Kaposi’s sarcoma-associated herpesvirus promotes mesenchymal-to-endothelial transition by resolving the bivalent chromatin of PROX1 gene

Increasing evidence suggests that Kaposi’s sarcoma (KS) arises from Kaposi’s sarcoma-associated herpesvirus (KSHV)-infected mesenchymal stem cells (MSCs) through mesenchymal-to-endothelial transition (MEndT). KSHV infection promotes MSC differentiation of endothelial lineage and acquisition of tumorigeneic phenotypes. To understand how KSHV induces MEndT and transforms MSCs to KS cells, we investigated the mechanism underlying KSHV-mediated MSC endothelial lineage differentiation. Like embryonic stem cells, MSC differentiation and fate determination are under epigenetic control. Prospero homeobox 1 (PROX1) is a master regulator that controls lymphatic vessel development and endothelial differentiation. We found that the PROX1 gene in MSCs harbors a distinctive bivalent epigenetic signature consisting of both active marker H3K4me3 and repressive marker H3K27me3, which poises expression of the genes, allowing timely activation upon differentiation signals or environmental stimuli. KSHV infection effectively resolves the bivalent chromatin by decreasing H3K27me3 and increasing H3K4me3 to activate the PROX1 gene. vIL-6 signaling leads to the recruitment of MLL2 and SET1 complexes to the PROX1 promoter to increase H3K4me3, and the vGPCR-VEGF-A axis is responsible for removing PRC2 from the promoter to reduce H3K27me3. Therefore, through a dual signaling process, KSHV activates PROX1 gene expression and initiates MEndT, which renders MSC tumorigenic features including angiogenesis, invasion and migration.


Introduction
Kaposi's sarcoma-associated herpesvirus (KSHV), also termed human herpesvirus type 8 (HHV8), has been proven to be the etiological agent of Kaposi's sarcoma (KS) [1], primary effusion lymphoma (PEL) [2], and multicentric Castleman's disease (MCD) [3]. Recently, KSHV was also found to be associated with childhood osteosarcoma in Xinjiang Uyghur population [4]. However, the lack of comprehensive understanding about KSHV infection and tumorigenesis hampers our ability to treat KSHV-associated cancers. For example, KS is atypical cancer. The onset of KS as simultaneously multiple lesions (in the absence of obvious metastasis), the multiclonal malignant nature (especially in the late stages), and the presence of unusual inflammatory cell infiltration and neoangiogenesis in the very early stage of KS, distinguish KS from other orthodox tumors. The origin of KS spindle cells remains elusive. The current leading model is that KS spindle cells may derive from endothelial cell lineage as KS cells express endothelial markers. However, KS cells are poorly differentiated and also express other markers such as smooth muscle, macrophage, and mesenchymal markers, suggesting that KS cells do not faithfully represent the endothelial cell lineage [5]. Recently, we found a series of evidence suggesting that KS originates from oral mesenchymal stem cells (MSCs) through KSHV-induced mesenchymal-to-endothelial transition (MEndT) [6].
MSCs have been identified as a population of hierarchical postnatal stem cells with the potential to self-renew and differentiate into osteoblasts, chondrocytes, adipocytes, cardiomyocytes, myoblasts, and neural cells [7,8]. The oral cavity contains a variety of distinct MSC populations, including dental pulp stem cells (DPSCs), periodontal ligament stem cells (PDLSCs), and gingiva/mucosa-derived mesenchymal stem cells (GMSCs) [9][10][11]. Among these MSCs, PDLSCs and GMSCs have the potential to directly interact with oral cavity saliva, microbiota, and virus, therefore having a great chance to be infected by KSHV in the oral cavity. Our recent study found that AIDS-KS spindle cells express Neuroectodermal stem cell marker (Nestin) and oral MSC marker CD29, suggesting an oral/craniofacial MSC lineage of AIDSassociated KS [6]. Furthermore, MSCs were found highly susceptible to KSHV infection, and the infection effectively promotes multiple lineage differentiation, especially endothelial differentiation. When implanted in mice, KSHV-infected MSCs were transformed into KS-like spindle-shaped cells with other KS-like phenotypes [6,12]. These findings provided evidence that KS derives from KSHV-infected MSCs. In addition, others identified PDGFRA+ MSCs as KS progenitors [13].
We sought to understand how KSHV infection regulates the MEndT process to promote malignant transformation. It is known that the regulation of stem cell fate determination and differentiation is mainly at the epigenetic level, which has been intensively studied in embryonic stem cells [14]. Embryonic stem cells undergo global remodeling during early stem cell development that commits them to the desired lineage. The lineage commitment of stem cells is controlled by epigenetic mechanisms, and histone modification appears to be the most important layer in such regulation [15]. Some critical transcription factor gene promoters in embryonic stem (ES) cells harbor a distinctive histone modification signature that combines the activating histone H3K4me3 mark and the repressive H3K27me3 mark. These bivalent domains are considered to poise the expression of developmental genes, allowing rapid activation upon developmental signals [16,17]. Increasing evidence suggests that mesenchymal stem cell differentiation, like embryonic development regulation, is also controlled at the epigenetic level [18]. Since MSCs face the same challenge as embryonic stem cells-multi-lineage differentiation and stem cell fate determination, it is possible that similar epigenetic mechanism found in embryonic stem cells may also apply to MSCs. The homeobox gene Prospero homeobox 1 (PROX1) is a master regulator gene that controls lymphatic vessel development and endothelial differentiation [19] and is up-regulated by KSHV infection in MSCs [6]. Elucidation of the regulation mechanism of PROX1 gene expression by KSHV is vital for comprehension of KSHV-mediated MEndT that leads to KS.
In the current study, we found that the PROX1 gene in MSCs exhibits a "bivalent" epigenetic signature consisting of both H3K4me3 and H3K27me3 marks. We investigated how the bivalent chromatin structure is established and how KSHV infection resolves the bivalent domain to activate PROX1 and MEndT genes. This study elucidates the mechanism underlying KSHV-mediated MEndT at the transcription level.

