Kaposin-B Enhances the PROX1 mRNA Stability during Lymphatic Reprogramming of Vascular Endothelial Cells by Kaposi's Sarcoma Herpes Virus

Kaposi's sarcoma (KS) is the most common cancer among HIV-positive patients. Histogenetic origin of KS has long been elusive due to a mixed expression of both blood and lymphatic endothelial markers in KS tumor cells. However, we and others discovered that Kaposi's sarcoma herpes virus (KSHV) induces lymphatic reprogramming of blood vascular endothelial cells by upregulating PROX1, which functions as the master regulator for lymphatic endothelial differentiation. Here, we demonstrate that the KSHV latent gene kaposin-B enhances the PROX1 mRNA stability and plays an important role in KSHV-mediated PROX1 upregulation. We found that PROX1 mRNA contains a canonical AU-rich element (ARE) in its 3′-untranslated region that promotes PROX1 mRNA turnover and that kaposin-B stimulates cytoplasmic accumulation of the ARE-binding protein HuR through activation of the p38/MK2 pathway. Moreover, HuR binds to and stabilizes PROX1 mRNA through its ARE and is necessary for KSHV-mediated PROX1 mRNA stabilization. Together, our study demonstrates that kaposin-B plays a key role in PROX1 upregulation during lymphatic reprogramming of blood vascular endothelial cells by KSHV.


Introduction
Kaposi's sarcoma (KS) is causally associated with human herpes virus (HHV)-8, also called KS-associated herpes virus (KSHV) [1]. KSHV develops various-sized KS tumors that are structurally accompanied by aberrant angiogenesis of slit-like vessels frequently containing red blood cells and inflammatory cells [2,3]. KS tumor cells characteristically appear spindle-shaped and are believed to be derived from endothelial cells. KS tumor cells were initially proposed to originate from blood vascular endothelial cell (BEC) because of their expression of BEC-specific antigens [4,5,6,7,8,9,10]. Later, however, KS tumor cells were also found to express lymphatic endothelial cell (LEC)-specific markers such as VEGF receptor-3 (VEGFR-3/flt4) and podoplanin [11,12,13,14,15,16,17,18,19], arguing for their lymphatic origin. Recently, we and others have demonstrated that KSHV reprograms the transcriptional profile of BECs to resemble LECs by upregulation of PROX1, the master regulator for the LEC-differentiation [20,21,22,23]. PROX1, the mammalian homolog of the Drosophila neuronal cell fate regulator Prospero, is a homeodomain transcription factor essential for development of a variety of organs, including the lymphatic system [24,25], the liver [26], the lens [27,28], the brain [29,30,31,32], the ear [33,34,35,36] and the heart [37,38]. During early lymphatic development, endothelial cells in the cardinal vein exhibit a mixed phenotype of both BECs and LECs. A subset of venous endothelial cells begins to express PROX1 and migrates out to form the initial lymphatic vessels [24,25]. This lymphatic differentiation process is found to be arrested in PROX1 knockout mice, which fail to develop the lymphatic system [24,25]. We and others found that ectopic expression of PROX1 induces lymphatic reprogramming of post-developmental BECs [39,40]. Therefore, PROX1 is thought to override the BEC phenotype by repressing BEC-specific markers and to induce lymphatic phenotypes by upregulating LEC-specific genes, functioning as the master control regulator for LEC differentiation.
Controlling mRNA stability is an important post-transcriptional regulatory process, which allows a rapid adjustment of the copy number of mRNAs by involving a sequence element called AUrich element (ARE) [41,42,43,44]. AREs are usually 50-150 nucleotide long and locate in the 39-untranslated region (UTR) of mRNAs with a short half life, serving as an mRNA-destabilizing determinant by promoting degradation of mRNAs. Notably, ARE-containing mRNAs are found to represent as much as ,8% of total mRNAs encoded in human cells and are involved in many essential biological processes such as signal transduction, cell growth and differentiation, immune responses, hematopoiesis and apoptosis [43,44]. AREs are grouped into three classes based on the number and distribution of the core AUUUA pentamers [43,45,46]. Class I ARE genes contain several dispersed copies of the AUUUA motif within the AU-rich region and include c-myc, c-fos, cyclins A, B1 and D1 and interferon-c. Class II ARE genes have at least 2 overlapping UUAUUUA(U/A)(U/A) motifs and include tumor necrosis factor (TNF)-a, interleukin (IL)-1b, IL-2, IL-3, granulocyte/macrophage colony-stimulating factor (GM-CSF), Cox-2 and VEGF. Finally, less characterized class III AREs do not contain the canonical AUUUA motif and are found in genes such as c-jun, GLUT, p53 and hsp70. Interestingly, while many cytokine-encoding mRNAs harbor the class II AREs, mRNAs encoding cell cycle regulators and transcription factors contain the class I and occasionally class III AREs [43]. Several ARE-binding proteins have been reported to either destabilize or stabilize ARE-containing mRNAs [43,47]. Notably, HuR, embryonic lethal abnormal vision (ELAV)-like RNA-binding protein, is one of the best characterized ARE-binding proteins and stabilizes labile ARE-containing mRNAs such as c-fos, MyoD, p21, cyclins A, B1 and D1, TNF-a, GM-CSF and VEGF [42,43,48,49,50,51,52,53,54]. Predominantly present in the nuclei, HuR shuttles between the nucleus and cytoplasm in response to various internal and external stimuli, and its mRNA-stabilizing function has been attributed to its cytoplasmic localization [46,50,52,55,56].
Importantly, the KSHV latent gene kaposin-B has been shown to activate the p38/MK2 pathway and to stabilize various cytokine mRNA containing AREs [57,58]. Kaposin-B can directly bind to MK2 and promote its kinase activity through its DR2 repeats and, in response to lipopolysaccharide (LPS), kaposin-B and MK2 were shown to be exported to cytoplasm [57,58]. Kaposin-B and MK2/p38 proteins have been shown to enhance the stability of ARE-containing mRNAs such as GM-CSF and IL-6, leading to an enhanced production of cytokines and signaling proteins [57,58]. However, the molecular mechanism underlying the kaposin-B/MK2-mediated stabilization of the ARE-containing mRNA remains to be better defined.
While study of endothelial cell fate reprogramming by KSHV has provided important insights into KS oncogenesis, the molecular mechanism underlying KSHV-mediated PROX1-upregulation has only begun to be elucidated. An interesting recent report has shown that Akt activation through gp130 receptor may play an important role in KSHV-induced lymphatic reprogramming [59]. Here, we found that PROX1 harbors an unusually long 39-UTR that contains the canonical ARE, which functions as a PROX1 mRNA-destabilizing determinant. Moreover, we discovered that HuR protein physically binds and stabilizes PROX1 mRNA and that cytoplasmic localization of HuR protein is activated by kaposin-B. Together, our data demonstrate that kaposin-B plays a key role in KSHV-mediated PROX1 upregulation.

