Perspective: fluorescence lifetime imaging and single-molecule spectroscopy for studying biological condensates

1.1 Why should we study biological condensates, intrinsically disordered proteins and their interactions with membranes?

Biological condensates, often formed through liquid-liquid phase separation (LLPS), provide a way for cells to organize biochemical reactions without the need for a surrounding membrane [2]. Many of the proteins involved in forming condensates are intrinsically disordered (IDPs), meaning they lack a stable three-dimensional structure. IDPs contribute to the dynamic and reversible nature of condensates. Membrane surfaces often serve as platforms for phase separation, influencing the formation and function of condensates [3]. The interaction between IDPs, biological condensates, and membranes plays a critical role in many cellular processes. Studying biological condensates, IDPs, and their interactions with membranes is crucial for advancing our understanding of fundamental cellular processes and their implications in health and disease [4].

1.2 Moving towards functional studies

Early work in the field focused on identifying a variety of biological condensates in cells, characterizing their physical properties, how they form from their molecular constituents, and how they are regulated.

Recently, the focus shifted to understanding the biological functions of condensates in cells. They have been implicated in a range of processes, covered by Alberti et al. [5], Shin and Brangwynne [2] among others:

1. Nucleating or activating processes

2. Buffering concentrations

3. Generating forces, e.g. membrane bending

4. Filtration, e.g. nuclear pore

5. Sensing environmental stimuli

6. Inactivating biochemical reactions

7. Regulating signaling by sequestering molecules

8. Acting as an organizational hub

Alternatively, Lyon et al. [6] proposed to group these functions into classes covering different length scales:

1. Molecular-scale functions (Å – nm): enhancement or suppression of biochemical reactions; regulation of macromolecular folding state

2. Mesoscale functions (µm): vectorial organization of biochemistry; establishing mesoscale architecture

3. Cellular-scale functions (µm – m): facilitating specific cellular localization; sensing and switching; buffering stochastic cellular noise

Biological condensates and IDPs have also been found to play roles in neurodegenerative diseases [7], cancer, and infectious diseases, as reviewed e.g. in Alberti and Dormann [4].

1.3 How can we experimentally approach these topics using new microscopy techniques?

Many experimental approaches exist to investigate various aspects of phase separation, the behavior of IDPs and their interactions with membranes. Among them, light microscopy techniques are essential. They can be applied across scales from in vitro studies to live cells and tissues, to observe processes that take nanoseconds, seconds or hours.

As Alberti et al. [5] point out, one should experimentally verify if a phase separation process observed in vitro for a particular biomolecule is indeed relevant to the biological process under study.

The goal of this perspective is to highlight where time-resolved and single-molecule fluorescence microscopy techniques in particular can offer new insights or complement other methods. I will especially focus on applications for Fluorescence Lifetime Imaging Microscopy (FLIM), time-resolved Förster Resonance Energy Transfer (FLIM-FRET), Fluorescence Correlation Spectroscopy (FCS), anisotropy imaging and time-resolved anisotropy measurements in investigating the biological functions of condensates, their role in pathology, the link between LLPS and IDPs, and the interactions between condensates, IDPs and membranes.

2.1 What is LLPS and what are condensates?

The same molecules can form different phases, e.g. water can exist as ice, water, or vapor. These phases differ in the higher-order organization of the water molecules. Transitions between phases are possible, such as condensation, i.e. the formation of dew drops.

More complex fluid mixtures can demix in a physicochemical process called LLPS, e.g. forming oil droplets in water. This results in a coexisting dilute and dense phase of a particular molecule, where the dense phase is often called condensate.

In a biological context, I will use the term condensate to refer to any membraneless intracellular assembly that has a physically defined boundary generated by surface tension, following Shin and Brangwynne [2] and Forman-Kay et al. [8].

2.2 When does LLPS occur?

Generally speaking, LLPS occurs under conditions for which interactions between like neighbors, i.e. macromolecule/macromolecule and water/water, are more energetically favorable than those between unlike neighbors, i.e. macromolecule/water interactions in a fully mixed solution [5], [9]. In their review, Hyman [9] provide detailed background information on LLPS and a comprehensive description of the physics of liquid states.

