key: cord-0042287-xgi0y7ep
authors: Tyler, Kenneth L.
title: Host and viral factors that influence viral neurotropism I. Viral cell attachment proteins and target cell receptors
date: 2003-03-17
journal: Trends Neurosci
DOI: 10.1016/0166-2236(87)90098-1
sha: fc6c3e1ae219e14ae0779f83ee027255ad056d4b
doc_id: 42287
cord_uid: xgi0y7ep

Since the time of Charcot, over a century ago, it has been recognized that the clinical signs and symptoms produced by viral illnesses involving the CNS are due in large part to the specific regions of the nervous system injured. As the list of viruses capable of infecting the CNS expanded rapidly during the earlier part of the 20th century, so too did the recognition that individual viruses exhibited specific affinities for particular regions of the nervous system, and often for particular cell types within these regions. In this part of our review we will focus specifically on the role played by the interaction between specific viral proteins (‘cell attachment proteins’) and receptors on target cells in determining the tropism of certain viruses for the CNS. In a subsequent article, the role of host genes, the site of entry and route of spread of virus, and the contribution of tissue-specific viral genes (enhancers) to viral neurotropism will be reviewed.

Since the time of Charcot, over a century ago, it has been recognized that the clinical signs and symptoms produced by viral illnesses involving the CNS are due in large part to the specific regions of the nervous system injured. As the list of viruses capable of infecting the CN5 expanded rapidly during the earlier part of the 20th century, so too did the recognition that individual viruses exhibited specific affinities for parb'cular regions of the nervous system, and often for particular cell types within these regions. In this part of our review we will focus specifically on the role played by the interaction between specific viral proteins ('cell attachment proteins') and receptors on target cells in determining the tropism of certain viruses for the CNS. In a subsequent article, the role ofhostgenes, the site of entry and route of spread of virus, and the contribution of tissue.specific viral genes (enhancers) to viral neurotropism will be reviewed.

It has been nearly thirty years since J. J. Holland and L. C. McLaren showed that there was a strong correlation between the susceptibility of specific tissues to pathologic injury by polivirus in vivo and the capacity of tissue homogenates to bind virus in vitro. They also showed that if the receptor-binding step was bypassed by transfection of cells with poliovirus RNA, then many receptor-negative cells supported one cycle of poliovirus replication 1'2. From these studies they concluded that the presence or absence of the appropriate cellular receptor was the major determining factor in cell and tissue susceptibility to poliovirus infection. However, they clearly recognized that the virus had to be capable of reaching the target cell population in the infected host, and that for cell injury to occur, the virus had to successfully complete steps in the viral replication cycle that followed receptor binding. Recent studies on the binding of radiolabelled poliovirus type 1 (Mahoney) to human regional CNS tissue homogenates appear to conflict somewhat with this earlier view by indicating that poliovirus binding activity within the CNS is much more widespread than the restricted distribution of pathologic lesions would lead one to predict 3. This suggests that factors other than receptor distribution must play a significant role in determining poliovirus neurotropism.

The poliovirus receptor is an integral membrane present in about 3000 copies on HeLa cells 4. All three serotypes of poliovirus compete for attachment to a common receptor, which is different from that utilized by other enteroviruses. The receptor is encoded by a gene present on human chromosome 19 (Ref. 5) , and susceptibility to poliovirus infection can be transmitted to previously insusceptible cells (e.g. mouse L-cell fibroblasts) by transfection of these cells with specific DNA fragments (presumably encoding the poliovirus receptor) derived from susceptible HeLa cells 6.

