Members of the endocannabinoid system are distinctly regulated in inflammatory bowel disease and colorectal cancer

Preclinical studies have demonstrated that the endocannabinoid system (ECS) plays an important role in the protection against intestinal inflammation and colorectal cancer (CRC); however, human data are scarce. We determined members of the ECS and related components of the ‘endocannabinoidome’ in patients with inflammatory bowel disease (IBD) and CRC, and compared them to control subjects. Anandamide (AEA) and oleoylethanolamide (OEA) were increased in plasma of ulcerative colitis (UC) and Crohn’s disease (CD) patients while 2-arachidonoylglycerol (2-AG) was elevated in patients with CD, but not UC. 2-AG, but not AEA, PEA and OEA, was elevated in CRC patients. Lysophosphatidylinositol (LPI) 18:0 showed higher levels in patients with IBD than in control subjects whereas LPI 20:4 was elevated in both CRC and IBD. Gene expression in intestinal mucosal biopsies revealed different profiles in CD and UC. CD, but not UC patients, showed increased gene expression for the 2-AG synthesizing enzyme diacylglycerol lipase alpha. Transcripts of CNR1 and GPR119 were predominantly decreased in CD. Our data show altered plasma levels of endocannabinoids and endocannabinoid-like lipids in IBD and CRC and distinct transcript profiles in UC and CD. We also report alterations for less known components in intestinal inflammation, such as GPR119, OEA and LPI.

Despite a wealth of preclinical data demonstrating that the ECS controls the well-being and normal functioning of the GI tract, little evidence exists whether this also holds true for the human GI tract. Since the use of medicinal Cannabis has been introduced in several countries 15 , knowledge on the human ECS is of great importance for a potential therapy with cannabinoids. Surveys [16][17][18][19] and observational/prospective studies 20-23 revealed that IBD patients often self-medicate with Cannabis to alleviate abdominal pain and diarrhea. So far, only one small prospective human trial has investigated the effect of tetrahydrocannabinol (THC) in CD patients, who showed some kind of benefit 22 . Another trial investigated the effect of cannabidiol (CBD), a Cannabis constituent with little or no affinity to CB receptors 24 in CD patients, revealing no beneficial effect 25 despite the fact that CBD improves inflammation in animal models of IBD 5 . A survey by Storr et al. showed that Cannabis use for more than six months was even a strong predictor for surgery in CD patients 18 . Therefore, in contrast to preclinical studies, the role of the ECS in humans is not so clear. Several studies describe deregulation of ECS receptors, enzymes and endocannabinoids in colonic biopsies and whole tissue from IBD patients [26][27][28] . A change in ECS components in human CRC tissue samples was reported, indicating overproduction of endocannabinoids 29,30 as well as high 30 or low 31 presence of CB 1 . High presence of CB 1 was indicative of a poorer prognosis in stage II microsatellite stable tumour patients 32 and stage IV CRC patients 33 . Also CB 2 was shown to be a marker for poor prognosis in CRC patients 34 although studies in mice have suggested a protective role for CB 1 31,35 and CB 2 8 . In contrast, Ligresti et al. did not observe differences in CB 1 and CB 2 expression between normal and CRC tissue 29 . These contradicting results warrant closer examination of the ECS and the endocannabinoidome in both humans and mice.
In a recent study we could demonstrate that GPR55 plays a pro-oncogenic role in experimental CRC and described a significant association between high GPR55 expression and shortened relapse-free survival of CRC patients 35 . In our present study we have chosen to investigate levels of endocannabinoids, endocannabinoid-like lipids and lysophosphatidylinositol (LPI), the endogenous ligand of GPR55, in plasma of IBD and CRC patients to see whether deregulation of these metabolites is also detectable in the blood. This is of importance as they may become useful as biomarkers of disease. In addition, we took biopsies from patients with UC and CD, and from control subjects to measure gene expression of various receptors and enzymes of the ECS and the endocannabinoidome using NanoString technology. As many commercial antibodies against GPR55 and CB 2 receptors lack specificity in immunohistochemical stainings 36,37 , we chose to investigate their localization in colonic biopsies of UC and CD patients and in colon cancer tissue by a novel and specific in situ hybridisation (ISH) technique (RNAscope) that we have successfully used in murine models of intestinal inflammation 37 .

