This study provides evidence of VMV compartmentalization within the CNS and other tissues and confirms its existence in the mammary gland as previously described between blood and colostrum  amongst seropositive naturally infected small ruminants. The sheep had Visna-like severe lesions in the CNS, with non-suppurative meningo-encephalitis . According to the specific feature of this outbreak [6, 7], most animals presented lesions in the spinal cord and 50 percent in the brain. Most of them also presented lung lesions. VMV sequence distribution differed between animals, as revealed by phylogenetic analysis. The different compartmentalization patterns may be explained by differences in host genetic susceptibility, anatomical features, the genome of the infecting viruses, as well as high rates of mutation and replication, generating a quasispecies , and variants which may escape from surveillance . Compartmentalization may involve genetic drift and founder effects , and/or being a consequence of different selective pressures .
Respiratory secretions containing infected alveolar macrophages can be responsible for SRLV transmission, especially in intensive rearing systems such as those applied in the Assaf breed of the animals under study . Consequently, blood, lymph, and then target tissues, such as L, MG, CPx and SC of CNS, become infected . Following purifying selection (as observed in this study) the virus evolves and compartmentalizes, producing a quasispecies. Thereafter, the virus may go back through lymphoid and blood cells to other tissues. In line with this scheme and using a novel methodology previously applied to investigate geographical origin of sequences , we found as expected, that intra-host ancestor sequences were not located in the CNS or other tissues, but in BAL or PBMC (when available). PBMC sequences were present in different clades, displaying the broadest variation per tissue, which is compatible with migration of virus from blood to tissues and vice versa.
However, particular ancestor/primary sequences could not be always identified. This may be due to the unavailability of PCR amplification in some samples. Thus, clones analyzed in the ancestor tissues (BAL and PBMC) may have not included the complete spectrum of ancestor viral sequences. Besides, all samples were obtained at a single time point (necropsy) even though the time points of infection by a particular virus may have not been simultaneous at different body sites. Also, there may have been re-infection events from inside or outside the body (like sheep 333, apparently with two main infections). Finally, body sites other than BAL (such as nasal and conjunctival mucosa) may have been the initial source of infection.
The virus under study belonged to the genotype A, also involved in other Visna infections . Like in CNS studies of HIV  and in contrast with findings on FIV  infections, we were unable to find any CNS-specific signature pattern. In these lentiviruses, V3-V5 regions of SU encoding neutralizing epitopes may mutate, leading to viral resistance to existing antibodies also affecting cell tropism [18, 30], including neurotropism and neurovirulence [14, 43]. Amongst the SRLV sequences under study (TM and V4-V5), V4 - claimed to have a function analogous to that of HIV-V3 region [32, 33] - was the most variable. Although variable and constant regions within the SRLV SU protein may trigger the production of neutralizing antibodies , the main sites/epitopes involved in this process have not been identified. Additional factors or genetic regions, such as those encoding viral proteins (Vif) that interact with host molecules (APOBEC; ), might be involved in neuroinvasion.
Of the 16 MG clones of animal 698, 12 showed a non-synonymous transition C to T at the same position, inserting a stop codon along the env region encoding the SU protein (Figure 1). The finding of stop codons along with codifying sequences suggests the emergence of quasispecies, the presence of a high rate of viral evolution and thus the existence of defective interfering particles/heteroclites. However, numerically realistic model studies suggest that these particles are unlikely to survive or influence viral dynamics .
Work on HIV and simian immunodeficiency virus (SIV) has shown that viral evolution in the brain is peculiar in that this immunologically privileged site allows a low level of viral replication  before disease onset , but it may be considered a "sanctuary" where the virus can evolve freely, avoiding the effects of adaptive immunity and/or therapy . In line with this, the least conserved Env amino acid sequence was found in SC and the most conserved in L (data not shown). The diagnostic LTR-PCR used in this study only yielded consistent amplifications when using CNS as DNA source, since all animals were seropositive and yielded PCR-specific amplicons in CNS (Table 1). This strongly suggests that CNS is a preferred site of infection in the Visna affected animals involved. SRLV replication occurred in brain, since viral antigens p25 and gp130 were detected in CNS by immunohistochemistry and two fully replicative viral isolates, able to replicate in SCP cells, were obtained from SC and cerebrospinal fluid from two of the diseased sheep used (Nos. 697 and 698, respectively) . This is compatible with studies on HIV infections, where viral replication in the CNS occurs in diseased individuals, leading then to divergent trends in different compartments .
This study shows that different SRLVs can coexist in the CNS from a Visna diseased sheep. This was clearly observed in animal 333, infected by two viruses (333a and 333b sequences), one closer to the remaining outbreak sequences than the other. Under the hypothesis that the strain's genetic makeup (in the SU genetic regions analyzed) and induced pathology are associated, the set of sequences most similar to other sequences in the outbreak would be causing the nervous clinical signs. The CNS 333a sub-set sequences were essentially from CPx (n = 7) and to a lower extent from SC (n = 2), whereas 333b ones were similarly represented in CPx (n = 9) and SC (n = 11). This would be consistent with a recent infection in the case of 333a, in which the virus from infected CPx cells (blood-CSF barrier) reached only a few SC sites. In the case of 333b sequences, the infection appeared to have been longer established, as it extended to various SC sites after CPx initial infection. Whether different viruses infected different cells of the same tissue (such as choroid plexus) or different viruses coinfected the same cell type remains unknown. In any case, this study demonstrates in VMV infections that, like in those by SIV, the physical continuity of the brain tissue does not necessarily result in phylogenetic proximity . Disruption of the blood-brain or the blood-CSF (CPx) barriers may have taken place, as proposed in FIV and HIV infections [43, 48]. In HIV infections, CPx may allow a bidirectional productive lentiviral infection of CNS and peripheral organs, having often a mixture of viral sequences [43, 46]. Similarly in this study, VMV-infected cells may have crossed the CPx barrier to colonize CSF and SC, and gone back from there to blood through blood-brain or blood-CSF (CPx) barriers.