The nature of the poliovirus cell attachment protein(s) has been partially clarified by the availability of high-resolution X-ray crystallographic studies of the three-dimensional structure of both poliovirus 7 and other picornaviruses, including human rhinovirus 14 (HRV14) 8 and mengovirus ~. On the outer surface of HRV14, protrusions formed by part of the outer capsid protein VP1 are surrounded by a deep (25A) 'canyon', and in mengovirus by a deep 'pit'. The canyon and pit walls are composed primarily of stretches of VP1 and to a lesser extent by portions of the carboxy terminus of outer capsid protein VP3. It has been suggested that stretches of VP1 and VP3 along the floor of the canyon may provide the site of viral attachment to cellular receptors 8. Although polioviruses do not use the same cellular receptor as HRV14, their crystallographic structure also shows 'peaks' on the outer surface of the virus composed of outer capsid protein VP1. These peaks are surrounded by a broad 'valley' analogous to the HRV14 'canyon' and the mengovirus pit '7. Poliovirus receptor binding sequences may be located in this region, and would presumably differ from the homologous sequences in HRV14, thereby accounting for the differences in receptor specificity and cell tropism.

Rabies virus provides another example of how cellular receptors and viral cell attachment proteins may interact to determine neurotropism. Following the bite of a rabid animal, rabies virus is inoculated into the subcutaneous tissues and muscles of the unfortunate victim. Virus replication in striated muscle cells may serve to amplify the size of the initial inoculum. In experimental models of rabies infection, virus rapidly localizes in neuromuscular and neurotendinal spindles and at the neuromuscular junction, before spreading via nerves to reach the spinal cord ~°. Lentz leagues demonstrated that in chick embryo myotubes, isolated murine diaphragm-phrenic nerve preparations, and muscles from mice inoculated with rabies virus in vivo, that rabies-virus binding sites appeared to co-localize with areas of acetylcholinesterase staining and o~-bungarotoxin binding 1~'12. Pre-treatment of cultured chick myotubes with either ~-bungarotoxin or (+)turbocurare inhibited rabies infection, and pre-treatment of fixed chick myoblasts with a monoclonal antibody to the ~-subunit of the acetylcholine receptor (AChR) blocked binding of radiolabelled rabies virus. These findings led the authors to propose that a rabies-specific host cell receptor was located at the neu romuscular junction, and that it was either identical to or closely associ- 11 12 ated with the nicotinicAChR ' . Cells lacking AChR can be infected by rabies virus in vitro ~3 and in vivo ~4, suggesting that host cell structures other than the AChR can also serve as receptors for rabies virus. One such receptor may be a membrane phospho-or glyco-lipid ~3. Vesicular stomatitis virus (VSV), another rhabdovirus, competes with rabies virus for binding to some cell lines, and independent evidence exists that phosphatidyl serine may serve as a VSV receptor. Gangliosides may also play a role in rabies virus binding, as preliminary studies indicate that the insertion of exogenous gangliosides into the membrane of rabies-insusceptible asialyo-CER cells results in the restitution of susceptibility to rabies virus infection 15.

The rabies envelope glycoprotein (G) appears to serve as the viral cell attachment protein. Binding may depend on the conformation of the protein in the viral envelope, as isolated G does not compete with whole virus in in-vitro binding studies 13. Lentz and his colleagues have drawn attention to the fact that portions of the amino acid sequence of G (amino acids 151-238) are homologous to regions of the amino acid sequence of certain curaremimetic snake neurotoxins. They hypothesized that this region of the rabies glycoprotein could contain the recognition site for viral binding to the AChR 16.

Another approach to identifying biologically important regions of G has been to isolate and characterize rabies virus variants that are resistant to neutralization by anti-G monoclonal antibodies. Biological characterization of a large number of these variants combined with partial amino acid sequence data have indicated that an arginine residue located at position 333 is an important marker of the pathogenic viral phenotype 17'18. Variants with a number of amino acid substitutions at this position (glutamine, glycine, isoleucine) are avirulent or have attenuated pathogenicity 17,~8. Conversely, variants isolated to date with amino acid substitutions at positions 330, 336, 338 are fully pathogenic ~8. The apathogenic rabies variants appear to compete for the same cellular receptor as the wild-type virus (in vitro) ~3. However, studies with one variant in vivo indicated that its rate of spread within the CNS, the number of neurons it infected, and the degree of cellular necrosis it induced were all considerably less than that of its wild-type parent, although the differences in pathology were felt to be quantitative rather than qualitative ~9. In another in-vivo study with two variants, it was found that unlike the wild-type virus, both variants were able to infect the lens of the rat eye after inoculation into the anterior chamber, indicating that qualitative differences in tropism may indeed exist between variants and wild-type rabies virus ~4. When compared to the wild-type virus, the variants also appeared to be incapable of spreading from the eye to the CNS via either parasympathetic oculomotor or pretectal retinopetal fibres, although they retained the capacity to spread via the sympathetic fibres in the ophthalmic nerve 14.

