Prevalence and Clinical Features of Inflammatory Bowel Diseases Associated with Monogenic Variants, Identified by Whole-exome Sequencing in 1000 Children at a Single Center

Abstract

Background & Aims

A proportion of infants and young children with inflammatory bowel diseases (IBD) have subtypes associated with a single gene variant (monogenic IBD). We aimed to determine the prevalence of monogenic disease in a cohort of pediatric patients with IBD.

Methods

We performed whole-exome sequencing analyses of blood samples from an unselected cohort of 1005 children with IBD, 0–18 y old (median age at diagnosis, 11.96 y) at a single center in Canada and their family members (2305 samples total). Variants believed to cause IBD were validated using Sanger sequencing. Biopsies from patients were analyzed by immunofluorescence and histochemical analyses.

Results

We identified 40 rare variants associated with 21 monogenic genes among 31 of the 1005 children with IBD (including 5 variants in XIAP, 3 in DOCK8, and 2 each in FOXP3, GUCY2C, and LRBA). These variants occurred in 7.8% of children younger than 6 y and 2.3% of children 6–18 y old. Of the 17 patients with monogenic Crohn’s disease, 35% had abdominal pain, 24% had non-bloody loose stool, 18% had vomiting, 18% had weight loss, and 5% had intermittent bloody loose stool. The 14 patients with monogenic ulcerative colitis or IBD unclassified received their diagnosis at a younger age, and their most predominant feature was bloody loose stool (78%). Features associated with monogenic IBD, compared to cases of IBD not associated with a single variant, were age of onset younger than 2 y (odds ratio [OR], 6.30; P=.020), family history of autoimmune disease (OR, 5.12; P=.002), extra-intestinal manifestations (OR, 15.36; P<.0001), and surgery (OR, 3.42; P=.042). Seventeen patients had variants in genes that could be corrected with allogeneic hematopoietic stem cell transplantation.

Conclusions

In whole-exome sequencing analyses of more than 1000 children with IBD at a single center, we found that 3% had rare variants in genes previously associated with pediatric IBD. These were associated with different IBD phenotypes, and 1% of the patients had variants that could be potentially corrected with allogeneic hematopoietic stem cell transplantation. Monogenic IBD is rare but should be considered in analysis of all patients with pediatric onset of IBD.

Lay Summary

We performed genetic analyses of children with inflammatory bowel diseases and identified genetic variants that cause this disease. Knowledge of these variants can be used to predict disease progression and select therapy.

Graphical Abstract

Inflammatory bowel diseases (IBD), comprised of Crohn’s disease (CD), ulcerative colitis (UC) and IBD-unclassified (IBD-U), are common chronic relapsing inflammatory conditions of the gastrointestinal tract that affect both children and adults^{1, 2}. Recent studies have demonstrated that IBD is a global disease with a growing prevalence in developed countries and accelerating incidence in newly industrialised countries³. It has been estimated that IBD develops during childhood or adolescence in over 25% of patients⁴. Canadian health administrative data showed that IBD incidence continues to increase rapidly throughout the pediatric age range with the greatest increase in children diagnosed prior to 5 years of age⁵.

Genome wide association studies (GWAS) have identified over 230 risk loci^6–8 explaining only 20 to 25% of genetic heritability in complex adult-onset IBD⁹. As with other complex disorders, host genetics are recognised to play a more predominant role in younger children termed very early onset IBD (VEOIBD; defined as diagnosis before 6 years of age)^{10, 11}. Next-generation sequencing (NGS)¹², including whole exome sequencing (WES), is now used clinically to investigate monogenic causes in complex diseases^{13, 14}, including VEOIBD¹⁵. Utilising WES, a number of genetic disorders associated with VEOIBD^{16, 17} (termed ‘monogenic IBD’ for this study; Table S1 for detailed list of monogenic IBD genes) have been identified. Monogenic IBD genes are categorized into those resulting in epithelial¹⁸ and/or immune^{16, 17} defects and knowledge of tissue specific gene expression and protein function is critical in determining precision treatment approaches¹⁵. Curative allogeneic hematopoietic stem cell transplantation (HSCT) is now considered standard-of-care for select patients with functionally validated immune related monogenic IBD defects¹⁷.

The prevalence of monogenic IBD across the full age-range (0 to 18 years) of pediatric IBD is unknown. Therefore, the primary aim of this study was to use WES to interrogate the currently known monogenic IBD genes to determine the prevalence in a single center cohort of 1005 pediatric IBD patients diagnosed before 18 years of age and identify phenotypic characteristics predictive of monogenic disease.

Methods

Detailed Methods and Clinical Descriptions are provided in Supplemental Material (Supplemental Tables S1-11 and Figures S1–6).

