Keywords

Introduction

The tumour microenvironment may be a suitable site for biomarker expression that relies on signalled changes from neoplastic cells. Complex bi-directional interactions occur and involve cell-intrinsic and cell-extrinsic mechanisms. Most attention has been focussed on the reciprocal signalling between cancer stem cells and tumour infiltrating immune cells, but other components of the tumour microenvironment play important roles in tumour initiation and progression. These include endothelial and pericytic cells, stromal fibroblasts, extracellular matrix macromolecules and dendritic cells [1]. Signalling molecules released from neoplastic cells can induce changes in adjacent tissue compartments, resulting in tissue changes than may be exploited for developing novel biomarkers.

Biomarkers

Biomarkers are now an essential part of clinical practice. In pathology, molecular testing is increasingly used routinely alongside immunohistochemistry for diagnosis [2]. Such testing ranges from single molecular markers detected by fluorescent or chromogenic in situ hybridisation through to whole genome sequencing, which currently can be completed within a two-week turnaround in the United Kingdom. In addition to using biomarkers for diagnosis, predictive biomarkers for drug response form a substantial part of the pathologist’s workload in a cancer centre. Both diagnostic and predictive biomarkers must be validated and ideally accredited before they can be used outside a clinical trial setting.

Prognostic biomarkers are also an expanding field and levels of evidence for their use vary. Many prognostic biomarkers are available commercially that are supported by published evidence but which have not been adopted into clinical guidelines for routine use. Often the reason for this is that prognostic biomarkers are typically developed in the laboratory using retrospective samples from clinical trials or biobanks. Testing in a prospective setting is normally required by regulators and health economic modelling must also be undertaken to demonstrate benefit by change of clinical practice. For example, if a biomarker can be used to ‘rule out’ disease progression in a cancer, then unnecessary follow up and expensive investigations can be avoided, with significant health cost savings. Even more importantly perhaps, patients can be relieved of anxiety and there are wider societal impacts as well as reduction in time off work, which may be impossible to include in health economic modelling. Developers of prognostic biomarkers can face the challenge that expensive prospective clinical trial data is expected by clinicians, regulators and the expert groups that formulate guidelines [3]. Real world data cannot be obtained until the biomarker is introduced and used for a considerable period. For these reasons, some prognostic biomarkers are offered commercially for years before achieving accreditation for routine clinical use.

The need for reliable prognostic biomarkers is exemplified by cutaneous melanoma, where in the American Joint Committee on Cancer (AJCC) stage I and II disease, only 20% of patients will undergo progression but all patients must be managed and followed up in the same way. In a recent Delphic survey of melanoma experts, thresholds for clinical follow up, cross sectional imaging and adjuvant therapy were determined, and are an early step towards individualising management recommendations based on risk [4]. Validated prognostic biomarkers have the potential to offer personalised management based on better knowledge of actual risk.

There is good evidence that patients are increasingly seeking information about their cancers and are keen to find prognostic tests that can contribute to a fuller understanding of their individual risk [5]. In a recent survey presented by Miley L-B et al., at the 2022 Fall Clinical PA & NP Conference, over 90% of patients with cutaneous melanoma surveyed wanted prognostic information about their melanoma at the time of diagnosis. Over 75% wanted to increase their knowledge and over 45% wanted a test that could inform treatment decisions. Provision of good quality information and guidance by the clinical team is key as patients may access unreliable sources by web based searching. It is important that the possible benefits and harms of taking a personalised risk test using a prognostic biomarker are explained and reference to the source scientific data may be required to provide a balanced view.

Several prognostic markers based on gene expression signature (GES) for cutaneous melanoma are commercially available [3]. The signatures include genes from the melanoma and its microenvironment. The biological mechanisms that underpin GES tests are often not known. It is likely that the different levels of gene expression detected rely on cross-talk between the melanoma cells, immune cells and stromal cells. An emerging prognostic biomarker for cutaneous melanoma (AMBLor) that relies on detecting changes in the microenvironment is described below. The test is based on finding changes in autophagy regulation expression in the epidermis overlying cutaneous melanoma.

