A novel zebrafish intestinal tumor model reveals a role for cyp7a1-dependent tumor-liver crosstalk in tumor's adverse effects on host

The nature of host organs and genes that underlie tumor-induced physiological disruption on host remains ill-defined. Here, we establish a novel zebrafish intestinal tumor model that is optimized for addressing this issue, and find that hepatic cyp7a1, the rate-limiting factor for synthesizing bile acids (BAs), is such a host gene. Inducing krasG12D by Gal4 specifically expressed in the posterior intestine resulted in formation of an intestinal tumor classified as dysplasia. The local intestinal tumor caused systemic detrimental effects on host including liver inflammation, hepatomegaly, growth defects, and organismal death. Whole-organismal level gene expression analysis and metabolite measurements revealed that the intestinal tumor reduced total BAs levels via down-regulation of hepatic cyp7a1. Genetically rescuing cyp7a1 expression in the liver restored the BAs synthesis and ameliorated tumor-induced liver inflammation, but not other tumor-dependent phenotypes. Thus, we found a previously unknown role of cyp7a1 as the host gene that links the intestinal tumor, hepatic cholesterol-BAs metabolism, and liver inflammation in tumor-bearing fish. Our model provides an important basis to discover host genes responsible for tumor-induced phenotypes and to uncover mechanisms underlying how tumors adversely affect host organisms.


Introduction
Furthermore, as is the case for zebrafish, most animal tumor models develop tumors at an 98 adult stage, thereby preventing us from investigation into how tumors affect growing, juvenile 99 vertebrates. For these reasons, a novel zebrafish tumor model is required. 100 In the current study, we successfully generated a novel intestinal tumor model. 101 Careful characterization of this model led to the identification of four tumor-induced 102 phenotypes including systemic inflammation, hepatomegaly, growth defects, and organismal 103 death, which are seen even in human cancer patients. Anomalies in gene expression and 104 metabolism were found in both the intestinal tumor and the distant liver upon whole-105 organismal level transcriptome analysis. On the basis of these, we found that a tumor-liver 106 crosstalk, which can be defined by reduced expression of hepatic cyp7a1 accompanied with 107 altered cholesterol-bile acids flux, promote infiltration of neutrophils to the liver (liver 108 inflammation) in tumor-bearing fish. 109 110 Results 111

pInt-Gal4-driven kras G12D expression causes outgrowth of posterior intestine, leading to 112 formation of the intestinal tumor 113
In order to generate a zebrafish model of tumorigenesis with early onset, we sought for Gal4 114 line(s) capable of driving gene expression to a single organ (ie. organ specificity) at an early 115 stage of zebrafish development. To this end, we crossed a set of Gal4 lines (Asakawa and 116 Kawakami, 2008; Asakawa et al., 2008) with a line Tg(5×UAS:EGFP-P2A-kras G12D ) 117 generated in this study with the Tol2 system ( Fig. 1A and Table S1) (Kawakami, 2004;118 Kawakami et al., 1998). Tg(5×UAS:EGFP-P2A-kras G12D ) harbored a mutated kras, kras G12D , 119 one of the most prevalent driver oncogenes in human malignant tumors ( Fig. 1A and Table  120 S1) (Schubbert et al., 2007). Expression of kras G12D  (pInt-Gal4), anterior intestinal cells (aInt-Gal4), brain (Brain-Gal4), and liver (Liver-Gal4) 126 ( Fig. 1). From these, pInt-Gal4 was chosen for further characterization due to its ability to 127 cause efficient outgrowth of posterior intestinal cells upon kras G12D expression ( Fig. 1B-1E). 128 aInt-Gal4 was also able to cause outgrowth of anterior intestinal cells (Figs. 1F-1I). However, 129 outgrowth of intestinal cells by aInt-Gal4 was less dramatic when compared to that by pInt-130 Gal4. Moreover, expression of aInt-Gal4, despite specific after 5 dpf, was somewhat non-131 specific during 2-4 dpf, leading to abnormal growth of epidermal cells in a temporal manner 132 ( Fig. S1A-1D). 