Successful Identification and Treatment of Cancer of Unknown Primary Originating From Gastric Cancer Using Comprehensive Genomic Profiling and Immune Checkpoint Inhibitor Therapy: A Case Report.

1
Introduction
Cancer of unknown primary (CUP) is defined as a histologically confirmed malignancy without an established primary site after pathological evaluation and radiographic examinations [1]. It is a complex and challenging condition where one or more metastatic tumors are present, but the location of the original tumor is unknown. This may be due to the primary cancer being undetectable by current imaging methods or having regressed by the time metastatic tumors are discovered. The diagnostic process begins with a thorough clinical evaluation, which includes a detailed patient history and physical examination. This is followed by imaging studies to locate and characterize metastases. Common imaging modalities include computed tomography (CT), magnetic resonance imaging, and positron emission tomography‐CT. Despite these techniques, the primary cancer is often not visualized, which complicates the diagnosis and treatment planning. CUP accounted for < 5% of all cancers. With the advent of gene expression profiling and comprehensive genomic profiling (CGP), 1%–2% of new cancer diagnoses are designated as CUP [2]. Gene expression profiling and next‐generation sequencing (NGS) may identify a potential site of origin in patients with CUP. Specific gene expression profiles have been well recognized in most cancers according to their site of origin, which reflects the different expression profiles present in their normal tissues of origin [3, 4]. The National Cancer Institute recommends routine testing to investigate gene expression levels and epigenetic factors [5]. Over the past decade, CGP has been developed to predict the tumor site of origins in patients with CUP.
Relative to patients with known primary tumors, the prognosis for patients with CUP is unfavorable, with a median overall survival rate of < 12 months [6]. The initial presentation of CUP typically involves symptoms related to metastatic spread, rather than the primary tumor. Common sites for metastases include the lymph nodes, liver, lungs, peritoneum, and bones. The symptoms experienced depend on the location and extent of metastasis but may include weight loss, pain, and general fatigue. Because the primary tumor is not initially detected, treatment is often empirical and based on the manifestations of metastatic disease. Most CUP cases are adenocarcinomas, which are often poorly differentiated or undifferentiated tumors. There are less frequent instances of squamous cell carcinomas and neuroendocrine tumors. Trials comparing empirical chemotherapy with molecularly tailored therapies have not demonstrated significant improvements in patient outcomes. Additionally, it is difficult to select an appropriate systemic therapy to match a specific type of cancer. Therefore, a molecular diagnosis that identifies the origin is expected to inform treatment selection and improve the prognosis. However, mutation profiling alone is not sufficient to guide the personalized treatment, given that targeted therapies against a particular driver mutation can act differently in different tumor types. This is mainly due to insufficient evidence supporting the improvement in the prognosis of CUP managed by the approach. With the advent of NGS technology, massively parallel sequencing of tumor DNA offers great opportunities to identify actionable variations and enables targeted therapy in oncologic practice. Effective treatment strategies for CUP may benefit from a combination of molecular profiling and traditional clinical approaches to more accurately tailor therapies and improve patient outcomes [7].
We herein report the case of a patient with CUP in which a precise diagnosis of gastric cancer origin enabled personalized treatment with targeted therapy, based on a comprehensive genomic analysis. The patient received immune checkpoint inhibitor (ICI) monotherapy and has achieved a complete response.