Patient Population

The study was conducted with Research Ethics Board (REB #1000024905) approval at the Hospital for Sick Children (SickKids). The SickKids IBD Center is the only IBD primary care and referral center for the Greater Toronto Area (GTA) and most children in the GTA with IBD are both diagnosed and followed at SickKids. All IBD patients (and families) diagnosed and/or followed at SickKids under 18 years of age were eligible for inclusion regardless of disease severity or age of diagnosis and enrolled over a 13-year period (2003–2015; Tables 1, S2–S4 for cohort characteristics). Inclusion criteria: diagnosed with IBD between 0 and 18 years of age and followed at the Hospital for Sick Children. Exclusion criteria: IBD with known chromosomal abnormalities, diagnosed syndromic disease, previous diagnosed primary immunodeficiency, referral for a second opinion from other regions of Canada, diagnosed with other forms of monogenic intestinal disease, and refusal to consent for genetics for any reason.

Table 1:

Phenotypic Characteristics of Probands in the Sequenced Cohort and the Monogenic Cohort.

	Total pediatric IBD Cohort	Non-monogenic pediatric IBD Cohort	Monogenic pediatric IBD Cohort	Odds Ratio (95% CI)	Bonferroni p Value
Total number (n=)	1005	974	31
Median age at diagnosis (IQR)	11.96 (8.96 – 14.21)	12.04 (9.05 – 14.25)	10.83 (3.45 – 12.53)		0.12
Age at diagnosis 0–1.9 years (Infantile)	2.9%	2.5%	13%	6.30 (1.98 – 20.08)	0.020
Age at diagnosis 2–5.9 years (VEOIBD)	11.2%	10.7%	22%	2.62 (1.06 – 6.46)	0.405
Age at diagnosis 6–9.9 years (EOIBD)	17.8%	18.3%	10%	0.67 (0.20 – 2.32)	1
Age at diagnosis 10–17.9 years	68.1%	68.5%	55%
Sex (% male)	60%	59.6%	71%	1.65 (0.78 – 3.82)	1
Family history of IBD	32%	31.7%	42%	1.55 (0.74 – 3.19)	1
First degree family history of IBD	14%	14%	7%	0.41 (0.07 – 1.40)	1
Family history of autoimmune disease	7%	6.3%	26%	5.12 (2.07 – 11.48)	0.002
Disease type	CD 59% UC/IBD-U 41%	CD 60% UC/IBD-U 40%	CD 55% UC/IBD-U 45%
Any EIM	11%	9.3%	61%	15.36 (7.31 – 33.52)	<0.0001
>1 EIM	3.3%	2.6%	26%	13.20 (5.13 – 31.51)	<0.0001
Progression to biologic therapy	38%	38.2%	32%	0.77 (0.34 – 1.62)	1
Progression to surgical therapy	9.8%	9.2%	26%	3.42 (1.40 – 7.57)	0.042

Comparisons are made between monogenic and non-monogenic groups. Odds ratio estimates and their 95% confidence intervals were computed using logistic regression models (refer to supplemental materials). Figures in bold indicate statistical significance (p<0.05 after Bonferroni correction). CI, confidence interval; EIM, extra-intestinal manifestation; IQR, interquartile range.

DNA Extraction

Genomic DNA was extracted from peripheral venous blood samples collected in EDTA. DNA concentration was estimated using the Qubit 2.0 Fluorometer and a 260:280 ratio calculated using a NanoDrop spectrophotometer. The average DNA yield obtained was 150 ug/ml, and approximately 2 ug of each patient DNA was extracted for next generation sequencing.

Whole Exome Sequencing (WES)

WES was carried out using high-quality genomic DNA that was mechanically fragmented by adaptive focused acoustic technology to a mean size of approximately 150 base pairs. The sheared libraries were prepared for exome capture using a custom-designed, highly automated approach to produce a library appropriately indexed for pooled exome capture and sequencing. The exomes were captured with the NimbleGen VCRome 2.1 reagent. The probe set targets 42 Mb of DNA covering the Vega, CCDS, and RefSeq gene models, microRNAs, and some regulatory regions. Captured samples were PCR amplified to ensure that exome enrichment and genome de-enrichment were successful. Samples that pass quality control were sequenced on the Illumina HiSeq 2500 platform using paired-end 75 bp reads and two indexing reads¹⁹.

Whole Exome Sequencing and Bioinformatics Pipeline

WES was performed in collaboration with the Regeneron Genetics Center (RGC). Regeneron Pharmaceuticals did not contribute to the data analysis, interpretation of findings, or writing of the manuscript. Genes were selected based on known association with monogenic IBD (outlined and referenced in Table S1). Genes identified as risk or without validated monogenic association were not included in this analysis. The data files were processed using the Care4Rare (C4R) bioinformatics pipeline at SickKids, which is comprised of three components; alignment, variant calling and annotation. Sequencing reads were aligned to human reference genome (GRCh38/hg38) using BWA-mem (Burrows-Wheeler Aligner, ver. 0.7.12) followed by indel realignment using Genome Analysis Toolkit (GATK - ver. 3.5), marking PCR duplicates using Picard and base recalibration by GATK. A BED file corresponding to the library preparation capture kit was used in the pipeline to limit the analysis to exonic intervals. The following variant callers were run on the BAM files of each family to produce family-based VCF files: GATK HaplotypeCaller²⁰ (ver 3.5), Vardict²¹ (ver. 1.4.6), Varscan²² (ver. 2.3.9), Samtools²³ (ver. 1.3), and Freebayes²⁴(ver. 1.0.0). An ensemble approach using bcbio.variation.recall (https://github.com/bcbio/bcbio-nextgen) was then used where only variants called by two or more variant callers were retained. Annovar (annovar/2016.02.01) was used to annotate the VCF files (Table S6A)²⁵. These files were subsequently normalized with the vt tool²⁶ and further annotated with Ensembl VEP²⁷ (ver. 82). VCF²⁸ anno added the annotations in Table S5A–B. The vcf2db program (https://github.com/quinlan-lab/vcf2db) was then used to generate GEMINI²⁹ (-compatible databases from family-based VCF files. GEMINI was used to query the different inheritance models in Table S5C using the filters in Table S5D. Further quality assessment measures are outlined in the Supplemental Materials (Figure S1A–C, S2).

