymal cells were located peri-sinusoidally (Fig. 5c), identifying this population 393

as hepatic stellate cells (HSC)64. RGS5+ cells were absent from the fibrotic niche (Fig. 394

5c). Cluster Mes(3) was distinguished by PDGFRA expression, in addition to high 395

levels of fibrillar collagens and pro-fibrogenic genes such as TIMP1 and CCL2 (Fig. 396

5b, d, Extended Data Fig. 9a). PDGFRα+ cells expanded significantly in cirrhotic livers 397

(Fig. 5a, e, f) and were spatially mapped to the fibrotic niche (Fig. 5f), enabling 398

annotation as scar-associated mesenchymal cells (SAMes). 399

To study heterogeneity within the SAMes population, further hierarchical clustering 400

was performed on this population (Extended Data Fig. 9b). Two populations of SAMes 401

were identified (Extended Data Fig. 9c, Supplementary Table 16), both of which 402

expanded in cirrhotic livers (Extended Data Fig. 9d). OSR1, a marker of fibroblast 403

subpopulations and regulator of extracellular matrix production65, distinguished a 404

subpopulation of SAMes (Extended Data Fig. 9c), labelled a population of periportal 405

cells in healthy liver (Extended Data Fig. 9e) and a subpopulation of scar-associated 406

cells in the fibrotic niche (Extended Data Fig. 9f). These OSR1+ cells are likely to 407

represent portal fibroblasts. 408

In rodent models, HSC differentiate into scar-producing myofibroblasts following 409

parenchymal liver injury66–68. We interrogated HSC differentiation in human liver 410

tissue using pseudotemporal ordering and RNA velocity analyses between HSC and 411

SAMes clusters, demonstrating a clear trajectory from HSC to SAMes (Extended Data 412

Fig. 9g). Assessment of gene co-expression modules along the HSC-to-SAMes 413

differentiation continuum indicated upregulation of fibrogenic genes including 414

COL1A1, COL1A2, COL3A1, TIMP1, CCL2 and downregulation of genes including 415

15

RGS5, IGFBP5, ADAMTS1 and GEM, which are known to be downregulated in murine 416

HSC in response to liver injury67 (Extended Data Fig. 9h). 417

Resolving the multi-lineage interactome in the fibrotic niche 418

Having defined the scar-associated macrophage, endothelial and mesenchymal 419

populations, we confirmed the close topographical association of these cells within the 420

fibrotic niche (Fig. 6a, b). To interrogate how these newly-defined cellular 421

subpopulations regulate fibrosis and to identify tractable therapeutic targets, we 422

performed an unbiased ligand-receptor interaction analysis between these scar-423

associated populations. We used CellPhoneDB, a repository of curated ligand-receptor 424

interactions integrated with a statistical framework. We calculated statistically 425

significant ligand-receptor pairs, based on expression of receptors by one lineage and 426

ligands by another, using empirical shuffling69. 427

Numerous statistically significant paracrine and autocrine interactions were detected 428

between ligands and cognate receptors expressed by SAM, SAEndo and SAMes within 429

the fibrotic niche (Supplementary Table 17). We focused further functional analyses on 430

interactions with SAMes, the fibrillar collagen-producing population. Both SAM and 431

SAEndo populations demonstrated multiple interactions that could regulate SAMes 432

function (Extended Data Fig. 10a). In keeping with our data demonstrating that human 433

and murine SAM promote fibrillar collagen expression in HSC (Extended Data Fig. 5d, 434

Extended Fig. 6g), SAM expressed a number of epidermal growth factor receptor 435

(EGFR) ligands which are known to regulate mesenchymal cell activation70,71. 436

Additionally, SAM expressed mesenchymal cell mitogens TNFSF12 and PDGFB25,51, 437

signaling to cognate receptors TNFRSF12A and PDGFRA on SAMes (Fig 6c). We 438

confirmed localization of these ligand-receptor pairs within the fibrotic niche of human 439

cirrhotic human liver (Fig. 6d, e). Both TNFSF12 and PDGF-BB induced primary 440

human HSC proliferation, which was inhibited by blockade of TNFSF12A and 441

PDGFRA respectively (Fig. 6f, g). Importantly, conditioned media from primary 442

human SAM promoted primary human HSC proliferation ex vivo (Fig. 6h), 443

demonstrating a functional role for SAM in regulating SAMes expansion. 444

16

SAEndo expressed high levels of non-canonical Notch ligands JAG1, JAG2 and DLL4 445

interacting with Notch receptor NOTCH3 on SAMes (Fig. 6i). Primary endothelial cells 446

from cirrhotic human liver demonstrated increased expression of JAG1 (Fig. 6j), whilst 447

NOTCH3 was identified on PDGFRα+ SAMes within the fibrotic niche (Fig. 6k). Co-448

culture of primary human HSC and endothelial cells from cirrhotic livers promoted 449

fibrillar collagen production by HSC, which was inhibited by addition of the Notch-450

signalling inhibitor Dibenzazepine (DBZ) (Fig. 6l, m). Furthermore, knockdown of 451

NOTCH3 expression in primary human HSC resulted in reduced fibrillar collagen 452

expression (Fig. 6n), confirming that Notch-signalling promotes a fibrogenic 453

mesenchymal cell phenotype45. 454

SAMes expressed a number of ligands demonstrating statistically significant 455

interactions with receptors on SAM and SAEndo (Extended Data Fig. 10b). In 456

particular, SAMes expressed chemokines such as CCL2, which regulates monocyte-457

macrophage recruitment and phenotype (Extended Data Fig. 10c, d). Furthermore, 458

immunoregulatory ligands such as IL34, CSF1 and CX3CL1 expressed by SAMes 459

(Extended Data Fig. 10c) are known to modulate macrophage function, survival and 460

proliferation72,73, potentially explaining the increased proliferation rate observed in 461

human SAM (Extended Data Fig. 4j, k). Intrahepatic angiogenesis is associated with 462

both degree of liver fibrosis and portal hypertension, a major clinical consequence of 463

liver cirrhosis74. Our fibrotic niche interactome analysis confirmed a number of pro-464

angiogenic interactions, with both SAMes (Extended Data Fig. 10c, d) and SAM 465

(Extended Data Fig. 10e-g) expressing angiogenic ligands, with cognate receptors 466

expressed by SAEndo. Additionally, SAEndo expressed CSF1 and CD200 (Extended 467

Data Fig. 10h-j), suggestive of an immunomodulatory role. This is emphasized by the 468

highly significant interactions detected between Notch ligand expression by SAEndo 469

and NOTCH2 expression by SAM (Extended Data Fig. 10i, j). Vascular Notch ligand 470

expression regulates monocyte-derived macrophage differentiation and macrophage 471

function in tissue repair75, congruent with our data demonstrating upregulation of the 472

transcription factor regulon for the Notch-target HES1, during differentiation from 473

monocytes to SAM (Fig. 3h, Extended Data Fig. 5f, g). Hence, SAEndo are likely to 474

regulate immune function as well as leucocyte recruitment to the fibrotic niche 475

(Extended Data Fig. 8e). 476

17

In summary, our unbiased dissection of the key ligand-receptor interactions between 477

novel scar-associated macrophages, endothelial and mesenchymal cells in the fibrotic 478

niche, highlights TNFRSF12A, PDGFRA and Notch signaling as important regulators 479

of mesenchymal cell function and hepatic fibrogenesis. Therapeutic targeting of these 480

intra-scar pathways represents a rational approach for the discovery of novel antifibrotic 481

treatments for patients with chronic liver disease. 482

Discussion 483

The fibrotic niche has not previously been defined in human liver. Here, using scRNA-484

seq and spatial mapping, we resolve the fibrotic niche of human liver cirrhosis, 485

identifying novel pathogenic subpopulations of TREM2+CD9+ pro-fibrogenic 486

macrophages, ACKR1+ and PLVAP+ endothelial cells and PDGFRα+ collagen-487

producing myofibroblasts. We dissect a complex, pro-fibrotic interactome between 488

multiple novel scar-associated cells and identify highly relevant intra-scar pathways 489

that are potentially druggable. This multi-lineage single cell dataset of human liver 490

cirrhosis should serve as a useful resource for the scientific community, and is freely 491

available for interactive browsing at http://www.livercellatlas.mvm.ed.ac.uk 492

Despite significant progress in our understanding of the molecular pathways driving 493

liver fibrosis in rodent models, a lack of corollary studies in diseased human liver tissue 494

has hindered translation into effective therapies, with currently no FDA or EMA-495

approved antifibrotic treatments available. Our multi-lineage ligand-receptor analysis 496

demonstrates the complexity of interactions within the fibrotic niche, highlighting 497

exemplar pathways such as TNFRSF12A, PDGFR and NOTCH signalling as key 498

regulators of mesenchymal cell function in fibrotic human liver. These data provide a 499

conceptual framework for more rational studies of antifibrotic therapies in both pre-500

clinical animal models and translational systems such as human liver organoid 501

cultures5,76,77. Further, this unbiased multi-lineage approach should inform the design 502

of combination therapies which will very likely be necessary to achieve effective 503

antifibrotic potency5,6. 504

Macrophages and endothelial cells are known to regulate liver fibrosis in rodent 505

models14,20,24,48,49. However, little is known regarding the heterogeneity and precise 506

molecular definitions of these cell types in human liver disease. Our data demonstrates 507

18

both the accumulation of discrete monocyte-derived macrophage and endothelial cell 508

populations in the fibrotic niche of cirrhotic livers, but also the persistence of spatially 509

distinct, non-scar associated resident Kupffer cells and liver sinusoidal endothelial 510

cells. This single-cell approach has important implications for therapy development; 511

facilitating targeting of pathogenic cells without perturbing homeostatic function. 512

In this era of precision medicine, where molecular profiling guides the development of 513

highly targeted therapies, we used scRNA-seq to resolve the key non-parenchymal cell 514

subclasses inhabiting the fibrotic niche of human liver cirrhosis. Application of our 515

novel scar-associated cell markers could potentially inform molecular pathology-based 516

patient stratification, which is fundamental to the prosecution of successful antifibrotic 517

clinical trials. Our work illustrates the power of single-cell transcriptomics to decode 518

the cellular and molecular basis of human organ fibrosis, providing a conceptual 519

framework for the discovery of relevant and rational therapeutic targets to treat patients 520

with a broad range of fibrotic diseases. 521

Note to referees: 522

http://www.livercellatlas.mvm.ed.ac.uk will be made freely available. For review 523

purposes, log-in details are username: Edinburghlivercellatlas; password: Cirrhosis; 524

please refresh your browser if the webpage does not load at first attempt. 525

19

Methods 526

Study subjects 527

Local approval for procuring human liver tissue and blood samples for scRNA-seq, 528

flow cytometry and histological analysis was obtained from the NRS BioResource and 529

Tissue Governance Unit (Study Number SR574), following review at the East of 530

Scotland Research Ethics Service (Reference 15/ES/0094). All subjects provided 531

written informed consent. Healthy background non-lesional liver tissue was obtained 532

intraoperatively from patients undergoing surgical liver resection for solitary colorectal 533

metastasis at the Hepatobiliary and Pancreatic Unit, Department of Clinical Surgery, 534

Royal Infirmary of Edinburgh. Patients with a known history of chronic liver disease, 535

abnormal liver function tests or those who had received systemic chemotherapy within 536

the last four months were excluded from this cohort. Cirrhotic liver tissue was obtained 537

intraoperatively from patients undergoing orthotopic liver transplantation at the 538

Scottish Liver Transplant Unit, Royal Infirmary of Edinburgh. Blood from patients with 539

a confirmed diagnosis of liver cirrhosis were obtained from patients attending the 540

Scottish Liver Transplant Unit, Royal Infirmary of Edinburgh. Patients with liver 541

cirrhosis due to viral hepatitis were excluded from the study. Patient demographics are 542

summarised in Extended Data Fig. 1a. Isolation of primary hepatic macrophage 543

subpopulations and endothelial cells from healthy and cirrhotic livers for cell culture 544

and analysis of secreted mediators was performed at the University of Birmingham, 545

UK. Local ethical approval was obtained (Reference 06/Q2708/11, 06/Q2702/61) and 546

all patients provided written, informed consent. Liver tissue was acquired from 547

explanted diseased livers from patients undergoing orthotopic liver transplantation, 548

resected liver specimens or donor livers rejected for transplant at the Queen Elizabeth 549

Hospital, Birmingham. For histological assessment of NAFLD biopsies, anonymised 550

unstained formalin-fixed paraffin-embedded liver biopsy sections encompassing the 551

complete NAFLD spectrum were provided by the Lothian NRS Human Annotated 552

Bioresource under authority from the East of Scotland Research Ethics Service REC 1, 553

reference 15/ES/0094. 554

Human tissue processing 555

20

For human liver scRNA-seq and flow cytometry analysis, a wedge biopsy of non-556

ischaemic fresh liver tissue (2-3 grams) was obtained by the operating surgeon, prior to 557

interruption of the hepatic vascular inflow. This was immediately placed in HBSS 558

(Gibco) on ice. The tissue was then transported directly to the laboratory and 559

dissociation routinely commenced within 20 minutes of the liver biopsy. To enable 560

paired histological assessment, a segment of each liver specimen was also fixed in 4% 561

neutral-buffered formalin for 24 hours followed by paraffin-embedding. Additional 562

liver samples, obtained via the same method, were fixed in an identical manner and 563

used for further histological analysis. For human macrophage cell sorting and 564

endothelial cell isolation, liver tissue (40 grams) was used from cirrhotic patients 565

undergoing orthotopic liver transplantation or control samples from donor liver or liver 566

resection specimens. 567

Animals 568

Adult male C57BL/6JCrl mice aged 8-10 weeks were purchased from Charles River. 569

Mice were housed under specific pathogen-free conditions at the University of 570

Edinburgh. All experiments were performed in accordance with UK Home Office 571

regulations. Liver fibrosis was induced with 4 weeks (9 injections) of twice-weekly 572

intraperitoneal carbon tetrachloride (CCl4) at a dose of 0.4 μl/g body weight, diluted 573

1:3 in olive oil as previously described15. Liver tissue was harvested 24 hours following 574

the final CCl4 injection, a time of active fibrogenesis15. Comparison was made to age-575

matched uninjured mice. 576

Preparation of single-cell suspensions 577

For human liver scRNA-seq, liver tissue was minced with scissors and digested in 578

5mg/ml pronase (Sigma-Aldrich, P5147-5G), 2.93mg/ml collagenase B (Roche, 579

11088815001) and 1.9mg/ml DNase (Roche, 10104159001) at 37°C for 30 minutes 580

with agitation (200–250 r.p.m.), then strained through a 120μm nybolt mesh along with 581

PEB buffer (PBS, 0.1% BSA, and 2mM EDTA) including DNase (0.02mg/ml). 582

Thereafter all processing was done at 4oC. The cell suspension was centrifuged at 400g 583

for 7 minutes, supernatant removed, cell pellet resuspended in PEB buffer and DNase 584

added (0.02mg/ml), followed by additional centrifugation (400g, 7 minutes). Red blood 585

21

cell lysis was performed (BioLegend, 420301), followed by centrifugation (400g, 7 586

minutes), resuspension in PEB buffer and straining through a 35μm filter. Following 587

another centrifugation at 400g for 7 minutes, cells were blocked in 10% human serum 588

(Sigma-Aldrich, H4522) for 10 minutes at 4oC prior to antibody staining. 589

For both human liver macrophage flow cytometry analysis and cell sorting and mouse 590

liver macrophage flow cytometry, cell sorting and scRNA-seq, single-cell suspensions 591

were prepared as previously described, with minor modifications78. In brief, liver tissue 592

was minced and digested in an enzyme cocktail 0.625 mg/ml collagenase D (Roche, 593

11088882001), 0.85 mg/ml collagenase V (Sigma-Aldrich, C9263-1G), 1 mg/ml 594

dispase (Gibco, Invitrogen, 17105-041), and 30 U/ml DNase (Roche, 10104159001) in 595

RPMI-1640 at 37°C for 20 minutes (mouse) or 45 minutes (human) with agitation 596

(200–250 r.p.m.), before being passed through a 100μm filter. Following red blood cell 597

lysis (BioLegend, 420301), cells were washed in PEB buffer and passed through a 598

35μm filter. Before the addition of antibodies, cells from human samples were blocked 599

in 10% human serum (Sigma-Aldrich, H4522) and mouse samples were blocked in 600

anti-mouse CD16/32 antibody (1:100; Biolegend, 101302) and 10% normal mouse 601

serum (Sigma, M5905) for 10 minutes at 4 °C. 602

For human PBMC scRNA-seq, 4.9ml peripheral venous blood samples were collected 603

in EDTA-coated tubes (Sarstedt, S-MonovetteÒ 4.9ml K3E) and placed on ice. Blood 604

samples were transferred into a 50ml Falcon tube. Following red cell lysis (Biolegend, 605

420301), blood samples were then centrifuged at 500g for 5 minutes and supernatant 606

was removed. Pelleted samples were then resuspended in staining buffer (PBS plus 2% 607

BSA; Sigma-Aldrich) and centrifugation was repeated. Samples were then blocked in 608

10% human serum (Sigma-Aldrich, H4522) in staining buffer on ice for 30 minutes. 609

Cells were then resuspended in staining buffer and passed through a 35μm filter prior 610

to antibody staining. 611

Flow cytometry and cell sorting 612

Incubation with primary antibodies was performed for 20 minutes at 4°C. All 613

antibodies, conjugates, lot numbers and dilutions used in this study are presented in 614

Supplementary Table 18. Following antibody staining, cells were washed with PEB 615

22

buffer. For both human macrophage flow cytometry analysis and cell sorting, cells were 616

then incubated with streptavidin-BV711 for 20 minutes at 4°C (Biolegend 405241; 617

Dilution 1:200). For human and mouse cell sorting (FACS) and mouse flow cytometry 618

analysis, cell viability staining (DAPI; 1:1000 dilution) was then performed, 619

immediately prior to acquiring the samples. 620

Human cell sorting for scRNA-seq was performed on a BD Influx (Becton Dickinson, 621

Basel, Switzerland). Viable single CD45+ (leucocytes) or CD45- (other non-622

parenchymal cells) cells were sorted from human liver tissue (Extended Data Fig. 1b) 623

and viable CD45+ CD66b- (PBMC) cells were sorted from peripheral blood (Extended 624

Data Fig. 1c) and processed for droplet-based scRNA-seq. 625

To generate conditioned media from cirrhotic liver macrophage subpopulations, cells 626

were sorted on a BD FACSAriaTM Fusion (Becton Dickinson, Basel, Switzerland). 627

