Open Access
Open Peer Review

This article has Open Peer Review reports available.

How does Open Peer Review work?

On the genetic involvement of apoptosis-related genes in Crohn's disease as revealed by an extended association screen using 245 markers: no evidence for new predisposing factors

  • Sonja EN Wagenleiter1,
  • Peter Jagiello2,
  • Denis A Akkad1,
  • Larissa Arning1,
  • Thomas Griga3,
  • Wolfram Klein1 and
  • Jörg T Epplen1Email author
Journal of Negative Results in BioMedicine20054:8

https://doi.org/10.1186/1477-5751-4-8

Received: 07 July 2005

Accepted: 30 November 2005

Published: 30 November 2005

Abstract

Crohn's disease (CD) presents as an inflammatory barrier disease with characteristic destructive processes in the intestinal wall. Although the pathomechanisms of CD are still not exactly understood, there is evidence that, in addition to e.g. bacterial colonisation, genetic predisposition contributes to the development of CD. In order to search for predisposing genetic factors we scrutinised 245 microsatellite markers in a population-based linkage mapping study. These microsatellites cover gene loci the encoded protein of which take part in the regulation of apoptosis and (innate) immune processes. Respective loci contribute to the activation/suppression of apoptosis, are involved in signal transduction and cell cycle regulators or they belong to the tumor necrosis factor superfamily, caspase related genes or the BCL2 family. Furthermore, several cytokines as well as chemokines were included. The approach is based on three steps: analyzing pooled DNAs of patients and controls, verification of significantly differing microsatellite markers by genotyping individual DNA samples and, finally, additional reinvestigation of the respective gene in the region covered by the associated microsatellite by analysing single-nucleotide polymorphisms (SNPs). Using this step-wise process we were unable to demonstrate evidence for genetic predisposition of the chosen apoptosis- and immunity-related genes with respect to susceptibility for CD.

Introduction

Crohn's disease (CD) is a chronic inflammatory disorder characterized by destructive processes in the intestinal wall. Interactions between genetic and environmental factors potentially lead to an imbalance between the luminal bacterial flora, and the innate as well as the adaptive immune systems [1, 2]. Epidemiological and genome wide studies have lead to the identification of factors establishing genetic involvement in CD [1, 3, 4]. Despite of fundamental findings, namely the variation in the CARD15 receptor and their association with CD, the causative instances regulating the exaggerated mucosal response remained elusive. The proposed pathomechanisms of CD are manifold. The dysregulated response of the innate immune system is supposed to present a crucial step in the pathogenesis of CD [5]. This fact has been confirmed genetically by several CD associations of genes such as CD14, TLR4 and in some instances the interaction of their variations with CARD15 [6, 7]. In regard to the polarized T helper (Th) response, the adaptive immune system appears affected in CD as well [810]. Moreover, several studies implicated a role of programmed cell death in CD [1115]. Apoptosis mediates 'self-tolerance', the elimination of autoreactive immune compartments. In addition, the thoroughly controlled termination of a physiological immune response is due to the process of programmed cell death. In CD mucosal T cells show less susceptibility to apoptosis [16]. In this context TNFα protein exerts multiple physiological effects, and anti-TNFα therapeutic strategies (e.g. infliximab) are effective in (maintaining) remission of CD [17]. In several studies it has been revealed that treatment of CD patients with infliximab leads to an activation of T cells rendering them susceptible for apoptosis [18, 19]. Interestingly, the effect of this treatment may not be due to neutralisation of soluble TNFα (and its binding to the TNFRs), but rather it may be caused by its affinity to membrane-bound TNFα putatively changing the ratio of anti- and pro-apoptotic mediators towards induction of apoptosis [18, 20]. Although the mechanisms of the causal role of T cells responses in CD remain to be determined in detail, there is substantial clinical evidence that suggests a role for uncontrolled activated T lymphocytes in the pathogenic process of CD [2124]. Nevertheless, it is uncertain, whether a genetic basis for a decreased activation/apoptosis of T lymphocytes in CD patients exists, and whether increased anti-apoptotic markers, found in T cells of these patients are due to the mucosal inflammation, secondarily [18].

In such a complex situation we used extended association screening (EAS) with markers representing 245 apoptosis- and (innate) immunity-related genes. The majority of the investigated markers have been successfully utilized in respective studies before [25, 26]. Our population based linkage mapping comprises a 3-stage analysis with pooled DNA in the initial phase and subsequently individual genotyping. In order to confirm such results, several tagging SNPs of the adjacent gene represented by the marker were analysed. Here, we investigated the role of distinct biological pathways for the susceptibility of CD.

