Elucidating the role of the brain extracellular matrix in promoting an invasive cancer phenotype and developing tools for dual-modality in vivo imaging

Student thesis: Doctoral ThesisDoctor of Philosophy


Cancer metastasis, due to aberrant cell migration, dissemination and colonisation at secondary sites, accounts for more than 90% of cancer-related deaths. This is particularly poignant in the context of brain metastasis,which is the most common cause of CNS tumours and often results in a poor patient prognosis. Brain metastases predominantly arise from primary lung, breast and melanoma tumours. To date, brain metastasis research has centred around understanding the molecular determinants that drive brain metastasis and how interactions with brain-resident cells drive or promote CNS metastasis. In terms of evaluating the role of the non-cellular component of the brain, in vitro studies have primarily utilised HA-based hydrogels supplemented with brain ECM components such as laminin, fibronectin and collagen and biomimetic materials, to demonstrate how matrix composition alters cell behaviour and phenotype. In this context, the laminin and fibronectin content of HA hydrogels has been shown to alter drug efficacy in in vitro melanoma brain metastasis models in which cell migration and adhesion were evaluated as functions of invasiveness and metastatic potential. However, many existing in vitro models focus on laminin and fibronectin in the ECM but do not evaluate the contribution from hyaluronic acid (HA), despite it being the major component of the brain ECM. Furthermore, studies that have focused on the role of HA in the brain ECM have primarily done so using a mechanobiological approach, focusing on the mechanical effects of matrix stiffness on cell migration, spreading, adhesion and invasiveness or on the mechanics of cell stiffness adaptation to different ECM compositions .In this context, a previous study by Staunton et al. demonstrated that, although there was no clear link between malignancy, ECM composition and mechanical phenotype, breast cancer cells were able to mechanically adapt, via changes in actomyosin contractility, to different ECM compositions such as HA and laminin. This project builds upon these initial observations of cell adaptation and cytoskeletal remodelling in response to the brain ECM, by exploring how HA modulates cancer cell phenotype and behaviour from the perspective of cell invasiveness, with a focus of mode of individual cell migration and matrix adhesion.

In this study, collagen I-only and HA-supplemented collagen I matrices were utilised to recapitulate the ECM at the invasive front of primary tumours and the ECM in the brain, respectively. This simplified model allowed the contribution of the HA component of the brain, in conferring and promoting an invasive phenotype, to be evaluated. Melanoma and breast cancer cells with differing metastatic biology were evaluated for mode of migration and levels of actomyosin contractility on these matrices, as it has previously been demonstrated in vitro on a thick collagen matrix and in vivo, that actomyosin contractility plays a vital role in individual cell migration and metastasis. Furthermore, cells with high levels of actomyosin contractility via pMLC2 have previously been shown to be enriched at the invasive front of melanoma and liver tumours.

Data in this study demonstrate that the invasive phenotype of rounded-amoeboid melanoma and breast cancer cells observed on collagen, was not altered on a HA-supplemented collagen matrix, with high levels of actomyosin contractility via pMLC2 and membrane blebbing. Moreover, cells with a rounded-amoeboid phenotype had an increased cell adhesion rate and/or overall adhesion when seeded on collagen with HA, versus collagen only.In some heterogeneous populations of cells, HA promoted a mesenchymal-to-amoeboid transition, resulting in an increased number of cells with a rounded-amoeboid phenotype, although adhesion to this matrix was not significantly altered compared to adhesion to a collagen-only matrix. To further understand matrix adhesion in these cell lines, CD44 and Ezrin/Radixin/Moesin (ERM) expression was also evaluated, as CD44 is the main HA receptor which interacts with actin cytoskeleton adaptor proteins, via ERM proteins, to transduce extracellular signals and promote cell migration. Although no correlation was observed between cell invasiveness, metastatic potential and CD44/ERM expression or activation in these cell lines, CD44 expression and distribution was altered when cells were seeded on the different matrices, supporting the role of CD44 as a matrix-sensing molecule.

