NOD2/CARD15 gene variants have not been universally associated with stricturing behaviour in Crohn's disease. Other behaviour modifying genes could explain these results.
To study the combined influence of NOD2/CARD15 variants and 4G/4G genotype of type-1 plasminogen activator inhibitor (PAI-1) gene on Crohn's disease behaviour.
One hundred and seventy Crohn's disease patients were studied prospectively, with a mean follow-up of 7± 6 years. Disease behaviour was registered by using two criteria: the Vienna classification and a non-hierarchical classification based on the behavioural Vienna categories.
In the multivariate analysis for stricturing behaviour according to the Vienna categories, only absence of colonic disease (OR, 4.0; 95% CI: 1.49–11.1; P = 0.006) was an independent predictive factor. However, in the multivariate analysis for stricturing disease applying a non-hierarchical criteria, ileal disease (OR, 4.19; 95% CI: 1.30–13.5; P = 0.01), and carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype (OR, 5.02; 95% CI: 1.44–17.48; P = 0.01) were independent predictive factors. In the multivariate analysis for penetrating behaviour, the 4G/4G PAI-1 (OR, 3.10; 95% CI: 1.54–6.23; P = 0.001) and male sex (OR, 2.44; 95% CI: 1.30–4.60; P = 0.005) were independent predictive factors irrespective of criteria applied.
Combined PAI-1 and NOD2/CARD15 genotyping predict complicated Crohn's disease. Patients with these variants could benefit from early interventions.
A growing body of evidence supports a key role for genetic factors in Crohn's disease (CD) (MIM 266600).1, 2 Three mutations within NOD2/CARD15 gene have been the more consistent and significant genetic factors implicated in CD susceptibility.3–5 However, phenotypic implications of these mutations in a major event of disease course, such as CD behaviour, have not been universally established. Diverse studies have shown that carriage of variants predisposes to stricturing disease6–8 and may protect against a penetrating complication.7, 9 However, other studies have found a positive association with penetrating phenotype with or without stricturing disease.10–13 A meta-analysis showed that CD patients carrying variants of NOD2/CARD15 had only a ‘modestly’ increased risk of stenosing CD (OR = 1.94, 95% CI: 1.61–2.34).14 The variability of these results may be due to several factors including ethnic variation, duration of CD, environmental factors and specially the classification used and phenotypic definitions. Nevertheless, being CD a complex polygenic disease,1, 2 other genes may be involved as behaviour modifiers in CD.
In a recent study, we have described that the 4G/4G genotype of the promoter polymorphism of the type-1 plasminogen activator inhibitor (PAI-1) gene, which is associated to increased PAI-1 levels,15, 16 is a strong determinant of penetrating behaviour in CD patients.17 The 4G/4G variant of PAI-1 has been associated with higher risk of thrombotic events15, 16, 18, 19 and it has been proposed that appearance of microvascular thrombi within inflammatory lesions may predispose to the formation of deep ulcerations and fistula. In addition, the 4G/4G variant may also have a profibrotic effect by preventing collagen degradation.20–22 Therefore, it is conceivable that this genetic variant could also predispose to stricturing behaviour in CD.
We hypothesise that different combinations of genotypes of PAI-1 and NOD2/CARD15 genes could account for the variability of the results in previously reported genotype–phenotype studies. The importance of identifying stable factors as genes that predispose to specific behaviour phenotypes is based on the need to select accurately patients for preventive and/or early treatment interventions.
Our aim was to study the combined influence of genotypes of both NOD2/CARD15 and PAI-1 genes on CD behaviour.
