Background: Tertiary lymphoid structures (TLSs) play key roles in tumour adaptive immunity. However, the prognostic value and molecular properties of TLSs in oesophageal squamous cell carcinoma (ESCC) patients have not been studied.
Methods: The prognostic values of the presence and maturation status of tumour-associated TLSs were determined in 394 and 256 ESCC patients from Sun Yat-sen University Cancer Center (Centre A) and the Cancer Hospital of Shantou University Medical College (Centre B), respectively. A deep-learning (DL) TLS classifier was established with haematoxylin and eosin (H&E)-stained slides using an inception-resnet-v2 neural network. Digital spatial profiling was performed to determine the cellular and molecular properties of TLSs in ESCC tissues.
Results: TLSs were observed in 73.1% of ESCCs from Centre A via pathological examination of H&E-stained primary tumour slides, among which 42.9% were TLS-mature and 30.2% were TLS-immature tumours. The established DL TLS classifier yielded favourable sensitivities and specificities for patient TLS identification and maturation evaluation, with which 55.1%, 39.5% and 5.5% of ESCCs from Centre B were identified as TLS-mature, TLS-immature and TLS-negative tumours. Multivariate analyses proved that the presence of mature TLSs was an independent prognostic factor in both the Centre A and Centre B cohorts (p < .05). Increased proportions of proliferative B, plasma and CD4+ T helper (Th) cells and increased B memory and Th17 signatures were observed in mature TLSs compared to immature ones. Intratumoural CD8+ T infiltration was increased in TLS-mature ESCC tissues compared to mature TLS-absent tissues. The combination of mature TLS presence and high CD8+ T infiltration was associated with the best survival in ESCC patients.
Conclusions: Mature TLSs improve the prognosis of ESCC patients who underwent complete resection. The use of the DL TLS classifier would facilitate precise and efficient evaluation of TLS maturation status and offer a novel probability of ESCC treatment individualization.
Keywords: deep learning; digital spatial profiling; oesophageal squamous cell carcinoma; prognosis; tertiary lymphoid structure.
© 2022 The Authors. Clinical and Translational Medicine published by John Wiley & Sons Australia, Ltd on behalf of Shanghai Institute of Clinical Bioinformatics.