KSHV infection is required for PROX1 upregulation in vascular endothelial cells
We and others have previously demonstrated that KSHV induces lymphatic reprogramming of vascular endothelial cells by upregulating PROX1 and that this PROX1-upregulation occurs in KSHV-infected cells in vitro [20,21,22,23]. In this study, we further investigated the correlation between KSHV infection and PROX1 upregulation both in vitro and in vivo. For the in vitro study, we infected cultured human dermal BECs with KSHV for 7 days and performed immunofluorescent studies for PROX1 and LANA/ ORF73, a KSHV viral protein that marks a latent KSHV-infection. We found that PROX1 was upregulated predominantly in LANApositive, KSHV-infected BECs, but not in LANA-negative, uninfected neighboring BECs ( Figure 1A-D). Our study revealed that ,78% of the cells (n = 390) was double negative for PROX1 and LANA, and ,18% double positive ( Figure 1E), strongly correlating PROX1 upregulation with de novo KSHV-infection.
We next stained KS biopsy sections with anti-PROX1 and LANA antibodies to analyze co-expression of PROX1 and LANA in KS tumor cells ( Figure 1F-I). We found that the majority of cells in KS tumors were infected with KSHV based on the characteristic LANA-speckles in KSHV-infected nuclei in vivo ( Figure 1G) and that most of the LANA-positive, KSHV-infected cells upregulated PROX1. Out of 160 cells, ,81% of the cells was double positive for PROX1 and LANA, ,12% double negative, and only 3,4% cells single-positive ( Figure 1J), a finding consistent with the in vitro data. Together, our studies demonstrate that PROX1 upregulation occurs only in KSHV-infected cells.

The KSHV latent gene kaposin-B induces PROX1upregulation in LECs
We next investigated how KSHV induces PROX1-upregulation in endothelial cells. Since KSHV upregulates PROX1 in the latent stage, when only a few viral genes are known to be expressed, we hypothesized that one or more KSHV latent genes may be responsible for the activation of PROX1 expression and thus tested their ability to upregulate PROX1 in various endothelial cell backgrounds such as LECs, BECs and human umbilical venous endothelial cells (HUVECs). Real time RT-PCR (qRT-PCR) analyses revealed that ectopic expression of the KSHV latent gene kaposin-B in LECs significantly upregulated PROX1 (6,7-fold) (Figure 2A). In comparison, kaposin-B did not notably induce PROX1 expression in either BECs or HUVECs (Figure 2A). From the same set of experiment, we investigated the effect of kaposin-B in the regulation of other lymphatic genes (podoplanin, VEGFR-3, LYVE-1, FGFR-3, SLC and p57) in LECs, BECs and HUVECs ( Figure S1). Interestingly, we found that although kaposin-B alone did not seem to induce the lymphatic reprogramming as extensively as KSHV [20,21,22,23], kaposin-B alone was able to partially modulate the