Patients and methods
One hundred and seventy unrelated Caucasian CD patients regularly visited in a single inflammatory bowel disease (IBD) unit of a tertiary referral centre in Spain were included in this study between December 2002 and January 2004. Ninety-one patients were males and 79 females. Mean follow-up was 7 ± 6 years with a median of 6 years. These CD patients belong to a study recently published addressed to establish an association between NOD2/CARD15 variants and surgery requirements in CD.23 In addition, 186 Caucasian unrelated healthy blood donors matched by sex and age were studied as control subjects. Patients and controls of Asian, African or Jewish ethnicity were excluded from this study, as well as CD patients with mixed ethnicities in previous known generations. The diagnosis of CD was based on standard clinical, radiological, endoscopic and histological criteria.24 All patients gave their written-informed consent to participate in the study after approval of the project by the local ethical committee. All clinical data were obtained from a prospective clinical database of our unit. At the initial visit, a complete clinical evaluation was performed. This evaluation included gender, age at diagnosis of CD, location and behaviour of CD, extra-intestinal manifestations of CD, previous appendectomy, smoking habit and family history of IBD. Location of CD was determined according to X ray, endoscopic or surgical findings. Location of CD was registered in not exclusive groups, according to the following criteria: upper gastrointestinal tract, for any disease location proximal to the terminal third of ileum; terminal ileum, if evidence of disease existed in the terminal third of ileum with or without spill over into the caecum; colonic disease if evidence of disease existed in the colon. As classification was not exclusive, patients could belong to more than one group. For initial behaviour of CD, the Vienna categories definitions were used:25 penetrating, for intra-abdominal or perianal fistulas, inflammatory masses, and/or abscess; stricturing, for occurrence of luminal narrowing (radiological or endoscopic) with prestenotic dilatation or obstructive signs, without concomitant penetrating disease; and inflammatory, for the rest of CD patients. Anal fissures, perianal ulcers without fistulas or abscess and post-operative complications like perforation or fistula formation were not included as penetrating lesions in the analysis. Smokers were defined as patients who smoked more than seven cigarettes per week. Ex-smoker was defined as a patient who gave up smoking at least one year before diagnosis. Non-smokers were defined as patients who had never smoked or smoked <7 cigarettes per week.26
Patients were visited at least twice a year. The following clinical characteristics were registered during follow-up: disease activity, smoking habit, use of steroids, use of immunosuppressants, use of anti-TNF-blocking agents, surgery requirements and changes in disease phenotype. Each steroid course was registered to obtain a mean number of steroid courses per year of follow-up. For disease activity, flare of CD was defined as presence of digestive and/or general symptoms that led to a change in the patient treatment by the physician in charge (MA, MS and JP). A mean number of flares per year of follow-up were also calculated for every patient. Behaviour of CD was always re-evaluated when patients presented a new flare of the disease. Behaviour of CD during follow-up was registered by using two criteria: the Vienna classification,25 and a non-hierarchical criterion applied to the behavioural Vienna categories.23 This non-hierarchical criterion was applied with the intention to avoid the loss of information, in particular for stricturing phenotype. Therefore, the presence of intestinal stricturing was also registered if stricturing occurred at a different time that the penetrating lesion, so that patients could belong to both categories, if they occur at different time points of CD evolution.
NOD2/CARD15 and PAI-1 genotyping
In all CD patients and control subjects, DNA was collected and genotyped for the three main variants of NOD2/CARD15 gene and for the PAI-1 promoter polymorphism. In brief, Genomic DNA from whole blood samples was isolated by using QIAmp DNA Blood Mini Kit (QIAgen, Hilden, Germany). For NOD2/CARD15 gene, specific sequences containing missense mutation R702W and G908R and frameshift mutation L1007fsinsC were amplified by polymerase chain reaction (PCR), using previously published specific primers and conditions7 The PCR products were purified by using QIAquick PCR Purification Kit (QIAgen, Hilden, Germany), sequenced using an ABI BigDye Terminators v. 1.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and run on an ABI 3100 automatic sequencer. For the PAI-1 promoter polymorphism, we used a protocol previously described based on a PCR technique and endonuclease digestion.27 For DNA amplification, a mutated oligonucleotide, which inserts a site for the Bsi YI enzyme in the product of amplification, was used. After amplification, PCR product was digested by Bsi YI restriction enzyme (Roche Diagnostics, Basel, Switzerland). The digested product was analysed by 4% agarose gel electrophoresis (Metaphor; FMC Bioproducts, Rockland, ME, USA) and visualized under ultraviolet light after staining with ethidium bromide. The expected size of the digested fragments was a single band of 98 base pairs (bp) for the 4G allele and two bands of 77 bp and 22 bp for the 5G allele. The investigators who determined the genotypes were blinded to the clinical characteristics of the patients. CD patients categorized as carrying NOD2/CARD15 variants were anyone with at least one variant on this gene. For the PAI-1 4G/5G polymorphism, we categorized CD patients in 4G/4G and other genotypes (4G/5G or 5G/5G). For the study of combined influence of NOD2/CARD15 variants, we categorized patients in four subgroups: patients carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype, patients carrying only NOD2/CARD15 variants but not the 4G/4G PAI-1 genotype, patients carrying only the 4G/4G PAI-1 genotype but not NOD2/CARD15 variants, and patients not carrying neither NOD2/CARD15 variants nor the 4G/4G PAI-1 genotype.
