Celiac disease

Celiac disease

Zorro Manrique, Maria Magdalena

DOI:

10.33612/diss.122712049

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):

Zorro Manrique, M. M. (2020). Celiac disease: From genetic variation to molecular culprits. University of Groningen. https://doi.org/10.33612/diss.122712049

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.

from genetic variation to molecular culprits

Maria Magdalena Zorro Manrique

Celiac disease: from genetics variation to molecular culprits.

ISBN: 978-94-034-2558-0 (e-book) ISBN: 978-94-034-2559-7 (print)

Copyright © 2020, María Magdalena Zorro Manrique

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means - electronic, mechanically, by

photocopying, recording or otherwise - without express written permission form the author and, when appropriate, the publisher holding the copyrights of the published articles.

Cover and layout design by Claudia González Arévalo claudia@tangroop.com

Printed by ProefschriftMaken

https://www.proefschriftmaken.nl

from genetic variation to molecular culprits

PhD thesis

to obtain the degree of PhD at the University of Groningen

on the authority of the

Rector Magnificus Prof. C. Wijmenga and in accordance with

the decision by the College of Deans.

This thesis will be defended in public on Wednesday 22 April 2020 at 11.00 hours

by

Maria Magdalena Zorro Manrique

born on 4 September 1979

in Puerto Wilches, Colombia

Prof. C. Wijmenga Dr. S. Withoff

Co-supervisor Dr. I.H. Jonkers

Assessment Committee

Prof. A. van den Berg

Prof. J.D. Laman

Prof. F. Koning

Kieu T.T. Le

Joram Mooiweer

Table of contents

Chapter 2 The role of celiac disease associated protein LPP in

intestinal barrier function. 21

Chapter 3 Systematic prioritization of candidate genes in disease loci identifies TRAFD1 as a master regulator of IFNg signalling in celiac disease.

43

Chapter 4 Tissue alarmins and adaptive cytokine induce dynamic and distinct transcriptional responses in tissue-resident intraepithelial cytotoxic T lymphocytes.

77

Chapter 5 Small and long regulatory RNAs in the immune system

and immune diseases. 107

Chapter 6 Knockdown of the lncRNA RP11-291B21.2 interferes

with the activation of CD8

T cells. 143

Chapter 7 Discussion. 171

Appendices ^{Summary. 189}

Sammenvatting 193

Resumen. 197

Acknowledgements. 201

Curriculum Vitae. 204

List of publications. 205

CHAPTER 1

General introduction

Chapter 1

General introduction

Celiac disease (CeD) is a common immune-mediated disorder triggered by intake of grain- derived gluten proteins that affects 1-2% of individuals in the western world. Although the precise cause of CeD is unknown, multiple environmental and genetic factors have been found to contribute to the development of this complex disease

.

In most CeD patients gluten ingestion triggers a strong immune response that provokes the activation of gluten-specific CD4

T cells, production of anti-transglutaminase 2 (TG2) antibodies by B cells and lymphocyte infiltration in the small intestine, and all these processes contribute to the villous atrophy and crypt hyperplasia characteristic of the disease. As a result, CeD patients have a variable degree of small intestinal inflammation and a broad range of manifestations that include diarrhea, abdominal pain and malabsorption

. In addition to the classical gastrointestinal symptoms, CeD patients can also present extra-intestinal manifestations such as anemia

, osteoporosis

, dermatitis herpetiformis

and neurological disorders

. Because of this wide range of symptoms, it is thought that many cases of CeD go undiagnosed

.

Currently, a lifelong gluten-free diet (GFD) is the only available treatment for CeD. Avoidance of gluten contributes to the recovery of the intestinal mucosa, the reduction of levels of anti-TG2 antibodies and eventually to reduction of symptoms

. In spite of its benefits, a GFD can be challenging to maintain due to multiple factors, including social restrictions, nutrient deficiencies and high costs

. In addition, a small sub-group of patients (1-5%) fail to respond to a GFD, and these individuals are at risk of developing a severe condition called refractory celiac disease (RCD) that is characterized by a remarkable infiltration of intraepithelial cytotoxic T cells (IE-CTLs) with abnormal phenotype. In RCD these intestinal abnormalities persist and can eventually contribute to the development of enteropathy- associated T cell lymphoma

.

Although gluten exposure is the most important environmental factor in CeD pathogenesis,

recent evidence implicates other environmental factors in disease development. Intestinal

viral infections and bacterial microbiota have been linked to CeD as possible environmental

triggers

. In a recent study Bouziat et al. suggested that reovirus infection induces

a proinflammatory response with a concomitant loss of oral tolerance to gluten

, while

the involvement of the bacterial microbiome has been suggested by studies reporting gut

microbiome dysbiosis in CeD patients as compared to healthy individuals

. Interestingly,

these CeD-associated changes in microbiota composition have been shown to affect the

processing of gluten peptides

, which may affect gluten presentation to gluten-specific

CD4

T cells and thereby increasing the inflammatory response. Thus, in addition to

gluten, environmental factors such as the gut microbiome and virome may contribute to the

environmental component of the risk of developing CeD, although their respective contributions

to CeD development are still unclear

. In contrast, genetic risk factors have been estimated to contribute approximately 50% of CeD risk, making them the major predisposing factors currently known for CeD

.

Genetic risk factors

The strongest genetic factor associated to CeD risk is the human leukocyte antigen (HLA) region. More than 90% of CeD patients carry either the HLA-DQ2 (HLA-DQ2.5) or the HLA- DQ8 allele, and these alleles appear to account for up to 40% of the genetic risk of developing the disease

. However, although the absence of these alleles in individuals means they will not get the disease, their presence alone cannot predict who will develop CeD because these alleles are present in approximately 30-40% of the general population. This suggests that, while the HLA-DQ2 and -DQ8 alleles are necessary for CeD development, additional genetic factors are required

. To date, genome-wide association studies (GWAS) have identified 42 non-HLA genetic variants to be associated with CeD (Fig. 1)

. Due to the modest effect of each non-HLA variant on overall disease risk, these together account for approximately 15% of heritability

. Interestingly, most of these genetic variants are also shared with other

Figure 1. Manhattan plot showing the results of association for 39 of 42 non-HLA CeD risk loci. Known loci (black), novel loci (blue) and risk loci with multiple signals (underlined) are depicted. The vertical line represents the genome-wide significant threshold (p value 5x10^-8). Adapted from Trynka, G. et a, 2012²⁰.

23 2221 2019 18 17 1615 14 13 12 11 10 9

8 7 6

5 4

3 2 1

0 10 20

–log 10 P

30 40 50

Chromosome

HCFC1,TMEM187, IRAK1 UBE2L3, YDJC ICOSLG UBASH3A PTPN2 SOCS1, PRM1,PRM2 CIITA and others CLK3 and others ZFP36L1 SH2B3, ATXN2 ETS1TREH, DDX6 POU2AF1 and other ZMIZ1 PFKFB3, PRKCQ PVT1ELMO1 TAGAP OLIG3, TNFAIP3 PTPRK BACH2 IRF4

KIAA1109, ADAD1, IL2, IL21 LPP

SCHIP1, IL12A ARHGAP31 CCR1-3, LTF CCR4, GLB1 CD28, CTLA4, ICOS STAT4 UBE2E3,ITGA4 IL18R1, IL18RAP PLEK, FBX048 PUS10 C10rf106 RGS1 FASLG,TNFSF18 RUNX3

C1orf93, MMEL1, TTC34

Chapter 1

immune-mediated disorders such as rheumatoid arthritis

(RA) and type I diabetes

(T1D), indicating the presence of a common etiological component in immune-mediated complex diseases.

