The E3 ubiquitin ligase UBR5 interacts with TTC7A and may be associated with very early onset inflammatory bowel disease

Very early onset inflammatory bowel disease (VEOIBD) denotes children with onset of IBD before six years of age. A number of monogenic disorders are associated with VEOIBD including tetratricopeptide repeat domain 7A (TTC7A) deficiency. TTC7A-deficiency is characterized by apoptotic colitis in milder cases with severe intestinal atresia and immunodeficiency in cases with complete loss of protein. We used whole exome sequencing in a VEOIBD patient presenting with colitis characterized by colonic apoptosis and no identified known VEOIBD variants, to identify compound heterozygous deleterious variants in the Ubiquitin protein ligase E3 component N-recognin 5 (UBR5) gene. Functional studies demonstrated that UBR5 co-immunoprecipitates with the TTC7A and the UBR5 variants had reduced interaction between UBR5 and TTC7A. Together this implicates UBR5 in regulating TTC7A signaling in VEOIBD patients with apoptotic colitis.


Results
Patient summary. A boy of Spanish ancestry, born to healthy non-consanguineous parents, presented with diarrhea, rectal bleeding and rectal prolapse at age 2 years and 9 months (Supplementary Figure 1A and Table 1 for blood work analysis). He had no extra intestinal manifestations of disease. Colonoscopy and histopathological examination demonstrated patchy inflammatory cell infiltrates with apoptosis leading to a diagnosis of IBD-Unclassified (Supplementary Figure 1B). He was initially treated with intravenous steroids but eventually required oral tacrolimus due to poor response. His treatment was changed to rectal 5-ASA and azathioprine after response to tacrolimus. At 4 years and 10 months age, his disease flared (Pediatric Ulcerative Colitis Activity Index 39,40 [PUCAI] score of 75 points) with no response to intravenous steroids and was switched to infliximab. Currently, he is maintained with 10 mg/kg/4 weeks infliximab (IFX) and rectal 5-ASA (due to a proctitis unresponsive to IFX) and oral 5-ASA with his most recent endoscopy showing only mild proctitis (Supplementary Figure 1C). Genetic analysis. Analysis of whole exome sequencing of DNA samples from the families did not identify any known monogenic disorder associated with VEOIBD (see Fig. 1A for analysis strategy and Supplementary Table 2 for full list of potential variants). However, we identified biallelic compound heterozygous nonsynonymous variants in the UBR5 gene of the patient (Supplementary Table 2 for further information regarding UBR5 mutation including MAF and damaging scores). A variant in exon 4 of UBR5 [hg38.g.chr8:102360605(G>T); NM_015902:c.250C>A] resulting in a proline to threonine substitution at amino acid position 84 (p.P84T) was inherited from the unaffected mother ( Fig. 1B,C). This variant is rare (gnomAD 41 minor allele frequency (MAF) = 0.000081) and predicted to be deleterious by bioinformatic algorithms, including a PHRED scaled combined annotation dependent depletion 42  . This variant has a CADD score of 24.6 and is not present in gnomAD. Both affected residues, P84 and L1405, are highly conserved across species (Fig. 1D-G respectively). While the p.P84T variant is not located in a known domain, the p.L1405V variant is predicted to affect the second nuclear localization sequence of UBR5 43 (Fig. 1H). A number of VEOIBD databases were searched for other patients with potential UBR5 variants but none were identified. Functional studies. Immunohistochemistry. As Hemotoxylin and Eosin staining of the patient's colonic biopsies (Supplementary Figure 1B) demonstrated increased apoptosis, we further examined apoptosis using immunohistochemistry (IHC) staining for cleaved (Cl) Caspase 3. Dual labeling of colon sections in a healthy control and patient with IBD and with known variants, we observed minimal cleaved(cl)-Caspase 3 positive cells ( Fig. 2A,B). In contrast, patients with UBR5 or TTC7A variants had increased cleaved-Caspase 3 positive cells (Fig. 2C,D). In our patient with UBR5 variants, we observed increased cleaved-Caspase 3 positive cells in the lamina propria (Fig. 2D). In contrast, in the TTC7A patient section (Fig. 2C), the cleaved-Caspase 3 positive cells are only in the epithelium. There is no observed difference in the intensity and architecture for β-catenin between UBR5 patient section (Fig. 2D) and control sections (healthy control + IBD patient sections) ( Fig. 2A,B). However, there is a disruption in epithelial layer architecture for β-catenin in TTC7A patient section (Fig. 2C).