The mammalian reoviruses have provided another example of a group of viruses in which the nature of the cell attachment protein and the viral receptor have been extensively investigated. Reoviruses of serotype 1 and 3 infect different populations of cells within the brain, pituitary gland, and retina. Using reassortant viruses containing different combinations of genes derived from type 1 and 3, it has been shown that the viral S1 dsRNA segment, which encodes the outer capsid protein, sigma 1, determines the capacity of these viruses to infect discrete populations of CNS cells 2°'21. In order to study the tropism of reovirus type 3 for neuronal cells, a number of sigma 1 antigenic variants were isolated based on their ability to escape neutralization with anti-sigma 1 monoclonal antibodies 22. These variants all have markedly attenuated neurovirulence and restricted CNS tropism when compared to the parental type 3 virus 22-24. Comparison of the predicted amino acid sequence of the sigma 1 proteins of the variants with that of the wild-type parental virus indicates that each of the variants has a single amino acid substitution (amino acid 340 or 419) 24. In-vivo investigation of the properties of a reassortant virus containing a $1 dsRNA segment derived from one of these variants clearly shows that the single amino acid substitution in the sigma 1 protein is responsible for the attenuated neurovirulence and restricted CNS tropism of the variants 25.

The nature of a receptor on cultured lymphoid and neuronal cells for reovirus type 3 has recently been identified 26. The receptor is a monorneric glycoprotein containing sialic acid residues, which has a molecular mass of 67 kDa. The receptor appears to be structurally similar to the mammalian 13-adrenergic receptor 27. This conclusion is based on three lines of evidence: (1) anti-reovirus receptor antibody immunoprecipitates purified I]-adrenergic receptor; (2) the molecular masses, isoelectric points, and trypsin digests of reovirus receptor and ~-adrenergic receptor are identical; and (3) purified reovirus receptor binds I~-antagonists, and this binding is blocked by I~-agonists. Obviously a number of questions concerning the reovirus type 3 receptor remain to be answered. Can cells lacking 13adrenergic receptor be infected by reovirus type 3 (suggesting that more than one cellular receptor may exist)? Does the in-vivo tropism type 3 depend on binding to the 13 receptor? Does reovirus type 3 bind to a pharmacologically active ligand binding site, and do 13-agonists inhibit viral binding?

to infection and destruction of oligodendrocytes. Passive immunization of mice with certain anti-E2 MAbs can also alter the course of infection from encephalitis to dernyelinating disease 3°. These findings suggest that the E2 peplorner glycoprotein is a major determinant of pathogenesis for neurotropic coronaviruses and that alterations in this protein may result in changes in viral neurotropisrn (e.g. sparing of neurons).

The receptor for neurotropic coronaviruses on CNS cells has not been identified, although a plasma membrane receptor for MHV-A59 on intestinal epithelial and liver cells has been partially characterized 31. This receptor is a monorneric protein with a molecular mass of 100-110 kDa. The presence or absence of this receptor appears to account for some of the differences in susceptibility to MHV-A59 infection that occur in inbred strains of mice 31. Since differences in the susceptibility of some inbred mouse strains have been mapped to chromosome 7 (Ref. 32), it will be interesting to see if this chromosome encodes the MHV-A59 receptor.

