Novel CARMIL2 loss-of-function variants are associated with pediatric inflammatory bowel disease

CARMIL2 is required for CD28-mediated co-stimulation of NF-κB signaling in T cells and its deficiency has been associated with primary immunodeficiency and, recently, very early onset inflammatory bowel disease (IBD). Here we describe the identification of novel biallelic CARMIL2 variants in three patients presenting with pediatric-onset IBD and in one with autoimmune polyendocrine syndrome (APS). None manifested overt clinical signs of immunodeficiency before their diagnosis. The first patient presented with very early onset IBD. His brother was found homozygous for the same CARMIL2 null variant and diagnosed with APS. Two other IBD patients were found homozygous for a nonsense and a missense CARMIL2 variant, respectively, and they both experienced a complicated postoperative course marked by severe infections. Immunostaining of bowel biopsies showed reduced CARMIL2 expression in all the three patients with IBD. Western blot and immunofluorescence of transfected cells revealed an altered expression pattern of the missense variant. Our work expands the genotypic and phenotypic spectrum of CARMIL2 deficiency, which can present with either IBD or APS, aside from classic immunodeficiency manifestations. CARMIL2 should be included in the diagnostic work-up of patients with suspected monogenic IBD.


Results
Identification of patients with biallelic CARMIL2 variant. In total, 4 patients from 3 unrelated families were identified with homozygous variants of CARMIL2 (Fig. 1A). Patient 1 was identified through whole exome sequencing (WES), performed during the evaluation for suspected monogenic IBD at the University Hospital of Padova. Patient 2, the eldest brother of Patient 1, was recognized by family segregation analysis and then further sequenced by WES. Patient 3 and Patient 4 were identified by screening for biallelic CARMIL2 variants from WES data in a cohort of 1005 pediatric IBD patients enrolled at SickKids, Toronto, as described in the "Methods" 29 .
Clinical features of patients with biallelic CARMIL2 variant. Patient 1 was born in Italy to consanguineous (first cousins) healthy Moroccan parents, fourth-born of four male sons. At 3.25 years he was diagnosed with colonic Crohn's disease, phenotype A1aL2B1G1 according to the Paris classification ( Fig. 1B,C, Table 1) 25 . Symptoms started when he was 2.6 years old. The patient was induced with prednisone, and after clinical remission he was maintained with azathioprine. Follow-up endoscopy performed after 18 months revealed histologic persistent mild total colonic inflammation in biopsies while in clinical remission. At the age of 5 years, the patient developed Streptococcus pneumoniae pneumonia complicated by sepsis, despite being vaccinated against pneumococcal disease. Past medical history included eczema in the first months of life. Anti-thyroglobulin antibodies and anti-thyroid peroxidase antibodies were significantly raised, while thyroid function tests and thyroid ultrasound were unremarkable. Patient 2 is the eldest brother of Patient 1 ( Table 1). The boy was diagnosed with Addison's disease at 12 years of age, and shortly after with Hashimoto's thyroiditis. No other endocrinopathies were detected. He suffered from difficult-to-treat eczema from a young age. He had no candidiasis, but at the age of 17 he was clinically diagnosed with extended Tinea corporis skin infection. Laboratory tests were significant for negative anti-adrenal antibodies and positive anti-thyroid autoantibodies. Plasma assay for very long-chain fatty acids was normal, thereby excluding peroxisomal disorders. Abdomen magnetic resonance imaging revealed bilateral adrenal hypoplasia. Sanger sequencing of the AIRE gene was non informative. Thus, Patient 2 was diagnosed with autoimmune polyendocrine syndrome type 2 (APS-2), or Schmidt syndrome, whose distinctive feature is Addison's disease, associated with at least one among autoimmune thyroid disease and type 1A diabetes mellitus. There was neither history of significant or recurrent infectious episodes, nor evidence of bowel disease and fecal calprotectin was normal.