Kaposi's sarcoma lesions express PROX1 and KSHV infection of mesenchymal stem cells induces PROX1 expression
It was reported that among 30 oral Kaposi sarcomas (KS) lesions, twenty-eight (93.3%) were positive for PROX1 [20]. To confirm that KSHV infection is responsible for the induction of PROX1 expression, we examined three cases of KS, one oral and two skin lesions for expression of PROX1 and KSHV antigen LANA using immunohistochemistry analysis. The result showed that all three samples were PROX1-positive and LANA positive ( Fig 1A). Periodontal ligament stem cells (PDLSCs) were infected with KSHV and the expression of PROX1 in response to KSHV infection was examined by double-staining immunofluorescence assay (IFA) and Western analysis. IFA showed that PROX1 expression was consistently correlated with KSHV infection (LANA expression) (Fig 1B), and Western blot analysis showed that PROX1 expression was upregulated by KSHV ( Fig 1C). Furthermore, mock-and KSHVinfected PDLSCs were implanted in immunocompromised mice under the kidney capsule. After four weeks, the implants were subjected to an immunohistochemistry study for PROX1 expression. As shown in Fig 1D, PROX1 expression was found to be significantly elevated in KSHV-infected PLDSCs in comparison to mock-infected PDLSCs. Overall, PROX1 expression is upregulated both in vivo and in vitro in response to KSHV infection in MSCs.
KSHV infection in PDLSCs suggest that PROX1 plays a critical role in KSHV-mediated MEndT and KS development. To investigate this hypothesis, we determined the role of PROX1 in MEndT by using an shRNA-mediated gene silencing approach. An shRNA specific to PROX1 was introduced into PDLSCs by a lentiviral vector, and then the transduced PDLSCs were infected by KSHV. The knockdown efficiency of the shRNA was validated by Western blot (Fig 2A). The contribution of PROX1 to endothelial lineage differentiation and neovascularity was evaluated by Western blot to test endothelial markers expression and a Matrigel tubulogenesis assay. Results showed that the expression level of lymphatic endothelial markers (VEGFR3, LYVE-1, and PDPN) and pan-endothelial marker (VEGFR2) decreased in PDLSCs transduced with shPROX1 compared to those with control shRNA (Fig 2B). Correspondingly, shRNA-mediated PROX1 inhibition resulted in the loss of tubulogenesis (Fig 2C). Transwell-matrigel assay was used to assess cell invasion driven by PROX1. The result showed that shRNA-mediated PROX1 inhibition significantly reduced invasion ability compared to control shRNA ( Fig 2D). Furthermore, when implanted in mice under kidney capsule, KSHV-

PLOS PATHOGENS
Epigenetic regulation of PROX1 by KSHV infected PDLSCs underwent an MEndT process revealed by the expression of endothelial markers CD31, VCAM1, and PDPN, while MEndT did not occur in PROX1-knockdown cells as they failed to express those endothelial markers (Fig 2E), suggesting that PROX1 is essential for the initiation of MEndT induced by KSHV infection.

KSHV promotes MEndT through epigenetic regulation of the PROX1 gene that possesses a bivalent chromatin structure
We sought to understand how KSHV infection regulates the MEndT process. Evidence suggests that mesenchymal stem cell differentiation is controlled at the epigenetic level [18]. KSHV-mediated MSCs reprogramming likely takes place at this level. Histone modifications play critical roles in regulating the expression of developmental genes in mammalian development [21,22]. To explore epigenetic regulation of KSHV-mediated MEndT, we examined histone modifications on the chromatin of several MEndT genes in mock-and KSHV-infected PDLSCs using Chromatin immunoprecipitation (ChIP) assay. Results revealed an interesting feature that the PROX1 promoter locus in mock-infected PDLSCs heavily consists of both

