Neuro-invasion by a ‘Trojan Horse’ strategy and vasculopathy during intrauterine flavivirus infection

Abstract

The central nervous system (CNS) is a major target of several important human and animal viral pathogens causing congenital infections. However, despite the importance of neuropathological outcomes, for humans in particular, the pathogenesis, including mode of neuro-invasion, remains unresolved for most congenital virus infections. Using a natural model of congenital infection with an RNA virus, bovine viral diarrhoea virus in pregnant cattle, we sought to delineate the timing and mode of virus neuro-invasion of and spread within the brain of foetuses following experimental respiratory tract infection of the dams at day 75 of pregnancy, a time of maximal risk of tissue pathology without foetal death. Virus antigen was first detected in the foetal brains 14 days postinfection of dams and was initially restricted to amoeboid microglial cells in the periventricular germinal layer. The appearance of these cells was preceded by or concurrent with vasculopathy in the same region. While the affected microvessels were negative for virus antigen, they expressed high levels of the type I interferon-stimulated protein ISG15 and eventually disappeared in parallel with the appearance of microcavitary lesions. Subsequently, the virus spread to neurons and other glial cells. Our findings suggest that the virus enters the CNS via infected microglial precursors, the amoeboid microglial cells, in a ‘Trojan horse’ mode of invasion and that the microcavitary lesions are associated with loss of periventricular microvasculature, perhaps as a consequence of high, unrestricted induction of interferon-regulated proteins.

The central nervous system (CNS) is a major target of several important human and animal pathogens causing congenital infections, often with devastating consequences to the foetus or newborn (Dammann & Leviton 1998; Volpe 2008). The mechanisms involved in neuropathology caused by infectious agents, and in particular viruses, can broadly be separated into three types, which are not mutually exclusive: (i) direct effect of the infectious agent on neuronal and glial cells because of infection of these cells, that is, direct cytopathogenicity, (ii) inflammatory, destructive effects and (iii) developmental derangements, that is, teratogenic effects (Lee & Bowden 2000; Volpe 2008). It may be difficult to separate these different pathways because destructive processes, whether virus cytopathogenicity or inflammation or both, affecting the developing brain often cause coincident cell or tissue loss and subsequent anomalous development. Despite the importance of neuropathological outcomes of intrauterine virus infections in humans (Dommergues et al. 1996; Volpe 2008), there are few appropriate models other than non-human primate models, which are costly and inherently fraught with ethical problems (Kinman et al. 2004; Jayaraman et al. 2007). However, congenital infection in either cattle (Bielefeldt-Ohmann 1995) or sheep (Swasdipan et al. 2001, 2002) with the non-arthropod-borne flavivirus bovine viral diarrhoea virus (BVDV) lends itself as a natural, highly reproducible model in which the mechanisms involved in the neuropathology of congenital infection can be further characterized (Dommergues et al. 1996; Dammann & Leviton 1998; Bielefeldt-Ohmann et al. 2008; Montgomery et al. 2008).

Bovine viral diarrhoea virus, along with classical swine fever virus and border disease virus of sheep and a small number of isolates from undomesticated species, makes up the genus Pestivirus in the family Flaviviridae. The virus family includes several notable human pathogens, including the viruses responsible for dengue fever, hepatitis C, yellow fever and Japanese encephalitis. Pestiviruses are positive-strand RNA viruses with a genome of approximately 12.5 kb in length. The RNA is translated into a single viral polyprotein that is processed by both viral and host proteases to either 11 or 12 virus proteins depending on the virus biotype. Two genetically distinct types (types 1 and 2) of BVDV occur, and within both genotypes, BVDV occurs as two distinct biotypes: non-cytopathic (ncpBVDV) and cytopathic (cpBVDV), with the former being the more common and the latter associated with mucosal disease in persistently infected cattle (Bielefeldt-Ohmann 1995). Establishment of viral persistence follows intrauterine infection with ncpBVDV in the first half of gestation (Bielefeldt-Ohmann 1995). The mechanisms underlying the establishment of viral persistence are still only poorly understood, but evasion of both the innate immune defence, notably the interferon response, and the adaptive immune response, the latter because of immaturity at the time of infection, is thought to be key factors. Nevertheless, we previously demonstrated a robust interferon response in persistently infected foetuses, as evidenced by expression of type I interferon-inducible molecules (Shoemaker et al. 2009). One of these products, ISG15, is a ubiquitin-like protein with antiviral activity as well as other profound effects on cell functions, notably in the CNS (Kunz et al. 2006).

