Article Text
PDF
Abstract
Background Over 50% of patients with relapsed or refractory large B-cell lymphoma (r/r LBCL) receiving CD19-targeted chimeric antigen receptor (CAR19) T-cell therapy fail to achieve durable remission. Early identification of relapse or progression remains a significant challenge. In this study, we prospectively investigate the prognostic value of dynamic circulating tumor DNA (ctDNA) and track genetic evolution non-invasively, for the first time in an Asian population of r/r patients undergoing CAR19 T-cell therapy.
Methods Longitudinal plasma samples were prospectively collected both before lymphodepletion and at multiple timepoints after CAR19 T-cell infusion. ctDNA was detected using a capture-based next-generation sequencing which has been validated in untreated LBCL.
Results The study enrolled 23 patients with r/r LBCL and collected a total of 101 ctDNA samples. Higher pretreatment ctDNA levels were associated with inferior progression-free survival (PFS) (p=0.031) and overall survival (OS) (p=0.023). Patients with undetectable ctDNA negative (ctDNA–) at day 14 (D14) achieved an impressive 3-month complete response rate of 77.8% vs 22.2% (p=0.015) in patients with detectable ctDNA positive (ctDNA+), similar results observed for D28. CtDNA– at D28 predicted significantly longer 1-year PFS (90.9% vs 27.3%; p=0.004) and OS (90.9% vs 49.1%; p=0.003) compared with patients who remained ctDNA+. Notably, it is the first time to report that shorter ctDNA fragments (<170 base pairs) were significantly associated with poorer PFS (p=0.031 for D14; p=0.002 for D28) and OS (p=0.013 for D14; p=0.008 for D28) in patients with LBCL receiving CAR T-cell therapy. Multiple mutated genes exhibited an elevated prevalence among patients with progressive disease, including TP53, IGLL5, PIM1, BTG1, CD79B, GNA13, and P2RY8. Notably, we observed a significant correlation between IGLL5 mutation and inferior PFS (p=0.008) and OS (p=0.014).
Conclusions Our study highlights that dynamic ctDNA monitoring during CAR T-cell therapy can be a promising non-invasive method for early predicting treatment response and survival outcomes. Additionally, the ctDNA mutational profile provides novel insights into the mechanisms of tumor-intrinsic resistance to CAR19 T-cell therapy.
- Immunotherapy
- Lymphoma
- Chimeric antigen receptor - CAR
Data availability statement
Data are available upon reasonable request. The datasets used in this study are available from the corresponding author on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
The circulating tumor DNA (ctDNA) demonstrates substantial promise in the prognostic risk stratification of patients undergoing chemotherapy for large B-cell lymphoma (LBCL).
WHAT THIS STUDY ADDS
This prospective cohort study represents the first exploration of the clinical significance of serial ctDNA profiling based on next-generation sequencing in Asian populations receiving CD19-targeted chimeric antigen receptor (CAR19) T-cell therapy for LBCL.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
The integration of dynamic ctDNA detection with imaging methods holds the potential to offer a more precise prediction of treatment response and survival outcomes. Additionally, ctDNA profiling can serve as a guide for timely interventions or personalized salvage treatments, particularly for individuals at risk of early relapse or those experiencing failure in CAR T-cell therapy.
Introduction
CD19-targeted chimeric antigen receptor (CAR19) T-cell therapy has emerged as an efficacious treatment for patients with relapsed or refractory large B-cell lymphoma (r/r LBCL). However, over 50% of patients experience relapse or progression.1–4 Although positron emission tomography (PET) is recommended for response assessments and disease monitoring, approximately one-third of patients initially identified as having a partial response (PR) or stable disease (SD) at 1 month postinfusion ultimately achieve complete response (CR) over time.1 4 This highlights the importance of dynamic monitoring and challenge of distinguishing hypermetabolic residual tumors from inflammation or infections. Since tissue biopsies are often not feasible in practice, there remains a critical need for a novel approach to precisely monitor responses and detect progressive disease (PD) early.