Categorical variables were compared by using the chi-squared test, applying the Yates correction when necessary. Continuous variables were expressed as mean ± s.d. and median, and compared using anova with Bonferroni as a post hoc test or unpaired Student's t-test, when necessary. For the univariate analysis of factors influencing CD behaviour, the following variables were evaluated: separate and combined NOD2/CARD15 and 4G/4G gene status, gender, CD location, history of appendectomy, family history of IBD, smoking habit and extraintestinal manifestations of IBD. The remaining continuous variables as age at diagnosis, time of follow-up, number of previous flares, number of steroid courses were dichotomized by splitting the samples using median values. Variables reaching a P-value equal or <0.05 in the univariate analysis were entered into the multivariate logistic regression analysis. Probability curves of developing penetrating disease, or stricturing disease after CD diagnosis in patients with non-penetrating or non-stricturing phenotype, respectively, at the time of diagnosis, were calculated according to the Kaplan–Meier method and compared using the log-rank test. Data on patients who did not develop structuring or penetrating behaviour at the end of the study were censored. All P-values were two-sided, and a value of <0.05 was considered to indicate a statistically significant difference. Analysis was carried out using the statview software package (SAS Institute Inc., Cary, NC, USA).
The 170 CD patients included in the study had the following phenotypic features (Table 1): the age at CD diagnosis was 32 ± 14 years, with a median of 28 years. Sixteen patients (9%) had lesions in the upper gastrointestinal tract, 128 (75.2%) in the terminal ileum and 107 (63%) in the colon. Behaviour of CD at the time of diagnosis was inflammatory in 123 cases (72%), stricturing in 16 cases (9%) and penetrating in 31 cases (18%).
|Crohn's disease patients [n = 170]|
|Mean age at diagnosis of CD, years||32 ± 14|
|Male sex, n (%)||91 (53)|
|Location of CD, n (%)|
|Upper gastrointestinal||16 (9)|
|Terminal ileum||128 (75)|
|Behaviour of CD, n (%)|
|Appendectomy, n (%)||21 (12)|
|Smoking habit, n (%)|
|Active smoker||87 (51)|
|Family history of IBD, n (%)||18 (11)|
|Extra-intestinal manifestations, n (%)||55 (32)|
|NOD2/CARD15 gene variants, n (%)||52 (31)|
|4G/4G PAI-1 genotype||47 (28)|
NOD2/CARD15 and PAI-1 gene status
For NOD2/CARD15 gene, at least one mutation was detected in 52 CD patients (30%) compared with 29 (16%) of control subjects (OR, 2.43; 95% CI: 1.42–3.98; P = 0.001). In CD patients, eight were homozygous or compound heterozygous and 44 were simple heterozygous for these mutations; for each one of the three variants of this gene, the frequencies were as follows: 30 (18%) for R702W, 8 (5%) for G908R and 18 (11%) for L1007fsinsC. No homozygote or compound heterozygous was found among control subjects and the rates for each one of the NOD2/CARD15 variants were as follows: 19 (10%) for R702W, seven (4%) for G908R and three (2%) for L1007fsinsC. Distribution of promoter PAI-1 polymorphisms were similar in CD patients and controls. Among CD patients forty-seven (28%) carried the 4G/4G genotype compared with 60 (32%) in control subjects (OR, 0.80; 95% CI: 0.51–1.26; P = 0.3).