PLOS PATHOGENS
Epigenetic regulation of PROX1 by KSHV repressive mark H3K27me3 and active mark H3K4me3. KSHV-infection dramatically increased H3K4me3 and decreased H3K27me3 in the PROX1 promoter ( Fig 3A). This chromatin structure pattern, known as bivalent chromatin structure, has been reported in the genes of key transcriptional factors responsible for development in embryonic stem cells [16]. The bivalent domain structure was found to be restricted to the PROX1 promoter, not seen in the coding region of PROX1 (Supporting Information, S1 Fig). In contrast, other MEndT relevant genes, such as PML, TGF-β3, and TGF-βRII, obtained elevated H3K4me3 in the chromatin of their regulatory regions in response to KSHV infection ( Fig 3A).
Given the unique nature of the bivalent domain on the PROX1 gene, we would confirm that the observed bivalent structure truly reflects the simultaneous presence of both H3K4me3 and H3K27me3 on the same chromosome rather than the presence of two subpopulations with different epigenetic characters. To this end, we performed a sequential ChIP assay in which PDLSCs chromatin was immunoprecipitated first with the Lys27 tri-methyl antibody and secondly with the Lys4 tri-methyl antibody ( Fig 3B). This sequential purification is designed to retain only chromatin that concomitantly carries both the active and repressive histone modifications. Then real-time PCR was used to detect PROX1 promoter DNA on the bivalent nucleosomes. As shown in Fig 3C, H3K4me3 and H3K27me3 were found to co-exist on the same PROX1 promoter chromatin, accurately representing a true bivalent epigenetic state. Overall, our result suggests that the bivalent epigenetic feature may silence the PROX1 gene and its downstream cascade in MSCs but keep them poised for activation. KSHV PDLSCs were subjected to a Chromatin immunoprecipitation (ChIP) assay with antibodies against H3K4me3 and H3K27me3. The extracted DNA was analyzed by qPCR for the PROX1 promoter region and other MEndT-related genes, including PML, TGF-βR2 and TGF-β3. PROX1 gene in mock-infected PDLSCs heavily consists of both active mark H3K4me3 and repressive mark H3K27me3, exhibiting a distinctive bivalent chromatin domain. (B) Schematic illustration of the sequential ChIP assay. Crosslinked chromatin from MSCs was immunoprecipitated with an anti-H3K27me3 antibody followed by a second immunoprecipitation with an anti-H3K4me3 antibody. (C) The extracted DNA was analyzed by qPCR with primers of the PROX1 promoter, and the result showed that H3K4me3 and H3K27me3 co-exist in the same nucleosome in the PROX1 promoter. Error bars represent SD. Statistics analysis was performed using one-way ANOVA test, and P-value was calculated by GraphPad Prism. P-value < 0.05 was considered significant ( � P<0.05, �� P<0.01 and ��� P<0.001) and N.S represented no significance.

KSHV resolves the PROX1 bivalent chromatin by recruiting MLL2 and SET1 complexes to and removing PRC2 from the PROX1 promoter
To understand how KSHV manipulates PROX1 epigenetic regulation, we aimed to investigate the mechanism of establishing the bivalent domain in the PROX1 promoter and resolving it by KSHV. The central players in setting up and maintaining bivalent chromatin structure are the trithorax group (TrxG) and polycomb group (PcG) proteins [17]. The former consists of SET1A/B and MLL1-4 complexes, catalyzing the trimethylation of histone H3 Lys4 [23]. The latter forms multi-subunit polycomb-repressor complexes (PRC) 1 and 2, responsible for the trimethylation of histone H3 Lys27. PRC2 is particularly crucial for H3K27me3 in many developmental genes. To identify the tri-methyltransferase complexes that are recruited to PROX1 chromatin and altered in response to KSHV infection, ChIP assays were performed on mockand KSHV-infected PDLSCs with antibodies against CFP1, MLL1, 2, respectively. The results showed that (i) prior to KSHV infection, MLL1 had been recruited to the PROX1 promoter region; (ii) KSHV infection did not change MLL1 content in the chromatin but recruited MLL2 and CFP1 to the promoter (Fig 4A-4C). CFP1, a CXXC finger protein 1 and specific component of the SET1A/B complex, serves as global regulation on H3K4 methylation. Thus, the elevation of CFP1 (SET1/COMPASS) suggests that KSHV infection increases the total level of H3K4me3 on the PROX1 promoter, and the recruitment of MLL2 (MLL2/COMPASS) resolves the bivalent domain of the PROX1 gene. Besides, no change in UTX (MLL3/4 COM-PASS) was detected in the PROX1 gene after KSHV infection (Fig 4D). RBBP5 and WDR5, shared by all six methyltransferase complexes (SET1A, SET1B, and MLL1-4), did not respond significantly to KSHV infection with a slight increase of RBBP5 ( Fig 4E) and a minor decrease of WDR5 ( Fig 4F). Therefore, we conclude that KSHV infection results in the recruitment of MLL2 and SET1 complexes to the PROX1 promoter region to elevate active histone marker H3K4me3.
On the other hand, the histone methyltransferase responsible for the deposition of the H3K27me3 mark on a bivalent domain is the PRC2 complex [24][25][26]. Our result showed that KSHV infection led to the removal of EZH2, the catalytic subunit of the PRC2 complex, from the PROX1 promoter, causing the loss of the H3K27me3 mark in the PROX1 region ( Fig 4G). Altogether, KSHV-mediated recruitment of MLL2 and SET1 and removal of PRC2 resolve the bivalent domain of the PROX1 promoter.

