Systemic Distribution of West Nile Virus Infection: Postmortem Immunohistochemical Study of Six Cases

Abstract

Rare cases of West Nile virus (WNV)‐associated inflammation outside the central nervous system (CNS) have been reported. We evaluated the systemic distribution of WNV in postmortem tissues during encephalitis in six patients using immunohistochemistry. WNV antigens were detected in neurons of CNS (all 6 cases), kidney (4 cases), lungs (2 cases), pancreas (2 cases), thyroid (2 cases), intestine (2 cases), stomach (1 case), esophagus (1 case), bile duct (1 case), skin (1 case), prostate (1 case) and testis (1 case). In systemic organs epithelial cells were infected. In none of the six cases were viral antigens identified in hepatocytes, heart, adrenal gland, nerves, skeletal muscles, bone, vessels and fat. All cases in which viral antigens were identified in systemic organs in addition to CNS were severely immunocompromised transplant recipients. With the exception of testis and brain, most foci of infection were not associated with inflammation. While the absence of inflammation may in part be due to patient immunosuppression or to possible transient nature of any host response, compartmentalization of viral antigen to the luminal region of epithelial cells may sequester WNV from immune recognition. Comparison of our findings with previous reports suggests that patients with WNV encephalitis can have widespread systemic infection.

INTRODUCTION

West Nile virus (WNV) was first isolated from the blood of a febrile woman in the West Nile province of Uganda in 1937. Historically, this mosquito‐borne flavivirus has been associated with periodic epidemics of febrile illness and sporadic encephalitis in Africa, the Mediterranean basin, Eastern Europe, southwest Asia and Australia (14). The virus first appeared in the western hemisphere during an outbreak of encephalitis in New York City in 1999 (35). Since then, it has spread dramatically across North America and southward into Latin America and the Caribbean (17).

The transmission of WNV in North America increases during warmer months, with peak activity from July through October (36). From 1999 through December 31, 2006, 23 886 cases of human WNV disease were reported in the USA (to the Centers for Disease Control, Atlanta, Georgia) with onset of illness as early as April and as late as December. Of these 23 886 cases, 9800 were reported to be neuro‐invasive disease, 13 439 were West Nile fever, and 647 had unspecified or other clinical presentations. There were 934 reported deaths because of WNV over the same period (Updates on reported human cases can be found at http://www.cdc.gov/ncidod/dvbid/westnile/index.htm).

The majority (about 80%) of human infections with WNV are clinically silent, and 20% of infections result in self‐limited fever. Typically, less than 1% of infected patients develop neuro‐invasive disease, which manifests as an overlapping spectrum of meningitis, encephalitis, and/or myeloradiculitis/poliomyelitis/flaccid paralysis (16, 50). Immunosuppressed and elderly persons are at higher risk of neuro‐invasive disease and long‐term sequelae (22, 36). Diabetes, hypertension and cerebrovascular disease have also been implicated as other possible risk factors for neuro‐invasive disease (17, 34). The higher incidence of WNV encephalitis during the current US infection suggests the introduced strain is more neurovirulent. Systemic infection is characterized by the acute onset of fever, headache, fatigue, malaise, muscle pain and weakness; gastrointestinal symptoms and a transient macular rash on the trunk and extremities are sometimes reported (17, 58). The clinical severity of WNV encephalitis ranges from mild disorientation to coma and death, and many patients manifest movement disorders, including severe tremors and Parkinsonism (50, 55). In approximately 13% of patients with neuro‐invasive WNV disease, WNV infection of spinal motor neurons causes acute asymmetric flaccid paralysis, similar to that seen with poliomyelitis (24).

The distribution of WNV in human tissues during infection has not been systematically described. Therefore, the cellular tropism of the virus in humans remains to be fully evaluated. Reports of WNV in non‐CNS tissues are rare. The presence of WNV in non‐CNS tissues of US patients was first reported in blood in 2002 by Huang et al, and since then only a few other cases have been reported. The 52 reported cases of fatal neuro‐invasive WNV disease in the USA are reviewed and summarized in the Supplementary Table S1.