To expand on these initial observations, a further aim of this project was to evaluate the invasive phenotype and metastatic potential of these breast cancer and melanoma cells in an in vivo context. The ability to track metastasising cells in in vivo metastasis models, using dual-modality imaging, is advantageous in this context, as it provides tools for evaluating metastatic dissemination to specific niches in vivo whilst also providing experimental endpoints for pre-clinical studies, by allowing confirmation that metastases are present in different tissues. This, in turn, reduces the need for large cohorts of animals to study metastatic progression over time.

To address this, this project describes the successful development of fluorescent protein and SSTR2 expressing reporter gene constructs for in vivo dual-modality imaging, alongside the establishment of pre-clinical in vivo metastasis models, which provide the initial tools for tracking metastasising cells in vivo.This reporter gene approach is particularly advantageous for evaluating metastatic spread to the brain given the low endogenous expression of SSTR2 in upper body regions, unlike existing strategies such as hNIS and D2R imaging. A2058 and 5555 melanoma cell lines, characterised in vitro for an invasive phenotype, and MDA-MB-231.BrM2 breast cancer cells, which have a known tropism for the brain, were transduced with a TagRFP/SSTR2 reporter gene construct and characterised for reporter gene expression, localisation and function in vitro. Subsequent in vivo validation demonstrated their utility for 10PET/CT imaging using 68Ga-DOTA-TATE, with PET signal confirmed to be specific for SSTR2 via blocking studies. Alongside, development of in vivo brain metastasis models involved the intracardiac (i/c) injection of GFP expressing A2058, 5555 and MDA-MB-231.BrM2 cells, which resulted in brain metastasis in each instance as well as metastasis to additional extra-cranial sites including the lungs, liver and kidney. Combining reporter gene imaging with these in vivo metastasis models demonstrated their applicability as imaging tools for cancer metastasis, as brain, lung and liver metastases were observed following i/c injection of the TagRFP/SSTR2 expressing variants of these cells. Future work would therefore focus on imaging TagRFP/SSTR2 melanoma and breast cancer metastases in the brain using 68Ga-DOTA-TATE PET/CT, as the final stage of model validation.

The initial characterisation of the reporter gene expressing cells and in vivo brain metastasis models as pre-clinical tools also provided a range of tissue for ex vivo characterisation of an invasive cellular phenotype and, validation of in vitro data. Immunohistochemical analysis showed that cells at the invasive front of primary A2058.TagRFP/SSTR2, 5555.TagRFP/SSTR2 and MDA-MB-231.BrM2.TagRFP/SSTR2 tumours were enriched for pMLC2 versus cells at the tumour body, with moderate or high levels of pMLC2 also observed in MDA-MB-231.BrM2 metastases in the lungs, liver, lymph nodes, kidneys, ovaries and brain, following intracardiac injection of cells. Moreover, although a clear link was not observed between cell phenotype and CD44 expression in vitro, cells at the invasive front of primary tumours were enriched for CD44, compared to the tumour body. High levels of CD44 were also observed in metastases in the brain and at a number of extra-cranial sites. Evaluation of the pattern of metastatic colonisation at different secondary sites showed a large proportion of individual cells in the brain compared to in the lungs and liver, in which metastases predominantly appeared as larger clusters. These individual cells were shown to have high levels of pMLC2 and CD44 with a rounded-amoeboid morphology, thus supporting the hypothesis that the brain ECM supports an invasive phenotype in breast cancer and melanoma cells. This characterisation also supports the initial in vitro work from this project which suggests that HA in the ECM may have an important role in conferring this phenotype. Overall, this in vitro and in vivo data supports the rationale for targeting the amoeboid cancer cell and/or its interactions with HA based matrices to block metastatic dissemination.
Date of Award1 Nov 2020
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorGilbert Fruhwirth (Supervisor), Vicky Sanz Moreno (Supervisor) & Philip Blower (Supervisor)

Cite this