X-Linked Lymphoproliferative Syndrome Presenting as Adult-Onset Multi-Infarct Dementia

Abstract

Pathogenic hemizygous variants in the SH2D1A gene cause X-linked lymphoproliferative (XLP) syndrome, a rare primary immunodeficiency usually associated with fatal Epstein-Barr virus infection. Disease onset is typically in early childhood, and the average life expectancy of affected males is ∼11 years. We describe clinical, radiographic, neuropathologic, and genetic features of a 49-year-old man presenting with central nervous system vasculitis that was reminiscent of adult primary angiitis but which was unresponsive to treatment. The patient had 2 brothers; 1 died of aplastic anemia at age 13 and another died of diffuse large B-cell lymphoma in his sixties. Exome sequencing of the patient and his older brother identified a novel hemizygous variant in SH2D1A (c.35G>T, p.Ser12Ile), which encodes the signaling lymphocyte activation molecule (SLAM)-associated protein (SAP). Molecular modeling and functional analysis showed that this variant had decreased protein stability, similar to other pathogenic missense variants in SH2D1A. The family described in this report highlights the broadly heterogeneous clinical presentations of XLP and the accompanying diagnostic challenges in individuals presenting in adulthood. In addition, this report raises the possibility of a biphasic distribution of XLP cases, some of which may be mistaken for age-related malignancies and autoimmune conditions.

INTRODUCTION

X-linked lymphoproliferative syndrome (XLP1 [MIM 308240]) is a rare, X-linked recessive, primary immunodeficiency characterized by an abnormal response to Epstein-Barr virus (EBV). XLP is attributed to pathogenic variants in the SH2D1A gene (Src homology 2 domain-containing gene 1A), which encodes the signaling lymphocyte activation molecule (SLAM)-associated protein (SAP) expressed in T cells, natural killer (NK) cells, and invariant NK cells (1). In the absence of SAP, SLAM family members become potent inhibitors of the immune response, causing widespread dysregulation, including suppression of CD8 T cells, NK-cell activation, and defective antibody production (2).

Men with XLP may present with mononucleosis, hemophagocytic lymphohistiocytosis, hypogammaglobulinemia, and lymphomas (3). Rare manifestations include aplastic anemia and central nervous system (CNS) vasculitis (1). The incidence in males is about 1–3 per million, and the median age at presentation is 3–4 years (4). Median life expectancy for male patients is 11 years (range: 2–69 years) (4). Only about half of patients survive to adulthood (4, 5).

We describe a 49-year-old man presenting with ultimately fatal CNS symptoms and progressive multifocal brain lesions; his brain biopsy and autopsy showed lymphocytic meningoencephalitis and perivasculitis (Fig. 1A). Exome sequencing of the patient and his deceased brother, who presented in his sixties with diffuse large B-cell lymphoma (DLBCL), revealed a novel hemizygous variant in SH2D1A (chrX:123480527G>T, c.35G>T, p.Ser12Ile), hereafter termed S12I. The phenotypes, family history, genetic analysis, and structural and functional assessment of the mutant SAP protein were consistent with a diagnosis of XLP.

FIGURE 1.

