PDF
Citation
Print
Introduction
Follicular lymphoma (FL) is the most common indolent non-Hodgkin lymphoma (NHL), representing 20%-30% of all cases with indolent NHLs [1,2]. Although involvedsite radiotherapy is recommended with curative intent for patients with localized stage I or II disease, there are no established treatment strategies for most patients with advanced-stage FL [3]. Based on extensive data supporting their efficacy in FL patients, anti-CD20 monoclonal antibodies such as rituximab used in combination with cytotoxic chemotherapies for patients with advanced-stage disease, immunochemotherapies such as bendamustine plus rituximab (BR) or rituximab plus cyclophosphamide, vincristine, and prednisone with or without doxorubicin (R-CHOP or R-CVP) have become the commonly used frontline induction treatments for symptomatic, advanced-stage FL patients [4-9]. In addition, rituximab maintenance after induction treatment is one of the options for patients with advanced-stage cancer, according to the sustained progression-free survival (PFS) benefit shown in the long-term results of the PRIMA study [10]. However, early relapse or progression remains a problematic issue. Progression of disease within 24 months (POD24) has been proposed as a prognostic indicator for FL [11,12]. Furthermore, the occurrence of large-cell transformation is another important issue in the treatment of FL patients because patients with large-cell transformation tend to follow an aggressive clinical course [13-15]. Although patients with initially high tumor burden might be at risk of early relapse or transformation, there are no studies leading to the establishment of a test predicting large-cell transformation and POD24. Therefore, the prediction and estimation of risk of early relapse as well as large-cell transformation might help to establish a risk-adapted treatment strategy for patients with FL.
Circulating tumor DNA (ctDNA) is a small extracellular fraction of DNA derived from tumor tissues or circulating tumor cells. As next-generation sequencing technology can be applied to analyze ctDNA [16], the feasibility of ctDNA in plasma as a predicting marker for early relapse or poor prognosis has been studied in various subtypes of lymphomas such as diffuse large B-cell lymphoma, mantle cell lymphoma, and Hodgkin’s lymphoma [17-19]. Nevertheless, there are limited data about the clinical relevance of ctDNA sequencing in patients with FL receiving systemic immunochemotherapy. Considering the high frequency of bone marrow involvement in FL, tumor-derived ctDNA might represent the genetic nature of the primary tumor and ctDNA sequencing could have a role as a biomarker in FL patients [20]. Therefore, we conducted ultra-deep targeted sequencing of ctDNA using blood samples obtained from FL patients who were registered in our prospective cohort study. In this study, we analyzed the mutation profiles of ctDNA serially collected at the time of diagnosis, at response evaluation and at relapse/progression to evaluate the clinical relevance of ctDNA detection in the treatment of FL patients receiving systemic immunochemotherapy.
Materials and Methods
1. Patients
This study analyzed patients with newly diagnosed, grade 1-3A FL in a cohort of lymphoma patients who were prospectively enrolled in our prospective cohort study between April 2017 and February 2020 (NCT03117036). Plasma cellfree DNA (cfDNA) was collected at the time of enrollment, at response assessment, and at relapse after written informed consent was obtained (Fig. 1). Analysis was performed in patients who fulfilled the following criteria: (1) patients who received rituximab-containing immunochemotherapy and rituximab maintenance; (2) patients who had archived plasma cfDNA samples available for ctDNA profiling; (3) patients for whom baseline and follow-up data, including at relapse and survival, were available; (4) patients who were pathologically confirmed with FL grade 1-3A after a review of tumor tissue samples obtained at the time of diagnosis.
2. Case report forms and treatments
The case report form contained patient information and related clinical and laboratory data as follows: Ann Arbor stage, Follicular Lymphoma International Prognostic Index (FLIPI), Eastern Cooperative Oncology Group (ECOG) performance status (PS), B-symptoms, serum lactate dehydrogenase (LDH), β-2 microglobulin (B2M), number of nodal sites, presence of organomegaly, and bone marrow biopsy results [21]. Patients received three kinds of regimen as frontline treatment: BR; R-CHOP (rituximab, cyclophosphamide, vincristine, doxorubicin, and prednisolone); R-CVP (rituximab, cyclophosphamide, vincristine, and prednisolone). According to the Lugano response classification, the treatment response was evaluated by computed tomography (CT) and 18F-fluorodeoxyglucose positron emission tomography CT (PET-CT) [22]. Patients achieving complete (CR) or partial response (PR) to the frontline treatment received rituximab (375 mg/m2 intravenous infusion or 1,400 mg subcutaneous injection) every 2 months for 2 years as maintenance therapy. During rituximab maintenance, surveillance evaluations were performed to monitor disease relapse. The cut-off date for update of disease and survival status was May 31, 2022.