Circulating tumor DNA (ctDNA) exhibits significant potential for tumor surveillance across various solid and hematologic malignancies.5–10 ctDNA quantification enables the prognostic risk stratification of patients undergoing first-line or salvage chemotherapy for LBCL before and during treatment.10–14 In particular, previous studies have indicated that both baseline and interim ctDNA measurements by the next-generation sequence (NGS) method were valuable in identifying high-risk patients undergoing chemoimmunotherapy.15–17 Nevertheless, limited data exist on its applicability in patients receiving CAR T-cell therapy. Frank et al demonstrated that undetectable ctDNA on day 28 (D28) after axicabtagene ciloleucel (axi-cel) infusion was associated with improved survival by tracking the clonotype of immunoglobulin gene.18 Additionally, an alternative approach involving low-pass whole-genome sequencing of ctDNA to detect somatic copy number alterations has been explored.19 20 However, these methods have not adequately addressed the key issue of treatment resistance determinants.
ctDNA-based target sequencing has facilitated the non-invasive exploration of tumor biology in patients with diffuse LBCL (DLBCL) undergoing chemotherapy.12 15 16 A recent Stanford study used plasma samples from patients with r/r LBCL undergoing axi-cel infusion to profile both the tumor and microenvironment, identifying genes such as PAX5, IRF8, and CD274 that correlated with treatment resistance.21 However, these studies were mainly retrospective and focused on axi-cel-treated patients. Further validation and exploration are crucial to uncover resistance patterns in CAR T-cell therapy.
Our study employed a capture-based target sequencing panel to evaluate the utility of ctDNA in a prospective cohort of patients with r/r LBCL undergoing CAR19 T-cell therapy. This is the first study conducted in an Asian population aimed at predicting treatment responses and survival outcomes in CAR19 T-cell therapy through dynamic ctDNA monitoring. Additionally, we analyzed ctDNA mutation profiles and the genomic clonal evolution patterns in patients resistant to therapy.
Methods
Patient information and sample collection
Patients with r/r LBCL who underwent CAR19 T-cell therapy were enrolled in two registered clinical trials (NCT04586478 and CTR20211683), conducted at the Lymphoma and Myeloma Center, Institute of Hematology & Blood Diseases Hospital. Eligible participants met the following inclusion criteria: (1) aged >18 years; (2) had a diagnosis of DLBCL, high-grade B-cell lymphoma (HGBL), or transformed DLBCL; (3) availability of pre-CAR and post-CAR T-cell infusion blood samples; and (4) completion of at least one treatment response assessment. Peripheral blood samples were prospectively collected both prior to lymphodepletion and 14, 28, 60, 90, and beyond 120 days or at progression after CAR19 T-cell infusion. Baseline characteristics were collected from medical records. Response assessments were conducted in accordance with the Lugano criteria, using PET and/or CT scans. These assessments were performed at specific intervals during the study period, including at D30 and D90 postinfusion, followed by assessments every 3 months thereafter. The study was approved by the Ethics Committee of the Institute of Hematology & Blood Diseases Hospital, Chinese Academy of Medical Science & Peking Union Medical College (KT2020116–EC–1), and written informed consent was obtained from all participants before sample collection.
