Background
Since the discovery of the human herpesvirus 6A (HHV-6A) and HHV-6B, a characteristic cytopathic effect induced in cells has been described. For instance, infected cells are characterized by the occurrence of enlarged cells, first described as “balloon-like” syncytia [
1]. Presence of balloon-like cells has often been used as an indicator of HHV-6 infection of primary cells cultures, ex-vivo infection of primary cells, and in-vitro infection of various cell lines [
1‐
4]. However, visual inspection of cell cultures is subjective and it can be difficult to establish clear criteria to decide if a cell is infected. Quantitative or semi-quantitative methods widely used to determine HHV-6 infection include immunofluorescence microscopy (IFA), flow cytometry, and molecular methods. In IFA, antibodies specific for HHV-6 proteins allow the identification of infected cells; this method has been used since early days of HHV-6 research [
2,
3]. HHV-6 antisera or monoclonal antibodies to various virus proteins have been used, including antibodies recognizing glycoproteins B (gB), gH, gQ, and protein U27 among others [
5‐
8]. A shortcoming of this method is that it requires a lengthy sample processing with multiple incubation and washing steps, as well as the scanning of randomized fields in an adequate microscope to obtain a representative sample adequate for quantitative analysis. In addition, IFA-based methods are subjected to person-to-person variability in the read-out, often requiring readings by two or more experienced investigators. An alternative is the analysis of cells stained with the fluorescent-labeled antibodies by flow cytometry [
9‐
11]. This method circumvents the subjectivity of the microscopy methods, although still requires multiple incubations and washing steps and adequate instrumentation. Molecular methods to detect HHV-6 infection include detection of expression of viral genes by reverse-transcription PCR [
11‐
13] and measurement of the amount of viral genome in infected cells or cell culture supernatants by quantitative PCR. Quantitative-PCR (qPCR)-based methods are very sensitive and are widely used [
14‐
18]. However, they require specialized equipment, reagents, and sample processing, and do not provide information about the actual number or frequency of infected cells or the amount of infectious virus produced, and so represent only an indirect measure of the infection.
The search of a fast and simple method to objectively monitor HHV-6 infection in-vitro on a daily basis led us to explore the use of an imaging-based automated cell counter able to measure the size of cells in a cell suspension. Measurements of the average size of cells (usually as cell diameter) using automated cell counters have been used for some time to monitor infections, in particular in cell culture systems set up for recombinant protein production using insect cells. For example, infection of Sf-9 cells with Baculovirus expression vectors results in changes in average size during the progress of infection, and monitoring the average size of the cells has been useful to estimate the degree of infection [
19], to predict the yield of recombinant protein produced [
20], and also to estimate virus titers [
21]. The work presented in this paper consists of a series of analyses that explore the feasibility of using a cell counter to routinely monitor HHV-6B infections in SupT1 cells. Results suggest that this method is fairly reliable to estimate infections when high doses of virus and/or long times are used; when compared to other methods like IFA, flow cytometry and qPCR, it is faster and simpler.
Discussion
This paper reports the development and evaluation of an assay useful for day-to-day follow-up of HHV-6B infections in-vitro. It was demonstrated that the cytopathic effect observed in HHV-6B infections (occurrence of enlarged cells) can be measured using an imaging-based automated cell counter, which provides systematic and objective measurements of the size of cells in the cultures. As the occurrence of enlarged cells was associated to productive infections in this particular system, this method provides an objective, fair, and fast way to monitor infections. The cell counter measures the diameters of cells identified in 8 different fields of a slide, providing the count of live and dead cells and their sizes, as well as processed data like cell density, average cell size and cell viability. The dynamic range for the average size of cultures was relatively narrow, with the settings of the counter selected for counting individual cells of up to 50 μm; conveniently, this parameter can be easily adjusted which broaden the applicability to other systems.