2
Case
A 72‐year‐old woman was admitted to Asahikawa Medical University Hospital with epigastric pain and underwent esophagogastroduodenoscopy (EGD) in June 2020. A duodenal tumor and gastric ulceration were discovered, and histological evaluation of the biopsy specimen indicated poorly differentiated adenocarcinoma and benign inflammation, respectively. She had a medical history of total hysterectomy for uterine myoma in her 40s and a family history of hepatocellular carcinoma in her brother. The Eastern Cooperative Oncology Group (ECOG) performance status was 0. Immunohistochemistry (IHC) was positive for CAM5.2, cytokeratin 7 (CK7), caudal type homeobox 2 (CDX2), and carcinoembryonic antigen (CEA), and negative for thyroid transcription factor‐1 (TTF‐1) and cytokeratin 20 (CK20), indicating pancreaticobiliary malignancy. CT showed an extra‐duodenal mass growing to the common bile duct and the gall bladder without distal metastasis to other organs (Figure 1a). Colonoscopy showed no colorectal tumors except for a small (4 mm) adenoma in the rectum. Laboratory investigations showed anemia (white blood cells, 5.56 × 103/mm3; red blood cells, 2.68 × 103/mm3, platelets 30.8 × 103/mm3), slightly elevated lactose dehydrogenase (344 U/L; normal level 105–210 U/L), and tumor marker levels of 71 U/mL of Carbohydrate antigen 19‐9 (CA19‐9, normal level ≦ 37 U/mL), 2.6 ng/mL of CEA (normal level ≦ 5.0 ng/mL), and 36.4 ng/mL of cytokeratin 19 fragment (CYFRA, normal level ≦ 3.5 ng/mL). Endoscopic ultrasound fine‐needle biopsy of the tumor located in the pancreaticobiliary groove revealed poorly differentiated adenocarcinoma. Re‐examination with EGD revealed a laterally spreading gastric malignancy at the gastric angle, and a histological examination revealed well‐differentiated adenocarcinoma. The final diagnosis was dual cancer of unknown origin of pancreaticobiliary cancer and early gastric cancer. Gemcitabine and tegafur/gimeracil/oteracil (TS‐1) were chosen as neoadjuvant chemotherapy regimens, followed by pancreatoduodenectomy with distal gastrectomy (Figure 1b). Histological examination of the surgically resected specimens revealed poorly differentiated adenocarcinoma [T4a, N1, M1, pStage IVB], possibly of gallbladder origin, consistent with CUP, and well‐differentiated adenocarcinoma [T1b (SM2), N0, M0, pStage IA] of the stomach (Figure 2). The origin of CUP was suspected from biliary tract and adjuvant chemotherapy with TS‐1 was initiated; however, lymph node metastases around superior mesenteric artery (SMA) progressed 5 months later (Figure 3a). Gemcitabine with TS‐1 therapy was re‐attempted as the next regimen and a comprehensive cancer genomic panel (FoundationOne CDx, Foundation medicine, MA) was performed. Using formalin‐fixed paraffin‐embedded samples from CUP, many actionable genetic changes were found, and the microsatellite status was found to be high with a high tumor mutational burden (26.48 mutations/Mb). The analysis identified many