Identification of KSHV components that are responsible for epigenetic regulation of PROX1 gene expression
To elucidate the mechanism underlying KSHV altering the epigenetic features of the PROX1 gene, we sought to identify the viral genes that control PROX1 activation. Inspection of KSHV transcription profiles of KS lesions [27] and KSHV-infected PDLSCs [6] revealed a class of KSHV genes expressed in both KS lesions as well as KSHV-infected MSCs. We ectopically expressed these KSHV genes, including latent genes LANA, K12, and vFLIP and lytic genes K1, ORF45, ORF50, K8, PAN, vIL-6, and vGPCR in PDLSCs. Among these viral genes, vIL-6 and vGPCR significantly promoted PROX1 expression ( Fig 5A).

PLOS PATHOGENS
Epigenetic regulation of PROX1 by KSHV proliferation, migration, and tumor formation [31]. To explore if vIL-6 and vGPCR promote PROX1 expression at the epigenetic level, we examined the effect of vIL-6 and vGPCR expression on histone modifications in the PROX1 gene in PDLSCs. We found that vIL-6 expression led to a significant increase in H3K4me3 marks but had no effect on H3K27me3 content, while vGPCR expression caused a dramatic decrease in H3K27me3 marks in the PROX1 promoter but did not change the level of H3K4me3 (Fig 5B and 5C). Co-expression of vIL-6 and vGPCR reproduced the alterations of H3K4me3 and H3K27me3, closely resembling to what was observed in KSHV-infected PDLSCs, and complete resolution of the bivalent domain ( Fig 5D). Furthermore, vIL-6 expression deposited H3K4me3 via recruiting MLL2 histone methyltransferases to the promoter, and vGPCR removed H3K27me3 from the bivalent domain by downregulating PRC2 complex (EZH2) in the PROX1 promoter (Fig 5E and 5F). Taken together, KSHV resolves the PROX1 bivalent domain through two pathways, namely vIL-6 and vGPCR signaling, to alter active H3K4me3 and repressive H3K27me3, respectively, and consequently regulate PROX1 gene expression and MEndT process.

KSHV-mediated vIL-6 and VEGF secretions regulate MSC differentiation by resolving the bivalent chromatin structure of the PROX1 gene
KS is an angiogenesis tumor, and pathological neoangiogenesis is a hallmark of the cancer. We previously reported that the treatment of MSCs with the conditioned medium of KSHVinfected MSCs sufficiently renders cells increased angiogenesis activity and MEndT, indicating that KSHV can induce MSC MEndT in a paracrine or autocrine manner [6,32]. We examined the conditioned medium of KSHV-infected PDLSCs for its effect on PROX1 expression and found that the treatment of PDLSCs with the conditioned medium dramatically upregulated PROX1 expression in PDLSCs, especially 48 hours post-infection (48 hpi) (Fig 6A). Furthermore, the treatment with the conditioned medium effectively altered the histone modification pattern closely resembling what was observed in KSHV-infected PDLSCs, i.e., increased H3K4me3 and decreased H3K27me3 in the PROX1 promoter region (Fig 6B). The effects of these viral components on PROX1 expression were assayed by Western blot. ChIP assays were performed with PDLSCs ectopically expressing vIL-6 (B), vGPCR (C), or both (D) for their effects on H3K4 and H3K27 trimethylation in the PROX1 promoter. The effects of expression of vIL-6 (E) and vGPCR (F) on CFP1, MLL2, and EZH2 in the PROX1 gene were examined by ChIP assays. Error bars represent SD. Statistics analysis was performed using the one-way ANOVA test. P-value < 0.05 was considered significant ( � P<0.05, �� P<0.01 and ��� P<0.001) and N.S represented no significance. https://doi.org/10.1371/journal.ppat.1009847.g005

