Abstract
Background/Aim: Although radiation therapy (RT) is an effective and safe treatment when administered locally for various stages of hepatocellular carcinoma (HCC), adequate biomarkers that are predictive of therapeutic efficacy have not been identified. We evaluated the clinical utility of circulating cell-free DNA (cfDNA) to predict treatment response of patients with HCC treated with RT. Patients and Methods: We prospectively recruited 37 patients diagnosed with HCC between March 2019 and May 2020. All patients were treated with RT as salvage therapy. Whole peripheral blood was collected twice, one before RT (baseline; V1) and another aliquot one week after the end of RT (V2). We determined whether cfDNA genomic copy number variations (CNVs) could predict treatment outcome. An I-score was calculated from the plasma cfDNA that reflected CNVs of cfDNA, which is evidence of genomic instability. Results: The I-score at V1 exhibited a strong correlation with the planning target volume (PTV) (coefficient=0.65) and was a predictive marker for progression-free survival (PFS). In particular, a mean I-score value at V1 of ≥6.3 had a significant positive correlation with PFS (p=0.017). Compared with patients who had a complete response (CR) following RT, non-CR patients had a higher mean I-score value at V2 ≥6.2 (p=0.034). Furthermore, I-score values at V1 and V2 and the delta I-score ratio were significantly associated with a pre-RT alpha-fetoprotein level ≥200 among non-CR patients. Conclusion: I-score values calculated from plasma cfDNA represent a potential biomarker for predicting treatment outcomes in patients with advanced HCC receiving RT.
Hepatocellular carcinoma (HCC) is the fourth most common cause of cancer-related deaths worldwide (1). The underlying etiology of HCC involves either viral infection (hepatitis B and hepatitis C) or non-viral causes (alcohol, non-alcoholic fatty liver disease, or non-alcoholic steatohepatitis) (2). Although screening, prevention, and treatment for these risk factors have been widely available, mortality from HCC continues to rise (3).
Treatment options for patients with HCC relies on a multidisciplinary approach based on an individual patient characteristics. Liver transplantation, resection, and radiofrequency ablation (RFA) are standard treatments options with a curative goal for primary HCC (4). However, there is a significant proportion of patients who are not eligible for these treatments. Catheter-based therapies [i.e., transarterial chemoembolization (TACE) or transarterial radioembolization (TARE)] are often administered to gain local control (4). Radiation therapy (RT) is also an effective and safe local treatment for various stages of HCC including unresectable or advanced disease, and its use has increased dramatically over the last two decades (4, 5). RT may also be the treatment of choice for patients who have contraindications to other treatment modalities or impaired portal vein blood flow (2, 4).
Plasma circulating cell-free DNA (cfDNA) is a noninvasive biomarker for cancer diagnosis and it has the potential to reveal genetic information regarding cancer heterogeneity and treatment failure (6-9). Several studies reported the usefulness of cfDNA as a predictive marker for treatment response after tumor resection or sorafenib administration (10, 11); however, the data is insufficient with respect to the prognostic role of cfDNA and clinical factors in advanced HCC treated with RT.
Therefore, we compared cfDNA, which reflects genomic instability, before and after RT to determine a correlation between a cfDNA-derived parameter (I-score) and treatment outcomes. We also evaluated the relationship between cfDNA and clinical factors in the application of cfDNA as a biomarker for predicting response to RT in patients with HCC.
Patients and Methods
Study design and patient eligibility. This was a prospective, single-institution cohort study approved by the Institutional Review Board (IRB No. H.1901-005-999). All patients provided written informed consent prior to enrollment. We recruited patients with a diagnosis of advanced HCC that were scheduled to be treated with RT between March 2019 and May 2020. Inclusion criteria were as follows: 1) Patients aged 20 years or older with an Eastern Cooperative Oncology Group performance status of 0-2, 2) histologically or radiologically proven advanced HCC confined to the liver without clinical evidence of distant metastasis, and 3) completion of planned salvage RT. We excluded patients who had other malignancies within 2 years before enrollment or who received RT for palliative therapy. Patients who had received chemotherapy or immune checkpoint blockade treatment for at least 1 month before or after RT were also excluded.
Whole peripheral blood was collected from each patient twice, once before beginning RT (baseline; V1) and again, one week after RT (V2). Treatment response and recurrence were evaluated every 2-3 months after the end of RT using multi-phase computed tomography (CT) and/or magnetic resonance imaging (MRI). Forty-five patients were initially recruited; however, seven withdrew their consent prior to treatment and one was dropped because of insufficient blood samples.