Materials and methods

Patients

One hundred and fifty eight well-characterized patients with a clinical, endoscopical and histological diagnosis of CD were included. This patient cohort has been reported before [27, 28]. All patients were of German origin and the diagnosis of CD was adjusted according to the diagnostic criteria of the European Community Workshop on Inflammatory Bowel Diseases (IBD). As controls a group of healthy northern German (NoG) and western German (WeG) origin were analysed. In the initial step a group of ~100 NoG individuals were used. In order to exclude population stratification, genotyping of chosen SNPs was performed in 180–460 NoG and WeG individuals.

Pooling of DNA

The DNA concentration from each individual of the patient and control cohorts was quantified by spectrophotometry, carried out four times, and then diluted accordingly to 100 ng/μl. In a second step the DNA was diluted to a concentration of 65 ng/μl and once more measured by spectrophotometry. Finally, DNA diluted to 50 ng/μl was adjusted to a final amount of 1000 ng for each individual in a pool of 50 persons. In the initial stage, marker analyses were performed with two patient and two control subpools, respectively.

Tailed primer PCR

Tailed primer PCR was performed as described before [25]: An 18 bp-tail was added to each sense oligonucleotide. PCR reaction included three oligonulceotides, two of which were target specific. The third one consists of the same sequence as the abovementioned tail that was additionally fluorescence-labelled.

Microsatellite markers

Intragenic microsatellite or markers located in the immediate vicinity (<50 kb) of the specific gene were included. Information on the oligonucleotide sequences and location of markers are given at the website (Additional file 1; see also Tab. 1). As reported before, only markers with equal "intra-subgroup" allele distributions with ≥ 2 alleles were considered in subsequent analyses [25]. Significantly associated markers were genotyped individually in order to exclude false-positive results due to possible pooling artefacts. All in all, 245 microsatellite markers representing distinct genes were analysed on an ABI377 slab-gel system (Applied Biosystems, Darmstadt, Germany).
Table 1

Genes investigated for CD association as represented by an intra- or juxtagenic microsatellite marker (for additional information see URL: http://www.ruhr-uni-bochum.de/mhg/marker_information_SENW.pdf)

apoptosis related

REQ

TNFSF12

CTLA4

Casp10

IL4

 

RNF7

TNFSF14

DAP

Casp14

IL4R

 

SMAC

TNFSF15

DAPK1

Casp2

IL6

AIF

TIAF1

TNFSF18

FADD

Casp3

IL8

APR3

TIAL1

TNFSF4

IKBKG

Casp4

IRF1

BCLG

TP73

TNFSF5

MADD

Casp5

NRG1B

BFAR

VDR

TNFSF6

MAP2K6

Casp6

PRL

CIDEB

 

TNFSF7

MAP3K14

Casp7

PRLR

CYBB

Bcl2 related

TNFSF8

MAP3K5

Casp8

 

CYP51

 

TNFSF9

MAP4K4

CASP8AP2

chromosome 6

DAD1

BCL2A1

TOSO

NFKB1

Casp9

 

DAP3

Bag1

 

NFKB2

 

No.1

DATF1

BAK

innate immunity

NSMAF

apoptosis suppressor

No.4

DAXX

BAX

 

PAWR

 

No.5

DEDD

BCL2

BPI

PIAS3

 

No.6

DHCR24

BCL2L1

CD14

PTEN

API5

No.7

EIF4G2

BCL2L11

CD5L

RARB

BIRC1

No.8

FASTK

BCL2L13

DEFB119/ DEFB121

RIPK1

BIRC2

D6S1014

FLIP

BID

DEFB127

RIPK2

BIRC3

D6S1959

FRZB

BIK

HBD1

RIPK3

BIRC4

D6S273

GSK3B

BNIP3L

IFNB1

RXRB

BIRC6

 

GSR

MCL1

LY64

STK17A

BIRC8

others

GZMA

 

LY86

STK17B

  

GZMB

TNF superfamily

LY96

TANK

cytokine chemokines

BPHL/TUBB

HLCS

 

NCF1

TRADD

 

TAPBPR

NME3

LTB (TNFSF3)

NCF4

Traf3

 

VEGF

NOL3

LTBR (TNFRSF3)

PGLYRP

Traf4

AXL

LGALS3

NOS1

TNFa

PLA2G4A

Traf5

CSF1R

BDNF

NOS2A

TNFRSF10A

PLUNC

Traf6

CSF2

NGFB

NOX1

TNFRSF10B

SerpinA1

 