Sample processing and DNA sequencing
Peripheral blood samples were collected in cell-free DNA (cfDNA) storage tubes (Cat# 218962, Streck Inc., Omaha, Nebraska, USA) and then processed by centrifugation within 24 hours to isolate plasma and peripheral blood mononuclear cells (PBMCs) and stored at −80℃. Oral mucosa sample for each patient was also collected for normal control when accessible. cfDNA and genomic DNA (gDNA) were extracted from plasma and oral mucosa cells or PBMCs using the MagMAX CellFree DNA Isolation Kit (ThermoFisher Scientific) and QIAamp DNA Tissue & Blood Kit (Qiagen), respectively, according to the manufacturer’s instructions. The quantity of each DNA sample was measured with the Qubit 2.0 Fluorometer using the Qubit dsDNA HS Assay kit (Life Technologies), and the quality of the extracted cfDNA was assessed using an Agilent 2100 bioanalyzer and the DNA high-sensitivity kit (Agilent Technologies). Serial plasma cfDNA and gDNA samples were sequenced using a custom capture-based target sequencing panel covering 188 lymphoma-related genes (Onco-LymScan panel), as previously described.14
Sequence data analysis
After removing the sequence adapters and low-quality regions using Trimmomatic (V.0.36), sequencing reads were mapped to the hg19 reference genome using BWA (V.0.7.10). Picard was used to distinguish the PCR duplicates, and GATK (V.3.2) was applied for local realignment. Sequencing data from paired tumor-germline samples were used to identify somatic mutations. Single-nucleotide variants and insertions/deletions (Indels) were called using SAMtools (V.0.1.1722), and copy number variations were detected by CONTRA. Structural variation calling was performed by Crest (V.1.0.25). ANNOVAR, Oncotator, and Vep were used to annotate all mutations for genes and functions. All final called variants were verified with the integrative genomics viewer browser (Broad Institute, USA). Mutations in plasma cfDNA were retained when the allele fraction was ≥0.1% and the mutated read number was ≥3.
ctDNA measurements
The ctDNA level for each sample was measured quantitatively by calculating the haploid genome equivalents (hGE) per mL of plasma (hGE/mL), and the result was expressed as a base-10 logarithm (Log hGE/mL). ctDNA from each postinfusion sample was also qualitatively reported to be negative or positive. Specifically, mutations identified from the baseline plasma were selected as disease biomarkers, and subsequently, we sought whether these basal mutations could be traced in the postinfusion plasma samples. ctDNA positive (ctDNA+) was defined as any baseline mutation was redetected, while ctDNA negative (ctDNA–) was defined when all baseline mutations achieved a clearance. Moreover, the ctDNA fragment sizes for each sample were determined by the median value of the length of the reads calculated from the binary alignment/map format.
Statistical analysis
Progression-free survival (PFS) was calculated from the time of CAR T-cell infusion until progression, death, or last follow-up. Overall survival (OS) was calculated from the time of CAR T-cell infusion until death or last follow-up. PFS and OS probabilities were estimated using the Kaplan-Meier method, and the survival of groups was compared using the log-rank test. Regression analysis of covariates was conducted by Cox proportional hazard model. Continuous variables were compared using t statistics or Mann-Whitney U test, and categorical variables were assessed by Fisher’s exact tests. P<0.05 was considered statistically significant. Analyses were performed by GraphPad Prism 8.0 (San Diego, California, USA) and R software package (V.4.2.1; http://www.r-project.org).
Results
Patient characteristics and outcomes
Between July 2020 and August 2022, 23 eligible patients with r/r LBCL undergoing CNCT19 or axi-cel therapy were included. A total of 101 ctDNA samples were collected longitudinally, both pre-CAR19 and post-CAR19 T-cell infusion (figure 1A). Detailed baseline characteristics are summarized in table 1. The median age was 45 years (range, 23–63 years), with 73.9% being male. Before the infusion, 43.5% (10/23) of patients had stage III-IV disease, 54.2% (14/23) of the patients had elevated lactate dehydrogenase (LDH) levels, 65.2% (15/23) exhibited coexpression of MYC and BCL2 (double expressor, DE), and 23.8% (5/21) were double-hit/triple-hit lymphoma. The median line of prior therapy was 3 (range, 2–6), and 12 out of 23 patients had primary refractory disease. Around D30 (±3) postinfusion, a CT scan was performed for most patients, except for eight patients who underwent PET assessments. And the majority of patients (n=21) had PET evaluations at either D90 or D60. With a median follow-up of 12.9 months (range, 3.1–25.7), the overall best response rate was 95.6%, with a CR rate of 56.5%. The 1-year PFS and OS were 56.2% and 71.8%, respectively.
