Merkel Cell Polyomavirus Large T Antigen Induces Cellular Senescence for Host Growth Arrest and Viral Genome Persistence through Its Unique Domain

Senescent cells accumulate in the host during the aging process and are associated with age-related pathogeneses, including cancer. Although persistent senescence seems to contribute to many aspects of cellular pathways and homeostasis, the role of senescence in virus-induced human cancer is not well understood. Merkel cell carcinoma (MCC) is an aggressive skin cancer induced by a life-long human infection of Merkel cell polyomavirus (MCPyV). Here, we show that MCPyV large T (LT) antigen expression in human skin fibroblasts causes a novel nucleolar stress response, followed by p21-dependent senescence and senescence-associated secretory phenotypes (SASPs), which are required for MCPyV genome maintenance. Senolytic and navitoclax treatments result in decreased senescence and MCPyV genome levels, suggesting a potential therapeutic for MCC prevention. Our results uncover the mechanism of a host stress response regulating human polyomavirus genome maintenance in viral persistency, which may lead to targeted intervention for MCC.


Introduction
Senescence refers to stable cell cycle arrest thought to be an antiproliferative cellular defense mechanism against cancer [1] and viral replication [2]. This process can be induced through a variety of cellular stress responses, such as the DNA damage response (DDR), the nucleolar stress response (NSR), and replicative stresses [1,3], leading to p53 activation and the subsequent upregulation of p21. p21 can induce sustained G1 cell cycle arrest and senescence, thereby preventing cell cycle progression and proliferation [1]. Senescent cells have an altered phenotype compared to their non-senescent counterparts, including the induction of senescence-associated secretory phenotype (SASP) genes, upregulation of anti-apoptotic genes [4], and downregulation of various antiviral signaling pathways [5,6].
Several studies have shown that viral infections can induce cellular senescence as part of the antiviral response [2,7]. Vesicular stomatitis virus titers were reduced when permissive cells were driven into senescence by bleomycin treatment [2]. The infection of fibroblast cells with measles virus induced cell fusion and cellular senescence, indicating the potential of senescence to limit viral replication [8]. On the other hand, there is growing evidence of viruses using senescent cells to their advantage. The replication efficiency of influenza virus and varicella zoster virus (VZV) was higher in senescent cells compared to non-senescent cells [6]. Senescent cells have also been shown to upregulate the entry receptor for dengue virus and severe acute respiratory syndrome coronavirus 2 (SARS-CoV2), yielding enhanced viral entry and infection [9,10]. Because senescent cells are more resistant to apoptosis and have reduced antiviral signaling, this cellular environment could be conducive for persistent viral infections.
These senescent phenotypes could be exploited by MCPyV, a double-stranded DNA tumor virus that causes a persistent infection. MCPyV is a ubiquitous pathogen that can cause Merkel cell carcinoma (MCC), a rare but aggressive skin cancer that mainly affects immunocompromised patients [11]. The MCPyV large tumor antigen (LT) is required for viral genome replication [12] and the restriction of cell proliferation [13,14]. This protein is unique from LT antigens of other human polyomaviruses in that it contains a disordered MCPyV unique region (MUR) domain [15,16]. The MUR domain aids in establishing viral latency through ubiquitination by host E3 ubiquitin ligases [15,[17][18][19] that destabilize MCPyV LT and result in the inhibition of MCPyV DNA replication. Recently, MCPyV replication has been shown to activate host cell senescence, potentially by a variety of intrinsic and extrinsic stresses, although its exact mechanism remains unknown [20]. Here, we investigate the mechanism of MCPyV LT-induced senescence and the role of the MCPyV LT unique domain in a host cellular stress response during human polyomavirus infection, which may explain a critical interplay between the host and a viral pathogen leading to cancer.