Immunological work-up of Patients 1 and 2 revealed a marked increase in the proportion of naïve T and B cells, with a reduction in memory T and B cells and marginal zone B cells. Patient 1 but not patient 2 had an increase in the number of double negative T cells (CD3 + CD4-CD8-TCRαβ+) (Supplementary Table S1). Both patients had negative anti-EBV IgG and IgM and EBV-DNA on peripheral blood. Both patients had low titers of specific antibody against tetanus and diphtheria toxoids. Patient 3 is a female diagnosed with ulcerative colitis (UC), type pancolitis (A1bE4S1), at the age of 11 years old (Table 1). Remission was achieved after induction with prednisone, and maintained first with sulfasalazine, then with azathioprine. After relapse, she failed infliximab and at the age of 17 years underwent colectomy with loop ileostomy and J pouch. Her post-operative course was complicated by intestinal leakage causing intraabdominal sepsis, stricture at pouch anastomosis and pouchitis. Past medical history included reactive airway www.nature.com/scientificreports/ disease and an episode of pneumonia. Before IBD onset she had no other significant history of infections or immune-related diseases. She had no familial history of IBD. Patient 4 is a male diagnosed with UC, type pancolitis (A1bE4S1), at the age of 15 years (Table 1). The patient was induced with IV methylprednisolone, became steroid dependent first, then steroid refractory; azathioprine was thus introduced. He developed azathioprine-induced pancreatitis with pseudocysts, hepatic abscesses and C. difficile infection. Drain insertion was complicated by peri-sigmoid abscess, colonic-peripancreatic enterocutaneous fistula and toxic megacolon. At the age of 16 the patient underwent subtotal colectomy and ileostomy, then J pouch surgery. Past medical history was unremarkable. Particularly, he had no previous history of infections, or immune-related diseases. He had no familial history of IBD.
Genetic analysis of patients with biallelic CARMIL2 variant. All four patients were found to have novel biallelic variants of CARMIL2 (Fig. 2).
Patient 1 and 2 have a homozygous nonsense variant p.Cys155ValfsTer54 of CARMIL2, located on Exon 6 ( Table 2). This frameshift mutation is caused by a single nucleotide (cytosine) deletion resulting in a nonsynonymous substitution of cysteine (C) with valine (V) and premature termination of translation after 54 codons (predicted number of amino acids: 207, while wild type CARMIL2 consists of 1435 amino acids). Variant p.C155VfsTer54 is not reported in literature and no data are available on allele frequency on reference databases (ExAC, gnomAD, 1000 Genomes database, EVS) ( Table 2). Segregation analysis showed that the parents were heterozygous carriers of the variant. Patient 3 has a homozygous nonsense CARMIL2 variant p.Glu978*, located on exon 29 of CARMIL2 (Table 2) resulting in stop-gain nucleotide substitution and premature termination codon. CARMIL2 variant p.E978* is not reported in literature and is absent from population databases. Patient 3 was recruited as a "singleton", so parental DNA was not available for variant segregation analysis. Patient 4 has a homozygous missense variant p.Asp623Glu, located on exon 21 of CARMIL2 (Table 2). It has not been reported in literature, and its frequency is extremely low (< 0.001) in population databases. Variant p.D623E is predicted to be deleterious by the majority of the algorithms examined. Moreover, the aspartic acid at position 623 is located in a leucine-rich repeat (LRR) domain of CARMIL2 and is part of an evolutionarily www.nature.com/scientificreports/ www.nature.com/scientificreports/ constrained region (Supplementary Table S2). The patient was recruited as a "duo", and only the unaffected father's DNA was available for allele segregation analysis. The father was a heterozygous carrier for the same variant. Sanger sequencing confirmed the presence of the biallelic variant p.C155VfsTer54 in Patient 1 and 2 but failed for both the variants identified in Patient 3 (p.E978*) and Patient 4 (p.D623E), due to a guanine-cytosine-rich template ( Supplementary Fig. S1). However, careful reexamination of raw WES reads showed high coverage support for homozygous mutant calls in both candidates. Furthermore, since LoF is the pathogenic mechanism of CARMIL2 deficiency and tissue immunostaining proved reduced protein expression in both patients, confirmation by sequencing was deemed unnecessary 34 .

Immunostaining of bowel biopsies in biallelic CARMIL2 variant carriers with inflammatory bowel disease.
Immunostaining on sigmoid formalin-fixed paraffin-embedded (FFPE) sections was performed to assess the expression levels of CARMIL2 in the bowel of the three patients affected by IBD (Patient 1, 3 and 4; Fig. 3). In both normal and IBD control sigmoid sections CARMIL2 was strongly expressed in the enterocytes of mucosal layer and stromal cells. In line with The Human Protein Atlas, CARMIL2 was mainly localized in cell cytoplasm, but some nuclear expression was also observed in blood cells 1 . In the two patients with homozygous protein-truncating variants (p.C155VfsTer54 and p.E978*), CARMIL2 protein expression was nearly absent or remarkably down-regulated in both glandular layer and stromal area. In Patient 4, carrying the homozygous missense variant p.D623E, CARMIL2 expression in the mucosal layer was significantly weaker in comparison to controls, while infiltrated blood cells located in the stromal area showed focally positive staining. This range of staining intensity in different cell types is consistent with previously published immunohistochemistry images, showing stronger staining in lymphocytes than in intestinal epithelium 18 . Cytokeratin 18 staining pattern in all samples was as expected based on previous reports 35 , indicating a preserved tissue architecture.