CSF2RB

NGFR

NOX3

TNFRSF10C

SerpinB1

cell cycle regulators

CSF3

TrkC

NOX4

TNFRSF10D

SFTPA1

 

Dtk

 

P2RX1

TNFRSF11A

SLPI

CCND2

erbB3

positive control

P53AIP1

TNFRSF11B

STAT3

CDC2

GAS1

CARD15

PDCD10

TNFRSF12

TGFB1

CDKN1A

IGF1

 

PDCD2

TNFRSF17

TLR1

CDKN2A

IGF2R

 

PDCD5

TNFRSF18

TLR2

PAK1B

IL10

 

PDCD6

TNFRSF19

TLR3

RbAp48

IL10RA

 

PDCD6IP

TNFRSF19L

TLR4

Rb2/p130

IL10RB

 

PDCD8

TNFRSF1A

TLR5

RBP1

IL11RA

 

PLA2G10

TNFRSF1B

TLR7

RBP2

IL12A

 

PLA2G1B

TNFRSF21

TLR8

RBQ-1

IL12B

 

PLA2G6

TNFRSF4

TLR9

RBQ-3

IL12RB2

 

PTGS1

TNFRSF5

TLR10

TP53

IL13RA2

 

REQ

TNFRSF6 (FAS)

 

TP53INP1

IL18

 

RNF7

TNFRSF6B

signal transduction

 

IL18R

 

SMAC

TNFRSF7

 

caspase related

IL1RL1

 

TIAF1

TNFRSF8

Traf1

 

IL1B

 

TIAL1

TNFRSF9

BCL10

ADPRT

IL2

 

TP73

TNFSF10

CHUK

CARD4

IL24

 

VDR

TNFSF11

CRADD

Casp1

IL2RA

 

Statistics for initial comparisons of allele frequencies

Raw data from ABI377 profiles were analysed by the Genotyper software (ABI) producing a marker specific allele image profile (AIP) which includes different heights of peaks reflecting the allele frequencies. In order to test differences of the AIPs between CD patients and the controls, all peak heights were summarized for each pool and set to 100 %. The total allele count for each distinct allele was then estimated. Thereupon, the AIPs of the case and control pools were compared statistically by means of contingency tables. Hence, P values are nominal and approximate, because estimated rather than observed counts were used for allele frequencies. The significance level was set at p = 0.05. In order to focus the statistics on major alleles, all minor alleles with a frequency of less than 0.05 were summarized to a virtual allele. Subsequently, a second statistical analysis by means of contingency tables was undertaken. A third step for statistical testing each allele individually was accomplished (and the summation of all other marker alleles), whereby the respective value of the patient group was compared with those of the controls and subsequent χ2 analyses. Despite of evidence that correction for multiple comparisons might eliminate 'real positive' results [26], Q value correction was performed with a cut off of 5% for the initial screening procedure [29].

Nevertheless, for selecting markers for further investigations, non-corrected P values were simply ranked according to their evidence for association including all performed statistical procedures.

Individual genotyping

Markers with significantly different allele distributions between patients and controls were controlled by genotyping individual DNA samples of patients and controls in order to exclude false-positive results due to pooling artefacts. Individual genotyping was performed by capillary gel electrophoresis by using the BeckmanCoulter CEQ8000 genetic analysis system (Beckman Coulter, Germany). Results were analysed by comparing each microsatellite allele frequency from the CD cohort with the corresponding allele frequency of the control group by χ2 testing and corrected by the number of marker specific alleles according to Bonferroni (see Tab. 2 and URL: http://www.ruhr-uni-bochum.de/mhg/marker_information_SENW.pdf). Hardy-Weinberg equilibrium (HWE) was tested using the Genepop program http://wbiomed.curtin.edu.au/genepop.
Table 2

P values for microsatellite markers located intragenically or in the immediate vicinity of represented genes after the initial step and individual genotyping.

 

p values

gene (as represented by the respective marker)

after analysis with pooled DNA

after summation of alleles beneath 5%

after analyses of each single allele (most significant allele)

after individual genotyping 1 (p c value)

after correction by Q-value of pooled data

FLIP

0.2871

0.1936

0.0100

0.0044 (pc > 0.05; c = 9)

n.s.

BCL2A1

0.0948

0.0948

0.0275

n.s.

n.s.

BAG1

0.2541

0.2541

0.0163

n.s.

n.s.

BPI

0.0011

0.0011

0.0031

n.s.

n.s.

erbB3

0.0760

0.0932

0.0100

n.s.

n.s.