Pretreatment ctDNA level was associated with clinical characteristics and served as a prognostic biomarker. (A) Sample collection. Blood sample was collected before infusion and at 14, 28, 60, 90, and more than 120 days during the first year postinfusion. (B) Pretreatment circulating tumor DNA (ctDNA) level comparison between different International Prognostic Index groups. (C) Pretreatment ctDNA level comparison between patients with I-II and III-IV stages. (D) Association between pretreatment ctDNA level and en sites. (E) Correlation between pretreatment ctDNA levels and lactate dehydrogenase levels. (F) Univariate Cox proportional hazard models for progression-free survival (PFS) and overall survival (OS) according to baseline clinical characteristics. (G–H) Prognostic impact of pretreatment ctDNA level on PFS and OS. LBCL, large B-cell lymphoma; IPI, International Prognostic Index; LDH, lactate dehydrogenase; EN, extranodal; CRP, C reactive protein; ECOG PS, Eastern Cooperative Oncology Group performance status; GCB, germinal center B cell; ctDNA, circulating tumor DNA; hGE, haploid genome equivalents.
Baseline characteristics before infusion
Pretreatment ctDNA level correlates with clinical characteristics and outcomes
We initially quantified pretreatment ctDNA levels in our cohort, with 91.3% (21/23) of patients having detectable ctDNA. The median ctDNA concentration was 2.23 (range, not detected to 4.69) Log hGE/mL. We proceeded to investigate whether ctDNA levels correlated with conventional tumor burden indicators and observed that patients with an International Prognostic Index (IPI) <2 had significantly lower ctDNA compared with those with an IPI ≥2 (p=0.007; figure 1B). High ctDNA was also notably associated with advanced stage (p=0.014; figure 1C), extranodal sites ≥2 (p=0.027; figure 1D), and elevated pretreatment LDH level (Spearman r=0.469, p=0.024; figure 1E). These findings suggest that pretreatment ctDNA holds promise as an alternative marker for assessing tumor burden.
Furthermore, high pretreatment ctDNA, in addition to high IPI scores and DE, demonstrated an adverse impact on PFS (p=0.05) and OS (p=0.05) according to Cox proportional hazard analyses (figure 1F). Patients with low ctDNA (n=11) exhibited superior PFS (p=0.031; figure 1G) and OS (p=0.023; figure 1H) compared with those with high ctDNA (n=12). Specifically, the 1-year PFS and OS rates were 81.8% and 90.0%, respectively, for the low ctDNA group, as opposed to 33.3% and 46.7%, respectively, for the high ctDNA group. Given that DE status was another significant baseline prognostic marker for survival, we further combined pretreatment ctDNA with DE status for risk stratification, revealing a subset with the poorest outcomes characterized by all patients experiencing disease progression within 1 year (online supplemental figure 1A,B).
Supplemental material
The dynamics of ctDNA after CAR T-cell infusion predict response and survival
Following CAR19 T-cell infusion, the median ctDNA level dropped rapidly to 1.01, 0.93, 0.97, and 0.24 Log hGE/mL at D14, D28, D60, and D90, respectively. Patients with sustained response demonstrated consistently lower ctDNA levels during the initial month postinfusion (online supplemental figure 1C). We further investigated the associations between ctDNA status at early timepoints and treatment response. Notably, patients who achieved ctDNA– at D14 and D28 exhibited significantly higher 3-month CR rates of 77.8% and 72.7%, respectively, in contrast to the CR rates of 11.1% and 18.8%, respectively, observed in patients who remained ctDNA+ (p=0.015 for D14; p=0.030 for D28; figure 2A,B).
The predictive performance of circulating tumor DNA (ctDNA) status and fragment sizes for patient outcomes. (A–B) The comparison of 3-month complete response (CR) rates prediction between D14 and D28 ctDNA status. (C–D) Kaplan-Meier survival estimate for progression-free survival (PFS) and overall survival (OS) according to D28 ctDNA status. (E–F) Kaplan-Meier survival estimates for PFS and OS according to D14 ctDNA fragment sizes. (G–H) Kaplan-Meier survival estimates for PFS and OS according to D28 ctDNA fragment sizes. CR, complete response; ctDNA, circulating tumor DNA.
