Transcription factors and stress response gene alterations in human keratinocytes following Solar Simulated Ultra Violet Radiation

Ultraviolet radiation (UVR) from sunlight is the major effector for skin aging and carcinogenesis. However, genes and pathways altered by solar-simulated UVR (ssUVR), a mixture of UVA and UVB, are not well characterized. Here we report global changes in gene expression as well as associated pathways and upstream transcription factors in human keratinocytes exposed to ssUVR. Human HaCaT keratinocytes were exposed to either a single dose or 5 repetitive doses of ssUVR. Comprehensive analyses of gene expression profiles as well as functional annotation were performed at 24 hours post irradiation. Our results revealed that ssUVR modulated genes with diverse cellular functions changed in a dose-dependent manner. Gene expression in cells exposed to a single dose of ssUVR differed significantly from those that underwent repetitive exposures. While single ssUVR caused a significant inhibition in genes involved in cell cycle progression, especially G2/M checkpoint and mitotic regulation, repetitive ssUVR led to extensive changes in genes related to cell signaling and metabolism. We have also identified a panel of ssUVR target genes that exhibited persistent changes in gene expression even at 1 week after irradiation. These results revealed a complex network of transcriptional regulators and pathways that orchestrate the cellular response to ssUVR.


Results
Transcriptional profiling of gene expression of HaCaT cells in response to ssUVR. Immortalized human keratinocyte HaCaT cells were exposed to either a single, high dose of ssUVR (12 J/cm 2 ) or 5 repetitive doses of ssUVR (once every three days) at 3, 6 or 12 J/cm 2 (Fig. 1A). The range of doses used in the study is comparable to what a person would be exposed to when standing outside for 40 minutes (3 J/cm 2 ) to 2.5 hours (12 J/cm 2 ) at noon under a clear sky with a UV index of 6 or higher 24 . Our previous study found that HaCaT cells exposed to a single dose of 12 J/cm 2 exhibited a moderate increase of phosphorylated histone variant H2AX, a well-characterized marker of DNA damage, at 1 hour post irradiation, as well as limited cell death (less than 20%) at 24 hours after irradiation 24 . Irradiated cells were harvested 24 hours post irradiation and subjected to transcriptional profiling. To mimic human skin repetitively exposed to ssUVR, cells were exposed to 5 repetitive doses of ssUVR at 3, 6, 12 J/cm 2 and were collected at 24 hours post irradiation. To examine the persistence of gene expression changes, cells were exposed to 5 repetitive doses of ssUVR at 12 J/cm 2 , and were collected at 1 week after the last irradiation. Two biological replicates of controls (sham) or irradiated cells (UVR) were collected for each dose or time point.
Characterization of gene expression changes induced by ssUVR was carried out by RNA-seq. Using the Illumina HiSeq. 2500 platform, a total of 0.6 billion reads were obtained from 16 RNA samples with an average of 38.2 million reads per sample. More than 90% of reads were mapped to human genome (Hg38). Data analysis was performed using Biomedical Genomics Workbench version 3.5.3 (Qiagen). Differential gene expression was analyzed using Advance RNA-seq plug-in tool (Qiagen) by comparing each treated group versus corresponding control group. The genes with a false discovery rate (FDR) < 0.05 between control and treated group and an average mean of total counts no less than 10 were defined as differentially expressed genes (DEGs). A total of 5 DEG lists were generated to represent genes changed in the different treatments (Supplemental Table 1). The numbers and the direction of regulation (Up or down) of DEGs in different exposure groups are summarized in Fig. 1B. In cells exposed to a single ssUVR at 12 J/cm 2 , a total of 506 DEGs genes were identified, of which 223 genes were up-regulated and 283 down-regulated. In cells exposed to 5 repetitive doses of ssUVR at 12 J/cm 2 , the total number of DEGs was increased to 1322, including 611 up-regulated and 711 down-regulated genes. In contrast, cells collected at 1 week post irradiation exhibited a reduced number of DEGs (n = 821) compared to those at 24 hours post irradiation. The number of DEGs was positively correlated with the exposure frequency of ssUVR, but negatively correlated with the recovery time. Moreover, classification of the protein functions of DEGs using Ingenuity Pathway Analysis (IPA) revealed that enzymes, transcriptional regulators, kinases, and transporters are the top represented categories in all five DEG lists (Supplemental Fig. 1).
