The detection and sequencing of the mutated ctDNA is one of the irreplaceable clinical measures in the postoperative management of colorectal cancer (CRC) cases. However, we are curious to comprehend the essential traits of mutated genes comprising metastatic sites out of whole mutated genes in primary sites. In the current retrospective study, we conducted target resequencing of ctDNA using 47 plasma samples and established a cancer panel carrying the commonly mutated genes between primary and recurrent tumors. We found that mutated genes in ctDNA indicated immune-resistance traits with respect to the impaired ability to present neoantigens by loss of expression or binding affinity to HLA in the primary tumor. Compared with the estimated neoantigens from all mutated genes in primary tumors, the neoantigen peptides from commonly mutated genes on the panel showed abundant expression but no binding affinity to HLA. Therefore, ctDNA mutations can be frequently and postoperatively detected to identify recurrence; however, these mutated genes were derived from immune-tolerated clones owing to the loss of neoantigen presentation in primary CRC tumors.
In general, we use circulating tumor (ct)DNA as a liquid comprehensive genomic profile (CGP) assay, which is not inferior to CGP tissue analysis in gastrointestinal cancers1,2,3,4,5. In addition, we can trace mutations in ctDNA to monitor the minimum residual tumors for identification of recurrence at the subclinical level. However, in terms of liquid biopsy using ctDNA, we have to comprehend the characteristics of the detected mutation in ctDNA derived from epithelial cells in the primary tumor nests to form recurrence postoperatively.
Considering the essential characteristics of the mutated clones that were chronologically and sustainably detected from the primary site to the recurrent site continuously, we assumed two possibilities. First, the mutated genes in ctDNA may be detected abundantly in tumors with high mutation allele frequency (MAF) or clonally expanded mutated genes that cover the entire primary tumor. These highly mutated or clonally mutated genes may be derived from the dominant cancer cells to promote cancer progression in primary colorectal tumors6. We previously disclosed that driver mutated genes, such as canonical oncogenes and suppressor genes, dominated the entire primary tumor region as the neutral evolution manner in advanced CRC cases7,8. In addition, we previously reported a case of CRC in which mutated KRAS was detected in ctDNA from primary and metastatic tumors simultaneously9, indicating a continuously higher MAF during longitudinal radical treatment.
Another possibility is that the localized host tumor immune response in primary tumors may affect the sensitivity to detect ctDNA in the circulation system. The tumor immune response in cancer microenvironment is comprised of CD8+ cytotoxic T lymphocyte, FOXP3+ CD4+ T regulatory cells, dendritic cells, macrophages, and cytokines. The several former studies touched the association between detectability of mutated ctDNA fragments and the host immunity10,11,12,13, however, they could not reach at any definitive conclusions. In terms of the association between the host immunity and the detectability of mutated ctDNA, the current study focuses on the presentation ability of neoantigens derived from somatic mutations in ctDNA, which is determined by the following two factors: the binding affinity of the diverse estimated neoantigens of mutated genes to human leukocyte antigen (HLA) and expression of tumor-specific RNA transcribed from mutated alleles. Both factors were indispensable for presenting neoantigens derived from mutated genes in ctDNA among all mutated genes in the primary tumor.
This study conducted target sequencing of ctDNA from 47 points in the clinical course of six cases of CRC with postoperative recurrence (CRCR) using a customized cancer panel for target resequencing of commonly mutated genes between primary and recurrent sites14 (Table 1). We calculated the binding affinity to HLA (half maximal inhibitory concentration [IC50]) and tumor-specific RNA expression of mutated genes among all mutated genes in primary tumors by in silico analysis. In the current study, we disclose how tumor immune response can affect the detectability of mutated ctDNA in the primary tumor.