Autophagy

Autophagy is a fundamental cellular process that eliminates molecules and subcellular elements, including nucleic acids, proteins, lipids and organelles, through lysosome-mediated degradation to promote cellular homeostasis, differentiation, development and survival. The discovery of selective autophagy receptors demonstrated that autophagy is a highly selective cellular clearance pathway regulated by bi-directional cellular cross talk. AMBRA1, (autophagy/Beclin-1 regulator 1), is a key activating molecule in Beclin-1-regulated autophagy. It is a highly-conserved adapter protein that plays multiple roles in the autophagy signalling network [6].

In the normal epidermis, immunohistochemistry has shown that AMBRA-1 is expressed in the cytoplasm of the keratinocytes. Expression is weakest in the basal layer and there is a gradient of increasing intensity through the prickle cell layer to the granular layer, where expression ends abruptly (Fig. 1.1). As the epidermis renews, daughter basal stem cells undergo amplification and then enter on a pathway of terminal differentiation. Autophagy appears to be essential to normal epidermal differentiation and as the terminally differentiating keratinocytes move away from the nutritional supply of the dermal stroma, recycling of their cytoplasmic components presumably becomes essential to normal maturation.

Fig. 1.1
An image depicts the immunohistochemistry expression of A M B R A 1 in normal epidermis.

Immunohistochemistry expression of AMBRA1 in normal epidermis. The protein is expressed in the cytoplasm only and shows a gradient with weakest staining in the basal layer increasing in intensity to the stratum corneum where staining ends abruptly

Development of AMBRA-1 as a Prognostic Biomarker

In a study of early stage cutaneous melanoma, it was found that AMBRA1 expression detected by immunohistochemistry in melanoma cells did not correlate with clinical outcomes. However, when the overlying epidermis in early stage melanomas was considered, it was observed that loss or reduction of AMBRA1 expression was frequently present. Retention of AMBRA-1 correlated with lack of disease progression, providing a potential biomarker of prognosis based on signalled changes occurring in the tumour microenvironment. Combining AMBRA1 with a second immunohistochemical marker, Loricrin (Fig. 1.2) to examine epidermis in AJCC Stage I melanomas resulted in an effective prognostic biomarker test [7]. In order to develop these observations into a prognostic biomarker for clinical use, a further multicentre study has recently been undertaken in a mixed cohort of 334 AJCC Stage I and 77 Stage II cutaneous melanomas from Roswell Park Cancer Centre, Buffalo USA (n = 241) and the Peter McCallum Cancer Centre, Melbourne, Australia (n = 170). Clinical follow up ranged from 60 to 287 months in these retrospective cohorts. Each cohort was powered to represent rates of metastasis of 10% for AJCC Stage I or up to 20% for Stage II disease. Results showed that a positive combined AMBRA1 and Loricrin test (AMBLor) with maintenance of either or both proteins, was associated with significantly increased disease-free survival of 97% compared to 87% for patients in which expression of both was lost (P = 0.01, 95% CI 0.9–0.42), and a negative predictive value of 97.14% (Fig. 1.3). The analysis was performed using newly created and validated humanised antibodies to AMBRA1 and Loricrin to ensure that consistent and quality controlled reagents would be available for future use. The antibodies can be used on the Ventana and Bond platforms which are the most widely used in pathology laboratories worldwide.

Fig. 1.2
An image of Loricrin is expressed in the stratum corneum as a continuous band in normal skin.

Loricrin is expressed in the stratum corneum as a continuous band in normal skin. Fine granular cytoplasmic is seen and nuclei can be labelled. Single cell gaps may be present

Fig. 1.3
A graph depicts the result of an interim analysis of A J C C stage one and two non-ulcerated cutaneous melanomas in cohorts from Melbourne and Buffalo.