Previously, we assessed the spread of ncpBVDV and the identity of the target cells in various organs and tissue types of bovine foetuses following experimental respiratory infection of the dams at day 75–80 of pregnancy (Bielefeldt-Ohmann et al. 2008; Smirnova et al. 2008; Shoemaker et al. 2009). Gross and histopathological changes were only seen in foetuses retrieved 115 days postinfection of the dam (PID), while viral antigen could be detected in most tissues from 14 days PID and onwards. The primary target organ for histopathological changes was brain and included leucomalacia, microcavitation and macrophage infiltration seen at 115 days PID (Bielefeldt-Ohmann et al. 2008; Montgomery et al. 2008). Viral antigen was detected in neurons, glial cells and infiltrating macrophages throughout all parts of the brain (Bielefeldt-Ohmann et al. 2008; Montgomery et al. 2008). However, neither the mode of neuro-invasion nor the basis for the microcavitation was elucidated in those studies. In this study, we sought to delineate the timing and mode of virus neuro-invasion and spread within the brain of foetuses following experimental respiratory infection of the dams with BVDV at day 75 of pregnancy and to unveil possible mechanisms involved in the development of the microcavitary lesions. The results suggest (i) a ‘Trojan Horse’ mode of virus entry into the CNS, that is, the transport of virus into the neuropil within infected cells, and (ii) a vasculopathy as a possible basis for the formation of periventricular cystic spaces.

Material and methods

Experimental outline

A total of 46 Hereford heifers, serologically negative for BVDV, were procured and transported to Colorado State University (CSU) facilities at the Foothills Campus, Fort Collins, where experimental work was conducted with the approval of the CSU Institutional Animal Care and Use Committee. Following acclimatization, the heifers were hormonally synchronized and subsequently artificially inseminated as per a previously established protocol (Bielefeldt-Ohmann et al. 2008; Smirnova et al. 2008). Pregnancies were confirmed by ultrasound examination 32–40 days after insemination. On day 75 of gestation, 23 of the heifers were challenged intranasally with 2 ml of virus stock at 4.4 log₁₀ TCID₅₀ of BVDV type 2 (Van Campen et al. 2000; Bielefeldt-Ohmann et al. 2008). The remaining 23 heifers received a sham challenge of tissue culture medium only and were subsequently kept in a separate paddock at a distance from the infected heifers. On days 7, 14 and 21 PID, four infected and four non-infected, randomly selected heifers were subjected to caesarean section and the foetuses retrieved. At the 115–117 days PID time point, seven non-infected and seven infected heifers were similarly treated, while at 170 days PID, four non-infected and only three infected foetuses were procured, as one virus-challenged heifer aborted spontaneously on day 150 PID because of causes unrelated to the experimental protocol, and both the heifer and the foetus were excluded from the study.

Caesarean sections, foetal blood and tissue collection

Foetuses were procured by caesarean section, using standard veterinary surgical techniques (Bielefeldt-Ohmann et al. 2008; Smirnova et al. 2008). Prior to severing the umbilical cord, foetal blood was collected by venepuncture of the cord. The foetus and all organs were weighed and samples collected for virus detection, mRNA, microarray and protein studies and histopathology (Smirnova et al. 2011). The calvarium of the skull was opened, and the foetal brains were submerged in 10% neutral-buffered formaldehyde, while remaining in situ in the skull for 24–48 h to firm up before further manipulations. Samples of all other organs were similarly fixed in 10% formaldehyde.

Virus RNA detection in foetal blood

Detection of BVDV RNA was performed as previously described (Smirnova et al. 2008, 2011). Briefly, total RNA was purified from samples of cord blood preserved with Tri reagent BD for blood derivatives (Sigma, St. Louis, MO, USA) using the RNeasy MinElute Cleanup kit (Qiagen, Valencia, CA, USA). Synthesis of cDNA was performed using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). Semi-quantitative real-time PCR (qRT-PCR) was performed with IQ SYBR green Supermix (Bio-Rad). Upon completion of qRT-PCR amplification, melting curve analysis was performed to assess the quality of amplification. Primers for BVDV were designed to target the conserved region of the 5′UTR sequence, nucleotides 190–376. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA was used as endogenous control for data normalization.

Tissue processing and histopathology

Intact brains were fixed for 48 h and transverse sections made for routine processing through to paraffin embedding. Where the samples could not be processed immediately, they were transferred into and stored in 70% ethanol for up to 1 week. The late gestation brains (the last two sampling points) were fixed in formaldehyde for an additional 48 h following trimming before processing to ensure uniform fixation.

