CD45-deficient severe combined immunodeficiency caused by uniparental disomy

Abstract

Analysis of the molecular etiologies of SCID has led to important insights into the control of immune cell development. Most cases of SCID result from either X-linked or autosomal recessive inheritance of mutations in a known causative gene. However, in some cases, the molecular etiology remains unclear. To identify the cause of SCID in a patient known to lack the protein-tyrosine phosphatase CD45, we used SNP arrays and whole-exome sequencing. The patient’s mother was heterozygous for an inactivating mutation in CD45 but the paternal alleles exhibited no detectable mutations. The patient exhibited a single CD45 mutation identical to the maternal allele. Patient SNP array analysis revealed no change in copy number but loss of heterozygosity for the entire length of chromosome 1 (Chr1), indicating that disease was caused by uniparental disomy (UPD) with isodisomy of the entire maternal Chr1 bearing the mutant CD45 allele. Nonlymphoid blood cells and other mesoderm- and ectoderm-derived tissues retained UPD of the entire maternal Chr1 in this patient, who had undergone successful bone marrow transplantation. Exome sequencing revealed mutations in seven additional genes bearing nonsynonymous SNPs predicted to have deleterious effects. These findings are unique in representing a reported case of SCID caused by UPD and suggest UPD should be considered in SCID and other recessive disorders, especially when the patient appears homozygous for an abnormal gene found in only one parent. Evaluation for alterations in other genes affected by UPD should also be considered in such cases.

SCID is a syndrome characterized by absent T- and B-lymphocyte function that is uniformly fatal in infancy without immune reconstitution (1, 2). Mutations in several different genes important for normal T-cell development, function, or survival have been shown to cause SCID, with a majority of reported cases caused by mutations in IL2RG, IL7RA, ADA, JAK3, RAG1, RAG2, or DCLRE1C (1, 2). Rare defects in six other genes have also been described, including two fatal cases caused by mutations in the gene encoding the CD45 protein tyrosine phosphatase (1, 3, 4).

Uniparental disomy (UPD) refers to the inheritance of two copies of a chromosome, or segment of a chromosome, from one parent. UPD was first observed in 1988 in a patient with cystic fibrosis who had inherited two maternal copies of chromosome 7 bearing a mutant CFTR allele (5, 6). Since that report, UPD has been found to underlie a number of diseases, including Prader–Willi, Angelman, and Beckwith–Wiedermann syndromes (7). UPD causes a genetic disorder either through inheritance of two mutant copies of a gene, thereby enabling a recessive mutation to manifest, or through inheritance of two silenced copies of an intact allele (8). UPD has not previously been reported as a mechanism of inheritance in SCID. In the present report, we are unique in describing an example of SCID caused by UPD, in which the first surviving CD45-deficient SCID patient inherited two complete copies of a single maternal Chr1 bearing a nonsense mutation in the exodomain of CD45. This complete isodisomy of Chr1 with resultant loss of heterozygosity (LOH) was present in both mesoderm-derived lymphocytes and ectoderm-derived buccal epithelial cells, suggesting that the duplication occurred before germ-layer specification.

Results

SCID Patient Lacks CD45 Expression.

Flow cytometric analysis of the patient’s peripheral blood lymphocytes at presentation at age 10 mo revealed normal numbers of B and NK cells but dramatically reduced numbers of T cells (Table 1). Moreover, the few T cells present were found to be nonfunctional as they were unresponsive to mitogenic stimulation (Fig. 1B). Serum IgM, IgA, and IgE were undetectable or low (Table 1). Cell-surface CD45 protein expression was lacking on all leukocytes following staining with Abs recognizing CD45 RA, CD45RO, or all CD45 isoforms (Fig. 1A and Table 1), consistent with the previous finding of absent bone marrow leukocyte CD45 expression at the referring hospital. CD45 is essential for T-cell development and T-cell receptor (TCR) signal transduction (9, 10), and has previously been identified in two fatal cases of SCID (3, 4). The patient in this report underwent a successful T-cell–depleted haploidentical maternal bone marrow stem-cell transplant without preconditioning or posttransplantation graft-versus-host disease (GVHD) prophylaxis at age 10 mo and currently has normal numbers of B and NK cells, as well as normal numbers of CD45-expressing, functional T cells at 5 y posttransplantation (Fig. 1B and Table 1).

