University of Groningen Genomic medicine in inflammatory bowel disease Voskuil, Michiel

(1)

Genomic medicine in inflammatory bowel disease

Voskuil, Michiel

DOI:

10.33612/diss.136307453

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from

it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date:

2020

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Voskuil, M. (2020). Genomic medicine in inflammatory bowel disease. University of Groningen.

https://doi.org/10.33612/diss.136307453

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the

author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately

and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the

number of authors shown on this cover page is limited to 10 maximum.

(2)

(3)

BLOOD AND ILEAL T CELLS FROM

PATIENTS WITH CROHN’S DISEASE

REVEALS TISSUE-SPECIFIC

CHARACTERISTICS AND DRUG TARGETS

Werna. T.C. Uniken Venema

1,2*

_{, Michiel D. Voskuil}

1,2*

_{, Arnau Vich Vila}

1,2

_{, Gerben van der Vries}

2 _{, Bernadien}

H. Jansen

1 _{, Bana Jabri}

3 _{, Klaas Nico Faber}

1,4

_{, Gerard Dijkstra}

1 _{, Ramnik J. Xavier}

5,6

_{, Cisca Wijmenga}

2 _{, Daniel}

B. Graham

5,6

_{, Rinse K. Weersma}

1 _{, Eleonora A.M. Festen}

1,2

Gastroenterology 2019 (modifi ed version)

* Both authors contributed equally to this work

1 _{Department of Gastroenterology and Hepatology, University of Groningen, University Medical Center Groningen, Groningen, the}

Netherlands;

2 _{Department of Genetics, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands;}

3 _{Department of Medicine and Committee on Immunology, University of Chicago, Chicago, Illinois, United States of America;}

4 _Department

(4)

Chapter 3 Abstract

Background and aims

Approximately 200 risk loci for Crohn’s disease (CD) have been identified in genome-wide

association studies – many of these lie in genes that regulate T-cell signaling. We performed RNA

sequencing analyses of single cells (scRNAseq) from the intestinal mucosa and peripheral blood

of patients with CD to identify cells that express CD risk genes and potential drug targets

Methods

We performed flow-cytometry and scRNAseq of 5,292 CD3-positive T cells isolated from peripheral

blood (PBL) and ileal biopsies from 3 patients with mild to moderate CD. Biopsies were dissociated

and separated into intraepithelial T lymphocytes (IEL) and lamina propria T lymphocytes (LPL). We

integrated T-cell transcriptomes with data from genome-wide association studies on CD risk loci

and drug target identification resources.

Results

Unsupervised clustering of scRNAseq data identified 6 distinct types of T cells, with different

distributions among IEL, LPL and PBL. The majority of IEL were cytotoxic T cells (58%), most cells

within the PBL reservoir were quiescent T cells (42%), while LPL were mostly T-helper 17 (Th17)

cells (73%). Within the intestinal mucosa, Th17 cells, cytotoxic T cells and a cluster containing a

mix of T-regulatory and quiescent T cells were significantly enriched for expression of CD risk

genes. Within PBL, a cluster containing a mix of effector and T-regulatory cells were significantly

enriched for expression of CD risk genes. Potential targets for drug repositioning include PDE4D,

which was upregulated in mucosal Th17 cells (and could be targeted by apremilast); ALOX5AP,

which was upregulated in mucosal cytotoxic T cells (and could be targeted by fiboflapon), and

ITGB2, which was upregulated in peripheral blood cytotoxic T cells (and could be targeted by

lifitegrast).

Conclusion

In scRNAseq analyses of T cells from patients with CD, we identified populations of intestinal

mucosal cells that are enriched for expression of CD risk genes. These findings provide insights

in cell types and expression patterns and identify drugs that might be used to treat patients with

CD.

(5)

Introduction

Crohn’s disease (CD) is a chronic inflammatory disease of the gastrointestinal tract that affects

around 2 in 1,000 individuals in the western world

[1]

_{. Over the last decade, research into the}

genetic background of CD has identified more than 200 genetic loci associated with a risk of

developing CD

[2]

_{. Analysis of the genes underlying these risk loci points to a major role for T cells}

in CD pathogenesis. Several T cell subtypes – mainly the T helper subtypes Th1, Th2, and Th17

– have been implicated in the characteristic CD pattern of enhanced inflammation

[3,4]

_{. While}

the origins of these T cells might lie in the peripheral blood, they are most likely to encounter

antigens and mature in gut-associated lymphoid tissue. This renders them both gut-resident and

gut-specific and thus different from their peripheral blood counterparts. Studies in CD patients

have shown that these intestinal T cells manifest hyperreactive immune responses and are more

resistant to apoptosis than those of healthy individuals, which increases their potential to damage

the intestinal mucosa

[5]

_.

In view of the unique role intestinal mucosal T cells play in CD, it is surprising that therapeutics

used to treat CD are neither specific nor selective for the pathogenic cell type at the disease site

[6]

_.

While new anti-inflammatory therapies have been developed over the past decades, only 30%

of CD patients show a long-term drug response

[7]

_{. Efforts to develop more effective treatments}

for CD should ideally target its disease site, its local effector cells, and the corresponding disease

pathways. In CD, the primary disease site is generally the mucosal layer of the terminal ileum,

where the mucosal T cells are the main effector cells. To design specific treatments, it is crucial

to understand which subtype of intestinal mucosal T cells are specifically involved in which

disease pathways and where in the ileal mucosa these T-cell subtypes are located. Single cell RNA

sequencing (scRNAseq) provides a powerful tool to characterize the cells, pathways and genes

involved in this complex disease

[8]

_.

In this study, we characterize T cell transcriptomes from inflamed ileal mucosal and peripheral

blood T cells (peripheral blood T lymphocytes; PBL) from patients with CD. Because of the

functionally divergent properties of T cells from the different mucosal layers, and because of the

limited exchange between these layers

[9]

_{, we separated intraepithelial T cells (intraepithelial T}

lymphocytes; IEL) from lamina propria T cells (lamina propria T lymphocytes; LPL).