In-vivo studies of persistent infection of susceptible strains of inbred mice with lymphocytic choriomeningitis virus (LCMV) have provided important insights into factors that influence neurotropism in this viral system 33'34. Some strains of inbred mice (e.g. C3H), infected at birth with LCMV (Armstrong), develop a syndrome of poor growth, retarded development and hypoglycemia, whereas others (e.g. BALB/WEHI, SWR/J) do not. Suscept- 

Information has accumulated about specific aspects of the viral cell attachment proteins of a number of other neurotropic viruses. The coronaviruses are enveloped RNA viruses, which include several neurotropic species [e.g. mouse hepatitis virus (MHV) type 3 and 4, strains A59 and JHM]. Variants of several of these viruses have been generated based on their ability to escape neutralization by monoclonal antibodies (MAb) directed against the envelope peplomer glycoprotein (E2) 28'29. Several of the E2 variants isolated to date have shown markedly attenuated neurovirulence. Instead of producing a severe acute encephalitis, a number of E2 variants produce demyelination due When different strains of LCMV are tested for their capacity to produce disease in susceptible C3H mice, striking variability is noted. Some LCMV strains (e.g. Armstrong) infect the majority of growth hormoneproducing cells, whereas others (e.g. WE) replicate in far fewer of these cells and do not alter growth hormone synthesis 33. In an effort to determine the genetic basis for this difference, reassortant viruses between LCMV Armstrong and LCMV WE were generated and studied for their ability to infect growth hormone-producing cells in the pituitary. The results of this study indicate that the LCMV small (S) RNA segment, which encodes the two viral surface glycoproteins, is responsible for determining the capacity of LCMV to infect these cells 34. Thus, the LCMV system provides another example of the importance of proteins on the surface of virions in determining viral neurotropism.

The use of mutants selected by their resistance to neutralizing monoclonal antibodies has also proved to be a fruitful approach to investigating the neurovirulence and neurotropism of mumps virus 35. Monoclonal antibodies against the hemagglutininneuraminidase (HN) surface glycoprotein have been used to select variants of the Kilham strain of mumps virus, at least one of which (M13) has significantly attenuated neurovirulence after intracerebral inoculation into suckling hamsters. In this experimental system, inoculated animals develop a syndrome characterized by lethargy, wasting, and incoordination followed after 2-3 weeks by the development of hydrocephalus. There is neuropathologic evidence of widespread cellular necrosis in the cerebrum and brain stem, severe ependymitis with obstruction of the aqueduct of Sylvius, leptomeningitis, and perb However, the severity of the ependymitis and subsequent obstructive hydrocephalus is identical to that of the wild-type virus. These results suggest that the mumps virus HN protein plays an important role in neurovirulence, and that this protein may also be involved in binding to or replication within 'neurons 35.

There has been an explosive growth in our understanding of the nature of the factors that influence the neurotropism of HIV (previously known as HTLVIII, LAY, ARV), the agent responsible for the acquired immunodeficiency syndrome (AIDS). Neuropathologists and neurologists quickly recognized that large numbers of patients with AIDS had evidence of central and peripheral nervous system dysfunction, including subacute encephalitis, vacuolar myelopathy, and polyneuropathy. Immunocytochemical procedures, in-situ hybridization, and culture of nervous system tissue obtained by biopsy or at autopsy, have provided unequivocal evidence for the neurotropism of HIV 36-4°. Using immunocytochemical staining techniques, HIV antigen has been localized primarily in the cortical grey matter, near microglial nodules, and both around and within the walls of cerebral blood vessels 38. The involvement of the CNS is widespread, although the concentration of HIV antigen-positive cells within 38 39 these regions is relatively low ' . HIV-infected cells, as defined by in-situ hybridization, immunocytochemistry, and electron microscopy, are typically monocytes, macrophages, giant cells, and capillary 38--40 endothelial cells . Viral antigen has not been detected in neurons in most studies, and is only rarely found in gila 4°.

A variety of studies have now established the fact that the HIV virus envelope glycoprotein (designated gp120 or gp110) is the cell attachment protein, and that the CD4 (T4) antigen is an essential component 41--45 of the HIV-cell receptor . The introduction of a functional cDNA encoding the CD4 antigen into a number of CD4 human lymphocyte and epithelial cell lines make these previously resistant cells susceptible to HIV infection, although the same does not appear to be true of non-human cell lines 43. In order to identify the exact epitope of the CD4 antigen with which HIV binds, a large number of anti-T4 monoclonal antibodies have been tested for their ability to block HIV infection 45. Competition binding experiments with these MAbs have indicated that the HIVbinding site on the CD4 antigen is relatively large, since antibodies to at least four distinct epitopes can interfere with virus infection. These experiments have also indicated that polymorphism of the CD4 antigen does not significantly alter the binding of HIV.