Functional validation of CARMIL2 variants. Next, we investigated the impact of the novel CARMIL2 variants on protein expression and cellular localization using cellular model systems. Western blotting was used to analyze the expression of 3xFLAG tagged CARMIL2 variants in HEK293Tcells (Fig. 4). Immunoblotting with anti-FLAG antibody (Fig. 4A) detected bands corresponding to the predicted molecular weight (MW) both for CARMIL2 wild type (WT) and for the protein-truncating variants, p.C155VfsTer54 and p.E978*. The MW of www.nature.com/scientificreports/ p.D623E variant was ~ 80 kDa, much lower than the WT, the only difference being the substitution of a single amino acid residue from aspartic acid to glutamic acid. This strongly suggested proteolysis of the p.D623E variant to a lower MW species. Immunofluorescence staining of transfected cells was performed in both HEK293T cells and HCT116 cells (Fig. 4B). Immunofluorescence imaging showed that the expression pattern of CARMIL2 WT was diffuse across the cytoplasm in both cell strains, in line with the existing knowledge of the protein, as reported in The Human Protein Atlas 1 . Similarly, CARMIL2 variant p.C155VfsTer54 exhibited cytoplasmic expression, albeit less homogeneous. Conversely, CARMIL2 variants p.E978* and p.D623E signal appeared as puncta structures throughout the cytoplasm of HEK293T cells. For variant p.D623E the puncta staining was consistent in HCT116 cells. Therefore, this missense variant appeared to form aggregates when expressed in different cell lines. Overall, different lines of evidence suggest that variant p.D623E is susceptible to mislocalization inside the cell and proteolysis, thus behaving as a null or non-functional allele.

Discussion
Our study confirms that CARMIL2 deficiency can manifest only with isolated IBD. Unlike more recent studies 17,18 reporting a very early IBD onset (6 out of 6 patients), two of our patients shared a later age of onset of IBD symptoms, namely at 11 and 15 years of age. Although the majority of studies on monogenic IBD have focused on the very early onset population, running the risk of selection bias, recent findings have pointed out that a genetic disorder should be considered in all patients with pediatric onset IBD 29 . Resistance to conventional lines of treatment and complicated disease course, similar to what was reported in Patients 3 and 4, should always prompt the execution of WES, in line with existing recommendations 22 . None of our patients presented overt clinical signs of immunodeficiency before the diagnosis of IBD. Patient 1 developed bacterial sepsis after the initiation of immunosuppression (i.e., azathioprine). Notably, one patient reported in the study of Magg et al. died due to septic complications at the age of 4 years while on treatment with azathioprine 17 . Patient 3 and 4 experienced severe infectious complications after surgery. The need for a surgical treatment appears to be a common feature of pediatric IBD associated to CARMIL2-deficiency, since it occurred in the majority of the www.nature.com/scientificreports/  Supplementary Fig. S2.  www.nature.com/scientificreports/ patients reported so far (6 out of 9 patients, including our cases) and was frequently due to a failure of medical treatment (3 out of 6 patients, including our cases) 17,18 . A recent study found that progression to surgery can be itself an indicator of monogenic etiology among pediatric IBD patients 29 . There is no consensus with respect to surgical timing, indications, and strategies in monogenic IBDs, including CARMIL2 deficiency 36 . Our results suggest that CARMIL2-deficient IBD patients harbor a significant risk of life-threatening immunosuppressionand surgery-related infections. A treatment option for several genetic immunodeficiencies underlying IBD-like phenotypes is hematopoietic stem cell transplantation (HSCT) 22 . However, it could be inappropriate or even harmful if an epithelial barrier defect coexists 23 . Herein, we confirm that CARMIL2 protein is expressed in gastrointestinal epithelium, but it remains to be clarified if this is relevant to IBD pathogenesis in deficient patients.
