Slco2a1 deficiency exacerbates experimental colitis via inflammasome activation in macrophages: a possible mechanism of chronic enteropathy associated with SLCO2A1 gene

Loss-of-function mutations in the solute carrier organic anion transporter family, member 2a1 gene (SLCO2A1), which encodes a prostaglandin (PG) transporter, have been identified as causes of chronic nonspecific multiple ulcers in the small intestine; however, the underlying mechanisms have not been revealed. We, therefore, evaluated the effects of systemic knockout of Slco2a1 (Slco2a1−/−) and conditional knockout in intestinal epithelial cells (Slco2a1ΔIEC) and macrophages (Slco2a1ΔMP) in mice with dextran sodium sulphate (DSS)-induced acute colitis. Slco2a−/− mice were more susceptible to DSS-induced colitis than wild-type (WT) mice, but did not spontaneously develop enteritis or colitis. The nucleotide-binding domain, leucine-rich repeats containing family, pyrin domain-containing-3 (NLRP3) inflammasome was more strongly upregulated in colon tissues of Slco2a−/− mice administered DSS and in macrophages isolated from Slco2a1−/− mice than in the WT counterparts. Slco2a1ΔMP, but not Slco2a1ΔIEC mice, were more susceptible to DSS-induced colitis than WT mice, partly phenocopying Slco2a−/− mice. Concentrations of PGE2 in colon tissues and macrophages from Slco2a1−/− mice were significantly higher than those of WT mice. Blockade of inflammasome activation suppressed the exacerbation of colitis. These results indicated that Slco2a1-deficiency increases the PGE2 concentration, resulting in NLRP3 inflammasome activation in macrophages, thus exacerbating intestinal inflammation.

is characterized by multiple intractable small-intestinal ulcers with manifestations including anaemia and hypoproteinaemia 10,11 , is caused by autosomal recessive loss-of-function mutations in SLCO2A1 12 . PGE 2 concentrations in patients with this genetic background should theoretically be higher than those in patients without this mutation because of lower PGE 2 metabolism. In fact, concentrations of PGE major urinary metabolites, which are stable metabolites derived from PGE 2 , in patients with chronic enteropathy, so-called chronic enteropathy associated with SLCO2A1 gene (CEAS), have been shown to be higher than those in patients with Crohn's disease 13 . Thus, given the protective effect of PGE 2 on mucosal injury, PGE 2 seems to exert contrasting effects on the intestinal mucosa, and the mechanisms underlying enteritis development under SLCO2A1 deficiency have not been clarified.
In the present study, we evaluated a potential role of SLCO2A1 in intestinal homeostasis using systemic as well as conditional Slco2a1-knockout mice. Deletion of Slco2a1 did not cause alteration of epithelial structure and spontaneous enteritis. However, SLCO2A1 protected against gut inflammation in an experimental colitis model. Further, we report a possible mechanism: increased PGE 2 because of Slco2a1 deficiency autocrinally causes inflammasome activation in macrophages, resulting in exacerbation of intestinal inflammation.

Slco2a1 deficiency does not result in spontaneous enterocolitis. To investigate the effect of Slco2a1
deficiency on the intestine, germline Slco2a1-knockout (Slco2a1 −/− ) mice 14 were evaluated for basal phenotype in comparison with wild-type (WT) mice. There were no significant differences in body weight between Slco2a1 −/− and WT mice at any time point evaluated (Fig. 1a). Slco2a1 −/− mice did not display any sign of enteritis as assessed by histological examination of ilea and colons at 8, 12, and 24 weeks of age (Fig. 1b,c, and Supplementary  Fig. S1a,b). mRNA levels of Il1b and Tnf, which are pivotal inflammatory cytokines 15 , in the colon were not affected by Slco2a1 knockout, although Il1b mRNA expression in the ileum of Slco2a1 −/− mice was decreased (Fig. 1d,e). These results demonstrated that Slco2a1 deficiency per se does not induce intestinal inflammation. Slco2a1 deficiency also did not affect the numbers of goblet cells and Paneth cells, which secrete mucins and antibacterial contents for mucosal barrier function 16,17 ( Supplementary Fig. S2a,b). Slco2a1 −/− mice are more susceptible to DSS-induced colitis. To investigate the role of SLCO2A1 in colitis development, we next examined the effects of Slco2a1 deficiency in experimental colitis models of Slco2a1 −/− and WT mice established by administering 3.5% DSS in the drinking water for 7 days. Slco2a1 −/− mice lost significantly more weight than did WT mice upon DSS-induced colitis (Fig. 2a). Histopathological examination of colons revealed more severe lesions, stronger neutrophil and lymphocyte infiltration into the mucosal and submucosal areas, and more extensive loss of crypts in Slco2a1 −/− mice than in WT mice, resulting in higher histological colitis grades (Fig. 2b,c).
