Single cell atlas of epithelial and stromal cell heterogeneity by lobe and strain in the mouse prostateMindy K Graham1,†, Roshan Chikarmane1,†, Rulin Wang1,†, Ajay Vaghasia1, Anuj Gupta1, Nicole Castagna1, Sarah Wheelan1, Charles Bieberich3, William G Nelson1,2, Angelo M De Marzo1,2, Srinivasan Yegnasubramanian1,21 Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University, School of Medicine 2 Department of Pathology, Johns Hopkins University, School of Medicine 3 Department of Biological Sciences, University of Maryland at Baltimore County† Contributed equally by authors MKG mgraha21@jhu.edu YV syegnasu@jhmi.edu

I. BACKGROUND

A

Bladder

SeminalVesicle

Ureter

AnteriorDorsal

Lateral

Ventral

Dissect each lobe

scRNA-seq

FVB/NJ (FVB)

C57BL/6J (BL6)

Histology

B

300 uM

100 uM100 uM

300 uM

10X

40X

Ventral

FVB

10X

40X

BL6

LateralAnterior

100 uM

300 uM

100 uM

300 uM

Dorsal

100 uM

300 uM

100 uM

300 uM

100 uM

300 uM

100 uM

300 uM

The mouse prostate is comprised of four lobes: anterior, dorsal, lateral, and ventral. While the lobes are known to di#er in tissue morphology and secretions, rigorous single-cell assessments of the transcriptional pro$le of both epithelial and stromal cell types for each lobe and in multiple mouse strains are lacking.

A) We dissected individual prostate lobes from two commonly used mouse strains, FVB/NJ (N = 2) and C57BL/6J (N = 3), and prepared single-cell RNA-sequencing (scRNA-seq) libraries and histology for each lobe. B) H&E staining of prostate lobes from each strain.

II. SINGLE-CELL RNA SEQUENCING OF MOUSE PROSTATES

Data dimensionality reduction and clustering analysis on scRNA-sequencing of mouse prostates revealed that epithelial cells possessed strain-speci$c di#erences, with luminal cells also displaying striking lobe-speci$c di#erences.

A) Mouse prostate scRNA-seq data (27,896 cells) were pre-processed using CellRanger and Seurat, and visualized using Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction and Louvain clustering, as implemented in Seurat (v 3.1.5). Cell types were identi$ed using previously characterized marker genes. B) Canonical marker genes used to identify cell types. C) Di#erential gene expression across clusters visualized as a heatmap of top 5 genes selected by p-value.

A

C

B

III. STRAIN AND LOBE SPECIFIC DIFFERENCES IN EPITHELIAL CELLS

Rpl29

Gm12

840

Scpe

p1Ms

4a5

Azgp

1

2210

407C

18Rik

Spink

8Wf

dc3

Scgb

2b20

Scgb

2b7

Eif3j2

Gm10

260

Gm10

709

Cap1Inm

tCc

dc80

Gdpd

3Cd

59a

Pcp4Itln1

Nupr1

S100

a10

Gprc5

aMg

st1Kr

t14Kr

t17

Basal

Luminal_Dorsal

Luminal_Anterior

Luminal_Lateral

Luminal_Ventral

Basal

Luminal_Dorsal

Luminal_Anterior

Luminal_Lateral

Luminal_Ventral

Genes

Percent Expressed0255075

0

1

2

Average Expression

FVB

BL6

TP63

0

2

4

Upstream Transcription Regulator (N = 1063)

Activation_StateActivated

Inhibited

NA

Basal vs Luminal

Krt15Sdc1

Lgals7Dusp6

AdmMarcksl1

Krt5Socs3Rasd1

Zfp36l2Arc

Epha2Rbms1

5330417C22RikDcxr

Fkbp11Prlr

Sdf2l1Tspan1

Agr29530003J23Rik

Mt3C1s2

Glb1l3A630095E13Rik

AY761184Sbp

Spink1SbplCbr2

Timp4Abo

Anxa1Gchfr

Rnf130

Gen

es

012

Average Expression

Percent Expressed0255075100

Strain

Lobe

Cell Type

Strain

BL6

FVB

Lobe

Anterior

Dorsal

Lateral

Ventral

Cell Type

Basal

Luminal

Immune

Fibroblast

Sm. Muscle

Pericytes

Endothelial

FVB & BL6

All Lobes

A

B C

D

E

A) Dotplot of top genes di#erentially expressed by strain in epithelial cells. B) To identify conserved genes expressed in epithelial cells independent of strain, we implemented Seurat’s integration analysis with the assumption that there are shared cell states present in both FVB and BL6 strains. Basal ecells across all lobes partitioned together, while luminal epithelial cells clustered in four groups by lobe. C) In the basal cluster, Tp63 is not as robustly expressed as other basal associated markers Krt5 and Krt15. D) Ingenuity Pathway Analysis of upstream regulators comparing the basal and epithelial clusters show that Tp63 is one of the top upregulated transcription factors. E) Dotplot of top di#erential genes expressed by lobe in epithelial. Plot also includes stromal cell types.

