Herpes Simplex Virus Encephalitis in Human UNC-93B Deficiency

See allHide authors and affiliations

Science  13 Oct 2006:
Vol. 314, Issue 5797, pp. 308-312
DOI: 10.1126/science.1128346


Herpes simplex virus-1 (HSV-1) encephalitis (HSE) is the most common form of sporadic viral encephalitis in western countries. Its pathogenesis remains unclear, as it affects otherwise healthy patients and only a small minority of HSV-1–infected individuals. Here, we elucidate a genetic etiology for HSE in two children with autosomal recessive deficiency in the intracellular protein UNC-93B, resulting in impaired cellular interferon-α/β and -λ antiviral responses. HSE can result from a single-gene immunodeficiency that does not compromise immunity to most pathogens, unlike most known primary immunodeficiencies. Other severe infectious diseases may also reflect monogenic disorders of immunity.

HSV-1 is a widespread virus that infects about 80% of young adults worldwide (1). HSE is a rare complication of HSV-1 infection, first described in 1941 (2), but it is the most common type of sporadic viral encephalitis in western countries (about 1 patient per 250,000 person-years) (3, 4). Mortality rates in HSE cases reached 70% before the advent of treatment with the antiviral drugs vidarabine in 1973 and acyclovir in 1981 (5). Most acyclovir-treated children survive, but many have neurological sequelae. The pathogenesis of this devastating viral illness is unclear, because it affects otherwise healthy patients. We hypothesized that HSE susceptibility may be inherited as a monogenic trait resulting in the specific impairment of immunity to HSV-1. This notion of pathogen-specific Mendelian immunodeficiency contrasts with the dominant paradigm, in which rare single-gene lesions confer vulnerability to multiple infections, whereas more common infections in otherwise healthy patients reflect polygenic predisposition (6). Consistent with our hypothesis, two related patients with HSE were reported in each of four unrelated families, with intervals of years between episodes affecting relatives in one or two generations (710). In addition, a recent genetic epidemiological survey of pediatric HSE in France reported a high frequency (13%) of affected consanguineous families [SOM text, note 1 (11)].

Most primary immunodeficiencies predispose subjects to multiple infectious diseases but not to HSE (12). However, two children with signal transducer and transcription activator (Stat-1) (13) and nuclear factor kappa B (NF-κB) essential modulator (NEMO) (14) deficiency were found to suffer from severe mycobacterial disease because of impaired IFN-γ–mediated immunity (15) and from HSE because of impaired interferon (IFN)-α/β– and -λ–mediated immunity (16). We therefore assessed IFN production by HSV-1–stimulated blood cells (17) from a series of otherwise healthy French children with sporadic HSE (SOM text, note 1). We detected two unrelated patients (P1 and P2), each born to first-cousin parents (SOM text, note 2). Both presented with HSE but showed no other evidence of unusual infectious disease and had efficiently controlled infections with at least nine viruses (SOM text, note 2). After 24 hours (fig. S1A) and 40 hours of HSV-1 stimulation, peripheral blood mononuclear cells (PBMCs) from these patients produced markedly lower levels of IFN-α and IFN-β, and marginally lower levels of IFN-λ than control cells from 50 healthy individuals (Fig. 1, A and B, and fig. S1B). However, these PBMCs produced normal levels of IFN-γ (Fig. 1A), tumor necrosis factor (TNF)–α, interleukin (IL)–1β, and IL-6 (fig. S1C). IFN-α (Fig. 1C), -β, and -λ (fig. S1, D and E) production in response to 10 other viruses (except possibly encephalomyocarditis virus) was also impaired. Blood leukocyte subsets—including IFN-producing dendritic cells (1821)—were, however, present in normal numbers (fig. S1F).

Fig. 1.