Sorted SAM (viable CD45+Lin-HLA-DR+CD14+CD16+CD163-TREM2+CD9+), TMo 628

(viable CD45+Lin- HLA-DR+CD14+CD16+CD163-TREM2-CD9-) and KC (viable 629

CD45+Lin- HLA-DR+CD14+CD16+CD163+CD9-) were plated in 12-well plates 630

(Corning, 3513) in DMEM (Gibco, 41965039) containing 2% FBS (Gibco, 10500056) 631

at 1x106 cells/ml for 24 hours at 37°C 5%CO2. Control wells contained media alone. 632

Conditioned media was collected, centrifuged at 400g for 10 minutes and supernatant 633

stored at -80°C. 634

For human macrophage flow cytometry analysis, following surface antibody staining, 635

cells were stained with Zombie NIR fixable viability dye (Biolegend, 423105) 636

according to manufacturers’ instructions. Cells were washed in PEB then fixed in IC 637

fixation buffer (Thermo-Fisher, 00-8222-49) for 20 minutes at 4°C. Fixed samples were 638

stored in PEB at 4°C until acquisition. Flow cytometry acquisition was performed on 639

6-laser Fortessa flow cytometer (Becton Dickinson, Basel, Switzerland). The gating 640

strategy is shown (Extended Data Fig. 4f, Fig. 2f). 641

Mouse macrophage cell sorting for scRNA-seq and co-culture experiments was 642

performed on a BD FACSAriaII (Becton Dickinson, Basel, Switzerland). For scRNA-643

seq, viable CD45+ Lin(CD3, NK1.1, Ly6G, CD19)- cells were sorted from healthy 644

(n=3) and CCl4-treated (n=3) mice and processed for droplet-based scRNA-seq. For 645

23

transwell co-culture, viable CD45+ Lin- CD11b+ F4/80+ TIMD4- CD9+ (mSAM) or 646

CD9- (mTMo) cells were sorted from CCl4-treated mice (Extended Data Fig. 6e). Flow 647

cytometry analysis on macrophages from healthy and CCl4-treated mice was also 648

performed on a BD FACSAriaII (Becton Dickinson, Basel, Switzerland), using the 649

same gating strategy (Extended Data Fig. 6e). All flow cytometry data was analysed 650

using Flowjo software (Treestar, Ashland, TN). 651

Luminex Assay 652

Detection of CCL2, Galectin-3, IL-1 beta, CXCL8 and Osteopontin (SPP1) proteins in 653

conditioned media from human liver macrophage subpopulations was performed using 654

a custom human luminex assay (R&D systems), according to the manufacturers 655

protocol. Data was acquired using a Bio-PlexÒ 200 (Bio-Rad, UK) and is presented a 656

median fluorescence intensity (MFI) for each analyte. 657

Cell Culture 658

Human hepatic stellate cell activation 659

Primary human hepatic stellate cells (HSC) were purchased (ScienCell, 5300) and 660

cultured in stellate cell medium (SteCM, ScienCell, 5301) on Poly-L-Lysine (Sigma, 661

P4832) coated T75 tissue culture flasks, according to the suppliers protocol. All 662

experiments were performed using cells between passage 3 and 5. For assessment of 663

fibrillar collagen gene expression, HSC were plated at 75,000 cells per well in 24 well-664

plates (Costar, 3524) in HSC media consisting of DMEM (Gibco, 21969-035) 665

supplemented with 20 µM HEPES (Sigma, H3375,), 2 mM L-Glutamine (Gibco , 666

25030-024), 1% Penicillin Streptomycin (Gibco, 15140-122, Gibco) and 2% Foetal 667

Bovine Serum (FBS, Gibco, 10270). HSC were serum starved overnight (in HSC media 668

without FBS), washed with PBS, then 250μl of conditioned media from primary human 669

macrophage subpopulations added for 24 hours. HSC were harvested for RNA. 670

Human hepatic stellate cell proliferation 671

For proliferation assays, following serum starvation HSC were harvested using TrypLE 672

Express (Gibco, Cat. no. 12604013), re-suspended in HSC media at 2.5x104/ml with 673

Incucyte NucLight Rapid Red (Essen Biosciences, 4717) at a dilution of 1 in 500 and 674

24

seeded into 384 well plates (Greiner Bio-One, 781090) at 25μl per well. HSC were then 675

treated with control media or (i) PDGF-BB (10ng/ml; Peprotech, 100-14B) or 676

TNFSF12 (100ng/ml; Peprotech, 310-06-5) with or without the PDGFR⍺ inhibitor 677

Crenolanib79 (1μM; Cayman chemicals, CAY1873), anti-TNRSF12A (2μg/ml; Life 678

Technologies, 16-9018-82, clone ITEM-4), mouse IgG2b kappa isotype control 679

antibody (2μg/ml; Life Technologies, 16-4732-82, clone eBMG2b) or vehicle control 680

as indicated or (ii) conditioned media from human hepatic macrophage subpopulations 681

as indicated. The final volume was 50μl for all conditions. Cells were then incubated 682

in an Incucyte ZOOM live cell analysis system (Essen Biosciences) humidified at 37°C 683

with 5% CO2 with imaging every 3 hours using the 10x optic for either 87 hours 684

(recombinant cytokines/inhibitors) or 39 hours (macrophage conditioned media). 685

Analysis was performed with the Incucyte proprietary analysis software (version 686

2018A) by using machine learning to distinguish the individual nuclei (stained red by 687

the NucLight Rapid Red dye) and perform nuclear counts of the images at each 3 hour 688

time point over the period of culture. Data are expressed as area under curve (AUC) 689

for % change in nuclear number from baseline versus time (hours), calculated in 690

GraphPad Prism (GraphPad Software, USA). 691

Gene knockdown in human hepatic stellate cells 692

Knockdown of NOTCH3 in human HSC was performed using siRNA. HSC were plated 693

at 75,000 cells per well in a 12 well plate (Costar, 3513) followed by serum starvation 694

overnight (in HSC medium without FBS). siRNA duplexes with Lipofectamine 695

RNAiMAX Transfection Reagent (ThermoFisher, 13778075) were prepared in 696

OptiMEM (ThermoFisher, 31985070) according to the manufacturer’s 697

recommendations, and used at a concentration of 50nM. Cells were exposed to the 698

duplex for 48 hours, in HSC media containing 2% FBS. Cells were harvested for RNA 699

and RT-qPCR. Knockdown efficiency was assessed by NOTCH3 RT-qPCR. The best 700

siRNA for knockdown was determined empirically using the FlexiTube GeneSolution 701

kit (Qiagen, GS4854). HSC treated with control siRNA (Qiagen, 1027280) and siRNA 702

for NOTCH3 (Qiagen, Hs_NOTCH3_3, SI00009513; knockdown 83%) were then 703

assessed for fibrillar collagen gene expression. 704

Mouse hepatic stellate cell activation 705

25

Primary murine HSC culture were isolated from healthy mice as described80. Briefly, 706

after cannulation of the inferior vena cava, the portal vein was cut to allow retrograde 707

step-wise perfusion with pronase (Sigma, P5147) and collagenase D (Roche, 708

11088882001) containing solutions, before ex vivo digestion in a solution containing 709

pronase, collagenase D and Dnase1 (Roche, 10104159001). HSC were isolated from 710

the digest solution by Histodenz (Sigma, D2158-100G) gradient centrifugation. HSC 711

were plated at a density of 400,000 cells per well in a 24 well plate (Costar, 3524) in 712

HSC media containing 10% FBS. Following overnight culture, cells were washed with 713

PBS and cultured in HSC media containing 2% FBS. For macrophage co-culture, 714

transwell inserts (0.4μm polyester membrane; Costar, 3470) were then placed above 715

adherent HSC. FACS-sorted CD9+ mSAM or CD9- mTMo from CCl4-treated mice 716

were resuspended in HSC media containing 2% FBS at 400,000 cells/ml and 200,000 717

cells added to the top of the transwell insert. Co-culture proceeded for 48 hours and 718

HSC were harvested for RNA. Quiescent HSC (harvested at start of co-culture) were 719

used as a control population. 720

Human liver endothelial cell isolation 721

Human liver endothelial cells (LEC) were isolated from cirrhotic explant livers and 722

non-fibrotic control donor liver as previously described81. Endothelial cells were 723

cultured on plasticware coated with rat-tail collagen (Sigma, C3867) in complete LEC 724

medium consisting of endothelial basal media (ThermoFisher, 11111044) containing 725

10% heat inactivated human serum (tcsBiosciences, CS100-500), 100U penicillin, 100 726

µg/mL streptomycin, 2mM glutamine (Sigma, G6784), VEGF (10 ng/mL; Peprotech, 727

100-20) and 10 ng/mL HGF (10 ng/mL; Peprotech, 100-39). LEC expression of 728

PLVAP, CD34, ACKR1 and JAG1 was assessed using flow cytometry. 729

Flow-based adhesion assays 730

Flow-based adhesion assays were performed as described56,81. Briefly, LEC from 731

healthy and cirrhotic liver were seeded into a rat-tail collagen coated Ibidi slide VI0.4 732

(Ibidi, 80606) at a density to give a monolayer and incubated overnight. Peripheral 733

blood was collected from healthy donors in EDTA-coated tubes. Peripheral blood 734

monocuclear cells (PBMC) were isolated using a lympholyte density gradient 735

(Cedarlane laboratories) then washed in PBS containing 1mM Ca2+, 0.5 mM Mg2+ and 736

26

0.15% bovine serum albumin (PBS/BSA). Monocytes were enriched from PBMC using 737

pan-monocyte isolation kit (Miltenyi biotech, 130-096-537) according the 738

manufacturer’s protocol. For flow-based adhesion assay, cells were resuspended at 106 739

cells per millilitre in endothelial basal media (ThermoFisher, 11111044) containing 740

0.15% BSA, then perfused over the LEC monolayer for 5min at 0.28ml/min. Non-741

adherent cells were washed off during 5min perfusion of 0.15% BSA human basal 742

endothelial medium and 10 random non-overlapping images were randomly recorded 743

from each channel. Total adherent (bright-phase; expressed as cell number/mm2/1 744

million cells perfused) and transmigrating cells (dark-phase; expressed as percentage 745

total adherent cells) on an LEC monolayer from each patient were counted and 746

quantified as previously described56. 747

Gene knockdown in endothelial cells 748

Knockdown of ACKR1 and PLVAP gene expression in human cirrhotic LEC was 749

performed using siRNA as previously described81. In brief, siRNA duplexes for 750

PLVAP, ACKR1, or negative control (Qiagen, 1027280) with Lipofectamine 751

RNAiMAX Transfection Reagent (ThermoFisher, 13778075) were prepared in 752

OptiMEM (ThermoFisher, 31985070) according to the manufacturer’s 753

recommendations, and used at a concentration of 25nM. Cells were exposed to the 754

duplex for 4 hours at 37oC after which time the media was replaced with endothelial 755

basal media containing 10% heat-inactivated human serum for 24 hours. The media 756

was then replaced with complete LEC media and incubated at 37oC with 5% CO2 for a 757

further 24 hours. Knockdown efficacy was assessed by flow cytometry and mean 758

fluorescence intensity (Extended Data Fig. 8f). The best siRNA for knockdown was 759

determined empirically using the FlexiTube GeneSolution kit (Qiagen, GS83483 760

(PLVAP) and GS2532 (ACKR1)). For flow-based adhesion assays, siRNAs for 761

PLVAP (Qiagen, Hs_PLVAP_1, SI00687547; knockdown 50.6%, ACKR1 (Qiagen, 762

Hs_Fy_5, SI02627667; knockdown 37.7%) or control siRNA were selected. 90,000 763

LEC from cirrhotic patients (n=6) were seeded into channels of a rat-tail collagen 764

coated Ibidi slide VI0.4, gene knockdown performed, followed by flow-based adhesion 765

assay as described above. 766

Endothelial and hepatic stellate cell co-culture 767

27

HSC (15,000 cells) were seeded into an Ibidi slide VI0.4 with and without primary 768

human LEC (15,000 cells) from individual cirrhotic patients (n=3) in LEC complete 769

medium. After 2h, all growth factor supplements were removed and cells were cultured 770

for a further 72 hours in endothelial basal media containing 10% heat-inactivated 771

human serum ± Notch-signalling inhibitor Dibenzazepine (Bio-Techne, 4489/10) or 772

vehicle (DMSO) control. Cells were fixed in 4% PFA for 30 minutes, permeabilised 773

with 0.3% Triton PBS for 5 minutes, blocked with 10% goat serum in PBS for 30 774

minutes followed by primary antibody incubation (mouse anti-PECAM1 and rabbit 775

anti-collagen 1; see Supplementary table 18) for 1 hour. Cells were washed in 0.1% 776

Triton PBS followed by addition of fluorescently-conjugated secondary antibodies 777

(1:500 dilution) for 1 hour. Cells were mounted with Pro-long Gold anti-fade DAPI, 778

images were taken on the Confocal Microscope Zeiss LSM780, and collagen 1 area 779

staining quantified using IMARIS. 780

RNA extraction and RT-qPCR 781

RNA was isolated from HSC using the RNeasy Plus Micro Kit (Qiagen, 74034) and 782

cDNA synthesis performed using QuantiTect Reverse Transcription Kit (Qiagen, 783

205313) according to the manufacturer’s protocol. Reactions were performed in 784

triplicate in 384-well plate format and were assembled using the QIAgility automated 785

pipetting system (Qiagen). RT-qPCR for human HSC was performed using PowerUp 786

SYBR Green Master Mix (ThermoFisher, A25777) with the following primers (all 787

Qiagen): GAPDH (QT00079247), COL1A1 (QT00037793), COL3A1 (QT00058233), 788

NOTCH3 (QT00003374). RT-qPCR for mouse HSC was performed using TaqMan 789

Fast Advanced Master Mix (ThermoFisher, 4444557) with the following primers: 790

Gapdh (ThermoFisher, Mm99999915_g1) and Col3a1 (ThermoFisher, 791

Mm00802300_m1). Samples were amplified on an ABI 7900HT FAST PCR system 792

(Applied Biosystems, ThermoFisher Scientific). Data was analysed using 793

ThermoFisher Connect cloud qPCR analysis software (ThermoFisher Scientific). The 794

2−ΔΔCt quantification method, using GAPDH for normalization, was used to estimate 795

the amount of target mRNA in samples, and expression calculated relative to average 796

mRNA expression levels from control samples. 797

28

Immunohistochemistry, immunofluorescence, smFISH 798

Formalin-fixed paraffin-embedded human liver tissue was cut into 4 μm sections, 799

dewaxed, rehydrated, then incubated in 4% neutral-buffered formalin for 20 minutes. 800

Following heat-mediated antigen retrieval in pH6 sodium citrate (microwave; 15 801

minutes), slides were washed in PBS and incubated in 4% hydrogen peroxide for 10 802

minutes. Slides were then washed in PBS, blocked using protein block (GeneTex, 803

GTX30963) for 1 hour at room temperature before incubation with primary antibodies 804

for 1 hour at room temperature. A full list of primary antibodies and conditions are 805

shown in Supplementary Table 18. Slides were washed in PBST (PBS plus 0.1% 806

Tween20; Sigma-Aldrich, P1379) then incubated with ImmPress HRP Polymer 807

Detection Reagents (depending on species of primary; rabbit, MP-7401; mouse, MP-808

6402-15; goat, MP-7405; all Vector Laboratories) for 30 minutes at room temperature. 809

Slides were washed in PBS followed by detection. For DAB staining, sections were 810

incubated with DAB (DAKO, K3468) for 5 minutes and washed in PBS before a 811

haematoxylin (Vector Laboratories, H3404) counterstain. For multiplex 812

immunofluorescence staining, following the incubation with ImmPress and PBS wash, 813

initial staining was detected using either Cy3, Cy5, or Fluorescein tyramide (Perkin-814

Elmer, NEL741B001KT) at a 1:1000 dilution. Slides were then washed in PBST 815

followed by further heat treatment with pH6 sodium citrate (15 minutes), washes in 816

PBS, protein block, incubation with the second primary antibody (incubated overnight 817

at 4oC), ImmPress Polymer and tyramide as before. This sequence was repeated for the 818

third primary antibody (incubated at room temperature for 1 hour) and a DAPI-819

containing mountant was then applied (ThermoFisher Scientific, P36931). All 820

immunofluorescence stains were repeated in a minimum of 3 patients and 821

representative images are displayed. 822

For AMEC Staining (only CLEC4M immunohistochemistry), all washes were carried 823

out with TBST (dH2O, 2oomM Tris, 1.5M NaCl, 1% Tween20 (all Sigma-Aldrich) 824

pH7.5) and peroxidase blocking was carried out for 30mins in 0.6% hydrogen peroxide 825

in Methanol. Sections were incubated with AMEC (Vector Laboratories, SK-4285) for 826

20 minutes and washed in TBST (dH2O, 200mM Tris, 1.5M NaCl, 1% Tween20 (all 827

Sigma-Aldrich)) before a haematoxylin (Vector Laboratories, SK-4285) counterstain. 828

29

For combined single-molecule fluorescent in situ hybridization (smFISH) and 829

immunofluorescence, detection of TREM2 was performed using the RNAscope® 2.5 830

LS Reagent Kit - BrownAssay (Advanced Cell Diagnostics (ACD)) in accordance with 831

the manufacturer’s instructions. Briefly, 5 μm tissue sections were dewaxed, incubated 832

with endogenous enzyme block, boiled in pretreatment buffer and treated with protease, 833

followed by target probe hybridization using the RNAscope® LS 2.5 Hs-TREM2 834

(420498, ACD) probe. Target RNA was then detected with Cy3 tyramide (Perkin-835

Elmer, NEL744B001KT) at 1:1000 dilution. The sections were then processed through 836

a pH6 sodium citrate heat-mediated antigen retrieval, hydrogen peroxidase treatment 837

and protein block (all as for multiplex immunofluorescence staining as above). MNDA 838

antibody was applied overnight at 4oC, completed using a secondary ImmPress HRP 839

Anti-Rabbit Peroxidase IgG (Vector Laboratories, MP7401), visualised using a 840

Flourescein tyramide (Perkin-Elmer, NEL741B001KT) at a 1:1000 dilution and stained 841

with DAPI. 842

Brightfield and fluorescently-stained sections were imaged using the slide scanner 843

AxioScan.Z1 (Zeiss) at 20X magnification (40X magnification for smFISH). Images 844

were processed and scale bars added using Zen Blue (Zeiss) and Fiji software82. 845

Cell counting and image analysis 846

Automated cell counting was performed using QuPath software83. Briefly, DAB-847

stained whole tissue section slide-scanned images (CZI files) were imported into 848

