*Review* **Ultrastructural Features of Membranous Replication Organelles Induced by Positive-Stranded RNA Viruses**

**Van Nguyen-Dinh and Eva Herker \***

Institute of Virology, Philipps-University Marburg, 35043 Marburg, Germany

**\*** Correspondence: eva.herker@uni-marburg.de

**Abstract:** All intracellular pathogens critically depend on host cell organelles and metabolites for successful infection and replication. One hallmark of positive-strand RNA viruses is to induce alterations of the (endo)membrane system in order to shield their double-stranded RNA replication intermediates from detection by the host cell's surveillance systems. This spatial seclusion also allows for accruing host and viral factors and building blocks required for efficient replication of the genome and prevents access of antiviral effectors. Even though the principle is iterated by almost all positive-strand RNA viruses infecting plants and animals, the specific structure and the organellar source of membranes differs. Here, we discuss the characteristic ultrastructural features of the virus-induced membranous replication organelles in plant and animal cells and the scientific progress gained by advanced microscopy methods.

**Keywords:** positive-strand RNA viruses; replication organelle; viral replication complex; membrane alterations; electron microscopy

**Citation:** Nguyen-Dinh, V.; Herker, E. Ultrastructural Features of Membranous Replication Organelles Induced by Positive-Stranded RNA Viruses. *Cells* **2021**, *10*, 2407. https://

Academic Editors: Thomas Hoenen and Allison Groseth

doi.org/10.3390/cells10092407

Received: 29 June 2021 Accepted: 2 September 2021 Published: 13 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

During infection, positive-strand RNA viruses utilize the host's cellular membranes to support every step of their replication cycle, i.e., virus entry, replication of the genome, and assembly and release of virions. These viruses induce (endo)membrane rearrangements in host cells to create a protective microenvironment for replication of their genomes and for subsequent production of new virions [1]. These endomembrane rearrangements form structures termed viral replication organelles (ROs), which are essential for virus replication. They are thought to shield viral replication intermediates from recognition and to protect them from the host cell defense systems, such as RNA silencing or interferon induction [2]. The ROs are confined membranous compartments generated by extensive alteration of (endo)membrane structures. While these membranous structures are essential for viral RNA replication, expression of single viral proteins is often enough to induce them, but size and detailed structural features may be different in the absence of virus replication.

These endomembrane re-arrangements can differ morphologically, from simple dilated membranous structures to very complex structures such as multi-vesicular bodies (Figure 1).

During the infection, viral proteins as well as hijacked host proteins target the (endo)membrane system of the host to remodel it. Through these virus-host interactions different RO structures are created, depending on the virus and the membrane source. The majority of ROs are vesicular structures. The simplest form are single membrane vesicles (SMVs), typically 50–200 nm in diameter with or without small pores that are 10–20 nm diameter which either link vesicles to each other or link the vesicle lumen to the external environment, i.e., the cytosol. These SMVs with pores, also called spherules, are believed to be generated from invaginations or evagination of the host organelle membranes. Multiple SMVs can be packed together in an organelle to form higher order vesicle packets (VPs). Slightly more complicated RO structures are double membrane vesicles (DMVs) normally

ranging from 200 to 400 nm in diameter. The biogenesis process of DMVs is not well understood. DMVs can be completely sealed with two membranous layers, the outer membrane can be connected to the organelle membrane the DMV originated from, and sometimes the inner vesicles share the same outer membrane and create a larger complex of DMVs. Some DMVs have open necks linking the internal lumen of DMVs to the external environment. More complex RO organizations such as multi-vesicular bodies (MVBs) also occur. These MVB structures are big vacuoles containing numerous small disordered membranous vesicles. Other, rarer membrane alterations are multi-membrane vesicles (MMVs), which are big multi-layered membranous particles of 300–400 nm in diameter, tubule-like structures of different diameters (20–50 nm) or zippered ER. Finally, massive unstructured membranous aggregates, which are called convoluted membranes (CMs), are frequently observed in virus-infected cells.

**Figure 1.** Membranous structures that occur in cell infected with positive-strand RNA viruses. Depicted are the most common membranous structures.

> Membrane sources are different membranous organelles such as the endoplasmic reticulum (ER), the Golgi apparatus, peroxisomes, mitochondria, and the plasma membrane, and in plants, chloroplasts, and tonoplasts. In the following sections, we discuss the ultrastructural features and membrane origins of ROs to integrate them into the cell biological context of the infected cell (Table 1).


**Table 1.** Membrane sources and morphologies of the replication organelles (ROs).

#### **2. Structure and Origin of Plant Positive-Strand RNA Virus Replication Organelles**

The ROs of positive-strand RNA viruses in plants are derived from different membranous organelles of the host including the ER, Golgi, peroxisomes, chloroplasts, and tonoplasts [34,35] (Figure 2).

**Figure 2. Structure and origin of plant positive-strand RNA virus replication organelles.** (**A**) 3D architecture of TuMVinduced complex membrane structures. Overview of a single slice of a tomogram of a TuMV-infected vascular parenchymal cell. (upper right) The 3D model shows a SMV with fibrillar material inside and with an adjacent intermediate tubular structure. (lower right) 3D model of a DMV with a core of electron-dense materials. Yellow, SMVs; light red, electron-dense materials; green, intermediate tubular structures; light blue, outer membranes of DMVs; dark blue, inner membranes of DMVs; dark red, the electron-dense materials inside DMVs [4]. (**B**) Dilated ER of BBSV-infected cells with SMVs (left) and 3D surface reconstruction of the tomogram corresponding to the intact spherules (right) depicting the outer ER membrane (yellow), BBSV-induced spherules (gray), and fibrillar materials inside the spherules (green). Scale bars 100 nm [3]. (**C**) Electron microscopy of MVB structures in PCV-infected BY-2 protoplasts. White arrows indicate clusters of vesicles. Single arrowheads correspond to MVB; MVB containing disordered membranous vesicles are indicated by black arrowheads, whereas those containing one row of vesicles that are surrounded by a single membrane are indicated by white arrowheads. White asterisks correspond to electron-dense material without detectable vesicles [5]. (**D**) TEM analysis and 3D reconstruction of MNSV-induced altered mitochondria. (left) TEM image of altered mitochondria. Numerous vesicles were observed on the external surface as well as internal large invaginations and internal dilations (star), or both. Yellow arrowheads indicate the pores connecting the lumen of the dilation to the surrounding cytoplasm. (right) 3D model of MNSV-induced altered mitochondria (blue, yellow, red, and purple) with large dilations inside and close interactions with lipid droplets (grey) and chloroplasts (green) [11]. (**E**) BSMV-induced chloroplast membrane rearrangement and 3D model of altered chloroplast membranes. (left) Tomogram slices of altered chloroplast membranes from leaves of BSMV-infected *N. benthamiana*. The arrowheads indicate the same spherules in different slices. (right) 3D model of remodeled chloroplast membranes induced by BSMV indicating the outer chloroplast membrane (cyan), inner chloroplast membrane (gray), and spherules derived from the outer membrane (yellow) [13]. (**F**) 3D visualization of remodeled tonoplasts in CMV-infected cells. (upper left) Tomogram slice of a CMV-infected *N. benthamiana* leaf cell. CMV-induced spherules are observed on a vacuolar membrane and in a MVB (arrowheads). The cell wall (CW), cytosol (Cy), and vacuole (Va) are indicated. Scale bar 500 nm. (lower left) 3D model depicting the vacuolar membrane (dark blue), MVBs (light blue), spherules on the vacuolar membrane and in the MVBs (yellow), and a membrane compartment (purple) with virus particles (red). (upper left) 3D model of the MVB with spherules open to the cytosol. (lower left) 3D model of the membrane compartment with virus particles. Scale bars 200 nm [14]. (**G**) 3D reconstruction of TBSV ROs in wild-type yeast cells characterized by peroxisome-peripheral MVBs depicting the MVB membranes (yellow), vesicle-like spherules (blue) located close to a mitochondrion (red) [9]. (**H**) Electron micrographs of the mesophyll cells of WYMV-infected wheat. The presence of

membranous inclusion body structures in the cytoplasm. The ER, membranous inclusion (MI), mitochondria (Mt), pinwheel inclusion (PW), and virus particles (VP) are labelled [6]. (**I**) A series of 2–7 appressed layers of double-membrane ER in yeast cells expressing both 2a pol and 1a of BMV, double-membrane ER layers are separated by regular, 50–60-nm spaces, the nucleus (Nuc) and cytoplasm (Cyto) are indicated. Scale bars 100 nm [7] Copyright (2004) National Academy of Sciences, U.S.A. The different parts were reproduced with permission.

#### *2.1. The Secretory Pathway Represents a Major Source for Membranes of Replication Organelles*

The secretory pathway of plant cells is frequently targeted by viruses as a source of membranes [34]. Like its mammalian counterpart, it is composed of a complex membrane network including the ER, the Golgi apparatus, the trans-Golgi network (TGN), and endosomes. This pathway is involved in the synthesis, modification, and transport of proteins, lipids, and polysaccharides [35]. Among those organelles, the ER is most frequently targeted by viruses for their productive replication. The ER is an extensive membrane network with specialized subdomains that occupies large parts of the cell and is the prime spot for lipid and protein synthesis. Increased protein (and lipid) synthesis occurs during infection with diverse viral species and can trigger ER stress responses.

During infection with positive-strand RNA viruses, the ER structure is often dramatically changed due to the interaction between viral and host membrane proteins to form the ROs. Some viruses, such as brome mosaic virus (BMV), tobacco mosaic virus (TMV), and red clover necrotic mosaic virus (RCNMV), induce massive ER proliferation forming ER aggregates either in the perinuclear region or randomly dispersed in the cytoplasm that are called convoluted membranes or membranous web [36–39]. Other viruses such as beet black scorch virus (BBSV) or tobacco necrosis virus (TNV-W) induce ER membrane dilations and invaginations that are rounded structures of up to 400 nm within the ER cisternae [3,40]. Along with the rearrangement of ER membranes, viruses also form higher order membrane structures called vesicle packets (VPs) containing small vesicle structures which are 50–100 nm in diameter [3]. Most of them are spherules composed of single or double membranes, called single or double membrane vesicles, SMVs or DMVs, respectively. Those vesicular structures are the areas where viruses replicate their genome [3]. The spherule structures in BBSV-infected cells are arranged along the VP membranes and are SMVs. Each spherule vesicle has a narrow neck (5–10 nm in diameter) linked to the VP membrane and thus connecting the spherule interior to the cytoplasm, suggesting that they are formed by invagination of ER membranes [3] (Figure 2B). Those VPs containing spherules with opened necks to the cytoplasm are also reported in other positive-strand RNA viruses that infect animal cells such as viruses in the *Flaviviridae* family [17,41].