Moreover, we explored the relationship between ctDNA dynamics and patient survival. Our analysis revealed that ctDNA– at D28 predicted significantly longer 1-year PFS (90.9% vs 27.3%; p=0.004) and OS (90.9% vs 49.1%; p=0.003) compared with patients who remained ctDNA+ (figure 2C,D). As for the earlier timepoint at D14, patients with ctDNA– exhibited a trend towards improved 1-year PFS (77.8% vs 33.3%, p=0.077; online supplemental figure 2A) and significantly superior OS (p=0.022; online supplemental figure 2B) than those with ctDNA+ status.
Supplemental material
While the ctDNA status is valuable for prognosis prediction, it relies on the accurate detection of somatic mutations. Therefore, we further investigated the clinical utility of a mutation-independent biomarker, specifically ctDNA fragment size. Survival analysis revealed that patients with long ctDNA fragment sizes (≥ 170 base pairs (bp)) at both D14 and D28 exhibited longer PFS (p=0.031 for D14; p=0.002 for D28; figure 2E and G) and OS (p=0.013 for D14; p=0.008 for D28; figure 2F and H) in comparison with those with shorter ctDNA fragment sizes (<170 bp). Additionally, we constructed a simple risk stratification model incorporating the two D28 ctDNA biomarkers—mutation status and fragment size—and stratifying patients into three distinct risk groups (online supplemental figure 2C,D). Patients with long ctDNA fragment size and ctDNA– exhibited excellent survival outcomes, with 1-year PFS exceeding 80%. In contrast, patients with both short fragment size and ctDNA+ experienced the worst outcomes (p<0.001 for PFS and OS).
Serial ctDNA monitoring mirrors radiographic response evaluation and adds specificity to PET
The individual responses to treatment, reflecting patients with diverse baseline ctDNA levels and postinfusion ctDNA clearance, are presented in figure 3A. After infusion, 15 patients were detected as ctDNA+ at least once, and 66.7% (10/15) of them experienced disease progression, whereas none of the patients with sustained ctDNA– did. The dynamic changes of ctDNA levels, coupled with clinical treatment responses assessed by PET and/or CT scans for each patient, are shown in online supplemental figure 3A–C. Our observations elucidate a coherent correlation between the different patterns of ctDNA dynamics and treatment responses.
Supplemental material
Progression assessment by circulating tumor DNA (ctDNA) monitoring and radiographic evaluation. (A) Swimmer plot showing the postinfusion ctDNA status and the clinical response evaluated by CT or positron emission tomography (PET). (B–C) Representative cases about ctDNA could assist radiographic evaluation for better progression prediction. PD, progressive disease; SD, stable disease; PR, partial response; CR, complete response; NE, not evaluable; PET, positron emission tomography; ctDNA, circulating tumor DNA.
We also compared the predictive capacity of ctDNA monitoring for disease progression with radiographic evaluation. Notably, the D28 ctDNA status demonstrated superior prognostic performance compared with PET evaluation conducted within 3 months postinfusion (online supplemental figure 3D). Among the 11 patients with D28 ctDNA+status, 72.7% (8/11) of them experienced disease progression eventually, whereas only 53.8% (7/13) of patients with positive PET scans progressed, indicating a higher rate of false positivity among patients undergoing PET evaluation. In contrast, only 1/11 (9.1%) patients with D28 ctDNA– status experienced disease progression, while 3/10 (30%) patients with PET positive progressed, suggesting ctDNA may be a more accurate method for monitoring minimal residual disease than PET, for instance, patient RY033, whose ctDNA status turned negative at D28, while PET indicated SD with a Deauville score of 5. Remarkably, this patient ultimately achieved a durable CR (figure 3B). Conversely, patient RY039 achieved CR according to PET evaluation at D28, while ctDNA remained positive, leading to subsequent disease progression at D90 (figure 3B). These cases underscore the potential of ctDNA evaluation to assist clinical decision-making when confronted with equivocal results.
