FOSL1 is a novel mediator of endotoxin/lipopolysaccharide-induced pulmonary angiogenic signaling

Systemic sepsis is a known risk factor for bronchopulmonary dysplasia (BPD) in premature infants, a disease characterized by dysregulated angiogenesis and impaired vascular and alveolar development. We have previoulsy reported that systemic endotoxin dysregulates pulmonary angiogenesis resulting in alveolar simplification mimicking BPD in neonatal mice, but the underlying mechanisms remain unclear. We undertook an unbiased discovery approach to identify novel signaling pathways programming sepsis-induced deviant lung angiogenesis. Pulmonary endothelial cells (EC) were isolated for RNA-Seq from newborn C57BL/6 mice treated with intraperitoneal lipopolysaccharide (LPS) to mimic systemic sepsis. LPS significantly differentially-regulated 269 genes after 6 h, and 1,934 genes after 24 h. Using bioinformatics, we linked 6 h genes previously unknown to be modulated by LPS to 24 h genes known to regulate angiogenesis/vasculogenesis to identify pathways programming deviant angiogenesis. An immortalized primary human lung EC (HPMEC-im) line was generated by SV40 transduction to facilitate mechanistic studies. RT-PCR and transcription factor binding analysis identified FOSL1 (FOS like 1) as a transcriptional regulator of LPS-induced downstream angiogenic or vasculogenic genes. Over-expression and silencing studies of FOSL1 in immortalized and primary HPMEC demonstrated that baseline and LPS-induced expression of ADAM8, CXCR2, HPX, LRG1, PROK2, and RNF213 was regulated by FOSL1. FOSL1 silencing impaired LPS-induced in vitro HPMEC angiogenesis. In conclusion, we identified FOSL1 as a novel regulator of sepsis-induced deviant angiogenic signaling in mouse lung EC and human fetal HPMEC.

Scientific RepoRtS | (2020) 10:13143 | https://doi.org/10.1038/s41598-020-69735-z www.nature.com/scientificreports/ prior work suggests that systemic LPS in mice programs a dysangiogenesis phenotype, which also results in alveolar remodeling 26 . Inflammation alters vascular development, as infants with BPD who experienced sepsis have increased risk of pulmonary hypertension 27 and biomarkers of angiogenesis such as VEGFA, KDR, endostatin, and angiopoietin-2 are altered in infants who develop subsequently BPD 28 . Despite the accumulating evidence for sepsis-induced aberrant angiogenesis to program vascular remodeling in the developing lung, the mechanisms by which LPS disrupts angiogenesis or programs dysangiogensis vascular phenotypes in the developing lung reamain unclear. Therefore, the objective of this study was to identify novel pathways by which systemic sepsis programs aberrant angiogenesis in the developing lung using an unbiased approach involving RNA-Seq and bioinformatics methods. By combining discovery-based RNA sequencing studies in a neonatal mouse model of sterile sepsis complemented by validation strategies using genetic manipulation in primary fetal human lung endothelial cell (EC) and an immortalized lung EC line we developed, we identify the transcription factor FOSL1 (FOS like 1, also known as FRA-1, FOS-related antigen 1), as a novel regulator of endotoxin-mediated dysangiogenesis in the developing lung.

Results
Lung endothelium transcriptome profiling identifies inflammatory, immune, and several pathways differentially regulated by systemic endotoxin in the developing lung at 6 h. To identify novel targets induced by systemic endotoxin, mouse lung EC were isolated from 4-day old (DOL-4) LPS-treated (2 mg/kg intraperitoneal, i.p) and control littermates (n = 3/group) after 6 h, and RNA-Seq was performed. Ingenuity Pathway Analysis (IPA) was used to analyze 24,392 genes, of which 269 were significantly up-or down-regulated with p and q < 0.05 (Fig. 1A). IPA-predicted physiological system development and functional categories induced by LPS were mainly involved in leukocyte recruitment and function, as expected for a pro-inflammatory stimulus (Table S1). Angiogenesis as a component of "Cardiovascular System Development and Function" was confirmed as predicted increased with LPS treatment, significant across the data set at 6 h (p = 2.84 × 10 -18 ) with + 3.422 z-score of activation (Fig. 1B).
