From cytology to systems pathology: Decoding the molecular heterogeneity of resectable non-small-cell lung cancer

DECIPHERING TUMOR HETEROGENEITY: FROM MICROSCOPIC IMAGING TO MACROSCOPIC GENOME CHARACTERISATION

Histological heterogeneity is a hallmark of NSCLC. Classical morphological features including lymphovascular invasion, necrosis, grading, and spread through air spaces (STAS) have independent prognostic significance. New evidence suggests that these patterns are due to differences in molecular modifications and tumor behavior. For example, STAS is associated with TP53 and Kirsten rat sarcoma (KRAS) mutations, which has similarities to the epithelial– mesenchymal transition and collective cell migration.^[1]

High-throughput genomic studies have shown that mixed-grade adenocarcinomas contain different driver clusters (epidermal growth factor receptor [EGFR], KRAS, Serine/Threonine Kinase 11 or STK11, and Kelch-like ECH-associated protei n 1 o r KEAP1) which define biological subtypes of the tumor as well as treatment response.^[2] The immune microenvironment adds an additional layer of complexity: programmed death-ligand 1 ( PD-L1) expression, tumor mutational burden, and spatial immune contexture affect response to checkpoint blockade.

The technologies of digital pathology and artificial intelligence now present an unprecedented ability to measure these morphological and molecular interrelationships. Deep learning algorithms are able to correlate histologic features that predict either genomic alterations or treatment response directly from digital slides.^[3,4] Such morpho-genomic correlations represent a change of paradigm – pathology as data science, transforming visual morphology to quantifiable biomarkers.

SYSTEMS PATHOLOGY: SYNTHESIS OF CYTOLOGY, HISTOLOGY, AND MOLECULAR DATA

Cytology has become an essential source of material for molecular analysis in NSCLC. Progress in fixation techniques, nucleic acid extraction, and targeted sequencing have brought to the point that even lots aspirates, bronchoalveolar lavages and effusions give reliable results for EGFR, anaplastic lipoma kinase (ALK), c-ros oncogene 1, receptor tyrosine kinase (ROS1), KRAS, and rearranged during transfection (RET) analyses from now on.^[5] Thus, cytology samples can act as a “bridge” from morphology to molecular profile when surgical specimens are not available.

In resectable cases, this integration allows patient stratification for neoadjuvant- or adjuvant treatment. Pivotal studies such as CheckMate 816 and AEGEAN confirmed a benefit of perioperative immunotherapy, but both also highlighted the importance of standard PD-L1 testing and genomic profiling even in small pre-operative specimens.^[6,7] Pathologists are therefore key to achieve tissue adequacy, standardize pre-analytical variables, and validate the molecular information obtained from cytologic preparations.

System pathology goes beyond individual biomarkers and combines cytological, histological, and molecular profiles into coherent diagnostic concepts. Multi-omics profiling with combined morphology, transcriptomes, and immune-histo-cyto phenotyping provides integrative risk profile NSCLC analysis.^[8] There are now computational models connecting cell shapes (morphology) to genomic signatures and to clinical outcome, leading toward predictive diagnostics.^[9,10]

This integration also directly relates to clinical decisions in surgery. For example, identification of STAS-positive adenocarcinomas in the pre-operative setting may affect radicality of resection or use acceptance and commitment therapy approaches. The conversation between surgeon and pathologist is, accordingly, drawing more heavily on data science through molecular proof and digital analytics.

FUTURE CONSIDERATIONS: THE ROAD TO PREDICTIVE AND ACTIONABLE PATHOLOGY

Migration toward the field of systems pathology depends on technological advances and organizational shift. Standardization of cytologic molecular testing, cross-validation of artificial intelligence tools across platforms and secure data transfer are imperative to ensure reproducibility. Merging morpho-molecular information into clinical practice would require close cooperation between pathologists, oncologists, and computational experts.

Novel technologies, such as liquid cytopathology in which circulating tumor DNA is used alongside cytologic examination to further minimize invasiveness and provide a continuous read-out of residual disease or treatment response, are on the horizon. With tissue getting to be ever costly, cytology may emerge as the optimal mediator of biology and clinical utility.

Finally, systems pathology is the natural progression of the pathologist’s role: from observer of morphology to integrator of data, and from diagnostician to an active player in personalized cancer management. In resectable NSCLC, for which tumor biology determines the optimal surgical and systemic approaches, these are no longer optional: molecular heterogeneity is now the cornerstone of precision oncology.

OPERATIONAL ROADMAP FOR SYSTEMS PATHOLOGY

Transitioning from descriptive to systems pathology includes a need for sustained capitulated path of technological innovation along with methodological standardization and inter-disciplinary governance.

First of all, a harmonized infrastructure for digital pathology must be established. Ensuring privacy and confidentiality of patient information and data is paramount, and shared repositories, standardized image formats, and federated learning platforms could make reproducible validation of AI-based diagnostic models possible.

Second, cytology is rarely (if at all) performed to enable molecular profiling, but it should regularly be integrated into diagnostic algorithms to enrich the full range of the spatial and temporal molecular signal. This includes stringent pre-analytical quality standards for sample collection and fixation, validated next-generation sequencing protocols on cytological material, and rapid turnaround times between molecular and morphological data.

Third, the integration of multi-omics – genomic, transcriptomic, and immune signatures that represent the underlying biology, in conjunction with digital morphology – should be the foundation of risk stratification within resectable NSCLC. Interoperable data architectures and bioinformatics pipelines will enable cross-modal correlation and translation to clinical grades in real time.

Finally, the establishment of multidisciplinary learning ecosystems (virtual consortia and translational pathology incubators) will also speed up the cultural transition to predictive and computational pathology. Such initiatives could provide ongoing education to cytopathologists, surgeons, and oncologists, promoting a linkage of biological complexity to clinical decision-making.

Theory-based systems pathology can transgress its theoretical paradigm and fundamentals with clinically-yielding metrics for precision oncology, where cytology vectors from a static diagnosis driver to a functional interface with patient data for patient outcomes through programmable dynamic cohorts driven by bioinformatics.

AVAILABILITY OF DATA AND MATERIALS

Not applicable – this is an editorial and does not include new data.

ABBREVIATIONS

STK11: Serine/threonine kinase 11

KEAP1: Kelch-like ECH-associated protein 1

PD-L1: Programmed death-ligand 1

ALK: Anaplastic lymphoma kinase

ROS1: c-ros oncogene 1, receptor tyrosine kinase

RET: Rearranged during transfection