Table 1.

Patient Immune phenotype and function

	Patient
	At presentation	Most recent	Controls
Serum Ig level*
IgG†	516	604	192–515
IgA	0	12	12–31
IgM	9	9	39–92
Lymphocyte subpopulation‡
CD45+	4 (0.4)	581 (51.9)	1,500–7,000
CD3+	31 (2.9)	596 (53.2)	1,111–5,183
Percent of CD3 that are CD45RO+ (%)	ND	(55.8)	(24.9–42.5)
Percent of CD3 that are CD45RA+ (%)	ND	(22.8)	(27.6–46.2)
CD4+	24 (2.3)	273 (24.4)	675–3151
Percent of CD4 that are CD45RO+ (%)	ND	(58.1)	(24.9–42.5)
Percent of CD4 that are CD45RA+ (%)	ND	(21.2)	(27.6–46.2)
CD8+	3 (0.3)	174 (15.5)	431–2012
Percent of CD8 that are CD45RO+ (%)	ND	(41.1)	(24.9–42.5)
Percent of CD8 that are CD45RA+ (%)	ND	(32.5)	(27.6–46.2)
TCRαβ+	21 (2)	468 (41.8)	1,855–3,199
CD20+	854 (80.9)	516 (46.1)	144–671
CD16+	71 (6.7)	31 (2.8)	152–709
Proliferative stimulus§
Medium	128	228	693 ± 825
Candida	ND	3,539	5,937–59,291
Tetanus	ND	8,486	13,004–68,696

ND, not determined.

*Values are expressed as mg/dL (IgG, IgA, IgM) or U/mL (IgE). Normal values are the 95% confidence intervals for 9- to 12-mo-old control subjects.

^†The patient was receiving intravenous immune globulin when the IgG levels were measured.

^‡Values are expressed as cells/mm³ or (percentage of lymphocytes). Control values are the 95% confidence intervals for 1,550 normals.

^§Values are cpm [³H]thymidine incorporation. Controls values are the mean ± SD of responses in 167 normals.

Fig. 1.

Patient CD45 expression and Immune data. (A) The CD45 expression on whole-blood samples from the patient and healthy volunteer as measured by four-color flow cytometry on electronically gated CD14⁻ cells. The anti-CD45 Ab recognizes all CD45 isoforms (clone HI30, solid lines). A FITC-labeled isotype control Ab is indicated by dashed lines. (B) The patient’s T-cell function was measured by [³H]thymidine incorporation on PBMC exposed to the indicated stimulus and plotted over time. (C) The expression of mRNA encoding CD45 as measured by SYBR green real-time PCR on cDNA prepared from sorted patient and normal volunteer PBMC. Levels of CD45 mRNA were normalized to those of β-actin for each sample and are presented as arbitrary unit (a.u.) of fold change calculated using the 2^–ΔΔCT method. Values are mean ± SEM of triplicate determinations. Con A, concanavalin A; pokeweed mitogen, PWM.

The loss of CD45 expression in the patient was not the result of a defect in transcription, because the level of CD45 mRNA in the patient was only slightly decreased relative to that of controls (Fig. 1C). Sanger sequencing revealed that the patient’s mother was heterozygous for a nonsense mutation at position 1618 (1618A > T) of the coding sequence in exon 14 of the CD45 gene that created a stop codon at amino acid 540 (K540X); however, no mutations were observed in the coding region of either paternal CD45 allele (Fig. 2). Surprisingly, the patient was homozygous for the 1618A > T mutation observed in the maternal allele (Fig. 2). This finding suggested that either one copy of paternal Chr1 bore a microdeletion eliminating the CD45 locus, or the patient inherited two copies of the mutant maternal CD45 allele.