Results
Differences in plasma endocannabinoid and endocannabinoid-like lipid levels between IBD/ CRC patients and control subjects. AEA, PEA and OEA were markedly increased in plasma of UC and CD (although PEA only slightly missed significance in UC (p = 0.0539) and CD (p = 0.0551) (Fig. 1a-c). The other main endocannabinoid, 2-AG, also showed an increase in IBD patients in comparison to control subjects, reaching significance (p = 0.0151) in CD, but not in UC (p = 0.2318; Fig. 1d). With regard to LPI (the endogenous GPR55 agonist), both LPI 18:0 and LPI with an arachidonic acid moiety (20:4) showed higher levels in IBD in comparison to control subjects ( Fig. 1f-g). In contrast, levels of LPI 16:0 were lower in UC patients vs. controls (Fig. 1e). We observed an increase of 2-AG, but not of AEA, PEA and OEA, in CRC, while only LPI 20:4 and not LPI 18:0 and 16:0 was higher in CRC plasma in comparison to controls ( Fig. 1e-g).
Consistent with the fact that they share similar synthetic and degradative pathways 38,39 plasma concentrations of AEA, PEA and OEA highly correlated in control and diseased individuals (except for CD patients) indicating concerted production and metabolism of these lipids (Fig. 2a). We also wanted to know whether some of the endocannabinoids would correlate with the severity of CD and UC and we found that only OEA levels correlated with the Harvey Bradshaw Index (HBI) index in CD patients but not with the total Mayo score (which describes severity of UC) (Fig. 2b). Levels of 2-AG negatively correlated with the HBI of CD patients, but not with the total Mayo score of UC patients (Fig. 2c). Correlations of endocannabinoids and endocannabinoid-like lipids with the other parameters mentioned in Tables 1 and 2 for UC and CD patients (endoscopic Mayo score, SES-CD score, white blood cell count, C-reactive protein, calprotectin) had p values > 0.05.
Intestinal mucosal biopsies from UC and CD patients clearly showed differential patterns of gene expression. In general, the fold increases/decreases were more pronounced in CD than UC patients. Most of the investigated transcripts showed tendencies of decrease in UC patients (Fig. 3a) in comparison to the ones from controls, whereas in CD patients (Fig. 3b), it was the opposite for all endocannabinoid-synthesizing and -degrading enzymes investigated. Among the synthesizing enzymes, mRNA levels of NAPE-PLD were decreased in UC, whereas in CD, we observed an increase of DAGLalpha. In contrast to the enzymes, expression of receptors like CNR1 (which encodes CB 1 ) and GPR119 were decreased in CD in comparison to controls whereas expression of CNR2 (which encodes CB 2 ), GPR18, and GPR55 were less affected. Notably, PPARdelta increased in CD. As an AEA responsive receptor, TRPV1 expression was higher in people with CD than in controls (Fig. 3b). For a complete list of fold changes and p values see Table S1 in the supplementary information.
Localization of GPR55 and CB2 in IBD and CRC patients. To demonstrate presence and distribution of a non-cannabinoid (but cannabinoid-responsive) G protein-coupled receptor and a cannabinoid receptor, we performed GPR55 and CB 2 mRNA staining by ISH in sections of colonic mucosal biopsies, as depicted in Fig. 4. In tissue of control subjects, low levels of GPR55 mRNA were detected in few epithelial cells and in cells of the lamina propria, predominantly lymphocytes (Fig. 4a). Biopsies of CD patients showed very low mRNA staining in the lamina propria (Fig. 4b). In UC patients, GPR55 mRNA was generally low, but locally higher in epithelial cells (Fig. 4c, arrows). In CRC tissue, low GPR55 gene expression was detected in epithelial tumour and few other, possibly inflammatory, cells (Fig. 4d). www.nature.com/scientificreports www.nature.com/scientificreports/ Low CB 2 receptor mRNA levels were found in epithelial cells of control subjects (Fig. 4e) and in the lamina propria of CD patients (Fig. 4f) as well as UC patients (Fig. 4g). Low CB 2 receptor mRNA was also detected in some tumour cells of tumour sections from CRC patients (Fig. 4h).