A section from each tissue block was stained with haematoxylin and eosin (HE), and selected samples were also stained with luxol fast blue/periodic acid Schiff (LFB–PAS) to assess myelination. The sections were examined without the knowledge of infection status of the foetus and changes recorded before the study key was opened. Digital microphotographs were taken using a Nikon Eclipse 51E microscope and a Nikon DS-Fi1 camera with a DS-U2 unit and NIS elements F software (Nikon Co, Ltd., Japan). Images are reproduced without manipulations other than cropping and adjustment of light intensity.

Immunohistochemistry (IHC)

Immunolabelling for BVDV non-structural protein p125/NS2-3 (Corapi et al. 1990), neurons, glial cells, inflammation markers and apoptosis was performed as previously described in detail (Bielefeldt-Ohmann et al. 2008; Tolnay et al., 2010). Table 1 lists the antibodies and antigen retrieval methods used. Double labelling for BVDV antigen and cellular phenotypic markers was performed as described (Bielefeldt Ohmann 1987; Miura et al. 2008). Digital microphotographs were used for recording of results as described earlier. The number of oligodendrocytes was quantified by counting OLIG2-positive cells in 10 non-overlapping fields of the interphase between the sub-ventricular and intermediate zone (selected in order to avoid nuclear overlaps; Volpe 2008) at both a frontal and caudal level of the lateral ventricles, using a 40× objective. The data are presented as mean ± SD and analyzed using Student's t-test for each time point PID. Differences at p < 0.05 were considered significant.

Table 1.

Antibodies used for immunohistochemistry

Antibody	Specificity	Source	Antigen retrieval
15C5 (mouse monoclonal)	BVDV p125/NS2-3	IDEXX	Proteinase K
Goat anti-Iba1 polyclonal	Microglia	Abcam	EDTA, pH 9
Rabbit anti-GFAP polyclonal	Astrocytes	DAKO	Proteinase K
Rabbit anti-OLIG2 polyclonal	Oligodendrocytes	Abcam	EDTA, pH 9
Rabbit anti-HIF1a polyclonal	HIF-1a*	Abcam	EDTA, pH 9
Rabbit anti-p38-MAPK polycl.	Phosho-p38-MAPKa	Cell Signal. Tech.	Citrate, pH 6
Rabbit anti-NOS2 polyclonal	iNOS/NOS2	Santa Cruz Biotech.	EDTA, pH 9
Rabbit anti-COX2 polyclonal	COX2	Santa Cruz Biotech.	EDTA, pH 9
Rabbit anti-caspase 3 polycl.	Activated caspase3	Abcam	Citrate, pH 6
Mouse anti-ISG15 monoclonal	Bovine ISG15	CSU	Proteinase K
Rabbit anti-VEGF polyclonal	VEGF	Lab Vision	EDTA, pH 9
Rabbit anti-ErbB4 polyclonal	ErbB4	Abcam	EDTA, pH 9

BVDV, bovine viral diarrhoea virus; CSU, Colorado State University; OLIG2, Oligodendrocyte precursor cells; VEGF, vascular endothelial growth factor.

^*Hypoxia-inducible factor 1 alpha.