Fig. 2.

Sequence analysis of the CD45 alleles of the patient and parents. The schematic depicts the domains of CD45, including the alternatively spliced exons that give rise to the RA and RO isoforms, the fibronectin III-like (FNIII) domain, transmembrane domain (TM), and phosphatase domains (PTP1 and -2). The position of the nonsense mutation is marked by an asterisk. The change in nucleotide sequence (Top) and amino acid sequence (Bottom) in the patient (Mut hCD45) are listed (Middle). The Sanger sequence trace on genomic DNA from parents and patient are depicted (Bottom).

SCID Is Caused by Duplication of the Mutant Maternal CD45 Allele Because of UPD of Chr1.

To distinguish these possibilities, SNP arrays were performed on genomic DNA from EBV lines derived from B lymphocytes of the parents and patient. These analyses demonstrated that there was no change in copy number across the CD45 locus on Chr1 (Fig. 3A), indicating that the patient had either inherited two entire copies of maternal Chr1 or two copies of the region encompassing the maternal CD45 mutation. The allele profiles from the whole-genome array analysis revealed LOH over the entire length of Chr1 (Fig. 3A), demonstrating isodisomy. Isodomy was found to be restricted to Chr1, as there was no evidence for LOH on any of the other chromosomes in the patient (Fig. 3B). Whole-genome array analysis revealed that isodisomy was also evident in other primary blood cell types (lymphocytes and neutrophils), which are derived from embryonic mesoderm, as well as in buccal epithelium, which is derived from embryonic ectoderm (Fig. 3C). Taken together, these data suggest that the events leading to the isodisomy of maternal Chr1 occurred very early in development, probably before the separation of the germ layers, and that isodisomy is likely manifested in most if not all tissues.

Fig. 3.

Copy number and allele peak analysis of various cell populations from the parents and patient. (A) Analysis of copy number and allele peaks in EBV-immortalized B cells is depicted. The y axis depicts DNA copy number (not log2 ratio) (Upper) and the allele peak (Lower). Allele peak panels normally show three “bands,” representing all homozygous (Top and Bottom bands) and heterozygous (Middle band) allele calls. Note that the copy number panels reveal two copies of each gene on Chr1, but show only two bands, with loss of the Middle (heterozygous) band (arrow) across the entire chromosome, indicative of UPD. (B) Chromosome coordinates for each autosome and the X chromosome are shown on the x axis. (C) Copy number and allele peak analysis are depicted for Chr1 in neutrophils, lymphocytes, and buccal epithelium from the patient. All analyses were performed using Affymetrix Cytogenetics Whole-Genome arrays as described in Materials and Methods.

Additional Genes on Chr1 Exhibit Changes in Coding Sequence.

Humans are estimated to have in excess of 200 recessive disease-causing mutations distributed throughout their genomes (>100 per haploid genome) (11). Because the patient inherited two complete copies of maternal Chr1, which represents ∼5% of the entire genome (both Chr1 homologs = 9%, maternal copy = 4.5% of entire genome), the patient would be predicted to have on the order of 18 recessive disease-causing alleles in both copies of Chr1 (9 per each parental copy). Importantly, because isodisomy in this patient appeared to arise before separation of the embryonic germ layers, these mutations would be expected to persist even after bone marrow transplantation in both nonlymphoid hematopoietic tissue and in nonhematopoietic tissues.

