Endothelial Cells Support Persistent Gammaherpesvirus 68 Infection

A variety of human diseases are associated with gammaherpesviruses, including neoplasms of lymphocytes (e.g. Burkitt's lymphoma) and endothelial cells (e.g. Kaposi's sarcoma). Gammaherpesvirus infections usually result in either a productive lytic infection, characterized by expression of all viral genes and rapid cell lysis, or latent infection, characterized by limited viral gene expression and no cell lysis. Here, we report characterization of endothelial cell infection with murine gammaherpesvirus 68 (γHV68), a virus phylogenetically related and biologically similar to the human gammaherpesviruses. Endothelial cells supported γHV68 replication in vitro, but were unique in that a significant proportion of the cells escaped lysis, proliferated, and remained viable in culture for an extended time. Upon infection, endothelial cells became non-adherent and altered in size, complexity, and cell-surface protein expression. These cells were uniformly infected and expressed the lytic transcription program based on detection of abundant viral gene transcripts, GFP fluorescence from the viral genome, and viral surface protein expression. Additionally, endothelial cells continued to produce new infectious virions as late as 30 days post-infection. The outcome of this long-term infection was promoted by the γHV68 v-cyclin, because in the absence of the v-cyclin, viability was significantly reduced following infection. Importantly, infected primary endothelial cells also demonstrated increased viability relative to infected primary fibroblasts, and this increased viability was dependent on the v-cyclin. Finally, we provide evidence for infection of endothelial cells in vivo in immune-deficient mice. The extended viability and virus production of infected endothelial cells indicated that endothelial cells provided a source of prolonged virus production and identify a cell-type specific adaptation of gammaherpesvirus replication. While infected endothelial cells would likely be cleared in a healthy individual, persistently infected endothelial cells could provide a source of continued virus replication in immune-compromised individuals, a context in which gammaherpesvirus-associated pathology frequently occurs.


Introduction
Endothelial cells create a physical barrier on the luminal surface of blood and lymphatic vessels. This barrier must be traversed by blood-borne pathogens and immune cells trafficking between tissues and the bloodstream. Many herpesviruses require systemic spread for persistence within a host, and therefore must cross such an endothelial cell barrier. To date, herpesviruses have been implicated as potential initiators of arterial injury, endothelial dysfunction, and local inflammation, possibly contributing to the pathogenesis of atherosclerosis [1][2][3][4]. Human cytomegalovirus (HCMV), a betaherpesvirus, infects endothelial cells in vivo. Studies have shown that infected endothelial cells play a role in HCMV dissemination and pathogenesis [5]. Endothelial cells exhibit regional specialization in gene expression and morphology depending on the local physiologic demands of their respective organs and tissues [6]. In light of this diversity, it is not surprising that endothelial cells from different tissues differ in their susceptibility to HCMV infection [7]. Kaposi's sarcoma-associated herpesvirus (KSHV), a human gammaherpesvirus, also infects endothelial cells in vivo. More importantly, KSHV is the causative agent of the endothelial cell neoplasm, Kaposi's sarcoma (KS). The murine gammaherpesvirus, cHV68, has been detected in aortic endothelium after infection of apoE deficient mice, as well as on the luminal surface of explanted aortas infected in vitro [8]. While a relationship between gammaherpesviruses and endothelial cells has been noted, the role of endothelial cells in chronic gammaherpesvirus infection and pathogenesis is ill-defined.
Gammaherpesviruses are a lymphotropic family of viruses associated with a broad spectrum of malignancies and lymphoproliferative diseases. These oncogenic viruses persist for the life of the host by establishing and maintaining a latent infection. Because gammaherpesviruses are extremely host-specific, studying the human viruses Epstein-Barr virus (EBV) and KSHV in nonhuman systems does not mimic natural infection. cHV68 is a natural pathogen of wild murid rodents and therefore provides a valuable small animal model of gammaherpesvirus infection [9][10][11][12][13]. This virus has important biological and genetic similarities to the human gammaherpesviruses, and results in a variety of pathologies in defined mouse models. Primary cHV68 infection is characterized by virus replication in lung epithelial cells, and the establishment of latency in B cells, dendritic cells, and macro-phages [12][13][14][15][16]. Persistent cHV68 infection also occurs in lung epithelial cells [17]. The role of endothelial cells in cHV68 infection has not been explored to date.
At the cellular level, gammaherpesvirus infection encompasses two broadly defined outcomes: productive, lytic replication and non-productive, latent infection. During lytic replication, viral DNA is amplified and host cell machinery is utilized for the production of viral progeny. Viral genes are actively transcribed and translated, contributing to new virus production. Ultimately, the cell is lysed and infectious virus released. In light of the fact that this process occurs quickly (24 to 48 hours in vitro) our understanding of virus-host interactions during primary infection is quite limited. During latency, viral DNA is not amplified, but instead is maintained as a nuclear episome in latently infected cells, and little viral gene transcription occurs. Latently infected cells remain intact and do not produce new infectious virus. Furthermore, it is thought that these long-lived latent cells may serve as the major mechanism by which gammaherpesviruses promote life-long infection of their hosts.
Here we characterize cHV68 infection of endothelial cells. To date, in vitro infection of most cells with cHV68 supports significant viral replication and results in complete cell lysis (data not shown, van Dyk). We have infected both primary endothelial cells and endothelial cell lines and demonstrated that they produce comparable amounts of virus as fibroblasts. However, whereas fibroblasts were mostly lysed by 36-48 hours post-infection, a significant percentage of infected endothelial cells remained intact. Analysis of endothelial cell lines revealed that these intact cells 1) were actively infected and undergoing the lytic transcriptional program, 2) continued to proliferate for a prolonged time after infection, 3) were altered in morphology and cell-surface protein expression, compared to uninfected endothelial cells, and 4) continued to release infectious virions as far as 30 days postinfection. In the absence of the cHV68 viral cyclin, endothelial cell viability was significantly decreased after infection, indicating that an active viral process was responsible for promoting the outcome of infection in endothelial cells. Of major significance, we also provided evidence of endothelial cell infection in primary cells and in vivo in immune-deficient mice. These data demonstrated that endothelial cells supported persistent, productive gammaherpesvirus infection, an outcome not previously reported for cHV68 or other gammaherpesviruses. Together, these data indicated the potential of the endothelial cell as a critical cell type in cHV68 pathogenesis, and open a new avenue of research into herpesvirus manipulation of endothelial cell biology.