MCPyV LT Mediates Cellular Senescence in Human Fibroblast Cells through Its Unique Domain
During our study of MCPyV LT-mediated cell growth regulation [15], we observed significant structural alterations in normal human fibroblasts BJ-hTERT expressing MCPyV LT, including enlarged nuclei and flattened morphology, which are features associated with cellular senescence. Because enlarged nuclei in MCC cases correlate with MCPyV positivity [21], we explored the possibility that senescence, mediated by MCPyV LT, is an integral part of the MCPyV life cycle. To determine if MCPyV LT expression alone induces cellular senescence, we created lentiviruses that encode either MCPyV LT or SV40 LT fused with 2A self-cleaving peptides (P2As) and enhanced green fluorescent proteins (eGFPs) [15] to transduce BJ-hTERT cells; an eGFP-P2A empty vector lentivirus served as a negative control ( Figure 1A). Because the MUR domain is unique to MCPyV and regulates cell proliferation [15], we asked if this domain had a role in regulating cell morphological changes and, therefore, senescence. To understand the effect of the MUR domain on MCPyV LT-induced morphological changes, a MCPyV LT mutant with a deletion of the MUR domain (MCPyV LT dMUR) [15] and a SV40 LT with the insertion of the MCPyV LT MUR domain (SV40LT+MUR) [15] were also examined. Only the MCPyV LT-expressing cells strongly stained positive for senescence-associated beta-galactosidase (SA-β-gal), the most common senescent marker, at 14 days post-transduction. Fourteen days post-transduction was chosen as we observed the most morphological changes at this timepoint. SA-β-gal staining was determined by both brightfield microscopic ( Figure 1B) and flow cytometric analyses ( Figure 1C). Approximately 50% of MCPyV LT-expressing cells were senescent, whereas only 5% or less of cells were senescent for the remaining LT constructs or empty vector. SA-β-gal-positive cells expressing MCPyV LT had a significant increase in nuclear area, approximately twice the size (>500 µm 2 ) of the empty vector and other LT constructs (225~250 µm 2 ) ( Figure 1D). Minimal staining or morphological changes were observed in cells expressing the empty vector, SV40 LT, and SV40LT+MUR. MCPyV LTdMUR expression greatly reduced SA-β-gal staining and nuclear size, indicating the potential of the MUR domain to promote senescence only in the context of MCPyV LT. Notably, the insertion of the MUR domain into SV40 LT did not promote senescence, suggesting that the MUR domain does not directly stimulate senescence and other factors are likely involved in inducing this phenotype. Analyses of mRNAs for a variety of senescence-associated secretory phenotype (SASPs), in which transcripts for ANKRD1, CSF2, CXCL1, CXCL2, EDN1, IL-6, and IL-8 were markedly upregulated by MCPyV LT expression and only modestly upregulated in MCPyV LTdMUR-expressing cells compared to the eGFP control ( Figure 1E), concluded that the MUR domain is required for the induction of senescence.
Senescence is commonly regulated through the activation of the p53/p21 pathway and is often characterized by prolonged G1 cell cycle arrest [1]. Immunoblot analyses revealed that only MCPyV LT led to p21 upregulation ( Figure 1F). MCPyV LT-positive cells also showed a substantial upregulation of cyclin E, a G1 phase marker [22][23][24][25], indicating G1 cell cycle arrest in a subset of the cell population. In addition, histone H3 phosphorylation at Ser10 (pHH3), a mitotic marker [26], is weakly expressed in the G2 phase [27], and MCPyV LT-positive cells displayed histone H3 phosphorylation at low levels, signifying G2 cell growth arrest [28,29]. In contrast, the MCPyV LTdMUR and SV40 LT+MUR mutants showed no significant p21 or cyclin E upregulation or pHH3 downregulation compared to the eGFP + control cells, implying that growth arrest was specific to wild-type MCPyV LT. Immunofluorescent analyses confirmed that enlarged MCPyV LT senescent cells exhibited high p21 expression ( Figure S1A) and lacked pHH3 expression ( Figure S1B). These results suggest that only MCPyV LT could activate p21, specifically requiring the MUR domain to induce growth arrest and cellular senescence.
Because Siebels et al. reported that MCPyV replication could induce cellular senescence in primary normal human dermal fibroblast (nHDF) cells [20], we also examined the ability of MCPyV LT to induce senescence in other normal human fibroblast cell lines, including, HFF-1, BJ, and neonatal nHDF cells. At 7 days post-transduction, an enlarged nuclear size (~400 µm 2 ) was observed in HFF-1 and BJ cell lines by MCPyV LT expression ( Figure S2A). Unexpectedly, a nonspecific background staining of SA-β-gal was detected in these cells ( Figure S2B). Additionally, senescent phenotypes, including a significantly enlarged nuclear size (>500 µm 2 ) ( Figure S3A) and a drastic upregulation of p21 expression ( Figure S3B) were induced by MCPyV LT expression at 7-and 10-days post-transduction in neonatal nHDF cells, although high levels of basal SA-β-gal activity were also observed ( Figure S3C). Due to the detection of the high nonspecific SA-β-gal staining pattern in these human fibroblast cell lines, we chose to continue utilizing BJ-hTERT cells to most accurately determine the senescent phenotypes induced by MCPyV LT. These results demonstrate that MCPyV LT expression alone can promote cellular senescence in human fibroblast cell lines.

