Intact Viral Particle Counts Measured by Flow Virometry Provide Insight into the Infectivity and Genome Packaging Efficiency of Moloney Murine Leukemia Virus

Gammaretroviruses, or, more specifically, murine leukemia viruses (MLVs), have been a longstanding model for studying retroviruses. Although being extensively analyzed and dissected for decades, several facets of MLV biology are still poorly understood. One of the primary challenges has been enumerating total intact virus particles in a sample. While several analytical methods can precisely measure virus protein amounts, MLVs are known to induce the secretion of soluble and vesicle-associated viral proteins that can skew these measurements. With recent technological advances in flow cytometry, it is now possible to analyze viruses down to 90 nm in diameter with an approach called flow virometry. The technique has the added benefit of being able to discriminate viruses from extracellular vesicles and free viral proteins in order to confidently provide an intact viral particle count. Here, we used flow virometry to provide new insights into the basic characteristics of Moloney MLV.

ABSTRACT

Murine leukemia viruses (MLVs) have long been used as a research model to further our understanding of retroviruses. These simple gammaretroviruses have been studied extensively in various facets of science for nearly half a century, yet we have surprisingly little quantitative information about some of the basic features of these viral particles. These include parameters such as the genome packaging efficiency and the number of particles required for a productive infection. The reason for this knowledge gap relies primarily on the technical challenge of accurately measuring intact viral particles from infected cell supernatants. Virus-infected cells are well known to release soluble viral proteins, defective viruses, and extracellular vesicles (EVs) harboring viral proteins that may mimic viruses, all of which can skew virus titer quantifications. Flow virometry, also known as nanoscale flow cytometry or simply small-particle flow cytometry, is an emerging analytical method enabling high-throughput single-virus phenotypic characterizations. By utilizing the viral envelope glycoprotein (Env) and monodisperse light scattering characteristics as discerning parameters of intact virus particles, here, we analyzed the basic properties of Moloney MLV (M-MLV). We show that <24% of the total p30 capsid protein measured in infected cell supernatants is associated with intact viruses. We calculate that about one in five M-MLV particles contains a viral RNA genome pair and that individual intact particle infectivity is about 0.4%. These findings provide new insights into the characteristics of an extensively studied prototypical retrovirus while highlighting the benefits of flow virometry for the field of virology.

IMPORTANCE Gammaretroviruses, or, more specifically, murine leukemia viruses (MLVs), have been a longstanding model for studying retroviruses. Although being extensively analyzed and dissected for decades, several facets of MLV biology are still poorly understood. One of the primary challenges has been enumerating total intact virus particles in a sample. While several analytical methods can precisely measure virus protein amounts, MLVs are known to induce the secretion of soluble and vesicle-associated viral proteins that can skew these measurements. With recent technological advances in flow cytometry, it is now possible to analyze viruses down to 90 nm in diameter with an approach called flow virometry. The technique has the added benefit of being able to discriminate viruses from extracellular vesicles and free viral proteins in order to confidently provide an intact viral particle count. Here, we used flow virometry to provide new insights into the basic characteristics of Moloney MLV.

INTRODUCTION

Murine leukemia viruses (MLVs) were first discovered as a result of their oncogenic effects over half a century ago (1–4). Shortly thereafter, it was revealed that infections of germ line cells led to the heritable and ubiquitous nature of endogenous retroviruses (ERVs) (5, 6). In this regard, ERVs account for 8 to 10% of the mouse genome (7). Typically, most mouse ERVs were found to be closely related to gammaretroviruses resembling MLV (7). The broad popularity of MLV as a model retrovirus began with its critical role in David Baltimore’s research that led to the shared discovery of the reverse transcriptase with Howard Temin, who was focusing on another retrovirus, Rous sarcoma virus (8, 9). The role of MLV as a prototypical retrovirus has been escalated by its genomic simplicity, originally thought to encode only the bare necessities for retroviral replication and infection: Gag, Pol, and Env (10). Dissection of the function of these three proteins and of the replication cycle of MLV has been invaluable for the field of retrovirology (11, 12).

Much of the knowledge acquired by studying MLV family members has been relatable to more complex retroviruses as well as those with more direct relevance to humans, such as human immunodeficiency virus (HIV) and human T-cell lymphotropic virus (HTLV). These findings include the processes of reverse transcription (RT), integration, viral genomic RNA dimerization, and selective viral RNA packaging (reviewed in references 13–15). The enhanced understanding of retroviruses has also created the opportunity of using MLV as a gene therapy vector in the laboratory, given its well-documented genome and versatility for stable gene delivery. In the clinic, however, HIV-based lentiviral vectors have become more favorable candidates for gene therapy in humans due to their reduced oncogenic potential compared to MLV vectors (16). Despite a wide array of data revolving around MLV and other retroviruses, there remain important knowledge gaps in the field. These include quantitative assessments of the physical viral titer, the viral genome packaging efficiency, and the number of particles required for a productive infection. These issues can easily be addressed if a method could rapidly discriminate and enumerate intact viral particles.

Generally, methods that quantify physical viral titers target individual viral components (i.e., nucleic acids or proteins) within a homogenized virus sample. While being invaluable tools, these types of approaches do not provide direct information on individual virus particles. Secreted proteins, degraded virions, and extracellular vesicles (EVs) represent sources of impurities within a viral sample. EVs are an important confounding factor, given their propensity to incorporate viral proteins and genomic content in a fashion similar to that of viruses (17–19). The concern of EV contamination in this regard is further emphasized when considering the significant overlap between EV and virus biogenesis (20–22). This is especially true for retroviruses, which have evolved to assemble and egress using synthesis and secretory pathways that are largely shared with EVs (23–25). In fact, our group and others have highlighted the incorporation of MLV proteins into EVs (26, 27). It is therefore clear that in order to effectively quantify viral samples, robust discrimination between EVs and intact viral particles must be achieved.