QuPath. Cell counts were carried out using the positive cell detection tool, detecting 849

haematoxylin-stained nuclei and then thresholding for positively-stained DAB cells, 850

generating DAB-positive cell counts/mm2 tissue. Identical settings and thresholds were 851

applied to all slides for a given stain and experiment. For cell counts of fibrotic septae 852

vs parenchymal nodules, the QuPath segmentation tool was used to segment the DAB-853

stained whole tissue section into fibrotic septae or non-fibrotic parenchymal nodule 854

regions using tissue morphological characteristics (Fig. 2j). Positive cell detection was 855

then applied to the fibrotic and non-fibrotic regions in turn, providing cell DAB-positive 856

cell counts/mm2 in fibrotic septae and non-fibrotic parenchymal nodules for each tissue 857

section. 858

30

Digital morphometric pixel analysis was performed using the Trainable Weka 859

Segmentation (TWS) plugin84 in Fiji software82. Briefly, each stained whole tissue 860

section slide-scanned image was converted into multiple TIFF files in Zen Blue 861

software (Zeiss). TIFF files were imported into Fiji and TWS plugin trained to produce 862

a classifier which segments images into areas of positive staining, tissue background 863

and white space84. The same trained classifier was then applied to all TIFF images from 864

every tissue section for a particular stain, providing a percentage area of positive 865

staining for each tissue section. For digital morphometric quantification of positive 866

staining of fibrotic septae versus parenchymal nodules, TIFF images were segmented 867

into fibrotic septae or non-fibrotic parenchymal nodule regions using tissue 868

morphological characteristics, followed by analysis using the TWS plugin in Fiji 869

software. 870

Histological assessment of NASH sections 871

Haematoxylin and eosin and picro-sirius red stained sections from each case were 872

whole-slide imaged using a NanoZoomer imager (Hamamatsu Photonics, Japan). 873

Images of stained sections were independently scored by a consultant liver transplant 874

histopathologist (T.J.K.) at the national liver transplant centre with experience in trial 875

scoring by applying the ordinal NAFLD activity score85. For observer-independent 876

quantification of picro-sirius red positive staining, images were split using ndpisplit86 877

into tiles of x5 magnification before the application of a classifier that had been trained 878

by the liver histopathologist using the machine learning WEKA plugin in FIJI82,84, as 879

previously described87. All analysis was undertaken blind to all other data. 880

Droplet-based scRNA-seq 881

Single cells were processed through the Chromiumä Single Cell Platform using the 882

Chromiumä Single Cell 3’ Library and Gel Bead Kit v2 (10X Genomics, PN-120237) 883

and the Chromiumä Single Cell A Chip Kit (10X Genomics, PN-120236) as per the 884

manufacturer’s protocol. In brief, single cells were sorted into PBS + 0.1% BSA, 885

washed twice and counted using a Bio-Rad TC20. 10,769 cells were added to each lane 886

of the 10X chip. The cells were then partitioned into Gel Beads in Emulsion in the 887

31

Chromiumä instrument, where cell lysis and barcoded reverse transcription of RNA 888

occurred, followed by amplification, fragmentation and 5′ adaptor and sample index 889

attachment. Libraries were sequenced on an Illumina HiSeq 4000. 890

Computational Analysis 891

Pre-processing scRNA-seq data 892

We aligned to the GRCh38 and mm10 (Ensembl 84) reference genomes as appropriate 893

for the input dataset, and estimated cell-containing partitions and associated UMIs, 894

using the Cell Ranger v2.1.0 Single-Cell Software Suite from 10X Genomics. Genes 895

expressed in fewer than three cells in a sample were excluded, as were cells that 896

expressed fewer than 300 genes or mitochondrial gene content >30% of the total UMI 897

count. We normalised by dividing the UMI count per gene by the total UMI count in 898

the corresponding cell and log-transforming. Variation in UMI counts between cells 899

was regressed according to a negative binomial model, prior to scaling and centering 900

the resulting value by subtracting the mean expression of each gene and dividing by its 901

standard deviation (En), then calculating ln(104*En+1). 902

Dimensionality reduction, clustering, and DE analysis 903

We performed unsupervised clustering and differential gene expression analyses in the 904

Seurat R package v2.3.088. In particular we used SNN graph-based clustering, where 905

the SNN graph was constructed using from 2 to 11 principal components as determined 906

by dataset variability shown in principal components analysis (PCA); the resolution 907

parameter to determine the resulting number of clusters was also tuned accordingly. To 908

assess cluster similarity we used the BuildClusterTree function from Seurat. 909

In total, we present scRNA-seq data from ten human liver samples (named Healthy 1-910

5 and Cirrhotic 1-5), five human blood samples (n=4 cirrhotic named Blood 1-4 and 911

n=1 healthy named PBMC8K; pbmc8k dataset sourced from single-cell gene 912

expression datasets hosted by 10X Genomics), and two mouse liver samples (n=3 913

uninjured and n=3 fibrotic). For seven human liver samples (Healthy 1-4 and Cirrhotic 914

1-3) we performed scRNA-seq on both leucocytes (CD45+) and other non-parenchymal 915

cells (CD45-); for the remaining three human livers (Healthy 5, Cirrhotic 4-5) we 916

performed scRNA-seq on leucocytes only (Extended Data Fig. 2e, g). 917

32

Initially, we combined all human scRNA-seq datasets (liver and blood) and performed 918

clustering analysis with the aim of isolating a population of liver-resident cells, by 919

identifying contaminating circulatory cells within datasets generated from liver digests 920

and removing them from downstream analysis. Specifically, we removed from our liver 921

datasets cells that fell into clusters 1 and 13 of the initial dataset in Extended Data Fig. 922

1d. 923

Using further clustering followed by signature analysis, we interrogated this post-924

processed liver-resident cell dataset for robust cell lineages. These lineages were 925

isolated into individual datasets, and the process was iterated to identify robust lineage 926

subpopulations. At each stage of this process we removed clusters expressing more than 927

one unique lineage signature in more than 25% of their cells from the dataset as 928

probable doublets. Where the cell proliferation signature identified distinct cycling 929

subpopulations, we re-clustered these again to ascertain the identity of their constituent 930

cells. 931

The murine scRNA-seq datasets were combined, clustered, and interrogated for cell 932

lineages in a similar manner as their human counterparts. 933

All heatmaps, t-SNE and UMAP visualisations, violin plots, and dot plots were 934

produced using Seurat functions in conjunction with the ggplot2, pheatmap, and grid 935

R packages. t-SNE and UMAP visualisations were constructed using the same number 936

of principal components as the associated clustering, with perplexity ranging from 30 937

to 300 according to the number of cells in the dataset or lineage. We conducted 938

differential gene expression analysis in Seurat using the standard AUC classifier to 939

assess significance. We retained only those genes with a log-fold change of at least 0.25 940

and expression in at least 25% of cells in the cluster under comparison. 941

Defining cell lineage signatures 942

For each cell we obtained a signature score across a curated list of known marker genes 943

per cell lineage in the liver (Supplementary Table 1). This signature score was defined 944

as the geometric mean of the expression of the associated signature genes in that cell. 945

Lineage signature scores were scaled from 0 to 1 across the dataset, and the score for 946

33

each cell with signature less than a given threshold (the mean of said signature score 947

across the entire dataset) was set as 0. 948

Batch effect and quality control 949

To investigate agreement between samples we extracted the average expression profile 950

for a given cell lineage in each sample, and calculated the Pearson correlation 951

coefficients between all possible pairwise comparisons of samples per lineage89. 952

Imputing dropout in T cell and ILC clusters 953

To impute dropout of low-abundance transcripts in our T cell and ILC clusters so that 954

we might associate them with known subpopulations, we down-sampled to 7,380 cells 955

from 36,900 and applied the scImpute R package v0.0.890, using as input both our 956

previous annotation labels and k-means spectral clustering (k=5), but otherwise default 957

parameters. 958

Analysing functional phenotypes of scar-associated cells 959

For further analysis of function we adopted the self-organising maps (SOM) approach 960

as implemented in the SCRAT R package v1.0.091. For each lineage of interest we 961

constructed a SOM in SCRAT using default input parameters and according to its 962

clusters. We defined the signatures expressed in a cell by applying a threshold criterion 963

(ethresh = 0.95 × emax) selecting the highest-expressed metagenes in each cell, and 964

identified for further analysis those metagene signatures defining at least 30% of cells 965

in at least one cluster within the lineage. We smoothed these SOMs using the 966

disaggregate function from the raster R package for visualisation purposes, and scaled 967

radar plots to maximum proportional expression of the signature. Gene ontology 968

enrichment analysis on the genes in these spots was performed using PANTHER 13.1 969

(pantherdb.org). 970

Inferring injury dynamics and transcriptional regulation 971

To generate cellular trajectories (pseudotemporal dynamics) we used the monocle R 972

package v2.6.192. We ordered cells in a semi-supervised manner based on their Seurat 973

clustering, scaled the resulting pseudotime values from 0 to 1, and mapped them onto 974

either the t-SNE or UMAP visualisations generated by Seurat or diffusion maps as 975

implemented in the scater R package v1.4.093 using the top 500 variable genes as input. 976

34

We removed mitochondrial and ribosomal genes from the geneset for the purposes of 977

trajectory analysis. Differentially-expressed genes along this trajectory were identified 978

using generalised linear models via the differentialGeneTest function in monocle. 979

When determining significance for differential gene expression along the trajectory, we 980

set a q-value threshold of 1e-20. We clustered these genes using hierarchical clustering 981

in pheatmap, cutting the tree at k=3 to obtain gene modules with correlated gene 982

expression across pseudotime. Cubic smoothing spline curves were fitted to scaled gene 983

expression along this trajectory using the smooth.spline command from the stats R 984

package, and gene ontology enrichment analysis again performed using PANTHER 985

13.1. 986

We verified the trajectory and its directionality using the velocyto R package v0.6.034, 987

estimating cell velocities from their spliced and unspliced mRNA content. We 988

generated annotated spliced and unspliced reads from the 10X BAM files via the 989

dropEst pipeline, before calculating gene-relative velocity using kNN pooling with 990

k=25, determining slope gamma with the entire range of cellular expression, and fitting 991

gene offsets using spanning reads. Aggregate velocity fields (using Gaussian smoothing 992

on a regular grid) and transition probabilities per lineage subpopulations were 993

visualised on t-SNE, UMAP, or diffusion map visualisations as generated previously. 994

Gene-specific phase portraits were plotted by calculating spliced and unspliced mRNA 995

levels against steady-state inferred by a linear model; levels of unspliced mRNA above 996

and below this steady-state indicate increasing and decreasing expression of said gene, 997

respectively. Similarly we plotted unspliced count signal residual per gene, based on 998

the estimated gamma fit, with positive and negative residuals indicating expected 999

upregulation and downregulation respectively. 1000

For transcription factor analysis, we obtained a list of all genes identified as acting as 1001

transcription factors in humans from AnimalTFDB94. To further analyse transcription 1002

factor regulons, we adopted the SCENIC v0.1.7 workflow in R95, using default 1003

parameters and the normalised data matrices from Seurat as input. For visualisation, we 1004

mapped the regulon activity (AUC) scores thus generated to the pseudotemporal 1005

trajectories from monocle and the clustering subpopulations from Seurat. 1006

35

Analysing inter-lineage interactions within the fibrotic niche 1007

For comprehensive systematic analysis of inter-lineage interactions within the fibrotic 1008

niche we used CellPhoneDB69. CellPhoneDB is a manually curated repository of 1009

ligands, receptors, and their interactions, integrated with a statistical framework for 1010

inferring cell-cell communication networks from single-cell transcriptomic data. In 1011

brief, we derived potential ligand-receptor interactions based on expression of a 1012

receptor by one lineage subpopulation and a ligand by another; as input to this algorithm 1013

we used cells from the fibrotic niche as well as liver sinusoidal endothelial cells and 1014

Kupffer cells as control, and we considered only ligands and receptors expressed in 1015

greater than 5% of the cells in any given subpopulation. Subpopulation-specific 1016

interactions were identified as follows: 1) randomly permuting the cluster labels of all 1017

cells 1000 times and determining the mean of the average receptor expression of a 1018

subpopulation and the average ligand expression of the interacting subpopulation, thus 1019

generating a null distribution for each ligand-receptor pair in each pairwise comparison 1020

between subpopulations, 2) calculating the proportion of these means that were "as or 1021

more extreme" than the actual mean, thus obtaining a p-value for the likelihood of 1022

subpopulation specificity for a given ligand-receptor pair, 3) prioritising interactions 1023

that displayed specificity to subpopulations interacting within the fibrotic niche. 1024

Canonical correlation analysis 1025

To compare human and murine populations of monocytic phagocytes, we used 1026

canonical correlation analysis as implemented in Seurat39. We map the genes in the 1027

human dataset to their murine orthologues using biomaRt, discarding any genes for 1028

which no orthologues can be found. We then calculate the shared low-dimensional 1029

subspace on the union of genes that are variably expressed in both datasets (n=159), 1030

and align using six canonical components as determined by evaluating the biweight 1031

midcorrelation. Results are visualised on t-SNEs as previously described. 1032

Deconvolution of whole liver microarray data 1033

To assess macrophage composition early-stage NAFLD, we performed deconvolution 1034

analysis on publicly available microarray data from annotated liver biopsy specimens 1035

taken across the NAFLD disease spectrum (GEO accession GSE48452)47. Tissue MP 1036

cells from our human scRNA-seq data were manually clustered into the main annotated 1037

36

MP populations. Signature gene expression profiles of SAM, TMo, KC were used to 1038

deconvolve the monocyte-macrophage composition of liver biopsy samples from 1039

GSE48452 using Cibersort96, as previously described46. The monocyte-macrophage 1040

composition of each biopsy sample was then compared to the associated histological 1041

and demographic features, available from the GEO database. 1042

Statistical Analysis 1043

To assess whether our identified subpopulations were significantly overexpressed in 1044

injury, we posited the proportion of injured cells in each cluster as a random count 1045

variable using a Poisson process, as previously described89. We modelled the rate of 1046

detection using the total number of cells in the lineage profiled in a given sample as an 1047

offset, with the condition of each sample (healthy vs cirrhotic) provided as a covariate 1048

factor. The model was fitted using the R command glm from the stats package. The p-1049

value for the significance of the proportion of injured cells was assessed using a Wald 1050

test on the regression coefficient. This methodology was also applied to assess 1051

significant changes in proportions of mononuclear phagocytes between healthy and 1052

cirrhotic liver tissue by flow cytometry. 1053

Remaining statistical analyses were performed using GraphPad Prism (GraphPad 1054

Software, USA). Comparison of changes between two groups was performed using a 1055

Mann-Whitney test (unpaired; two-tailed) or using a Wilcoxon matched-pairs signed 1056

rank test (paired; two-tailed). Comparison of changes between multiple groups was 1057

performed using a Kruskal-Wallis and Dunn, one-way ANOVA and Tukey or repeated 1058

measures one-way ANOVA and Tukey tests. Correlations were preformed using 1059

Pearson correlation and best fit line plotted using linear regression. P-values<0.05 were 1060

considered statistically significant. 1061

Data and materials availability 1062

Our expression data will be freely available for user-friendly interactive browsing 1063

online at http://www.livercellatlas.mvm.ed.ac.uk (log-in details for purposes of review; 1064

username: Edinburghlivercellatlas; password: Cirrhosis). CellPhoneDB is available at 1065

www.CellPhoneDB.org, along with lists of membrane proteins, ligands and receptors, 1066

37

and heteromeric complexes. All raw sequencing data will be deposited in the Gene 1067

Expression Omnibus (GEO; GEO NO. XXXX). 1068

Code availability 1069

R scripts enabling the main steps of the analysis are available from the corresponding 1070

authors on reasonable request. 1071

38

Acknowledgements 1072

This work was supported by an MRC Clinician Scientist Fellowship (MR/N008340/1) 1073

to P.R., a Wellcome Trust Senior Research Fellowship in Clinical Science (ref. 103749) 1074

to N.C.H., an AbbVie Future Therapeutics and Technologies Division grant to N.C.H., 1075

a CORE – Children's Liver Disease Foundation grant (ref. R43927) to N.C.H. and a 1076

Tenovus Scotland grant (ref. E18/05) to R.D. and N.C.H. R.V-T. was funded by EMBO 1077

and Human Frontiers long-term fellowships. C.J.W. was funded by a BBSRC New 1078

Investigator Award (BB/N018869/1). P.N.N., C.J.W. and N.T.L. are funded by the 1079

NIHR Birmingham Biomedical Research Centre at the University Hospitals 1080

Birmingham NHS Foundation Trust and the University of Birmingham. This paper 1081

presents independent research supported by the NIHR Birmingham Biomedical 1082

Research Centre at the University Hospitals Birmingham NHS Foundation Trust and 1083

the University of Birmingham. The views expressed are those of the author(s) and not 1084

necessarily those of the NHS, the NIHR or the Department of Health and Social Care. 1085

J.P.I. is funded by the NIHR Bristol Biomedical Research Centre, University Hospitals 1086

Bristol Foundation Trust and the University of Bristol. We thank the patients who 1087

donated liver tissue and blood for this study. We thank J. Davidson, C. Ibbotson, J. 1088

Black and A. Baird of the Scottish Liver Transplant Unit and the research nurses of the 1089

Wellcome Trust Clinical Research Facility for assistance with consenting patients for 1090

this study. We thank the liver transplant coordinators and surgeons of the Scottish Liver 1091

Transplant Unit and the surgeons and staff of the Hepatobiliary Surgical Unit, Royal 1092

Infirmary of Edinburgh for assistance in procuring human liver samples. We thank S. 1093

Johnston, W. Ramsay and M. Pattison (QMRI Flow Cytometry and Cell Sorting 1094

Facility, University of Edinburgh) for technical assistance with fluorescence activated 1095

cell sorting (FACS) and flow cytometry. We thank Jon Henderson for technical support 1096

and Gillian Muirhead for assistance with liver endothelial cell isolation. 1097

1098

Author Contributions 1099

P.R. performed experimental design, tissue procurement, data generation, data analysis 1100

and interpretation, and manuscript preparation; R.D. performed experimental design, 1101

data generation and data analysis; E.D., K.P.M., B.E.P.H., M.B., J.A.M. and N.T.L. 1102

39

performed data generation and analysis; J.R.P. generated the interactive online browser; 1103

M.E. and R.V-T. assisted with CellPhoneDB analyses and critically appraised the 1104

manuscript; T.J.K. performed pathological assessments and provided intellectual 1105

contribution; N.O.C., J.A.F. and P.N.N. provided intellectual contribution; C.J.W. 1106

performed tissue procurement, data generation, interpretation and intellectual 1107

contribution; J.R.W-K. performed computational analysis with assistance from J.R.P 1108

and R.S.T. and advice from C.P.P., J.M. and S.A.T.; J.R.W-K. also helped with 1109

manuscript preparation and C.P.P., J.M. and S.A.T. critically appraised the manuscript; 1110