To investigate the mechanisms underlying colitis exacerbation in Slco2a1 −/− mice further, we analysed gene expression profiles in colon tissues of Slco2a1 −/− and WT mice in the presence/absence of DSS-induced colitis by microarray analysis (Fig. 2d). Gene set enrichment analysis (GSEA) indicated that the most strongly enriched pathways in Slco2a1 −/− mice with DSS-induced colitis were related to the inflammatory immune response and leukocyte migration/activation (Fig. 2e,f). Comparison of the expression levels of enriched genes in the inflammatory response (Fig. 2g) and cytokine expression (Fig. 2h) revealed that the expression of inflammasome-related genes, such as Il1b, Il18, P2rx7, and macrophage-related chemokines, such as Ccl2 and Ccl4, was upregulated in Slco2a1 −/− compared to WT mice administered DSS. The proinflammatory cytokine interleukin (IL)-1β, which is cleaved and activated by inflammasome assembly 18 , and the nucleotide-binding domain, leucine-rich repeats containing family, pyrin domain-containing-3 (NLRP3) inflammasome reportedly play important roles in inflammation in DSS-induced colitis model mice 19 . Inflammasomes activation is involved in other experimental enteritis/colitis models 20,21 and is related to human inflammatory bowel disease (IBD) via genetic and environmental factors 22,23 . IL-1β production induced by inflammasome activation is increased in Slco2a1 −/− mice. We next focused on the role of the NLRP3 inflammasome in the exacerbated colitis in Slco2a1 −/− mice. mRNA expression of pro-inflammatory cytokines in colon tissues as assessed by RT-qPCR was increased in Slco2a1 −/− compared to WT mice on days 3 and 7 of DSS treatment (Fig. 3a)., Although Casp1 mRNA expression was decreased (Fig. 3b), protein levels of pro-and cleaved IL-1β, cleaved caspase-1 (CASP1), and NLRP3 were also significantly elevated in Slco2a1 −/− compared to WT mice with colitis ( Fig. 3c, Supplementary Fig. S3). These results indicated that NLRP3 inflammasome activation might be associated with the exacerbation of DSS colitis in Slco2a1 −/− mice. The NLRP3 inflammasome and subsequent secretion of mature IL-1β by macrophages reportedly are a critical mechanism of intestinal inflammation in the DSS-induced acute colitis mouse model 19 . Considering this fact together with the microarray data, we next focused on the inflammasome pathway in intestinal macrophages, and investigated whether inflammasome activation in macrophages could exacerbate the intestinal inflammation. To assess the inflammasome status in macrophages with genetic Slco2a1 deletion, we isolated colonic lamina propria (LP) macrophages from WT and Slco2a1 −/− mice with DSS-induced colitis. Consistent with the results in vivo, Il1b mRNA expression was significantly increased, whereas Casp1 mRNA expression was significantly decreased in Slco2a1 −/− mice (Fig. 3d). Nlrp3 mRNA expression was elevated, but not significantly. In line with the mRNA data for LP macrophages, the protein levels of NLRP3, cleaved caspase-1, and mature IL-1β were significantly increased in bone marrow-derived macrophages (BMDMs) from Slco2a1 −/− compared to the levels in BMDMs from WT mice (Fig. 3e, Supplementary Fig. S4). Accordingly, although administration of MCC950, a specific small-molecule inhibitor of NLRP3 inflammasome activity 24 , did not affect body weight changes (Fig. 3f), it reduced the severity of colitis in Slco2a1 −/− mice (Fig. 3g,h), diminished the Il1b and Tnf mRNA levels ( Fig. 3i) with suppressed formation of mature IL-1β and cleaved caspase-1 (Fig. 3j, Supplementary Fig. S5), suggesting that the NLRP3 inflammasome contributed to exacerbation of DSS colitis under Slco2a1 deletion.

