Comprehensive Characterization and Validation of the Tumor Microenvironment in Patients with Relapsed/Refractory Large B-Cell Lymphoma Identifies Subgroups with Greatest Benefit from CD19 CAR T-Cell Therapy
Bibliographic record
Abstract
Multiple immune therapies such as CD19 CAR T-cells and bispecific antibodies are now approved for patients with relapsed/refractory large B-cell lymphoma (rrLBCL). The efficacy of these therapies is likely influenced by lymphoma microenvironment (LME) characteristics, but these have not been comprehensively characterized in rrLBCL. We performed single-nucleus multiome (RNA + ATAC), bulk RNA sequencing, and whole exome sequencing (WES) of 120 biopsies from 115 patients with rrLBCL to capture both hematopoietic and non-hematopoietic cell (NHC) types. After stringent quality control, 970,239 cells were analyzed. Non-B-cell lineages were classified into 76 transcriptionally-distinct cell types using unsupervised clustering (21 T/NK subsets; 25 myeloid subsets; 30 NHC subsets), including many subpopulations that have not been previously characterized in lymphoma. LME archetypes were defined by non-negative matrix factorization (NMF) of non-B cell types, yielding 5 cellular modules: lymph-node 1 [LN1] and lymph-node 2 [LN2] characterized by lymph-node structural cell types, antigen presenting cells, and naïve and memory T cells; T-effector/exhausted [TEX] characterized by high frequencies of effector and exhausted CD8 T-cells; and fibroblast/macrophage 1 [FMAC1] and 2 [FMAC2] characterized by high frequencies of macrophage and fibroblast subsets including cancer associated fibroblasts (CAFs). The LN1 and LN2 modules and the FMAC1 and FMAC2 modules were each correlated and therefore considered collectively in tumor archetype construction, resulting in three major archetypes: LN (33% of tumors), TEX (25% of tumors) and FMAC (42% of tumors). The TEX archetype was significantly enriched for the activated B-cell (ABC) cell of origin subtype (P = 0.01) and germinal center B-cell (GCB) subtype occurred more frequently within the FMAC archetype (P = 0.08). The FMAC archetype was significantly enriched for “Dark-zone” signature (DZsig)-positive (P < 0.001), and the LN archetype significantly enriched for DZsig-negative (P=0.002) tumors. There was no significant association between archetype and LymphGen subtype. Cell-cell communication analysis revealed significant differences in predicted ligand-receptor pair interactions between archetypes, with the FMAC archetype being characterized by TGFB1 and PDGF signaling; the TEX archetype by PD1, CTLA4 and TIM3 signaling; and the LN archetype by CXCL12, IL7, CCL19 and CCL21 signaling. Among these biopsies, 17 were pre- and 13 post-CAR T cell therapy. Analysis of these cases generated the hypothesis that the LN archetype was associated with greater benefit from CAR T cell therapy. To test this, we leveraged our bulk RNA-sequencing data plus published Nanostring data from the ZUMA7 study of axicabtagene ciloleucel (axi-cel) in second line rrLBCL to develop a Naïve Bayes classifier for these archetypes. Evaluation of response data from ZUMA7 showed that the greatest benefit for axi-cel compared to chemotherapy was observed within the LN subtype (HR=0.2; P<0.0001), compared to the FMAC (HR=0.34; P<0.0001) and TEX (HR=0.65; P=0.12). As such, LN subtype patients had significantly longer PFS compared to FMAC and TEX subtype patients in the axi-cel arm (HR=0.5, P=0.01), with 1 year PFS of 70%, 46% and 37%, respectively. There was no significant difference in PFS between archetypes in the chemotherapy arm (HR=1.1; P=0.74). In conclusion, accurate construction of the rrLBCL LME using direct cell measurements of both hematopoietic and non-hematopoietic cells with single-nucleus genomics permits the identification of cell types, cell modules, pathways of cell-cell interaction, and LME archetypes with important implications for LBCL biology, and may present an opportunity for LME-guided selection of patients most likely to benefit from cellular therapy.
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 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| 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".