Impaired IFN-α, -β and -λ production by blood cells in response to viruses and TLR7, TLR8, and TLR9 agonists. (A and B) IFN-α, -β, -γ and -λ production in PBMCs left unstimulated or stimulated with HSV-1 for 40 hours and tested by enzyme-linked immunosorbent assay (ELISA). (C) IFN-α production by PBMC, assessed by ELISA, 40 hours after stimulation with intact viruses [see (11) for multiplicity of infection (MOI) values]. P2 was tested only once for human herpesvirus 6 (HHV-6). (D) IFN-β and -λ mRNA levels in PBMC, unstimulated or stimulated for 2 hours with LPS. β-glucuronidase (GUS) was used for calibration. Two experiments were carried out for P1 and a single experiment for P2. (E) IFN-α production, measured by ELISA, in PBMC in RPMI supplemented with 10% fetal bovine serum (FBS), in response to poly(I:C) (TLR3), LPS (TLR4), 3M-13 (TLR7), 3M-2 (TLR8), R-848 (TLR7/8), and CpG-C (TLR9), 40 hours after stimulation [see (11) for concentrations]. (F) IFN-α production in PBMC in serum-free medium (SFM) or in RPMI supplemented with 1% FBS, in response to poly (I:C). (G) IL-6 production, assessed by ELISA, in response to Pam2CSK4 (TLR2/6), Pam3CSK4 (TLR1/2), zymosan (TLR2/6), lipoteichoic acid (LTA) (TLR2), LPS (TLR4), flagellin (TLR5), and IL-1β, 24 hours after whole-blood stimulation [see (11) for concentrations]. (H) IL-10 production, assessed by ELISA, in response to TNF-α and phorbol-myristate-acetate-ionomycin, 24 hours after whole-blood stimulation. For experiments indicated with an asterisk (*), the mean values for each patient and control values were not significantly different. All comparisons were performed on log-transformed values. Experiments (A to C and E to H) were carried out at least thrice for P1 and twice for P2. Mean values ± SD are indicated.

Virus detection and IFN-α/β and -λ production by PBMCs may be mediated by surface-expressed Toll-like receptor 4 (TLR4) and, to a greater extent, intracellular TLR3, TLR7, TLR8, and TLR9, which can be triggered by nucleic acids mimicking viral products (17, 22, 23). The patients' cells responded normally to lipopolysaccharide (LPS), an agonist of TLR4, in terms of IFN-β and -λ mRNA production and the secretion of other cytokines (Fig. 1D and fig. S1G). However, these cells showed impaired IFN-α, -β, and -λ, IL-1β, TNF-α, and IL-6 production in response to all specific agonists of TLR7 (3M-13, R-848), TLR8 (3M-2, R-848), and TLR9 (CpG-C) tested (Fig. 1E and fig. S1G). Cells from both patients showed normal IFN and cytokine responses to the nonspecific TLR3 agonist polyinosinic-polycytidylic acid [poly(I:C)] under culture conditions, in which all control cells responded (Fig. 1, E and F, and fig. S1, G to I). Finally, the patients responded well to IL-1β and other TLR agonists (except possibly the TLR5 agonist flagellin, like some healthy controls) (Fig. 1G and fig. S1J) and to TNF-α (Fig. 1H). The two patients with HSE therefore had defects in their response to TLR7, TLR8, and TLR9 stimulation (not formally excluding the possibility of impaired TLR3 and TLR5 activation).

The cellular responses of the two patients resembled that described for mice lacking UNC-93B, an endoplasmic reticulum protein with 12 membrane-spanning domains involved in TLR3, TLR7, and TLR9 activation (TLR8 is absent in mice, and TLR5 response was not tested in mutant mice) (24). P1 was found to be homozygous for a four-nucleotide deletion (CTTT) at positions 1034 to 1037 in UNC93B1 exon 8 (1034del4) (Fig. 2A and fig. S2A) (25). P2 was homozygous for a single-nucleotide substitution at position 781 (G>A), the last nucleotide of exon 6 (781G>A). Full-length UNC93B1 mRNAs from P1 and P2 were barely detectable by reverse transcription polymerase chain reaction (Fig. 2B), and an alternatively spliced mRNA, lacking exon 6, was found in P2 (fig. S2B). Neither of the mutant alleles carried by P1 and P2 were found in a survey of 100 healthy European controls. The patients' parents and siblings who were tested were heterozygous for the corresponding mutant alleles (Fig. 2C); their cells responded normally to R-848, vesicular stomatitis virus (VSV), and HSV-1 (Fig. 2D and fig. S2C), and they did not develop HSE upon HSV-1 infection. The 1034del4 and 781G>A alleles were thus recessive for both clinical (HSE) and immunological (lack of viral and TLR responsiveness) phenotypes, strongly implicating UNC-93B deficiency as the probable cause of HSE in these two patients.

Fig. 2.

