Progress in Molecular Classification of Follicular Lymphoma: A Dual Typing Predictive Model Based on RNA Sequencing and Immunohistochemistry

Follicular Lymphoma (FL) is a B-cell malignancy characterized by a slow clinical progression, with a median overall survival extending up to 20 years. However, the heterogeneous nature of FL leads to significant differences in treatment response and prognosis for individual patients. Despite the success of traditional clinical-biological parameters (e.g., FLIPI, FLIPI-2, PRIMA-PI) in guiding therapy, these tools are inadequate for individualized precise diagnosis and selection of specific therapies. Hence, the demand for precision medicine is increasingly pressing.

In past studies, the molecular cell-of-origin (COO) classification of diffuse large B-cell lymphoma (DLBCL) has demonstrated substantial value in prognostic assessments of ABC and GCB subtypes. However, molecular classification within FL remains insufficiently defined. Preliminary studies have suggested that FL may exhibit distinct molecular signatures similar to normal germinal center (GC) B-cells and memory (MEM) B-cells, but no comprehensive classification model has yet validated its clinical significance.

In a study led by Camille Laurent and colleagues, a research team utilized RNA sequencing (RNA-seq) and immunohistochemistry (IHC) to molecularly subtype FL within the large-scale RELEVANCE clinical trial, establishing a predictive model rooted in cell-of-origin. This paper, published in Blood, breaks new ground for precision treatment in FL.

---

Study Origin and Design

The study involved a multidisciplinary collaboration of experts across various international institutions, including Toulouse Cancer-Oncopole, Bristol Myers Squibb, the Lymphoma Study Association, and the Marseille-Luminy Immunology Center. The primary study samples were derived from the RELEVANCE trial, a Phase III clinical trial comparing rituximab-chemotherapy (R-chemo) with rituximab-lenalidomide (R2) in 1,030 FL patients with a high tumor burden.

The study primarily analyzed tumor tissue samples from 433 patients, performing RNA sequencing, DNA whole-exome sequencing (WES), and IHC. A subset of samples also underwent fluorescence in situ hybridization (FISH). To validate the findings, the investigators used additional cohorts from the PRIMA and BCCA studies.

---

Methods: RNA Sequencing and Molecular Model Development

1. Core Experimental Workflow

The team first conducted an independent component analysis (ICA) on RNA-seq data from 324 patient samples, identifying 46 key gene clusters. Using unsupervised clustering techniques, two gene clusters, cc17 and cc21, were isolated as pivotal for FL molecular subtyping. These were associated with signals enriched in germinal center B-cells (GC B-cells) and memory B-cells (MEM B-cells), respectively. Based on these clusters, the researchers developed a 20-gene Linear Predictor Score (LPS), named FL20, capable of subtyping samples into GC-like and MEM-like subtypes.

Additionally, key features such as mutational characteristics, exomic sequences, immune cell infiltration patterns, and other factors were analyzed to validate the relationship between the subtypes, biological features, and clinical outcomes.

2. Gene Subtyping and Mutational Characteristics

An analysis of WES data from 242 samples revealed 89 significantly mutated genes, including well-documented FL drivers such as CREBBP, KMT2D, and EZH2. GC-like subtypes exhibited a mutational profile resembling GCB-DLBCL, with frequent mutations in genes like EZH2 and STAT6, indicating closer biological ties. MEM-like subtypes, on the other hand, showed a higher mutation rate in genes related to the MTOR pathway (e.g., ATP6AP1 and RRAGC mutations). This mutational pattern suggests notable differences in tumor evolution and treatment responsiveness between the subtypes.

3. Clinical Prognosis and Model Validation

In the treatment arms, GC-like subtypes had significantly longer progression-free survival (PFS) compared to MEM-like subtypes under R-chemo treatment. However, both subtypes had comparable PFS under R2 treatment. Further analysis showed that MEM-like patients exhibited significantly better survival outcomes with R2 treatment compared to R-chemo. This conclusion was successfully validated in independent PRIMA and BCCA cohorts.

---

Core Findings and Development of an IHC Algorithm

To facilitate clinical application of FL20, the team developed an IHC-based typing algorithm (called FLCM). Using four antibodies—FOX-P1, LMO2, CD22, and MUM1—FLCM allows routine stratification of FL patients into GC-like or MEM-like subtypes. The FLCM achieved 96.6% concordance with the RNA-seq-based FL20 typing and demonstrated strong inter-pathologist reproducibility.

Additionally, simplified versions of the FLCM algorithm (FL-MAB and FLC-MAB) using only 2 or 3 antibodies were developed. While these versions allow convenient implementation, they led to higher proportions of unclassified cases.

---

Significance and Implications

This study represents the first to stratify FL into GC-like and MEM-like subtypes at the transcriptomic level, establishing clear links to clinical prognosis. The study’s significance is multi-faceted:

1. Precision Treatment Guidance: MEM-like subtypes are unsuitable for R-chemo treatment but may benefit from R2 therapy or future-generation immunomodulators (e.g., more potent Aiolos/Ikaros degraders).

2. Broad Applicability of Subtyping: Both the FL20 model and IHC methods can be broadly applied across laboratories and clinical cohorts, enabling widespread use in tumor stratification.

3. Theoretical Breakthrough: The study addresses a gap in molecular COO-based classification for FL and clarifies that DLBCL’s subtyping methods (e.g., ABC, GCB classifications) do not directly apply to FL.

4. Tool Innovation: The FLCM algorithm provides an economical, efficient tool for clinical pathology, facilitating rapid and practical FL stratification.

---

Future Directions and Challenges

While the study demonstrated the reliability of FL20 and FLCM, their applicability may be influenced by the heterogeneity of GC-like/MEM-like distributions across cohorts. Additionally, the roles of non-coding region mutations, alternate therapies, and immune-chemotherapy regimens in FL warrant further exploration.

This research lays the groundwork for molecular classification and precision treatment of FL. Its findings have the potential to drive more individualized FL therapy selections, improving long-term patient survival outcomes.