actionable changes; ARID1A P225fs*175 (28.1%), ERRFL1 P339fs*111 (29.6%), KRAS G12D (46.2%), APC S1465fs*3 (50.2%), ATR I774fs*3 (24.7%), CDKN2A/B S7fs*19 (28.7%), CREBBP P1946fs30 (32.4%), FAM46C A232T (31.9%), FLCN H429fs*39 (32.1%), MLH1 A681T (2.2%), MLL2 P647fs*283 (53.4%) P648fs* (21.2%), MSH3 K383fs*32 (35.2%), QKI K134fs*14 (63.9%), RICTOR N1065S (51.2%), TP53 R273C (64.2%), and WHSC1 E1344fs*91 (31.6%). Furthermore, microsatellite instability (MSI) was confirmed using an MSI assay kit (FALCO Biosystems, Kyoto, Japan), and peak shifts of the five markers were found by capillary electrophoresis (Figure 4). These genetic findings suggested that CUP was a microsatellite unstable solid tumor; therefore, ICI was chosen as the next treatment for the recurrent lymph node metastasis. Pembrolizumab monotherapy (200 mg/body every 3 weeks) was administered. Lymph node metastasis shrank, and complete remission was maintained with no adverse event. Tumor marker levels of CA19‐9, CEA, and CYFRA were within normal range, EGD and colonoscopy presented no remnant cancers in the gastrointestinal tract. The treatment was stopped after 31 cycles (Figure 3b) and no recurrence was observed at 1 year after the cessation of treatment (Figure 5).
As a result of the cancer genomic panel, FLCN H429fs*39 and MLH1 A681T were suspected to be germline pathological variants (PGPV), and germline sequences of FLCN chr17: 17216395 and MLH1 chr3: 37048955 were assessed by NGS (Kazusa DNA Research Institute, Chiba, Japan). No germline variants were detected in this study. An MLH1 variant is associated with Lynch syndrome due to the MSI‐high status; eventually, the result excluded the possibility of Lynch syndrome. Her family history did not meet the Lynch‐Amsterdam criteria either. IHC of mismatch repair (MMR) protein, M1 (Roche Diagnostics, Basel, Switzerland), G210‐1129 (Roche Diagnostics), PU29 (Leica Microsystems), and EP52 (Agilent Technology, Santa Clara, CA) were used for MutL homolog 1 (MLH1), MutS homolog 2 (MSH2), MSH6, and postmeiotic segregation increased 2 (PMS2) staining, respectively. The expression of MLH1, MSH2, and PMS2 was lost in the carcinoma cells of CUP (Figure 6).
The CGP of the CUP tissue identified several pathogenic variants, including TP53, KRAS, and APC, which are frequently observed in gastrointestinal cancers. To further investigate the tumor origin, both the early gastric cancer and CUP specimens underwent IHC and target gene sequencing [8]. The well‐differentiated tubular adenocarcinoma showed a partial loss of MLH1, MSH2, and PMS2. Microdissection from each MLH1‐negative (area B) and MLH1‐positive area (area C) was conducted, and the samples were subjected to genetic testing (Figure 7 and Table 1). Both MLH1‐positive and ‐negative areas showed the same variations (i.e., KRAS G12D and APC S1465Wfs*3), indicating that these carcinomas share a common genotype. Eventually, the CUP was revealed from the origin of the gastric cancer cells.