Histopathologic examination of central nervous system (CNS) tissues from patients with WNV neuro‐invasive disease shows perivascular inflammation, microglial nodules, neuronophagia, and variable necrosis and neuronal loss, with pathologic changes concentrated in the brainstem, deep gray nuclei and anterior horns of the spinal cells (1, 3, 5, 6, 7, 9, 10, 12, 13, 15, 19, 20, 21, 25, 26, 28, 37, 40, 44, 45, 46, 47, 48, 52, 53). Spinal cord inflammation was seen in 17 of 23 patients who died with WNV neuro‐invasive disease (15). Inflammation was more prominent in the anterior horns than in the posterior horns of nine patients. Foci of demyelination, gliosis and occasional perivascular infiltrates may be found in persons with prolonged clinical courses (15).

Generally, the presence of WNV has been documented in persons who died early during the course of WNV infection (15, 21); however, persistence of WNV in human brain tissue samples 4 months after initial diagnosis has been recently documented (40). Using immunohistochemical (IHC) staining, WNV antigen has been demonstrated variably in brain, kidney, spleen, liver, lung, skin, bone marrow and intravascular mononuclear cells (1, 9, 10, 15, 19, 23, 37, 38, 40, 46, 53). IHC staining is more often positive than viral culture, showing WNV antigens in approximately 50% of fatal WNV neuro‐invasive disease cases. IHC staining is more commonly positive in patients who died during the first week of illness when viral antigen concentrations in CNS tissues are high (15), although these studies have been conducted primarily in immunocompetent persons. Persistence of viral antigen for considerably longer time periods has been demonstrated by IHC in patients with severe immunocompromising conditions, such as solid organ or bone marrow transplant recipients (15, 40). Viral antigens are found principally within neurons and neuronal processes, prominently in the brainstem, basal ganglia, thalami and spinal cord anterior horn cells, consistent with WNV’s neurotropic reputation. In general, viral antigens are focal and sparse, except in severely immunosuppressed patients where they are detected extensively throughout the CNS (15). However, there is one recent report suggesting glial cell infection (40). We have previously reported confocal florescent laser scanning microscopic visualization of WNV antigens in infected human brain tissue, but no infected systemic cells were visualized in screening sections of the lungs, heart, liver, spleen, kidneys and pancreas from this patient (37).

Rare cases of clinically and/or histologically confirmed WNV‐associated hepatitis, pancreatitis, myocarditis, cardiac dysrhythmia, myositis, rhabdomyolysis, orchitis, nephritis, chorioretinitis, uveitis, vitritis and optic neuritis have been reported (2, 10, 12, 15, 24, 27, 33, 37, 41, 42, 46, 53). There are few systematic surveys of WNV in human postmortem tissues, apart from the CNS, during neuro‐invasive WNV infection. WNV has been identified at autopsy in renal and splenic tissues by culture, IHC and reverse transcriptase polymerase chain reaction (RT‐PCR) (15, 37); in testicular tissue by electron microscopy (EM) (10, 53); and in liver, lung, bone marrow, intravascular mononuclear cells and skin by IHC (38). WNV infection of non‐CNS tissues has important implications for organ and blood donations. The aim of this study was to characterize the systemic distribution of WNV infection in human postmortem tissue in six patients and compare this with previously published reports.

MATERIALS AND METHODS

Case definition. A patient was considered to have neuro‐invasive WNV infection if there were clinical and histopathologic evidence of encephalitis and there was positive serology for WNV or presence of flavivirus antigens in brain tissue sections by IHC.

Histology. Hematoxylin and eosin (H&E) stained brain tissue sections were evaluated for inflammatory lesions in representative slides that were taken from the following CNS locations: cerebral cortex, hippocampus, thalamus, basal ganglia, midbrain, pons, medulla, cerebellum and spinal cord. The presence of meningitis and neuritis was also recorded. Histopathologic studies of non‐CNS tissues were also performed.