Results

Foetal BVDV infection and virus entry of brain

Bovine viral diarrhoea virus RNA was first detected at low levels in foetal blood of one foetus 14 days PID; however, by 21 days PID (day 97 of gestation), all foetuses had high levels of viral RNA in the blood, which persisted for the duration of the study (Table 2). By IHC, the BVDV non-structural protein p125/NS-2-3 was first detected in amoeboid macrophage-like cells, positive for Iba1, in the periventricular germinal zone (PVGZ) of two foetuses at day 14 PID (day 89 of gestation) (Figure 1a). At the day 21 PID time point (day 97 of gestation), all four foetuses of BVDV-challenged dams had detectable BVDV replication in the brain, restricted to amoeboid cells and few other precursor cell types in the PVGZ, as well as to macrophages in the meninges, occasional macrophage-like cells in the choroid plexus of the lateral and 4th ventricle and to cells focally in the rhombic lip of the metencephalon and in the external granular layer of the cerebellar anlage (Figure 1). The infected amoeboid, Iba1-positive cells appeared to enter the PVGZ zone from meninges via the vascular layer between the periventricular and the sub-ventricular germinal zones, corpus striatum and caudate nucleus and extend towards the apex of the lateral ventricles, exactly following the route described for seeding of the brain by microglial precursor cells (Rezaie & Male 1999; Billiards et al. 2006; Monier et al. 2007) (Figure 2 and data not shown). Notably, one of the day 14 PID foetuses still negative for viral antigen in brain also lacked Iba1-positive cells in the PVGZ. While only few cells expressing the oligodendrocyte-specific transcription factor Oligodendrocyte precursor cells (OLIG2) were positive for BVDV antigen by IHC (data not shown), there appeared to be a notable decrease in this cell type in the infected foetuses, detectable already at 14 PID (38.04 ± 16.9 per high-power field in control fetuses versus 22.6 ± 10.1 in BVDV-infected fetuses; p < 0.0001) and later (21 PID: 42.2 ± 17.5 per high-power field in control fetuses versus 27.4 ± 9.11 in BVDV-infected fetuses; p < 0.0001) compared to non-infected control foetuses (Figure 2). This cell loss appeared to be preceded by or simultaneous with a corresponding mild increase in cells positive for activated caspase 3 in the PVGZ (not shown). In no instance was viral RNA or antigen detected in blood or tissues of the foetuses derived from sham-infected dams, who also remained sero-negative throughout the course of the experiment.

Table 2.

Demonstration of bovine viral diarrhoea virus (BVDV) RNA and ISG15 mRNA in cord blood of bovine foetuses following sham or BVDV challenge of the dams on day 75 of gestation

Gestational age	Days PID	Challenge	BVDV RNA*	ISG15 mRNA†
Day 82 (11.7 w)	7	control	−(4/4)	+(4/4)
		BVDV	−(4/4)	+(4/4)
Day 89 (12.7 w)	14	control	−(4/4)	+(4/4)
		BVDV	+(1/1)‡	+++++(1/1)‡
Day 97 (14 w)	21	control	−(4/4)	+(4/4)
		BVDV	+++++(4/4)	++/+++(4/4)
Day 192 (26.5 w)	115	control	−(7/7)	+(7/7)
		BVDV	++ to ++++(7/7)	++/+++(7/7)
Day 245 (35 w)	170	control	−(4/4)	+(4/4)
		BVDV	++/+++(2/3)§	+ to +++(3/3)

PID, postinfection of the dam.

^*2^−ΔCT range: + = 0.001–0.9, ++ = 1.0–9.9, +++ = 10.0–24.9, ++++ = 25.0–49.9, +++++ = 50.0–65.0.

^†2^-ΔCT range: + = 0.001–0.09, ++ = 0.1–0.9, +++ = 1.0–4.9, ++++ = 5.0–9.9, +++++ = 10.0–15.0.

^‡Blood was successfully obtained from the umbilical cord of only one foetus from the challenge group at this time point.

^§one of three foetuses was virus-RNA negative in RT-PCR on cord blood, but had widespread infection in brain and other organs as determined by Immunohistochemistry (IHC).

Figure 1.

Virus antigen distribution in the brain of bovine foetuses 14 and 21 days postinfection of the dam (PID) at day 75 of gestation. (a) Only amoeboid glial cells in the periventricular germinal zone (PVGZ) are positive for bovine viral diarrhoea virus (BVDV) antigen at day 14 PID (day 89 of gestation; red-staining cells; examples indicated with arrows). (b) By 21 days PID (day 96 of gestation), large numbers of virus-infected amoeboid glial cells have arrived in the PVGZ. (c) The amoeboid glial cells stain positive for Iba-1 (black label; methyl green counterstain). Insert: Iba-1-positive cells are positive for BVDV antigen resulting in a blackish-red hue. (d) Clusters of amoeboid cells positive for BVDV antigen were present in the rhombic lip of the metencephalon and in the external granular layer of the cerebellar anlage (arrow) on day 21 PID. (e, f) High-power microphotographs of the cells in the two areas indicated in panel d. Immunohistochemistry. Bars equal: (a) 100 μm, (b) 200 μm, (c, e) 60 μm, (d) 500 μm, (f) 25 μm.

Figure 2.