Endothelial cells supported cHV68 replication
Endothelial cells are heterogeneous in nature, express unique surface antigens, and differ in their susceptibility and response to infection by various pathogens [6]. In light of this diversity, we investigated endothelial cell lines from various anatomic locations, and from BALB/c and C57BL/6 mice, for their ability to support cHV68 replication. We analyzed the outcome of cHV68 infection in endothelial cell lines from lung (CD3), brain capillary (MB114), and lymph node (SVEC) by single-step and multi-step growth assays. Single step growth proceeded with kinetics similar to previously published infection of the fibroblast cell line, NIH 3T12 cells [18] in each of the three endothelial cell lines analyzed (Fig. 1A, top panel). Though equivalent numbers of cells were plated per well, the differences amongst the endothelial cell lines in titer at time zero were not surprising given that cell lines grow at varying rates and to varying densities. A comparison of viral titers at 36 hours revealed that the three endothelial cells lines produced titers comparable to those achieved in 3T3 fibroblast cells. Additionally, multi-step growth in MB114 and CD3 endothelial cells was similar to growth in 3T3 fibroblasts (Fig. 1A, bottom  panel). These data demonstrated that cHV68 is capable of replication in endothelial cells infected in vitro.
Using a GFP-labeled virus, we next sought to determine if the percentage of infected endothelial cells was comparable to percentage of infected fibroblasts. Following infection of MB114 endothelial cells and 3T12 fibroblasts, cells were analyzed for expression of GFP from the cHV68 genome by flow cytometry. At 24 and 48 hours post-infection the majority of endothelial cells and fibroblasts expressed GFP (Fig. 1B). These data indicated that, like fibroblasts, endothelial cells were uniformly infected and supported the cHV68 lytic transcription program.
A population of endothelial cells remained intact following cHV68 infection and uniformly expressed lytic antigens Uninfected endothelial cells and fibroblasts grew as adherent monolayers ( Fig. 2A, top panel). However, in infected endothelial cell cultures, we observed a population of intact, phase-bright, and non-adherent cells as early as 96 hours post-infection, whereas only cellular debris remained in the infected fibroblast cultures (data not shown). At six days post-infection, we collected non-adherent cells and cellular remains by centrifugation. After washing, we resuspended this material in complete media for culture and analysis. The cells harvested from the infected endothelial cultures appeared as individual, intact cells suspended in culture, whereas the material harvested from the infected fibroblast cultures appeared as clumps of cellular debris ( Fig. 2A, bottom panel). Prior to centrifugation, we measured cell viability by trypan blue exclusion. Very few infected fibroblasts remained as viable, nonadherent cells at six days post-infection (Fig. 2B). However, a significant proportion of endothelial cells remained as viable, nonadherent cells at six days post-infection. When MB114 endothelial cells were treated with a non-toxic dose of phosphonoacetic acid

Author Summary
Various human diseases are associated with gammaherpesvirus infections, including neoplasms of lymphocytes (e.g. Burkitt's lymphoma) and endothelial cells (e.g. Kaposi's sarcoma). Gammaherpesvirus infection of cells usually results in either productive infection that is characterized by new virus production and rapid destruction of the host cell, or latent infection that is characterized by long-term carriage of viral DNA in intact cells that do not produce virus. Here, we characterize endothelial cell infection using a small animal model of gammaherpesvirus infection and disease. While infection of endothelial cells resulted in virus production as most cells do, infection of this cell type was unique in that cells remained intact and continued to proliferate. These intact, infected endothelial cells were significantly altered in appearance and gene expression compared to uninfected cells, changes predicted to impact endothelial cell growth and function. Endothelial cells were unique in their ability to support this type of persistent infection. These data demonstrate an additional mechanism, beyond latency, by which gammaherpesviruses may achieve long-term propagation, particularly in conditions of immune suppression. Our results suggest persistently infected cells as a therapeutic target for prevention/ treatment of chronic disease, and may provide a mouse model for future testing.
(PAA), an inhibitor of viral DNA replication and late gene synthesis, most cells remained adherent, and there were significantly fewer detached, viable cells (Fig. 2B). PAA alone had no effect on the viability, adherence, or phenotype of uninfected MB114 cells in culture (data not shown). These data revealed that a population of endothelial cells became non-adherent and remained viable at six days post-infection, whereas fibroblasts were destroyed by cHV68 infection. Additionally, this outcome of infection in endothelial cells was influenced by late viral gene expression, as most cells remained adherent in the presence of PAA.
Next we determined the status of virus replication in the viable and non-adherent cells. Cells collected at six days post-infection were analyzed by flow cytometry for expression of the cHV68 glycoprotein, gp150, which is expressed during the lytic transcription program on the surface of infected cells [19]. Cell viability was also determined using propidium iodide (PI), a cell impermeant dye that is excluded from intact cell membranes. In these experiments we also included the S11E cell line, a mouse B cell lymphoma line which harbors latent cHV68, and therefore is negative for gp150 surface expression [20]. The majority of MB114 endothelial cells and 3T12 fibroblasts were positive for surface gp150 expression (Fig. 2C). While 3T12 fibroblasts were mostly PI positive, MB114 endothelial cells excluded PI. These data indicated that endothelial cells which remained viable at six days post-infection were undergoing the lytic transcription program and were not latent or uninfected.