In contrast to BBSV, the ROs of peanut clump virus (PCV) in tobacco protoplasts form VPs containing multiple SMVs which are called multivesicular bodies (MVBs). These MVBs contain multiple disordered membranous vesicles of 80–200 nm in diameter often in one row of vesicles and surrounded by a single membrane [5] (Figure 2C). Interestingly, Turnip mosaic viruses (TuMV), do not only induce formation of SMVs but also of DMV–like structures that are found in the perinuclear cytoplasmic region [4] (Figure 2A). The DMVs formed during TuMV infection occur during the late stage of infection concomitantly with massive membrane arrangements leading to altered endomembrane structures such as dilated ER and membranous inclusion bodies [4]. Cytoplasmic or membranous inclusion bodies (MIBs) were observed not only in TuMV infection but also in cells infected with different viruses such as wheat yellow mosaic virus (WYMV). WYMV forms MIBs in infected wheat plants that are large, amorphous, crystalline lattice-like inclusion bodies in the cytoplasm. The periphery of these MIBs appears to be connected to the rough ER [6] (Figure 2H), but high-resolution structural information is not available yet.

However, other membranous structures beside spherular invaginations and vesicles can support RNA virus genome replication. For example, BMV in yeast cells can replicate the RNA at multilayer stacks of appressed double membranes [7] (Figure 2I). In infected cells, the relative expression levels and interactions between viral 1a and 2a-pol proteins

can change the structure of perinuclear membrane rearrangements associated with RNA replication from small spherular invaginations to large stacks of 2–7 appressed layers of double-membrane ER. Intriguingly, these membrane stacks are highly ordered with 50–60 nm spaces, which is exactly the same width as the diameter of the spherules. These karmellae-like, multilayer structures are composed of stacks of ER that arise around the nucleus by folding over continuous sheets of ramified, double-membrane ER with its enclosed lumen. The double-membrane layers contain 1a and 2a-pol proteins and support BMV RNA replication but were not observed in yeast cells when only 1a or only 2a-pol proteins were expressed. Individual expression of BMV 1a induces only perinuclear spherules while 2a-pol alone does not cause any membrane alterations [7].

#### *2.2. Peroxisomes and Mitochondria as Membrane Origins*

The plant peroxisome is a single membrane-bound organelle that is solely responsible for beta-oxidation of fatty acids and the glyoxylate cycle, reactive oxygen species and reactive nitrogen species metabolism, and is involved pathogen defense. It is also one of the main target organelles for viruses as a membrane source to form ROs, especially for viruses in the *Tombusviridae* family such as tomato bushy stunt virus (TBSV) or cucumber necrosis virus (CNV) [9,10]. TBSV replicates in peroxisome-derived MVBs both in plant and yeast cells that are often found in close proximity to mitochondria (Figure 2G) [9]. Those MVBs are interconnected through membranes and might be nascent peroxisomes whose maturation and detachment from the ER is blocked by viral factors. In *N. tabacum* cell lines, TBSV p33 protein targets to peroxisomes and induces clustering and the formation of peroxisomal ghosts, but not MVBs, when expressed on its own [42]. CNV infection induces peroxisome biogenesis to form ROs [10]. Following infection, the peroxisomal boundary membranes are highly vesiculated, leading to the formation of doughnut- or C-shaped MVBs with the central region containing cytoplasmic material. The interiors of these doughnut-shaped MVBs contain many single-membrane vesicle-like structures with 80–150 nm in diameter. These vesicles appear to be connected to the MVB boundary membrane through a neck, and they provide the sites for CNV genome replication [10]. If peroxules that form in response to oxidative stress, which often occurs during virus infection, are hijacked by viruses as well, is currently unknown.

Interestingly, members of *Tombusviridae* not only target the ER or peroxisomes but also the mitochondria to form ROs to support viral replication as exemplified by melon necrotic spot virus (MNSV) and Carnation Italian ringspot virus (CIRV) [11,12]. In MNSVinfected cells, the mitochondrial structure is dramatically altered, and these abnormal organelles are frequently found close-by lipid droplets and ER membranes [11] (Figure 2D). Ultrastructural changes include dilated cristae and a vesiculated outer membrane. This vesiculated membrane forms multiple single-membrane vesicles with 45–50 nm in diameter which surround the large dilations inside the mitochondria. These vesicles appear to be connected to the cytoplasm or to the internal lumen of the large dilations through necklike structures. Immuno-EM suggests that MNSV RNA and capsid proteins reside in the large dilations of abnormal mitochondria, suggesting that MNSV performs its genome replication as well as packaging in mitochondria and possibly within the interior of the vesicles [11].

#### *2.3. The Chloroplast and Tonoplast Are Plant-Specific Membrane Sources*

One organelle unique in plant cells that is also a target structure for many viruses is the chloroplast. Chloroplasts are membrane-rich organelles that conduct photosynthesis [43]. Barley stripe mosaic virus (BSMV) is a member of family *Virgaviridae* that alters chloroplast morphology during infection. In BSMV-infected plant cells, the membranes of the chloroplasts change dramatically with clusters of outer membrane-derived invaginated spherules (diameter ~50 nm with a neck of 11 nm) within inner membrane-derived packets (average diameter 112 nm) [13] (Figure 2E). The small spherules are linked via neck-like structures to the cytosol and immune-EM analysis revealed the presence of the viral RNA

and replication proteins, suggesting that these spherules are the site of BSMV genome replication. In addition, big cytoplasmic invaginations surrounded by double membranes that contained virions were observed inside the chloroplasts [13]. This suggests that in addition to RNA replication, viral assembly takes place within the chloroplast.

The semipermeable membrane surrounding the vacuole is the tonoplast, an organelle that plays an important role in osmotic regulation of turgor pressure and that is targeted by viral infection. Already in the 1980s, cucumber mosaic virus (CMV)-infected leaf cells were shown to harbor tonoplast-associated vesicular structures [44]. The latest findings revealed that vacuole membranes are remodeled and invaginated in cells infected with CMV or tobacco necrosis virus A Chinese isolate (TNV-AC) [14]. Membrane invaginations form spherules at the periphery of the vacuole that are 50–70 nm in diameter (Figure 2F). These spherules contain neck-like structures that connect their interior with the cytosol. Interestingly, in CMV-infected cells, besides the spherules located at the tonoplast membrane, peripheral spherule-containing MVBs were also observed. The spherules inside the MVBs are also open towards the cytoplasm with a neck-like structure and the interior of the MVB seems to be connected to the vacuole. In addition to spherule-containing MVBs, membrane compartments harboring viral particles are found in close proximity to the vacuole and the ROs [14].

#### **3. Structure and Origin of Animal Positive-Strand RNA Virus Replication Organelles**

Similar to plant viruses, genome replication of all positive-strand RNA viruses that infect animal cells is intimately associated with membranes. The viral ROs supporting the replication of the viral genomes are generated from different host cellular membranous organelles including the endoplasmic reticulum (ER), the Golgi apparatus, mitochondria, lysosomes, and the plasma membrane [1] (Figure 3).

**Figure 3. Structure and origin of animal positive-strand RNA virus replication organelles.** (**A**) TEM images of HeLa cells transfected with the TBEV DNA replicon. White arrowheads show dilated ER areas; black arrowheads denote replication-vesicle-like structures inside the dilated ER areas. Insets show magnifications of the indicated areas. Scale bars 1 µm [41]. (**B**,**C**) DENV-infected Huh7 cells. (left) Tomogram slice shows DENV-induced convoluted membranes (CM),

vesicles (Ve), and tubes (T) that form a network of interconnected membranes in continuity with ER membranes. (right) 3D surface model of the membranes in the boxed area. The outer (cytosolic) face of the continuous membrane network is depicted in yellow; the ER lumen is dark [15]. (left) Stacked virus particles are in ER cisternae that are directly connected to virus-induced vesicles (white arrow). (right) 3D surface model of the virus-induced structures in the boxed area showing the continuity of virus-and vesicle-containing ER cisternae. ER membranes are depicted in yellow, inner vesicle membranes in light brown, and virus particles in red [15]. (**D**) Proliferation of the ER in human neuronal cells infected with TBEV. TBEV particles and TBEV-induced vesicles are located inside the proliferated and reorganized cisternae of the rough ER. 3D reconstruction of lamellar whorls, which are surrounded by cisternae arising from the rough ER (blue) and accommodate tubule-like structures (green). Detailed image shows the connection between the envelope (yellow) of a TBEV particle with nucleocapsid (red) and a tubule-like structure (indicated with an arrow) inside the rough ER. Scale bars 50 nm [19]. (**E**) 3D model of the HCV replication organelles surrounding lipid droplets. Electron tomography suggests that DMVs arise from ER membranes that are tightly wrapped around lipid droplets. (Left) Single tomographic slice of an HCV-infected cell with lipid droplets that are tightly wrapped by ER membranes and that stain positive for E2 and NS5A as revealed by fluorescence microscopy (not shown). (right) 3D reconstruction of the membranes surrounding the lipid droplet. ER membranes and DMVs are shown in yellow; the phospholipid monolayer of the lipid droplet monolayer membrane is shown in cyan. Insets illustrate that the DMVs originate from the wrapping ER membrane. Scale bars 100 nm [45]. (**F**) High-resolution analysis of ER-DMV interconnectivity in SARS-CoV-2-infected Calu-3 cells. Tomogram slices depict a membrane connector or zippered ER (light green) in contact with a DMV (red). (right) Superposition of rendered DMV and ER. Scale bars 200 nm [25]. (**G**) Tomogram slices and 3D reconstructions of mitochondria in FHV-infected *Drosophila* cells. (Left) Tomogram slices showing FHV-induced spherule rearrangements of a mitochondrion. Labels denote outer mitochondrial membrane (OM) and inner mitochondrial membrane (IM). White arrowheads indicate the necks that connect spherules to the OM. Asterisks mark two spherules that connect via necks to the OM. A red arrow marks the ∼10 nm channel connecting the spherule interior to the cytoplasm. (upper right) 3D tomogram image with blue indicates OM, white indicates FHV spherules. (lower right) A close-up view of the connections between the OM and the spherules and 90◦ rotation of spherules showing the channels that connect the spherule interiors to the cytoplasm [29]. (**H**) (upper left) Tomogram slice of FHV spherules in a mitochondrion. Mitochondrial outer membrane (red), spherule membrane (blue), interior spherule filaments (black), and spherule openings (white) are indicated with arrowheads. Scale bar 100 nm. (lower left) 3D reconstruction of the spherule outlined in upper panel. Scale bars 50 nm. (right) Filaments are associated with FHV spherule pores. Tomographic slices with arrowheads pointing to the mitochondrial outer membrane (red), the spherule membrane (blue), the spherule opening (white), and the extruding filaments that likely represent viral RNA) (black). Scale bars 100 nm [30]. (**I**) 3D ET volumes of RUBV replication complex in BHK-21 cell. Tomogram slice (left) and the corresponding 3D model (right) of a CPV (yellow) surrounded by the rough ER (light green) and containing a number of vacuoles, vesicles, and a rigid straight sheet (brown) that is connected with the periphery of the CPV; mitochondria (red), vesicles and vacuoles (white) and cytoplasm (grey). Scale bars 200 nm [31]. (**J**) Poliovirus ROs in HeLa cells. (left) Viral replication structures are strongly associated with staining for a Golgi antigen, GM130. Scale bar 500 nm. (right) 3D reconstructions of poliovirus ROs at the early, intermediate, and late stages, 3, 4, and 7 hours post infection, respectively, each depicting central slices in tomographic volumes, central slices with segmented overlays, and segmented volumes, with blue indicating SMVs and yellow and green indicating inner and outer membranes of DMVs, respectively. Scale bars 100 nm [27]. (**K**) Plasma membrane invaginations and vacuole formation in SINV-infected BHK-21 cells. Scale bar 200 nm [32]. The different parts were reproduced with permission.