Immunohistochemistry.     Paraffin‐embedded sections were deparaffinized in Histoclear 3X5 minutes (National Diagnostics, Atlanta, GA) and rehydrated through graded alcohols (100%, 95% and 75% each for two minutes) and PBS for 5 minutes. Endogenous peroxidase was inactivated by immersing in 3% H₂O₂ for 20 minutes. Antigen unmasking was performed by bringing sections to a 95°C 1X Target Retrieval Solution (Dako Corporation, Carpinteria, CA, USA) for 30 minutes, cooling 30 minutes to room temperature, tissue sections were washed twice in H₂O and PBS, and tissue section were blocked with protein blocking agent (Thermo Shandon, 407501) for 30 minutes. Tissue sections were then incubated for 2 h at room temperature or over night at 4°C with Rabbit anti‐WNV (1:1000, Galveston, TX T35502) primary antibody (56). After being incubated with the primary antibody, slides were washed three times with PBS Tween 20 and goat anti‐Rabbit biotinylated link antibody (1:200, Dako Corporation) fluroprobe‐labeled streptavidin, and nuclear counterstain was sequentially applied. Positive controls included formalin‐fixed, paraffin‐embedded human CNS and non‐CNS tissue sections from patients with culture‐confirmed infections with WNV. Negative controls included CNS and non‐CNS tissue sections of patients with encephalitis caused by human immunodeficiency virus (HIV), or with meningitis caused by Streptococcus pneumoniae. Negative controls for each specimen also consisted of tissue sections incubated with normal mouse immunoglobulin or glial fibrillary acidic protein (GFAP) specific antisera rather than WNV specific antibodies. All the stained slides were independently evaluated by two authors (HBA and CAW), and then these two authors jointly assessed each slide to resolve whether a slide was positive or negative.

Confocal microscopy. Double or triple label confocal florescent laser scanning microscopy using antibodies for Pancytokeratin, CD3, CD68 and WNV was performed as previously described (37).

RESULTS

All six cases of neuro‐invasive WNV studied had serologic evidence of WNV infection. Table 1 lists the clinical and diagnostic characteristics for the six cases of neuro‐invasive WNV disease. Three of the six cases studied have been previously reported, and the respective citations are listed in Table 1. All six cases lived in states that reported animal or other human WNV infections in 2002. Their mean age was 57.3 years (range, 26 to 87 years); three (50%) were younger than 60 years of age, and four (66.7%) were male. Underlying medical conditions were reported in four patients, and three patients were severely immunocompromised transplant recipients. All patients presumably acquired the disease through mosquito bites.

Table 1.

Clinical and diagnostic characteristics of the six cases of neuro‐invasive WNV infection. Abbreviations: WNV = West Nile virus; WID = WNV illness duration; H&E = hematoxylin and eosin; IHC = immunohistochemistry; PCR = polymerase chain reaction; NR = not reported; M = male; F = female; ND = not done; CSF = cerebrospinal fluid.

Case	Citation	Age/sex (years)	Predisposing history	WID (days)	WNV serology positive samples	H&E inflammatory lesions positive organs	WNV IHC positive organs	WNV PCR positive organs
1*	37	87/M	None	11	CSF, Serum	Brain	Brain	Brain, CSF, spleen, kidney
2†	10, 53	43/M	Renal transplant, diabetes mellitus	14	Serum	Brain, kidney, testes, skeletal muscle	Brain, kidney, lung, pancreas, thyroid, small bowel, prostate, testes, skin, bile duct, stomach	Brain
3	NR	75/F	Diabetes mellitus, cerebrovascular disease, high blood pressure	8	CSF	Brain	Brain, kidney, esophagus	ND
4	NR	26/M	Small bowel transplant	142	CSF, Serum	Brain	Brain, kidney, lung, pancreas, thyroid, colon	ND
5	28	46/F	Non‐Hodgkin lymphoma, bone marrow transplant	17	CSF, Serum	Brain	Brain, kidney	Brain, blood
6	NR	66/M	None	18	CSF, Serum	Brain	Brain	ND

Summary of demographic and pathological findings in six cases of neuro‐invasive WNV presumed secondary to mosquito bite.

^* WNV culture was positive on the brain, CSF, spleen and kidney of Case 1.

^† WNV electron microscopy was positive in the testis of Case 2.