3. cfDNA extraction and library preparation
Whole blood samples were collected in cfDNA (BCT, Streck Inc., Omaha, NE) tubes. Plasma was prepared using three centrifugation steps with increasing centrifugal force. Cell-free nucleic acid was extracted from plasma using a QIAamp Circulating Nucleic Acid Kit (Qiagen, Santa Clarita, CA). Genomic DNA (gDNA) was isolated using a QIAamp DNA Mini Kit (Qiagen). DNA concentrations were quantified using an Infinite M200 Pro NanoQuant (Tecan, Mannedorf, Switzerland) and the fragment size distribution was measured using a 4200 TapeStation Instrument (Agilent Technologies, Santa Clara, CA). An AllPrep DNA/RNA Mini Kit (Qiagen) was used to extract gDNA from formalin-fixed paraffin-embedded tumor tissue. gDNA was quantified and fragmented in the same manner. Purified gDNA was sonicated (7 minutes, 0.5% duty, intensity of 0.1, and 50 cycles/ burst) into 150-200 bp fragments using a Covaris S2 (Covaris Inc., Woburn, MA). gDNA and plasma DNA libraries were created using a KAPA Hyper Prep Kit (Kapa Biosystems, Woburn, MA). For the library construction of plasma cfDNA, hybrid selection was performed using the customized panel. The panel comprised 61 genes and covered 171,988 bp across the human genome (Fig. 1) [23-27].
4. Sequencing data processing
All data were aligned to the hg19 reference using BWA-mem (v0.7.5). GATK v4.0.0 [28], and SAMTOOLS v1.6 [29] were used for base quality recalibration, cross-validation of the UID family, and sorting SAM and BAM files, respectively. After the sequencing data alignment step, discordant paired and off-target sequencing reads were removed. Picard (v2.9.4) was used to group reads into the same UMI families, and then a home-built Python (v2.7.10) script was used for error suppression. The error suppression method was based on a previous study [30].
5. Detection of ctDNA mutations
All bases were subjected to Phred quality filtering using a threshold Q of 30, and only positions where total depths were above 500× were considered for variant identification. To exclude germline mutations in analysis, non-reference alleles present at a frequency greater than 1% in the matched germline DNA were removed. After applying the error suppression method to the sequencing data, the following selection steps were used to eliminate remaining sequencing errors: (1) Variants not significantly greater than the error found in matched germline DNA (Binomial Bonferroni adjusted p-value < 0.01) were filtered out; (2) Variant candidates with a high strand bias (90% if supporting reads ≥ 20; Fisher exact test, p-value < 0.1 if supporting reads < 20) were removed; (3) If the z-statistic of the variant was not significantly higher than the background error obtained from gDNA (Bonferroni adjusted p-value < 0.05), it was excluded from the analysis. Allele frequency ≥ 0.15% and alternative allele count ≥ 5 were selected for candidate mutations, whereas a variant of ≥ 5% was selected for the case of indel. We employed VEP (v102) [31] to perform functional annotation and excluded synonymous mutations from the analysis. The subsequent analyses considered variants that were non-synonymous mutations, including splicing variants. The levels of ctDNA were quantified as genome equivalents (GEs) that were determined as the product of total cfDNA concentration and the maximal variant allele frequency (VAF) of somatic mutations [32]. Tumor mutation burden (TMB) was defined as the total sum of selected variants across the target panel regions after removing errors. The human GE, converted from VAF, was used for clinical interpretation of the detected variants. Furthermore, we classified the patients into high- and low-mutation burdens using the third quantile value of the TMB (< 10 or ≥ 10). For evaluation of concordance between mutation profiles of plasma ctDNA and primary tumor, we have used the archived mutation data from formalin-fixed paraffin-embedded tumor tissue samples where targeted genetic sequencing was performed using the panel, including 405 genes in practice as previously reported [33]. Thus, we compared the mutation profiles of ctDNA with that of primary tumor tissue in those patients with available tumor tissue sequencing data out of the total study population.