Author Summary
Kaposi's sarcoma (KS) is the most common cancer in HIVpositive patients and KS-associated herpes virus (KSHV) was identified as its causing agent. We and others have discovered that when the virus infects endothelial cells of blood vessels, KSHV reprograms the cell type resembling endothelial cells in lymphatic vessels. Although endothelial cells of the blood vascular system and of the lymphatic system share functional similarities, the cell type-reprogramming does not occur under a normal physiological condition. Therefore, cell-fate reprogramming by the cancer-causing virus KSHV provides an important insight into the molecular mechanism for viral pathogenesis. Our current study investigates the molecular mechanism underlying the KSHV-mediated cell fate reprogramming. We identified that a KSHV latent gene kaposin-B plays an important role in KSHV-mediated regulation of PROX1 to promote PROX1 mRNA stability. This study will provide a better understanding on the tumorigenesis and pathogenesis of KS with a potential implication toward new KS therapy.
expression of other LEC-signature genes in KSHV-infected endothelial cells. We next overexpressed kaposin-B in LECs and performed the semi-quantitative RT-PCR analyses against PROX1 and IL-6, a known kaposin-B target gene [57]. Like IL-6, PROX1 was also significantly upregulated by kaposin-B in LECs ( Figure 2B). In a separate experiment, a Flag-tagged kaposin-B was transfected into LECs and the steady-state level of PROX1 protein was determined by western blot analyses ( Figure 2C). Together, these data demonstrate that the KSHV viral gene kaposin-B can upregulate PROX1 expression in LECs, but not in BECs and HUVECs where the lymphatic-specific PROX1 is not expressed.

PROX1 mRNA has an unusually long 39 untranslated region
We then set out to investigate the molecular mechanism underlying kaposin-B-induced PROX1 upregulation. Kaposin-B has been demonstrated to upregulate various cytokine genes by stabilizing their mRNAs through AREs located in their 39-UTRs [57]. We thus examined the mRNA structure of the PROX1 gene. Although the open reading frame (ORF) of human or mouse PROX1 gene is about 2.2 kb long and encodes a 737-amino acidlong protein, reported northern blot analyses revealed that PROX1 transcript was as large as 8-kb in various tissues [60,61,62], suggesting that the PROX1 transcript has a long UTR at the 59-and/or 39 ends. In fact, we found that an 8-kb PROX1 transcript harboring an extended 39-UTR has been annotated in a public genome database (ENST00000366958). However, the corresponding Prox1 transcript from mouse has not been annotated in the same public database. To further confirm the presence of PROX1 transcript with an extended 39-UTR in both human and mouse, we aligned 7-kb downstream genomic sequences of human or mouse PROX1 against human or mouse expressed sequence tag (EST) databases and found that numerous EST sequences were mapped to the downstream of both human PROX1 and mouse Prox1 genes ( Figure 3A&B), indicating that this region is indeed transcribed as a part of the fifth exon of PROX1 mRNA in both species. We then investigated sequence conservation of this 39-UTR among Prox1 genes of other species. Analyses using a genome browser revealed a high DNA sequence homology in this extended 39-UTR, especially in the second half, of the Prox1 genes from primates, placental mammals or vertebrates covering 48-speices [63] ( Figure 3C).
Since PROX1 was expressed in multiple organs such as the brain, liver, muscle and heart [60,61,62] and most of EST fragments were derived from cDNA libraries of these organs, we asked whether LECs express PROX1 mRNA with a long 39-UTR and thus performed northern blot analyses by using three different 39-UTR probes and a PROX1 open reading frame (ORF) probe against RNAs isolated from human LECs. Indeed, both the 39-UTR and the ORF probes detected a single ,8-kb band ( Figure 3D), indicating that LECs express a ,8-kb long PROX1 mRNA. In addition, we performed the 39-rapid amplification of cDNA end (RACE) analyses and found that a majority of human LEC-PROX1 mRNA terminates at 5,414-bp downstream from the termination codon of PROX1 ( Figure 3E). We identified a classical poly-A signal sequence (AATAAA) at ,20-bp upstream of the termination site. Together, our data demonstrate that A functional AU-rich element is present in the PROX1 39-UTR and promotes PROX1 mRNA turnover Because kaposin-B stabilizes cytokine mRNAs through AREs in their 39-UTR [57], we asked whether PROX1 mRNA contains a canonical ARE in its unusually long 39-UTR. We performed bioinformatic analyses against the conserved vertebrate Prox1 39-UTRs and found an AU-rich region approximately 400-bp upstream of the mRNA termination site ( Figure 3F). Notably, this region contains three copies of the AUUUA core pentamer (its location marked in Figure 3A) that can be classified as a class I ARE [43,44,45,46]. Moreover, these three tandem copies of core pentamer were also found to be conserved in Prox1 mRNA of dog, mouse, human, chicken and zebrafish ( Figure 3F).
We next investigated whether the newly discovered PROX1-ARE can serve as an mRNA instability determinant by utilizing the classical b-globin mRNA stability reporter system (pTet-BBB) [64,65]. In this system, the tetracycline (Tet)-controlled promoter directs the inducible expression of the rabbit b-globin gene and a specific DNA sequence element such as a putative ARE can be cloned into the 39-UTR of the b-globin gene in order to evaluate its ability to destabilize otherwise stable b-globin mRNA [64,65]. We cloned a 40-bp fragment containing the PROX1 AUUUA core pentamer in pTet-BBB ( Figure 3G) and evaluated its effect on the stability of b-globin mRNA in NIH3T3/Tet-Off cells by northern blotting analyses. Indeed, the three AUUUA motifs from the PROX1 39-UTR significantly destabilized the stable b-globin mRNA as potently as the c-fos ARE that was used as a positive control ( Figure 3H), demonstrating that the 40-bp PROX1-ARE is sufficient to function as an instability determinant for PROX1 mRNA. Quantification of the northern blot bands revealed that while the unmodified b-globin mRNA shows a long half-life ( .210-minutes), the PROX1-ARE, like the c-fos-ARE, shortened the half-life of bglobin mRNA to ,60-minutes ( Figure 3I). Together, our data demonstrate that PROX1 mRNA contains an ARE functioning as an mRNA instability determinant in its unusually long 39-UTR and that this newly identified PROX1-ARE may play an important role in the post-transcriptional regulation of PROX1 expression.