CD risk genes are often dysregulated in the intestinal mucosa of CD patients

[10]

_{, but it is not}

known through which cell types these genes contribute to disease. We show which T cells

express which CD risk genes, knowledge which sheds more light on how CD risk genes lead to

disease. In addition, since it has been shown that genetically supported drug targets lead to more

successful drug development

[11]

_{, we assess which current CD therapies and candidates for drug}

repositioning target which T cell populations based on their mRNA expression.

(6)

Methods

For details of antibodies, reagents, kits and software used see Supplementary Table 1.

Patient selection and sample collection

For this study, we recruited one male and two female patients with mild to moderate CD (ages

36, 25, and 48, respectively). Written informed consent was obtained and ethical approval was

granted by the University Medical Center Groningen Ethics Committee. All three patients had

been diagnosed with histologically proven ileal CD at least five years earlier. None had undergone

surgical resections. For each patient, five pinch biopsies were collected from mildly inflamed

terminal ileum, the most common site of inflammation in CD and a clear anatomical landmark.

Biopsies were collected into RPMI 1640 medium on ice and processed immediately. 10 mL of

peripheral blood was collected subsequently in tubes containing 158 USP units lithium-heparin.

Tissue processing

Ileal biopsies were dissociated into single cells following a previously published protocol,

separating IEL and LPL fractions using EDTA+DTT and collagenase digestion

[9]

_{. RNAseq-mediated}

RNA degradation was limited by adding 15μl RNAse Superase

TM

_{before collagenase digestion.}

PBL were isolated using density gradient centrifugation with Lymphoprep

TM

_{solution. Tissue}

processing, except for collagenase incubation, was performed on ice.

Flow cytometry of single cell suspensions

All cell suspensions (IEL, LPL, and PBL) were stained with the same antibody panel comprising

propidium iodide and monoclonal antibodies to TCRαβ, CD3, CD8a, CD8b, CD19, CD45RO, CD62L

and CD326.

Cells were sorted using the MoFlo Astrios (Beckman Coulter, Inc.), using forward- and side-scatter

signals to define lymphocyte fraction and exclude unwanted events such as debris, non-viable

cells and doublets. CD3+TCRαβ+ lymphocytes were sorted one-cell-per-well into 384-well plates

containing cell lysis buffer, on ice (see ‘Single cell RNA library preparation and sequencing’) (Figure

1; step 1). To minimize cell perturbation, sorting was performed at low pressures. After sorting,

each cell was collected in a well containing lysis buffer consisting of dNTP mix, Oligo dT primer,

Triton X-100 and RNase inhibitor.

Single cell RNA library preparation and sequencing

Single cell transcriptome libraries were created using a protocol based on Smart-Seq2 library

preparation using 3’-paired-end sequencing

[12]

_{. After a 3-minute incubation/ligation step at 72°C,}

a TSO-primer containing Unique Molecular Identifiers (UMI’s) was bound to the poly-A tail of RNA

transcripts, after which these were reverse transcribed using SmartScribe reverse transcriptase.

Next, the barcoded complementary DNA (cDNA) were amplified using BC-PCR primers. To

eliminate short fragments, cDNA products were purified with 0.8X ratio Agencourt Ampure XP

beads. Following purification, tagmented libraries were constructed with the Nextera XT DNA

Chapter 3

(7)

preparation kit. To allow multiplexed high-throughput sequencing, cells were pooled with N7xx

Nextera primer barcoding, enabling in silico sorting. Following amplification, products were

cleaned with 0.6X ratio Agencourt Ampure XP beads. To check for size distribution of the cDNA

product, the product was measured repeatedly on a PerkinElmer LabChip GX high-sensitivity

DNA chip. After final quality check of amplified and purified products, pools were combined into

superpools (8 barcoded pools/superpool–672 cells/superpool) and sequenced on the Illumina

NextSeq500 sequencer with 400M 75bp paired-end reads.

scRNAseq de-multiplexing pipeline

We obtained sequencing data for 5,292 single cells, with an average of ~400,000 reads per cell

(Figure 1; step 2). Sequencing data was processed through a de-multiplexing pipeline. Reads

with a Hamming distance >1 and an average quality score ≤10 were excluded. Trimmed fastQ

files were aligned to the human reference genome (Ensemble release 75, GRCh37) using STAR

(v. 2.5.1b)

[13]

_{with default settings. Approximately 75% of reads mapped uniquely. Before gene}

quantification, Picardtools (v. 2.2.2)

[14]

_{was used to sort aligned reads and flag individual cells and}

UMI’s. Gene level quantification was performed using Dropseq and filtering read count on unique

UMI’s, resulting in a gene/cell UMI count matrix.

Quality control

UMI counts were processed using Seurat

[15]

_{. To correct for cell-to-cell gene detection differences,}

we log-normalized UMI counts per cell and excluded genes expressed in <3 cells. To control for

low quality cells and doublets, we excluded cells expressing <200 genes per cell, cells expressing

>2500 genes per cell, and cells with >5% mitochondrial genes (Figure 1; step 2). We performed a

principal component analysis (PCA) and found that patient/batch, tissue-of-origin, and number of

UMIs were the major drivers among the first 6 PCs. For our analyses, we regressed out the number

of UMIs and percentage mitochondrial expression. Interpatient differences were assessed in

differential expression analyses. Since differential expression between different tissues-of-origin

may partly be driven by altered gene expression induced by collagenase digestion, we excluded

127 genes previously found to be influenced by collagenase digestion

[16]

_.

(8)

Figur

e 1. F

lo

w char

t illustr

ating e

xper

imen

tal pr

oc

ess

. ED

TA and collagenase r

ef

er t

o tr

eatment of the cells with these substances t

o isolat

e the intraepithelial and lamina

pr

opr

ia

cells

, r

espec

tiv

ely

. Charac

ter

ization

of

T cell

subt

ypes

as descr

ibed in

the

‘R

esults

’. T

he

numbers in

brack

ets

refl

ec

t the number

of epit

opic

CD8αβ-positiv

e

T cells

measur

ed b

y fl

o

w-cyt

ometr

y. IEL: intraepithelial

T lymphoc

yt

e. LPL: lamina pr

opr

ia

T lymphoc

yt

e. PBL: per

ipheral blood

T lymphoc

yt

e. St

eps 1-2-3 cor

respond with the t

ex

t

in the

‘M

ethods

’ sec

tion. F

ACS: fl

uor

escence

-ac

tivat

ed cell sor

ting

.