At 12 days post-infection we determined the viability of cells collected and cultured at six days post-infection. PI staining of S11 cells, after six days in culture, revealed that 51.7% (62.5 SEM) of cells were PI negative and thus viable. PI staining of post-infection fibroblasts, treated in parallel with infected endothelial cells, revealed that only 14.5% (61.5 SEM) of the cells harvested remained intact. In contrast, PI staining of post-infection endothelial cells revealed that 92.5% (60.3 SEM) of the cells harvested were intact and viable (Fig. 2D). Analysis of the viral replication program in MB114 endothelial cells at 12 days postinfection revealed that the majority of cells continued to express gp150 (Fig. 3C). Therefore, infection of endothelial cells with cHV68 resulted in a population of cells that escaped lysis and remained intact as far as 12 days post-infection, while undergoing the lytic transcription program.

Intact endothelial cells were productively infected and underwent a continued lytic program
The observed viability of endothelial cells after cHV68 infection could be the result of escape from infection or latent infection of these cells. To determine the infection status of the intact endothelial cells, we performed RT-PCR analysis of viral gene transcripts. The cHV68 M2 gene transcript is synthesized during both lytic and latent infection, as are the polymerase III (pol III) transcripts encoded at the left end of the viral genome [21,22]. Latently infected S11 cells contained both pol III-1 and M2 transcripts (Fig. 3A). We detected these transcripts in 3T3 fibroblasts at 24 and 36 hours post-infection, as well as from the 3T3 cellular debris harvested at six days post-infection, while no cellular b-actin transcript was detected from infected 3T3 cells at 36 hours post-infection. No RNA could be detected from infected 3T3 debris at 12 days post-infection, whereas we recovered RNA from infected MB114 cells as far as 12 days post-infection. Pol III-1 and M2 transcripts were detected in infected MB114 cells at 24 and 36 hours post-infection and from the intact, cultured cells at six and 12 days post-infection. We detected b-actin RNA in infected MB114 cells at 24 and 36 hours post-infection, but not at six and 12 days post-infection. However, we detected another cellular mRNA transcript, cyclophilin A, at six days post-infection in 3T3 fibroblasts and MB114 cells (Fig. S2), and comparable detection of the cellular 18S rRNA transcript was observed in all conditions. Gradual loss of the cellular b-actin transcript in viable cells was not surprising given that b-actin has been documented to vary significantly in the setting of virus infection [23], and selective degradation of certain mRNAs frequently occurs during virus infection [24][25][26][27][28]. Analysis of viral gene transcripts indicated that those endothelial cells which remained intact after cHV68 infection contained viral gene transcripts as far as 12 days postinfection.
To determine whether the intact, infected endothelial cells were undergoing active viral replication, we analyzed infected cells for the expression of early and late lytic replication-associated M3 and gB. As predicted, 3T3 fibroblasts, which support lytic replication, expressed both M3 and gB (Fig. 3B). In contrast, S11 B cells, which contain latent cHV68, did not express either M3 or gB. Intact, infected MB114 endothelial cells not only expressed M3 and gB early (36 hours), but also at six and 12 days post-infection. A faster migrating gB-specific band was also detected at six days post-infection, and may indicate differential glycosylation or degradation. Additionally, infected MB114 endothelial cells were positive for surface protein expression of gp150 at both 12 and 24 days post-infection (Fig. 3C). Our analysis of lytic protein expression indicated that endothelial cells which remained intact after cHV68 infection expressed early and late lytic proteins, indicating that not only did cHV68 infected endothelial cells still contain virus, these cells also expressed viral proteins indicative of active viral replication, and not latency.
To further test the status of viral replication in endothelial cells, we infected MB114 endothelial cells with GFP-cHV68 and collected the intact, non-adherent cells at six days post-infection. GFP expression is driven by the CMV immediate early promoter and has been demonstrated to be expressed only during lytic infection [29]. At the time of harvest, these cells uniformly expressed GFP (Fig. 3D histogram) and continued to do so at 20 days post-infection, (Fig. 3D micrographs). Therefore, the nonadherent endothelial cells which remained intact after cHV68 infection did not escape infection and were not latently infected.

Intact, infected endothelial cells released new infectious virus particles
After establishing that the intact endothelial cells harvested at six days post-infection were indeed infected, we determined whether these cells produced mature virions. Transmission electron microscopy (TEM) of these cells revealed virions at various stages of maturation throughout the nucleus and cytoplasm of 100% of the 20 cells imaged at 12 days post- The culture conditions we established for maintaining the nonadherent cells included centrifugation and resuspension in fresh media every six days (Fig. S1). To test the contribution of nonadherent cells to virus production, aliquots of the supernatants were collected for measurement of cell-free virus. A culture of latently infected S11 cells did not yield any detectable cell-free virus after six days in culture by plaque assay (Fig. 4B). In contrast, after six days in culture (12 days post-infection), cell-free supernatant from infected MB114 cells yielded significant viral titer, and this titer was significantly higher than that of supernatant taken from infected 3T3 cultures. Therefore, endothelial cells continued to release new infectious virions for at least 12 days post-infection. Infected endothelial cell supernatant contained 100-fold more virus than that of a lysed, infected, fibroblast culture.

Intact endothelial cells undergo transient proliferation in culture after cHV68 infection
After the initial observation that cHV68 infection of several endothelial cell lines from different anatomic locations resulted in a population of cells that escaped lysis, we focused our analysis on this outcome in MB114 endothelial cells. Beginning at the time of collection (day six post-infection), we counted intact, trypan blue excluding cells every three days. Infected MB114 cells remained viable for at least 30 days post-infection (Fig. 5A).