In this report, we demonstrate that small-particle flow cytometry performed on viruses, an approach called flow virometry (FVM), can be used as a rapid, high-throughput, and effective method to quantify intact viral particles released by an infected cell. To validate this approach, we employed the use of a model retrovirus, Moloney murine leukemia virus (M-MLV). The laboratory strain used here has been modified to contain a superfolder green fluorescent protein (sfGFP) within the proline-rich region of the viral envelope glycoprotein (Env). We confirm our previously reported results that Env is an effective discrimination marker to distinguish virus from EVs based on fluorescence intensity and side-scattered (SSC) light profiles by FVM (26, 28). We then compared virion titer quantifications by FVM to those by other common analytical techniques and demonstrate the importance of discriminating intact virions from EVs and from free viral proteins. Finally, to further highlight the value of FVM, we correlated intact MLV particle counts to the infection rate and to the absolute amount of viral genomic RNA (gRNA). These parameters then enabled us to calculate the infectivity of M-MLV and its genome packaging efficiency, which to this point was unknown.

RESULTS

Generation of 10 unique virus producer cell clones.

When retroviruses infect a cell population, a large proportion of cells will harbor proviruses integrated into unique positions in their genome (29, 30). Furthermore, transduced cells may harbor one or several integrated proviral genomes. These factors may affect both host and viral gene expression and result in cell-to-cell variabilities in regard to virus gene expression, virus production efficiency, and virus infectivity. For this study, we used M-MLV that expresses sfGFP on its surface as a fusion protein with Env (MLV-sfGFP). MLV-sfGFP viral particles were used to infect NIH 3T3 cells, and infected fluorescent cells were then isolated by single-cell sorting by flow cytometry. Ten fluorescent virus-producing cell clones (clone 1 [C1] to C10) were then randomly selected, expanded, and used throughout this study. Virus-containing supernatants from these clonal cell populations were directly used in our experiments without further enrichment and constitute the viral clone stocks.

Assessment of the MLV particle titers by ELISAs.

Enzyme-linked immunosorbent assays (ELISAs) were conducted to measure the amounts of two viral protein constituents, the viral capsid p30 protein and Env, in each of the viral clone stocks. ELISAs are routinely used in the field to normalize the virus input. Through the use of commercially available p30 and GFP ELISA kits, we found that each of the 10 MLV-sfGFP clones displayed levels of p30 and Env within less than an order of magnitude of each other (Fig. 1A). With knowledge of the molar mass of these proteins, a concentration of molecules can be calculated (Fig. 1B). Guided by literature values for the numbers of p30 and Env molecules in a single MLV virion (Table 1), ranges of inferred viral particle concentrations within these stocks were estimated (Fig. 1C). For our calculations, we used 2,187 p30 molecules and 300 Env molecules per virion, both of which constitute previously reported values (Table 1) (31, 32). If all p30 and Env molecules measured by the ELISA were virus associated, the average virus titer would be 2.55 × 10¹⁰ ± 0.06 × 10¹⁰ particles/ml based on p30, or 2.53 × 10¹⁰ ± 0.10 × 10¹⁰ particles/ml based on Env. However, this is likely to be an overestimate due to the presence of soluble or EV-associated viral proteins. The ratio of p30 to Env is also relatively constant among all 10 clones tested, averaging 7.4 ± 0.8, except for one outlier, at 5.4 ± 0.3 (Fig. 1D).

FIG 1.

Viral titer measurement by an ELISA. (A) The protein content in viral supernatants was assessed using a commercially available ELISA to quantify both p30 and Env (GFP) levels. (B) The concentrations in panel A were converted to protein molecules using the corresponding molecular masses (Table 1). (C) Numbers of protein molecules in panel B were converted to an inferred virus titer based on the stoichiometry outlined in Table 1. Averages of 2,187 p30 molecules and 300 Env molecules per virion are assumed based on the literature. (D) Ratio of p30 to Env protein molecules for each viral stock.

TABLE 1.

Reported p30 capsid and Env contents within an MLV particle

Viral protein	Molecular mass (kDa)a	No. of proteins/virus particle (reference)
Capsid (p30)	30.61	1,100–1,800 (66)
		1,860–2,514 (31)
Envelope glycoprotein (Env)	73.30	60–210 (67)
		240–360 (32)
		∼300 (68)

^aMolecular mass without posttranslational modifications.

Viral RNA genome content.

As an alternative to protein content, measuring viral genomic content is another typical approach to estimate quantities of virus. MLV has been well defined to specifically package 0 or 2 copies of its viral RNA genome (gRNA), with no specific affinity for monomeric viral gRNA (33–36). Here, we sought to determine the relative abundance of viral gRNA in each viral clone stock by amplifying two different regions of the viral genome: the packaging signal (Psi) and Env. Viral gRNA was isolated from an aliquot of each stock using a commercially available silica-based extraction column with carrier RNA to increase recovery. To account for possible nucleic acid loss during the purification process, a set amount of a commercial RNA standard was spiked into the lysis buffer before each extraction. The purified nucleic acid was then reverse transcribed into a cDNA that was then used as a template for probe-based droplet digital PCR (ddPCR) to determine the absolute copy number of viral gRNAs.

While the total isolation efficiency was monitored using the internally spiked RNA standard, the efficiency of each individual step was also monitored. Free RNA spiked into culture supernatants was rapidly degraded, while monitoring the isolation efficiency was made possible by spiking the RNA standard directly into viral lysis buffer where RNase activity had been inactivated (Fig. 2A). In our system, we determined the RNA extraction efficiency to be approximately 50%, while RT maintained an efficiency of over 70%. Tracking the spiked RNA standard allowed us to determine the total procedural efficiency of viral cDNA synthesis for each independent sample, which averaged around 33% (Fig. 2A). Through this determination of the total isolation efficiency, the absolute number of viral gRNA copies was calculated for both Psi and Env amplicons (Fig. 2B). Similar to the ELISA results, the level of viral gRNA in each stock was quite consistent among each of the 10 clones tested, with on average 2.19 × 10⁹ ± 0.22 × 10⁹ genomes/ml.