133 pInt-Gal4 expression judged by EGFP expression was detectable from 4 dpf (days 134 post-fertilization) ~ 5 dpf ( Fig. 2A-2B). Outgrowth of posterior intestinal cells by pInt-  driven kras G12D expression was evident at 5 dpf ( Fig. 2A-2B). Oncogene expression was 136 confirmed by qPCR ( Fig. 2C and Table S1). Moreover, 100% of fish harboring both pInt-137 Gal4 and 5×UAS:EGFP-P2A-kras G12D exhibited the outgrowth phenotype at 5 dpf (Fig. S2A). 138 Thus, at this stage, we were able to phenotypically discriminate tumor-bearing fish. The 139 number of intestinal cells determined by DAPI-staining in kras G12D -expressing fish was 140 significantly increased compared to that in the controls expressing EGFP under the regulation 141 by pInt-Gal4 (Fig. 2D-2J). In the previous study, Wallace et al. show that the mitotic rate of 142 intestinal epithelial cells is high (~ 40%) through 3 dpf, dropping at 4 ~ 5 dpf (< 5 %) 143 (Wallace et al., 2005). Despite the assumption that the majority of intestinal cells are post-144 mitotic at 5 dpf, we counted the number of mitotic cells by pH3 (phosphorylated histone H3)-145 staining ( Fig. 2K-2S) and BrdU-incorporation experiments at this time point (Fig. S2B-S2J). 146 The number of pH3-positive mitotic cells (Fig. 2K-2S) and BrdU-incorporated cells (Fig.  147 S2B-S2J) were consistently higher in kras G12D -expressing fish than in the sibling controls, 148 strongly suggesting that pInt-Gal4-driven kras G12D  supporting these findings ( Fig. 3I-3L). Based on these atypia phenotypes, it was likely that 156 pInt-Gal4-driven kras G12D expression in the posterior intestine led to dysplasia, a type of 157 tumor. Despite the disorganized structure of posterior intestine, the intestinal lumen was not 158 completely disrupted (Fig. 3I-3L). Consistent with this, food was present in the intestinal 159 lumen of tumor-bearing fish following feeding ( Fig. S3A-S3B). 160 We did not observe visible invasion and dissemination of EGFP positive cells in 161 our experimental window (Fig. 3A-3L Microscopic analyses showed considerable increase for the number neutrophils at 187 the whole-organismal level in tumor-bearing fish at 7 dpf ( Fig. 4A-4H). Immunostaining with 188 anti-Lyz antibody revealed that neutrophils were accumulated in the intestinal tumor when 189 compared to the normal intestine ( Fig. 4I-4O). During the cause of the experiments, we noted 190 that neutrophils had also infiltrated the liver (Fig. 4P-4Q). In order to better visualize tumor-191 induced liver inflammation, mCherry was expressed specifically in the liver using the liver 192 specific fabp10a promoter (Tg(fabp10a:mCherry)) ( Fig. 4P-4Q) (Her et al., 2003). We 193 counted the number of EGFP-positive neutrophils in the liver expressing mCherry. As a result, 194 we found that the number of neutrophils in the intestines of tumor-bearing fish was greater 195 than that in the sibling controls (12 ± 2.3 vs 30 ± 6.0, p = 0.0062: Fig. 4P were significantly smaller than the sibling controls (Figs. 5A and S4), the difference 209 observable from 7 dpf. The results varied among clutches at 7 dpf, whereas the growth defect 210 phenotype was very consistent at 9 dpf ( Fig. S4A-S4B). The growth defect phenotype was 211 identified in the complete absence of foods (i.e. exogenous nutrient): although zebrafish 212 larvae are able to eat from 5-6 dpf, yolk-derived nutrient inherited from the mother keep fish 213 alive without visible abnormalities at least until 9 dpf. This enabled us to ignore experimental 214 variations on zebrafish behaviors related to eating and on nutrient absorption rate in the 215 intestine in explaining the growth defect phenotype. Based on these analyses, we concluded 216 that the local intestinal tumor caused a systemic growth defect. 