#### *3.1. The ER Is the Main Hub for Animal Virus RO Fomation*

Among the different membrane-bounded organelles, the ER represents the main membrane source for many positive-strand RNA virus ROs in animal cells [46]. The *Flaviviridae* family is one positive-strand RNA virus family that is well-known for ER-based RO formation [16,46,47]. In cells infected with dengue virus (DENV) tick-borne encephalitis virus (TBEV), West Nile virus (WNV), or Zika virus (ZIKV), the ER structure is dramatically altered owing to viral genome translation and replication. These viruses induce the formation of different membranous structures in the cytoplasm: vesicle packets (VPs) inside the ER, convoluted membranes (CMs) (Figure 3B), which are peculiar membranous aggregates with unknown function [18,47,48], and, in some cases, dilated ER, which are enlarged rough ER cisternae filled with granular material e.g. in TBEV infected cells [41,49]. In Hela cells transfected with a TBEV DNA replicon, the dilated ER cisterna grow to big cytoplasmic vacuoles containing small spherule-like structures 80–100 nm in diameter, which have open necks towards the cytoplasm (Figure 3A) [41]. The most prominent membranous structures derived from the ER in flavivirus-infected cells are the VPs that are the sites of viral genome replication and thus represent the ROs [17,41,50]. Early immuno-EM studies in DENV-infected insect cells indicated that VPs (or smooth membrane structures, SMS) are the site of DENV RNA replication [50]. These VPs are ER-derived membranous structures that are dilated ER cisterna containing single-membrane vesicles (SMVs) with a diameter of 80–150 nm [15] (Figure 3C). These SMVs originate from the invagination of the ER membrane into the ER lumen, have a spherule structure with small, 10–15 nm diameter necks opening to the cytoplasm. Necks were also observed linking SMVs inside the VPs in WNV-infected cells [16]. Densely packed viral particles are frequently within the ER in close proximity to VPs [15] (Figure 3C). Interestingly, an electron tomography (ET) study of TBEV-infected human neuronal cells investigated the proliferating ER in infected cells and found additional tubule-like structures of different diameters (20–50 nm) inside ER cisternae [19] (Figure 3D). In some instances, these tubule-like structures have direct contacts with viral particles inside these proliferated ER cisterna [19]. The function of these tubule-like structures is thus far unknown; they may represent membranous structures involved in viral replication, abnormal cellular structures arising due to altered membrane metabolism, or a feature of cellular process to limit the viral infection [51].

Among the members of *Flaviviridae*, hepatitis C virus (HCV) is somewhat unique regarding the prototypical RO structures. In HCV-infected hepatocyte cells, ER membranes are intensively rearranged to form the membranous web (MW). The MW contains vesicles of different morphologies, mainly SMVs or DMVs, embedded in a matrix of membranes which are sometimes close to or wrap tightly around lipid droplets [21,22]. HCV infection as well as expression of single HCV proteins induce different types of membranous vesicles in cells [20–22]. While NS3/4A and NS4B induces only SMVs, NS5A induces MMVs and infrequently DMVs [22]. However, expression of the complete replicase complex (NS3-NS5B) is needed for formation of DMVs that are indistinguishable from the ones observed in infection [22]. In HCV-infected cells there are vesicles in clusters containing SMVs of variable sizes (100–200 nm in diameter), sometimes sticking together and harboring internal invaginations, and SMVs of a homogeneous size (~100 nm in diameter) that are clustered together and sometimes arrayed around lipid droplets. However, the most prominent vesicular structure induced by HCV are DMVs, likely representing the ROs. The DMVs are heterogeneous in size, with an average diameter of 200-400 nm, and are morphologically similar to membrane alterations identified in cells infected with coronaviruses [23] or picornaviruses [27]. These vesicles are characterized by two closely apposed membranes. EM/ET analysis revealed that most of the DMVs are generated from the ER and some of them are still connected to ER sheets via their outer membrane [22]. Although most of DMVs are completely closed structures and it is still unknown why HCV would induce these closed structures, a small percentage of them (8–10%) [22] has an opening neck towards the cytosol. The opened and closed DMVs thus may reflect the different stages of DMV "maturation", early and late, respectively [22]. An immunolabeling study of purified DMVs revealed an enrichment for viral proteins as well as dsRNA suggesting that DMVs indeed play an important role for viral RNA replication [52]. Viral RNA amplification may occur inside DMVs, which would allow the exit of newly synthetized viral genomes as long as the DMV is open, but replication might also occur on the outer surface of DMVs [22,52]. A more recent study using correlative light and electron microscopy (CLEM) indicated that DMVs emerge from ER membranes which are tightly wrapped around lipid droplets [45] (Figure 3E). EM/ET analysis of HCV-infected cell revealed two types of lipid droplets: lipid droplets that are tightly wrapped by the ER and that stain positive for the HCV glycoprotein E2 and nonstructural protein NS5A by immunofluorescence microscopy as well as lipid droplets that are not wrapped by ER and that do not stain positive for E2 and NS5A. These data suggest that HCV proteins trigger wrapping of ER membranes around lipid droplets. This tightly closed contact between DMVs and ER-wrapped lipid droplets may enable short-distance trafficking of viral RNA from replication vesicles to assembly sites at lipid droplet–associated ER membranes [45]. Later during HCV infection,

multi-membrane vesicles (MMVs) with an average diameter 350–400 nm are generated, likely originating from DMVs through secondary enwrapping events [22].

DMVs are observed not only in HCV infection but also during infection with other positive-strand RNA viruses, such as members of *Nidovirales*, including coronaviruses and arteriviruses [1]. DMVs are well-known typical ROs of coronaviruses [23,24]. A new study employing 3D reconstructions using FIB-SEM (focused ion beam milling combined with scanning EM) to determine morphological alterations induced in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-infected human lung epithelial cells revealed extensive fragmentation of the Golgi apparatus, alteration of the mitochondrial network, and recruitment of peroxisomes to viral ROs, which are clusters of DMVs [25]. In the SARS-CoV-2-infected cells, the ER network was altered intensively to generate the ROs, which consist predominantly of DMVs with an average diameter of 250–350 nm. Theses DMVs were tightly connected with the ER network linking the outer membrane to ER-derived structures such as ER connectors. Similar to DMVs in HCV-infected cells, the DMVs in SARS-CoV-2-infected cells are mostly closed DMV structures. However, DMV-DMV contacts were observed in SARS-CoV-2-infected cells, either through funnel-like junctions between two DMVs or fused DMVs consisting of multiple vesicles sharing the same outer membrane [25]. As described above, SARS-CoV-2 induces the formation ER connectors between the DMVs and ER tubules [25] (Figure 3F). These membranous structures were also described as zippered ER in gamma- or betacoronaviruses, such as infectious bronchitis virus (IBV) or Middle East respiratory syndrome coronavirus (MERS-CoV) [24,26]. The zippered ER or ER connectors lack luminal space, suggesting that they are formed through zippering or collapsing of ER cisternae. However, in contrast to SARS-CoV-2, electron tomograms showed that IBV-induced spherules are tethered to zippered ER and that there is a small pore connecting the interior of the spherule with the cytoplasm [26]. Of note, in a recent study of ZIKV, zippered ER structures were also observed in infected cells [17]. 3D reconstruction of regions containing zippered ER in ZIKV-infected cells revealed that the collapsed ER was connected to regions containing invaginated replication vesicles [17,47].

#### *3.2. Further Down the Secretory Route, the Golgi Apparatus Supports RO Formation*

Many viruses rely on the secretory route through the Golgi apparatus for maturation and release of viral progeny. However, some viruses also employ Golgi membranes to establish their ROs for viral RNA replication, e.g. poliovirus or coxsackieviruses, which are members of *Picornaviridae* family. Membrane alterations in poliovirus-infected cells include the formation of SMVs and DMVs [28]. A recent publication employing immuno-EM with subsequent diaminobenzidine (DAB) labeling suggested that membrane rearrangements in poliovirus-infected cells may occur in a sequential manner [27] (Figure 3J). In the early stage of infection, small clusters of SMVs appear. Later in infection, they are replaced by either round or irregularly shaped DMVs. Interestingly, the small clusters of SMVs of 100–200 nm in diameter strongly stained positive for a Golgi antigen, GM130, a cis-Golgi marker, but not for calnexin, an ER marker. These data suggest that the ROs of polioviruses may originate from the Golgi apparatus. However, it is too early to exclude a role of the ER for biogenesis of these ROs as ER-proteins might be dislocated during RO formation. dsRNA, i.e. viral RNA replication intermediates, as well as metabolically labeled viral RNA were detected in both SMVs and DMVs of poliovirus ROs, suggesting that both structures are relevant sites for poliovirus RNA synthesis [27].