In order to validate RNA-seq results, 4 genes were selected (PMEPA1, DDIT4, TXNIP, KRT4) based on their expression changes. While PMEPA1 is the top upregulated gene in cells exposed to either a single dose ssUVR or 5 repetitive ssUVR, DDIT4, TXNIP and KRT4 are among top 10 downregulated genes in both treatments (Supplemental Table 1). As shown in Fig. 1C, although the fold change of each gene from RT-qPCR is slightly different to those obtained from RNA-seq, the trend of gene regulation remains the same. Thus, the RNA-seq results indicate that ssUVR alters the global gene expression profile.
Functional analysis of DEGs in HaCaT cells exposed to single ssUVR. To uncover the molecules and biological pathways associated with cellular responses to ssUVR, Gene Ontology (GO) for the significant DEGs was analyzed using Database for Annotation, Visualization and Integrated Discovery (DAVID), an NCBI SCIENTIfIC REPORTS | 7: 13622 | DOI:10.1038/s41598-017-13765-7 web-based functional annotation tool. The top 10 highly represented biological processes of down-regulated genes, shown in Fig. 2A, are mainly related to mitotic cell cycle regulation and glycolysis, with each group containing 4/10 and 2/10 GO terms respectively. The four GO terms that are related to mitotic cell cycle regulation are mitotic nuclear division (p = 1.71 × 10 -17 ), mitotic cytokinesis (p = 2.93 × 10 −10 ), sister chromatid cohesion segregation (p = 4.02 × 10 −8 ) and spindle organization (p = 2.63 × 10 −6 ); while the two GO terms that are related to glycolysis are glycolytic process (p = 2.72 × 10 −8 ) and canonical glycolysis (p = 1.86 × 10 −6 ). In contrast, GO terms from genes that were up-regulated by ssUVR are quite diverse. The top three GO terms for the upregulated genes are rRNA processing (p = 6.32 × 10 −6 ), stress response (p = 3.31 × 10 −5 ) and extracellular matrix organization (p = 3.31 × 10 −5 ) ( Fig. 2A).
Ingenuity Pathway Analysis (IPA) was used to identify major pathways that were enriched in the DEG sets. Consistent with DAVID analysis, pathways related to cell cycle regulation were overrepresented (Fig. 2B). The top two canonical pathways are G2/M phase DNA damage checkpoint regulation (AURKA, BORA, CCNB1,  CCNB2, CDC25C, CDK1, CKS2, CKS1B, PLK1, TOP2A) and mitotic roles of Polo-like kinase (CCNB1, CCNB2,  CDC20, CDC25C, KIF23, PLK1, PRC1, PTTG1, RAD1). In addition, the genes involved in G1/S checkpoint regulation, regulation of cellular mechanics by calpain protease, and ATM signaling were also overrepresented. Interestingly, the majority of genes related to cell cycle regulation are down-regulated, especially those involved in mitotic-related spindle assembly checkpoint, such as BUB1, BUB1B, CCNB1, CCNB2, CDC20, CDK1, PLK1, PTTG1. Other highly enriched pathways are those related to cell adhesion and important cell signaling pathways, such as Neuregulin signaling, NF-kB signaling, IL-1 signaling, death receptor signaling and p38 MAPK signaling.