Genomic analysis of serial ctDNA monitoring reveals mutation resistance to CAR T-cell therapy and deciphers clonal evolution patterns in r/r patients
In light of our earlier observations regarding the prognostic value of quantitative, kinetic, and structural characteristics of ctDNA in LBCL, we then elucidated the baseline somatic mutational landscape of the patients in our cohort. The most frequently mutated genes at baseline were TP53 (52%), IGLL5 (48%), KMT2D (35%), PIM1 (26%), and MYC (22%) (online supplemental figure 4A). Notably, we observed a significantly higher mutation frequency in the IGLL5 gene within the PD group (n=10) compared with the non-PD group (n=13) (80% vs 23.1%; p=0.012; figure 4A). Furthermore, certain genes exhibited an elevated prevalence among PD patients, including TP53 (70.0% vs 38.5%), PIM1 (40.0% vs 15.4%), BTG1 (30.0% vs 7.7%), CD79B (30.0% vs 0%), GNA13 (30.0% vs 0%), P2RY8 (30.0% vs 0%), and MYD88 (27.3% vs 7.7%). We identified mutations in several genes as being associated with inferior PFS, including IGLL5 (HR=6.347), CD79B (HR=6.137), P2RY8 (HR=6.293), ETV6 (HR=5.172), and KLHL6 (HR=5.522) (figure 4A), and the adverse effect on OS was similar (online supplemental figure 4B). Of particular note, when employing Kaplan-Meier analysis on genes mutated in at least four cases, patients harboring IGLL5 mutations exhibited both inferior PFS (p=0.008; figure 4B) and OS (p=0.014; figure 4C). These results showed a heavier mutation load for patients with progression and indicated that IGLL5 may be a novel biomarker for predicting unfavorable outcomes.
Supplemental material
Mutational profile revealed IGLL5 gene mutation associated with poor survivals. (A) Mutation profile comparison between progressive disease (PD) patients and non-PD patients (left) and prognostic effect of the top mutated genes on progression-free survival (PFS) (right). Genes associated with poor PFS are labeled in red, and IGLL5 gene is labeled with an asterisk with a significantly higher mutation frequency in the PD groups than in non-PD groups by Fisher’s exact test. (B–C) Kaplan-Meier survival estimate for PFS and overall survival according to IGLL5 gene mutation status. PFS, progression-free survival.
Subsequently, we explored clonal evolution that underwent pressure from CAR19 T cells by comparing seven paired samples (5/7 were germinal center B-cell-like subtype) collected at preinfusion and progression stages. Recurrent gene mutations, such as IGLL5 and TP53, were identified and found to be consistently present at both timepoints across most cases (figure 5A). And we found three distinct clonal evolution patterns indicative of a notable degree of intratumoral heterogeneity. In figure 5B, as referred to a stable clonal pattern, all mutations in patient RY038 at relapse evolving from genetically same clones present preinfusion, but the variant allele fraction (VAF) of all genes became higher. Notably, low VAF of mutations in MYC p.Q50L, IGLL5 p.P20L, and KMT2D p.I5304Kfs*29 genes was observed at a D28 timepoint prior to relapse. In figure 5C, patient RY045 exhibited a stable clone with mutation loss pattern. The new clone of ETV6X10_splice at relapse evolved from a small subclone at preinfusion, and the subclone of TP53 p.P278A was cleared after treatment. Figure 5D illustrates a branching clonal pattern for patient RY046, where a novel clone bearing XPO1 p.M297V and FAT1 p.P3493A mutations evolved from a distinct clone at preinfusion, and the original subclone of DDX3X p.E366* mutation disappeared. Both patients RY038 and RY046 achieved only a PR response and rapidly progressed at 1.5 and 2.2 months, respectively, after treatment, while patient RY045 achieved a CR with a PFS of 4.4 months. It is of note that the co-occurrences of TP53 and IGLL5 mutations were really common in these patients with PD, suggesting a potential synergistic role of these two genes in driving tumor development and resistance to CAR19 T-cell therapy.
Somatic mutation dynamics and clonal evolution patterns for patients resistant to chimeric antigen receptor (CAR) T-cell therapy. (A) Oncoprint showing the landscape of somatic mutations in samples before and after progression on CAR T-cell therapy. (B) Clonal evolution patterns for three representative progressed patients.