IHC staining for UBR5 and TTC7A in a healthy control showed that in the colon, UBR5 localized mainly in the nuclei of immune cells with minimal epithelial expression (Fig. 3A). In colonic sections from an IBD patient without UBR5 or TTC7A variants and a TTC7A-deficiency patient (previously described 44 Table 3). Interestingly, colonic sections from our UBR5 variant patient demonstrated a different pattern of localization in the patient, as compared to healthy controls and IBD patients with strong signal intensity for UBR5 aggregated in the epithelial indicative of intra-epithelial lymphocytes (Fig. 3D,E).
Identification of TTC7A as a binding partner of UBR5. Previously, tandem mass spectrometry (MS) using TTC7A WT and the E71K variants as bait identified UBR5 as a potential interactor of TTC7A 1 . To further validate these finding, we used co-immunoprecipitation (co-IP) studies in HEK 293 T cells and showed that UBR5 did co-IP with TTC7A ( Fig. 4A) and TTC7A with UBR5 (Fig. 4B). TTC7A showed reduced co-IP with the identified UBR5 L1405V mutant protein, as compared with UBR5 WT (Fig. 5A,B). However, TTC7A showed increased co-IP with a catalytically dead UBR5 E3 ligase C2768A mutant in the HECT domain. Further analysis of the previously identified TTC7A VEOIBD mutants (E71K, Q526X, and A832T) 1 also showed that UBR5 had reduced co-IP to the missense variants but increased co-IP with the Q526X truncation (Fig. 6A,B). These results validate the previously identified interaction between TTC7A and UBR5 and indicate a role of UBR5 in TTC7A signaling 1

Discussion
We have recently demonstrated that 3% of pediatric IBD patients have monogenic forms of IBD 12 . Like the case presented here, many of these patients present at a very early age with severe disease that is difficult to treat 12 . The young age of our patient, the severity of disease requiring biologic therapy, and the presence of apoptosis on biopsy made this patient a strong candidate for genetic analysis to determine a potential monogenic cause of www.nature.com/scientificreports/ disease. We first screened for variants in genes known to be associated with VEOIBD, and as this patient was male we focused on X-linked genes associated with intestinal epithelial apoptosis including FOXP3 and XIAP, and also autosomal recessive genes including LRBA, ARPC1B, and TTC7A (see 12,45 for a complete list of genes associated with apoptosis). However, neither variants in these genes nor other genes associated with VEOIBD were identified. Therefore, we examined novel candidate genes and prioritized variant that were rare and damaging based on known biological function, animal models, and known interaction with previously identified VEOIBD genes. UBR5 was selected as a potential gene candidate based on evidence from our previous study utilizing tandem mass spectrometry to identify potential binding partners for TTC7A 1,18 . Bi-allelic deleterious variants in TTC7A were identified as a causal gene for severe intestinal and immune disease with high penetrance 1,13-16,18,21-24,46-49 . Many TTC7A-deficient patients present with clinical features associated with monogenic IBD or VEOIBD 1,14,15,18,21,22,47 . A key pathological feature of TTC7A deficiency is increased intestinal epithelial cell apoptosis 1,13,18,47 that is not commonly found in typical non-genetic forms of IBD. Although, our patient did not have the severe immunodeficiency and intestinal stricturing disease observed in severe loss-offunction TTC7A mutations, he did have apoptotic colonic disease associated with the less severe form of the disease caused by hypomorphic TTC7A mutations.