3 Crohn’

s disease patients

251 (28)

919 (248)

803 (167)

229 (57)

579 (347)

14 (1)

334 (77)

207 (50)

64 (35)

17 (6)

17 (5)

636 (210)

REG1A REG1B EPCAM

IL17A

IL21,IRF4 NR4A2,

CREM

C

TL

Q

uiesc

en

t

Eff

ec

tor/T

reg

Tr

eg/Q

uiesc

en

t

Th17

C

TL

REG1A/1B

Tr

eg/Q

uiesc

en

t

Th17

REG1A/1B

FACS for TCRαβ+

AND CD3+ cells

scRNAseq + quality control

Clustering + differential expression analysis

Tr

eg/Q

uiesc

en

t

C

TL

Tr

eg/Q

uiesc

en

t

TNFRSF4

FOXP3,CCR7 NOG,SELL

REG1A REG1B EPCAM

IL17A IL21,IRF4 NR4A2,

CREM

Tr

eg/Q

uiesc

en

t

Eff

ec

tor/T

reg

CD160 GZMA/B GNL

Y

CCL4/5

CCR7 IL6ST LEF1 TCF7

CD160

_TNFRSF4

ITGAE IL32

CCR7

TNFRSF4 LTB,MT

A2

CMTM6

EDT

A

Collagenase

CD160 GZMA/B GNL

Y

CCL5

CD160 GZMA/B GNL

Y

CCL5

TNFRSF4

FOXP3,CCR7 NOG,SELL

Figur

e 1:

F

lo

w char

t illustr

ating pr

oc

edur

e of

T-cell isola

tion and single

-c

ell RNA sequencing

total (

CD8

αβ+

)

Per

ipher

al blood

IEL

LPL

2,604 PBL

1,344 IEL

1,344 LPL

2,202 PBL

994 IEL

874 LPL

In

testinal muc

osal biopsies

St

ep 1

St

ep 2

St

ep 3

total 5,292

total 4,070

(9)

Integrating flow-cytometry and RNA expression data

Flow-cytometry data was analyzed using Kaluza software (Beckman Coulter, Inc.). Positivity for any

epitope was determined by visual inspection of plots with staining intensities. This adjudication

process was cross-checked by a flow-cytometry expert. We used Summit software (Beckman

Coulter, Inc.) to extract intensity data per cell into a CSV file format. Data on positivity for epitope

per cell was merged with metadata of the Seurat data file.

Clustering

Cells were clustered using a K-nearest-neighbor-based method

[15]

_{(Figure 1; step 3). Clustering}

was done with the PCs that were significant in a null-distribution determined through JackStraw

analysis. We varied clustering resolution between 0.6 and 1.2, and settled on 0.6 because higher

resolution produced clusters defined by marginal differential gene expression. Different subsets

(based on tissue-of-origin and the CD8αβ epitope) were clustered separately to check for

subset-specific cell types and/or states. t-SNE reduction was used for visualization purposes.

Differential expression analysis

Differential expression analysis was done using the MAST function in Seurat (Figure 1; step 3).

Differential expression was calculated between tissues-of-origin and cell types. To classify only

genes that are expressed at least moderately, we applied a filter of genes expressed in at least 1%

of the cells. After Bonferroni correction for multiple testing, results were considered significant

at an adjusted P-value < .05. Regressing out the effects of individual patient and sex did not

induce major differences in gene expression signatures: 94% of the T-cell-subtype-specific

gene expression signatures overlapped between analyses (Supplementary Table 2). To capture

subtle changes in expression, we used the non-regressed data for subsequent analyses. Both

upregulated and downregulated differential expression were calculated (Supplementary Tables

3A+B), however interpretation of the latter should be done with caution given the zero-inflated

nature of scRNAseq data.

Pathway analysis

Pathway analysis was performed with the ReactomePA package

[17]

_{in R, using the differentially}

expressed genes as input. A pathway was considered up- or downregulated at a

multiple-testing-corrected P-value < .05 (Supplementary Tables 4A-L).

Tissue specific T cell transcriptomes

We compared tissue- and disease-specific gene expression signatures to a previously published

micro-array-based transcriptomic survey of T cells from peripheral blood and the terminal ileum

[9]

_.

In that study, tissue was derived from healthy individuals and dissociated using the same protocol

we use (separating IEL and LPL). We compared the genes differentially expressed between tissues

in our dataset to genes showing > 1.4-fold change between tissues (n=92) in this previously

published dataset.

(10)

Validation of REG1A/1B expressing cells using immunofluorescence

We detected a small group of cells that represent a previously undefined cell population expressing

both CD3 and REG1A/1B. We used immunofluorescence on CD ileal biopsies to demonstrate their

presence.

CD risk genes

Through literature search, we identified 179 CD risk genes (Supplementary Table 5). From each

genome-wide association study (GWAS) risk locus, we selected the gene implicated through

coding variants or functional variants. If multiple genes were implicated within a locus, the genes

implicated through functional evidence were selected. Three loci remained in which two genes

were implicated, and all were included in further analysis. If no functional or coding data was

available, the most likely implicated gene (generally implicated indirectly through functional

studies) was selected. Since scRNAseq detects overexpression more robustly than underexpression,

and a previous study showed that CD risk genes dysregulated between CD intestinal mucosa and

healthy intestinal mucosa are generally upregulated, we focus on upregulated genes

[10]

_.

Drug target genes

We aligned IBD drug target genes extracted from OpenTargets

[18]

_{(n = 2,712) and Drugbank}

[19]

_{, to}

differentially expressed genes, and subsequently selected genes encoding proteins targeted by

drugs currently available for (clinical trials in) humans.

Comparison with scRNAseq dataset of T cells from healthy subjects

Due to limited publicly available data on human single cells, we could only compare peripheral

blood CTL to one existing healthy blood CTL dataset

[20]

_{. We randomly sampled 1,000 times 251}

healthy CTL, combining them with 251 CD CTL, using Seurat’s canonical correlation analysis.

We then calculated differential expression using 11,673 genes expressed in both datasets and

assessed the number of differentially expressed CD risk genes. Furthermore, we studied publicly

available scRNAseq data of epitopic CD4-positive naïve T cells for their mRNA expression of genes

typically associated with epitopic CD8αβ-positive T cells

[20]

_.

