In this study we demonstrate that in neurons of hippocampal slices MV efficiently spreads, not by the release of extracellular virus or by syncytium formation, but retrogradely by localized cell-to-cell contact at synapses. Since rat brain is devoid of the human CD46 receptor or its homologue (32
), it is remarkable that infection of rat hippocampal tissue occurs in the apparent absence of CD46. The initial penetration of MV into hippocampal neurons could be due either (i) to the mode of inoculation (injection into slices), which could mechanically introduce virus into cells, or (ii) to alternative receptors or CD46 homologues present on immature rat neurons. The first possibility is unlikely because MV, unlike other viruses (which are taken up by the endosomal pathway), must fuse at the cell surface and not intracellularly to promote infection. The second possibility is supported by the fact that MV can infect murine cells (12
) and brains of neonatal rats (for a review, see reference 26
) and (upon passaging from one neonatal brain to another) can adapt to efficiently infect adult rat brains, suggesting that immature rat neurons possess alternative receptors that can be used by MV.
The mode of MV spread in neurons from rat hippocampal slices agrees with the conclusions of Rall and colleagues (24
), who showed that MV can spread in cultures of dissociated hippocampal neurons from CD46-transgenic mice in the presence of anti-CD46 antibodies that inhibit MV binding and infection. The fact that, in contrast to CaCo2 and Vero cells, infected hippocampal neurons both in vitro and in situ do not release infectious MV particles strongly suggests that in nervous tissue MV does not follow the classical route of infection. We thus conclude that in hippocampal neurons MV assembly is deficient and that the virus is transmitted by cell-to-cell contact. In support of this conclusion, it has been previously shown that MV can spread from persistently infected, nonneuronal cells to neighboring cells in a contact-dependent manner (16
). Other support comes from the findings of Duprex et al., who demonstrated, by employing the same MV expressing GFP as used in this study, that infection in human and murine neuroblastoma cells spreads by cell-to-cell contact along pathways (12
). Interestingly, we found MV to propagate to neurons rather than glial cells. Thus, our results and the former reports using different systems support the hypothesis that interneuronal viral transmission occurs via synapses.
The spread of MV infection requires that ribonucleoprotein be transferred from one infected neuron to another. It has been shown previously that F protein requires H protein to trigger extensive cell-to-cell fusion (8
). The lack of H protein on the surface of spine-like structures from dendrites and soma (Fig. ) explains the absence of syncytium formation in the brain tissue. However, this does not exclude the possibility that F protein could promote microfusions, either alone, at sites where it is concentrated, or with the help of minor amounts of H protein (not detectable by our confocal techniques) localized at these sites. It is, therefore, likely that infectious ribonucleoprotein translocates via such microfusions at synapses. Such a mode of transmission requires no release of infectious particles and is in line with the fact that infectious virions could not be recovered from the brains of SSPE patients (25
). It may, therefore, represent the natural route of MV transmission within the brain.
Our data demonstrate that viral envelope proteins and ribonucleoproteins are sorted to dendrites. These findings are consistent with those of an electron microscopic study of MV-infected mouse brain, which showed that MV ribonucleoprotein was aligned along postsynaptic endings (45
), and suggest retrograde MV transmission. Since, in addition, (i) it is known that 76% of the axons from CA3 pyramidal cells form synapses onto a given CA1 pyramidal cell whereas only 8% of the CA1 pyramidal cell axons form synapses onto a given CA3 pyramidal cell (10
) and (ii) we show that MV spreads much more efficiently from the CA1 to the CA3 region than in the reverse direction (Fig. ; Table ), MV appears to spread in an exclusively retrograde manner.
Based on the present data, the sorting of MV proteins in epithelial cells and that in neurons are different. In epithelial cells, the envelope glycoproteins F and H are targeted intrinsically to the basolateral domain, but M protein retargets a large percentage of F and H protein to the apical domain (34
). Here, in neurons, M protein is not required for the formation of infectious particles, and M protein does not retarget F and H proteins from dendrites to axons, as would be expected if the analogy to polarized epithelial cells (11
) were correct. Moreover, we show that the F protein, but not the H protein, is enriched at the dendritic surface, indicating a differential transport of F and H proteins in neurons. It appears that functional F rather than H protein is required for neuronal MV propagation as (i) MV isolated from the brain autopsy samples of SSPE victims forms F proteins with highly conserved, functional ectodomains (required for fusion) and highly altered cytoplasmic tails, whereas the M protein is either absent or highly compromised (1
); (ii) due to steep mRNA expression gradients, the ratio of F and H protein is considerably increased in SSPE patient brains (40
); and (iii) mutations (deletions or other alterations) were always found in the cytoplasmic tail of F protein from SSPE patient isolates, improving the fusogenic function of the F protein (7
). It thus appears that only functional F protein is needed for MV transmission in the brain. Moreover, interactions between the envelope glycoproteins themselves and with the M protein may not be required in neurons and even appear deleterious for efficient MV transmission in the brain. Thus, mutations abolishing or minimizing such interactions, especially in the M and H proteins, might be of selective advantage for MV spread in the central nervous system.
Finally, synapse-specific retrograde spreading has been previously shown for the Bartha strain of pseudorabies virus (6
) and for rabies virus (9
). Due to this feature, pseudorabies virus has been successfully employed to trace neuronal circuits in rats (21
). Our results now suggest that, at least in vitro, MV can be utilized in neuroanatomical studies. In this context, it is interesting that under our conditions MV propagates to pyramidal cells rather than interneurons or glial cells. We have previously shown that MV—compared to the more neurotropic Semliki Forest virus—does not show such a high preference for neurons during the initial infection of hippocampal slices, resulting in 62% of the GFP-positive cells being neurons (13
). Here, we now demonstrate that the propagation of MV from originally infected cells to neighboring cells in hippocampal tissue occurs only retrogradely through synapses, i.e., to neurons. In addition, as MV spreads to pyramidal cells, which contain the excitatory neurotransmitter glutamate, rather than to interneurons, which have inhibitory neurotransmitters, it might differentiate between excitatory (glutamatergic) and inhibitory (γ-aminobutyratergic) synapses.