Landscape of mutated ctDNA using target sequencing in six CRC cases
We applied ten primary tumors and ten postoperative recurrence sites to extract genomic DNA for whole-exome sequencing (WES) analysis, which was reported in our previous study14. We selected 443 commonly mutated genes between 10 primary sites and 10 recurrent (metastatic) sites to establish a cancer panel for target resequencing (Fig. S1). In addition, we added 35 significant canonical mutated genes. Out of those 35 genes, twenty-seven genes were overlapped with the 443 mutated genes from the current 10 cases. Therefore, as a consequence, 451 mutated genes were on the panel (Fig. S2). Unfortunately, we could not collect an adequate amount of plasma from four cases shaded areas in Table 1, such as CRCR2, CRCR3, CRCR6, and CRCR10, therefore, we excluded them from the target sequence analysis as the liquid biopsy. The clinical courses of the six cases, CRCR1, CRCR4, CRCR5, CRCR7, CRCR8, and CRCR9 involving 47 samples from primary or metastatic tumors are presented in Fig. 1. For example, in CRCR7, we detected 63 mutated genes out of 65 mutations on the panel (average AF of 63 genes: 0.146) at 4 M (➀) and 63 of 65 mutations (average AF of 63 genes: 0.172) at the diagnosis of metastasis (➁) (Fig. 1).
Verification of ctDNA to capture commonly mutated genes between primary and recurrent tumors
In this retrospective study, it was essential to verify the accuracy of the current assay for implementing the target sequence of plasma ctDNA. We found that this assay system could capture mutated ctDNA genes using a cancer panel that carrying the commonly mutated genes between primary and recurrence sites. As shown in Fig. 2, CRCR1, CRCR4, CRCR7, CRCR8, and CRCR9 have candidate target genes with mutations that were detected repeatedly in ctDNA for tumor tracing throughout the postoperative clinical course. In CRCR7, the PEX5 gene15 was clearly captured multiple times by commonly mutated genes in primary and recurrent tumors.
Comparison of neoantigen presentation ability between mutated genes in ctDNA and all mutated genes in primary tumors
We focused on the ability to present neoantigens derived from commonly mutated genes between primary and recurrent sites compared with whole mutated genes in primary tumors. Presenting neoantigens to activate the tumor immune response requires simultaneous estimation of the binding affinity of the mutated allele to HLA and the expression of the cancer-specific mutated allele. Major histocompatibility complex (MHC) restriction was examined by predicting the binding affinity of single nucleotide variants (SNVs) to HLA (using the analytical pipeline NetMHCpan)16,17 (Fig. S3). We extracted an altered read from the tumor RNA BAM file and measured the expression of tumor-specific mutated genes among all mutated genes in the primary sites as the scheme.
In CRCR7P, we found that tumor-specific mutated PEX5 RNA expression was significantly higher than the expression of all genes in the primary site (Table 2); however, there were no peptide PEX5 fragments within the high range of binding affinity (IC50 < 50 nM) among the estimated 6740 peptide fragments from 104 mutated genes. Therefore, the altered PEX5 must not be presented as a neoantigen peptide. In CRCR1P, mutated OR10A6 showed a higher binding affinity (20 [5.29%] of 378 peptides) to HLA than that of other mutated genes (p < 0.0001). On the other hand, the mutated OR10A6 gene was not presented as a neoantigen (Table 2); therefore, OR10A618must not be presented as a neoantigen peptide. As shown in Table 2, representative mutated genes that could be chronologically traced by ctDNA showed either low binding affinities with HLA or low expression of mutated genes in a mutually exclusive manner. We plotted ctDNA mutated genes to demonstrate the minimized binding affinity to HLA and low expression of mutated transcripts in ctDNA (Fig. 3). Therefore, we assumed that chronologic ctDNA-detected mutated genes were derived from immune-tolerant cancer cells rather than cytolytic activity-inducing collapsed cancer cells.
Comparison of neoantigen presentation ability between commonly mutated genes and all mutated genes in primary tumors
We summarized the results of both factors to determine the ability to present neoantigen peptides in five cases (Tables 3 and 4). As shown in Table 2, CRCR4P, CRCR7P, CRCR8P, and CRCR9P showed significantly higher expression of ctDNA-detected mutated genes compared with other mutated genes in the primary tumor. However, these four cases showed no binding affinity to HLA (Table 4); therefore, none of the mutated genes in the four cases were presented as neoantigens. Furthermore, highly mutated genes were observed in both the primary and recurrent sites and were expressed as transcripts; however, they could not bind to HLA. Consequently, they could not be presented as neoantigens that activate the tumor immune response.