Interim analysis of AJCC stage I and II non-ulcerated cutaneous melanomas (n = 411) in cohorts from Melbourne and Buffalo. There were 70 cases in the low risk group and 341 cases lost expression of both AMBRA1 and Loricrin, leaving their AJCC risk unchanged. Only two patients progressed in the low risk group after a minimum of five years of follow up. A negative predictive value of 97.14% for progression prediction was found. In the group maintaining one or both markers, 44 cases progressed out of 341 studied. Analysis of additional cohorts using a blinded prospective-retrospective study design is ongoing

The AMBLor test can only be performed on non-ulcerated Stage I and II cutaneous melanomas, that have been removed with a small margin of normal surrounding skin. The marginal skin serves as an excellent positive control and the pathologist interprets the test by comparison of the protein expression in the epidermis overlying the melanoma with that in the marginal skin. Additional negative and positive normal skin batch controls are also used. A training programme has been developed for pathologists and an interpretation guide is available.

Further larger validation cohorts using a ‘prospective-retrospective’ study design in which the biomedical scientists and pathologists are blinded to the clinical outcomes are currently being evaluated. If successful, UKCA and CE marking for the antibody test will be sought.

AMBLor has successfully passed the second stage of the National Institute for Clinical Excellence (NICE) accreditation process in the United Kingdom through a MedTech Innovation Briefing [8]. A prospective clinical trial to ascertain whether having a low risk AMBLor result would change clinical decision making and what the impact of a positive or negative test result would be on patient anxiety is about to be launched.

Sentinel Node Biopsy in Melanoma

The use of sentinel lymph node biopsy (SLNB) in melanoma is controversial. Current guidelines such as the National Comprehensive Cancer Network (NCCN) recommendations do not recommend SLNB for patients where the risk of metastatic disease is less than 5%, unless there is significant uncertainty about the local staging [9]. In a recent addition to these guidelines, gene expression profiling (GEP) is included but should not guide clinical decision making in this group. In higher risk groups a GEP prognostic test may be used to inform the likelihood of a positive SLNB on an individual basis. The ability of AMBLor to predict SNLB status is not known but will be tested in a prospective clinical trial. Such a biomarker could help to inform clinical decision making. It should be remembered that SLNB has a morbidity and identification of cases where the procedure is inappropriate would benefit a subset of patients.

Cross-talk Mechanism in Melanoma

Some insight into the mechanisms by which melanoma is able to deregulate autophagy in the epidermis has been gained by the discovery of a paracrine mechanism mediated through the Transforming Growth Factor beta (TGFβ) pathway [10]. Using semi-quantitative immunohistochemistry, it was demonstrated that increased TGFβ2 in the melanoma cells was associated with loss or significant reduction of AMBRA1 in the epidermis overlying the melanoma and with ulceration. Further, TGFβ2 treatment of keratinocytes in culture resulted in downregulation of AMBRA1, which is followed by downregulation of loricrin and claudin-1. It can be speculated that the TGFβ2 paracrine mechanism is the cause of spontaneous ulceration in melanoma, which is known to be an adverse prognostic feature, as reflected in AJCC staging. Loss or reduction of AMBRA1 expression in the epidermis overlying melanoma is interpreted in the AMBLor test by comparison with the normal epidermis at the margin (Fig. 1.4).

Fig. 1.4
A pair of images of an immunohistochemistry stain for A M B R A 1 demonstrating cellular interactions in melanoma Keratinocytes near melanoma cells express less A M B R A 1.