Radiologic and pathologic findings of CNS lymphocytic vasculitis and disease progression. (A) Pedigree of the family carrying a novel SH2D1A variant. Standard pedigree symbols are used; WES indicates individuals that had exome sequencing. Red arrow indicates the proband; d = age at death; numbers inside the pictograms = number of children. (B) Initial magnetic resonance scan of the proband (early in the disease course) shows bilateral multifocal areas of abnormal signal in the cortex, with relatively little edema on fluid-attenuated inversion recovery (FLAIR) and diffusion-weighted images (DWI) (B1 and B2). There is patchy enhancement in the abnormal areas (B3). (C) With disease progression, new areas of abnormality were detected 1 month later on FLAIR (C1, C2), DWI (C3), and postgadolinium (C4) images. Site of the brain biopsy is indicated on the left frontal lobe (arrow). (D) Nine months later, after treatment with corticosteroids and chemotherapy, no new abnormalities were detected on FLAIR images (D1, D2), and enhancement was decreased (D3). (E) Progression of disease is visible 18 months later as new areas of abnormality on FLAIR (E1, E2), DWI (E3), and postgadolinium (E4) images. (F) The final imaging study, 20 months later, shows extensive new areas of abnormality on FLAIR images, with evidence of significant cortical edema (F1, F2), and new abnormalities in the diffusion study (F3, F4). (G) Postmortem brain had focal rust discoloration of the dorsal leptomeninges and a 0.7-cm lesion in the superior frontal gyrus (consistent with the biopsy site [Bx]). The sulci and gyri revealed no cortical atrophy over the convexity or the medial temporal lobe. Multiple areas of cortical softening were evident, particularly in the posterior and the medial cortical areas (e.g. cuneus and precuneus; paracentral gyrus) and in the convexity watershed between the anterior and middle cerebral arteries. (H–J) Sequential sections through the supratentorial tissues showed that the ventricles were undilated and undisplaced. The cortical gray mantle had multiple areas of softening, some of which were hemorrhagic (hematoma [Hem], white arrows). Lesions in the occipital lobe had liquefactive changes, less so in the more-anterior lesions. A gradient of decreasing severity was observed from the occipital pole to the frontal pole, with more lesions in the parasagittal region than the lateral convexity. The temporal lobe was largely spared, except in the most posterior areas. (K, L) Extensive perivascular inflammation in the leptomeninges was immunopositive for CD3 (T-lymphocytes). (M) Fewer cells were immunopositive for CD20 (B-lymphocytes). (N) Few mitotically active cells were detected with Ki-67 immunostaining. (O, P) Perivascular inflammation was evident with CD45 immunostains in leptomeningeal arterioles. (Q) Small arterioles in leptomeninges with vacuolation and degeneration of smooth muscle cells in the media. (R) Many lipid-laden macrophages (MP), duplicated elastic lamina (arrows) in the intima (1-μm section; toluidine blue stain). AD indicates adventitia; LU, vascular lumen; MP, macrophages; SMC, smooth muscle cells. (S–Z) Electron micrographs showing a large vacuole [V] present in the outermost layer of media of a small leptomeningeal artery. Adventitia (AD) is at the upper right corner of panels S and T. Red arrow points to the area enlarged in panel U, which shows multilaminar membranous material lining the vacuole [V]. (V) Cellular organelles and membranous material accumulate in smooth muscle cytoplasm between the vacuole [V] and the basal lamina (BL). (W) Only the membranous material remains in degenerating smooth muscle cells. The surrounding BL contains banded collagen fibers. (X–Z) Autophagic vacuoles are shown in smooth muscle cells of a cerebral arteriole. Within the vacuoles are serpiginous, lamellar structures and nonspecific, myelin-like fingerprint bodies.

MATERIALS AND METHODS

This study was approved by the Mayo Clinic Institutional Review Board.

Human Tissue Sample Processing

A brain-only autopsy was performed after receiving informed consent from the legal next-of-kin; the organ was sent to the Mayo Clinic brain bank (Jacksonville, Florida). The right hemibrain was frozen at −80°C; the left was fixed in formaldehyde. DNA was extracted from frozen cerebellar tissue. Peripheral blood from the patient’s brother was obtained from the Mayo Clinic Florida Familial Cerebrovascular Diseases Registry (Jacksonville, Florida). Additional details can be found in the Supplementary Data.

Exome Sequencing

Exome sequencing was performed with the Sure Select V5+UTR exome capture kit (Agilent Technologies Santa, Clara, CA), with 2.1 µg of genomic DNA from the patient and his brother. Sequencing was performed on the HiSeq platform (Illumina, San Diego, CA). Alignment and variant calling was performed with GenomeGPS workflow (Mayo Clinic Bioinformatics Core). Causal variants were identified using Ingenuity Variant Analysis software (Qiagen, Redwood City, CA) (Supplementary Data).

Molecular Dynamics Simulations

Nanoscale Molecular Dynamics software was used for implicit solvent molecular dynamics simulations with the CHARMM27 (Chemistry at HARvard Macromolecular Mechanics) force field (6). Conformations of SH2D1A and the bound SLAM peptide were extracted from simulations, and their interaction was scored by using ZRANK (7) after standardizing the structures with Rosetta (Supplementary Data).