Virus entry of the brain and effect on oligodendrocytes. (a) Entry of bovine viral diarrhoea virus (BVDV)-infected amoeboid cells via the vascular layer of the periventricular and subventricular germinal zone indicated by arrows. Insert a: BVDV antigen–positive cells in the choroid plexus (star in main photograph). Insert b: BVDV antigen–positive cells in the periventricular layer on the left of the main photograph. (b) Oligodendrocyte precursor cells (OLIG2 positive; brown nuclei) in the peri- and subventricular germinal zone of a control foetus on day 89 of gestation. (c) OLIG2-positive cells (brown nuclei) occur in decreased numbers in the peri- and subventricular germinal zone of a BVDV-infected foetus 21 days postinfection of the dam (PID) (day 96 of gestation). Haematoxylin and eosin (HE) (a) and immunohistochemistry (b, c, inserts). Bars equal: (a) 400 μm, (b, c and insets a, b) 150 μm.

In summary, the results point to a neuro-invasive mechanism involving productively infected, migratory cells entering the developing brain with subsequent spread of the virus to other cell types and possibly collateral adverse effects on oligodendrocyte precursors.

Neuropathology and virus spread within the brain

No gross pathology other than mild to moderate congestion or apparent oedema of the meninges was noted in any of the foetuses, and in most instances, these changes were considered results of the slightly protracted surgical procedure because of cord blood sampling. Histologically marked microvascular changes were noted in the PVGZ of the two BVDV antigen–positive foetuses on day 14 PID, presenting as very pronounced hypertrophy of endothelial cells, microthrombi and perivascular oedema (Figure 3). This was accompanied by rarefaction of a zone of the PVGZ between the ependymal lining and the microvascular network (Figure 3), but haemorrhage was not evident, nor was there any fibrin deposition. The changes were most prominent in the dorsal part of the PVGZ and extended variably towards the corpus striatum (Figure 3). The endothelial cells of affected capillaries were negative for BVDV antigen, but the vessels were clearly associated with accumulations of BVDV-positive amoeboid microglial cells in the surrounding parenchyma (Figure 3) and an increase in cells expressing vascular endothelial growth factor (VEGF; Figure 3d). Of the four foetuses sampled at 21 days PID (day 96 of gestation), one had similar vascular changes as well as microthrombi in the choroid plexus (Figure 4), while in the other three foetuses, the affected PVGZ vessels appeared to have been lost and rarefaction or microcavities were noted in their place, often containing BVDV antigen-positive cells and minimal floccular material in the cystic spaces (Figure 4d).

Figure 3.

Vasculopathy in the periventricular germinal zone of bovine viral diarrhoea virus (BVDV)-infected foetuses. (a) Marked proliferation and hypertrophy of tortuous vascular segments in the germinal zone (partly boxed in) of a foetus 14 days postinfection of the dam (PID). (b) Higher power microphotograph of the area boxed in panel A showing hypertrophic vascular segments with microthrombosis (insert) and lined by plump endothelial cells. (c) Localization of BVDV-infected amoeboid glial cells (red-staining cells, red arrow indicates one example) relative to the hypertrophic vascular segments (black arrow), which are negative for BVDV antigen. (d) Expression of vascular endothelial growth factor (VEGF) by amoeboid glial cells and few other cell types in the periventricular germinal zone of a BVDV-infected foetus 14 days PID. Insert: lack of expression of VEGF in a non-infected control foetus. Haematoxylin and eosin (HE) (a, b) and immunohistochemistry (c, d). Bars equal: (a) 350 μm, (b, c) 150 μm, (d) 50 μm.

Figure 4.

Vascular lesions and microcyst formation in bovine viral diarrhoea virus (BVDV)-infected foetuses. (a) Microthrombi were also present in the choroid plexus of a BVDV-infected foetus 21 days PID. (b, c) Loss of the hypertrophic vascular segments and formation of cystic spaces in the periventricular zone. The cysts contain floccular material and few cells, which (d) are positive for BVDV antigen and Iba1 (insert: double labelling for BVDV and Iba1). Haematoxylin and eosin (HE) (a–c) and immunohistochemistry. Bars equal: (a) 85 μm, (b, c) 100 μm, (d) 150 μm.

By 115 days PID (day 192 of gestation), virus antigen was detectable in neurons and glial cells, mainly microglial cells, throughout all parts of the foetal brains. Histopathological changes included microcavitation, mainly in periventricular areas (Figure 5), and variable but generally mild rarefaction of the white matter (hypomyelination) of some cortical gyri accompanied by infiltration of macrophage-like cells. Similar changes were also present in the day 170 PID foetuses (day 245 of gestation), only more severe. Despite the presence of virus-infected cells in the cerebellum, neither gross nor histological changes were seen in this part of the brain at any time point PID.

Figure 5.