Combined distribution of NOD2/CARD15 gene variants and 4G/4G genotype of PAI-1 gene in CD patients and controls is shown in Table 2. As expected, CD patients have a higher rate of carriage of NOD2/CARD15 gene variants evidenced by a slightly higher frequency of carrying jointly both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype (20 of 170 [12%]) compared with controls (10 of 186 [5%]; OR, 2,38; 95% CI: 1.07–5.26; P = 0.04) and a higher frequency of patients carrying only NOD2/CARD15 variants but not the 4G/4G PAI-1 genotype (32 of 170 [19%]) compared with controls (19 of 186 [10%]; OR, 2,04; 95% CI: 1.11–3.84; P = 0.02).
|Combined NOD2/CARD15 and PAI-1 gene status||CD (n = 170) n (%)||Controls (n = 186) n (%)|
|NOD2/CARD15 variants & 4G/4G PAI-1 genotype||20 (12)||10 (5)|
|NOD2/CARD15 variants & non-4G/4G PAI-1 genotype||32 (19)||19 (10)|
|NOD2/CARD15 wt & 4G/4G PAI-1 genotype||27 (16)||50 (27)|
|NOD2/CARD15 wt & non-4G/4G PAI-1 genotype||91 (54)||107 (58)|
Genotype–phenotype correlation at CD diagnosis
Correlations of NOD2/CARD15 variants and PAI-1 genotypes, each with the clinical characteristics of CD at diagnosis, are shown in Table 3. For NOD2/CARD15 variants, a significant association was observed only with ileal disease (OR, 3.36; 95% CI: 1.31–8.58; P =0.008). None of other clinical characteristics at the time of CD diagnosis including CD behaviour was associated with NOD2/CARD15 variants. For PAI-1 promoter polymorphism, there were not clinical characteristics of CD associated to 4G/4G genotype at the time of diagnosis.
|NOD2/CARD15 Variants (n = 52)||NOD2/CARD15 Wild type (n = 118)||4G/4G of PAI-1 gene (n = 47)||Non-4G/4G genotype (n = 123)|
|Median age at diagnosis||27||29||30||27|
|IQR (25–75), years.||20–39||20–35||23–52||22–39|
|<28 years at diagnosis||27 (51%)||58 (49%)||26 (55%)||59 (47%)|
|Male sex||28 (53%)||73 (61%)||28 (59%)||63 (51%)|
|Family history of IBD||4 (7%)||14 (11%)||6 (12%)||12 (9%)|
|Never smoked||24 (46%)||45 (38%)||24 (46%)||45 (38%)|
|Previous appendectomy||7 (13%)||14 (11%)||7 (14%)||14 (11%)|
|Location of CD|
|Upper gastrointestinal||7 (13%)||9 (7%)||4 (8%)||12 (9%)|
|Terminal ileum||46 (88%)*||82 (69%)*||36 (76%)||92 (74%)|
|Colon||30 (57%)||77 (65%)||27 (57%)||80 (65%)|
|Behaviour of CD|
|Penetrating||11 (21%)||20 (16%)||9 (19%)||22 (16%)|
|Stricturing||6 (11%)||10 (8%)||4 (8%)||12 (8%)|
|Inflammatory||35 (67%)||88 (74%)||34 (72%)||89 (74%)|
|Extra-intestinal manifestations||12 (23%)||43 (36%)||12 (25%)||43 (34%)|
Analysis of factors influencing CD behaviour over time
Main CD features over time of our population are shown in Table 4. Behaviour of CD changed over time in a significant proportion of patients. At the end of follow-up, 73 patients (43%) maintained a purely inflammatory phenotype, whereas 71 patients (42%) had a penetrating behaviour. According to the Vienna criteria, 26 patients (15%) had a final stricturing phenotype; when applying the non-hierarchical criteria, 54 patients (32%) had stricturing lesions at any time during follow-up.