Phase diagrams are commonly used to describe the environmental conditions for which LLPS takes place. They show the concentration of a particular molecule on the x-axis and the environmental parameter under study, e.g. temperature, pH, ATP concentration, etc. on the y-axis. For a certain range of environmental conditions, and concentrations above the saturation concentration, the mixture exists either in a one-phase or a two-phase state, which are separated by the coexistence line. Beyond the critical point, the mixture does not phase separate anymore.

Phase diagrams of a system under study can be generated experimentally in vitro by systematically changing the concentration of the constituents under constant environmental conditions or by varying the environmental parameter of interest for a given concentration.

In living cells, generating phase diagrams is more challenging, because usually neither the concentration nor the local environment can be controlled directly and varied systematically: The fluorescence intensity can be converted to absolute concentrations using FCS, as described e.g. in Bracha et al. [10] and Sanders et al. [11]. Environmental parameters can be monitored live using fluorescent sensors. For example, Jain et al. [12] discovered that stress granule assembly and dynamicity require ATP by employing a GFP-based ATP sensor. Kim et al. [13] showed that ATP binding to CAPRIN1, an RNA-binding protein found in stress granules, P bodies, and messenger RNA transport granules, increases its phase separation propensity by monitoring TNP-ATP fluorescence.

There are several classes of fluorescent sensors: intensity-based ones that change their spectral properties, fluorescence lifetime-based ones that vary their decay behavior, and FRET-based sensor constructs that alter the FRET efficiency. Measuring the relative fluorescence intensity distribution in several channels requires careful controls, as the intensity can be influenced by photobleaching, autofluorescent background, and other instrumental factors. The fluorescence lifetime is generally independent of the intensity and thus offers a robust read-out. By now, a large library of fluorescence lifetime-based sensor fluorophores exists, e.g. for pH [14], temperature [15], viscosity [16], glucose [17], phosphate ions [18], and chloride ions [19]. Similarly, many different FRET sensors have been developed. While the FRET efficiency can be determined from donor and acceptor intensity, the variation of the donor lifetime on the other hand is a very reliable measure for the FRET efficiency as well. With FLIM or FLIM-FRET, respectively, these sensors can be read out quantitatively in real time from live cells using only a single spectral detection channel. Furthermore, these measurements are very robust, as the lifetime is insensitive to excitation intensity, scattering, photobleaching and other common artifacts.

2.3 Commonly used fluorescence methods

On the one hand, FLIM can be used for highly multiplexed imaging, as it can discriminate several fluorophores per spectral channel based on different lifetime decay behavior. Thereby, multiple molecular constituents of condensates can be observed simultaneously.

On the other hand, FLIM can be employed to quantitatively and robustly read out fluorescent sensors that are sensitive to specific environmental parameters. Thus, the influence of these parameters on phase separation behavior, condensate function, and more can be investigated. The sensors can be either organic fluorophores that change their lifetime or FRET-based sensor constructs. In the latter case, FLIM-FRET imaging of the donor lifetime can reliably quantify the FRET efficiency.

Moreover, FLIM-FRET imaging and smFRET can report on direct molecular interactions of labeled molecules. These could be e.g. proteins binding to form condensates, ligands or substrates binding to enzymes, or proteins wetting membranes and thereby coming into direct contact with labeled lipids.

Furthermore, FLIM-FRET as well as smFRET can measure absolute distances and distance changes in the range of 1–10 nm in cells and thereby shed light on the conformational dynamics of IDPs inside condensates.

Observation of Fluorescence Recovery After Photobleaching (FRAP) indicates that the labeled species is mobile. Based on the pattern of recovery, FRAP can also evaluate if condensates are spatially homogeneous, as one would expect for a liquid.

FCS measurements can be taken at selected points of interest in the sample. On the one hand, FCS can quantify molecular concentrations, a critical parameter for phase separation. On the other hand, FCS can measure diffusion speed and distinguish different modes of diffusion. Depending on the context, variations in diffusion behavior can inform e.g. on

1. local variations of viscosity inside and outside condensates, which slow down labeled macromolecules

2. liquid versus gel-like or glassy solid states, where in the latter molecular motion is arrested

3. binding and complex formation, upon which the larger complex moves more slowly

4. altered lipid diffusion in membranes after wetting by a condensate.

Single particle tracking (SPT) provides valuable data on molecular motion of individual molecules. SPT is often used to distinguish between different types of diffusion. It can also be employed to obtain the viscosity of condensates [20].