Although GWAS have been very successful in associating genomic loci with disease, there is a need for follow-up studies aimed at pinpointing causal genetic variants (single nucleotide polymorphisms, SNPs) and genes in these loci

. There are still some limitations that prevent successful identification of causal SNPs

. First, due to linkage disequilibrium, many adjacent SNPs are co-inherited and are likely to have a similar association or correlation, which complicates prioritisation

. Second, in only four of the CeD-associated loci – MMEL1, SH2B3, NCF2 and IRAK – do the SNPs affect protein encoding regions

. Most of the risk SNPs associated with CeD (and complex immune-mediated diseases in general) fall in non- coding regions of the genome, including intergenic regions, and their consequence is not understood

. In complex diseases it has consistently been found that these variants are enriched in regulatory domains controlling gene expression, including promoter and enhancer elements characterized by regions of open chromatin and specific histone modifications

. This indicates that, rather than altering protein function, risk SNPs associated with immune diseases control the expression of genes encoding proteins or non-coding RNAs.

To overcome the difficulties discussed above, complementary methods are applied to move from SNP associations to the downstream consequences on gene expression and the regulation of biological pathways. Zooming in on GWAS loci via fine-mapping, for instance, can identify smaller regions that encompass smaller groups of variants with the highest probability of causality

. The functional impact of fine-mapped SNPs can then be tested by assessing the correlation between expression and genotype (quantitative trait locus (eQTL) analysis)

; the interaction between GWAS SNPs and other genomic regions (chromatin interaction conformation assays (3C,4C))

; the enrichment-overlap with functional elements such as enhancers and promoters (using data generated by the ENCODE

or Epigenome roadmap projects)

and the potential alterations of transcription factor binding sites

. These approaches have confirmed that CeD-associated SNPs affect gene expression rather than change the amino acid sequence of proteins

. Additionally, application of computational approaches such as gene set enrichment analysis and gene network analysis to genes differentially expressed in CeD can elucidate the biological pathways and tissues where these risk genes play a role in disease pathophysiology

, and the prioritized set of candidate genes that results can then be further validated by in vitro and in vivo assays

.

From SNPs to disease mechanisms

It has been hypothesized that the intestinal barrier is compromised in CeD

, thereby facilitating

the passage of gluten peptides into the lamina propria, where they are deaminated by tissue

TG2. This deamination process strongly increases the binding affinity of gluten peptides to

HLA-DQ2 or -DQ8 molecules on the surface of antigen presenting cells (APCs) such as

dendritic cells, B cells and macrophages

. The gluten peptides presented in the context of HLA-DQ2 or -DQ8 are recognized by gluten-specific CD4

T cells, leading to activation of the inflammatory response characteristic of CeD. Upon activation, CD4

T cells release cytokines such as IFNg and IL-21

. IL-21 may provide inflammatory signals to B cells and intraepithelial cytotoxic T cells (IE-CTLs). B cell activation results in the secretion of antibodies against gluten and TG2

. However, the clinical implications and pathogenic role of these antibodies are still unclear. IFNg promotes the Th-1 immune response, activation of APCs and licensing of IE-CTLs to kill intestinal epithelial cells (Fig. 2)

.

A recently performed genetic study integrated different layers of genomic information, including eQTL analysis, cell-type-specific enhancer enrichment, functional annotation of GWAS SNPs and co-expression analyses

. In addition to confirming what was already known about the involvement of the immune system in CeD, this study provided genetic evidence for

Figure 2. Main players in CeD pathogenesis. A compromised intestinal barrier in predisposed individuals allows the passage of gluten peptides into the lamina propria, where the gluten peptides are deaminated by TG2, which enhances their affinity for HLA-DQ2 or HLA-DQ8 molecules on the surface of APC (dendritic cells, B cells, macrophages). Presen- tation to gluten-specific CD4⁺ T cells results in activation and proliferation of these cells, which then release IFNg and IL-21. These cytokines provide signals that enhance the cytolytic properties of IE-CTLs and promote the differentiation of B cells towards plasma cells that produce anti-gluten and anti-TG2 antibodies. Some of the environmental factors (gluten, microbiome, infections) that can influence the disease onset are indicated. Cytokines/inflammatory molecules expressed by intestinal epithelial cells are depicted (IL-15, IFN-1), as are intestinal epithelial cells (IEC), intraepithelial cytotoxic lymphocytes (IE-CTLs), antigen presenting cells (APC).

Microbiome Gluten

IL-15 TG2 IFN-1

IE-CTL IEC

IL-21 IFNg

Anti-gluten Anti-TG2

HLA DQ2/8

IL-21 IFNg

?

Gluten specific

CD4

T Cell

APC

B cell

Chapter 1

the involvement of the adaptive immune system via the IFNg signaling pathway (although the IFN locus itself has not been associated with CeD) and a role for B cells in CeD. Moreover, the same study prioritized several genes (LPP, C1orf106) in CeD-associated loci that might contribute to decreased intestinal barrier function

. Disturbance of intestinal permeability could not only facilitate the passage of gluten into the lamina propria but also that of infectious agents. This would boost the presentation of gluten peptides to CD4

T cells, resulting in the release of pro-inflammatory signals, and these processes would contribute to a stronger immune response

. Although this evidence suggests that barrier dysfunction can contribute to CeD onset

, it is still unclear whether this is a primary defect and a cause that contributes directly to disease onset or of it is a consequence of the inflammatory environment in the gut of CeD individuals

.

In addition to dysregulation of the adaptive immune system and adaptive cytokines, innate cytokines such as IL-15 and IFN type I (IFN-1) are upregulated in intestinal epithelial cells of CeD patients

. These cytokines enhance the cytolytic and proinflammatory properties of CTLs and dendritic cells

, respectively. Simultaneously, an induction of stress-induced non-classical major histocompatibility complex class I molecules is observed on the surface of epithelial cells, which are recognized by natural killer receptors expressed on IE-CTLs.

This interaction licenses IE-CTLs to kill intestinal epithelial cells, thus contributing to villous atrophy

. These findings demonstrate the involvement of both adaptive cytokines (IFNg and IL-21) and tissue-derived cytokines (IL-15 and IFN-1) in activation of CTLs. To date, little is known about the signaling processes elicited in IE-CTLs by these cytokines

.