TP73

0.5928

0.3535

0.0302

n.s.

n.s.

TLR9

0.3004

0.3004

0.0300

n.s.

n.s.

TNFRSF17

0.0012

0.0014

0.0014

0.0012 (pc < 0.01; c = 6)

n.s.

CARD15

0.0083

0.0247

0.0054

0.0050 (pc < 0.04; c = 7)

n.s.

P values were generated using three different procedures as described in the methods' section. Briefly, data were analysed by means of contingency tables, initially comparing allele distributions represented by the AIF (after analyses with pooled DNA), then after summation of alleles < 5% in order to focus on the major alleles and, finally, after comparison of each single allele between the control and patient cohorts. For analysing the results of the individual genotyping χ2 testing was utilised.

1Genotyping was performed with the same individuals used in the pooling procedure, and, when remaining significant, further individuals were added to the analyses (FLIP: CD = 134, controls = 150; TNFRSF17: CD = 147, controls = 135; CARD15: CD = 144, controls = 165).

SNP genotyping

SNPs in genes as represented by significantly associated markers after individual genotyping were investigated by analysis of restriction fragment length polymorphisms (RFLP; see Tab. 3). As the marker representing the TNFRSF17 gene is located in ~1 MBp distance to the MHC class II transactivator (MHC2TA) gene, a functional variation (rs3087456, [30]) of MHC2TA was genotyped by RFLP analyses in 147 CD patients and 463 healthy controls from the abovementioned control populations (see Tab. 3). The results were evaluated by means of χ2 -and HWE testing. Linkage disequilibrium (LD) between the marker alleles and the polymorphism was calculated by the Genepop program.
Table 3

Investigated SNPs in genes as represented by significantly differing microsatellites of the individual genotyping step.

Gene

rs#

Allele 01/02

Oligonucleotides (sense/antisense)

RE

TM (°C)

Allele: fragment length (bp)

FLIP

Rs7583529

A/C

GGTGATTATTCGGACCCCA/AACTACAGATCCCGTGTGGAG

TseI

57

01: 155

02: 103/52

 

Rs2041765

T/C

GAACAAGGAGAGAACCTGGAC/GAGCTGGAAGGCACAGTACA

MboII

56

01: 309

02: 188/121

TNFRSF17

Rs3743591

A/G

ATAAGCAGTTTCTGTTTCAGATGT/CTCTACAAGAATTCCAGAGCA

BceAI

55

01: 223

02: 147/76

 

Rs11570139

C/T

GCCCTGATATTTACACCCTGT/CAGCCATCTGCAACATGAT

CaiI

54

01: 269

02: 161/108

 

Rs373496

T/C

AGGAACTGAAACTCACAATAGC/CAGCTCATTATCTGTCTGATGTT

AluI

55

01: 247

02: 100/90/54/3

MHC2TA

Rs3087456

G/A

* 1 GTGAAATTAATTTCAGAGC TGT/CTCAGCTTCCCCAAGGAT

BfmI

58

01: 268

02: 231/37

Analyses were performed by using the RFLP method. The table depicts information on the used SNPs as well as RFLP/PCR conditions. * 1 A 5'-tail was added to the mismatch (bold letter) sense primer (5'-CATCGCTGATTCGCACAT-3'). PCR was performed with a third oligonucleotide with the equal sequence as the tail. RE: restriction enzyme; TM: melting temperature (used for annealing in PCR).

Results

Initial step

Microsatellites representing 245 genes involved in apoptosis regulation (see Tab. 1) were investigated by using EAS. None of the markers presented with significant intra-subgroup differences confirming the homogeneity of the pools. The statistical evaluation of the microsatellite frequencies in the CD patient and the control cohorts revealed 9 significantly different allele distributions of intra- or juxtagenic markers for FLIP, BCL2A1, BAG1, BPI, erbB3, TP73, TLR9, TNFRSF17 and CARD15 (summarized in Tab. 2).

Individual genotyping

Individual genotyping confirmed significant P values only for the 3 markers FLIP (p = 0.0044, pc > 0.05, in HWE), TNFRSF17 (p = 0.0012, pc < 0.01, in HWE) and the positive control CARD15 (p = 0.0050, pc < 0.04, in HWE). The additional associations for the other markers were rejected (see Tab. 2 and Additional file 1). There were no differences analysing CARD15+ and CARD15- patients.