While FRAP, FCS and SPT all investigate molecular dynamics, they have different advantages and limitations (Table 1): FRAP is excellent for measuring diffusion on a larger scale, especially for moderately fast or slow-moving molecules in relatively simple systems. Its simplicity makes it a go-to technique for many researchers. On the other hand, FCS offers superior temporal resolution and sensitivity to molecular heterogeneity, making it ideal for studying fast dynamics or environments where different diffusion modes coexist. SPT is more complex but allows for tracking individual molecules with high precision, making it ideal for studying heterogeneity and detailed movement patterns (Table 1).

For more information on these methods, readers can refer to the following literature:

1. FRAP, FLIP, FLAP, FRET and FLIM: [21]

2. FRAP, FRET, FCS: [22]

3. FRET, FLIM, FCS, and FRAP: [23]

4. FLIM, FRET, FCS, FRAP, SPT, Transient state (TRAST) imaging: [24]

5. FLIM, fret, smFRET, FCS, anisotropy: [25]

6. FCS: [26]

4.1 How do condensates form?

For condensates to fulfill their biological function, they need to provide a unique internal environment favoring macromolecule/macromolecule interactions, which is different from the solvent environment favoring macromolecule/water interactions [8]. Macromolecules that are likely to undergo LLPS in solution share a common feature, namely multivalency of interactions. There are two classes of protein architectures that aid the formation of networks of multivalent interactions [2]:

1. Proteins with multiple folded domains which transiently interact with short linear motifs in other proteins, such as the SH3 domain which binds to proline-rich motifs [39]

2. IDPs or intrinsically disordered protein regions (IDRs) with multiple “sticker” motifs, that interact either with other IDRs or RNA [40]

The prion-like intrinsically disordered domain of FUS, an archetypal phase-separating protein, serves as a model system to study phase separation. SmFRET studies by Joshi [41] revealed two conformationally distinct subpopulations coexisting in the monomeric form. Furthermore, they showed that structural unwinding of the peptide chain as illustrated in Figure 1 promotes intermolecular contacts between IDRs, which form a dynamic network and drive phase separation.

Figure 1:

A schematic depicting the structural unwinding of compact conformers into partially extended conformers upon phase separation. Adapted from Joshi [41] under a Creative Commons Attribution 4.0 International License.

The “stickers-and-spacers” framework from research on associative polymers is used to explain how multivalent interactions drive phase transitions that create condensates, as introduced by Choi et al. [42]. Stickers, which form transient, weak bonds, can be single amino acid residues or short motifs. They are separated by non-interacting spacer sequences.

Martin [43] investigated how the amino acid sequence of the archetypal PLD or low-complexity domain (LCD) from heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1) (A1-LCD) determines its phase behavior. They found that uniform patterning of aromatic residues promotes LLPS and inhibits aggregation. Furthermore, they demonstrated that the multivalence (i.e. the number) of interaction motifs, in this case aromatic residues in PLDs, determines the extent of temperature-dependent compaction of individual molecules in dilute solutions. The number of aromatic residues also determines the coexistence curves (binodals).

The researchers used FCS to measure the diffusion and concentrations of the protein before LLPS, in the dilute and dense phase. This enabled them to obtain experimentally derived binodals. Based on their experimental findings, combined with computational results, they developed a sticker-and-spacers model that can predict the phase behavior of PLDs on the basis of their sequence.

Besides predicting the phase separation behavior from a protein’s amino acid sequence, the stickers-and-spacers model can even be employed to rationally engineer condensates of RNA and disordered sticker-spacer polypeptides with desired viscoelastic properties, as demonstrated by Alshareedah [44].

Another model that is used to explain phase separation behavior is that of scaffolds and clients, as introduced by Banani et al. [45]. Scaffold molecules drive phase separation, while clients partition into condensates. In native condensates, many molecules show an intermediate behavior, from being absolutely required for condensation, to not altering condensate properties [6]. Molecules that have more interactions are likely more scaffold-like, whereas molecules with lower connectivity are likely more client-like, as shown e.g. for stress granule assembly by Yang et al. [46] and for P bodies by Xing et al. [47].