To identify such alleles, exome sequencing was performed on DNA samples from mother, father, and patient. Using the maternal alleles for UPD analysis, we identified 36 homozygous SNPs, all of which were located on Chr1 (Table 2). Conversely, no paternal alleles of UPD from any chromosome were identified. Consistent with the Sanger sequencing analysis, the maternal PTPRC (CD45) nonsense allele was homozygous in the patient. The identified maternal UPD SNP alleles are distributed throughout Chr1, in agreement with the results of the whole-genome array indicating UPD of the entire maternal Chr1. Among the 36 SNP alleles on Chr1, we found 7 other mutated genes in addition to PTPRC bearing nonsynonymous SNPs that were predicted to have deleterious effects on gene function (Table 2). Although several of these were found in genes of unknown function, others have been shown to function in cell-cell contact (LGALSG), binding bacterial cell wall components (PGLYRP3), replication of DNA (ORC1L), and in binding seratonin (HTR1D) (12–15). Collectively, these SNPs appear to have no deleterious effects on early development, because aside from SCID caused by the CD45 mutation, the 5-y-old male patient appears to be otherwise phenotypically normal. Nevertheless, it remains possible these mutations may manifest later.

Table 2.

Analysis of homozygous mutations resulting from UPD of chromosome 1

Gene	Chromosome: position	Function	Mutation type/position	Uniprot	Change in AA Seq.	SVM prediction*
HTR1D	1:23392180	Seratonin receptor	Missense	P28221	R374W	Deleterious (96%)
C1ord94	1:34440323	Unknown	Missense	Q6P1W5	N251I	Deleterious (85%)
ORC1L	1:52627549	Origin recognition complex	Missense	Q13415	A372V	Deleterious (55%)
PGLYRP3	1:151546220	Peptidoglycan recognition	Missense	Q96LB9	R68Q	Deleterious (60%)
PTPRC	1:196954019	CD45 phosphatase	Nonsense	P08575	K540*	Deleterious*
DDX59	1:198886387	Dead-box polypeptide 59; unknown	Missense	Q5T1V6	L368P	Deleterious (55%)
SIPA1L2	1:230628143	Signal-induced Proliferation associated like 2; unknown	Missense	Q9P2F8	S1482W	Deleterious (80%)
LGALS8	1:234773522	Galectin family; adhesion	Missense	O00214-2	P239S	Deleterious (80%)

*Prediction of the likelihood a mutation will be deleterious to protein function using SVM. The number in parenthesis quantifies the likelihood that a mutation will disrupt protein function. A detailed description is available in Materials and Methods.

Discussion

We are unique in describing a case of SCID caused by UPD. This child is also the first reported surviving CD45-deficient SCID patient. The first reported SCID patient lacking CD45 expression died from a CMV infection 55 d after a matched unrelated bone marrow transplant and was subsequently shown to be homozygous for an in-frame 6-bp deletion in exon 11 of the CD45 gene that caused deletion of residues E339 and Y340 in the extracellular domain of the protein (3, 16). Both of her consanguineous parents were heterozygous for this deletion. The other reported patient with CD45-deficient SCID inherited a large deletion in the 3′ end of the CD45 gene from his mother. The patient’s second CD45 gene mutation was a G > A substitution at the donor splice site of IVS 13, leading to aberrant splicing that was not present in either parent (4). The child died from a B-cell lymphoma at age 2 y. The patient in the present report underwent successful bone marrow transplantation without pretransplant chemotherapy and with a rigorously T-cell–depleted maternal bone marrow stem-cell transplant that led to chimerism in several lineages (eosinophils, neutrophils, and B cells) in addition to T cells.

UPD has been reported to cause disease either through inheritance of two silent, imprinted alleles from the same parent (e.g., Prader–Willi Syndrome) or through inheritance of a duplicated recessive, mutant allele (8). The case of SCID described here resulted from inheritance of two entire copies of a maternal Chr1 bearing a nonsense mutation in the exodomain of CD45, which eliminated its expression and abrogated T-cell function. Because both SCID (<1/40,000 births) and UPD (1/3,500 births) are rare, it is unlikely that our identification of UPD underlying SCID is an isolated occurrence (17).