Epigenetic regulation of PROX1 by KSHV
In light of that KSHV-encoded vIL-6 and vGPCR can resolve the bivalent chromatin domain of the PROX1 gene, and so does the conditioned medium of KSHV-infected MSCs, we proposed a model for the paracrine regulation of epigenetic property in PROX1 gene as follows. It has been reported that vGPCR upregulates VEGF production to induce angiogenesis and lymphatic reprogramming [31,33]. Thus, vGPCR-induced VEGF, together with vIL-6, may be necessary and sufficient to resolve the bivalent chromatin and activate the PROX1 gene. We found that ectopic expression of vGPCR indeed induced VEGF-A expression in PDLSCs, as revealed by RT-qPCR and Western blot analysis (Fig 6C). When PDLSCs were treated by human interleukin-6 (hIL-6), vIL-6 supernatant, and VEGF-A, we observed the upregulation of PROX1 expression and PDPN expression (Fig 6D). The effects of vIL-6 and VEGF-A on histone modification of the PROX1 gene were analyzed by treating PDLSCs with the supernatant of vIL-6-transfected 293T cells (48 hours post-transfection) and VEGF-A, followed by ChIP assay for H3K4me3 and H3K27me3. Interestingly, the treatment with vIL-6 supernatant led to increased H3K4me3, and VEGF-A caused decreased H3K27me3 in the PROX1 promoter (Fig 6E and 6F). Therefore, we conclude that KSHV induces PROX1 activation through a dual signaling process, i.e., activating vIL-6 signaling to increase active histone marker H3K4me3 and using vGPCR-VEGF-A axis to reduce repressive histone marker H3K27me3, to resolve the bivalent chromatin structure and activate PROX1 gene expression.
Then, the contribution of vIL-6 signaling and vGPCR-VEGF axis to the MEndT process and tumorigenic properties were evaluated. PDLSCs were transfected with vectors for vIL-6 and vGPCR and MEndT was monitored by the expression of lymphatic endothelial and pan-endothelial markers PDPN, LYVE-1, and VEGFR2 with Western analysis. Result showed that the endothelial markers were significantly elevated in response to vIL-6 and vGPCR (Fig 6G). In addition, in vIL-6 or/and vGPCR-overexpressed PDLSCs, both angiogenesis activity, represented by the formation of capillary-like tubules (Fig 6H), and malignant invasive feature, revealed by a Transwell-matrigel invasion assay (Fig 6I), were significantly elevated. Furthermore, a neutralizing antibody against VEGF-A can efficiently block endothelial marker expression and inhibit cell angiogenesis (tubule formation) and malignant invasion abilities (Supporting Information, S3 Fig). Taken together, vIL-6 and vGPCR-VEGF axis are crucial for MEndT and malignant transformation potential.

vIL-6 signaling and vGPCR-VEGF axis are indispensable for epigenetic regulation of PROX1 and acquisition of tumorigenic properties
To assess the significance of vIL-6 and vGPCR-VEGF axis-mediated epigenetic regulation of PROX1, we silenced these two viral genes individually or simultaneously in KSHV-infected The conditioned medium-treated cells were subjected to a ChIP assay for histone modifications (H3K4me3 and H3K27me3) in the PROX1 promoter. (C) vGPCR induced the production of VEGF-A, revealed at mRNA and protein levels. (D) Human interleukin-6 (hIL-6), viral interleukin-6 (vIL-6) (supernatant of HEK293T cells transfected with a vIL-6 expression vector), and VEGF were added in PDLSC cultures, and cells were analyzed by Western blot with PROX1 and PDPN antibodies. ChIP assays were performed with vIL-6-(E) and VEGF-treated PDLSCs (F) for H3K4me3 and H3K27me3 modifications in the PROX1 promoter. (G) The expression of endothelial differentiation markers PDPN, LYVE-1, and VEGFR2 in vIL-6, vGPCR, or vIL-6/ vGPCR expressing PDLSCs was evaluated by Western blot. (H) vIL-6, vGPCR, and vIL-6/vGPCR expressing PDLSCs were placed on Matrigel, and tubulogenesis was examined by Tubule formation assay and quantified by measuring the total segment tube length using the ImageJ software. (I) vIL-6, vGPCR, and vIL-6/vGPCR expressing PDLSCs (1.5x10 4 cells/well) were seeded in the upper chamber of Transwell with Matrigel in serum-free α-MEM. α-MEM containing 20% FBS was used as a stimulus for chemotaxis in the lower chamber. Cells migrated to the lower chamber were stained with crystal violet and counted. Quantification was performed by the ImageJ software. The error bars represent SD. Statistics analysis was performed using t-test in GraphPad Prism. P-values < 0.05 were considered significant ( � P<0.05, �� P<0.01 and ��� P<0.001) and N.S represented no significance. https://doi.org/10.1371/journal.ppat.1009847.g006

Epigenetic regulation of PROX1 by KSHV
PDLSCs by using small interference RNA (siRNA). Knockdown of vIL-6 and vGPCR, each moderately inhibited PROX1 expression compared to control siRNA (si-NC) cells (Fig 7A and  7C). The inhibition of vIL-6 reduced H3K4me3 marks in the PROX1 promoter but did not affect the H3K27me3 marker (Fig 7B), while knockdown of vGPCR expression increased H3K27me3 in the promoter but had no effect on H3K4me3 (Fig 7D). Double knockdown of vIL-6 and vGPCR dramatically reduced PROX1 expression (Fig 7E) by decreasing active mark H3K4me3 and elevating repressive mark H3K27me3 in the PROX1 promoter, counteracting the effect of KSHV infection in the PROX1 promoter (Fig 7F).
Then vGPCR-VEGF axis signaling was blocked in vGPCR and vIL6-expressing PDLSCs using an anti-VEGF-A neutralizing antibody and the effect of ablation of vGPCR axis on cell oncogenic phenotypes was examined. As demonstrated by tubule formation and Transwellmatrigel invasion assays, ectopical expression of vGPCR or co-expression of vGPCR and vIL-6 significantly elevated PDLSC angiogenesis and malignant invasion capabilities, which were completely abolished when the cells were incubated with anti-VEGF-A antibody (Fig 7G and  7H).