217 It is well-known that tumor-bearing animals waste muscle and fat, resulting in a whether the growth defect phenotype could be attributed to cachexia, Oil Red O staining for 222 neutral TGs and lipids was performed. Stronger staining was detected for the liver and brain 223 at 9 dpf, a pattern of which was not prominently different between tumor-bearing fish and the 224 sibling controls (Fig. 5B-5E). This suggested that the intestinal tumor at this stage did not 225 have a strong impact on the systemic lipid level. In addition, HE staining did not find obvious 226 loss of host tissues such as muscles at 9 dpf ( Fig. 5F-5G Next we asked if the intestinal tumor worsens mortality of zebrafish. We counted 232 the number of dead and live fish every day and found that the survival rate of tumor-bearing 233 fish (less than 50% at 14 dpf) was significantly lower than that of the sibling controls 234 (approximately 80%; Fig. 5I). This phenotype was not due to a complete defect in swimming 235 ability and/or a complete loss of appetites in tumor-bearing fish, because tumor-bearing fish performed. Zebrafish at 7 dpf were roughly dissected into the three parts, the liver, the 251 intestinal tumor or normal intestine, and the rest part of body ( Fig. 6A and Table S2-S6). We 252 were particularly focused on the liver since the liver was preferentially inflamed by the 253 intestinal tumor (Fig. 4), despite a lack of visible metastasis to the liver in our experimental 254 setting. A set of genes potentially affected by the intestinal tumor (Table S2- tumor-induced phenotypes, liver inflammation, hepatomegaly, and the growth defect, were 305 rescued by the fabp10a:mCherry-P2A-cyp7a1 transgene (Fig. 8). We found that cyp7a1 306 overexpression did not significantly rescue the growth defect phenotype (Fig. 8A). As was the 307 case for Fig. 5A, the results to some extent varied depending on clutches: in one clutch, we 308 observed a trend for the rescue while not in a different clutch. Upon pooling data from 309 multiple clutches, we concluded that cyp7a1 overexpression did not consistently and 310 significantly rescue the growth defect phenotype. Moreover, tumor-induced hepatomegaly 311 (0.028 ± 0.0013 mm 2 (control) vs 0.038 ± 0.0016 mm 2 (tumor), p = 0.00016: Fig This study has two major advances. First, we established the novel zebrafish intestinal tumor 338 model, which is optimized for studying body-wide tumor-organ interaction in vivo. Second, 339 using the model, we discovered a tumor-liver interaction that mediates enhanced recruitment 340 of neutrophils to the liver in tumor-bearing fish, via down-regulation of a cholesterol-341 metabolizing gene cyp7a1 as a critical host gene. 342 343

Establishment of a novel intestinal tumor model in zebrafish 344
The zebrafish intestinal tumor model we have newly established harbors several strengths for 345 studying tumor-organ interaction at the whole-organismal level (Fig. 7). The combination of 346 pInt-Gal4 and UAS-controlled kras G12D induces epithelial tumor formation in the posterior 347 intestine at as early as 5 dpf, when zebrafish are small and completely transparent (Figs. 1-3). 348 Yet, zebrafish larvae after 5 dpf are able to swim and eat and therefore it is likely that  Thomas et al., 2008), was decreased at as early as 5-7 dpf in tumor-421 bearing fish (Figs. 6C and 6D). This reduction was concordant with the reduced total BAs 422 levels (Fig. 6E), which was not due to the decreased body size, as we did not find any 423 correlation between body length and the amount of bile acids in each individual (Fig. S6). 424 Indeed, rescuing cyp7a1 expression in the liver by means of the fabp10a promoter 425 significantly restored total BAs levels in tumor-bearing fish (Fig. 7). Intriguingly, this rescue 426 was associated specifically with buffered liver inflammation (Fig. 8): the number of 427 neutrophils in the liver was increased in the presence of the intestinal tumor (Fig 4P-R), 428 which was significantly ameliorated by overexpression of cyp7a1 in the liver (Fig. 8E-8G).  Table S1. 500

RNA isolation, cDNA synthesis, and quantitative PCR (qPCR) 502