We found that the commonly mutated genes between primary and recurrent tumors indicated the expression of these transcripts, although there is no binding affinity to HLA. Therefore, these mutated genes were not induced neoantigens in the activation of the tumor-immune system. We assumed that the frequently mutated genes in recurrent tumors were derived from immune-tolerated clones in primary tumors without neoantigen presentation. Our previous study supports this finding. We compared the expression of tumor immune response-related genes, such as CD8, CD4, PD-1, LAG3, A2aR, and TIM-3, between primary and metastatic sites using the same RNA seq data from the same sample set used in the current study14. We found abundant expression of an immune exhausted indicator, TIM-3, in metastatic sites compared with that in primary sites in an in-house study as well as The Cancer Genome Atlas data14. In our previous study, we found that postoperative recurrence requires immune tolerance in the cancer microenvironment of colorectal cancer.
Meanwhile, we conducted targeted sequencing of ctDNA using the cancer panel comprising 416 commonly mutated genes between primary and recurrent sites. As a result, several genes, such as OR10A in CRCR1P and PEX5 in CRCR7P, revealed immune-tolerated findings without presentation of the neoantigens due to the mutually exclusive findings in either the loss of expression of mutated genes or lack of binding affinity to HLA. Immune tolerance induced by the loss of neoantigen presentation may be essential for clones to form recurrences. As we described above, commonly mutated clones between primary and recurrent sites indicated immune-resistant owing to the diminished binding affinity of neoantigen to HLA. In addition, the expression of immune exhausted genes, such as TIM-3 was more abundant in the recurrent than primary sites in our previous study14. Wang et al.19 reported that Tim-3 inhibited the MHC-I-restricted antigen presentation not in cancer cells but in macrophages in vitro and in vivo. Regarding the cause of the reduced binding affinity to HLA in CRC, the loss of MHC class I expression plays a pivotal role in presenting processed antigens to T lymphocytes, including tumor antigens in colorectal cancer cases20, and LOH of HLA class I genes and B2M mutations have also been reported to be an indicator of poor prognosis21,22. Therefore, we assumed that most mutated genes in primary and recurrence sites detected by ctDNA have derived from the immune-resistant clones with the loss of MHC class I expression.
The limited number of target genes in each cancer panel was a limitation of the current study. We could not compare the detectability of ctDNA among the three groups, such as primary and recurrence commonly mutated genes, primary site-specific mutated genes, and recurrent site-specific mutated genes, owing to the limited number of plasma samples. In addition, we did not examine the binding affinity of estimated neoantigens to MHC-class II HLAs. Further study is required to elucidate the complete significance of the mutation in the plasma ctDNA. In addition, the detectability of mutated ctDNA preoperatively was low. We usually implement the target re-sequencing analysis using cancer panels of Foundation one, Gardant 360, and others carrying canonical driver genes. However, in the current study, we established and applied the cancer panel carrying commonly mutated genes between primary and recurrence tumors to comprehend the involvement of the host immunity during the evolutional process from primary to recurrence sites. Therefore, we could not detect the mutated ctDNA in the preoperative plasma samples.
In conclusion, recurrence required immune tolerance derived from the loss of neoantigen presentation ability, which was caused either by reduced cancer-specific mutated gene expression or by low binding affinity to HLA in CRC cases. The estimated neoantigen peptide derived from commonly mutated genes between primary and recurrent tumors showed no binding affinity to HLA compared with all mutated genes at primary sites.
Materials and methods
Enrolled patients and plasma samples
We used WES and RNA sequencing on ten primary tumors and ten postoperative metastatic tumors (the first one of metastases in each case) from ten cases of CRC from our previous study14 and established a cancer panel in the current study (Fig. S3). Therefore, we collected and examined 47 plasma samples from six cases of CRC: CRCR1, CRCR4, CRCR5, CRCR7, CRCR8, and CRCR9 (Table 1).
The study design was approved by the institutional review boards and ethics committees of the hospitals to which the patients were admitted (the Kyushu University Hospital Institutional Review Board [protocol number 609-06] and Cancer Institute Hospital Institutional Review Board [protocol number 2010-1058]). This study was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all study participants.