The formation of condensates, particularly large ones in vitro, can simply be observed with time-lapse transmission or Differential Interference Contrast (DIC) microscopy. Whether molecules of interest partition into a condensate is usually verified with multicolor fluorescence microscopy, checking for colocalization between them and a known scaffold molecule. Here, FLIM can prove advantageous again, by enabling multiplexing of several fluorescent markers per spectral channel, as exemplified by Starling et al. [48].

Colocalization in an image with confocal resolution does not prove direct molecular interactions, binding between molecules or complex formation. To verify these, a range of techniques such as FCS, time-resolved fluorescence anisotropy, FLIM-FRET, Nuclear Magnetic Resonance (NMR), circular dichroism (CD), or single molecule tracking can be used. Binding affinities can also be measured with NMR or smFRET.

To elucidate the composition of a condensate, proximity labeling mass spectrometry can be combined with fluorescence microscopy, as demonstrated by Marmor-Kollet et al. [49] for stress granules.

4.2 What about condensate size?

Biologists are familiar with many structurally defined macromolecular complexes such as the ribosome, proteasome, and mitochondrial complex 1, which are stably folded and consist of proteins only or proteins and RNA. These are clearly different from condensates such as the nucleolus and stress granules, which consist of thousands of molecules with a highly dynamic relative organization. However, many condensates in vivo, like transcriptional foci, signaling puncta and ribonucleoprotein (RNP) assemblies, depicted in Figure 2, are small, usually 100 nm–300 nm in diameter, and contain only tens to hundreds of their constituents.

For these small condensates, several conceptual and practical questions still remain, e.g.:

1. Do condensates have to reach a certain size to create a unique internal environment that is different from the outside solvent [8]?

2. At which size or molecule number does a new function arise?

3. How do properties of small condensates with few molecules compare to those of much larger, often micron-sized condensates generated in vitro?

For example, a parameter that is very important in small condensates is surface tension, which can lead to a different organization of molecules at the interface compared to the center of the condensate and could thus influence function [6].

To approach these questions experimentally, one can employ super-resolution microscopy approaches, such as Stimulated Emission Depletion (STED) or Single Molecule Localization Microscopy (SMLM) imaging. The combination of STED and FLIM [50] as well as of SMLM and FLIM [51] for multiplexed imaging has been demonstrated. These techniques can visualize the spatial distribution of multiple constituents simultaneously with a resolution of tens of nm. But they require special fluorophores and labeling strategies or buffer conditions. In this regard, the newly developed combination of Image Scanning Microscopy (ISM) with FLIM (ISM-FLIM) offers the advantage of working with conventional fluorophores, placing no additional requirements on labeling, as shown by Castello [52] and Sisamakis et al. [53] among others. So, in addition to advanced multiplexing, reading out lifetime-based or FRET-based environmental sensors with sub-diffraction resolution becomes possible. Yet, ISM-FLIM is limited to a resolution of about 120 nm.

Beyond imaging, STED can also be combined with FCS to measure concentrations and investigate diffusion behavior by Lanzanò [54] and complex formation [55] in sub-diffraction volumes. STED-FCS allows one to freely tune the size of the observation volume to tens of nm in diameter and can thus even report on different modes of diffusion at smallest spatial scales, as demonstrated by Schneider [56].

4.3 What about the internal structure of condensates? How does it relate to composition, dynamics and function?

Biological condensates contain multiple molecular constituents, which may be immiscible. These can form inhomogeneous multiphase assemblies with complex multiphase or multilayer architectures that are not spherical anymore [8]. For example, droplets with lower surface tension (ConA) will engulf those with higher surface tension (ConB), if the surface tension of ConA/solvent < ConA/ConB < ConB/solvent. Several examples are well documented:

1. the nucleolus [57] and stress granules [12] that have a core–shell architecture

2. nuclear speckles exhibit a distinct architecture [58]

3. RNP bodies such as processing bodies and P granules show internal structuring [59], [60], [61]

For visualizing the distribution of multiple constituents in such coexisting multiphase condensates, spectral as well as lifetime-based multiplexed confocal and super-resolution imaging approaches as described above are powerful tools. Lyon et al. [6] point out that correlative cryo-super resolution fluorescence imaging and cryo-electron tomography can also prove beneficial in this regard.

