The humble H&E stain for discovering melanoma prognostic features: pathetically inadequate in the molecular era or a wastefully unexplored treasure trove?
Can histopathologists be drivers of biomarker discovery?
In melanoma care there are critical points where biomarkers might enhance care (Deacon, Frontiers in Med, 2021): determine whether a pigmented lesion requires biopsy; improve histological distinction between melanoma and naevus; decide whether the patient should have a sentinel node biopsy and/or to receive adjuvant treatment; and to decide what type systemic treatment to give and determine likelihood of response
As a histopathologist I spend hours per day evaluating H&E stained sections and I am fascinated by the idea that such a simple stain could be used to identify novel histological patterns that might be the basis for new and powerful morphological biomarkers. But in the age of molecular pathology I wonder if this idea is merely quaint. I believe that this stain is a treasure trove of untested morphological patterns because it is a visual snapshot that integrates everything that is occurring up to and including the moment of biopsy, as shaped by the functioning of the genome, transcriptome and proteome. Most excitingly, histopathologists are best placed to mine this visual information. Assuming there really is something to be gained from H&E based biomarkers, what sort of questions should we be addressing?
Histopathologists are particularly invested in finding ways to distinguish naevi from melanomas. We are all too aware that there is a substantial minority of melanocytic lesions that are histologically ambiguous. A major problem is the lack of an ideal gold standard for malignancy. One can easily create a study sample of unequivocal naevi and melanomas, but no one needs a biomarker for these. We need to separate histologically ambiguous tumours into correct biological categories, which by definition have no histological gold standard. Metastasis as the gold standard is not sensitive for malignant biology. For example, a 65 year old with a 5.0 mm Clark level V, ulcerated SSMM with 4 mitoses per mm2 would have only a 43% risk of being sentinel node positive (see https://www.melanomarisk.org.au/SNLForm). In a study of disease free survival with 700 participants (von Schuckmann, JAMA Dermatology 2019) that included IB to IIC disease, 94 (13.4%) developed metastasis within 2 years. Even among the 49 patients with T4b, only 16 (32.7%) developed metastasis. Thus, metastasis is not present in the majority of melanoma patients with frankly malignant histology. This means that most biologically malignant tumours with ambiguous histology might have no metastasis. Instead of metastasis, another possible gold standard is a molecular feature such as genomic instability. The lack of a readily available gold standard means that this is a very challenging area. Personally I have avoided conducting studies of diagnostic biomarkers on ambiguous tumours for two reasons: (i) the difficult issue of defining a true gold standard for malignant biology and (ii) the difficulty gathering a sufficient sample size to make compelling generalisations.
On the other hand, the H&E stain could play a central role in prognostic biomarker discovery, such as features that predict sentinel node status or stratify high risk SLNB negative patients for adjuvant treatment. The National Comprehensive Cancer Network sets a threshold of 5% SLNB+ risk for considering sentinel node biopsy. This means that all melanomas that are T2 are eligible while those T1 tumours greater than 0.8 mm are potentially eligible with additional risk factors. At present, approximately 80% of all sentinel node biopsies are negative (Morton NEJM 2006). Better stratification of risk would potentially reduce the number needed to biopsy to find the minority that will be SLNB positive. H&E biomarkers stand a reasonable chance of success in this discovery setting because clear gold standard outcomes exist for prognostic features. Notably, a biomarker that only requires an H&E stain will be extremely cheap, simple and can be reported at the same time as all the other H&E-based features such as BT, ulcer, etc. This greatly lowers the barrier to translation. One difficult issue is that a new prognostic biomarker needs to be used alongside or as part of TNM staging. Yet, in order to be considered by a committee like AJCC there needs to be a large archive of pathology reports that include the new feature for statistical evaluation. This is only likely to occur if such a feature is recommended or ideally mandated by national guidelines. In the UK this would be the RCPath dataset for histological reporting and clinical guidelines from the BAD.
In melanoma, BT, Clark level, ulceration, microscopic satellitosis, mitotic rate, tumour regression, tumour infiltrating lymphocytes and growth phase were identified and developed by trailblazers in the field of melanoma pathology from the 1960s onwards. But maybe the golden age of H&E biomarker discovery has passed and there is nothing new to discover. My impression is that H&E discovery is now considered redundant in the face of molecular features. But in fact there are currently no widely used molecular biomarkers for melanoma prognosis. One of the better known is DecisionDx, a gene expression panel that stratifies melanomas into one of two risk groups. However, this has been deemed to require more evidence before widespread routine use (Grossman, JAMA Dermatol 2020). So the field remains wide open for new features. I am inspired by icons of melanoma practice and research to hunt for novel morphological prognostic patterns. But what sort of patterns can we still discover?
What novel features could be investigated that require only H&E stain
Candidate patterns are only limited by one’s imagination but the ones that occur to me are those that advance or improve existing features, histological correlates of molecular alterations or entirely novel patterns. Studies of H&E based prognostic patterns, beyond established features, are depressingly sparse. Recent investigations include the assessment of novel patterns of TIL infiltration that are more nuanced than the classic absent, non-brisk and brisk patterns of Clark. For example, here in Leicester we investigated a novel way to score TILs in primary melanoma (Saldanha, Am J Surg Pathol 2017). Some investigators have looked at improving BT by trying to combine it with macroscopic measurements to estimate tumour volume (e.g. Voss, Dermatology 2014). Histological and clinical features have been correlated with the underlying molecular progression pathways and this conceptual framework is used in the 4th edition of the WHO skin tumour book. This has spawned the idea of the so called BRAFoma (Viros, PLoS Med 2008). My own work looked at enhancing BT, first by looking at the local density of cells at the position of BT measurement, which we called “Breslow Density”, then combing this with BT into a “Targeted Burden Score” (Saldanha, Am J Surg Pathol 2018) and finally by looking at the estimated area of invasive cells on the same slide where BT was measured (Saldanha, JAMA Dermatol 2019). In addition, my colleagues and I showed that invasive width was a stronger prognostic feature than BT (Saldanha, Am J Surg Pathol 2020)
Barriers
The concept of looking at novel histological features is simple enough but there are substantial barriers to doing this in practice, which include ethics committee and research governance approval, accessing and assembling a suitable retrospective cohort, attaching follow up data and getting statistical support. Finally, the pathway from idea to validation for a new biomarker is long and difficult. I am yet to complete this journey for any feature I have investigated.
So can Histopathologists play a part in melanoma basic research?
I think the role for most histopathologists is to be alert to the possibility of novel histological patterns during daily practice and then consider preliminary testing of any hypotheses on a small set of samples in a case-control format. This would lead to preliminary proof of concept for a promising idea. As an example, none of my own studies cited above had any major funding and studies got underway through the combined effort from myself, a consultant colleague (Dr Mark Bamford), keen senior trainees and medical students. I appreciate that many histopathologists are far too hard pressed to take up research activities but I hope this blog article at least lays out the possibilities and pitfalls and provides food for thought about how dermatopathologists might remain at the forefront of wringing every last drop of opportunity from the humble H&E stain.
Biography
Professor Gerald Saldanha
MB ChB, PhD
I am a consultant histopathologist at the University Hospitals of Leicester NHS Trust and Honorary Professor in the University of Leicester. My Diagnostic role is in dermatopathology. In research, after completing a PhD I focused on melanoma molecular biomarkers that can be assessed in formalin-fixed paraffin-embedded tissue. I have now turned my attention towards histological prognostic features for melanoma that can be identified using only a H&E stain. I still hold the ever vanishing hope that one day I will be able to balance my diagnostic and research roles effectively!