FIG 2.

Measurement of viral gRNA and of the infectious titer. (A) The efficiency of RNA column extraction and reverse transcription was monitored using an RNA standard. (B) Using the efficiencies from panel A, the absolute number of viral genomes was determined using a ddPCR strategy targeting the packaging signal (Psi) or Env sequences. (C) Transducing units (TU) were measured for each viral stock clone. Virus was analyzed from 10 independent cell clones producing MLV-sfGFP. Each data point is representative of results from two independent experiments. P values were calculated by paired Student’s t test. **, P ≤ 0.01.

MLV infectious titer.

Not all viruses released from a transduced cell are infectious. Some viruses are devoid of genomes, and others may harbor inactivating mutations or may be defective. Furthermore, given that the supernatant of infected cells contains soluble viral proteins, degraded virions, virus-like particles (VLPs), and EVs containing viral components, determining the true viral titer can be very challenging. As such, the infectious titer is often used for practical and functional reasons. Here, NIH 3T3 cells were infected with each of the viral clones and monitored for GFP fluorescence at 24 h postinfection by flow cytometry. By knowing the number of target cells infected and the volume of the supernatant used, it is simple to determine the overall concentration of infectious units or transducing units (TU) in the sample. Our data show that the 10 clones exhibited an average infectious titer of 2.18 × 10⁷ ± 0.20 × 10⁷ TU/ml (Fig. 2C).

Characterization of sfGFP, Env-GFP, and MLV-sfGFP producer cells and their supernatants.

The accuracy of the various viral quantification methods is in large part limited by the ability to enumerate intact viral particles. Importantly, given the “bulk” nature of these analytical methods, it is extremely challenging to discern the influence of soluble or EV-associated viral constituents. As a result, little robust information is currently available, for instance, about the genome packaging efficiency of M-MLV or the actual infectivity of released particles. This is because such calculations require an exact enumeration of intact viral particles in a sample (i.e., the physical titer). However, we have shown in a recent publication that Env expressed on the surface of M-MLV produced by cell transfection is a robust discrimination marker that can distinguish intact MLVs from EVs by FVM (28).

Here, we wanted to determine if this discrimination between EVs and viruses holds true if Env is expressed by a retroviral vector in transduced cells, in contrast to episomal plasmid-expressed Env, as we previously reported (28). To investigate this issue, we cloned sfGFP or the Env-GFP fusion protein into a retroviral vector to stably transduce NIH 3T3 cells. Infected fluorescent cells were then sorted to purity by flow cytometry. For sfGFP, we established cell populations expressing low and high levels of fluorescence by cell sorting. Cells labeled “Control” in the figures represent untransduced NIH 3T3 cells.

Transduced cells and their supernatants were analyzed by SDS-PAGE and blotted for GFP, β-tubulin, and p30. As expected, transduced cells showed the appropriate sizes for GFP (∼27 kDa), unprocessed Env-GFP (gPr85-GFP) (∼115 kDa), and the furin-cleaved surface (SU) moiety of Env-GFP that contains the fluorescent reporter protein (gp70-GFP) (∼100 kDa) (Fig. 3A). The Env-GFP-transduced cells also displayed an additional band of an unknown nature (∼60 kDa). The expression of p30 was not detectable in either the cell lysates or the supernatants, as would be expected (data not shown). Importantly, we did not detect GFP expression by Western blotting in the supernatants of transduced sfGFP or Env-GFP cells, not even when the supernatants were concentrated by ultracentrifugation (data not shown), which is not completely unexpected (37).

FIG 3.

Comparative uptake of sfGFP and Env-GFP by MLV and EVs. (A and B) SDS-PAGE analysis of sorted NIH 3T3 cells transduced with a retroviral vector expressing sfGFP or Env-GFP (A) or 10 infected cell clones producing replicative MLV-sfGFP (B). (C) Viral supernatants from the cell clones were analyzed by SDS-PAGE and probed for p30 and Env-GFP contents. C1 to C10 depict unique chronically infected MLV-sfGFP cell clones isolated by single-cell sorting. (D) Supernatants from each of the 10 clonal producer cells were analyzed by FVM. The gated region highlights GFP-positive particles. Each dot plot is representative of data from two independent experiments. (E) Histograms illustrating the GFP fluorescence intensity profiles from each of the 10 clones analyzed in panel D. a.u., arbitrary units; MFI, mean fluorescence intensity; CV, coefficient of variation.

We then analyzed each of the 10 MLV-sfGFP-producing cell clones and their supernatants by Western blotting (Fig. 3B and C). Our results show that the cell lysates of all 10 clones also exhibit unprocessed and processed Env-GFP of the same size as that of the Env-GFP control and p30 capsid protein in proportions similar to those of their respective Env-GFP (Fig. 3B). Similarly, supernatants from the various clones all had detectable levels of Env-GFP (SU) and proportional levels of their respective p30 (Fig. 3C). Despite differences in virus protein expression, egress viruses from all 10 cell clones nearly completely resolved from instrument noise and nonfluorescent particles in the sample by FVM (Fig. 3D), and all viruses exhibited a highly homogenous fluorescence intensity (Fig. 3E).

Discrimination of MLV-sfGFP from EVs by FVM.