|Crohn's disease patients [n = 170]|
|Behaviour over time of CD|
|According to Vienna criteria||26 (15%)|
|Non-hierarchical criteria||54 (32%)|
|IQR (25–75), years||3–10|
|Active smoker||70 (41%)|
|Number of flares per year||1.6 ± 1.5|
|Number of steroid courses per year||0.52 ± 0.60|
|Steroid dependency||59 (35%)|
|Use of Anti-TNF antibody||33 (19%)|
|Surgery requirement||70 (41%)|
Correlations of NOD2/CARD15 and PAI-1 genes, each with CD behaviour over time, are shown in Table 5. For NOD2/CARD15 variants, we did not find a significant association with stricturing phenotype applying the Vienna classification (OR, 1.01; 95% CI: 0.40–2.49; P = 0.9). Instead, a trend towards a higher rate of carrying NOD2/CARD15 variants was observed for the development of stricturing phenotype when using the non-hierarchical criteria (OR, 1.97; 95% CI: 1.0–3.90; P = 0.05). As for the PAI-1 gene variants, a significant association was observed between penetrating behaviour (OR, 3.10; 95% CI: 1.54–6.23; P =0.001), including non-perianal penetrating behaviour (OR, 2.46; 95% CI: 1.16–5.20; P = 0.01), and perianal penetrating behaviour (OR, 2.4; 95% CI: 1.1–5.20l; P = 0.01) with the 4G/4G genotype.
|NOD2/CARD15 Variants (n = 52)||NOD2/CARD15 wild type (n = 118)||4G/4G PAI-1 genotype (n = 47)||Non 4G/4G genotype (n = 123)|
|Total||26 (50%)||45 (38%)||29 (61%)‡||42 (34%)|
|Non-perianal||16 (30%)||24 (20%)||17 (36%)‡||23 (18%)|
|Perianal||15 (28%)||25 (21%)||17 (36%)‡||23 (18%)|
|Vienna criteria||8 (15%)||18 (15%)||6 (12%)||20 (16%)|
|Non-hierarchical criteria||22 (42%)*||32 (27%)||19 (40%)||35 (28%)|
|IQR (25–75), years||3–10||3–10||3–10||3–10|
|Active smoker||29 (55%)||45 (38%)||26 (55%)||48 (39%)|
|Number of flares/year||0.66 ± 0.69||0.77 ± 0.82||0.59 ± 0.70||0.79 ± 0.80|
|Immunosuppressants||37 (71%)†||61 (51%)||29 (61%)||69 (56%)|
|Anti-TNF antibody||12 (23%)||21 (17%)||11 (23%)||22 (17%)|
|Surgery requirement||28 (53%)†||42 (35%)||24 (51%)||46 (37%)|
Combined influence of NOD2/CARD15 and PAI-1 gene variants on CD behaviour is shown in Table 6. A penetrating phenotype was observed in a higher proportion of patients carrying the 4G/4G PAI-1 genotype: in 12 out of 20 patients (60%) carrying both the 4G/4G PAI-1 genotype and NOD2/CARD15 variants, and in 17 out of 27 patients (62%) carrying only the 4G/4G PAI-1 genotype but not NOD2/CARD15 variants, compared with 14 out of 32 patients (43%) carrying only NOD2/CARD15 variants but not the 4G/4G PAI-1 genotype, and 28 out of 91 patients (30%) carrying neither NOD2/CARD15 variants nor the 4G/4G PAI-1 genotype (χ2 = 12,2; P = 0.006).