#### *3.3. Mitochondria, Lysosomes, and the Plasma Membrane Are Involved in RO Formation*

Interestingly, the flock house virus (FHV), a member of the family *Nodaviridae* targets the mitochondria to form ROs supporting their RNA replication. In FHV-infected *Drosophila* cells, the mitochondrial outer membrane is dramatically altered [29] (Figure 3G). The virus induces the formation of invaginations at the outer mitochondrial membrane into the spherule structures with an average diameter of 50 nm. All spherules are outer mitochondrial membrane invaginations with their lumen connected to the cytoplasm

through a small pore of 10 nm in diameter, which is sufficient for ribonucleotide import and product RNA export [29]. A recent cryo-electron tomography study showed the presence of electron-dense structures within the spherules, which likely corresponds to the viral RNA as the volume correlated well with viral RNA length [30] (Figure 3H). This study additionally revealed the structure and symmetry of the proteins that form the pore complex. These pore complexes were frequently associated with long cytoplasmic electron-dense trails, likely representing exported viral RNA [30].

The lysosome is another cellular organelle which is a favorite target for some positivestrand RNA viruses such as rubella virus (RUBV) and members of *Togaviridae*, including Semliki Forest virus (SFV) and sindbis virus (SINV) [53–55]. These viruses alter lysosome and endosome structures to form cytopathic vacuoles (CPVs) that represent the viral ROs [54]. In RUBV-infected cells, the rough ER, mitochondria, and the Golgi are clustered around CPVs, which are linked to the cytosol and enclose vesicular structures [31] (Figure 3I). These organelles contain active ROs from which replicated RNA is transported to virion assembly sites at Golgi membranes. These CPVs have a quite variable diameter of 600–2000 nm. Electron tomography and 3D reconstruction revealed that CPVs enclose a variety of different membrane structures such as stacked membranes, rigid membrane sheets, small vesicles, and larger vacuoles that are connected through membrane contacts with each other and functionally connected to the endocytic pathway. CPVs have additional membrane contact sites to other cellular organelles such as the rough ER and Golgi vesicles, but not to nearby mitochondria. Immunogold labeling confirmed the presence of replicase complex proteins and dsRNA inside CPVs, suggesting that RNA synthesis occurs on or in vesicles within the CPVs [31].

As mentioned above, alphaviruses, such as SFV, SINV and WEEV, are known to induce formation of CPVs in infected cells, which are modified lysosomes and endosomes and the sites of viral RNA replication. Interestingly, in SINV-infected cells spherules containing dsRNA and nonstructural protein (nsP) are initially formed at the plasma membrane [32,33]. Immunofluorescence microscopy and EM revealed that at early times of infection, viral nsPs as well as dsRNA replication intermediates locate to spherules the plasma membrane [33] (Figure 3K). These spherules form as evaginations at the plasma membrane and the presence of plasma membrane-associated dsRNA and ns proteins suggest that they represent ROs. Later in infection, these spherules are internalized by endocytosis; trafficking and maturation to CPVs is dependent on phosphatidylinositol 3-kinase activity and the cytoskeleton [33], highlighting the often complex nature of viral RO formation.

#### **4. Recent Technical Developments and Challenges**

For multiplication, viruses need to infect a suitable host cell to be able to replicate their genome, to produce and release new infectious virions, and thus continue the next round of the infectious cycle. The interactions of viruses with their hosts are highly dynamic, diverse and complex, and occur on multiple levels. It is important to elucidate the molecular mechanisms of these virus-host interactions in order to understand virus replication cycles and how viruses affect and alter the cell biology of their host to support viral replication. This knowledge is not only important for better understanding of the biology of viruses but also to support control of viral infections, to predict their effect on ecology and human health, and to design effective antiviral strategies against chronic and emerging viral infectious diseases.

"Seeing is believing", we clearly trust observations that we can visualize. Microscopy, especially high-resolution light/fluorescence and electron microscopy (EM) are important tools for visualizing structures of viral and host cell components and thus for the generation of general concepts governing virus-host interactions. Indeed, EM and virus research developments are deeply intertwined since the invention of EM [56,57]. EM is one of the critical methods to elucidate how viruses replicate in the microstructure environments of the infected cell in order to produce new virions [58]. In general, EM techniques encompass

two main applications: transmission EM (TEM) and scanning EM (SEM), which each are different microscopic techniques [59]. The resolution of SEM is lower than that of TEM. In contrast, SEM provides a larger sample scanning ability or a bigger field of view for both surface and volume. Therefore, TEM is the favorite method to study small structures in detail, whereas SEM applications help to expand the sample scales.

The combination of EM with advanced light microscopy techniques termed correlative light and electron microscopy (CLEM) provides even more detailed information as it allows to analyze the dynamics and localization of viral and/or host protein-protein interactions in the context of detailed structural aspects of the intracellular environment. In this method proteins are visualized through fluorescent tags or antibodies using light microscopy in order to find rare biological events or to identify specific structures prior to characterizing the structures and their surroundings at high-resolution using EM. The current full spectrum of state-of-the-art microscopic techniques covers an extensive range of scales, resolutions, and information. Many of the methods mentioned together with the viral RO structures in this review, such as electron (cryo)tomography, CLEM, volume SEM, or 3D TEM have thrived and were further advanced within only two decades, especially since cryo-EM was discovered and developed in the 1980s [60].

The newly advanced electron tomography (ET), including volume SEM and cryotomography, has been a useful method in elucidating the 3D volume architecture of viral ROs. Volume SEMs such as serial block face SEM and focused ion beam milling (FIB)-SEM have been used to explore virus-host interaction with the nanometer resolution in wider and thicker volume samples including tissues. Furthermore, advanced cryo-FIB-SEM techniques are applied on cryo-stage specimens, which can help to avoid the artifacts of conventional EM sample preparation due to chemical fixation and staining processes and can also help to improve the stabilization of native structures in the specimen [61]. Although currently cryotomography of FIB-milled cryo-lamellae is the outstanding method in ET, the area that can be investigated is restricted to a very small and thin cellular region (the cryo-lamella) [62]. Difficulties in sample preparation combined with the need for highly demanding technical skills and high equipment costs are further limitations that are needed to be solved with technology developments in the future [62]. On the contrary to volume SEM, cryotomography methods can yield magnificent structural details with molecular-level resolution of the viral ROs in the cryo-native condition [63]. Cryotomography is currently one of the most powerful methods for investigation and characterization of the biological structures of viral ROs from the macro-structural morphology to the nano-organization of detailed protein structures which were presented in many current studies on viral ROs discussed in this review. Furthermore, current cryo-CLEM application, which combines cryo-light microscopy and cryo-EM opens a new way in investigating the molecular mechanisms of virus-host interactions more specifically and more accurate under cryogenic conditions [63]. However, similar to cryo-FIB-SEM, only a small area of the targeted cellular structure can be processed for investigation and the processing of cryotomography requires highly developed technical skills, limiting the popularity of 3D-cryoEM. For cryo-CLEM, the limited resolution of cryo-light microscopy, mostly based on wide-field light microscope also decreases the accuracy of this technique when it comes to localization of specific structural protein or events [64].

Of course, one main obstacle when investigating virus-infected specimens is the need for inactivation, especially for human pathogenic viruses. Thus, these samples require strong fixation that may cause artifacts. Alternatively, all steps including the image acquisition under cryo conditions have to be performed under biosafety containment, which is difficult to implement. Thus, we may need to rely on non-pathogenic model viruses for some of the advanced microscopy techniques.

#### **5. Conclusions**

Positive-stranded RNA viruses dramatically remodel intracellular membranes into distinct RO structures that support the synthesis of viral RNA. ROs provide optimal

micro-environments for viral genome replication and shield replication intermediates such as double-stranded RNA (dsRNA) from detection by innate immune sensors. Many questions about the biogenesis process viral ROs remain unanswered, i.e., for many viruses we do not have detailed information on host factors such as proteins and specific lipids that contribute to RO formation. Likewise, the dynamic nature of how and where and when during infection viral proteins required for RO formation interact with host proteins to remodel intracellular membranes into viral ROs and to stabilize the RO morphology remains to be determined. For many viruses pores connecting the RO interieur with the cytosol are observed but how viral proteins interact with host membrane proteins to stabilize these pore structures are still poorly understood [65]. The crown-shaped molecular complexes of some of the pores unveiled in recent studies of positive-stranded RNA viruses have provided us an overview of the protein complex organization of these pores [30,66,67]. However, how flexible the pores are and how the pore proteins regulate the transit of proteins and nucleotides/viral RNA from and to ROs and, possibly, coordinate it with other processes in the viral replication cycle is still poorly investigated. For other viruses, closed ROs have been observed frequently. If they are inactive/old ROs or just open up intermittently is still unclear. Elucidating how the viral replication complexes work on a molecular level and integrating biochemical knowledge with structural information gained by EM analysis are challenging goals for the future.

**Author Contributions:** Writing—original draft preparation, V.N.-D. and E.H.; writing—review and editing, V.N.-D. and E.H.; visualization, V.N.-D. and E.H; funding acquisition, E.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Deutsche Forschungsgemeinschaft (DFG HE 6889/2 and HE 6889/5) and the LOEWE Center DRUID (Novel Drug Targets against Poverty-related and Neglected Tropical Infectious Diseases).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We would like to apologize to all colleagues whose work we did not cite. The different parts were reproduced with permission or under the Creative Commons CC BY license: Figure 2 (A) [4], (B) [3], (C) [5], (D) [11], (E) [13], (F) [14], (G), [9], (H) [6], (I) [7] Copyright (2004) National Academy of Sciences, U.S.A; Figure 3 (A) [41], (B-C) [15], (D) [19], (E) [45], (F) [25], (G) [29], (H) [30], (I) [31], (J) [27], (K) [32].