Discussion
Although a considerable amount of knowledge has been gained from experimental models of intestinal inflammation and CRC about the protective effects of cannabinoids, there is still a wide gap between preclinical evidence and the role of the ECS in human GI diseases. So far, little evidence exists whether cannabinoids are beneficial in human IBD and CRC, as large clinical trials are still missing. Our study was, therefore, designed to measure changes within the members of the ECS and members now regarded as part of the endocannabinoidome (e.g. GPR55, GPR119, PPARs, PEA, OEA, LPI etc.) 11 in human IBD and CRC to evaluate potential targets for pharmacological therapy. We also chose to measure endocannabinoids and several endocannabinoid-like lipids in plasma, a more convenient procedure to perform when determining biomarkers.    Supplementary Table S1. www.nature.com/scientificreports www.nature.com/scientificreports/ The theory states that the ECS is upregulated in inflammatory conditions to reinstate homeostasis 57 . According to our findings, this may be particularly true for AEA, PEA, and OEA, which were increased in IBD, but not in CRC. In line with our results, an increase of AEA (but not of 2-AG) and PEA was already described in colonic mucosal biopsies of UC patients in comparison to controls 58,59 . We also observed an increase of 2-AG in CD and, consistent with reports by Sailler et al. 60 , in CRC. Ligresti et al. also measured increased levels of AEA in colorectal tissue of CRC patients 29 . OEA has not yet been investigated in models of intestinal inflammation but seems to be part of a concerted upregulation of endocannabinoids and endocannabinoid-like lipids. That OEA shares pathways of synthesis and degradation with PEA and AEA 38,39 , is indicated by the close correlation of AEA, PEA and OEA levels (Fig. 2a). The increase in OEA is of considerable interest as it correlates with the HBI in CD patients. This endocannabinoid-like lipid is a most potent GPR119 agonist 61 . GPR119 has been linked with satiety and the brain-gut axis 62 , rather than with inflammation, but an article recently indicated a role in the prevention of steatohepatitis 63 . GPR119 may also mediate release of GLP-1 64 , a peptide that was shown to increase in DSS colitis 65 and to promote anti-inflammatory mechanisms 66 . In our study, GPR119 was found downregulated in IBD patients, which may have been caused by the increased presence of OEA.
Next to endocannabinoids and endocannabinoid-like lipids, we were also interested in LPI, the endogenous ligand for GPR55, in order to understand how this ligand is regulated in IBD and CRC. Elevated levels of LPI have been shown in ovarian cancer 67 and in a small cohort of colon cancer patients 68 . Stearoyl-LPI (18:0) and arachidonoyl-LPI (20:4) are the most abundant fatty acids in LPI, but LPI 20:4 was the most potent among them to produce responses in GPR55-HEK293 cells 69 . In mouse models of intestinal inflammation and CRC, GPR55 was demonstrated to act as a pro-inflammatory and pro-carcinogenic receptor 35,45 . The strong increase of LPI 18:4 and 20:4 in IBD and CRC may indicate an active LPI-GPR55 axis in these diseases. This concept is strongly supported by a new study showing increased LPI levels in mice with chronic and acute intestinal inflammation 70 . GPR55 transcripts, however, were not upregulated in the IBD cohorts probably indicating, similar to GPR119, receptor downregulation by increased presence of the ligand.
Evaluation of transcripts by NanoString analysis revealed differential regulation of endocannabinoidsynthesizing/-degrading enzymes in IBD patients. Most of all, expression levels of NAPE-PLD and DAGLalpha differed between UC and CD. The 2-AG producing enzyme DAGLalpha was increased in CD patients, which is in line with the high 2-AG levels measured in plasma. The purpose of high 2-AG availability may be to ameliorate inflammation, as suggested by Alhouayek et al. 40 and our own findings (2-AG levels are lower in cases of a high HBI; Fig. 2b). In contrast, UC patients showed decreased NAPE-PLD expression but high levels of AEA, PEA and OEA. Although our results on NAPE-PLD expression are supported by immunohistochemical findings by Marquez et al. 27 , it is not quite clear why endocannabinoids and endocannabinoid-like lipids were increased while their synthesizing enzyme was downregulated. An allosteric feedback inhibition of NAPE-PLD by AEA (or PEA and OEA) or a reduced degradation of the endocannabinoids/-like lipids would be likely explanations although it is possible that N-acylethanolamine-producing enzymes other than NAPE-PLD were responsible for the increase of AEA, PEA and OEA 71 . ABHD6, a serine hydrolase known to regulate 2-AG efficacy at CB 2 receptors 72 was decreased in UC, but its role in the GI tract has not yet been elucidated. Recent reports state that ABHD6 may control 2-AG levels in macrophages and that inhibition of ABHD6 increase LPI 20:4 levels in LPS-activated macrophages 70,73 . www.nature.com/scientificreports www.nature.com/scientificreports/ Most of the CB and non CB receptor transcripts, in particular, those of CB 1 and GPR119, were downregulated in IBD (not significantly in UC). Although we should keep in mind that many of the CB and non CB receptors are expressed in lamina propria cells, which are sensitive to steroids and immunosuppressants, CB 1 and GPR119 are mainly localized to epithelial, enteric neuronal and enteroendocrine cells of the gut 74,75 suggesting that they should be little affected by immunosuppressants. Thus, one explanation for the decrease in CB 1 gene expression could be that CB 1 was downregulated as a response to high levels of its ligand AEA which could have led to desensitization, a common feature in GPCRs with high ligand exposure 76 .