|NOD2/CARD15 variants||NOD2/CARD15 wild type||P value|
|4G/4G PAI-1 genotype (n = 20)||Non 4G/4G genotype (n = 32)||4G/4G PAI-1 genotype (n = 27)||Non 4G/4G genotype (n = 91)|
|Penetrating||12 (60%)||14 (43%)||17 (62%)||28 (30%)||0.006|
|Vienna criteria||4 (20%)||4 (12%)||2 (7%)||16 (17%)||NS|
|Non-hierarchical criteria||13 (65%)||9 (28%)||6 (22%)||26 (28%)||0.007|
|IQR (25–75), years||3–12||3–9||3–9||4–10|
|<5.6 years of follow-up||10 (50%)||17 (53%)||16 (59%)||42 (46%)||NS|
|Number of flares per year||0.41 ± 0.36||0.91 ± 0.85||0.88 ± 1.0||0.74 ± 0.78||NS|
|<0.96 flares per year||11 (55%)||12 (37%)||10 (37%)||48 (52%)||NS|
|Active smoker||12 (60%)||13 (40%)||9 (33%)||36 (39%)||NS|
|Use of immunosuppressants||15 (75%)||22 (68%)||14 (51%)||47 (51%)||NS|
|Use of Anti-TNF antibody||5 (25%)||7 (21%)||6 (22%)||15 (16%)||NS|
|Surgery requirement||15 (75%)||13 (40%)||9 (33%)||33 (36%)||0.01|
Stricturing phenotype was not associated with any combination of NOD2/CARD15 and PAI-1 gene status, according to the Vienna criteria (Table 6; χ2 = 2, 1; P = 0.5). Interestingly, when applying a non-hierarchical criterion, a higher proportion of stricturing disease was exclusively observed in patients carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype. In this group, 13 out of 20 patients (65%) developed intestinal stenosis, compared with nine out of 32 patients (28%) carrying only NOD2/CARD15 variants but not the 4G/4G PAI-1 genotype, six out of 27 patients (22%) carrying only the 4G/4G PAI-1 genotype but not NOD2/CARD15 variants, and 26 out of 91 patients (28%) not carrying neither NOD2/CARD15 variants nor the 4G/4G PAI-1 genotype (χ2 = 11.9; P = 0.007) (Table 6). The association between carrying jointly both NOD2/CARD15 variants and 4G/4G PAI-1 genotype and stricturing lesions is reinforced by comparison of only this group of patients with all rest of CD patients (OR, 4.93; 95% CI: 1.84–13.24; P = 0.001).
Univariate and multivariate analysis for CD behaviour
In the univariate analysis for penetrating behaviour, including all variables listed in statistical analysis section, male sex (OR, 2.44; 95% CI: 1.30–4.60; P =0.005) and the 4G/4G PAI-1 genotype (OR, 3.10; 95% CI: 1.54–6.23; P = 0.001) were associated with this phenotype. A trend towards a higher rate of carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype (OR, 2.31; 95% CI: 0.89–5.99; P = 0.08) was observed in patients with a penetrating behaviour. In the multivariate analysis, male sex (OR, 2.43; 95% CI: 1.26–4.68; P = 0.007) and 4G/4G genotype of PAI-1 gene (OR, 3.45; 95% CI: 1.41–8.39; P = 0.006) were independent predictive factors for penetrating behaviour of CD.
In the univariate analysis for stricturing behaviour according to the Vienna classification, including all variables listed in the statistical analysis section, ileal disease (OR, 4.61; 95% CI: 1.04–20.43; P = 0.04) and absence of colonic disease (OR, 5.0; 95% CI: 2.04–12.5; P = 0.0005) were associated with this phenotype. A limit statistical significance was observed with extra-intestinal manifestations (OR, 0.33; 95% CI: 0.10–1.00; P = 0.05). In the multivariate analysis, only absence of colonic disease (OR, 4.0; 95% CI: 1.49–11.1; P = 0.006) was an independent predictive factor for stricturing behaviour of CD. However, in the univariate analysis for stricturing disease applying the non-hierarchical criteria, ileal disease (OR, 6.09; 95% CI: 2.04–18.1; P = 0.001), absence of colonic disease (OR, 2.85; 95% CI: 1.44–5.55; P = 0.002), a follow-up longer than 5.6 years (OR, 2.16; 95% CI: 1.11–4.20; P = 0.02) and carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype (OR, 4.93; 95% CI: 1.84–13.2; P = 0.001) was positively associated with this phenotype. A trend towards a higher rate of carrying NOD2/CARD15 variants (OR, 1.97; 95% CI: 1.0–3.90; P = 0.05) was also observed in patients with a stricturing disease. In the multivariate analysis, only ileal disease (OR, 4.19; 95% CI: 1.30–13.5; P = 0.01) and carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype (OR, 5.02; 95% CI: 1.44–17.48; P = 0.01) were independent predictive factors for stricturing phenotype of CD. Carriage of NOD2/CARD15 gene variants was not an independent predictive factor for stricturing behaviour (OR, 0.84; 95% CI: 0.34–2.07; P = 0.7).