Slco2a1 deficiency in macrophages, but not intestinal epithelial cells, alters susceptibility to DSS-induced colitis.
To investigate whether Slco2a1 deficiency in macrophages is associated with exacerbation of colitis, we developed mice with macrophage-specific deletion of Slco2a1 (Lysozyme M-cre;Slco2a1 fl/ fl , hereafter referred to as Slco2a1 ΔMP ). On only experimental day 4, Slco2a1 ΔMP mice treated with DSS exhibited more severe reductions in body weight than did Slco2a1 fl/fl littermates, although weight loss was not different between the two groups on experimental day 7 (Fig. 4a). Histological inflammatory changes were also significantly increased in Slco2a1 ΔMP mice (Fig. 4b,c). On the other hand, mice with intestinal epithelial cell-specific deletion of Slco2a1 (Villin-cre;Slco2a1 fl/fl , hereafter referred to as Slco2a1 ΔIEC ) showed less body weight reduction and milder intestinal inflammation than did Slco2a1 fl/fl littermates ( Fig. 4d-f).
Increased PGE 2 secretion by macrophages in the colon in Slco2a1 −/− mice might be a potential mechanism of exacerbation of DSS-induced colitis. As Slco2a1 encodes a PGT 6 , we next hypothesized that inflammasome activation in macrophages could be caused by alteration of the PGE 2 concentration by Slco2a1 deficiency. In colon tissue of DSS colitis mice on day 7, PGE 2 concentrations were significantly higher in Slco2a1 −/− than in WT mice ( Fig. 5a), whereas the concentrations did not differ between Slco2a1 ΔIEC and Slco2a1 fl/fl littermates (data not shown). Ptgs1 and 2 encoding PG synthesizing enzyme COX-1 and 2, respectively, and Hpgd encoding PG-inactivating 15-PGDH were downregulated in Slco2a1 −/− mice (Fig. 5b). Furthermore, PGE 2 concentrations in colon tissue were significantly higher in Slco2a1 ΔMP administered DSS than in Slco2a1 fl/fl littermates ( Supplementary Fig. S6a), although mRNA levels of Ptgs1, Ptgs2, Abcc4, and Hpgd were did not differ between both mice ( Supplementary Fig. S6b). Therefore, we next explored PGE 2 production by Slco2a1-deficient macrophages, which were identified as key players in DSS colitis exacerbation. PGE 2 concentrations in extracellular fluids of BMDMs after lipopolysaccharide (LPS) stimulation from Slco2a1 −/− mice were significantly higher than those in the WT counterparts, although intracellular concentrations did not differ (Fig. 5c,d). mRNA levels of Ptgs1, Ptgs2, Abcc4, and Hpgd were significantly decreased in LP macrophages from Slco2a1 −/− mice with DSS-induced colitis when compared with their WT counterparts (Fig. 5e). The decreased mRNA expression of these genes and the increased extra-to-intracellular PGE 2 concentration ratio in Slco2a1-deficient BMDMs indicated that Slco2a1 deficiency might contribute to increase in PGE 2 concentration via inhibition of transport of extracellular PGE 2 into the cytoplasm.
The PGE 2 /EP2 signalling pathway has been recently reported to boost pro-IL-1β production in BMDMs stimulated with LPS, resulting in mature IL-1β production via NLRP3 activation 25 . Together with our data, this finding suggests that Slco2a1 deficiency might lead to a decline in PGE 2 metabolism and thus, a high concentration of PGE 2 around macrophages, subsequently resulting in exacerbated intestinal inflammation via inflammasome activation in macrophages. In support of this hypothesis, although indomethacin treatment led to considerable bodyweight loss in both WT and Slco2a1 −/− mice given DSS compared to no treatment (Fig. 5f), Slco2a1 −/− mice administered DSS and treated with indomethacin exhibited significantly lower histological colitis scores than WT mice administered DSS and indomethacin, and Slco2a1 −/− mice administered DSS alone (Fig. 5g,h). To assess the mechanistic link between Slco2a1 deficiency and inflammasome activation via PGE 2 pathway, BMDMs from WT and Slco2a1 −/− mice were stimulated with LPS and PGE 2 /COX inhibitor indomethacin. In line with (2020) 10:4883 | https://doi.org/10.1038/s41598-020-61775-9 www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ the results in vivo, pro-IL-1β and caspase-1 induced by exogenous PGE 2 were cleaved by LPS stimulation. Those mature IL-1β and cleaved caspase-1 were boosted in BMDMs from Slco2a1 −/− mice compared to BMDMs from WT mice (Fig. 5i, Supplementary Fig. S7). The upregulated expression of mature IL-1β and cleaved caspase-1 were supressed by the COX inhibitor indomethacin, especially dramatically in BMDMs from Slco2a1 −/− mice, in support of in vitro experiment (Fig. 5i, Supplementary Fig. S7).

