2.1. Flaviviruses
The first publications on DENV-induced membrane modifications in mammalian cells (marked cytoplasmic vacuolization and accumulation of virions in crystalloid arrays surrounded by membranes) date from the seventies [
9,
10]. Several years later, studies on DENV-infected mosquito cells were published in which the visualization of virions was reported [
11,
12]. In 1997, Grief and coworkers described DENV-induced membrane alterations of various morphologies, including convoluted membranes (CMs) and vesicle packets (VPs) [
13]. Subsequent immuno-EM studies indicated that VPs and CMs may represent the site of DENV replication and RNA translation/polyprotein processing, respectively [
14]. The most complete characterization of DENV-induced intracellular membrane rearrangements elucidated their 3D architecture as well as their spatial connection with viral assembly sites [
15]. TEM of resin-embedded infected cells revealed a complex collection of convoluted and vesicular structures, including CMs that were usually surrounded by multiple vesicles, often appearing as longitudinal vesicle arrays. By using electron tomography (ET), the latter were found to correspond to ER tubules containing 80–90 nm single-membrane vesicles (Ve) that result from the invagination of the ER membrane into the ER lumen. By conventional EM, these vesicles appeared as double membrane vesicles, likely corresponding to the VPs described earlier [
13]. Immuno-EM confirmed that the vesicles visible in resin-embedded cells were induced by DENV infection and contained all NS proteins. However, only NS3 was detectable within the CMs, which could be due to lower affinity of the antibodies or poor accessibility of the other NS proteins in the CMs. Double-stranded RNA (dsRNA) detected by immunostaining appeared as discrete electron-dense structures inside or on the cytosolic surface of a subset of vesicles, suggesting that dsRNA might be present only in some of the vesicles at a given time point. Furthermore, the vesicles contain rather uniform pores of ~10 nm diameter towards the cytosol (
Figure 2A). Thus, both the topology of the vesicles and the immunolabeling results support the idea that the vesicles might be the site of RNA replication. Moreover, these results showed that replication factories are a continuous membrane network that provides a platform for the transport of viral proteins and genomes between sites of RNA replication, ribosome-containing compartments (RNA translation) and virus assembly sites. In fact, virus budding sites were found in close proximity to the pores of the replication vesicles. This topological link may ensure efficient production and delivery of viral RNA for the assembly of infectious virus progeny. Consistent with these findings, a very recent publication using ET showed that these virally modified structures were also observed in DENV-infected mosquito cells, with one exception: CMs were absent from DENV-infected C6/36 mosquito cells [
16]. In addition, after multiple rounds of virus replication, tubular structures were also observed in the vicinity of VPs. These structures might represent a hallmark of chronically infected insect cells, since these structures are also induced by TBEV in tick cells (see below).
The first reports on WNV-infected cells described the visualization of virions [
17]. An extensive characterization of Kunjin Virus—the Australian variant of WNV (WNV
KUN)—infected cells has been carried out more recently [
18,
19,
20]. Three well-defined structures were found, corresponding to large CMs, paracrystalline arrays (PCs) and VPs that appeared as membrane sacs containing small vesicles (Ve) [
18,
21]. Based on immunolocalization studies, a distinct redistribution of the
trans-Golgi network (TGN) and colocalization of TGN markers with dsRNA has been observed, suggesting that the replication factories of WNV
KUN were derived from the TGN [
22]. Three-dimensional reconstructions of the WNV
KUN replication sites revealed an intimate association of the rough ER (rER) with the bounding membrane of the VPs [
20] (
Figure 2B), resembling the vesicles observed in DENV-infected cells. These results argue for an additional role of the rER in the formation of the WNV
KUN replication factories. Similar to DENV, individual necks were observed in the vesicles as well as the majority of the viral RNA, as detected by immunolabeling with a dsRNA-specific antibody, resided within these vesicles [
18,
20,
22,
23,
24]. In most cases, viral RNA spanned the breadth of the vesicles and was juxtaposed to the necks open to the cytoplasm [
20].
Figure 2.