MCPyV LT-Mediated Senescence Is DDR-Independent
Senescent growth arrest is often triggered by a persistent DNA damage response [4]. A previous study has shown that MCPyV LT can induce double-stranded breaks and activate the host cell's DDR through p53 activation [13]. To determine whether MCPyV LT induces DDR-dependent senescence, we looked for the presence of gamma-H2AX (γ-H2AX), a biomarker for DNA double-strand breaks [30]. A subset of MCPyV LTexpressing senescent cells with enlarged nuclei (>500 µm 2 ) were positive for γ-H2AX staining, although some senescent cells were also negative for γ-H2AX staining despite morphologic changes ( Figure 2A). Moreover, MCPyV LT-expressing cells with a normal nuclear size (~200 µm 2 ) were still able to induce a DDR response. Although eGFP, MCPyV LTdMUR, SV40 LT, and SV40 LT+MUR-expressing cells were significantly lower for SAβ-gal staining ( Figure 1B,C), 20-30% of cells were positive for γ-H2AX foci formation regardless of the stimulus ( Figure S4). These results indicate that wild-type MCPyV LT likely mediates cellular senescence through a DDR-independent manner.
Phosphorylation of p53 at serine 15 (S15), an upstream event prior to p21 upregulation, can be induced by the NSR [31]. S15 phosphorylation of p53 was noted in MCPyV LTpositive cells marked by the staining of numerous foci in the nucleus, but it was substantially reduced in MCPyV LTdMUR ( Figure 2E). SV40 LT constructs exemplified phospho-p53 staining ( Figure 2E), but did not induce p21 expression, as shown in Figures 1F and S1A as previously reported in BJ-hTERT and U2OS cells [13,36]. Taken together, these results indicate that MCPyV LT-mediated senescence is associated with the NSR, leading to the activation of the p53/p21 pathway. is DDR-independent. Immunofluorescent staining for γ−H2AX, a DNA damage marker, in senescent (GFP + , nuclear size > 500 µm 2 ) and non−senescent (GFP + , nuclear size =~200 µm 2 ) MCPyV LT−positive BJ−hTERT cells. Scale bar = 60 µm. (B) MCPyV LT reorganizes the nucleolus. MCPyV LT induced a ring−like perinucleolar distribution of NPM1 within the nucleolus, a characteristic of canonical NSR. Immunofluorescent analysis of NPM1 is shown (red). Scale bars = 20 µm. (C) MCPyV LT uniquely alters NPM1 structure. The percentage of NPM1 nucleolar ring−positive cells for the empty vector and each LT construct was recorded. Statistical significance was determined using the one−way ANOVA test (**** p < 0.0001, ns = not significant). Standard error bars represent mean value with standard error, n = 30 cells with 3 replicates. (D) Cells with NPM1 reorganization induced by MCPyV LT display senescent phenotypes. Nuclear size (µm 2 ) of MCPyV LT−expressing cells with or without NPM1 nucleolar rings. Unpaired Student's t-test (**** p < 0.0001, ns = not significant). Standard error bars represent mean value with standard deviation, n = 25 cells per condition with 3 replicates. (E) MCPyV LT expression triggers the NSR and induces p53 phosphorylation through the MUR. Phospho−p53 (S15) is detected in MCPyV LT−expressing cells.

MCPyV LT-Induced Senescence Is p21-Dependent
Because p21 has a key role in cellular senescence, we hypothesized that the loss of p21 expression would block MCPyV LT-induced senescence. To determine whether the MCPyV LT-mediated modulation of p53/p21 is required for the onset of senescence, a knockdown of p21 was performed by short hairpin RNA (shRNA) transduction in BJ-hTERT cells expressing MCPyV LT using two shRNA lentiviral constructs, shp21.1 and shp21.2. Transduction with scrambled shRNA (Scr) was used as a control. The knockdown of p21 was confirmed at the mRNA and protein levels in MCPyV LT-expressing cells by RT-qPCR ( Figure 3A) and immunoblotting ( Figure 3B), respectively. Knockdown using shp21.1 drastically decreased p21 mRNA (~95%) and protein (~75%) levels, while knockdown with shp21.2 reduced p21 mRNA expression to a lesser degree (~40%) and did not decrease p21 protein levels. Dual knockdown with shp21.1 and shp21.2 (shp21.1/2) reduced p21 to similar levels to knockdown with shp21.1 alone. Because it was the most efficient at knocking down p21, the shp21.1 construct (p21KD) was used for the remaining experiments (Figures 4-6). Knockdown of p21 after lentiviral transduction of MCPyV LT reversed cell cycle arrest, as shown by decreased cyclin E and increased pHH3 expression ( Figure 3B). Moreover, p21 knockdown cells no longer exhibited an enlarged and flattened morphology compared to the MCPyV LT-expressing cells transduced with the scrambled shRNA control ( Figure 3C). The nuclear area of MCPyV LT-expressing cells was significantly reduced from 580 µm 2 to~190 µm 2 , approximately the nuclear size of the eGFP control cells ( Figure 3D). SA-β-gal staining confirmed that p21 knockdown largely inhibited MCPyV LT-mediated senescence ( Figures 3E and S5) and NSR ( Figure 3F), concluding that MCPyV LT-induced senescence is p21-dependent.