Coverage Analysis

In order to assess the read depth of the exonic intervals for the 67 VEOIBD genes, the average coverage values from the sample_interval_summary file, generated by the GATK DepthOfCoverage tool, were extracted for all 2,309 samples. Using custom Python and R scripts, coverage box-plots were generated for each VEOIBD exonic interval (Figure S1C), median read depth 78.0.

Sequencing Quality Assessment

Sequencing Quality Assessment is outlined in Figures S1–2. Transition to transversion (Ts/Tv) ratios per sample were obtained using the bcftools³⁰ (ver 1.6) stats tool. The number of variants called according to functional consequence estimated using Ensembl VEP version 75. Functional variants included: stop gained, splice donor and acceptors, frameshift, missense, inframe insertion and deletions, initiator codon and splice region variants. The distribution of total number of variants per sample along with the number of heterozygous and homozygous variants were calculated. Sex analysis was performed, by investigating the number of homozygous and heterozygous variants in X chromosome for both male and females. When assessing the number of heterozygous variants in X chromosome, five outliers were identified; one was incorrectly coded in the database and four were subsequently excluded from further analysis. The coefficient of inbreeding (COI) was calculated using plink (version 2.0) software. We applied the “Plink –make-king” tool to the project-level merged VCF file containing all the variants of the entire dataset. Only autosomes were considered by the program. The spouse pairs showing a high COI value (>0.7) were reviewed for consanguinity. The genotype quality (GQ) of variants were extracted from merged VCF file using a custom script and the distribution was plotted as a box plot.

Sanger Sequencing

Selected variants predicted to be pathogenic and assessed as deleterious by annotation tools outlined above were verified by Sanger sequencing in probands and relatives if sufficient DNA was available. Primers were designed using primerBLAST and PCR was performed in our research laboratory. The Sanger sequencing service was provided by ACGT (Toronto).

Genotype Phenotype Analysis

Following high quality filtering each patient deemed to have a protein coding or splice variant that was high quality and rare (maf<0.01) were reverse phenotyped. This meant that clinical data was extracted from the database with deep phenotyping performed on any outstanding clinical details via access to an electronic medical record system. Immune and pathology work up was clinically driven, and results were accessed on the electronic medical system.

Paris Classification

Accurate phenotype classification is essential in determining the utility of genotype-phenotype correlation. The Paris Classification (Table S4) was developed by a group of experts in pediatric IBD in 2009³¹. This was an update on the previously published Montreal Classification of IBD³². The Paris Classification considers age at diagnosis (A1a<10years, A1b 10–17years), location (L1 distal 1/3 ileum +/− limited cecal disease; L2 colonic, L3 ileocolonic, L4 isolated upper disease) and behaviour of disease (B1 non-stricturing, nonpenetrating; B2 stricturing, B3 penetrating, p perianal disease modifier), along with consideration of linear growth impairment (G0 no evidence of growth delay, G1 growth delay). This aims to capture the more dynamic features of the pediatric IBD phenotype resulting in uniform standards for defining IBD phenotypes.

Statistical Methods

Descriptive statistics were provided with medians and interquartile range (IQR) for continuous variables. Mann-Whitney U test was used for non-normally distributed continuous variables. Chi-squared or Fisher’s exact test was applied for categorical variables. Categorical variables were compared by calculating an odds ratio (OR) using logistic regression models. For the analysis of age in pediatric IBD cohorts, the age group 10–17.99 years was used as the baseline level, to which other age groups were compared using logistic regression. Results were considered statistically significant when p<0.05 after Bonferroni correction for multiple testing. Statistical analyses were performed by using the SPSS 22.0 software (Chicago, IL), as well as the R function glm and R package nnet for logistic regression modeling.

Results

Cohort Characteristics

In total, 2305 (99.8%) participants (1005 pediatric IBD patients, and 1300 parents and siblings) were analysed (4/2309 individuals failed quality control; Figure 1A, Table S2). Forty-nine percent of pediatric IBD patients were part of complete trios (patient and both of their parents) including 26 quads (trio plus sibling; Figure 1B) and 77% of patients had at least one first degree family member sequenced (including 105 affected first-degree family members; Table S2–3). Pediatric IBD patients had a 1.5:1 male to female ratio and were diagnosed with CD (601, 60%) and UC/IBD-U (404, 40%) (Table S4). The median age at diagnosis was 11.96 years and the median age at symptom onset was 10.65 years (Table 1; see age distribution of the patients in Figure 1B, Table S4). Principal component analysis of the cohort demonstrated broad ethnic diversity including European, East Asian, or South Asian ethnicity (Figure 1C, Figure S2, Table S5E).