By using immunohistochemistry and immunoassays, several studies performed in mucosal tissue of IBD patients all described upregulation of CB 1 26,28 and, therefore, they seem to be in contrast to our results. However, one of the reasons why we opted for evaluating transcripts instead of protein is the fact that many commercially available CB 1 77 and CB 2 36 antibodies are not specific. In addition, as CB 1 78 and CB 2 79 receptors can be internalized and recycled, receptor protein regulation may very well differ from regulation of transcripts. Additionally, ISH stainings did not indicate upregulation of CB2 mRNA (Fig. 4f).
mRNA of PPAR receptors, in particular of PPARalpha and PPARgamma, were also measured as they have been frequently implicated in anti-inflammatory actions of cannabinoids in intestinal epithelial cells 80 . Notably, they were regulated differentially in UC and CD.
Since we chose to study levels of endocannabinoids/-like lipids in plasma, the study has certain limitations and caveats. Plasma levels of endocannabinoids and endocannabinoid-like lipids can be associated with body mass index (BMI) and HDL-cholesterol levels 81 , however, this has been only shown for 2-AG and not for AEA, PEA and OEA in obese people with high, but not low intra-abdominal adiposity 81 . Also LPI levels were shown to be increased in obese people 82 . According to that study, obese people also have increased levels of LPI 16:0 which we did not measure in our disease cohorts. The mean BMIs of our UC (23.7 ± 4.0 kg/m 2 ) and CD (23.3 ± 3.6 kg/m 2 ) cohorts do not lie in the obesity range, indicating that obesity can be excluded as a confounding factor. Another caveat is the age difference between CRC patients and control subjects. But similar to our results, previous studies also showed increased 2-AG levels in colonic mucosal biopsies 29 and plasma 60 from CRC patients indicating that the differences we have seen in our samples are likely unrelated to age differences.
In summary, our data have highlighted the role of the 'endocannabinoidome' and in particular of the ECS in human IBD and CRC showing a distinct, but not general regulation of their components in IBD and CRC. The increase of endocannabinoids and endocannabinoid-like lipids in IBD seems to be a concerted action, but only OEA and 2-AG correlated, although oppositely, with disease severity in CD patients. In addition, regulation of endocannabinoid-synthesizing/-degrading enzymes is different to CB receptor expression. The study also suggests that LPI may have a proinflammatory role in human IBD, supported by recent data of murine intestinal inflammation 70 . Although frequently suggested as a homeostatic mechanism to injury in the GI tract, we failed to see an upregulation of CB receptor mRNA in the colon of IBD patients by NanoString. As many herbal and synthetic cannabinoid (also non-psychotropic) products are available, larger clinical trials are worthwhile to test their efficacy in GI diseases Methods patients. Adult patients with confirmed UC (n = 17; mean age ± SD: 45 ± 13.1; 8 males/9 females), CD (n = 15; mean age ± SD: 40.3 ± 9.5; 11 males/4 females), CRC (n = 12; mean age ± SD: 75.5 ± 10.3; 10 males/2 females) and control subjects (n = 19; mean age ± SD: 34.6 ± 15.1; 6 males/13 females) were included in the study. UC and CD patients were recruited from the IBD clinic of the Department of Internal Medicine at the Medical University of Graz; CRC patients were recruited as part of the OncoTrack project (http://www.oncotrack.eu/) by the General Hospital Graz West and the St. John of God Hospital Graz, control subjects were recruited from the endoscopy unit of the Department of Internal Medicine and the Division of Pharmacology, Medical University of Graz. For the measurement of endocannabinoids and LPI in plasma, blood was collected in Vacuette EDTA tubes (Greiner-Bio-One, Austria), centrifuged at 1500xg for 15 min, aliquoted and frozen at −80 °C until use.