Microcysts in the peri- and subventricular zones of bovine viral diarrhoea virus (BVDV)-infected foetuses 170 days postinfection of the dam (PID). (a) Microcyst (MC) in the subventricular zone apparently lacking a continuous cellular lining, in contrast to the adjacent blood vessel (BV) and the lateral ventricle (LV). (b) Immunohistochemistry (IHC) for GFAP demonstrates that the microcysts are surrounded by a layer of condensed astrocytic dendrites (brown mesh work). The GFAP-negative cells internal to this dendrite web appear to be a mixture of oligodendrocytes (c; arrow in insert) and (d) Iba1-positive macrophages or microglial cells, with the latter also floating in the cyst lumen (arrow) and expressing BVDV antigen (red cells in the consecutive section shown in insert). Haematoxylin and eosin (HE) (a) and immunohistochemistry (b–d). Bars equal: (a) 350 μm, (b) 60 μm, (c) 150 μm, (d) 150 μm.

The microcavities seen in foetuses 115 and 170 days PID, and located in the periventricular neuropil corresponding to the day 14 and day 21 PID lesions, did not appear to have a continuous cellular lining. Rather, they were surrounded by a layer of highly condensed astrocytic dendrites and occasional interspersed oligodendrocytes (Figure 5), all of which were negative for BVDV antigen, while occasional Iba1-positive cells within the lumen of these cavities expressed BVDV antigen (Figure 5d and insert).

Taken together, the data suggest that microcavitation in the PVGZ of infected foetuses may be associated with and possibly preceded by microvascular proliferation and thrombosis.

Interferon and inflammatory responses in the brain

Corresponding to the mRNA findings for the cord blood (Table 2), intense ISG15 expression was detected in the brain of foetuses from BVDV-challenged dams as early as 14 days PID. Notably, ISG15 protein expression was also detected in the two foetuses negative for BVDV antigen at that time point. Initially, vascular endothelial cells, including those of the PVGZ microvasculature, appeared to be the main interferon responders (Figure 6), but more diffuse and very intense ISG15 expression was seen from early in infection and throughout the remaining study period in most of the infected foetuses. All control foetuses were negative for ISG15 expression (Figure 6).

Figure 6.

Expression of innate immune response products in the brain of bovine viral diarrhoea virus (BVDV)-infected foetuses. (a) ISG15 expression is largely restricted to endothelial cells of the microvasculature in the foetal brain 14 days postinfection of the dam (PID). Insert: lack of ISG15 expression in a control foetus. (b) By 115 days PID, ISG15 expression is seen in most cells throughout the brain of a BVDV-infected foetus, while similar-aged control foetuses are negative (insert). (c, d) NOS2 expression in the brain, including the choroid plexus, of a BVDV-infected foetus 14 days PID. Immunohistochemistry. Bars equal: (a) 150 μm, (b) 350 μm, (c, d) 300 μm.

In contrast to the intense interferon response, there appeared to be minimal or no detectable up-regulation of phospho-p38-MAPK in the infected PVGZ or in cells in and around the hypertrophic microvessels or the microcavitations (data not shown). However, NOS2 (iNOS) was up-regulated in the brain of foetuses from day 14 PID, most notably in the two foetuses with detectable neuro-infection (Figure 6). By day 21 PID, the expression of NOS2 had comparatively diminished and was undetectable at the later time points. Corroborating earlier findings (Bielefeldt-Ohmann et al. 2008), there was no significant up-regulation of HIF-1α in brains of the infected foetuses at any time point PID, except for the choroid plexus, where macrophage-like cells in the interstitium and intravascularly expressed cytoplasmic and variably nuclear HIF-1α (data not shown). However, in no case did this correspond to the presence of BVDV antigen in this particular compartment.

In summary, BVDV infection induced a sustained, strong interferon response in the foetal brains, while other inflammatory mediators were more transiently or not detectably expressed at the time points examined.

Discussion

Three key findings emerged from this study: (i) BVDV appears to invade the foetal brain via infiltrating leucocytes, specifically the microglial precursor cells, called amoeboid glial cells, (ii) vasculopathy around the time of viral invasion followed by microvascular loss, presumably by necrosis, either leads to or is concurrent with formation of microcavities or cysts that persist and may enlarge over the subsequent weeks of gestation and (iii) a very marked interferon type I response with expression of ISG15 occurs at the time and even prior to detectable neuro-invasion by virus and which persists throughout gestation. Notably, ISG15 protein expression in the two foetuses negative for BVDV antigen at 14 days PID suggests that they might have been infected but that the virus had not yet reached the brain. Alternatively, the early interferon was of maternal origin, as initially vascular endothelial cells, including those of the PVGZ microvasculature, appeared to be the main interferon responders.