During this course of infection, the number of intact cells in the culture increased from 7.08610 5 (60.16 SEM) to 8.51610 6 (60.28 SEM) during the first six days of culture, a ten-fold increase. A corresponding increase in the percentage of membrane viable cells also occurred during this time. Subsequent to day 12 post-infection, the total number of intact cells began to decrease, as did the overall viability of the culture (Fig. 5A). The observed increase in both intact cell number and percent viability during the first six days of culture, concurrent with positive gp150 staining ( Fig. 2C and Fig. 3C), indicated that endothelial cells were intact and proliferating in culture while maintaining the lytic transcription program.
To further examine the proliferation observed during the first six days of culture, we stained MB114 cells with carboxyfluoroscein (CFSE) prior to infection. CFSE is a fluorescent molecule used to measure cell proliferation, in that each time a cell divides, the two daughter cells contain half the CFSE of the parent cell. S11 cells demonstrated a drop in CFSE signal intensity after six days in culture, consistent with cell division. At the time of harvest (day six post-infection), MB114 cells also demonstrated a drop in CFSE signal intensity consistent with multiple cell divisions (Fig. 5B). After six more days in culture (12 days post-infection) CFSE signal intensity dropped further, consistent with continued cellular proliferation. Thus, endothelial cells which remained intact after cHV68 underwent multiple rounds of proliferation.
Note that by 30 days post-infection, however, the viability of the culture was quite low in comparison to earlier time points (Fig. 5A). To further test the fate of the surviving endothelial cells, we examined viability of infected cells by staining with annexin V and PI (Fig. 5C). MB114 cells were 92.4% (60.6 SEM) viable (annexin V negative, PI negative) at two days post-infection and 72.5% (614.5 SEM) viable at four days post-infection. At six days postinfection all MB114 cells were non-adherent and 31.0% (610.74 SEM) of these non-adherent cells were viable. In contrast to MB114 cells, dying cells dominated infected 3T12 cultures at six days post-infection These data indicate that many MB114 cells died early during infection, and in agreement with our proliferation data ( Fig. 5A and 5B) the population of viable cells increased to 86.4% (60.6 SEM) of the culture by 12 days postinfection. However, by 30 days post-infection dying cells finally dominated the infected culture. These data support that the subset of cells surviving and proliferating at six days post-infection were unique their extended survival, but did not remain viable indefinitely. Of significance, during the course of this infection most of the cells expressed the lytic glycoprotein gp150 on their surface at both early and late time points (Fig. 2C and Fig. 3C). Moreover, the percentage of cells expressing gp150 did not change markedly with time ( Fig. 2C and Fig. 3C). These data indicate that the vast majority of viable, infected endothelial cells were undergoing active viral replication, and that these cells were neither latently infected nor had escaped infection.
Intact endothelial cells were altered in size, shape, and surface marker expression after cHV68 infection To investigate the morphologic differences between uninfected endothelial cells and the endothelial cells harvested at six days post-infection, we quantified cell size and internal granularity by flow cytometric determination of forward versus side scatter. Uninfected MB114 cells exhibited a broad range of light scatter, indicative of a cell population diverse in size and internal granularity. In contrast, infected cells were very uniform in forward and side scatter (Fig. 6A). PAA treatment immediately following MB114 infection resulted in significantly fewer cells becoming non-adherent (Fig. 2B). Like the uninfected MB114 cells, MB114 cells treated with PAA following infection exhibited a broad range of forward and side scatter (Fig. 6A). Thus, cHV68 infection of endothelial cells resulted in a uniform population of cells with altered morphology, and this outcome was dependent on productive viral infection and late viral gene expression.
To further investigate the differences between uninfected endothelial cells and the intact cells harvested six days after infection, we compared the expression of cell-surface proteins on these two cell populations by flow cytometry. Because the cells harvested at six days post-infection persisted in culture as nonadherent cells, we chose to investigate surface expression of the adhesion markers intercellular adhesion molecule-1 (ICAM-1, CD54) and vascular cell adhesion molecule-1 (VCAM-1, CD106). Additionally, we analyzed cells for surface expression of Thy1, a cell surface protein which is upregulated on the surface of activated endothelial cells and functions in cell-cell interactions [30][31][32]. Uninfected MB114 cells expressed ICAM-1, VCAM-1, and Thy1 (Fig. 6B). In contrast, infected endothelial cells expressed neither ICAM-1 nor VCAM-1, but did express Thy1. Based on these data, cHV68 infected endothelial cells downregulate cell surface expression of two adhesion molecules, while maintaining surface expression of an activation marker.
In the absence of the viral cyclin, endothelial cell viability after cHV68 infection is reduced We demonstrated that cHV68 infection of endothelial cells resulted in a population of intact cells which persisted in culture as non-adherent cells and continued to produce new virus. Our analysis of endothelial cell infection in the presence of PAA indicated that this outcome is influenced by viral DNA replication and/or late gene synthesis ( Fig. 2B and Fig. 4A). To begin dissecting the mechanism of persistent cHV68 in endothelial cells infected in vitro, we tested the role of the cHV68 viral cyclin (v-cyclin) in this system. We measured expression of the cHV68 vcyclin in infected MB114 endothelial cells and infected fibroblasts by RT-PCR and western analysis. Notably, the cHV68 v-cyclin gene transcript and protein were detectable in infected endothelial cells and fibroblasts ( Fig. 7A and Fig. S3A).