Discussion
This study showed that Slco2a1 deficiency led to high concentration of PGE 2 metabolic pathway, resulting in exacerbated intestinal inflammation via inflammasome activation in macrophages (Fig. 6), which might be a possible mechanism of CEAS.
Various studies have reported that PGs, which are synthesized from arachidonic acid by COXs, and selective prostanoid receptor agonists exert anti-inflammatory and mucosal protective effects in experimental colitis by inhibiting inflammatory cytokines and inducing mucus secretion in intestinal epithelial cells 26,27 . Clinical reports of NSAID-induced intestinal mucosal injury and basic evidence of genetic COX deletion-and COX inhibitor-exacerbated colonic injury in several models of colitis 28-30 support the beneficial effect of PGE 2 on intestinal epithelial cells.
On the other hand, studies reporting on roles of molecules related to PGE 2 metabolism in inflammatory conditions are limited. The PGE 2 metabolic pathway involves SLCO2A1, which mainly imports PGE 2 into the cytoplasm, and the metabolic enzyme 15-PGDH, which oxidizes PGE 2 to 15-keto-PGE 2 31 . 15-PGDH expression on intestinal epithelial cells is downregulated in patients with active IBD 1 , whereas 15-PGDH reportedly was upregulated by inflammatory cytokines, such as IL-6, in a prostate cancer cell line 32 . These findings indicate that the PGE 2 metabolic pathway also contributes to an altered response to inflammatory stimuli, depending on the cell type. In fact, Hpgd knockout or pharmacologic inhibition of 15-PGDH increased tissue PGE 2 levels and lowered the susceptibility to DSS colitis in mice 33 .
The present study revealed that germline Slco2a1-deficient mice were more susceptible to DSS-induced colitis than WT mice, but did not spontaneously develop enteritis or colitis. This result was opposite to what we expected based on findings in Hpgd-knockout mice 33 and the beneficial effects of PGs and selective prostanoid receptor agonists in the intestine 26,27 . Therefore, we conducted an unbiased microarray assay to identify mechanisms for www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ the exacerbation of experimental colitis in Slco2a1-knockout mice. Intriguingly, microarray analysis of colon tissues revealed that inflammasome-related genes, such as Il1b, Il18, and P2rx7, and macrophage-related chemokines, such as Ccl2 and Ccl4, were upregulated in Slco2a1-knockout mice administered DSS when compared with WT mice treated with DSS. The cytokine IL-1β has been implicated as a central mediator of the inflammatory processes in patients with IBD and in experimental colitis [34][35][36] . Inactive pro-IL-1β is cleaved by caspase-1, which is activated within the inflammasome, to generate the mature, active form 37 . The inflammasome is formed by NLRP3, the adaptor protein apoptosis-associated speck-like protein containing a caspase-recruitment domain, and caspase-1 18 . Protein levels of mature IL-1β and cleaved caspase-1 were increased in colon tissues of Slco2a1-knockout mice treated with DSS, which led us to hypothesize that NLRP3 inflammasome activation might contribute to the exacerbation of DSS colitis in Slco2a1-deficient mice. This hypothesis was proven true by administering mice MCC950, a specific inhibitor of the NLRP3 inflammasome 24,38 , as MCC950 administration reduced the severity of colitis in Slco2a1-deficient mice via suppressing the expression of mature IL-1β and cleaved caspase-1.
Several studies have shown that macrophages are the main site of IL-1β production in the colon in IBD patients [39][40][41] . The NLRP3 inflammasome and subsequent secretion of mature IL-1β by macrophages has been reported as a main mechanism of intestinal inflammation in model mice 19,25 . In line with these findings, LP macrophages and BMDMs isolated from Slco2a1-deficient mice showed stronger inflammasome activation than WT mice. Macrophage-specific, but not intestinal epithelial cell-specific Slco2a1 knockout increased the susceptibility to DSS-induced colitis, partially phenocopying systemic Slco2a1-deficient mice, which indicated that Slco2a1 in macrophages might play an important role in maintaining mucosal homeostasis in conditions of injury.