Sample collection and preparation
Genomic DNA and RNA were extracted from freshly frozen tumor samples and adjacent normal intestinal mucosa using an AllPrep DNA/RNA Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions.
Establishment of the cancer panel
We focused on the fundamental dynamics of the ctDNA fraction during the clinical course of CRC. The genome sequences of ten primary tumors and ten metastatic tumors were extracted, and exome sequencing was conducted (Table 1). According to the manufacturer's instructions, DNA was captured using a SureSelect Human All Exon 50 Mb kit (Agilent Technologies, Santa Clara, CA, USA). Captured DNA was sequenced using a HiSeq 2500 (Illumina K.K., Tokyo, Japan) with the paired-end 75–100-bp read option.
The commonly mutated gene of MAF in the primary site and the metastatic site was selected in each case for carrying on the customized cancer panel. In terms of establishing a cancer panel, we used ten primary sites and ten metastatic sites in our previous study (Table 1). We applied 451 mutated genes for the bespoke cancer panel (Fig. S3) established from commonly mutated genes between ten primary and ten metastatic sites. However, because of the inadequate amount of blood samples, we did not conduct a target sequence of plasma samples of CRCR2, CRCR3, CRCR 6, and CRCR10.
Next-generation sequencing library construction
Indexed Illumina next-generation sequencing (NGS) libraries were prepared from plasma DNA. Plasma DNA was used for library construction without additional fragmentation. Genomic DNA was sheared before library construction using a Covaris S2 instrument (Woburn, MA, USA) to obtain 200-bp fragments. According to the manufacturer's protocol, NGS libraries of plasma DNA were constructed using the KAPA Hyper Prep Kit (Kapa Biosystems, Wilmington, MA, USA). A sequencing library was prepared using the KAPA Hyper Prep Kit (Kapa Biosystems) and SureSelect Target Enrichment System (Agilent Technologies). End repair and A-tailing reactions were performed in 60-µL reaction volumes. The mixtures were then incubated at 20 °C and 65 °C for 30 min each. Adapter ligation was performed using 110-µL volumes, and samples were incubated at 16 °C for 16 h using a SureSelect Adapter (Agilent Technologies). After postligation cleanup, the ligated fragments were amplified in a 50-µL solution containing 2 × KAPA HiFi HotStart ReadyMix and 10 × KAPA Library Amplification Primer Mix (Kapa Biosystems). We used the following cycling protocol: 98 °C for 45 s, 14–16 cycles (depending on the input DNA mass) of 98 °C for 15 s, 65 °C for 30 s, 72 °C for 30 s, and 72 °C for 5 min (1 cycle). Library purity, library concentration, and fragment length were determined using a 2100 Bioanalyzer (Agilent Technologies).
Plasma DNA extracted from CRC patient samples was captured using a SureSelectXT Custom 1 Kb–499 kb, 16 (Agilent Technology) according to the manufacturer’s instructions. A panel of 451 genes was designed and validated in this study. Captured DNA was sequenced using a HiSeq2000 (Illumina K.K.) to generate paired-end (75–100 bp) reads for each sample. Targeted deep sequencing was performed for all samples using a multigene panel, with a mean sequencing depth of 3810×.
We used WES data from our previous study14. The sequence data were processed using an in-house pipeline (https://genomon-project.github.io/GenomonPagesR/). The sequencing reads were aligned to the National Center for Biotechnology Information Human Reference Genome Build 37 hg19 with BWA version 0.7.8 using the default parameters. Polymerase chain reaction duplicates were removed using the Picard method. Mutation calling was performed using the EBCall algorithm23 with the following parameters: (1) mapping quality score ≥ 20; (2) base quality score ≥ 15; (3) both the tumor and normal depths ≥ 10; (4) variant reads in tumors ≥ 4; (5) variant allele frequencies (VAFs) in tumor samples ≥ 0.02; and (6) VAFs in normal samples ≤ 0.01.
We used RNA sequencing data from our previous study14; however, we applied RNA seq data from six primary sites and six metastatic sites (black boxes in Table 1). Approximately three billion single-end reads were generated using an Illumina HiSeq 2500 system, as previously described24.