Western Blots and Immunoprecipitation Assays

Wild type (WT) SAP and the S12I variant were expressed in HEK293 cells by transfecting them with pCI2.HA-SAP (WT) or the S12I mutant expression vector (8). Cells were transfected by using Lipofectamine 2000 transfection reagent (Thermo Fisher Scientific, Waltham, MA) and 2.5 μg of DNA. The next day, cells were either left untreated or were treated with a proteasome inhibitor (20 µM of MG132) for 8 hours. Cells were subsequently lysed, centrifugation clarified, separated by SDS-PAGE, and immunoblotted. For immunoprecipitation experiments, HEK293T cells were infected with recombinant vaccinia virus expressing chimeric FLAG-tagged F.CD4.2B4 and were either stimulated or unstimulated with pervanadate (PV) (9). Cells were lysed and cleared by centrifugation. Lysates were incubated with anti-FLAG-bound goat-anti-mouse agarose beads and 2 µg of the indicated GST-SAP fusion protein, rotated for 1 hour at 4°C, and washed. Proteins were eluted from the beads by boiling in 4× SDS sample buffer, separated by SDS-PAGE, and immunoblotted. Additional experimental details can be found in the Supplementary Data.

RESULTS

Clinical Description of the Family

The proband was a 49-year-old man of European descent who presented with numbness in 3 toes on his right foot and then numbness in digits 4 and 5 of his left hand. We questioned whether he had contracted Lyme disease while traveling in Central America 1 year prior. Radiographic progression of CNS lymphocytic vasculitis, despite empiric Lyme treatment, led to biopsy of a left frontal lobe lesion (Fig. 1B1–4, C1–4). The biopsy showed T-cell infiltration with lymphocytic meningoencephalitis, perivasculitis, serofibrinous exudation, focal necrosis, and extensive arachnoid fibrosis (Fig. 1K–R). Treatment was initiated with high-dose prednisone and multiple cycles of cyclophosphamide. His symptoms improved for 3 months and he appeared to be in clinical and radiographic remission (Fig. 1D1–4). He subsequently had tingling and loss of function in his right hand, diffuse weakness, and dysarthria. Follow-up imaging showed new lesions (Fig. 1E1–4). Cyclophosphamide chemotherapy was restarted. He developed difficulty with daily tasks and required a walker. The patient reported a history of mononucleosis at age 18 years, but testing was negative for Lyme disease, EBV, cytomegalovirus, herpes simplex virus, JC virus, HIV, St. Louis encephalitis virus, and West Nile virus. Cerebrospinal fluid (CSF) analysis was performed at this time and showed: glucose 66 mg/dL; protein 35 mg/dL; 0 red blood cells per microliter; <1 white blood cell per microliter. CSF testing for cryptococcal antigen, Treponema pallidum (Venereal Disease Research Laboratory, VDRL), Lyme, histoplasma, and cysticercosis was also negative. There were no oligoclonal bands and the CSF IgG index was 0.64. One month later, repeat CSF analysis showed 0 red blood cells per microliter and 2 white blood cells per microliter and cytology was negative for blasts or malignant cells. Serum IgG concentration was 498 mg/dL (reference range: 767–1590 mg/dL).

Magnetic resonance images showed cortical and subcortical T2 hyperintensities in the frontal, temporal, parietal, and occipital areas, consistent with subacute to chronic laminar necrosis, possibly attributable to infarcts. Symptoms and lesions progressed despite chemotherapy and high-dose corticosteroids (Fig. 1F1–4) and the patient died at age 51 years.

The proband had a brother who had leukemia at age 12 years and died from aplastic anemia a year later. No additional medical history was available for this individual. The proband’s other sibling was a 59-year-old man who worked in forestry. He had symptoms of Lyme disease after tick bites and received extended courses of antibiotics, including doxycycline, biaxin, and amoxicillin. He presented with intermittent fevers, dyspnea, and pleural effusions at age 64. Multiple tissue and serologic studies for Bartonella were negative. Inguinal lymph node biopsy yielded a diagnosis of DLBCL. At age 65, he had fevers and worsening dyspnea, prompting hospitalization. The patient developed severe systemic inflammatory response syndrome despite antibiotics and vasopressors and he died at age 65. Autopsy showed DLBCL involving multiple organs, including the celiac, portal, periaortic, peripancreatic, and mediastinal lymph nodes. Additional clinical findings are described in the Supplementary Data.