Senescent MCPyV LT-Expressing Cells Are Arrested in G2 Phase
p21-induced senescence commonly leads to G1 cell cycle arrest [1]. To verify the cell cycle status and further characterize the growth arrest modulated by MCPyV LT, cell cycle and cell proliferation analyses were performed. Flow cytometric analysis showed that MCPyV LT induced G2 cell cycle arrest, as seen by the enrichment of cells in the G2 phase ( Figures 4A and S6) as previously reported [13,20]. Further analysis of MCPyV LT G1and G2-arrested cells revealed that these cells were larger than the control and MCPyV LTdMUR-expressing cells, suggesting that these growth-arrested cells were likely senescent ( Figure 4B). G2-arrested cells were larger than their G1-arrested counterparts, implying that G2-arrested cells were senescent. This possibility was confirmed by the observation that approximately 70% of MCPyV LT G2-arrested cells were positive for SA-β-gal. In contrast, only~20% of MCPyV LTdMUR G2 cells were positive for SA-β-gal ( Figure 4C). As expected, the knockdown of p21 in MCPyV LT-expressing cells induced cell cycle re-entry ( Figure 4A), reduced cell size to eGFP control levels ( Figure 4B), and decreased SA-β-gal expression to near MCPyV LTdMUR levels ( Figure 4C).
MCPyV LT expression has been shown to reduce cell proliferation through either the C-terminus or the MUR domain [13][14][15]. To identify if cellular senescence was a potential cause of growth inhibition through the MUR domain, a HoloMonitor (PHI Lab), a non-invasive live-cell imager and counter was utilized [37]. MCPyV LT-expressing cells proliferated slower than their MCPyV LTdMUR counterparts ( Figure 4D). Additionally, the incorporation of 5-ethynyl-2 -deoxyuridine (EdU) by MCPyV LT-expressing cells was markedly reduced compared to the MCPyV LTdMUR mutant ( Figures 4E and S7). Knockdown of p21 in MCPyV LT-expressing cells reversed growth inhibition induced by MCPyV LT expression ( Figure 4D,E). Although p21-dependent senescence often leads to G1 cell cycle arrest [1], our results indicate that MCPyV LT-mediated p21 activation mainly facilitated G2-arrested cellular senescence, negatively regulating cell growth.

MCPyV LT-Induced Senescence Alters Host Gene Expression Profile
Senescent cells are characterized by the enhanced secretion of cytokines known as SASPs [1]. To further characterize MCPyV LT-induced senescence, RNA sequencing was performed ( Figure 5). Multidimensional analyses indicated that replicates were transcriptionally similar and clustered together ( Figure S8A). Of note, cells expressing eGFP or dMUR had similar transcriptomes, further demonstrating the importance of the MUR domain on LT function ( Figure S8A,B). Additionally, p21 knockdown drastically altered gene expression. Volcano plots illustrated many differentially expressed genes between MCPyV LT-and LTdMUR-expressing cells and in MCPyV LT-expressing cells with p21KD ( Figure S8C). A heatmap containing the top 100 differentially expressed genes (DEGs) was generated, comparing eGFP, MCPyV LT, MCPyV LTdMUR, and MCPyV LT+p21KD ( Figure 5A, Supplementary Dataset S1). Numerous SASPs, including CXCL1, CXCL2, CXCL5, IL-1a, IL-1b, IL-6, IL-8, SAA1, and SAA2, were identified on this heatmap and MCPyV LT led to a modest upregulation of some of these SASPs compared to the MCPyV LTdMUR and eGFP control. Unexpectedly, p21KD led to a dramatic increase in the expression of inflammatory cytokines, potentially due to cell death induced by p21KD after lentiviral transduction.
To further investigate the changes in gene expression due to MCPyV LT, the MUR domain, and p21-dependent senescence, we generated heatmaps of the top 50 differentially expressed genes (DEGs), comparing MCPyV LT to eGFP, LTdMUR, and p21KD, respectively ( Figure 5B, Supplementary Dataset S1). SASPs and senescence genes commonly found as a top DEG were examined. Log 2 fold change of genes and their corresponding p values compared to eGFP were plotted ( Figure 5C). Cyclin E transcripts were significantly upregulated compared to the empty vector and the dMUR mutant (p value < 0.05). Significant SASPs that were upregulated in MCPyV LT-expressing cells compared to eGFP and dMUR mutant cells included SAA1, SAA2, CXCL1, and CXCL5. Pathway analyses indicated that MCPyV LT-expressing cells were enriched in senescence-related pathways and downregulated in rRNA and ribosomal metabolic pathways compared to p21KD cells ( Figure 5D), reinforcing the conclusion that MCPyV LT promotes the NSR and activates p21-mediated senescence [38]. Genes found in the enriched senescent pathways included numerous histone variants (Supplementary Dataset S1), which have been previously associated with senescence [39]. Some of the expressed aging/senescence-related genes found highly upregulated in MCPyV LT + cells (SULF2 [40], namely MFAP5 [41], KISS1 [42], and APOE [43,44]) were inversely regulated by p21KD ( Figure 5C,D). Black dashed line represents PE positive control. Statistical significance was determined using the one−way ANOVA test (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant). Standard error bars represent mean value with standard error. A total of 10,000 events were analyzed for each sample, n = 3. (D) MCPyV LT restricts cell proliferation by inducing senescence. Non−invasive live cell imaging of transduced BJ−hTERT cells was conducted using a phase holographic imaging system (HoloMonitor M4) to measure cell proliferation rates. (E) MCPyV LT inhibits cell proliferation. Cell proliferation assay. EdU incorporation (pink) was measured by fluorescence microscopy. Scale bar = 60 µ m.