Overall, CARMIL2 should be included in the diagnostic work-up of patients with suspected monogenic IBD regardless of the age at disease onset and of the presence of overt clinical signs of immune deficiency. In fact, the identification of CARMIL2 deficiency has the potential to influence treatment choice and might improve disease prognosis, especially by means of prevention and prompt recognition of infectious complications. Alongside phenotypic differences, CARMIL2 deficiency is characterized by marked genotypic heterogeneity. The known variants associated with CARMIL2 deficiency have different impact on the protein (e.g. nonsense, frameshift, missense, etc.) and they are spread along the gene (Fig. 2). The only consistency seems to be the lack of mutations in the C-terminus portion of the gene. One could speculate that mutations occurring in the C-terminus, close to end of the coding sequence, may preserve some level of protein function. Therefore, failure to produce the expected phenotype would lead to missed diagnosis. The lack of obvious genotypic-phenotypic correlation, as well as the inter-and intrafamilial clinical heterogeneity, even among carriers of the same CARMIL2 variant, seems to point towards a contribution of additional environmental, genetic, or epigenetic-modifying factors in determining the clinical manifestations of CARMIL2 deficiency.
Bioinformatic tools can predict the damaging effects of mutations, but might overestimate them, hence the importance of experimental validation. Different lines of evidence from the experiments we performed showed that p.D623E variant behaves as a non-functional allele. Particularly, immunofluorescence of transfected cells resulted in a granular pattern. Endoplasmic-reticulum-associated protein degradation (ERAD) of p.D623E protein variant might explain this finding. In fact, the residue 623 of CARMIL2 belongs to an LRR domain, a structural motif with a horseshoe shape, with an interior parallel beta sheet, an exterior array of helices, and an hydrophobic core containing many leucine residues 37 . An amino acid substitution in this region may lead to the exposure of normally hidden hydrophobic patches, usually buried inside the protein to keep the lowest energy state. Exposed patches might lead to protein aggregation, or they could be recognized as a substrate by ERAD, a cellular pathway that targets misfolded or mutated proteins, which are retained inside the endoplasmic reticulum (ER) and targeted for ubiquitination and subsequent proteolytic degradation 38,39 . ER retention and impaired trafficking could explain the staining pattern observed in transfected cells, the proteolysis the lower than predicted MW. Protein overexpression might make the process more obvious by engulfing the cell. Another, non-exclusive possibility is that substitution from aspartic acid to glutamic acid leads the protein to be recognized by a glutamic-acid-specific protease. This is the first report showing the association between CARMIL2 deficiency and autoimmune endocrinopathies. A monogenic etiology has been proved in a subset of patients with APS-2 and several lines of evidence support CARMIL2 as a candidate gene [30][31][32][33]40,41 . First, the significance of rare LoF variants of CARMIL2 in human disease has already been established, consistently with a recessive disease model 10 . This is confirmed by the modest LOUEF (loss-of-function observed/expected upper bound fraction) score (0.552 for CARMIL2, indicating a degree of intolerance to LoF variants) shown using gnomAD (Genome Aggregation Database) data 42 . Secondly, biallelic CARMIL2 LoF variants fully segregate with an immune dysregulation disorder in multiple independent families with autoantibodies and absence of antigen-specific antibodies to bacterial vaccines 10,14 . This is consistent with the known B cell phenotype of human CARMIL2 deficiency. Additionally, skin manifestations have been described in the majority of CARMIL2-deficient patients, and eczema is part of other monogenic APSs [9][10][11][12][13][14]17,18,30 . Thirdly, the function of CARMIL2 is consistent with the known pathogenesis of APS, being expressed in immune cells and endocrine glands and being essential for development of regulatory T cells and for Th17 differentiation, similarly to other genes mutated in monogenic APS 1, 8,9,[30][31][32][33] . Moreover, CARMIL2 interactome include DOCK8, that is mutated in patients presenting with autoimmune disorders, eczema and compromised Treg function 9,43 . Notably, CARMIL2-deficient mice and humans did not develop any obvious organ-specific autoimmune disorder, despite a reduction in Tregs 8,9 . This might depend on the coincident defect in effector T cells 9 . Furthermore, CARMIL2-deficient mice did not display any intrinsic B cell defect [8][9][10] . Therefore, knockout mice might not be a suitable model for studying human CARMIL2 deficiency. Remarkably, enteropathy and/or IBD-like intestinal inflammation, autoimmune endocrinopathy and skin manifestations are common in IPEX and other syndromes presenting with IPEX-like features, caused by monogenetic defects affecting Treg function 22,30,44 . Hence, CAR-MIL2 deficiency should be added to the increasing group of IPEX-like syndromes, as already proposed 18 .