Analysis of genetic influence on changes in CD behaviour over time
In the Kaplan–Meier analysis, we observed that carrying the 4G/4G genotype PAI-1 genotype was the only gene status associated with an earlier development of penetrating phenotype (P < 0.0001) (Figure 1). According to the Vienna classification, there was not an association between any gene status and stricturing behaviour over time. However, when applying a non-hierarchical behaviour classification, NOD2/CARD15 variants had a borderline association with an earlier development of stricturing disease (P = 0.03) (Figure 2), and when the probability curves of developing stricturing disease according the four subgroups of NOD2/CARD15 and PAI-1 genes were analysed, we observed that only the subgroup of patients carrying NOD2/CARD15 gene variants and the 4G/4G PAI-1 genotype was associated with an earlier development of stricturing lesions (P = 0.02) (Figure 3). This observation is reinforced in the analysis comparing patients carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype with all rest of CD patients. Patients carrying both genes develop earlier a stricturing phenotype (P = 0.002, figure not shown).
Identification of genes involved in CD susceptibility has been an important advance in understanding the molecular genetics of CD3–5, 28 and has promoted the research on the clinical implications that could directly result from this discovery. This is the first study supporting the hypothesis previously put forward about of key role of gene–gene interactions in the determination of a major feature of CD course such as behaviour phenotype of the disease. Our results suggest that NOD2/CARD15 and PAI-1 genes may influence CD behaviour, which would allow predicting the future course of the disease and ultimately may help in developing preventive clinical strategies.
The phenotypic definitions applied to the behaviour categories are a key issue in the interpretation of the results, as it is shown in the current study. CD is a clinically heterogeneous disorder and there is not a standard classification of CD phenotypes.25, 29–31 We applied two criteria to categorize behaviour phenotype of CD: the hierarchical Vienna criteria,25 and a non-hierarchical criteria,23 with the intention to allow an accurately genotype–phenotype assessment.
Analysis of factors influencing stricturing disease
The Vienna classification is the classic and more recognized classification of CD.25 However, this classification has been criticized in various reports because it does not account for all the observed variations of CD phenotype.32–34 This classification applies a hierarchical criterion for disease behaviour. Thus, patients with both stricturing and penetrating disease are only classified as penetrating, and may conceal a significant number of patients with stricturing disease behind the penetrating phenotype. The other classification used in this study referring to CD behaviour is non-hierarchical or exclusive as is the Vienna classification, with the intention to evaluate each category separately, in particular for stricturing phenotype. In fact, 16% of patients in our study population (28 cases) were categorized as having stricturing disease only when applying the non-hierarchical criteria, but not when applying the Vienna criteria; over a twofold proportion of stricturing disease was found by using the non-hierarchical criteria compared with the Vienna criteria.
Thus, when applying the Vienna classification to our CD population, we did not find an association between stricturing phenotype and NOD2/CARD15 gene variants or any of four subgroups of combined NOD2/CARD15 and the 4G/4G PAI-1 gene status. However, when applying a non-hierarchical criterion, patients carrying both the 4G/4G PAI-1 genotype and NOD2/CARD15 variants had a higher frequency of development of stricturing disease, which appeared earlier in the course of CD. The independent association between stricturing disease with carriage of both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype, but not with carrying NOD2/CARD15 variants in the absence of 4G/4G PAI-1 genotype, supports this observation.
