Youssef Yakkioui

In this thesis, we present our focus of biomarker research in skull base chordomas. Biomarkers are defined as biological substances or characteristics, either qualitative or quantitative, that outlines a certain (patho)physiological state and is often used as a surrogate to guide clinicians to indicate disease severity, treatment response and prognosis. Here we summarize the most important findings of the chapters.

The first manuscript in chapter 2, presents a of the entity called chordoma. As many important historical and recent findings shaped our current thoughts and ideas of chordoma, we accentuate certain discoveries that are a mainstay in how we view the pathophysiological state in chordomas today. First, we admire the human embryology by describing the formation of the notochord, an embryonic elongated rod of cells that (amongst others) guides the formation of the vertebral axis and skull base by secretion of important signaling factors, the vertebral axis and the skull base. To illustrate the beliefs that form our current concept of chordomagenesis, the etiology of chordomas, we highlight the historical development of several paradigms and what we now consider the physiological foundation. By displaying a hypothesized model of these mechanisms, we inform the reader of the possible scenarios for chordomas to arise. The molecular groundwork responsible for tumor development and subsequent tumor behavior, has been very diverse over the past two decades, and primarily performed by single center studies. Here, we discuss (epi-)genetic findings, with the predominantly observation of chromosomal instability in chordoma cells and a proposed single crisis event that is advocated to precede it, called chromothripsis. This genetic backbone of the disease is very heterogeneous, possibly clarifying the array of aberrant molecular pathways in chordoma cells. Brachyury, as a relatively unique hallmark and more stable signaling factor in chordomas, is discovered in 2006 and is classified as a notochordal factor with associations to several pathological mechanisms of the disease. This transcription factor, holds much promise as a biomarker and potential target for therapy. With the discovery of notochordal marker brachyury, the authors argue that there might be potential in further investigating of other notochordal transcription factors. The paucity of studies performed on the level of epigenetics and other unexplored research fields is also touched on.

In chapter 3, the authors support the notion of the notochord being a crucial tissue for chordoma research. In detail, we describe a process by which we isolate notochordal tissue from the human embryonic intervertebral discs. Tissue collected from aborted fetuses with the gestational age of 9, 11, and 13 weeks were collected. After dissection, the spinal column was fixed in 3.7% formalin for 48 hours before being embedded in paraffin by standard procedure. Immunohistochemically stained tissue of the fetal vertebral spine for brachyury expression, resulted in strong staining of the notochord, but also weak staining of the intervertebral disc and vertebral body. Using laser capture microdissection, the notochordal section is isolated and good quality of RNA distracted from this section is confirmed. The authors discuss isolation of other gestational ages and suggest this protocol to be the new standard in isolation of notochordal tissue.

Using the collected intervertebral discs discussed in the previous chapter, in chapter 4, we report on technical specifications and requirements for isolating and comparing protein samples in proteomic analysis. Total protein extraction, as well as specific TiO2-based phosphopeptides enrichment and hydrazyde-based glycopeptides purification are described in detail. For quantitative differential protein expression, isobaric labeling with isobaric tag for relative and absolute quantitation (iTRAQ) is combined with liquid chromatography–mass spectrometry (LC-MS). The vast amount of data gathered from these analysis, represent a challenge for the investigator. The authors suggest the usage of several bioinformatics tools, depending on the preferred application. Software using gene enrichment with tabular results (DAVID, gProfiler, etc.) and gene enrichment with graphical presentation (String, Cytoscape) might prove to be helpful. However, as these results depict the association of genes, with protein-protein libraries not readily available, validation of such results remains a crucial aspect.

Chapter 5 sets out to use such a proteomic assessment to answer a pivotal question in chordoma, and to a lesser extent in chondrosarcomas; “Can we find biomarkers relating to tumor recurrence and do we have potential therapeutics to target them?”. Here, we have pooled tumor lysates from patients in specific groups: primary chordomas, primary chordomas that recurred, primary chondrosarcomas, and primary chondrosarcomas that recurred. With the use of TiO2 phosphopeptide enrichment protocols, similar to ones described in chapter 4, the four groups of iTRAQ labeled phosphopeptides were compared after tandem mass spectrometry analysis. Differential expression of proteins between the tumors that have recurred plotted against tumors that haven’t, demonstrates a set of biomarkers that are potentially associated with recurrence. In chordoma that recurred, increased phosphorylation of the discovered peptide could be tracked down to an increase of kinase activity in the tyrosine-protein kinases feline sarcoma oncogene (FES) and feline encephalitis virus-related kinase (FER). With regards to drug sensitivity, the set of phosphorylated chordoma proteins of recurrent patients were predicted to be targeted by Nilotinib and Imatinib. Future studies have to validate this proof of concept.