Glial nodules with variable loss of neurons and perivascular cuffing by mononuclear cells were observed in all six cases. In general, mononuclear inflammation and loss of neurons were most prominent in the gray matter of the thalamus, medulla, pons and midbrain, whereas the cerebral cortex and cerebellum showed fewer inflammatory nodules and less perivascular inflammatory cell cuffing. Inflammation was seen in all six spinal cord samples and was considered extensive in four cases. Inflammation was usually seen in both anterior and posterior horns, but inflammation was more prominent in the anterior horn in all cases. Lymphocytic meningitis was observed in two of the six cases. Lymphocytic neuritis of cranial or spinal nerves was present in one of the six cases of neuro‐invasive WNV infection studied. The histology of all non‐CNS tissues (heart, lung, liver, spleen, kidney, pancreas, stomach, intestine, bone, adrenal gland, testes, blood vessels and peripheral nerves) studied from the six cases with neuro‐invasive WNV infection were unremarkable, with the exception of the renal transplant patient (Case 2) whose testes, psoas muscle and kidney showed interstitial lymphocytic inflammatory infiltrates.

The distribution of WNV in systemic and CNS tissues is summarized in Table 2. Figure 1 illustrates examples of the immunofluorescence staining in one case with the most abundant immunofluorescence positivity (Case 2). Positive IHC staining of CNS tissue was obtained in all six cases. With respect to the various regions of the CNS, WNV antigens were detected by IHC in the brainstem (5 cases), spinal cord (5 cases), cerebellum (5 cases), thalamus (4 cases), basal ganglia (2 cases), hippocampus (2 cases) and neocortex (1 case). IHC staining showed viral antigens throughout the cytoplasm of neurons and neuronal processes, usually associated with mononuclear inflammation of variable intensity. In most cases, infected cells were sparse, with staining in only one or two neurons per section. However, cases with severe immunosuppression showed regions with extensive viral antigens in most of the neurons of the thalamus, pons, medulla, midbrain, cerebellum and cerebral cortex, not always commensurate with inflammation.

Table 2.

Immunohistochemical results in tissues of six cases of neuro‐invasive WNV infection. Abbreviations: CNS = central nervous system; ND = not done; NA = not applicable; WNV = West Nile virus.

Organ	Case number						Summary
	1	2	3	4	5	6
CNS
Neocortex	Positive	Negative	Negative	Negative	Negative	Negative	1/6
Hippocampus	Positive	Negative	Negative	Negative	ND	Positive	2/5
Thalamus	Positive	Positive	ND	Positive	ND	Positive	4/4
Basal ganglia	Positive	Positive	Negative	Negative	ND	Negative	2/5
Brainstem	Positive	Positive	Positive	Negative	Positive	Positive	5/6
Cerebellum	Positive	Positive	Positive	Negative	Positive	Positive	5/6
Spinal cord	Positive	ND	Positive	Positive	Positive	Positive	5/5
Kidney
Glomeruli	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Tubules	Negative	Positive	Positive	Positive	Positive	Negative	4/6
Lung
Bronchi	Negative	Positive	Negative	Positive	Negative	Negative	2/6
Alveoli	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Pancreas
Exocrine glands	Negative	Positive	Negative	Positive	Negative	Negative	2/6
Duct epithelia	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Thyroid epithelia	ND	Positive	Negative	Positive	Negative	Negative	2/5
Intestinal epithelia	ND	Positive	Negative	Positive	Negative	ND	2/4
Prostate	ND	Positive	NA	ND	NA	ND	1/1
Testes
Epithelia	ND	Positive	NA	ND	NA	ND	1/1
Interstitial cells	ND	Negative	NA	ND	NA	ND	0/1
Skin
Epidermis	ND	Negative	ND	ND	Negative	ND	0/2
Eccrine glands	ND	Positive	ND	ND	Negative	ND	1/2
Liver hepatocytes	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Bile duct epithelia	Negative	Positive	Negative	Negative	Negative	Negative	1/6
Spleen
Red pulp	Negative	Negative	ND	Negative	ND	Negative	0/4
White pulp	Negative	Negative	ND	Negative	ND	Negative	0/4
Lymph node	ND	Negative	ND	Negative	ND	Negative	0/3
Heart	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Stomach	ND	Positive	ND	ND	Negative	ND	1/2
Esophagus	ND	ND	Positive	ND	ND	ND	1/1
Skeletal muscle	ND	Negative	ND	ND	Negative	ND	0/2
Bone
Osteocytes	ND	ND	ND	ND	Negative	Negative	0/2
Marrow	ND	ND	ND	ND	Negative	Negative	0/2
Peripheral nerves	Negative	Negative	ND	Negative	ND	ND	0/3
Blood vessels	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Adipose tissue	Negative	Negative	Negative	Negative	Negative	Negative	0/6
Adrenal gland	ND	Negative	Negative	Negative	Negative	ND	0/4

