Update on the heritability of inflammatory bowel disease: The importance of twin studies

Initial reports of familial clustering of inflammatory bowel disease (IBD), i.e., families with multiple relatives with Crohn's disease (CD) or ulcerative colitis (UC), suggested that these relatively uncommon disorders might be due to substantial genetic effects. Case series later showed that IBD risk to relatives of IBD probands (first person with IBD identified in a family) far exceeded that expected for relatively uncommon disorders. For example, in one of the first population-based studies of familial IBD, Orholm et al1 observed that the risk of CD to first-degree relatives of CD patients was 10-fold greater and the risk of UC to relatives of UC patients was 8-fold greater than the risk of CD or UC to first-degree relatives of healthy controls. Additionally, CD and UC appeared genetically related, as CD risk for a relative of a UC case was 2-fold greater and the opposite risk was 4-fold greater than that for individuals without an IBD family history. Nonetheless, detractors to theories of IBD having genetic origins noted that UC and CD incidence rates increased greatly—and steadily—each decade since first ascertained, far greater than could be accounted by greater disease recognition, and incidences differed widely by geography.2 Moreover, familial clustering might be due to environmental factors (e.g., predisposing microorganisms shared among relatives). However, long before the discovery of the now more than 40 established genes for CD and UC,3, 4 IBD case reports in monozygotic twins (MZ) suggested that IBD was unusually heritable, with 17 of 20 CD pairs and 5 of 11 UC pairs concordant for disease.5 The first study of IBD among unselected, population-based twins, the Swedish twin registry, provided the proof that IBD, particularly CD, was in large part due to major genetic risk factors—although the MZ twin concordance, with probands identified by discharge diagnoses in public hospitals and diagnoses confirmed by review of medical records, was far less than that suggested by the case reports in the literature.5 In 1988, Tysk et al5 identified 34 MZ and 43 same-sex dizygotic (DZ) twins with at least one diagnosis of CD or UC. Among the 18 MZ twins with a CD proband, eight were concordant for CD; among the 26 DZ twins with a CD proband, only one was concordant (Table 1). In contrast, the observed MZ and DZ concordance for UC was 6.3% and 0%, respectively. As noted by the authors, the CD heritability of liability estimate of 1.0 (95% confidence interval [CI] 0.80–1.0) was higher and the UC estimate of 0.53 (95% CI 0.24–0.82) similar to concurrent estimates reported for diabetes (0.77), hypertension (0.57), or asthma (0.29). IBD Reported in Published Studies of Unselected Twin Cohorts IBD Reported in Published Studies of Unselected Twin Cohorts The remarkably greater MZ compared to DZ concordance for CD can only be explained by genetic factors: while both twin types are born at the same time and had all been reared in the same environment, MZ twins are essentially human clones of each other, having originated from the same zygote and thus having inherited the identical genetic material from both their mothers and fathers. DZ twin pairs are genetically as related as non-twin sibling pairs and will inherit the identical copy of any given gene from their mother or their father only 50% of the time. Additionally, the incomplete concordance among MZ twins suggested that within the same general environment, specific environmental or other acquired risk factors either trigger CD or UC in the affected twin, or perhaps for some factors, “protect” the unaffected twin. For example, in an investigation that utilized Swedish and Danish IBD twin cohorts, smoking was 5-fold protective of UC among discordant MZ twins.6 The Swedish twin registry findings of 1988 were a major incentive to proceed with molecular genetics research to identify genes for CD (which in turn has been one of the leading complex disorders to have susceptibility genes identified), and might have discouraged research to identify genes for UC. The twin cohorts that followed (Table 1) either supported the high CD concordance for MZ twins or observed a substantially reduced MZ twin concordance rate than the original Swedish twin study, whereas UC was observed to be more genetic.7,–10 Had the landmark, Swedish population-based twin study overestimated heritability of CD and underestimated that of UC, or was CD more genetic in the Scandinavian population as suggested by the Danish study that identified probands by questionnaires with a high rate of return mailed to a large twin registry? If anything, a longitudinal follow-up limited to the original Swedish IBD twins suggested that MZ concordance for both CD and UC was underestimated due to failure to diagnose or identify IBD: a co-twin suspected of having CD in the original study was formally diagnosed in 1992 (26 years after his brother was diagnosed with CD in 1966); a co-twin with insufficient evidence to diagnose UC in the original study had sufficient evidence at follow-up; and a co-twin denied UC on the original questionnaire was later found to have a definite diagnosis of UC, made in a Swedish hospital not covered by the original survey.11 One criticism of the twin study was that it had not been age-standardized, and the original Swedish study only included a relatively small window of time to ascertain IBD cases (i.e., from age 6 to 26 years for the youngest participants born in 1958). In this month's issue of Inflammatory Bowel Diseases, Halfvarson provides a major service to IBD epidemiology by reevaluating the original twin cohort for both IBD cases missed during the original study period and extending the time of observation for an additional 15 years (through 2001).12 The original cohort of same-sex twins born between 