In addition to canonical pathway analysis, upstream regulator analysis (IPA) was used to predict the activation or inhibition of upstream regulators based on the Ingenuity Knowledge Base. Table 1 lists the transcription factors that were predicted to be either activated or inhibited in upstream regulator analysis. HaCaT cells exposed to single ssUVR exhibited significant inhibition of FOXM1, ATF6, FOXO1, MITF, ATF4, and EHF. FOXM1 is a transcription factor regulating the expression of genes essential for DNA damage repair, cell cycle progression, and mitosis 27,28 . ATF4 and ATF6 are transcription factors that regulate expression of genes involved in endoplasmic reticulum (ER) stress response 29 . While ATF6 acts as a sensor for ER stress and controls the expression of genes required for the unfolded protein response, ATF4 plays important role in restoring ER homeostasis and regulation of genes involved in autophagy [29][30][31] . Moreover, KDM5B, TP63, SMAD4, CDKN2A, NUPR1, Notch1 and STAT3 are predicted to be highly activated regulators (Table 1).
Taken together, the analyses of enriched pathways and upstream regulators revealed a significant inhibition of cell cycle progression and the expression of stress-related genes are among the most important events occurred at 24 hours after ssUVR.
Functional analysis of DEGs in HaCaT cells exposed to 5 repetitive ssUVR. In HaCaT cells exposed to 5 repetitive doses of ssUVR, the number of DEGs was correlated to the increased intensity of ssUVR. While cells exposed to 3 J/cm 2 had 159 genes significantly changed, the numbers increased to 991 in cells exposed to 6 J/cm 2 and 1322 in those with 12 J/cm 2 (Supplemental Table 1). Among all of the genes changed after 5 repetitive exposures to ssUVR, 60% of the DEGs identified overlapped in two or three dose groups. The ratio of overlapping genes to total changed genes were 92% in the group exposed to 3 J/cm2, 89% in those exposed to 6 J/cm2, and 65% in cells exposed to 12 J/cm 2 (Fig. 3A). Moreover, 92 out of 112 genes that overlapped in all three dose groups exhibited a dose-dependent fold change (FC). Among these genes, dose-dependent changes of 10 representative genes (5 up-and 5 down-regulated genes) are illustrated in Fig. 3B. Thus, the results suggest that not only the number of DEGs but also the fold change of each DEG was correlated to increased intensity of ssUVR.
To study the functional relevance, the DEGs identified in the highest dose of ssUVR (12 J/cm 2 ) were subjected to DAVID annotation and IPA analysis. As shown in Fig. 4A, while cell cycle regulation remains the major GO term category in down-regulated genes, including the terms mitotic nuclear division (p = 1.06 × 10 −9 ), chromosome segregation (p = 4.16 × 10 −5 ), and mitotic cytokinesis (p = 7.68 × 10 −5 ); the rest of top 10 GO terms represent a diverse biological function, including cell-cell adhesion (p = 2.18 × 10 −17 ), type I interferon signaling (p = 8.01 × 10 −8 ), DNA damage response (p = 1.69 × 10 −5 ), epithelial cell differentiation (p = 2.63 × 10 −7 ), and epidermis development (p = 7.53 × 10 −5 ). Surprisingly, the majority of highly enriched GO terms in the up-regulated genes are related to protein translation, including rRNA processing (p = 1.66 × 10 −20 ) and ribosomal biogenesis (p = 6.83 × 10 −6 ). Canonical pathway analysis revealed that integrin-linked kinase (ILK) signaling, an important pathway mediating cell-cell and cell-matrix interaction 32 , was the top overrepresented canonical pathway for this exposure group (Fig. 4B). Several members of integrin receptors (ITGA2, ITGAV, ITGB1, ITGB5 and ITGB6) were increased in ssUV irradiated cells. Moreover, the calculated z-score predicted a significant inhibition of genes involved in interferon signaling as well as nuclear receptor-mediated pathways (LXR/RXR and PPAR/RXR). Interestingly, a significant downregulation of both type I and type II IFN pathways in cells exposed to repetitive ssUVR was observed. Both STAT1 and STAT2 were reduced in cells exposed to ssUVR, accompanied by the downregulation of IFN target genes, including IRF1 and IFITM1-3 (Supplemental Fig. 2). Lastly, upstream regulator analysis revealed a significant inhibition of TP73 and Nrf2 activity along with the activation of c-Myc, ATF3 and CBX5, which represented altered upstream transcriptional regulation in the cells exposed to 5 repetitive ssUVR ( Table 2). Comparisons of genes and pathways in response to different ssUVR. To examine the persistence of the changes in gene expression by ssUVR exposure, gene expression