Our results suggest that the higher rate of stricturing phenotype previously observed in CD patients carrying NOD2/CARD15 variants6, 8, 13 would be restricted to the subgroup of patients carrying these variants jointly with 4G/4G PAI-1 genotype, and would explain some of the inconsistencies of genotype–phenotype studies in patients carrying NOD2/CARD15 variants, including the modest higher risk of stenosing CD observed in a meta-analysis (OR = 1.94, 95% CI: 1.61–2.34).14 The basis of the association between carrying jointly, both 4G/4G PAI-1 genotype and NOD2/CARD15 variants, and stricturing disease is not clear. Presence of the 4G/4G PAI-1 genotype may contribute to develop a fibrotic intestinal response in patients carrying NOD2/CARD15 variants. It has been suggested that NOD2/CARD15 variants may determine an increased production of transforming growth factor-β and increased collagen deposit by smooth muscle cells and fibroblasts in the intestine.6 In this setting, presence of a hypercoagulability state favouring ischaemia as the presence of 4G/4G PAI-1 genotype may enhance this abnormal immune response, as ischaemia is a well-known factor inducing the development of fibrosis in various organs35, 36 including the intestine.37 In addition, the 4G/4G PAI-1 genotype is associated with increased expression of PAI-1.15, 16 PAI-1 is a profibrotic protein thought to play major roles in the development of extracellular matrix integrity and may act preventing collagen degradation.20–22 In experimental models using PAI-1 knock-out animals, absence of PAI-1 has been shown to reduce aldosterone-induced collagen deposits in the kidney, and bile duct ligation-induced hepatic fibrosis.38, 39 Therefore, it is conceivable that genetic variants associated with increased PAI-1 levels could predispose to an increase in collagen, fibrosis and stenosis in a particular inflammatory process.
Analysis of factors influencing penetrating phenotype
These results confirm our initial observation that 4G/4G PAI-1 genotype is a key factor implicated in the development of early penetrating disease,17 including both non-perianal and perianal penetrating disease. This implication is not influenced by NOD2/CARD15 gene status. As expected, this result is not changed by the criteria used because of predominance of penetrating phenotype over other phenotypes in the hierarchical Vienna classification.
The 4G/4G genotype is associated with high levels of PAI-1 and a higher risk of thrombosis15, 16, 18, 19 and this could be facilitated by inflammatory states.40 In such conditions, it is possible that microthrombi appear in the intestinal microvasculature, followed by bowel ischaemia and necrosis, thus facilitating the development of deep ulcers, bowel fistulae and perforation. This hypothesis is supported by the demonstration of microthrombi and other vascular lesions in the histological examination of bowel specimens from CD patients.41–43 Moreover, this result is in keeping with previous observations that smoking, another known factor involved in thrombosis and hypercoagulability state, is also involved in the development of penetrating phenotype of CD.9
Given the previous observation that a significant proportion of CD patients have both stricturing and penetrating disease,6 our findings suggest that carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype may be a key determinant of this combined phenotype. Therefore, we suggest that this subgroup may specifically benefit from an aggressive early treatment with the intention to avoid these complications. However, as a relative low number of patients remain in each subgroup according to genetic status or behaviour phenotype, future confirmatory studies will be necessary.
Our results are in keeping with the previous observations showing that NOD2/CARD15 is a susceptibility gene for CD,3–5 and that the 4G/4G PAI-1 genotype in CD patients is not different compared with controls.17. Despite a slight higher frequency of patients carrying both NOD2/CARD15 variants and the 4G/4G PAI-1 genotype in CD patients compared with controls, the border line statistical significance observed does not clearly establish an evidence of epistasis between these two genes in terms of disease susceptibility.
Overall, our study supports the importance of gene–gene interactions as a determinant of CD behaviour, as those patients carrying both 4G/4G PAI-1 genotype and NOD2/CARD15 variants presented a higher frequency of stricturing phenotype, whereas NOD2/CARD15 gene variants per se would not be a determinant of stricturing phenotype. In addition, the current results confirm that the 4G/4G PAI-1 genotype is a determinant of penetrating behaviour irrespective of NOD2/CARD15 gene status. Combined PAI-1 and NOD2/CARD15 genotyping may have a useful clinical application determining subgroups of patients that will develop a severe course of the disease, who may benefit from a more aggressive therapeutic strategy to prevent the appearance of complications. Confirmation of these results in future studies including larger number of patients is required before interventional studies based on these genotyping are proposed.
Declaration of funding interests: this work was supported by Ministerio de Ciencia y Tecnología Grant SAF2002-02211, Instituto de Salud Carlos III Grant C03/02, MARATO TV3 Grant 003110, and Asociación Española de Gastroenterología Grant 2004.