Representative images of membrane rearrangements induced by different members of the family
Flaviviridae. (
A) Dengue Virus (DENV); (
B) West Nile Virus (WNV); (
C) Tick-borne Encephalitis Virus (TBEV); (
D) Hepatitis C Virus (HCV). Slices through tomograms of infected cells (
on the left) and 3D top and lateral (90° rotation) views of the same tomograms (
on the right) are depicted, showing the characteristic virus-induced structures. The replication vesicles (Ve) of DENV, WNV and TBEV (genus
Flavivirus) correspond to invaginations of ER membranes that remain connected to the cytosol via 10 nm-pores (highlighted with white arrows in the 3D lateral views), forming vesicle packets (VPs). The replication factory of HCV (genus
Hepacivirus) is primarily composed of double membrane vesicles (DMVs) that seem to be formed asER protusions connected to ER membranes via neck-like structures (highlighted with white arrows in the 3D lateral view). The ER is shown in yellow (DENV, TBEV and HCV) or in red (WNV) and the replication organelles in brown (DENV, TBEV and HCV) or in white (WNV). The outer and inner membranes of DMVs are depicted in different shades of brown (outer membrane in dark brown and inner membrane in light brown).
Figure 2B is reproduced with permission from [
20].
Figure 2.
Representative images of membrane rearrangements induced by different members of the family
Flaviviridae. (
A) Dengue Virus (DENV); (
B) West Nile Virus (WNV); (
C) Tick-borne Encephalitis Virus (TBEV); (
D) Hepatitis C Virus (HCV). Slices through tomograms of infected cells (
on the left) and 3D top and lateral (90° rotation) views of the same tomograms (
on the right) are depicted, showing the characteristic virus-induced structures. The replication vesicles (Ve) of DENV, WNV and TBEV (genus
Flavivirus) correspond to invaginations of ER membranes that remain connected to the cytosol via 10 nm-pores (highlighted with white arrows in the 3D lateral views), forming vesicle packets (VPs). The replication factory of HCV (genus
Hepacivirus) is primarily composed of double membrane vesicles (DMVs) that seem to be formed asER protusions connected to ER membranes via neck-like structures (highlighted with white arrows in the 3D lateral view). The ER is shown in yellow (DENV, TBEV and HCV) or in red (WNV) and the replication organelles in brown (DENV, TBEV and HCV) or in white (WNV). The outer and inner membranes of DMVs are depicted in different shades of brown (outer membrane in dark brown and inner membrane in light brown).
Figure 2B is reproduced with permission from [
20].

In cells infected with TBEV, one of the most important tick-transmitted viruses in Europe and Asia, virus particles and membrane-connected vesicles were also observed inside the ER [
25], similar to what was described for DENV and WNV
KUN. The viral dsRNA was only detected inside the vesicular structures within rER, suggesting that TBEV rearranges internal cell membranes to generate a compartment that protects viral RNA from detection by cytoplasmic pathogen recognition receptors (PRRs) [
26,
27,
28]. This localization of dsRNA might suffice to delay the onset of the IFN response [
25]. For TBEV [
25] and WNV
KUN [
19] it was shown that treatment with brefeldin A (BFA), a drug which disrupts the Golgi apparatus, did not interfere with viral replication. However, this treatment rendered WNV
KUN sensitive to the antiviral action of the IFN-induced protein MxA, indicating that BFA might have disrupted the membranous WNV
KUN replication compartments, thus leading to exposure of dsRNA and its detection by PRRs. In contrast, treatment of TBEV-infected cells with BFA neither affected viral replication, nor the level of IFN production. These findings indicate that TBEV dsRNA might be stored inside BFA-resistant membrane vesicles that robustly protect the viral RNA from recognition by cellular sensors.
Vector-borne flaviviruses like DENV and TBEV must replicate in both mammalian and arthropod cells. A few comparative studies have been published describing virus-induced structures such as cytoplasmic membrane proliferations and vesicle formation, also in insect cells [
11,
16,
21,
29,
30,]. A detailed comparative ultrastructural analysis of TBEV-induced modifications revealed that the extent of membrane expansion and the abundance of vesicles were lower in insect cells [
31]. Single-membrane vesicles, ranging in diameter from 60–100 nm were frequently found within proliferated ER areas, often occurring in large groups contained within ER cisternae. Pore-like openings connected these vesicles to the cytoplasm and to other vesicles. Apart from these vesicles, in tick-infected cells elongated vesicles or tubules were found that were much more prevalent in persistently than in acutely infected cells. These tubules were only occasionally noted in infected mammalian cells, similar to what was found with DENV-infected cells [
15]. The tubular structures had a cross-sectional diameter of 60–100 nm, similar to the one of vesicles, reached up to 800 nm in length, were closed at the ends and often arranged in fascicle-like bundles, shrouded with the ER membrane. However, no pores between the tubules or towards the cytoplasm were observed [
31]. The function of these tubules is unclear and it is not known whether they represent
bona fide features of replication factories, aberrant structures as a result of incorrect membrane remodeling, or the result of a cellular process to restrict infection [
31]. In any case, the tubules might be a feature of persistent infection, eventually linked to the high number of defective virus particles, because the lack of pores could prevent proper replication or packaging of the viral genome [
32,
33,
34]. Further studies are required to shed light on the biogenesis and biological significance of these membranous tubular structures.
