Next-generation Sequencing-based Test for Resectable Colorectal Cancer in Real-world Clinical Practice

This study aimed to evaluate the signicance of Next-generation sequencing (NGS)-based gene panel testing in resectable colorectal cancers (CRC)s by analyzing real-world data collected prospectively from patients. Methods: Patients with CRC who underwent surgery from July 2018 to February 2020 at our institution were included, and correlations between various NGS data and clinicopathological ndings were evaluated. Overall, 107 patients were included in this study. The tumor stage was I in 28 cases (26.2%), II in 40 cases (37.4%), III in 32 cases (29.9%), and IV in 7 cases (6.5%). Actionable gene alterations were found in 97.2% of the cases. Co-alteration analysis suggested that either TP53- or APC-related alterations were more frequently found in early-stage tumors (stage I). The copy number alteration count was signicantly lower in right side colon tumors than in tumors in other locations (P < 0.05). Homologous recombination deciency (HRD) was more often identied in stage IV tumors than in stage I or II tumors (P < 0.05). Moreover, high HRD status was suggested to be useful for identifying high-risk stage II tumors (P < 0.05). In this study, data HRD was identied a useful result of gene panel testing with novel utility in clinical practice.


Introduction
Malignant tumors are traditionally diagnosed and classi ed based on the organ of origin and histological type, and treatments are selected according to such classi cation. However, it has become clear in recent years that malignant tumors are caused by the accumulation of various genetic mutations. Hence, treatment strategies against malignant tumors place more emphasis on targeting such genetic mutations. The concept of "precision medicine," in which genetic mutations in individual malignant tumors are analyzed and individualized treatment targeting those mutations is employed, has also been gaining ground in recent years [1]. In the case of colorectal cancer (CRC), it has been reported that tumors are caused by carcinogenic pathways involving various genetic mutations, such as mutations caused by (NGS) studies, such as those of The Cancer Genome Atlas (TCGA), have been performed and have revealed that CRC can be classi ed into some molecular subtypes based on genomic events [3].
NGS-based genomic testing is currently used in clinical settings for the practice of precision medicine.
Gene panel tests such as MSK-IMPACT TM (New York, NY, USA) and Foundation One® CDx (Cambridge, MA, USA) have been approved by the U.S. Food and Drug Administration (FDA), and their use is spreading to countries worldwide, including Japan [4]. However, the indications for these gene panel tests are limited to locally advanced or metastatic solid tumors for which standard treatment has been completed or advanced solid tumors for which no standard treatment is available. Therefore, currently, very few patients can bene t from testing. In fact, while actionable gene mutations are identi ed in 37-86% of solid cancer patients, only 10-20% of them are actually identi ed as targetable mutations [5][6][7]. Extending the indications of gene panel testing to early-stage tumors may reveal its utility, but no such studies have been conducted thus far.
In our institute, in-house NGS-based gene panel testing, which analyzes 160 oncogenes, is performed for all resectable solid cancer patients in a clinical trial setting. The novelty of this trial is that the gene panel testing is performed immediately after the primary curative surgery, the timing that is not available for other insurance covered testings thus far. In addition, our in-house gene panel testing has cost effectiveness, which is cheaper than other tests. This trial was expected to explore any advantage of genetic information in the decision of treatment after curative surgery.
Here, we report the CRC results: real-world data collected prospectively from all patients with CRC who underwent primary curative surgery at our hospital, including patients with early-stage cancers. The aim of this study was to investigate any additional information on genetic changes during CRC progression and to explore the signi cance of gene panel testing at primary surgery, which will lead to further expansion of the testing. Next-generation sequencing Tumor tissue was collected from surgical specimens of CRC patients who provided consent to undergo comprehensive genomic testing. The details of the panel have been previously reported [4,9,10]. Brie y, genomic DNA was extracted from 10-µm-thick formalin-xed para n-embedded (FFPE) tissue sections of tumor specimens using the Maxwell RSC FFPE Plus DNA Kit (Cat. AS1720, Promega, Madison, USA) according to the manufacturer's instructions. DNA quality was checked by calculating the DNA integrity number (DIN) using an Agilent 4200 TapeStation (Agilent Technologies, Waldbronn, Germany); all analytes had DIN≥2.0. Libraries were generated from 80 (DIN≤2.5) or 160 (DIN>2.5) ng of DNA per sample using the Human Comprehensive Cancer Panel, GeneRead DNAseq Panel PCR kit V2, GeneRead DNA Library I Core Kit, and GeneRead DNA Library I Amp Kit (Qiagen, Hilden, Germany), and the library quality was assessed using the Agilent D1000 ScreenTape (Agilent Technologies). Targeted amplicon exome sequencing was performed using a 160 cancer-related gene panel as previously described [4,9,10]. The targeted regions of all 160 genes were speci cally enriched using oligonucleotide probes. The enriched libraries were sequenced with a paired-end (150 bp×2) sequencing method using the NextSeq sequencing platform (Illumina, San Diego, CA, USA), resulting in a mean depth of 500. The sequencing data were analyzed using the GenomeJack bioinformatics pipeline (Mitsubishi Space Software Co., Ltd., Tokyo, Japan) (http://genomejack.net/) as previously described [11]. The proportion of tumor cells ranged from 5 to 80% (median 45%). Tumor mutation burden (TMB) was de ned as the number of nonsynonymous and synonymous mutations in the target. The estimated copy number (CN) of the tumor cells was calculated by the following formula: estimated CN=(measured CN−2)/proportion of tumor cells+2. Homologous recombination de ciency (HRD) was evaluated by determining the "HRD score". The score was calculated using an algorithm similar to the LOH score in Myriad mychoice® CDx (Salt Lake City, UT, USA). Although the LOH score is calculated by the sum of LOH, telomeric allelic imbalance, and large-scale state transitions, the latter two factors cannot be calculated in target-gene panel sequences due to the limited number of genes. Thus, a unique method of counting copy number alterations (CNAs) has been used to ensure measurement sensitivity. In detail, the score was de ned as the percentage of detected breakpoints in the whole genome and differences in the CNA status of adjacent probe genes. CNA status includes 3 categories: loss, neutral and ampli cation. LOH regions spanning ≥90% of a whole chromosome or chromosome arm are considered to be due to non-HRD mechanisms [12]. Thus, chromosomes with fewer than 2 probe genes (#8, #18, #21 and X in this test) were excluded from the calculation of the HRD score. In addition, chromosomes with the same CNA status on a single chromosome were also excluded. In this study, an "actionable" gene alteration was de ned as pathogenic variants and copy number alterations (CN>4 or homozygous deletions (HD) or LOH). The annotated and curated analysis report was discussed at a genome expert conference consisting of medical oncologists, molecular oncologists, pathologists, medical geneticists, clinical laboratory technicians, bioinformaticians, genetic counselors, pharmacists, and nurses.