The identification of the CD4 antigen as the HIV receptor in lymphoid cells led to the search for this antigen in the CNS. CD4 mRNA sequences can be detected in homogenates of human brain tissue by northern blot analysis 43 . This mRNA appears to be of two size classes (3 kb and 1.8 kb), the larger of which is of similar size to CD4 mRNA derived from Tlymphocytes. The rough distribution of the CD4 mRNA appears to parallel the distribution of HIVinfected cells, with larger amounts in the cortex and less in the cerebellum and brainstem. The exact population(s) of cells in the CNS that contain the CD4 mRNA has not yet been clearly determined.

Attempts have been made to identify the specific region(s) of the HIV envelope glycoprotein that are important in the binding of HIV to the T4 antigen. Comparison of HIV isolates from different patients, as well as sequential isolates from the same individual, have led to the identification of 'constant' and 'variable' regions within the HIV glycoprotein 46'47. Systematic correlation of variations in the gp120 protein with differences in tissue tropism (e.g. neurotropism) of HIV isolates has not yet been described, so it remains unclear whether the observed variations in the gp120-cell attachment protein result in viruses with altered affinity for tissues such as the CNS.

A number of regions of the HIV envelope glycoprotein (gp120) have been found to show amino acid sequence homologies with various biologically interesting molecules (see below). Several of the regions are rather short (e.g. those homologous to interleukin 2A and VIP), some of the regions appear to fall within the 'highly variable '46'47 regions of the gp120 sequence (e.g. the region homologous to VIP and part of the region homologous to the immunoglobulin oc-heavy chain constant region), and in one case (IL 2A) the homology is to the transmembrane region of the protein. Data supporting any significant biological role for these homologies, either on the tropism of HIV or its effects on the nervous and immune systems, is essentially nonexistent, although this has not hampered speculation on these subjects.

One of the basic principles of neurotropism emerges from the several examples cited above. For many viruses, the interaction between specific cell attachment proteins on the surface of the virus particle and specific receptor proteins on the outer membrane of host cells is a fundamental determinant of viral neurotropism. In the subsequent article in this series we shall examine how a variety of other viral and host factors, including host genes, the site of viral entry, the route of viral spread, and the action of tissue-specific viral genes all exemplify other important determining principles of viral neurotropism. 

Proc. NatlAcad. 5ci. USA

Proc. Natl Acad. 5ci. USA

Proc. Natl Acad. 5ci. USA

We welcome In a recent article 1, McBurney and comments from our Neering discussed the possible readers. Short mechanisms involved in the maincommunications stand the best chance tenance of low levels of free of publication, and calcium in neurons. Largely by the editor reserves the analogy with other cell types, right to take extracts they defined four mechanisms: from thelongerones, uptake of calcium into mitochon-dria, uptake into endoplasmic reticulum, calcium extrusion by the plasma membrane and calcium binding by cytosolic proteins. They did not, however, take account of two uniquely neuronal sites of calcium sequestration. In presynaptic nerve terminals, synaptic vesicles may be a major site of storage for calcium. The use of cytochemical techniques for the detection of calcium stores has demonstrated such a role for synaptic vesicles 2'3, and isolated brain synaptic vesicles have been shown to accumulate calcium through the action of a Ca 2+ ATPase 4.At many synapses, the postsynaptic site of calcium sequestration may be a unique neuronal organelle, the spine apparatus 5.In the dendritic spines of many central neurons, the spine apparatus (Fig. 1) is the only organelle present. Cytochernical studies have again pin-pointed the spine apparatus as a potential calcium store 2'3. In addition, it has been speculated that the spine apparatus of dendritic spines could be the site of localization of receptors for inositol 1,4,5trisphosphate 6, which in other systems releases calcium from intracellular stores. It is likely that neurons possess calcium regulatory mechanisms similar to those in other cell types, but also others with uniquely neuronal features.