HuR physically binds to PROX1-ARE and upregulates PROX1 expression
We next investigated the molecular mechanism underlying the regulation of PROX1 mRNA stability through its ARE and searched for ARE-binding factors that may interact with the PROX1-ARE. The nuclear protein HuR, also known as ELAVL1, has been shown to be one of the best characterized ARE-binding proteins that bind various cytokine/chemokine mRNAs to increase their stability [66]. Therefore, we evaluated the possibility of HuR binding to the PROX1-ARE and promoting the mRNA stability. Toward this aim, a HuRexpressing vector was transfected into primary LECs and PROX1 expression was determined. Indeed, our semi-quantitative RT-PCR analyses showed that HuR overexpression resulted in upregulation of PROX1 as well as a known HuR-target gene VEGF [49,67] ( Figure 4A). This HuR-mediated PROX1 upregulation was also confirmed by using quantitative qRT-PCR ( Figure 4B) and western blot ( Figure 4C) analyses.
We next asked whether HuR protein can physically interact with the AU-rich region of the PROX1 39-UTR and performed co-immunoprecipitation (co-IP) for a protein-RNA complex of endogenous HuR protein and PROX1 mRNA from primary LECs by using an anti-HuR antibody as previously described [68]. Precipitated protein-RNA complex was de-crosslinked, reversetranscribed and PCR-amplified by using two neighboring sets of PROX1 PCR primers; Probe-1 and Probe-2. While Probe-1 primer pair detects the PROX1-ARE region, Probe-2 primer pair, 42-bp away from Probe-1, binds at 25-bp downstream from the end of PROX1 mRNA ( Figure 4D). Importantly, whereas RT-PCR using Probe-1 amplified a corresponding product, RT-PCR using Probe-2 did not yield any product, indicating that the endogenous HuR protein is physically associated with the PROX1-ARE region and also that the Probe-1 product was not due to possible genomic DNA contamination in our co-immunoprecipitation assays ( Figure 4D). As controls, we could detect the association of HuR protein with VEGF and UBE2N mRNAs [49,68], but not with GAPDH mRNA ( Figure 4D).
We next performed a RNA electrophoresis mobility-shift assay (RNA EMSA) to corroborate the molecular interaction between HuR protein and PROX1-ARE mRNA. We in vitro transcribed a RNA EMSA probe spanning the PROX1-ARE region and then investigated if this RNA probe can make a stable RNA-protein complex with recombinant HuR protein and/or another known ARE-binding protein tristetraprolin (TTP) [47]. Indeed, while GST alone or GST-TTP protein did not make any detectable RNAprotein complex, GST-HuR recombinant protein formed a stable complex with the PROX1-ARE RNA probe (Figure 4 E, lanes 1-4). It is interesting to find that HuR, but not TTP, binds to PROX1- ARE RNA probe, although both are known to have an affinity to the ARE motif [69]. Moreover, the complex formation between PROX1-ARE RNA probe and GST-HuR protein could be inhibited by an unlabeled specific competitor (PROX1-ARE RNA probe), but not by an unlabeled non-specific competitor (yeast total RNA), indicating a specific molecular interaction between HuR protein and PROX1-ARE region ( Figure 4E, lanes 5,14). Taken together, our data demonstrate that HuR protein physically interacts with PROX1 mRNA through the AU-rich region.