Autosomal recessive UNC-93B deficiency and impaired fibroblastic cell responses to viruses and TLR3 agonist. (A) Schematic representation of UNC93B1 gene structure. Human UNC93B1 has 11 exons (Roman numerals), encoding a protein with 12 predicted transmembrane domains (shown in gray). 1034del4 (P1) results in a frameshift that creates a premature stop codon in exon 9 at position 1339–1341. 781G>A(P2)generates an alternatively spliced mRNA, lacking exon 6, and the resulting frameshift creates a premature stop codon in exon 9 at position 1249–1251. (B) UNC93B1 cDNA in an EBV-transformed B cell line from a healthy control (C) and P1, after full-length PCR amplification. The internal amplification control was GAPDH. A shorter UNC93B1 cDNA product is shown for PBMC from C and P2. (C) Family pedigrees, with allele segregation in two families. The patients, in black, are homozygous for the mutation. All other family members tested, in white, are heterozygous. (D) IFN-α production, assessed by ELISA, in PBMC from P1 and his mother (M), father (F), and siblings (S1 and S2), in response to 40 hours of stimulation with VSV, poly(I:C), and R-848. (E) IFN-λ production, measured by ELISA, in SV40-transformed fibroblast lines from a control (C) and P1, upon stimulation with poly(I:C), HSV-1, or VSV for 24 hours. (F) IFN-λ production, measured by ELISA, in SV-40–transformed fibroblast lines from a control (C) and P1, upon stimulation with poly(I:C) for 18 hours [see (11) for concentrations]. The patient's fibroblasts were tested 48 hours after transient transfection, with an insertless vector (mock) or a pCDNA3 expression vector containing the wild-type UNC93B1 cDNA. Mean values ± SD are indicated.

Human skin–derived fibroblasts express TLR3 and respond to poly(I:C) (26). Moreover, the nonresponsiveness of fibrosarcoma cell line P2.1 to poly(I:C) results from TLR3 deficiency (27). We thus studied fibroblasts from P1 and showed that they failed to produce IFN-β and IFN-λ in response to poly(I:C), HSV-1, or VSV (Fig. 2E and fig. S2D) but responded well to TNF-α. The impaired fibroblast response to poly(I:C) contrasted with the normal response of the patient's blood cells, probably reflecting the strict TLR3 dependence of the poly(I:C) response in fibroblasts and the use of TLR3-independent pathways in PBMCs. This is consistent with the role of mouse UNC-93B in the TLR3-dependent response to poly(I:C) of peritoneal macrophages (24). We then used the patient's fibroblasts to validate the pathogenic role of the mutant UNC93B1 alleles. Upon transient transfection with a wild-type UNC93B1 allele, but not with mock vector, P1 fibroblasts regained normal IFN-β and IFN-λ secretion in response to poly(I:C) (Fig. 2F and fig. S2E). Similarly, P1 fibroblasts stably transfected with a wild-type UNC93B1 allele regained the ability to produce IFN-β upon VSV infection (fig. S2F). These complementation studies confirmed that the UNC93B1 mutant alleles were responsible for the impaired fibroblast responses to both poly(I:C) and viruses.

We further characterized the impact of UNC-93B deficiency by studying the responses to poly(I:C) in fibroblasts from P1 and a patient with interleukin-1 receptor–associated kinase (IRAK)–4 deficiency [with resulting impaired responses to IL-1β and agonists of TLR7, TLR8, and TLR9, but not TLR3 agonist poly(I:C) (17, 28)]. Poly(I:C) failed to induce the production of IFNβ and IFNλ mRNAs in P1 cells (fig. S3A), and dimerization of the intracellular viral response protein interferon regulatory factor (IRF)–3, mitogen-activated protein kinase (MAPK) p38 phosphorylation (fig. S3, B and C), and the DNA-binding activity of NF-κB (fig. S3D) were impaired. In P1 cells, NF-κB and MAPK p38 were activated normally in response to IL-1β and TNF-α. We then tested the response of Epstein-Barr virus (EBV)–transformed B cells to TLR7 and TLR8 agonists (17). With each of the cytokines examined, including IFN-α, -β, -γ, and -λ, only TNF-α was reproducibly induced in all control EBV-B cells after stimulation with the TLR7/8 agonists 3M-2, 3M-13, and R-848. Unlike control cells, those from P1 and the IRAK-4–deficient patient failed to secrete TNF-α (fig. S3E) and did not show the normal degradation of IRAK-1 in response to TLR7/8 stimulation (fig. S3F). Human UNC-93B is thus involved in the response to TLR3, TLR7, and TLR8 agonists and operates upstream from IRF-3, NF-κB, MAPK, and IRAK-1.