The supernatants from MLV-sfGFP-infected cell clones and from the sfGFP- and Env-GFP-transduced cells were directedly analyzed by FVM without further enrichment, as previously described (28, 38). To minimize particle coincidence during sample acquisition, MLV-sfGFP samples were diluted 4-fold more than samples from sfGFP- and Env-GFP-transduced cells (Fig. 4A). Low-frequency and dispersed fluorescence events were detected in supernatants from the sfGFP cells, which correlated with their respective levels of cytosolic sfGFP expression (Fig. 4A and B). These events likely correspond to EVs that incorporate cytosolic sfGFP as cargo. Env-GFP-expressing cells also display low-frequency fluorescence events with heterogenous SSC patterns. These are likely EVs with Env-GFP on their surface (26). In contrast, MLV-sfGFP-infected cell clones release large numbers of highly monodisperse particles that are roughly 100-fold more abundant and 2-fold brighter than Env-GFP particles released from producer cells (Fig. 4B and C).

FIG 4.

Enumeration and analysis of MLV and EV particles by FVM. (A) Supernatants produced from uninfected NIH 3T3 cells (Control), sorted sfGFP- and Env-GFP-transduced cells, and MLV-sfGFP producer clone 6 were analyzed by FVM. The gated region highlights GFP-positive particles. (B and C) The particles gated in panel A were analyzed for the number of GFP⁺ (B) particles and the GFP mean fluorescence intensity (MFI) (C). (D) The producer cells from panel A were analyzed by flow cytometry for differences in GFP mean fluorescence intensities. Dashed lines indicate background levels established with the control sample. Virus was analyzed from 10 independent clones producing MLV-sfGFP; each data point is representative of results from two independent experiments.

Viral incorporation of Env is highly consistent.

Several studies have investigated MLV assembly and Env incorporation into the viral envelope (39, 40). Since the number of virion-associated Env molecules may affect virus infectivity, evaluating Env levels on individual virions is of significant interest in virology. However, until now, Env expression has primarily been calculated based on the total protein content in virus lysates (41). Given that the fluorescence intensity of each virus is directly proportional to Env-GFP expression on its surface, it is now possible to measure relative levels of Env expression on individual viruses by FVM.

FVM analysis shows that the fluorescence intensity of the viruses is not directly related to Env expression levels in the infected cells (Fig. 4C and D). Also, MLV-sfGFP viruses acquire higher levels and much more homogeneous densities of Env on their surface than EVs (Fig. 4C). This is exemplified by two MLV-sfGFP-infected clones, each with lower or higher relative fluorescence levels than the average in the group (Fig. 4D), whereas all viral stocks display very similar Env-sfGFP incorporation levels (Fig. 4C). These data clearly highlight the tightly regulated nature of Env incorporation into these viruses.

Intact particle enumeration by FVM can be achieved by antibody staining.

FVM to characterize viruses may have limited use when virions do not exhibit intrinsic fluorescence, such as MLV-sfGFP. Here, we attempted to stain fluorescent and nonfluorescent viruses using a phycoerythrin (PE)-conjugated anti-GFP antibody to compare intact particle counts. We generated a nonfluorescent MLV-sfGFP by mutating a single amino acid within the sfGFP sequence (MLV-DsfGFP). The mutated MLV exhibits no detectable GFP fluorescence but is clearly discernible by SSC intensity measured with a 405-nm laser (405-SSC) (Fig. 5A). Using a PE–anti-GFP antibody, we stained virus stocks of MLV-sfGFP and MLV-DsfGFP prior to direct FVM analysis (Fig. 5B). Appropriate concentrations of the antibody were previously established (38). Stained MLV-sfGFP and MLV-DsfGFP viruses displayed similar profiles of scattered light and fluorescence intensity. Our results show that not only does this display the potential of FVM for characterizing surface antigens on viral particles, but it also enables an equally accurate way of enumerating nonfluorescent virions (Fig. 5C).

FIG 5.

Viral particle quantification using antibody staining. (A) FVM analysis comparing fluorescent virus (MLV-sfGFP) (clone 6) to nonfluorescent virus (MLV-DsfGFP). (B) Both viral supernatants were labeled with a fluorescent phycoerythrin (PE) antibody targeting an exposed epitope on GFP. (C) Quantification of virus stocks from three independent experiments. P values were calculated by paired Student’s t test. n.s., not statistically significant (P > 0.05).

Intact viral particle counts reveal a large abundance of free viral protein and noninfectious virions.

A valuable feature of FVM is its ability to provide an intact viral particle count, provided that EVs are discernible. Here, we aimed to correlate intact viral particle counts to viral gRNA and virus infectivity. We then used this information to compare the overall accuracy of FVM in parallel with some of the various other analytical methods used to measure virus titers.

The same 10 viral stocks were quantified in parallel by FVM and nanoparticle tracking analysis (NTA) (Fig. 6A). The data were then directly compared to those of the other bulk analytical methods used in this study. While NTA revealed a slightly higher particle count, the results were not statistically different from those of FVM (Fig. 6B). This highlights that the main particles in these viral stocks, from these specific producer cells, are indeed viral particles. The ELISA consistently overestimated the viral particle concentration, while viral genomic and TU analyses reported much lower values. With this in mind, the physical viral particle counts obtained from the FVM analysis were used to establish their relationship to viral constituents (e.g., protein and viral gRNA). With considerations given to previously reported levels of p30 and Env expected in individual MLV particles (Table 1), the ELISA data indicate that each viral stock contains a surprisingly large amount of protein, over 76%, that is not associated with intact viral particles (Fig. 6C and Table 2).

FIG 6.

MLV particle infectivity and viral gRNA packaging efficiency determined by FVM. (A) All 10 MLV-sfGFP stocks were quantified by FVM and nanoparticle tracking analysis (NTA). Absolute viral counts obtained by each method are compared. Virus counts for the ELISA reflect the total number of viruses obtained if all the p30 was associated with intact virus particles. (B) Zoomed-in view from panel A to compare results obtained by FVM, NTA, and genome pair analysis. (C) Relationship of virus-associated to free viral protein (capsid p30 or Env) determined by using the information from panel A, Fig. 1B, and Table 1. (D) Viral gRNA packaging efficiency calculated from the information in panel A and Fig. 2B. (E) Number of virions required for a productive infection using information from panel A and Fig. 2C. (F) Number of viral gRNA-containing virions required for a productive infection. Each data point is representative of results from two independent experiments. P values were calculated by paired Student’s t test. **, P ≤ 0.01; ****, P ≤ 0.0001; n.s., not statistically significant (P > 0.05).