Positive, many WNV immunoreactive cells with complete filling of cytoplasm present; Negative, no WNV immunoreactive cells with complete filling of cytoplasm present.

Figure 1.

Examples of H&E (A,C,E,G,I,K) and immunofluorescent distribution of WNV antigen (Green; B,D,F,H,J,L) are demonstrated from Case 2. Kidney (Figure 1A,B) shows abundant viral antigen in select distal renal tubules. WNV antigen is particularly focused on the epithelial surface independent of the minimal inflammatory response present. Lung tissue (C,D) shows moderate autolysis and minimal inflammation despite abundant WNV antigen on apical surface of bronchial epithelium. Exocrine pancreas (E,F) showed no inflammation despite abundant regions of epithelial WNV infection. In the liver (G,H) viral antigen was limited to select regions of the bile duct epithelium. For the most part the skin (I,J) showed no inflammation despite abundant WNV antigen within eccrine glands. The testes (K,L) were severely inflamed and showed abundant WNV antigen within portions of the seminiferous tubules. H&E = hematoxylin and eosin; WNV = West Nile virus.

CD68 immunostaining revealed profuse microglial activation in the gray and white matter accentuated in regions delineated above. Gilial fibrillary acid protein (GFAP) immunostaining highlighted moderate global astrocytosis. CD20 and CD3 immunostaining revealed sparse perivascular B‐lymphocytes and variable parenchymal infiltration by T‐lymphocytes. CD4 and CD8 immunophenotypic characterization of the T‐lymphocytes revealed few CD4 positive T‐helper cells with a marked predominance of infiltrating CD8 positive T‐suppressor/cytotoxic cells. Confocal florescent laser scanning microscopy using antibodies for CD3, CD68 and WNV revealed diffuse positive cytoplasmic viral immunofluorescence of many neurons in the substantia nigra, spinal anterior horn cells, pontine neurons, neurons of the CA2 and 3 of the hippocampus, the entorhinal cortex and neurons of the thalamus with rare neurons in the caudate nucleus and arcuate nucleus. Triple label staining for WNV and leukocyte markers CD68 and CD3 confirmed H&E impression of T‐lymphocyte infiltration and macrophage engulfment of infected neurons (Figure 2).

Figure 2.

Top images: Paraffin section of human brain with WNV encephalitis immunostained for macrophages (CD68‐Red) and WNV (Green), show infected neurons in region of macrophage infiltration. Occasional neurons appear to be engulfed by ameboid macrophages. Bottom images: Paraffin section of human brain with WNV encephalitis immunostained for T‐cells (CD3‐Red) and WNV (Green), show infected neurons with infiltrating CD3 T‐cells. Occasional T‐cell nodules surround fragmented neurons. WNV = West Nile virus.

