Abstract Hepatoid adenocarcinoma of the stomach (HAS) is a rare subtype of gastric cancer (GC) that histologically resembles hepatocellular carcinoma (HCC). Despite its low incidence, HAS had a poor 5-year survival rate. Currently, the linkages between clinicopathological and genomic features of HAS and its therapeutic targets remain largely unknown. Herein, we enrolled 90 HAS patients and 270 stage-matched non-HAS patients from our institution for comparing clinicopathological features. We found that HAS had worse overall survival and were more prone to develop liver metastasis than non-HAS in our cohort, which was validated via meta-analysis. By comparing whole-exome sequencing data of HAS (n=30), non-HAS (n=63), and HCC (n=355, The Cancer Genome Atlas), we identified a genomic landscape associated with unfavorable clinical features in HAS, which contained frequent somatic mutations and widespread copy number variations. Notably, signaling pathways regulating pluripotency of stem cells affected by frequent genomic alterations might contribute to liver metastasis and poor prognosis in HAS patients. Furthermore, HAS developed abundant multiclonal architecture associated with liver metastasis. Encouragingly, target analysis suggested that HAS patients might potentially benefit from anti-ERBB2 or anti-PD-1 therapy. Taken together, this study systematically demonstrated a high risk of liver metastasis and poor prognosis in HAS, provided a clinicogenomic landscape underlying these unfavorable clinical features, and identified potential therapeutic targets, laying the foundations for developing precise diagnosis and therapy in this rare but lethal disease. Keywords: hepatoid adenocarcinoma of the stomach, liver metastasis, prognosis, whole-exome sequencing, clinicogenomic landscape 1. Introduction Hepatoid adenocarcinoma of the stomach (HAS) is a rare subtype of gastric cancer (GC) that histologically resembles hepatocellular carcinoma (HCC), accounting for 0.38-1.6% of all GC [55]^1. HAS was recognized as a highly malignant carcinoma, featuring rapid progression, high liver metastasis propensity, and poor prognosis [56]^2^, [57]^3. Limited by a scarcity of large-scale clinical research and a critical knowledge gap regarding genomic features in HAS, little consensus on its standardized diagnostic and therapeutic strategies has been achieved. Currently, under standard management for conventional gastric cancer (CGC), the 5-year survival rate of advanced-stage HAS patients was only 9% [58]^2. Since firstly reported in 1985 [59]^4, several HAS cases have been reported worldwide, primarily in China [60]^5 and Japan [61]^6. Due to extremely similar clinicopathological features between metastatic HAS and HCC, such as elevated serum alpha-fetoprotein (AFP) level and metastatic liver lesion mimicked HCC-like morphologically, metastatic HAS patients were easily misdiagnosed as HCC in clinical practice [62]^7^, [63]^8. Controversially, most case reports revealed that HAS patients had an inferior prognosis than non-HAS patients [64]^5^, [65]^9^, [66]^10, whereas a few studies did not observe a significant difference of prognosis between HAS and non-HAS patients[67]^11^, [68]^12. So far, most clinical studies focusing on HAS were limited to case reports, case series, or small-sample studies. Therefore, a large-scale study with systematic analysis is urgently required to investigate and validate the clinicopathological characteristics and prognosis of HAS. Characterization of molecular landscape of HAS is a crucial step to deepen the understanding of its clinicopathological features and develop precise therapeutic strategies. Nevertheless, current evidence regarding the molecular characteristics of HAS was mainly established on a limited number of studies. A previous study identified a list of recurrently mutated genes and copy number gains in HAS using targeted sequencing on a panel of 483 cancer-related genes [69]^13. A recent multi-omics study demonstrated the molecular features associated with hepatoid differentiation in a non-metastatic HAS population [70]^14. Despite recent advances in molecular characterization, the molecular mechanism underlying tumor metastasis and unfavorable prognosis in HAS remains unclear. Herein, we constructed a large clinical HAS cohort and performed comparative analysis and meta-analysis to uncover the distinct clinicopathological characteristics and prognosis of HAS compared with non-HAS. Then, using whole-exome sequencing (WES) analysis, we compared multidimensional genomic features between HAS, non-HAS, and HCC. Importantly, we performed potential targets screening in HAS. These data provide a comprehensive clinicogenomic landscape of HAS, which deepen the understanding of how its unfavorable clinical features are linked to the genomic profiles. In addition, we also identified potential therapeutic targets in this rare but lethal disease. 