Clinically Significant Shared and Distinct Genomic Alterations in Chinese and Western Patients with Intrahepatic Cholangiocarcinoma

Background The goal of this study is to disclose the clinically significant genomic alterations in patients with of the Western Methods A total of 86 Chinese patients were enrolled in this study. Samples from those patients were sequenced for a panel of pan-cancer genes. Results were compared to a public dataset from a cohort of Western patients. The comparison between the two populations was conducted in the driver genes, actionability, and TMB. Results The Chinese and Western cohorts had 38 and 12 driver genes, respectively. Seven driver genes ( IDH1 , KRAS , TP53 , BAP1 , PBRM1 , ARID1A , and NRAS ) were shared by the two cohorts. For both cohorts, half of the patients had actionable mutations. The two cohorts shared most of the actionable genes but differed much in the frequency. Though KRAS mutations were at the first and second actionable rank respectively for Chinese and Western populations, they were still at a relatively low level of actionable evidence. Four driver genes ( SPTA1 , ARID2 , TP53 , and GATA1 ) were found significantly correlated with the tumor mutation burden. Conclusions The revealed genomic alterations with clinical significance could help to improve the treatment of intrahepatic cholangiocarcinoma.


Introduction
In contrast to the Chinese population, intrahepatic cholangiocarcinoma (iCCA) is rare in the Western population (Singal et al., 2011). However, the incidence of iCCA was increasing in the Western population. Though at a low incidence, iCCA regularly had a poor prognosis. The only potentially curative treatment for iCCA is surgery. In a cohort study, half of the patients who underwent curative resection had a 5-year survival rate at 19% and a median survival of 27.6 months (Dhanasekaran et al., 2013). With the advent of precision medicine, targeted therapy and immunotherapy would change the treatment of iCCA, especially for those unresectable or resistant patients.
The first step of precision medicine is to identify the genotype of patients. Zou et al. (2014) report a landscape of intrahepatic cholangiocarcinoma in the Chinese population by whole-exome sequencing.
They identified 8 driver genes in the 103 iCCA patients. Those genes were TP53, KRAS,IDH1,PTEN,ARID1A,EPPK1,ECE2 and FYN. Another study in the Western population conducted by Lowery et al.
(2018) compared intrahepatic and extrahepatic cholangiocarcinomas with a panel of 410 genes. By sorting the frequency of mutations in patients, they got another eight commonly mutated genes in intrahepatic and extrahepatic cholangiocarcinoma including IDH1, TP53, ARID1A, BAP1, KRAS, PBRM1, SMAD4 and ATM. Besides, there was also a 30-patients study comparing different types of biliary tract cancer with a 22-gene panel by Hogdall et al. (2020). They identified three significantly mutated genes including ARID1A, TP53, and KRAS in CCA (cholangiocarcinoma). All the above studies concentrated on the mutations in one population, neglecting the difference between populations. In considering the significant incidence difference between the Chinese and Western populations, it would be interesting to know the clinically significant shared and distinct genomic alterations.
In this study, we sequenced 86 samples from Chinese iCCA patients. Since different researches used different methods to call the driver genes, directly comparing the results from the multiple pieces of research could be problematic. Through the same analysis pipeline, driver genes, actionability and tumor mutation burden were estimated and then compared.

Mutation Calling
A total of 579 pan-cancer genes (Yuansuo® from Origimed incorporation, Shanghai, China) were captured by targeted amplification. Adaptors were trimmed from raw DNA reads by cutadapt (https://github.com/marcelm/ cutadapt, version 1.18) and de-duplicated with an in-house pipeline.
High-quality reads were mapped to the UCSC hg19 reference sequences using BWA MEM (version 0.7.9a) (Li and Durbin, 2009)

Tumor Mutational Burden Analysis
Tumor mutational burden (TMB) was calculated as the number of mutations multiplied by a adjust factor. The adjust factors were 1.1152 and 0.9738 for the two assays from the MSK dataset, and 0.7875 for the ORI dataset. The TMB difference between mutated and wild type patients was compared by the Mann-Whitney U test. The false discovery rate (FDR) was used to adjust the pvalues.