Data availability statement
Data are available at: https://humandbs.biosciencedbc.jp/en/hum0120-v4#target2. Our sequence data are available as NBDC Research ID; hum0120.v4. In terms of mutated ctDNA, we can obtain target sequence data of ctDNA (JGAS000549). In addition, whole exome sequences of 10 primary sites and metastatic sites (9 liver tumors and 5 lung tumors) were available at: Tumor tissues (DRA011183) and non-tumor tissue non-tumor tissues (JGAD000311).
HLA genotyping (Hayashi method)
For HLA genotyping from whole-genome sequencing data, the Bayesian ALPHLARD method was used, which was designed to perform accurate HLA genotyping from short-read data and predict the HLA sequences of the sample. The latter function enables the identification of somatic mutations by comparing the HLA sequences of the tumor and matched normal samples. The statistical formulation for the posterior probability can be described as follows:
where R = (R1, R2) is the pair of HLA types (reference sequences), S = (S1, S2) is the pair of sample HLA sequences, X = (× 1, × 2,…) is a set of sequence reads, and I = (I1, I2,…) is a set of variables using one or two values (jth element; Ij, indicating that the jth read xj is generated from SI j). On the right-hand side of the equation, the left term indicates the likelihood of the sequence reads when the HLA and reference sequences are fixed. The middle and right times are the priors. The parameters, HLA sequences, and HLA types were determined using the Markov Chain Monte Carlo procedure.
Prediction of potential N-acetylglucosamine peptides
Using the Neoantimon package in R, the HLA types of individual patients were obtained (Fig. S3). To identify potential N-acetylglucosamine (NAG) peptides, we used a nonrelapse-based automated pipeline, available at https://github.com/hase62/Neoantimon. Using WES data, this pipeline can easily and automatically construct mutated, and wild-type peptides, including the mutation position, calculation of binding affinity to MHC molecules (using netMHCpan4.0), and integration of the total and tumor-specific RNA expression data based on VAFs calculated from RNA sequence data at the mutation position.
Institutional review board statement
The study design was approved by the institutional review boards and ethics committees of the hospitals to which the patients were admitted (the Kyushu University Hospital Institutional Review Board [protocol number 609-06] and Cancer Institute Hospital Institutional Review Board [protocol number 2010-1058]). This study was conducted in accordance with the principles of the Declaration of Helsinki.
Informed consent statement
Written informed consent was obtained from all study participants.
We used the Mann–Whitney U test or Fisher’s exact tests to test the associations between variables. Data analyses were performed using JMP 14 (SAS Institute, Cary, NC, USA) and R software version 3·1·1 (R Foundation for Statistical Computing, Vienna, Austria).
Chabon, J. J. et al. Circulating tumour DNA profiling reveals heterogeneity of EGFR inhibitor resistance mechanisms in lung cancer patients. Nat. Commun. 7, 11815 (2016).
Huang, A. et al. Detecting circulating tumor DNA in hepatocellular carcinoma patients using droplet digital PCR is feasible and reflects intratumoral heterogeneity. J. Cancer 7, 1907–1914 (2016).
Pectasides, E. et al. Genomic heterogeneity as a barrier to precision medicine in gastroesophageal adenocarcinoma. Cancer Discov. 8, 37–48 (2018).
Ueda, M. et al. Somatic mutations in plasma cell-free DNA are diagnostic markers for esophageal squamous cell carcinoma recurrence. Oncotarget 7, 62280–62291 (2016).
Pantel, K. & Alix-Panabieres, C. Liquid biopsy and minimal residual disease—latest advances and implications for cure. Nat. Rev. Clin. Oncol. 16, 409–424 (2019).
Niida, A. et al. Modeling colorectal cancer evolution. J. Hum. Genet. 66, 869–878 (2021).
Saito, T. et al. A temporal shift of the evolutionary principle shaping intratumor heterogeneity in colorectal cancer. Nat. Commun. 9, 2884 (2018).
Uchi, R. et al. Integrated multiregional analysis proposing a new model of colorectal cancer evolution. PLoS Genet. 12, e1005778 (2016).