Discussion
We previously reported that KSHV infection of human MSCs triggers an MEndT process that may lead to Kaposi's sarcoma development [6]. The current study sought to elucidate the mechanism of KSHV-mediated MEndT. The central event in KSHV-initiated MEndT is viralmediated gene expression reprogramming that re-routes the differentiation process of MSCs. In this study, we found that the expression of the PROX1 gene, a master regulator of endothelial lineage differentiation and MEndT, is controlled at the epigenetic level through a bivalent chromatin domain that is usually seen in embryonic development transcription factor genes [16]. KSHV infection activates PROX1 by resolving the bivalent domain through a dual signaling process, i.e., vIL-6 signaling increases active histone marker H3K4me3, and vGPCR-VEGF axis decreases repressive histone marker H3K27me3. As a consequence, the activation of PROX1 initiates endothelial lineage differentiation and MEndT, leading to MSC acquisition of tumorigenic features such as proliferation, angiogenesis, and migration/invasion (schematically depicted in Fig 8).
Postnatal stem cells have multi-lineage differentiation potential into osteoblasts, adipocytes, chondrocytes, or myocytes [34]. PROX1 is a homeodomain transcription factor essential for the development of a variety of organs, including the lymphatic system, lens, heart, liver, pancreas, and central nervous system [35,36]. During early lymphatic development, endothelial cells in the cardinal vein exhibit a mixed phenotype of BECs and LECs. A subset of venous endothelial cells begins to express PROX1 and migrates out to form initial lymphatic vessels [19]. [37]. KSHV infection of BECs and LECs reprograms both types of cells to move away from their original cell identities and acquire the opposite cell type (LEC and BEC, respectively) properties, in accompanying with up-and down-regulation of PROX1 expression, respectively [38,39]. Further study delineated the opposing regulation of PROX1 by KSHV to direct cell differential fate through IL-3R (up-regulation) and Notch (down-regulation) signaling [40]. In our previous study, we found that KSHV-infection of oral MSCs can up-regulate PROX1 and, as a consequence, promote endothelial-lineage differentiation or MEndT, while KSHV-infection of terminally differentiated LECs down-regulates PROX1 [6]. Taken together, PROX1 is a key regulator in KSHV-mediated MEndT that reprograms MSCs leading to KS sarcomagenesis. In the current study, we confirmed the critical roles of PROX1 in developing KS tumorigenic phenotypes

PLOS PATHOGENS
Epigenetic regulation of PROX1 by KSHV (Fig 2). Therefore, the regulation of PROX1 expression by KSHV is a key event in transforming a KSHV-infected MSC to Kaposi's sarcoma. In addition, recent studies reveled another role of KSHV-promoted PROX1 activation in KS tumorigenesis that PROX1 enhances KSHV lytic replication and sustains the population of KSHV infected cells that would otherwise be lost as KS cells divide [41,42]. Overall, the revelation of epigenetic regulation of PROX1 in MSCs by KSHV in this study provided a novel insight into the MEndT process and KS tumorigenesis.
Herein, the PROX1 gene was found to reside in bivalent chromatin where activating histone H3 Lys4 trimethylation (H3K4me3) mark and repressive histone H3 Lys27 trimethylation (H3K27me3) mark co-exist on the same nucleosome. In 2006, a subset of promoters associated with both activating H3K4me3 and repressive H3K27me3 marks, referred to as "bivalent" chromatin domain, was discovered in mouse embryonic stem cells (ESCs) [16]. Such a signature bivalent chromatin appears to be more specific to the promoter of developmentally important transcription factors. The synchronous existence of H3K4me3 and H3K27me3 posits these transcription factor genes in a poised state, enabling them to be rapidly activated upon suitable developmental cues and environmental stimuli [17]. Our finding that the The knockdown (KD) efficiency was verified with quantitative RT-PCR and the effect on PROX1 expression was assayed by Western blot. (B) KSHV-PDLSCs transduced with si-vIL-6 or si-NC were subjected to ChIP assays to evaluate H3K4me3 and H3K27me3 in the PROX1 promoter. (C) vGPCR expression was silenced with a specific siRNA (si-vGPCR) in KSHV-infected PDLSCs. The knockdown efficiency was verified with quantitative RT-PCR and the effect on PROX1 expression was assayed by Western blot. (D) KSHV-PDLSCs transduced with si-vGPCR or si-NC were subjected to ChIP to evaluate H3K4me3 and H3K27me3 in the PROX1 promoter. (E) KSHV-infected PDLSCs co-transduced with si-vIL-6 and si-vGPCR were assayed for PROX1 expression by Western blot, and (F) the histone modifications in the PROX1 promoter by ChIP assay. (G and H) vGPCR or vIL-6/vGPCR expressing PDLSCs were incubated with a neutralizing antibody against VEGF-A (α-VEGF-A). The effects of the neutralizing antibody on cell angiogenesis and malignant invasion were examined by tubule formation assay (G) and Transwell-matrigel invasion assay (H), respectively. Statistics analysis was performed using the one-way ANOVA test, and P-value was calculated by GraphPad Prism. P-values < 0.05 were considered significant ( � P<0.05, �� P<0.01 and ��� P<0.001) and N.S represented no significance. https://doi.org/10.1371/journal.ppat.1009847.g007