Finally, we explored the pathogenesis of HSE in the UNC-93B–deficient patients by investigating the possible role of this protein in the cell-autonomous control of viruses. Four hours after infection, UNC-93B– and Stat-1–deficient (13) fibroblasts sustained remarkably high rates of VSV replication (Fig. 3, A and B). Most control cells remained alive 24 hours after infection, whereas the rates of cell death for UNC-93B– and Stat-1–deficient cells increased similarly with viral titer (Fig. 3C). Similar results were obtained with HSV-1, which is less cytopathic and induces IFN less strongly than VSV in human fibroblasts (Fig. 3D). Thus, UNC-93B–deficient and Stat-1–deficient cells derived from patients displaying vulnerability to HSE displayed high rates of cytolysis after viral infection. However, in UNC-93B–deficient cells treated with recombinant IFN-α2b before viral infection, the cellular phenotype was fully complemented in terms of both viral titer (Fig. 3E) and cell viability (Fig. 3, F and G). Thus, the higher level of cell death in UNC-93B–deficient cells was a consequence of enhanced viral growth, itself resulting from impaired IFN-α/β production upon viral infection. These findings for fibroblasts may extend to neurons, providing a plausible pathogenic mechanism for HSE, but do not exclude the possible involvement of hematopoietic cells involved in antiviral immunity.

Fig. 3.

The rapid viral replication and high mortality rates of UNC-93B–deficient fibroblastic cells are IFN-α–dependent. (A) VSV viral titer, estimated on Vero cells, in SV40-transformed fibroblast cell lines from a healthy control (C) and P1, at various time points after infection with VSV. Titration was carried out twice, and a representative experiment is shown. (B) Intracellular VSV-G protein levels in SV40-transformed fibroblast cell lines from a healthy control (C), P1, and a Stat-1–deficient patient (Stat-1–/–) 5 hours after infection with VSV. This experiment is representative of three, with two different SV40-transformed control fibroblast cell lines. Nuclei were stained with Hoechst stain. (C and D) Live cell percentage, estimated by resazurin oxidation/reduction, for SV40-transformed fibroblast cell lines from a healthy control (C), P1, and a Stat-1–deficient patient (Stat-1–/–), 24 and 96 hours after infection with various MOI of VSV (C) and HSV-1 (D), respectively. (E) VSV viral titer, estimated on Vero cells, for SV40-transformed fibroblast cell lines from a healthy control (C), P1, and a Stat-1–deficient patient (Stat-1–/–), 30 minutes and 6 hours after infection. Fibroblast cells were used untreated or after treatment with recombinant IFN-α2b 18 hours before infection. (F and G) Percentage of live cells, estimated by resazurin oxidation/reduction, for the SV40-transformed fibroblast cell lines from a healthy control (C) and P1, 24 and 96 hours after infection with various MOI of VSV (F) and HSV-1 (G), respectively. Fibroblast cells were used untreated or after treatment with recombinant IFN-α2b 18 hours before infection. Mean values ± SD are indicated.

We have identified autosomal recessive UNC-93B deficiency as a genetic etiology of HSE in otherwise healthy patients. We have previously shown that the human TLR7-, TLR8-, and TLR9-IFN-α/β and -λ signaling pathways are IRAK-4–dependent and redundant for immunity to most viruses, including HSV-1 (17). Our findings thus indicate that the UNC-93B–dependent production of IFN-α/β and -λ controls HSV-1 by TLR3-dependent and/or TLR-independent pathways. Unexpectedly, UNC-93B appears to be redundant for protective immunity to most other microbes, including viruses. UNC-93B–deficient mice were susceptible to multiple infections under experimental conditions (24), whereas our patients with HSE, infected under natural conditions—the hallmark of the human model (29)—were otherwise healthy. This discovery broadens our view of primary immunodeficiencies, which should no longer be seen as restricted to children with multiple infectious diseases (6). Severe infectious diseases in the general population do not necessarily reflect complex genetic predisposition but may show Mendelian inheritance (30). Finally, our results have important therapeutic implications, as at least some HSE patients would probably benefit from recombinant IFN-α treatment, as suggested by mouse models of HSE (31), just as patients with mycobacterial disease and genetic defects resulting in low endogenous IFN-γ levels benefit from life-saving IFN-γ treatment (15).

Supporting Online Material

SOM Text

Figs. S1 to S3

References and Notes

View Abstract

Navigate This Article