For gene expression experiments, we often pooled multiple fish in a single tube. This was for 503 obtaining sufficient amount of high-quality RNAs especially when dissection was performed, 504 and for lowering the risk to select outliers from the clutch. Given that single female generally 505 produces more than 50 embryos, selecting e.g. 3 ~ 5 fish from a clutch may give rise to 506 unwanted bias in sample collection. Pooling multiple fish, 3 ~ 10, depending on the size of 507 clutches, in a single tube, and treat it as one biological replicate, is useful to reduce these risks. 508 Total RNA was isolated using TRIzol (Thermo Fisher SCIENTIFIC) or RNeasy Mini Kit 509 (QIAGEN). cDNA was synthesized using SuperScript III First-Strand Synthesis System 510 (Thermo Fisher SCIENTIFIC) or Transcriptor First Strand cDNA Synthesis Kit (Roche). The 511 obtained cDNAs were 5-or 10-fold-diluted and subjected into qPCR experiments by using 512 LightCycler480 Instrument II system and SYBR Green Master Mix (Roche). The obtained 513 data were analyzed using the delta-Ct method. The primers used are listed in Table S1. Tg(UAS:EGFP) +/Tg larvae from the same clutch were used. At 5 dpf, larvae ware fixed in 4% 559 PFA in PBS at 4°C for overnight. Larvae were washed with PBS for five times and treated 560 with 3% hydrogen peroxide in 0.5% sodium hydride at room temperature to bleach pigments. 561 After removing pigments, larvae ware washed with PBS and then transferred into methanol, 562 and stored at −30°C until staining. Larvae were washed with 0.5% PBT for 5 times.  Table S2. To identify differentially expressed genes (DEGs), 619 we first focused on the well-annotated protein-coding genes. RPKM scores were used to 620 calculate the ratio tumor/control. In this calculation, 1 was added to all RPKM scores to 621 ignore the scores below "1", and to make analyses more stringent. Recognizing that our 622 dissection cannot prevent cross-contamination, genes showing more than 0.8-fold-enrichment 623 and > 0 RPKM in the tissue of interest were further considered. The obtained ratios were used 624 to sort genes to find potential DEGs. As an initial screening to identify reliable DEGs, we 625 focused on a set of genes showing more than 3-fold changes in the RNA-seq experiments. 626 Considering possible differences among clutches, the RNA-seq experiment was followed by 627 qPCR validation with samples prepared from different clutches. Thus, the RNA-seq 628 experiment functioned as a screening to identify DEGs. Data visualization was done mostly 629 using ggplot2 (http://ggplot2.org/). In main figures, we show genes consistently validated by 630 qPCR. In our experience with our dataset, the validation rate was high for genes with more 631 than 3-fold changes in the intestine-derived samples. In the liver and rest part of the body, "3-632 fold criteria" was not enough to obtain a high validation rate (i.e. genes showing more than 3-633 fold changes such as pklr failed to be validated by qPCR (data not shown)). dpf. The data harbors 12 biological replicates. Error bars represent ± s.e.m. Statistical 935 significance was tested using student's t-test (unpaired, one-tailed). 936 (S) Liver size of the sibling controls and tumor-bearing fish at 7 dpf. Liver size was measured 937 from Tg(fabp10a:mCherry) images using ImageJ software. The data harbors 12 biological 938 replicates. Error bars represent ± s.e.m. Statistical significance was tested using student's t-test 939 (unpaired, one-tailed). number of fish used is 163 (7 dpf control fish), 155 (7 dpf tumor-bearing fish), 154 (9 dpf 945 control fish) and 154 (9 dpf tumor-bearing fish). Data from three independent clutches are 946 pooled. Data from each clutch are shown in Fig. S4. Error bars represent ± s.e.m. Statistical 947 significance was tested using student's t-test (unpaired, two-tailed). 948 data harbors 3 biological replicates, each containing 7 fish for 5 dpf and 5 fish for 7 dpf, 968 respectively. Error bars represent ± s.e.m. Statistical significance was tested using student's t-969

Figure S6 Correlation between body length and total BAs levels 1063
Correlation between body length and total bile acids levels at 9 dpf (n = 20 per a group). 1064 1065   Fold expression (Control = 1) RNA-seq for pooled samples as a screening (Table S2-