E.M.H., D.J.M. and S.J.W. procured human liver tissue and critically appraised the 1111

manuscript. J.P.I., F.T. and J.W.P. provided intellectual contribution and critically 1112

appraised the manuscript; N.C.H. conceived the study, designed experiments, 1113

interpreted data and prepared the manuscript. 1114

Author Information 1115

The authors declare no competing financial interests. 1116

40

References 1117

1. Marcellin, P. & Kutala, B. K. Liver diseases: A major, neglected global public 1118

health problem requiring urgent actions and large-scale screening. Liver Int. 1119

(2018). doi:10.1111/liv.13682 1120

2. Hernandez-Gea, V. & Friedman, S. L. Pathogenesis of Liver Fibrosis. Annu. 1121

Rev. Pathol. Mech. Dis. 6, 425–456 (2011). 1122

3. Angulo, P. et al. Liver Fibrosis, but No Other Histologic Features, Is 1123

Associated With Long-term Outcomes of Patients With Nonalcoholic Fatty 1124

Liver Disease. Gastroenterology 149, 389-397.e10 (2015). 1125

4. Ramachandran, P. & Henderson, N. C. Antifibrotics in chronic liver disease: 1126

tractable targets and translational challenges. The Lancet Gastroenterology and 1127

Hepatology 1, 328–340 (2016). 1128

5. Friedman, S. L., Neuschwander-Tetri, B. A., Rinella, M. & Sanyal, A. J. 1129

Mechanisms of NAFLD development and therapeutic strategies. Nature 1130

Medicine 24, 908–922 (2018). 1131

6. Trautwein, C., Friedman, S. L., Schuppan, D. & Pinzani, M. Hepatic fibrosis: 1132

Concept to treatment. Journal of Hepatology 62, S15–S24 (2015). 1133

7. Yukinori Koyama and David A. Brenner. Liver inflammation and fibrosis. J. 1134

Clin. Invest. 5, 55–64 (2017). 1135

8. Stubbington, M. J. T., Rozenblatt-Rosen, O., Regev, A. & Teichmann, S. A. 1136

Single-cell transcriptomics to explore the immune system in health and disease. 1137

Science 358, 58–63 (2017). 1138

9. Giladi, A. & Amit, I. Single-Cell Genomics: A Stepping Stone for Future 1139

Immunology Discoveries. Cell 172, 14–21 (2018). 1140

10. Halpern, K. B. et al. Single-cell spatial reconstruction reveals global division of 1141

labour in the mammalian liver. Nature 542, 352–356 (2017). 1142

11. MacParland, S. A. et al. Single cell RNA sequencing of human liver reveals 1143

distinct intrahepatic macrophage populations. Nat. Commun. 9, 4383 (2018). 1144

12. Searle, B. C., Gittelman, R. M., Manor, O. & Akey, J. M. Detecting sources of 1145

transcriptional heterogeneity in large-scale RNA-seq data sets. Genetics 204, 1146

1391–1396 (2016). 1147

13. Wynn, T. A., Chawla, A. & Pollard, J. W. Macrophage biology in 1148

41

development, homeostasis and disease. Nature 496, 445–455 (2013). 1149

14. Duffield, J. S. et al. Selective depletion of macrophages reveals distinct, 1150

opposing roles during liver injury and repair. J. Clin. Invest. 115, 56–65 (2005). 1151

15. Ramachandran, P. et al. Differential Ly-6C expression identifies the recruited 1152

macrophage phenotype, which orchestrates the regression of murine liver 1153

fibrosis. Proc. Natl. Acad. Sci. 109, E3186–E3195 (2012). 1154

16. Zimmermann, H. W. et al. Functional Contribution of Elevated Circulating and 1155

Hepatic Non-Classical CD14+CD16+ Monocytes to Inflammation and Human 1156

Liver Fibrosis. PLoS One 5, e11049 (2010). 1157

17. Beattie, L. et al. Bone marrow-derived and resident liver macrophages display 1158

unique transcriptomic signatures but similar biological functions. J. Hepatol. 1159

65, 758–768 (2016). 1160

18. Minutti, C. M. et al. Local amplifiers of IL-4Rα–mediated macrophage 1161

activation promote repair in lung and liver. Science (80-. ). 356, 1076–1080 1162

(2017). 1163

19. Baek, J.-H. et al. IL-34 mediates acute kidney injury and worsens subsequent 1164

chronic kidney disease. J. Clin. Invest. 125, 3198–3214 (2015). 1165

20. Pradere, J.-P. et al. Hepatic macrophages but not dendritic cells contribute to 1166

liver fibrosis by promoting the survival of activated hepatic stellate cells in 1167

mice. Hepatology 58, 1461–1473 (2013). 1168

21. Petrasek, J. et al. IL-1 receptor antagonist ameliorates inflammasome-1169

dependent alcoholic steatohepatitis in mice. J. Clin. Invest. 122, 3476–3489 1170

(2012). 1171

22. Syn, W. K. et al. Osteopontin is induced by hedgehog pathway activation and 1172

promotes fibrosis progression in nonalcoholic steatohepatitis. Hepatology 53, 1173

106–115 (2011). 1174

23. Henderson, N. C. et al. Galectin-3 regulates myofibroblast activation and 1175

hepatic fibrosis. Proc. Natl. Acad. Sci. 103, 5060–5065 (2006). 1176

24. Seki, E. et al. CCR2 promotes hepatic fibrosis in mice. Hepatology 50, 185–1177

197 (2009). 1178

25. Wilhelm, A. et al. Interaction of TWEAK with Fn14 leads to the progression of 1179

fibrotic liver disease by directly modulating hepatic stellate cell proliferation. J. 1180

42

Pathol. 239, 109–121 (2016). 1181

26. Scott, C. L. & Guilliams, M. The role of Kupffer cells in hepatic iron and lipid 1182

metabolism. J. Hepatol. (2018). doi:10.1016/j.jhep.2018.02.013 1183

27. Guilliams, M. et al. Dendritic cells, monocytes and macrophages: A unified 1184

nomenclature based on ontogeny. Nature Reviews Immunology 14, 571–578 1185

(2014). 1186

28. Scott, C. L. et al. Bone marrow-derived monocytes give rise to self-renewing 1187

and fully differentiated Kupffer cells. Nat. Commun. 7, 10321 (2016). 1188

29. Gomez Perdiguero, E. et al. Tissue-resident macrophages originate from yolk-1189

sac-derived erythro-myeloid progenitors. Nature 518, 547–551 (2015). 1190

30. Hoeffel, G. et al. C-Myb+ Erythro-Myeloid Progenitor-Derived Fetal 1191

Monocytes Give Rise to Adult Tissue-Resident Macrophages. Immunity 42, 1192

665–678 (2015). 1193

31. Yona, S. et al. Fate Mapping Reveals Origins and Dynamics of Monocytes and 1194

Tissue Macrophages under Homeostasis. Immunity 38, 79–91 (2013). 1195

32. Mass, E. et al. Specification of tissue-resident macrophages during 1196

organogenesis. Science 353, aaf4238–aaf4238 (2016). 1197

33. Karlmark, K. R. et al. Hepatic recruitment of the inflammatory Gr1 + 1198

monocyte subset upon liver injury promotes hepatic fibrosis. Hepatology 50, 1199

261–274 (2009). 1200

34. La Manno, G. et al. RNA velocity of single cells. Nature 1 (2018). 1201

doi:10.1038/s41586-018-0414-6 1202

35. Klein, A. M. Technique to measure the expression dynamics of each gene in a 1203

single cell. Nature 560, 434–435 (2018). 1204

36. Friedman, S. L. & Arthur, M. J. P. Activation of cultured rat hepatic lipocytes 1205

by Kupffer cell conditioned medium. Direct enhancement of matrix synthesis 1206

and stimulation of cell proliferation via induction of platelet-derived growth 1207

factor receptors. J. Clin. Invest. 84, 1780–1785 (1989). 1208

37. Yang, L. et al. Vascular Endothelial Growth Factor Promotes Fibrosis 1209

Resolution and Repair in Mice. Gastroenterology 146, 1339-1350.e1 (2014). 1210

38. Zimmermann, H. W. et al. Interleukin-8 is activated in patients with chronic 1211

liver diseases and associated with hepatic macrophage accumulation in human 1212

43

liver fibrosis. PLoS One 6, e21381 (2011). 1213

39. Butler, A., Hoffman, P., Smibert, P., Papalexi, E. & Satija, R. Integrating 1214

single-cell transcriptomic data across different conditions, technologies, and 1215

species. Nat. Biotechnol. 36, 411–420 (2018). 1216

40. Scott, C. L. et al. The Transcription Factor ZEB2 Is Required to Maintain the 1217

Tissue-Specific Identities of Macrophages. Immunity 49, 312-325.e5 (2018). 1218

41. Haldar, M. et al. Heme-mediated SPI-C induction promotes monocyte 1219

differentiation into iron-recycling macrophages. Cell 156, 1223–1234 (2014). 1220

42. Bradley, E. W., Ruan, M. M. & Oursler, M. J. Novel pro-survival functions of 1221

the Kruppel-like transcription factor Egr2 in promotion of macrophage colony-1222

stimulating factor-mediated osteoclast survival downstream of the MEK/ERK 1223

pathway. J. Biol. Chem. 283, 8055–64 (2008). 1224

43. Fang, F. et al. The Early Growth Response Gene Egr2 (Alias Krox20) Is a 1225

Novel Transcriptional Target of Transforming Growth Factor-β that Is Up-1226

Regulated in Systemic Sclerosis and Mediates Profibrotic Responses. Am. J. 1227

Pathol. 178, 2077–2090 (2011). 1228

44. Zhang, Q. et al. Notch signal suppresses toll-like receptor-triggered 1229

inflammatory responses in macrophages by inhibiting extracellular signal-1230

regulated kinase 1/2-mediated nuclear factor κB activation. J. Biol. Chem. 287, 1231

6208–6217 (2012). 1232

45. Ni, M. et al. Novel Insights on Notch signaling pathways in liver fibrosis. Eur. 1233

J. Pharmacol. 826, 66–74 (2018). 1234

46. Schelker, M. et al. Estimation of immune cell content in tumour tissue using 1235

single-cell RNA-seq data. Nat. Commun. 8, 2032 (2017). 1236

47. Ahrens, M. et al. DNA Methylation Analysis in Nonalcoholic Fatty Liver 1237

Disease Suggests Distinct Disease-Specific and Remodeling Signatures after 1238

Bariatric Surgery. Cell Metab. 18, 296–302 (2013). 1239

48. Ding, B. Sen et al. Divergent angiocrine signals from vascular niche balance 1240

liver regeneration and fibrosis. Nature 505, 97–102 (2014). 1241

49. Xie, G. et al. Role of differentiation of liver sinusoidal endothelial cells in 1242

progression and regression of hepatic fibrosis in rats. Gastroenterology 142, 1243

918-927 e6 (2012). 1244

44

50. Weinl, C. et al. Elk3 Deficiency Causes Transient Impairment in Post-Natal 1245

Retinal Vascular Development and Formation of Tortuous Arteries in Adult 1246

Murine Retinae. PLoS One 9, e107048 (2014). 1247

51. Borkham-Kamphorst, E. & Weiskirchen, R. The PDGF system and its 1248

antagonists in liver fibrosis. Cytokine and Growth Factor Reviews 28, 53–61 1249

(2016). 1250

52. Liu, S. B. et al. Lysyl oxidase activity contributes to collagen stabilization 1251

during liver fibrosis progression and limits spontaneous fibrosis reversal in 1252

mice. FASEB J. 30, 1599–1609 (2016). 1253

53. Ikenaga, N. et al. Selective targeting of lysyl oxidase-like 2 (LOXL2) 1254

suppresses hepatic fibrosis progression and accelerates its reversal. Gut 66, 1255

1697–1708 (2017). 1256

54. Thiriot, A. et al. Differential DARC/ACKR1 expression distinguishes venular 1257

from non-venular endothelial cells in murine tissues. BMC Biol. 2017 151 15, 1258

45 (2017). 1259

55. Pruenster, M. et al. The Duffy antigen receptor for chemokines transports 1260

chemokines and supports their promigratory activity. Nat. Immunol. 10, 101–8 1261

(2009). 1262

56. Shetty, S., Weston, C. J., Adams, D. H. & Lalor, P. F. A Flow Adhesion Assay 1263

to Study Leucocyte Recruitment to Human Hepatic Sinusoidal Endothelium 1264

Under Conditions of Shear Stress. J. Vis. Exp. (2014). doi:10.3791/51330 1265

57. Géraud, C. et al. GATA4-dependent organ-specific endothelial differentiation 1266

controls liver development and embryonic hematopoiesis. J. Clin. Invest. 127, 1267

1099–1114 (2017). 1268

58. Knolle, P. A. & Wohlleber, D. Immunological functions of liver sinusoidal 1269

endothelial cells. Cell. Mol. Immunol. 13, 347–353 (2016). 1270

59. Fatima, A. et al. Foxc1 and Foxc2 deletion causes abnormal 1271

lymphangiogenesis and correlates with ERK hyperactivation. J. Clin. Invest. 1272

126, 2437–2451 (2016). 1273

60. Harada, K. et al. Identification of targets of Prox1 during in vitro vascular 1274

differentiation from embryonic stem cells: functional roles of HoxD8 in 1275

lymphangiogenesis. J. Cell Sci. 122, 3923–3930 (2009). 1276

45

61. Rocha, A. S. et al. The Angiocrine Factor Rspondin3 Is a Key Determinant of 1277

Liver Zonation. Cell Rep. 13, 1757–1764 (2015). 1278

62. Corada, M. et al. Sox17 is indispensable for acquisition and maintenance of 1279

arterial identity. Nat. Commun. 4, 2609 (2013). 1280

63. Cuervo, H. et al. Endothelial notch signaling is essential to prevent hepatic 1281

vascular malformations in mice. Hepatology 64, 1302–1316 (2016). 1282

64. Bahrami, A. J. et al. Regulator of G-Protein Signaling-5 Is a Marker of Hepatic 1283

Stellate Cells and Expression Mediates Response to Liver Injury. PLoS One 9, 1284

e108505 (2014). 1285

65. Vallecillo-García, P. et al. Odd skipped-related 1 identifies a population of 1286

embryonic fibro-adipogenic progenitors regulating myogenesis during limb 1287

development. Nat. Commun. 8, 1218 (2017). 1288

66. Henderson, N. C. et al. Targeting of αv integrin identifies a core molecular 1289

pathway that regulates fibrosis in several organs. Nat. Med. 19, 1617–1624 1290

(2013). 1291

67. De Minicis, S. et al. Gene Expression Profiles During Hepatic Stellate Cell 1292

Activation in Culture and In Vivo. Gastroenterology 132, 1937–1946 (2007). 1293

68. Hinz, B. et al. Recent Developments in Myofibroblast Biology. Am. J. Pathol. 1294

180, 1340–1355 (2012). 1295

69. Vento-Tormo, R. et al. Single-cell reconstruction of the early maternal–fetal 1296

interface in humans. Nat. 2018 5637731 563, 347 (2018). 1297

70. Liang, D. et al. Inhibition of EGFR attenuates fibrosis and stellate cell 1298

activation in diet-induced model of nonalcoholic fatty liver disease. Biochim. 1299

Biophys. Acta - Mol. Basis Dis. 1864, 133–142 (2018). 1300

71. Minutti, C. M. et al. A Macrophage-Pericyte Axis Directs Tissue Restoration 1301

via Amphiregulin-Induced Transforming Growth Factor Beta Activation. 1302

Immunity 50, 645-654.e6 (2019). 1303

72. Mouchemore, K. A. & Pixley, F. J. CSF-1 signaling in macrophages: 1304

pleiotrophy through phosphotyrosine-based signaling pathways. Crit. Rev. 1305

Clin. Lab. Sci. 49, 49–61 (2012). 1306

73. Ensan, S. et al. Self-renewing resident arterial macrophages arise from 1307

embryonic CX3CR1 + precursors and circulating monocytes immediately after 1308

46

birth. Nat. Immunol. 17, 159–168 (2016). 1309

74. Thabut, D. & Shah, V. Intrahepatic angiogenesis and sinusoidal remodeling in 1310

chronic liver disease: New targets for the treatment of portal hypertension? J. 1311

Hepatol. 53, 976–980 (2010). 1312

75. Krishnasamy, K. et al. Blood vessel control of macrophage maturation 1313

promotes arteriogenesis in ischemia. Nat. Commun. 8, 952 (2017). 1314

76. Brenner, D. A. Of mice and men and NASH. Hepatology (2018). 1315

doi:10.1002/hep.30186 1316

77. Broutier, L. et al. Human primary liver cancer–derived organoid cultures for 1317

disease modeling and drug screening. Nat. Med. 23, 1424–1435 (2017). 1318

1319

References (Methods) 1320

78. Bain, C. C. et al. Long-lived self-renewing bone marrow-derived macrophages 1321

displace embryo-derived cells to inhabit adult serous cavities. Nat. Commun. 7, 1322

ncomms11852 (2016). 1323

79. Heinrich, M. C. et al. Crenolanib Inhibits the Drug-Resistant PDGFRA D842V 1324

Mutation Associated with Imatinib-Resistant Gastrointestinal Stromal Tumors. 1325

Clin. Cancer Res. 18, 4375–4384 (2012). 1326

80. Mederacke, I., Dapito, D. H., Affò, S., Uchinami, H. & Schwabe, R. F. High-1327

yield and high-purity isolation of hepatic stellate cells from normal and fibrotic 1328

mouse livers. Nat. Protoc. 10, 305–315 (2015). 1329

81. Patten, D. A. et al. SCARF-1 promotes adhesion of CD4+ T cells to human 1330

hepatic sinusoidal endothelium under conditions of shear stress. Sci. Rep. 7, 1331

17600 (2017). 1332

82. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. 1333

Nat. Methods 9, 676–682 (2012). 1334

83. Bankhead, P. et al. QuPath: Open source software for digital pathology image 1335

analysis. Sci. Rep. 7, 16878 (2017). 1336

84. Arganda-Carreras, I. et al. Trainable Weka Segmentation: A machine learning 1337

tool for microscopy pixel classification. Bioinformatics 33, 2424–2426 (2017). 1338

85. Kleiner, D. E. et al. Design and validation of a histological scoring system for 1339

nonalcoholic fatty liver disease. Hepatology 41, 1313–1321 (2005). 1340

47

86. Deroulers, C. et al. Analyzing huge pathology images with open source 1341

software. Diagn. Pathol. 8, 92 (2013). 1342

87. Kendall, T. J. et al. Hepatic elastin content is predictive of adverse outcome in 1343

advanced fibrotic liver disease. Histopathology 73, 90–100 (2018). 1344

88. Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. & Regev, A. Spatial 1345

reconstruction of single-cell gene expression data. Nat. Biotechnol. 33, 495–1346