IV. UNIQUE LUMINAL POPULATIONS THAT CLUSTER INDEPENDENT OF LOBE AND STRAIN

Basa

lLu

mina

lStr

oma

Lumi

nal P

sca+

Lumi

nal F

oxi1+

Wfdc2Anxa1Ahnak

Tacstd2Gprc5a

PscaPglyrp1

Krt7Edn1

Pmaip1Ly6a

Tspan8Atp6v1g3

Foxi1Serpinb9

LdhbS100a1

CochCyp4v3

PamFam189a2

Atp6v1aGde1

Hepacam2Tfcp2l1

Serpinb6bRgs5Clic6Cds1

Atp6v0d2Cald1

Gen

es

Percent Expressed

0255075100

0

1

Average Expression

0.02

0.01

0.00

0.01

0.03 0.02 0.01 0.00 0.01

DC1

DC

2

Anterior

Dorsal

Lateral

Ventral

Luminal Psca+

Basal

DC1

DC

2

0.0

0.5

1.0

1.5

DPT

Path and TipsTips

Anterior

Dorsal

Lateral

Ventral

Luminal Psca+

Basal

0.0 0.5 1.0 1.5 2.0

Diffusion Pseudotime 0.02

0.01

0.00

0.01

0.03 0.02 0.01 0.00 0.01DC1

DC

2

0

1

2

3

Ly6a

0.02

0.01

0.00

0.01

0.03 0.02 0.01 0.00 0.01DC1

DC

2

0

1

2

3

4

5Psca

0.02

0.01

0.00

0.01

0.03 0.02 0.01 0.00 0.01DC1

DC

2

0

1

2

3

4Krt5

A E

B C

D

F

A) UMAP highlighting a mixed population of luminal epithelial cells that are not lobe or strain speci$c, and are either Foxi1+ or Psca+. B) A di#usion plot show that the Psca+ cells are positioned between basal and di#erentiated lobe-speci$c luminal cells. C) Plotted in di#usion pseudotime, with basal cells at time 0, the Psca+ cells are earlier than the lobe-speci$c luminal epithelial cells. D) Di#usion plot with di#usion pseudotime, with tips indicated for basal cells at the earliest time point, and progressing to anterior and dorsal as one tip, and lateral and ventral as a another tip. E) The mixed clusters have several unique marker genes that make them distinct from other cell types. F) In the di#usion plot, Ly6a, a stem-cell associated marker is expressed in both Psca+ cells and Krt5+ basal cells.

V. STROMAL AND IMMUNE CELLS OF MOUSE PROSTATE

Macrophage

T CellsDendritic

0

5

0 2UMAP_1

UM

AP_2

EndothelialFibroblast (1)

Fibroblast (2)

Pericytes

Smooth Muscle

0

5

10

0 10UMAP_1

UM

AP_2

Lumi

nal

Basa

lMa

croph

age

Myelo

idT c

ellsEn

dothe

lialFib

roblas

t (1)

Fibrob

last (2

)

Peric

ytes

Smoo

th Mu

scle

C1qbC1qcC1qaCcl4

ApoeCcl3

Cd68Gm13546

Mcemp1Ly75

Cbfa2t3Flt3Kit

Gimap6Cd3gIl2rb

Tnfrsf18Sh2d2a

Ctla2aCd3eCd2

Cd3dCdh5

EngPecam1

Aqp1Cd93

Flt1Lmo2S1pr1

Cyp1a1PlvapEgfl7

Efna1Adgrf5EmcnCyyr1Rnd1Rorb

F2rAdm

Rdh10Fzd1

Sult1e1LumOgn

Fbln1DptC3

CtgfDpep1Clec3bPcp4l1S1pr3Nrarp

Gucy1a3Notch3Pde3a

Rgs4Mgst3

Gja4SncgCnn1Actg2

Sh3bgrMyh11

Fbxo32Pcp4

Fbxl22Igfbp2Acta1

Feat

ures

Percent Expressed0255075

0

1

2

Average Expression

0.00

0.05

0.10

0.15

0.20

0.25

Anterior Dorsal Lateral Ventrallobe

fract

ion

Fibroblast 1

0.00

0.02

0.04

0.06

Anterior Dorsal Lateral Ventrallobe

fract

ion

Fibroblast 2

0.00

0.02

0.04

Anterior Dorsal Lateral Ventrallobe

fract

ion strain

BL6

FVB

Smooth Muscle

200 uM

50 uM

FVB

100 uM

25 uM

10x

40x

Anterior Dorsal Lateral Ventral

BL6

10x

40x

CCR2

0

2

Upstream Regulators (N = 836)