Kaposin-B promotes PROX1 mRNA stability through HuR
Our findings of kaposin-B-induced PROX1 upregulation and HuR-binding to PROX1-ARE directed us to ask whether HuR and/or kaposin-B upregulate PROX1 by enhancing PROX1 mRNA stability. Toward this question, we overexpressed HuR or kaposin-B in primary LECs and quantified the steady-state level of PROX1 mRNA by qRT-PCR. Indeed, the ectopic expression of HuR or kaposin-B delayed the turnover of PROX1 mRNA in LECs and increased the half-life of PROX1 mRNA from ,60 minutes in the control LECs to ,180 minutes in LECs overexpressing HuR or kaposin-B ( Figure 5A). We then asked whether HuR is required for kaposin-B-mediated PROX1 upregulation by knockdown of HuR in kaposin-Bexpressing LECs. We found that HuR-knockdown significantly inhibited kaposin-B-mediated upregulation of PROX1 mRNA and protein determined by qRT-PCR and western analyses, respectively ( Figure 5B&C). Moreover, we confirmed that this reduction in kaposin-B-mediated PROX1 upregulation is due to a decrease in PROX1 mRNA stability upon knockdown of HuR ( Figure 5D). Together, our data demonstrate that kaposin-B upregulates PROX1 by promoting its mRNA stability through HuR. Kaposin-B stimulates cytoplasmic localization of HuR protein While HuR protein mainly resides in the nucleus, various cell stress signals activate cytoplasmic accumulation of HuR [52]. We next asked if kaposin-B activates localization of HuR protein to the cytoplasm where mRNA stability is regulated. Indeed, our immunofluorescent analyses revealed that ectopic upregulation of kaposin-B stimulated cytoplasmic mobilization of HuR protein ( Figure 6A). Moreover, we harvested the cytoplasmic and nuclear fractions from control vs. kaposin-B-overexpressing LECs to quantify the amount of mobilized HuR by kaposin-B. Consistent with the immunostaining data, a significant amount of HuR protein was found to be exported to the cytoplasm ( Figure 6B). Therefore, our data demonstrate that cytoplasmic accumulation of HuR protein is activated by kaposin-B, which may play an important role in PROX1 upregulation.
The p38/MK2 kinase pathway is required for cytoplasmic accumulation of HuR protein and PROX1 mRNA stabilization Kaposin-B has been shown to activate the p38/MK2 pathway and stabilize various cytokine mRNAs [57,58]. We further examined this previous observation in LECs and found that the ectopic expression of kaposin-B activated phosphorylation of p38 and MK2 proteins ( Figure 7A). We then investigated whether activation of the p38/MK2 pathway is required for kaposin-Bmediated PROX1 upregulation. Notably, previous studies have shown that p38 MAPK promotes cytoplasmic accumulation of HuR in different cell types [70,71,72]. Therefore, we asked if activated MK2 can promote cytoplasmic localization of HuR protein in LECs and transfected LECs with vectors expressing a wild type, constitutively active (EE) or dominant negative (K76R) form of MK2 protein [73]. We found that whereas the constitutively active (EE) MK2 protein stimulated cytoplasmic accumulation of HuR, wild type or dominant negative MK2 protein did not ( Figure 7B). Moreover, the expression of PROX1 was upregulated by constitutively active (EE) MK2 protein, but not by wild type or dominant negative MK2 protein, in LECs determined by western analyses (Figure 7C,D). Importantly, this upregulation of PROX1 by MK2 (EE) protein was abrogated by siRNA-mediated knockdown of HuR ( Figure 7C,D), indicating that HuR is required for the kaposin-B/p38/MK2 pathwaymediated PROX1 upregulation. We also confirmed these findings by using quantitative qRT-PCR measuring PROX1 mRNA level ( Figure 7E). Interestingly, we found that HuR knockdown slightly decreases PROX1 expression compared to control siRNA ( Figure 7D,E). We think that this is because HuR protein is present at a basal level in the cytoplasm of LECs ( Figure 6B) and may stabilize PROX1 mRNA under the normal condition and thus knockdown of HuR resulted in decrease of PROX1 expression. This speculation is supported by our endogenous HuR co-immunoprecipitation data ( Figure 4D) demonstrating a stable complex formation between HuR protein and PROX1 mRNA in normal primary LECs.
We next asked whether MK2 (EE) protein-mediated PROX1 upregulation is due to PROX1 mRNA stabilization and thus studied the regulation of the PROX1 mRNA half-life by MK2 (EE) protein. Indeed, MK2 (EE) promoted PROX1 mRNA stability by increasing mRNA half-life by more than 60 minutes in LECs ( Figure 7F). Moreover, we found that this increase in PROX1 stability by MK2 (EE) could be abrogated by knockdown of HuR ( Figure 7G). Taken together, our data demonstrate that

The role of HuR in KSHV-mediated PROX1 upregulation
Our data above demonstrate the essential contribution of HuR in PROX1 upregulation by kaposin-B. We next asked whether HuR plays an important role in KSHV-mediated PROX1 upregulation in primary human BECs and HUVECs. We first confirmed the KSHV-mediated PROX1 mRNA upregulation in BECs and HUVECs ( Figure 8A) and also PROX1 protein expression by KSHV in HUVECs ( Figure 8B). We next investigated the cytoplasmic accumulation of HuR protein in KSHV-infected cells and found that KSHV infection resulted in a significant cytoplasmic localization of HuR protein in HUVECs ( Figure 8C). Consistently, we also observed HuR cytoplasmic localization in KSHV-infected BECs (data now shown). We then asked whether HuR plays a role in the KSHV-mediated PROX1 upregulation by knockdown of HuR by siRNA in KSHV-infected BECs and HUVECs. Importantly, knockdown of HuR significantly decreased the half-life of PROX1 mRNA in KSHVinfected BECs and HUVECs ( Figure 8D,E). In comparison, we were not able to measure the half-life of PROX1 mRNA in uninfected BECs or HUVECs due to their low/absent expression of PROX1. Taken together, our data demonstrate that KSHV infection stimulates cytoplasmic localization of HuR protein and that HuR plays an important role in KSHV-mediated PROX1 upregulation.