To determine the role of the v-cyclin in endothelial cell infection, we infected MB114 cells with either wildtype cHV68 or a v-cyclin deficient cHV68 (v-cyclin.STOP cHV68) [18]. Cells were harvested at six days post-infection and cultured as described previously. As per previously published reports in fibroblasts, we determined that the v-cyclin was dispensable for cHV68 growth in endothelial cell lines by multi-step growth assays (data not shown Because viability of MB114 cells after infection was impaired in the absence of the v-cyclin, we next determined the effect of this viral gene on persistent viral replication in viable endothelial cells. Cell free supernatants from infected MB114 cells were titered by plaque assay (Fig. 7C). Infections with both wildtype and vcyclin.STOP virus resulted in persistent viral production up to 30 days post-infection. Total viral titers from v-cyclin.STOP infections were significantly less than wildtype infections at 12 and 24 days, however, an equivalent amount of virus was produced per cell as in wildtype cHV68 infections. Cell surface protein expression of ICAM-1, VCAM-1, and Thy1 was the same following infection with either wildtype or v-cyclin.STOP virus ( Fig. 6B and Fig. S3B), thus the v-cyclin was not required for the surface phenotype of infected endothelial cells and the surface phenotype of infected endothelial cells did not predict survival. Therefore, upon infection, the primary function of the v-cyclin was to promote the viability of endothelial cells.
Primary endothelial cells were infected in vivo, and ex vivo demonstrated prolonged viability while supporting cHV68 growth Given that immortalized endothelial cell lines are inherently different from endothelial cells in vivo, we next determined the outcome of cHV68 infection in primary endothelial cells. Primary endothelial cells were isolated from C57/BL6 mouse lungs and characterized as per previously published methods (Fig. S4).
First, we determined if primary lung endothelial cells could support growth of wildtype cHV68 by multi-step growth assay (Fig. 8A). Because of the apparent role for the v-cyclin in infection of endothelial cell lines, we also analyzed growth of v-cyclin.STOP virus in primary endothelial cells. Infection proceeded with kinetics comparable to previously published infection of NIH 3T12 fibroblasts [18]. Growth curves did not differ between wildtype and v-cyclin.STOP virus. These data revealed that primary endothelial cells supported growth of cHV68, irrespective of the vcyclin.
Second, we analyzed the percent of cells that remained viable after cHV68 infection at a low MOI. Immediately following removal of the viral inoculum (t = 0), mouse embryonic fibroblasts (MEFs) and primary lung endothelial cells did not differ significantly in viability (p.0.05) (Fig. 8B). However, at 48 hours post-infection with wildtype cHV68, primary lung endothelial cells were significantly more viable than MEFs (p,0.001), and remained significantly more viable at 96 (p,0.05) and 120 hours post-infection (p,0.001). Notably, endothelial cells infected with the v-cyclin deficient cHV68 had reduced viability at 48, 96, and 120 hours post-infection. Given that cells were infected at a very low MOI, it was not surprising that a proportion of MEFs (28.5%611.1 SEM) were viable at 96 hours post-infection. However, unlike endothelial cells, MEFs were mostly lysed by 120 hours post-infection with wildtype virus. These data demonstrate that primary endothelial cells surpass MEFs in viability following cHV68 infection, and that this outcome is promoted by the v-cyclin.
Lastly, we examined lung endothelial cells following acute cHV68 infection in vivo. Our in vitro studies revealed that viable, infected endothelial cells were not latently infected, but uniformly supported a lytic viral program (Fig. 2C, Fig. 3C and 3D). Because persistent viral infection and lytic viral antigen expression is likely to be cleared by an intact immune response, we analyzed endothelial cell infection in lung tissues of immune-deficient CD8-alpha null mice. At six days post-intranasal infection, lung cells enriched for those bearing the endothelial cell marker CD31 were analyzed alongside the remaining CD31 depleted lung cells. Given the possibility that a small percentage of non-endothelial cells can transiently express CD31 (i.e. macrophages and neutrophils), we depleted the CD31 positive cell population of these potential contaminating cells and analyzed resultant cell populations by flow cytometry (Fig. S5B). PCR analysis with single copy sensitivity for cHV68 gene50 (Rta) was performed to determine the frequency of viral genome positive cells in the CD31-enriched and remaining CD31-depleted lung cell populations (Fig. 8C). Data from this analysis revealed that approximately 1/102 CD31-enriched cells were viral genome positive, whereas 1/525 remaining CD31depleted lung cells contained viral DNA. Thus, these data support that a surprisingly large proportion of lung endothelial cells were viral genome positive following in vivo infection.
To further demonstrate that these viral genome positive cells are actively infected, and to exclude the possibility of endocytic uptake of virus or abortive infection, we performed RT-PCR analysis on endothelial enriched and depleted lung cells (Fig. 8D). The transcript for the endothelial cell marker CD31 was robustly detected in the endothelial enriched lung cells from each of the four mice analyzed, whereas a very low level of this transcript was detected from the CD31 depleted lung cells of only one out of the four mice analyzed. In combination with flow cytometric analysis (Fig. S5B), these data support that our enrichment strategy was effective in isolating CD31+ cells from total lung cells. Additionally, the viral pol III-1 transcript was detected from both endothelial cell enriched and depleted lung cells of cHV68 infected mice, and was absent from lungs of mock infected mice. Detection of viral gene transcripts from lung cells indicates virus infection of these cells. Notably, within each infected mouse, pol III-1 detection was comparable between endothelial enriched and CD31-depleted lung cell populations, suggesting that detection of this viral transcript in the endothelial cell population was not due a few contaminating infected non-endothelial cells. Taken together, these data demonstrate cHV68 infection of endothelial cells in vivo.

Discussion
Herpesviruses have been implicated as potential initiators of a variety of endothelial pathologies [1][2][3][4]. Given the intimate interactions observed between herpesviruses and endothelial cells, and the systemic spread of cHV68 during infection, we characterized the outcome of endothelial cell infection with cHV68, a small animal model for the human gammaherpesviruses. While cHV68 replicated comparably in endothelial cells and fibroblasts up to 36 hours post-infection, infected cultures of endothelial cells had a high percentage of viable, non-adherent cells that remained following infection. Of significance, these viable, non-adherent cells had not simply escaped infection, but instead were actively infected, as determined by the presence of multiple markers of the lytic replication program. While the absolute number of viable, infected endothelial cells varied among different cell lines and primary cells, the prolonged viability of endothelial cells in comparison to fibroblasts was remarkably consistent in endothelial cell lines from distinct anatomic locations, as well as in primary endothelial cells. Furthermore, optimal survival of both endothelial cell lines and primary endothelial cells was dependent on the presence of the cHV68 v-cyclin, indicating that this outcome was a process actively promoted by virus infection.