Additionally, there appears to be a substantial loss of oligodendrocyte precursors early in the infection, apparently from causes other than direct virus infection of these cells, and this may lead to the subsequent rarefaction of white matter in the core of many gyri (Bielefeldt-Ohmann et al. 2008 and present study). This may be comparable to periventricular leukomalacia in humans, where a presumed loss of premyelinating oligodendrocytes is at the core of the myelination defect (Volpe 2008). In contrast, despite widespread infection of neurons in later stages of intrauterine life, there is no notable cell loss, nor do any of these changes incite an inflammatory response other than mild macrophage infiltration in meninges, choroid plexus and in areas of white matter rarefaction.

Neuro-invasion by viruses may occur by one of at least four, not mutually exclusive, routes: (i) infection of or trans-cytosis through the microvascular endothelial cells and other components of the blood–brain barrier (BBB), (ii) transport of the virus across the BBB by infected leucocytes, the ‘Trojan Horse’ strategy, (iii) viremic dissemination to the olfactory bulb and (iv) axonal retrograde transport from infected peripheral neurons (Salinas et al. 2010). In the case of BVDV neuro-invasion following maternal infection at the very end of the first trimester (11- to 12-week gestation), the ‘Trojan Horse’ strategy appears to be used, with the virus having infected the microglial precursor cells in the periphery. This interpretation is supported by the finding that no Iba1-positive amoeboid glial cells were detected in the two foetuses derived from virus-challenged dams and which were BVDV antigen negative at 14 days PID, while BVDV antigen and Iba1-positive cells coincided in the other two foetuses from this time point, as well as in all four foetuses at 21 days PID. This finding does not preclude other modes of neuro-invasion following maternal infection at other time points of gestation. Whether BVDV productively infects endothelial cells remains a controversial issue, although IHC studies suggest that cells associated with the BBB, either astrocytes, microglial cells, pericytes or the endothelial cells themselves, may be infected at later stages of foetal persistent infection (Bielefeldt-Ohmann et al. 2008; Montgomery et al. 2008; this study).

The present study did not allow for a delineation of how the virus spreads within the foetal CNS. At 21 days PID, the virus appeared to be restricted to the amoeboid glial cells and possibly few other cells in the germinal layers of the neopallium and the cerebellar anlage, while at the next sampling time point, 115 days PID, virus antigen was already present in many neurons, microglial cells and occasional astrocytes and rare oligodendrocytes throughout all compartments of the brain. It may be hypothesized that the precursor cells of these cell types are infected in the periventricular germinal layers or the germinal layer of the cerebellum prior to migration to their final destinations. However, only further studies with more sampling time points in that intervening period between 21 and 115 days PID will be able to cast light on this question. Such analysis should also address the possible effect of virus infection on the migration pattern of the precursor cells (Lazarini et al. 2003). In this context, it was notable that only the myeloid/macrophage subpopulations giving rise to the amoeboid glial cells and the macrophages in the meninges were infected, while the majority of macrophages residing in the choroid plexus appeared to be virus-antigen negative. This finding raises the question of whether the infection takes place in foetal haematopoietic tissue, either bone marrow, liver or spleen, or it happens en route to the CNS. The answer to this question may lie in the newly identified, ontogenically distinct derivation of microglia from the myeloid lineage (Ginhoux et al. 2010).