Brain Autopsy

The proband’s brain had extensive cortical infarction and focal hemorrhages of varying histologic age in the cerebral cortex and subcortical white matter (Fig. 1G–J). The occipital lobe had an acute hematoma (Fig. 1J). The cerebral white matter had Wallerian degeneration that was evident in descending fiber tracts at lower levels of the neuraxis, including the medullary pyramid. Sections of parietal and occipital lobes (Fig. 1H–J) had parenchymal and perivascular inflammation (Fig. 1K). The infiltrate was composed of small, atypical T-lymphocytes and many histiocytes. All affected areas had astrocytosis, activated microglia, and lipid- and pigment-laden macrophages (Fig. 1R). Cellular infiltrates had only sparse mitotic activity, ruling out a neoplastic process (Fig. 1N). B-lymphocytes were sparse and mostly in perivascular locations; in contrast, T-lymphocytes were more diffuse (Fig. 1L, M).

Small arteries had distinctive pathologic alterations (Fig. 1K, O, Q, R), with vacuolation of smooth muscle cells, as well as disruption and duplication of internal elastic lamina (Fig. 1Q, R). In advanced lesions, elastic lamina, and smooth muscle cells were replaced by collagen (Fig. 1R). Many vessels had lipid-laden macrophages and deposits of cholesterol (Fig. 1R). Noninflammatory arteriopathic changes were observed in the cerebral cortex, basal ganglia, thalamus, brainstem, and cerebellum. The leptomeninges over the base of the brain were translucent. The available blood vessels at the base of the brain showed a normal adult configuration of the circle of Willis and no atherosclerosis. Small arteries were most often affected, whereas veins and venules were not. Arteriopathy was not associated with necrosis, fibrinoid changes, or multinucleated giant cells. Thioflavin-S fluorescent microscopy showed mild amyloid angiopathy in leptomeningeal vessels. There was no colocalization of arteriopathy with amyloid angiopathy.

In summary, pathologic findings were characterized by widespread arteriopathy and focal chronic encephalitis and vasculitis. The arteriopathy featured prominent lysosomal pathology on ultrastructural studies, suggesting a lysosomal storage disease such as Fabry disease, which can be associated with stroke; however, lysosomal pathology was not detected in endothelial cells, and the ultrastructural features did not resemble those of Fabry disease (Fig. 1S–Z). Another arteriopathy in the differential diagnosis was cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy, but the patient had no family history of stroke and pathognomonic granular osmiophilic bodies were absent on electron microscopy (Fig. 1S–Z).

Exome Sequencing

Exome sequencing of both brothers revealed a shared novel hemizygous variant in SH2D1A (chrX:123480527G>T, c.35G>T, p.Ser12Ile) that was confirmed by Sanger sequencing (Fig. 2A, B). The S12I variant was not listed in the Genome Aggregation Database, but several known pathogenic missense variants are near the S12I variant in exon 1 (Fig. 2C). The Ser12 residue is conserved across mammals and in zebrafish (Fig. 2D). S12I falls within an N-terminal α helix that faces outward into the phosphotyrosine-binding pocket (Fig. 2E). In silico prediction algorithms suggested pathogenicity, and S12I was classified as a variant of uncertain significance. No other pathogenic variants were observed (e.g. in NOTCH3) that would account for the patient’s clinical presentation.

FIGURE 2.

