Author Correction: Stimulation of hepatocarcinogenesis by activated cholangiocytes via Il17a/f1 pathway in kras transgenic zebrafish model

It has been well known that tumor progression is dependent on secreted factors not only from tumor cells but also from other surrounding non-tumor cells. In the current study, we investigated the role of cholangiocytes during hepatocarcinogenesis following induction of oncogenic krasV12 expression in hepatocytes using an inducible transgenic zebrafish model. Upon induction of carcinogenesis in hepatocytes, a progressive cell proliferation in cholangiocytes was observed. The proliferative response in cholangiocytes was induced by enhanced lipogenesis and bile acids secretion from hepatocytes through activation of Sphingosine 1 phosphate receptor 2 (S1pr2), a known cholangiocyte receptor involving in cholangiocyte proliferation. Enhancement and inhibition of S1pr2 could accelerate or inhibit cholangiocyte proliferation and hepatocarcinogenesis respectively. Gene expression analysis of hepatocytes and cholangiocytes showed that cholangiocytes stimulated carcinogenesis in hepatocytes via an inflammatory cytokine, Il17a/f1, which activated its receptor (Il17ra1a) on hepatocytes and enhanced hepatocarcinogenesis via an ERK dependent pathway. Thus, the enhancing effect of cholangiocytes on hepatocarcinogenesis is likely via an inflammatory loop.


Results
Increase of cholangiocytes upon induction of oncogenic kras V12 expression in hepatocytes of kras V12 transgenic zebrafish larvae. To visualize the response of cholangiocytes upon kras V12 induction in hepatocytes. 3-dpf (day postfertilization) kras V12 transgenic (shorted as kras + in this report) larvae were induced by doxycycline (Dox) for 5 days to initiate hepatocarcinogenesis. Two specific cholangiocyte markers, www.nature.com/scientificreports/ Alcam and Cytokeratin 18 (Ck18), were used for identifying cholangiocytes in the liver sections. As shown in Fig. 1A-D, both Alcam and Ck18 stained cholangiocytes showed significant increases in the kras + group in comparison with the wildtype (WT) control group upon induction of oncogenic kras V12 expression by Dox. Cholangiocyte density was determined following Dox induction from 8 to 96 h. A shown in Fig. 1E,F, Alcam staining revealed significant increases in cholangiocyte density in the kras + group compared to the WT group at all time points and a significant increase was observed as early as 8 h after the initiation of Dox treatment.

Acceleration of tumor growth via S1pr2 induction. Several cholangiocyte receptors have been
reported to control cholangiocyte proliferation under different physiological and pathological conditions 8,9,11,40 . These receptors include secretin receptor (Sctr) 41 , estrogen receptor (Esr), insulin receptor (Insra and Insrb), glucagon-like-peptide receptor 2 (Glpr2), nerve growth factor receptor (Ngfr) and sphingosine 1 phosphate receptor 2 (S1pr2) 40,[42][43][44][45] , etc. Expression of these cholangiocyte receptor genes in cholangiocytes and hepatocytes of WT and kras + adult zebrafish after Dox induction were analyzed by RT-qPCR. As shown in Supplementary Fig. S1, majority of these receptor genes showed little changes in expression in cholangiocytes and hepatocytes following oncogenic kras V12 expression in hepatocytes, but s1pr2 expression had striking increases in both hepatocytes (28 fold) and cholangiocytes (> 120 fold). This is reminiscence of several previous reports that s1pr2 is highly expressed in cholangiocarcinoma and bile duct cell diseases 43,46 . Thus, in view of high s1pr2 expression in cholangiocytes and its correlation with cholangiocarcinoma and cholestatic liver disease, we chose s1pr2 as a promising candidate receptor to study its role during HCC progression in our zebrafish model. To analyze the role of S1pr2 during hepatocarcinogenesis, both an agonist, taurocholate (TCA), and an antagonist, JTE-013, were used 42 . kras + and WT larvae were treated with TCA or JTE-013 in conjunction with Dox from 3 to 8 dpf. TCA led to a further increase in liver size compared to the group treated with Dox alone ( Fig. 2A,B). JTE-013 had an opposite effect, as the liver size became significantly smaller than that in the Dox alone group ( Fig. 2A,B). To confirm the link between S1pr2 and cholangiocytes, Alcam staining was used to determine cholangiocyte density in the same treatment groups. As shown in Fig. 2C and quantified in Fig. 2D, cholangiocyte density in kras + larvae was significantly increased in the TCA/Dox group compared to that in the control (Dox alone) group, while the JTE-013/Dox group showed a lower cholangiocyte density than the control group. In comparison, there was no significant change of cholangiocyte density in WT larvae by the two chemicals ( Fig. 2A,B). In addition, co-staining for Alcam and pERK (a known downstream marker for S1pr2 activation in intrahepatic cholangiocarcinoma (ICC) 44 , indicating elevation and inhibition in pERK signal (nucleus-localized) in cholangiocytes by TCA and JTE-013 respectively (Fig. 2E,F).