MCPyV LT-Induced Senescence Alters Host Gene Expression Profile
Senescent cells are characterized by the enhanced secretion of cytokines known as SASPs [1]. To further characterize MCPyV LT-induced senescence, RNA sequencing was performed ( Figure 5). Multidimensional analyses indicated that replicates were transcriptionally similar and clustered together ( Figure S8A). Of note, cells expressing eGFP or dMUR had similar transcriptomes, further demonstrating the importance of the MUR domain on LT function ( Figure S8A,B). Additionally, p21 knockdown drastically altered gene expression. Volcano plots illustrated many differentially expressed genes between MCPyV LT-and LTdMUR-expressing cells and in MCPyV LT-expressing cells

MCPyV LT-Mediated Senescence Maintains MCPyV Genome
As senescent cells are characterized by a stable cell cycle arrest and are resistant to apoptosis [4], we asked if MCPyV could exploit cellular senescence to establish a persistent viral infection. To test the hypothesis that senescence is required for the longevity of MCPyV infection, and to do so in the context of an actual MCPyV molecular clone, we transfected BJ-hTERT cells with a full-length MCPyV genome (pMCPyV-MC) in which the early genes were tagged with eGFP-P2A ( Figure 6A). Senescent MCPyV cells were present at 21 days post-transfection, whereas knockdown with p21 greatly abolished SAβ-gal staining (GFP + /SA-β-Gal + ) as determined by both flow cytometry and microscopic analyses ( Figures 6A,B and S9).
To determine if senescence is required for MCPyV genome maintenance, we reversely co-transfected BJ-hTERT cells with the MCPyV genome and scrambled shRNA lentiviral vector (mCherry-Puro shScr) or p21 shRNA vector (mCherry-Puro shp21.1). Cells were initially transfected with shp21.1 to inhibit senescence as we sought to determine how the loss of senescence impacts the long-term persistence of MCPyV. We observed MCPyV genome maintenance using flow cytometry by detecting GFP levels over time ( Figure 6C). Because MCPyV LT induced G2 cell cycle arrest ( Figure 4A), we specifically examined the G2 population. Cells were gated on GFP and mCherry (GFP + mCherry + ) to ensure that only cells that had taken up both MCPyV genome and shRNA plasmid were examined ( Figure 6C). At day 14 post-transfection, p21 knockdown reduced the percentage of GFP + (MCPyV + ) cells by~25% compared to scrambled control cells. Additionally, p21 knockdown could consistently reduce MCPyV genome levels at both 10-and 14-days post-transfection ( Figure 6D), suggesting that MCPyV genome persistence is dependent on cellular senescence.
Previous studies have identified senolytics as drugs that can selectively clear senescent cells to reduce disease morbidity [1,10,45]. Therefore, we investigated whether senolytic treatment could potentially act as a therapeutic agent to prevent MCC. At fourteen days post-transfection, cells were treated with navitoclax, dasatinib and quercetin (D+Q), or DMSO for two days. To determine if these senolytics could reduce the senescent cell burden, a SA-β-Gal assay was conducted. Navitoclax treatment [46] could markedly reduce SAβ-Gal staining, whereas D+Q treatment did not show a significant change in SA-β-Gal staining ( Figures 6E and S10). Moreover, navitoclax could reduce viral genomes levels, decreasing MCPyV DNA levels by~70% ( Figure 6F), demonstrating the importance of cellular senescence on MCPyV genome persistence.