Statistical analysis
All statistical analyses were performed with GraphPad Prism 9 (GraphPad Software, Inc., San Diego, CA, USA). The Mann-Whitney U test and chi square test were applied as appropriate. The signi cance level was set as 0.05. Co-alteration analysis results were plotted using Circos, a Perl language-based tool used to represent visual data in a circular form [13]. The Circos plot was generated by using ClicO FS Circular Layout Interactive Converter Free Services. The details of usage have been previously reported [14].

Patient characteristics
The characteristics of CRC patients who were included in this analysis are shown in Table 1 Figure 1a and  Figure 1b shows the co-alteration analysis results according to tumor stage. In stage I tumors, the majority of co-alterations were either TP53or APCrelated, whereas other combinations were identi ed in advanced tumors.

Tumor mutation burden
Regarding TMB, there was no signi cant difference in any of the comparisons (Figure 3a-c

Homologous recombination de ciency
The HRD score was marginally associated with tumor stage (Figure 4b): stage I 0 (0-4.5) v.s. stage II 0 (0-7.5) v.s. stage III 0 (0-8. 3) v.s. stage IV 1.5 (0-9.8) (p = 0.05). There were statistically signi cant differences between stages I and IV (p = 0.02), between stages II and IV (p = 0.005) and between stages III and IV (p = 0.03). We next attempted to use the HRD score as a recurrence risk predictor in stage II patients (Figure 4c). When an HRD score ≥ 1 was de ned as a high HRD score, the HRD score correlated well with the high-risk stage II tumors: low risk 0% v.s. high risk 30% (p = 0.049). On the other hand, the HRD score was not associated with histological type or tumor location (Figure 4a