Enrichment for CD risk gene expression

We first tested whether CD risk genes are overrepresented in genes differentially expressed

between CD CTL and healthy CTL. Empirical P-values of overrepresentation of CD risk genes in

genes differentially expressed between CD and healthy CTL were derived from null-distributions

generated from 10

5 _{rounds of random sampling from all the genes overlapping between sets.}

We then assessed whether T cell subtypes showed enrichment for expression of CD risk genes.

We created null-distributions by randomly selecting sets the size of the differentially expressed

genes per cell type from the total set of 12,531 genes. Empirical P-values of overrepresentation of

CD risk genes within cell-type-specific differentially expressed gene sets were derived from the

null-distributions generated from 10

5 _{rounds of random sampling.}

(11)

Data and software availability

The scripts used can be found at https://github.com/WeersmaLabIBD/SingleCell. Summit

can be obtained from Beckman Coulter. The raw data is available at https://ega-archive.org

(EGAS00001002702).

Results

We analyzed 5,292 T cells from paired samples of peripheral blood (PBL) and ileal mucosal (IEL and

LPL) from three CD patients (Figure 1). After quality control, 4,070 T cells remained for analysis.

These cells expressed 966 genes per cell, on average, and 41,134 distinct genes in total (Figure 1;

step 2). Differential expression analyses were performed, and results were considered statistically

significant after adjustment for multiple testing using Bonferroni correction (significance

threshold of adjusted P-values < .05). t-SNE plots depicting data marked for the individual and sex

are shown in Supplementary Figure 1.

We first compared tissue populations. IEL, LPL, and PBL show markedly different expression

profiles, as visualized with t-SNE (Figure 2). PBL vs mucosal T cells differentially express 515 genes,

225 of which were overexpressed in mucosal T cells (Supplementary Tables 6A+B).

Figure 2. t-SNE plot representing T cell transcriptomes from different compartments. PBL: peripheral blood T

lymphocytes. IEL: intraepithelial T lymphocytes. LPL: lamina propria T lymphocytes.

−40

−20

0

20 −20

0

20

40 t-SNE 1

t-SNE 2

PBL

IEL

LPL

3

(12)

Comparing these results to T cell gene expression signatures from peripheral blood and the

terminal ileum of healthy individuals

[9]

_{, we observe an overlap of 49 out of 88 genes differentially}

expressed between PBL and mucosal T cells. Within mucosal T cells, IEL cells are characterized by

overexpression of 152 genes, including CCL5, CD160 and CXCR6, while LPL cells are characterized

by overexpression of 130 genes, including CD69, TNFAIP3 and PTGER4 (Supplementary Tables

7A+B).

Identification of T cell subtypes based on marker gene expression

Next, we investigated the presence of T cell subtypes in our dataset, which we defined based

on known genetic markers. Distributions of T cell subtypes in PBL versus mucosal T cells show

divergent patterns (Figure 3A+B).

PBL subtype characteristics: predominantly a quiescent profile

Within the PBL population, we find Cytotoxic T cells (CTL), Quiescent cells, T regulatory cells

on an effector memory background (Effector/Treg), and Treg cells on a quiescent background

(Treg/Quiescent) (Figure 3A, Supplementary Figure 2, Supplementary Table 8). We defined a

quiescent profile based on expression of CCR7 and an effector profile based on expression of

CD160 (Supplementary Figure 3A). PBL predominantly have a quiescent profile. Most of these

cells were characterized as Quiescent T cells based on expression of CCR7, IL6ST, SELL, NOG and

TCF7, representing a mix between central memory and naïve T cells (Supplementary Table 8,

Supplementary Figure 3B). Based on the CD45RO surface epitope, we found that PBL are mainly

memory, rather than naïve T cells (Supplementary Figure 3C). A second cluster of cells with a

quiescent profile express TNFRSF4 and were characterized as Treg/Quiescent cells. These Treg/

Quiescent T cells express the genes for LTB, which is used by Tregs to migrate across endothelial

cells. Treg/Quiescent T cells also express the gene for a transcription factor that regulates T helper

cell function (MTA2) and HNRNPH1, which reduces the ability of iTregs to suppress proliferation

of activated T cells

[21]

_{(Supplementary Figure 3D). Among PBL with an effector background, we}

identified a third cluster of cells, the Effector/Treg cluster, showing expression of TNFRSF4 and

ITGAE. These Effector/Treg T cells express the gene for IL32, which can induce both pro- and

anti-inflammatory cytokine production and FOXP3 expression.

These cells express genes for multiple transcription factors, many with a transcriptional repressor

function, including PRDM1

[22]

_{and CMTM6 (which encodes a regulator of PLD-1 expression)}

(Supplementary Figure 3E). Finally, the fourth blood T cell population, composed of effector

memory cells, was characterized as CTL based on high expression of marker genes such as

EOMES, PRF1 and TBX21. These CTL also express cytotoxic enzymes, such as GNLY, GZMH and

GZMB, and genes for cytokines with a strong chemoattractant function, such as CCL4 and CCL5

(Supplementary Table 8, Supplementary Figure 3F).

Chapter 3

(13)

Mucosal T cell subtype characteristics: predominantly an effector profile

The mucosal T cells could be divided into four T cell subtypes: T-helper 17 (Th17) cells, Treg/

Quiescent cells, CTL, and a group of REG1A/1B-expressing cells (Figure 3B, Supplementary

Figure 2, Supplementary Table 8).

In the intestinal mucosa, we identified a cluster of CTL based on expression of aforementioned

genes as well as XCL1, XCL2 and ITGA1

[23,24]

_{. Mucosal CTL are mainly derived from the IEL, with}

only a few cells deriving from the LPL (Figure 1, Supplementary Figure 4A). A cluster of mucosal

quiescent cells was characterized by expression of TNFRSF4, FOXP3, CCR7, IL6ST, NOG and SELL. We

defined these cells as Treg/Quiescent cells, which are mainly seen in LPL but also form a fraction

of IEL (Figure 1, Supplementary Figure 4B). We identified a large population of effector memory

cells, present mainly in LPL, characterized by expression of IL17A, IL22 and IRF4, and defined this

as Th17 cells. These cells also express genes for transcription factors typical for Th17 cell function.