After preliminary identification of significant changes of gene expression, the IPA Knowledge Base (KB) of ~ 5.9 million citations was used to efficiently filter RNA-Seq gene expression data. We targeted those genes that were previously not known to be altered with LPS treatment and showed significant changes. This yielded 97 genes that were significantly up-and down-regulated with p and q < 0.05 (Table S2). We validated 6/97 targets, identified by RNA-Seq with RT-PCR (Real-Time PCR); Fosl1, Adam8, Cxcr2, Plac8, Fcgr3, and Hpx in primary murine pulmonary EC isolated under similar experimental conditions as RNA-seq data in independent mouse samples (n = 6/group) (Fig. 1C). Genes were chosen for validation based on below described IPA analysis linking 6 h and 24 h sequencing data.
Lung endothelium transcriptome profiling identifies angiogenic or vasculogenic genes differentially regulated in the neonatal lung with systemic endotoxin at 24 h. To identify novel LPSinduced angiogenesis and vasculogenesis pathways, RNA-Seq was performed as above on mouse lung EC isolated from LPS-treated and control DOL-4 mice 24 h after LPS exposure. IPA identified 1,934 genes which were significantly up-or down-regulated with p and q < 0.05 ( Fig. 2A). IPA-predicted physiological system development and function categories induced by LPS were mainly involved in leukocyte recruitment and function (Table S3). Angiogenesis as a component of "Cardiovascular System Development and Function" was confirmed as predicted increased with LPS treatment, significant across the data set at 24 h (p = 2.68 × 10 -26 ) with + 2.474 z-score of activation (Fig. 2B). As before, IPA was used to efficiently filter RNA-Seq gene expression data, identifying genes known to be important in angiogenesis or vasculogenesis. This yielded 235 genes (see Table S4). Using Pubmed searches, we identified Adam8, Cxcr2, Rnf213, and Hpx as genes previously implicated in regulating angiogenesis or vasculogenesis in developmental or pathological conditions, but not reportedly downstream of LPS or its receptor Toll Like receptor 4. We validated expression of these genes in independent mouse lung EC samples obtained 24 h after i.p. LPS injection (Fig. 2C).

Discovery of novel transcriptional links between LPS and angiogenesis/vasculogenesis. As
we were investigating novel links between genes induced by LPS at 6 h that can transcriptionally upregulate angiogenic/vasculogenic genes at 24 h, we used IPA to make connections between genes at 6 h not known to be induced by LPS, and genes at 24 h known to be involved in angiogenesis or vasculogenesis (Fig. 3A). Only genes found to have ≥ 50% increase in expression were pursued. Connections between 6 and 24 h genes were based solely on transcriptional regulation in the IPA. Non-transcriptional interactions such as protein-protein interactions or phosphorylation, etc. were initially explored but not pursued, and are described in the Supplementary material (Table S5). Additionally, to maximize impact, only 6 h genes with ≥ 3 downstream targets at 24 h were investigated, while genes with < 3 downstream targets are shown in the Supplementary material (Table S5).
Six genes (F10, Fcgr2a, Fosl1, Ndrg1, Selplg, Zbtb16) were induced 6 h after LPS in mouse lung EC, and found to have ≥ 3 transcriptional relationships with 24 h downstream angiogenic/vasculogenic targets. Three of these targets (F10, Ndrg1, Selplg) were eliminated based on literature review, as these were not novel LPS/TLR targets . This left us with Fcgr2a, Fosl1, and Zbtb16 as potential target genes induced early with LPS that transcriptionally regulated 24 h angiogenic genes (Fig. 4). Fcgr2a is not a transcription factor, Zbtb16 was not validated by RT-PCR, so these were excluded. This left Fosl1 as the sole remaining target.