Results obtained from testing the method in the T lymphoblast cell line SupT1, susceptible to infection by HHV-6B strain Z29, indicate that it can indeed be used to differentiate non-infected and infected cells and cultures. However, analysis of the size of individual cells showed some overlap between distributions of non-infected and infected populations. It is possible that, as a result of a non-synchronic infection of cells, the fraction of cells that were not infected, or which were not yet displaying cytopathic effect, might account for some degree of this overlap; this implies at least two populations coexisting in the same culture. In addition, non-infected cells undergo slight variations in size, likely related to the cell-cycle stage, also contributing to the overlap, although to a lesser degree. The overlap is greatly reduced when the average size of cells is used for the analysis rather than the distribution individual cell sizes. At the times studied (4 and 7 dpi) enlarged cells are likely predominant, driving the average to higher values and reducing the overlap. Indeed, there were minimal doses of virus and minimal times post-infection before which the infected cultures could not be assessed by the method. It is possible that under conditions of low virus input, non-infected and relatively fast-dividing cells outnumbered the few infected cells initially generated, resulting few enlarged cells and neglectable changes in overall average size of the cells. Low amounts of virus produced under these conditions would not generate enough infected cells to induce a measurable shift in the population’ size distribution. Likewise, at earlier time-points, the increase in average size and/or the number of enlarged cells might not be big enough to generate a significant shift in the population’ size distribution. As infections progress, the proportion of infected cells and the actual size of these cells would have increased enough to significantly shift the average size of the cells.
A ROC statistical analysis was used to define a practical cutoff value for deciding whether there was infection or not. The performance of the method was better when the average size of cells rather than individual cells’ size distributions were used. The selection of the cutoff value was done by maximizing the product specificity and sensitivity and considering how much loss of specificity (how many false positives) or loss in sensitivity (how many false negatives) can be tolerated. Note that these depend on the particular aim of an experiment. The consensus cutoff values selected for the analysis involving average size of cells were associated with 4% or less false negatives and false positives, which was considered appropriate. In the case of monitoring production of virus stocks for example, not reaching 100% sensitivity might be acceptable, as infected cultures that cannot be clearly differentiated from the non-infected cultures are likely to represent low-level infections, and resulting virus stocks likely to contain low amount of virus. In the case of measuring the infectivity of a virus stock, both the highest specificity and sensitivity are desirable, because wrong assignment of positive or negative wells would affect the determination. The performance of the method when individual size of cells was used was not as good as for average size of the population, which might be a reflection of the nature of system (inherent size variation in non-infected cells and non-synchronic infections) resulting in a mix of cells at different states. The cutoff values selected when using individual size of cells were associated with higher percentages of false negatives (23-34%), which indicates that cells below the cutoff value might be wrongly assigned as not infected. This is consistent with the results of the linear regression analysis, in which only 51% the data on antigen expression in cells was explained by the percentage of cells above the cutoff.
The size-based method described here was particularly useful when characterizing medium and high level infections; unfortunately, low-level infections did not generate enough enlarged cells such that the shift in the size distribution could not be differentiated by this method. For instance, in procedures such as generation of infectious virus stocks or production of infected cells (for preparation of total cell lysates for protein analyses, gene expression assays, etc.), the size-based method can be used to follow the infections. In these cases, inability of detect low-level infections does not reduce the practicality of the method, given that low-level infections do not result in good stocks able to further propagate the virus nor in highly infected cell preparations, where any change in gene or protein expression in infected cells would not be masked by the expression in the predominant non-infected cells.
The analyses and cutoff values described in this work were obtained using the system of SupT1 cells and HHV-6B strain Z29 virus. Other cell lines and virus, such as SupT1.CIITA [
23], MOLT-3 and Jurkat E6 infected with HHV-6B strain Z29 and HSB-2 cells infected with HHV-6A strain GS, also showed measurable occurrence of enlarged cells, which should make them suitable for analysis by this method. However, different combinations of cell types and virus strains have specific characteristics. Basal (non-infected) size distributions were different for the different cell lines, as were infection susceptibilities. Thus, each combination of cell and virus need to be individually standardized. In setting up the assays, some factors discussed above should be considered to successfully differentiate non-infected and infected cultures, in particular the time after infection when data is collected, the initial amount of virus used for infection and confirmation of a productive infection. HHV-6 is known to infect many cells type with infection progressing to different endpoints in different cells and with different degrees of viral production [
26]. It is possible that certain cell types might be infected by HHV-6B without exhibiting cytopathic effects or size increases upon infection. The method described here would not be appropriate for such situations.