Examples include NR4A2, which orchestrates Th17-mediated auto-immune inflammation through

IL21, and CREM, whose splice variant ICER supports Th17 differentiation

[25,26]

_{(Supplementary}

Figure 4C).

Finally, we identified a much smaller group of cells that could not be defined based on published

genetic markers. These cells express high levels of REG1A/1B, genes known to be expressed in

the exocrine pancreas, but also recently reported in metaplastic Paneth cells in IBD patients

[27]

_.

However, REG1A/1B cells show no upregulation of known Paneth cell marker genes

[28]

_{. This}

population is mainly present in the IEL, and shows co-expression of CD3E/G and EPCAM (Figure

1, Supplementary Figure 4D). Immunofluorescence staining of ileal mucosa of CD patients

confirmed the presence of these cells in the mucosa, expressing both CD3 and REG1A proteins

(Supplementary Figure 5).

(14)

Figure 3A. t-SNE plot representing transcriptomes of different T cell subtypes in peripheral blood. CTL: Cytotoxic

T lymphocytes. Quiescent: Quiescent lymphocytes. Effector/Treg: T regulatory lymphocytes on an effector memory

background. Treg/Quiescent: T regs on a quiescent background.

Figure 3B. t-SNE plot representing transcriptomes of different T cell subtypes in intestinal mucosa. CTL: Cytotoxic T

lymphocytes. REG1A/1B: cell highly expressing REG1A/1B. Th17: T-helper 17 cells. Treg/Quiescent: T regulatory lymphocytes

on a quiescent background.

−20

0

20 −20

−10

0

10

20 t-SNE 1

t-SNE 2

Peripheral blood

CTL

Quiescent

Effector/Treg

Treg/Quiescent

−30

−20

−10

0

10

20 −20

0

20 t-SNE 1

t-SNE 2

Intestinal mucosa

CTL

REG1A/1B

Th17

Treg/Quiescent

IEL

LPL

Chapter 3

(15)

Strikingly, all cell-subtype-clusters consisted of both epitopic CD8αβ-positive and - negative cells

(Supplementary Figures 3C+4E). All cell-type-clusters remained visible when separating cells

according to epitopic CD8αβ expression (Supplementary Figure 6A+B). Moreover, CD8A and

CD8B transcripts were expressed in both epitopic CD8αβ-positive and -negative cells. Notably,

healthy epitopic CD4-positive naïve T cells from a publicly available scRNAseq dataset also express

both CD4 and CD8A/B transcripts (Supplementary Figure 7 A-C). All cell types were present in all

three patients (Supplementary Table 9).

Pathways enriched in T cell subtypes

Pathway analysis of genes differentially expressed between T cell compartments revealed that

mucosal T cells are enriched for pathways indicative of active T cell migration, such as ‘cell surface

interactions at the vascular wall’ (Supplementary Table 10A). PBL are characterized by general

cell homeostasis functions such as ‘eukaryotic translation initiation’ and ‘peptide chain elongation’

(Supplementary Table 10B).

Within the intestinal mucosa, the IEL seem to actively attract immune cells and show an

enrichment of the ‘chemokine receptors bind chemokines’ and ‘immunomodulatory interaction

between lymphoid and non-lymphoid cells’ pathways (Supplementary Table 10C). LPL are

enriched for ‘interleukin-4 and 13 signaling’. Interleukin-4 is thought to induce proliferation,

cytokine production and promotion of T cell survival (Supplementary Table 10D).

Cell-type-specific pathway analysis revealed that genes differentially expressed in mucosal CTL

are enriched for pathways such as ‘PD-1 signaling’, ‘regulating T cell exhaustion’ and ‘chemokine

receptors bind chemokines’ (Supplementary Table 10E), indicating that these cells are involved

in immune cell migration and regulation of inflammation. In contrast, mucosal Th17 cells are

enriched for pathways such as ‘cellular responses to heat stress’ and the ‘HSF1 pathway’, indicating

active responses to cellular stress (Supplementary Table 10F). Peripheral blood CTL are involved

in ‘Neutrophil degranulation’ and ‘Platelet degranulation’, indicating a role in anti-microbial

inflammation (Supplementary Table 10L).

CD risk gene expression

Cells that express genes known to be associated to CD risk are most likely to play a role in CD

disease mechanisms

[4]

_{. Therefore, we hypothesized that genes upregulated in T cells from CD}

patients should be enriched for CD risk genes compared to T cells from healthy individuals. We

compared single-cell gene-expression signatures from peripheral blood CTL from CD patients

to publicly available peripheral blood CTL from a healthy donor (Supplementary Table 11).

Upregulated genes in healthy CTL contained ~1% CD risk genes, while upregulated genes in

CTL from CD patients contained ~2% CD risk genes, which suggests enrichment of CD risk genes

in CD. However, when performing permutations to correct for non-normal distribution, this

enrichment was not significant (P = .079).

(16)

Overrepresentation of CD risk gene expression in T cell compartments and subtypes

Second, we explored enrichment of CD risk gene expression within our dataset. Using

permutation analysis, we found that both IEL and LPL expressed significantly more CD risk genes

than expected by chance (P = .004 and P < 1.0 x 10

-5

_{, respectively), suggesting that cells within}

these compartments play a role in CD inflammation. PBL were not enriched for CD risk gene

expression (P = .46).

Identifying the T-cell-subtypes that overexpress CD risk genes can enhance our understanding of

how these cell types are relevant to disease. Th17 cells show the highest number of overexpressed

CD risk genes, and are most specifically enriched for CD risk gene expression (P < 1.0 x 10

-5

_).

Th17 cells show upregulation of CD risk genes such as CCL20, PTGER4, IFNG and TNFAIP3, which

are involved in chemotaxis, regulation of T cell inflammatory responses, and NOD2-directed

autophagy

[29]

_{. Mucosal CTL and Treg/Quiescent and peripheral blood Effector/Treg cells are}

also significantly enriched for CD risk gene expression (P = .028, P = .025, and P = 8.1 x 10

-4

_,

respectively). Mucosal CTL show upregulation of the CD risk genes PTPN22 and SOCS1, which are

regulators of immune signaling pathways

[30]

_{. Mucosal Treg/Quiescent cells show upregulation of}

CXCR5. Peripheral blood Effector/Treg show upregulation of, among others, PRDM1 and TNFRSF4.