SNP genotyping

SNP markers (Tab. 3) were genotyped located in the respective genes in the vicinity of the microsatellites representing TNFRSF17 and FLIP. Thus, SNPs were analyzed spread across the genes representing haplotypes as predisposed by the 'LD Select' method reported before [31]. RFLP analyses did not reveal any association of the selected SNPs, neither by comparing the CARD15+ nor the CARD15- patients with the control group.

Comparison of TNFRSF17 microsatellite alleles

The genotypes of the TNFRSF17 microsatellite alleles were compared between the patient and control cohorts. Analyses revealed evidence either for a predisposing (allele 3) and a protective allele (2) or linkage between these alleles and the marker alleles, respectively.

Genotypes including allele 2 are overrepresented in the control cohort, whereas those with the apparently predisposing allele 3 are more frequent in the CD cohort, thus confirming the results of individual genotyping (see Fig. 1).
Figure 1

Genotype frequencies of the microsatellite representing the TNFRSF17 gene. Only genotypes with a frequency of > 0.01 are included. Alleles of the respective microsatellite are indicated as numbers in the X-axis according to their length in bp. For example: 1–1 (read from the number below the numerical series and the first number of the numerical series) means homozygous genotype for microsatellite allele number one and 1–4 heterozygous genotype for allele 1 and 4. Genotypes comprising allele 2 are over-represented within the control group (47% vs. 29%; pc = 0.0042 with c = 2), whereas allele 3 genotypes are more frequent in the patient cohort (58% vs. 52% pc = 0.3130; c = 2). Therefore, allele 2 might imply a protective effect and/or allele 3 a predisposing effect on CD. Interestingly, the genotype 2–3 is more prevalent in the control group. This result can be interpreted by a different effect size of allele 2 (↑) as compared to allele 3, or the significant difference of this microsatellite is only due to linkage of allele 2 with a protective factor.

MHC2TA analyses

The analyses of the functionally significant polymorphism rs3087456 revealed a marginal association in our CD patients when allele or genotype frequencies were compared between the combined control (WeG and NoG did not differ in allele frequencies) and the patient cohorts (see Tab. 4). Analyses for LD between TNFRSF17 and MHC2TA alleles, however, did not reveal any significant deviations from equilibrium.
Table 4

Allele and genotype frequencies of the functional MHC2TA polymorphism (rs3087456).

 

Allele frequencies

p value

OR (CI)

Genotype frequencies

p value

CD (n = 147)

C

0.32

0.05

1.33 (0.90–2.01)

CC

0.08

0.54

 

T

0.68

  

CT

0.48

0.06

     

TT

0.44

0.03

controls (n = 463)

C

0.26

  

CC

0.07

 
 

T

0.74

  

CT

0.39

 
     

TT

0.54

 

OR: odds ratio; CI: 95% confidence interval

Discussion

The pathomechanisms of CD are still not exactly understood, albeit certain CARD15 variations appear especially frequent in CD patients; thus genetic involvement is proven. These genetic predisposition factors, however, are neither sufficient nor explain they the pathogenesis in all CD patients. In this study we present an association screen mainly for apoptosis and immunity related genes by microsatellite markers as investigated in a 3-step approach.

Our initial analyses revealed 9 significantly different allele distributions of intra- or juxtagenic markers for FLIP, BCL2A1, BAG1, BPI, erbB3, TP73, TLR9, TNFRSF17 and CARD15 (see Tab. 2). Yet, after correction by Q-value, none of those markers remained significant. On the other hand, a recent study raised the question, whether the correction for multiple comparisons should be applied at all in EAS [26]. For example, in these analyses a previously significantly associated microsatellite (representing the TNFα gene), which has been used as a positive control such as CARD15, would have been rejected by the correction procedure. Therefore, it remains conceivable that the abovementioned markers represent rather hints for additional predisposing factors/loci with low effect size.

The most promising markers (reflected by a significant p-value) were included in further analyses regardless of the correction procedure. Individual genotyping rejected most markers found to be significantly different in the initial step of our approach and only three markers remained significant representing the TNFRSF17, FLIP, CARD15 genes (Tab. 2). Obviously, pooled and individual genotyping yield somewhat contradictory results. Eight microsatellites revealed significantly differences between the patient and control cohorts after the pooling procedure, whereas individual genotyping results in the confirmation of 'only' 2 markers. These conspicuous differences might be due to several artefacts caused by analyses with pooled DNA. For example, a typical artefact is the length-dependent amplification of short alleles or the presence of null-alleles. Additionally, consistency of the analyses by a slab-gel system might reflect a further hindrance in this subtle procedure. Nevertheless, individual genotyping eliminates false positive results due to pooling artefacts and, in case of significant results, enables the thorough analyses of the marker alleles in detail (see Fig.1). In order to confirm the aforementioned positive results further markers (SNPs, Tab. 3) were genotyped located in the respective genes in the vicinity of the microsatellites representing TNFRSF17 and FLIP. Yet, RFLP analyses did not reveal any association of the selected SNPs and, therefore, the microsatellite data were not confirmed. On the other hand, these SNPs might not represent regions properly that encompass regulatory elements.