A recent study identified 80 nm-diameter vesicles within the ER lumen of TBEV-infected BHK-21 cells and in cells transfected with a TBEV replicon [
35]. ET revealed that these vesicles are invaginations of the ER within a highly organized network of interconnected membranes with half of vesicles containing pore-like connections to the cytoplasm (
Figure 2C). However, no pore-like openings were observed between adjacent/neighboring vesicles, in contrast to what has been described for cells infected with Langat Virus (LGTV), a naturally attenuated tick-borne flavivirus [
31] or in WNV
KUN-infected cells [
20]. Interestingly, in TBEV replicon cells, the number of pore-containing vesicles was slightly larger (~75%) and they were found in much more fragmented ER tubules as compared to TBEV-infected cells. However, despite more extensive ER rearrangements in replicon cells, they contained fewer vesicles, consistent with the lower level of viral replication [
36,
37,
38].
Conventional EM analysis of neurons infected with Murray Valley Encephalitis Virus (MVEV) revealed several ultrastructural features, including proliferation of ER and Golgi complex membranes as well as the appearance of membrane-bound spherical vesicles (75–145 nm diameter) [
39], similar to those observed for the related flaviviruses Japanese Encephalitis Virus (JEV) [
40,
41] and St. Louis Encephalitis Virus (SLEV) [
42]. In the latter case, cylindrical membranous structures (or tubules) were also observed [
43]. The presence of vesicles was also detected in monkey liver cells infected with Yellow Fever Virus (YFV) [
44]. These findings indicate that all members of the genus
Flavivirus utilize the ER as a source of membranes for the formation of their replication factories, whereas assembly of new virions seems to occur at ER sacs in close proximity to the replication sites [
15,
35], thus creating an optimized membranous environment to support efficient viral replication and assembly. Maturation of the newly synthesized virions takes place in the Golgi apparatus, where flavirirus virions are often observed [
15].
2.2. Hepaciviruses
In stark contrast to flaviviruses, HCV, the prototype of the genus
Hepacivirus, provokes an alternative rearrangement of intracellular membranes, originally designated “membranous web” (MW). This term referred to compact vesicle accumulations embedded into a membranous matrix [
45] as detected in cells inducibly expressing the HCV polyprotein. By using different EM methods, we and others have recently found that the MW is primarily composed of double membrane vesicles (DMVs) [
46,
47,
48]. The fact that the kinetics of their appearance correlates with HCV replication suggests that these structures play an important role for viral RNA amplification [
48]. Indeed, immunolabeling of purified DMVs revealed an enrichment for viral proteins as well as dsRNA [
46,
49]. Importantly, DMVs contain enzymatically active viral replicase [
49] and they originate from ER membranes, similar to what has been found for other members of the family
Flaviviridae. ET analysis showed that most of the DMVs remain connected to the ER via their outer membrane [
48] (
Figure 2D). Although DMVs are primarily closed structures, ~10% of them have an opening towards the cytosol. Late in infection, multi-membrane vesicles (MMVs) with an average diameter of 390 nm are generated, likely originating from DMVs by secondary enwrapping events [
48].
By using Huh7.5 cells infected with the highly replicative HCV strain JFH-1, Ferraris and coworkers observed three different types of membrane alterations: vesicles in clusters (ViCs), contiguous vesicles (CVs) and DMVs [
47]. The ViCs were small single-membrane vesicles of variable size (100–200 nm), grouped together in well-delimited areas. Most of them had an internal invagination. The CVs were small single-membrane vesicles, present in large numbers and widely distributed throughout the cytoplasm, with a more homogeneous size (around 100 nm). They were tightly associated to each other and tended to form a collar around lipid droplets (LDs). DMVs were heterogeneous in size (150–1000 nm) and had a thick, electron-dense membrane consisting of two closely apposed membranes. The increase of CVs’ number correlated with an increase of intracellular HCV RNA levels, arguing for a possible role of CVs in the early stages of viral replication. The presence of NS5A in CVs, as demonstrated by immunogold staining, is consistent with this hypothesis. Alternatively, CVs might constitute the membranous platform for viral assembly. In fact, the core protein is present in these structures (16%) as well as on the LD surface (81%). However, so far visualization of virus particles in infected cells has not been possible, making this hypothesis difficult to prove. While most of the dsRNA signal was located within DMVs or at DMV membranes, ViCs were free of viral components and RNA and these structures as well as CVs were very rarely observed in cells with a subgenomic JFH-1 replicon [
46] or absent in cells infected with a JFH-1 variant designated Jc1 [
48]. The first 3D reconstruction of a complete HCV-infected cell revealed that all these three membrane structures were tightly connected and closely associated with LD clusters [
47].