Working model for KSHV/kaposin-induced PROX1 expression and lymphatic reprogramming
Based on our data presented here, we build a working model for the molecular mechanism underlying KSHV-mediated PROX1 upregulation (Figure 9). When KSHV infects vascular endothelial cells, the virus may employ two or more mechanisms for PROX1 upregulation: one may be a transcriptional activation of PROX1 possibly through Akt activation [59] and the other a posttranscriptional PROX1 mRNA stabilization by kaposin-B, which activates the p38/MK2 pathway. Activated MK2 by kaposin-B stimulates the nuclear export and cytoplasmic accumulation of HuR protein. Cytoplasmic HuR protein binds to the AU-rich region in the 39-UTR of PROX1 mRNA and slows down PROX1 mRNA turnover, thus increasing the steady-state level of PROX1 mRNA in KSHV-infected cells.

Discussion
KS tumor cells were reported to be derived from endothelium about 40 years ago [74]. However, the exact histogenetic origin of KS had remained uncertain because KS cells were found to express mixed cell-lineage markers of BECs and LECs [75]. Previously, we and others demonstrated that KSHV induces lymphatic reprogramming of vascular endothelial cells by upregulating the master control gene of lymphatic differentiation, PROX1 [20,21,22,23]. This finding of endothelial cell fate reprogramming by KSHV has provided an important insight into the pathology of KS and KSHV. Nonetheless, the question how the virus induces the host cell fate change remained to be answered.
Our previous study revealed that the KSHV latent gene LANA only marginally induced expression of PROX1 (1.93-fold) [20]. Interestingly, recent two exciting studies have established a molecular connection between the function of LANA and PROX1 gene regulation: Di Bartolo et al. showed that KSHV LANA inhibits TGF-b signaling through epigenetic silencing of TGF-b type II receptor [76] and Oka et al. demonstrated that inhibition of TGF-b signaling upregulates PROX1 by ,2 fold [77]. In spite of this intriguing molecular association, the degree of KSHVinduced PROX1 upregulation ( .8 fold) shown by us and others [20,21,22,23] prompted us to hypothesize that LANA could not be the major activator for PROX1 upregulation in KSHV-infected cells and that another mechanism should also be present for KSHV-mediated PROX1 upregulation. This rationale directed us to search for an additional activator(s) among KSHV latent genes.
In this study, we investigated the role of kaposin-B, a latent gene of KSHV, in the KSHV-mediated PROX1 upregulation and found that kaposin-B promotes mRNA stability of PROX1. We defined the structure of PROX1 mRNA and identified a class Itype ARE in its 39-UTR, through which PROX1 expression can be post-transcriptionally regulated by physiological or pathological stimuli. At this point, it seems that kaposin-B targets both class I (e.g., PROX1) and class II (e.g., GM-CSF [57]) mRNAs and the specificity determinant for kaposin-B targets needs to be further defined. Moreover, we discovered that the ARE-binding protein HuR is exported to the cytoplasm by kaposin-B and also by KSHV infection, and increase PROX1 mRNA stability. These findings are consistent with our observation from in vitro cell cultures and KS tumor samples that PROX1 upregulation occurs in only KSHV-infected cells, not in neighboring uninfected cells (Figure 1).
Kaposin-B-induced mRNA stabilization appears an attractive model for PROX1 upregulation since kaposin-B has been shown to upregulate other cytokine mRNAs such as GM-CSF and IL-6 [57]. However, it needs to be highlighted that kaposin-B is not the sole component in the molecular mechanism underlying PROX1upregulation by KSHV because mRNA stabilization inevitably requires pre-existing mRNA and PROX1 mRNA is not expressed in BECs [24,25]. Therefore, other factors/stimuli are needed for , or a dominant negative-MK2 (K76R) vector was transfected into LECs for 16 hours and the nuclear (N) and cytoplasmic (C) fractions were isolated and subjected to western blot analyses against HuR, lamin A/C (nuclear marker) and tubulin (cytoplasm marker). (C) HuR is required for MK2-mediated PROX1 upregulation. Expression of PROX1 protein was determined by western blot analyses in LECs that were transfected with a control (CTR), a wild type-MK2 (WT), a constitutively active-MK2 (EE) or a dominant negative-MK2 (K76R) vector. After 16 hours, transfected cells were divided into two groups and then transfected again with either luciferase siRNA (siCTR) or HuR siRNA (siHuR). Protein level of PROX1, Myc-tag, HuR and b-actin was determined after 16 hours. (D) Relative intensity of PROX1 bands in panel C was measured and charted in a graph. (E) MK2-mediated PROX1 upregulation is due to an increased expression of PROX1 mRNA. The steady-state level of PROX1 mRNA from the same set of experiment as panel (C) was determined by qRT-PCR. (F) MK2 activation increases Prox1 mRNA stability. A control (CTR) or a constitutive-MK2 (MK2 (EE)) vector was transfected into LECs for 16 hours and the steady-state level of PROX1 mRNA was determined by qRT-PCR analyses at the indicated time point (minutes) after Actinomycin-D (ActD) (2 