In some instances, the LD of distinct microsatellite alleles covers long genetic distances, thus further gene variations might be in linkage with these alleles. Since the significantly associated 'TNFRSF17' marker is located at the IBD8 region with 1MBp distance to the major histocompatibility class (MHC) II transactivator (MHCIITA), a previously reported functional variation of the MHC2TA gene was analysed (see Tab. 4; [30]). MHC2TA regulates the expression of human leukocyte antigen (HLA) genes regulating the adaptive immune system by presenting antigens to CD4+ T cells, thereby re-activating these cells. The HLA region has been implicated in IBD [32]. In addition to the localisation of MHC2TA at IBD8 and the associated marker in the adjacent region, the putative biological relevance of the functional rs3087456 polymorphism for CD motivated us to genotype this variation. The analyses did reveal a marginal association in our CD patients when allele or genotype frequencies were compared between the combined control and patient cohorts (see Tab. 4). Yet there was no evidence for LD between TNFRSF17 and MHC2TA alleles. In order to validate these data further patient cohorts comprising more individuals must be scrutinised. In addition, other genes that might be linked with the 'TNFRSF17' marker must be analysed (at least 15 RefSeq genes in the region are encompassed by the microsatellite marker and MHC2TA).

In conclusion, this study did not reveal overt evidence for CD predisposition factors in apoptotic (and immune) pathways. Certainly, our approach depends on the LD between the investigated microsatellites and putative predisposing or protective alleles, depending on functional relevance to the disease. Thus, in some instances microsatellites might not be entirely representative for the adjacent genes. Furthermore, the investigated genes only cover part of the factors which coordinate programmed cell death. Yet, future information about haplotype blocks may facilitate more far-fetched interpretations of our analyses.

Declarations

Authors’ Affiliations

(1)
Department of Human Genetics, Ruhr-University
(2)
Institute for Clinical Molecular Biology, University Schleswig-Holstein
(3)
Department of Gastroenterology, University Hospital Bergmannsheil