UPD can occur through a variety of mechanisms including mistakes in meiosis that produce aneuploid gametes or postfertilization errors in a normal zygote (mitotic disjunction) and these distinct causes produce UPD with different inheritance patterns (17). The case of UPD described here manifested as complete duplication of an entire maternal allele of Chr1 and was present in tissues arising from both embryonic mesoderm (blood) and ectoderm (buccal epithelium). Accordingly, this was unlikely to be the result of a postfertilization error, which typically produces chimeric representation of UPD involving chromosomal fragments. The presence of Chr1 duplication in at least two of the three embryonic germ layers also suggests that this event occurred very early in development, perhaps involving the gametes. It is estimated that nearly 20% of eggs and 3–4% of sperm are aneuploid, raising the possibility that fertilization events involving aneuploid gametes occur relatively frequently (18). Accordingly, the isodisomy we observed could have been caused by fertilization of a disomic gamete (egg) by a nullisomic male gamete, a process known as gamete complementation, or by monosomy rescue, when a monosomic gamete is fertilized by a nullisomic gamete, following which the chromosome is duplicated (17). In cases involving the division of cells with a normal karyotype, the occurrence of UPD is generally considered to be a de novo genetic alteration with negligible risk of recurrence (8). Nevertheless, the mother of our patient had a miscarriage in the year before his birth of a 7-wk-old embryo with trisomy of all chromosomes (Fig. 4). This finding suggests a predisposition to aneuploidy in one of the parents. Sequencing of the maternal exome revealed no obvious mutations in genes regulating meiosis; notably, however, the father was found to be heterozygous for a K326X nonsense mutation in BRCA2. BRCA2 mutations have been associated with increased UPD in breast cancer (19). Accordingly, it will be of interest to carefully investigate both parental exome sequences for potential mutations in genes regulating meiosis, as well as to determine if the father’s sperm cells exhibit increased levels of aneuploidy compared with that of controls. Taken together, these findings raise the possibility that in some cases, a predisposition to UPD may exist.

Fig. 4.

Triploid karyotype in 7-wk-old embryo. Chromosome spreads were produced from tissue obtained from an embryo spontaneously aborted at 7 wk of gestation. The miscarriage occurred in the year before the birth of the SCID patient.

In this patient, SCID was successfully treated by bone marrow transplantation. Nevertheless, our analysis revealed that the nonlymphoid blood cells as well as other mesoderm- and ectoderm-derived tissues retained UPD of the entire maternal Chr1. Therefore, this patient may be at risk for additional consequences resulting from the complete LOH of this chromosome. Indeed, exome sequencing revealed the presence of seven additional genes on Chr1 bearing homozygous mutations predicted to be deleterious to function (Table 1). Moreover there are 15 genes on Chr1 that are predicted to be affected by imprinting (20, 21); therefore, some of these genes may be silenced in the majority of the patient’s tissues. Thus, it would seem prudent for those providing care to such patients be aware that additional complications might arise, and monitor accordingly.

Our findings are unique in representing a reported case of SCID caused by UPD. Although UPD has been implicated in the development of a number of inherited syndromes, it is not widely considered as an inheritance mechanism and has only been described in isolated cases of five other primary immune-deficiency syndromes, including C4 deficiency (22), cartilage-hair hypoplasia without associated SCID (23), Chediak–Higashi syndrome (24), familial hemophagocytic lymphohistiocytosis (25), and IFN-γ receptor 1 deficiency (26). Disease in our patient was because of isodisomy of Chr 1, and of the 23 other reported cases of this type of UPD (27), only one was associated with immune deficiency (24).