502 (2015). 1347

89. Haber, A. L. et al. A single-cell survey of the small intestinal epithelium. 1348

Nature (2017). doi:10.1038/nature24489 1349

90. Li, W. V. & Li, J. J. An accurate and robust imputation method scImpute for 1350

single-cell RNA-seq data. Nat. Commun. 9, 997 (2018). 1351

91. Camp, J. G. et al. Multilineage communication regulates human liver bud 1352

development from pluripotency. Nature 546, 533–538 (2017). 1353

92. Trapnell, C. et al. The dynamics and regulators of cell fate decisions are 1354

revealed by pseudotemporal ordering of single cells. Nat. Biotechnol. 32, 381–1355

386 (2014). 1356

93. McCarthy, D. J., Campbell, K. R., Lun, A. T. L. & Wills, Q. F. Scater: Pre-1357

processing, quality control, normalization and visualization of single-cell RNA-1358

seq data in R. Bioinformatics 33, 1179–1186 (2017). 1359

94. Zhang, H. M. et al. AnimalTFDB 2.0: A resource for expression, prediction 1360

and functional study of animal transcription factors. Nucleic Acids Res. 43, 1361

D76–D81 (2015). 1362

95. Aibar, S. et al. SCENIC: Single-cell regulatory network inference and 1363

clustering. Nat. Methods 14, 1083–1086 (2017). 1364

96. Newman, A. M. et al. Robust enumeration of cell subsets from tissue 1365

expression profiles. Nat. Methods (2015). doi:10.1038/nmeth.3337 1366

1367

1368

48

Figure 1: Single cell atlas of human liver non-parenchymal cells. 1369

a, Overview: extraction of non-parenchymal cells (NPC) from healthy or cirrhotic 1370

human liver, followed by fluorescence-activated cell sorting (FACS) into leucocyte 1371

(CD45+) and other NPC fractions (CD45-) for droplet-based 3' single-cell RNA-seq. b, 1372

t-SNE visualisation: clustering (colour) 66,135 non-parenchymal cells (points; n=5 1373

healthy and n=5 cirrhotic human livers). c, t-SNE visualisation: injury condition 1374

(colour; healthy versus cirrhotic). d, t-SNE visualisation: cell lineage (colour) inferred 1375

from expression of known marker gene signatures. Endo, endothelial cell; ILC, innate 1376

lymphoid cell; Mast, mast cell; Mes, mesenchymal cell; MP, mononuclear phagocyte; 1377

pDC, plasmacytoid dendritic cell. e, Scaled heatmap (red, high; blue, low): cluster 1378

marker genes (top, colour coded and numbered by cluster and colour coded by 1379

condition) and exemplar genes and lineage annotation labelled (right). Cells columns, 1380

genes rows. 1381

FACS

Leucocytes (CD45+)

Other NPCs (CD45-)

Single-cell RNA-seq

Healthy

Cirrhotic

aFigure 1

210-1-2

Relative expression

TF

FCN1CSTAMNDAC1QCMS4A7CD163HMOX1CSF2RA

LILRA4PTCRASCT

KLRF1PRF1MKI67UBE2CBIRC5NUSAP1TRACLTBCD79ATNFRSF17FCRL5TPSAB1

RAMP2CLEC14AEGFL7HSPG2

FCN3CRHBPCLEC4GMYL9TAGLNACTA2TIMP1COL3A1COL1A2

ALB

APOC2

KNG1TTRORM1

KRT19CXCL6

IL1BVCANCLEC7A

CYBBMSR1CPVL

FOLR2CLEC10A

LRRC26MAP1ACLEC4C

KLRC1FGFBP2TOP2ARRM2AURKBASPMCD2SIT1MS4A1

POU2AF1JSRP1

CPA3

ECSCRPLVAPVWFEMCN

OIT3LIFRCLEC4MRGS5

CSRP2PDGFRB

LUMPDGFRACCL19

F2PRAP1HSD17B6ITIH1CYP4A11AGXT

CXCL1SOX9

Cell lineage

MP

pDC

ILC

Cycling

T cellB cellPlasma cellMast cell

Endothelia

Mesenchyme

Epithelia

ConditionCluster1 2 3 4 5 6 7 8 9 10 11 12 13 14 16 17 18 19 21

e2015

b Cluster1

11

9

10

2

8

7

6

5

4

3

21

20

19

18

17

16

15

14

13

12

21

20

19

18

17

16

15

1413

1211

10

9

8

7

6

5

4

32

1

Cell lineage

B cell

Endothelia

Epithelia

ILC

Mast cell

Mesenchyme

MP

pDC

Plasma cell

T cell

Cycling

dPlasma cell

B cell

T cellEpithelia

MP ILC

Cycling

MastMes

Endo

pDC

c Healthy

Cirrhotic

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

49

Figure 2: Identifying scar-associated macrophage subpopulations. 1382

a, t-SNE visualisation: clustering 10,737 mononuclear phagocytes (MP) into 10 clusters 1383

(colour and number). Annotation of clusters (identity). TMo, tissue monocyte; SAM, 1384

scar-associated macrophage; KC, Kupffer cell; cDC, conventional dendritic cell. b, t-1385

SNE visualisation: annotating MP cells by injury condition (colour). c, Fractions of MP 1386

subpopulations in healthy (n=5) versus cirrhotic (n=5) livers, Wald test. d, Scaled 1387

heatmap (red, high; blue, low): MP cluster marker genes (top, colour coded by cluster 1388

and condition), exemplar genes labelled (right). Cells columns, genes rows. e, Violin 1389

plots: scar-associated macrophage and tissue monocyte cluster markers. f, 1390

Representative flow cytometry plots: quantifying TREM2+CD9+ MP fraction by flow 1391

cytometry in healthy (n=2) versus cirrhotic (n=3) liver, Wald. g, Representative 1392

immunofluorescence micrograph, cirrhotic liver: TREM2 (red), CD9 (white), collagen 1393

1 (green), DAPI (blue). Scale bar, 50μm. h, Automated cell counting: TREM2 staining, 1394

healthy (n=10) versus cirrhotic (n=9) liver, Mann-Whitney. i, Automated cell counting: 1395

CD9 staining, healthy (n=12) versus cirrhotic (n=10) liver, Mann-Whitney. j, 1396

Topographically assessing scar-associated macrophages: exemplar tissue segmentation 1397

(left), stained section morphologically segmented into fibrotic septae (orange) and 1398

parenchymal nodules (purple)). TREM2+, CD9+, TIMD4+ and MARCO+ automated 1399

cell counts (right) in parenchymal nodules versus fibrotic septae, Wilcoxon. Error bars, 1400

s.e.m.; * p-value<0.05; ** p-value < 0.01; *** p-value < 0.001. 1401

d e

f

0 103 104-103

TREM2-APC

CD

9-P

E-V

io77

0

0

103

104

105

-103

0 103 104-103

TREM2-APC

Healthy Cirrhotic g

c

210-1-2

S100A12

NBEAL1

TREM2C1QC

CD5LCD163VCAM1MARCO

MNDA

IFITM2APOBEC3ARGS2

CD9

TIMD4MERTKNR1H3

ARL4CFCER1ACD1ELGALS2

ConditionCluster

SIGLEC1

2mm2mm

TRE

M2+ c

ells

/mm

2

400

0

600

200

Fibroticseptae

Nodules

CD

9+ cel

ls/m

m2

200

0

300

100

Fibroticseptae

Nodules

TIM

D4+ c

ells

/mm

2

200

0

300

100

Fibroticseptae

Nodules

MA

RC

O+ c

ells

/mm

2

100

0

150

50

Fibroticseptae

Nodules

CD

9+ cel

ls/m

m2 l

iver

300

0

100

200

Healthy Cirrhotic

Healthy Cirrhotic

CD

9

Cirrhotic

Healthy Cirrhotic

TRE

M2+ c

ells

/mm

2 liv

er

200

0

50

100

150

Healthy

TREM

2

Collagen 1CD9TREM2 DAPI

h i

j

Healthy Cirrhotic

% o

f MP

s

20

0

10

25

5

15

***

bFigure 2

** **

** ***** *

Cluster

TREM2

CD9

MNDAGen

e ex

pres

sion

TREM2

CD9

S100A12

MNDA

C1QC

TMo(1)

TMo(2)

SAM(1)

SAM(2)

KC(1)

KC(2)cD

C2cD

C1

TMo(3)

5.92±1.5% 19.7±3.2%

FCN1

TMo(1)

TMo(2)

SAM(1)

SAM(2)

KC(1)

KC(2)cD

C2cD

C1

TMo(3)

Cluster

Frac

tion

of M

P ce

lls

0.4

0.3

0.2

0

0.1

****** ***HealthyCirrhotic

Healthy

Cirrhotic

ClusterMP(1)MP(2)

MP(4)MP(5)

Cycling

MP(7)

MP(3)

MP(6)

MP(8)MP(9)

12345678910

2

10

8

76

5

1

3

4

IdentityTMo(1)TMo(2)

SAM(1)SAM(2)

KC(2)

TMo(3)

KC(1)

cDC2cDC1

9

a

t-SNE1

t-SN

E2

50

Figure 3: Fibrogenic phenotype of scar-associated macrophages. 1402

a, Self-Organising Map (SOM; 60x60 grid): smoothed scaled metagene expression of 1403

mononuclear phagocyte (MP) lineage. 20,952 genes, 3,600 metagenes, 44 signatures. 1404

A-F label metagene signatures overexpressed in one or more MP subpopulations. b, 1405

Radar plots (left): metagene signatures A-C showing distribution of signature 1406

expression across MP subpopulations, exemplar genes (middle) and Gene Ontology 1407

(GO) enrichment (right). c, Diffusion map visualisation, blood monocytes and liver-1408

resident MP lineages (23,075 cells), annotating monocle pseudotemporal dynamics 1409

(purple to yellow). RNA velocity field (red arrows) visualised using Gaussian 1410

smoothing on regular grid. Below: Annotation of MP subpopulation, injury condition. 1411

d, UMAP visualisation, blood monocytes and liver-resident MP lineages, annotating 1412

monocle pseudotemporal dynamics (purple to yellow). RNA velocity field (red arrows) 1413

visualised using Gaussian smoothing on regular grid. Below: Annotation of MP 1414

subpopulation, injury condition. e, Unspliced-spliced phase portraits (top row), cells 1415

coloured as in d, for monocyte (MNDA), SAM (CD9) and KC marker genes (TIMD4). 1416

Cells plotted above or below the steady-state (black dashed line) indicate increasing or 1417

decreasing expression of gene, respectively. Spliced expression profile for genes 1418

(middle row; red high, blue low). Unspliced residuals (bottom row), positive (red) 1419

indicating expected upregulation, negative (blue) indicating expected downregulation 1420

for genes. MNDA displays negative velocity in SAM, CD9 displays positive velocity in 1421

monocytes and SAM, TIMD4 velocity is restricted to KC. f, UMAP visualisation, 1422

transition probabilities per SAM subpopulation, indicating for each cell the likelihood 1423

of transition into either SAM(1) or SAM(2), calculated using RNA velocity (yellow 1424

high; purple low; grey below threshold of 2x10-4). g, Scaled heatmap (red, high; blue 1425

low): cubic smoothing spline curves fitted to genes differentially expressed across 1426

blood monocyte-to-SAM (right arrow) and blood monocyte-to-cDC (left arrow) 1427

pseudotemporal trajectories, grouped by hierarchical clustering (k=3). Gene co-1428

expression modules (colour) labelled right. h, Cubic smoothing spline curve fitted to 1429

averaged expression of all genes in module 1, along monocyte-SAM pseudotemporal 1430

trajectory, selected GO enrichment (right) and curves fit to exemplar genes (below). 1431

a

A

CB

bc

D E

F

Figure 3

d

Genes Gene Ontologyb

-log10(P-value)

TMo(1)cDC1TMo(2)

KC(2)SAM(1)

cDC2TMo(3)

KC(1)SAM(2)

Sign

atur

e A

A

cDC1

cDC2

Sign

atur

e B

KC(2)SAM(1)

KC(1)SAM(2)

B

cDC1

KC(2)SAM(1)

cDC2

KC(1)SAM(2)Si

gnat

ure

C

TREM2FN1SPP1LTC4SCCR2CLEC5A

MMP14

CST6HTRA1SPARCL1FMNL3

TNFSF12

IL1BMNDALGALS3AREGPLAURCXCR4

IL13RA1

FCGR2AFCGR2BCD86CSTA

IFNGR2

MARCOVCAM1CD5LTIMD4FCGR3ACD163MERTK

CD302NR1H3IL18SCARB2MSR1FOLR2MRC1

Positive Regulation of Inflammatory ResponseAngiogenesis

Extracellular Structure Organization

Myeloid Leukocyte Activation

C

Proteoglycan Biosynthetic ProcessExtracellular Matrix Organization

0 2 4

Antigen Processing and Presentation

Regulation of T Cell ActivationResponse to Interferon-gamma

0 5 10

Vesicle-mediated TransportResponse to BacteriumResponse to LipidReceptor-Mediated Endocytosis

0 10 20

Phagocytosis

Iron Ion Homeostasis

TransitionProbability

Blood monocytes

SAM

Tissuemonocytes

KC

cDC

Condition

210-1-2

Module 1

Module 2

g

TMo(1)TMo(2)

TMo(3)

TMo(1)TMo(2)

TMo(3)

Module 3

Blood monocytes SAMcDC

0.0

0.5

1.0Pseudotime

Blood monocytes

SAM

Tissuemonocytes

KC

cDC

Condition

2.0x10-4

7.0x10-4

4.5x10-3

SAM(1) SAM(2)

e

Spliced

Uns

plic

ed

TIMD4

1.5

0.60

ResidualPos.

Neg.Neut.

Exp.High

Low

CD9

0.4

1.50

MNDA

0.5

2.50

0.0

0.5

1.0Pseudotime

f

Gene Ontologyh

-log10(P-value)0 5 10 15

Regulation of Mononuclear Cell Proliferation

Regulation of Fibroblast Proliferation

Response to WoundingPhagocytosis

Extracellular Matrix Organization

Regulation of Angiogenesis

Wound Healing

Response to Chemokine

Regulation of Myeloid Cell DifferentiationAntigen Processing and Presentation

1.0

Module 1

0 0.5Pseudotime

Sig

natu

re

expr

essi

on

-2.5

0

2.5

Scal

ed g

ene

expr

essi

on

Pseudotime

-2.5

0

2.5

0 0.5 1.0

TREM2

-2.5

0

2.5

0 0.5 1.0

PDGFB

0 0.5 1.0

CD9

0 0.5 1.0

LGALS3

0 0.5 1.0

HES1

0 0.5 1.0

CXCR4

0 0.5 1.0

VEGFA

0 0.5 1.0

SPP1 CCL2

0 0.5 1.0

CXCL8

0 0.5 1.0

DM1

DM

2

UMAP1

UM

AP

2

UMAP1

UM

AP

2

UMAP1

UM

AP

2

51

Figure 4: Identifying scar-associated endothelial subpopulations. 1432

a, t-SNE visualisation: clustering 8,020 endothelial cells, annotating injury condition. 1433

b, Fractions of endothelial subpopulations in healthy (n=4) versus cirrhotic (n=3) livers, 1434

Wald. c, Scaled heatmap (red, high; blue, low): endothelial cluster marker genes (colour 1435

coded top by cluster and condition), exemplar genes labelled right. Cells columns, 1436

genes rows. d, Representative immunofluorescence micrograph, healthy versus 1437

cirrhotic liver: CD34 (red), CLEC4M (white), PLVAP (green), DAPI (blue). e, 1438

Representative immunofluorescence micrographs, healthy versus cirrhotic liver: 1439

RSPO3, PDPN, AIF1L, VWA1 or ACKR1 (red), CD34 (white), PLVAP (green), DAPI 1440

(blue). f, Digital morphometric pixel quantification: CLEC4M staining healthy (n=5) 1441

versus cirrhotic (n=8), PLVAP staining healthy (n=11) versus cirrhotic (n=11), ACKR1 1442

staining healthy (n=10) versus cirrhotic (n=10), Mann-Whitney. g, Scaled heatmap 1443

(red, high; blue, low): endothelial cluster marker transcription factor regulons (colour 1444

coded top by cluster and condition), exemplar regulons labelled right. Cells in columns, 1445

regulons in rows. Scale bars, 50μm. Error bars, s.e.m.; * p-value < 0.05, ** p-value < 1446

0.01, *** p-value < 0.001. 1447

baHealthyCirrhotic

0.6

0.4

0.2

0Frac

tion

of e

ndot

helia

l cel

ls

Endo(1

)

Endo(2

)

Endo(4

)

Endo(5

)

Endo(6

)

Endo(7

)

Endo(3

)

Cluster

*** ******

Figure 4

ClusterEndo(1)

Endo(5)

Endo(3)Endo(2)

Endo(4)

Endo(6)Endo(7)

1234567

43

7

6

5

2

1

Healthy

Cirrhotic

2

1

0

-1

-2

dConditionCluster

ACKR1VWA1PLVAP

SOX17

CLEC4GCLEC4MSTAB2

CD14

CCL21PROX1PDPN

CPECD320BMPERRSPO3WNT2LHX6HTR2B

AIF1L

Hea

lthy

Cirr

hotic

PLVAPCLEC4MCD34 DAPI

e

Healthy Cirrhotic

CLE

C4M

CLE

C4M

% a

rea *

Healthy Cirrhotic

20

0

5

10

15

25

PLVA

P

PLV

AP

% a

rea

0

5

10

15

20

Healthy Cirrhotic

***

AC

KR

1

AC

KR

1 %

are

a

0

0.05

0.10

0.15

0.20

Healthy Cirrhotic

**

2

1

0

-1

-2

ConditionCluster

IRF1

HOXA9

GATA4NR5A2

STAT2

RARB

MEF2C

ATF3

HOXD8HOXD9FOXC2

GATA6

PPARASOX17KLF2

FOXP1ELK3JDP2

f g

Cirr

hotic

Hea

lthy

e

MRC1

LYVE1

c

PLVAPCD34PDPN DAPI PLVAPCD34AIF1L DAPIPLVAPCD34RSPO3 DAPI PLVAPCD34ACKR1 DAPIPLVAPCD34VWA1 DAPI

t-SNE1

t-SN

E2

52

Figure 5: Identifying scar-associated mesenchymal cell populations. 1448

a, t-SNE visualisation: clustering 2,318 mesenchymal cells (Mes) into 4 clusters (colour 1449

and number). Annotation of clusters (identity). VSMC, vascular smooth muscle cell; 1450