References

  1. Russel MG, Pastoor CJ, Janssen KM, van Deursen CT, Muris JW, van Wijlick EH, Stockbrugger RW: Familial aggregation of inflammatory bowel disease: a population-based study in South Limburg, The Netherlands. The South Limburg IBD Study Group. Scand J Gastroenterol Suppl. 1997, 223: 88-91.PubMedGoogle Scholar
  2. Hugot JP, Thomas G: Genome-wide scanning in inflammatory bowel diseases. Dig Dis. 1998, 16 (6): 364-369. 10.1159/000016893.View ArticlePubMedGoogle Scholar
  3. Hugot JP, Chamaillard M, Zouali H, Lesage S, Cezard JP, Belaiche J, Almer S, Tysk C, O'Morain CA, Gassull M: Association of NOD2 leucine-rich repeat variants with susceptibility to Crohn's disease. Nature. 2001, 411 (6837): 599-603. 10.1038/35079107.View ArticlePubMedGoogle Scholar
  4. Ogura Y, Bonen DK, Inohara N, Nicolae DL, Chen FF, Ramos R, Britton H, Moran T, Karaliuskas R, Duerr RH: A frameshift mutation in NOD2 associated with susceptibility to Crohn's disease. Nature. 2001, 411 (6837): 603-606. 10.1038/35079114.View ArticlePubMedGoogle Scholar
  5. Rogler G: Update in inflammatory bowel disease pathogenesis. Curr Opin Gastroenterol. 2004, 20 (4): 311-317. 10.1097/00001574-200407000-00003.View ArticlePubMedGoogle Scholar
  6. Franchimont D, Vermeire S, El Housni H, Pierik M, Van Steen K, Gustot T, Quertinmont E, Abramowicz M, Van Gossum A, Deviere J: Deficient host-bacteria interactions in inflammatory bowel disease? The toll-like receptor (TLR)-4 Asp299gly polymorphism is associated with Crohn's disease and ulcerative colitis. Gut. 2004, 53 (7): 987-992. 10.1136/gut.2003.030205.View ArticlePubMedPubMed CentralGoogle Scholar
  7. Klein W, Tromm A, Griga T, Folwaczny C, Hocke M, Eitner K, Marx M, Duerig N, Epplen JT: Interaction of polymorphisms in the CARD15 and CD14 genes in patients with Crohn disease. Scand J Gastroenterol. 2003, 38 (8): 834-836. 10.1080/00365520310003147.View ArticlePubMedGoogle Scholar
  8. Pallone F, Blanco Gdel V, Vavassori P, Monteleone I, Fina D, Monteleone G: Genetic and pathogenetic insights into inflammatory bowel disease. Curr Gastroenterol Rep. 2003, 5 (6): 487-492.View ArticlePubMedGoogle Scholar
  9. Okazawa A, Kanai T, Watanabe M, Yamazaki M, Inoue N, Ikeda M, Kurimoto M, Ishii H, Hibi T: Th1-mediated intestinal inflammation in Crohn's disease may be induced by activation of lamina propria lymphocytes through synergistic stimulation of interleukin-12 and interleukin-18 without T cell receptor engagement. Am J Gastroenterol. 2002, 97 (12): 3108-3117. 10.1111/j.1572-0241.2002.07107.x.View ArticlePubMedGoogle Scholar
  10. Singh UP, Singh S, Iqbal N, Weaver CT, McGhee JR, Lillard JW: IFN-gamma-inducible chemokines enhance adaptive immunity and colitis. J Interferon Cytokine Res. 2003, 23 (10): 591-600. 10.1089/107999003322485099.View ArticlePubMedGoogle Scholar
  11. Souza HS, Tortori CJ, Castelo-Branco MT, Carvalho AT, Margallo VS, Delgado CF, Dines I, Elia CC: Apoptosis in the intestinal mucosa of patients with inflammatory bowel disease: evidence of altered expression of FasL and perforin cytotoxic pathways. Int J Colorectal Dis. 2005, 20 (3): 277-286. 10.1007/s00384-004-0639-8.View ArticlePubMedGoogle Scholar
  12. Xia B, Yu YH, Guo QS, Li XY, Jiang L, Li J: Association of Fas-670 gene polymorphism with inflammatory bowel disease in Chinese patients. World J Gastroenterol. 2005, 11 (3): 415-417.View ArticlePubMedPubMed CentralGoogle Scholar
  13. Peppelenbosch MP, van Deventer SJ: T cell apoptosis and inflammatory bowel disease. Gut. 2004, 53 (11): 1556-1558. 10.1136/gut.2004.040824.View ArticlePubMedPubMed CentralGoogle Scholar
  14. Brannigan AE, O'Connell PR, Hurley H, O'Neill A, Brady HR, Fitzpatrick JM, Watson RW: Neutrophil apoptosis is delayed in patients with inflammatory bowel disease. Shock. 2000, 13 (5): 361-366.View ArticlePubMedGoogle Scholar
  15. Seidelin JB, Nielsen OH: [Apoptosis in chronic inflammatory bowel disease. The importance for pathogenesis and treatment]. Ugeskr Laeger. 2003, 165 (8): 790-792.PubMedGoogle Scholar
  16. Ina K, Itoh J, Fukushima K, Kusugami K, Yamaguchi T, Kyokane K, Imada A, Binion DG, Musso A, West GA: Resistance of Crohn's disease T cells to multiple apoptotic signals is associated with a Bcl-2/Bax mucosal imbalance. J Immunol. 1999, 163 (2): 1081-1090.PubMedGoogle Scholar