Another question that remains open is whether the infection of neurons is productive or abortive. The answer is most likely abortive, because virus antigen appears to be restricted to the Golgi region in the vast majority of infected neurons at 115 days PID and later time points. One mechanism involved in this replication restriction could be the interferon type I response, as evidenced by the widespread and intense ISG15 expression, while at the same time explaining why we did not see substantial loss of neurons by either apoptosis or necrosis (Bielefeldt Ohmann & Babiuk 1988; Jeon et al. 2009) or infiltration of leucocytes as manifestation of an innate inflammatory response (Kim et al. 2008; Chen et al. 2010), but leaves unanswered what the effect of the infection might be on the synaptogenesis and functionality of the neurons (Campbell et al. 1999; Hans et al. 2004; Gonzalez-Dunia et al. 2005; Kunz et al. 2006). In contrast, many oligodendrocytes were lost, most likely at the precursor stage, even though few appeared to be infected. It is known from other infection models, most notably HIV infections in humans, that oligodendrocytes and their precursors are exquisitely sensitive to many factors elaborated by activated microglial cells and astrocytes (Streit et al. 1999; Haynes et al. 2003; Block et al. 2007; Segovia et al. 2008). As the infected microglial cells were clearly activated, as evidenced by their expression of NOS2, VEGF (Figures 3d & 6c,d) and COX2 (not shown), and as astrocytes likewise demonstrated reactivity in the context of the microcavitation, it appears likely that factors secreted by these cells caused adverse effects on the oligodendrocyte lineage at the same time as activation of microglia and astrocytes may have reduced their production of trophic factors for the oligodendrocyte precursors (Nicholas et al. 2001; Haynes et al. 2003; Rhodes et al. 2006; Block et al. 2007; Segovia et al. 2008). Furthermore, the marked interferon response in the neuropil may have added to the insults to the oligodendrocytes (Ritchie et al. 2002; Kunz et al. 2006; Trebst et al. 2007), with the collective effect at later stages of development being hypo- or demyelination (Segovia et al. 2008; Bielefeldt-Ohmann et al. 2008, and this study).

Interestingly, despite early and relatively widespread virus infection in the germinal layer of the developing cerebellum (Figure 1), neither gross nor histological lesions were noted in this part of the brain in the present study or in our earlier studies (Bielefeldt-Ohmann et al. 2008; Montgomery et al. 2008). Cerebellar hypoplasia is one of the hallmarks of natural congenital infection with BVDV, albeit with variable penetrance (Bielefeldt-Ohmann 1984, 1995). It remains an unresolved question whether BVDV-induced cerebellar hypoplasia is a consequence of (i) infection with certain BVDV strains, (ii) a very time-dependent event not yet pin-pointed, (iii) a breed-dependent factor, (iv) environmental co-factors or (v) some combination of these, but the fact is that this particular aspect of the neuropathology spectrum is very difficult to reproduce experimentally.

While the results of our earlier study focusing on day 115 PID (Bielefeldt-Ohmann et al. 2008) did not reveal any significant up-regulation of HIF-1α expression and nuclear translocation in the persistently infected foetuses and failed to support the hypothesis that the microcavitations in the PVGZ and white matter of the BVDV-infected foetuses were caused by hypoxia, similar to those of white matter lesions in cerebral palsy of humans (Volpe 2008; Trollmann & Gassmann 2009), the possibility remained that hypoxia might be a proximal cause occurring much sooner after maternal infection. However, our current study also failed to support a role for hypoxia in the development of microcavities and cystic spaces in the periventricular germinal layers and white matter. Rather, an initial burst of neovascularization, followed by microthrombosis, perivascular oedema and then complete loss of these vascular structures without any signs of haemorrhage or inflammatory response, appeared to be either the genesis of the microcavitations or a concurrent event. Angiogenesis and vasculogenesis are generally thought to be regulated by HIF-1 and the downstream production of VEGF (Coulon et al. 2010; Rey & Semenza 2010); however, HIF-1-independent pathways may occur under some circumstances such as the TRIF/TBK/IRF3 pathway (Korherr et al. 2006). Other possible mechanisms include interferon-induced CXCL12/SDF-1 production by microglial cells (Bajetto et al. 2001; Okamato et al. 2005) or direct VEGF induction by the virus. Viruses such as respiratory syncytial virus and another flavivirus, dengue virus, can directly induce VEGF expression (De Silva et al. 2006; Psarras et al. 2006; Azizan et al. 2009). We noted expression of VEGF in amoeboid microglial cells of BVDV-infected foetuses on days 14 and 21 PID, supporting the idea that either interferon, CXCL12/SDF-1 or the virus directly induces vasculogenic factors and proliferation of microvasculature in a zone rich in stem cells. However, without a constant blood flow in these vascular segments, the vessels will disappear by unknown mechanisms (Rubanyi et al. 1990). While direct evidence of this scenario, as a pathway to formation of microcavities in place of lost microvascular segments, is lacking so is any alternative explanation at present, and it thus provide a working hypothesis for future studies, where the natural model of BVDV infection in cattle (Bielefeldt-Ohmann 1995; Hansen et al. 2010) or sheep (Swasdipan et al. 2001, 2002) could be exploited.

Author contributions

All authors contributed substantially to (i) study design, data acquisition and/or data analysis and (ii) drafting and/or revision of manuscript, and all authors have approved the final, submitted version. Other contributors are listed in the acknowledgement.