PLOS PATHOGENS
Epigenetic regulation of PROX1 by KSHV transcription factor PROX1 gene is also posited in the distinctive bivalent chromatin in oral MSCs suggests that lineage differentiation transcription factor genes in postnatal stem cells are regulated at the epigenetic level in the same mechanism as that used in embryonic stem cells. In this study, we investigated how KSHV resolved the bivalent chromatin domain for PROX1 activation and revealed that KSHV uses a dual signaling process, i.e., activating vIL-6 signaling to increase active histone marker H3K4me3 and using vGPCR-VEGF-A axis to reduce repressive histone marker H3K27me3, to resolve the bivalent chromatin structure and activate PROX1 gene expression. This is an interesting finding and suggests that there is a "two-factorauthentication" mechanism in the resolution of bivalent chromatin. The biological significance for such a two-factor-authentication has not yet been understood, but we speculate that genes in bivalent domain chromatin are poised for a quick response to an external stimulus. Still the dual signaling process may grant a more efficient and accurate (with an additional layer of security) response to activate the gene.
vIL-6 and vGPCR were found to play essential roles in resolving the bivalent chromatin domain and activating the PROX1 gene. vGPCR exerts this function through producing VEGF-A. Thus, vIL-6 and VEGF-A regulate PROX1 expression through an autocrine or paracrine mechanism. This finding explained our previous observations that MSCs can acquire KS tumorigenic properties by being incubated with conditioned media of KSHV-infected MSCs [32], and MSCs could undergo MEndT via paracrine regulation [6]. This finding also implies that KSHV may not only regulate PROX1 expression in its infected MSCs, but also modulate uninfected cells in the environment for PROX1 activation. In fact, it is well known that Kaposi sarcoma secrets abundant inflammation cytokines and growth factors including IL-1, IL-6, basic fibroblast growth factor (bFGF), platelet-derived growth factor (PDGF), tumor necrosis factor (TNF), interferon-gamma (IFN-γ), and vascular endothelial growth factor (VEGF), as well as KSHV-encoded vIL-6 [38,[43][44][45]. In the early stage, the KS lesion is not a real sarcoma, but an angiohyperplastic-inflammatory lesion mediated by inflammatory cytokines and angiogenic factors. With the development of KS, spindle-shaped tumor cells accumulate and become dominant, triggered or amplified by KSHV infection [46]. Our result is consistent with the notion that KS is a cytokine disease and further suggests that cytokine autocrine or paracrine plays a central role in establishing KS spindle-shaped tumor cells in KS lesions.

Ethics statements
The use of human samples and PDLSCs in this study was approved by the Medical Ethics Review Board of Sun Yat-sen University (approval no. 2015-028). Written informed consent was provided by study participants. The animal experiments were approved by the Animal Ethics Review Board of Sun Yat-sen University (approval no. SYSU-IACUC-2018-000162) and carried out strictly following the Guidance suggestion of caring laboratory animals, published by the Ministry of Science and Technology of the People's Republic of China.

Cell culture
The isolation of human PDLSCs has been described previously [6]. PDLSCs were maintained in alpha minimal essential medium (α-MEM) with 10% heat-inactivated fetal bovine serum (FBS), 2 mM L-glutamine and antibiotics. The second to tenth passages of PDLSCs were used in this study. Human embryonic kidney HEK293T cells were purchased from American Type Culture Collection (ATCC) and cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and antibiotics. iSLK.219 cells, cultured in DMEM with 10% FBS and antibiotics, were gifted from Dr. Ke Lan at Wuhan University.

KSHV preparation and infection
iSLK.219 cells carrying rKSHV.219 were induced for lytic replication by 1 μg/mL doxycycline and 1 mM sodium butyrate. Five days post-induction. The culture media were filtered through a 0.45 um filter and ultra-centrifuged with Beckman Coulter OptimaTM L-100XP at 100000 g for 1 h. The pellet was resuspended in 1/100 volume (culture media) of 1x PBS and stored at -80˚C until use. PDLSCs were seeded at 2x10 5 cells per well in 6-well plates. Cells were infected with KSHV in the presence of polybrene (4 μg/mL) at an MOI of 50 (viral genome copy equivalent). After centrifugation at 2500 rpm for 60 min at room temperature, the cells were incubated at 37˚C with 5% CO 2 for 2 hours. Then, the inoculum was removed by changing culture medium.