Treatment. Prior to referral for RT, all patients had received prior treatment, such as surgery, RFA, or TACE. After local treatment, such as RFA or TACE, there was an interval of at least 3 weeks until blood collection and RT. All patients underwent external beam RT as salvage therapy. Either intensity-modulated radiation therapy (IMRT) or stereotactic body radiation therapy (SBRT) was administered at the physician’s discretion. The SBRT dose was 50 Gy either in five fractions (fx’s) (n=10) or 4 fx’s (n=2). In the case of IMRT, the following regimens were used; 50 Gy/10 fx’s (n=13), 50 Gy/25 fx’s (n=4), 50 Gy/20 fx’s (n=3), 55 Gy/25 fx’s (n=2), 40 Gy/16 fx’s (n=2), or 45 Gy/15 fx’s (n=1).
Sample collection and extraction of cfDNA. Peripheral blood samples (10 ml) were collected at the aforementioned timepoints. The blood was centrifuged at 1,600×g at 4°C for 10 min, followed by a secondary centrifugation using the same parameters. After removing the remaining or contaminating cells, only the supernatant was collected, aliquoted (1ml samples) and stored at −80°C. Plasma cfDNA was extracted using the Chemagic cfDNA 2 K kit (PerkinElmer, Waltham, MA, USA). The DNA eluent was quantified using a Qubit 2.0 Fluorometer and a Qubit dsDNA HS (High Sensitivity) assay kit (Cat# Q32851, Life Technologies, Carlsbad, CA, USA).
Library preparation for whole-genome sequencing and calculation of genomic instability. DNA libraries were prepared using a TruSeqNano kit (Cat# FC-121-4003, Illumina Inc., San Diego, CA, USA). Approximately 5 ng of cfDNA was used to perform end-repair, adenylation, and adaptor ligation. After checking the size distribution of the final libraries using high sensitivity D1000 Screen Tape (Agilent Technologies, Santa Clara, CA, USA), the pooled libraries were analyzed using a NextSeq 500 instrument (Illumina, Inc.). The generated reads were aligned to the hg19 human reference genome using the BWA-mem algorithm (0.7.5.a) (10). Polymerase chain reaction duplicates were eliminated using Picard tools (version 1.96) (13). We calculated the relative frequency of the sequencing reads mapped to each bin and modified the GC content bias using the LOESS algorithm (14). A Z-score was calculated using the mean and standard deviation of each bin. To express the extent of genomic instability, we adopted the concept of an I-score (13), which is defined as the sum of the absolute values of the Z-scores of all bins, in which the Z-score value was greater than or less than 2. The I-score formula was as follows: (6)
A higher I-score indicates higher chromosomal instability as a surrogate marker of whole-genome instability (11). In addition, the delta I-score ratio was defined as follows:
Statistical analysis. We determined whether the I-score was related to treatment outcome [progression-free survival (PFS)]. Kaplan-Meier curves for the treatment outcomes were compared using a log-rank test. Fisher’s exact tests were used for categorical variables and Wilcoxon signed-rank tests were performed for continuous variables. A Pearson’s coefficient was calculated to determine the correlation between I-score and planning target volume (PTV). Prognostic factors associated with PFS were evaluated using univariate and multivariate analyses with the backward stepwise Cox regression model. Disappearance of measurable lesions in the follow-up CT and/or MRI performed at least 2-3 months after RT was defined as a complete response (CR). All statistical tests were two-sided and performed using STATA/MP, version 15.0 (StataCorp, College Station, TX, USA) with significance defined as p<0.05. Bar graphs were generated using PRISM version 9.1.1 (GraphPad Software, San Diego, CA, USA).
All statistical tests were performed, and residual plots were depicted by using STATA/MP version 15.0 (StataCorp, College Station, TX)
Results
The median follow-up was 18.9 months (range=5.1-30.4 months). The median age at RT was 61 years (range=41-80 years). Most patients had a hepatitis B virus infection (75.7%) with an Eastern Cooperative Oncology Group performance status of 0 (86.5%) and 35 patients (94.5%) had Child-Pugh Class A liver function. Detailed patient characteristics are presented in Table I.
Baseline characteristics of all patients.
A total of 74 paired samples from 37 HCC patients were analyzed by whole-genome sequencing and an I-score was calculated. The median baseline I-score (I-score at V1) was 5.79 (range=4.54-12.1) and that at 1 week after RT (I-score at V2) was 5.74 (range=4.49-11.61). The I-score at V1 exhibited a strong correlation with the PTV (coefficient=0.65, Figure 1) and it was significantly associated with disease progression after RT (p=0.048, Figure 2A). However, the I-score at V2 showed no significant association with progression (p=0.370, Figure 2B) and a similar result was observed for the delta I-score ratio (p=0.303). Using an I-score cutoff value of V1 >6.2785, which corresponds to the top 95% among subjects without cancer, the area under the receiver operating characteristics curve was 0.708, with a sensitivity of 50.0% and a specificity of 90.9%.