Discussion
Taken together, our results show that MCPyV infection, mainly MCPyV LT expression, can activate a nucleolar stress response in human fibroblast cells. This, in turn, activated p53 transcriptional activity, inducing p21 expression. p21 upregulation subsequently led to G2 cell cycle arrest, promoting cellular senescence. MCPyV was able to exploit these senescent cells by maintaining its viral genome for long-term survival ( Figure 6G).
MCPyV-induced senescence was recently reported by Siebels et al. [20]. Although their MCPyV LT interaction study revealed an interaction with KRAB-associated protein 1 (KAP1), a senescence-related factor [47] in primary normal human dermal fibroblasts, Siebels et al. observed neither senescence nor p21 transcriptional activation assessed in primary nHDFs in 2-4 days of LT expression alone. Our data suggest that LT expression alone induces host cell senescence with significant morphologic changes in human fibroblasts, including BJ-hTERT (Figure 1), HFF-1, BJ, and nHDF ( Figures S2A and S3A) at least 7 days post-transduction. An upregulation of p21 was also observed in neonatal nHDF cells at 10 days post-transduction ( Figure S3B). These data suggest that MCPyV LT expression alone can gradually promote cellular senescence at later time points (≥7 days post-transduction). As an unexpected high SA-β-gal-positive background was observed in early passages of non-immortalized skin fibroblast cells ( Figures S2B and S3C), we used BJ-hTERT, a skin fibroblast cell line commonly used to study aging and senescence [48][49][50], for all our experiments to determine accurate gene changes in MCPyV LT-induced senescence. Although BJ-hTERT cells express human telomerase reverse transcriptase (hTERT), which is known to inhibit senescence, it has been shown that hTERT has no effect on stress-induced senescence [51], but only inhibits stress-related apoptosis and necrosis [52].
In a latent viral infection, while the full viral genome is retained in the host cell, its late gene expression is dramatically restricted and only few viral antigens and no viral particles are produced. To qualify as latency, this quiescent state of infection must display persistence and reversibility [53]. Protein-mediated viral latency is the only known mechanism for human polyomavirus latency, shown by successful reversibility from a quiescent state to a productive infection that is regulated by MCPyV LT proteostasis [17]. Accompanied by protein interaction networks [15,17,20,54], we show that an additional key nucleolar stress pathway, triggered by MCPyV LT overexpression, plays a pivotal role in persistent viral infection.
Although viral replicative machinery has been associated with modulating the nucleolus function, very few DNA viral oncoproteins have been specifically characterized to trigger changes in nucleolar distribution [38,55]. Canonically, the induction of the nucleolar stress response is caused by a disruption of ribosomal biogenesis [32]. Though our RNA-Seq data suggest MCPyV LT can downregulate rRNA synthesis, the mechanism of NSR induction by MCPyV LT needs further investigation. Certainly, other stress responses in different cell types, such as DDR, shortened telomeres, replicative stresses, and/or apoptosis, will combinedly influence the extent of senescence in the early and late stages of the infection process. Nonetheless, it is important that both persistency and reversibility are mainly modulated by the MCPyV LT unique domain that is potentially involved in a series of host interactions and the regulation of its own proteostasis [15,17,56,57].
Most of the known human cancer viruses have the ability to antagonize p53 by expressing viral oncoproteins that promote cellular proliferation by abrogating p53-induced cell cycle arrest in response to DNA damage [58]. Human polyomavirus LTs have been known to interact with p53, which immediately blocks DDR and innate immune sensing for the host's survival in the initial infection process [59]. Unlike other human polyomaviruses, MCPyV LT does not seem to possess the ability to interact and directly inhibit p53 [14], but rather activate p53 phosphorylation and function [13]. This activation of p21 transcription by p53, uniquely induced by MCPyV LT, may establish a distinct niche of MCPyV as a human-cancer-causing virus among human polyomaviruses. Low mutation rates of p53 [54,60] and an intact p53/p21 pathway [61] in MCCs consistently support our results that the activation of p21-dependent senescence potentially plays a role in MCPyV infection.
It should be noted that the LT-induced senescence response in human fibroblasts has been simultaneously evolving throughout the passage of time, potentially due to changes in miRNA and gene expression profiles [62], suggesting a variety of gene expression changes can alter and allow host cellular responses to support latent infection periods. Specifically, we observed that late passages of BJ-hTERT cells supported enhanced senescence and reduced cell death with LT expression and p21 knockdown conditions compared to early passaged cells. We also noticed a significant level of mitotic cell death as an early event induced by p21 knockdown, although cells recover over time. This p21 knockdownmediated cell death may be caused by the re-entry of senescent cells back into the cell cycle, which enhances the activation of DDR as shown by others [63].
Senescence has been characterized as an irreversible exit of the cell cycle [64]. Given that certain viruses depend on host cell proliferation for viral replication, viral oncogenemediated cell senescence has been suggested to be an antiviral mechanism for host cells. Our results define a central role of senescence as a potential contributing factor to viral oncogenesis that has not been previously described. Despite this cell-intrinsic tumor suppression, senescent cells have also been implicated as active contributors to tumorigenesis by extrinsically promoting many hallmarks of cancer, including evasion of the immune system [65]. Through the induction of cellular senescence, MCPyV potentially creates a precancerous environment in which it can avoid the immune system that enhances cell survival and viral persistence. The senolytic drug navitoclax is a Bcl-2 anti-apoptotic family protein inhibitor. Early studies on navitoclax have shown promising results to suppress MCC growth [66,67], suggesting the Bcl-2 pathway as a potential therapeutic target to prevent MCPyV infection and MCC development. Because senescent cells are heterogeneous [1], senolytics that utilize different mechanisms of action are able to target senescent cells and the detrimental effects of navitoclax on LT-related cellular senescence remain to be further investigated.
Although whether latent viral infection can ultimately lead to oncogenesis by inducing host senescence and other stress environments needs further investigation, such cellular responses surely contribute to maintaining cellular homeostasis during virus infections. Our results uncover the mechanism of a host stress response in regulating human polyomavirus genome maintenance in a persistent viral infection, which may lead to targeted interventions for MCC.