Based on ctDNA information, the M7-FLIPI score was estimated with a weighted summation of high-risk FLIPI, ECOG-PS > 1, and occurrence of at least one non-silent mutation among any seven genes (EZH2, ARID1A, MEF2B, EP300, FOXO1, CREBBP, and CARD11) through the calculator presented at the German Low-Grade Lymphoma Study Group official internet site (https://www.german-lymphoma-alliance.de). In addition, the low- and high-risk groups (< 0.8 vs. ≥ 0.8) were defined according to the same cut-off values in the previous study [24].
6. Statistical analysis
Descriptive statistics were determined as proportions and medians, and the intergroup comparisons for categorical variables were assessed by Fisher’s exact test. PFS was calculated as the date from diagnosis to disease progression or death related to any cause. Overall survival (OS) was defined as the period from the date of diagnosis to death or the last date of follow-up. Survival curves were described using Kaplan-Meier estimates and were compared between groups using the log-rank test. Statistical analyses were performed using IBM PASW ver. 25.0 software program (IBM SPSS Inc., Armonk, NY) and survival analysis package in R software (Survival 3.4-0).
Results
1. Characteristics and outcomes of patients
The characteristics of 40 patients are summarized in Table 1. The median age was 52 years (range, 29 to 81 years), and 65.0% (n=26) were less than 60 years old. There were more males (n=25, 62.5%) than females (n=15, 37.5%). Eighty-seven point five percent of patients showed good general condition on ECOG-PS 0 (n=35). Only five patients (12.5%) experienced B-symptoms at diagnosis. In laboratory analysis, 92.5% of patients (n=37) showed elevated LDH levels, and 47.4% (n=18) presented elevated B2M. Twenty-three patients (59.0%) presented with grade 1-2 FL, and 16 (41%) showed 3A FL. Advanced-stage FL was estimated at 82.5% (n=33), high-risk FLIPI showed 65% (n=26), and a high value of m7-FLIPI was documented at 30.0% (n=12). No significant difference in clinical characteristics was observed between the patients who received BR or R-CHOP/R-CVP. The overall response rate (CR and PR) was 92.5% (37/40) and CR rate was 65.0% (26/40). At a median follow-up of 49.2 months (95% confidence interval [CI], 47.0 to 51.4 months), relapse or progression occurred in 14 patients (n=14/40, 35.0%), and eight patients (n=8/14, 57.1%) died due to disease progression. Three of the 14 patients with disease progression were definitively diagnosed with transformed FL (S1 Fig.). The median PFS and OS did not reach the median value (S2 Fig.). Of 14 patients who experienced disease progression, six received the BR regimen, and eight were treated with R-CVP or R-CHOP. Although the follow-up period of patients treated with BR was relatively shorter than R-CVP/R-CHOP, there was no difference in PFS (NR [not reached] vs. 108.6 months, p=0.987) and OS (NR vs. 128.6 months, p=0.921) according to the type of immunochemotherapy (S2 Fig.).
2. Mutation profiles of ctDNA at diagnosis
To detect ctDNA mutations, we performed targeted deep sequencing in plasma cfDNA and white blood cell (WBC) gDNA from 40 patients. The target panel was designed as a pool of RNA baits covering 61 FL-related genes and targeting 171,988 bp of the human genome. The total reads generated from the plasma cfDNA and WBC gDNA samples were 54.5 and 30.7 million reads on average. After filtering unaligned reads and polymerase chain reaction duplicates from sequencing data, the mean of unique sequencing coverage depths of plasma cfDNA and WBC gDNA were 5283.3 and 6172.8 (S3 Table).