1886 to 1958 is now less dependent on attained age, as the minimal attained age of the original cohort would be 43 years. Additional twins born between 1959 to 1980 were also evaluated and reported separately. The study was rigorous, as 28% of the 207 probands suspected of having IBD were excluded after review of medical notes due to other diagnoses explaining disease or (among 2%) missing medical records. Twin pairs where one or both twin had died, but MZ or DZ status could be confirmed, were also included. Interestingly, similar to other longitudinal studies, 5% of probands originally diagnosed as CD had the diagnosis later changed to UC and vice versa. Three new MZ pairs concordant for UC, both probands and co-twins, were added to the original cohort. Why the probands or co-twins were not identified in the original report is unclear, as all diagnoses were made between 1958–1973, but could have come from a number of reasons (e.g., diagnoses were made at hospitals originally not part of the inpatient database, cases were identified only after a later admission—perhaps after 1987 when all public hospital discharges were made available, some diagnoses were made as outpatients, previously had incomplete records or a change in diagnosis, etc.). Compared to the original study,5 with the co-twins identified in the longitudinal update of the original discordant pairs,11 and the three new concordant MZ pairs but only 17 new discordant UC pairs found, the MZ concordance for UC was now much greater in the updated study, and is more consistent with UC having a genetic origin—although the MZ pair concordance of 15% was not significantly greater than the DZ concordance of 6%.12 In contrast, for CD, none of the 15 new MZ pairs with a CD proband were concordant. With inclusion (from the prior longitudinal follow up study)11 of the ninth concordant CD pair that was discordant in the original report, the net CD pair concordance among MZ twins born between 1886–1958 birth cohort was now less than two-thirds of the original report (Table 1).12 Halfvarson astutely pointed out that compared to CD among twins in the original report, in the updated report of the same birth cohort, all of the new twins (all discordant) had CD diagnosed on average 7 years later (with a range extended from age 59 to age 79) and with greater delay from onset (average of 4 years from onset in new cases compared to zero years in original cohort), suggesting more mild disease; both older onset and mild disease tends to be associated with a less genetic phenotype. An additional reason for decreased CD concordance in MZ pairs is that the extended period of ascertainment likely introduced more cases with L2 disease (colon-only). Halfvarson noted that L2 disease was much greater among twins diagnosed from 1981 to 2000 as compared to 1961–1980 (45% to 11%, P < 0.004), although this is reported for all cases ascertained including those born from 1886 to 1980. Colon-only disease has been found to be less frequent in familial CD and, in the total cohort, uncommon among the MZ twin probands (1/13) with the co-twin concordant for CD. NOD2 mutations, the highest penetrance CD risk gene identified, is a significant risk factor for ileal involvement.13 For all CD cases, the original Swedish twin report may have been underrepresented with colon-only CD, as this phenotype became incorporated under the rubric of CD only after Lockhart-Mummery and Morson's report in 1964.14 Furthermore, colon-only CD has been steadily increasing in population-based studies, perhaps due to real change in phenotype distribution, aging population (as colonic site is more frequent in later-onset CD), or increased ascertainment following introduction of flexible colonoscopy in the 1970s and its gradual acceptance as a standard procedure in evaluating possible IBD. Although the extended cohort might have ascertained more cases with colon-only CD, relative to other studies, there might be some bias in both the original study and the updated report against ascertainment of mild CD, as probands were only identified by inpatient diagnoses. As shown in Table 1, the updated MZ concordance rates seem to reflect those from all other original twin studies (see last row).7,–10 Two of the other published, unselected IBD twin studies had CD and UC twin pairs identified by questionnaires sent to British and German patient support groups. Expected biases of greater participation from MZ twins were not apparent and high participation rates were estimated. The Danish study from 2000 (with diagnoses clarified in 2005),9 like the Swedish study, investigated a national twin population but identified patients by IBD questionnaires, with potential self-referral bias. However, the participation rate was high, and by not being restricted to inpatients, it may have identified a broader population of patients, especially UC (which less often requires hospitalization and might be reflected in the higher percentage of UC versus CD cases). Nonetheless, the window of time for disease development in the Danish study (like the original Swedish twin study) was relatively short, the population youthful (attained age only 12 to 41), and MZ twin concordance for CD was high. Most likely, CD concordance would be similarly reduced with a longer duration of observation and inclusion of persons that develop non-youthful IBD. Importantly, all of the studies suffer from a relatively modest number of twins, and hence differences among concordance rates might not be considered so reliable. Yet with this “corrected” long-term observed Swedish twin cohort added to the other published twin studies, tempered by taking into account the inherent biases and the small sample sizes of each study, a good mosaic of where CD and UC twin concordance lies is now becoming apparent. It is noteworthy that a larger British study, not included in Table 1 and published