With respect to the non‐CNS tissues, WNV antigens were detected by immunofluorescence in the tubules of kidney (4 cases) (Figure 1A,B), bronchi of lung (2 cases) (Figure 1C,D), pancreas (2 cases) (Figure 1E,F), bile duct (1 case) (Figure 1G,H), skin (1 case) (Figure 1I,J), testis (1 case) (Figure 1K,L), thyroid (2 cases), intestine (2 cases), stomach (1 case), esophagus (1 case) and prostate (1 case). Except for the kidney and testes, most viral antigens were detected in non‐CNS locations with minimal inflammation. While this may be due to the subject’s immunosuppression, the distribution of viral antigens in the infected epithelium demonstrated a clear polarity for the apical surface, away from the basement membrane and thus distal to immune surveillance (Figure 3). In the testes where WNV antigen did not show a polar distribution, inflammation was more pronounced. In all six cases, no viral antigens were identified in; liver, spleen, heart, adrenal gland, lymph node, peripheral nerve, skeletal muscle, bone, blood vessel and fat. Occasional individual cells showed non‐specific staining in lymph nodes and spleen; however, these could not readily be distinguished from rare positive foci in negative control immunohistochemistry. These rare non‐specific staining was considered negative, as it did not demonstrate the characteristic staining of the complete filling of cell cytoplasm consistent with the positive staining seen on the positive control WNV case. WNV antigens were detected in 10 different non‐CNS organs in the renal transplant patient (Case 2), whose testes, kidney and psoas muscle showed interstitial lymphocytic inflammatory infiltrates. Of the other five cases, WNV antigens were detected in five different non‐CNS organs in one case (Case 4), two different non‐CNS organs (Case 3) in another case, one non‐CNS organ in one more case (Case 5), and in no non‐CNS organ in the remaining two cases with no underlying medical conditions (Cases 1 and 6). Cases with relatively severe immunosuppression following transplantation showed viral antigens in more systemic organs in addition to the CNS.

Figure 3.

Paraffin section of intestine immunostained for epithelial cells (Pancytokeratin‐Red) and WNV (Green) shows infected epithelial cells containing viral antigen predominantly in cytoplasm on the apical side of the nuclei. WNV = West Nile virus.

DISCUSSION

This study confirms widely disseminated systemic WNV infection in patients with fatal neuro‐invasive WNV disease. Since the introduction of WNV to the USA, there has been concern about the spectrum and outcomes of this disease in immunosuppressed patients. Severe WNV disease and fatalities are more frequent in the elderly population, in whom underlying chronic, possibly immunocompromising, conditions are more common (22, 40). However, host factors that increase the risk of developing neuro‐invasive WNV disease with a fatal outcome, other than age and immunosuppression, have yet to be clearly identified (17, 34).

Computed tomography of the head is generally normal and not helpful in the diagnosis of neuro‐invasive WNV, except to exclude other etiologies of encephalopathy. Brain magnetic resonance imaging (MRI) may frequently be normal, even in severe encephalitis (8, 49). However, when abnormalities are noted on MRI, they include prominent signal abnormalities in the deep gray matter, predominantly in the basal ganglia, thalamus, brain stem and cerebellum of patients with encephalitis, and in the anterior spinal cord in patients with poliomyelitis‐like syndrome (50). These lesions are particularly notable on T2‐ and fluid attenuation inversion recovery (FLAIR) sequences (8, 49, 50). The location of these MRI abnormalities corresponds to regions of the CNS with prominent inflammation on histology and virus localization by IHC of reported cases of neuro‐invasive WNV disease (1, 3, 5, 6, 7, 9, 10, 12, 13, 15, 19, 20, 21, 25, 26, 28, 37, 40, 44, 45, 46, 47, 48, 52, 53), as confirmed in this series.

In patients with WNV neuro‐invasive disease, WNV‐specific IgM is almost always detectable in serum and/or cerebrospinal fluid (CSF) by the time CNS symptoms begin (31), as confirmed in this series. However, the utility of WNV seropositivity as a diagnostic tool is limited in immunosuppressed patients, where IgM antibody appears late in the course of the infection (28, 53). Real‐time RT‐PCR is the most sensitive nucleic acid amplification test for WNV, able to detect as few as 50 viral RNA copies per milliliter (approximately 0.1 PFU/mL), which is approximately 1000‐fold more sensitive than culture (39). The sensitivity of RT‐PCR among 28 patients with serologically confirmed neuro‐invasive WNV disease was 57% in CSF and 14% in serum (30). Using RT‐PCR, WNV RNA has been detected in serum, blood, CSF, brain, spleen and kidney (18, 51, 54).