Results
The landscape of somatic mutations among the Chinese and Western populations A total of 86 Chinese patients with primary intrahepatic cholangiocarcinoma were enrolled in this study (ORI dataset). The characteristics of these patients were listed in Table 1. The median diagnosed age of Chinese patients was 59, ranging from 18 to 83 years old. For comparison, we also included a cohort dataset (MSK) curated by Zehir et al. (2017), which consisted of 87 iCCA patients from the Western population. The median diagnosed age of Western patients was 65 years old, ranging from 37 to 79. The Chinese cohort had a younger age than the Western cohort at diagnosis (p-value = 0.024, Mann-Whitney U test).

Driver Genes In Both Populations
To remove the possible false positive, we utilized MutSigCV (Lawrence et al., 2013) to identify the driver gene in iCCA. The same full genome coverage file was used as a control for both populations.
The calculation identified 38 and 12 significantly mutated genes for ORI and MSK datasets, respectively ( Fig. 2A) (Supplemental Table 1). Seven mutated genes (KRAS, TP53, BAP1, IDH1, PBRM1, ARID1A, and NRAS) were shared by them. There were also 31 ORI-specific driver genes and 5 MSK-specific driver genes. Most shared driver genes had a higher mutation allele frequency in both cohorts (Fig. 2B). Gene ontology analysis of the biological processes revealed enriched functions in macromolecule modification, regulation of cell proliferation and positive regulation of metabolic process for the ORI cohort (Fig. 2C). MSK cohort was enriched with the glyoxylate cycle, regulation of neuron death and regulation of cell proliferation (Fig. 2D). KEGG pathway analysis identified a significant pathway of melanoma, generic cancer and endometrial cancer in the ORI cohort (Supplemental Fig. 1) but none in the MSK cohort.

Clinical Actionability Of Genomic Variations
In order to reveal the treatment potential, mutations were annotated with the drug actionability into six levels of evidence proposed by OncoKB (Chakravarty et al., 2017). Each mutation was assigned the highest level according to the actionability proof strength. Then we summarized the best treatment for each patient. For the ORI cohort, 46.98% of patients had actionable mutations (Fig. 3A).
For the MSK cohort, 40.56% of patients had actionable mutations (Fig. 3B). Among them, 23.26% and 34.83% of patients had standard care biomarkers predictive of response to an FDA-approved drug (Level_2A). Next, we studied whether the mutations in each gene had an equal actionability for both populations. The actionability for each mutation was summarized for each gene. The best highest actionable level was chosen for each gene in a patient. Then we counted the number of patients for each gene by the highest levels. ORI cohort had more actionable mutations in KRAS, PIK3CA, NF1, and CDKN2A (Fig. 3C). MSK cohort had more actionable mutations in IDH1, KRAS, PIK3CA, NRAS, IDH1 and ATM (Fig. 3D). Both cohorts had a similar set of actionable genes, differing only in their frequency (Fig. 3CD).

Driver Genes Correlated With High Tumor Mutation Burden
Tumor mutation burden was found a significant biomarker for immunotherapy efficacy in many other pieces of research. Next, we investigated the TMB distribution in both cohorts. First, the TMB was calculated based on the number of missense mutations across the covered genome length. The median TMB of the ORI cohort was 3.1, ranging from 0 to 50.2. The median TMB of the MSK cohort was 2.95, ranging from 0.98 to 27.88. No significant difference was found between the averages of the two cohorts (Fig. 5A). There were 11 and 5 patients with TMB > 10 mut/Mb for the ORI and MSK cohorts, respectively. Among them, two patients without actionable genes had TMB > 10 mut/Mb in each cohort. TMB was significantly correlated with KMT2D, MUC16, SPTA1, ARID2, FAT4, FREX2, KMT2C, ACVR2A, LRP1B, and NF1 in ORI cohort (Fig. 5B-K) and with TP53 and GATA1 in MSK cohort (Fig. 5LM). Among those genes, SPTA1, ARID2, TP53, and GATA1 were found as driver genes in their specific cohorts according to the analysis above.