Our results indicate that UPD is readily detectable by SNP array and, although our findings represent a unique instance of SCID caused by UPD, we suggest that UPD should be regarded as a novel third mode of inheritance in cases of SCID, as well as any other inherited recessive disorder in which the patient appears homozygous for an abnormal gene found in only one parent. If UPD is present, one should then consider performing high-throughput sequence analyses to determine if additional genes affected by UPD exhibit deleterious homozygous mutations or are predicted to be affected by imprinting and monitor patients accordingly. Finally, although UPD is generally regarded as a de novo genetic alteration, the finding that the mother of the patient had a previous miscarriage that exhibited triploidy (Fig. 4) suggests that in some cases of UPD, a predisposition to chromosome missegregation may exist.

Materials and Methods

Subject.

The patient presented at age 6 mo with severe GE reflux and failure to thrive and developed Pneumocystis jiroveci pneumonia at age 10 mo. He was noted to have pancytopenia, hypogammaglobulinemia, normal B cells, and NK cells but almost no T cells on flow cytometry, and he had an absent lymphocyte proliferative response to phytohemagglutinin (PHA). A bone marrow biopsy showed normal cellularity but no detectable CD45 on his leukocytes. He was referred to Duke University Medical Center, where the diagnosis of CD45 deficient SCID was confirmed (Fig. 1 A and B, and Table 1), and he received an unconditioned, rigorously T-cell–depleted haploidentical maternal bone marrow stem-cell transplant without posttransplantation GVHD prophylaxis. An adenovirus infection was diagnosed on admission with positive stool and respiratory secretion cultures, and he was treated with Cidofivir for the first 4 mo. Shortly after the medication was stopped he developed fever, leukocytosis (43,800/mm³), eosinophilia (24,528/mm³), elevated serum IgG (1,740 mg/dL) and IgE (1,700 IU/mL) levels, and diffuse lymphadenopathy with negative routine cultures. An axillary lymph-node biopsy performed on day 139 posttransplantation showed necrotizing granulomatous lymphadenitis with CD45⁺ (bone marrow donor) myeloid (neutrophils and eosinophils) and lymphoid (primarily B cells, but also T cells) cells with no evidence of malignancy or GVHD and negative stains for microorganisms.

Karyotype analysis of the biopsied node showed engraftment with 10% female donor cells and the patient’s peripheral blood T-cell proliferative response to PHA of 71,316 cpm at the time also demonstrating the presence of T-cell function. Serum immunofixation studies showed a monoclonal IgG-κ component with a lambda light chain. A repeat stool adenovirus culture was positive at 5 mo posttransplantation and Cidofivir therapy was reinitiated and continued until 8 mo posttransplantation and cultures have remained negative to date. The patient subsequently developed diarrhea at 7 mo posttransplantation and Clostridium difficile toxin was found in colonoscopy samples. He responded to metronidazole treatment.

The patient’s T-cell proliferative responses normalized at 4 mo posttransplantation (Fig. 1B). His fevers, elevated white cell counts, increased serum Ig levels, and adenopathy resolved by 12 mo posttransplantation and he is currently doing well clinically 5 y posttransplantation with normal T-cell function (Fig. 1B and Table 1) while receiving subcutaneous IgG replacement therapy. The patient was the child of unrelated Caucasian parents. There was no known family history of immunodeficiency, and the patient has a healthy older sister and younger brother. However, his mother had a miscarriage in the year before the patient’s birth of a 7-wk embryo with triploidy of all chromosomes (Fig. 4).

Immunologic Phenotype Analysis.

Serum Ig levels were determined by nephelometry. Standard four-color flow cytometry of whole blood was performed with labeled Abs to CD3ε, CD4, CD8, CD14, CD16, CD20, CD25, CD45 (clone HI30 recognizing all isoforms), CD45 RA, CD45 RO, CD56, CD62L, TCRαβ, and TCRγδ purchased from BD Biosciences. Lymphocyte proliferation was assessed by measuring [³H]thymidine incorporation into mononuclear cells following culture with optimal concentrations of the indicated stimuli as previously described (28). This analysis is not part of a clinical trial. All studies were performed with the approval of the Duke University Health System’s Institutional Review Board for Clinical Investigations, and written informed consent by the patient’s parents. Patient immune phenotype and function data were analyzed by J.L.R., R.H.B., and R.E.P. CD45 expression data analysis was performed by J.L.R., J.S., and X.-P.Z.