Before 2001, attempts to culture WNV from postmortem tissues in the USA had been unsuccessful. Subsequently, WNV has been cultured from a blood sample obtained from a patient who died during the early stages of infection, before an antibody response could be mounted (21). But WNV is rarely isolated from the blood of patients with neuro‐invasive WNV disease, because IgM is present by the time neurologic symptoms develop and viremia is usually absent. WNV has been more frequently cultured from human CSF (37, 54). Recently, the virus has been isolated from postmortem human brain tissue of mostly immunosuppressed patients with high viral loads (9, 23, 37, 44). The other postmortem human tissues, apart from brain, from where the virus has been cultured are spleen and kidney (37). WNV is best isolated in cell culture or suckling mice and identified by indirect immunofluorescence assay with specific monoclonal antibodies or by RT‐PCR. Visualization of WNV particles by EM is rare, and when found, virions have been seen within the endoplasmic reticulum of neurons in the brain (9, 15, 37, 44) and testis (10, 53).

One report of WNV antigens in acutely fatal WNV hemorrhagic fever suggested WNV infection of endothelial cells (38). We did not observe endothelial infection in our subjects. This may be because our subjects survived longer and the virus may have been cleared from the endothelium or the difference may be accounted for on technical grounds. For obvious reasons human autopsies are seldom perfused to eliminate vascular blood. The kinetics of WNV viremia is such that by the time patients with encephalitis die, viremia is very low or absent. The reported WNV hemorrhagic fever autopsy showed very high terminal viremia (10⁷ virions per milliliter of plasma) (38), thus it is possible that this vascular viremia generated the impression of endothelial infection as it has in other viral infections with high viremia (57, 59).

Because four of our six patients with WNV encephalitis were immunosuppressed, the distribution of WNV described here may not reflect systemic infection in non‐immunosuppressed patients with asymptomatic infection. Nevertheless, the pronounced epithelial tropism and polar distribution of WNV suggests a cautionary note regarding the potential for asymptomatic human carriers. WNV may evade the immune system after primary acute infection by budding toward the apical surface of glandular epithelium. This polar budding may permit individuals to be chronic carriers without immune or inflammatory clearance of virus. While excretion of virus in bile and urine is unlikely to transmit to new hosts, the presence of virus in the testes leaves open the possibility of sexual transmission. The presence of virus chronically in sweat glands may foster further arthropod dissemination. Our observation of a persistence of viral antigens in systemic organs and the CNS after the initial onset of symptoms (up to 142 days in our small bowel transplant patient, Case 4) confirms recent reports in transplant recipients (15, 40). As IHC positivity indicates productive viral infection, our observation suggests that WNV can persist in certain systemic organs and CNS for almost 5 months after the initial infection in patients with severe immunocompromising conditions. This may have treatment implications for transplantation patients who contract WNV. Additionally, the chronic persistence of the apical budded WNV in the kidney and skin without any inflammation or symptoms creates therapeutic challenges for kidney and skin transplantation programs.

Numerous animal models of WNV infection have been described (4, 29, 43, 60). For economic and experimental design reasons, rodents have been most commonly studied. The various animal models in immunocompetent animals show remarkable similarity to human infection, with most animals remaining asymptomatic, but developing a brief viremia that quickly abates with production of specific neutralizing antibody. Recent studies in hamsters (56) have shown chronic or persistent WNV infection may occur. Intriguingly, the persistently infected hamsters have shown WNV infection in organs similar to what we describe here (eg, kidney and lung). The mechanism of viral persistence in immunocompetent animals remains to be fully elucidated; however, some selected viral mutations may facilitate the chronic infection (11). In contrast, experimental infection of immunosuppressed hamsters with WNV (32) produced a fulminant encephalitis with multiple organ involvement and death, as observed in our immunosuppressed patients.

ACKNOWLEDGEMENTS

This project has been funded in whole or in part with Federal funds from the National Institute of Allergy and Infectious Disease, National Institutes of Health, Department of Health and Human Services, under Contract N01 50027 HHSN266200500027C.

H. Vinters is supported in part by the Daljit D. and Elaine Sarkaria Chair in Diagnostic Medicine.

Supporting information

Table S1. Reported cases of neuroinvasive WNV deaths in the United States.

Supporting info item