Cutaneous T-cell lymphoma (CTCL), primarily mycosis fungoides (MF) and Sézary syndrome (SS), has long been characterized as a neoplasm of mature memory T cells, based on monoclonal T-cell receptor (TCR) rearrangements and tissue-resident memory (TRM)/central memory (TCM) T-cell phenotypes. This review synthesizes converging population-genetic, multi-omic, and single-cell evidence to argue that this characterization is incomplete and that a progenitor-based model better accounts for the full spectrum of disease biology. We present evidence that initiating mutations arise in hematopoietic stem or early lymphoid progenitor survive thymic selection, and diversify after TCR assembly, resulting in branched evolution across both blood and skin. In SS, paired analyses reveal > 200 shared variants between CD34+ progenitors and Sézary cells, as well as signal-joint T-cell receptor excision circle (sjTREC) positivity, providing direct progenitor-level evidence. In MF, convergent signals, multiple malignant clonotypes per lesion, greater blood–skin than skin–skin clonotype overlap, and compartment-specific CNV subclones, implicate hematogenous seeding and reseeding. Population-scale lymphoid clonal hematopoiesis and lymphoid-pattern mosaic chromosomal alterations define a compatible antecedent state. Spatial single-cell atlases and trajectory analyses map site-conditioned programs in skin, including Th2-skewed cytokines, microbial responses, and UV signatures, that select and expand subclones and explain inter- and intra-patient heterogeneity. This framework reconciles mature immunophenotypes with upstream initiation and clarifies why compartment-focused therapies often reshape rather than eradicate disease. It yields testable predictions and actionable implications: trials should pair multicompartment cytoreduction with strategies that attenuate progenitor-derived reservoirs, restore immune balance, and repair skin barrier dysfunction. A progenitor-initiated, niche-adapted model provides a coherent scaffold for more durable control in CTCL.
Zhang et al. (Tue,) studied this question.