With any potential novel VEOIBD variants, functional studies are required to demonstrate a potentially causative defect. Here we used co-IP experiments to validate our genetic studies and showed that variants in both TTC7A and UBR5 reduce co-IP. Previous studies of Ubr5 -/mice demonstrated widespread apoptosis by E9.5 embryonic development stage 50 . UBR5 is a HECT E3 ubiquitin ligase that is found to be mutated or amplified in various cancer types (including colorectal cancer 51 ), to inhibit intestinal apoptosis 35 and UBR5 knockdown resulted in increased apoptosis in ovarian cells 52 . Pathological examination of biopsies from patients with both TTC7A and UBR5 variants demonstrate that UBR5 has low expression in the healthy gut but is highly upregulated in TTC7A-deficiency patient indicating a possible compensatory mechanism. Also, in our VEOIBD patient with UBR5 variants, there was increased apoptosis in both epithelial and immune cells. As TTC7A is expressed in both epithelial and immune cells, this may be due to dysregulation of TTC7A-PI4K signalling but as UBR5 has a role in DUBA signalling in T-cells, this may contribute to disease progression 53 . It is interesting to speculate that our patient may benefit from treatment with Leflunomide as it was recently shown in a preclinical study as a potential therapy for TTC7A-deficiency 54 .
Our studies suggest that UBR5 may be associated with VEOIBD; however, there are a number of important limitations. First, despite searching a number of VEOIBD databases, we were unable to identify a second patient with bi-allelic variants in UBR5. UBR5 does not have any homozygous loss-of-function variants on gnomad (https ://gnoma d.broad insti tute.org/) suggesting that loss of function is detrimental and often it takes time to identify additional patients. Second, our functional studies using tandem mass-spec and co-immunoprecipitation experiments have shown that TTC7A and UBR5 appear to interact; however, we did not determine the precise role of UBR5 in TTC7A signaling. Therefore, further study into the function of this putative interaction and the relevant cell type(s) will be critical in our understanding of the disease pathogenesis.

Methods
Next-generation sequencing and data analysis. WES was performed in collaboration with the Regeneron Genetics Center (RGC) on this proband and his unaffected parents who were enrolled and consented in our NEOPICS partnership (https ://www.neopi cs.org/). Exome capture was carried out using the NimbleGen VCRome 2.1 and sequencing was done using an Illumina HiSeq 2500 platform with paired-end 75 bp reads. Sequencing reads were aligned to human reference genome (GRCh38). Variants were called using the Genome Analysis Toolkit (GATK) (pmid:20644199) and the generated VCF files were subsequently annotated with snpEff (pmid:22728672). Polymorphisms reported in public databases with Minor Allele Frequency (MAF) > 1% and synonymous variants were filtered out. Potential pathogenicity protein-coding variants were prioritized using evolutionary conservation and various prediction tools (SIFT, PolyPhen2, Mutation Taster) from dbNSFP 55 . Inheritance modeling was carried out using GEMINI software 56 (https ://gemin i.readt hedoc s.io/ en/lates t/) to identify variants that fit autosomal recessive, de novo, and X-linked inheritance patterns.
Patient data availability. The identified UBR5 variants of our patient will be submitted to the ClinVar 57 database (https ://www.ncbi.nlm.nih.gov/clinv ar/) upon publication. Information on the raw whole-exome sequencing data will not be published to protect research participant privacy. Sanger sequencing. Sanger sequencing was performed in the patient and parents to validate the com- The following primers were used to sequence P84T: forward TGG TAG AGT TTG CAG GAT TGG (sense), and reverse TGA TAA CTG ACT CCT CTG CTACT (anti-sense). The following primers were used to sequence L1405V: forward CCA AGG ACT GTG GGA CAA A (sense), and reverse CTC TTG CCA CTG AAC GTA GAA (anti-sense).