The present study showed that Slco2a1 deficiency led to an increased PGE 2 concentration in colon tissues. In-vitro experiments using BMDMs corroborated that PGE 2 concentrations in extracellular fluids of BMDMs from Slco2a1-deficient mice were higher than those in their WT counterparts, possibly through suppression of PGE 2 metabolism. A recent study demonstrated that PGE 2 /EP2 signalling pathway facilitated pro-IL-1β production in BMDMs stimulated with LPS, resulting in mature IL-1β by the NLRP3 inflammasome activation 25 . Increased PGE 2 production was related with macrophage infiltration in the intestine in mice with DSS-induced colitis 42 . Taking these findings together, it can be considered that the NLRP3 inflammasome in macrophages was autocrinally activated by the elevated concentrations of PGE 2 , leading to exacerbation of colitis in Slco2a1-deficient mice. In support of this, exacerbated colitis in mice lacking Slco2a1 was improved by treatment with the COX inhibitor indomethacin as the concentration of PGE 2 around macrophages decreased. On the other hand, in WT mice, colitis was worsened by indomethacin administration, which is consistent with finding by Okayama et al. 43 . These implied that the severity of colitis would depend on the balance between PGE 2 mediated anti-inflammatory effect on epithelium and pro-inflammatory effect on activated macrophage. Indomethacin treatment led to considerable body weight loss in both WT and Slco2a1-deficient mice treated with DSS when compared to non-treated animals, but this was presumed to be because of mucosal injury of the stomach or small intestine caused by indomethacin [43][44][45] . www.nature.com/scientificreports www.nature.com/scientificreports/ Our observation that increased PGE 2 levels exacerbated colitis is in contrast to findings in Hpgd-deficient mice by Zhang et al. 33 , who reported that deletion of 15-PGDH increased tissue PGE 2 levels and protected mice from experimental colitis. A possible explanation could be the difference in the dominant locations of SLCO2A1 and 15-PGDH expression. Zhang et al. concluded that the phenotype of Hpgd-deficient mice is due to 15-PGDH deletion in colonic epithelial cells, which are the primary site of 15-PGDH expression in the colon 46 . On the other hand, our study revealed that the phenotype of Slco2a1-deficient mice is mainly due to Slco2a1 deletion in macrophages. In fact, intestinal epithelial-specific Slco2a1-deficient mice has lower susceptibility to experimental colitis as with the model in the study by Zhang et al. Although suppressed PGE 2 metabolism in macrophages might contribute to the enhanced PGE 2 concentration, Slco2a1 ΔMP only partially reproduced the systemic knockout phenotype in the present study. As SLCO2A1 is mainly expressed on vascular endothelial cells in the intestinal tracts of mice 47 and humans 12 , a decrease in the clearance of PGE 2 from vascular endothelial cells might also contribute to the increase in the PGE 2 concentration around macrophages in the intestine. Further studies will clarify the role of SLCO2A1 on endothelial cells in intestinal homeostasis.
Although a recent study has demonstrated that SLCO2A1 function is lost in patients with CEAS 12 , the current study revealed that Slco2a1-deficient mice did not spontaneously develop enteritis. This might be because Slco2a1/SLCO2A1 has different functions in vivo in mice and humans. Patients with CEAS might have a "second-hit factor" other than lack of SLCO2A1, such as other genetic factors or certain pathogens in the intestinal lumen. Considering the clinical impact, these mice model may not be the best model to analyse CEAS which affects mainly small intestine. Further studies using another enteritis model are needed to clarify the pathogenesis of CEAS.
In summary, Slco2a1 deficiency increased the PGE 2 concentration around macrophages, possibly by suppressing PGE 2 metabolism, resulting in activation of the NLRP3 inflammasome in macrophages and thus exacerbating intestinal inflammation in an experimental colitis model. Our findings shed some light on the pathogenesis of the intestinal inflammation associated with SLCO2A1 and might provide a novel therapeutic target. Dextran sodium sulphate (DSS)-induced colitis and treatments. Acute colitis was induced by 3.5% (w/v) DSS (molecular weight, 5000; Fujifilm, Osaka, Japan) added to the drinking water for 7 days. Seven-to nine-week-old gender-matched mice were included in each group. Body weight was recorded daily. Mice were sacrificed on day 3 or 7, and the colons were collected. To inhibit the NLRP3 inflammasome in vivo, mice received an intraperitoneal injection of 50 mg/kg MCC950 (AdipoGen Life Sciences, San Diego, CA) (or PBS as a control) every other day from one day before to day 7 of DSS administration. To inhibit PGE 2 production in vivo, indomethacin (Nacalai Tesque, Kyoto, Japan) was dissolved in ethanol (10 mg/mL) and was added to the drinking water at 1 mg/kg per day, concurrent with DSS treatment.