Sugimachi, K. et al. Serial mutational tracking in surgically resected locally advanced colorectal cancer with neoadjuvant chemotherapy. Br. J. Cancer 119, 419–423 (2018).
Cristiano, S. et al. Genome-wide cell-free DNA fragmentation in patients with cancer. Nature 570, 385–389 (2019).
Lam, W. K. J. et al. Sequencing-based counting and size profiling of plasma Epstein-Barr virus DNA enhance population screening of nasopharyngeal carcinoma. Proc. Natl. Acad. Sci. USA 115, E5115–E5124 (2018).
Luo, H. et al. Circulating tumor DNA methylation profiles enable early diagnosis, prognosis prediction, and screening for colorectal cancer. Sci. Transl. Med. 12, 524 (2020).
Shoda, K. et al. Monitoring the HER2 copy number status in circulating tumor DNA by droplet digital PCR in patients with gastric cancer. Gastric Cancer 20, 126–135 (2017).
Sakimura, S. et al. Impaired tumor immune response in metastatic tumors is a selective pressure for neutral evolution in CRC cases. PLoS Genet. 17, e1009113 (2021).
Lauer, C., Volkl, A., Riedl, S., Fahimi, H. D. & Beier, K. Impairment of peroxisomal biogenesis in human colon carcinoma. Carcinogenesis 20, 985–989 (1999).
Andreatta, M. & Nielsen, M. Gapped sequence alignment using artificial neural networks: Application to the MHC class I system. Bioinformatics 32, 511–517 (2016).
Hoof, I. et al. NetMHCpan, a method for MHC class I binding prediction beyond humans. Immunogenetics 61, 1–13 (2009).
Duroux, R., Mandeau, A., Guiraudie-Capraz, G., Quesnel, Y. & Loing, E. A rose extract protects the skin against stress mediators: A potential role of olfactory receptors. Molecules 25, 4743 (2020).
Wang, Z. et al. Tim-3 promotes listeria monocytogenes immune evasion by suppressing major histocompatibility complex class I. J. Infect. Dis. 221, 830–840 (2020).
Anderson, P., Aptsiauri, N., Ruiz-Cabello, F. & Garrido, F. HLA class I loss in colorectal cancer: Implications for immune escape and immunotherapy. Cell Mol. Immunol. 18, 556–565 (2021).
Montesion, M. et al. Somatic HLA class I loss is a widespread mechanism of immune evasion which refines the use of tumor mutational burden as a biomarker of checkpoint inhibitor response. Cancer Discov. 11, 282–292 (2021).
Tikidzhieva, A. et al. Microsatellite instability and Beta2-microglobulin mutations as prognostic markers in colon cancer: Results of the FOGT-4 trial. Br. J. Cancer 106, 1239–1245 (2012).
Van Loo, P. et al. Allele-specific copy number analysis of tumors. Proc. Natl. Acad. Sci. USA 107, 16910–16915 (2010).
Magi, A. et al. EXCAVATOR: Detecting copy number variants from whole-exome sequencing data. Genome Biol. 14, R120 (2013).
This research used the supercomputing resources provided by the Human Genome Center, Institute of Medical Science, University of Tokyo (http://sc.hgc.jp/shirokane.html). We thank M. Kasagi, S. Sakuma, M. Murakami, T. Fukuda, N. Mishima, and T. Kawano for their assistance.
This project was supported by AMED, P-CREATE 20 cm0106475h0001(e-Rad ID: 20317791); the Takeda Science Foundation 2020; JSPS KAKENHI (20H05039, 19H03715, 19K09220), Grant-in-Aid for Scientific Research on Innovative Areas (15H05912); Priority Issue on Post-K computer (hp170227, hp160219); Project for Cancer Research and Therapeutic Evolution (19 cm0106504h0004); and a research grant from the Princess Takamatsu Cancer Research.
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Nagayama, S., Kobayashi, Y., Fukunaga, M. et al. Mutated genes on ctDNA detecting postoperative recurrence presented reduced neoantigens in primary tumors in colorectal cancer cases. Sci Rep 13, 1366 (2023). https://doi.org/10.1038/s41598-023-28575-3