Peripheral blood CTL showed only marginally significant enrichment for CD risk gene expression

(P = .050) (Supplementary Table 8). Mucosal REG1A/1B, peripheral blood Quiescent and Treg/

Quiescent cells did not differentially express CD risk genes.

Drug targets in CTL and Th17 cells

We next investigated which T-cell-subtypes are targeted by existing CD drugs or candidates

for drug repositioning towards CD. We focused on CTL and Th17 cells because of their

well-characterized expression signatures and their central role in CD pathogenesis

[31,32]

_{(Table 1). We}

identified a number of genes significantly overexpressed in these T-cell-subtypes that are targeted

by drugs under study or approved for use in humans. Th17 cells show upregulation of IL17A, the

gene product of which is targeted by Secukinumab

[33]

_{. ITGAE, a gene upregulated in mucosal}

CTL and encoding a subunit αE of integrin complex αEβ7, is a target for Etrolizumab

[34]

_{. S1PR5}

is upregulated in peripheral blood CTL, and is a known drug target for Ozanimod

[35]

_{. Potential}

targets for drug repositioning include PDE4D in mucosal Th17 cells, a target for Apremilast

[36]

_,

under investigation for UC treatment; ITGB2 in peripheral blood CTL, which is a target for

Lifitegrast

[37]

_{, approved for keratoconjunctivitis sicca; and ALOX5AP in mucosal CTL, which is a}

target for Fiboflapon

[38]

_{, under investigation for asthma (Table 1).}

Discussion

Using scRNAseq, we have characterized over 4,000 T cell transcriptomes from paired sets of ileal

and peripheral blood T cells from CD patients. A better understanding of which cell types make

up the functional fabric of different tissues will provide insight into the mechanisms underlying

disease. This insight will considerably improve the process of drug development

[11]

_{, and therefore}

major investments are being made to develop murine and human cell atlases

[39,40]

_{. However, these}

cell atlases are focused on healthy tissue, and location- and disease-specific scRNAseq datasets

Chapter 3

(17)

O

ver

expr

ession

of CD

risk genes

and genes

enc

oding

pot

en

tial

drug

tar

gets

in

ileal

muc

osal

Th17

cells

and

cyt

ot

oxic

T

lymphoc

yt

es

. L

egend:

Ileal

mucosal T

ipheral blood c

yt

ot

oxic

T lymphoc

yt

es (

C

TL) and ileal mucosal C

TL sho

w the highest number of sig

nificantly o

ver

expr

essed CD r

isk genes

. CD r

isk genes:

T

nificantly o

ver

expr

essed

CD r

isk genes

ar

e sho

wn in the first

ro

w of

this table

. P

er cell

type

, the t

op 5 most

sig

nificantly o

ver

expr

essed

kno

wn CD drug tar

tes

for

drug

repositioning

ar

e sho

wn. S

elec

tion of CD r

isk genes is explained in

‘M

ethods: CD r

isk genes

’. ‘K

no

wn

CD

drug

tar

get

’ r

ef

ers t

o genes encoding tar

rently appr

ov

ed or under in

vestigation f

or tr

eatment of CD in humans

. ‘C

andida

te f

or drug r

epositioning

’ r

ef

ers t

o genes encoding tar

gets f

or drugs cur

vestigation in humans f

or other diseases than CD

. (also see

‘M

ethods: Drug tar

get genes

’).

CD

: Cr

ohn

’s disease;

Th17:

T helper 17 cells;

C

TL

: c

yt

e; *e

.g

. A

dalimumab

, Etaner

cept, or Natalizumab; **e

.g

. Alefacept or Alemtuzumab; ***indir

ec

t tar

get.