Animals. Slco2a1
Histopathological scoring of enterocolitis. Tissues were flushed with PBS and fixed 4% paraformaldehyde phosphate buffer solution for 5 days prior to paraffin embedding. Sections (4 μm) were then cut and stained with haematoxylin and eosin (H&E) and alcian blue/periodic acid-Schiff (AB/PAS) using an Alcian Blue-PAS Stain Kit (ScyTek Laboratories, Logan, UT) following the manufacturer's instructions. Ileum or colon sections were evaluated using a histological scoring system 49 or histological grading of colitis 50,51 . Goblet and Paneth cells were quantified in at least 100 crypts of each mouse.
Isolation and stimulation of murine macrophages. LP macrophages were isolated from the mouse colons as described previously 52 . CD11b-positive cells among LP cells were purified by magnetic-activated cell sorting using CD11b MicroBeads (130-049-601; Miltenyi Biotec, Bergisch Gladbach, Germany) following the manufacturer's instructions.

RNA extraction and quantitative reverse-transcription polymerase chain reaction (RT-qPCR).
Total RNA was extracted from tissues and cells using an ISOGEN II kit (Nippon Gene, Tokyo, Japan) following the manufacturer's protocol. A High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific) was used per the manufacturer's protocol to convert the total RNA to complementary DNA. qPCRs were run in duplicate using SYBR Select Master Mix (Thermo Fisher Scientific) in an Applied Biosystems 7500 Fast Real-Time PCR system, and data were analysed with the in-built software (Thermo Fisher Scientific). Thermal cycling conditions were as follows: 50 °C for 2 min and 95 °C for 2 min, 40 cycles of denaturation at 95 °C for 3 s and annealing at 60 °C for 30 s. The primers used are listed in Supplementary Table S1. Primers were designed by using Primer-BLAST in the NCBI genome browser. Gene expression was normalized to the expression level of Hprt mRNA.
Colon tissue explant cultures. Colon sections (1 cm) from individual mice were washed with PBS to remove faecal contents and were then cultured in 24-well plates in 500 μL of RPMI 1640 (Nacalai Tesque) containing 2% FBS and antibiotics at 37 °C and in the presence of 5% CO 2 . After a 24-h incubation, the culture medium was collected and centrifuged. The supernatants were transferred to new tubes for enzyme-linked immunosorbent assay (ELISA) of PGE 2 . Microarray analysis. Total RNA extracted from colon tissues was purified using an RNAqueous-Micro Kit (Ambion, Austin, TX) for microarray analysis. The quality and quantity of purified RNA were assessed using an Agilent Technologies 2100 Bioanalyzer and NanoDrop spectrophotometer, respectively, and the RNAs were applied to a SurePrint G3 Mouse Gene Expression 8 × 60 K (Agilent, Inc., Santa Clara, CA). Raw data were extracted using Agilent Feature Extraction Software (v11.0.1.1). The raw data for each gene were automatically summarized in the Agilent feature extraction protocol to generate a raw data text file, providing expression data for each gene probed on the array. Data for array probes that had a Flag A were filtered out. Selected gProcessed-Signal values were log-transformed and normalized by the quantile method. Hierarchical cluster analysis was performed using complete linkage and Euclidean distance as a measure of similarity. For GSEA 55 , the GSEA v3.0 software (Broad institute, Massachusetts Institute of Technology and Regents of the University of California) was performed with gene set database c5.all.v6.2.symbols.gmt. The number of random sample permutations was set at 1000. Statistical analysis. The data are presented as means ± standard errors of the mean (SEMs) or medians plus interquartile ranges (IQRs). Significance of differences among groups was determined by Student's t-test, Welch's t-test, or Mann-Whitney U test. P < 0.05 was considered significant.