Comprehensive molecular characterization of cutaneous squamous cell carcinoma reveals determinants of metastatic progression

Cutaneous squamous cell carcinoma (cSCC) is the second most common form of cancer worldwide. While most cases are localized and not life-threatening, a small but critical subset—roughly 2% to 5% of patients—develop metastases (the spread of cancer to distant organs). For decades, clinicians have relied on standardized staging systems, such as the AJCC or the BWH models, to decide who needs aggressive surgery and who only requires monitoring.

However, these traditional methods fall short because they rely heavily on the physical characteristics of the tumor once it has been fully removed. This creates a fundamental problem in clinical timing. By the time a pathologist can accurately stage a tumor based on its depth and appearance in an excised specimen, the window for preventative, pre-operative decision-making has often closed. We lack a way to look at a small initial biopsy and say with certainty whether that specific tumor possesses the molecular machinery required to spread.

The limitations of clinical staging

Current prognostic frameworks for cSCC are primarily clinico-pathological. They assess risk by looking at the tumor's physical dimensions, its location, and its histological differentiation (how closely the cancer cells resemble healthy skin cells). As noted in the study, these systems are inherently reactive. They require a complete pathological assessment of the excised tumor. This renders them unsuitable for use with the small biopsy samples typically taken during initial diagnosis.

Furthermore, these methods suffer from observer variability. Different pathologists may interpret the "aggressiveness" of a tumor's appearance differently. This leads to inconsistent care across medical centers. Because cSCC is often perceived as a low-risk cancer due to its low overall metastatic rate, it has historically been understudied in large-scale molecular atlases. This has left a significant gap in our ability to identify the biological drivers that push a tumor from a localized lesion to a systemic, metastatic disease.

Mapping the molecular architecture of progression

To fill this gap, the researchers assembled the D-SQUAME study. This is a massive nationwide cohort of 19,120 patients. They employed a nested case-control (NCC) design. This design allowed them to efficiently compare a balanced number of tumors that progressed to metastasis (cases) against those that did not (controls) [Figure 1A]. This design is mathematically potent. It allows researchers to extract meaningful signals from rare events without needing to sequence the entire population.

The study's approach to identifying the determinants of metastasis followed a multi-layered hierarchy:

1. Genomic Profiling: The team performed whole-exome sequencing (WES; a method to sequence all protein-coding regions of DNA) on 147 tumors. They mapped the "driver" mutations—the specific genetic alterations that provide a selective advantage to cancer cells. They found a median mutation burden of 44 mutations per megabase [Figure 1B]. This means for every million units of DNA, there are 44 mutations. This burden was dominated by UV radiation signatures.
2. Transcriptomic Clustering: Using RNA-sequencing on 378 tumors, they looked at gene expression (the activity levels of genes). They discovered that tumors naturally cluster into three distinct states: Differentiated, Basal-like, and Mesenchymal-like [Figure 4A].
3. Integrative Modeling: The researchers used a regularized Cox regression model (a statistical method for analyzing time-to-event data) with "late feature integration." They analyzed both the genetic mutations and the expression patterns separately. They then combined their predictive strengths into a single mathematical engine.

The goal was to move beyond mere observation. They sought a predictive tool that could function on the limited material available in a standard biopsy.

Decoding the drivers of metastatic competence

The results reveal that metastasis is not a random event. It is tied to specific biological programs. At the genomic level, the authors identified 38 driver genes. They found that metastatic progression was significantly associated with activating mutations in the RAS signaling pathway. It was also associated with loss-of-function alterations in the SWI/SNF chromatin remodeling complex (proteins that manage how DNA is packaged). These associations were clearly visible in the data [Figure 2A, Figure 3D].

Interestingly, the median cSCC is driven by a high number of "hits." The authors report a median of 9 hits across 6 genes. This is more complex than the 4–5 hits typically seen in other cancers.

The most striking finding lies in the transcriptomic landscape. The authors report that the "Mesenchymal-like" cluster is strongly linked to the shortest metastasis-free survival [Figure 4B]. These tumors exhibit poorly differentiated programs and high metabolic activity, known as the Warburg effect. Conversely, "Differentiated" tumors show much more favorable outcomes. These tumors maintain programs typical of mature skin cells.

By distilling these complex signals, the researchers developed the SCCore-GEP. This is a 23-gene expression signature. This signature acts as a molecular compass. It points toward high-risk tumors by measuring the balance between epidermal differentiation and stemness-related migration. When validated in independent cohorts, the SCCore-GEP achieved a weighted C-index of 0.83 [Figure 5B]. The C-index measures how well a model ranks risks, where 1.0 is a perfect prediction. This result represents a significant leap in predictive accuracy over current standards.

Assessing the boundaries of the signature

While the SCCore-GEP represents a major advancement, several caveats remain. First, the discovery cohort utilized formalin-fixed paraffin-embedded (FFPE; a method of preserving tissue in wax) blocks. These blocks were more than 10 years old. While the researchers applied rigorous quality control, the age of the samples can introduce technical noise in RNA integrity.

Second, the study notes that cSCC often has low "neoplastic cellularity." This means the tumor cells are frequently diluted by a dense infiltration of immune cells and reactive tissue. This dilution makes it harder to isolate the pure signal of the cancer itself. This is a persistent hurdle in skin cancer genomics.

Finally, the study's temporal context is worth noting. The discovery cohort was collected between 2007 and 2009. The molecular profiles captured may not account for the selective pressures introduced by modern immunotherapies. As the clinical landscape shifts, the relationship between tumor signatures and metastatic outcome may evolve.

The verdict: A new standard for risk stratification

The evidence strongly supports the implementation of molecular risk stratification in cSCC. The SCCore-GEP is not merely a marginal improvement. It is a fundamentally different kind of tool. Unlike clinical staging, which requires the tumor to be removed and fully dissected, this 23-gene signature can be deployed directly on biopsy material.

This capability allows for "upstaging" patients. This means identifying individuals in traditionally low-risk categories who actually harbor high-risk molecular profiles. The authors demonstrate that this signature can capture up to 75% of metastatic events within these low-risk groups [Figure 5D]. For practitioners, this means the difference between a conservative follow-up and a life-saving, aggressive surgical intervention. The SCCore-GEP is ready for clinical translation.

Figures from the paper