Creation of immortalized human pulmonary lung endothelial cell line (HPMEC-im). We proceeded to establish functional relationships between FOSL1 and downstream targets in an in vitro model. We RT-PCR assessment of novel genes activated by LPS linked to angiogenesis and vasculogenesis genes in HPMEC-im. Five genes identified from our 6 h RNA sequencing data were assessed with RT-PCR in HPMECs, including FOSL1, ADAM8, and FCGR3, as well as CXCR2 and CXCL10 which were included for validating LPS-induced cytokine expression (Fig. 5B). FOSL1 had multiple multiple connections to differentially-expressed 24 h angiogenesis genes, and it was the only transcription factor identified. Therefore, FOSL1 (FOS like 1, Fra-1, AP-1 transcription factor subunit) was further explored as an early EC gene upregulated with LPS that could potentially induce expression of downstream genes involved in angiogenesis and vasculo- . Data presented at mean ± standard deviation, *p < 0.001 between LPS-treated and saline control baseline by one-sample, two-tailed Student's t test as all data were normally distributed. Graphs created in SPSS 26 for Mac (https ://www.ibm.com/ analy tics/spss-stati stics -softw are) and image assembled in Adobe Illustrator 24.1.0 for Mac (https ://www.adobe .com/produ cts/illus trato r.html). , did not seem relevant to angiogenesis (e.g., Cd44), or were not specific for angiogenesis (e.g., Tnf, Il-1β). Therefore, the list of 24 h angiogenic/ vasculogenic genes was expanded beyond those only predicted by IPA to be linked to 6 h genes. We then focused our search to genes (a) not reported to be downstream of LPS, (b) with at least a twofold increase in expression, and (c) not known to be linked to FOSL1 in the IPA-KB. This resulted in a list of 15 potential genes of interest: Adam8, Aif1, Fermt3, Hif-3α, Il-18bp, Lrg1, Myl7, Prkcb, Prok2, Rnf213, S1pr5, Sema4d, Spink5, Syk, and Sp100a. As these novel relationships were outside of the validated IPA-KB, JASPAR analysis was performed to identify transcriptional relationships from FOSL1, and indicated eight of these genes may have FOSL1-binding sites in the human promoter (Fig. 3B), including ADAM8, HIF-3α, LRG1, PRKCB, PROK2, SPINK5, AND SP100A (Fig. 4). RNF213 and HPX were also included based on previous preliminary Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. Of these, four angiogenesis/vasculogenesis genes   6A). Finally, combined FOSL1 overexpression and LPS treatment resulted in significant increases in gene expression compared to controls without FOSL1 transfection or LPS treatment. HIF-3α expression did not change with FOSL1 overexpression, and served as negative control. IL-1α expression, which is known to be negatively regulated by FOSL1 29 , served as an additional control to validate FOSL1-dependent LPS-mediated gene expression in HPMEC. These data demonstrate that FOSL1 transcriptionally regulates native, and synergistically, LPS-sensitive gene expression of identified targets.
To investigate whether FOSL1 directly binds to the promoter region of identified targets after LPS treatment we performed chromatin immunoprecipitation (ChIP). We noted that FOSL1 robustly bound to the promoters of ADAM8, HPX, LRG1, PROK2, and RNF213 at baseline, indicating baseline regulation of these targets by FOSL1 (Fig. 6B). FOSL1 binding to the promoter regions of ADAM8, PROK2, and RNF213 increased by 86.7 to 100.4%, respectively, 7 h after LPS treatment in HPMEC-im, suggesting LPS-induced transcriptional activation of genes in HPMEC-im (Fig. 6C). Increased expression of downstream angiogenic genes with FOSL1 overexpression, and  Taken in conjunction with prior overexpression data in HPMEC-im, these data imply that induction of target genes after LPS treatment is FOSL1-dependent. We next examined the impact of FOSL1 in regulating angiogenic sprouting in primary HPMEC by pursuing a matrigel-based angiogenesis assay in HPMEC. LPS induced robust in vitro angiogenesis in matrigel as evidenced by increased branches and networks (confluence of ≥ 2 branches). LPS-induced angiogenesis was strongly suppressed by FOSL1 silencing (Fig. 7C-E). These data demonstrate that FOSL1 regulates LPS-induced in vitro angiogenesis.