Discussions And Conclusion
This study presents real-world NGS sequencing data obtained from samples prospectively collected from CRC patients who were eligible for curative surgery. Although there have been many reports on unresectable advanced or metastatic CRCs, the results of this study, in which all the cases were resectable and over 60% were stage I or II cancer, are important for understanding the potential signi cance of genetic testing. According to this study, TMB and CNA count are associated with pathological histology and tumor location and represent the biological features of tumors. A combination of altered genes represents tumor progression. In contrast, the HRD score is better associated with tumor stage and represents tumor progression, suggesting its possible utility in clinical practice.
Here, we employed an in-house targeted amplicon exome sequencing-based panel including 160 cancerrelated genes that has been validated in several solid tumors, such as ovarian cancer and pancreatic cancer [4,9,10]. The detection rate of actionable genes in CRC was higher than 90% with this panel, which is comparable with that with other NGS-based oncogene panels [6, 15,16]. The major driver genes in CRC, including APC, KRAS, and SMAD4, also had mutation rates comparable to those in other studies [17][18][19]. The slightly lower mutation rate in TP53 may be because our study included more early-stage cancers than other studies. Interestingly, co-alteration analysis showed that most of the stage I tumors had alterations with either TP53 or APC co-alterations. On the other hand, more advanced staged tumors (stages III and IV) had various gene combinations. This nding may explain ACS, in which APC and TP53 alterations leading to adenoma formation are the rst to manifest, and other gene alterations accumulate during malignant transformation [20][21][22]. Therefore, our results indicate that stage I tumors are closer to adenomas.
Chromosomal aberrations are present in approximately 85% of CRCs, generally involving losses in 8p, 17p and 18q and gains in 7p, 7q, 8q, 13q and 20q [23]. It has been suggested that these events occur during ACS [24]. In particular, gains of 20q were rst found through banding analysis in CRC and have been observed in more than 65% of CRC cases, which suggests that the genes encoded on 20q have a key role in contributing to the phenotype of CRC when overexpressed [25,26]. In this study, less frequent gains of genes in 13q and 20q were found in MUC than in other CRC types. This nding may suggest that the developmental pathways of mucinous and non-mucinous tumors are different and that the development of mucinous tumors does not rely on ACS.
In this study, the CNA count was related to tumor location. Consistent with other reports, right side colon tumors had lower CNA counts than tumors in other locations [27]. This nding may suggest that cellular genomic instability is more pronounced in left-side tumors than in right-side tumors. In addition, the CNA count tended to increase as the tumor stage advanced. No clear results were proposed ascertaining the existence of a gradual increase in the CNA count between early, invasive and metastatic CRC [28], but it has been suggested that progression from invasive cancer to metastasis is accompanied by an increase in the CNA count [29,30]. A high CNA count means that the number of genes with ampli cation and loss is high, most likely representing genomic instability. Our results suggest that genomic instability increases with cancer progression.
Although immunotherapy has proven to be effective in treating cancers and is being approved for various types of cancer, including CRC, the number of patients who can bene t from it is still limited [31]. TMB is an emerging biomarker of sensitivity to immune checkpoint inhibitors and has been shown to be more signi cantly associated with the response to PD-1 and PD-L1 blockade immunotherapy than PD-1 or PD-L1 expression [32]. In CRCs, TMB is reported to be higher in right side colon tumors than in left side tumors [33]. Although not statistically signi cant, this study also showed that the TMB was relatively higher in right side colon tumors. The distribution of TMB and the subset of patients with high TMB have not been well characterized and are issues to be elucidated in the future.
HRD has received much attention, primarily in breast cancer treatment, since an underlying mechanism of breast cancer formation has been largely attributed to the HRD pathway [34]. Indeed, the importance of the breast and ovarian cancer susceptibility proteins BRCA1 and BRCA2 has been well documented [35,36]. Genomic tests, such as Myriad myChoice® CDx (Myriad Genetics, Inc., Salt Lake City, UT, USA), which detects BRCA1 and BRCA2 mutants, have been approved by the FDA and are used to detect biomarkers for PARP inhibitor treatment [37]. Unfortunately, the relationship between CRC and HRD has not yet been fully studied. A few reports have shown that brain metastases of CRC and locally advanced rectal carcinomas exhibit elevated mutational signatures of HRD [38,39]. In this study, a higher HRD score was clearly correlated with tumor progression, and moreover, it was suggested to correlate well with the highrisk stage II classi cation. Considering postoperative chemotherapy, this nding could be utilized for patient and regimen selection. In fact, a high HRD is associated with susceptibility to platinum agents in ovarian cancers [40].
In conclusion, real-world NGS sequencing data from resectable CRC represent signi cant biological features of cancer progression. Evaluating HRD in each tumor is considered to be useful in clinical practice as a novel readout of gene panel testing.

Declarations
Funding: a) Homologous recombination de ciency (HRD) score in each histological type. b) HRD score in each tumor stage. c) Percentages of high HRD score cases in low-and high-risk stage II cases. d) HRD score in