4.4 How do IDPs behave and function inside condensates?

Condensates are greatly enriched in IDPs or proteins containing IDRs. These do not have a stable 3D structure but exist as conformational ensembles. Many experimental techniques exist to characterize these conformational ensembles and their dynamics, as reviewed extensively by Abyzov et al. [62]:

1. The overall size of the conformational ensemble can be determined with fluorescence anisotropy, dual-focus FCS, NMR, Small-Angle X-ray Scattering (SAXS), Dynamic Light Scattering (DLS), and Size Exclusion Chromatography (SEC)

2. Specific distances between amino acids can be obtained with smFRET, NMR, and cross-linking mass spectrometry

3. Secondary structures can be investigated with NMR (chemical shifts), and CD

In particular, FCS and smFRET, a “spectroscopic ruler”, are well-established, versatile tools for in vitro studies. Depending on the experimental context in which this ruler is applied, it can yield answers to many different questions, for example:

1. How do oppositely charged IDPs interact during phase separation? Chowdhury [63] used smFRET to prove that the formation of highly disordered complexes of linker histone H1 and chaperone prothymosin α is driven by counterion release. Rai [64] used FCS and anisotropy measurements to show that electrostatic interactions between tau and prion protein lead to the formation of nanoclusters inside condensates.

2. How does the high viscosity inside condensates affect IDP dynamics? Galvanetto [65] employed smFRET and nsFCS to demonstrate that linker histone H1 and chaperone prothymosin α remain highly dynamic on the molecular scale despite the high macroscopic viscosity of the condensate.

3. How do IDPs aggregate inside condensates? Wen [66] use smFRET and FCS to study the conformational changes of tau protein during phase separation. They observe an extension of tau, and the formation of nanoscale clusters, which can promote aggregation into amyloid fibrils. Ray [67] employ FRET to monitor the fibrillization of α-Synuclein inside condensates. In a similar vein, Ray et al. [68] performed time-resolved fluorescence anisotropy decay measurements which revealed the rotational motion of α-syn and thereby the increased rigidity of the molecules inside α-syn droplets compared to those outside. Droplets formed under different conditions and the aging of droplets over time could be observed.

4. How do IDPs interact with RNA? Cubuk [69] applied smFRET and FCS to characterize the SARS-CoV-2 nucleocapsid (N) protein, which is responsible for viral genome packaging. It undergoes LLPS with RNA and forms single-genome condensates instead of multi-genome droplets.

Exploring the conformational dynamics of IDPs in condensates in cells is more challenging, requiring non-invasive techniques like fluorescence microscopy.

In particular, FLIM-FRET can report on molecular distances and distance changes, as exemplified by the Lemke lab’s studies of the nuclear pore complex (NPC) [70]: Hundreds of IDPs, referred to as FG-NUPs, create a permeability barrier in the central channel of the NPC to regulate transport. How these highly dynamic FG-NUPs function to achieve specific yet efficient transport was not clear. The team used FLIM-FRET to obtain distance distributions of FG-NUPs inside the NPCs, which can yield chain dimensions according to polymer scaling laws as shown in Figure 3.

Figure 2:

Scales of biomolecular condensates and component biomolecules. Taken from Forman-Kay et al. [8] under a Creative Commons License (Attribution-NonCommercial 4.0 International), as described at http://creativecommons.org/licenses/by-nc/4.0/.

Figure 3:

Scaling law of the NUP98 FG domain. Taken from Yu [70] under a Creative Commons Attribution 4.0 International License.

4.5 How to measure enzymatic activity inside condensates?

One of the biological functions of condensates is to enhance or suppress biochemical reactions by concentrating a specific set of molecules. Furthermore, multiphase condensates can vectorially organize several reaction steps due to their multilayer architecture. This has been shown e.g. for ribosome biogenesis in the nucleolus in vitro [31], [71] and in vivo [72], [73].

Still, measuring biochemical activity inside condensates in cells to understand the effect of condensate formation on enzyme activity remains experimentally challenging. Fluorescence microscopy can verify the localization of an enzyme under study inside the condensate. FCS can also report on its local concentration compared to the surrounding solution, as well as on the concentration of fluorescent substrates or ligands. Moreover, novel sensor probes can reveal the presence of small molecule enzyme substrates and products. Here, FLIM imaging of lifetime-based sensors or FLIM-FRET imaging of FRET sensors offer a quantitative readout that is more robust than intensity-based reporters and require only a single spectral channel. This complements high resolution imaging mass spectrometry, which can also quantify small molecules inside condensates, as Pareek et al. [74] demonstrated for de novo purine biosynthesis. Furthermore, a variety of fluorescence-based enzyme assays that measures enzyme activity exists, based e.g. on fluorophores released from the substrate or on different spectra of substrate and product.