Cell Sorting.

Patient and healthy volunteer cells used for real-time PCR and indicated cytochip analyses were isolated from frozen peripheral-blood mononuclear cells (PBMC) using a FACS Vantage SE (Becton Dickinson) by sorting cells incubated with unlabeled mouse antibody to HLA-DR11 (Lifespan Biosciences) followed by incubation with normal goat serum and staining with RPE-goat anti-mouse Ig (Lifespan Biosciences). Sorting was required because available patient frozen PBMC had been obtained 3 mo postbone marrow transplantation but before T-cell development and contained 9.3% CD45⁺ maternal (HLA-DR11⁻) lymphocytes. Sorted patient (HLA-DR11⁺) cells stained brighter for HLA-DR11 plus secondary antibody than with secondary antibody alone. Similarly sorted HLA-DR11⁻ normal volunteer cells were collected to control for potential effects of cell sorting in subsequent experiments.

Real-Time PCR Analysis of CD45 mRNA Expression.

RNA isolated from sorted patient and healthy volunteer PBMC (TRIzol; Invitrogen) was used for cDNA synthesis (Superscript III; Invitrogen). CD45 and β-actin transcripts were then quantified by real-time PCR with a Mastercycler realplex and SYBR green master mix (Eppendorf). Three sets of CD45 primers (sequences available on request) were used for amplification of regions of exons 1–2 and exons 13–14 upstream of the patient exon 14 mutation, and exons 21–23 downstream of the patient mutation.

CD45 Sequence Analysis.

RNA isolated from PBMC (RNeasy Mini Kit; Qiagen) was used for synthesis of cDNA templates (Superscript II; Invitrogen) that were PCR-amplified with primer pairs (sequences available upon request) that generated overlapping products spanning the full length CD45 variant 1 transcript coding sequence (NM002838.3). Genomic DNA templates isolated from whole blood (DNeasy Tissue Kit; Qiagen) were also PCR-amplified with primer pairs (sequences available upon request) spanning alternatively spliced CD45 exons 4 through 6 and surrounding intron splice sites. PCR products were purified (Qiaex II Gel Extraction Kit; Qiagen) and used as templates in sequencing reactions (Big Dye Terminator Cycle Sequencing System; PerkinElmer Life Sciences). Sequencing reactions representing both strands were analyzed using an ABI 377 Prism DNA (PerkinElmer) instrument and software. The detected 1618A > T mutation was further evaluated by sequence analysis of genomic DNA obtained from the patient and his parents using exon 14 specific primers. Nucleotide numbers refer to the published cDNA sequence of full length CD45 variant 1 transcript (NM0002838.3) with start codon = 1. CD45 Sanger sequencing data were analyzed by J.L.R.

Cytogenetics Whole-Genome Array.

Total genomic DNA (100 ng) from each test sample was whole genome-amplified. Individual amplified DNA samples were each purified, fragmented, biotin labeled, and hybridized to an Affymetrix Cytogenetics Whole-Genome 2.7 M array, according to the manufacturer’s protocol. The hybridized arrays were washed using an Affymetrix GeneChip Fluidics 450 apparatus and then scanned with a GeneChip Scanner 3000 7G. Probe hybridization intensities were analyzed using Affymetrix GeneChip Command Console, and DNA copy number and allele analysis were performed using Affymetrix Chromosome Analysis Suite software, which compares all hybridization intensities against a built-in reference data set (Cytogenetics_Array.na31.v1.REF_MODEL). All tiny focal copy number peaks were checked against the Database of Genomic Variants of the Hospital for Sick Children, Toronto, to rule out the possibility that a focal peak is a known copy number variant (polymorphism). The Cytogenetics Whole-Genome array was performed and analyzed by J.P. and J.R.T. All data were submitted to the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) repository, accession no. GSE35674.