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Olga Dolnik , Gesche K. Gerresheim and Nadine Biedenkopf \***

Institute for Virology, Philipps-University Marburg, 35043 Marburg, Germany; dolnik@staff.uni-marburg.de (O.D.); gesche.gerresheim@staff.uni-marburg.de (G.K.G.) **\*** Correspondence: nadine.biedenkopf@staff.uni-marburg.de; +49-(0)-64212825307

**Abstract:** Infections by negative strand RNA viruses (NSVs) induce the formation of viral inclusion bodies (IBs) in the host cell that segregate viral as well as cellular proteins to enable efficient viral replication. The induction of those membrane-less viral compartments leads inevitably to structural remodeling of the cellular architecture. Recent studies suggested that viral IBs have properties of biomolecular condensates (or liquid organelles), as have previously been shown for other membraneless cellular compartments like stress granules or P-bodies. Biomolecular condensates are highly dynamic structures formed by liquid-liquid phase separation (LLPS). Key drivers for LLPS in cells are multivalent protein:protein and protein:RNA interactions leading to specialized areas in the cell that recruit molecules with similar properties, while other non-similar molecules are excluded. These typical features of cellular biomolecular condensates are also a common characteristic in the biogenesis of viral inclusion bodies. Viral IBs are predominantly induced by the expression of the viral nucleoprotein (N, NP) and phosphoprotein (P); both are characterized by a special protein architecture containing multiple disordered regions and RNA-binding domains that contribute to different protein functions. P keeps N soluble after expression to allow a concerted binding of N to the viral RNA. This results in the encapsidation of the viral genome by N, while P acts additionally as a cofactor for the viral polymerase, enabling viral transcription and replication. Here, we will review the formation and function of those viral inclusion bodies upon infection with NSVs with respect to their nature as biomolecular condensates.

**Keywords:** negative strand RNA viruses (NSV); viral inclusion bodies; biomolecular condensates; liquid-liquid phase separation (LLPS); viral replication; nucleoprotein; phosphoprotein

#### **1. Introduction**

As viruses are obligatory intracellular parasites, their replication cycle relies on essential processes in the infected host cell. Viruses thereby exploit and remodel the cellular architecture by inducing structural, functional, or biochemical changes to enable efficient viral replication.

During infection, many viruses induce the formation of distinct and specialized intracellular compartments that facilitate viral replication. Those specialized intracellular compartments are very heterogenous and designated as viral inclusions, inclusion bodies (IBs), viroplasms, virosomes, or viral factories and present a hallmark of viral infection [1,2]. Some of those compartments are connected directly to membranes, such as the endoplasmatic reticulum (ER) in Hepatitis C virus [3,4], dengue virus [5] or severe acute respiratory syndrome coronavirus (SARS CoV) 2 [6,7] infections, lysosomes in Semliki forest virus infection [8] or mitochondria (Flock House virus) [9]. These single- and double-membrane vesicles, convoluted membranes or tubular structures are a typical feature of infection by positive strand RNA viruses [10–15]. In contrast, viral inclusions during infections with many negative-sense RNA viruses are membrane-less but still localize in special cytoplasmic areas (summarized in Figure 1).

**Citation:** Dolnik, O.; Gerresheim, G.K.; Biedenkopf, N. New Perspectives on the Biogenesis of Viral Inclusion Bodies in Negative-Sense RNA Virus Infections. *Cells* **2021**, *10*, 1460. https://doi.org/ 10.3390/cells10061460

Academic Editors: Thomas Hoenen and Allison Groseth

Received: 19 May 2021 Accepted: 8 June 2021 Published: 10 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Figure 1.** Overview on viral replication organelles in cells upon infection with different RNA viruses. Viral replication compartments associated with membranes are depicted in green, membrane-less compartments are indicated in blue. Those with liquid phase properties are depicted as droplets. DMV, double membrane vesicles. ER, endoplasmatic reticulum. IAV, Influenza A virus. MTOC, microtubule organizing center, ERGIC, endoplasmtic-reticulum–Golgi intermediate compartment.

> Recent investigations could demonstrate that many of those IBs share common properties with liquid organelles or biomolecular condensates. Those active, biochemically functional and membrane-less cellular compartments have become an emerging interest during the last decade. Biomolecular condensates display the properties of liquids and are highly dynamic and regulated structures in the cell involved in many different biological processes [16,17]. The underlying biophysical mechanism is, in most cases, regulated by liquid-liquid phase separation (LLPS), a mechanism similar to a water-in-oil-mixture [18,19]. The emerging investigations of these highly dynamic structures lead also to a paradigm change with respect to viral IB formation and function. Typical features of biomolecular condensates like dynamics, fusion activity, and reversibility are also characteristic for viral IB formation [19–22].

> Here, we will review the current state of viral IB formation and function in infections with negative-sense RNA viruses, especially with respect to the emerging field of viral inclusions with properties of biomolecular condensates.

#### *1.1. Biomolecular Condensates*

In the last years, the physical properties of cellular molecules as a key factor for cellular organization gained more and more interest [16,17,22–24]. The mechanism and localization of biochemical processes have been for long years attributed solely to membranesurrounded organelles. However, studies over the last decade have proven evidence of a cellular compartmentalization lacking a lipid boundary. Those membrane-less cellular structures are very heterogeneous in size and composition. Although they share similarities with the surrounding cytoplasm, they present separated, sometimes impenetrable cytoplasmic organelles that show a high dynamic plasticity and assemble/disassemble

rapidly [16,25–27]. Owing to their biophysical properties that they share with liquids (like droplet fusion, surface tension, etc.), these biochemical functional compartments have been referred to as liquid (droplet) organelles or, more common in cellular biology, as biomolecular condensates [16–19,27–30]. Biomolecular condensates have been observed in different cells among eukaryots, bacteria, yeast, and archae [23,31–34]. They are ubiquitous observed across cellular compartments. In the cytoplasm, they are represented by stress granules, P-bodies, G-bodies [23,35–38] or in the nucleus, by nucleolus or Cajal bodies [39–41], for example. The main biophysical mechanism underlying the formation of biomolecular condensates is the LLPS, a mechanism similar to a water-in-oil mixture that leads to the separation of two (or even more) phases when left unperturbed [42,43]. The transition from soluble molecules to condensates (saturation concentration), liquid crystals, or aggregates is strongly regulated by thermodynamical factors like temperature, concentration, valency, and interaction strength between molecules [28,44–47]. The interphase of the different phases results intracellularly in the membrane-less boundary of biomolecular condensates that allow penetration by molecules with similar properties, while it excludes molecules with dissimilar features [20,22,43]. In cell biology, LLPS originates from protein:protein, protein:RNA, or RNA:RNA interactions that lead to the remodeling of a soluble phase into a condensated, dense phase. A key factor here is the multivalency of the molecules itself: multiple inter- and intramolecular connections that can lead to the formation of condensates with multiple interaction partners [28,29]. Certain properties of protein:protein interfaces have already been shown to drive protein phase separation: arginine-glycine-glycine/arginine-glycine (RGG/RG) motifs [48], chargecharge interactions and intrinsically disordered regions (IDRs) [49–51]. Interestingly, many IDR-containing proteins also have RNA-interaction interfaces. Under high concentration and molecular crowding, structured protein domains have also been described to drive LLPS [29,52]. Accordingly, posttranslational modifications such as phosphorylation or methylation also have a big impact on the formation of biomolecular condensates [53–56]. Expression of RNA and proteins, changes in their ratio, as well as RNA-scaffolded assembly of proteins all contribute to condensation and dynamics of LLPS in cells [35,57–59].

#### *1.2. Viral Replication Cycle of Negative Strand RNA Viruses (NSV)*

The members of negative strand RNA viruses (NSV) comprise viruses that have a single-stranded, negative-sense RNA genome. They can be divided into virus families that have segmented genomes such as Orthomyxoviruses, Arenaviruses, and Bunyaviruses, or into non-segmented negative-sense RNA viruses (nsNSV, also termed *Mononegavirales*). The latter comprises several virus families (for example, *Paramyxoviridae, Bornaviridae, Rhabdoviridae, Pneumoviridae, Filoviridae*) with high relevance of individual representatives as human pathogens such as Measles virus (MeV), Nipah virus (NiV), Rabies virus (RABV), Respiratory Syncytial Virus (RSV), or Marburg and Ebola virus (MARV, EBOV) [60]. These viruses share a common architecture of their genomes. The RNA genome length varies between 12 (RSV, VSV) and 19 (filoviruses) kb in length and contains essential untranslated regions (UTRs) at their 30 - (*leader*) and 50 - (*trailer*) terminal ends important for viral transcription, replication, and encapsidation [61–65]. While the number of genes encoded by nsNSV varies among its families (from 5 to 10), the organization and relative position of the structural genes is highly conserved: 3'- *leader*- Nucleoprotein N- Phosphoprotein P- Matrix protein (M)- Glycoprotein (G)- RNA-dependent RNA polymerase (L for large protein)- *trailer*-50 (Figure 2A,B).

**Figure 2.** (**A**) Schematic diagram of a NSV particle. (**B**) General genome organization of a NSV. (**C**) Replication cycle of an NSV (based on a filoviral replication cycle). After entry into the cell **(1)** and release of the nucleocapsid into the cytoplasm **(2)**, primary viral transcription **(3)** is initiated by the integrated viral polymerase complex. Viral mRNAs are translated by the host translation machinery **(4)**. Synthesized viral proteins support new rounds of viral transcription **(5)**, replication **(6)** and nucleocapsid assembly **(7)**. Nucleocapsids are transported **(8)** to the cell periphery where they assemble to virions **(9)** and bud from the plasma membrane **(10)**.