To further characterize the effects of S1pr2 on hepatocarcinogenesis, molecular markers for proliferation, apoptosis and fibrosis were examined by immunohistochemical staining after TCA and JTE-013 treatments. Cell proliferation, as determined by PCNA staining, were further increased by TCA and reduced, though not significantly, by JTE-013 in the kras + livers (Fig. 3A,B). Based on Caspase 3a staining, cell apoptosis was reduced by TCA and greatly enhanced by JTE-013 in the kras + livers (Fig. 3C,D). Consistently by Laminin and Collagen staining ( Fig. 3E-H), these fibrosis markers showed great increases by TCA and decreases by JTE-013 in the kras + livers. Thus, the agonist TCA enhanced liver cell proliferation and fibrosis and reduced apoptosis while the antagonist JTE-013 showed exactly the opposite effects, indicating the general positive role of S1pr2 in liver cell proliferation and fibrosis and a negative role in liver cell apoptosis.

Crosstalk between oncogenic hepatocytes and cholangiocytes.
It is well known that triglyceride accumulation in the liver, or non-alcoholic steatohepatitis (NASH), increases the risk for development of HCC and ICC 47 . We hypothesized that in the kras V12 transgenic model, oncogenic hepatocytes develop NASH by accumulating triglycerides, which are converted to bile acids by Cholesterol 7 alpha-hydroxylase (Cyp7a1), the main enzyme in the classical pathway of bile acid synthesis 48 . Bile acids may promote cholangiocyte proliferation through S1pr2 activation. To test the hypothesis, we examined triglyceride accumulation in the liver upon kras induction in hepatocytes. As shown in Supplementary Fig. S2A,B, Oil red O staining of liver triglycerides showed a significant increase in triglyceride accumulation in the kras + group than those in the WT group at all time points from 8 to 96 h following the induction of oncogenic kras V12 expression. This finding confirmed that kras V12 induced hepatocarcinogenesis is accompanied with hepatic steatosis. Consistent with this, cyp7a1 mRNA was 22-folds higher in kras + hepatocytes than WT hepatocytes ( Supplementary Fig. S2C).
To further test our hypothesis, a differential feeding experiment was carried out with 10% cholesterol (to induce NASH in the liver), 10% glucose, normal feed or starvation from 5 to 12 dpf (see Method). In both kras + and WT groups, the liver size was significantly larger in the cholesterol feeding group than those in the other three groups (Fig. 4A,B). Hepatic triglyceride accumulation was also higher in the cholesterol group than in the other three groups (Fig. 4C,D). Thus, kras + larvae may have a priority for either de novo synthesis or storage of fatty acids to support tumor growth and proliferation.
To confirm the link between hepatic triglyceride accumulation and S1pr2, s1pr2 mRNA was measured in different feeding groups. Indeed, s1pr2 mRNA was significantly higher in the cholesterol group than in the other groups (Fig. 4E), indicating that NASH induction may also activate s1pr2 expression. Furthermore, when total bile acid was determined, we found that its concentration was significantly higher in the cholesterol group than those in the normal feeding group in both kras + and WT larvae, confirming that the increase in hepatic triglycerides is associated with an increase in the total bile acid content (Fig. 4F). Finally, there was an increase in both cholangiocyte density (Fig. 4G,H) and PCNA + cells (Fig. 4I,J) in the cholesterol group than that in the normal feeding group in both kras + and WT larvae, thus providing an additional clue that fatty liver has an impact on cholangiocyte proliferation probably through bile acid activation of S1pr2. The activated cholangiocytes in turn influence liver cell proliferation and carcinogenesis. www.nature.com/scientificreports/ www.nature.com/scientificreports/ www.nature.com/scientificreports/ Molecular feedback mechanism of cholangiocytes to oncogenic hepatocytes. Cholangiocytes may influence hepatocyte carcinogenesis by secreting pro-inflammatory cytokines. To investigate this possibility, cholangiocytes and hepatocytes were isolated by FACS for RNA extraction. Selected cytokine genes were analyzed for their expression in cholangiocytes