AUTHORS
Jisu Shin, MPH, HeeJin Cheon, MD, PhD2, Katharine Moosic, PhD3, Bryna Shemo, MA4, Emma Gray, MD5, Timothy William Gaul6, Reese Burgener7, Kathryn Akers8, Nathan Lee Roberts, MD9, David J Feith, PhD10, Aakrosh Ratan, PhD11, Thomas P Loughran, MD12
AFFILIATIONS
1 University of Virginia School of Medicine, Charlottesville, Virginia, United States, University of Virginia, Charlottesville, Virginia, United States
2 University of Pennsylvania, Philadelphia, Pennsylvania, United States
3 University of Virginia School of Medicine, Charlottesville, Virginia, United States
4 University of Virginia School of Medicine, Charlottesville, Virginia, United States
5 University of Virginia School of Medicine, Charlottesville, Virginia, United States
6 University of Virginia School of Medicine, Charlottesville, Virginia, United States
7 University of Virginia School of Medicine, Charlottesville, Virginia, United States
8 University of Virginia School of Medicine, Charlottesville, Virginia, United States
9 University of Virginia School of Medicine, Charlottesville, Virginia, United States, UVA Comprehensive Cancer Center, Charlottesville, Virginia, United States
10 University of Virginia School of Medicine, Charlottesville, Virginia, United States, UVA Comprehensive Cancer Center, Charlottesville, Virginia, United States
11 University of Virginia School of Medicine, Charlottesville, Virginia, United States, University of Virginia, Charlottesville, Virginia, United States, UVA Comprehensive Cancer Center, Charlottesville, Virginia, United States
12 University of Virginia School of Medicine, Charlottesville, Virginia, United States, UVA Comprehensive Cancer Center, Charlottesville, Virginia, United States
ABSTRACT
Introduction: T-cell large granular lymphocyte leukemia (T-LGLL) is a rare clonal lymphoproliferative disorder of cytotoxic CD8+ T cells, typically diagnosed in the mid-sixties. Disease severity ranges from asymptomatic to severe cytopenias requiring treatment. Co-occurring autoimmune diseases such as rheumatoid arthritis (RA) are common. Age at diagnosis spans several decades and has been linked to survival, yet its clinical and molecular correlates remain undefined. We integrated hematologic, treatment, survival, and transcriptomic data to characterize age at diagnosis as an axis of disease heterogeneity in T-LGLL.
Methods: We retrospectively analyzed up to 364 T-LGLL patients from the University of Virginia LGL Leukemia Registry. Complete blood count (CBC) trajectories with age (n=208 untreated patients, CBC within 1 year of diagnosis) were compared with an NHANES healthy reference population (n=8,727) using Fisher's test for equality of correlations and age/sex-adjusted z-scores. A 1,000-iteration subsampling analysis confirmed adequate power. Time to first treatment (n=304, 226 events) was analyzed using Kaplan-Meier estimation, log-rank tests, and Cox models with sequential adjustment for sex, STAT3 mutation status, and RA. Overall survival (n=271, 23 deaths) was analyzed by Cox regression. A logistic model with stepwise AIC selection tested whether baseline CBC parameters predicted eventual treatment need. CD8+ T-cell RNA-sequencing (n=82) was analyzed by weighted gene co-expression network analysis (WGCNA) to identify age-correlated modules, benchmarked against healthy CD8+ naive T cells (DICE database, n=89) with bootstrap resampling to confirm robustness.
Results: Median age at diagnosis was 53 years (range 13-84) in this cohort. 16.5% (n=59) of patients had RA and 37.8% (n=136) harbored STAT3 mutations. Relative to NHANES, hemoglobin and hematocrit showed a weak positive correlation with age in the healthy reference population that was reversed to a significant negative correlation in T-LGLL (cocor p=2.9x10-8 and 1.0x10⁻⁸), an effect most pronounced in males. Absolute lymphocyte count and platelet count, which normally decline with age, showed no age association in T-LGLL (subsampling confirmed this was not a power artifact, detection rate >99%). To distinguish altered age-associated CBC patterns from overall disease effects, we standardized each patient’s CBC values to age- and sex- expected NHANES values. All nine CBC parameters significantly displaced from expected values, with neutropenia the most severely depressed parameter (median z=−2.17) and lymphocyte count relatively elevated (z=+0.80). Consistent with the correlation analysis above, age- and sex-standardized erythroid deficits worsened with age in males but not females. STAT3 mutation status, not age, was the strongest predictor of time to first treatment (median 11.1 vs 34.4 months in mutated vs wild-type, log-rank p=0.03, HR=1.33, p=0.035 after adjustment). Older age at diagnosis was associated with shorter overall survival (HR=1.06 per year, p=0.002), but this association was not significant after Human Mortality Database-based population-mortality adjustment. RA comorbidity showed a numerically shorter time to treatment that did not reach significance (p=0.20). A baseline logistic model retaining hemoglobin, RDW, and ANC predicted eventual treatment need (Tjur's R²=0.31). Age was not a predictor of treatment need, and first-line treatment response did not differ significantly by age (p=0.57). WGCNA of RNA-seq data identified a ribosomal/translation module (201 genes, strongly enriched for the KEGG Ribosome pathway) that declined markedly with age in healthy CD8+ naive T cells but was significantly attenuated in T-LGLL cells (paired Wilcoxon p=3.8x10-24, robust to cohort-size-matched DICE subsampling), alongside a protocadherin-γ cluster module that increased in expression with age in T-LGLL patients (r=0.24-0.29).
Conclusions: Age at diagnosis in T-LGLL is associated with distinct hematologic and transcriptomic features, including disease-specific disruption of normal age-related CBC trajectories, association of STAT3 mutation status with earlier treatment initiation, and relative preservation of ribosomal gene expression programs in leukemic CD8+ T cells. These findings support age at diagnosis as a biologically meaningful axis of disease heterogeneity that may inform future risk stratification in T-LGLL.
DISCLOSURES
Katharine Moosic, PhD: No relevant conflicts of interest to declare.