In the work described in this thesis, we have studied how CeD-associated genetic variation can be translated to molecular culprits (genes, pathways and relevant cell types). One of the most exciting discoveries of genomics research using next generation sequencing methodologies has been the existence of a novel class of genes that are transcribed but not translated: long non-coding RNAs (lncRNAs). Our group found genetic evidence that, in loci associated with

Figure 3. Overview of the cell types and methods used in this thesis. A) Human-derived primary and immortalized cell lines and culture systems (2D and 3D). B) Tools to target genes and mimic inflammation. C) Lab approaches. D) Data analysis strategies, in silico tools, datasets and population cohorts employed to conduct this study

Cell types and culture

systems Conventional and novel

lab methods In silico

Methods and cohorts Gene and environmental

perturbations IELs

BiopsiesGut

T cells 2D/3D Cultures

Monocytes Caco-2 PBMCs

Cell lines

Biopsy- derived intestinal cells

Anti CD3/CD28 Stimulation

Cytokine stimulation Rescue

SiRNA shRNA

WB qPCR

Olink

FACSsorting

Single cell RNA-

seq

Chip-seq

Bioinformatic analysis Co-expression

Analysis

Data clustering

Available datasets

Population Cohorts

ACGTA CGTAA GCTTA

RNA-seqBulk

complex diseases including CeD, lncRNAs are affected by disease-associated SNPs

. Thus, it is not only non-coding regulatory elements, but also lncRNAs, that may contribute to the deregulation caused by CeD-associated variants.

In the research leading to this thesis I made use of in silico approaches in combination with in vitro experiments in human-derived cells exposed to conditions that resemble the inflammatory environment in the intestine of CeD patients in order to prioritize genes and lncRNAs potentially involved in CeD and to further understand their role in disease onset and pathogenesis (Fig. 3).

Aim and outline of the thesis

The aim of this thesis is to gain further insights into the function of some of the CeD candidate genes and the molecular pathways and mechanisms that play a role in CeD. The candidate genes have been selected by in silico and omics approaches in cell types known to play a role in CeD pathophysiology.

As was described above, it has been hypothesized that CeD patients have an inherent barrier function defect that could facilitate gluten transport into the lamina propria. Previously, LPP has been identified as one of multiple CeD-associated genes that may be involved in cell-cell interaction and intestinal barrier homeostasis. In CeD biopsies, LPP expression is reduced when compared to normal biopsies. In Chapter 2, we examined whether a reduction in LPP does lead to decreased barrier homeostasis using the Caco-2 cell line, which is widely used in barrier function and pharmacology research, as an in vitro model. We generated a stable LPP knockdown cell line of the parental Caco-2 cell line and evaluated the effect on proliferative capacity, permeability and lumen formation in 2D and 3D culture environments. Moreover, we evaluated the transcriptional response in this cell line under standard culture conditions and upon challenge with IFNg, a cytokine known to be involved in the pathogenesis of CeD.

During the course of my thesis research, population cohort studies became available that

could be used for eQTL analysis. Additionally, novel statistical genetics approaches were

developed in our lab or published by others that could be applied to the data generated in

the cohort studies to prioritize culprit SNPs and genes in disease-associated loci. In Chapter

3 we applied a systematic approach to integrate eQTL data from the BIOS cohort (total RNA

transcriptomics from whole blood of 4000 participants from general population cohorts)

with CeD association data derived from the most recent CeD GWAS meta-analysis

. We

applied four different in silico approaches (LD-overlap, Bayesian co-localization, Mendelian

randomization and DEPICT) to prioritize potential causal genes, resulting in the identification

of 126 positional and functional candidate genes. Co-expression and pathway analysis were

applied to prioritize the main cell types and biological pathways in which these genes are

most likely to play a role. TRAFD1, one of the prioritized genes, was selected for functional

follow-up.

Chapter 1

The pro-inflammatory response elicited by gluten in CeD patients ultimately converges on IE-CTLs that are consequently licensed to kill epithelial cells. Although several key CeD- associated cytokines are known to affect IE-CTLs, little is known about the transcriptional programs triggered by these cytokines. In Chapter 4 we generated TCRab

CD8

cytotoxic T cell lines from human small intestinal epithelium to study the dynamics of the transcriptional response and of genome-wide H3K27 acetylation (H3K27ac) changes in response to stimulation with tissue-derived cytokines, also known as alarmins (IFNb and IL-15) or a T cell–derived cytokine (IL-21). These three cytokines have not only been associated with tissue destruction in CeD, but also with other autoimmune disorders with tissue-specificities such as RA, inflammatory bowel disease (IBD) and T1D. The data we generated was analyzed in depth to describe the biological pathways that are triggered in IE-CTLs in response to tissue- derived cytokines versus T cell–derived cytokines. We further studied the relation between gene expression and epigenomic (H3K27ac) profiles to get a better understanding of the potential mechanism of gene regulation. Finally, we tested the potential enrichment of genes responding to cytokine stimulation in risk loci associated with autoimmune diseases (GWAS data) to identify genes that might contribute to immune deregulation in IE-CTLs.

The next generation sequencing and genomics revolution has led to the discovery of novel classes of genes and novel insights into disease biology. A significant portion of the SNPs associated with complex immune-mediated diseases have been shown to overlap with DNA motifs that control the expression or binding sites of micro-RNAs (miRNAs) or to intersect gene regulatory motifs of lncRNAs. Chapter 5 is an overview of the general features of these two classes of non-coding RNAs in terms of synthesis, structure and the potential mechanism by which they modulate gene expression or other processes in the cell. This review focusses on the role of these genes in different cells of the adaptive and the innate immune system, and on their involvement in immune mediated disorders such as CeD, IBD and multiple sclerosis. Chapter 6 focuses on a single lncRNA, lncRNA RP11-291B21.2, that is strongly modulated in response to TCR activation in CD8

T cells, including IE-CTLs. We describe the expression pattern of this lncRNA in different CD8

T cell populations derived from blood and infer its potential biological function using single cell RNA-seq data and co-expression network analysis. Finally, knockdown experiments performed in IE-CTL cell lines and RNA sequencing data were interrogated to provide additional clues for pinpointing this lncRNA’s function in IE-CTLs, which are the effector cell type in CeD.

Chapter 7 summarizes the major findings of this thesis research project and sets them in

a broader perspective by discussing their implications in the context of CeD and immune-

mediated disorders in general. The main drawbacks and limitations of our experimental

approaches are described, as are directions and suggestions for how to dig deeper into the

biological contributions of the risk genes in the pathogenesis of complex diseases.

References

1. Lionetti, E. & Catassi, C. The role of environ- mental factors in the development of celiac disease: what is new?. Dis. 3, 282–293 (2015).

2. Caja, S., Mäki, M., Kaukinen, K. & Lindfors, K.

Antibodies in celiac disease: Implications Be- yond Diagnostics. Cell Mol Immunol. 8, 103-9 (2011).

3. Freeman, H. J. Iron deficiency anemia in celiac disease. World J. Gastroenterol. 21, 9233–8 (2015).

4. Mustalahti, K., Collin, P., Sievänen, H., Salmi, J. & Mäki, M. Osteopenia in patients with clini- cally silent coeliac disease warrants screening.

Lancet. 354, 744-5 (1999).

5. Fry, L. Dermatitis Herpetiformis. Baillieres.

Clin. Gastroenterol. 9 (2), 371–93 (1995).

6. Hadjivassiliou, M. et al. Gluten sensitivity: from gut to brain. Lancet Neurol. 9, 318-30 (2010).

7. Fasano, A. et al. Prevalence of celiac disease in at-risk and not-at-risk groups in the united states: a large multicenter study. Arch. Intern.

Med. 163, 286-92 (2003).