Diagnosis of UC and CD was established by standard clinical, endoscopic and histologic criteria 83,84 . All UC and CD patients with active disease also underwent contemporaneous colonoscopy to assess endoscopic disease activity. For UC patients, disease activity were assessed using the total Mayo score and endoscopic activity by the endoscopy Mayo subscore 85 . For CD patients, activity was assessed clinically by the Harvey Bradshaw Index (HBI) and endoscopically by the Simple Endoscopic Score for Crohn's Disease (SES-CD) [86][87][88] . Endoscopic biopsies were collected from inflamed intestinal mucosal area from UC and CD patients. Colonoscopy was performed as part of the clinical workup because of active disease. Normal colonic biopsies were obtained from a control group with a normal colonoscopy recruited from patients undergoing colonoscopy as part of the colon cancer screening program and from patients with diagnostic workup of occult or overt gastrointestinal bleeding. All subjects were recruited from the endoscopy unit of the Department of Internal Medicine, Medical University of Graz. Those with significant comorbidities, intercurrent illness such as infections, and pregnant women were excluded from the study. Collected biopsies were immediately frozen or fixed in 10% phosphate buffered formalin for histochemical analysis. All subjects suffering from UC and CD were on some sort of medication. Characteristics and medication of CD and UC patients are shown in Tables 1 and 2. Characteristics of CRC patients are shown in Table 3. UICC classifications are according to Sobin 89 .
Ethical approval and informed consent. Ethical approval was granted by the ethics committee of the Medical University of Graz and confirmed by the ethics committee of the St John of God Hospital Graz (protocol numbers: 24-281 ex 11/12; 23-015 ex 10/11; 17-291 ex 05/06 and 23-002 ex 10/11). Procedures were carried out in accordance with international guidelines. All participants provided written, informed consent. All samples and medical data used in this study were irreversibly anonymized.
www.nature.com/scientificreports www.nature.com/scientificreports/ Determination of endocannabinoids, endocannabinoid-like lipids and LPI in plasma by LC-MS measurement. The determination of the endocannabinoids AEA and 2-AG, the structurally related compound 1-arachidonoyl glycerol (1-AG), and of the endocannabinoid-like lipids OEA and PEA was performed by liquid-liquid-extraction in combination with ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/ MS). Reference substances and internal standards were provided by Cayman Chemical (Ann Arbor, MI, USA). For all analytes isotopically labeled analogues were used as internal standards. Working solutions were prepared in acetonitrile as mixture of all analytes and internal standards, respectively.
The LC-MS/MS system consisted of a triple quadrupole mass spectrometer QTRAP 6500+ (Sciex, Darmstadt, Germany) equipped with a Turbo Ion Spray source operated in positive electrospray ionization mode and an Agilent 1290 Infinity LC-system with binary HPLC pump, column oven and autosampler (Agilent, Waldbronn, Germany). The chromatographic separation was performed using an Acquity UPLC BEH C18 2.1 × 100 mm column (particle size of 1.7 µm, Waters, Eschborn, Germany) in combination with an UHPLC Fully Porous C18 column guard (Phenomenex, Aschaffenburg, Germany). A gradient elution using water with 0.0025% formic acid (solvent A) and acetonitrile with 0.0025% formic acid (solvent B) was applied.
Plasma samples were thawed in a refrigerator and processed on ice all the time. The sample (100 µL) was mixed with 20 µL acetonitrile and 20 µL IS working solution. The mixture was vortexed and centrifuged. Afterwards the mixture was extracted once with 200 µL ethyl acetate/hexane (9/1, v/v). The upper phase was removed after vortexing and centrifuging and evaporated at 45 °C under a gentle stream of nitrogen. The sample was reconstituted in 50 µL acetonitrile. Phosphate-buffered-saline (PBS) was used as surrogate matrix for the preparation of calibration standards and quality control samples. A volume of 100 µl PBS was spiked with 20 µL of a working solution and processed as stated before starting with the addition of 20 µL IS working solution.