Discussion
In this first prospective cohort of r/r LBCL patients treated with CAR19 T-cell therapy in Asian populations, we conducted a capture-based target sequencing panel to explore the clinical significance of serial ctDNA profiling. Our investigation illustrates that the quantitative and fragmentation characteristics within ctDNA dynamically predict treatment responses and survival outcomes following CAR19 T-cell therapy. Furthermore, our findings revealed novel gene mutations that exhibited correlations with tumor-intrinsic resistance and distinctive clonal evolution patterns.
As a non-invasive liquid biopsy method, ctDNA detection serves as an alternative to traditional radiographic assessments without radiation exposure. Compared with so far reports, our cohort also showed pretreatment high ctDNA levels aligned with established prognostic markers,4 22 23 like advanced stage, extranodal involvement, and elevated LDH. Likewise, high pretreatment ctDNA levels were correlated with inferior PFS and OS. These findings underscore the utility of pretreatment ctDNA profiling for risk stratification in patients considered for CAR T-cell therapy.
After CAR T-cell infusion, ctDNA levels displayed a substantial reduction within the initial 2 weeks. In line with previous research by Deng et al,24 the significant decline in VAF during the first week was correlated significantly with ongoing CR at the 3-month follow-up assessed via PET. In our own cohort, individuals with ctDNA– at both D14 and D28 achieved an impressive 3-month CR rate exceeding 70% based on PET evaluations. Moreover, the dynamics of ctDNA clearance held significant predictive value for survival. Earlier studies have indicated that a 2.5-log decrease in ctDNA levels after one or two treatment cycles is associated with favorable outcomes in treatment-naïve DLBCL patients undergoing chemotherapy.11 25 However, the optimal timepoint and level for ctDNA assessment in the context of CAR T-cell therapy remain undefined. Studies by Frank et al18 and Sworder et al21 emphasized the prognostic importance of a 2.5-log decrease or non-detectable ctDNA at D28 following CAR T-cell infusion, which correlated with a notably extended event-free survival. Consistently, our study revealed that ctDNA– at D28 was highly effective in predicting disease recurrence and progression. Additionally, an earlier timepoint, D14, similarly proved valuable in predicting OS in our patient population.
Concomitantly, shorter ctDNA fragment size has been reported to be indicative of higher tumor burden,26 and the baseline fragmentation patterns are associated with distinct survival outcomes.27 Remarkably, we are the first to report that shorter ctDNA fragments (<170 bp), at D14 and D28, are significantly correlated with worse PFS and OS in r/r LBCL patients treated with CAR T-cell therapy. Furthermore, the risk stratification performance of D28 ctDNA status combined with fragment size outperformed models based solely on mutations or fragments. Recent studies have indicated that initiating combination PD1 or PD-L1 antibody treatment early (≤2 weeks) does not enhance CAR T-cell efficacy. Instead, it carries the risk of delaying the peak expansion of CAR T cells.28 29 Considering that some patients initially classified as having a PR or SD at 1 month deepen into CR later, we assert that ctDNA profiling at D28 holds significant clinical predictive value. It aids in early identification of resistant patients, thereby assisting clinicians in implementing timely interventions.
Early PET evaluation 30 days postinfusion has been demonstrated to be associated with unfavorable outcomes.30 However, distinguishing between a metabolically active residual tumor and the ongoing eradication of cancer or inflammation in cases of a positive 1-month PET scan can be challenging. Recently, Dean et al31 provided evidence that concurrent ctDNA assessment enhanced the specificity of 1-month PET scans following CAR T-cell therapy in a prospective multicenter cohort. Nevertheless, in our institution, as in the majority of cell therapy centers, we traditionally perform metabolic response evaluations 3 months after CAR T-cell infusion due to the substantial cost and radiation exposure associated with PET scans. When we compared the D28 ctDNA status with 3-month PET, we observed more patients with ctDNA+ experienced disease progression compared with those with positive PET (72.7% vs 53.8%). Though these assessments were not conducted at matched timepoints, the two methods in several patients who underwent PET scans at D28 postinfusion were further compared. And our results supported that serial ctDNA tracking may serve as an adjunctive tool to add sensitivity and specificity to PET evaluations.