Exome Sequencing and Bioinformatic Analysis.

DNA libraries were prepared from 2 μg genomic DNA using modified Illumina Genomic PE Sample Prep Kit protocol (Illumina) where all DNA purification steps were performed with AMPure SPRI bead purification (Bechman Coulter Genomics). Coding sequences were captured using the Agilent SureSelect Target Enrichment System with the Human All Exon Kit targeting 50 Mb of sequence. The captured DNA libraries were PCR-amplified using the supplied paired-end PCR primers and sequenced in one lane of an Illumina Genome Analyzer IIx. Sample preparation and sequencing was performed by Expresssion Analysis. Sequence reads were mapped to the reference genome (hg18; http://www.ncbi.nlm.nih.gov/) using the SamTools package (29) and BWA aligner (30). Duplicated reads were removed with Picard (http://picard.sourceforge.net/). Recalibration of base quality and indel realignment were performed using the GATK package (31). Single-nucleotide variants and indel variants were identified using the Unified Genotyper caller of GATK package from multiple samples, including samples from other projects. Mutations were annotated with SeattleSeq Annotation (http://gvs.gs.washington.edu/SeattleSeqAnnotation/). A SQL database was created from the annotated dataset. For maternal UPD analysis, alleles that fulfill the following requirements were identified: (i) heterozygous in mother’s sample; (ii) homozygous for wild-type allele in father’s sample; (iii) homozygous for wild-type allele in other unrelated samples in the dataset; (iv) homozygous for alternative allele in child’s sample. For paternal UPD analysis, alleles that fulfill the following requirements are identified: (i) heterozygous in father’s sample; (ii) homozygous for wild-type allele in mother’s sample; (iii) homozygous for wild-type allele in other unrelated samples in the dataset; (iv) homozygous for alternative allele in child’s sample. Manual examination was conducted with TViewer (http://tviewer.sourceforge.net/) of SamTools to identify high confidence mutations from the raw sequence data. Analysis of the exome sequence data were performed by A.L., B.L., and S.P. All data were submitted to the NCBI Database of Genotypes and Phenotypes (dbGaP) repository, accession no. phs000479.v1.p1.

To predict the phenotype of missense mutations, we obtained sequence-based and structure-based information on each mutation. The mutations were mapped to Uniprot sequences (32), and these sequences were obtained from the Uniprot Web site. We used PSI-BLAST (33) to search for homologs of the Uniprot sequence in the Uniref100 database and to calculate a position-specific scoring matrix (PSSM) for each sequence. The homologs with greater than 35% sequence identity to the queries were aligned with the program Muscle (34) and the program AL2CO (35) was used to calculate a conservation score for each sequence position. We used the program Disopred (36) to predict whether each mutation position was present in ordered regions or intrinsically disordered regions of the proteins. We also used the Pfam Web site to determine whether the mutations were located in defined domains according to Pfam (37). Finally, mutations were mapped to protein structures using profile-profile alignment search methods developed by us (38).

For the mutations with available structures, we used a support vector machine (SVM) trained on 4,600 disease-associated mutations in 1,491 human proteins and 4,600 mutations between human proteins and primate orthologs. The features used included surface accessible area in the biological assemblies and monomers of known structures, the PSSM scores of the wild-type and mutant amino acids, and the conservation scores. For those mutations not present in any structure and predicted to be ordered, we used a separate SVM without the structural features. For those mutations without structures and predicted to be disordered, we used a third SVM trained on 463 mutations in disordered regions of 261 different human proteins. Table 2 shows the predicted phenotypes of the mutations. If the deleterious probability is greater than 50%, then the prediction is that the mutation is deleterious; otherwise, the mutation is predicted to be neutral. The analysis predicting the likelihood that mutations are deleterious to protein function was performed by Q.W. and R.L.D.