Exome sequencing reveals a novel SH2D1A variant and molecular modeling and functional characterization predict altered signaling lymphocyte activation molecule (SLAM) binding with decreased SLAM-associated protein (SAP) stability. (A) A novel hemizygous c.35G>T (NM_002351.4) S12I variant in SH2D1A was detected in the proband and his brother. Exome sequence reads showing the variant were visualized in the Integrative Genomics Viewer. (B) Sanger sequencing confirmed the presence of the SH2D1A alteration in both individuals. (C) Protein diagram (NP_002342.1) showing clustering of pathogenic missense alterations around the S12I variant. Only missense variants in the first 25 amino acids are shown. (D) Amino acid conservation across species at and around the S12I variant. (E) Crystal structure showing SAP bound to SLAM peptide (shown in white sticks), with the 2 critical residues identified (black). The relationship of variants in the 3-dimensional structural region (colored spheres) to SLAM binding is as follows: R32Q and C42W, loss of binding; T68I, diminished binding; and T53I, normal binding. (F) Quantification of the buried surface area of the SLAM peptide by SH2D1A. The median wild type (WT) value is indicated by the horizontal line. Medians of each variant are indicated within boxplots and surrounded by violin plots. Asterisks indicate statistical significance; *p < 0.001; **p < 0.0001. Color indicates variant category, with shading within a category for visual distinction only. ExAC indicates Exome Aggregation Consortium; MAF, minor allele frequency; SASA, solvent-accessible surface area. (G) The interactions observed in simulations were scored using ZRANK and visualized as described in panel F. ZRANK scores are pseudo-energies, with lower scores indicating stronger interaction. Only pathogenic variants and select rare variants markedly altered binding. S12I was among the variants predicted to destabilize the interaction. Morra et al (10) studied selected variants and showed that T53I had normal SLAM binding [N], T68I diminished [D], and R32Q and C42W lost SLAM binding [L]. (H) Immunoblot of SAP showing loss of protein stability with the S12I variant, with rescue of protein levels after treatment with the proteasome inhibitor MG132. WT protein levels were stable even with MG132 treatment, likely because of increased stability. (I) FLAG-CD4-2B4 is immunoprecipitated in all 6 lanes (anti-FLAG immunoblot). The anti-pTyr blot on top shows that FLAG-CD4-2B4 is hyperphosphorylated in the pervanadate-stimulated lanes, as expected. The GST blot shows that GST does not interact with the phosphorylated FLAG-CD4-2B4, but both the WT and S12I GST-SAP proteins interact more strongly with the tyrosine-phosphorylated FLAG-CD4-2B4, suggesting that the GST-SAP12I variant has a functional SH2 domain that allows it to recognize and bind tyrosine-phosphorylated 2B4. The input blots show equivalent levels of GST and GST-SAP in all lanes.

Molecular Modeling of SH2D1A Variants

We generated molecular models of WT SH2D1A and our novel variant, S12I. To evaluate the potential pathogenicity of S12I in silico, we identified positive and negative comparators from population and disease databases. We calculated summary information from each simulation to determine structural and dynamic features shared between S12I and pathogenic variants. We calculated the surface area of the SLAM peptide as an indication of altered conformation (Fig. 2F). We then used the ZRANK algorithm to score interactions between SH2D1A and the SLAM peptide (Fig. 2G). Two pathogenic variants (C42Y and T53I) and 1 rare variant (A66P) had stronger predicted interactions. Five pathogenic variants, 2 rare variants, and S12I had weaker predicted interactions. Our predictions corroborated previous functional studies (10).

Functional Characterization of S12I Variant in SAP

Certain missense variants can affect the stability of SAP, with reduced half-life leading to disease. To determine the effect of the S12I variant, we expressed hemagglutinin (HA)-tagged SAP WT and S12I variant in HEK293 cells. SAP immunoblotting showed decreased protein stability for the S12I variant, with rescue of protein levels after treatment with the proteasome inhibitor MG132 (Fig. 2H). Transfection of 2-fold excess HA-S12I mutant plasmid (5 µg) showed no increased expression in MG132-untreated cells (data not shown).