Taken together, these findings indicate a fundamental role of DMVs in HCV replication. An in-depth comparison of the study by Ferraris and coworkers [
47] and our publication [
48] suggests that CVs might be also DMVs for several reasons: first, CVs have electron dense tightly apposed membranes; second, by using correlative light and electron microscopy, we also detected DMV accumulations around LDs, reminiscent of the CVs described by Ferraris and coworkers [
47]; third, taking into consideration the density of content and morphology, some of the structures described as DMVs by Ferraris and colleagues might correspond to MMVs according to our nomenclature. This might account for the differences in size between the DMVs reported in both studies (up to 1000 nm
versus 150 nm, respectively). Alternatively, the difference might be due to the use of distinct virus strains (JFH-1 and Jc1) that differ in their capacity to produce infectious virus particles by ~3 orders of magnitude [
50], which might also explain the presence of ViCs only in JFH-1 infected cells.
2.3. Pestiviruses
Much less about membranous replication factories is known for pestiviruses. TEM-based studies from the times in which the genus
Pestivirus was still belonging to the family
Togaviridae reported that pestivirus-infected cells exhibited ultrastructural modifications of rER and contained small numbers of virus-like particles (VLPs) [
51,
52]. Gray and Nettleton (1987) reported that Border Disease Virus (BDV)-infected cells contained several profiles of ER and many dense lamellar bodies, which when transversely sectioned appeared as multiple rows of tubules, 33 nm in diameter [
52]. These lamellae were often found in association with rER and in one occasion VLPs appeared to be budding within them. Bovine Viral Diarrhea Virus (BVDV)-infected cells contained rER modified into tubules, in which electron-dense VLPs were present. More recent studies on BVDV-infected cells revealed cytoplasmic vacuolization and VLPs in dilated ER cisternae [
53,
54]. In addition, membrane structures consisting of vesicles of various sizes enclosed in much larger vesicles have been reported [
55]. These structures that morphologically resemble multivesicular bodies (MVBs) are distinct from the HCV-induced membranous web and more reminiscent of the flavivirus-induced VPs.
Studies on the morphogenesis of pestiviral particles were hampered by a low rate of virion production. In a recent study, Schmeiser and colleagues have overcome this problem by using high multiplicity of infection in MDBK cells with a distinct virus strain, the Giraffe-1 strain [
56]. Obtained results define the ER as the site of pestivirus particle assembly, where budding of virions was observed. Virus particles were also found inside the lumen of the Golgi and in vesicles associated with the Golgi compartment, suggesting that virus egress occurs via the conventional secretory pathway. Interestingly, replication kinetics of pestiviral RNA did not correlate with distinct membrane rearrangements and only slight dilatation of the ER lumen was noticed. The absence of significant membrane rearrangements argues for a major difference between pestiviruses and other members of the
Flaviviridae family. Interestingly, the authors detected the capsid protein and dsRNA, the marker for viral replication intermediates, mainly in MVBs, indicating that pestiviruses are either using this compartment for replication or that viral RNA and proteins are transferred to this compartment for degradation. Similar assumptions have been made for HIV [
57] and Marburg virus, a member of the
Filoviridae family [
58,
59]. Alternatively, pestiviral RNA and protein in MVBs might be intermediates of the entry process, prior to fusion of the envelope with the endosomal membrane. Indeed, particles inside MVBs matching the morphological criteria of pestivirus virions were detected [
56]. However, MVBs of non-infected cells also contain vesicles for lysosomal degradation termed intraluminal vesicles (ILVs) that display a very similar morphology to pestiviral virions. Thus, unambiguous discrimination between ILVs and pestivirus particles will require detailed immunolabeling approaches.