Correlation between I-score at V1 and planning target volume of radiotherapy (Coefficient 0.65 by Pearson’s correlation test).
Comparison of I-score distribution relative to disease progression. (A) I-score at V1 (p=0.048), (B) I-score at V2 (p=0.370). I-score at V1 was significantly associated with disease progression after radiotherapy.
For the analysis of prognostic factors for PFS, patients were grouped using the mean value of the I-score at V1 (6.3) and V2 (6.2). In a multivariate analysis, a higher I-score at V1 with a cutoff of 6.3 had a significantly higher hazard ratio (HR) compared with a lower I-score for PFS (HR=2.69, 95%CI=1.19-6.04, p=0.017) (Table II, Figure 3). However, the I-score at V2 showed borderline significance (HR=2.03, 95%CI=0.88-4.68, p=0.098) only in the univariate analyses, whereas no meaningful results were found in the multivariate analyses.
Univariable and multivariable analyses for progression-free survival.
I-score at V1 ≥6.3 was associated with a significantly worse progression-free survival than an I-score at V1 <6.3 (p=0.013).
Seventeen patients (46.0%) had a complete response after RT (CR group). The CR group showed significantly improved results for PFS (p=0.005, Figure 4A). Although no statistical significance was found, there were differences in the delta I-score ratio between the CR and non-CR groups (p=0.377, Figure 4B). The patients in the CR group clearly showed a negative delta I-score ratio compared with the non-CR group with a positive delta I-score ratio. In a subgroup analysis of prognostic factors for the non-CR group, the I-score at V2 with a cutoff of 6.2 was a significant factor that predicted non-CR following RT (HR=3.45, 95%CI=1.10-10.85, p=0.034) (Table III). A pre-RT alpha-fetoprotein (aFP) value of 200 or greater prior to RT was an adverse factor for CR with borderline significance (HR=3.03, 95%CI=0.98-9.43, p=0.055). To correlate the I-score and aFP to predict non-CR after RT, we performed additional analyses. For non-CR patients, the I-score at V1 (cutoff 6.3) and V2 (cutoff 6.2), and the delta I-score ratio were significantly associated with aFP levels (Table IV). Furthermore, in patients with an AFP ≥200, the delta I-score ratio was associated with non-CR with borderline significance (p=0.051, Figure 5).
Comparison of progression-free survival (PFS) and Delta I-score ratio according to the complete response (CR) status. A) CR after radiotherapy (RT) was associated with PFS (p=0.005). B) The delta I-score ratio of the CR group showed a negative value compared with the positive value of the non-CR group (p=0.37).
Univariable and multivariable analyses of non-complete response after radiation therapy (RT).
Analyses of correlation between I-scores and alpha-fetoprotein (aFP) in predicting complete response (CR).
In patients with an alpha-fetoprotein (aFP) ≥200, the delta I-score ratio was associated with a non-complete response with borderline significance (p=0.051).
Discussion
cfDNA is thought to be released from cancer cells as they undergo apoptosis or necrosis because of the high cell turnover and the hypoxic tumor microenvironment (14, 16). Numerous studies have demonstrated that cancer tissues and cfDNA exhibit similar molecular genetic mutations (17-20) and the amount of cfDNA is proportional to tumor burden (21). Liquid biopsy is a method to obtain the genetic information of cancer using cfDNA in the blood. The analysis of cfDNA by liquid biopsy is useful for predicting prognosis and treatment response and evaluating minimal residual disease during surveillance. Cancer cells are characterized by genomic copy number instability because of chromosomal instability (CIN) as well as tumor-specific point mutations, which is a feature commonly found in various solid cancers (22). CIN has recently been studied in relation to cancer diagnosis and the prediction of prognosis and treatment response. The CIN score is a statistically quantified value of the copy number alterations in cfDNA from cancer patients compared with a normal control group by performing whole-genome sequencing. Genomic instability and CIN score are proportional (23).