HSC, hepatic stellate cell; SAMes, scar-associated mesenchymal cell. Annotation by 1451

injury condition (colour). b, Scaled heatmap (red, high; blue, low): Mes cluster marker 1452

genes (top, colour coded by cluster and condition), exemplar genes labelled (right). 1453

Cells columns, genes rows. c, Representative immunofluorescence micrograph, healthy 1454

versus cirrhotic liver: RGS5 (red), MYH11 (white), PDGFRα (green), DAPI (blue). d, 1455

Violin plots: fibrillar collagen expression in mesenchymal cell clusters. e, Fractions of 1456

mesenchymal subpopulations in healthy (n=4) versus cirrhotic (n=3) livers, Wald. f, 1457

Digital morphometric pixel quantification: PDGFRα staining, healthy (n=11) versus 1458

cirrhotic (n=11) liver (top), Mann-Whitney. Topographically assessing PDGFRα 1459

staining, stained cirrhotic sections (n=10) morphologically segmented into fibrotic 1460

septae and parenchymal nodules (bottom), Wilcoxon. Scale bars, 50μm. Error bars, 1461

s.e.m.; * p-value<0.05; ** p-value < 0.01; *** p-value < 0.001. 1462

ba

e

HealthyCirrhotic

***

VSMC

Mesenchymal cluster

Frac

tion

of

mes

ench

ymal

cel

ls

0.4

0.2

0

0.6

0.8

HSC

Hea

lthy

Cirr

hotic

PDGFRαf

Figure 5

PD

GFR

α %

are

a

0

5

10

15

20

Healthy Cirrhotic

***

1

2

3 4

Cirrhotic

Healthy

ClusterMes(1)

Mes(3)

Mes(2)

Mes(4)

1

2

3

4

ConditionCluster

210-1-2

MYH11

COL3A1COL1A1

KRT19PDPNMSLN

RGS5

TIMP1PDGFRACCL2

WT1

SAMes

Mesoth

elia

30

0

50

10

Fibroticseptae

Nodules

**P

DG

FRα

% a

rea

20

40

cPDGFRαMYH11RGS5 DAPI

HealthyPDGFRαMYH11RGS5 DAPI

Cirrhotic

Condition

IdentityVSMC

SAMes

HSC

Mesothelia

VSMCMesenchymal cluster

HSC SAMes Mesothelia

COL3A1

COL1A1

Gen

e ex

pres

sion

dt-SNE1

t-SN

E2

53

Figure 6: Multi-lineage interactions in the fibrotic niche. 1463

a to b, Representative immunofluorescence micrographs, fibrotic niche. a, TREM2 1464

(red), PLVAP (white), PDGFRα (green), DAPI (blue). b, TREM2 (red), ACKR1 1465

(white), PDGFRα (green), DAPI (blue). c, Dotplot: ligand-receptor interactions 1466

between scar-associated macrophages (SAM) and scar-associated mesenchyme 1467

(SAMes). X-axis, ligand (red) and cognate receptor (blue); y-axis, cell populations 1468

expressing ligand (red) and receptor (blue); circle size, p-value; colour (red, high; 1469

yellow, low), means of average ligand and receptor expression levels in interacting 1470

subpopulations. d to e, Representative immunofluorescence micrographs, fibrotic 1471

niche. d, TREM2 (red), PDGFB (white), PDGFRα (green), DAPI (blue), arrows 1472

TREM2+PDGFB+ cells. e, TNFRSF12A (red), TNFSF12 (white), PDGFRα (green), 1473

DAPI (blue), arrows TNFRSF12A+PDGFRα+ cells. f to h, Hepatic stellate cell (HSC) 1474

proliferation assay: y-axis, area under curve (AUC) of % change in HSC number over 1475

time (hours), one-way ANOVA and Tukey. f, Vehicle, PDGF-BB (10ng/ml), 1476

Crenolanib (1μM), all n=3. g, Control, TNFSF12 (100ng/ml), anti-TNFRSF12A 1477

(2μg/ml), isotype control (2μg/ml), all n=3. h, Conditioned media from hepatic 1478

macrophages; SAM, tissue monocytes (TMo), Kupffer cells (KC), control, all n=2. i, 1479

Dotplot: ligand-receptor interactions between scar-associated endothelial cells 1480

(SAEndo) and SAMes. X-axis, ligand (red) and cognate receptor (blue); y-axis, cell 1481

populations expressing ligand (red) and receptor (blue); circle size, p-value; colour (red, 1482

high; yellow, low), means of average ligand and receptor expression levels in 1483

interacting subpopulations. j, Endothelial cell JAG1 flow cytometry: healthy (n=3) or 1484

cirrhotic (n=9) liver, representative histogram (top), mean fluorescence intensity (MFI, 1485

bottom), Mann-Whitney. k, Representative immunofluorescence micrographs, fibrotic 1486

niche. NOTCH3 (red), DLL4 (white), PDGFRα (green), DAPI (blue), arrows 1487

NOTCH3+PDGFRα+ cells. l to m, Cirrhotic endothelial cell and HSC co-culture, Notch 1488

inhibitor Dibenzazepine (DBZ; 10μM). l, Representative immunofluorescence 1489

micrographs, Collagen 1 (magenta), PECAM1 (green), DAPI (blue). m, Digital pixel 1490

analysis; collagen 1 area, n=3 per condition, RM one-way ANOVA and Tukey. n, HSC 1491

gene knockdown: control (n=7) or NOTCH3 (n=7) siRNA, qPCR of stated gene, 1492

expression relative to mean expression of control siRNA, Mann-Whitney. Scale bars, 1493

50μm. Error bars, s.e.m.; * p-value<0.05; ** p-value < 0.01; *** p-value < 0.001. 1494

PDGFRαPLVAPTREM2 DAPI PDGFRαACKR1TREM2 DAPIa b

c

d

Figure 6

PDGFRαDLL4NOTCH3 DAPI

SAM→SAMesinteractions

AU

C (%

cha

nge

in H

SCnu

mbe

r ove

r tim

e)

0

3,000

1,000

2,000

*******

TNFSF12 - + + +

Isotype - - - +Anti-TNFRSF12A - - + -

AU

C (%

cha

nge

in H

SCnu

mbe

r ove

r tim

e)

0

8,000

4,000

6,000

2,000

Vehicle + + -

Crenolanib - - +PDGF-BB - + +

*******

0 1 2 3

1.00.50.0

-log10(P-value)

Mean

TNFS

F12-

TNFR

SF1

2AIL

-1R

N-IL

1 R

ecep

tor

IL-1

B-IL

1 R

ecep

tor

PD

GFD

-PD

GFR

BP

DG

FC-P

DG

FRA

PD

GFB

-PD

GFR

BP

DG

FB-P

DG

FRA

PD

GFB

-LR

P1

TGFB

1-E

GFR

ER

EG

-EG

FRH

BE

GF-

EG

FRA

RE

G-E

GFR

SAM(1)-SAMes

SAM(2)-SAMes

PDGFRαPDGFBTREM2 DAPI

PDGFRαTNFSF12TNFRSF12A DAPI

0

1,500

500

1,000

AU

C (%

cha

nge

in H

SCnu

mbe

r ove

r tim

e)

Contro

lSAM

TMo

Conditioned mediaKC

*****

**

JAG1-PE

0

40

80

Cou

nt

0 103 104 105 106

MFI

*

Health

y

Cirrhoti

c

5,000

0

10,000

15,000

20,000

JAG1SAEndo→SAMesinteractions

PDGF NotchIL-1

SAEnd(1)-SAMes

SAEnd(2)-SAMes

HAEnd-SAMes

0 1 2 3

1.00.50.0

-log10(P-value)

Mean

TGFB

1-E

GFR

PD

GFB

-LR

P1

PD

GFB

-PD

GFR

AP

DG

FB-P

DG

FRB

PD

GFC

-PD

GFR

AP

DG

FD-P

DG

FRB

IL-1

B-IL

1 R

ecep

tor

IL-1

RN

-IL1

Rec

epto

rTN

FSF1

2-TN

FRS

F12A

GA

S6-

AX

LD

LL4-

NO

TCH

3JA

G2-

NO

TCH

3JA

G1-

NO

TCH

3D

LL4-

NO

TCH

2JA

G2-

NO

TCH

2JA

G1-

NO

TCH

2D

LL4-

NO

TCH

1JA

G2-

NO

TCH

1JA

G1-

NO

TCH

1

EGFR PDGF IL-1

0

400

200

300

100Col

lage

n 1

area

cov

erag

e

HSC HSC+Endo

HSC+Endo+DBZ

**e

f

g

h

i j

k l

m n

COL1 PECAM1 DAPI COL1 PECAM1 DAPI

HSC + Endo HSC + Endo+ DBZ

ControlNotch3

siRNA**

Rel

ativ

e H

SC

gene

exp

ress

ion

0

1.0

1.5

0.5

GeneNOTCH3 COL1A1 COL3A1

*

54

Extended Data Figure 1: Strategy for isolation of human liver non-parenchymal 1495

cells. 1496

a, Patient demographics and clinical information. b, Representative flow cytometry 1497

plots: gating strategy for isolating leucocytes (CD45+) and other non-parenchymal cells 1498

(CD45-) from healthy and cirrhotic liver. c, Representative flow cytometry plots: gating 1499

strategy for isolating peripheral blood mononuclear cells (PBMC). d, t-SNE 1500

visualisation: clustering 103,568 cells (n=5 healthy human livers, n=5 cirrhotic human 1501

livers, n=1 healthy PBMC, n=4 cirrhotic PBMC), annotating source (PBMC versus 1502

liver) and cell lineage inferred from known marker gene signatures. Endo, endothelial 1503

cell; ILC, innate lymphoid cell; Mast, mast cell; Mes, mesenchymal cell; MP, 1504

mononuclear phagocyte; pDC, plasmacytoid dendritic cell. e, Dotplot: annotating 1505

PBMC and liver dataset clusters by lineage signatures. Circle size indicates cell fraction 1506

expressing signature greater than mean; colour indicates mean signature expression 1507

(red, high; blue, low). f, Violin plots: number of unique genes (nGene) and number of 1508

total Unique Molecular Identifiers (nUMI) expressed in PBMC. g, Pie charts: 1509

proportion of cell lineage per PBMC sample. h, Box and whisker plot: agreement in 1510

expression profiles across PBMC samples. Pearson correlation coefficients between 1511

average expression profiles for cell in each lineage, across all pairs of samples. Black 1512

bar, median value; box edges, 25th and 75th percentiles; whiskers, full range. 1513

e

0.000.250.500.75

Fraction of cells

0.000.250.500.751.00

Cluster1 2 3 4 5 6 87 9 10 11 12 13 14 15 16 1817 19 20 21

MP

pDC

ILC

T cell

B cellPlasma cell

Mast cell

Endothelia

Mesenchyme

Hepatocyte

Cholangiocyte

Line

age

sign

atur

e

Signature score

d

Healthy liver (n=5)

Cirrhotic liver (n=5)

Blood (n=4)

Age (yrs) 57.4±7.9 56.6±5.8 63.2±3.8

Gender (M:F) 4:1 3:2 3:1

Aetiology of liver disease

N/A 2xNAFLD2xALD1xPBC

3xNAFLD1xHH

Haemoglobin (g/l) 145±14 106±17 131±2.1

White Cell Count (x109/l) 8.2±2.2 5.9±1.7 3.7±1.5

Platelets (x109/l) 300±91 137±56 73±38

Prothrombin Time (s) 11.6±1.1 19.6±3.8 16.0±3.6

Creatinine (µmol/l) 76.4±14.5 96.8±28.6 74.5±9.7

Na+ (mmol/l) 141±2.6 131±7.0 139±2.1

Bilirubin (µmol/l) 10±5.2 79.6±83.5 36.3±20.0

ALT (IU/l) 27.8±19.3 77.8±80.7 96.2±121.0

ALP (IU/l) 122±47 140±80 203±153

MELD Score 6.6±0.5 17.3±4.5 11.7±4.3

NAFLD:Non-alcoholic fatty liver disease; ALD:Alcohol-related liver dsease; PBC:Primary biliary cholangitis; HH:Hereditary haemochromatosis; ALT:Alanine transaminase; ALP:Alkaline Phosphatase; MELD:Model for End-Stage Liver Disease

a

c Blood

PBMC

CD66b-FITC

SSC

-A

0 1030

20K

40K

60K

0

20K

40K

60K

CD45-APC0 103 104

SS

C-A

b Liver

Leucocytes

100

101

102

103

104

SS

C-A

DAPI100 101 102 103 104

CD45-APC

SS

C-A

100

101

102

103

104

0 103

Other non-parenchymal

cells

1234567891011

12131415161718192021

ClusterBloodLiver

Source Cell lineage

pDCMP

ILCT cellB cellPlasma cellMast cellEndotheliaMesenchymeEpithelia

g hMP

pDC

ILC

T cell

B cell

Plasma cell

Pearson correlation between samples1.000.950.900.800.65 0.70 0.75 0.85

f

Blood

0

10,000

20,000

30,000

40,000

nUM

I

Blood

nGen

e

1,000

2,000

3,000

4,000

5,000

0

Blood 2

PBMC 8K

Blood 1

Blood 4

Blood 3

Cell lineage

pDC

MPILCT cellB cellPlasma cell

Extended Data Figure 1

Cel

l lin

eage

21

20

19

18

17

16

151413

11

10

9

8

7

6

5

4

31

12

2

Plasma cellB cell

T cell

Epithelia

MPILCMast

Mes

EndopDC

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

55

Extended Data Figure 2: Quality control and annotation of human liver-resident 1514

cells. 1515

a, t-SNE visualisation: lineage signature expression across liver-resident cell dataset 1516

(red, high; blue, low). b, Dotplot: annotating liver-resident cell clusters by lineage 1517

signature. Circle size indicates cell fraction expressing signature greater than mean; 1518

colour indicates mean signature expression (red, high; blue, low). c, Violin plot: number 1519

of unique genes (nGene) expressed across liver-resident cell lineages in healthy versus 1520

cirrhotic livers. d, Violin plot: number of total Unique Molecular Identifiers (nUMI) 1521

expressed across liver-resident cell lineages in healthy versus cirrhotic livers. e, Pie 1522

charts: proportion of cell lineage per liver sample. f, Box and whisker plot: agreement 1523

in expression profiles across liver samples. Pearson correlation coefficients between 1524

average expression profiles for cell in each lineage, across all pairs of samples. Black 1525

bar, median value; box edges, 25th and 75th percentiles; whiskers, range. g, t-SNE 1526

visualisation: liver-resident dataset per liver sample; Cirrhotic samples annotated by 1527

aetiology of underlying liver disease; NAFLD, Non-alcoholic fatty liver disease; ALD, 1528

Alcohol-related liver disease; PBC, Primary biliary cholangitis. 1529

a

g

1 2 3 4 5 6 87 9 10 11 12 13 14 15 16 1817 19 20 21Cluster

MPpDCILC

T cell

B cellPlasma cell

Mast cellEndothelia

Mesenchyme

HepatocyteCholangiocyte

Cycling 0.000.250.500.751.00

0.000.250.500.751.00

c

MPpD

C ILC T cell

B cell

Plasma c

ell

Mast c

ell

Endoth

elia

Mesen

chym

e

Epithe

lia

Cyclin

g

0

2,000

4,000

6,000

8,000

nGen

e

eCell lineage

B cell

Endothelia

Epithelia

ILC

Mast cell

Mesenchyme

MPpDC

Plasma cell

T cell

Cycling

Healthy 1 Healthy 4Healthy 2 Healthy 5Healthy 3

Cirrhotic 1 Cirrhotic 4Cirrhotic 2 Cirhotic 5Cirrhotic 3

Leucocytes and other NPCs Leucocytes only MP

pDC

ILC

T cell

B cell

Plasma cell

Mast cell

Endothelia

Mesenchyme

Epithelia

Pearson correlation between samples1.000.950.900.800.65 0.70 0.75 0.85

HealthyCirrhotic

f

MPpD

C ILC T cell

B cell

Plasma c

ell

Mast c

ell

Endoth

elia

Mesen

chym

e

Epithe

lia

Cyclin

g

nUM

I

0

20,000

40,000

60,000

80,000

100,000d

Healthy 1

Cirrhotic 1NAFLD

Healthy 2

Cirrhotic 2ALD

Healthy 4

Cirrhotic 4NAFLD

Healthy 5

Cirrhotic 5PBC

Cirrhotic 3ALD

Healthy 3

MP

B cell

Mesenchyme

pDC

Plasma cell

Hepatocyte

ILC

Mast cell

Cholangiocyte

T cell

Endothelia

Cycling

0.000.250.500.751.00

b

HealthyCirrhotic

HealthyCirrhotic

Fraction of cells

Sig. score

Sig. score

Extended Data Figure 2

Line

age

sign

atur

e

Cell lineage Cell lineageC

ell l

inea

ge

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

56

Extended Data Figure 3: Annotating human liver lymphoid cells. 1530

a, t-SNE visualisation: clustering 36,900 T cells and innate lymphoid cells (ILC), 1531

annotating injury condition. cNK, cytotoxic NK cell. b, Fractions of T cell and ILC 1532

subpopulations in healthy (n=5) versus cirrhotic (n=5) livers, Wald. c, t-SNE 1533

visualisations: selected genes expressed in the T cell and ILC lineage. d, Scaled 1534

heatmap (red, high; blue, low): T cell and ILC cluster marker genes (colour coded top 1535

by cluster and condition), exemplar genes labelled right. Cells columns, genes rows. e, 1536

t-SNE visualisations: downsampled T cell and ILC dataset (7,380 cells) pre- and post-1537

imputation; annotating data used for visualisation and clustering, inferred lineage, 1538

injury condition. No additional heterogeneity was observed following imputation. f, t-1539

SNE visualisation: clustering 2,746 B cells and plasma cells, annotating injury 1540

condition. g, Scaled heatmap (red, high; blue, low): B cell and plasma cell cluster 1541

marker genes (colour coded top by cluster and condition), exemplar genes labelled 1542

right. Cells columns, genes rows. h, Fractions of B cell and plasma cell subpopulations 1543

in healthy (n=5) versus cirrhotic (n=5) livers, Wald. Error bars, s.e.m.; *** p-value < 1544

0.001. 1545

c

a b

2

1

0

-1

-2

LTBIL7RCD52VIMCCL5CD8AGZMHNCR3KLRG1GPR171

CCL3XCL1IL2RBXCL2CD160KLRD1

CLIC3LAT2CCL3L3GNLYPRF1FGFBP2SPON2

CEBPD

ConditionCluster

0.6

0.4

0.2

0

Frac

tion

of T

cel

ls a

nd IL

Cs

Cluster

CirrhoticHealthy

d

CD4

23

10

NCAM1

23

10

B cell(1)