Activation_StateActivated

Inhibited

Fibroblast (1) vs Fibroblast (2)

Basa

lEn

dothe

lialFib

roblas

t 1Fib

roblas

t 2Im

mune

Lumi

nal A

nterio

r

Lumi

nal A

nterio

r_Dors

al

Lumi

nal D

orsal

Lumi

nal F

oxi1

Lumi

nal L

ateral

Lumi

nal P

sca

Lumi

nal V

entra

l 1

Lumi

nal V

entra

l 2

Peric

ytes

Smoo

th Mu

scle

Basal

Endothelial

Fibroblast 1

Fibroblast 2

Immune

Luminal Anterior

Luminal Anterior_Dorsal

Luminal Dorsal

Luminal Foxi1

Luminal Lateral

Luminal Psca

Luminal Ventral 1

Luminal Ventral 2

Pericytes

Smooth Muscle

Receptor Cluster

Liga

nd C

lust

er

0

10

20

30

# LR PairsParacrine

20

40

60

80

# LR PairsAutocrine

Autoc

rine

Paracrine

ITGA4

ITGB2

SLC40A1

CCR2

CD247

CC

R5

KLRD

1

CD36I

L2RAC

D69CD

3DCD3

GCCR

1C5AR

1C3AR1

CSF1R

FN1

SPON2

TGM2

CP

CCL7

B2M

CCL11

CC

L2 THBS

2

ICAM

1LG

ALS1

VCAM

1

C3

PLAT

THBS1

COL1A1

CSF1

COL1A2

0.91

0.58

scoreFibroblast 2

Immune

A B

C

D

E

F GA) UMAP of immune clusters and stromal clusters. Immune clusters are primarily comprised of macrophages with some T-cells and dendritic cells. B) Dot plot of di#erentially expressed genes for each stromal cluster. C) Scatter plot of $broblast 1, $broblast 2, and smooth muscle cells for each lobe show that each lobe (in both BL6 and FVB) is enriched for di#erent stromal cell types. D) H&E images of each lobe for each mouse. The histological di#erences in each lobe may in part be di#erences in the composition of stroma. E) IPA upstream regulator analysis indicating that CCR2 is upregulated in $broblast 2 cluster compared to $broblast 1. F) Matrix heatmap showing the number of interactions for each cluster. For $broblast 2, the highest number of interactions are autocrine, followed by paracrine interactions between $broblast 2 (ligand) and immune cells (receptor). G) Chord plot of the top paracrine ligand and receptor interactions between $broblast 2 (ligand) and immune cells (receptor).

VI. SUMMARY AND CONCLUSIONS

In these foundational single-cell studies of strain and lobe-speci$c di#erences in the mouse prostate, we have uncovered previously uncharacterized cell types and nominated unique molecular markers of multiple cell types for a more granular in situ examination of mouse prostate tissues.

Our scRNA-seq analysis of normal mouse prostates revealed that lobe and strain-speci$c di#erences are primarily restricted to epithelial cell types, with stromal cell types being largely conserved across strain and lobe. However, the composition of stromal cell types di#ers among lobes with a particular enrichment of $broblast-1 cells in the anterior lobe.

The combination of lobe-speci$c di#erences in luminal cells and the stromal composition in the prostate likely contribute to the histological di#erences observed between mouse prostate lobes. Additionally, the strain and lobe-speci$c di#erences in epithelial cells may also be a principal driver in the di#erences observed between mouse strains in the progression of prostate lesions in various prostate cancer models and why certain lobes of the prostate are predisposed to prostate cancer, such as the Hi-MYC and TRAMP models.

Overall, the $ndings of this study help establish the fundamental cell types residing in the normal mouse prostate of common mouse strains and serve as a reference to better understand how genetic alterations in transgenic mouse models are impacted by the normal biology of cells in the prostate.

Sources of Funding: Maryland Cigarette Restitution Fund, NIH/NCI U01CA196390, P30CA006973