TABLE 2.

Moloney MLV infectivity, virus-associated p30/Env, and viral gRNA packaging efficiency^a

Parameter	Avg value (%) ± SD
Virus-associated p30	23.8 ± 2.2
Virus-associated Env	23.9 ± 3.8
Viral gRNA packaging efficiencyb	18.2 ± 1.1
Overall virus infectivityc	0.37 ± 0.03
Infectivity of genome-containing virusesd	1.98 ± 0.13

^aPer individual virion, as determined by FVM, an ELISA, and ddPCR. Results are displayed as averages ± standard errors for viral stocks analyzed in duplicate from 10 unique MLV-sfGFP clones. Pairwise statistical analyses were conducted for each viral stock.

^bProportion of intact virus particles containing two copies of the viral genome. Based on the literature, we consider a retrovirus that packages either 0 or 2 copies of the viral gRNA (33–36, 53). The value represents the combined average packaging efficiency for the Psi and Env amplicons.

^cProportion of intact viral particles that generate a single TU in NIH 3T3 cells.

^dProportion of genome-containing intact viral particles that generate a single TU in NIH 3T3 cells.

Similarly, considering the necessity for the pairing of viral RNA genomes for efficient packaging into the capsid, the absolute genome count can be used to determine the packaging efficiency (33–36). We illustrate that there is between a 16 and 21% (18% average) viral gRNA packaging efficiency for M-MLV, depending on whether Psi or Env sequences are used as targets for the amplification (Fig. 6D and Table 2).

Assessment of intact particle counts by FVM can also be related to TU values to determine the number of virions required for a single productive infection. The individual virion infectivity seems surprisingly low, with roughly 200 to 400 particles being required for a single productive infection of NIH 3T3 cells (Fig. 6E and Table 2). By using the average packaging efficiency of Psi and Env, we determined that approximately 50 genome-containing viral particles are required to productively infect a single NIH 3T3 cell (Fig. 6F and Table 2).

sfGFP insertion into Env does not alter virion stability.

An interesting finding from Fig. 6C was the revelation of a large amount of soluble viral protein in infected cell supernatants. To determine if the insertion of sfGFP within Env influences virion stability, we sought to compare soluble and virus-associated p30 ratios of MLV-sfGFP to those of native M-MLV (referred to as the wild type [WT]) and MLV-V5. The latter virus has a minimally modified Env containing a 14-amino-acid V5 epitope tag insertion in place of sfGFP (238 amino acids). Additionally, an important difference to note is that unlike MLV-V5 and MLV-sfGFP producer cells, which are clones generated through single-cell sorting, the cells producing native WT MLV are a heterogenous and unsorted population of infected NIH 3T3 cells.

Due to the lack of a commercially available fluorophore-conjugated MLV Env antibody, we were unable to directly stain WT MLV. Despite this shortcoming, the enumeration of unmodified and unlabeled virus is still possible due to its distinct and uniform SSC characteristics by FVM (Fig. 7A). Staining of surface epitopes within Env was conducted for both sfGFP and V5 (Fig. 7B). We then calculated, as described above, inferred particle counts by a p30 ELISA (Fig. 7C) and intact particle counts for both SSC analysis and epitope tag staining (Fig. 7D). The WT virus displayed a >4-fold-lower p30 content than the clones (Fig. 7C), and as a result, there were fewer intact viruses in the same proportion by SSC analysis (Fig. 7D). When comparing virus-associated with free p30 in infected cell supernatants, we notice that the WT virus has more virus-associated p30 (28%) than MLV-sfGFP (17%) and MLV-V5 (18%) (Fig. 7E). Overall, all three types of viruses produce a large amount of p30 that is not associated with intact virus particles, while sfGFP and V5 insertions in MLV Env do not appear to exert a major impact on virus particle stability.

FIG 7.

Impact of Env insertions on MLV stability. (A) FVM analysis of fluorescent MLV-sfGFP and nonfluorescent WT MLV and MLV-V5 viruses. (B) Viral supernatants were labeled with an antibody targeting an exposed epitope on Env (top, anti-GFP-PE; bottom, anti-V5-PE). (C) The viral protein concentration as determined by an ELISA was converted to an inferred virus titer based on the stoichiometry outlined in Table 1, as described in the legend of Fig. 1. (D) Physical titer quantification of virus stocks based on SSC analysis and PE staining. (E) Relationship between virus-associated and free capsid p30 protein based on data from panels C and D. The data represent results of four technical replicates from one experiment. **, P ≤ 0.01; n.s., not statistically significant (P > 0.05).

DISCUSSION

Most analytical techniques in virology are incapable of differentiating intact virus form free soluble or EV-associated viral components in a sample. It is now well established that EVs associate with viral proteins and nucleic acids in a variety of systems (17–22, 26). These phenomena emphasize the importance of developing techniques to separate EVs from viral particles, either physically or analytically. One such method to achieve this, as demonstrated here, is the use of FVM. We and others have shown the abilities of FVM to detect and enumerate viral particles (28, 37, 42–48). In this work, we took the technology further to quantify intact M-MLV particles and learn more about the biology of this important virus for the field of virology.