Immunohistochemistry for AMBRA1 showing cellular interactions in melanoma. The keratinocytes adjacent to melanoma cells show reduced expression of AMBRA1 a compared to expression in the normal skin in the surrounding margin b where cytoplasmic expression is intense. This is mediated through the TGFβ2 pathway

Autophagy and Squamous Cell Carcinoma of the Head and Neck

Oropharyngeal Squamous Carcinoma

Although oropharyngeal squamous cell carcinomas (OPSCC) that actively transcribe high-risk human papillomavirus (HPV) have a more favourable prognosis that their HPV-negative counterparts, the mechanism remains undefined. In vitro studies have shown that HPV-positive OPSCC cells exhibit reduced macroautophagy/autophagy activity, mediated by the ability of HPV-E7 to interact with AMBRA1, to compete with its binding to BECN1 and to trigger its calpain-dependent degradation [11]. Further, pharmacological inhibition of autophagy and downregulation of AMBRA1 have been shown to sensitize HPV-negative OPSCC cells to the cytotoxic effects of cisplatin. Immunohistochemical analysis using a tissue microarray (TMA) showed that AMBRA1 expression appears reduced in HPV-positive compared to HPV-negative OPSCCs [11]. The data suggest that AMBRA1 may be a key target of HPV resulting in impairment of autophagy. This leads to the proposition that targeting of autophagy could be a possible therapeutic strategy for improving the response of HPV-negative OPSCC to chemotherapy [11].

Whole genome sequencing (WGS) of both the viral and somatic genomes in HPV positive OPSCC has revealed a complex picture, in which HPV may play different roles in different tumours. Although WGS appears to reveal subgroups within HPV positive OPSCCs, the patterns are currently too complex to translate WGS into a biomarker for clinical use. The finding that AMBRA-1 expression is downregulated in HPV positive OPSCC, raises the possibility that it may be a useful prognostic marker. In our larger (unpublished) cohort study based on whole sections of OPSCC evaluated by p16 immunohistochemistry and HPV in situ hybridisation, no clear prognostic pattern emerged, however. Currently, re-evaluation of the cohort for changes in stromal and endothelial AMBRA1 expression using AI is underway.

Cutaneous Squamous Cell Carcinoma

Currently there are no validated prognostic biomarkers in routine clinical use for prediction of metastatic behaviour of cutaneous squamous cell carcinoma (cSCC). Staging systems are of limited clinical utility with regard to identification of primary squamous cell carcinomas that have metastatic potential. The staging system proposed by the Brigham and Women’s has great utility [12] but biomarkers that could identify low risk primary cSCC with a very high negative predictive value could relieve patient anxiety and save considerable health service resource.

The evidence for a role in autophagy in cSCC is mostly from laboratory studies and the data suggest that autophagy may have a protective effect, allowing malignant keratinocytes to escape apoptosis [13]. Our initial studies examining AMBRA1 in a cohort of cSCC by semi-quantitative immunohistochemistry have shown a trend towards expression in the neoplastic keratinocytes associating with adverse outcome, but statistical significance has not been achieved (data unpublished).

A promising approach is to utilise Artificial Intelligence (AI) in combination with AMBRA1 expression and this has shown improved prognostic power in our preliminary study. In a recent study of whole slide images, AI has shown the ability to distinguish between rapidly metastatic and non-metastatic primary cSCC with an area under receiver operator curve (AUROC) of 0.747 [14]. Combining the AI with known adverse factors in a risk model increased the AUROC to 0.917. The risk factor model with AI predicted high 5-year disease specific survival (DSS) for patients with cSCC with 0 or 1 RFs (100 and 95.7%) and poor DSS for patients with cSCCs with 2 or 3 RFs (41.7 and 40.0%). Perhaps the most intriguing finding is that the AI system appears to recognise morphological features in routine sections of primary cSCC that are associated with metastasis that pathologists do not identify. Whether these features are in the neoplastic keratinocytes or the tumour micro-environment is unknown.

Conclusion

The biological behaviour of neoplasms depends not only on the properties of the neoplastic cells but on their interactions with cells in the microenvironment. The development of the AMBLor test is an example of the way that detecting changes in the tumour microenvironment can be exploited to predict prognosis in a cancer. In the future, it seems likely that biomarker tests will be incorporated into risk models that combine AI and clinical features to provide personalised management plans.