3
Discussion
The primary origin of this case was not primarily identified due to the large duodenal tumor mass at the porta hepatis. No primary tumor was larger than the mass. A few endoscopic and histological examinations revealed a gastric tumor, but it was suspected to be a coincidentally occurring gastric cancer. Since the depth of the laterally spreading cancer was expected to be the submucosa, it was diagnosed as an early‐stage cancer. Initially, we planned to treat the CUP with neoadjuvant chemotherapy. Pancreatoduodenectomy with distal gastrectomy and cholecystectomy was performed to resect the CUP tumor and early gastric cancer together. Histologically poorly differentiated and well‐ to moderately differentiated adenocarcinoma were the final diagnoses of the CUP tumor and gastric tumor, respectively. Metastatic tumors usually have a poorly differentiated phenotype and sometimes undergo genetic changes after anti‐cancer chemotherapy. We hypothesized that gastric malignant cells might cause greater metastasis to the portal hepatis with invasion into the duodenum, gallbladder, and bile duct. Genetic testing using both CGP and target sequencing was useful for identifying the origin of the CUP in our case (Table 2).
Some guidelines are available for the diagnosis and treatment. IHC analysis, which is used to analyze protein expression patterns, may provide a better understanding of the original carcinoma in the case with CUP [9, 10]. For instance, certain markers are associated with specific cancers (e.g., NKX3.1 for prostate cancer or SATB2 for colorectal cancer). Many IHC markers are recommended to indicate the primary cancer [9]. However, IHC is often not definitive, particularly in poorly differentiated or undifferentiated tumors. Therefore, genetic testing in the diagnosis of CUP is challenging. Advances in molecular diagnostics are expected to improve the ability to identify the primary sites of CUP. CGP allows for a comprehensive analysis of both DNA and RNA from tumor samples [11]. It can be used to identify actionable mutations and genetic alterations that may suggest the origin of the cancer. However, although NGS can detect potential therapeutic targets, it does not always provide a definitive primary site [7]. Treatment options for CUP are determined by clinical, pathological, and molecular genetic testing. Site‐directed therapy was decided according to the guidelines. If available, biomarker‐driven therapy can be chosen (e.g., BRAF V600E, NTRK gene fusion, MSI‐high, or HER2‐positive), which is expected to improve the patient prognosis. A comprehensive, multidisciplinary approach is essential, both for an accurate diagnosis and effective treatment planning. Advances in molecular biology and imaging techniques hold promise for improving the diagnostic accuracy and management of CUP.
Genetic alterations in gastric cancers have been well studied, and the variant frequencies were obtained from public databases. According to the cBioPortal website (Figure S1), KRAS and APC variants occur in 15% and 12% of stomach adenocarcinoma, respectively. In colorectal cancer, the same database reports 40% of KRAS and 65% of APC variants. Pancreatic cancer shows 83% of KRAS and 2% of APC, whereas gallbladder cancer shows frequencies of 10% and 3%, respectively. These data suggest colorectal cancer as a likely origin; however, total colonoscopic examination revealed no evidence of colorectal malignancy. Concurrent KRAS and APC variants are uncommon in gastric cancer. Therefore, a combination of targeted sequencing is useful for identifying cancers. CGP using CUP material showed variations in KRAS G12D and APC S1465Wfs*3 followed by the target NGS analysis. In particular, the APC variant S1465W was not coincidental because only 10 gastric cancer patients with the single nucleotide variation were deposited in cBioPortal (Figure S2). Therefore, CUP must be derived from gastric carcinoma. Tumor progression was suspected to have occurred as follows: The gastric tumor remained superficial within the original gastric tissue, while the metastatic cancer cells in the porta hepatis exhibited poor histology and rapid growth. Over time, the gastric tumor partially lost its MMR protein expression, and the metastatic tumor showed MSI‐high status. Gastric cancer develops into an MSI‐high genotype [12]. The recurrence of lymph node metastasis was eliminated by pembrolizumab treatment. As demonstrated in the KEYNOTE‐158 phase II multicohort study, a meaningful and durable benefit with a long‐term response to pembrolizumab is expected against MSI‐high cancers [13, 14, 15]. MSI‐high cancers have been reported in 1.6%–1.8% of CUP cases, and the efficacy of ICI treatment in this setting remains insufficiently studied [16, 17]. Kato et al. [18] reported a successful response to a combination of nivolumab and trametinib in a CUP patient harboring MLH1 and KRAS mutations. Since CUP cases with MSI‐high or deficient MMR are rare and ICI administration in such cases has seldom been documented, our case—demonstrating a favorable ICI response—may be of particular clinical relevance. PD‐L1 status has been reported as a predictor of overall survival in CUP patients treated with ICI [17]. Further studies are warranted to identify predictive biomarkers for ICI sensitivity and improved survival.
A case with a large mass at the portal hepatis was primarily diagnosed as CUP, and a comprehensive approach identified it as a gastric cancer origin. Both the genetic testing and the genotype‐matched treatment with ICI successfully induced complete remission of metastatic cancer.

Author Contributions
T.S. designed this case report and performed a whole study. S.Y. and M.T. contributed to histological diagnosis and immunohistochemical study. H.T. contributed to write this manuscript. Y.O. and Y.M. performed genetic analysis. K.T., K.A., N.U., and S.K. were involved in the patient's diagnosis and treatment. K.M. processed the experimental data and performed the analysis. M.F. supervises this research.

Ethics Statement
The genetic analysis in the present case was approved by Asahikawa Medical University Hospital Ethical Committee (approval number: 19067).

Consent
Informed consent for participation and publication was obtained from the patient.

Conflicts of Interest
Y.O. and Y.M. received research funding from Hitachi High‐Tech Corporation.

Supporting information

Data S1: Supporting Information.