Importantly, this work was conducted with 10 clonal expansions of single-cell-sorted, chronically infected NIH 3T3 cells to capture possible diversity between individual virus particles. Furthermore, demonstrating that MLV can be distinguished from other cell-derived particles was critical. Env-GFP-transduced cells were key to demonstrating that the relative frequency of Env accumulation on EVs is only about 1% compared to MLV. However, one must keep in mind that the diversity, abundance, and composition of EVs are likely to vary under different infection conditions or even in other cell types. Furthermore, it is clear that MLV incorporates a consistent amount of Env into each viral particle, more than twice that occurring in EVs. Cytosolic or Env-fused GFP expressed at similar or even higher levels in cells than MLV-sfGFP does not produce EVs with the fluorescence intensity seen on the virus particles. Combined with our previously reported virus microfiltration and NTA data (28), we are very confident that the population visualized by FVM is, in a very large majority, intact MLV particles.

Given that the flow cytometer used in this work is set to threshold from SSC, only particles of a given size and light scattering properties will trigger a perceived and recorded event. While there is understandably some smaller EVs and various other types of particles that are below our limit of detection, free proteins or small aggregates will also go undetected. This is an important advantage of FVM because the physical removal of these particles or even staining antibodies from samples becomes unnecessary. Additionally, MLV is confidently identified by both its surface expression of Env and its typical monodisperse SSC profile, which is granted by the highly controlled structural composition of the virus. These specific features thus greatly facilitate virus particle gating confidence in FVM applications. Even among the 10 MLV-sfGFP producer cell clones analyzed, there is a disparity in Env expression of nearly 100-fold. Yet this does not translate into a discernible variation in Env accumulation on individual virions, highlighting the tightly regulated incorporation of Env into MLV (Fig. 3E and Fig. 4D).

Through a direct comparison of viral particle counts by FVM and protein levels measured by an ELISA, we identified that there is an abundance of free viral protein in the chronically infected cell supernatants. We then addressed valid concerns that the sfGFP sequence inserted into Env could be altering the stability of MLV. By using a WT virus and a virus with a short V5 epitope tag, we confirmed that this observation is indeed a general feature of MLV (Fig. 7). Other groups have also investigated the effects of insertions in MLV Env and Gag and did not find these specific modifications to significantly alter particle structure or infectivity (23, 49, 50). However, we observed that cells chronically infected with the WT virus produced fewer viruses, and these viruses displayed slightly more virus-associated p30 in their supernatants. It should be noted that the method that we used to generate the cells chronically infected with the WT virus is slightly different from the one that we used to produce the infected clones. Given that there is no direct way to monitor infection by the WT virus in live cells due to the absence of a reporter or a commercial fluorescent Env antibody, we infected NIH 3T3 cells at a lower multiplicity of infection (MOI) than what was used with the clones and allowed the infection to blindly spread over 6 weeks. The infected cell population that resulted was used for the experiment presented in Fig. 7. As such, there is a possibility that not all cells were in fact infected, or maybe the transcription of the virus in some cells was silenced. But, more likely, fewer cells were multiply infected, which could easily explain the lower levels of virus produced overall.

Furthermore, given that the ratio of p30 to Env remains relatively constant in the supernatants of all viral clones (Fig. 1C), we posit that the vast majority of this free protein is composed of degraded virions, as opposed to secreted soluble viral proteins. MLV has a relatively short reported half-life of 4.5 h at 37°C, in contrast to >200 h at 4°C (51). To obtain sufficient amounts of virus for all the different types of analyses in this study, we harvested virus after 72 h. This may have led to the accumulation of protein from degraded virions.

By relating the genome pair analysis with intact MLV particle enumeration, we calculated a viral genome packaging efficiency of between 16 and 21%, depending on the amplicon (Psi or Env). Viral genomes released by damaged particles are likely a minor source of experimental bias given that free RNA was entirely and rapidly degraded in our system. But yet again, viral genomic RNA is coated by nucleoproteins, and ruptured capsids may still offer some level of protection against nucleases. If this were the case, the true genome packaging efficiency of MLV would be lower than what we have calculated. EVs could also potentially be packaging viral RNAs, but we clearly demonstrate that there is a minimal presence of EVs in our samples (Fig. 4 and 6). A small difference of 5% in packaging efficiency is seen when considering Psi (20%) and Env (15%) sequences. Possible explanations include that (i) the amplification efficiency of the Psi sequence is slightly better than that of Env or (ii) a minor Psi-positive (Psi⁺) and Env-negative (Env⁻) splicing variant of the viral gRNA is recruited into virions. Although copackaging of full-length viral gRNA and a subgenomic RNA has been reported, both these RNAs contain Psi and Env sequences and would be unlikely to contribute to the discrepancy in packaging efficiency (52). Regardless of the small 5% difference, both Psi and Env amplicons indicate that a large proportion of virions are lacking a viral RNA genome pair.

The low gRNA packaging efficiency that we have measured for MLV is in sharp contrast with the ∼95% packaging efficiency observed for HIV-1 gRNA (53). However, it must be noted that the study mentioned above was carried out with HIV-1 containing a heavily modified retroviral genome with several gene deletions and the insertion of 18 to 24 stem-loops that bind to a bacteriophage viral coat protein-fluorophore chimera. This system was used because it has the distinct benefit of single-copy resolution and direct visualization of genomic RNA within a virion. In perspective, our present study uses a minimally modified MLV genome that enabled us to correlate viral gRNA to intact particle counts by FVM to provide quantitation of genome packaging efficiency. As a next step, analysis of HIV-1 genome packaging efficiency by FVM would be an interesting way to confirm previously reported results.

A point of major intrigue that was revealed by this study is why does MLV have such a poor gRNA packaging efficiency and produce such a large abundance of VLPs. Our group also previously made similar observations concerning VLPs when studying the MLV glycosylated Gag accessory protein (26). One possibility is that MLV uses genome-deficient virions, degraded virions, and EVs expressing viral proteins as decoys to overwhelm humoral immune responses. Or perhaps it is the opposite. Maybe these VLPs act in a fashion similar to that of defective interfering particles (DIPs) that have been recognized for a variety of other viruses to attenuate pathogenicity by stimulating viral neutralization and antiviral immune responses for the benefit of host survival (54–58). Either of these concepts could be factors contributing to the low pathogenicity of MLV in vivo.