Oellerich et al. conducted a study to predict the response to chemotherapy or radiotherapy using CIN scores in 24 patients representing esophageal cancer, colorectal cancer, non-Hodgkin’s lymphoma, pancreatic ductal carcinoma, and non-small cell lung cancer (23). A decrease of 50% or more in CIN score was considered responsive to treatment or stable. Among them, the treatment response for at least 15 patients could be predicted by a change in CIN at approximately 3-8 weeks prior to cancer treatment response confirmation using standard radiologic imaging tests. In addition, the absolute change in the CIN value was significantly lower in the responding group compared with the disease progression group. Roylance et al. performed a study to examine whether CIN scores can predict treatment response in patients with ER-positive breast cancer and ovarian cancer and found that the CIN score was higher in patients who exhibited resistance to treatment (24). These results are consistent with the hypothesis that CIN is associated with tumor progression and metastasis. As previously mentioned, several studies reported cfDNA as a predictive marker for treatment response in patients with HCC who received surgery or sorafenib (10, 11). A study by Atsushi et al. found that cfDNA may represent a predictive marker for disease progression after surgery in patients with HCC (10). Oh et al. recently reported that cfDNA is a potential biomarker that predicts outcome in advanced HCC patients treated with first-line sorafenib (11). For RT, Park et al. revealed that post-RT cfDNA levels were negatively correlated with treatment outcome (25). In addition, after examining the dynamic changes in blood cfDNA levels in patients with stage I-IV HCC, Huang et al. found that cfDNA levels decreased following tumor removal (26). Nonetheless, there have been insufficient studies revealing an association between cfDNA and clinical factors in terms of predicting response and prognosis in patients with HCC who underwent RT.
In the present study, we adopted an I-score calculated from plasma cfDNA as a surrogate marker for genomic instability. Calculated I-scores, which intuitively infer chromosomal instability, were used instead of quantifiable total cfDNA concentrations. We found that the I-score at V1 was a predictive marker for disease progression and PFS and particularly a mean I-score value at V1 had a statistically significant meaning for prediction. Compared with the CR group, patients exhibiting a non-CR were significantly associated with higher I-score values at V2. Based on the results, we confirmed that cfDNA before RT was a predictive marker for original tumor burden in HCC, which may be the reason why the I-score at V1 was a significant factor for tumor metastases and patient survival. The I-score at V2, after 1 week of RT, was unrelated to disease progression and PFS, indicating that it did not reflect overall prognosis, because it was a transient state in which damaged or necrotic cells were mixed immediately after RT. Rather, the I-score at V2 was significant as a prognostic factor in patients who did not show a CR after RT. The I-score immediately after RT may reflect the local control of treatment rather than the overall prognosis. Furthermore, in patients who did not show a CR after RT, all I-score indices (I-score at V1 & V2, delta I-score ratio) were associated with the pre-RT aFP level, and aFP levels >200 showed borderline significance as a prognostic factor for non-CR. Considering that the sample size was limited in this prospective study, it is noteworthy that an association between cfDNA and treatment outcome was also confirmed with a clinical factor, such as a tumor marker. Therefore, tumor response prediction using cfDNA may be particularly useful when aFP before RT is above a certain level in actual clinical practice.
Currently, the evaluation of treatment response after RT is largely dependent on radiographic findings; however, radiologic imaging takes at least two or three months or even longer after RT to confirm the tumor response. In addition, it has the disadvantage of exposing cancer patients to additional radiation and the sensitivity of identifying minimal residual disease is low (27). Therefore, we determined whether a parameter that reflects genomic instability, such as cfDNA (I-score), is useful for clinical practice. Although there was a limitation to the number of samples needed to draw firm conclusions, it is meaningful that cfDNA results can be compared through blood sampling before and immediately after treatment. Further studies are needed to validate our results using a larger patient cohort. Additionally, if HCC-specific methylation patterns are identified through methylation sequencing of cfDNA, the usefulness of cfDNA for diagnosis and prognosis would be significantly increased (28).
Conclusion
The I-score calculated from plasma cfDNA represents a potential biomarker that predicts treatment outcome in patients with advanced HCC receiving RT. Further studies are needed to determine whether cfDNA is a useful biomarker for RT-treated patients with HCC in a clinical setting.
Footnotes
Authors’ Contributions
Conceptualization: Hyun-Cheol Kang; Data curation: Dong-Yun Kim, Jae Sik Kim, Eui Kyu Chie, Hyun-Cheol Kang; Formal analysis: Dong-Yun Kim, Eun-Hae Cho, Hyun-Cheol Kang; Investigation: Dong-Yun Kim, Hyun-Cheol Kang; Methodology: Dong-Yun Kim, Eun-Hae Cho, Eui Kyu Chie, Hyun-Cheol Kang; Project administration: Dong-Yun Kim, Eun-Hae Cho, Jae Sik Kim, Eui Kyu Chie, Hyun-Cheol Kang; Supervision: Eui Kyu Chie, Hyun-Cheol Kang; Writing – original draft: Dong-Yun Kim; Writing – review & editing: Dong-Yun, Eun-Hae Cho, Jae Sik Kim, Eui Kyu Chie, Hyun-Cheol Kang.
Conflicts of Interest
The Authors have no conflicts of interest to declare in relation to this study.
- Received June 5, 2023.
- Revision received July 4, 2023.
- Accepted July 5, 2023.
- Copyright © 2023, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.