Figure 1: Flowchart of WES pediatric IBD study.

A) Flowchart of WES in pediatric IBD study. 1,644,648 variants were called. Following variant prioritization 1,124,679 were ExAC MAF < 0.01 and 379,588 high/med impact severity. Variants then underwent GEMINI Pedigree Analysis and inheritance modelling. In the known monogenic IBD genes 5 were autosomal dominant (AD); 6 autosomal recessive (AR); 9 compound heterozygotes (CH) (note CH in Table 2 is denoted by AR a/b); and 11 X-linked recessive (XL) inheritance.

B) Familial inheritance-based analysis by age group of the sequenced cohort. Bar graph displays the sequencing of the cohort – based on singletons (probands alone); incomplete trios; complete trios; complete quads by age group. A ‘trio’ is defined as a sequenced proband and both parents. An ‘incomplete trio’ is a proband and any relative. A ‘complete quad’ is a proband, both parents and a relative.

C) Principal Component Analysis. Ethnic diversity demonstrated amongst the cohort.

Identification of Rare Damaging Variants in Monogenic IBD Genes

GEMINI analysis and initial filtering of the WES data was based on rare, protein coding variants, deleterious predictions and damaging scores (including CADD score >15³³) and a secondary manual filtering based on confirmatory inheritance pattern, segregation, concurrence with clinical features associated with phenotypes of known genetic disease, as described in Table S1, and pathogenicity was based on the ACMG classification³⁴. In 31 (3%) of the 1005 pediatric IBD patients, we identified 40 distinct rare damaging variants in 21 of the known 67 monogenic IBD genes (Table 2). Of the 31 patients with monogenic IBD variants, 23 (74%) patients were sequenced as trios, 4 (13%) as incomplete trios, and 4 (13%) as singletons. All 40 predicted pathogenic monogenic IBD variants were orthogonally validated using Sanger sequencing (data not shown) and transmission was confirmed to be either autosomal dominant (AD), autosomal recessive (AR; all bi-allelic), or X-linked recessive (XL). Functional de novo variants in the IBD monogenic genes were not identified in any patient in this cohort. Functionally, 67% were missense variants and 32% were predicted loss-of-function (LOF) alterations, either stop-gained, frameshift, splice-site, or inframe indels (Table 2, Figure 2A). As shown in Figure 2B, among the 31 children harboring variants in known monogenic IBD genes, those most represented were XIAP (5/31, 16%), DOCK8 (3/31, 10%), ARPC1B, FOXP3, GUCY2C and LRBA (2/31, 6%). Overall, 3% of the 1005 pediatric IBD patients were suspected to have disease-causing variants in monogenic IBD genes.

Table 2:

Variants identified among monogenic IBD genes.