Measuring viral titers has always been a primary requirement, a time-consuming chore, and a challenge in virology. For practical reasons, it is the infectious titer that is usually measured, which corresponds to the infectious component of the total viral population in a sample. However, it has now become clear that the physical titer of the virus population is an equally valuable parameter. Noninfectious viral particles, especially in living animal model studies, can stimulate immune reactions to eliminate the virus (i.e., vaccines), divert them away from the infectious viruses (i.e., immunological decoys), or even attenuate virus pathogenicity for the benefit of the host and pathogen (i.e., DIPs). This is why classical titer measurements such as the MOI or the 50% tissue infectious dose (TCID₅₀), which correlates infectious titers to cytopathic effects, describe only part of the features of a given virus stock. Intact particle counts by FVM therefore offer a quick and simple way to measure the overall physical titer, which can then be more accurately correlated to infectivity and cytopathic effects.

Precise enumeration of intact viral particles has also enabled us to calculate the number of particles required for a single productive infection. This varied between 200 and 400 particles for our 10 clones, resulting in about 0.37% infectivity on average (Table 2). Albeit seemingly low, these values fall in line with the reported ranges for retroviral particles per infectious unit (P/IU) observed in various other systems (59). This being said, we should emphasize that our experiments were carried out under controlled and optimized conditions, with the use of the polycation Polybrene and spinoculation to enhance infections. Regardless of these technical efforts and because of mobility limitations of free virus in a liquid suspension that are governed by Brownian motion, there is a distinct possibility that some of the viruses in our assays still did not come into contact with a potential target cell (60). This scenario would result in an underestimation of the actual number of infectious units in our samples. Thus, measurements of infectivity will vary considerably depending on the cells used and the experimental conditions. In sum, FVM now provides a new and easy way of monitoring the infectious titer under different experimental conditions. Importantly, we also demonstrate that intrinsically fluorescent viruses are not essential for use in FVM. Native viruses can be enumerated by SSC analysis but, preferably, for more accurate results, should be stained using a fluorophore-conjugated commercial antibody (for interlaboratory consistency) targeting highly expressed epitopes on the surface of the virus.

Further development of FVM and much-needed improvements to the fluorescence detection sensitivity of flow cytometers will likely enable the scientific community to address new and unanswered questions about individual virus heterogeneity inside a virus population, in the same way as it is routinely done for cells using standard flow cytometry. As illustrated in this work, FVM can provide rapid and high-throughput information about physical and infectious virus titers, packaging efficiency, the presence of defective or interfering particles, and surface antigenic uptake. The minimal sample processing and time required to obtain results may even make this technique suitable for clinical diagnostics, especially given its higher sensitivity than other common diagnostic methods (37).

MATERIALS AND METHODS

Cells.

Human embryonic kidney epithelium cells (293T) and mouse embryonic fibroblasts (NIH 3T3) were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Wisent) that was supplemented with 10% fetal bovine serum (FBS) (Corning), 100 U/ml penicillin, and 100 μg/ml streptomycin (Wisent) and propagated in an incubator at 37°C with 5% CO₂.

Plasmids and viruses.

The plasmid encoding M-MLV-sfGFP was generated through restriction-free cloning to replace enhanced green fluorescent protein (eGFP) in the M-MLV vector, which has been described by our group previously (26, 61) (forward [FWD] primer 5′-TTCAGTCACCAAACCACCCAGTGGGAGCAAGGGCGAGGAACTGTTCACC-3′ and reverse [REV] primer 5′-CGGTACGTACGCACCGGTGGACTTGTACAGCTTGTACAGCTCGTCCATGCCGTGGG-3′). This plasmid originated from the pMOV-eGFP vector, which has been described previously (62, 63). To create MLV-DsfGFP, an R96C mutation was created within sfGFP (FWD primer 5′-GCTACGTGCAGGAATGCACCATCAGCTTCAAGGACGACGGC-3′ and REV primer 5′-CTTGAAGCTGATGGTGCATTCCTGCACGTAGCCCTCGGGC-3′). For M-MLV-V5, which was created in this study, restriction-free cloning was conducted to insert the V5 tag within the proline-rich region of MLV Env (FWD primer 5′-GTGGGGGTATACGCGTGGGCAAGCCGATTCCCAATCCTCTGCTTGGCCTCGATTCAACTACGCGTGGGGCGATCGCGCCCGGG-3′ and REV primer 5′-CCGAAGAGCAAAATCATTAGGAGTACAATGAGGGGTCCC-3′).

Generation of MLV-sfGFP producer clones and virus production.

The principal virus used throughout this study is M-MLV with sfGFP inserted into the proline-rich region of Env, creating a fusion protein (Env-sfGFP) with sfGFP being exposed on the surface of the virus (26, 28, 38, 61, 62). We call this virus MLV-sfGFP throughout our study for simplicity. Virus was produced from chronically infected NIH 3T3 cells as previously described (26, 28, 61). However, in order to monitor possible heterogeneity between individual virus particles, these chronically infected cells were single-cell sorted based on sfGFP expression and expanded. For this investigation, viruses released from 10 different stable monoclonal expansions of chronically infected cells were randomly selected and scrutinized.