To summarize chapter 6, the authors examined cell cycle biomarkers in chordoma patients treated in our institution. Twenty-five formalin fixed paraffin-embedded chordoma specimens are immunohistochemically stained with antibodies for the cell-cycle markers protein 53 (p53), cyclin dependent kinase 4 (CDK4) and murine double minute 2 (MDM2). Extent of the staining was semi-quantitatively scored on number of positive cells (nuclear expression) categorized as 0 (Negative), 1 (1–10%), 2 (10–50%) and 3 (>50%) and quantitatively scored using optical density measurements. Clinical variables, (e.g. age, gender and overall survival) and pathological outcomes scores (e.g. necrosis, mitotic index and Ki67 scores (MIBN-LI)) were also recorded. All three cell cycle markers showed a significant correlation with MIBN-LI, although the MIBN-LI along with patient age, gender, tumor location failed to show a relation with survival. Expression of CDK4 (p=0.02) and PSP (p<0.01), however, were both significantly correlated to poor overall survival. Also, histologically observed necrosis (p<0.05) and a dedifferentiated tumor subtype (p<0.01) were related to adverse patient outcome. These results advocate a direct or indirect link of these biomarkers with worse outcome in patients with chordoma tumors. The final research chapter, chapter 7, concerns a novel hypothesis of viral involvement in the etiology of chordomas. As previously mentioned, a recent assumption in the etiology of chordomas, is chromothripsis. Viral involvement is frequently heralded as an event preceding this phenomenon. To examine the theory of viral involvement in chordoma, the authors report on a study in which the presence of multiple oncogenic viruses were assessed in chordoma and chondrosarcoma patients. As the most outstanding finding, the presence of parvovirus BN9 (PVBN9) DNA was detected in 4 of 18 chordomas (22%) and in 1 of 15 chondrosarcomas (7%). Immunohistochemical analysis recognized the VP2 capsid protein of PVBN9 in 44% of cases (14 of 32), suggesting an even higher percentage of tumors affected by the virus. DNA from other viruses such as human herpes virus 7 (HHV-7) and Epstein-Barr virus (EBV/HHV4) were present in 6 of the 18 (33%) and 4 of the 18 (22%) chordoma samples. Although a thorough genome-wide search of viral DNA would increase understanding of this hypothesis, this study potentiates a viral involvement in the etiology or pathophysiology of chordoma. In the final chapter, the authors address the important findings and place them into a general scope of chordoma research. Firstly, a general description of the clinical challenges is briefly touched on to highlight the need for biomarker research. We challenge certain important etiologic and pathophysiological concepts, such as the notochordal origin and the significance of brachyury expression and chromothripsis. Due to lack of typical notochordal biomarkers, we finally accept a notochordal origin for these tumors as “most applicable” until a notochordal expression profile is recognized. Also, we advocate the established use of methylation profiling in unraveling the (epi)genetic fingerprint that points to the hallmark of chordomagenesis. We further highlight the amount of biomarker research performed and the paucity of these results being validated on larger data sets. This by itself addresses the great challenge most chordoma researchers stumble on; the unfortunate small number of patients and cases when compares to other more frequent occurring tumors, such as breast and colon cancer. Most of the biomarker study-result, including ours, therefore must be placed into a different perspective. However, it is still required of researchers, especially in a small research area like chordomas, to use a holistic understanding of the disease, including the paucity of high quality data, and to apply this knowledge for future studies. Amalgamating unexplored zones of tumor biology and multiple fields of cell biology, would help discover molecular (prognostic) biomarkers that might advance as molecular targets for medical therapy. This way, we hope to lift the burden that physicians and patient face in the clinical management of these tumors. Valorisation Knowledge valorisation is described as ‘the process of creating value from knowledge, by making knowledge suitable and/or available for social (and/or economic) use and by making knowledge suitable for translation into competitive products, services, processes and new commercial activities’. (definition taken from the report of the National Valorisation Committee, Leonie van Drooge, Rens Vandeberg et al., Waardevol: Indicatoren voor Valorisatie. Den Haag, Rathenau Instituut, 2011). Introduction The global burden of human cancer worldwide is rapidly growing and is predicted to become the single leading cause of death in every country of the world in the 21st century. According to the latest release of the Global Cancer Observatory (GCO, GLOBOCAN 2018), it is estimated that there have been 18.1 million new cancer cases and 9.6 million cancer deaths in 2018. As we progress and improve our understanding of the epidemiology of cancers, the multitude of cancer-causing variables increases which captures the extraordinary diversity of this challenging disease. Due to research initiatives that are attempting to