mg/ml) treatment. (G) HuR is necessary for MK2-induced PROX1 stability. Constitutively active MK2 (EE) was overexpressed in LECs for 16 hours and the cells were then transfected again with luciferase siRNA (MK2 (EE)+siCTR) or HuR siRNA (MK2 (EE)+siHuR) for 16-hours, followed by Actinomycin-D (ActD) administration (2 mg/ml). PROX1 mRNA level at the indicated time points was determined by qRT-PCR analyses. Data are represented by mean and standard deviation (SD) and three independent experiments were performed to yield similar results. Asterisks in panels F and G present p-value less than 0.05. doi:10.1371/journal.ppat.1001046.g007 the initial transcriptional activation of the PROX1 mRNA synthesis. This hypothesis is further supported by two of our findings. First, PROX1-upregulation by kaposin-B was much prominent in LECs where PROX1 mRNA is already present, in comparison to BEC and HUVEC-backgrounds where PROX1 expression is fairly low, if any ( Figure 2A). Second, PROX1 expression in BECs and HUVECs was more strongly activated by the entire virus (KSHV), compared to by kaposin-B alone (Figures 2A & 8A), again suggesting that kaposin-B alone is unable to efficiently activate PROX1 expression in BECs and HUVECs. Importantly, Morris et al has recently shown that Akt activation through gp130 is required for KSHV-mediated PROX1 upregulation and lymphatic reprogramming [59]. Considering the fact that KS tumors have been associated with numerous cytokines, chemokines and diffusible factors in their microenvironments [78,79,80,81], it is highly likely that multiple KS-associated viral and/or cellular factors may activate the gp130/Akt pathway to prime the initial activation of PROX1 transcription [59]. We hypothesize that this transcriptional activation may require a subsequent secondary post-transcriptional mechanism involving kaposin-B to achieve PROX1 upregulation. This two-step mechanism is also consistent with the fact that PROX1-upregulation is limited to KSHV-infected cells. Taken together, Figure 9 illustrates our hypothesis that both steps (transcriptional activation and mRNA stabilization) may be necessary to achieve PROX1-upregulation and lymphatic reprogramming of blood vascular endothelial cells by KSHV.  On the other hand, kaposin is a unique KSHV latent gene considering distinct features in its transcription and translation. Kaposin transcripts are the most abundantly expressed viral mRNA throughout all stages of KS progression determined by in situ hybridization assays and a complex translational program directs production of multiple isoforms of kaposin gene product, termed kaposin A, B and C [82,83]. Kaposin-B uses a non-conventional CUG start codon and consists of a series of tandem repeats of hydrophobic 23-amino acids, named DR1 and DR2 [82,83]. DR2 can directly bind to MK2 and, when overexpressed, DR2 domain alone can dominantly inhibit the mRNA-stabilization function of the whole kaposin-B protein [57,58]. It was also found that the DR1/DR2 repeats are more abundantly expressed in lytic-or TPAtreated cells [84] and that DR2 is reiterated three to five times in different stains of KSHV [58], suggesting a significant variation in the expression level and DR2 repeat number of kaposin-B. Accordingly, it will be interesting to investigate whether kaposin-B-induced PROX1 mRNA stabilization is more prominent during viral reactivation. Moreover, it is possible that both the expression level and structure of kaposin-B may affect its mRNA-stabilizing function through different ARE-binding proteins (including HuR) and target a different set of cellular mRNAs. It would be also exciting to study if kaposin-B may have other functions in addition to its role in mRNA stabilization.
Our current study brings up numerous questions. To date, the pathological role of PROX1 in KS development and progression has not been defined. Is the lymphatic phonotype more favorable for KSHV infection and propagation? Is lymphatic reprogramming a by-product or a goal of PROX1 upregulation by KSHV? Some insights may be obtained from interesting findings that PROX1 was shown to increase the invasion of endothelial tumor cells [85] and that PROX1 promotes the transition from benign to highly dysplastic phenotype in colon cancer [86]. While these studies support the oncogenic roles of PROX1, many other studies demonstrate the opposing role of PROX1 as a tumor suppressor [87,88,89,90,91,92]. Therefore, cell type and tissue microenvironment may be crucial for PROX1 to play the oncogenic versus tumor suppressive role and further studies will be necessary to better understand the role of PROX1 in KS tumor development. Moreover, PROX1 has been reported to be important for cell-fate specifications in a broad range of cells including lymphatics [24,25], liver [26], lens [27,28], brain [29,30,31,32], the ear [33,34,35,36] and the heart [37,38] during development and has been associated with post-developmental processes such as cell cycle regulation [28,39,93] and inflammation [94]. It will be very interesting to investigate whether PROX1 is post-transcriptionally regulated for any of its functions during and after development.