The RNA is tightly encapsidated in a non-covalently manner by the nucleoprotein (N, NP) that forms together with the other viral nucleocapsid or accessory proteins a helical Ribonucleoprotein complex (RNP) [61,66–71]. NSV are enveloped viruses that integrate their surface protein(s) (G, GP, Hemagglutinin H, or fusion protein F) into the host-derived membrane (Figure 2A). A layer of viral matrix protein(s) (M, VP40) represents the matrix that connects the membrane with the nucleocapsid. The replication cycle of nsNSV takes place in the cytoplasm of the host cell, with the exception of Bornaviruses that have a nuclear phase during their replication [72]. Entry of the virus is mediated by the attachment and binding of the surface protein to its receptor and fusion of the viral with the cellular membrane [73–75] (Figure 2C). Subsequently, the viral RNP is released into the cytoplasm of the cell. The RNP serves as template for viral RNA synthesis that starts (owing to the negative-sense genomic RNA) with primary viral transcription accomplished by the incorporated viral polymerase complex [76,77]. The RNA-dependent RNA polymerase (RdRp) L forms together with the phosphoprotein (P, VP35) the viral polymerase complex that enables mRNA synthesis of the viral genes [78,79]. Some representatives of the nsNSV encode additional viral nucleocapsid or accessory proteins that are essential viral transcription factors often regulated by phosphorylation (for example, VP30 for filoviruses or M2-1 for RSV) [77,80,81]. mRNA synthesis starts at the 30 -end of the genomic RNA and results in short, uncapped leader RNAs and 50 -capped, 30 -polyadenylated mRNAs [82,83]. Transcription of the monocistronic mRNAs is assumed to follow a start-stop mechanism regulated by highly conserved gene start and gene end sequences located in UTRs [78,84]. Polyadenylation of the viral mRNAs by the viral polymerase slows down transcription at the gene ends that may result in dissociation of the RdRp from the template. The result is

a descending gradient of viral mRNAs from the first (N) to the last (L) gene, suggesting that the RdRp initiates transcription predominantly at the 30 -end of the viral genome and not from internal genes [76,82,85–89]. Following transcription, the cellular translation machinery translates the mRNAs into new viral proteins. Replication is carried out via the synthesis of a full-length antigenome in positive orientation that serves as a template for replication of the negative-sense genomic RNA. The switch from viral transcription to viral replication, when the RdRp ignores the transcription start and stop signals to synthesize the full-length antigenome, is not completely understood. It is suggested that the amount of newly synthesized N plays an important role to enable encapsidation of the nascent full-length antigenomic RNA during viral replication. N is synthesized as a monomer but starts to oligomerize quite rapidly and forms nucleocapsid-like structures, also with cellular RNA [90–92]. To prevent encapsidation of cellular RNA by N, N is kept soluble by the interaction with P [93]. The N0P complex, allows a concerted and regulated encapsidation of the viral RNA template [94]. However, different pools of polymerase complexes complemented by cellular and/or viral co-factors are also discussed to define either transcriptase or replicase activity of the RdRp [95]. Simultaneously with viral replication, genomic RNA serves again as template for further rounds of viral transcription accomplished by the newly synthesized polymerase complex components (secondary transcription).

Encapsidated genomic full-length RNA assembles together with the other nucleocapsid proteins to mature nucleocapsids that are condensed and transported along the cytoskeleton to the sites of viral budding at the cell periphery. The surface protein G co-localizes at the budding sites with the matrix protein M that drives the incorporation of the nucleocapsids into virions that are subsequently released from the plasma membrane [96–98].

#### *1.3. Characteristics of Viral Inclusion Bodies (IBs)*

New insights into the attributes of biomolecular condensate formation have also led to a reconsideration of viral IB formation in the virology field. High similarities of viral IB formation with biomolecular condensates driven by LLPS are obvious. Many viral IBs upon infection with nsNSVs have a high dynamic plasticity, they assemble/dissemble rapidly during infection, grow in size and appearance, and allow transport of exclusive molecules from in- or outwards. A major driving force in NSV IB formation is the expression of N and P proteins that are suggested as the basic scaffold in IB formation during infection. Two types of N-P interactions involving different interaction domains have been described: A monomeric N0P complex preventing association of N with cellular RNA, and nucleocapsid-associated P upon N oligomerization following its binding to genomic and antigenomic RNA [99]. All these steps involve multiple protein:protein and protein:RNA interactions that are mediated by highly conserved IDRs in both oligomeric proteins, P and N. All these attributes would contribute to the multivalent interactions underlying LLPS. Furthermore, several studies have demonstrated that de novo RNA synthesis occurs in viral IBs [100–105]. While the cellular protein synthesis itself is often shut down due to the viral infection, viral protein synthesis starts on a large scale. Simultaneously with an excess of viral protein expression, the viral RNA is subsequently replicated, encapsidated by the nucleoprotein and packaged with the nucleocapsid proteins.

All these different steps lead inevitably to strong changes in viral protein:protein interactions or protein:RNA interfaces, that might also contribute to LLPS in the viral IB and its surrounding. Understanding the biophysical mechanisms of viral IB biogenesis and regulation will also contribute to understanding the role and function of IBs for viral multiplication.

#### **2. Viral Inclusions Formed upon Infection with Non-Segmented Negative Strand RNA Viruses (nsNSV)**

IB formation and changes in the phase separation due to viral infections might lead to the induction of essential subsequent steps of the viral life cycle like, viral RNA synthesis, encapsidation, assembly of nucleocapsid, and their transport to the cellular periphery. In the last years, there is significant new information about the replication of individual

nsNSV in correlation with LLPS, which we will review in greater detail (summarized in Figure 3).

**Figure 3.** Summary of inclusion body (IB) formation upon infection with NSV. Small IBs indicate minimal required viral proteins for IB formation, while larger IBs represent mature IBs as biomolecular condensates formed by LLPS. \* IB formation by LLPS suspected. Different steps of the viral life cycle taking place in IBs as indicated. In red cellular proteins that localize to IBs. RABV, Rabies virus [106]. VSV, vesicular stomatitis virus [103,107]. RSV, Respiratory syncytial virus [101,108]. MeV, Measles virus [109,110]. NiV, Nipah virus [111]. EBOV, Ebola virus [104,105]. MARV, Marburg virus [112,113]. IAV, Influenza A virus [114,115]. IBAG, IB associated granules. IBperi, perinuclear IB. IBpm, IB plasma membrane. NC, nucleocapsid. HSP70/90, heat shock protein 70/90. FAK, focal adhesion kinase. G3BP, Ras GTPase-activating protein-binding protein 1. TIA1, T-cell restricted intracellular antigen 1. TIAR, TIA1-related protein. PCBP2, Poly(RC) Binding Protein 2. p65, NF-κB subunit p65. p65 MAPK, p38 mitogen-activated protein kinase. OGT, O-linked N-acetylglucosamine transferase. MAVS, mitochondrial antiviral-signaling protein. MDA5, melanoma differentiation-associated protein 5. WDR5, WD repeat protein 5. Tsg101, tumor susceptibility gene 101. IQGAP1, Ras GTPase-activating-like protein 1. NXF1, Nuclear RNA export factor 1. CAD, carbamoyl-phosphate synthetase 2, aspar-tate transcarbamylase, and dihydroorotase. SRPK1, Serine-arginine protein kinase. PP2A-B56, protein phosphatase 2 B56 subunit. eIF4G, Eukaryotic translation initiation factor 4 G. eIF3, Eukaryotic initiation factor 3. PABP, Poly(A)-binding protein. Rab11, Ras-related protein Rab-11.

#### *2.1. Rhabdoviridae: IBs of RABV and VSV*

A prototype of IB formation upon infection with nsNSV are the Negri bodies that are formed in neurons upon infection with RABV [116]. These cytoplasmic IBs were named after their discoverer Aldechi Negri in 1903 and present a hallmark of rabies diagnosis in the central nervous system (CNS). Negri bodies have been described as places of viral transcription and replication [102]. Components of the viral replication machinery are hence localized in Negri bodies as well as the matrix protein M. Apart from that, cellular proteins like HSP70 and focal adhesion kinase (FAK) are recruited to those IBs [117–119]. Negri bodies were the first viral IBs that have been demonstrated to present organelles with liquid properties [106]. Using fluorescently labelled RABV together with live-cell imaging and FRAP (fluorescence recovery after photobleaching) technologies, the nature of Negri bodies as biomolecular condensates formed by LLPS was demonstrated. Negri bodies are spherical structures with fusion capacity; they show transit with vesicles and can be in reversible form once they encounter a physical barrier [106,116]. The highly dynamic formation of Negri bodies was shown by applying a hypertonic shock to the RABV-infected cells that resulted in the dis- and reappearance of Negri bodies in only 15 min. Interestingly, at later time points of infection, the shape of Negri bodies was changed and they were associated with membranes, most likely derived from the ER [106,116]. The minimal requirement for Negri body formation as a biomolecular condensate was the recombinant expression of N and P alone. However, the typical pinching off events seen from Negri bodies (most likely RNPs) were missing upon N-P expression, suggesting further viral or cellular factors that contribute to the final nature of Negri bodies. The key domains of P that mediate Negri body appearance in complex with N were narrowed down by mutational approaches to the dimerization domain, the amino-terminal part of its second intrinsically disordered domain (IDD2) as well as the C-terminus. IDDs in general have no stable three-dimensional structure, but instead show a high degree in flexibility that can result in binding to other proteins or RNA, as well as in post-translational modifications like phosphorylation [120]. In this regard, IDD1 and IDD2 of P are flanking a dimerization domain (DD) and, like the C-terminus, are phosphorylated. However, phosphorylation of P did not impact Negri body formation [106]. While it was previously shown that stress granules, also liquid organelles, are formed in close proximity to Negri bodies, fusion events or exchange of proteins between both could not be demonstrated, suggesting that both cellular compartments present separate phases within the cytoplasm [106,121].

VSV IBs appear first around 4 h post infection and are also the major site of VSV RNA synthesis. Primary viral transcription, however, is suggested to take place in the cytoplasm prior to IB formation [103]. VSV IBs were recently shown by live-cell imaging to present liquid organelles, whose formation is dependent on LLPS [107]. Disrupting the microtubule cytoskeleton with nocodazol resulted in round inclusions containing eGFP-P labeled VSV. Those IBs were able to fuse by random motion supporting the hypothesis of intrinsic surface tension of VSV IBs, a characteristic feature of LLPS. In contrast to other members of the nsNSV, besides the expression of P and N, IB formation additionally requires the expression of the VSV polymerase L [103,107]. This was tested by recombinant expression of the proteins and complementary by depleting viral protein expression in VSVinfected cells using puromycin, as global protein synthesis inhibitor, or protein-specific PPMOs (peptide-conjugated morpholino oligomers). In addition, the inhibition of M protein expression using a specific PPMO had no effect on the formation or properties of the IBs [103,107]. Using an inactive mutant of L, L G174A, revealed that IB formation is independent of viral RNA synthesis, suggesting that the nature of the protein:protein interaction is the driving force of VSV IB formation via LLPS.