Discussion
The significantly distinct incidence of iCCA between the Chinese and Western populations had driven us to disclose their genetic and actionable difference. Previously, driver genes were often identified by frequency among the populations previously. As suggested by Dees et al. (2012)  Seven genes (KRAS, TP53, IDH1, ARID1A, PBRM1, NRAS, and BAP1) were shared between ORI and MSK cohort. Their mutation prevalence was higher than the median in both cohorts. We compared them to the results from other multiple publications (Zou et al., 2014, Lawrence et al., 2013. The sharing genes included KRAS, TP53, IDH1, and ARID1A, which were driver genes in both ORI and MSK cohorts. The other three genes (PBRM1, NRAS, and BAP1) were not reported in the research of Zou et al. (2014) and Lowery et al. (2018). Actually, NRAS is a very important gene involving many cancer-related signaling pathways, such as MAPK, mTOR, and PI3K-Akt signaling pathways. Currently, NRAS mutation is actionable at level 3B. Patients with NRAS mutation could be treated with Binimetinib or Binimetinib + Ribociclib.
Most iCCA patients can benefit from targeted therapy. About 50 ~ 60% of patients had actionable mutations in the ORI cohort and MSK cohort. Among the actionable mutations, the MSK cohort had a higher actionable level than the ORI cohort, which implied the big potential in the treatment of iCCA for the Chinese population. The major gap between the actionability between the two cohorts lied in the two major actionable genes, KRAS and IDH1. ORI cohort had a higher percentage of KRAS mutation at actionable level 4 and less percentage of IDH1 mutation at the actionable level 2B.
Considering the low evidence proof in KRAS mutations in both populations, drugs targeting KRAS mutations should be improved urgently.
K601E mutation in BRAF was recommended to be treated with PLX8394.
Mann-Whitney U test had revealed that TMB from both cohorts was not significantly different, which was consistent with the results of an abstract (Abdel-Wahab et al., 2019). The higher incidence of iCCA in the Chinese population did not come from a higher mutation rate. Immunotherapy could have similar efficacy in both populations. There were several successful cases of immunotherapy for iCCA. taking anti-PD-1 immunotherapy after chemotherapy resistance. In both cohorts, two patients who had no actionable mutation but had TMB > 10mut/Mb could benefit from immunotherapy. Besides, the identified genes that were highly correlated with TMB could be biomarkers for immunotherapy prognosis. For example, MUC16 (CA-125) is a very long protein (14,500 amino acids), which is easier to mutate in cancer cells. MUC16 was reported to associate with a higher TMB and a better immunotherapy outcome in Gastric Cancer (Li et al., 2018).
This study also had limitations. For example, TMB was calculated from different gene panels for the ORI and MSK cohort, which could weaken the results of the comparison. But according to a more restrict comparison, the conclusion still held between the Chinese and Western populations (Abdel-Wahab et al., 2019).
In summary, this study compared the Chinese and Western populations in the driver genes, actionability and TMB for iCCA patients. Shared and distinct driver genes were identified. KRAS and IDH1 mutations appeared in about 30% of patients with a significant frequency bias in the two

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests Funding This work was supported by grants from the clinical medicine innovative plan of Jinan (201907073) YJZ and WR designed the study. SX, YG, and YWZ analyzed and interpreted the patient data. SX, YG, XZ, NF, ZZ, GR, YJZ, and WR collected the patient information and samples. SX, YG, YWZ, ZS, XZ, NF, ZZ, GR, YJZ, and WR drafted the manuscript. All authors read and approved the final manuscript.        Genes associated with tumor mutation burden (A) The tumor mutation burden (TMB) distribution in both cohorts was plotted. (B-K) Ten genes were significantly correlated with TMB in the ORI cohort (FDR<0.01). (LM) Ten genes were significantly correlated with TMB in the ORI cohort (FDR<0.1).

Figure 5
Genes associated with tumor mutation burden (A) The tumor mutation burden (TMB) distribution in both cohorts was plotted. (B-K) Ten genes were significantly correlated with TMB in the ORI cohort (FDR<0.01). (LM) Ten genes were significantly correlated with TMB in the ORI cohort (FDR<0.1).

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download. suplementaltable1.docx