To produce virus, chronically infected NIH 3T3 cell clones were seeded at a density of 2 × 10⁵ cells per well in a 6-well plate. Exactly 2 ml of medium was provided for an incubation period of 72 h. Supernatants were collected, cleared of cellular debris, and passed through a 450-nm cartridge filter for downstream analysis. Infected producer cell clones were analyzed by flow cytometry or SDS-PAGE to measure relative protein levels. For SDS-PAGE analysis, the following antibodies were used: R187 anti-p30 (ATCC CRL-1912), anti-EGFP (JL-8; Clontech), horseradish peroxidase (HRP)-conjugated anti-rat IgG (Sigma), HRP-conjugated anti-mouse IgG (Cell Signaling), and HRP-conjugated anti-β-tubulin (Abcam). For transducing unit (TU) calculations, cells were infected at several multiplicities of infection (MOIs) using spinoculation and Polybrene as previously described (26). Flow cytometry analysis was conducted at 24 h postinfection to mitigate the impact of viral spreading for a more accurate measurement of TU per milliliter in the initial stock. TUs were calculated using the following equation: transducing units/ml = number of infected cells/volume of viral supernatant (ml).

Generation of stably transduced sfGFP- and Env-GFP-expressing cells.

To generate NIH 3T3 cells stably expressing sfGFP and Env-GFP, we used the M-MLV-based pMXs-Puro retroviral system (Cell BioLabs) to transduce the cells. The coding sequence of sfGFP or Env-GFP was inserted into the multiple-cloning site of pMX-Puro. Virus particles were produced by cotransfection of the packaging plasmids pOGP (Gag-Pol) and pMDG (vesicular stomatitis virus G protein [VSV-G]), as described previously (64). Transfection of 293T cells was conducted with polyethylenimine (PEI) as described previously (65). Briefly, 5 × 10⁵ cells were seeded 24 h prior to transfection. A total of 2 μg of plasmid DNA was transfected at a ratio of 1:0.7:0.3 (pMXs-Puro/pOGP/pMDG). Cells were incubated to produce virus for 72 h before the supernatant was collected and cleared of cellular debris by centrifugation and microfiltration. NIH 3T3 cells were infected as described above. After 72 h of expansion, the bulk of the fluorescent sfGFP- and Env-GFP-expressing infected cells were sorted by flow cytometry. These cells were then expanded and used as controls for this work.

Gag and Env quantifications by ELISAs.

Commercially available ELISA kits were purchased from Cell BioLabs to analyze the levels of p30 (MuLV core antigen ELISA kit, catalog number VPK-156) and Env-GFP (GFP ELISA kit, catalog number AKR-121). The manufacturer’s instructions were closely followed. These kits provided a reliable concentration of protein content by mass for p30 and GFP. The molar mass of p30 (Table 1) was used to determine moles of p30 in each sample. For theoretical viral particle titer calculations, an average of 2,187 p30 molecules were assumed to constitute a single MLV virion (31). The molar mass of the GFP standard within the GFP ELISA kit (catalog number AKR-121), which contains a His tag, was considered to determine moles of GFP and, therefore, Env-sfGFP within each sample. Similarly, to infer the viral particle titer, an average of 300 Env-sfGFP molecules were assumed to be expressed on the surface of a single virion (32).

Viral genome quantification.

The QIAamp viral RNA minikit (Qiagen) was used to isolate RNA from viral supernatants according to the manufacturer’s instructions, with the slight modification that an RNA standard (VetMAX) was added to the lysis buffer prior to initial lysis. RNA was eluted in nuclease-free water and reverse transcribed using QScript with RNase H activity. RNase H was kept to avoid the synthesis of multiple copies of cDNA per RNA molecule.

The cDNA was diluted and analyzed by droplet digital PCR (ddPCR) of the packaging signal (Psi) and Env-sfGFP sequence targets within each sample (Psi FWD primer 5′-TGGGGGCTCGTCCGGGAT-3′, Psi REV primer 5′-CCGGGTGTTCCGAACTCGTCAGTTC-3′, Psi probe 5′-ACCCCTGCCCAGGGACCACCGACCCACC-3′, Env-sfGFP FWD primer 5′-CCGACAAGCAGAAGAACGGC-3′, Env-sfGFP REV primer 5′-CTTGTACAGCTCGTCCATGCCG-3′, and Env-sfGFP probe 5′-CCCCGTGCTGCTGCCCRACAACCACTACC-3′). Additionally, the RNA standard was monitored in each column to evaluate the extraction efficiency for each individual sample. A primer-probe mixture was provided and used according to the manufacturer’s instructions (VetMAX). Data were analyzed using QuantaSoft and extrapolated based on the determined extraction efficiency and dilutions used for the assay.

Nanoparticle tracking analysis.

Nanoparticle tracking analysis (NTA) was carried out using the ZetaView PMX110 multiple-parameter particle tracking analyzer (Particle Metrix) in size mode using ZetaView software version 8.02.28. Samples were diluted in phosphate-buffered saline (PBS) to ∼10⁷ particles/ml. The system was calibrated using 105- and 500-nm polystyrene beads, and videos were then recorded and analyzed at all 11 camera positions, with a 2-s video length, a camera frame rate of 30 fps, a shutter of 70, and a temperature of 21°C.

Flow virometry analysis.

A detailed description of the instrument settings and the MLV particles used in this study by FVM was reported recently (28, 38). FVM was conducted using a Beckman Coulter CytoFLEX S instrument using 405-nm SSC-H as a threshold parameter (threshold typically at 1,500 arbitrary units). All virus samples were passed through 0.45-μm cartridge filters, unless otherwise stated, and analyzed without additional concentration or enrichment. Antibody staining was carried out with an aliquot of virus using PE-conjugated anti-GFP (catalog number FM264G; BioLegend) at a final concentration of 0.2 μg/ml; this was then diluted for analysis. Optimal antibody concentrations for staining were previously established (28, 38). PE-conjugated polyclonal anti-V5 (catalog number ab72480; Abcam) was used at a concentration of 0.8 μg/ml. Nonviral supernatants were analyzed at a 1-in-250 dilution, while viral supernatants were diluted 1 in 1,000, using 100-nm-filtered PBS. Samples were acquired for 1 min on the slow-flow setting for a total of 10 μl of each diluted sample. Concentrations were then extrapolated based on dilution factors.