#### *2.2. Pneumoviridae: IBs of RSV*

RSV IBs have been described as spherical cytoplasmic structures where viral transcription and replication occurs and to which all viral proteins of polymerase complex, N, P, L, M2-1 are recruited to enable viral RNA synthesis [101,122–124]. Besides the components of the viral polymerase complex, the nonstructural protein NS2 and the matrix protein M are recruited to RSV IBs [125,126]. RSV IBs also recruit cellular proteins involved in translation initiation, like the poly A binding protein PABP, translation initiation factor eIF4G [101], protein phosphatase 1 (for regulating RSV transcription mediated by M2-1 phosphorylation) [81] or heat shock proteins HSP90 and HSP70 [127,128]. Additionally, cellular proteins involved in nucleocapsid assembly and -transport like actin, actin-associated proteins and rhoGTPases like rac1, rhoA and cdc42 colocalize in IBs [129–131].

While genomic RNA could be detected in RSV IBs [124,132], a recent study confirmed additionally viral mRNA synthesis to be present in RSV IBs, independent of their size [101]. Live-cell imaging and pulse chase analyses with a fluorescently labelled recombinant RSV (M2-1 GFP fusion protein) underlined the dynamics of IB formation during the RSV replication cycle. A very intriguing finding of this study was the identification of a subcompartment inside the IBs by super-resolution microscopy, called IBAGs (for IBassociated granules), where newly synthesized viral mRNA accumulated together with the viral transcription activator M2-1, while N, P, L, and genomic RNA were excluded [101]. Formation of IBAGs was strongly dependent on viral RNA synthesis as their number increased during the viral replication cycle from 12 h p.i. on. Interestingly, while nascent viral mRNA and the cellular proteins PABP or eIF4G involved in translation initiation co-localized in IBAGs, other components of the cellular translation machinery, like the ribosomal subunit proteins S6 or L4, did not concentrate on IBs at all. As pulse-chase experiments could demonstrate that newly synthesized viral mRNA only transits through IBAGs, it is suggested that they might present rather transient mRNA storage sites but not sites of viral mRNA translation that most likely occurs in the cytoplasm. IBAGs share similarities with cellular stress granules that are formed by LLPS [133], although IBAGs do not contain typical stress granule proteins like G3BP or TIA-1. The minimal requirement of RSV IB formation is, like for RABV, expression of N and P alone [108,134]. The assembly of IBs was shown to be dependent on the RNA binding- and oligomerization capability of N and P, as N mutation towards a N0P complex was not sufficient to induce IB assembly in transfected cells [108]. With respect to P, it was demonstrated that the oligomerization domain as well as its C-terminus were essential for IB formation. FRAP experiments on expressed mCherry-tagged N and P proteins could demonstrate *in cellula* as well as in vitro that the formation of RSV IBs occurs by LLPS mediated by N- and P interactions [108].

#### *2.3. Paramyxoviridae: IBs of MeV and NiV*

In MeV-infected cells, all the components of the MeV polymerase complex N, P, and L, as well as C colocalize in cytoplasmic IBs where also viral RNA synthesis takes place [135–139]. However, IB formation is initiated by the recombinant expression of N and P alone, even in the absence of viral replication [109]. Extensive studies during the last years have been made to identify N and P domains that contribute to their interaction. Both proteins show a high plasticity with structured and disordered domains. N has been described to consist of a folded domain (NCORE) responsible for RNA binding, with two terminal arms followed by a highly flexible region called NTAIL [140–143]. The tetrameric P contains a long intrinsically disordered (PTAIL) and a shorter disordered domain (PLOOP) [140,144,145]. PLOOP is terminated by a small C-terminal three-helix bundle (XD) that has been shown to interact with RNA-associated NTAIL and, for parainfluenzavirus 5 also with L [144,146]. In contrast, interaction in the N0P complex is mediated via the C-terminal domain of P [147].

Recent studies could demonstrate that MeV IBs represent biomolecular condensates formed by LLPS [109]. Characteristics of biomolecular condensates and LLPS, like a highly dynamic exchange between materials inside the IB with its surrounding, were observed by live-cell imaging upon MeV infection. IB formation was highly dynamic from small spherical structures to large inclusions. Interestingly, while smaller IBs were ubiquitously distributed in the cytoplasm, larger IBs appeared at the perinuclear region. By inhibition of dynein, a motor protein, the formation of perinuclear larger IBs was reduced suggesting that small cytoplasmic IBs transported dynein-dependent along microtubules to the cell nucleus to fuse towards larger IBs. Two further important assets of biomolecular condensates could be detected for MeV IB: the recruitment of cellular proteins (for example, eGFP- or mCherrytagged WD-repeat protein (WDR) 5) [138], and recovery from photobleaching [109]. LLPS was initiated by the interaction of the C-terminal disordered region of N and P. N mutants that were unable to bind RNA could still form N- and P-mediated IBs. This suggests that RNP complexes, an often described driving force for LLPS [35,57–59], do not contribute to

MeV IB formation [109], in contrast to RSV [108]. In vitro experiments using co-expressed N0P complexes and different mutants thereof confirmed that phase separation in vitro is also mediated by P and N interactions [110]. Interestingly interaction of PXD and NTAIL has been previously described to mediate the transport of the polymerase complex to the nucleocapsid prior to RNA synthesis [144,145,148–150]. Preventing the interaction between P and N by an N S491L mutation, a mutation that reduced viral transcription in cells [148,151], resulted in a complete abrogation of LLPS *in vitro*. The same was true upon mutation of P S86A and S151A indicating that P phosphorylation also contributes to phase separation. Interestingly, adding RNA to N-P droplets in vitro leads to the recruitment of RNA to droplets and triggered the encapsidation of RNA by N to nucleocapsid-like structures. The rate of encapsidation in those droplets measured by real-time NMR was enhanced when compared to the dilute phase [110]. These data confirmed the formation of nucleocapsid-like structures in these droplets and suggested a role of LLPS for the maturation of MeV nucleocapsids.

IB formation is also found upon infection by Mumps virus [152], Parainfluenzavirus 3 [153,154] and 5 [155] and NiV [111] and also initiated mainly by N and P expression. However, whether those IBs have properties of biomolecular condensates that contribute to efficient viral replication is so far not clear.

NiV, a highly pathogenic member of the *Paramyxoviridae,* differs from the other nsNSV by the induction of the formation of two distinct types of IB during infection. While one type is localized as spherical structures in the perinuclear region (IBperi), the second type characterized by a square shape is found at the plasma membrane (IBpm) [111]. Both types show not only different localization in the cell but also differ in their kinetics of formation and their content of proteins. While IBperi are rapidly formed by N and P proteins upon transfection or early in infection, the matrix protein M is only found inside IBpm, suggesting that they present places of virion assembly and budding. However, fusion events could not be observed between IBpm, neither in transfected nor in infected cells suggesting a transport of nucleocapsids through the cell from one IB to another. Another very interesting finding was that IBperi did not contain positive-sense RNA (mRNA or antigenomic RNA), suggesting that they represent no places of viral RNA synthesis, which is in contrast to many other nsNSV inclusions bodies. From this study, it is suggested that viral RNA synthesis takes place in a network of membrane-like reticular structures close to the ER [156], which is supported by the detection of nucleocapsids outside the IBperi. Whether LLPS and phase separation plays a role in the formation of IBperi and IBpm during NiV replication is so far not clear.

#### *2.4. Filoviridae: IBs of MARV and EBOV*

For Filoviruses, it is not clear whether IBs represent virus induced liquid-like compartments characterized by LLPS, as published so far for other NSV [106,107,110]. However, the current literature provides some evidence that this mechanism of compartmentalization, resulting in high functional dynamic and flexibility during the replication cycle, might be applied by filoviruses as well.

Live-cell imaging and time course studies showed that first small IBs appear in the perinuclear region of filovirus-infected cells [104,105] (and own unpublished data for MARV). The small IBs grow with time, they can fuse with each other, and in addition undergo fission events generating smaller IBs from bigger ones [104] (and own unpublished data for MARV). These observations were made using recombinant viruses expressing fluorophore-tagged nucleocapsid proteins like L-mCherry or VP30-GFP and might suggest LLPS processes during IB formation [104,112]. Earlier studies using single protein expression showed that the nucleoprotein NP alone induces the formation of IBs in transfected cells. All other nucleocapsid-associated proteins VP35 (a P analogue), viral nucleocapsid proteins VP30 and VP24, as well as L are diffusely distributed in the cytosol upon single expression and become IBs localized when co-expressed with NP [157–159]. The nucleocapsid proteins are important for the formation and structure of infectious nucleocapsids

and possess in addition a wide range of functions in the filoviral replication cycle. VP35 is the co-factor of the viral polymerase L and inhibits IFN-signaling [160,161]. VP30 is a phosphorylation-dependent viral transcription factor necessary to initiate the formation of viral mRNAs [77]. VP24 is important for the formation and condensation of nucleocapsids and inhibits viral transcription and replication as well as innate immune response by interfering with interferon-mediated signaling [162–165].

It was recently shown that the C-terminal domain of EBOV NP is necessary for IB formation and that co-expression of VP35 can rescue IB formation upon expression of a C-terminal deleted NP [166]. This experiment suggests that IB formation and other functions of NP and VP35 involved in transcription and replication of viral RNA are separated processes, since RNA synthesis could not be rescued in this setting. Functional separation of different protein forms, for example, due to modifications like phosphorylation or different protein:protein complexes can occur by LLPS. Here, MARV and EBOV VP30 phosphorylation represent an example of how this modification changes protein:protein and protein:RNA interactions and influences its functions and localization in IB [77,167–170].

Viral RNA is the second important component detected in filovirus induced IBs [104,105,113]. Since early in infection, when primary transcription of viral mRNA takes place, IBs are not detectable, at later time points when protein translation starts and secondary mRNA transcription is initiated, IB formation colocalizes with *de novo* RNA synthesis and large IBs coincident with RNA replication [104,105]. Interestingly, IBs with different compositions of viral nucleocapsids proteins like L and VP35 were detected, suggesting the existence of different subsets of IBs with different functional properties [104]. The regulation of transcription and replication in filovirus IBs is still not understood and it has to be worked out if and how LLPS might favor one or the other process by formation of subcompartments, as shown for RSV IBAGs [101].