After removing putative false positives according to our statistical criteria, we obtained the list of somatic mutations from plasma cfDNA samples (see Materials and Methods). The mutations of ctDNA were detected in all baseline samples that were obtained at the time of diagnosis (100%, 40/40). The mutation profiles of ctDNA were compared with that of primary tumor tissue in 12 patients who had tissue samples available for analysis. The comparison showed 51.5% (34/66) sensitivity and 50.7% (34/67) positive predictive value of ctDNA of the somatic mutation library from tissue (Fig. 2A). The mutation profiles of cfDNA in 40 patients showed that the most-detected somatic mutation was KMT2C (n=20, 50%), CREBBP (n=18, 45%), KMT2D (n=18, 45%), MEF2B (n=10, 25%), and ARID1A (n=9, 23%) (Fig. 2B). According to the type of frontline treatment, there was no significant difference among mutation profiles. The frequency of M7-FLIPI genes was as follows: CREBBP (n=18, 45%), ARID1A (n=9, 23%), MEF2B (n=10, 25%), EZH2 (n=8, 20%), EP300 (n=6, 15%), CARD11 (n=5, 13%), and FOXO1 (n=4, 10%) (Fig. 2B).
The median ctDNA concentration in the 40 patients was 2.50 human GEs per mL (hGE/mL) (range, 1.38 to 5.45 hGE/mL), and the mean total ctDNA concentration was estimated at 2.62 hGE/mL (2.62±0.88 hGE/mL). Thus, we dichotomized patients into low- and high-ctDNA groups based on the median (< 2.50 hGE/mL or ≥ 2.50 hGE/mL), mean values (< 2.62 hGE/mL or ≥ 2.62 hGE/mL) of ctDNA. Moreover, we classified the patients into high- and low-mutation burdens using the third quantile value of TMB (< 10 or ≥ 10) (Table 2). The ctDNA concentrations were significantly associated with TMB (p < 0.001) (Fig. 3A). However, the ctDNA concentrations were unrelated to the treatment response (Fig. 3B and C). Furthermore, the mutation burdens of seven genes in m7-FLIPI were not different between responders and nonresponders to frontline treatment (p > 0.999) (Fig. 3D).
3. Mutations of ctDNA and survival outcome
The patients with high TMB (≥ 10) showed shorter PFS (p=0.040) and OS (p=0.042) than patients with low TMB (< 10) (Fig. 4A). The patients with high median ctDNA concentration (≥ 2.50 hGE/mL) were associated with poor OS (p=0.041), but not with PFS (p=0.240) (Fig. 4B). Likewise, patients with high mean ctDNA concentration (≥ 2.62 hGE/mL) showed a poor OS (p=0.045), but not PFS (p=0.094) (Fig. 4C). However, the survival outcomes in patients who had the mutational profile on m7-FLIPI were not affected in terms of PFS (p=0.690) and OS (p=0.500) (Fig. 4D).
Among seven genes of M7-FLIPI (EZH2, ARID1A, MEF2B, EP300, FOXO1, CREBBP, and CARD11), only those with a CREBBP mutation showed a tendency to poor PFS (p=0.078) without an association with OS (p=0.110) (Fig. 4E). When we analyzed the association of mutations of each gene in our panel, TP53 mutation was significantly related with worse PFS (p=0.019), and it also showed a tendency to poor OS (p=0.058) (Fig. 4F). In addition, patients having both mutations of CREBBP and TP53 had a tendency to earlier disease progression than those patients without them, regardless of frontline treatment regimens (Fig. 4G). Thus, the mutations of CREBBP and TP53 were correlated with POD24 in the univariate (HR, 3.983; 95% CI, 1.655 to 9.586; p=0.002) and multivariate (HR, 2.433; 95% CI, 0.628 to 9.419; p=0.011) analysis (S4 Table). The CREBBP gene was found in 18 of the 40 patients (45%), and TP53 was found in five patients (13%). When we compared the mutation profiles between ctDNA and tumor tissue in 12 patients who had tissue samples available, CREBBP had a sensitivity of 33.0%, but due to the small sample size in this study, it is difficult to determine the tissue mutation concordance rate for TP53.
4. Longitudinal analysis of ctDNA
Longitudinal analysis of patients achieving CR showed a decrease in ctDNA concentration at the time of CR; mutations detected at diagnosis disappeared at CR in the P016, P020, and P031 (Fig. 5A). On the other hand, patients with relapse or progression showed an increase or reappearance of mutations (Fig. 5B). Of the 14 patients who exhibited disease progression throughout the entire follow-up period, samples were analyzed in 13 cases, except for a patient who had experienced disease progression recently. When we analyzed the dynamics of ctDNA mutations, a patient showed a reappearance of ctDNA mutations that were detected at diagnosis (P002). This patient eventually relapsed during rituximab maintenance. By contrast, in two patients (P011 and P014) who failed to respond to their frontline treatment, the mutations of ctDNA either remained at the same level or increased at the time of disease progression. The longitudinal clonal evolution mutation profiles of the remaining ten patients are shown in S5 Fig., with the exclusion of the three patients presented.