Biological Category	Patient	Sex	Age at Diagnosis	Chrom	Gene	Inheritance model	aa mutation	Impact	ExAC maf	CADD 1–3 Phred score	Family History	Reported Inheritance	Causal Evidence
Epithelial barrier & response defects	1	M	15	chr2	ALPI	AR a	A360V	missense	0.0006363	26.8	+	AR	G, F, P
						AR b	Q439X	stop gained	0.0001295	29.7
	2	F	11.91	chr3	COL7A1	AR a	R1696C	missense	0.0003	28.7	−	AR/AD (milder)	P
						AR b	R523W	missense	None	25.5
	3	M	12.3	chr12	GUCY2C	AD	G549S	missense	0.000045	33	+	AD	N/A
	4	F	15.1	chr12	GUCY2C	AD	F525L	missense	0.000067	25.5	−	AD	N/A
	5	M	11.38	chr3	SLCO2A1	AR a	R314W	missense	0.0000164	26.2	−	AR	P
						AR b	M 539V	missense	0.0003	27.1
	6	F	0	chr2	TTC7A	AR a	E71K	missense	0.0000082	34	−	AR	G, F, P
						AR b	Q526X	stop gained	0	36		AR
T and B cell differentiation defects	7	M	2.75	chr7	ARPC1B	AR	A105V	missense	0	25.1	−	AR	G, F, P
	8**	M	3.5	chr7	ARPC1B	AR	V91WfsX121	frameshift	0	none	−	AR	G, F, P
	9	M	15.52	chrX	BTK	XL	N484K	missense	0	23	−	XL	F, P
	10	M	5.5	chrX	DKC1	XL	T49M	missense	0	26	−	XL	G, F, P
	11	F	12.38	chr9	DOCK8	AR a		splice	0.0000411	35	−	AR/AD (milder)	G, P
						AR b	S97L	missense	0.0005	26.4
	12	F	14.73	chr9	DOCK8	AR a	R947C	missense	0.0007	22.6	+	AR/AD (milder)	P
						AR b	R532Q	missense	0.0000414	26.4
	13	F	14.65	chr9	DOCK8	AR a	R59H	missense	0.0021	26.7	+	AR/AD (milder)	P
						AR b	Y19C	missense	0	22.1
	14	M	1.63	chr4	LRBA	AR	V737I	missense	0.0002	23.2	−	AR	P
	15**	M	9.33	chr4	LRBA	AR*	E1916X	missense	0	None	−	AR	G, F, P
	16	F	2.5	chr2	STAT1	AD	V389A	missense	0	19.69	−	AD	G, F, P
Hyper- & Auto-inflammatory disorders	17	M	14	chr10	HPS1	AR a	E9D	missense	0.0016	20.4	+	AR	p
						AR b	Y81F	missense	0.0000695	23.4
	18	M	12.75	chr1	PIK3CD	AD	V616A	missense	0.0000083	27.1	+^	AD	N/A
	19**	M	8.4	chrX	SH2D1A	XL	M1T	start Met loss	0	23.4	−	XL	G, F, P
	20	M	11.86	chrX	XIAP	XL	R29K	missense	0	25.7	+	XL	P
	21	M	9.1	chrX	XIAP	XL	S278X	stop gained	0	37	+	XL	G, P
	22	M	10.83	chrX	XIAP	XL	A216T	missense	0	25.6	+	XL	P
	23	M	10.67	chrX	XIAP	XL	ΔE349	inframe deletion	0.0004	None	−	XL	G, P
	24**	M	0	chrX	XIAP	XL	V298fsX306	frameshift	0	None	−	XL	G, F, P
	25**	M	3.4	chrX	CYBB	XL	A84A	splice	0.000012	19.34	−	XL	G, F, P
Regulatory T cells and immune regulation	26	M	12	chr2	CTLA4	AD		splice	0	28	−	AD	G,, P
	27	M	3.05	chrX	FOXP3	XL	L198P	missense	0	24.4	+	XL	F, P
	28	M	12.68	chrX	FOXP3	XL	H121Y	missense	0.0000782	21	+	XL	F, P
	29**	F	0.17	chr21	IL10RB	AR	W3X	stop gained	0	27.6	+^	AR	G, F, P
Others	30	F	4.42	chr2	HSPA1L	AR	K495EfsX13	frameshift	0.0003	None	−		G, P
	31	M	12.26	chr1	MASP2	AR a	D244N	missense	0.000068	28.7	+	AR	N/A
						AR b	W513C	missense	8.24E-06	24.7

Age at diagnosis (years); Chrom/chr - chromosome; Inheritance model; AD - autosomal dominant; AR - autosomal recessive; ARa/ARb - compound heterozygous; XL - X-linked recessive; aa - amino acid; CADD - Combined Annotation Dependent Depletion; Family History: + - yes, − no

^{^}indicates first degree.

Causal Evidence: G - Genetic, F - Functional, P - Pathology.

^**- Patient had Allogeneic Hematopoietic stem cell transplant (HSCT).

Figure 2: Characteristics of identified monogenic IBD population.

A) Variant Type. Variant types identified using WES in the pediatric IBD cohort.

B) Gene Variants identified on WES analysis. Graph demonstrates gene variants with the most common highlighted in bold.

C) Age at Diagnosis as per Variant Class. Epithelial barrier response defects; T- and B- cell differentiation defect; Hyper- and Auto-inflammatory disorders; Regulatory T-cells and immune regulation; and other. Colors represent the color used in Table 2.

D) Age at diagnosis of IBD and disease phenotype. Crohn’s disease (CD), Ulcerative Colitis (UC), IBD-unclassified (IBD-U). Graph displays monogenic pediatric IBD variant diagnosed in VEOIBD age group (35%) indicated by blue line and over 10 years of age (55%) (Paris 1b) indicated by red line.

E) Odds ratio analysis of phenotypic features of identified monogenic IBD. Odds ratios and their confidence intervals were computed using logistic regression models (see Supplemental Materials). Age (in years); EIM = extra-intestinal manifestation; FHx – family history; AI = autoimmune. Each horizontal line represents the 95% confidence interval for odds ratio (red dot). Dashed line indicates expected value of 1.0. *represents statistical significance (p<0.05 after Bonferroni correction).

F) Age at Hematopoietic stem cell transplant. Age of patient at time of HSCT (in years) with identified monogenic IBD variant in ARPC1B (P8), IL10RB (P29), LRBA (P15), SH2D1A (P19), XIAP (P24). Green line refers to age of diagnosis of IBD. Red line refers to age of monogenic diagnosis. Blue line refers to age of HSCT.