8. Makharia, G. K. Current and emerging therapy for celiac disease. Front. Med. 1-6 (2014).

9. Leffler, D. A. et al. Factors that influence adher- ence to a gluten-free diet in adults with celiac disease. Dig. Dis. Sci. 53, 1573–81 (2008).

10. Rubio-Tapia, A. & Murray, J. A. Classification and management of refractory coeliac disease.

Gut 59, 547–57 (2010).

11. Sánchez, E. et al. Influence of environmental and genetic factors linked to celiac disease risk on infant gut colonization by bacteroides species. Appl. Environ. Microbiol. 77, 5316–23 (2011).

12. Verdu, E. F., Galipeau, H. J. & Jabri, B. Novel players in coeliac disease pathogenesis: role of the gut microbiota. Nat. Rev. Gastroenterol.

Hepatol. 12, 497 (2015).

13. Bouziat, R. et al. Reovirus infection triggers in- flammatory responses to dietary antigens and development of celiac disease. Science. 356, 44–50 (2017).

14. Sánchez, E., Donat, E., Ribes-Koninckx, C., Fernández-Murga, M. L. & Sanz, Y. Duode- nal-mucosal bacteria associated with celiac disease in children. Appl. Environ. Microbiol.

79, 5472–5479 (2013).

15. Wacklin, P. et al. The duodenal microbiota composition of adult celiac disease patients is associated with the clinical manifestation of the disease. Inflamm. Bowel Dis. 19, 934–941 (2013).

16. Caminero, A. et al. Duodenal bacteria from pa- tients with celiac disease and healthy subjects distinctly affect gluten breakdown and immu- nogenicity. Gastroenterology 151, 670–683 (2016).

17. Withoff, S., Li, Y., Jonkers, I. & Wijmenga, C.

Understanding celiac disease by genomics.

Trends Genet. 31, 295-308 (2016).

18. Liu, E., Rewers, M. & Eisenbarth, G. S. Genet- ic testing: who should do the testing and what is the role of genetic testing in the setting of celiac disease?. Gastroenterology 128, 33-7 (2005).

19. Jabri, B. & Sollid, L. M. T cells in celiac dis- ease. J. Immunol. 198, 3005-3014 (2017).

20. Trynka, g. Et al. Dense genotyping identifies and localizes multiple common and rare vari- ant association signals in celiac disease. Nat.

Genet. 43, 1193–201 (2011).

21. Zhernakova, A. et al. Meta-analysis of ge- nome-wide association studies in celiac dis- ease and rheumatoid arthritis identifies four- teen non-HLA shared loci. Plos Genet. 2, e1002004 (2011).

22. Walker, N. M. et al. Shared and distinct genetic variants in type 1 diabetes and celiac disease.

N. Engl. J. Med. 359, 2767-77 (2008).

23. Gallagher, M. D. & Chen-Plotkin, A. S. The post-GWAS era: from association to function.

Am. J. Hum. Genet. 102, 717–730 (2018).

24. Edwards, S. L., Beesley, J., French, J. D. &

Dunning, M. Beyond GWAS: Illuminating the dark road from association to function. Am J Hum Genet. 93, 779-97 (2013).

25. Bonvouloir, N., Lemieux, N., Crine, P., Boileau, G. & Desgroseillers, L. Molecular cloning, tis- sue distribution, and chromosomal localization of MMEL2, a gene coding for a novel human member of the neutral endopeptidase-24.11 family. DNA Cell Biol. 20, 493–498 (2001).

26. Mori, T. et al. Lnk/SH2B3 controls the produc- tion and function of dendritic cells and regu- lates the induction of IFN-y–producing T cells.

J. Immunol. 193, 1728–1736 (2014).

27. Singh, A. et al. The IRAK-ERK-p67phox-nox-2 axis mediates TLR4, 2-induced ROS produc- tion for IL-1β transcription and processing in monocytes. Cell. Mol. Immunol. 13, 745–763 (2016).

28. Kumar, V. et al. Human disease-associat- ed genetic variation impacts large intergenic non-coding RNA expression. Plos Genet. 9, e1003201 (2013).

29. Zhang, F. & Lupski, J. R. Non-coding genetic variants in human disease. Hum. Mol. Genet.

24, R102–R110 (2015).

Chapter 1

30. Maurano, M. T. et al. Systematic localization of common disease-associated variation in regu- latory DNA. Science. 337, 1190–1195 (2012).

31. Nicolae, D. L.et al. Trait-associated SNPs are more likely to be eQTLs: annotation to en- hance discovery from GWAS. Plos Genet. 6, e1000888 (2010).

32. Dekker, J., Rippe, K., Dekker, M. & Kleckner, N. Capturing chromosome conformation. Sci- ence 295, 1306–11 (2002).

33. Consortium, T. E. P. An integrated encyclope- dia of DNA elements in the human genome.

Nature 489, 57–74 (2012).

34. Himmelstein, D. S. & Baranzini, S. E. Hetero- geneous network edge prediction: a data inte- gration approach to prioritize disease-associ- ated genes. PLOS Comput. Biol. 11, e1004259 (2015).

35. Van Elburg, R. M., Uil, J. J., Mulder, C. J. J.

& Heymans, H. S. A. Intestinal permeability in patients with coeliac disease and relatives of patients with coeliac disease. Gut. 3, 354-7 (1993).

36. Molberg, Ø. et al. Tissue transglutaminase se- lectively modifies gliadin peptides that are rec- ognized by gut-derived t cells in celiac disease.

Nat. Med. 4, 713-7 (1998).

37. Chladkova, B. et al. Gliadin fragments promote migration of dendritic cells. J. Cell. Mol. Med.

15, 938–48 (2011).

38. Bodd, M. et al. HLA-DQ2-restricted gluten-re- active T cells produce IL-21 but not IL-17 or IL-22. Mucosal Immunol. 3, 594–601 (2010).

39. Nilsen, E. M. et al. gluten specific, HLA-DQ restricted t cells from coeliac mucosa produce cytokines with Th1 or Th0 profile dominated by Interferon gamma. Gut. 37, 766-76 (1995).

40. Lindfors, K. et al. Coeliac Disease. Nat. Rev.

Dis. Prim. 5, 3 (2019).

41. Kumar, V. et al. Systematic annotation of celiac disease loci refines pathological pathways and suggests a genetic explanation for increased Interferon-gamma levels. Hum. Mol. Genet.

24, 397–409 (2015).

42. Schumann, M., Siegmund, B., Schulzke, J. D.

& Fromm, M. Celiac disease: role of the epi- thelial barrier. Cell Mol Gatroenterol Hepatol.

3, 150-162 (2017).

43. Mention, J. J. et al. Interleukin 15: a key to dis- rupted intraepithelial lymphocyte homeostasis and lymphomagenesis in celiac disease. Gas- troenterology. 125, 730-45 (2003).

44. Monteleone, G. et al. Role of interferon alpha in promoting T helper cell type 1 responses in the small intestine in coeliac disease. Gut. 48, 425–9 (2001).

45. Mattei, F., Schiavoni, G., Belardelli, F. & Tough, D. F. IL-15 is expressed by dendritic cells in re- sponse to type I IFN, double-stranded RNA, or lipopolysaccharide and promotes dendritic cell activation. J. Immunol. 167, 1179-87 (2001).

