The study team performed single-cell RNA sequencing on baseline peripheral blood mononuclear cells from 22 patients, yielding 135,271 quality-controlled cells for analysis. Patient-level pseudo-bulk comparisons identified inflammatory and metabolic pathway enrichment in patients who did not reach pathological complete response. Specifically, investigators reported enrichment of TNF-alpha/NF-kappaB signaling in cytotoxic T/NK and B cells from non-pCR patients (global q ≤ 3.94e-7). The cytotoxic T/NK compartment also showed higher activity of MYC targets and mTORC1 signaling in non-pCR patients (both global q = 7.00e-5). Short sentences punctuate the signal: baseline inflammation tracked with worse pathologic outcome. Longer analytic clauses follow—these transcriptional shifts suggest systemic immune activation and metabolic reprogramming that may reflect dysfunctional antitumor immunity before therapy.
T-cell transcriptional patterns differed by outcome. The cohort that achieved pCR maintained signatures consistent with preserved naive and memory features. The authors noted a higher proportion of CX3CR1-expressing cells within a CD4 memory/activation-like subset in pCR cases; this association reached q = 0.0157 with an AUC of 0.949. That is a strong discrimination signal in this dataset. The source did not report mechanistic assays to prove causality; the observation remains associative and requires replication.
B-cell baseline programs also diverged. pCR was associated with a nominally higher naive-like B-cell program score (P = 0.029; q = 0.059), supported by elevated IGHM and IGHD expression, while XBP1 expression was higher in non-pCR patients. The authors interpreted these differences as potential markers of B-cell maturation states that correlate with eventual pathological outcome. The original report did not provide functional validation of these B-cell transcriptional signatures.
Flow cytometry was applied to 125 peripheral-blood specimens collected at baseline, during treatment and before surgery. Those longitudinal data exposed compartment- and time-dependent remodeling, with the most notable effects observed in B-cell populations. At the pre-surgery time point, the balance of CD38 expression on switched-memory B cells differed between outcome groups: 62.6% in pCR patients versus 50.0% in non-pCR patients (P = 0.0086; q = 0.0667). The effect size (Cliff's delta = 0.651) and diagnostic performance (AUC = 0.825) were reported by the study team.
Temporal patterns were not uniform. Early increases in unswitched-memory B-cell CD38 balance were numerically greater in non-pCR patients, whereas the pre-surgery switched-memory B-cell CD38 balance was higher in pCR patients. These dynamics suggest that treatment timing and B-cell compartment identity matter when interpreting peripheral immunophenotypes. The authors emphasized that the CD38-defined differences remained at least nominally significant in leave-one-patient-out sensitivity analyses, indicating some robustness to single-patient influence. The source did not report specific chemotherapy backbone details beyond stating pembrolizumab-based neoadjuvant chemoimmunotherapy; dosing and schedule were not detailed in the preprint.
T-cell phenotypes changed as well but with a contrasting pattern: pCR-associated T-cell features included preservation of naive/memory programs and the CX3CR1 signal in CD4 memory/activation-like cells. The report did not present interventional data to link these circulating T-cell states mechanistically to intratumoral immune activity or to survival outcomes beyond their association with pCR.
Analyses combined nonparametric testing, effect-size metrics, false-discovery rate correction and sensitivity checks. The scRNA-seq pseudo-bulk results included highly significant global q values for inflammatory pathways (TNF-alpha/NF-kappaB q ≤ 3.94e-7) and for metabolic programs (MYC, mTORC1 q = 7.00e-5 in cytotoxic T/NK). The CX3CR1 proportion in CD4 memory/activation-like cells associated with pCR (q = 0.0157; AUC = 0.949), signaling high discriminatory ability in this set of samples. For flow cytometry, pre-surgery switched-memory B-cell CD38 balance differed by outcome (P = 0.0086; q = 0.0667; Cliff's delta = 0.651; AUC = 0.825). The authors also reported a nominal B-cell naive-like program score difference (P = 0.029; q = 0.059). Short declarative sentences present the metrics. Longer sentences outline interpretation and caveats: effect sizes and AUCs indicate potential biomarker utility but require prospective testing in independent cohorts.
The source applied leave-one-patient-out analyses to test robustness; some associations remained nominally significant. The dataset sizes underlying scRNA-seq (22 patients) and flow cytometry (125 specimens from 64 patients) limit broader generalizability and the authors framed these findings as candidate signals rather than validated biomarkers.
The investigators declared no competing interests. They reported ethics approvals from the Provincial Research Ethics Committee of Malaga and the Santiago-Lugo Research Ethics Committee (CEI-SL; approval number 2024/036) and confirmed adherence to the Declaration of Helsinki. Raw data were deposited in Zenodo (DOI provided in the source). The authors emphasized that circulating transcriptional states and CD38-defined B-cell dynamics are nominations for prospective validation; the preprint did not present independent cohort validation or mechanistic proof. The source did not provide explicit chemotherapy dosing or schedule details in the text. Open questions that remain in the dataset include whether the peripheral signatures track with long-term outcomes beyond pCR, how they relate to intratumoral immune states, and whether they retain performance across diverse patient populations. These are the specific next steps the report indicates for further work.