Discussion
Although sepsis is known to promote the vascular and alveolar remodeling underlying bronchopulmonary dysplasia, the mechanisms by which inflammation can modulate pulmonary angiogenesis have not been fully defined. In this unbiased, hypothesis-generating study, we performed transcriptome profiling on lung EC in a rodent model of systemic sepsis, and identified FOSL1 as a novel early target gene not previously described to be transcriptionally regulated by LPS. This novel transcriptional regulator was then linked to the angiogenesis or vasculogenesis genes LRG1, RNF213, and Sp100, which have previously not been reported to be downstream of LPS or TLR signaling. Complementary overexpression and silencing approaches in a HPMEC-im line we generated as well as primary fetal HPMEC validated the LPS-FOSL1-angiogenesis/vasculogenesis pathway by revealing dependence of basal angiogenesis and vasculogenesis gene expression on FOSL1. Finally, impaired in vitro angiogenesis in FOSL1-deficient ECs strongly implicates the transcription factor FOSL1 as a central mediator of inflammation-induced pulmonary EC angiogenesis. Previous studies in premature infants and experimental models of BPD suggest that dysregulation of angiogenesis is important for BPD pathogenesis 3,18 . For example, in rodent models of BPD, hyperoxia leads to decreased angiogenesis through ERK1/2-dependent signaling 19 . Similarly, hyperoxia variably induces expression of angiogenic genes such as the angiopoietins, platelet derived growth factors, and vascular endothelial growth factor in rats 20 , culminating in a BPD-like phenotype. Angiopoietin-1 has been shown to ameliorate experimental BPD in a rodent model 21 , while angiopoietin-2 worsens hyperoxia-mediated lung injury in nitric oxide-synthase 2-deficient mice 30 , and increased angiopoietin-2 is associated with BPD in preterm infants 31 . Our group has previously shown that LPS treatment of newborn rodents recapitulates the alveolar simplification and impaired ; and (C) assessment of angiogenesis or vasculogenesis genes after treatment of primary HPMECs (n = 3) and HPMEC line (n = 5) with LPS for 24 h (p values determined by one-sample, two-tailed Student's t test; all data were normally distributed except 24 h primary HPMEC LRG1 which was assessed with one-sample Wilcoxon signed rank test). Data presented at mean ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001 between LPS-treated and saline control baseline. Graphs created in SPSS 26 for Mac (https ://www. ibm.com/analy tics/spss-stati stics -softw are) and image assembled in Adobe Illustrator 24.1.0 for Mac (https :// www.adobe .com/produ cts/illus trato r.html). In this study, we identified FOSL1 as a novel regulator of the pulmonary angiogenic response. FOSL1 is one of several subunits that binds a Jun-family member to form the AP-1 complex. FOSL1 has important effects in osteoclasts and disorders of bone via Wnt signaling [34][35][36] and mediates anti-inflammatory actions of some natural compounds [37][38][39] . FOSL1 is also a proto-oncogene, affecting the ability of cancer cells to invade and metastasize in sarcomas 40 , squamous cells carcinomas 41 , melanoma 42 , gastric cancer 43 , pancreatic cancer 44 , and lung cancer 44 . With respect to the endothelium, it is known that Fosl1 −/− embryos die and develop vascular defects in extraembryonic tissues 45 . Accordingly, embryonic stem cells lacking FOSL1 differentiate into ECs but do not form primitive capillaries or tube-like structures and FOSL1 is a required for HUVEC assembly into vessels 45 . FOSL1 has also been reported as modulating angiogenesis-vasculogenesis and β3 integrin and endothelial cell adhesion 46 , but this study is the first report of FOSL1 acting as a transcriptional regulator of angiogenesis-related genes. Additionally, we extend existing work 46 by including knock-down and over-expression experiments. While an interaction between LPS and FOSL1 in modulating gefitinib-induced interstitial lung disease has been reported 47 , whether LPS directly regulates FOSL1-dependent signaling and the role of FOSL1 in mediating LPS-induced expression of key downstream angiogenic targets has not been previously investigated.

Scientific RepoRtS
We identified PROK2, RNF213, and LRG1 as novel targets which downstream of LPS, transcriptionally regulated by FOSL1, and are implicated in angiogenesis/vasculogenesis. www.nature.com/scientificreports/ in oxygen-induced retinopathy 49 , while RNF213 50 is required for normal vascular development. Of these, LRG1 is the only overlapping reported finding in the literature where FOSL1 knockdown in HUVECs decreased LRG1 to the same degree that we found in our immortalized HPMEC line. The current study places this understanding within the context of neonatal sepsis and implicates FOSL1 not only in embryonic development, but also potentially as a mediator of impaired vascular development in the neonatal lung. We further explored this relationship between LPS, FOSL1, and angiogenesis/vasculogenesis with primary HPMEC and an immortalized HPMEC line we derived from primary HPMEC. Not only did FOSL1 overexpression increase sensitivity to LPS stimulation, FOSL1 overexpression alone induced angiogenic or vasculogenic gene expression, indicating basal regulation. Additionally, using a ChIP assay, we demonstrated specific binding of FOSL1 to promoters of PROK2, ADAM8, RNF213, HPX, and