This conserved outcome of endothelial cell infection is particularly striking given that endothelial cells display phenotypic heterogeneity in structure and function depending on anatomic location [33][34][35]. Based on the persistent infection observed in diverse endothelial cell lines, and heightened viability of primary endothelial cells following infection, we propose that cHV68 may have evolved machinery to specifically promote persistent infection in endothelial cells.
Endothelial cells serve as a natural site of infection and possible viral reservoir of HCMV [36][37][38][39][40], suggesting a role for HCMVinfected endothelial cells in viral spread and persistence. Additionally, recent reports implicate circulating endothelial progenitor cells as potential reservoirs of KSHV and possible precursors of KS spindle cells [41,42]. However, the specific mechanisms by which infected endothelial cells contribute to the pathogenesis of these human viruses remains unclear. Murine cHV68 pathogenesis involves dissemination from the lung to lymph nodes, spleen, and peritoneum [12,15]. In light of this systemic spread, cHV68 likely encounters an endothelial cell barrier.
The human gammaherpesvirus KSHV causes a serious endothelial cell malignancy, KS, which predominantly occurs in immunecompromised individuals (e.g. AIDS patients). KS tumors are comprised of distinctive spindle cells of endothelial origin and a variable inflammatory infiltrate [43][44][45][46]. KSHV is detected primarily in the endothelial component of the lesion, and though most of these cells harbor latent KSHV, a subset of them enter the lytic cycle [47,48]. While there is precedent for a mixed infection type within KS tumors (i.e. both lytic and latent infection), the history of the lytically infected cells remains elusive, though recent reports point to circulating endothelial progenitor cells [41,42]. In vitro, endothelial cell infection with KSHV is predominantly latent [43,[49][50][51]. However, when these cells are transferred into mice, they show evidence of lytic gene expression and virus production [52].
Because viral DNA replication and/or late gene synthesis were important for endothelial cell outcome of cHV68 infection (Fig. 2B and Fig. 6A), our findings indicate that an active viral process occurred within the cells to yield these changes. Consistent with this hypothesis, we identified that the cHV68 v-cyclin is required for optimal endothelial cell viability after cHV68 infection in vitro. The cHV68 v-cyclin promotes cell cycle progression in primary lymphocytes and can function as an oncogene in transgenic mice [18]. While the v-cyclin is critical for reactivation from latency, to date the v-cyclin has been dispensable in all assays of lytic replication in vitro [53,54]. Given the lytic nature of endothelial cell infection, the contribution of the v-cyclin to optimal endothelial cell viability after cHV68 infection indicates a role for the v-cyclin outside of its requirement in reactivation from latency. Additionally, the apparent role of the v-cyclin in this prolonged infection of both endothelial cell lines and primary lung endothelial cells may explain the slight decrease in lung titers that resulted in mice infected with low doses of v-cyclin.STOP cHV68 and other vcyclin mutant viruses compared to wildtype cHV68 [55]. Although the precise mechanism by which the v-cyclin promotes endothelial cell viability is unknown at this time, it is possible that the v-cyclin provides growth cues that allow for anchorageindependent growth of endothelial cells after cHV68 infection.
While our initial analysis focused on the role of the v-cyclin, it is very likely that additional viral genes facilitate persistent endothelial cell infection. Lead candidates include the anti-apoptotic viral bcl-2 gene M11 and the viral GPCR (ORF 74), whose homologs in KSHV have been implicated in endothelial cell survival and transformation [56][57][58][59]. Additional candidates for optimal endothelial cell infection are the ribonucleotide reductase homologs, ORF 60 and 61 of cHV68. Although the role of these genes in cHV68 infection of endothelial cells is currently untested, the MCMV ribonucleotide reductase homolog is required for in vivo replication and pathogenesis of this endothelial cell-tropic virus [60].
One of the most noticeable alterations of the persistently infected endothelial cells described here is the extent of change in their cellular morphology and properties. Though we did not test the oncogenic potential of these cells, endothelial cells achieved anchorage-independent growth, a property frequently associated with oncogenic transformation. These cells also underwent significant changes in protein expression on the cell surface, with down-regulation of the cellular adhesion proteins ICAM-1 (CD54) and VCAM-1 (CD106). While there was decreased expression of ICAM-1 and VCAM-1, changes in cell surface expression were not global, since Thy1 expression remained positive on these cells (Fig. 6B). It is worth noting that infection of endothelial cells with KSHV results in down-regulation of MHC class I, PE-CAM (CD31), and ICAM-1 (CD54), but not LFA-3 (CD58) or Fas (CD95), and the viral genes K3 and K5 have been demonstrated to regulate this outcome [61][62][63][64][65][66]. The contribution of the mK3 gene of cHV68 to persistent endothelial cell infection remains untested. Down-regulation of certain surface markers but not others during viral infection indicates a specific phenomenon, rather than global down-regulation, and provides further evidence that an active process was responsible for the observed endothelial cell outcome of cHV68 infection.
While cHV68 may have evolved mechanisms that promote persistent infection of endothelial cells (e.g. the v-cyclin), it is also possible that endothelial cells have a cellular program that limits the cellular lysis typically observed during productive infection. This putative cellular adaptation may be particularly important in limiting destruction of blood vessel integrity during viral infection. A recent report of host cell response to cHV68 in three different cell types (fibroblasts, endothelial precursor cells and macrophages) identified 148 genes whose expression was altered in endothelial precursor cells, but not macrophages or fibroblasts [67]. Taken together with the unique endothelial cell outcome of cHV68 infection reported here, these data make a compelling argument for cell type specific responses to productive cHV68 infection.