  17. Hommes DW, van Deventer SJ: Targeting tumor necrosis factor-alpha in inflammatory bowel disease: why, how, and when?. Curr Opin Gastroenterol. 2003, 19 (4): 350-357. 10.1097/00001574-200307000-00005.View ArticlePubMedGoogle Scholar
  18. ten Hove T, van Montfrans C, Peppelenbosch MP, van Deventer SJ: Infliximab treatment induces apoptosis of lamina propria T lymphocytes in Crohn's disease. Gut. 2002, 50 (2): 206-211. 10.1136/gut.50.2.206.View ArticlePubMedPubMed CentralGoogle Scholar
  19. Van den Brande JM, Braat H, van den Brink GR, Versteeg HH, Bauer CA, Hoedemaeker I, van Montfrans C, Hommes DW, Peppelenbosch MP, van Deventer SJ: Infliximab but not etanercept induces apoptosis in lamina propria T-lymphocytes from patients with Crohn's disease. Gastroenterology. 2003, 124 (7): 1774-1785. 10.1016/S0016-5085(03)00382-2.View ArticlePubMedGoogle Scholar
  20. Sands BE: Why Do Anti-Tumor Necrosis Factor Antibodies Work in Crohn's Disease?. Rev Gastroenterol Disord. 2004, 4 (Suppl. 3): S10-S17.PubMedGoogle Scholar
  21. Sturm A, Leite AZ, Danese S, Krivacic KA, West GA, Mohr S, Jacobberger JW, Fiocchi C: Divergent cell cycle kinetics underlie the distinct functional capacity of mucosal T cells in Crohn's disease and ulcerative colitis. Gut. 2004, 53 (11): 1624-1631. 10.1136/gut.2003.033613.View ArticlePubMedPubMed CentralGoogle Scholar
  22. Sartor RB: Current concepts of the etiology and pathogenesis of ulcerative colitis and Crohn's disease. Gastroenterol Clin North Am. 1995, 24 (3): 475-507.PubMedGoogle Scholar
  23. Doering J, Begue B, Lentze MJ, Rieux-Laucat F, Goulet O, Schmitz J, Cerf-Bensussan N, Ruemmele FM: Induction of T lymphocyte apoptosis by sulphasalazine in patients with Crohn's disease. Gut. 2004, 53 (11): 1632-1638. 10.1136/gut.2003.037911.View ArticlePubMedPubMed CentralGoogle Scholar
  24. Elson CO, Sartor RB, Tennyson GS, Riddell RH: Experimental models of inflammatory bowel disease. Gastroenterology. 1995, 109 (4): 1344-1367. 10.1016/0016-5085(95)90599-5.View ArticlePubMedGoogle Scholar
  25. Jagiello P, Gencik M, Arning L, Wieczorek S, Kunstmann E, Csernok E, Gross WL, Epplen JT: New genomic region for Wegener's granulomatosis as revealed by an extended association screen with 202 apoptosis-related genes. Hum Genet. 2004, 114 (5): 468-477. 10.1007/s00439-004-1092-z.View ArticlePubMedGoogle Scholar
  26. Wieczorek S, Jagiello P, Arning L, Dahmen N, Epplen JT: Screening for candidate gene regions in narcolepsy using a microsatellite based approach and pooled DNA. J Mol Med. 2004, 82 (10): 696-705. 10.1007/s00109-004-0569-5.View ArticlePubMedGoogle Scholar
  27. Klein W, Tromm A, Griga T, Fricke H, Folwaczny C, Hocke M, Eitner K, Marx M, Duerig N, Epplen JT: A polymorphism in the CD14 gene is associated with Crohn disease. Scand J Gastroenterol. 2002, 37 (2): 189-191. 10.1080/003655202753416867.View ArticlePubMedGoogle Scholar
  28. Klein W, Tromm A, Griga T, Fricke H, Folwaczny C, Hocke M, Eitner K, Marx M, Duerig N, Epplen JT: Interleukin-4 and interleukin-4 receptor gene polymorphisms in inflammatory bowel diseases. Genes Immun. 2001, 2 (5): 287-289. 10.1038/sj.gene.6363779.View ArticlePubMedGoogle Scholar
  29. Storey JD, Tibshirani R: Statistical significance for genomewide studies. Proc Natl Acad Sci U S A. 2003, 100 (16): 9440-9445. 10.1073/pnas.1530509100.View ArticlePubMedPubMed CentralGoogle Scholar
  30. Swanberg M, Lidman O, Padyukov L, Eriksson P, Akesson E, Jagodic M, Lobell A, Khademi M, Borjesson O, Lindgren CM: MHC2TA is associated with differential MHC molecule expression and susceptibility to rheumatoid arthritis, multiple sclerosis and myocardial infarction. Nat Genet. 2005, 37 (5): 486-494. 10.1038/ng1544.View ArticlePubMedGoogle Scholar
  31. Carlson CS, Eberle MA, Rieder MJ, Yi Q, Kruglyak L, Nickerson DA: Selecting a maximally informative set of single-nucleotide polymorphisms for association analyses using linkage disequilibrium. Am J Hum Genet. 2004, 74 (1): 106-120. 10.1086/381000.View ArticlePubMedGoogle Scholar
  32. Yap LM, Ahmad T, Jewell DP: The contribution of HLA genes to IBD susceptibility and phenotype. Best Pract Res Clin Gastroenterol. 2004, 18 (3): 577-596. 10.1016/j.bpg.2004.01.003.View ArticlePubMedGoogle Scholar

Copyright

© Wagenleiter et al; licensee BioMed Central Ltd. 2005

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.