LRG1 downstream targets at baseline, and increased binding to PROK2, ADAM8, RNF213 promoters after LPS treatment. That all genes did not demonstrate additional binding with LPS treatment could represent differences in time kinetics of binding after LPS treatment. Differences in the degree of LPS-induced target gene expression between mice and HPMEC could result from the influence of other lung cell types and cytokines in vivo, timing of i.p LPS in vivo vs. direct LPS in vitro, and encapture of several RNA transcripts by RNA sequencing as against qRT-PCR. Similary, differences between primary HPMEC and HPMEC-Im may have arisen because of faster proliferation in HPMEC-im and the potential for multiple gene copies in HPMEC-im. However, the direction and amplitude of gene expression changes with LPS were not hugely different considering these variables. A prior study in adult mice with constitutive over-expression of FOSL1 resulted in rapid death after high-dose LPS treatment (50 mg/kg) with massive inflammation and chemokine response 47 . However, this study did not report any findings related to expression of angiogenesis genes or angiogenic phenotype 47 . In previous studies from our lab, a 25-fold lower dose of i.p. LPS, as in this study, did not result in significant lethality, but a more nuanced alteration of the angiogenic/vasculogenic phenotype with decreased lung EC population 51 . The dosing we used is more consistent with the clinical scenario in preterm infants, where most infants survive Gramnegative sepsis. Herein, we propose aberrant pulmonary angiogenesis as a possible mechanism contributing to this response. Conversely, mice deficient in Fosl1 are less susceptible to LPS-induced lung injury and mortality 52 , with lower lung injury responses, fewer neutrophils, less NFκB and cJun-AP1 binding 52 , and lower IL-1β and MIP-1α expression and higher IL-10 expression 53 . Our in vivo results in HPMEC is consistent with these study in showing that pro-inflammatory genes induced with LPS are suppressed or increased after FOSL1 silencing or overexpression, respectively. Our data support enhanced inflammation-induced angiogenesis/vasculogenesis as an important mechanism of altered FOSL1 expression. Manipulation of FOSL1 expression may therefore lead to modulation of response to inflammation. The lack of readily-available Fosl1 +/− mice (Fosl1 −/− mice have embryonic lethality) and inducible EC-specific conditional Fosl1 knockout mice precluded us from performing experiments to confirm our in vitro findings. Validation of our in vitro findings in mice with Fosl1 deficiency in vivo would have strengthened our conclusions. Work by Schreiber et al. demonstrating a role for Fosl1 in placental vascular development 54 is however, consistent with our results.
This study has several strengths. Use of unbiased bioinformatics tools to generate hypotheses allowed us to quickly identify novel targets, which could then be explored with standard laboratory techniques in both murine and human cells, including a newly created and validated HPMEC line. Additionally, demonstration of direct binding of FOSL1 to promoter genes, and loss-and gain-of-function approachs to examine relationships between LPS, FOSL1 and target angiogenesis/vasculogenesis pathway genes confirmed regulation. Finally, validation of FOSL1 as a mediator of sepsis-induced aberrant angiogenesis in human EC enhances clinical relevance. A limitation is our initial reliance on the IPA knowledge base, which is periodically, not continuously, updated; however, confirmation of IPA findings with more up-to-date literature searches was undertaken. Examining how FOSL1 gets activated by LPS, investigating protein level changes of target genes, and their individual roles in LPS-induced angiogenesis are interesting avenues for future research. Finally, while embryonic lethality of Fosl1 −/− mice precluded in vivo mechanistic validation of our results, studies in mice with conditionally-deleted endothelial Fosl1 would strengthen our data.
In conclusion, we have identified FOSL1 as a novel, early, central target in inflammation-induced angiogenesis in primary lung EC and HPMECs, linking sepsis with expression of multiple genes involved with aberrant angiogenesis in the developing lung, including ADAM8, LRG1, RNF213, HPX, CXCR2, PROK2, and SP100, which may contribute to development of BPD. Taken together, these data indicate a strong, central role of FOSL1 in regulating lung EC angiogenesis in the context of inflammation. While hyperoxia and sepsis are associated with vascular remodeling and BPD, they may alter lung angiogenesis potentially differently. Hyperoxia disrupts angiogenesis, while sepsis may program a dysangiogenesis phenotype, wherein there might be increased but non-directional dysmorphic angiogenesis [22][23][24][25] . Our prior work and the results of this study that systemic LPS may program dysangiogenesis lung phenotypes that may also impair normal vascular development. The role of FOSL1 in mediating sepsis-induced pulmonary dysangiogenesis and vascular remodeling using EC-specific conditional FOSL1 knock down mice as well as the role of FOSL1 in regulating inflammatory angiogenesis in other organs are topics or future research. Our data support testing whether short-term inhibition of EC-FOSLI can serve as a therapeutic target to prevent development of vascular defects in BPD.