changes were analyzed in cells that were collected at one week after the 5 th ssUVR exposure (12 J/cm 2 ). A total of 821 genes were altered with treatment, including 505 up-regulated genes and 316 down-regulated genes (Supplemental Table 1). Interestingly, about 400 DEGs from the 1 week recovery group were overlapped with those from cells collected 24 hours post irradiation, suggesting that nearly half of ssUVR-induced gene alterations persist for several cell generations. Cross-comparison of the 3 DEG lists yielded a total of 167 genes that were commonly altered in all three sample sets (Fig. 5A). After removing 8 genes that were regulated in an opposite direction, we obtained a list of 159 DEGs (84 up-regulated and 75 down-regulated) that were altered by both single and repetitive ssUVR and retained the persistent change even at 1 week after irradiation (Supplemental Table 2). A comparative analysis of the three DEG list was performed using the IPA comparison analysis tool in order to further investigate the alteration of genes or pathways in cells exposed to single or 5 repetitive ssUVR as well as cells undergoing a 1-week recovery. As shown in Fig. 5B, the comparison analysis of upstream regulators revealed two distinct groups of transcription factors. The first group is overrepresented in cells exposed to single ssUVR, and included TP63, E2F4, FOXM1, NUPR1 and HIF1A. Among these TFs, TP63 and E2F remained highly represented in cells with 5x ssUVR, while others were less prominent. In contrast, the second group of TFs were more likely changed only after repetitive ssUVR. The transcription factors in this group contain MYC, KDM5B, E2F3, JUN and SP1. Interestingly, these TFs seemed less represented in cells following 1 week recovery, suggesting a relatively transient TF activity after ssUVR. The shift in expression of transcriptional regulators from group I to group II may reveal the transition of cellular response from early cell cycle arrest and stress response toward to a more sophisticated adaptation. Comparative analysis of canonical pathways also revealed the transition of cellular Figure 3. Differentially expressed genes in cells exposed to 5 repetitive ssUVR at 3, 6, 12 J/cm 2 . (A) Venn diagram of differentially expressed genes in cells exposed to 3, 6, and 12 J/cm ssUVR. (B) Illustration showing dose-dependent changes of 5-upregulated and 5-downregulated genes in cells exposed to 5 repetitive ssUVR. response from cell cycle regulation toward metabolic changes mediated by AHR/NRF2 (Fig. 5C). It is worth noting that, unfolded protein response, CDK5 signaling and ERK/MAPK signaling were overrepresented in the 1 week recovery group.

Discussion
Despite many studies on UVA/B/C radiation-induced changes in gene expression, the molecules and pathways involved in the cellular response to low dose sunlight are incompletely understood. In the present study, we investigated the genes and pathways altered in response to single dose or repetitive solar simulated UV irradiation of HaCaT keratinocytes using RNA-seq and subsequent functional annotation tools. These results reveal that ssUVR is able to up-or down-regulate genes with diverse cellular functions in a dose-dependent manner. While cells exposed to single dose of ssUVR displayed a significant inhibition in genes related to cell cycle progression, especially mitotic related genes, cells that were exposed to repetitive doses of ssUVR exhibited changes in genes involved in signal transduction, cell structure and metabolism. In addition, we identified a list of ssUVR target genes, which were altered early in the response to ssUVR and maintained the changes for several cell generations post irradiation.
Mammalian cells have developed a sophisticated response to protect themselves against DNA damage induced by UV radiation. Multiple checkpoints are activated during the cell cycle to pause cell division at a specific phase until the damage repair is completed or bypassed 33 . While G1/S and G2/M checkpoints control the entry into S-phase and M-phase respectively, intra-S and mitotic spindle checkpoints maintain the fidelity of DNA replication and chromosome segregation. These checkpoints are controlled by distinct yet interconnected pathways and gene networks. Our study revealed that DEGs related to cell cycle regulation were overrepresented in keratinocytes exposed to a single dose of ssUVR. Although the changes covered all four checkpoints and cell cycle phases, the majority of down-regulated genes were related to G2/M arrest and mitotic spindle checkpoint regulation.