2. Methods 2.1 Patient cohort and data collection This study was approved by the Institutional Review Board of the First Affiliated Hospital of Zhejiang University School of Medicine and conducted in compliance with the guidelines of the Declaration of Helsinki. A total of 12622 GC patients were screened from patients who received biopsies/surgery and underwent standardized pathological diagnoses in the First Affiliated Hospital, Zhejiang University School of Medicine from January 2009 to June 2020. Based on WHO classification of tumors of the digestive system (2019)[71]^15, HAS was morphologically defined as a tumor composed of large polygonal eosinophilic neoplastic cells (hepatoid differentiation), regardless of the percentage of hepatocyte-like regions or serum AFP level. Two certificated pathologists (M.K. and C.Z.L.) independently confirmed the pathological diagnoses. Any discrepancies were then resolved by consulting another experienced pathologist (X.D.T.). This study enrolled 90 HAS patients and 973 CGC patients with annotated clinicopathological and follow-up data. Subsequently, 270 non-HAS patients were screened from CGC via a manner of 3:1 stage-matched with HAS for comparative analysis. All patients provided written informed consent. We obtained clinicopathological information from patient medical records in our institutional database. The clinical information included gender, tumor size, tumor location, serum levels of AFP, carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA199), first-metastasis site, computed tomography (CT) images, and therapy. Pathological features such as TNM stage (American Joint Committee on Cancer (AJCC), 8th edition), vascular invasion, hematoxylin & eosin (H&E)-stained micrographs, and ERBB2 immunohistochemistry (IHC) results were also collected. The baseline clinicopathological characteristics were displayed in Table [72]1 and [73]Table S1. Patients were followed up via telephone, letters, and medical records. Overall survival (OS) time was defined as the interval from the date of diagnosis to the date of death or the last follow-up point. During follow-up, the site and time of the first metastasis were recorded. Synchronous metastasis was defined as distant metastasis at diagnosis or within six months during follow-up, whereas metachronous metastasis occurred after six months during follow-up [74]^16. We retrospectively collected fresh-frozen tumor tissues with paired tumor adjacent normal tissues from 30 HAS cases and 63 non-HAS cases for WES. In addition, WES data with annotated clinicopathological information of 355 TCGA patients diagnosed as HCC (TCGA-LIHC) were downloaded from the cBioportal database ([75]https://www.cbioportal.org/). Table 1. Clinicopathological characteristics of HAS and stage-matched non-HAS patients Characteristic HAS (N = 90) Stage-matched non-HAS (N = 270) P Age -years 0.420 ≤ 60 33(36.7) 112(41.5) > 60 57(63.3) 158(58.5) Gender 0.265 Female 24(26.7) 89(33.0) Male 66(73.3) 181(67.0) T stage 0.114 T1/2 22(24.4) 47(17.4) T3/4 48(53.3) 166(61.5) Unknown 20(22.2) 57(21.1) N stage 0.362 N0 19(21.1) 40(14.8) N1 12(13.3) 42(15.6) N2 19(21.1) 52(19.3) N3 20(22.2) 79(29.3) Unknown 20(22.2) 57(21.1) M stage > 0.999 M0 59(65.6) 177(65.6) M1 31(34.4) 93(34.4) AJCC stage > 0.999 I 9(10.0) 27(10.0) II 16(17.8) 48(17.8) III 34(37.8) 102(37.8) IV 31(34.4) 93(34.4) Size -cm 0.251 ≤ 5.0 47(52.2) 161(59.6) > 5.0 23(25.6) 56(20.7) Unknown 20(22.2) 53(19.6) Location 0.088 Antrum 43(47.8) 146(54.1) Body 22(24.4) 77(28.5) Cardia 25(27.8) 46(17.0) Unknown 0(0) 1(0.4) Vascular invasion < 0.001 Positive 44(48.9) 81(30.0) Negative 23(25.6) 124(45.9) Unknown 23(25.6) 65(24.1) Serum AFP level -ng/mL < 0.001 ≤ 20.0 21(23.3) 255(94.4) > 20.0 69(76.7) 13(4.8) Unknown 0(0) 2(0.7) Serum CEA level -ng/mL 0.009 ≤ 5.0 47(52.2) 191(70.7) > 5.0 37(41.1) 77(28.5) Unknown 6(6.7) 2(0.7) Serum CA199 level -U/mL 0.341 ≤ 37.0 69(76.7) 207(76.7) > 37.0 15(16.7) 61(22.6) Unknown 6(6.7) 2(0.7) ERBB2 IHC status 0.003 -/+ 44(48.9) 158(58.5) ++/+++ 34(37.8) 54(20.0) Unknown 12(13.3) 58(21.5) Surgery type 0.062^d Distal gastrectomy 35(38.9) 143(53.0) Proximal gastrectomy 2(2.2) 3(1.1) Total gastrectomy 34(37.8) 75(27.8) No surgery 19(21.1) 49(18.1) Chemotherapy^a 0.256 Received 67(74.4) 191(70.7) Not received 20(22.2) 79(29.3) Unknown 3(3.3) 0(0) Target therapy^b 0.228^d Received 3(3.3) 8(3.0) Not received 84(93.3) 262(97.0) Unknown 3(3.3) 0(0) Immunotherapy^c < 0.001^d Received 6(6.7) 1(0.4) Not received 81(90.0) 269(99.6) Unknown 3(3.3) 0(0) [76]Open in a new tab ^a Chemotherapy regimen was mainly based on the 5-fluorouracil (5-FU) plus platinum combination, including fluorouracil, leucovorin plus oxaliplatin (FOLFOX), oxaliplatin plus S-1 (SOX), and capecitabine plus oxaliplatin (XELOX). Other regimens included S-1, docetaxel, and paclitaxel plus S-1 (SPA). ^b Target therapy regimen included trastuzumab, bevacizumab, and apatinib. ^c Immunotherapy regimen included sintilimab and nivolumab. ^d Statistical analysis was conducted using Fisher's exact test, and other categorical data were using the chi-square test. The cases with unknown data were not included in the statistical analysis. 2.2 Meta-analysis We conducted a systematic literature search in PubMed, Web of Science, Embase, Scopus, Cochrane Library, and CNKI database up to January 2021, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The search terms were as follows: (“adenocarcinoma” OR “adenocarcinomas” OR (“malignant” AND “adenoma”) OR “malignant adenoma”) AND “hepatoid” AND (“Stomach” OR “Stomachs” OR “Gastric”). We also searched articles in the references if they were