Our study had some limitations, including that overexpression experiments might have been unreliable for protein-truncating variants p.C155VfsTer54 and p.E978*. In fact, it can be assumed that in vivo these variants are targeted by nonsense-mediated mRNA decay 45 . For this reason, the more intense immunofluorescence signal of p.C155VfsTer54 compared to WT CARMIL2 in cultured cells probably does not reflect what happens in vivo, and could have been determined by higher transfection efficiency due to a shorter peptide.
In summary, the phenotypic spectrum of CARMIL2 deficiency is broader than previously known, ranging from severe immunodeficiency to IBD and organ-specific autoimmunity. CARMIL2 should be part of the diagnostic evaluation of patients with suspected monogenic IBD, even in the absence of obvious signs of immunodeficiency. Genetic diagnosis may be vital in monogenic IBD, to guide specific treatment, prevent surgery or unnecessary therapies, anticipate complications, and help genetic counseling 22 www.nature.com/scientificreports/ deficiency can present exclusively with APS. Further studies are needed to better define IBD immunopathogenesis in CARMIL2 deficiency and the potential therapeutic utility of HSCT.

Methods
Helsinki guidelines. All human experiments followed the Helsinki Guidelines. Informed consent was obtained from the participants parents and the study had local ethics board approval at The Hospital for Sick Children (SickKids), Toronto, Canada (Research Ethics Board: REB1000024905).
Patients. Patient 1 and his brother, Patient 2, were recruited from the Pediatric Gastroenterology, Hepatology and Liver Transplantation Unit at the Pediatric Clinic of the University Hospital of Padova. Patient 3 and Patient 4 were discovered through screening of a large cohort of pediatric IBD patients recruited at SickKids as previously described 29 . Children (age < 18) undergoing evaluation for IBD were enrolled over a 13-year period (2003-2015). Diagnosis of IBD was made according to the recommendations of the Porto criteria 49 . Patients with known primary immunodeficiency, chromosomal abnormalities, syndromic disease or diagnosed with other forms of monogenic intestinal disease were excluded. After obtaining the caregivers' consent, clinical data were recorded, and biological samples were processed and stored in a biobank. WES analyses of blood samples was thus carried out as detailed below. For Patient 1 and Patient 2, the phenotype review was carried out at University Hospital of Padova using electronic medical record software Galileo (NoemaLife, Dedalus). For the SickKids cohort, following variant prioritization 50 , each patient deemed to have pathogenetic biallelic CARMIL2 variants was reversed phenotyped, using clinical data from databases and electronic medical record systems. For all IBD patients phenotype was classified according to the pediatric modification of the Montreal classification, also known as Paris classification 25 .
Whole exome sequencing. For Patient 1 and 2, WES and segregation analysis were performed. Genomic DNA was extracted and purified from whole blood. DNA was prepared for sequencing with the kit SureSe-lectXT All Exon V5 kit (Agilent). Exome sequencing was performed with next-generation sequencing (NGS) technology on the Illumina HiSeq 2500 platform, using paired-end 100 bp read. DNA sequences were mapped and analyzed using as reference genome assembly GRCh38 (Genome Reference Consortium Human Build 38).
For the SickKids cohort, banked genomic DNA isolated from whole blood collected by venipuncture using a Qiagen Puregene Blood Core Kit was processed for exome capture using the NimbleGen VCRome 2.1 design. Captured libraries were sequenced on the Illumina HiSeq 2500 platform using paired-end 75 bp reads at the Regeneron Genetics Center (RGC), Tarrytown, NY, USA. Exome sequencing coverage was 30 × or greater for > 85% of the bases targeted. For the SickKids cohort, WES data from 2307 participants (1005 index patients, and 1302 parents and siblings) were analyzed using the FORGE (Finding of Rare Disease Genes) pipeline. Raw 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). Five variant callers (GATK HaplotypeCaller ver. 3.5, Vardict ver. 1.4.6, Varscan ver. 2.3.9, Samtools ver. 1.3, and Freebayes ver. 1.0.0) were run on the Binary Alignment Map (BAM) files of each family to produce family based Variant Call Format (VCF) files when at least 2 of the 5 agreed on a called variant. Inheritance modeling on family level VCF files was performed using the GEMINI tool to query for rare (Minor Allele Frequency [MAF] < 0.01) protein coding variants that fit autosomal recessive, compound heterozygous, de novo, autosomal dominant, and X-linked inheritance models filters. VarSeq software (Golden Helix) was used to import, filter, and do inheritance modeling on the variants from each trio. www.nature.com/scientificreports/ For candidate patients identified within the SickKids pediatric IBD cohort, sequencing was carried out at The Centre for Applied Genomics (TCAG, The Hospital for Sick Children, Toronto, Canada), that provides highquality capillary-based fluorescent sequencing on dual ABI 3730XL instruments. Different sets of primers were used to troubleshoot sequencing reactions.