Ultrastructural analysis using electron microscopy identified IBs in filovirus-infected cells that contained nucleocapsids with different electron-densities [91,162,163,171,172]. Ectopic expression of nucleocapsid proteins revealed that thin-walled helices are formed in the presence of NP [171]. Thick-walled helices with high electron density can only be observed in the presence of NP, VP35, and VP24 [171]. The thick-walled helices are mainly located in the periphery of IBs, at the plasma membrane during viral budding and in extracellular virus particles [171]. It is presumed that the thin-walled helices represent RNPs, which serve as templates for the viral polymerase, and the thick-walled electron dense helices represent mature and transport-competent nucleocapsids in infected cells or nucleocapsid-like structures in transfected cells [162,171,173]. Therefore, a proper ratio of NP and VP35 in IBs seems to control the morphogenesis of nucleocapsids in EBOV-infected cells [174].

The formation of transport-competent nucleocapsids that have to be transported from the IBs to the budding sites seems to be highly dependent on VP24 functions in RNP condensation, which in turn blocks EBOV genome replication [162,165,175–177]. Ejection of transport-competent nucleocapsids from MARV and EBOV IBs correlates with high dynamics and the nature of described biomolecular condensates or liquid-like viral factories, which exchange material with the surrounding cytosol, as reviewed by Su and colleagues [112,178,179]. In addition, the transport of EBOV and MARV nucleocapsids from IBs to budding sites depends on actin polymerization, and the dynamic of IB assembly and disassembly is dependent on microtubules, representing a further characteristic described for liquid-like viral factories of other NSVs [104,112,178,180].

Filovirus IBs are not membrane-enclosed, as shown in many ultrastructural images; however, often located in close proximity to different cellular membrane compartments like ER, endosomal vesicles, and mitochondria [181–183]. Which and how host cell factors contribute to filovirus IB formation is not known. Several cellular proteins, like Tsg101, IQ-GAP1, NXF1, CAD and SRPK1, and others have been identified inside IBs, being important for different steps of the filovirus replication cycle [172,177,184–187]. Interestingly, it was published recently that ER contact sites regulate the dynamics of membrane-less organelles like P-bodies [188]. It is therefore also likely that filoviral IBs contact the different cellular

compartments to enable material exchange, for example, viral and cellular proteins, and RNA, to favor different viral replication steps (transcription and translation, replication, assembly, condensation and transport of nucleocapsids). It remains to be analyzed if the required ATP provided by mitochondria and the necessary translation of viral and cellular proteins in close proximity to IBs might be covered and orchestrated by the mechanisms of liquid-to-solid transitions [22,189,190].

#### **3. Viral Inclusions Formed upon Infection with Segmented Negative Strand RNA Viruses (sNSV)**

IAV, a member of the *Orthomyxoviridae*, belongs to the segmented NSVs containing eight segments of RNPs inside the virion. The fact that most virions contain precisely eight segments of each type indicates that genome packaging in IAV infection is a highly regulated process [171,191,192]. It is suggested that the whole genome assembly of the eight segments takes place before transport to the plasma membrane, where the final assembly of the virion takes place [171,193,194]. In contrast to most other members of the NSVs, IAV replicate their genome in the nucleus. The eight viral RNP segments exit the nucleus and accumulate in IBs in a perinuclear region that enlarge in the course of infection [114,115,195]. Since IAV replication takes place in the nucleus, IBs are no sites of viral RNA synthesis. However, a recent study could demonstrate that IAV IB formation displays characteristics of liquid organelles or biomolecular condensates. Their formation in close proximity of the ER exit sites is spatially regulated, dependent on Rab11-GTPase and shows continuous cycling events of vesicles between the ER and the Golgi apparatus [115,196]. As expression of a single viral RNP could already initiate the formation of viral inclusions, viral IBs obviously occur before the assembly of whole genome RNP complexes. Sharing properties of biomolecular condensates, it is supposed that IAV IBs segregate viral RNPs from the cytosol to increase their concentration at hotspots that, in turn, facilitate the recruitment of other viral RNPs to allow assembly of whole IAV genome complexes [115]. Given the special feature of IAV genome reassortment, it is likely that IAV IB formation plays an important role in the assembly of newly reassorted IAV genomes.

The genus *Bunyavirales* contains viruses with either bi- or tripartite genomes containing the L, (M,) and S segments. In contrast to other bunyaviruses, the nonstructural proteins (NSs) of Severe fever with thrombocytopenia syndrome (SFTS) virus were able to form viral IBs upon transfection and infection, whichwas dependent on NSs self-interaction. It could be further demonstrated that those NSs-induced IBs contain the nucleoprotein and are places of viral RNA synthesis [197]. Interestingly, a colocalization of the SFTS IBs with lipid droplets was observed, and inhibition of lipid metabolism negatively affected SFTS replication.

For Bunyamwera virus, viroplasms have been described as tubular structures associated with the Golgi apparatus and the rough ER that are places of viral RNA synthesis and assembly [198]. For Junín virus (an Arenavirus), the nucleoprotein N was shown to induce the formation of discrete cytosolic IBs that may present viral transcription and replication centers. In contrast to most other nsNSV, those structures were associated with membranes and contained lipid metabolites [199].

However, whether these structures share biochemical properties with biomolecular condensates is so far not clear.

#### **4. Role of NSV IBs in Antiviral Response**

Given the spatial segregation of viral IB from the surrounding cytoplasm, it is also conceivable that IB formation may function as an additional viral escape strategy to avoid recognition by intracellular components of the antiviral defense machinery. Activation of pattern recognition receptors (PPRs) like RIG-1 and MDA5 recognizing cytosolic dsRNA leads to the activation of type 1 interferon and inflammatory responses combating viral infection [200–202]. A key determinant of antiviral activity are the viral phosphoproteins that are also key regulators for viral IB formation. The P proteins and their analogues have been described to block, for example, phosphorylation of the interferon regulatory factor 3 (IRF3) or IRF 7 [203–208], bind to dsRNA, and prevent RIG-I signaling or PKR activation [207,209,210].

Preventing activation of cell-intrinsic defense by IB formation could either be enabled by sterical exclusion or by concentrated sequestration of antiviral sensors avoiding activation of downstream pathways [211].

RSV antagonizes the innate immune response by sequestering cellular proteins involved in antiviral response activities into the IBs, such as NF-κB subunit p65, p38 mitogenactivated protein kinase (MAPK), O-linked N-acetylglucosamine transferase (OGT), mitochondrial antiviral-signaling protein (MAVS), and MDA5 [132,212,213]. Sequestration of MAPK p38 and OGT was suggested to suppress MK2 activity and formation of stress granules [213]. The cellular proteins were recruited to the IBs most likely via their interaction with N or P, suggesting an immune evasion strategy independent of the immunomodulatory RSV proteins NS1, NS2, or SH [212]. However, although the NF-κB subunit p65 was recruited to IBs, there was no co-localization with N and P suggesting that p65 localization might be regulated by other multivalent interactions within IBs [212].

Stress granules (SG), also liquid organelles with a role in antiviral activity [133,214] have been found in close proximity to RABV as well as VSV IBs [106,121,215]. While active fusion events between both biomolecular condensates could not be observed for RABV, the SG marker protein G3BP was found in some of the RABV IBs [106]. The function of G3BP localization in RABV IBs is unknown but may point towards the direction that LLPS may exclude antiviral proteins inside viral IBs to block antiviral downstream effectors [106]. For VSV IBs, in contrast, some SG proteins such as T-cell restricted intracellular antigen 1 (TIA1), TIA1-related protein (TIAR) or Poly(RC) Binding Protein 2 (PCBP2) co-localized to IBs [215]. The same is true for the EBOV IFN antagonist VP35 that can disrupt SG formation by sequestration of SG proteins into EBOV IBs (eIF4G, eIF3, PABP, and G3BP-1, but no TIA-1) to block innate immune responses [186,187].

For SFTS virus, a bunyavirus, it was demonstrated that sequestering of antiviral factors like IRF7, RIG-I, or STAT2 into viral IBs via the interaction with NSs leads to the suppression of IFN-alpha and -beta signaling pathways [216–219].

#### **5. Conclusions**

Over the last decade, the understanding of the intracellular architecture has changed tremendously by the discovery that intracellular membrane-less compartments represent liquid organelles or biomolecular condensates formed by LLPS. This also led to a paradigm change in the field of virology, especially with respect to the underlying mechanism of viral IB formation and maturation. For many NSVs, the liquid properties of IBs could be already demonstrated, with strong evidence that expression of N and P proteins are mostly the minimal requirement for IB formation (Figure 3). This could be attributed to their special protein architecture that includes multiple disordered regions and RNA-binding domains, hence multivalent interaction interfaces that contribute to LLPS. While RNA synthesis does take place in some of the NSV IBs, the structural role of RNA synthesis for LLPS formation and contribution to IB maturation is not fully understood, as well as the assembly of viral nucleocapsids in or from matured IBs. One may speculate that the molecular crowding of viral (and also cellular) proteins upon viral infection initiates the formation of IBs above a certain threshold, laying the foundation for the induction of further steps of the viral life cycle, possibly also driven by LLPS. In that regard, nucleocapsid assembly may be triggered as a result of the environmental changes induced by N and P expression and RNA synthesis.

Different cellular proteins interacting and co-localizing with viral proteins inside IBs have been identified so far. How they contribute to IB formation and LLPS is until now elusive. It is also feasible that many more cellular proteins might be recruited towards IBs due to their similar physicochemical properties, and maybe not all of them by their direct interaction with a viral protein. These interactions might be transient and require more live cell imaging and time laps studies and the use of super-resolution techniques. Research on the composition of IBs in cells will be an exciting field in the future, although challenging, since the liquid properties will make IBs purification difficult. The role of IBs in innate immunity, and how sequestration of cellular antiviral proteins into viral IBs may contribute actively to counteract antiviral activity will be also of great interest in the next years.

Future research on the biogenesis of viral IB formation and the underlying biophysical mechanism will help to understand how IBs promote viral replication, and may lay the foundation of the development of future antivirals, leading to the disassembly of viral IBs or that that may block viral RNA synthesis in place.

**Author Contributions:** All listed authors (O.D., G.K.G., N.B.) have written, edited, and reviewed the manuscript. G.K.G. provided the visualization. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Philipps-University Marburg: O.D. (BSL4 facility core staff) and via a junior research group fund from the FCMH (Research Campus of Central Hessen) to N.B. (junior group leader), G.K.G. (post-doctoral position).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** In this section, you can acknowledge any support given which is not covered by the author contribution or funding sections. This may include administrative and technical support, or donations in kind (e.g., materials used for experiments).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**