Cell cultures and transfection
Human primary dermal blood vascular endothelial cells (BECs) and lymphatic endothelial cells (LECs) were isolated from anonymous neonatal human foreskins and cultured as previously described [95] with an approval of the University of Southern California Internal Review Board (PI: YK Hong). Primary human umbilical venous endothelial cells (HUVECs) were purchased from Lonza (Basel, Switzerland), and cultured in EGM-2 medium (Lonza). NIH3T3 cells containing tTA (Tet-Off), named B2A2, were kindly provided by Dr. Ann-Bin Shyu (University of Texas Houston Health Science Center) [64,65,96]. NIH3T3 cells were transfected by using Lipofectamine 2000 (Invitrogen) and primary endothelial cells were transfected by electroporation (Nucleofactor II, Amaxa Biosystems).

Rapid amplification of cDNA ends (RACE)
39-RACE assay was performed by following the manufacturer's instruction (First Choice RLM Race kit, Applied Biosystems). Total RNA was isolated from human primary LECs and subjected to 39 RACE by using two sets of PCR primers (GGATTGGTCT-CAGCGCTACC, GCGAGCACAGAATTAATACGACT; AA-CTGAACTGATAAAGTCAATTTTTG, CGCGGATCCGAA-TTAATACGACTCACTATAGG). Amplified PCR products were cloned in the pGEM-T Easy vector (Promega) and sequenced to define the end of PROX1 mRNA.

KS and KSHV production
De-identified anonymous KS specimens were obtained from the AIDS and Cancer Specimen Resource (ACSR) with an approval of the University of Southern California Internal Review Board (PI: YK Hong). KSHV was purified from BCBL-1 cells by following a previous description [23] with a minor modification. Briefly, BCBL-1 cells were cultured to the density of 5,10 million cells/ml and then activated with TPA (20 ng/ml) and sodium butyrate (NaB, 3 mM). After 24,48 hours, TPA/NaB-containing media was replaced with normal media and cells were incubated for additional 5 days. Culture media was then collected and filtered through 0.45-mm filters and centrifuged for 20 minutes at 4uC at 8,000 rpm to remove cell debris. Supernatant was centrifuged for 5 hours at 4uC at 11,000 rpm to pellet the virus, which was then resuspended in endothelial cell media. Infectivity was measured by immunohistochemistry for LANA after infection for 5 days.
RNA EMSA and RNA/protein immunoprecipitation GST-HuR and GST-TTP fusion proteins were isolated as previously described [98]. Expression vectors for GST-HuR and GST-TTP were kindly provided by Drs. Henry Furneaux (University of Connecticut) [51] and Gilles Pages (University of Nice-Sophia Antipolis, France) [99], respectively. For RNA EMSA experiments, radio-labeled RNA transcripts (200 Kcpm/reaction) was mixed with 200 ng GST fusion proteins or GST alone in a previously described binding buffer [100]. The reaction mixture was incubated for 30 minutes at 30uC and treated for 15 minutes at room temperature with 100 U of Ribonuclease T1 (Roche). For competition assays, specific (unlabeled PROX1-ARE RNA probe) or nonspecific (yeast total RNA) competitors were incubated for 15 minutes at 30uC with the proteins in the binding buffer before the addition of the labeled transcripts. The reaction mixtures were resolved on 8% native polyacrylamide gels in 0.56 Tris borate-EDTA (TBE) buffer. Gels were dried and exposed to X-ray film. RNA/protein immunoprecipitation assay was performed essentially as described [96] by using protein A/G-Sepharose beads preincubated with anti-HuR (3A2, SC-5261, Santa Cruz Biotechnology). RNA was isolated from supernatants and reverse-transcribed with Superscript II (Invitrogen). The transcripts were amplified by PCR by using the following primers: PROX1 (probe-1), ATCC-TAATTTATTTGATGAAGGTG, TGCACATACATTCAGT-TTAAGAGG; PROX1 (probe-2), TCAGTGCCGTTTGAT-TTTCTTAAA, GGAACA TCTTTCCTTGTTCTTAGA; VEG-F, TCCAATCTCTCTCTCCCTGAT, CGGATAAACAGTAG-CACCAAT [67]; and UBE2N, TACCCAATGGCAGCCCC-TAA, TTCCACTGCTCCGCTACATCA [68]. The resulting PCR products were analyzed by 2% agarose gels.

Quantitative real time and conventional RT-PCR
Real-time RT-PCR (qRT-PCR) was performed by using TaqMan EZ RT-PCR Core Reagent (Applied Biosystems). For dual-labeled probe-based qRT-PCR reactions, each reaction was multiplexed for both target gene and the internal control b-actin for normalization. Conventional RT-PCR was performed by using Superscript II (Invitrogen) and Taq polymerase (New England Biolabs). Primer sequences will be provided upon request.