We have provided evidence for cHV68 infection of endothelial cells in vivo at early times post-infection. Interestingly, infected cells (including endothelial cells) were not abundant in acutely infected lung tissue, as little to no GFP was detected by flow cytometric analysis of unfractionated lung cells after infection with a GFPmarked virus (data not shown), but required sensitive PCR methods for detection. Our in vitro analysis demonstrated a change in surface phenotype of infected endothelial cells which could correspond to substrate detachment and release into circulation. This is an intriguing idea in light of published reports of circulating endothelial cells and progenitor endothelial cells in virus infection. Our in vitro studies implicated endothelial cells as a persistent source of virus production, however, the extent to which endothelial cells contribute to cHV68 persistence in vivo, and to what degree it might be influenced by host immune status, are important issues that remains to be addressed. While it is unlikely that an intact immune response would permit such long-term expression of viral antigens in vivo, we hypothesize that persistently infected endothelial cells could provide a significant source of continued virus replication in immune-compromised individuals (i.e. AIDS patients), a context in which gammaherpesvirusassociated pathology frequently occurs.
Given the unusual properties of these viable and infected endothelial cells (e.g. anchorage-independent growth and altered cell surface protein expression), it will be important to critically address the potential role of endothelial cells as a reservoir for infection in vivo in both immune-competent and immune-deficient individuals.
In conclusion, our data provide evidence for prolonged gammaherpesvirus infection in endothelial cells. This outcome appears to be the result of a specific interaction between cHV68 and endothelial cells, as it is promoted by a viral gene (the cHV68 v-cyclin) and, to date, is unique to endothelial cells. These data further refine the concept of gammaherpesvirus infection and demonstrate that a gammaherpesvirus can undergo robust productive replication in the context of prolonged host cell viability. Identification of intermediate outcomes of gammaherpesvirus infection, such as the one we have described here, has major implications for our understanding of the nature of gammaherpesvirus infection as it relates to specific cell types. Such a course of infection provides an additional mechanism, beyond latency, by which gammaherpesviruses can achieve longterm propagation.
All infections were carried out at a multiplicity of infection (MOI) of 5 plaque forming units (PFU) per cell. Inoculum was removed after one hour of infection at 37uC, cell monolayers rinsed three times with sterile phosphate buffered saline (PBS), and complete media added. Intact and non-adherent cells were collected at six days post-infection, at which time cells and media were collected, counted to determine post-infection viability by trypan blue exclusion, and then centrifuged for 10 minutes at 2086g. Cell pellets were washed twice in sterile PBS and then resuspended in complete RPMI at a concentration of 5610 5 viable cells/mL for continued culture. Cells were counted every three days of culture, and adjusted to a concentration of 5610 5 viable cells/mL. Every six days of culture, cells were counted and centrifuged for 10 minutes at 2086g to remove cell-free cHV68 and cellular debris. Pellets were washed twice in sterile PBS and resuspended in RPMI at a concentration of 5610 5 viable cells/ mL. Culture scheme is depicted in supplement (Fig. S1).

Viral titer by plaque assay
To measure virus replication, infected samples were analyzed by plaque assay at various times post-infection. To measure cell-free virus titer from infected cultures, supernatants were collected every six days of culture and analyzed by plaque assay. Samples were thawed, serially diluted, and plated onto NIH 3T12 cells in 12 well plates in triplicate. Infection was performed at 37uC for one hour. Cells were overlaid with a 1:1 mix of DMEM and carboxymethylcellulose plus fungizone (final concentration 250 ng/mL). Plates were incubated for seven days at 37uC. On day seven, carboxymethylcellulose was removed, and plates were stained with 0.35% methylene blue in 70% methanol and rotated for 15-20 minutes, before rinsing with water and counting on a light box. All titers were determined in parallel with a laboratory standard.

Flow cytometry
For propidium iodine (PI) viability studies, cells were washed twice in PBS (five minutes, 10006g), resuspended in a 0.5 mg/mL PI solution, and incubated for 15 minutes. Following incubation, cells were centrifuged (five minutes, 10006g) and washed in a solution of PBS, 2% fetal calf serum, and 0.1% NaN 3 (buffer A). Cells were fixed in 1% paraformaldehyde, and analyzed by flow cytometry. For two parameter viability analysis, cells were washed in 1X annexin V binding buffer (BD Bioscience, San Jose, CA), resuspended in 0.5 mg/mL PI and 5 mL annexin V-FITC (BD Bioscience), incubated for 15 minutes, washed in binding buffer, fixed in 1% paraformaldehyde, and analyzed by flow cytometry.
For carboxyfluorescein (CFSE) proliferation studies, MB114 cells and Sll cells were washed twice in PBS (five minutes, 10006g) and resuspended in a solution of PBS and 2% fetal calf serum (buffer B) at a concentration of 1610 6 cells per mL. An equal volume of 4 mM CFSE in buffer B was added to the cell suspension (2 mM final concentration) and pipet mixed. After three minutes the labeling reaction was quenched with an equal volume of fetal calf serum for 30 seconds, and buffer B was added for a total volume of 50 mL. Cells were centrifuged (five minutes, 10006g), resuspended in buffer B, and aliquots collected for day 0 analysis. Remaining, labeled cells were resuspended in complete media and cultured. Labeled MB114 cells were infected at an MOI = 5 PFU/ cell. Infected cells were harvested as described at day six postinfection and analyzed by flow cytometry at six and 12 days post-infection. Stained and unstained S11 cells were analyzed at the same time as a positive control.