Next, we determined whether residual S12I mutant protein had defective recognition and binding to tyrosine-phosphorylated 2B4 (pTyr-2B4). HEK293T cells were infected with a recombinant vaccinia virus expressing FLAG epitope-tagged CD4 2B4, which has an extracellular FLAG CD4 through the transmembrane domain linked to the cytosolic tail of 2B4. Infected cells were either unstimulated or stimulated with PV to artificially increase tyrosine phosphorylation. The clarified cell lysate was added to anti-FLAG beads containing GST or GST-SAP fusion proteins. As shown in Figure 2I, both WT and S12I GST-SAP proteins interacted with the tyrosine-phosphorylated FLAG-CD4-2B4. From these studies, the S12I variant appeared to have a functional SH2 domain that allowed it to recognize and bind pTyr-2B4; however, loss of protein stability likely caused premature degradation and decreased protein levels, resulting in the disease phenotype.

DISCUSSION

CNS vasculitis is a rare manifestation of XLP (Supplementary DataTable S1). Most cases occur with EBV infection, and Dutz et al have suggested that a defective immune response to EBV can result in systemic vasculitis (11). However, EBV infections are not essential for developing XLP or CNS vasculitis (12, 13). Most cases of vasculitis occur later in childhood (median age, ∼16 years; range: 5–31 years) (Supplementary DataTable S1). Our patient presented precipitously in adulthood after an undiagnosed illness. Similar to other cases of XLP-related CNS vasculitis, our patient quickly declined despite immunosuppression.

Our case is similar to adult primary angiitis of the CNS (PACNS), which has no known genetic cause (14, 15). PACNS involves vessels in the cerebral cortex and leptomeninges, resulting in cognitive impairment, stroke, and transient ischemic attack (14, 15). PACNS is rare (∼2 per 1 000 000 person-years) and has a 2:1 male predominance, with a median age at diagnosis of 50 years (14, 15). The cause of PACNS is unknown, but infectious agents may trigger disease (14, 15). In addition, rheumatic diseases such as polyarteritis nodosa may overlap clinically with PACNS (14, 15). Cerebral amyloid angiopathy can also trigger vasculitis (14, 15). The mild amyloid angiopathy in our patient appeared to be incidental.

The nature of the infiltrating inflammatory cells has rarely been studied in PACNS. One immunohistochemical study showed infiltration by CD45R0-positive T cells in and around small cerebral vessels, implicating memory T cells in the pathogenesis (16). It further suggested that PACNS can result from an antigen-specific immune response occurring in the wall of cerebral arteries, possibly from activation by pathogen-derived antigens. This mechanism is reminiscent of the CNS vasculitis described in XLP cases, and the clinical overlap between these disorders suggested shared underlying pathology. We are unable to determine whether the changes in this patient represent a primary arteriopathy or vascular changes secondary to inflammation. Ultrastructural changes were observed in affected regions that had other pathologic changes, but were also present in unaffected normal appearing brain tissue.

The proband’s brother presented in his sixties with DLBCL, which has been described previously with XLP. The mean age at diagnosis of lymphoma in the setting of XLP is 15 years (range: 2–40 years) (5). In patients presenting with malignant lymphoproliferative disease, tumors are primarily non-Hodgkin lymphomas (including Burkitt lymphoma) and arise predominantly in extranodal sites of the abdomen and cervical region (1, 4, 5). One study reported a high frequency of pathogenic SH2D1A variants in young males with high-grade mature B-cell lymphomas, including DLBCL (17). In DLBCL, variation in genes involved in DNA repair and immune function, including SH2D1A, have been proposed as factors driving susceptibility (18).

Rare manifestations of XLP include aplastic anemia and bone marrow failure (19, 20). Aplastic anemia in patients with XLP is typically diagnosed after EBV infection (1, 20). The proband’s other brother died of aplastic anemia, consistent with XLP; however, the mutation status of this individual is unknown.

Although the age of disease onset in XLP varies, the unusually late presentation in both patients suggests genetic or environmental modifiers. Observer bias may lead to nondetection and underestimation of prevalence in adults. Few patients presenting later in life have been described, but other primary immunodeficiencies may have atypical phenotypes, including later age of onset, in the context of hypomorphic variants. We suggest the possibility of a biphasic distribution of XLP—with some patients having onset in early childhood and others presenting later in adulthood. XLP in adults can be mistaken for other age-related malignancies and autoimmune conditions. Adult male patients with treatment-refractory CNS vasculitis should consider genetic testing for XLP, particularly if there is a family history suggestive of the disorder.