Data acquisition was done using Analyst Software 1.6.3 and quantification was performed with MultiQuant Software 3.0.2 (both Sciex, Darmstadt, Germany), employing the internal standard method (isotope dilution mass spectrometry). Calibration curves were calculated by linear regression with 1/x weighting.
Further information on the used gradient, mass spectrometric parameters, precursor to product ion transitions (m/z), internal standards as well as the lower and upper limit of quantification can be found in the Supplementary Table S2. For LPI measurement, lipid was extracted according to Matyash et al. 90 . Liquid chromatography-mass spectrometry (LC-MS) measurements for LPI quantification and data processing were performed with slight modifications as previously described in Triebl et al. 91  RNAscope in situ hybridisation (ISH). Tissue was fixed and dehydrated, embedded in paraffin, cut in 5 μm sections, and finally dried and stored before submitting it to RNAscope ISH.  www.nature.com/scientificreports www.nature.com/scientificreports/ RNAscope is an advanced ISH method where two adjacent probes (so-called ZZ probes) are needed to bind to the target sequence to develop a signal. This method provides the possibility of detecting a low number of mRNAs in peripheral tissue due to a decreased background noise. Twenty ZZ probes for human CB 2 (targeting bases 141-1193 of NM_001841.2) and GPR55 (targeting bases 2-1057 of NM_005683.3) (purchased from Advanced Cell Diagnostics [ACD] Newark, USA) were used to detect the corresponding mRNAs in intestinal mucosal biopsies of controls, IBD and CRC patients. The ISH was performed according to the manufacturer's protocol by using RNAscope 2.5 HD brown kit (ACD).
In brief, tissue sections were baked at 60 °C for 1 h, de-waxed, rehydrated and treated with H 2 O 2 for 10 min. Target retrieval was performed by cooking of tissue in retrieval buffer using the Brown FS3000 food steamer for 15 min, each step followed by washes in H 2 O dest. The next day, sections were digested with Protease Plus at 40 °C for 20 min, washed, followed by incubation with the corresponding probes at 40 °C for 2 h. The procedure was continued according to the manufacturer's protocol. CB 2 and GPR55 mRNAs were stained using 3,3'-diaminobenzidine (DAB; provided by ACD). Sections were counterstained with 1:5 dilutions of Gill's II hematoxylin, washed, dried and mounted with Vectamount mounting medium (Vector Laboratories). Sections from patients and control subjects were put on one slide for comparison. Brightfield images were taken using an Olympus BX41 microscope and an Olympus UC 90 digital camera connected with Olympus CellSense standard 1.17 imaging software (Olympus, Vienna, Austria). Contrast, brightness and color balance of images were adjusted using Corel Photo Paint.

statistics.
Mass spectrometry data analysis. Data are shown as mean +/− standard deviation (SD).
Statistical analysis was performed using GraphPad Prism 7.05 (GraphPad Software, La Jolla, CA, USA). Cohorts of control subjects and patients were compared by unpaired two-tailed Student's t-test. For data with varying SDs, two-tailed t-test with Welsh correction was applied. Normal distribution of the data was tested using d' Agostino and Pearson omnibus normality test. Single outliers were detected using Grubb's test. For LPI 20:4 measurement in control subjects, two outliers were detected using the ROUT method (Q = 5%) and excluded. For 2-AG measurement, 3 samples were below/above the detection limit and were, therefore, omitted. P values < 0.05 were considered significant.
Correlations. Data were subjected to normality test (d' Agostino & Pearson omnibus). In case of normal distribution, Pearson correlation was performed (deployed for all correlations shown). For groups without normal distribution, nonparametric correlation (Spearman) was applied.
NanoString data analysis. RCC raw data files were imported to nSolver 4.0 Software (NanoString Technologies, Seattle, WA USA) and raw data pre-processing and normalization was performed according to standard procedures (positive control and codeset normalisation using reference genes). Log2 normalized data were imported to Partek Genomic Suite Software v6.6 (Partek Inc, St Louis, MO) for statistical analysis. One-Way ANOVA was performed between sample groups. Genes with p values < 0.05 and fold changes of at least 1.5 were considered to be significantly regulated.