Given the substantial clinical and biological heterogeneity of LBCL, researchers have diligently sought new, effective biomarkers for tumor microenvironment to identify non-responders to CAR T-cell therapy. For the first and largest longitudinal study reported by Sworder et al,21 the authors have identified specific genes such as PAX5, IRF8, and CD274 that correlated with treatment resistance. Employing an NGS-based approach, our serial profiling of ctDNA also offered fresh insights into the molecular mechanisms in patients with r/r LBCL. Although TP53 did not exhibit the previously reported negative impact,32 we identified mutations in certain genes, including IGLL5, CD79B, P2RY8, EVT6, and KLHL6, which are associated with a higher risk of disease progression. Notably, pretreatment mutations in the IGLL5 gene displayed a significant correlation with inferior survival outcomes. It is worth mentioning that IGLL5 mutations appear to be more prevalent in lymphomas originating from the germinal center. Previously, IGLL5 has been recognized as a pivotal driver gene in Burkitt lymphoma,33 and it has been significantly enriched in follicular lymphoma and primary gastrointestinal DLBCL, with correlations established between its presence and shorter survival.34 35 A case report by Kurosawa et al has also highlighted IGLL5 as the founder mutation preceding the development of DLBCL transformation in splenic B-cell lymphoma/leukemia.36 Given the high co-occurrence rates of IGLL5 and TP53 mutations in paired pretreatment and progressive samples (5/7 were germinal center B-cell (GCB) subtype), it is rational to hypothesize that these mutations may act synergistically in driving the development of malignant GCB cells and resistance to CAR T-cell therapy. However, it is essential to underscore that the clinical prognostic value of this finding awaits validation through larger-scale studies in the future.
While our study has yielded valuable insights, it is important to acknowledge several limitations. First, the sample size is small, and the observed clinical performance of ctDNA should be subject to further investigation and validation in larger cohorts. Second, the absence of paired tumor specimens for our patients is a limitation, although our method has demonstrated a median concordance rate of 77.8% in newly diagnosed DLBCL.14 Lastly, in our study, the PET evaluation is determined by the Deauville score, and there is potential for improving specificity by applying the total metabolic tumor volume in future research.
In summary, the dynamic analysis of ctDNA during CAR T-cell therapy emerges as a promising non-invasive tool for predicting outcomes of r/r LBCL and provides crucial insights into the mechanisms of tumor-intrinsic resistance. In anticipation of future clinical practice, there are two key aspects to consider. First, the integration of dynamic ctDNA detection with imaging methods holds promise in stratifying patients for risk assessment. Second, the early and precise prediction of patient efficacy through ctDNA analysis can guide timely interventions, such as the combination of PD-1 inhibitors. Additionally, the development of tailored salvage treatments for individuals experiencing CAR T-cell therapy failure stands to benefit from the incorporation of mutation profiles identified through ctDNA. These approaches aim to enhance the personalized and effective management of patients undergoing CAR T-cell therapy.
Data availability statement
Data are available upon reasonable request. The datasets used in this study are available from the corresponding author on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the Institute of Hematology & Blood Diseases Hospital, Chinese Academy of Medical Science & Peking Union Medical College (KT2020116–EC–1). Participants gave informed consent to participate in the study before taking part.
Acknowledgments
We appreciate the effort of the physicians in enrolling patients and thank all the patients involved for their collaboration during clinical practice.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
HZ, WL and XW contributed equally.
Contributors Concept and design: DZ, LQ, and JW; Methodology: XW, CW, TM, HZ, and WL; Patient recruitment: WL, HL, DS, TX, WH, WS, SY, GA, and YX; Collection and assembly of the clinical data: HZ and WL; Analysis and interpretation of data: HZ, WL, XW, and CQ; Manuscript writing: HZ, WL, and XW; All authors revised the manuscript. Guarantor: DZ.
Funding This work was supported by the CAMS Innovation Fund for Medical Sciences (CIFMS 2022-I2M-1-022, 2021-I2M-1-041, 2020-I2M-C&T-B-085).
Competing interests There are no competing interests.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.