Transition from G2 phase into mitosis is driven by CDK1 (Cdc2) and its regulator cyclin B. During G2 phase, CDK1 is inactivated by phosphorylation of T14 and Y15 by protein kinases WEE1 and MYT1 34,35 . Activation of CDK-Cyclin B involves dephosphorylation of CDK1 by CDC25 phosphatase 36 and phosphorylation of Cyclin B by PLK1 37 . PLK1 is a serine/threonine protein kinase that plays important roles in mitosis as well as DNA damage response 38,39 . During G2/M transition, PLK1 is activated by Aurora A kinase and its cofactor Bora 40 . Upon detection of DNA damage, activation of the ATM/ATR/CHK1/CHK2 signaling cascade leads to inhibition of CDC25C and BORA/ Aurora A/PLK1, which subsequently inactivates CDK1-Cyclin B leading to G2/M cell cycle arrest 38,39 . Although most of the regulatory events occur through phosphorylation and dephosporylation, it was previously reported that CDC25C protein levels were reduced in HaCaT cells exposed to UVB 41 . Our results showed a significant down-regulation of these genes in ssUV-irradiated HaCaT cells. It is worth noting that HaCaT carries p53 mutations in both alleles and therefore does not have a functional p53 protein. In cells that do not have functional p53, cell arrest at the G2/M boundary has been reported to be mediated by reduced CDC25C and subsequent inactivation of the cyclin B1-CDK1 complex 41 . The down-regulation of CDC25C, cyclin B1, cyclin B2 and CDK1 identified in our study may be specific to mitotic arrest in HaCaT cells exposed to ssUVR. Despite the lack of functional p53, the changes of many genes involved in the mitotic process identified in our HaCaT cells exhibited similar regulation when compared to UV irradiated skin cell in vivo 11 .
In HaCaT cells exposed to repetitive ssUVR, the most inhibited canonical pathway, predicted by IPA analysis, was interferon signaling. Interferon signaling pathways are crucial for antiviral and antibacterial defense 45 . There are three types of interferons; Type I interferons contain IFN-α, IFN-β, IFN-ε, IFN-κ, and IFN-ω; Type II included IFN-γ and type III contains IFN-λ 45 . Type I IFNs bind to IFN-α receptor I and II, which subsequently activate STAT1 and STAT2 by phosphorylation and activation of janus kinase 1 (JAK1) and tyrosine kinase 2 (TK2) 45,46 . Activated STAT1 and STAT2 complexes translocated to the nucleus and bound to IFN-stimulated response elements, activating type I IFN-response genes. Type II IFNγ bound to cell surface receptor IFNγR and activated STAT1 homodimers and respective downstream target genes. Many studies have shown that type I interferons are key players in normal skin and are normally downregulated in three major types of skin cancer: squamous carcinoma, basal cell carcinoma, and melanoma [46][47][48] . Our results show that there was a significant downregulation of type I and type II IFN pathways in cells exposed to repetitive ssUVR. Both STAT1 and STAT2 were reduced in UV radiated cells, accompanied by downregulation of IFN target genes, including IRF1 and IFITM1-3. IRF1 and IRF6 were also down-regulated in the irradiated cells. This result is consistent with a previous study that UVR is able to down-regulate IFNγ activated STAT1 phosphorylation in both human and mouse keratinocytes 49,50 . IRF-6 is highly expressed in skin, and is crucial for skin and limb formation as well as craniofacial development. Depletion of IRF6 strongly affected keratinocyte proliferation and differentiation 51 . UVR-induced downregulation of IFN pathways and their respective target genes were not observed in cells exposed to single ssUVR, suggesting that the changes may be a late response that requires repetitive UV exposure. In addition to interferon signaling, multiple nuclear receptors were also down-regulated by ssUVR, such as RARr, RXRa and RXRb. Both LXR/RAR and PPAR/RXR activation were predicted to be inhibited following ssUVR exposure. Moreover, genes involved in cell migration and cytoskeleton reorganization were induced, such ILK signaling and CDC42 signaling. Therefore, compared to a strong inhibition of cell cycle progression in cells exposed to single ssUVR, cells exposed to repetitive ssUVR displayed a variety of changes in cell signaling and metabolism.