Functional Validation

While in silico pathogenicity of variants identified using WES was based on CADD score³³, inheritance pattern, impact, frequency, and ACMG classification³⁴, we attempted to further validated each variant through retrospective clinical assessment of the probands, functional testing, and/or pathological examination of biopsy samples (Tables 2 and S6–8, and Figures S3–6 for detailed review of each patient). All patients had clinical evidence supporting the known monogenic disease phenotype (clinical features^{5, 16} are outlined in Table S1 and detailed for each patient in Table S8), although some patients had milder or incomplete forms of the disease as previously demonstrated in patients with monogenic-forms of IBD¹⁵. The majority of patients had multiple levels of support including 16 patients with genetic support of causality with either known ClinVar pathogenicity and/or LOF variants. Functional testing including protein expression, immunological testing, and biochemical assays were carried out in 14 patients, including 3 patients without further genetic evidence. To provide additional support for the patients without available clinical testing, samples were examined for both known histological features of disease³⁵ and protein expression/localization based on known RNA/protein expression defined by Protein Atlas (www.proteinatlas.org). We identified either pathological features associated with monogenic disease and/or altered protein staining in all 27 patient samples examined including 8 patients without other supporting genetic or functional evidence (Figure S3–6). Therefore, only 4 patients with variants in 3 genes (GUCY2C, PIK3CD, and MASP2) did not have supporting evidence as these 3 genes had neither clinical functional tests nor validated antibodies available to examine pathology. Together this further supports the in silico prediction in 27/31 patients and the role of these genes in the development of IBD.

Implications of Genetic Diagnoses

Demographic and phenotypic clinical characteristics of the probands with putative monogenic IBD are summarized in Table 3 and Table S8 for detailed phenotyping. Figure 2C–D and Table 1 showed that in the monogenic group the median age at diagnosis was 10.83 years and median age of symptom onset of 9.69 years and 20/31 (64%) patients were diagnosed at greater than 6 years of age. In the monogenic IBD variant group, 71% were male, 17 (55%) were diagnosed with Crohn’s disease, and 14 (45%) were UC/IBD-U (Figure 2D). The presenting clinical features for monogenic CD patients (n = 17) were abdominal pain (35%), non-bloody loose stool (24%), vomiting (18%), weight loss (18%) and intermittent bloody loose stool (5%) (Table 3). The monogenic UC/IBD-U patients (n = 14) were diagnosed at a younger age and the most predominant presenting clinical feature was bloody loose stool (78%) (Table S3). Features associated with monogenic disease in comparison to the remaining pediatric IBD cohort were age of onset of disease < 2 years (OR 6.30, P = 0.022), family history of autoimmune disease (OR 5.12, P = 0.002), extra-intestinal manifestations (EIM) of IBD (OR 15.36, P < 0.0001) and surgery (OR 3.42, P = 0.046) (Figure 2E and Table 1).

Table 3:

Phenotypic Features of Identified Monogenic IBD Cohort

Monogenic cohort (n=31)	Crohn’s disease	Ulcerative Colitis/IBD-Unclassified
Disease class (n=)	17	14
Sex (% male)	82%	57%
Median age at diagnosis (years) (IQR)	12 (9.5–12.7)	3.9 (0.9–11.5)
Age at diagnosis 0–5.9 years (VEOIBD)	2 (6%)	9 (29%)
Age at diagnosis 6–9.9 years (EOIBD)	2 (6%)	1 (4%)
Age at diagnosis 10–17.9 years	13 (42%)	4 (13%)
Most common presenting feature	Abdominal pain (35%)	Bloody loose stool (78%)
Extra-intestinal manifestation	76%	42%
Family history - IBD	58%	21%
Family history - autoimmune disease	29%	21%
Personal history – autoimmune disease	23%	35%
Consanguinity	0	14%
Ethnicities	70% Caucasian 18% South Asian 6% East Asian 6% Mixed	37% Caucasian 21% African 21% Mixed 14% East Asian 7% South Asian
Abnormality on immune work up	18%	28%
Serology	17% positive ASCA 6% positive ANA	36% ANCA positive 14% ANA positive
Endoscopy	35% L3 29% L3L4a 12% L3L4b 12% L2 12% L1	42% E4 28% E3 7% E2 23% limited due to disease severity
Histopathology	35% granulomatous inflammation	28% apoptosis
Therapies	29% Surgery 53% Biologic agent 12% HSCT 6% Other	21% Surgery 14% Biologic agent 28% HSCT 37% Other
Outcome		1 death

Phenotypic features of monogenic IBD cohort (n=31); 17 (54%) with monogenic Crohn’s disease, 14 (45%) with monogenic Ulcerative Colitis/IBD-U. ASCA – anti-Saccharomyces cerevisiae antibodies. ANCA – anti-neutrophil cytoplasm antibodies. ANA – anti-nuclear antibody.

In total, 17/31 monogenic IBD patients (> 1% of total cohort) had variants in genes known to be amenable to allogeneic stem cell transplant (ARPC1B, IL10RB, LRBA, SH2D1A, XIAP, CYBB, CTLA4, STAT1, BTK, and FOXP3; Table S9). All 17 patient/families via their most responsible-physicians (including those transitioned to adult care) were informed of the genetic variants and patients are undergoing clinical genetic validation and counselling where appropriate as some patients had milder disease and transplant may not be a recommended or preferable treatment option. Six patients (19% of monogenic IBD cohort) already had allogeneic HSCT (Table 2 and S8: Patient 8 - ARPC1B, Patient 29 - IL10RB, Patient 15 - LRBA, Patient 19 - SH2D1A, Patient 24 - XIAP; Patient 25 - CYBB; all unpublished; Patient 25 was transplanted elsewhere and not included in the subsequent analysis). The median age at transplant in these 5 patients was 11.5 years (IQR 0.45–12.62) and there was often a delay between onset of disease and ultimate genetic diagnosis and HSCT (Figure 2F). Of note, monogenic epithelial IBD or combined epithelial-immune defects (i.e. TTC7A deficiency and the one death in this study population; Patient 6³⁶) may not respond to HSCT or biologic therapies and may present specific therapeutic challenges³⁷.