46. Jabri, B. & Sollid, L. M. Tissue-mediated con- trol of immunopathology in coeliac disease.

Nat. Rev. Immunol. 9, 858-70 (2009).

47. Jabri, B. & Abadie, V. IL-15 functions as a dan- ger signal to regulate tissue-resident T cells and tissue destruction. Nat Rev Immunol. 15, 771-83 (2015).

48. Ricaño-Ponce, I. et al. Refined mapping of au- toimmune disease associated genetic variants with gene expression suggests an important role for non-coding RNAs. J. Autoimmun. 68, 62–74 (2016).

49. Zhernakova, D. V. et al. Identification of con- text-dependent expression quantitative trait loci in whole blood. Nat. Genet. 49, 139-145 (2017).

50. Ricaño-Ponce, I. et al. Immunochip meta-anal-

ysis in european and argentinian populations

identifies two novel genetic loci associated with

celiac disease. Eur. J. Hum. Genet. (2019).

3Depts. of Gastroenterology, Infectious Diseases and Rheumatology, Charité–Universitätsmedizin, Berlin, Germany.

4Institute of anatomy, University of Jena, Jena, Germany.

CHAPTER 2

The role of celiac disease associated protein LPP in intestinal barrier function.

Maria Zorro

, Vinod Kumar

, Lucja M. Jarosz

, Luz Maria Medrano

, Joram Mooi- weer

, Michael Schumann

, Jan F. Richter

, Cisca Wijmenga

, Sven C.D. van IJzen- doorn

, Iris Jonkers

, Sebo Withoff

.

In progress

Chapter 2

Abstract

Intestinal epithelial barrier function is critical for homeostasis of digestive and immunological processes. Deregulation of the epithelial barrier has been associated with several diseases, in particular celiac disease (CeD). Although the majority of genes associated with CeD are related to immune function, genome-wide association studies, eQTL mapping and pathway analyses have also implicated a number of genes that play a role in cell–cell interaction.

We hypothesized that these “non-immune” CeD genes exert their effect by deregulation of intestinal barrier homeostasis and therefore investigated the potential role of LPP, one of the “non-immune” genes most strongly associated to CeD, in barrier function. Using Caco- 2 cells as a model of the intestinal barrier, we evaluated the functional impact of reduced LPP expression. Cells with LPP knockdown (KD) display a reduction in proliferation and their capacity to form lumen-like structures in 3D culture assays. Microscopic analysis revealed increased permeability at focal sites in transwell assays in KD but not in control cell lines.

Finally, RNA-seq analysis revealed that exposure to Interferon-g leads to downregulation of metabolic pathways and a concomitant upregulation in immune genes in the KD cell line, suggesting that LPP might also play a role in the immune response elicited by epithelial cells.

The downregulation of LPP observed in intestinal biopsies of CeD patients supports our

hypothesis that LPP downregulation contributes to the deregulation of epithelial cell function

in CeD both at barrier- and immune-response-level.

Introduction

Celiac disease (CeD) is a chronic autoimmune disorder triggered by ingestion of gluten in genetically susceptible individuals. In CeD, the oral intake of gluten causes an abnormal activation of both the innate and adaptive immune response that results in cell damage and villus atrophy in the small intestine

. Although the HLA-DQ2 and -DQ8 alleles are the most important genetic risk factors, genome-wide association studies (GWAS) have uncovered an additional 42 loci outside the HLA-locus that also contribute to CeD susceptibility

. Many of the associated loci encompass genes that regulate T and B cell biology, as well as genes involved in the IFNg signaling pathway. This is not surprising given that deregulation of the immune system and IFNg signaling are involved in disease onset

. More recently, however, eQTL mapping and pathway and co-expression analyses of GWAS loci have also prioritized genes involved in cell–cell interaction and intestinal barrier function as potential causal genes contributing to CeD susceptibility. We hypothesized that these non-immune genes, exemplified by LPP and C1orf106, contribute to CeD pathophysiology by disrupting cell–cell interaction, leading to increased permeability, also called ‘leaky gut’, one of the hallmarks of CeD

. Barrier function may therefore already be impaired in CeD individuals before development of CeD

, which is in line with a role for genetic factors in barrier function.

It has been shown by others that CeD biopsies exhibit alterations in the expression of tight junction proteins

, actin cytoskeleton disturbances

and increased intestinal epithelial cell apoptosis

. It is currently unclear if these alterations cause CeD or are a consequence of disease-associated inflammation. Interestingly, one of the strongest non-HLA CeD-associated genetic variants (odds ratio ~1.26)

mapped to a 2.8 kb region in intron 2 of the LPP gene (chr3: 188117070–188119901, NCBI build 37)

. Although the function of LPP remains poorly characterized, transcriptional analysis of intestinal biopsies showed that its expression is downregulated in CeD patients compared to healthy individuals

. A recent study demonstrated that LPP co-localizes with E-cadherin in adherent junctions and focal adhesions. Other researchers have described alterations in intracellular distribution of E-cadherin in cell junctions upon LPP knockdown

, and LPP knockdown led to disturbances in the intracellular distribution pattern of LPP and focal adhesion proteins after exposure to gliadin peptides, the degradation products of dietary gluten

. These data are in line with our in silico predictions indicating a putative role for LPP in cell–cell interaction and support the hypothesis that downregulation of LPP expression contributes to enhanced intestinal permeability in CeD patients.

Here, we aim to show that reduced expression of LPP leads to decreased barrier homeostasis in Caco-2 cells, a well characterized in vitro model for studying intestinal barrier function.

We generated Caco-2 cells that stably expressed a short hairpin RNA (shRNA) sequence

targeting LPP mRNA. We then assessed the effects of LPP knockdown on the proliferative

Chapter 2

capacity of the cells, on the permeability of Caco-2 monolayers in 2D transwell experiments, and on lumen formation in 3D spheroid cultures. Furthermore, we evaluated the transcriptional response after a challenge with IFNg. What we found was that reduced LPP expression led to decreased cell proliferation and polarization. Moreover, LPP knockdown increased the local permeability and the response to inflammatory signals. Our results provide new insights into a potential role for LPP in both epithelial integrity and in immune function.

Materials and Methods

Cell culture. A human Caco-2 cell line (colonic epithelial cancer cells; ATCC), a gift from Dr. Sven van IJzendoorn from the Cell Biology department of the University Medical Center Groningen, was cultured at 37

C in maintenance media containing MEM medium (GIBCO), 10% fetal bovine serum (Sigma-Aldrich), 1% non-essential amino acids (Invitrogen) and 100 U/ml penicillin-streptomycin (GIBCO). For passaging, cells were treated for 5 minutes with 1X trypsin (Invitrogen) and passaged twice per week.

Caco-2 LPP knockdown, scrambled and LPP rescue cell lines. LPP Knockdown cell lines (KD) were developed by transfecting pRS vectors containing four different shRNA that targeted LPP (see details in Supplementary Fig. 1) using the Amaxa cell nucleofector Kit T for Caco-2 (Lonza), according to the manufacturer’s instructions. This generated four different sublines. Untransfected Caco-2 cells (wild type (WT)) or cells transfected with a non-targeting scrambled shRNA (SCR) were generated as controls. To generate stable cell lines, transfected cells were selected by culturing for 2 months in increased concentrations of puromycin (GIBCO), up to a final concentration of 10 mg/ml. Knockdown efficiency was determined by RT-qPCR and western blot (every ~3-4 weeks). An LPP-rescue cell line was created by expressing an LPP cDNA sequence engineered to be resistant to the LPP shRNA.