Another interesting observation was the capacity of the size-based method to measure the effect of various methods of virus inactivation. It is known that heat and UV-inactivation have an effect in the infectivity of HHV-6, interfering with processes that require the establishment of a productive infection, as shown by greatly reduced antigen expression and cytopathic effect [
32,
33]. However, these inactivation methods do not eliminate the ability of the virus to induce other processes, for example cytokine and chemokine secretion [
34,
35], suggesting that stable components of the virion are preserved and can still be taken up by cells and exert effects. Under the conditions employed in our experiments, there was a significant reduction of the amount of infectious virus after inactivation; although not complete, it was enough to result in low-level infections, which did not induce changes in the average size. Bearing this in mind, the size-base method potentially could be used as a first screening tool to test antiviral methods.
In our work, we used a particular microscope-based imaging cell counting system but, in principle, the size-based method would not be restricted to imaging instruments and other counting methods should work as well. Flow-based cell counters, like Coulter counters and flow cytometers, also collect size-based data, and potentially could also be used for this type of analysis. In a flow cytometer for example, the forward scatter data (FSA) could be used for quantitative size-based assessment of infections. A preliminary assessment of this method showed a shift in the population of infected cells, as observed in FSA histograms (Additional file
4). However, we observed that it was important to carefully and consistently select the voltages to ensure that most of the population of infected cells is included in the analysis and to allow comparisons among measurements performed at different days. A second important point to consider in the use of a flow-based cytometer in evaluation of infection-induced size increases is the increased susceptibility of giant cells to fragmentation, which has been noted in a previous study [
36]. In that report, researchers used a Coulter counter, and found that, enlarged cells that they could identify by visual inspection using a microscope could not be seen in the size histogram obtained by the Coulter counter, leading them to suggest that the big refractile infected cells had been fragmented during the process [
36]. In this regard, a non-flow imaging cell counter such as the one use in our work reported here, has the advantage that the sample manipulation is minimal, favoring the preservation of the bigger cells to a certain degree.
In comparisons of measurements of average size with measurements of the amount of virus DNA in supernatant or the expression of viral antigens in cells, similar trends were observed, suggesting that all methods depict the same phenomenon in a similar way. This corroborates that the average cell size-based method is a valid method to follow HHV-6B infections; it has the advantage of being simple, fast, and providing an objective measurable result, all of which can be important when rapid decision is required.
Of the methods that we evaluated, the qPCR-based method was the most sensitive, allowing detection of 8 ± 2 copies of DNA in a single reaction, with a dynamic range between ~104 to ~1011 copies of DNA per mL. This allowed detection of small increases in the amount of virus DNA produced when lower amounts of virus were used to infect the cells and/or when earlier times were analyzed. In fact, qPCR method could detect amounts of virus that did not induce measurable shifts in the average size of the cell cultures, consistent with low-level infections. However, this method can be used only for populations of cells, not for evaluation of infection on a cell-by-cell basis.
The flow cytometry method using fluorescence detection of viral antigens has the advantage of providing the frequency of cells expressing that viral antigen within the whole population, which is an important metric when studying viral infections in-vitro. Infected cells could be identified by flow cytometry when relatively low doses of virus and early time points were analyzed, even after accounting for non-specific binding of antibodies to non-infected cells, and by the increased auto-fluorescence of infected cells. A drawback of this method is that infected cells might be mistakenly excluded from the analysis as a result of several factors including the acquisition voltages and gating strategies excluding the largest cells, uptake of viability dyes by dying infected cells, or increased fragility of larger cells. These factors could be especially important when collecting data at longer time points after infection when the viability of the cells has decreased and when there is a high proportion of larger cells. We seem to have observed this effect in Fig.
2d and e, where infections performed with the highest concentration of virus exhibited reduced gB expression at 7 dpi relative to 4 dpi, whereas the other markers of infection (size, cell death, and viral DNA copies) all increase during this time period. This might be the result of the largest and/or most fragile cells being excluded from the analysis.
Finally, besides monitoring infections, the method was used to determine the relative infectivity of virus stocks. In the end-point assay, the average size method performed well in differentiating among dilutions. However, the fact that low-level infections do not induce significant changes in the average size could lead to underestimation of viral titers, especially compared to more sensitive methods like qPCR and flow cytometry. In spite of this, in terms of infectivity, the size-based TCID50 provided better equivalence among virus stocks than copies of viral DNA.