AUTHORS
Joo Song, MD, Hongli Li, MS2, Michael LeBlanc, PhD3, Jibin Zhang, PhD4, Lisa Rimsza, MD, PhD5, Deborah Stephens, DO6, Alexey V. Danilov, MD, PhD7, Jennifer E. Amengual, MD8, Jonathan W. Friedberg, MD9, Sonali Smith, MD, FASCO10
AFFILIATIONS
1 City of Hope, Duarte, United States
2 Fred Hutchinson Cancer Center, Seattle, Washington, United States
3 Fred Hutchinson Cancer Center, Seattle, Washington, United States
4 City of Hope, Duarte, United States
5 University of Arizaon, Tuscon, Arizona, United States
6 University of North Carolina, Chapel Hill, California, United States
7 City of Hope, Duarte, United States
8 Columbia University, New York City, New York, United States
9 University of Rochester, Rochester, New York, United States
10 University of Chicago, Chicago, Illinois, United States
ABSTRACT
Background
The genetic classifiers of diffuse large B-cell lymphoma (DLBCL) — the Shipp/Chapuy DLBclass clusters (C1–C5) and the Wright/Schmitz LymphGen classes — were derived almost entirely from advanced-stage cohorts. Whether limited-stage disease (~30% of DLBCL) shares this architecture is unknown. SWOG S1001 (Persky, JCO 2020) prospectively treated PET-determined limited-stage DLBCL, providing a cohort in which to define its molecular spectrum and benchmark it against advanced-stage disease.
Methods
Whole-exome sequencing and PureCN copy-number profiling were performed on pre-treatment S1001 biopsies. A genetic-sample matrix was used to generate paired DLBclass and LymphGen calls. Of 108 eligible samples, 91 were assigned both subtypes and comprise the analysis cohort. Subtype frequencies were compared with the Chapuy et al (Blood 2024) advanced-stage reference (n=699) by two-sided Fisher exact test. Cell-of-origin was assigned by Hans IHC and/or Lymph3Cx GEP. Progression-free survival (PFS; 20 events) and overall survival (OS; 15 deaths) were analyzed by subtype as event rates with timing binned into 0–2, 2–5, and >5-year intervals; patient-level time-to-event data are pending.
Results
The 91 patients were representative of the parent trial: median age 61 years (range 23–85), 52 (57%) male, ECOG 0 in 65 (71%), stage I in 57 (63%) and stage II in 33 (36%), and extranodal disease in 43 (47%). COO was GCB in 55 (60%) and ABC/non-GCB in 19 (21%).
DLBclass distribution (limited stage vs advanced stage): C1/BN2-like 16% vs 20% (p=0.49), C2/A53-like 24% vs 24% (p=1.0), C3/EZB-like 21% vs 16% (p=0.29), C4/ST2-like 25% vs 14% (p=0.009), C5/MCD-like 13% vs 26% (p=0.009). Median DLBclass confidence was 0.69 (IQR 0.55–0.89); 44/91 (48%) of calls exceeded 0.7, lowest in C1 (median 0.64) and highest in C2 (0.84). LymphGen independently reproduced the shift: ST2 enriched (15% vs 5%; p=0.0003), MCD depleted (4% vs 12%; p=0.03).
The 12 limited-stage C5 cases were genetically atypical: only 3/12 carried MYD88 L265P and none carried CD79B mutations (0/12; 0/91 cohort-wide), so no case showed the MYD88 L265P + CD79B co-mutation that defines advanced-stage MCD/C5.
C5/MCD-like disease was less frequent in limited stage yet retained its adverse prognostic association, with the highest progression-free survival (PFS) event rate (5/12, 42%) and mortality (4/12, 33%) of any cluster, versus 19% and 14% in the remaining clusters (PFS p=0.13; OS p=0.11) — directionally consistent but not significant at 20 events. C4/ST2-like and C1/BN2-like had the lowest event rates (13% PFS each). Interim positron emission tomography (PET) positivity (11/89, 12%) showed no subtype enrichment, and no PET-evaluable C5 case was positive (0/11), so PET did not capture the C5 signal. 9/20 PFS events (45%) occurred beyond 5 years, concentrated in C3/EZB-like (5/5 events >5 yr), whereas all C5 events occurred within 5 years.
Conclusions
To our knowledge this is the largest prospectively treated, uniformly PET-determined limited-stage DLBCL cohort with genomic profiling. Limited-stage DLBCL has a molecular spectrum proportionally different from advanced-stage disease: the favorable C4/ST2 axis is enriched while the high-risk C5/MCD genotype is rare and, when present, lacks the canonical MYD88 L265P + CD79B co-mutation. Nonetheless, C5/MCD-like cases concentrated adverse events and were undetected by interim PET, while C3/EZB-like cases drove the characteristic late-relapse tail. These hypothesis-generating results, limited by modest event numbers, support incorporating genetic classifiers into limited-stage DLBCL trials.
DISCLOSURES
Joo Song, MD: No relevant conflicts of interest to declare.