In brief, shRNA-resistant cDNA was created by introducing nine silent point mutations in the LPP shRNA targeting region of LPP cDNA using the QuikChange II site-directed mutagenesis kit (Stratagene). The recombinant LPP cDNA was cloned into the pLenti-C-Myc-DDK-IRES- Neo expression vector (ORIGENE) and transfected into the LPP KD cell line with the strongest reduction in LPP (KD cell line containing the shRNA #2) using the Amaxa cell nucleofector Kit T (Lonza), according to the instructions of the supplier. Selection was started 48 hours (hrs) post transfection by adding increasing concentrations of G418 sulfate (Geneticin, GIBCO) to the culture medium, up to a final concentration of 400 mg/ml. The level of LPP was tested by real time (RT)-qPCR and western blot. The rescue cell line was passaged twice a week in maintenance medium with puromycin (10 mg/ml) and Geneticin (400 mg/ml).

RT-qPCR. Total RNA was extracted using the mirVana™ miRNA isolation kit (Ambion)

according to the manufacturer’s protocol. cDNA was prepared by RevertAid H Minus First Strand

cDNA Synthesis Kit (Thermo Scientific). We used primers for LPP, HPRT, MX1 and IL15R

as follows: LPP 5’ CCAACAATGTCTCACCCATC, LPP 3’ ACTACCGGGGCAAACTTTTT,

MX1 5’ ATCCAGCCACCATTCCAAGG, MX1 3’ TTTGCGATGTCCACTTCGGA, IL15R 5’ AGAGAGCCTCTCCCCTTCTG, IL15R 3’ TTGAAGGCATCAGCTGGGAG. Reverse transcription was performed using the Fast-Real Time PCR system 7900 ht (Applied Biosystems) using the SYBR green mix (Bio-Rad). Data were analyzed with the SDS software package V2.3 (Applied Biosystems). Expression values were normalized to the endogenous control (HPRT) to calculate the relative LPP, MX1 or IL15R expression level using the 2

method. Relative LPP values were then converted to percentages by setting the SCR as a reference (100%).

Western blot. Caco-2 cells were homogenized in lysis buffer (PBS containing 2% SDS and 1X complete protease inhibitor cocktail (Roche)). Protein concentration was determined with the bicinchoninic acid BCA Protein Assay Kit (Thermo Scientific). 20 mg of protein was loaded on SDS-polyacrylamide electrophoresis gel and transferred to nitrocellulose membrane (Bio- Rad). Blots were probed overnight with mouse anti-LPP antibody (1:1,000, clone 8B3A11, ABCAM) or for 1 hr with mouse anti-actin antibody (1:5,000, clone C4M, MP Biomedicals), followed by incubation with goat anti-mouse horseradish-peroxidase–conjugated secondary antibodies (1:10,000, Jackson Immuno Research). Bands were visualized using Lumi-Light Western Blotting Substrate (Roche). Images were captured by the Chemidoc MP imaging system (Bio-Rad). The signal intensity of each band of interest was quantified using Image Lab™ software (Bio-Rad). Specific intensities were normalized using the actin signal and expressed as a percentage by setting the SCR cell line as reference (100%).

3D cultures. Cells were suspended in DMEM supplemented with 2% v/v matrigel (BD Biosciences) and seeded at a density of 6000 cells/well onto matrigel pre-coated 8-well Lab- Tek Chambered coverglasses (Nunc). After 1 week, bright field microscopy pictures were taken with the Evos FL cell imaging system (Thermo Fisher Scientific). The number and size of spheroids and the presence of a lumen in the center of each spheroid was quantified manually using the ImageJ program (NIH). At least 100 spheroids were counted per condition.

To further analyze spheroid morphology, their nuclei and actin were stained (see details in the section below).

Immunofluorescence microscopy. Cells were fixed in 4% paraformaldehyde at room

temperature for 20 or 30 minutes (for 2D or 3D cultures, respectively), and permeabilized

with 0.2% Triton-X100 (Sigma-Aldrich) for 10 or 20 minutes at 37

C (for 2D or 3D cultures,

respectively). Nonspecific background was blocked with 3% FBS at 37

C. Cells were incubated

with primary mouse anti-LPP (1:250, clone 8B3A11, Abcam) or E-cadherin (1:100, clone

5H6L18, Invitrogen) at 37

C for 1.5 hrs. The cells were washed with PBS and incubated with

secondary Alexa-Fluor488 or cy3-conjugated goat anti-mouse or anti-rabbit antibodies (1:250,

Jackson ImmunoResearch Laboratories). Actin was labeled with TRITC-phalloidin (10mg/ml,

Sigma-Aldrich) and the nuclei with DAPI (1mg/ml, SIGMA) at 37

C for 30 min. Images were

acquired with a Leica TCS SP8 confocal microscope (Leica) and analyzed using Leica LAS

Chapter 2

Cell proliferation. Caco-2 cells were plated at 250,000 cells/ml in 25 cm

flasks (Greiner) and grown for 72 hrs. Next, cells were trypsinized and quantified microscopically using trypan blue (GIBCO) staining.

Transepithelial resistance. Caco-2 cells were seeded in 0.4 mm polycarbonate transwells in 12-well plates (Corning) at a density of 50,000 cells/well and maintained for 21 days.

The medium was replaced 2-3 times per week. Transepithelial resistance (TEER) values were measured twice per week with an EVOM2 Ohm meter (World Precision Instruments), according to the manufacturer’s instructions.

Paracellular permeability assay. The permeability across Caco-2 monolayers was assessed by measuring the flux of fluorescein isothiocyanate–labeled dextran (FD4, Sigma- Aldrich) from the apical to the basal compartment of the transwell. After 21 days, the medium in the basal compartment was replaced by Hank’s balanced salt solution (HBSS) to mimic physiological or “normal conditions”, or by 2.5 mM EDTA in calcium- and magnesium-free PBS (“stress conditions”). 500ml of FD4 solution (250 mg/ml) were added in each apical compartment. Aliquots were taken from the basal compartment at 30 min intervals over a 2.5-hr period at 37

C in sterile conditions. The amount of FD4 in the basal compartment was determined in a BioTEK Synergy HT plate reader (Biotek) using a calibration curve. The apparent permeability coefficient (P

) was calculated by the following formula: P

= dQ/

dt×1/A×Co, where P

= apparent permeability in cm/s, dQ/dt = the rate of appearance of FD4 in the basal compartment in μg/s, Co = the initial FD4 concentration in the apical compartment in μg/ml and A = the surface area of the insert in cm

.

Statistical Analysis. GraphPad Prism V7.0 (GraphPad software, San Diego, CA) was used to assess statistical differences across groups. Detailed information (type of statistical tests applied, p values and number of replicates) is provided in the text of the figure legends.