For example, Testa [75] directly visualized local pH changes inside condensates that concentrate the enzyme urease, as shown in Figure 4. Urease hydrolyzes urea, producing carbon dioxide and ammonia, a strong base. Therefore, urease activity can be observed with a pH-sensitive fluorophore. Beyond this, the researchers found that pH increase inside the condensates increases with their radius, which indicates transport limitations.

Figure 4:

Visualization of the pH change in the droplets containing 1 μM urease at different times using two different concentrations of substrate (50 and 100 mM urea) along with the control (no substrate). Scale bars are 50 μm. Adapted from Testa [75] under a Creative Commons Attribution 4.0 International License.

In turn ongoing biochemical reactions affect the properties the condensate. Solutes shift the phase diagram, and thus stabilize either the condensate or mixed phase. Reaction products change the surface tension and viscosity, which can result in chemo-mechanical coupling that enables condensates to swim, as revealed by Jambon-Puillet [76].

5.1 How do membranes and condensates interact with each other?

An emerging focus of research in this field is the interaction between cellular membranes, condensates and IDPs. Condensates can adhere to and partially or completely wet membranes, depending on the interfacial tensions between condensates, membranes and the cytosol. Thereby condensates influence membrane curvature and tension during processes such as endocytosis, cell migration and cell division.

Measuring the tension of cellular membranes is a complex task that used to require special techniques: Micropipette aspiration [83], [84] and optical tweezers [85], [86] measure the membrane tension locally in a labor-intensive experiment. Atomic Force Microscopy (AFM) determines the elastic properties, but can only access the cell membrane, not internal organelles [87].

FLIM is the new technique of choice to rapidly visualize and quantify changes in membrane tension non-invasively using the mechanosensitive Flipper-TR probe first introduced by Colom et al. [88]. Flipper-TR integrates into the hydrophobic core of lipid bilayers and changes its fluorescence emission and lifetime in response to mechanical changes.

Membrane deformation upon wetting is involved in different biological processes such as budding, nanotube formation, compartmentalization and fission, e.g.

1. Bergeron-Sandoval [89] described puncta of endocytic coat proteins enabling actin independent endocytosis in budding yeast.

2. Pombo-García [90] found that “membrane prewetting by condensates promotes tight-junction belt formation”.

Condensates also influence the lipids in membranes upon wetting in various ways, e.g.:

1. Lee [3] found that condensates reduce lipid mobility and reorganize lipid partitioning, employing multicolor imaging and FRAP.

2. Mangiarotti [91] observed increased lipid packing due to contact with condensates as well as increased dehydration, using hyperspectral and FLIM imaging of sensor fluorophores.

The local order of lipids can be investigated with FLIM using appropriate sensor fluorophores, as well as with anisotropy imaging of fluorophores that insert themselves into the lipid bilayer. Lipid diffusion speed and diffusion mode inside membranes directly below or next to condensates can be monitored with FCS techniques. Notably, STED-FCS is well suited to study hindered diffusion of lipids in membranes with high spatial resolution, as described in detail by Sezgin [92], Benda [93] among others.

In turn, there is growing evidence that membranes facilitate the formation of condensates involved in various biological processes:

1. For example, Snead et al. [94] found membrane recruitment of the protein Whi3 to the ER promotes RNP condensate formation, using FCS to measure Whi3 diffusion in solution or at membranes.

2. Kaur and Lee [95] used FLIM to study the interaction of α-synuclein with lipid membranes, by labeling α-synuclein with the environmentally sensitive fluorophore NBD, which serves as a direct membrane-binding reporter. The fluorescence lifetime distributions of this fluorophore obtained from different puncta in FLIM images showed that α-syn adopts a multitude of membrane-associated conformers, challenging currently available structural models.

3. Favetta [96] employed smFRET to study how phosphorylation induces a conformational change and alters the function of SARS-CoV-2 N protein. Phosphorylation switches between two membrane-associated behaviors of N protein, being either adhered to membrane surfaces or being tightly bound to RNA when engulfed into new virions.