increase our understanding, preventative measures and novel processes and services are increasingly developed to tackle this conundrum. This section is dedicated to scientifically valorizing the research result of the current thesis and should be viewed through a biomedical scope. Fundamental research relevance The ‘publish and/or perish’ era we are currently participating in, results in minor appreciation for the fundamental understanding of diseased mechanisms and focusses excessively on clinical applicability. When viewed historically, great advances in medicine and biomedical science are the collective achievement of thousands by scientists for whom the primary goal was to understand fundamental mechanisms of biology. Funding institutions and project managers aiming primarily at swift results, preferring observing to understanding, often neglect this concept. As a result, we start to observe less and less and creativity is replaced by productivity. In chordoma, understanding the molecular processes responsible for tumor growth, cell and molecular heterogeneity and treatment resilience, are pivotal in the quest for finding a potential target for disease prediction and therapy. In order to provide insight into the majority of molecular data available we have published a comprehensive review of these molecules in our second chapter. One of the interesting mechanisms discussed is “chromothripsis”. A phenomenon deemed responsible for the chromosomal instability often witnessed in chordoma cells. Investigating viral involvement in chordoma was a rational follow up as it is stated to relate to chromothripsis. In chapter 7, we therefore studied the presence of oncogenic viruses in healthy and tumorous tissue derived from chordoma patients. The observation indicates a potential involvement of primarily parvovirus BN9 in the disease, broadening the scope of potential biomarkers for potential clinical application. Clinical relevance The body of work in this dissertation has the primary focus to increase the scientific understanding of the tumor’s biological behavior and to help unravel the complex heterogeneity that accompanies clinical research in chordoma. The essential foundation of clinical management of all patients, right after communicating the diagnosis, is guidance of patients during the disease course. For most physicians, a vigorous challenge lies within this aspect of care where patients, after processing the devastating diagnosis, generally inquire about the prognosis in order to assess their expected quality of life for the remainder of their life. For many experienced physicians the main dilemma lies within the vast heterogeneity of the prognosis. This complexity is, amongst others, primarily derived from the often-observed distinct treatment sequence and, more likely, biological predisposition. Identifying biological substances (biomarkers) that relate to a specific disease state, susceptibility to treatment or outcome often capture this biological tendency for a malignant or benign outcome. It should therefore come to no surprise that hunt on biomarkers in current biomedical research is one of the leading subjects. Discovered biomarkers in other cancers (e.g. glioma) are utilized to enhance the diagnosis and moreover to predict treatment outcome and prognosis. In the design of the study reported on in chapter 6 the authors set out to investigate clinical and molecular markers related to worse outcome. As displayed, were patient characteristics are not reliable markers for the prediction of disease course, molecular markers p53 and CDK4 are related to a more malignant disease course. As we were the first to report on the expression of CDK4 in chordoma and its relation to worse prognosis, it is interesting to note that recently, this negative association has been validated in a larger cohort of 75 patients further cementing this claim. Currently, a clinical trial using a CDK4 inhibitor, Palbociclib, is recruiting patients with advanced disease in chordoma. As this typical “from bench to bedside” exemplifies an increase of clinically relevant experiments performed today, one must note that often these single biomarker based research initiatives often fail to be reproduced in other datasets as important aspects, such as dose-response are often overlooked. Identifying a set of biomarkers often results in a more comprehensive understanding of the diseases state. And so, the authors of this thesis in chapter have set out to investigate a more diverse set of proteins (proteome) with its relation to different chordoma like tissues. With the previously mentioned notochordal proposed origin and closely related chondrosarcoma’s, a comparative study design was implemented. Data of such a design yielded significant findings, not included in this thesis, with a potential to serve clinical biomarkers. However, computational analysis using algorithms to examine large sets of data with respect to certain clinical aspects bear the risk of overlooking the vulnerability of overfitting the data. This occurs when a multitude of potential predictive biomarkers are applied to differentiate a small number of outcome events. The apparent risk of failure to reproduce such findings with other clinical dataset remains a difficult problem. A socio-economic perspective As the incidence