Stimulation experiments. Caco-2 cells were cultured in 0.4 mm polycarbonate Transwells in 12-well plates (Corning) at a density of 75,000 cells/well. The medium was replaced 2-3 times per week. After 2 weeks, the cells were treated in the basal compartment with 60 ng/ml of IFNg (PeproTech) for 3, 8 or 40 hrs. Untreated cells were included as controls. Following incubation, cells were gently washed, collected in cell lysis buffer from the mirVana™ miRNA isolation kit (AMBION) and stored at -80

C for further use. To pre-assess the responsiveness of Caco-2 cells to IFNg stimulation, relative levels of MX1 and IL15R mRNA were measured by RT-qPCR.

RNA sequencing. RNA from unstimulated and IFNg-stimulated cells was isolated with the

mirVana™ miRNA isolation kit (Ambion), following the manufacturer’s protocol. Sample

concentration and integrity was measured using the Nanodrop 1000 Spectrophotometer

(Thermo Scientific) and the High-sensitivity RNA analysis kit (Bio-Rad EXPERION). The

libraries were prepared with the Truseq RNA preparation kit (Illumina), using the manufacturer’s protocol. The pooled libraries were sequenced on a HiSeq2500 instrument (Illumina), using default parameters (single read, 1x50bp). The trimmed fastQ files where aligned to build 37 of the human reference genome using Hisat (version 0.5.1)

, allowing for two mismatches and sorted using SAMtools V1.2

. Gene quantification was performed by HTSeq/0.6.1p1

using mode=union, stranded=no. The Ensembl V75 database was used for gene annotation.

Differentially expressed genes (DEGs; those with |log2 FC>1| and adjusted p value ≤ 0.01) were extracted by the DESeq2 R package

. Principal component analysis (PCA) plots and heatmaps were generated by R base functions and the pheatmap and ggplot2 packages.

Results

LPP knockdown impairs proliferation and lumen formation in Caco-2 cells

Given the potential role of LPP in cell–cell adhesion, we wanted to know if LPP downregulation interferes with cell proliferation and cell polarity. Therefore, we knocked down LPP in Caco-2 cells using shRNA molecules. We tested four different shRNAs against different exons of LPP (Supplementary Fig. 1) and observed that the cells transfected with shRNA #2 presented the most efficient and consistent knockdown (Supplementary Fig. 2). We therefore selected this cell line for further characterization and coined it KD. Our Caco-2 KD cell line displayed a significant decrease of LPP at mRNA-level (~75%) and protein-level (~90%) compared to the controls (WT or SCR, Fig. 1A-C). Furthermore, a significant reduction (~50%) of proliferative capacity was observed in KD compared to controls (Fig. 1D). Of note, the treatment with the non-targeting scrambled shRNA did not affect LPP expression or cell proliferation, indicating that LPP remains at physiological levels and that the transfection conditions were not toxic to the cells.

To determine the potential effect of LPP knockdown on cell polarization, we grew the cells for 1 week in the presence Matrigel, a commercial hydrogel containing cellular matrix proteins, to induce spontaneous formation of 3D spheroids with a central lumen

. We noted that all three cell lines (WT, SCR and KD) developed spheroids similar in numbers (~10 spheroids/

microscopic field, Fig. 1E) and size (~100 mM, Fig. 1F). However, the percentage of spheroids with a hollow central lumen was significantly lower in KD compared to controls (~15% vs

~50%; Fig. 1G). To exclude the possibility that reduced proliferation of the KD cells simply caused a delay in the lumen formation, rather than an impairment of the process, we followed up several of the 3D cultures for 2 weeks but did not observe any changes (data not shown).

This suggests that the impairment of lumen formation is independent of proliferative capacity.

Many types of epithelial cells are polarized, which means that they present three domains:

an apical domain facing the lumen, a basal domain underlying the epithelial cells from the

connective tissue and a lateral domain between the apical and basal domains

. Microscopic

analysis of the spheroids indicated that most control spheroids consisted of a single layer

of cells lining a central lumen. The WT and SCR spheroids were characterized by nuclei

Chapter 2

positioned towards the basal side of the spheroid and a thick layer of actin at the luminal side (Fig. 1H, I). In contrast, the position of the nuclei and the deposition of actin fibers appeared to be random in most of the KD spheroids (Fig. 1J), indicating defects in apico-basal cell polarity in the KD cells leading to failure of spheroid formation.

LPP knockdown does not affect the barrier function as measured by TEER and FD4 flux We evaluated TEER, a measurement of the electrical resistance across a cell monolayer that reflects tight junction integrity

, and observed a time-dependent increase in TEER in the control cell lines, with values after 3 weeks significantly higher in the SCR cell line compared to WT. In contrast, the KD cells displayed minimal TEER increase during the first 2 weeks, followed by a strong increase after 3 weeks (Fig. 2A). However, this increase was not statistically different between SCR and KD. This dramatic change in TEER in the KD cell line could be caused by a delay in monolayer formation during the first 2 weeks. Surprisingly, we performed nucleic staining after 1, 2 and 3 weeks and observed confluent monolayer formation in all cases at all time points (data not shown), indicating that the lower TEER in

Figure 1. LPP knockdown impairs proliferation and lumen formation in Caco-2 cells. Caco-2 controls (WT, SCR) or LPP knockdown (KD) cell lines were grown under standard conditions (A-D) or in the presence of Matrigel (E-J). After 3 days, LPP expression was analyzed by western blot (A-B) or qPCR (C). Cell proliferation was assessed by trypan blue exclusion (D). Spheroid characteristics: number (E), size (F), percentage of hollow lumens (G). Confocal images of representative 3D cultures (H-J) showing nuclei (DAPI) and actin distribution (phalloidin). Scale bar represents 100 mM. Data are expressed as a percentage by taking the SCR control cell line as reference (100%). Bar plots show mean

± SEM. Graphs are representative of at least three independent experiments. p values were obtained using a one-tailed Student t test, **** p<0.0001.

the KD cell cultures is not caused by a decrease in proliferation or incomplete confluency.

It is important to note that the large variability in the replicate data, particularly in the KD cells, that is illustrated by the high coefficient of variation (CV) (Fig. 2B) makes it difficult to conclusively state that TEER is affected by LPP knockdown. In contrast, the variability decreased gradually to acceptable levels (CV<20%)

in the controls. Together this implies that additional experiments are needed to assess tight junction integrity more accurately.

Next, we evaluated the paracellular permeability in the cell cultures by measuring the flux of FD4 from the apical compartment to the basal compartment of the transwell system.

We could not detect FD4 in the basal compartment after 2.5 hrs under normal conditions (HBSS) in any of the Caco-2 cell lines (data not shown), implying that there were no defects in paracellular permeability under “physiological conditions”. Conversely, the flux of FD4 towards the basal compartment increased from 0 to ~1x10

cm/seg after treatment with EDTA (Fig.

2C), which is known to disrupt tight junctions

. We noted that the permeability is significantly higher in the KD and WT compared to the SCR cell line. However, due to the differences in permeability among the controls, it is difficult to conclude that there is a substantial increase in the permeability in the KD cells under the conditions that we tested. Finally, even though the permeability assay we performed has been extensively used in industry and research to provide an estimation of global permeability

, this strategy does not consider the presence of focal places where the tight junctions and the permeability are defective. We therefore performed a pilot experiment using a novel method called the sandwich

that assesses the

Figure 2. LPP knockdown does not affect barrier function of Caco-2 cell monolayers. Caco-2 cells were expanded for 21 days in transwell supports. TEER values were measured twice per week (A). The coefficients of variation of TEER measurements demon- strate the variability of the assay (B). Permeability (apparent per- meability coefficient (P_app)) of Caco-2 monolayers under stress conditions (EDTA) (C). Plots show the mean ± SD (if applicable).