Senescence Assay
The detection of SA-β-Gal activity was performed using a colorimetric senescence β-Galactosidase histochemical staining kit (G-Biosciences, St. Louis, MO, USA) or a fluorescent Cell Meter™ Cellular Senescence Activity Assay Kit with Xite™ Red beta-D-galactopyranoside (AAT Bioquest, Pleasanton, CA, USA) according to the manufacturer's instructions. Briefly, the cells were washed with cold PBS and then fixed with 4% paraformaldehyde in PBS for 15 min. Cells were then permeabilized for 15 min with 0.1% Triton in PBS or a permeabilization buffer (Thermo Scientific, Rockford, IL, USA). After incubation with the SA-β-Gal detection solution for 12 h (G-Biosciences) or 45 min (AAT Bioquest) at 37 • C, the cells were stained with DAPI (Thermo Scientific) for microscopic analyses or Hoechst 33342 (Invitrogen) for flow cytometric analyses. For LT-transduced BJ-hTERT cells, senescence assays were performed at 14 days or 21 days post-transduction for samples with or without p21 knockdown, respectively. Normal and primary fibroblast cell lines were stained for SA-β-Gal at 7 days post-transduction. MCPyV genometransfected cells were transduced with shp21.1 at 14 days post-transfection. At 21 days post-transfection, a senescence assay was conducted. Images were captured using a RE-VOLVE4 fluorescent microscope (Echo Laboratories). All the samples were stained at least in triplicate. A total of 50 cells were counted for each LT construct. The nuclear area was calculated using Echo Pro software (Echo Laboratories). A positive PE signal was detected using the PE-Annexin V (BD Biosciences, San Jose, CA, USA) or CS&T Research Beads (BD Biosciences, 655050). For PE-positive control samples, BJ-hTERT cells were washed twice with PBS and then heat-killed by incubation at 95 • C for 3 min. A 1:1 cell suspension of heat-killed and non-heat-killed cells were stained according to the manufacturer's instructions. A total of 10,000 events were recorded in triplicate for each sample.

SASP Analysis by RT-qPCR
For RT-qPCR, total RNA was isolated using the Monarch Total RNA Miniprep Kit (New England Biolabs, Ipswich, MA, USA) according to the manufacturer's instructions. mRNA levels of SASPs ( Figure 1E) were measured by RT-qPCR using an iTaq Universal One-Step RT-qPCR Kit (Bio-Rad, Hercules, CA, USA) with primer pairs as previously described [68]. Quantitative analyses were performed using the comparative ∆∆Ct method by detecting ribonuclease P (RNase P) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as reference genes using the primer pairs listed in Table S1. All RT-qPCR experiments included melting curve analyses to confirm the specificity of the amplicons (95 • C for 15 s, 60 • C for 20 s, and 95 • C for 15 s).