M

uc

os

al

Th

17 c

ells

P

er

iphe

ra

l bl

oo

d

CTL

M

uc

os

al

CTL

G

en

e

Func

ti

on

G

en

e

Func

ti

on

G

en

e

Func

ti

on

k g

en

e

C

L2

0 C

hem

oa

ttr

ac

ta

nt

fo

r

va

ri

ou

s i

m

un

e c

ells

C

TSW

R

egu

lat

io

n

of

T

c

el

l c

yto

ly

ti

c

ac

tiv

it

y

PL

C

G

2 Tr

an

sm

em

br

an

e

si

gn

al

in

g

of

im

m

un

e

sy

st

em

re

ce

pt

ors

D

NA

JB

4 H

eat

s

hock

pr

ot

ei

n,

in

vo

lved

in

p

ro

te

in

fo

ld

in

g

LS

P1

A

dh

es

io

n

an

d

tr

an

s

en

do

th

el

ial

m

ig

ra

tio

n

PT

PN

22 N

egat

ive

regu

lat

or

of

TC

R

si

gna

ling

, p

os

it

iv

e r

eg

ula

to

r o

f

TL

R

s

ig

na

ling

IF

N

G

C

yt

oki

ne

, i

nv

olv

ed

in

ad

ap

ti

ve/

in

nat

e

im

m

un

it

y

PRK

CB

A

po

pt

osi

s

re

gu

la

ti

on

SO

C

S1

C

yt

oki

ne

-i

nd

uci

bl

e

negat

ive

re

gu

la

ti

on

o

f cyt

ok

in

e

si

gn

al

in

g

IR

F4

R

egu

lat

io

n

of

m

uco

sal

Th

17 cel

l d

iff

er

en

ti

at

io

n

PT

PR

C

T c

ell a

nt

ig

en

r

ec

ep

to

r s

ig

na

lin

g

re

gu

la

ti

on

M

A

P3K

8 T-he

lp

er

c

ell

di

ffe

re

nt

ia

ti

on

a

nd

IF

N

g

ex

pre

ssi

on

G

en

e

D

rug

o

r

co

m

po

und

G

en

e

D

rug

o

r

co

m

po

und

G

en

e

D

rug

o

r

co

m

po

und

C

D

dr

ug

IL

17 A

Secu

ki

nu

m

ab

33 S1P

R5

O

zan

im

od

35 C

D

3D

/E

/G

NI

-0

40

1

54 FCG

R3

A

Ig

G

c

las

s

m

A

bs

*

IT

GA

E

Etro

liz

um

ab

34 C

C

R9

V

erc

irn

on

55 TG

FBR

1 M

on

ger

sen

56 ***

da

te

fo

r

dr

ug

ti

on

in

g

D

NA

JB

1 A

pat

or

sen

45 C

C

L5

H

ep

ar

in

co

m

po

unds

51 , 5

2 C

C

L5

H

ep

ar

in

co

m

po

unds

51 , 5

2 SIK

1 D

ab

raf

an

ib

46 FG

R

D

as

at

in

ib

53 AD

RB1

ST

D

-10

1-D1

57 H

SP9

0A

A

1 N

ed

ocr

om

il

47 FCG

R3

A

Ig

G

c

las

s

m

A

bs

**

A

LO

X

5A

P

Fi

bo

fla

po

n

38 PT

G

ER

4 R

iv

en

pr

os

t; L

im

ap

ro

st

;

D

in

op

ro

sto

n

48 ,4

9,

50 IT

G

B2

Lifit

eg

ra

st

37 SL

A

M

F7

El

ot

uzu

m

ab

58 PD

E4

D

A

pr

em

ilas

t

36 AD

RB2

Be

ta

-2

ad

ren

er

gi

c

rec

ep

to

r

bl

ocker

s

MD

M4

A

LR

N

-69

24

59

3

(18)

study focused on improving understanding of the key role of T cells in CD pathology

[4,41]

_{. This}

will improve our understanding of disease in human subjects with CD and provide a reference

for future studies. Our study also provides insight into which cells are targeted by current CD

therapies and how drugs may be repositioned towards CD.

Our study is limited by several factors. The data is derived from three patients, which may reduce

the amount of variation covered. We found, however, that most cell-type-specific gene-expression

signatures remained after correcting for patient differences. We sampled intestinal biopsies from

mildly to moderately inflamed intestinal mucosa, which allowed us to successfully isolate sufficient

numbers of T cells from a largely intact mucosa. Unfortunately, fluorescent-labeled antibodies to

epitopic CD4 were not included in our antibody panel. Hence, we cannot reliably identify certain

T cell subtypes such as epitopic CD4/CD8-double positive cells. Finally, laboratory procedures

may cause cellular perturbation, which leads to alterations in gene expression. We minimized

experimental perturbations by directly processing our samples using extensively published

protocols and by excluding genes whose expression is known to be affected by collagenase

digestion.

We compared our peripheral blood and terminal ileum T cell data from CD patients to similar

T cell data from healthy individuals, and observed many similarities in tissue-specific

gene-expression signatures

[9]

_{. Although disease-status and technical differences may influence gene}

expression, the overlap in tissue-specific expression signatures indicates that these signatures are

strong enough to be present across diseases and techniques.

We show that T cell transcriptomes are compartment-specific. Our pathway analyses show that

mucosal T cells are enriched for T cell activation and trafficking, while PBL show upregulation of

more general cell homeostasis pathways. In our dataset, we detected six different T cell clusters:

mainly effector-type T cells in the intestinal mucosa and quiescent-type cells in peripheral blood.

We interpreted the gene expression signatures of each cluster with known T cell marker genes,

and determined the approximate T cell subtype composition of each cluster. Imposing strict

classical immune classifications based on scRNAseq data is difficult, since some well-known

T-cell-subtype cell markers (such as cytokines or chemokines) are hard to detect at transcriptome

level. ScRNAseq, however, does have the unique ability to identify previously unknown cell types

or cell-type-specific characteristics, since it enables hypothesis-free data collection and analysis.

We have identified a hitherto undescribed cluster of cells expressing both CD3EG and REG1A/

REG1B, and confirmed its presence in CD-affected mucosa from the terminal ileum.

While classifying our cells into T cell subtypes based on transcriptomes, we expected epitopic

CD8αβ-positive and -negative T cells to separate different T cell subtypes. Clustering cells based

on transcriptome similarities, however, primarily subdivides cells into their different effector

functions, rather than into classical epitopic CD8αβ-positive and -negative cell types.

(19)

We hypothesize that cell types enriched for CD risk gene expression are likely candidates to play

a role in CD pathogenesis

[4]

_{. Indeed, our CD peripheral blood CTL, which were not significantly}

enriched for CD risk gene expression, do show upregulation of a higher number of CD risk

genes compared to healthy peripheral blood CTL, although this trend was not significant after

permutation. Unfortunately, we are not aware of any publicly available scRNAseq datasets from

IEL or LPL from healthy individuals, even though these T cell compartments do show enrichment

for CD risk genes in our analyses.

Mucosal Th17, CTL, and Treg/Quiescent cells and peripheral blood Effector/Treg cells are enriched

for transcribed CD risk genes in our dataset. This emphasizes the cell-specific context in which CD

risk genes may contribute to pathogenesis, consistent with findings from recent efforts to map

potential causal CD genetic risk variants to specific cell types

[3]

_{. Studying which CD risk genes}

are expressed by which cell types can help our understanding of how these genes contribute

to disease

[4]

_{. For example, the upregulated CD risk genes PTPN22 and SOCS1 could exert their}

risk effect by stimulating cytotoxicity in mucosal CTL through the IFNα pathway, while CCL20,

PTGER4, IFNG and TNFAIP3, which were upregulated in Th17 cells, could exert their effect through

chemotaxis pathways.

We focused on upregulated genes in our analyses. Moreover, previous bulk RNAseq data has

shown that nearly all the CD risk genes dysregulated between CD intestinal mucosa and healthy

intestinal mucosa are upregulated in CD, suggesting that most CD risk genes exert their risk

through overexpression

[10]

_.

A problem in CD is that most current therapies are neither tissue-specific nor selective, and

that many potential new drugs fail in early stages of drug development. It has been established

that selecting genetically supported targets can double success rates of drug development by

prioritizing which drugs to advance to clinical development phases, and the expression of drug

targets are often sufficiently predictive to be used for this early clinical development

[11]

_{. We hope}

our results will contribute to a better understanding of the working mechanisms of currently

available drugs, e.g. Ozanimod targeting S1PR5 in peripheral blood CTL.