For forward versus side scatter analysis, MB114 cells harvested at six days post-infection were analyzed. Uninfected MB114 cells were detached from flasks with 0.5 mM EDTA and analyzed in parallel with infected cells. In certain experiments, to determine the effect of viral DNA replication and/or late gene synthesis on forward and side scatter properties, MB114 cells were treated with 200 mg/mL phosphonoacetic acid (PAA) after one hour of infection. At six days post-infection, non-adherent cells were harvested and counted by trypan blue exclusion to determine the percent of cells infected that were viable and non-adherent at time of harvest. Cells which remained adherent to the flask were then detached with 0.5 mM EDTA, combined with the harvested nonadherent cells, and analyzed by flow cytometry for forward and side scatter properties. Effective block of late gene synthesis was confirmed by plaque assay titer of treated cells. Viral titer was reduced 96.2% after 24 hours and 99.9% after six days as compared to untreated cells. 10 micron control beads were run in parallel with each experiment (Beckman Coulter, Fullerton, CA).

Western analysis
Cells were collected and boiled in Laemmli buffer for 10 minutes. Total protein concentration of each cell extract was determined by Lowry assay (DC protein assay kit, Bio-Rad, Hercules, CA). 10-20 mg of total protein per extract was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Proteins were electrotransferred (ThermoFisher Scientific, Portsmouth, NH) to PVDF membranes (Millipore, Billerica, MA) and blocked in PBS with 0.05% Tween-20 and 5% nonfat milk for one hour at room temperature. Western blots for cHV68 protein expression were incubated with 10 mg/mL monoclonal mouse anti-cHV68 gB [76] (kind gift of Dr. Phillip Stevenson), monoclonal mouse anti-cHV68 M3 at 1:50, polyclonal rabbit anti-cHV68 v-cyclin at 1:2000 (kind gifts of Dr. Herbert Virgin), and monoclonal antibody to mouse b-actin at 1:1000 (Sigma Chemical, St. Louis, MO). Blots were washed for 45 minutes in PBS with 0.05% Tween-20, then incubated with donkey antimouse or donkey anti-rabbit antibodies at 1:6000 for one hour. Blots were then washed in PBS containing 0.1% Tween-20 for 45 minutes. Proteins were visualized using an ECL Plus Western blotting detection kit (Amersham Pharmacia Biotech).

Transmission electron microscopy
Cells were pelleted (five minutes, 2086g), resuspended in 1 mL PBS, and pelleted again (five minutes, 2086g). Supernatant was removed and 2.5% glutaraldahyde solution (adjusted to pH 7.4 using HCl and to 400 mOsm using CaCl 2 ) added. Samples were processed and imaged by Dr. Gary Mierau, The Children's Hospital Department of Pathology/Laboratory Services, Aurora, CO. Briefly, samples were post-fixed in 2% cacodylate buffered osmium tetroxide (pH 7.4), dehydrated in a graded series of alcohols, and embedded in epoxy resin. Sections, approximately 80 nm in thickness, were stained with uranyl acetate and lead citrate prior to examination at 60 kV with a Zeiss EM-10 transmission electron microscope (Carl Zeiss Inc, Thornwood, NY).
Limiting dilution nested PCR detection of cHV68 genome-positive cells The frequency of lung cells from CD8-alpha knock-out mice containing cHV68 genome was determined using a previously described nested PCR assay (LD-PCR) with single-copy sensitivity to detect gene 50 of cHV68 [77]. Briefly, freshly isolated cells were counted, resupsended in isotonic solution, and then diluted in 10 4 uninfected NIH 3T12 cells prior to serial dilution plating. Plated cells were lysed overnight in proteinase K, followed by two rounds of nested PCR. Reactions were performed on 12 replicates per dilution per sample and products resolved on a 2% agarose gel and identified by ethidium bromide staining. PCR sensitivity was quantitated using 10, 1, or 0.5 copies of a gene 50 containing plasmid (pBamH I N) diluted in 10 4 uninfected NIH 3T12 cells.

Statistical analysis
Data were analyzed using GraphPad Prism software (GraphPad Software, San Diego, CA). Data were analyzed using the paired Student's t test to determine statistical significance. Limiting dilution data were subjected to nonlinear regression analysis, and the frequency of genome-positive cells was calculated using the Poisson distribution to assume that the cell number at which 63.2% of events were detected corresponded to the occurrence of a single event.

Supporting Information
Text S1 Endothelial cell specific markers included CD31 and CD54. CD80 and CD86 were included as non-endothelial cell specific markers, though a previous report has identified low level CD80 expression on primary murine lung endothelial cells [71]. Fluorescence was determined relative to unstained cells (grey). Cell morphology and surface expression was similar to previously characterized primary endothelial cells [71]. Found at: doi:10.1371/journal.ppat.1000152.s006 (0.89 MB TIF) Figure S5 Isolation and characterization of murine lung endothelial cells infected in vivo. (A) Primary murine lung endothelial cells were isolated for analysis of infection in vivo as described in Materials and Methods. Lung cells were stained for the endothelial cell marker CD31, as well as with a cocktail of PE labeleled antibodies specific to the following cell types (which includes potential contaminating infected cells): macrophages, granulocytes, CD8 + T lymphocytes, CD4 + T lymphocytes, and B lymphocytes. CD31 positive cells were enriched from total lung cells and then finally depleted of cells stained with the PE cocktail. (B) Flow cytometric analysis of lung cell separation following in vivo infection was performed on total lung cells and PE depleted/ CD31+ enriched cells. Left and middle panels show representative data (1610 6 cells per stain) from an infected mouse (Fig. 8D). Gates were set on unstained cells, and data are representative of the four mice analyzed in Fig. 8D. Splenocytes (right panel) were included as a staining control. Found at: doi:10.1371/journal.ppat.1000152.s007 (0.20 MB TIF)