Immunofluorescence histochemical staining of bowel biopsies. Bowel samples were fixed in neutral buffered formalin without methanol and embedded in paraffin using routine protocols (formalin-fixed paraffin-embedded, FFPE). Tissue samples, including normal control and IBD control, were retrieved from the Unit of Anatomical Pathology, Medicine Department, University Hospital of Padova and from the Division of Pathology, The Hospital for Sick Children, Toronto. Only FFPE with well-preserved tissue architecture were chosen, to avoid false negative staining. As negative control and disease control, a non-IBD patient sample and an unrelated IBD patient sample were used, respectively. Immunofluorescent histochemical staining was performed on sigmoid FFPE sections as previously described 53 . Briefly, paraffin-embedded sections were deparaffinized using xylene and rehydrated with different percentages of ethanol. Antigen retrieval was achieved with high-pressure cooking in EDTA-borax buffer made with 1 mM EDTA, 10 mM borax, 10 mM boric acid and 0.001% ProClin 300 (Supelco) at pH 8.5. To block non-specific staining, the slides were incubated for 1 h at room temperature (RT) in 4% bovine serum albumin (BSA) and 20% donkey serum in phosphate-buffered saline (PBS). Slides were incubated with primary antibodies, including anti-CARMIL2 antibody produced in rabbit (Sigma, HPA041402) and anti-cytokeratin 18 mouse monoclonal antibody (Abcam), overnight at 4 °C. On the following day, stained slides were washed 3 times for 5 min with PBS. Secondary antibodies, namely Rhodamine Red-X (RRX) Affin-iPure F(ab') 2 Fragment Donkey Anti-Rabbit IgG (Jackson ImmunoResearch Laboratories) and Fluorescein (FITC) AffiniPure F(ab') 2 Fragment Donkey Anti-Mouse IgG (Jackson ImmunoResearch Laboratories), were incubated at RT in darkness for 2 h, then slides were washed 3 times for 10 min in darkness. RedDot2 Far-Red Nuclear Stain (Biotium) was used for nuclear counterstaining at a dilution of 1:200. Finally, sections were mounted overnight with Vectashield antifade mounting medium (Vector Laboratories). Immunostained slides were imaged using a Leica confocal laser scanning microscope (Leica, TCS-SP8) and LAS-AF software (Leica Microsystems). Image processing, including color resolution, color separation, and merging of fields, was done using Adobe Photoshop CS5 software (Adobe Systems Incorporated).
Plasmids. Plasmid containing an insert with human CARMIL2 sequence with 3 tandem FLAG epitopes (DYKDDDDK) on N-terminus was a gift from John Cooper (Washington University, St. Louis, Addgene plasmid #118740) 5 . Mutations of CARMIL2 were generated using site-directed mutagenesis by ACGT Corp. (Toronto, Canada). Plasmid were expanded through transformation and harvest from Escherichia coli DH10B bacteria as per standard protocols and purified using EZ-10 Spin Column Plasmid DNA Miniprep Kit (Bio Basic, BS614) and PureLink HiPure Plasmid Maxiprep Kit (Invitrogen, Thermo FisherScientific, K210007). Positive control plasmid FLAG-TRAF6 was provided by Muise laboratory (SickKids).
Cell culture and transfection. HEK293T cells and HCT116 cells were provided by Muise laboratory and maintained in DMEM (Wisent Inc.) containing 10% of heat-inactivated fetal bovine serum (FBS) and antibiotic-antimycotics at 37 °C in 5% CO 2 . HEK293T cells were chosen for the ease of transfection, while HCT116 cells were selected for being a colon cancer cell line, thus closer to intestinal cells. For western blot, HEK293T cells were grown in a Falcon 6-well plate (Corning) to reach 60-70% confluence and transiently transfected with 1 µg of construct DNA (3xFLAG-CARMIL2 WT or mutation, or FLAG-TRAF6 as positive control) per each well using PolyJet (SignaGen Laboratories) transfection reagent, according to the manufacturer's instructions. Cells were collected for lysis and protein analysis 48 h after transfection. For immunofluorescence, HEK293T cells and HCT116 cells were transiently transfected for 48 h as described above using 0.5 µg of construct DNA (3xFLAG-CARMIL2 WT or mutation) per well.