Distinct Molecular Subtypes of Classic Hodgkin Lymphoma Identified By Comprehensive Noninvasive Profiling
Bibliographic record
Abstract
Introduction: The scarcity of malignant Reed-Sternberg cells has hampered comprehensive genomic profiling of classic Hodgkin lymphoma (cHL) as might inform personalized therapeutic strategies. Given that profiling of circulating tumor DNA (ctDNA) has shown utility in non-Hodgkin lymphoma genotyping and risk stratification, we employed a noninvasive approach in cHL to overcome challenges imposed by low tumor fraction and improve risk stratification. Patients & Methods: We profiled 478 plasma and 26 tumor samples from 304 patients diagnosed with cHL, 98% of whom were enrolled prior to anti-lymphoma therapy. Median age was 29 (range 4-86), 37% had advanced stage (III/IV) disease, and among the subset with early stage (I/II) disease (63%), 91% had unfavorable GHSG risk. We applied CAPP-Seq and Whole Exome Sequencing (WES) to genotype plasma and tumor samples and used 'phased variant enrichment and detection sequencing' (PhasED-Seq) for detection of measurable residual disease (MRD). Whole exome genotypes were generated using a novel gradient boosting model from mutation and cell-free DNA fragmentomic features. We combined mutation calls with genome-wide copy number profiles to define distinct cHL genetic subtypes by lexical clustering through Latent Dirichlet Allocation. To functionally characterize truncating interleukin 4 receptor (IL4R) mutations, we generated a set of recombinant mutant constructs by site directed mutagenesis, and measured phosphorylation levels of IL4R's proximal downstream target STAT6 following ligand stimulation using flow cytometry. Results: Among 16 patients evaluable for paired tumor and blood specimens, analysis of shared mutations detected in both analytes revealed plasma variant allele fractions (AF) to exceed tumor AFs in 75% of cases (Fig A). The average enrichment exceeded 6-fold, demonstrating noninvasive genotyping to be superior to bulk tumor tissue genotyping for most patients. When compared to patients with diffuse large B-cell lymphoma (DLBCL), median plasma AF in cHL were significantly higher (2.3% vs 1.2%, P=0.03), and cHL tumors shed ~2.75x more ctDNA per mL malignant tumor volume (13.8 vs 5.0 haploid genome equivalents (hGE), P<0.0001). We nominate a candidate mechanism driving this striking variation in ctDNA shedding. To comprehensively profile the coding genomic landscape of cHL, we performed plasma WES (360x median coverage) of 119 pretreatment samples with sufficiently high AF allowing us to identify several novel recurrent lesions and to noninvasively define genetically distinct cHL clusters. Among these newly identified recurrent somatic lesions, we identified a novel class of truncating IL4R mutations in ~10% of cHL patients. These IL4R mutations were distinct from those observed in primary mediastinal B-cell lymphoma (PMBL), with cHL mutations typically disrupting IL4R's intracellular immunoreceptor tyrosine-based inhibitory motif (ITIM) domain and conferring cytokine dependent gain of function phenotypes in vitro through enhancement of IL13, but not IL4 signaling (n=48, P<0.05). Strikingly, IL13 expression was substantially higher in cHL tumors than non-Hodgkin lymphomas, and IL13 amplifications (5q31.1) were enriched in IL4R mutant cases (P<0.001), suggesting an underlying autocrine loop. Finally, unlike hotspot IL4R mutations in PMBL, gain-of-function phenotypes of cHL mutations were blockable by antibodies targeting surface IL4R (n=5, P<0.01), which may therefore serve as a precision therapy target. Among 244 treatment-naïve adult patients, pretreatment ctDNA levels predicted progression-free survival (PFS) both as a continuous (HR 2.1, P=0.02) or a dichotomous variable (HR 3.3, P=0.003). Importantly, associations of pretreatment ctDNA levels and outcomes were independent of stage-based and unfavorable risk groups (both P<0.05). Among patients evaluable for MRD, we observed rapid molecular response to therapy, including after ABVD or Bv-AVD. Specifically, MRD negativity rates at C(ycle)1 D(ay)15 and C3D1 were 38% and 90%, respectively. Importantly, ctDNA detection at both C1D15 and C3D1 were prognostic for PFS (P=0.03 and P=0.002, Fig B). Conclusions: Using a noninvasive approach, we overcome known challenges in cHL profiling and describe several molecularly distinct HL subtypes as defined by genotypes, ctDNA levels, and MRD with diagnostic, prognostic, and therapeutic potential. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".