UV radiation, particularly at shorter wavelength, is able to modulate the skin neuroendocrine activity 6 . Several studies have reported an increase of cortisol synthesis by UVB/UVC, which was associated with the increased levels of CRH, POMC, 11βHSD1, CYP11A1, CYP11B1, MC1R, MC2R and decreased 11βHSD2 [52][53][54][55] . Some genes were altered in a similar trend in our study, but the changes were not statistically significant (data not shown). It is worth noting that the UV source used in our study consists of 95% UVA and 5% UVB. This small amount of UVB may not be sufficient to induce significant gene changes related to cortisol synthesis, as UVA alone had no effect on local cortisol production 6,53 .
Our previous study demonstrated a massive reduction of histone lysine acetylation in human keratinocytes exposed to ssUVR 24 . As histone acetylation is normally associated with active gene expression, it is quite surprising that the numbers of up-regulated and down-related genes are equally represented in most comparison conducted in our study, with the exception of cells exposed to repetitive ssUVR at 3 J/cm 2 . The data suggests that histone hypoacetylation may have a very limited impact on ssUVR-altered gene expression. This is in line with a recent study using HEK293 cells which demonstrated limited correlation between genomic distribution of H4K16 acetylation and transcriptional regulation 56 . It is possible that ssUVR-induced histone hypoacetylation leads to changes in chromatin structure and contributes to both DNA damage sensing and successful DNA repair, rather than the global changes of gene expression. However, we do not exclude the possibility that altered histone acetylation may mediate ssUVR-induced expression changes in a gene-specific manner. For example, aquaporin 3 (AQP3), an important cell membrane protein that is crucial for water and glycerol transportation as well as skin hydration 57,58 , was downregulated in cells exposed to either single ssUVR (−1.9 fold) or repetitive ssUVR (−2.9 fold) (Supplemental Table 2). These results were consistent with the finding of AQP3 downregulation and decrease of water permeability in keratinocytes exposed to UVB 59 , Interestingly, AQP3 levels remained low in cells undergo 1-week recovery (Supplemental Table 2), suggesting a possible epigenetic change mediates AQP3 down-regulation by ssUVR. This was further supported by a recent study on HDAC3 suppression of AQP3 expression in mouse epidermal keratinocytes 60 .
In summary, our present study compared changes in gene expression in human keratinocytes exposed to single or repetitive ssUVR. In addition, changes present one week after the final dose of repetitive irradiation were measured. Using RNA-seq combining with IPA comparative analysis, DEG lists were generated and pathways and their connected upstream regulators were analyzed to look for changes in expression resulting in cellular abnormalities. The results indicate that ssUVR-modulated the expression of genes with diverse cellular functions. While single ssUVR caused a significant inhibition of genes involved in cell cycle progression, repetitive ssUVR lead to extensive changes of genes related to many important cell signaling pathways, cell adhesion, and metabolism. These data suggest that a complex network of transcriptional regulators and pathways orchestrate the cellular response to ssUVR. Moreover, we identified 159 genes that maintained their initial changes after several cell passages, suggesting a potential epigenetic mechanism that may mediate ssUVR-induced transcriptome changes. Further analysis of the epigenetic landscape in cells exposed to single or repetitive ssUVR will provide new insights into the mechanism underlying solar UV radiation-induced skin damage and carcinogenesis.