RNA expression of drug target genes in target cells is merely a prioritization strategy and not a

guarantee of success, as illustrated by IL17A, a well-studied drug target for Secukinumab that

is upregulated in mucosal Th17 cells. Secukinumab was, however, shown not to be effective in

IBD

[33]

_{. A possible explanation for this failure is that IL17A production is not limited to Th17 cells}

and that the role of IL17A is therefore not unambiguous

[42]

_{. Indeed, targeting the IL23/IL17 axis}

in other ways by Ustekinumab and Mirkizumab has shown clinical success after repositioning

towards IBD

[43,44]

_.

(20)

We saw a variation in T cell subtype composition between patients. We hypothesize that this

reflects true biology, at least in part, and that differences in immune cell composition between

patients contribute to the phenotypic variation between patients, possibly explaining why

current medical therapies only work for subsets of patients. Cell-type-specific expression data of

drug targets will therefore be useful for determining which patient subgroups are most likely to

benefit from specific drugs and which previously failed drugs might still have potential in certain

subgroups.

In conclusion, we have demonstrated that multiple ileal mucosal and one peripheral blood T cell

subtype from CD patients are enriched for CD risk gene expression. T cell subtypes known to be

involved in CD pathogenesis provide promising targets for future cell-type-specific therapies in

CD patients. Location- and disease-specific scRNAseq datasets are still scarce, and this dataset

can be a reference for future research. This kind of detailed characterization of cells is crucial

to furthering our understanding of molecular processes leading to health and disease, and to

identifying potential targets for drug development.

Acknowledgements

The authors thank the patients-participants of the 1000IBD cohort for contributing blood and

intestinal biopsy samples; Tjasso Blokzijl, Desiree Brandenbrug-Weening, Jelkje de Boer, Kim de

Lange, Tim Raine, and Pieter van der Vlies for laboratory support; Patricia Rogers, Wayel Abdulahad,

and Geert Mesander for technical assistance with fluorescence-activated cell sorting; Rudi

Alberts for data management support; the UMCG Genomics Coordination center, the UG Center

for Information Technology and their sponsors BBMRI-NL and TarGet for storage and computer

infrastructure; and Timothy Tickle and Mark Daly for contributing to the scientific discussion.

This article was edited for language and formatting by Kate Mc Intyre, Scientific Editor in the

Department of Genetics, University Medical Center Groningen.

Author contributions

WTCUV and MDV contributed equally to this work and are shared first authors. WTCUV and MDV

and EAMF participated in conception, design and coordination of the study. GD, RKW and EAMF

recruited patients. WTCUV, MDV and BHJ performed the laboratory experiments. WTCUV, MDV,

GvdV, AVV and EAMF performed the data analyses and data interpretation. BJ, DBG, GD and KNF

provided support with data interpretation. All authors assisted in the writing and reviewing of the

manuscript, and approved of sending it out for publication.

Conflicts of interest

WTCUV, MDV, AVV, GvdV, BHJ, BJ, KNF, GD, RJX, CW, DBG, RKW and EAMF have no (potential)

financial relationships with any organizations that might have an interest with the submitted

work and no other relationships or activities that could appear to have influenced the submitted

work.

(21)

Funding

CW is supported by a European Research Council (ERC) advanced grant (FP/2007-2013/ERC

grant 2012-322698), a Netherlands Organization for Scientific Research (NWO) Spinoza prize

(NWO SPI 92-266), the NWO Gravitation Netherlands Organ-on-Chip Initiative (024.003.001),

the Stiftelsen Kristian Gerhard Jebsen foundation (Norway) and a RuG investment agenda grant

Personalized Health. RKW is supported by an NWO VIDI grant (016.136.308). EAMF is supported

by a Dutch Digestive Foundation Career Development grant (CDG 14-04). WTCUV is supported

by the Foundation “De Drie Lichten”, The Netherlands, and the Boehringer Ingelheim Fonds. The

Department of Genetics, Section Research and Development, University of Groningen, University

Medical Center Groningen contributed to laboratory expenses for this study.

Supplementary data

Supplementary data to chapter 3 “Single cell RNA sequencing of blood and ileal T cells from patients

with Crohn’s disease reveals tissue-specific characteristics and drug targets” are available upon request

(m.d.voskuil@umcg.nl) and in part online (open access) at: https:/www.gastrojournal.org/article/

S0016-5085(18)35203-X/fulltext. Supplementary Figure 0, combining multiple supplementary

figures, can be found in this thesis.

Supplementary Figure 0. t-SNE plots representing T cells pooled from 3 individual patients with

CD.

Supplementary Figure 1. t-SNE plot representing transcriptomes of T cells pooled from three

individual patients.

Supplementary Figure 2. t-SNE plot representing transcriptomes of different T cell subtypes from

both peripheral blood and intestinal mucosa.

Supplementary Figure 3. t-SNE plots featuring expression of cellular marker genes in peripheral

blood lymphocytes.

Supplementary Figure 4. t-SNE plots featuring expression of cellular marker genes in peripheral

blood lymphocytes.

Supplementary Figure 5. Representative immunofluorescence staining of intestinal mucosa

from patients with Crohn’s disease.

Supplementary Figure 6. t-SNE plot representing transcriptomes of different T cell subtypes from

both peripheral blood and intestinal mucosa.

Supplementary Figure 7. Gene expression of CD8A, CD8B and CD4 transcripts.

Supplementary Table 1. Specification for materials and methods (available online).

Supplementary Table 2. Overlap in T cell expression signature without and with regression for

individual patients and gender.

Supplementary Table 3. Downregulated genes per T cell compartment and T cell subset.

Supplementary Table 4. Downregulated pathways per T cell compartment and T cell subset.

Supplementary Table 5. Crohn’s disease risk genes.

Supplementary Table 6. T cell gene expression signatures specific to peripheral blood and

intestinal mucosa.

(22)

Supplementary Table 8. T cell expression signatures specific to T cell subsets (available online).

Supplementary Table 9. Distribution of different T cell subtypes per patient.

Supplementary Table 10. Pathway enrichment per T cell compartment and T cell subset.

Supplementary Table 11. Differentially expressed genes between CD peripheral blood CTL and