in abstract form, that included 249 twin pairs (48 CD and 52 UC MZ twin pairs) found MZ concordance rates (33.3% and 13.4%, respectively) similar to that of the Swedish updated cohort and all combined published studies.15 Rigorous, unselected twin studies can provide exciting insights into the genetic nature of diseases and the updated Swedish study does not disappoint.12 As reported very rarely in case reports, an MZ twin pair having both diagnoses of CD (ileal site) and UC (fulminant colitis causing death) was identified among all 21 IBD concordant twins (13 CD, 7 UC, 1 CD-UC) born between 1886 to 1980.12 Hence, it is now recognized that due to unknown factors a person having a genetic predisposition to CD might also develop UC, but this only happens very rarely—less than 5% in the present study. What might be considered more remarkable is that CD location for 9 of the 13 MZ twin pairs (ileal-L1, colon-only L2, or ileal-colonic L3) was identical. Moreover, among those with discordant location, in two cases the proband was L1 and the co-twin was L3 (thus sharing ileal involvement) and in two cases the proband was L3 and the co-twin was L2, (thus sharing colonic involvement). Thus, all concordant MZ pairs shared overlapping sites of CD involvement. A previous, more extensive report in Inflammatory Bowel Diseases on CD clinical features among twins (that utilized both the Swedish and Danish cohorts), observed that 11 of 17 concordant MZ pairs were concordant for CD site, with only one CD pair having no site overlap (L4 upper GI/jejunal/L3 ileal-colonic).16 Presently, only NOD2 has been established as a risk gene for ileal CD, but a previous study found no overrepresentation of NOD2 mutations among concordant versus discordant twins.9 The near complete overlap of disease site among concordant MZ co-twins suggests that more effort should be put forth to identify genes or sets of genes that control CD disease location. CD behavior showed unexpected concordance among MZ twins in that previous study as well (11/16 after 10 years, P = 0.01). In contrast to CD, extent of UC involvement in the updated Swedish study did not show a tendency for concordance among the MZ twin pairs; only one of four was concordant for extent.12 Three twins had disease limited to proctitis at diagnosis (two remained as proctitis on follow up), none concordant for disease site: this goes against a tendency for some IBD investigators to exclude proctitis in studies of UC to decrease heterogeneity. Another interesting observation was that there were proportionately more males among the MZ twin concordant cases than discordant cases for both CD and UC. Although this did not reach significance, it is worth evaluating in a larger series, as male sex might be a risk factor for more genetic CD: CD in children has been identified as having a greater male preponderance.17 Greater concordance in males than females could point to X-chromosome gene effects. Well-designed twin studies provide major opportunities not only for giving us a more accurate (and in this case very timely) picture of disease heritability, but also for dissecting the nature of a disease. The original Swedish twin cohort data5 has been used to estimate the contributions of the confirmed susceptiblity genes and loci to overall CD and UC heritability,3 hence the data from the now robust, updated Swedish twin cohort should now be used to more accurately estimate CD and UC heritability for these purposes.12 Examination of IBD genetic risk factors have shown that persons who carry multiple CD risk variants in multiple genes (i.e., NOD2, IL23R, ATG16L1, and IBD5) become increasingly at greater risk of CD than those with fewer total number of risk variants.18, 19 We might find that discordant twins have a lower CD “genetic load” than concordant twins. Other explanations for the large excess of concordance in MZ twins (more than twice DZ) include gene–gene interactions and/or increased environmental sharing in MZ twins, and these should be explored.20 Eventually, all IBD susceptibility genes should be evaluated in twins, a relatively simple task with present technology; we might find some IBD genes have very high penetrances or modify disease expression. For example, MZ twins that are homozygous for disruptive mutations in the interleukin 10 receptor subunit proteins, IL10R1 or IL10R2—mutations that result in a rare form of severe CD in infancy—would be expected to be 100% concordant for CD.21 MZ twins both disease concordant and discordant are very useful for evaluating potential epigenetic factors: while MZ twins carry the same copies of all genes, DNA methylation patterns and other epigenetic modifications of genes and their expression differs over time and with different environmental factors. Twin studies are very useful to help dissect environmental factors and microbiome influences, and initial studies that utilized Scandinavian twin cohorts have been performed.6, 22 One challenge to IBD twin studies of specific genetic and environmental risk factors is that IBD is relatively uncommon, and well-powered studies to evaluate modest risk factors might be impractical. Nonetheless, about 0.3% of births are MZ twins and the proportion of DZ twins is even greater. Hence, a concerted effort to identify and enroll IBD twins from multiple countries, with detailed environmental risk factors assessed and biological materials obtained, might yield the necessary study material sufficient to power these types of studies. Genetic analyses of combined cohorts of IBD cases and controls from several countries have recently made possible the discovery of numerous low-risk IBD genes; hence, a similar effort to gather and methodically study large IBD twin cohorts of over 1000 twin pairs might similarly produce a bounty of unique and highly useful genetic and environmental information concerning the nature of IBD.