Methods
Cell culture. Immortalized human HaCaT keratinocytes were obtained from Dr. Chuangshu Huang at NYU School of Medicine. The cells were cultured in DMEM medium (4.5 g/L glucose) containing 10% fetal bovine serum and 1% penicillin/streptomycin at 37 °C and 5% CO 2 .
Solar simulated UV radiation. The UV source and simulated UV radiation are the same as previously described 24 . Solar simulated UV radiation was performed using a modified Hand Foot II phototherapy instrument (National Biological Corporation, Beachwood, OH) with 8 UVA lamps (HOUVALITE F24T12/BL/HO [PUVA], National Biological Corporation). The lamp emission is filtered by a single glass plate, resulting in an average intensity of 3.4-3.5 mw/cm 2 for UVA and 0.18 mw/cm 2 for UVB at cell culture surface, which equivalent to a spectrum of 95% UVA and 5% UVB. UVA intensity was measured with a UVX digital handheld radiometer using the UVX-36 UVA sensor (UVP, Upland, CA). UVB intensity was measured with an ILT1400 Photo detector using the SEL240 UVB sensor (International Light, Newburyport, MA). The intensity of the lamp output was measured before every exposure and the exposure times were calculated to deliver the desired doses. The lowest dose (3 J/cm 2 ) and the highest dose (12 J/cm 2 ) used in this study required the approximate exposure time of 15 min and 57 min, respectively. For cells exposed to single ssUVR, 3.5 × 10 6 cells were seeded in 10-cm cell culture dish at the day before irradiation. Prior to exposure, cells were washed twice with PBS and exposed to 12 J/cm 2 ssUVR in 10 ml of Hank's Balanced Salt Solution (HBSS, life technologies, Grand Island, NY) supplemented with 4 mM glucose. After irradiation, the HBSS was replaced with fresh medium and cells were cultured for 24 hours. Control samples were treated identically but covered with foil during irradiation (Sham). To maintain a constant temperature, a table fan was used to reduce heat production during the exposure. For repetitive radiation, cells were exposed to 5 repetitive doses of ssUVR at 3, 6, 12 J/cm 2 as described above, except they were allowed to recover for 3 days prior to the next round of radiation. To examine the persistence of gene expression changes, cells were exposed to 5 repetitive doses of ssUVR at 12 J/cm 2 , and were harvested at one week after final irradiation.
RNA isolation and sequencing library preparation. Total RNA was extracted from sham and ssUVR-exposed cells using the Trizol reagent (Life Technologies, Gaithersburg, MD). RNA-Seq libraries were prepared using Illumina TruSeq RNA-sample preparation kit according to the manufacture's instruction. Sequencing was performed at the NYU School of Medicine Genome Technology Center using Illumina HiSeq. 2500 by multiplexed single-read run with 50 cycles.

RNA-Seq data analysis.
Raw sequence data (Fastq) were loaded into Biomedical Genomics Workbench Version 3.5.3 (Qiagen) for data analysis. The raw Fastq files were trimmed to remove any remaining adaptors and ambiguous nucleotides. The trimmed sequence files were aligned to human genome (Hg38) allowing two mismatches. Reads mapped to the exons of a gene were summed at the gene level. Gene expression levels were quantified as total read per million (TPM). Differential gene expression was analyzed using Advance RNA-seq plug-in tool (Qiagen) by comparing each treated group versus corresponding control group. TMM (trimmed mean of M values) normalization was performed to adjust library sizes before differential expression analysis. The genes with a false discovery rate (FDR) < 0.05 between control and treated group and an average mean of total counts no less than 10 were defined as differentially expressed genes (DEGs).
Gene ontology and pathway analysis. Functional annotation was analyzed with the Gene Ontology (GO) classification system using the Database for Annotation, Visualization and Integrated Discovery (DAVID), an NCBI web based functional annotation tool. Gene network, top pathways and upstream regulators were analyzed by Ingenuity Pathway Analysis (Qiagen).
Data Availability Statement. The datasets generated during the current study have been deposited in NCBI Gene Expression Omnibus (GEO accession number GSE102676).