Data represent one of at least three individual experiments.

Chapter 2

permeability in focal sites by immobilizing labeled molecules close to their site of passage.

Here we observed more locations with local disruptions of permeability in the KD cells compared to controls after exposure to EDTA (Supplementary Fig. 3). Although this finding implies a marked susceptibility of the KD cell line to alterations in both tight junction formation and permeability in response to environmental stress, further replications should be done to confirm this result.

Non-physiological overexpression of LPP in the KD cell line does not fully reconstitute Caco-2 cell proliferation and does not rescue lumen formation.

To confirm that the impaired proliferation and polarization we observe in Caco-2 cells is due to LPP reduction, we generated a rescue cell line by expressing LPP cDNA in the KD subline insensitive to the shRNA. As expected, we observed a significant increase in LPP expression in the rescue cell line compared to the KD subline (Fig. 3A, B). Furthermore, we noted a moderate but significant increase in the proliferation rate of the rescue cell line compared to the KD cell line (Fig. 3C), indicating that LPP rescue at least partially restores cell proliferation.

Figure 3. Non-physiological overexpression of LPP in the KD cell line does not fully reconstitute Caco-2 cell pro- liferation and does not rescue lumen formation. Caco-2 WT, SCR, KD or rescue cell lines were grown under standard culture conditions. After 3 days, LPP expression was analyzed by western blot (A) or qPCR (B). Cell proliferation was quantified by trypan blue exclusion (C). Representative confocal images of 3D (D, F, G) or 2D (E) cultures were taken to assess the percentage of hollow lumens (D), the cell morphology and the distribution of LPP (E-F), actin or E-cadherin (G). Bar plots are depicted as percentage by setting the SCR control as reference (100%). Bar plots show mean ± SEM.

Data is representative of at least three independent experiments. ***p < 0.001, **** p <0.0001; one-tailed Student’s t-test.

One possible explanation for the incomplete restoration of proliferative capacity in the KD upon reintroduction of LPP could be that expression of LPP rescue cDNA is variable in the KD cell line. Indeed, immune-staining of LPP showed that LPP levels were very heterogeneous and that some cells presented much higher levels than control cells (Fig. 3E, F). Thus, the high expression of LPP in a subpopulation of cells might explain the overall observation of a two-fold increase in LPP mRNA and protein levels (Fig. 3A, B) in the rescue cell line. Lastly, we assessed the polarization capacity of the rescue cell line by staining rescue spheroids for actin and E-cadherin. This experiment showed that the rescue cells, like the KD cells, still display an altered distribution of actin and E-cadherin throughout the cells (Fig. 3G). Taken together, these results indicate that our attempt to rescue the knockdown phenotype resulted in overexpression of LPP in some cells of the population, but not in all. As a consequence, the polarization defect in the KD cell line could not be rescued, suggesting that LPP expression may be tightly regulated to ensure that it localizes properly in the cell–cell junctions and to allow adequate proliferation and polarization of the intestinal cells.

IFNg induces a distinct transcriptional signature in Caco-2 cell lines and an enhanced inflammatory response upon LPP knockdown

To explore the major transcriptional changes upon LPP knockdown occurring under physiological and inflammatory conditions, we evaluated the gene expression responses of unstimulated and IFNg-stimulated Caco-2 cells at different time points (3, 8 and 40 hrs; Fig.

4A) using RNA-seq. The optimal IFNg dose was determined in a previous titration assay experiment where we selected the minimal concentration that induced a stable decrease in TEER in the WT cell line as a proxy for alterations in tight junction integrity (Fig. 4B). IFNg stimulation led to a time-dependent increase in the expression of the IL-15 receptor gene (IL15R) and the MX1 gene (Fig. 4C, D), two genes highly expressed in intestinal biopsies of CeD patients

, which indicates that our model responds to CeD-related inflammatory signals. Of note, MX1 expression was higher (~40x) in the KD cell line compared to controls after 3, 8 and 40 hrs of IFNg stimulation. MX1 transcripts decreased after 3 hrs of stimulation in the KD cell line, but remained ~10x higher compared to WT and SCR. MX1 plays a pivotal role in the antiviral immune response

and in inflammatory-mediated disease etiology

. The remarkable induction of this gene that we observed in the KD cell line suggests that reduced expression of LPP contributes to an exacerbated inflammatory response in intestinal epithelial cells.

PCA of the global gene transcription response indicates that all the cell lines in the panel

respond to IFNg in a time-dependent fashion (Fig. 4E). Although the PCA clearly discriminates

between the WT, SCR or KD cell lines, it should be noted that the distance between the

samples shows major differences between the controls at baseline (unstimulated) and

after IFNg stimulation, suggesting that the overall transcriptome in the SCR cell line does

not resemble the WT. These discrepancies were surprising given the previously described

similarity between the controls in phenotypic characteristics such as proliferation rate and

Chapter 2

To test whether the transcriptomic differences in the controls were due to alterations in LPP expression, we plotted the expression profile of LPP over time (Fig. 4F). In general, we observed similar expression patterns in the control cells. Despite small fluctuations, the LPP level was 1.2 (log2) times lower in the KD cell line compared to controls. Together these data suggest that all the cell lines are responsive to IFNg, which is a disease-associated inflammatory stimulus or stressor.

LPP knockdown affects inflammatory and metabolic genes in response to IFNg

To identify the main biological pathways affected by LPP knockdown, we performed an unsupervised hierarchical cluster analysis of all the IFNg-dependent DEGs. A total of 2,212 DEGs were identified and clustered based on log2-fold changes relative to unstimulated controls. In agreement with the PCA, clustering analysis revealed that each cell line exhibits a characteristic response upon stimulation over time and that the WT and SCR cell lines show striking differences despite their phenotypic resemblance and consistent LPP expression in time (Fig. 5A). We identified 15 clusters, of which seven were similar between the control cell lines but different in the KD at one or more time points (clusters 1, 2, 5, 6, 7, 9 and 12;

indicated with black boxes in Fig. 5A). These seven clusters represent pathways that are uniquely perturbed upon LPP knockdown. Pathway analysis on these DEGs revealed that these clusters encompass genes associated with a range of biological processes (Fig. 5B).

Figure 4. IFNg induces a distinct transcriptional signature in Caco-2 cell lines and an enhanced inflammatory response upon LPP knockdown. Caco-2 cell lines were seeded in transwells. After 2 weeks, cells were left untreated (unstimulated (0 hrs)) or incubated with IFNg 50, 200, 500 ng/ml (A) or 60 ng/ml (C-F). (A) IFNg titration using TEER (arbitrary units) as proxy. Each concentration was tested on duplicated. (B) Experimental setup for C-F. Relative expres- sion of MX1 (C) and IL-15R (D) determined by qPCR. Bars indicate the mean ± SEM of technical triplicates in one of the duplicates of the assay. (E) PCA showing the top variable gene expression patterns determined by RNA-seq in the different cells lines and time points. (F) Line plot depicting LPP expression pattern (Log 2 counts, each point in duplicate).