of cancer is increasing, a substantial socio-economic burden arises as, according to the World Health Organization (WHO) healthcare associated costs rising when plotted against per capita GDP. Latest indications (2018, CBS) indicate a weight of approximately 100 billion euro in the Netherlands, where most of the spending is made in the insured spectrum (Zorgverzekeringswet). Consideration on allocation of health care resources are based largely on cost-effectiveness analysis in which terms as quality-adjusted life-year (QALY) are the mainstay. Applying QALY’s in decision support for allocation in cancer treatment are shown to contain important limitations. However, the high cost of (novel) cancer treatments on society is increasing and will become unaffordable, making further progression of the current standard impossible. Most of these costs in oncology are attributed to the high price of cancer drugs and the overuse of drugs in anticancer treatment regiments. Potent solutions to this problem have been proposed, with the most potent being a shift towards treatments tailored specifically to the genetic makeup of an individual’s tumor. This way expenditure of (new) anticancer drugs that confer only a minor level of clinical benefit are eliminated. In chordoma, standard primary treatment regimen for patients lacks any drug therapy, as chordomas are deemed unresponsive to conventional chemotherapy. Due to this limitation, a significant proportion of treated patients have a propensity to show tumor recurrence. Although recent treatment guidelines on this, frequently observed, recurrent growth of the tumor do reserve the possibility of implementing targeted therapy in the form of anticancer drugs in advanced disease, the scientific evidence favoring such a treatment remains limited. Reasons for this predicament are the relatively low incidence of chordoma and the heterogenic markup of the patient characteristics and cell pathophysiology. High costs, as a result of efforts of medical oncologists to treat patients with advanced disease and subsequent morbidity accompanying such a treatment, would be substantially diminished if tumor recurrence and implementation and accuracy of a patient tailored approach in advanced disease could be predicted. In this thesis, chapter 5 is dedicated to discovering potential biomarkers in the tumor cells of chordoma and chondrosarcoma patients. By comparing tumor protein lysates of primary tumor of patients with no recurrence and primary tumor of patients with recurrent disease, we were able to identify a spectrum of phosphorylated kinases that are indicative of recurrent disease. Two of the more frequently proposed antidrug therapies in chordoma Imatinib and Nilotinib hold promise by targeting these biological factors. Although these findings require validation in a larger cohort, applying these biomarkers will benefit optimization of individual health care for chordoma patients and minimize disease burden for the general population by reducing unnecessary treatments and their related morbidities. From proprietary data to a collective effect and future perspective As quoted by Isaac Newton (1642-1727): “If I have seen further than others, it is by standing upon the shoulders of giants”’, it is only with the help of previous research that we gain our insights in the underlying mechanism of the diseased process. For this to continue, it is imperative that scientist, as much as possible, confirm to the notion of shared knowledge to enable synergistic research. In this thesis, two research articles are dedicated to aiding future chordoma research endeavors by informing the reading on how to perform certain procedures. In chapter 3, the authors illustrate how to isolate human notochordal tissue with the help of laser captured microdissection. Here, advice and procedural strategies on extraction of tissue from human fetuses, with a gestation of 9, 11 and 13 weeks, are discussed. This is relevant for chordoma research, as this tissue is proposed to be essential in the etiology of this tumor. A growing body of studies performed on the etiology of chordoma will therefore benefit from this detailed description. Similarly, observation of trends toward increased investigation of the proteome in cancer research, has instigated our initiative to publish the methodology we have applied to research the chordoma proteome. Although the results of this final study are outside of the scope of this thesis, the promising outcome supports a more frequent use of this field of research. As the title of the current manuscript already implies and recent standards in clinical care are progressively leaning towards the use of molecular biomarkers to guide our understanding and management of the disease. The author expects that single biomarker-based study designs will be replaced by a broader set of biomarkers derived from an integrated assessment of the disease state (e.g. genomics, transcriptomics, proteomics as well as metabolomics and nutriomics) plotted against a well-documented clinical databases. This large datasets, “Big Data”, when balanced and validated accordingly, diminish concerns about clinical and statistical validity that hinder their application and will provide the clinician and patients with increased knowledge on how to improve the quality of life.