Sequential Chromatin immunoprecipitation (sequential ChIP)
Sequential ChIP was performed as previously described [16]. Briefly, the first step was standard ChIP with antibody against H3K27me3 and the sheared DNA fragments were controlled in the range of 400-500 base pairs. Eluted chromatin was diluted 50-fold and subjected to a second immunoprecipitation with antibody against H3K4me3. Immune complexes were eluted, cross-linking was reversed, and DNA was purified as described above.

Immunofluorescence assay
Cells were fixed with 3.6% formaldehyde in PBS for 10 min, permeabilized in 0.1% Triton X-100 for 15 min and blocked in 1% BSA for 1 h at room temperature. The samples were

Tube formation assay
48-well plates were pre-coated with matrigel (1:1 dilute with α-MEM without FBS, 100 μL/ well) and incubated at 37˚C for 1 h to allow gelation to occur. 1x10 5 PDLSCs were suspended in 200 μL α-MEM without FBS and placed to the top of the gel. The cells were incubated at 37˚C with 5% CO 2 for 6-8 h, and tube formation images were captured using a ZEISS microscope. The quantitation of the tube was carried out using the ImageJ software to measure the length of the total tube segments in the image. The average value was used for the histogram.

Cell invasion assay
Cell invasion assays were carried out in 24-well Transwell units. Briefly, polycarbonate filters with 12-mm pores were precoated with 100 μL of matrigel (1:10 diluted with α-MEM without FBS). 1.5x10 4 cells in serum-free media were placed in the top wells, and the bottom chambers were filled with 20% FBS medium. After 24 h incubation, the cells that had passed through the filter were stained with crystal violet. The number of invaded cells was counted from multiple randomly selected fields using ImageJ software. Photographs were taken and independent experiments were performed in triplicate. particles were prepared by transfecting HEK293T cells with pMSCVpuro-(K1, K12, ORF45, ORF50/RTA, ORF71/vFLIP, ORF73/LANA, K8, PAN, vIL-6, and vGPCR) and PIK plasmids at the ratio of 1:1. Media containing retroviruses were harvested at 48 and 72 h and used to transduce PDLSCs. Transduced PDLSCs were selected under 2 μg/mL puromycin for a week.

Supernatant transfer assay
The supernatant from KSHV-PDLSCs (48 hpi) was transferred to PDLSC culture at the ratio of 1:1. HEK293T cells were transfected with vIL-6 expression vector, and media were collected after 48 h post-transfection 48. The supernatant with vIL-6 was added to PDLSC culture at the ratio of 1:1.

Kidney capsule transplantation
PDLSCs were suspended with α-MEM media and seeded on 96-well plates precoated with 0.5% agarose at 2x10 4 cells per well. The spheroids were grown at 37˚C for 1 to 2 d with 5% humid CO 2 . The media containing spheroids were collected and transferred to 15 mL conical tubes, washed twice with 1x PBS, and centrifuged at 1000 rpm for 5 min. About 100 spheroids were placed on gelfoam scaffold for 1 d in medium. Recipient female nude mice (6-8 weeks, n = 3-5) were weighed and anesthetized with isoflurane. Kidney was exposed through an incision of the skin and muscle on the back of the mouse. The kidney capsule was opened with the fine tip of no.5 forceps. The spheroids/scaffold block was placed under the kidney capsule. Sutures were placed, and capsule-grafting products were harvested 28 d after transplantation.

Statistical analysis
Data were analyzed by two-tailed Student's t-test and one-way ANOVA in GraphPad Prism. P values < 0.05 were considered significant ( � P<0.05, �� P<0.01 and ��� P<0.001) and N.S represented no significance.
Supporting information S1 Fig. Histone modifications (H3K4me3 and H3K27me3) in the PROX1 coding region in mock-and KSHV-infected PDLSCs. Mock-and KSHV-infected PDLSCs were subjected to a Chromatin immunoprecipitation (ChIP) assay with antibodies against H3K4me3 and H3K27me3. The extracted DNA was analyzed by qPCR with primers of the PROX1 coding region.

S3 Fig. Blockade of vGPCR-VEGF-A axis using a neutralizing antibody against VEGF-A inhibits PROX1 activation and KSHV-promoted angiogenesis and invasion phenotypes.
(A) PDLSCs were treated with conditioned medium from KSHV-PDLSC. The expression of PROX1 and PDPN were analyzed by Western blot. (B) KSHV-infected PDLSCs (K-PDLSCs) were placed on Matrigel, and tubulogenesis was examined by Tubule formation assay and quantified by measuring the total segment tube length using the ImageJ software. (C) K-PDLSCs (1.5x10 4 cells/well) were seeded in the upper chamber of Transwell with Matrigel in serum-free α-MEM. α-MEM containing 20% FBS was used as a stimulus for chemotaxis in the lower chamber. Cells migrated to the lower chamber were stained with crystal violet. The error bars represent SD. Statistics analysis was performed using t-test in GraphPad Prism. Pvalues < 0.05 were considered significant ( � P<0.05, �� P<0.01 and ��� P<0.001) and N.S represented no significance. (TIF)