B cell(2)

Plasma cell(2)

Plasma cell(1)

ClusterHealthyCirrhotic

Condition

b

f g

h

Condition

2

1

0

-1

-2

ClusterHLA-DRA

LTBMS4A1TCL1A

FCER2CD79B

IGHA1MZB1IGHA2

SELMCD38

IGLC7SDC1IGLC3PDK1

IGHG3IGHGP

B cell(1

)

B cell(2

)

Plasma

cell(2

)Plas

ma

cell(1

)

Cluster

0.5

0.4

0.2

0

0.3

0.1

Frac

tion

of B

cel

ls a

ndPl

asm

a ce

lls

CirrhoticHealthy

HealthyCirrhotic

Condition

0

1

2

3

4

5

6 7

8

scImputecluster

012345678

Seurat t-SNE scImpute t-SNE

scImpute t-SNE scImpute t-SNE

e

Extended Data Figure 3

*** *** *** ***

2

10 9

8

7

6

5

1

4

3

11IdentityCD4+ T cell(1)CD4+ T cell(2)CD8+ T cell(1)CD8+ T cell(2)CD8+ T cell(3)Cycling T cellNK cell(1)NK cell(2)cNK cell

Cycling cNKCycling NK

11

12345678910

CD4+ T cell(1

)

CD4+ T cell(2

)

NK cell(1

)

NK cell(2

)

cNK ce

ll

CD8+ T cell(1

)

CD8+ T cell(2

)

CD8+ T cell(3

)

Seurat clusterCD4+ T cell(1)CD4+ T cell(2)CD8+ T cell(1)CD8+ T cell(2)CD8+ T cell(3)Cycling T cellNK cell(1)NK cell(2)cNK cell

Cycling cNKCycling NK

Seurat clusterCD4+ T cell(1)CD4+ T cell(2)CD8+ T cell(1)CD8+ T cell(2)CD8+ T cell(3)Cycling T cellNK cell(1)NK cell(2)cNK cell

Cycling cNKCycling NK

CD3D

23

10

4

GZMA

23

10

4

MKI67

23

10

CD8A

23

10

4

KLRF1

23

10

4

CXCR4

2310

45

FCGR3A

23

10

4

SELL

23

10

GZMB

2310

45

CCR7

23

10

CD69

IFNG

231

45

2310

45

0

HealthyCirrhotic

Condition

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

57

Extended Data Figure 4: Annotating human liver mononuclear phagocytes. 1546

a, t-SNE visualisation: clustering and selected genes expressed in mononuclear 1547

phagocyte (MP) lineage. b, Violin plots: Kupffer cell (KC) cluster markers. c, 1548

Representative immunofluorescence micrograph, healthy versus cirrhotic liver: TIMD4 1549

(red), CD163 (white), MARCO (green), DAPI (blue), arrows 1550

CD163+MARCO+TIMD4- cells. d, Automated cell counting: TIMD4 staining, healthy 1551

(n=12) versus cirrhotic (n=9) liver, Mann-Whitney. e, Automated cell counting: 1552

MARCO staining, healthy (n=8) versus cirrhotic (n=8) liver, Mann-Whitney. f, 1553

Representative flow cytometry plots: gating strategy for identifying KC, TMo and 1554

SAM. SAM are detected as TREM2+CD9+ cells within the TMo and SAM gate (see 1555

Fig. 2f). g, Representative immunofluorescence micrograph, cirrhotic liver: TREM2 1556

(red), MNDA (white), collagen 1 (green), DAPI (blue). h, Representative micrograph, 1557

cirrhotic liver: TREM2 (smFISH; red), MNDA (immunofluorescence; green), DAPI 1558

(blue). i, Representative immunofluorescence micrograph, cirrhotic liver: CD9 (red), 1559

MNDA (white), collagen 1 (green), DAPI (blue). j, Violin plots: cycling MP cluster 1560

markers. k, Fractions of cycling MP subpopulations in healthy (n=5) versus cirrhotic 1561

(n=5) livers, Wald. Scale bars, 50μm. Error bars, s.e.m.; * p-value < 0.05, *** p-value 1562

< 0.001. 1563

kj

f

SS

C-A

0

103

104

105

-103

0 103 104-103

CD45-BV650

Aut

oflu

ores

cenc

e

0

103

104

105

-103

HLA-DR-AF4880 103 104-103 105

CD

14-B

UV

395

0

103

104

105

-103

CD1C-PE/Dazzle™5940 103 104-103 105

CD

14-B

UV

395

0

103

104

105

-103

0 103 104-103

CD16-BUV737 CD163-BV4210 103 104-103 105

CD

9-P

E-V

io77

0

0

103

104

105

-103

Kupffercells

TMo and SAM

SS

C-A

0

103

104

105

-103

0 103 104-103

Lin-BV711(CD3, CD335, CD19, CD66b,

LILRA4, CD326)

g TREM2 MNDA DAPI

CD14

2310

4

CD68

2310

4

CD1C

2310

HLA-DRA

46

20

FCGR3A

2310

4

CSF1R

2310

CLEC9A

23

10

LYZ

46

20

MKI67

210

TIM

D4

Healthy Cirrhotic

MA

RC

O

Collagen 1MNDACD9 DAPI

MARCO

TIMD4

CD163

CD5L

VCAM1

Cluster

Gen

e ex

pres

sion

a b

Collagen 1MNDATREM2 DAPI

MARCOCD163TIMD4 DAPI

Cirr

hotic

Hea

lthy

c d

e

h i

Healthy Cirrhotic

TIM

D4+ c

ells

/mm

2 liv

er

0

100

200

300

Healthy Cirrhotic

MA

RC

O+ c

ells

/mm

2 liv

er

0

100

200

300

CirrhoticHealthy

CyclingSAM

CyclingcDC1

CyclingKC

CyclingcDC2

Cluster

0.010

0

0.005

0.015

Frac

tion

of M

P ce

lls

***

4

0

2

6 CD1C

Cyclin

g

cDC2

Cyclin

g

cDC1

Cyclin

g

KC Cyclin

g

SAM

4

0

2

6 MARCO

Cyclin

g

cDC2

Cyclin

g

cDC1

Cyclin

g

KC Cyclin

g

SAM

4

0

2

6 CLEC9A

Cyclin

g

cDC2

Cyclin

g

cDC1

Cyclin

g

KC Cyclin

g

SAM

4

0

2

6 TREM2

Cyclin

g

cDC2

Cyclin

g

cDC1

Cyclin

g

KC Cyclin

g

SAM

Gen

e ex

pres

sion

Extended Data Figure 4

*

2

10

9 8

76

5

43

1

TMo(1)

TMo(2)

SAM(1)

SAM(2)

KC(1)

KC(2)cD

C2cD

C1

TMo(3)

t-SNE1

t-SN

E2

58

Extended Data Figure 5: Phenotypic characterisation of mononuclear 1564

phagocytes in healthy and cirrhotic human livers. 1565

a, Self-Organising Map (SOM; 60x60 grid): smoothed mean metagene expression 1566

profile for mononuclear phagocyte (MP) subpopulations. b, Radar plots (left): 1567

metagene signatures D-F showing distribution of signature expression across MP 1568

subpopulations, exemplar genes (middle) and Gene Ontology (GO) enrichment (right). 1569

c, Luminex assay: quantification of levels of stated proteins in culture medium from 1570

FACS-isolated scar-associated macrophages (SAM, n=3), tissue monocytes (TMo, 1571

n=2), Kupffer cells (KC, n=2), and control (media alone, n=2). MFI, median 1572

fluorescence intensity. d, Hepatic stellate cell (HSC) activation assay: HSC treated with 1573

conditioned media from FACS-isolated SAM (n=3) or TMo (n=3); qPCR of stated 1574

genes, expression relative to mean expression of control HSC (n=6), Kruskal-Wallis 1575

and Dunn. e, Cubic smoothing spline curve fitted to averaged expression of all genes 1576

in module 2 from blood monocyte-SAM pseudotemporal trajectory, selected GO 1577

enrichment (right) and curves fit to exemplar genes (below). f, Cubic smoothing spline 1578

curve fitted to averaged expression of all genes in module 3 from blood monocyte-cDC 1579

pseudotemporal trajectory, GO enrichment (right) and curves fit to exemplar genes 1580

(below). g, Scaled heatmap (red, high; blue, low): transcription factor regulons across 1581

MP pseudotemporal trajectory and in KC. Colour coded top by MP cluster, condition 1582

and pseudotime, selected regulons labelled right. Cells columns, regulons rows. h, 1583

Violin plots: selected regulons expressed across MP clusters. 1584

baTMo(1) TMo(2) TMo(3)

SAM(1) SAM(2) KC(1)

KC(2) cDC2 cDC1

MP cluster

PseudotimeConditionCluster

HES1

EGR2

BCL11A

NFE2NFKB1

HOXB7

NR1H3SPIC

MAFB

STAT2

HIF1A

E2F2

EGR3

2

1

0

-1

-2

Cluster

NR1H3

SPIC

HES1

Reg

ulon

act

ivity

KC(2)

TMo(1)

TMo(2)

TMo(3)

SAM(1)

KC(1)

SAM(2)

EGR2

BCL11A

NFE2

0.08

0.04

0

0.15

0.050

0.10

0.03

0.010

0.02

0.075

0.0250

0.050

0.15

0.050

0.10

0.050

0.10

hg

Extended Data Figure 5Genes Gene Ontology

cDC1

KC(2)SAM(1)

cDC2

KC(1)SAM(2)

Sign

atur

e D

D

cDC1

KC(2)

cDC2

KC(1)SAM(2)

Sign

atur

e E

E

-log10(P-value)

cDC1

KC(2)SAM(1)

cDC2

KC(1)SAM(2)Si

gnat

ure

F

F

HES4

IFNLR1CSF3R

PLOD1CASP9

PTGER3CHI3L2S100A9S100A12VCAN

RHOF

DUSP4IDO1CLEC9AITGB7

FLT3

BATF3PLCD1XCR1

ENPP1CXCR3

TLR10

MAP4K4SLC11A1PTX3BST1S100ZHRH2

TREM1IL3RANRG1

IL17RA

CLEC4ECD300C

Inflammatory response Response to lipopolysaccharideNADP metabolic process

Leukocoyte chemotaxis20 4

T cell migration

Positive regulation of apoptotic processRegulation of cell communication

Positive regulation of cytokine production

20 4

Leukocyte migration

Defense response

Cell communicationRegulation of kinase activity

50 10

fe

TMo(1)TMo(2)

TMo(3)

TMo(1)TMo(2)

TMo(3)

TMo(1)TMo(2)

TMo(3)

SAM(1)

Gene Ontology

-log10(P-value)0 7 14 21

Defense response

Response to bacterium

Leukocyte chemotaxis

Leukocyte degranulation

Leukcoyte migration

1.0

Module 2

0 0.5Pseudotime

Sig

natu

re

expr

essi

on

-2.5

0

2.5

Scal

ed g

ene

expr

essi

on

Pseudotime

-2.5

0

2.5

0 0.5 1.0

S100A12

-2.5

0

2.5

0 0.5 1.0

LYZ

-2.5

0

2.5

0 0.5 1.0

FCN1

-2.5

0

2.5

0 0.5 1.0

SELL

-2.5

0

2.5

0 0.5 1.0

SLC11A1

-2.5

0

2.5

0 0.5 1.0

MNDA

Gene Ontology

-log10(P-value)0 2 4 6

Regulation of T cell activation

Dendritic cell differentiation

Interleukin-12-mediated signaling pathway

Regulation of lymphocyte proliferation

Fc receptor signaling pathway

1.0

Module 3

0 0.5Pseudotime

Sig

natu

re

expr

essi

on

-2.5

0

2.5

Scal

ed g

ene

expr

essi

on

Pseudotime

-2.5

0

2.5

0 0.5 1.0

CD1C

-2.5

0

2.5

0 0.5 1.0

ITGB7

-2.5

0

2.5

0 0.5 1.0

CLEC9A

-2.5

0

2.5

0 0.5 1.0

BATF3

-2.5

0

2.5

0 0.5 1.0

IDO1

-2.5

0

2.5

0 0.5 1.0

FCER1A

MFI

Population

c

30

0

60

90

Contro

lSAM

TMo KC

IL-1ß

30

0

60

90 CCL2

Contro

lSAM

TMo KC

90

80

100

110 SPP1

Contro

lSAM

TMo KC

20

0

3,000

6,000 LGALS3

Contro

lSAM

TMo KC

20

0

5,000

10,000 CXCL8

Contro

lSAM

TMo KC

d

2

0

8

10

4

6

Rel

ativ

e H

SC

gene

exp

ress

ion

COL1A1

Contro

lSAM

TMo

Conditioned Media

*

0

30 COL3A1

Contro

lSAM

TMo

**20

10

59

Extended Data Figure 6: Characterisation of macrophages in mouse liver 1585

fibrosis. 1586

a, t-SNE visualisation: clustering 3,250 mouse mononuclear phagocytes (mMP) into 5 1587

clusters (colour and number). Annotation of mouse clusters (identity). mTMo, tissue 1588

monocyte; mSAM, scar-associated macrophage; mKC, Kupffer cell; mcDC, 1589

conventional dendritic cell. b, t-SNE visualisation: annotating mMP cells by injury 1590

condition (colour); healthy (uninjured) or fibrotic (4 weeks carbon tetrachloride 1591

treatment). c, Scaled heatmap (red, high; blue, low): mMP cluster marker genes (top, 1592

colour coded by cluster and condition), exemplar genes labelled (right). Cells columns, 1593

genes rows. d, t-SNE visualisations: selected genes expressed in mMP. e, 1594

Representative flow cytometry plots: gating strategy for identifying mKC, CD9- mTMo 1595

and CD9+ mSAM. f, Quantifying mouse macrophage subpopulations by flow 1596

cytometry: healthy (n=6) and fibrotic (n=8) mouse liver, macrophage subpopulation (x-1597

axis) as a percentage of total viable CD45+ cells (y-axis), Mann-Whitney. g, Hepatic 1598

stellate cell activation assay: co-culture of hepatic stellate cells (HSC) from uninjured 1599

mouse liver and FACS-isolated macrophage subpopulations (Mθ) from fibrotic mouse 1600

liver (left). Co-culture with CD9- mTMo (n=8) or CD9+ mSAM (n=8), qPCR of 1601

Col3a1 in HSC, expression relative to mean expression of quiescent HSC (right), 1602

Wilcoxon. h, t-SNE visualisation: clustering 3,250 mouse mononuclear phagocytes 1603

(mMP) and 10,737 human mononuclear phagocytes (hMP) into 5 clusters (colour and 1604

number) using canonical correlation analysis (CCA). Annotation of cross-species 1605

clusters (identity). i, t-SNE visualisations: human and mouse macrophage 1606

subpopulation annotation. j, t-SNE visualisations: selected genes expressed in cross-1607

species clusters. Error bars, s.e.m.; ** p-value < 0.01; *** p-value < 0.001. 1608

ca bExtended Data Figure 6

1

2

45

3

ClusterClustermMP(1)

mMP(2)

mMP(4)

mMP(5)

mMP(3)

Identity1

2

3

4

5

mTMo

mSAM

mcDC2

mcDC1

mKC

Mouse Healthy

Fibrotic

0 103 104-103 105

CD

11b-

AP

C-C

y70

103

104

105

-103

F4/80-PE-Cy7 CD9-PE0 103 104-103 105

SS

C-A

0

103

104

105

-103

CD9+CD9-

TIMD4-AF6470 103 104-103 105

CD

11b-

AP

C-C

y7

0

103

104

105

-103

mKC

Mono-Mac

e

f

mKC

% o

f CD

45+

Cel

ls

60

0

20

40

CD9+ mSAM

CD9- mTMo

***HealthyFibrotic

*** *** g

210-1-2

ConditionCluster

Trem2

Cd74Cd209a

Vcam1

Spp1

Cd5l

Clec9a

PlaurLyz2

Timd4

S100a4

Cd163Marco

Xcr1

Ccl3Lgals3

Cst3

Human & Mouse

ClusterClusterh&mMP(1)

h&mMP(2)

h&mMP(4)

h&mMP(5)

h&mMP(3)

Identity1

2

3

4

5

TMo

SAM

cDC

Cycling

KC1

2

4

5 3

Mouse SAMHuman SAM

Human KC Mouse KC

TREM2

MARCO SPP1

CD9 TIMD4

LGALS3

Mθ

HSC

HSC GeneExpression

d

h i j

Rel

ativ

e H

SC

G

ene

Exp

ress

ion

CD9+CD9-

8

0

12

4

Mθ Subpopulation

Col3a1**

Cd9 Trem2

Lyz2

Timd4

Cd209a Xcr1

20

4

2310

210

46

20

20

42310

46

20

46

20

46

20

40

8

531

531

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

60

Extended Data Figure 7: Scar-associated macrophage expansion in human 1609

NASH 1610

a to d, Deconvolution: publicly available whole liver microarray data (n=73) assessed 1611

for frequency of scar-associated macrophages (SAM), Kupffer cells (KC) and tissue 1612

monocytes (TMo) using Cibersort algorithm. a, Macrophage composition: x-axis, GEO 1613

accession number; y-axis, fraction of monocyte-macrophages; Top, annotated by liver 1614

phenotype; NASH, Non-alcoholic steatohepatitis. b, Frequency of SAM in control 1615

(n=14), heathy obese (n=27), steatosis (n=14) and NASH (n=18) livers, Kruskal-Wallis 1616

and Dunn. c, Frequency of SAM in patients with histological NAFLD activity score 1617

(NAS) of 0 (n=37), 1-3 (n=19) and 4-7 (n=17) (left). Frequency of SAM in patients 1618

with histological fibrosis score of 0 (n=46), 1 (n=20) and 2-4 (n=5) (right), Kruskal-1619

Wallis and Dunn. d, Frequency of SAM in female (n=58) and male (n=15) patients 1620

(left). Frequency of SAM in patients aged 23-39 (n=22), 40-49 (n=29) and 50-80 (n=22) 1621

(middle). Frequency of SAM in patients with a body mass index (BMI) of 17-30 (n=18), 1622

31-45 (n=28) and 46-70 (n=27) (right). All Kruskal-Wallis and Dunn. e, CD9 and 1623

TREM2 staining in NASH liver biopsy sections (left). Automated cell counting (right): 1624

CD9 staining, NAS 1-3 (n=13) versus NAS 4-8 (n=21). TREM2 staining, NAS 1-3 1625

(n=12) versus NAS 4-8 (n=16), Mann-Whitney. f, Correlation of automated cell counts 1626

with picrosirius red (PSR) digital morphometric pixel quantification in NAFLD liver 1627

biopsy tissue; CD9 staining (top; n=39); TREM2 staining (bottom; n=32); Pearson 1628

correlation and linear regression. Scale bars, 50μm. Error bars, s.e.m.; * p-value < 0.05, 1629