Cell Cycle Analysis
BJ-hTERT cells were fixed, permeabilized, and then stained with Hoechst 33342 (Invitrogen) to determine cell cycle stages. Cells were analyzed using an LSR Fortessa flow cytometer (BD Biosciences) and 10,000 events were recorded per replicate. FlowJo (TreeStar, Ashland, OR, USA) software was utilized for data analyses. To examine viable cells, forward scatter (FSC) area versus the side scatter (SSC) area density plots were analyzed. Doublets were removed through the FSC area versus the FSC height gating. Cell cycle analyses were conducted on GFP-positive cells to examine only the transduced cells.

EdU Incorporation Assay
Cells were pulse labelled with EdU (final concentration of 10 µM, Invitrogen) 24 h before fixation. EdU incorporation was detected using the Click-iT Plus EdU Alexa 647 Fluor Imaging Kit (Invitrogen) according to the manufacturer's instructions. Cells were then stained with DAPI (Thermo Scientific) and visualized with a fluorescence microscope (REVOLVE4, Echo Laboratories).

Cell Proliferation Assay
Cells were plated on 6-well, TC plates (Thermo Scientific) at a confluency of approximately 30%, and they were incubated with 10% serum. After 24 h, cells were recorded using a phase holographic imaging system (HoloMonitor M4 laser microscope, Phase Holographic Imaging, Lund, Sweden) [37]. Images were recorded every 30 min for 48 h. Average cell proliferation was quantified using HStudio M4 software (version 2.7.1).

RNA Sequencing Analysis
RNA sequencing and gene reads were aligned to the hg38 human genome. Raw gene counts were used for integrated differential expression and pathway (iDEP) analyses and were transformed and normalized using iDEP [69]. All figures and sequence analyses were generated from iDEP or Zymo Research. Genes below 0.5 counts per million (CPM) were excluded and the false-discovery rate (FDR) for differential gene expression was kept below 0.1 (FDR < 0.1) with a minimum fold change of 1.5. RNA sequence data are available at the NCBI Gene Expression Omnibus (GEO) under the following accession number: GSE189291.

Flow Cytometric Analysis of MCPyV Genome
The MCPyV genome [70] with eGFP (pMCPyV-MC, Figure S11) was inserted into a minicircle (MC) parental backbone plasmid pTubb3-MC (Addgene #87112) using SmaI/Bsr GI/BstEI sites and isolated according to the manufacturer's instructions (System Biosciences, Palo Alto, CA, USA). BJ-hTERT cells were reverse-transfected with pMCPyV-MC and either mCherry-Puro shScr or mCherry-Puro shp21.1 to inhibit senescence and to determine its impact on long-term viral maintenance. At 10-and 14-days post-transfection, cells were harvested and subjected to flow cytometry to measure GFP and mCherry levels. The detection of mCherry indicated the expression of p21 shRNA constructs and the observation of GFP indicated the presence of the MCPyV genome. Cells were first gated on mCherry, which was followed by the gating of GFP to ascertain how p21 knockdown affected MCPyV persistence. The gating of GFP and then mCherry was used to ensure MCPyV-infected cells had taken up shRNA constructs. MCPyV viral genome levels of BJ-hTERT reversely co-transfected with mCherry-Puro shScr or mCherry-Puro shp21.1 were also calculated utilizing qPCR with 20 ng of DpnI digested DNA, isolated using Quick-DNA kit (Zymo Research, Irvine, CA, USA). qPCR was carried out with PowerUpTM SYBR Green Master Mix (Applied Biosystems, Foster City, CA, USA) using a StepOnePlusTM system (Applied Biosystems) according to the manufacturer's protocol. Quantitative analyses were performed using the comparative ∆∆Ct method by detecting RNase P as a reference gene and three MCPyV detection primer pairs (Table S1). All qPCR experiments included melting curve analyses to confirm the specificity of the amplicons (95 • C for 15 s, 60 • C for 20 s, and 95 • C for 15 s).

Senolytic Treatment
pMCPyV-MC or pTubb3-MC (Addgene #87112) transfected cells were maintained for 14 days to yield the establishment of senescent phenotypes. Cells were plated onto 24-well or 12-well plates for the SA-β-Gal assay or qPCR analysis, respectively, and grown for one day. Navitoclax (250 nM), D+Q (25 nM dasatinib plus 250 nM quercetin) (Selleckchem), or DMSO (Sigma-Aldrich) were added to the cells for two days. Senolytic-treated cells were then subjected to a SA-β-Gal assay or qPCR to detect viral genome levels, in which 25 ng of undigested DNA was analyzed.

Statistical Analysis
The figures show the average values and the error bars show the standard error or standard deviation. One-way analysis of variance (ANOVA), multiple Student's t-tests, or unpaired Student's t-test was used to determine statistical significance, with p < 0.05 considered significant using GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA, USA). Unless stated, all experiments were tested at least three times.