Methods
Cell culture and reagents. Human primary pulmonary microvascular endothelial cells (HPMEC) were purchased from a commercial vendor (ScienCell, Carlsbad, CA) and used as previously described 32 . These HPMECs were derived from lungs of newborn infants, and only cells between passages 3 and 4 were used for all experiments. Primary HPMEC used for our experiments were derived from the lungs of 18 weeks female fetus. HPMECs were grown in endothelial cell medium supplemented with 5% fetal bovine serum (FBS), antibiotics,  51 . We did not observe any mortality in our experiments. The pups were left with the dam and closely monitored for signs of distress for half a hour, and then they were observed briefly once more on the same day for the 24 h experiments. All animal experiments were performed on mouse pups on DOL 4. Mice were sacrificed using a i.p. injection of pentobarbital (100 mg/kg) and were exsanguinated after cessation of heartbeat. The lungs were harvested for EC isolation.
Isolation of murine endothelial cells. Each endothelial preparaton was from pooled lungs of two C57BL/6 pups (4 days old). 3-4 biological replicates, 6-8 mice/condition was used for experiments. The protocol for the isolation of mouse lung EC was described previously 55 . Briefly, the lungs from euthanized pups were minced with scissors in DMEM and transferred to pre-warmed 1 mg/mL collagenase solution and rotated at 37 °C for 45 min 55 . The digested mixture was passed through a 14 g cannula several times, strained through a 70 µm cell strainer, and washed with a 20% FBS solution 55 . The supernatant was then centrifuged at 400 × g for 5 min and the pellet was resuspended with 0.1% bovine serum albumin 55 . The endothelial cells were isolated with anti-PECAM-1 antibody-conjugated Dynabeads (Thermo-Fisher, Rockford, IL) on a rocker for 15 min at room temperature as per the manufacturer's protocol. RNA was extracted following standard protocol as described below.
Generation of HPMEC cell line. Primary HPMEC (male, 19 week fetus) were immortalized using SV40 large T antigen. Lentiviral transduction (Applied Biological Materials (ABM), Richmond, BC, Canada). These endothelial cells were cultured in endothelial cell media as above and only cells less than passage ten were used.
To validate immortalized HPMEC (HPMEC-im), we cultured cells on coverslips and stained them with antibodies of endothelial markers, such as CDH5, ERG and VWF. The slides were used for IF staining using the primary antibodies (Abcam, Cambridge, MA) with the corresponding Alexa Fluor secondary antibodies (Thermo-Fisher). The slides mounted in Prolong Gold with DAPI (Thermo-Fisher), which stains the nucleus. We also verified IM HMPEC with Alexa Fluor 488 AcLDL uptake experiment. Images were taken at 63× magnification using a Zeiss LSM510 confocal microscope with an attached camera.
RNA isolation and cDNA generation. Total RNA was extracted from HPMECs and mouse lung EC or tissue using the PureLink RNA Mini Kit (Thermo-Fisher) following the manufacturer's instructions as previously described 32 . cDNA was synthesized from 1 μg of RNA using a SuperScript IV First-Strand Synthesis Kit (Thermo-Fisher) per the manufacturer's recommendations.

RNA-Seq.
Total RNA was extracted from mouse lung EC using the mirVana miRNA isolation kit (Thermo-Fisher). RNA sequencing was performed on a HiSeq 1,500 system (Illumina, San Diego, CA). Mapping of RNA-Seq reads and transcript assembly and abundance estimation were conducted using the Tuxedo Suite software package (Broad Institute, Cambridge, MA) as previously described [56][57][58] . DAVID was used for biological function analysis. Heatmaps were generated from heatmapper.ca/expression and are scaled by row and hierarchically clustered by average linkage distance via Euclidean measurement. Full sequencing data (BioProject ID PRJNA600007) are available online at https ://www.ncbi.nlm.nih.gov/biopr oject /60000 7.

JASPAR analysis.
The promotor sequences (about 2 kB region upstream of transcription start site) were downloaded from NCBI genome browser. FOSL1 binding sites on the promoters were predicted by JASPAR 2018 61 with a threshold of 0.9.