Discussion

In this single center cohort study of 1005 pediatric IBD patients we utilized WES to determine a 3% prevalence of damaging variants in genes linked to monogenic IBD. A number of studies have shown an estimated prevalence of monogenic IBD between 0 to 70%^38–44 (reviewed in Table S10). However, these studies are difficult to compare as they examine only subsets of monogenic IBD genes and use a number of sequencing methodologies including genetic panels, WES, and mixed methodologies. Moreover, these studies may have significant selection bias as the highest rates of monogenic disease are within cohort studies of patients referred with severe disease and/or very young age of onset increasing the likelihood to identify monogenic disease. Prior to our study, few studies have examined older pediatric patients in a systematic way. In Toronto, Canada, there are very few Community Pediatric Gastroenterologists; therefore, the vast majority of pediatric IBD patients in the Greater Toronto Area (catchment area population of approximately 6 million) are diagnosed and followed until 18 years of age at SickKids and makeup the cohort described here. This patient cohort is a major strength of this study as it is a large heterogeneous, multi-ethnic, well-characterised, unselected cohort of children diagnosed from a single pediatric IBD center and patients/families were enrolled regardless of age of diagnosis and disease severity. Another strength of our pediatric IBD cohort was that the majority of patients had at least one family member sequenced allowing for family-based genetic analysis.

The frequency of monogenic variants in VEOIBD (7.8%) reinforces that exome sequencing should become part of standard-of-care for this group of patients diagnosed with IBD, especially children diagnosed under two years of age. Pediatric gastroenterologists may screen for monogenic forms of IBD in very young children; however, previous studies have not ascertained the prevalence of these genes across the entire pediatric age range. As described, most monogenic IBD studies have focused only on very young children or young children with the most severe forms of disease^{38, 40, 45, 46}, while this study examined an unbiased cohort of patients and extends the age of onset of monogenic IBD throughout the pediatric age range. Another key finding of our study was that 64% (20/31) of the monogenic pediatric IBD patients presented after 6 years of age. We found an unexpected prevalence of 2.3% of monogenic variants in all children aged 6 years and older. For these older children the phenotypic features including extra-intestinal manifestations of IBD and family history of autoimmune disease may be used to select patients for consideration of whole exome sequencing analysis.

For each monogenic disorder associated with chronic IBD, the bowel inflammation often has variable penetrance and is only one component of a disease that may manifest with a wide spectrum of phenotypes^{15–18, 47–49} (Table S1). Detailed phenotyping of each patient with a monogenic IBD variant (Tables S8A–U, Figures S4–6) suggests that within each group of genes the phenotypic variation is likely due to the genetic heterogeneity of disease causing variants, genetic disease modifiers, and undetermined environmental factors. Interestingly, Huang et al. demonstrated that patients with Chronic Granulomatous Disease (CGD) who developed IBD had a higher polygenic risk score (PRS) for IBD GWAS variants when compared to CGD patients without IBD⁵⁰. We similarly developed a polygenic risk score and compared non-IBD controls (n=7492), Toronto cohort patients with monogenic IBD variant carriage (n=31), and Toronto cohort IBD patients without monogenic IBD variant carriage (n=974). However, we did not identify any significant differences in PRS between Toronto cohorts maybe due to the small number of patients with each type of monogenic disease (data not shown). It is interesting to speculate that IBD GWAS risk variants coupled to environmental factors are driving the IBD presentation in some patients with monogenic forms of disease. Overall, our findings suggest a wide heterogeneity in monogenic IBD clinical presentations with earlier age of onset, a family history of autoimmune disease, extra-intestinal manifestations of IBD, and surgery as indicators of monogenic disease (Table 1, Figure 2E). However, these features are common in pediatric IBD and detailed phenotyping of the monogenic cohort may have resulted in an over-representation in these patients.

Special consideration should be given to specific gene expression of epithelial versus immune monogenic forms of IBD (Table 2, Figure 2C). The intestinal epithelial barrier is composed of a layer of columnar cells that function as a gateway between the gut lumen and the lamina propria. Of those epithelial defects purported to be associated with monogenic IBD¹⁸, we identified ALPI, COL7A1, TTC7A, GUCY2C, and SLCO2A1 in our cohort. Variants in these genes may cause perturbations in the epithelial barrier leading to immune dysregulation resulting in IBD¹⁸. This group of patients present a clinical challenge as there are no specific epithelial treatments available that improve the epithelial barrier dysfunction and HSCT is not a viable treatment option. However, a recently published preclinical study has identified Leflunomide as a potential therapy for TTC7A-deficiency⁵¹.