Discussion
In this study, we analyzed ctDNA mutations profiles using serially collected plasma cfDNA from newly diagnosed FL patients to explore the feasibility of ctDNA mutations as a biomarker for the management of patients. The sequencing of plasma cfDNA with baseline samples identified ctDNA mutations in all cases, and high baseline TMB (≥ 10) correlated with higher ctDNA concentration and poor PFS. Among various genes where mutations were detected, the presence of CREBBP and TP53 mutations were associated with the POD24, implying poor prognosis of FL. In a previous study, a TP53 mutation was reported to be associated with poor prognosis in FL patients although the frequency was around 10%- 20% [27]. The frequency of mutations in CREBBP is around 60%-70% and known to be related to the genomic evolution in tumor cells [24,26]. Thus, our findings suggest that the treatment strategy for FL patients could be modified according to the ctDNA mutation profiles including TP53 and CREBBP mutations at diagnosis. In addition, we evaluated the role of ctDNA mutations as a tool for disease monitoring. Although the number of analyzed patients was small, the persistence or reappearance of ctDNA mutations correlated with disease relapse or progression in PET-CT (Fig. 5). This result suggests there is a potential role for ctDNA mutations as a tool for disease prognosis and monitoring in FL patients.
To our knowledge, this is the first study to evaluate the feasibility of ctDNA mutations in newly diagnosed Asian FL patients who received immunochemotherapy and rituximab maintenance. All patients were from our prospective cohort, and serially collected plasma cfDNA samples were made available to us for analysis. Nevertheless, our results should be interpreted with caution due to several limitations. First, there were limitations in the sensitivity and specificity of detecting ctDNA mutations reflecting the mutation profiles of primary tumors in our study because the concordance of mutations between tumor tissue and ctDNA was confirmed in only half of the genes examined (Fig. 2). This low rate of concordance in our study might be associated with the following issues. First, we only compared the mutation profiles of ctDNA with mutations of primary tumor tissue in 12 patients out of 40 patients due to lack of available data for analysis because we only had 12 primary tumor tissue samples for sequencing. Considering that mutations in ctDNA were detected in all baseline samples at diagnosis (100%, 40/40), we might be able to provide more robust data if we could perform the comparison between blood ctDNA and tumor tissue in all patients. Moreover, we focused on patients who underwent rituximab maintenance after induction immunochemotherapy out of all patients with FL, which might also influence the relatively small number of tissue samples in our study population. Secondly, the limitations of our methodology could lead to this result because our panel included a limited number of genes. The targeted sequencing of tumor tissue was conducted in clinical practice using the panel with more than 400 genes at our hospital. We used the data for comparison with that of known ctDNA mutation profiles. However, the number of genes in our ctDNA analysis was much smaller than that of LymphomaSCAN because our study used only 61 genes to detect ctDNA mutations, including essential genes of M7-FLIPI. In our previous studies evaluating the role of ctDNA mutation profiles in various subtypes of NHL including diffuse large B-cell lymphoma and peripheral T-cell lymphomas, the restricted range of genes available for the analysis has been the limitation [34-37]. Furthermore, the analytical performance of our platform might have limitations in the detection of ctDNA mutations. Indeed, the concordance rate increased in our recent study for natural killer/T-cell lymphoma using the different platform with more than 160 genes [38]. Therefore, the feasibility of ctDNA analysis in the clinical application could be influenced by the performance of analytical platforms and the number of genes included in the panel.
In conclusion, the analysis of ctDNA mutations in plasma at diagnosis and during follow-up might help to both predict outcome and monitor disease status of patients with advanced FL considering its potential as a tool for predicting the occurrence of relapse as well as large-cell transformation earlier than it becomes clinically evident. However, there are technical limitations in the low sensitivity of ctDNA analysis results due to heterogeneous mutational presentations. The feasibility of ctDNA measurement should be further improved before it becomes an appropriate and clinically relevant test in FL patients.
XML Download