by RT-qPCR. We found that il17a/f1 (interleukin 17a/f1), a proinflammatory cytokine gene, was about four fold up-regulated in kras + cholangiocytes compared to that in WT cholangiocytes (Fig. 5A). In comparison, expression of other tested cytokine genes including tnfα (tumor necrosis factor alpha), nfap (nuclear factor activating protein) il5 (interleukin 5) and il12b (interleukin 12b) showed much less increase in kras + cholangiocytes. Furthermore, the induction of il17a/f1 in kras + cholangiocytes was much higher than that in kras + hepatocytes (Fig. 5B). Interestingly, the induction of Il17a/f1 receptor (il17ra1a) mRNA expression appeared to be high in both hepatocytes (3.9 fold) and cholangiocytes (6.5 fold) in kras + fish (Fig. 5C). Thus, Il17a/f1 secreted from cholangiocytes may exert its effect on hepatocytes upon oncogenic kras induction in hepatocytes. Finally, to link the changes with S1pr2, we analyzed the expression of il17a/f1 in zebrafish larvae after treating them with S1pr2 agonist (TAC) or antagonist (JTE-013) to activate or suppress cholangiocytes. We indeed noticed up-and down-regulation of il17a/f1 expression by TAC and JTE-013 respectively (Fig. 5D). This observation support that the tumorigenic effect of cholangiocytes on hepatocytes might be via il17a/f1. To further investigate the stimulating role of cholangiocytes upon oncogenic hepatocytes via the IL17a/f1 pathway, we analyzed the expression pattern of downstream markers of Il17a/f1 receptor by immunostaining of specific downstream marker pERK. As presented in Fig. 5E and quantified in Fig. 5F, pERK immunostaining signal was significantly expressed in the liver of kras + zebrafish larvae upon TCA activation of cholangiocytes. In contrast, inhibiting cholangiocytes via JTE-013, led to subsequent decrease in pERK than the control group. Determent of HCC progression by Il17a/f1 morpholino knockdown. IL17A family has been shown to be involved in several types of cancers [49][50][51] . To validate the effect of il17a/f1 on tumor development and infiltration of immune cells (neutrophils and macrophages) to the liver, il17a/f1 was specifically knocked down via two different morpholinos: one targeted at il17a/f1 translation start site (Trs_Mo) and the other at an il17a/f1 splicing site (Spl_Mo). A control morpholino (Ctr_Mo) targeting at human beta-globin gene was also used. As shown in Supplementary Fig. S3, both Trs_Mo and Spl_Mo reduced liver size significantly in the kras + larvae compared to uninjected or Ctr_Mo injected groups. Three independent microinjection experiments were carried out for these mopholinos and all experiments showed consistent results.
To validate the splicing blocking morpholino Spl_Mo, which targeted the first intron-exon junction, a pair of PCR primers flanking the first intron were used to monitor an anticipated 200-bp fragment of the splice blocked target. After gel electrophoresis of RT-PCR products from embryos injected or uninjected with morpolinos, a band of 200 bp size only appeared in the embryos injected with Spl_Mo morpholino but not in the control Ctr_MO and uninjected group (Fig. 6A,B).
To further validate the effect of knockdown of il17a/f1, two immune cell reporter transgenic lines were employed, lyz + for dsRed + neutrophils and mpeg + for mCherry + macrophages. spl-il17a/f1 morpholino was injected into kras + , kras + /lyz + , lyz + , kras + /mpeg + , mpeg + and WT embryos at one-cell stage and injected embryos were analyzed for liver size, dsRed + neutrophil counts and density, mCherry + macrophage count and density within the liver. As shown in Fig. 6C,E, by 6 dpf, there was an overall decrease of liver size in spl-MO injected larvae, compared to those in kras + larvae injected with Ctr_MO and uninjected group. Furthermore, we noticed that il17a/f1 knockdown led to decreases of both number and density of infiltrated neutrophils to the liver (Fig. 6F,G). Similarly, decreases of liver size, number and density of infiltrated macrophages to the liver by il17a/f1 knockdown were also observed (Fig. 6D,J-L). Collectively, our data have shown that cholangiocytes could accelerate HCC progression through Il17a/f1 cytokine which could be a potential target for cancer therapy.