We also identified monogenic immune IBD genes involved in a number of cellular processes outlined in Table 2 and Figure 2C and each patient discussed in detail in Figures S4–6. For example, we identified variants in T- and B-cell differentiation including BTK (Figure S4) and DOCK8. DOCK8 has diverse roles in the immune system including regulation of the actin cytoskeleton and can present from infancy to adulthood with variable symptoms including severe infections, atopy, autoimmunity, cancer, and IBD⁵². In this study, 3 patients had bi-allelic DOCK8 variants with clinical features consistent with DOCK8-deficiency (Table S8H, Figure 3G) including eczema and food allergy although they did not have truncating variants classically associated with this disease. FOXP3 is a transcription factor that is specifically expressed in regulatory T-cells which play a critical role in T-cell tolerance^{53, 54}. Variants in FOXP3 can lead to X-linked, immune dysregulation, polyendocrinopathy, and enteropathy, IPEX syndrome. In this study, 2 male patients with chronic colitis and other autoimmune and extra-intestinal features were identified (Table S8I, Figure S5). Variants in XIAP cause an X-linked recessive disorder with a widely reported age of onset and diverse phenotypes including infantile onset and predisposition to hemophagocytic lymphohistiocytosis (HLH) and lymphoproliferative syndrome (XLP). Zeissig et al reported XIAP variants in 4% of all male pediatric CD patients⁵⁵. Here we demonstrated that 1% (5/391) of male pediatric CD patients had XIAP variants of which one patient with a V298fsX306 XIAP variant was successfully transplanted and currently has no active disease (Patient 24, Tables 2 and S8U, Figure S6A).

A major difficulty in utilizing genetics in clinical care of children with IBD is the lack of standardized functional testing. This is a critical step in precision medicine, especially when recommending major alternative treatment strategies such as allogeneic bone marrow transplant in patients with primary immunodeficiencies (PID) genes associated with monogenic IBD¹⁶ or palliation in patients with severe forms of TTC7A-deficiency⁵⁶ or PLVAP-deficiency⁵⁷. There are a few genes where the protein product can be easily assayed. For example, in IL10R deficiency STAT3 phosphorylation can be measured after IL10 stimulation⁵⁸ and reactive oxygen species (ROS) can be measured in Chronic Granulomatous Disease, although disease-causing variants in NCF4 may have normal ROS production⁵⁹. While for others genes, only experimental biochemical assays are available in selected research labs for example XIAP⁶⁰, TTC7A³⁶, and ARPC1B⁶¹. Protein expression of a number of monogenic IBD genes associated with primary immunodeficiencies can be measured using flow cytometry-based assays, such as LRBA, FOXP3, XIAP¹⁷; however, missense variants which may result in normal gene expression and deleterious protein function will not be identified using this methodology. In an attempt to further validate causative variants, we utilized a combination of genetic, functional and/or pathological approaches; however, further standardized testing is necessary for all patients with monogenic IBD variants and critical for those with variants potentially amenable to allogeneic HSCT. Furthermore, when functional testing is not available, as with most of the variants described here, patients must be fully informed of the inherent risks of genetic interpretation on therapeutic decisions.

There are limitations with the WES methodology used in this study with approximately 5% of exons are poorly covered¹² (see Figure S1B for exon coverage of monogenic IBD genes). In this study, this limitation was illustrated by the poor WES coverage of exon 1 in SH2D1A. Using manual review of raw WES data and Sanger sequencing validation, we identified a variant (M1T) resulting in the loss of the start methionine in exon 1 of SH2D1A in a patient with severe ileitis, colitis, and growth failure (Patient 19, Tables 2 and S8Q, Figure S6A). The identification of this SH2D1A variant resulted in curative HSCT for this patient. Also WES does not cover non-coding yet potentially functional regions of the genome and has limited capacity to identify copy number changes and structural variants¹². Furthermore, there has been a rapid increase in the discovery of monogenic IBD genes⁴⁷ and we anticipate that many more genes will be discovered. Therefore, this study likely underestimates the contribution of monogenic gene disorders in pediatric IBD.

Overall, this single pediatric IBD center study supports a 3% prevalence of damaging variants in genes linked to monogenic IBD. Most importantly, this study demonstrates that 1% of monogenic pediatric IBD patients have variants in genes associated with primary immunodeficiency that are potentially curable through allogeneic HSCT (Table S9). We believe this data supports the diagnosis of monogenic disease beyond the very early onset IBD population especially in children with a family history of autoimmune diseases and those with evidence of extra-intestinal manifestations of IBD. Molecular identification of disease-causing variants in monogenic disease genes can inform patient management and improve outcomes by targeting definitive and personalized treatment strategies.

What you need to know.

BACKGROUND AND CONTEXT

A proportion of infants and young children with inflammatory bowel diseases (IBD) have subtypes associated with a single gene variant driving disease (monogenic IBD).

NEW FINDINGS

In whole-exome sequencing analyses of more than 1000 children with IBD, we found that

3% had rare variants in genes previously associated with pediatric IBD. These caused different phenotypes, and

1% of the patients had variants that could be corrected with gene therapy and hematopoietic stem cell transplantation

LIMITATIONS

Further studies are needed to determine what genetic factors might cause the other 97% of cases of pediatric IBD.

IMPACT

Monogenic IBD is rare but should be considered in analysis of all patients with pediatric onset of IBD. Knowledge of genetic factors can be used in prognosis and selection of therapy.