** p-value < 0.01. 1630

GEO accession numberb

aExtended Data Figure 7

CD

9+ cel

ls/m

m2 l

iver

4-8NAS score

1-3

50

0

100

150

200 *

25PSR % area

0 5 10 15 20

CD

9+ cel

ls/m

m2 l

iver

0

100

200

300

p<0.0001r=0.8199

TREM

2+ cel

ls/m

m2 l

iver

4-8NAS score

1-3

50

0

100

150

25PSR % area

0 5 10 15 20

TREM

2+ cel

ls/m

m2 l

iver

0

100

200

300

p=0.0003r=0.5952

Frac

tion

of m

onoc

yte-

mac

roph

ages

0

0.5

1.0

Normal Healthy obese Steatosis NASH

KCSAMTMo

0

0.3

0.1

0.2

Frac

tion

of m

onoc

yte-

mac

roph

ages

4-71-30NAS score

***

0

0.5

0.1

0.4

Frac

tion

of m

onoc

yte-

mac

roph

ages

0.3

0.2

***

2-410Fibrosis score

0

0.3

0.1

0.2

Frac

tion

of m

onoc

yte-

mac

roph

ages

50-8040-4923-39Age

0

0.3

0.1

0.2

Frac

tion

of m

onoc

yte-

mac

roph

ages

46-7031-4517-30BMI

0

0.3

0.1

0.2

Frac

tion

of m

onoc

yte-

mac

roph

ages

MaleFemaleSex

d

TREM

2C

D9

NASH

Contro

l

Health

y

obes

e

Steatos

is

NASH0

0.3

0.1

0.2

Frac

tion

of m

onoc

yte-

mac

roph

ages

*

Liver Phenotype

c

e f

61

Extended Data Figure 8: Phenotypic characterisation of endothelial cells in 1631

healthy and cirrhotic human livers. 1632

a, t-SNE visualisations: clusters and selected genes expressed in endothelial lineage. b, 1633

Violin plots: endothelial cluster marker genes. c, Digital morphometric pixel 1634

quantification: PLVAP stained cirrhotic sections (n=10) morphologically segmented 1635

into fibrotic septae and parenchymal nodules (top); ACKR1 stained cirrhotic sections 1636

(n=10) morphologically segmented into fibrotic septae and parenchymal nodules 1637

(bottom), Wilcoxon. d, Flow cytometry: endothelial cells from healthy (n=3) or 1638

cirrhotic (n=7) liver, representative histogram for stated marker (top), mean 1639

fluorescence intensity (MFI, bottom), Mann-Whitney. e, Flow-based adhesion assay: 1640

peripheral blood monocytes were assessed for adhesion (left) and % of adherent cells 1641

which transmigrate (right); endothelial cells from healthy (n=5) or cirrhotic (n=4) liver, 1642

Mann-Whitney. f, Endothelial cell gene knockdown: cirrhotic endothelial cells treated 1643

with siRNA to PLVAP (n=6), ACKR1 (n=5) or control siRNA (n=6). Representative 1644

flow cytometry histograms for stated marker (top); comparison to isotype control 1645

antibody. Flow-based adhesion assay (bottom), peripheral blood mononuclear cells 1646

assessed for adhesion (bottom left) and % of adherent cells which transmigrate (bottom 1647

right) following siRNA treatment of endothelial cells, Kruskal-Wallis and Dunn. g, 1648

Self-Organising Map (SOM; 60 x 60 grid; top left): smoothed scaled metagene 1649

expression of endothelia lineage. 21,237 genes, 3,600 metagenes, 45 signatures. A-E 1650

label metagene signatures overexpressed in one or more endothelial subpopulations. 1651

SOM: smoothed mean metagene expression profile for each endothelial subpopulation 1652

(top right). Radar plots (bottom left): metagene signatures A-E showing distribution of 1653

signature expression across endothelial subpopulations, exemplar genes (bottom 1654

middle) and Gene Ontology (GO) enrichment (bottom right). h, t-SNE visualisation: 1655

endothelia subpopulation annotation, injury condition. i, t-SNE visualisation: 1656

endothelial lineage annotated by monocle pseudotemporal dynamics (purple to yellow; 1657

grey indicates lack of inferred trajectory). RNA velocities (red arrows) visualised using 1658

Gaussian smoothing on regular grid. Error bars, s.e.m.; * p-value<0.05; ** p-value < 1659

0.01. 1660

b

0.0

0.5

1.0Pseudotime

ERG

210

46

PECAM1

210

3

ICAM2

210

3

CD34

210

3

VWF

210

34

CDH5

210

3

ENG

210

34

d

e

h

g

Extended Data Figure 8

Endo(7)

Gen

e ex

pres

sion

Endo(1)

Endo(2)

Endo(3)

Endo(4)

Endo(5)

Endo(6)

Cluster

CLEC4M

CD34

PDPN

RSPO3

AIF1L

PLVAP

VWA1

ACKR1

CLEC4G

LSEC

ClusterEndo(1)1

Hepatic artery

Endo(5)5Cluster

Central vein

Endo(4)4Cluster

Scar-associated

Endo(6)Endo(7)

67

Cluster

Lymphatic

Endo(2)2Cluster

HealthyCirrhotic

Condition

Endo(1) Endo(3)Endo(2)

Endo(4) Endo(6)Endo(5)

Endo(7)

A

C B

E

D

g

Genes Gene Ontology

Sign

atur

e C Endo(1)

Endo(2)

Endo(3)

Endo(4) Endo(5)

Endo(6)

Endo(7)

C

AIF1LCD200CXCR4SOX17EPHA4CXCL12

VEGFCJAG2DLL4JAG1

EDNRBHBEGF

LAMC1DYSFPGFITGA2

LOXL2LAMB1

PDGFDCX3CL1PDGFBLOXCOL15A1COL4A1

Sign

atur

e A Endo(1)

Endo(2)

Endo(3)

Endo(4) Endo(5)

Endo(6)

Endo(7)

A

ACKR1MDK

CTGF

FN1EDN1

MIF

S100A10HYAL2IL3RAIL32ITGA6FABP4

Sign

atur

e B Endo(1)

Endo(2)

Endo(3)

Endo(4) Endo(5)

Endo(6)

Endo(7)

B

Sign

atur

e D Endo(1)

Endo(2)

Endo(3)

Endo(4) Endo(5)

Endo(6)

Endo(7)

D

CLEC4MFCGR2BITGA9STAB1SELPSCARB2

MRC1CD36

GATA4IL33

CD14

TGFBR3

Endo(1)Endo(2)

Endo(3)

Endo(4) Endo(5)

Endo(6)

Endo(7)

E

Sign

atur

e E PROX1

PTX3

MMRN1SEMA3DIL7

HOXA7

IGF1TBX1PDPN

CCL21TFF3

HOXD8

-log10(P-value)

Endothelium development

Regulation of Notch signaling pathway

Animal organ developmentEndothelial cell differentiation

Angiogenesis

0 4 8 12

Vasculature development

Wound healing

Extracelluar matrix organizationResponse to wounding

Angiogenesis

0 5 10 15

Peptide biosynthetic process

Antigen processing and presentationMyeloid leukocyte activation

Activation of immune response0 20 40 60

Non-canonical Wnt signaling pathway

Vesicle-mediated transportBlood vessel development

EndocytosisImmune response

0 6 12 18Receptor-mediated endocytosis

Lymphangiogenesis

Positive regulation of cell migrationLymph vessel development

Regulation of cell adhesion

0 3 6 9

Regulation of cell motility

1

27

6

5

43

Nor

mal

ised

Cou

ntM

FIa

*Adhesion

Health

y

Cirrhoti

c

20

0

40

60

80

Adh

eren

t cel

ls p

er m

m2

per 1

06 cel

ls p

erfu

sed

10

0

20

40

50

30

Transmigration

Health

y

Cirrhoti

c

% o

f Adh

eren

t Cel

ls

c

PLV

AP

% A

rea

40

0

60

20

Fibroticseptae

Nodules

**

AC

KR

1 %

Are

a

0

0.01

Fibroticseptae

Nodules

**

0.02

0.05

0.04

0.03

Nor

mal

ised

Cou

nt

*

Health

y

Cirrhoti

c

200

0

400

600

800

PLVAP

PLVAP-APC0 103 104

0

40

80

20

60

100

Contro

l

PLVAP

ACKR1

siRNA

10

0

20

40

30

% o

f Adh

eren

t Cel

ls *TransmigrationAdhesion

Adh

eren

t cel

ls p

er m

m2

per 1

06 cel

ls p

erfu

sed

20

0

40

60

80

100

Contro

l

PLVAP

ACKR1

siRNA

0

20

100

40

60

80

100 101

PLVAP-APC

PLVAPsiRNA

ControlsiRNA

Isotype

0

20

100

40

60

80

100 101

ACKR1-PEVio770

ACKR1siRNA

ControlsiRNA

Isotype

f

i

5,000

0

10,000

15,000

20,000

Health

y

Cirrhoti

c

CD34

0 104 105

CD34-BV6500

40

80

20

60

100

Health

y

Cirrhoti

c

500

0

1,000

1,500

2,000

ACKR1

0 104 105103

ACKR1-PEVio7700

40

80

20

60

100

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

62

Extended Data Figure 9: Characterisation of mesenchymal cells in healthy and 1661

cirrhotic human livers. 1662

a, t-SNE visualisations: selected genes expressed in mesenchymal lineage. b, t-SNE 1663

visualisation: clustering 319 scar-associated mesenchymal cells (SAMes), into 2 further 1664

subclusters (colour and number). c, Scaled heatmap (red, high; blue, low): SAMes 1665

subcluster marker genes (top, colour coded by cluster and condition), exemplar genes 1666

labelled (right). Cells columns, genes rows. d, Fractions of SAMes subpopulations in 1667

healthy (n=4) versus cirrhotic (n=3) livers, Wald. e, Representative 1668

immunofluorescence micrograph, portal region of healthy liver: OSR1 (red), Collagen 1669

1 (green), DAPI (blue). e, Representative immunofluorescence micrograph, fibrotic 1670

niche of cirrhotic liver: OSR1 (red), Collagen 1 (green), DAPI (blue). g, t-SNE 1671

visualisation: Hepatic stellate cell (HSC) and SAMes clusters annotated by monocle 1672

pseudotemporal dynamics (purple to yellow). RNA velocity field (red arrows) 1673

visualised using Gaussian smoothing on regular grid. h, Scaled heatmap (red, high; blue 1674

low): cubic smoothing spline curves fitted to genes differentially expressed across 1675

HSC-to-SAMes pseudotemporal trajectories, grouped by hierarchical clustering (k=2). 1676

Colour coded by pseudotime and condition (top). Gene co-expression modules (colour) 1677

and exemplar genes labelled right. Scale bars, 50μm. Error bars, s.e.m.; *** p-value < 1678

0.001. 1679

1680

aExtended Data Figure 9

MYH11

COL3A1

RGS5

COL1A1

KRT19

TIMP1

PDPN

PDGFRA

23

10

531

46

20

10

2

46

20

2310

20

4

46

20

cb

12

3 4

3A

3B

ClusterSAMes A

SAMes B3B

3A

210-1-2

ConditionCluster

IGFBP3EDNRB

OGNCRABP2CTHRC1OSR1

d

0.0

0.5

1.0Pseudotime

h

ConditionPseudotime

CSRP2RGS5

FABP4FABP5

ID1IGFBP5

HealthyCirrhotic

SAMes cluster

Frac

tion

of

mes

ench

ymal

cel

ls

0.10

0.05

0

0.15

0.25

SAMes A

Condition

SAMes B

0.20

*** *** OSR1 Collagen1 DAPI OSR1 Collagen1 DAPI

Cirr

hotic

Hea

lthy

e f

PDGFRALOXL1CCL2 COL3A1

COL1A1

TIMP1

COL1A2

ADAMTS1GEM

IGF1

Module 1

Module 2

HSC SAMesg

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

t-SNE1

t-SN

E2

RBP1

63

Extended Data Figure 10: Characterisation of the cellular interactome in the 1681

fibrotic niche. 1682

a, Circle plot: potential interaction magnitude from ligands expressed by scar-1683

associated macrophages (SAM) and endothelial cells (SAEndo) to receptors expressed 1684

on scar-associated mesenchyme (SAMes). b, Circle plot: potential interaction 1685

magnitude from ligands expressed by SAMes to receptors expressed on SAM and 1686

SAEndo. c, Dotplot: ligand-receptor interactions between SAMes, SAM and SAEndo. 1687

X-axis, ligand (red) and cognate receptor (blue); y-axis, ligand (red) and receptor (blue) 1688

expressing cell populations; circle size, p-value; colour (red, high; yellow, low), means 1689

of average ligand and receptor expression levels in interacting subpopulations. d, 1690

Representative immunofluorescence micrographs, fibrotic niche in cirrhotic liver. Top; 1691

CCL2 (red), CCR2 (white), PDGFRα (green), DAPI (blue), arrows CCL2+PDGFRα+ 1692

cells. Bottom; ANGPT1 (red), TEK(white), PDGFRa (green), DAPI (blue), arrows 1693

ANGPT1+PDGFRa+ cells. e, Circle plot: potential interaction magnitude from ligands 1694

expressed by SAM to receptors expressed on SAEndo. f, Dotplot: ligand-receptor 1695

interactions between SAM and SAEndo. X-axis, ligand (red) and cognate receptor 1696

(blue); y-axis, ligand (red) and receptor (blue) expressing cell populations; circle size, 1697

p-value; colour (red, high; yellow, low), means of average ligand and receptor 1698

expression levels in interacting subpopulations. g, Representative immunofluorescence 1699

micrographs, fibrotic niche in cirrhotic liver. TREM2 (red), FLT1 (white), VEGFA 1700

(green), DAPI (blue), arrows TREM2+VEGFA+ cells. h, Circle plot: potential 1701

interaction magnitude from ligands expressed by SAEndo to receptors expressed on 1702

SAM. i, Dotplot: ligand-receptor interactions between SAEndo and SAM. X-axis, 1703

ligand (red) and cognate receptor (blue); y-axis, ligand (red) and receptor (blue) 1704

expressing cell populations; circle size, p-value; colour (red, high; yellow, low), means 1705

of average ligand and receptor expression levels in interacting subpopulations. j, 1706

Representative immunofluorescence micrographs, fibrotic niche in cirrhotic liver. Top; 1707

TREM2 (red), CD200 (white), CD200R (green), DAPI (blue), arrows 1708

TREM2+CD200R+ cells. Bottom; TREM2 (red), DLL4 (white), NOTCH2 (green), 1709

DAPI (blue), arrows TREM2+NOTCH2+ cells. Scale bars, 50μm. 1710

1711

Extended Data Figure 10

VEGFAFLT1TREM2 DAPI

SAM→SAEndinteractions

Ephrin Angiogenic

CC

L2-A

CK

R1

CX

CL8

-AC

KR

1

GH

RL-

TBX

A2R

IL15

-IL15

RA

IL-1

RN

-IL1

Rec

epto

r

TGFB

1-TG

FBR

3

PD

GFC

-FLT

4

VE

GFB

-AD

RB

2

VE

GFB

-FLT

1

VE

GFB

-NR

P1

VE

GFA

-FLT

1

VE

GFA

-KD

R

VE

GFA

-NR

P2

VE

GFA

-NR

P1

EFN

B1-

EP

HB

4

EFN

B1-

EP

HA

4

EFN

A4-

EP

HA

4

EFN

A4-

EP

HA

2

Ligand-Receptor pair

SAM(1)-HAEnd

SAM(2)-HAEnd

SAM(1)-SAEnd(1)

SAM(2)-SAEnd(1)

SAM(1)-SAEnd(2)

SAM(2)-SAEnd(2)

0 1 2 3

1.00.50.0

-log10(P-value)

Mean

dPDGFRαCCR2CCL2 DAPI

Ligand-Receptor pair

GA

S6-

AX

L

CX

CL1

2-C

XC

R4

CC

L2-C

CR

2C

CL1

1-C

CR

2C

CL2

6-C

CR

2C

CL2

6-C

CR

1C

CL1

9-C

CR

L2C

X3C

L1-C

X3C

R1

CS

F1-C

SF1

RIL

-34-

CS

F1R

PTN

-PLX

NB

2

CX

CL1

2-A

CK

R3

CC

L2-A

CK

R1

CX

CL1

-AC

KR

1C

XC

L8-A

CK

R1

WN

T4-F

ZD6

HG

F-M

ET

TGFB

3-TG

FBR

3TG

FB1-

TGFB

R3

VE

GFC

-FLT

4V

EG

FC-K

DR

VE

GFA

-KD

RV

EG

FA-F

LT1

VE

GFA

-NR

P1

VE

GFA

-NR

P2

VE

GFB

-AD

RB

2

VE

GFB

-NR

P1

VE

GFB

-FLT

1

AN

GP

T1-T

EK

AN

GP

T2-T

EK

ChemokinesCSF1R TGFß Angiogenic0 1 2 3

1.00.50.0

-log10(P-value)

Mean

SAMes-SAM(1)

SAMes→SAM and SAEndinteractions

SAMes-SAM(2)

SAMes-HAEnd

SAMes-SAEnd(1)

SAMes-SAEnd(2) PDGFRαTEKANGPT1 DAPI

e

b

c

CD200RCD200TREM2 DAPI

NOTCH2DLL4TREM2 DAPI

SAEnd→SAMinteractions

Ligand-Receptor pair

JAG

1-N

OTC

H1

JAG

2-N

OTC

H1

DLL

4-N

OTC

H1

JAG

1-N

OTC

H2

JAG

2-N

OTC

H2

DLL

4-N

OTC

H2

GA

S6-

AX

L

CX

CL1

2-C

XC

R4

CC

L2-C

CR

2

CX

3CL1

-CX

3CR

1

CS

F1-C

SF1

R

CD

200-

CD

200R

1

EFN

B1-

EP

HB

2

EFN

B2-

EP

HB

2

PD

GFB

-LR

P1

MD

K-L

RP

1

PR

OS

1-A

XL

GA

S6-

ME

RTK

ICA

M1-

ITG

AL

NO

V-N

OTC

H1

ICA

M1-

aMb2

com

plex

ICA

M1-

aLb2

com

plex

SAEnd(1)-SAM(2)

SAEnd(1)-SAM(1)

SAEnd(2)-SAM(1)

SAEnd(2)-SAM(2)

HAEnd-SAM(2)

HAEnd-SAM(1)

NotchChemokines TAM Integrin

0 1 2 3

1.00.50.0

-log10(P-value)

Mean

f

g

h i j

a