Discussion
Liver is the largest internal organ, consisting of 70% hepatocytes and 15% cholangiocytes. Hence cholangiocytes represent the second largest population of cellular entity of the liver and they are important to maintain liver homeostasis after hepatocyte loss and inflammation [52][53][54] . In our study, we used an established transgenic zebrafish model to overexpress kras V12 oncogene in hepatocytes to initiate hepatocarcinogenesis and investigated the interaction between oncogenic hepatocytes and neighboured cholangiocytes. Our data showed that there was a rapid and consistent increase of cholangiocytes from 8 h post Dox treatment. Cholangiocyte number and density continued to increase over the 96 h of Dox treatment compared to hepatocytes in WT larvae. Our data also showed a firm correlation between cholangiocyte density and liver size increase after kras V12 activation in hepatocytes. Further pharmacological experiments were performed to confirm the bond between the two main types of liver cells, hepatocytes and cholangiocytes.
The pharmacological amenability of cholangiocytes to different drugs depend primarily on receptor of choice on cholangiocytes surface. Under different physiological conditions such as inflammation and cholestasis, cholangiocytes express specific receptors to promote their proliferation, such as S1pr2 44 , estrogenic receptors 8 , muscarinic receptor 55 and secretin receptor 9 . Among those receptors, S1pr2 showed the most dramatic increase in cholangiocytes after kras V12 induction in hepatocytes. Also, the same receptor has been reported to be activated in human cholangiocarcinoma cell lines [42][43][44] as well as after bile duct ligation in mouse models 56 . By using specific agonist (TCA) or antagonist (JTE-013) to S1pr2, we demonstrated that liver tumorigenesis became further enhanced or deterred respectively. Upon cholangiocyte activation, hepatocytes proliferation and fibrosis were found to be increased while apoptosis was decreased; these are signs of enhanced HCC progression. Consistent with the above observations, it has been reported that activated cholangiocytes can accelerate liver fibrosis by different mechanisms, firstly by secreting profibrotic factors such as connective tissue growth factor (CTGF). www.nature.com/scientificreports/ www.nature.com/scientificreports/ Secondly laminin synthesis can occur in cholangiocytes as in rat cholestatic models. Finally, cholangiocytes can induce fibrosis directly in hepatocytes by promoting epithelial mesenchymal transition (EMT) or indirectly by promoting other hepatic cells 8,9,11 . Hepatocytes and cholangiocytes interact reciprocally in the liver under different physiological conditions 57-59 . One of the earliest studies to dissect the interaction between hepatocytes and cholangiocytes was performed in a rat model of hepatocarcinogenesis induced by ethionin, Novikoff et al. showed that multiple intercellular junctions connect hepatocytes, cholangiocytes and bile canalicular structure. These junctions are responsible for the intercellular transport of specific soluble factors that can cause changes in the histology and ultrastructure of both cell types 39 . Here we attempted to go further to identify the molecular crosstalk between the two types of cells in our kras V12 zebrafish model. Upon kras V12 induction in adult hepatocytes for 7 days, a dramatic increase in the expression of cyp7a1 mRNA was observed in oncogenic hepatocytes compared to WT hepatocytes. Cyp7a1 is the main gene in the classical pathway of bile synthesis from cholesterol precursor. Recent reports indicate that the bile acids can directly activate s1pr2 60 . Consistent with this, we also found that the concentration of total bile acids was also increased in kras + larvae, especially when they were fed with high cholesterol diet (Fig. 4). The increase of total bile acids correlates with the increase of the size of oncogenic kras + livers. Thus, there should be a connection of the bile acids in activating cholangiocytes during liver tumorigenesis.
Our data also showed that il17a/f1 was up-regulated in cholangiocytes upon oncogenic activation of hepatocytes while its receptor, il17ra1a, was up-regulated in hepatocytes; thus, the effect of cholangiocytes on hepatocytes is likely via Il17a/f1 pathway. Consistent with this, staining of Il17a/f1 downstream marker pERK showed a positive correlation with cholangiocyte activation by its agonist TCA (Fig. 2E,F). To further validate the role of Il17a/f1 released from cholangiocytes on promoting liver tumorigenesis, we performed il17a/f1 morpholino knockdown experiments on kras + zebrafish larvae. Indeed, knockdown of il17a/f1 caused a deferment of liver tumorigenesis as judged by decreased liver size (Fig. 6). Overall, the crosstalk between hepatocytes and cholangiocytes was mostly based on the correlated expression of il17a/f1 cytokine gene and its receptor gene (il17ra1a) and a preliminary il17a/f1 experiments. In future, further works to specifically localize Il17a/f1 and Il17ra1a in cholangiocyte and hepatocytes and characterization of their function are needed. However, consistent with our current observations, previous reports also indicated elevated expression of IL17 in human HCC samples 51 and its role in enhancing HCC inflammatory environment [61][62][63][64][65] . Previously, we also confirmed that inflammatory immune cells play a prominent role during HCC progression 35 .