Plasmid constructs. pCMV-Tag2B EDD and C2768A were a gift from Darren Saunders & Charles Watts
(Addgene plasmid # 37188 and 37189 respectively) 43 . These plasmids were modified by deleting the His tag on the C-terminus of the cDNA for the experiment in Fig. 5. UBR5 mutant plasmids were created from the pCMV-Tag2B EDD WT plasmid by ACGT Corporation (Toronto). pEGFP-C1 EDD was also a gift from Darren  Statistical analysis. Co-IP: Odyssey FC (LI-COR Biosciences), a chemiluminescence scanner, was used for imaging of the western blot. Densitometry of the western blot was obtained by ImageStudioLite (LI-COR Biosciences) and quantified by making the ratio of the IP band for a protein in a sample relative to the lysate band in the same sample. For each sample, the IP/lysate ratio obtained for TTC7A for each sample was divided with IP/lysate ratio for UBR5 from the same sample. For each sample, the values obtained for TTC7A/UBR5 ratios of IP/lysate were made relative to TTC7A WT + UBR5 transfected sample. Vice versa was done for Fig. 5A to generate Fig. 5B. GraphPad Prism 5.0 software (GraphPad Software, San Diego, CA) was used to create the graph and student's t-test was performed for statistical analysis of the variables of interest. All experiments were n = 3 unless otherwise stated.

IF histochemical staining on formalin-fixed paraffin-embedded (FFPE) sections.
Colon-sigmoid mucosa tissue samples were retrieved from the Division of Pathology, The Hospital for Sick Children. These samples include healthy control, IBD control with normal GI histology and without validated variants from infants. Patient's biopsies from Spain (UBR5 patient) and BC (TTC7A patient 44 ) infants were obtained with Ethics approval and informed consent as previously summarized 58 .
Dual immunofluorescent histochemical staining. The details for IF staining on FFPE section procedure are published 58 . Briefly, as a first step, paraffin was removed using Xylene, and afterwards rehydrated with different  Table 1 for detail) polyclonal antibody and anti-cyto-structure mouse monoclonal antibody (Abcam Inc. Toronto, Ontario, see Table 1) incubation was performed overnight at 4 °C. On the following day, stained slides were washed three times for 5 min with 1X PBS. Secondary antibody, donkey anti-rabbit IgG Fab2 fragment-Rhodamine conjugate mixed with donkey anti-mouse IgG Fab2 fragment-FITC conjugate (Jackson Immuno Research Lab, West Grove, PA) incubation was performed at room temperature in darkness for 2 h, and slides were washed afterwards three times for 10 min in darkness. As a nuclear counterstain reagent, RedDot2 far red fluorescence (Biotium Inc. Fremont CA) was used at a dilution of 1:200. Finally, sections were mounted overnight with Vector shield fluorescence mounting medium (Vector Labs, Burlington, ON).
Confocal microscopy. Double/triple-immunostained sections were imaged using a Leica confocal laser scanning microscope (model TCS-SP8) and LAS-AF software (Leica Microsystems, Wetzlar, Germany), as previously reported 59 . The variable excitation wavelengths of the krypton/argon laser were 488 nm for fluorescein isothiocyanate conjugate, 568 nm for Texas Red complex, and 695 nm for Alexa Fluor 680 conjugate/RedDot 2 www.nature.com/scientificreports/ (nuclear counterstaining). Image processing, including color resolution, color separation, and merging of fields, were carried out using Adobe PhotoShop CS5 software (Adobe Systems Incorporated, San Jose, CA, USA).
Morphometric and colocalization analysis. The NIH ImageJ software was used with appropriate algorithms to analyze the degree of co-occurrence and correlation for TTC7A and UBR5 in the images. A total of 100 cells were selected that showed positive staining for both, TTC7A and UBR5, from 5 different areas within each slide. The JACoP plugin 60 for ImageJ was applied in these cells for obtaining Pearson and Mander's coefficients. Pre-  www.nature.com/scientificreports/ viously reported algorithms 61 were used to determine co-localization and the associated statistics are reported in Supplementary Table 3. Linear regression/correlation and the t-test were used for the statistical/correction analysis to report on colocalization.
Helsinki guidelines. All human experiments followed the Helsinki Guidelines.
Informed consent. Informed consent was obtained from the participant's parents and the study had local ethics board approval at Hospital Regional Universitario de Málaga, Málaga, Spain and Hospital for Sick Children, Toronto, Canada (Research Ethics Board: REB1000024905).