In summary, our data suggested that cholangiocytes play an important role in promoting HCC development through an inflammatory loop. While hepatocytes increase bile acids synthesis and lipogenesis to satisfy its demanding need for energy, cholangiocytes respond positively to hepatic bile acids and induce pro-inflammatory environment through Il17a/f1 secretion and other cytokines. This in turn accumulates more inducing signals for hepatic carcinogenesis. A proposed model for the interaction between oncogenic hepatocytes and cholangiocytes upon kras V12 induction in hepatocytes is presented in Fig. 7. In future, it will be interesting to investigate whether www.nature.com/scientificreports/ the stimulating role of Il17a/f1 is only specific to kras-induced cancer or universally to most other cancers. Understanding of this should provide valuable information for development of effective therapeutic approaches.
Chemical treatment. All  Induction of zebrafish NASH by cholesterol feeding. Supplement-enriched diets were prepared as previously described 29 . Briefly, cholesterol (Sigma) was dissolved in dimethyl sulfoxide (DMSO) to make a 10% solution, of which 400 μl was added to 0.5 g of standard zebrafish larval food (dried algae). The diet was left to dry overnight, grounded to powder form and provided to treated larvae on daily basis. The same was followed for 10% glucose enriched diet. Finally, starvation group was deprived from feeding throughout the experimental period.
Morpholino knockdown of il17a/f1. Two morpholino oligonucleotides targeting an RNA splice site (Spl-Mo, GTT CAC TTC AGC TAT ACT CAC CAT A) and the translation site (Trs-Mo, CGG AGG TTT AAC GCT GAT GACAT) of il17a/f1 were designed and synthesized by Gene Tools (Philomath, OR). A standard control morpholino (Ctr_Mo, 5′-CCT CTT ACC TCA GTT ACA ATT TAT A-3′) targeting a human beta-globin intron (Gene Tools, Philomath, OR) was also used as a negative control. Aliquots of morpholino (1 mM) and 1% (wt/ vol) phenol red in Danieau solution were injected into embryos at the 1-cell stage. Dox was added to all larvae from 3 to 6 dpf. To further validate the effect of splicing morpholino, RNA was isolated from 100 to 150 5-dpf larvae that were microinjected with Sp-Mo. A forward primer (ATG TCA TCA GCG TTA AAC CTCC) and reverse primer (ATG TAA GTC CAT GGA GAG ATGG) flanking the first intron were used to check for the presence of a 200-bp fragment from the first intron ( Supplementary Fig. S3A).
Determination of total bile acids. Concentration of total bile acids was determined in 10-dpf larvae by using a total bile acid quantification kit (Crystal Chem, 80470) according to manufacture protocol.
Photography and image analyses. Zebrafish larvae after each treatment were collected, anesthetized immediately in 0.08% Tricaine (E10521; Sigma) and immobilized by using 3% methylcellulose (M0521; Sigma) before proceeding to imaging. Zebrafish larvae were photographed individually with an Olympus microscope (Olympus, Tokyo, Japan).
Immunostaining and cytological analyses. Following the end of treatment, zebrafish Larvae were fixed in 4% PFA dissolved in PBS, embedded in bacterial-agar, and cryo-sectioned at 8-μm thickness using a cryotome. This was followed by immunohistochemical staining. Most of the primary antibodies used were derived from rabbits, including anti-proliferating cell nuclear antigen (PCNA) (FL-261, Santa Cruiz), anti-caspase 3 (Abcam), anti-collagen I (ab23730; Abcam) and anti-laminin (L9393, Sigma). Anti-Alcam (ANZN-8, ZIRC, USA) was derived from the mouse. Anti-rabbit or anti-mouse secondary antibodies were purchased from Thermo Fisher Scientific.
Isolation of hepatocytes, neutrophils, macrophages and cholangiocytes by fluorescence activated cell sorting. Fabp10+, lyz+, and mpeg+ transgenic zebrafish lines in wild-type and kras+ background were used for fluorescence-activated cell sorting (FACS) isolation of hepatocytes, neutrophils, and macrophages respectively. 7-10 adult livers were pooled and dissociated into single cells in the presence of 0.05% trypsin (T1426; Sigma) by using a 40-μm mesh (352340; Fisher Scientific, Pittsburgh, PA) as previously indicated 69 . Hepatocytes (fabp10+) were isolated based on dsRed expression, neutrophils (lyz+) were isolated based on DsRed expression, and macrophages (mpeg+) were isolated based on mCherry expression. For isolation of cholangiocytes, Alcam antibody was used as an accepted cell surface marker for cholangiocytes together with the Alexa Fluor secondary antibody (Thermo Scientific).